diff options
Diffstat (limited to 'src/VBox/Runtime/r3/posix')
54 files changed, 21334 insertions, 0 deletions
diff --git a/src/VBox/Runtime/r3/posix/Makefile.kup b/src/VBox/Runtime/r3/posix/Makefile.kup new file mode 100644 index 00000000..e69de29b --- /dev/null +++ b/src/VBox/Runtime/r3/posix/Makefile.kup diff --git a/src/VBox/Runtime/r3/posix/RTFileQueryFsSizes-posix.cpp b/src/VBox/Runtime/r3/posix/RTFileQueryFsSizes-posix.cpp new file mode 100644 index 00000000..886c3844 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTFileQueryFsSizes-posix.cpp @@ -0,0 +1,78 @@ +/* $Id: RTFileQueryFsSizes-posix.cpp $ */ +/** @file + * IPRT - File I/O, RTFileFsQuerySizes, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_FILE + +#include <errno.h> +#include <sys/types.h> +#include <fcntl.h> +#include <sys/statvfs.h> + +#include <iprt/file.h> +#include <iprt/assert.h> +#include <iprt/errcore.h> +#include <iprt/log.h> +#include <iprt/string.h> + + +RTR3DECL(int) RTFileQueryFsSizes(RTFILE hFile, PRTFOFF pcbTotal, RTFOFF *pcbFree, + uint32_t *pcbBlock, uint32_t *pcbSector) +{ + struct statvfs StatVFS; + RT_ZERO(StatVFS); + if (fstatvfs(RTFileToNative(hFile), &StatVFS)) + return RTErrConvertFromErrno(errno); + + /* + * Calc the returned values. + */ + if (pcbTotal) + *pcbTotal = (RTFOFF)StatVFS.f_blocks * StatVFS.f_frsize; + if (pcbFree) + *pcbFree = (RTFOFF)StatVFS.f_bavail * StatVFS.f_frsize; + if (pcbBlock) + *pcbBlock = StatVFS.f_frsize; + /* no idea how to get the sector... */ + if (pcbSector) + *pcbSector = 512; + + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/RTFileSetAllocationSize-posix.cpp b/src/VBox/Runtime/r3/posix/RTFileSetAllocationSize-posix.cpp new file mode 100644 index 00000000..534754ac --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTFileSetAllocationSize-posix.cpp @@ -0,0 +1,87 @@ +/* $Id: RTFileSetAllocationSize-posix.cpp $ */ +/** @file + * IPRT - RTFileSetAllocationSize, linux implementation. + */ + +/* + * Copyright (C) 2016-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_FILE +#include <iprt/file.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include <iprt/errcore.h> + +#include <dlfcn.h> +#include <errno.h> +#include <unistd.h> +#include <fcntl.h> + +/** + * The posix_fallocate() method. + */ +typedef int (*PFNPOSIXFALLOCATE) (int iFd, off_t offStart, off_t cb); + +RTDECL(int) RTFileSetAllocationSize(RTFILE hFile, uint64_t cbSize, uint32_t fFlags) +{ + AssertReturn(hFile != NIL_RTFILE, VERR_INVALID_PARAMETER); + AssertReturn(!(fFlags & ~RTFILE_ALLOC_SIZE_F_VALID), VERR_INVALID_PARAMETER); + AssertMsgReturn(sizeof(off_t) >= sizeof(cbSize) || RT_HIDWORD(cbSize) == 0, + ("64-bit filesize not supported! cbSize=%lld\n", cbSize), + VERR_NOT_SUPPORTED); + + if (fFlags & RTFILE_ALLOC_SIZE_F_KEEP_SIZE) + return VERR_NOT_SUPPORTED; + + int rc = VINF_SUCCESS; + PFNPOSIXFALLOCATE pfnPosixFAllocate = (PFNPOSIXFALLOCATE)(uintptr_t)dlsym(RTLD_DEFAULT, "posix_fallocate"); + if (RT_VALID_PTR(pfnPosixFAllocate)) + { + int rcPosix = pfnPosixFAllocate(RTFileToNative(hFile), 0, cbSize); + if (rcPosix != 0) + { + if (errno == EOPNOTSUPP) + rc = VERR_NOT_SUPPORTED; + else + rc = RTErrConvertFromErrno(errno); + } + } + else + rc = VERR_NOT_SUPPORTED; + + return rc; +} +RT_EXPORT_SYMBOL(RTFileSetAllocationSize); diff --git a/src/VBox/Runtime/r3/posix/RTHandleGetStandard-posix.cpp b/src/VBox/Runtime/r3/posix/RTHandleGetStandard-posix.cpp new file mode 100644 index 00000000..30e54331 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTHandleGetStandard-posix.cpp @@ -0,0 +1,144 @@ +/* $Id: RTHandleGetStandard-posix.cpp $ */ +/** @file + * IPRT - RTHandleGetStandard, POSIX. + */ + +/* + * Copyright (C) 2012-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <errno.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <sys/ioctl.h> +#include <fcntl.h> +#ifdef _MSC_VER +# include <io.h> +#else +# include <unistd.h> +#endif + +#include "internal/iprt.h" +#include <iprt/handle.h> + +#include <iprt/file.h> +#include <iprt/pipe.h> +#include <iprt/assert.h> +#include <iprt/errcore.h> +#include <iprt/log.h> + +#include "internal/socket.h" + + + +RTDECL(int) RTHandleGetStandard(RTHANDLESTD enmStdHandle, bool fLeaveOpen, PRTHANDLE ph) +{ + /* + * Validate and convert input. + */ + AssertPtrReturn(ph, VERR_INVALID_POINTER); + int fd; + switch (enmStdHandle) + { + case RTHANDLESTD_INPUT: fd = 0; break; + case RTHANDLESTD_OUTPUT: fd = 1; break; + case RTHANDLESTD_ERROR: fd = 2; break; + default: + AssertFailedReturn(VERR_INVALID_PARAMETER); + } + + /* + * Is the requested descriptor valid and which IPRT handle type does it + * best map on to? + */ + struct stat st; + int rc = fstat(fd, &st); + if (rc == -1) + return RTErrConvertFromErrno(errno); + + rc = fcntl(fd, F_GETFD, 0); + if (rc == -1) + return RTErrConvertFromErrno(errno); + bool const fInherit = !(rc & FD_CLOEXEC); + + RTHANDLE h; + if (S_ISREG(st.st_mode)) + h.enmType = RTHANDLETYPE_FILE; + else if ( S_ISFIFO(st.st_mode) + || (st.st_mode == 0 && st.st_nlink == 0 /*see bugs on bsd manpage*/)) + h.enmType = RTHANDLETYPE_PIPE; + else if (S_ISSOCK(st.st_mode)) + { + /** @todo check if it's really a socket... IIRC some OSes reports + * anonymouse pips as sockets. */ + h.enmType = RTHANDLETYPE_SOCKET; + } +#if 0 /** @todo re-enable this when the VFS pipe has been coded up. */ + else if (isatty(fd)) + h.enmType = RTHANDLETYPE_PIPE; +#endif + else + h.enmType = RTHANDLETYPE_FILE; + + /* + * Create the IPRT handle. + */ + switch (h.enmType) + { + case RTHANDLETYPE_FILE: + /** @todo fLeaveOpen */ + rc = RTFileFromNative(&h.u.hFile, fd); + break; + + case RTHANDLETYPE_PIPE: + rc = RTPipeFromNative(&h.u.hPipe, fd, + (enmStdHandle == RTHANDLESTD_INPUT ? RTPIPE_N_READ : RTPIPE_N_WRITE) + | (fInherit ? RTPIPE_N_INHERIT : 0) + | (fLeaveOpen ? RTPIPE_N_LEAVE_OPEN : 0)); + break; + + case RTHANDLETYPE_SOCKET: + rc = rtSocketCreateForNative(&h.u.hSocket, fd, fLeaveOpen); + break; + + default: /* shut up gcc */ + return VERR_INTERNAL_ERROR; + } + + if (RT_SUCCESS(rc)) + *ph = h; + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/RTMemProtect-posix.cpp b/src/VBox/Runtime/r3/posix/RTMemProtect-posix.cpp new file mode 100644 index 00000000..41400a87 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTMemProtect-posix.cpp @@ -0,0 +1,105 @@ +/* $Id: RTMemProtect-posix.cpp $ */ +/** @file + * IPRT - Memory Allocation, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/alloc.h> +#include <iprt/assert.h> +#include <iprt/param.h> +#include <iprt/errcore.h> +#include <iprt/string.h> + +#include <errno.h> +#include <sys/mman.h> + + +RTDECL(int) RTMemProtect(void *pv, size_t cb, unsigned fProtect) RT_NO_THROW_DEF +{ + /* + * Validate input. + */ + if (cb == 0) + { + AssertMsgFailed(("!cb\n")); + return VERR_INVALID_PARAMETER; + } + if (fProtect & ~(RTMEM_PROT_NONE | RTMEM_PROT_READ | RTMEM_PROT_WRITE | RTMEM_PROT_EXEC)) + { + AssertMsgFailed(("fProtect=%#x\n", fProtect)); + return VERR_INVALID_PARAMETER; + } + + /* + * Convert the flags. + */ + int fProt; +#if RTMEM_PROT_NONE == PROT_NONE \ + && RTMEM_PROT_READ == PROT_READ \ + && RTMEM_PROT_WRITE == PROT_WRITE \ + && RTMEM_PROT_EXEC == PROT_EXEC + fProt = fProtect; +#else + Assert(!RTMEM_PROT_NONE); + if (!fProtect) + fProt = PROT_NONE; + else + { + fProt = 0; + if (fProtect & RTMEM_PROT_READ) + fProt |= PROT_READ; + if (fProtect & RTMEM_PROT_WRITE) + fProt |= PROT_WRITE; + if (fProtect & RTMEM_PROT_EXEC) + fProt |= PROT_EXEC; + } +#endif + + /* + * Align the request. + */ + cb += (uintptr_t)pv & PAGE_OFFSET_MASK; + pv = (void *)((uintptr_t)pv & ~PAGE_OFFSET_MASK); + + /* + * Change the page attributes. + */ + int rc = mprotect(pv, cb, fProt); + if (!rc) + return rc; + return RTErrConvertFromErrno(errno); +} diff --git a/src/VBox/Runtime/r3/posix/RTMpGetCount-posix.cpp b/src/VBox/Runtime/r3/posix/RTMpGetCount-posix.cpp new file mode 100644 index 00000000..f48d6f98 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTMpGetCount-posix.cpp @@ -0,0 +1,89 @@ +/* $Id: RTMpGetCount-posix.cpp $ */ +/** @file + * IPRT - RTMpGetCount, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/mp.h> +#include <iprt/assert.h> + +#include <unistd.h> +#if !defined(RT_OS_SOLARIS) +# include <sys/sysctl.h> +#endif + + +RTDECL(RTCPUID) RTMpGetCount(void) +{ + /* + * The sysconf way (linux and others). + */ +#if defined(_SC_NPROCESSORS_MAX) || defined(_SC_NPROCESSORS_CONF) || defined(_SC_NPROCESSORS_ONLN) + int cCpusSC = -1; +# ifdef _SC_NPROCESSORS_MAX + int cMax = sysconf(_SC_NPROCESSORS_MAX); + cCpusSC = RT_MAX(cCpusSC, cMax); +# endif +# ifdef _SC_NPROCESSORS_CONF + int cConf = sysconf(_SC_NPROCESSORS_CONF); + cCpusSC = RT_MAX(cCpusSC, cConf); +# endif +# ifdef _SC_NPROCESSORS_ONLN + int cOnln = sysconf(_SC_NPROCESSORS_ONLN); + cCpusSC = RT_MAX(cCpusSC, cOnln); +# endif + Assert(cCpusSC > 0); + if (cCpusSC > 0) + return cCpusSC; +#endif + + /* + * The BSD 4.4 way. + */ +#if defined(CTL_HW) && defined(HW_NCPU) + int aiMib[2]; + aiMib[0] = CTL_HW; + aiMib[1] = HW_NCPU; + int cCpus = -1; + size_t cb = sizeof(cCpus); + int rc = sysctl(aiMib, RT_ELEMENTS(aiMib), &cCpus, &cb, NULL, 0); + if (rc != -1 && cCpus >= 1) + return cCpus; +#endif + return 1; +} + diff --git a/src/VBox/Runtime/r3/posix/RTPathUserDocuments-posix.cpp b/src/VBox/Runtime/r3/posix/RTPathUserDocuments-posix.cpp new file mode 100644 index 00000000..ee8c8036 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTPathUserDocuments-posix.cpp @@ -0,0 +1,63 @@ +/* $Id: RTPathUserDocuments-posix.cpp $ */ +/** @file + * IPRT - RTPathUserDocuments, posix ring-3. + */ + +/* + * Copyright (C) 2011-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/path.h> +#include <iprt/errcore.h> +#include <iprt/assert.h> + +RTDECL(int) RTPathUserDocuments(char *pszPath, size_t cchPath) +{ + /* + * Validate input + */ + AssertPtrReturn(pszPath, VERR_INVALID_POINTER); + AssertReturn(cchPath, VERR_INVALID_PARAMETER); + + int rc = RTPathUserHome(pszPath, cchPath); + if (RT_FAILURE(rc)) + return rc; + + rc = RTPathAppend(pszPath, cchPath, "Documents"); + if (RT_FAILURE(rc)) + *pszPath = '\0'; + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/RTPathUserHome-posix.cpp b/src/VBox/Runtime/r3/posix/RTPathUserHome-posix.cpp new file mode 100644 index 00000000..daa30ea1 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTPathUserHome-posix.cpp @@ -0,0 +1,173 @@ +/* $Id: RTPathUserHome-posix.cpp $ */ +/** @file + * IPRT - Path Manipulation, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_PATH +#include <stdlib.h> +#include <limits.h> +#include <errno.h> +#include <unistd.h> +#include <sys/stat.h> +#include <sys/time.h> +#include <stdio.h> +#include <sys/types.h> +#include <pwd.h> + +#include <iprt/path.h> +#include <iprt/env.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include "internal/path.h" +#include "internal/fs.h" + + +#ifndef RT_OS_L4 +/** + * Worker for RTPathUserHome that looks up the home directory + * using the getpwuid_r api. + * + * @returns IPRT status code. + * @param pszPath The path buffer. + * @param cchPath The size of the buffer. + * @param uid The User ID to query the home directory of. + */ +static int rtPathUserHomeByPasswd(char *pszPath, size_t cchPath, uid_t uid) +{ + /* + * The getpwuid_r function uses the passed in buffer to "allocate" any + * extra memory it needs. On some systems we should probably use the + * sysconf function to find the appropriate buffer size, but since it won't + * work everywhere we'll settle with a 5KB buffer and ASSUME that it'll + * suffice for even the lengthiest user descriptions... + */ + char achBuffer[5120]; + struct passwd Passwd; + struct passwd *pPasswd; + memset(&Passwd, 0, sizeof(Passwd)); + int rc = getpwuid_r(uid, &Passwd, &achBuffer[0], sizeof(achBuffer), &pPasswd); + if (rc != 0) + return RTErrConvertFromErrno(rc); + if (!pPasswd) /* uid not found in /etc/passwd */ + return VERR_PATH_NOT_FOUND; + + /* + * Check that it isn't empty and that it exists. + */ + struct stat st; + if ( !pPasswd->pw_dir + || !*pPasswd->pw_dir + || stat(pPasswd->pw_dir, &st) + || !S_ISDIR(st.st_mode)) + return VERR_PATH_NOT_FOUND; + + /* + * Convert it to UTF-8 and copy it to the return buffer. + */ + return rtPathFromNativeCopy(pszPath, cchPath, pPasswd->pw_dir, NULL); +} +#endif + + +/** + * Worker for RTPathUserHome that looks up the home directory + * using the HOME environment variable. + * + * @returns IPRT status code. + * @param pszPath The path buffer. + * @param cchPath The size of the buffer. + */ +static int rtPathUserHomeByEnv(char *pszPath, size_t cchPath) +{ + /* + * Get HOME env. var it and validate it's existance. + */ + int rc = VERR_PATH_NOT_FOUND; + const char *pszHome = RTEnvGet("HOME"); /** @todo Codeset confusion in RTEnv. */ + if (pszHome) + + { + struct stat st; + if ( !stat(pszHome, &st) + && S_ISDIR(st.st_mode)) + rc = rtPathFromNativeCopy(pszPath, cchPath, pszHome, NULL); + } + return rc; +} + + +RTDECL(int) RTPathUserHome(char *pszPath, size_t cchPath) +{ + int rc; +#ifndef RT_OS_L4 + /* + * We make an exception for the root user and use the system call + * getpwuid_r to determine their initial home path instead of + * reading it from the $HOME variable. This is because the $HOME + * variable does not get changed by sudo (and possibly su and others) + * which can cause root-owned files to appear in user's home folders. + */ + uid_t uid = geteuid(); + if (!uid) + rc = rtPathUserHomeByPasswd(pszPath, cchPath, uid); + else + rc = rtPathUserHomeByEnv(pszPath, cchPath); + + /* + * On failure, retry using the alternative method. + * (Should perhaps restrict the retry cases a bit more here...) + */ + if ( RT_FAILURE(rc) + && rc != VERR_BUFFER_OVERFLOW) + { + if (!uid) + rc = rtPathUserHomeByEnv(pszPath, cchPath); + else + rc = rtPathUserHomeByPasswd(pszPath, cchPath, uid); + } +#else /* RT_OS_L4 */ + rc = rtPathUserHomeByEnv(pszPath, cchPath); +#endif /* RT_OS_L4 */ + + LogFlow(("RTPathUserHome(%p:{%s}, %u): returns %Rrc\n", pszPath, + RT_SUCCESS(rc) ? pszPath : "<failed>", cchPath, rc)); + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/RTSystemQueryOSInfo-posix.cpp b/src/VBox/Runtime/r3/posix/RTSystemQueryOSInfo-posix.cpp new file mode 100644 index 00000000..6dc2f975 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTSystemQueryOSInfo-posix.cpp @@ -0,0 +1,100 @@ +/* $Id: RTSystemQueryOSInfo-posix.cpp $ */ +/** @file + * IPRT - RTSystemQueryOSInfo, POSIX implementation. + */ + +/* + * Copyright (C) 2008-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/system.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/errcore.h> + +#include <errno.h> +#include <sys/utsname.h> + + +RTDECL(int) RTSystemQueryOSInfo(RTSYSOSINFO enmInfo, char *pszInfo, size_t cchInfo) +{ + /* + * Quick validation. + */ + AssertReturn(enmInfo > RTSYSOSINFO_INVALID && enmInfo < RTSYSOSINFO_END, VERR_INVALID_PARAMETER); + AssertPtrReturn(pszInfo, VERR_INVALID_POINTER); + if (!cchInfo) + return VERR_BUFFER_OVERFLOW; + + /* + * Handle the request. + */ + switch (enmInfo) + { + case RTSYSOSINFO_PRODUCT: + case RTSYSOSINFO_RELEASE: + case RTSYSOSINFO_VERSION: + { + struct utsname UtsInfo; + if (uname(&UtsInfo) < 0) + return RTErrConvertFromErrno(errno); + const char *pszSrc; + switch (enmInfo) + { + case RTSYSOSINFO_PRODUCT: pszSrc = UtsInfo.sysname; break; + case RTSYSOSINFO_RELEASE: pszSrc = UtsInfo.release; break; + case RTSYSOSINFO_VERSION: pszSrc = UtsInfo.version; break; + default: AssertFatalFailed(); /* screw gcc */ + } + size_t cch = strlen(pszSrc); + if (cch < cchInfo) + { + memcpy(pszInfo, pszSrc, cch + 1); + return VINF_SUCCESS; + } + memcpy(pszInfo, pszSrc, cchInfo - 1); + pszInfo[cchInfo - 1] = '\0'; + return VERR_BUFFER_OVERFLOW; + } + + + case RTSYSOSINFO_SERVICE_PACK: + default: + *pszInfo = '\0'; + return VERR_NOT_SUPPORTED; + } + + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/RTSystemQueryTotalRam-posix.cpp b/src/VBox/Runtime/r3/posix/RTSystemQueryTotalRam-posix.cpp new file mode 100644 index 00000000..151a12f5 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTSystemQueryTotalRam-posix.cpp @@ -0,0 +1,61 @@ +/* $Id: RTSystemQueryTotalRam-posix.cpp $ */ +/** @file + * IPRT - RTSystemQueryTotalRam, windows ring-3. + */ + +/* + * Copyright (C) 2010-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/system.h> +#include "internal/iprt.h" + +#include <iprt/errcore.h> +#include <iprt/assert.h> +#include <iprt/string.h> + + + + +RTDECL(int) RTSystemQueryTotalRam(uint64_t *pcb) +{ + AssertPtrReturn(pcb, VERR_INVALID_POINTER); + return VERR_NOT_IMPLEMENTED; +} + +RTDECL(int) RTSystemQueryAvailableRam(uint64_t *pcb) +{ + AssertPtrReturn(pcb, VERR_INVALID_POINTER); + return VERR_NOT_IMPLEMENTED; +} diff --git a/src/VBox/Runtime/r3/posix/RTTimeNow-posix.cpp b/src/VBox/Runtime/r3/posix/RTTimeNow-posix.cpp new file mode 100644 index 00000000..b00394c8 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTTimeNow-posix.cpp @@ -0,0 +1,61 @@ +/* $Id: RTTimeNow-posix.cpp $ */ +/** @file + * IPRT - RTTimeNow, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_TIME +#define RTTIME_INCL_TIMEVAL +#include <sys/time.h> +#include <time.h> + +#include <iprt/time.h> + + +/** + * Gets the current system time. + * + * @returns pTime. + * @param pTime Where to store the time. + */ +RTDECL(PRTTIMESPEC) RTTimeNow(PRTTIMESPEC pTime) +{ + struct timeval tv; + gettimeofday(&tv, NULL); + return RTTimeSpecSetTimeval(pTime, &tv); +} + diff --git a/src/VBox/Runtime/r3/posix/RTTimeSet-posix.cpp b/src/VBox/Runtime/r3/posix/RTTimeSet-posix.cpp new file mode 100644 index 00000000..8d848b50 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTTimeSet-posix.cpp @@ -0,0 +1,60 @@ +/* $Id: RTTimeSet-posix.cpp $ */ +/** @file + * IPRT - RTTimeSet, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_TIME +#define RTTIME_INCL_TIMEVAL +#include <sys/time.h> +#include <time.h> +#include <errno.h> + +#include <iprt/time.h> +#include "internal/iprt.h" + +#include <iprt/errcore.h> + + +RTDECL(int) RTTimeSet(PCRTTIMESPEC pTime) +{ + struct timeval tv; + if (settimeofday(RTTimeSpecGetTimeval(pTime, &tv), NULL) == 0) + return VINF_SUCCESS; + return RTErrConvertFromErrno(errno); +} + diff --git a/src/VBox/Runtime/r3/posix/RTTimeZoneGetCurrent-posix.cpp b/src/VBox/Runtime/r3/posix/RTTimeZoneGetCurrent-posix.cpp new file mode 100644 index 00000000..6d0d84d0 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/RTTimeZoneGetCurrent-posix.cpp @@ -0,0 +1,262 @@ +/* $Id: RTTimeZoneGetCurrent-posix.cpp $ */ +/** @file + * IPRT - RTTimeZoneGetCurrent, POSIX. + */ + +/* + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/time.h> +#include "internal/iprt.h" + +#include <iprt/env.h> +#include <iprt/file.h> +#include <iprt/path.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/errcore.h> +#include <iprt/types.h> +#include <iprt/symlink.h> +#include <iprt/stream.h> + +#if defined(RT_OS_DARWIN) || defined(RT_OS_SOLARIS) +# include <tzfile.h> +#else +# define TZDIR "/usr/share/zoneinfo" +# define TZ_MAGIC "TZif" +#endif + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +#define PATH_LOCALTIME "/etc/localtime" +#if defined(RT_OS_FREEBSD) +# define PATH_TIMEZONE "/var/db/zoneinfo" +#else +# define PATH_TIMEZONE "/etc/timezone" +#endif +#define PATH_SYSCONFIG_CLOCK "/etc/sysconfig/clock" + + +/** + * Checks if a time zone database file is valid by verifying it begins with + * TZ_MAGIC. + * + * @returns IPRT status code. + * @param pszTimezone The time zone database file relative to + * <tzfile.h>:TZDIR (normally /usr/share/zoneinfo), + * e.g. Europe/London, or Etc/UTC, or UTC, or etc. + * + * @note File format is documented in RFC-8536. + */ +static int rtIsValidTimeZoneFile(const char *pszTimeZone) +{ + if (pszTimeZone == NULL || *pszTimeZone == '\0' || *pszTimeZone == '/') + return VERR_INVALID_PARAMETER; + + int rc = RTStrValidateEncoding(pszTimeZone); + if (RT_SUCCESS(rc)) + { + /* construct full pathname of the time zone file */ + char szTZPath[RTPATH_MAX]; + rc = RTPathJoin(szTZPath, sizeof(szTZPath), TZDIR, pszTimeZone); + if (RT_SUCCESS(rc)) + { + /* open the time zone file and check that it begins with the correct magic number */ + RTFILE hFile = NIL_RTFILE; + rc = RTFileOpen(&hFile, szTZPath, RTFILE_O_READ | RTFILE_O_OPEN | RTFILE_O_DENY_WRITE); + if (RT_SUCCESS(rc)) + { + char achTZBuf[sizeof(TZ_MAGIC)]; + rc = RTFileRead(hFile, achTZBuf, sizeof(achTZBuf), NULL); + RTFileClose(hFile); + if (RT_SUCCESS(rc)) + { + if (memcmp(achTZBuf, RT_STR_TUPLE(TZ_MAGIC)) == 0) + rc = VINF_SUCCESS; + else + rc = VERR_INVALID_MAGIC; + } + } + } + } + + return rc; +} + + +/** + * Return the system time zone. + * + * @returns IPRT status code. + * @param pszName The buffer to return the time zone in. + * @param cbName The size of the pszName buffer. + */ +RTDECL(int) RTTimeZoneGetCurrent(char *pszName, size_t cbName) +{ + int rc = RTEnvGetEx(RTENV_DEFAULT, "TZ", pszName, cbName, NULL); + if (RT_SUCCESS(rc)) + { + /* + * $TZ can have two different formats and one of them doesn't specify + * a time zone database file under <tzfile.h>:TZDIR but since all + * current callers of this routine expect a time zone filename we do + * the validation check here so that if it is invalid then we fall back + * to the other mechanisms to return the system's current time zone. + */ + if (*pszName == ':') /* POSIX allows $TZ to begin with a colon (:) so we allow for that here */ + memmove(pszName, pszName + 1, strlen(pszName)); + /** @todo this isn't perfect for absolute paths... Should probably try treat + * it like /etc/localtime. */ + rc = rtIsValidTimeZoneFile(pszName); + if (RT_SUCCESS(rc)) + return rc; + } + else if (rc != VERR_ENV_VAR_NOT_FOUND) + return rc; + + /* + * /etc/localtime is a symbolic link to the system time zone on many OSes + * including Solaris, macOS, Ubuntu, RH/OEL 6 and later, Arch Linux, NetBSD, + * and etc. We extract the time zone pathname relative to TZDIR defined in + * <tzfile.h> which is normally /usr/share/zoneinfo. + * + * N.B. Some OSes have /etc/localtime as a regular file instead of a + * symlink and while we could trawl through all the files under TZDIR + * looking for a match we instead fallback to other popular mechanisms of + * specifying the system-wide time zone for the sake of simplicity. + */ + char szBuf[RTPATH_MAX]; + const char *pszPath = PATH_LOCALTIME; + if (RTSymlinkExists(pszPath)) + { + /* the contents of the symink may contain '..' or other links */ + char szLinkPathReal[RTPATH_MAX]; + rc = RTPathReal(pszPath, szLinkPathReal, sizeof(szLinkPathReal)); + if (RT_SUCCESS(rc)) + { + rc = RTPathReal(TZDIR, szBuf, sizeof(szBuf)); + AssertRC(rc); + if (RT_SUCCESS(rc)) + { + Assert(RTPathStartsWith(szLinkPathReal, szBuf)); + if (RTPathStartsWith(szLinkPathReal, szBuf)) + { + /* <tzfile.h>:TZDIR doesn't include a trailing slash */ + const char *pszTimeZone = &szLinkPathReal[strlen(szBuf) + 1]; + rc = rtIsValidTimeZoneFile(pszTimeZone); + if (RT_SUCCESS(rc)) + return RTStrCopy(pszName, cbName, pszTimeZone); + } + } + } + } + + /* + * /etc/timezone is a regular file consisting of a single line containing + * the time zone (e.g. Europe/London or Etc/UTC or etc.) and is used by a + * variety of Linux distros such as Ubuntu, Gentoo, Debian, and etc. + * The equivalent on FreeBSD is /var/db/zoneinfo. + */ + pszPath = PATH_TIMEZONE; + if (RTFileExists(pszPath)) + { + RTFILE hFile = NIL_RTFILE; + rc = RTFileOpen(&hFile, PATH_TIMEZONE, RTFILE_O_READ | RTFILE_O_OPEN | RTFILE_O_DENY_WRITE); + if (RT_SUCCESS(rc)) + { + size_t cbRead = 0; + rc = RTFileRead(hFile, szBuf, sizeof(szBuf), &cbRead); + RTFileClose(hFile); + if (RT_SUCCESS(rc) && cbRead > 0) + { + /* Get the first line and strip it. */ + szBuf[RT_MIN(sizeof(szBuf) - 1, cbRead)] = '\0'; + size_t const offNewLine = RTStrOffCharOrTerm(szBuf, '\n'); + szBuf[offNewLine] = '\0'; + const char *pszTimeZone = RTStrStrip(szBuf); + + rc = rtIsValidTimeZoneFile(pszTimeZone); + if (RT_SUCCESS(rc)) + return RTStrCopy(pszName, cbName, pszTimeZone); + } + } + } + + /* + * Older versions of RedHat / OEL don't have /etc/localtime as a symlink or + * /etc/timezone but instead have /etc/sysconfig/clock which contains a line + * of the syntax ZONE=Europe/London or ZONE="Europe/London" amongst other entries. + */ + pszPath = PATH_SYSCONFIG_CLOCK; + if (RTFileExists(pszPath)) + { + PRTSTREAM pStrm; + rc = RTStrmOpen(pszPath, "r", &pStrm); + if (RT_SUCCESS(rc)) + { + while (RT_SUCCESS(rc = RTStrmGetLine(pStrm, szBuf, sizeof(szBuf)))) + { + static char const s_szVarEq[] = "ZONE="; + char *pszStart = RTStrStrip(szBuf); + if (memcmp(pszStart, RT_STR_TUPLE(s_szVarEq)) == 0) + { + char *pszTimeZone = &pszStart[sizeof(s_szVarEq) - 1]; + + /* Drop any quoting before using the value, assuming it is plain stuff: */ + if (*pszTimeZone == '\"' || *pszTimeZone == '\'') + { + pszTimeZone++; + size_t const cchTimeZone = strlen(pszTimeZone); + if (cchTimeZone && (pszTimeZone[cchTimeZone - 1] == '"' || pszTimeZone[cchTimeZone - 1] == '\'')) + pszTimeZone[cchTimeZone - 1] = '\0'; + } + + rc = rtIsValidTimeZoneFile(pszTimeZone); + if (RT_SUCCESS(rc)) + { + RTStrmClose(pStrm); + return RTStrCopy(pszName, cbName, pszTimeZone); + } + } + } + RTStrmClose(pStrm); + } + } + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/allocex-r3-posix.cpp b/src/VBox/Runtime/r3/posix/allocex-r3-posix.cpp new file mode 100644 index 00000000..b8f1a1ad --- /dev/null +++ b/src/VBox/Runtime/r3/posix/allocex-r3-posix.cpp @@ -0,0 +1,120 @@ +/* $Id: allocex-r3-posix.cpp $ */ +/** @file + * IPRT - Memory Allocation, Extended Alloc Workers, posix. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define RTMEM_NO_WRAP_TO_EF_APIS +#include <iprt/mem.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include <iprt/errcore.h> +#include <iprt/string.h> +#include "../allocex.h" + +#include <sys/mman.h> + + +DECLHIDDEN(int) rtMemAllocEx16BitReach(size_t cbAlloc, uint32_t fFlags, void **ppv) +{ + AssertReturn(cbAlloc < _64K, VERR_NO_MEMORY); + + /* + * Try with every possible address hint since the possible range is very limited. + */ + int fProt = PROT_READ | PROT_WRITE | (fFlags & RTMEMALLOCEX_FLAGS_EXEC ? PROT_EXEC : 0); + uintptr_t uAddr = 0x1000; + uintptr_t uAddrLast = _64K - uAddr - cbAlloc; + while (uAddr <= uAddrLast) + { + void *pv = mmap((void *)uAddr, cbAlloc, fProt, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + if (pv && (uintptr_t)pv <= uAddrLast) + { + *ppv = pv; + return VINF_SUCCESS; + } + + if (pv) + { + munmap(pv, cbAlloc); + pv = NULL; + } + uAddr += _4K; + } + + return VERR_NO_MEMORY; +} + + +DECLHIDDEN(int) rtMemAllocEx32BitReach(size_t cbAlloc, uint32_t fFlags, void **ppv) +{ + int fProt = PROT_READ | PROT_WRITE | (fFlags & RTMEMALLOCEX_FLAGS_EXEC ? PROT_EXEC : 0); +#if ARCH_BITS == 32 + void *pv = mmap(NULL, cbAlloc, fProt, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + if (pv) + { + *ppv = pv; + return VINF_SUCCESS; + } + return VERR_NO_MEMORY; + +#elif defined(RT_OS_LINUX) +# ifdef MAP_32BIT + void *pv = mmap(NULL, cbAlloc, fProt, MAP_PRIVATE | MAP_ANONYMOUS | MAP_32BIT, -1, 0); + if (pv) + { + *ppv = pv; + return VINF_SUCCESS; + } +# endif + + /** @todo On linux, we need an accurate hint. Since I don't need this branch of + * the code right now, I won't bother starting to parse + * /proc/curproc/mmap right now... */ +#else +#endif + return VERR_NOT_SUPPORTED; +} + + +DECLHIDDEN(void) rtMemFreeExYyBitReach(void *pv, size_t cb, uint32_t fFlags) +{ + RT_NOREF_PV(fFlags); + munmap(pv, cb); +} + diff --git a/src/VBox/Runtime/r3/posix/dir-posix.cpp b/src/VBox/Runtime/r3/posix/dir-posix.cpp new file mode 100644 index 00000000..558085ab --- /dev/null +++ b/src/VBox/Runtime/r3/posix/dir-posix.cpp @@ -0,0 +1,733 @@ +/* $Id: dir-posix.cpp $ */ +/** @file + * IPRT - Directory manipulation, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_DIR +#include <errno.h> +#include <unistd.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <fcntl.h> +#include <dirent.h> +#include <dlfcn.h> +#include <stdio.h> + +#include <iprt/dir.h> +#include "internal/iprt.h" + +#include <iprt/alloca.h> +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include <iprt/mem.h> +#include <iprt/param.h> +#include <iprt/path.h> +#include <iprt/string.h> +#include "internal/dir.h" +#include "internal/fs.h" +#include "internal/path.h" + +#if !defined(RT_OS_SOLARIS) && !defined(RT_OS_HAIKU) +# define HAVE_DIRENT_D_TYPE 1 +#endif + + +RTDECL(bool) RTDirExists(const char *pszPath) +{ + bool fRc = false; + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat s; + fRc = !stat(pszNativePath, &s) + && S_ISDIR(s.st_mode); + + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTDirExists(%p={%s}): returns %RTbool\n", pszPath, pszPath, fRc)); + return fRc; +} + + +RTDECL(int) RTDirCreate(const char *pszPath, RTFMODE fMode, uint32_t fCreate) +{ + RT_NOREF_PV(fCreate); + + int rc; + fMode = rtFsModeNormalize(fMode, pszPath, 0, RTFS_TYPE_DIRECTORY); + if (rtFsModeIsValidPermissions(fMode)) + { + char const *pszNativePath; + rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat st; + if (mkdir(pszNativePath, fMode & RTFS_UNIX_MASK) == 0) + { + /* If requested, we try make use the permission bits are set + correctly when asked. For now, we'll just ignore errors here. */ + if (fCreate & RTDIRCREATE_FLAGS_IGNORE_UMASK) + { + if ( stat(pszNativePath, &st) + || (st.st_mode & 07777) != (fMode & 07777) ) + chmod(pszNativePath, fMode & RTFS_UNIX_MASK); + } + rc = VINF_SUCCESS; + } + else + { + rc = errno; + /* + * Solaris mkdir returns ENOSYS on autofs directories, and also + * did this apparently for NFS mount points in some Nevada + * development builds. It also returned EACCES when it should + * have returned EEXIST, which actually is within the POSIX + * spec (not that I like this interpretation, but it seems + * valid). Check ourselves. + */ + if ( rc == ENOSYS + || rc == EACCES) + { + rc = RTErrConvertFromErrno(rc); + if (!stat(pszNativePath, &st)) + rc = VERR_ALREADY_EXISTS; + } + else + rc = RTErrConvertFromErrno(rc); + } + } + + rtPathFreeNative(pszNativePath, pszPath); + } + else + { + AssertMsgFailed(("Invalid file mode! %RTfmode\n", fMode)); + rc = VERR_INVALID_FMODE; + } + LogFlow(("RTDirCreate(%p={%s}, %RTfmode): returns %Rrc\n", pszPath, pszPath, fMode, rc)); + return rc; +} + + +RTDECL(int) RTDirRemove(const char *pszPath) +{ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + if (rmdir(pszNativePath)) + { + rc = errno; + if (rc == EEXIST) /* Solaris returns this, the rest have ENOTEMPTY. */ + rc = VERR_DIR_NOT_EMPTY; + else if (rc != ENOTDIR) + rc = RTErrConvertFromErrno(rc); + else + { + /* + * This may be a valid path-not-found or it may be a non-directory in + * the final component. FsPerf want us to distinguish between the two, + * and trailing slash shouldn't matter because it doesn't on windows... + */ + char *pszFree = NULL; + const char *pszStat = pszNativePath; + size_t cch = strlen(pszNativePath); + if (cch > 2 && RTPATH_IS_SLASH(pszNativePath[cch - 1])) + { + pszFree = (char *)RTMemTmpAlloc(cch); + if (pszFree) + { + memcpy(pszFree, pszNativePath, cch); + do + pszFree[--cch] = '\0'; + while (cch > 2 && RTPATH_IS_SLASH(pszFree[cch - 1])); + pszStat = pszFree; + } + } + + struct stat st; + if (!stat(pszStat, &st) && !S_ISDIR(st.st_mode)) + rc = VERR_NOT_A_DIRECTORY; + else + rc = VERR_PATH_NOT_FOUND; + + if (pszFree) + RTMemTmpFree(pszFree); + } + } + + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTDirRemove(%p={%s}): returns %Rrc\n", pszPath, pszPath, rc)); + return rc; +} + + +RTDECL(int) RTDirFlush(const char *pszPath) +{ + /* + * Linux: The fsync() man page hints at this being required for ensuring + * consistency between directory and file in case of a crash. + * + * Solaris: No mentioned is made of directories on the fsync man page. + * While rename+fsync will do what we want on ZFS, the code needs more + * careful studying wrt whether the directory entry of a new file is + * implicitly synced when the file is synced (it's very likely for ZFS). + * + * FreeBSD: The FFS fsync code seems to flush the directory entry as well + * in some cases. Don't know exactly what's up with rename, but from the + * look of things fsync(dir) should work. + */ + int rc; +#ifdef O_DIRECTORY + int fd = open(pszPath, O_RDONLY | O_DIRECTORY, 0); +#else + int fd = open(pszPath, O_RDONLY, 0); +#endif + if (fd >= 0) + { + if (fsync(fd) == 0) + rc = VINF_SUCCESS; + else + { + /* Linux fsync(2) man page documents both errors as an indication + * that the file descriptor can't be flushed (seen EINVAL for usual + * directories on CIFS). BSD (OS X) fsync(2) documents only the + * latter, and Solaris fsync(3C) pretends there is no problem. */ + if (errno == EROFS || errno == EINVAL) + rc = VERR_NOT_SUPPORTED; + else + rc = RTErrConvertFromErrno(errno); + } + close(fd); + } + else + rc = RTErrConvertFromErrno(errno); + return rc; +} + + +size_t rtDirNativeGetStructSize(const char *pszPath) +{ + long cbNameMax = pathconf(pszPath, _PC_NAME_MAX); +# ifdef NAME_MAX + if (cbNameMax < NAME_MAX) /* This is plain paranoia, but it doesn't hurt. */ + cbNameMax = NAME_MAX; +# endif +# ifdef _XOPEN_NAME_MAX + if (cbNameMax < _XOPEN_NAME_MAX) /* Ditto. */ + cbNameMax = _XOPEN_NAME_MAX; +# endif + size_t cbDir = RT_UOFFSETOF_DYN(RTDIRINTERNAL, Data.d_name[cbNameMax + 1]); + if (cbDir < sizeof(RTDIRINTERNAL)) /* Ditto. */ + cbDir = sizeof(RTDIRINTERNAL); + cbDir = RT_ALIGN_Z(cbDir, 8); + + return cbDir; +} + + +int rtDirNativeOpen(PRTDIRINTERNAL pDir, uintptr_t hRelativeDir, void *pvNativeRelative) +{ + NOREF(hRelativeDir); + NOREF(pvNativeRelative); + + /* + * Convert to a native path and try opendir. + */ + char *pszSlash = NULL; + char const *pszNativePath; + int rc; + if ( !(pDir->fFlags & RTDIR_F_NO_FOLLOW) + || pDir->fDirSlash + || pDir->cchPath <= 1) + rc = rtPathToNative(&pszNativePath, pDir->pszPath, NULL); + else + { + pszSlash = (char *)&pDir->pszPath[pDir->cchPath - 1]; + *pszSlash = '\0'; + rc = rtPathToNative(&pszNativePath, pDir->pszPath, NULL); + } + if (RT_SUCCESS(rc)) + { + if ( !(pDir->fFlags & RTDIR_F_NO_FOLLOW) + || pDir->fDirSlash) + pDir->pDir = opendir(pszNativePath); + else + { + /* + * If we can get fdopendir() and have both O_NOFOLLOW and O_DIRECTORY, + * we will use open() to safely open the directory without following + * symlinks in the final component, and then use fdopendir to get a DIR + * from the file descriptor. + * + * If we cannot get that, we will use lstat() + opendir() as a fallback. + * + * We ASSUME that support for the O_NOFOLLOW and O_DIRECTORY flags is + * older than fdopendir(). + */ +#if defined(O_NOFOLLOW) && defined(O_DIRECTORY) + /* Need to resolve fdopendir dynamically. */ + typedef DIR * (*PFNFDOPENDIR)(int); + static PFNFDOPENDIR s_pfnFdOpenDir = NULL; + static bool volatile s_fInitalized = false; + + PFNFDOPENDIR pfnFdOpenDir = s_pfnFdOpenDir; + ASMCompilerBarrier(); + if (s_fInitalized) + { /* likely */ } + else + { + pfnFdOpenDir = (PFNFDOPENDIR)(uintptr_t)dlsym(RTLD_DEFAULT, "fdopendir"); + s_pfnFdOpenDir = pfnFdOpenDir; + ASMAtomicWriteBool(&s_fInitalized, true); + } + + if (pfnFdOpenDir) + { + int fd = open(pszNativePath, O_RDONLY | O_DIRECTORY | O_NOFOLLOW, 0); + if (fd >= 0) + { + pDir->pDir = pfnFdOpenDir(fd); + if (RT_UNLIKELY(!pDir->pDir)) + { + rc = RTErrConvertFromErrno(errno); + close(fd); + } + } + else + { + /* WSL returns ELOOP here, but we take no chances that O_NOFOLLOW + takes precedence over O_DIRECTORY everywhere. */ + int iErr = errno; + if (iErr == ELOOP || iErr == ENOTDIR) + { + struct stat St; + if ( lstat(pszNativePath, &St) == 0 + && S_ISLNK(St.st_mode)) + rc = VERR_IS_A_SYMLINK; + else + rc = RTErrConvertFromErrno(iErr); + } + } + } + else +#endif + { + /* Fallback. This contains a race condition. */ + struct stat St; + if ( lstat(pszNativePath, &St) != 0 + || !S_ISLNK(St.st_mode)) + pDir->pDir = opendir(pszNativePath); + else + rc = VERR_IS_A_SYMLINK; + } + } + if (pDir->pDir) + { + /* + * Init data (allocated as all zeros). + */ + pDir->fDataUnread = false; /* spelling it out */ + } + else if (RT_SUCCESS_NP(rc)) + rc = RTErrConvertFromErrno(errno); + + rtPathFreeNative(pszNativePath, pDir->pszPath); + } + if (pszSlash) + *pszSlash = RTPATH_SLASH; + return rc; +} + + +RTDECL(int) RTDirClose(RTDIR hDir) +{ + PRTDIRINTERNAL pDir = hDir; + + /* + * Validate input. + */ + if (!pDir) + return VERR_INVALID_PARAMETER; + if (pDir->u32Magic != RTDIR_MAGIC) + { + AssertMsgFailed(("Invalid pDir=%p\n", pDir)); + return VERR_INVALID_PARAMETER; + } + + /* + * Close the handle. + */ + int rc = VINF_SUCCESS; + pDir->u32Magic = RTDIR_MAGIC_DEAD; + if (closedir(pDir->pDir)) + { + rc = RTErrConvertFromErrno(errno); + AssertMsgFailed(("closedir(%p) -> errno=%d (%Rrc)\n", pDir->pDir, errno, rc)); + } + + RTMemFree(pDir); + return rc; +} + + +/** + * Ensure that there is unread data in the buffer + * and that there is a converted filename hanging around. + * + * @returns IPRT status code. + * @param pDir the open directory. Fully validated. + */ +static int rtDirReadMore(PRTDIRINTERNAL pDir) +{ + /** @todo try avoid the rematching on buffer overflow errors. */ + for (;;) + { + /* + * Fetch data? + */ + if (!pDir->fDataUnread) + { + struct dirent *pResult = NULL; +#if RT_GNUC_PREREQ(4, 6) +# pragma GCC diagnostic push +# pragma GCC diagnostic ignored "-Wdeprecated-declarations" +#endif + int rc = readdir_r(pDir->pDir, &pDir->Data, &pResult); +#if RT_GNUC_PREREQ(4, 6) +# pragma GCC diagnostic pop +#endif + if (rc) + { + rc = RTErrConvertFromErrno(rc); + /** @todo Consider translating ENOENT (The current + * position of the directory stream is invalid) + * differently. */ + AssertMsg(rc == VERR_FILE_NOT_FOUND, ("%Rrc\n", rc)); + return rc; + } + if (!pResult) + return VERR_NO_MORE_FILES; + } + + /* + * Convert the filename to UTF-8. + */ + if (!pDir->pszName) + { + int rc = rtPathFromNative(&pDir->pszName, pDir->Data.d_name, pDir->pszPath); + if (RT_FAILURE(rc)) + { + pDir->pszName = NULL; + return rc; + } + pDir->cchName = strlen(pDir->pszName); + } + if ( !pDir->pfnFilter + || pDir->pfnFilter(pDir, pDir->pszName)) + break; + rtPathFreeIprt(pDir->pszName, pDir->Data.d_name); + pDir->pszName = NULL; + pDir->fDataUnread = false; + } + + pDir->fDataUnread = true; + return VINF_SUCCESS; +} + + +#ifdef HAVE_DIRENT_D_TYPE +/** + * Converts the d_type field to IPRT directory entry type. + * + * @returns IPRT directory entry type. + * @param Unix + */ +static RTDIRENTRYTYPE rtDirType(int iType) +{ + switch (iType) + { + case DT_UNKNOWN: return RTDIRENTRYTYPE_UNKNOWN; + case DT_FIFO: return RTDIRENTRYTYPE_FIFO; + case DT_CHR: return RTDIRENTRYTYPE_DEV_CHAR; + case DT_DIR: return RTDIRENTRYTYPE_DIRECTORY; + case DT_BLK: return RTDIRENTRYTYPE_DEV_BLOCK; + case DT_REG: return RTDIRENTRYTYPE_FILE; + case DT_LNK: return RTDIRENTRYTYPE_SYMLINK; + case DT_SOCK: return RTDIRENTRYTYPE_SOCKET; + case DT_WHT: return RTDIRENTRYTYPE_WHITEOUT; + default: + AssertMsgFailed(("iType=%d\n", iType)); + return RTDIRENTRYTYPE_UNKNOWN; + } +} +#endif /*HAVE_DIRENT_D_TYPE */ + + +RTDECL(int) RTDirRead(RTDIR hDir, PRTDIRENTRY pDirEntry, size_t *pcbDirEntry) +{ + PRTDIRINTERNAL pDir = hDir; + + /* + * Validate and digest input. + */ + if (!rtDirValidHandle(pDir)) + return VERR_INVALID_PARAMETER; + AssertPtrReturn(pDirEntry, VERR_INVALID_POINTER); + + size_t cbDirEntry = sizeof(*pDirEntry); + if (pcbDirEntry) + { + AssertPtrReturn(pcbDirEntry, VERR_INVALID_POINTER); + cbDirEntry = *pcbDirEntry; + AssertMsgReturn(cbDirEntry >= RT_UOFFSETOF(RTDIRENTRY, szName[2]), + ("Invalid *pcbDirEntry=%d (min %zu)\n", *pcbDirEntry, RT_UOFFSETOF(RTDIRENTRYEX, szName[2])), + VERR_INVALID_PARAMETER); + } + + /* + * Fetch more data if necessary and/or convert the name. + */ + int rc = rtDirReadMore(pDir); + if (RT_SUCCESS(rc)) + { + /* + * Check if we've got enough space to return the data. + */ + const char *pszName = pDir->pszName; + const size_t cchName = pDir->cchName; + const size_t cbRequired = RT_UOFFSETOF(RTDIRENTRY, szName[1]) + cchName; + if (pcbDirEntry) + *pcbDirEntry = cbRequired; + if (cbRequired <= cbDirEntry) + { + /* + * Setup the returned data. + */ + pDirEntry->INodeId = pDir->Data.d_ino; /* may need #ifdefing later */ +#ifdef HAVE_DIRENT_D_TYPE + pDirEntry->enmType = rtDirType(pDir->Data.d_type); +#else + pDirEntry->enmType = RTDIRENTRYTYPE_UNKNOWN; +#endif + pDirEntry->cbName = (uint16_t)cchName; + Assert(pDirEntry->cbName == cchName); + memcpy(pDirEntry->szName, pszName, cchName + 1); + + /* free cached data */ + pDir->fDataUnread = false; + rtPathFreeIprt(pDir->pszName, pDir->Data.d_name); + pDir->pszName = NULL; + } + else + rc = VERR_BUFFER_OVERFLOW; + } + + LogFlow(("RTDirRead(%p:{%s}, %p:{%s}, %p:{%u}): returns %Rrc\n", + pDir, pDir->pszPath, pDirEntry, RT_SUCCESS(rc) ? pDirEntry->szName : "<failed>", + pcbDirEntry, pcbDirEntry ? *pcbDirEntry : 0, rc)); + return rc; +} + + +/** + * Fills dummy info into the info structure. + * This function is called if we cannot stat the file. + * + * @param pInfo The struct in question. + * @param + */ +static void rtDirSetDummyInfo(PRTFSOBJINFO pInfo, RTDIRENTRYTYPE enmType) +{ + pInfo->cbObject = 0; + pInfo->cbAllocated = 0; + RTTimeSpecSetNano(&pInfo->AccessTime, 0); + RTTimeSpecSetNano(&pInfo->ModificationTime, 0); + RTTimeSpecSetNano(&pInfo->ChangeTime, 0); + RTTimeSpecSetNano(&pInfo->BirthTime, 0); + memset(&pInfo->Attr, 0, sizeof(pInfo->Attr)); + pInfo->Attr.enmAdditional = RTFSOBJATTRADD_NOTHING; + switch (enmType) + { + default: + case RTDIRENTRYTYPE_UNKNOWN: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL; break; + case RTDIRENTRYTYPE_FIFO: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_FIFO; break; + case RTDIRENTRYTYPE_DEV_CHAR: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_DEV_CHAR; break; + case RTDIRENTRYTYPE_DIRECTORY: pInfo->Attr.fMode = RTFS_DOS_DIRECTORY | RTFS_TYPE_DIRECTORY; break; + case RTDIRENTRYTYPE_DEV_BLOCK: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_DEV_BLOCK; break; + case RTDIRENTRYTYPE_FILE: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_FILE; break; + case RTDIRENTRYTYPE_SYMLINK: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_SYMLINK; break; + case RTDIRENTRYTYPE_SOCKET: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_SOCKET; break; + case RTDIRENTRYTYPE_WHITEOUT: pInfo->Attr.fMode = RTFS_DOS_NT_NORMAL | RTFS_TYPE_WHITEOUT; break; + } +} + + +RTDECL(int) RTDirReadEx(RTDIR hDir, PRTDIRENTRYEX pDirEntry, size_t *pcbDirEntry, + RTFSOBJATTRADD enmAdditionalAttribs, uint32_t fFlags) +{ + PRTDIRINTERNAL pDir = hDir; + + /* + * Validate and digest input. + */ + if (!rtDirValidHandle(pDir)) + return VERR_INVALID_PARAMETER; + AssertPtrReturn(pDirEntry, VERR_INVALID_POINTER); + AssertMsgReturn( enmAdditionalAttribs >= RTFSOBJATTRADD_NOTHING + && enmAdditionalAttribs <= RTFSOBJATTRADD_LAST, + ("Invalid enmAdditionalAttribs=%p\n", enmAdditionalAttribs), + VERR_INVALID_PARAMETER); + AssertMsgReturn(RTPATH_F_IS_VALID(fFlags, 0), ("%#x\n", fFlags), VERR_INVALID_PARAMETER); + size_t cbDirEntry = sizeof(*pDirEntry); + if (pcbDirEntry) + { + AssertPtrReturn(pcbDirEntry, VERR_INVALID_POINTER); + cbDirEntry = *pcbDirEntry; + AssertMsgReturn(cbDirEntry >= RT_UOFFSETOF(RTDIRENTRYEX, szName[2]), + ("Invalid *pcbDirEntry=%zu (min %zu)\n", *pcbDirEntry, RT_UOFFSETOF(RTDIRENTRYEX, szName[2])), + VERR_INVALID_PARAMETER); + } + + /* + * Fetch more data if necessary and/or convert the name. + */ + int rc = rtDirReadMore(pDir); + if (RT_SUCCESS(rc)) + { + /* + * Check if we've got enough space to return the data. + */ + const char *pszName = pDir->pszName; + const size_t cchName = pDir->cchName; + const size_t cbRequired = RT_UOFFSETOF(RTDIRENTRYEX, szName[1]) + cchName; + if (pcbDirEntry) + *pcbDirEntry = cbRequired; + if (cbRequired <= cbDirEntry) + { + /* + * Setup the returned data. + */ + pDirEntry->cwcShortName = 0; + pDirEntry->wszShortName[0] = 0; + pDirEntry->cbName = (uint16_t)cchName; + Assert(pDirEntry->cbName == cchName); + memcpy(pDirEntry->szName, pszName, cchName + 1); + + /* get the info data */ + size_t cch = cchName + pDir->cchPath + 1; + char *pszNamePath = (char *)alloca(cch); + if (pszNamePath) + { + memcpy(pszNamePath, pDir->pszPath, pDir->cchPath); + memcpy(pszNamePath + pDir->cchPath, pszName, cchName + 1); + rc = RTPathQueryInfoEx(pszNamePath, &pDirEntry->Info, enmAdditionalAttribs, fFlags); + } + else + rc = VERR_NO_MEMORY; + if (RT_FAILURE(rc)) + { +#ifdef HAVE_DIRENT_D_TYPE + rtDirSetDummyInfo(&pDirEntry->Info, rtDirType(pDir->Data.d_type)); +#else + rtDirSetDummyInfo(&pDirEntry->Info, RTDIRENTRYTYPE_UNKNOWN); +#endif + rc = VWRN_NO_DIRENT_INFO; + } + + /* free cached data */ + pDir->fDataUnread = false; + rtPathFreeIprt(pDir->pszName, pDir->Data.d_name); + pDir->pszName = NULL; + } + else + rc = VERR_BUFFER_OVERFLOW; + } + + return rc; +} + + +RTDECL(int) RTDirRewind(RTDIR hDir) +{ + PRTDIRINTERNAL pDir = hDir; + + /* + * Validate and digest input. + */ + if (!rtDirValidHandle(pDir)) + return VERR_INVALID_PARAMETER; + + /* + * Do the rewinding. + */ + /** @todo OS/2 does not rescan the directory as it should. */ + rewinddir(pDir->pDir); + pDir->fDataUnread = false; + + return VINF_SUCCESS; +} + + +RTDECL(int) RTDirRename(const char *pszSrc, const char *pszDst, unsigned fRename) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszSrc, VERR_INVALID_POINTER); + AssertPtrReturn(pszDst, VERR_INVALID_POINTER); + AssertMsgReturn(*pszSrc, ("%p\n", pszSrc), VERR_INVALID_PARAMETER); + AssertMsgReturn(*pszDst, ("%p\n", pszDst), VERR_INVALID_PARAMETER); + AssertMsgReturn(!(fRename & ~RTPATHRENAME_FLAGS_REPLACE), ("%#x\n", fRename), VERR_INVALID_PARAMETER); + + /* + * Take common cause with RTPathRename. + */ + int rc = rtPathPosixRename(pszSrc, pszDst, fRename, RTFS_TYPE_DIRECTORY); + + LogFlow(("RTDirRename(%p:{%s}, %p:{%s}): returns %Rrc\n", + pszSrc, pszSrc, pszDst, pszDst, rc)); + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/env-posix.cpp b/src/VBox/Runtime/r3/posix/env-posix.cpp new file mode 100644 index 00000000..60a9112e --- /dev/null +++ b/src/VBox/Runtime/r3/posix/env-posix.cpp @@ -0,0 +1,179 @@ +/* $Id: env-posix.cpp $ */ +/** @file + * IPRT - Environment, Posix. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#ifdef RT_OS_DARWIN +/* pick the correct prototype for unsetenv. */ +# define _POSIX_C_SOURCE 1 +#endif +#include <iprt/env.h> + +#include <iprt/alloca.h> +#include <iprt/assert.h> +#if defined(DEBUG) && defined(RT_OS_LINUX) +# include <iprt/asm.h> +#endif +#include <iprt/err.h> +#include <iprt/string.h> + +#include <stdlib.h> +#include <errno.h> + +#include "internal/alignmentchecks.h" + + +RTDECL(bool) RTEnvExistsBad(const char *pszVar) +{ + AssertReturn(strchr(pszVar, '=') == NULL, false); + return RTEnvGetBad(pszVar) != NULL; +} + + +RTDECL(bool) RTEnvExist(const char *pszVar) +{ + return RTEnvExistsBad(pszVar); +} + + +RTDECL(const char *) RTEnvGetBad(const char *pszVar) +{ + AssertReturn(strchr(pszVar, '=') == NULL, NULL); + + IPRT_ALIGNMENT_CHECKS_DISABLE(); /* glibc causes trouble */ + const char *pszValue = getenv(pszVar); + IPRT_ALIGNMENT_CHECKS_ENABLE(); + return pszValue; +} + + +RTDECL(const char *) RTEnvGet(const char *pszVar) +{ + return RTEnvGetBad(pszVar); +} + + +RTDECL(int) RTEnvPutBad(const char *pszVarEqualValue) +{ + /** @todo putenv is a source memory leaks. deal with this on a per system basis. */ + if (!putenv((char *)pszVarEqualValue)) + return 0; + return RTErrConvertFromErrno(errno); +} + + +RTDECL(int) RTEnvPut(const char *pszVarEqualValue) +{ + return RTEnvPutBad(pszVarEqualValue); +} + + +RTDECL(int) RTEnvSetBad(const char *pszVar, const char *pszValue) +{ + AssertMsgReturn(strchr(pszVar, '=') == NULL, ("'%s'\n", pszVar), VERR_ENV_INVALID_VAR_NAME); + +#if defined(_MSC_VER) + /* make a local copy and feed it to putenv. */ + const size_t cchVar = strlen(pszVar); + const size_t cchValue = strlen(pszValue); + char *pszTmp = (char *)alloca(cchVar + cchValue + 2 + !*pszValue); + memcpy(pszTmp, pszVar, cchVar); + pszTmp[cchVar] = '='; + if (*pszValue) + memcpy(pszTmp + cchVar + 1, pszValue, cchValue + 1); + else + { + pszTmp[cchVar + 1] = ' '; /* wrong, but putenv will remove it otherwise. */ + pszTmp[cchVar + 2] = '\0'; + } + + if (!putenv(pszTmp)) + return 0; + return RTErrConvertFromErrno(errno); + +#else + if (!setenv(pszVar, pszValue, 1)) + return VINF_SUCCESS; + return RTErrConvertFromErrno(errno); +#endif +} + + +RTDECL(int) RTEnvSet(const char *pszVar, const char *pszValue) +{ + return RTEnvSetBad(pszVar, pszValue); +} + +RTDECL(int) RTEnvUnsetBad(const char *pszVar) +{ + AssertReturn(strchr(pszVar, '=') == NULL, VERR_ENV_INVALID_VAR_NAME); + + /* + * Check that it exists first. + */ + if (!RTEnvExist(pszVar)) + return VINF_ENV_VAR_NOT_FOUND; + + /* + * Ok, try remove it. + */ +#ifdef RT_OS_WINDOWS + /* Use putenv(var=) since Windows does not have unsetenv(). */ + size_t cchVar = strlen(pszVar); + char *pszBuf = (char *)alloca(cchVar + 2); + memcpy(pszBuf, pszVar, cchVar); + pszBuf[cchVar] = '='; + pszBuf[cchVar + 1] = '\0'; + + if (!putenv(pszBuf)) + return VINF_SUCCESS; + +#else + /* This is the preferred function as putenv() like used above does neither work on Solaris nor on Darwin. */ + if (!unsetenv((char*)pszVar)) + return VINF_SUCCESS; +#endif + + return RTErrConvertFromErrno(errno); +} + +RTDECL(int) RTEnvUnset(const char *pszVar) +{ + return RTEnvUnsetBad(pszVar); +} + diff --git a/src/VBox/Runtime/r3/posix/errvars-posix.cpp b/src/VBox/Runtime/r3/posix/errvars-posix.cpp new file mode 100644 index 00000000..b6a98709 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/errvars-posix.cpp @@ -0,0 +1,88 @@ +/* $Id: errvars-posix.cpp $ */ +/** @file + * IPRT - Save and Restore Error Variables, POSIX Ring-3. + */ + +/* + * Copyright (C) 2011-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <netdb.h> +#include <errno.h> + +#include <iprt/errcore.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include "internal/magics.h" + + + +RTDECL(PRTERRVARS) RTErrVarsSave(PRTERRVARS pVars) +{ + pVars->ai32Vars[0] = RTERRVARS_MAGIC; + pVars->ai32Vars[1] = errno; + pVars->ai32Vars[2] = h_errno; + return pVars; +} + + +RTDECL(void) RTErrVarsRestore(PCRTERRVARS pVars) +{ + AssertReturnVoid(pVars->ai32Vars[0] == RTERRVARS_MAGIC); + h_errno = pVars->ai32Vars[2]; + errno = pVars->ai32Vars[1]; +} + + +RTDECL(bool) RTErrVarsAreEqual(PCRTERRVARS pVars1, PCRTERRVARS pVars2) +{ + Assert(pVars1->ai32Vars[0] == RTERRVARS_MAGIC); + Assert(pVars2->ai32Vars[0] == RTERRVARS_MAGIC); + + return pVars1->ai32Vars[0] == pVars2->ai32Vars[0] + && pVars1->ai32Vars[1] == pVars2->ai32Vars[1] + && pVars1->ai32Vars[2] == pVars2->ai32Vars[2]; +} + + +RTDECL(bool) RTErrVarsHaveChanged(PCRTERRVARS pVars) +{ + Assert(pVars->ai32Vars[0] == RTERRVARS_MAGIC); + + return pVars->ai32Vars[0] != RTERRVARS_MAGIC + || pVars->ai32Vars[1] != errno + || pVars->ai32Vars[2] != h_errno; +} + diff --git a/src/VBox/Runtime/r3/posix/fileaio-posix.cpp b/src/VBox/Runtime/r3/posix/fileaio-posix.cpp new file mode 100644 index 00000000..befa49fb --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fileaio-posix.cpp @@ -0,0 +1,1071 @@ +/* $Id: fileaio-posix.cpp $ */ +/** @file + * IPRT - File async I/O, native implementation for POSIX compliant host platforms. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_DIR +#include <iprt/asm.h> +#include <iprt/file.h> +#include <iprt/mem.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include <iprt/thread.h> +#include <iprt/semaphore.h> +#include "internal/fileaio.h" + +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) +# include <sys/types.h> +# include <sys/sysctl.h> /* for sysctlbyname */ +#endif +#if defined(RT_OS_FREEBSD) +# include <fcntl.h> /* O_SYNC */ +#endif +#include <aio.h> +#include <errno.h> +#include <time.h> + +/* + * Linux does not define this value. + * Just define it with really big + * value. + */ +#ifndef AIO_LISTIO_MAX +# define AIO_LISTIO_MAX UINT32_MAX +#endif + +#if 0 /* Only used for debugging */ +# undef AIO_LISTIO_MAX +# define AIO_LISTIO_MAX 16 +#endif + +/** Invalid entry in the waiting array. */ +#define RTFILEAIOCTX_WAIT_ENTRY_INVALID (~0U) + +/** No-op replacement for rtFileAioCtxDump for non debug builds */ +#ifndef LOG_ENABLED +# define rtFileAioCtxDump(pCtxInt) do {} while (0) +#endif + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** + * Async I/O request state. + */ +typedef struct RTFILEAIOREQINTERNAL +{ + /** The aio control block. FIRST ELEMENT! */ + struct aiocb AioCB; + /** Next element in the chain. */ + struct RTFILEAIOREQINTERNAL *pNext; + /** Previous element in the chain. */ + struct RTFILEAIOREQINTERNAL *pPrev; + /** Current state the request is in. */ + RTFILEAIOREQSTATE enmState; + /** Flag whether this is a flush request. */ + bool fFlush; + /** Flag indicating if the request was canceled. */ + volatile bool fCanceled; + /** Opaque user data. */ + void *pvUser; + /** Number of bytes actually transferred. */ + size_t cbTransfered; + /** Status code. */ + int Rc; + /** Completion context we are assigned to. */ + struct RTFILEAIOCTXINTERNAL *pCtxInt; + /** Entry in the waiting list the request is in. */ + unsigned iWaitingList; + /** Magic value (RTFILEAIOREQ_MAGIC). */ + uint32_t u32Magic; +} RTFILEAIOREQINTERNAL, *PRTFILEAIOREQINTERNAL; + +/** + * Async I/O completion context state. + */ +typedef struct RTFILEAIOCTXINTERNAL +{ + /** Current number of requests active on this context. */ + volatile int32_t cRequests; + /** Maximum number of requests this context can handle. */ + uint32_t cMaxRequests; + /** The ID of the thread which is currently waiting for requests. */ + volatile RTTHREAD hThreadWait; + /** Flag whether the thread was woken up. */ + volatile bool fWokenUp; + /** Flag whether the thread is currently waiting in the syscall. */ + volatile bool fWaiting; + /** Flags given during creation. */ + uint32_t fFlags; + /** Magic value (RTFILEAIOCTX_MAGIC). */ + uint32_t u32Magic; + /** Flag whether the thread was woken up due to a internal event. */ + volatile bool fWokenUpInternal; + /** List of new requests which needs to be inserted into apReqs by the + * waiting thread. */ + volatile PRTFILEAIOREQINTERNAL apReqsNewHead[5]; + /** Special entry for requests which are canceled. Because only one + * request can be canceled at a time and the thread canceling the request + * has to wait we need only one entry. */ + volatile PRTFILEAIOREQINTERNAL pReqToCancel; + /** Event semaphore the canceling thread is waiting for completion of + * the operation. */ + RTSEMEVENT SemEventCancel; + /** Head of submitted elements waiting to get into the array. */ + PRTFILEAIOREQINTERNAL pReqsWaitHead; + /** Tail of submitted elements waiting to get into the array. */ + PRTFILEAIOREQINTERNAL pReqsWaitTail; + /** Maximum number of elements in the waiting array. */ + unsigned cReqsWaitMax; + /** First free slot in the waiting list. */ + unsigned iFirstFree; + /** List of requests we are currently waiting on. + * Size depends on cMaxRequests and AIO_LISTIO_MAX. */ + volatile PRTFILEAIOREQINTERNAL apReqs[1]; +} RTFILEAIOCTXINTERNAL, *PRTFILEAIOCTXINTERNAL; + +/** + * Internal worker for waking up the waiting thread. + */ +static void rtFileAioCtxWakeup(PRTFILEAIOCTXINTERNAL pCtxInt) +{ + /* + * Read the thread handle before the status flag. + * If we read the handle after the flag we might + * end up with an invalid handle because the thread + * waiting in RTFileAioCtxWakeup() might get scheduled + * before we read the flag and returns. + * We can ensure that the handle is valid if fWaiting is true + * when reading the handle before the status flag. + */ + RTTHREAD hThread; + ASMAtomicReadHandle(&pCtxInt->hThreadWait, &hThread); + bool fWaiting = ASMAtomicReadBool(&pCtxInt->fWaiting); + if (fWaiting) + { + /* + * If a thread waits the handle must be valid. + * It is possible that the thread returns from + * aio_suspend() before the signal is send. + * This is no problem because we already set fWokenUp + * to true which will let the thread return VERR_INTERRUPTED + * and the next call to RTFileAioCtxWait() will not + * return VERR_INTERRUPTED because signals are not saved + * and will simply vanish if the destination thread can't + * receive it. + */ + Assert(hThread != NIL_RTTHREAD); + RTThreadPoke(hThread); + } +} + +/** + * Internal worker processing events and inserting new requests into the waiting list. + */ +static int rtFileAioCtxProcessEvents(PRTFILEAIOCTXINTERNAL pCtxInt) +{ + int rc = VINF_SUCCESS; + + /* Process new requests first. */ + bool fWokenUp = ASMAtomicXchgBool(&pCtxInt->fWokenUpInternal, false); + if (fWokenUp) + { + for (unsigned iSlot = 0; iSlot < RT_ELEMENTS(pCtxInt->apReqsNewHead); iSlot++) + { + PRTFILEAIOREQINTERNAL pReqHead = ASMAtomicXchgPtrT(&pCtxInt->apReqsNewHead[iSlot], NULL, PRTFILEAIOREQINTERNAL); + + while ( (pCtxInt->iFirstFree < pCtxInt->cReqsWaitMax) + && pReqHead) + { + RTFIELAIOREQ_ASSERT_STATE(pReqHead, SUBMITTED); + pCtxInt->apReqs[pCtxInt->iFirstFree] = pReqHead; + pReqHead->iWaitingList = pCtxInt->iFirstFree; + pReqHead = pReqHead->pNext; + + /* Clear pointer to next and previous element just for safety. */ + pCtxInt->apReqs[pCtxInt->iFirstFree]->pNext = NULL; + pCtxInt->apReqs[pCtxInt->iFirstFree]->pPrev = NULL; + pCtxInt->iFirstFree++; + + Assert( (pCtxInt->iFirstFree <= pCtxInt->cMaxRequests) + && (pCtxInt->iFirstFree <= pCtxInt->cReqsWaitMax)); + } + + /* Append the rest to the wait list. */ + if (pReqHead) + { + RTFIELAIOREQ_ASSERT_STATE(pReqHead, SUBMITTED); + if (!pCtxInt->pReqsWaitHead) + { + Assert(!pCtxInt->pReqsWaitTail); + pCtxInt->pReqsWaitHead = pReqHead; + pReqHead->pPrev = NULL; + } + else + { + AssertPtr(pCtxInt->pReqsWaitTail); + + pCtxInt->pReqsWaitTail->pNext = pReqHead; + pReqHead->pPrev = pCtxInt->pReqsWaitTail; + } + + /* Update tail. */ + while (pReqHead->pNext) + { + RTFIELAIOREQ_ASSERT_STATE(pReqHead->pNext, SUBMITTED); + pReqHead = pReqHead->pNext; + } + + pCtxInt->pReqsWaitTail = pReqHead; + pCtxInt->pReqsWaitTail->pNext = NULL; + } + } + + /* Check if a request needs to be canceled. */ + PRTFILEAIOREQINTERNAL pReqToCancel = ASMAtomicReadPtrT(&pCtxInt->pReqToCancel, PRTFILEAIOREQINTERNAL); + if (pReqToCancel) + { + /* The request can be in the array waiting for completion or still in the list because it is full. */ + if (pReqToCancel->iWaitingList != RTFILEAIOCTX_WAIT_ENTRY_INVALID) + { + /* Put it out of the waiting list. */ + pCtxInt->apReqs[pReqToCancel->iWaitingList] = pCtxInt->apReqs[--pCtxInt->iFirstFree]; + pCtxInt->apReqs[pReqToCancel->iWaitingList]->iWaitingList = pReqToCancel->iWaitingList; + } + else + { + /* Unlink from the waiting list. */ + PRTFILEAIOREQINTERNAL pPrev = pReqToCancel->pPrev; + PRTFILEAIOREQINTERNAL pNext = pReqToCancel->pNext; + + if (pNext) + pNext->pPrev = pPrev; + else + { + /* We canceled the tail. */ + pCtxInt->pReqsWaitTail = pPrev; + } + + if (pPrev) + pPrev->pNext = pNext; + else + { + /* We canceled the head. */ + pCtxInt->pReqsWaitHead = pNext; + } + } + + ASMAtomicDecS32(&pCtxInt->cRequests); + AssertMsg(pCtxInt->cRequests >= 0, ("Canceled request not which is not in this context\n")); + RTSemEventSignal(pCtxInt->SemEventCancel); + } + } + else + { + if (ASMAtomicXchgBool(&pCtxInt->fWokenUp, false)) + rc = VERR_INTERRUPTED; + } + + return rc; +} + +RTR3DECL(int) RTFileAioGetLimits(PRTFILEAIOLIMITS pAioLimits) +{ + int rcBSD = 0; + AssertPtrReturn(pAioLimits, VERR_INVALID_POINTER); + +#if defined(RT_OS_DARWIN) + int cReqsOutstandingMax = 0; + size_t cbParameter = sizeof(int); + + rcBSD = sysctlbyname("kern.aioprocmax", /* name */ + &cReqsOutstandingMax, /* Where to store the old value. */ + &cbParameter, /* Size of the memory pointed to. */ + NULL, /* Where the new value is located. */ + 0); /* Where the size of the new value is stored. */ + if (rcBSD == -1) + return RTErrConvertFromErrno(errno); + + pAioLimits->cReqsOutstandingMax = cReqsOutstandingMax; + pAioLimits->cbBufferAlignment = 0; +#elif defined(RT_OS_FREEBSD) + /* + * The AIO API is implemented in a kernel module which is not + * loaded by default. + * If it is loaded there are additional sysctl parameters. + */ + int cReqsOutstandingMax = 0; + size_t cbParameter = sizeof(int); + + rcBSD = sysctlbyname("vfs.aio.max_aio_per_proc", /* name */ + &cReqsOutstandingMax, /* Where to store the old value. */ + &cbParameter, /* Size of the memory pointed to. */ + NULL, /* Where the new value is located. */ + 0); /* Where the size of the new value is stored. */ + if (rcBSD == -1) + { + /* ENOENT means the value is unknown thus the module is not loaded. */ + if (errno == ENOENT) + return VERR_NOT_SUPPORTED; + else + return RTErrConvertFromErrno(errno); + } + + pAioLimits->cReqsOutstandingMax = cReqsOutstandingMax; + pAioLimits->cbBufferAlignment = 0; +#else + pAioLimits->cReqsOutstandingMax = RTFILEAIO_UNLIMITED_REQS; + pAioLimits->cbBufferAlignment = 0; +#endif + + return VINF_SUCCESS; +} + +RTR3DECL(int) RTFileAioReqCreate(PRTFILEAIOREQ phReq) +{ + AssertPtrReturn(phReq, VERR_INVALID_POINTER); + + PRTFILEAIOREQINTERNAL pReqInt = (PRTFILEAIOREQINTERNAL)RTMemAllocZ(sizeof(RTFILEAIOREQINTERNAL)); + if (RT_UNLIKELY(!pReqInt)) + return VERR_NO_MEMORY; + + pReqInt->pCtxInt = NULL; + pReqInt->u32Magic = RTFILEAIOREQ_MAGIC; + pReqInt->iWaitingList = RTFILEAIOCTX_WAIT_ENTRY_INVALID; + RTFILEAIOREQ_SET_STATE(pReqInt, COMPLETED); + + *phReq = (RTFILEAIOREQ)pReqInt; + + return VINF_SUCCESS; +} + + +RTDECL(int) RTFileAioReqDestroy(RTFILEAIOREQ hReq) +{ + /* + * Validate the handle and ignore nil. + */ + if (hReq == NIL_RTFILEAIOREQ) + return VINF_SUCCESS; + PRTFILEAIOREQINTERNAL pReqInt = hReq; + RTFILEAIOREQ_VALID_RETURN(pReqInt); + RTFILEAIOREQ_NOT_STATE_RETURN_RC(pReqInt, SUBMITTED, VERR_FILE_AIO_IN_PROGRESS); + + /* + * Trash the magic and free it. + */ + ASMAtomicUoWriteU32(&pReqInt->u32Magic, ~RTFILEAIOREQ_MAGIC); + RTMemFree(pReqInt); + return VINF_SUCCESS; +} + +/** + * Worker setting up the request. + */ +DECLINLINE(int) rtFileAioReqPrepareTransfer(RTFILEAIOREQ hReq, RTFILE hFile, + unsigned uTransferDirection, + RTFOFF off, void *pvBuf, size_t cbTransfer, + void *pvUser) +{ + /* + * Validate the input. + */ + PRTFILEAIOREQINTERNAL pReqInt = hReq; + RTFILEAIOREQ_VALID_RETURN(pReqInt); + RTFILEAIOREQ_NOT_STATE_RETURN_RC(pReqInt, SUBMITTED, VERR_FILE_AIO_IN_PROGRESS); + Assert(hFile != NIL_RTFILE); + AssertPtr(pvBuf); + Assert(off >= 0); + Assert(cbTransfer > 0); + + memset(&pReqInt->AioCB, 0, sizeof(struct aiocb)); + pReqInt->fFlush = false; + pReqInt->AioCB.aio_lio_opcode = uTransferDirection; + pReqInt->AioCB.aio_fildes = RTFileToNative(hFile); + pReqInt->AioCB.aio_offset = off; + pReqInt->AioCB.aio_nbytes = cbTransfer; + pReqInt->AioCB.aio_buf = pvBuf; + pReqInt->pvUser = pvUser; + pReqInt->pCtxInt = NULL; + pReqInt->Rc = VERR_FILE_AIO_IN_PROGRESS; + RTFILEAIOREQ_SET_STATE(pReqInt, PREPARED); + + return VINF_SUCCESS; +} + + +RTDECL(int) RTFileAioReqPrepareRead(RTFILEAIOREQ hReq, RTFILE hFile, RTFOFF off, + void *pvBuf, size_t cbRead, void *pvUser) +{ + return rtFileAioReqPrepareTransfer(hReq, hFile, LIO_READ, + off, pvBuf, cbRead, pvUser); +} + + +RTDECL(int) RTFileAioReqPrepareWrite(RTFILEAIOREQ hReq, RTFILE hFile, RTFOFF off, + void const *pvBuf, size_t cbWrite, void *pvUser) +{ + return rtFileAioReqPrepareTransfer(hReq, hFile, LIO_WRITE, + off, (void *)pvBuf, cbWrite, pvUser); +} + + +RTDECL(int) RTFileAioReqPrepareFlush(RTFILEAIOREQ hReq, RTFILE hFile, void *pvUser) +{ + PRTFILEAIOREQINTERNAL pReqInt = (PRTFILEAIOREQINTERNAL)hReq; + + RTFILEAIOREQ_VALID_RETURN(pReqInt); + RTFILEAIOREQ_NOT_STATE_RETURN_RC(pReqInt, SUBMITTED, VERR_FILE_AIO_IN_PROGRESS); + Assert(hFile != NIL_RTFILE); + + pReqInt->fFlush = true; + pReqInt->AioCB.aio_fildes = RTFileToNative(hFile); + pReqInt->AioCB.aio_offset = 0; + pReqInt->AioCB.aio_nbytes = 0; + pReqInt->AioCB.aio_buf = NULL; + pReqInt->pvUser = pvUser; + pReqInt->Rc = VERR_FILE_AIO_IN_PROGRESS; + RTFILEAIOREQ_SET_STATE(pReqInt, PREPARED); + + return VINF_SUCCESS; +} + + +RTDECL(void *) RTFileAioReqGetUser(RTFILEAIOREQ hReq) +{ + PRTFILEAIOREQINTERNAL pReqInt = hReq; + RTFILEAIOREQ_VALID_RETURN_RC(pReqInt, NULL); + + return pReqInt->pvUser; +} + + +RTDECL(int) RTFileAioReqCancel(RTFILEAIOREQ hReq) +{ + PRTFILEAIOREQINTERNAL pReqInt = hReq; + RTFILEAIOREQ_VALID_RETURN(pReqInt); + RTFILEAIOREQ_STATE_RETURN_RC(pReqInt, SUBMITTED, VERR_FILE_AIO_NOT_SUBMITTED); + + ASMAtomicXchgBool(&pReqInt->fCanceled, true); + + int rcPosix = aio_cancel(pReqInt->AioCB.aio_fildes, &pReqInt->AioCB); + + if (rcPosix == AIO_CANCELED) + { + PRTFILEAIOCTXINTERNAL pCtxInt = pReqInt->pCtxInt; + /* + * Notify the waiting thread that the request was canceled. + */ + AssertMsg(RT_VALID_PTR(pCtxInt), ("Invalid state. Request was canceled but wasn't submitted\n")); + + Assert(!pCtxInt->pReqToCancel); + ASMAtomicWritePtr(&pCtxInt->pReqToCancel, pReqInt); + rtFileAioCtxWakeup(pCtxInt); + + /* Wait for acknowledge. */ + int rc = RTSemEventWait(pCtxInt->SemEventCancel, RT_INDEFINITE_WAIT); + AssertRC(rc); + + ASMAtomicWriteNullPtr(&pCtxInt->pReqToCancel); + pReqInt->Rc = VERR_FILE_AIO_CANCELED; + RTFILEAIOREQ_SET_STATE(pReqInt, COMPLETED); + return VINF_SUCCESS; + } + else if (rcPosix == AIO_ALLDONE) + return VERR_FILE_AIO_COMPLETED; + else if (rcPosix == AIO_NOTCANCELED) + return VERR_FILE_AIO_IN_PROGRESS; + else + return RTErrConvertFromErrno(errno); +} + + +RTDECL(int) RTFileAioReqGetRC(RTFILEAIOREQ hReq, size_t *pcbTransfered) +{ + PRTFILEAIOREQINTERNAL pReqInt = hReq; + RTFILEAIOREQ_VALID_RETURN(pReqInt); + RTFILEAIOREQ_NOT_STATE_RETURN_RC(pReqInt, SUBMITTED, VERR_FILE_AIO_IN_PROGRESS); + RTFILEAIOREQ_NOT_STATE_RETURN_RC(pReqInt, PREPARED, VERR_FILE_AIO_NOT_SUBMITTED); + AssertPtrNull(pcbTransfered); + + if ( (RT_SUCCESS(pReqInt->Rc)) + && (pcbTransfered)) + *pcbTransfered = pReqInt->cbTransfered; + + return pReqInt->Rc; +} + + +RTDECL(int) RTFileAioCtxCreate(PRTFILEAIOCTX phAioCtx, uint32_t cAioReqsMax, + uint32_t fFlags) +{ + PRTFILEAIOCTXINTERNAL pCtxInt; + unsigned cReqsWaitMax; + + AssertPtrReturn(phAioCtx, VERR_INVALID_POINTER); + AssertReturn(!(fFlags & ~RTFILEAIOCTX_FLAGS_VALID_MASK), VERR_INVALID_PARAMETER); + + if (cAioReqsMax == RTFILEAIO_UNLIMITED_REQS) + return VERR_OUT_OF_RANGE; + + cReqsWaitMax = RT_MIN(cAioReqsMax, AIO_LISTIO_MAX); + + pCtxInt = (PRTFILEAIOCTXINTERNAL)RTMemAllocZ( sizeof(RTFILEAIOCTXINTERNAL) + + cReqsWaitMax * sizeof(PRTFILEAIOREQINTERNAL)); + if (RT_UNLIKELY(!pCtxInt)) + return VERR_NO_MEMORY; + + /* Create event semaphore. */ + int rc = RTSemEventCreate(&pCtxInt->SemEventCancel); + if (RT_FAILURE(rc)) + { + RTMemFree(pCtxInt); + return rc; + } + + pCtxInt->u32Magic = RTFILEAIOCTX_MAGIC; + pCtxInt->cMaxRequests = cAioReqsMax; + pCtxInt->cReqsWaitMax = cReqsWaitMax; + pCtxInt->fFlags = fFlags; + *phAioCtx = (RTFILEAIOCTX)pCtxInt; + + return VINF_SUCCESS; +} + + +RTDECL(int) RTFileAioCtxDestroy(RTFILEAIOCTX hAioCtx) +{ + PRTFILEAIOCTXINTERNAL pCtxInt = hAioCtx; + + AssertPtrReturn(pCtxInt, VERR_INVALID_HANDLE); + + if (RT_UNLIKELY(pCtxInt->cRequests)) + return VERR_FILE_AIO_BUSY; + + RTSemEventDestroy(pCtxInt->SemEventCancel); + RTMemFree(pCtxInt); + + return VINF_SUCCESS; +} + + +RTDECL(uint32_t) RTFileAioCtxGetMaxReqCount(RTFILEAIOCTX hAioCtx) +{ + PRTFILEAIOCTXINTERNAL pCtxInt = hAioCtx; + + if (hAioCtx == NIL_RTFILEAIOCTX) + return RTFILEAIO_UNLIMITED_REQS; + return pCtxInt->cMaxRequests; +} + +RTDECL(int) RTFileAioCtxAssociateWithFile(RTFILEAIOCTX hAioCtx, RTFILE hFile) +{ + NOREF(hAioCtx); NOREF(hFile); + return VINF_SUCCESS; +} + +#ifdef LOG_ENABLED +/** + * Dumps the state of a async I/O context. + */ +static void rtFileAioCtxDump(PRTFILEAIOCTXINTERNAL pCtxInt) +{ + LogFlow(("cRequests=%d\n", pCtxInt->cRequests)); + LogFlow(("cMaxRequests=%u\n", pCtxInt->cMaxRequests)); + LogFlow(("hThreadWait=%#p\n", pCtxInt->hThreadWait)); + LogFlow(("fWokenUp=%RTbool\n", pCtxInt->fWokenUp)); + LogFlow(("fWaiting=%RTbool\n", pCtxInt->fWaiting)); + LogFlow(("fWokenUpInternal=%RTbool\n", pCtxInt->fWokenUpInternal)); + for (unsigned i = 0; i < RT_ELEMENTS(pCtxInt->apReqsNewHead); i++) + LogFlow(("apReqsNewHead[%u]=%#p\n", i, pCtxInt->apReqsNewHead[i])); + LogFlow(("pReqToCancel=%#p\n", pCtxInt->pReqToCancel)); + LogFlow(("pReqsWaitHead=%#p\n", pCtxInt->pReqsWaitHead)); + LogFlow(("pReqsWaitTail=%#p\n", pCtxInt->pReqsWaitTail)); + LogFlow(("cReqsWaitMax=%u\n", pCtxInt->cReqsWaitMax)); + LogFlow(("iFirstFree=%u\n", pCtxInt->iFirstFree)); + for (unsigned i = 0; i < pCtxInt->cReqsWaitMax; i++) + LogFlow(("apReqs[%u]=%#p\n", i, pCtxInt->apReqs[i])); +} +#endif + +RTDECL(int) RTFileAioCtxSubmit(RTFILEAIOCTX hAioCtx, PRTFILEAIOREQ pahReqs, size_t cReqs) +{ + int rc = VINF_SUCCESS; + PRTFILEAIOCTXINTERNAL pCtxInt = hAioCtx; + + /* Parameter checks */ + AssertPtrReturn(pCtxInt, VERR_INVALID_HANDLE); + AssertReturn(cReqs != 0, VERR_INVALID_POINTER); + AssertPtrReturn(pahReqs, VERR_INVALID_PARAMETER); + + rtFileAioCtxDump(pCtxInt); + + /* Check that we don't exceed the limit */ + if (ASMAtomicUoReadS32(&pCtxInt->cRequests) + cReqs > pCtxInt->cMaxRequests) + return VERR_FILE_AIO_LIMIT_EXCEEDED; + + PRTFILEAIOREQINTERNAL pHead = NULL; + + do + { + int rcPosix = 0; + size_t cReqsSubmit = 0; + size_t i = 0; + PRTFILEAIOREQINTERNAL pReqInt; + + while ( (i < cReqs) + && (i < AIO_LISTIO_MAX)) + { + pReqInt = pahReqs[i]; + if (RTFILEAIOREQ_IS_NOT_VALID(pReqInt)) + { + /* Undo everything and stop submitting. */ + for (size_t iUndo = 0; iUndo < i; iUndo++) + { + pReqInt = pahReqs[iUndo]; + RTFILEAIOREQ_SET_STATE(pReqInt, PREPARED); + pReqInt->pCtxInt = NULL; + + /* Unlink from the list again. */ + PRTFILEAIOREQINTERNAL pNext, pPrev; + pNext = pReqInt->pNext; + pPrev = pReqInt->pPrev; + if (pNext) + pNext->pPrev = pPrev; + if (pPrev) + pPrev->pNext = pNext; + else + pHead = pNext; + } + rc = VERR_INVALID_HANDLE; + break; + } + + pReqInt->pCtxInt = pCtxInt; + + if (pReqInt->fFlush) + break; + + /* Link them together. */ + pReqInt->pNext = pHead; + if (pHead) + pHead->pPrev = pReqInt; + pReqInt->pPrev = NULL; + pHead = pReqInt; + RTFILEAIOREQ_SET_STATE(pReqInt, SUBMITTED); + + cReqsSubmit++; + i++; + } + + if (cReqsSubmit) + { + rcPosix = lio_listio(LIO_NOWAIT, (struct aiocb **)pahReqs, cReqsSubmit, NULL); + if (RT_UNLIKELY(rcPosix < 0)) + { + size_t cReqsSubmitted = cReqsSubmit; + + if (errno == EAGAIN) + rc = VERR_FILE_AIO_INSUFFICIENT_RESSOURCES; + else + rc = RTErrConvertFromErrno(errno); + + /* Check which ones were not submitted. */ + for (i = 0; i < cReqsSubmit; i++) + { + pReqInt = pahReqs[i]; + + rcPosix = aio_error(&pReqInt->AioCB); + + if ((rcPosix != EINPROGRESS) && (rcPosix != 0)) + { + cReqsSubmitted--; + +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) + if (errno == EINVAL) +#else + if (rcPosix == EINVAL) +#endif + { + /* Was not submitted. */ + RTFILEAIOREQ_SET_STATE(pReqInt, PREPARED); + } + else + { + /* An error occurred. */ + RTFILEAIOREQ_SET_STATE(pReqInt, COMPLETED); + + /* + * Looks like Apple and glibc interpret the standard in different ways. + * glibc returns the error code which would be in errno but Apple returns + * -1 and sets errno to the appropriate value + */ +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) + Assert(rcPosix == -1); + pReqInt->Rc = RTErrConvertFromErrno(errno); +#elif defined(RT_OS_LINUX) + pReqInt->Rc = RTErrConvertFromErrno(rcPosix); +#endif + pReqInt->cbTransfered = 0; + } + /* Unlink from the list. */ + PRTFILEAIOREQINTERNAL pNext, pPrev; + pNext = pReqInt->pNext; + pPrev = pReqInt->pPrev; + if (pNext) + pNext->pPrev = pPrev; + if (pPrev) + pPrev->pNext = pNext; + else + pHead = pNext; + + pReqInt->pNext = NULL; + pReqInt->pPrev = NULL; + } + } + ASMAtomicAddS32(&pCtxInt->cRequests, cReqsSubmitted); + AssertMsg(pCtxInt->cRequests >= 0, ("Adding requests resulted in overflow\n")); + break; + } + + ASMAtomicAddS32(&pCtxInt->cRequests, cReqsSubmit); + AssertMsg(pCtxInt->cRequests >= 0, ("Adding requests resulted in overflow\n")); + cReqs -= cReqsSubmit; + pahReqs += cReqsSubmit; + } + + /* + * Check if we have a flush request now. + * If not we hit the AIO_LISTIO_MAX limit + * and will continue submitting requests + * above. + */ + if (cReqs && RT_SUCCESS_NP(rc)) + { + pReqInt = pahReqs[0]; + + if (pReqInt->fFlush) + { + /* + * lio_listio does not work with flush requests so + * we have to use aio_fsync directly. + */ + rcPosix = aio_fsync(O_SYNC, &pReqInt->AioCB); + if (RT_UNLIKELY(rcPosix < 0)) + { + if (errno == EAGAIN) + { + rc = VERR_FILE_AIO_INSUFFICIENT_RESSOURCES; + RTFILEAIOREQ_SET_STATE(pReqInt, PREPARED); + } + else + { + rc = RTErrConvertFromErrno(errno); + RTFILEAIOREQ_SET_STATE(pReqInt, COMPLETED); + pReqInt->Rc = rc; + } + pReqInt->cbTransfered = 0; + break; + } + + /* Link them together. */ + pReqInt->pNext = pHead; + if (pHead) + pHead->pPrev = pReqInt; + pReqInt->pPrev = NULL; + pHead = pReqInt; + RTFILEAIOREQ_SET_STATE(pReqInt, SUBMITTED); + + ASMAtomicIncS32(&pCtxInt->cRequests); + AssertMsg(pCtxInt->cRequests >= 0, ("Adding requests resulted in overflow\n")); + cReqs--; + pahReqs++; + } + } + } while ( cReqs + && RT_SUCCESS_NP(rc)); + + if (pHead) + { + /* + * Forward successfully submitted requests to the thread waiting for requests. + * We search for a free slot first and if we don't find one + * we will grab the first one and append our list to the existing entries. + */ + unsigned iSlot = 0; + while ( (iSlot < RT_ELEMENTS(pCtxInt->apReqsNewHead)) + && !ASMAtomicCmpXchgPtr(&pCtxInt->apReqsNewHead[iSlot], pHead, NULL)) + iSlot++; + + if (iSlot == RT_ELEMENTS(pCtxInt->apReqsNewHead)) + { + /* Nothing found. */ + PRTFILEAIOREQINTERNAL pOldHead = ASMAtomicXchgPtrT(&pCtxInt->apReqsNewHead[0], NULL, PRTFILEAIOREQINTERNAL); + + /* Find the end of the current head and link the old list to the current. */ + PRTFILEAIOREQINTERNAL pTail = pHead; + while (pTail->pNext) + pTail = pTail->pNext; + + pTail->pNext = pOldHead; + + ASMAtomicWritePtr(&pCtxInt->apReqsNewHead[0], pHead); + } + + /* Set the internal wakeup flag and wakeup the thread if possible. */ + bool fWokenUp = ASMAtomicXchgBool(&pCtxInt->fWokenUpInternal, true); + if (!fWokenUp) + rtFileAioCtxWakeup(pCtxInt); + } + + rtFileAioCtxDump(pCtxInt); + + return rc; +} + + +RTDECL(int) RTFileAioCtxWait(RTFILEAIOCTX hAioCtx, size_t cMinReqs, RTMSINTERVAL cMillies, + PRTFILEAIOREQ pahReqs, size_t cReqs, uint32_t *pcReqs) +{ + int rc = VINF_SUCCESS; + int cRequestsCompleted = 0; + PRTFILEAIOCTXINTERNAL pCtxInt = (PRTFILEAIOCTXINTERNAL)hAioCtx; + struct timespec Timeout; + struct timespec *pTimeout = NULL; + uint64_t StartNanoTS = 0; + + LogFlowFunc(("hAioCtx=%#p cMinReqs=%zu cMillies=%u pahReqs=%#p cReqs=%zu pcbReqs=%#p\n", + hAioCtx, cMinReqs, cMillies, pahReqs, cReqs, pcReqs)); + + /* Check parameters. */ + AssertPtrReturn(pCtxInt, VERR_INVALID_HANDLE); + AssertPtrReturn(pcReqs, VERR_INVALID_POINTER); + AssertPtrReturn(pahReqs, VERR_INVALID_POINTER); + AssertReturn(cReqs != 0, VERR_INVALID_PARAMETER); + AssertReturn(cReqs >= cMinReqs, VERR_OUT_OF_RANGE); + + rtFileAioCtxDump(pCtxInt); + + int32_t cRequestsWaiting = ASMAtomicReadS32(&pCtxInt->cRequests); + + if ( RT_UNLIKELY(cRequestsWaiting <= 0) + && !(pCtxInt->fFlags & RTFILEAIOCTX_FLAGS_WAIT_WITHOUT_PENDING_REQUESTS)) + return VERR_FILE_AIO_NO_REQUEST; + + if (RT_UNLIKELY(cMinReqs > (uint32_t)cRequestsWaiting)) + return VERR_INVALID_PARAMETER; + + if (cMillies != RT_INDEFINITE_WAIT) + { + Timeout.tv_sec = cMillies / 1000; + Timeout.tv_nsec = (cMillies % 1000) * 1000000; + pTimeout = &Timeout; + StartNanoTS = RTTimeNanoTS(); + } + + /* Wait for at least one. */ + if (!cMinReqs) + cMinReqs = 1; + + /* For the wakeup call. */ + Assert(pCtxInt->hThreadWait == NIL_RTTHREAD); + ASMAtomicWriteHandle(&pCtxInt->hThreadWait, RTThreadSelf()); + + /* Update the waiting list once before we enter the loop. */ + rc = rtFileAioCtxProcessEvents(pCtxInt); + + while ( cMinReqs + && RT_SUCCESS_NP(rc)) + { +#ifdef RT_STRICT + if (RT_UNLIKELY(!pCtxInt->iFirstFree)) + { + for (unsigned i = 0; i < pCtxInt->cReqsWaitMax; i++) + RTAssertMsg2Weak("wait[%d] = %#p\n", i, pCtxInt->apReqs[i]); + + AssertMsgFailed(("No request to wait for. pReqsWaitHead=%#p pReqsWaitTail=%#p\n", + pCtxInt->pReqsWaitHead, pCtxInt->pReqsWaitTail)); + } +#endif + + LogFlow(("Waiting for %d requests to complete\n", pCtxInt->iFirstFree)); + rtFileAioCtxDump(pCtxInt); + + ASMAtomicXchgBool(&pCtxInt->fWaiting, true); + int rcPosix = aio_suspend((const struct aiocb * const *)pCtxInt->apReqs, + pCtxInt->iFirstFree, pTimeout); + ASMAtomicXchgBool(&pCtxInt->fWaiting, false); + if (rcPosix < 0) + { + LogFlow(("aio_suspend failed %d nent=%u\n", errno, pCtxInt->iFirstFree)); + /* Check that this is an external wakeup event. */ + if (errno == EINTR) + rc = rtFileAioCtxProcessEvents(pCtxInt); + else + rc = RTErrConvertFromErrno(errno); + } + else + { + /* Requests finished. */ + unsigned iReqCurr = 0; + unsigned cDone = 0; + + /* Remove completed requests from the waiting list. */ + while ( (iReqCurr < pCtxInt->iFirstFree) + && (cDone < cReqs)) + { + PRTFILEAIOREQINTERNAL pReq = pCtxInt->apReqs[iReqCurr]; + int rcReq = aio_error(&pReq->AioCB); + + if (rcReq != EINPROGRESS) + { + /* Completed store the return code. */ + if (rcReq == 0) + { + pReq->Rc = VINF_SUCCESS; + /* Call aio_return() to free resources. */ + pReq->cbTransfered = aio_return(&pReq->AioCB); + } + else + { +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) + pReq->Rc = RTErrConvertFromErrno(errno); +#else + pReq->Rc = RTErrConvertFromErrno(rcReq); +#endif + } + + /* Mark the request as finished. */ + RTFILEAIOREQ_SET_STATE(pReq, COMPLETED); + cDone++; + + /* If there are other entries waiting put the head into the now free entry. */ + if (pCtxInt->pReqsWaitHead) + { + PRTFILEAIOREQINTERNAL pReqInsert = pCtxInt->pReqsWaitHead; + + pCtxInt->pReqsWaitHead = pReqInsert->pNext; + if (!pCtxInt->pReqsWaitHead) + { + /* List is empty now. Clear tail too. */ + pCtxInt->pReqsWaitTail = NULL; + } + + pReqInsert->iWaitingList = pReq->iWaitingList; + pCtxInt->apReqs[pReqInsert->iWaitingList] = pReqInsert; + iReqCurr++; + } + else + { + /* + * Move the last entry into the current position to avoid holes + * but only if it is not the last element already. + */ + if (pReq->iWaitingList < pCtxInt->iFirstFree - 1) + { + pCtxInt->apReqs[pReq->iWaitingList] = pCtxInt->apReqs[--pCtxInt->iFirstFree]; + pCtxInt->apReqs[pReq->iWaitingList]->iWaitingList = pReq->iWaitingList; + } + else + pCtxInt->iFirstFree--; + + pCtxInt->apReqs[pCtxInt->iFirstFree] = NULL; + } + + /* Put the request into the completed list. */ + pahReqs[cRequestsCompleted++] = pReq; + pReq->iWaitingList = RTFILEAIOCTX_WAIT_ENTRY_INVALID; + } + else + iReqCurr++; + } + + AssertMsg((cDone <= cReqs), ("Overflow cReqs=%u cMinReqs=%u cDone=%u\n", + cReqs, cDone)); + cReqs -= cDone; + cMinReqs = RT_MAX(cMinReqs, cDone) - cDone; + ASMAtomicSubS32(&pCtxInt->cRequests, cDone); + + AssertMsg(pCtxInt->cRequests >= 0, ("Finished more requests than currently active\n")); + + if (!cMinReqs) + break; + + if (cMillies != RT_INDEFINITE_WAIT) + { + uint64_t TimeDiff; + + /* Recalculate the timeout. */ + TimeDiff = RTTimeSystemNanoTS() - StartNanoTS; + Timeout.tv_sec = Timeout.tv_sec - (TimeDiff / 1000000); + Timeout.tv_nsec = Timeout.tv_nsec - (TimeDiff % 1000000); + } + + /* Check for new elements. */ + rc = rtFileAioCtxProcessEvents(pCtxInt); + } + } + + *pcReqs = cRequestsCompleted; + Assert(pCtxInt->hThreadWait == RTThreadSelf()); + ASMAtomicWriteHandle(&pCtxInt->hThreadWait, NIL_RTTHREAD); + + rtFileAioCtxDump(pCtxInt); + + return rc; +} + + +RTDECL(int) RTFileAioCtxWakeup(RTFILEAIOCTX hAioCtx) +{ + PRTFILEAIOCTXINTERNAL pCtxInt = hAioCtx; + RTFILEAIOCTX_VALID_RETURN(pCtxInt); + + /** @todo r=bird: Define the protocol for how to resume work after calling + * this function. */ + + bool fWokenUp = ASMAtomicXchgBool(&pCtxInt->fWokenUp, true); + if (!fWokenUp) + rtFileAioCtxWakeup(pCtxInt); + + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/fileio-at-posix.cpp b/src/VBox/Runtime/r3/posix/fileio-at-posix.cpp new file mode 100644 index 00000000..72d3cc05 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fileio-at-posix.cpp @@ -0,0 +1,107 @@ +/* $Id: fileio-at-posix.cpp $ */ +/** @file + * IPRT - File I/O, RTFileReadAt and RTFileWriteAt, posix. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <errno.h> +#include <sys/types.h> +#include <unistd.h> + +#include "internal/iprt.h" +#include <iprt/file.h> + +#include <iprt/err.h> +#include <iprt/log.h> + + + +RTDECL(int) RTFileReadAt(RTFILE hFile, RTFOFF off, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + ssize_t cbRead = pread(RTFileToNative(hFile), pvBuf, cbToRead, off); + if (cbRead >= 0) + { + if (pcbRead) + /* caller can handle partial read. */ + *pcbRead = cbRead; + else + { + /* Caller expects all to be read. */ + while ((ssize_t)cbToRead > cbRead) + { + ssize_t cbReadPart = pread(RTFileToNative(hFile), (char*)pvBuf + cbRead, cbToRead - cbRead, off + cbRead); + if (cbReadPart <= 0) + { + if (cbReadPart == 0) + return VERR_EOF; + return RTErrConvertFromErrno(errno); + } + cbRead += cbReadPart; + } + } + return VINF_SUCCESS; + } + + return RTErrConvertFromErrno(errno); +} + + +RTDECL(int) RTFileWriteAt(RTFILE hFile, RTFOFF off, const void *pvBuf, size_t cbToWrite, size_t *pcbWritten) +{ + ssize_t cbWritten = pwrite(RTFileToNative(hFile), pvBuf, cbToWrite, off); + if (cbWritten >= 0) + { + if (pcbWritten) + /* caller can handle partial write. */ + *pcbWritten = cbWritten; + else + { + /* Caller expects all to be write. */ + while ((ssize_t)cbToWrite > cbWritten) + { + ssize_t cbWrittenPart = pwrite(RTFileToNative(hFile), (const char *)pvBuf + cbWritten, cbToWrite - cbWritten, + off + cbWritten); + if (cbWrittenPart < 0) + return cbWrittenPart < 0 ? RTErrConvertFromErrno(errno) : VERR_TRY_AGAIN; + cbWritten += cbWrittenPart; + } + } + return VINF_SUCCESS; + } + return RTErrConvertFromErrno(errno); +} + diff --git a/src/VBox/Runtime/r3/posix/fileio-posix.cpp b/src/VBox/Runtime/r3/posix/fileio-posix.cpp new file mode 100644 index 00000000..cd7d238f --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fileio-posix.cpp @@ -0,0 +1,934 @@ +/* $Id: fileio-posix.cpp $ */ +/** @file + * IPRT - File I/O, POSIX, Part 1. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_FILE + +#include <errno.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <sys/ioctl.h> +#include <fcntl.h> +#ifdef _MSC_VER +# include <io.h> +# include <stdio.h> +#else +# include <unistd.h> +# include <sys/time.h> +#endif +#ifdef RT_OS_LINUX +# include <sys/file.h> +#endif +#if defined(RT_OS_OS2) && (!defined(__INNOTEK_LIBC__) || __INNOTEK_LIBC__ < 0x006) +# include <io.h> +#endif +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) +# include <sys/disk.h> +#endif +#ifdef RT_OS_SOLARIS +# include <stropts.h> +# include <sys/dkio.h> +# include <sys/vtoc.h> +#endif /* RT_OS_SOLARIS */ + +#include <iprt/file.h> +#include <iprt/path.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include <iprt/thread.h> +#include "internal/file.h" +#include "internal/fs.h" +#include "internal/path.h" + + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +/** Default file permissions for newly created files. */ +#if defined(S_IRUSR) && defined(S_IWUSR) +# define RT_FILE_PERMISSION (S_IRUSR | S_IWUSR) +#else +# define RT_FILE_PERMISSION (00600) +#endif + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +#ifdef O_CLOEXEC +static int volatile g_fHave_O_CLOEXEC = 0; /* {-1,0,1}; since Linux 2.6.23 */ +#endif + + + +RTDECL(bool) RTFileExists(const char *pszPath) +{ + bool fRc = false; + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat s; + fRc = !stat(pszNativePath, &s) + && S_ISREG(s.st_mode); + + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTFileExists(%p={%s}): returns %RTbool\n", pszPath, pszPath, fRc)); + return fRc; +} + + +#ifdef O_CLOEXEC +/** Worker for RTFileOpenEx that detects whether the kernel supports + * O_CLOEXEC or not, setting g_fHave_O_CLOEXEC to 1 or -1 accordingly. */ +static int rtFileOpenExDetectCloExecSupport(void) +{ + /* + * Open /dev/null with O_CLOEXEC and see if FD_CLOEXEC is set or not. + */ + int fHave_O_CLOEXEC = -1; + int fd = open("/dev/null", O_RDONLY | O_CLOEXEC, 0); + if (fd >= 0) + { + int fFlags = fcntl(fd, F_GETFD, 0); + fHave_O_CLOEXEC = fFlags > 0 && (fFlags & FD_CLOEXEC) ? 1 : -1; + close(fd); + } + else + AssertMsg(errno == EINVAL, ("%d\n", errno)); + g_fHave_O_CLOEXEC = fHave_O_CLOEXEC; + return fHave_O_CLOEXEC; +} +#endif + + +RTR3DECL(int) RTFileOpen(PRTFILE pFile, const char *pszFilename, uint64_t fOpen) +{ + return RTFileOpenEx(pszFilename, fOpen, pFile, NULL); +} + + +RTDECL(int) RTFileOpenEx(const char *pszFilename, uint64_t fOpen, PRTFILE phFile, PRTFILEACTION penmActionTaken) +{ + /* + * Validate input. + */ + AssertPtrReturn(phFile, VERR_INVALID_POINTER); + *phFile = NIL_RTFILE; + if (penmActionTaken) + *penmActionTaken = RTFILEACTION_INVALID; + AssertPtrReturn(pszFilename, VERR_INVALID_POINTER); + + /* + * Merge forced open flags and validate them. + */ + int rc = rtFileRecalcAndValidateFlags(&fOpen); + if (RT_FAILURE(rc)) + return rc; +#ifndef O_NONBLOCK + AssertReturn(!(fOpen & RTFILE_O_NON_BLOCK), VERR_INVALID_FLAGS); +#endif +#if defined(RT_OS_OS2) /* Cannot delete open files on OS/2. */ + AssertReturn(!(fOpen & RTFILE_O_TEMP_AUTO_DELETE), VERR_NOT_SUPPORTED); +#endif + + /* + * Calculate open mode flags. + */ + int fOpenMode = 0; +#ifdef O_BINARY + fOpenMode |= O_BINARY; /* (pc) */ +#endif +#ifdef O_LARGEFILE + fOpenMode |= O_LARGEFILE; /* (linux, solaris) */ +#endif +#ifdef O_NOINHERIT + if (!(fOpen & RTFILE_O_INHERIT)) + fOpenMode |= O_NOINHERIT; +#endif +#ifdef O_CLOEXEC + int fHave_O_CLOEXEC = g_fHave_O_CLOEXEC; + if ( !(fOpen & RTFILE_O_INHERIT) + && ( fHave_O_CLOEXEC > 0 + || ( fHave_O_CLOEXEC == 0 + && (fHave_O_CLOEXEC = rtFileOpenExDetectCloExecSupport()) > 0))) + fOpenMode |= O_CLOEXEC; +#endif +#ifdef O_NONBLOCK + if (fOpen & RTFILE_O_NON_BLOCK) + fOpenMode |= O_NONBLOCK; +#endif +#ifdef O_SYNC + if (fOpen & RTFILE_O_WRITE_THROUGH) + fOpenMode |= O_SYNC; +#endif +#if defined(O_DIRECT) && defined(RT_OS_LINUX) + /* O_DIRECT is mandatory to get async I/O working on Linux. */ + if (fOpen & RTFILE_O_ASYNC_IO) + fOpenMode |= O_DIRECT; +#endif +#if defined(O_DIRECT) && (defined(RT_OS_LINUX) || defined(RT_OS_FREEBSD) || defined(RT_OS_NETBSD)) + /* Disable the kernel cache. */ + if (fOpen & RTFILE_O_NO_CACHE) + fOpenMode |= O_DIRECT; +#endif + + /* create/truncate file */ + switch (fOpen & RTFILE_O_ACTION_MASK) + { + case RTFILE_O_OPEN: break; + case RTFILE_O_OPEN_CREATE: fOpenMode |= O_CREAT; break; + case RTFILE_O_CREATE: fOpenMode |= O_CREAT | O_EXCL; break; + case RTFILE_O_CREATE_REPLACE: fOpenMode |= O_CREAT | O_TRUNC; break; /** @todo replacing needs fixing, this is *not* a 1:1 mapping! */ + default: + AssertMsgFailed(("fOpen=%#llx\n", fOpen)); + fOpen = (fOpen & ~RTFILE_O_ACTION_MASK) | RTFILE_O_OPEN; + break; + + } + if ( (fOpen & RTFILE_O_TRUNCATE) + && (fOpen & RTFILE_O_ACTION_MASK) != RTFILE_O_CREATE) + fOpenMode |= O_TRUNC; + + switch (fOpen & RTFILE_O_ACCESS_MASK) + { + case RTFILE_O_READ: + fOpenMode |= O_RDONLY; /* RTFILE_O_APPEND is ignored. */ + break; + case RTFILE_O_WRITE: + fOpenMode |= fOpen & RTFILE_O_APPEND ? O_APPEND | O_WRONLY : O_WRONLY; + break; + case RTFILE_O_READWRITE: + fOpenMode |= fOpen & RTFILE_O_APPEND ? O_APPEND | O_RDWR : O_RDWR; + break; + default: + AssertMsgFailedReturn(("RTFileOpen received an invalid RW value, fOpen=%#llx\n", fOpen), VERR_INVALID_FLAGS); + } + + /* File mode. */ + int fMode = (fOpen & RTFILE_O_CREATE_MODE_MASK) + ? (fOpen & RTFILE_O_CREATE_MODE_MASK) >> RTFILE_O_CREATE_MODE_SHIFT + : RT_FILE_PERMISSION; + + /** @todo sharing? */ + + /* + * Open/create the file. + */ + char const *pszNativeFilename; + rc = rtPathToNative(&pszNativeFilename, pszFilename, NULL); + if (RT_FAILURE(rc)) + return (rc); + + int fh; + int iErr; + if (!penmActionTaken) + { + fh = open(pszNativeFilename, fOpenMode, fMode); + iErr = errno; + } + else + { + /* We need to know exactly which action was taken by open, Windows & + OS/2 style. Can be tedious and subject to races: */ + switch (fOpen & RTFILE_O_ACTION_MASK) + { + case RTFILE_O_OPEN: + Assert(!(fOpenMode & O_CREAT)); + Assert(!(fOpenMode & O_EXCL)); + fh = open(pszNativeFilename, fOpenMode, fMode); + iErr = errno; + if (fh >= 0) + *penmActionTaken = fOpenMode & O_TRUNC ? RTFILEACTION_TRUNCATED : RTFILEACTION_OPENED; + break; + + case RTFILE_O_CREATE: + Assert(fOpenMode & O_CREAT); + Assert(fOpenMode & O_EXCL); + fh = open(pszNativeFilename, fOpenMode, fMode); + iErr = errno; + if (fh >= 0) + *penmActionTaken = RTFILEACTION_CREATED; + else if (iErr == EEXIST) + *penmActionTaken = RTFILEACTION_ALREADY_EXISTS; + break; + + case RTFILE_O_OPEN_CREATE: + case RTFILE_O_CREATE_REPLACE: + { + Assert(fOpenMode & O_CREAT); + Assert(!(fOpenMode & O_EXCL)); + int iTries = 64; + while (iTries-- > 0) + { + /* Yield the CPU if we've raced too long. */ + if (iTries < 4) + RTThreadSleep(2 - (iTries & 1)); + + /* Try exclusive creation first: */ + fh = open(pszNativeFilename, fOpenMode | O_EXCL, fMode); + iErr = errno; + if (fh >= 0) + { + *penmActionTaken = RTFILEACTION_CREATED; + break; + } + if (iErr != EEXIST) + break; + + /* If the file exists, try open it: */ + fh = open(pszNativeFilename, fOpenMode & ~O_CREAT, fMode); + iErr = errno; + if (fh >= 0) + { + if ((fOpen & RTFILE_O_ACTION_MASK) == RTFILE_O_OPEN_CREATE) + *penmActionTaken = fOpenMode & O_TRUNC ? RTFILEACTION_TRUNCATED : RTFILEACTION_OPENED; + else + *penmActionTaken = RTFILEACTION_REPLACED; + break; + } + if (iErr != ENOENT) + break; + } + Assert(iTries >= 0); + if (iTries < 0) + { + /* Thanks for the race, but we need to get on with things. */ + fh = open(pszNativeFilename, fOpenMode, fMode); + iErr = errno; + if (fh >= 0) + *penmActionTaken = RTFILEACTION_OPENED; + } + break; + } + + default: + AssertMsgFailed(("fOpen=%#llx fOpenMode=%#x\n", fOpen, fOpenMode)); + iErr = EINVAL; + fh = -1; + break; + } + } + + rtPathFreeNative(pszNativeFilename, pszFilename); + if (fh >= 0) + { + iErr = 0; + + /* + * If temporary file, delete it. + */ + if (fOpen & RTFILE_O_TEMP_AUTO_DELETE) + { + /** @todo Use funlinkat/funlink or similar here when available! Or better, + * use O_TMPFILE, only that may require fallback as not supported by + * all file system on linux. */ + iErr = unlink(pszNativeFilename); + Assert(iErr == 0); + } + + /* + * Mark the file handle close on exec, unless inherit is specified. + */ + if ( !(fOpen & RTFILE_O_INHERIT) +#ifdef O_NOINHERIT + && !(fOpenMode & O_NOINHERIT) /* Take care since it might be a zero value dummy. */ +#endif +#ifdef O_CLOEXEC + && fHave_O_CLOEXEC <= 0 +#endif + ) + iErr = fcntl(fh, F_SETFD, FD_CLOEXEC) >= 0 ? 0 : errno; + + /* + * Switch direct I/O on now if requested and required. + */ +#if defined(RT_OS_DARWIN) \ + || (defined(RT_OS_SOLARIS) && !defined(IN_GUEST)) + if (iErr == 0 && (fOpen & RTFILE_O_NO_CACHE)) + { +# if defined(RT_OS_DARWIN) + iErr = fcntl(fh, F_NOCACHE, 1) >= 0 ? 0 : errno; +# else + iErr = directio(fh, DIRECTIO_ON) >= 0 ? 0 : errno; +# endif + } +#endif + + /* + * Implement / emulate file sharing. + * + * We need another mode which allows skipping this stuff completely + * and do things the UNIX way. So for the present this is just a debug + * aid that can be enabled by developers too lazy to test on Windows. + */ +#if 0 && defined(RT_OS_LINUX) + if (iErr == 0) + { + /* This approach doesn't work because only knfsd checks for these + buggers. :-( */ + int iLockOp; + switch (fOpen & RTFILE_O_DENY_MASK) + { + default: + AssertFailed(); + case RTFILE_O_DENY_NONE: + case RTFILE_O_DENY_NOT_DELETE: + iLockOp = LOCK_MAND | LOCK_READ | LOCK_WRITE; + break; + case RTFILE_O_DENY_READ: + case RTFILE_O_DENY_READ | RTFILE_O_DENY_NOT_DELETE: + iLockOp = LOCK_MAND | LOCK_WRITE; + break; + case RTFILE_O_DENY_WRITE: + case RTFILE_O_DENY_WRITE | RTFILE_O_DENY_NOT_DELETE: + iLockOp = LOCK_MAND | LOCK_READ; + break; + case RTFILE_O_DENY_WRITE | RTFILE_O_DENY_READ: + case RTFILE_O_DENY_WRITE | RTFILE_O_DENY_READ | RTFILE_O_DENY_NOT_DELETE: + iLockOp = LOCK_MAND; + break; + } + iErr = flock(fh, iLockOp | LOCK_NB); + if (iErr != 0) + iErr = errno == EAGAIN ? ETXTBSY : 0; + } +#endif /* 0 && RT_OS_LINUX */ +#if defined(DEBUG_bird) && !defined(RT_OS_SOLARIS) + if (iErr == 0) + { + /* This emulation is incomplete but useful. */ + switch (fOpen & RTFILE_O_DENY_MASK) + { + default: + AssertFailed(); + case RTFILE_O_DENY_NONE: + case RTFILE_O_DENY_NOT_DELETE: + case RTFILE_O_DENY_READ: + case RTFILE_O_DENY_READ | RTFILE_O_DENY_NOT_DELETE: + break; + case RTFILE_O_DENY_WRITE: + case RTFILE_O_DENY_WRITE | RTFILE_O_DENY_NOT_DELETE: + case RTFILE_O_DENY_WRITE | RTFILE_O_DENY_READ: + case RTFILE_O_DENY_WRITE | RTFILE_O_DENY_READ | RTFILE_O_DENY_NOT_DELETE: + if (fOpen & RTFILE_O_WRITE) + { + iErr = flock(fh, LOCK_EX | LOCK_NB); + if (iErr != 0) + iErr = errno == EAGAIN ? ETXTBSY : 0; + } + break; + } + } +#endif +#ifdef RT_OS_SOLARIS + /** @todo Use fshare_t and associates, it's a perfect match. see sys/fcntl.h */ +#endif + + /* + * We're done. + */ + if (iErr == 0) + { + *phFile = (RTFILE)(uintptr_t)fh; + Assert((intptr_t)*phFile == fh); + LogFlow(("RTFileOpen(%p:{%RTfile}, %p:{%s}, %#llx): returns %Rrc\n", + phFile, *phFile, pszFilename, pszFilename, fOpen, rc)); + return VINF_SUCCESS; + } + + close(fh); + } + return RTErrConvertFromErrno(iErr); +} + + +RTR3DECL(int) RTFileOpenBitBucket(PRTFILE phFile, uint64_t fAccess) +{ + AssertReturn( fAccess == RTFILE_O_READ + || fAccess == RTFILE_O_WRITE + || fAccess == RTFILE_O_READWRITE, + VERR_INVALID_PARAMETER); + return RTFileOpen(phFile, "/dev/null", fAccess | RTFILE_O_DENY_NONE | RTFILE_O_OPEN); +} + + +RTR3DECL(int) RTFileClose(RTFILE hFile) +{ + if (hFile == NIL_RTFILE) + return VINF_SUCCESS; + if (close(RTFileToNative(hFile)) == 0) + return VINF_SUCCESS; + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileFromNative(PRTFILE pFile, RTHCINTPTR uNative) +{ + AssertCompile(sizeof(uNative) == sizeof(*pFile)); + if (uNative < 0) + { + AssertMsgFailed(("%p\n", uNative)); + *pFile = NIL_RTFILE; + return VERR_INVALID_HANDLE; + } + *pFile = (RTFILE)uNative; + return VINF_SUCCESS; +} + + +RTR3DECL(RTHCINTPTR) RTFileToNative(RTFILE hFile) +{ + AssertReturn(hFile != NIL_RTFILE, -1); + return (intptr_t)hFile; +} + + +RTFILE rtFileGetStandard(RTHANDLESTD enmStdHandle) +{ + int fd; + switch (enmStdHandle) + { + case RTHANDLESTD_INPUT: fd = 0; break; + case RTHANDLESTD_OUTPUT: fd = 1; break; + case RTHANDLESTD_ERROR: fd = 2; break; + default: + AssertFailedReturn(NIL_RTFILE); + } + + struct stat st; + int rc = fstat(fd, &st); + if (rc == -1) + return NIL_RTFILE; + return (RTFILE)(intptr_t)fd; +} + + +RTR3DECL(int) RTFileDelete(const char *pszFilename) +{ + char const *pszNativeFilename; + int rc = rtPathToNative(&pszNativeFilename, pszFilename, NULL); + if (RT_SUCCESS(rc)) + { + if (unlink(pszNativeFilename) != 0) + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativeFilename, pszFilename); + } + return rc; +} + + +RTR3DECL(int) RTFileSeek(RTFILE hFile, int64_t offSeek, unsigned uMethod, uint64_t *poffActual) +{ + static const unsigned aSeekRecode[] = + { + SEEK_SET, + SEEK_CUR, + SEEK_END, + }; + + /* + * Validate input. + */ + if (uMethod > RTFILE_SEEK_END) + { + AssertMsgFailed(("Invalid uMethod=%d\n", uMethod)); + return VERR_INVALID_PARAMETER; + } + + /* check that within off_t range. */ + if ( sizeof(off_t) < sizeof(offSeek) + && ( (offSeek > 0 && (unsigned)(offSeek >> 32) != 0) + || (offSeek < 0 && (unsigned)(-offSeek >> 32) != 0))) + { + AssertMsgFailed(("64-bit search not supported\n")); + return VERR_NOT_SUPPORTED; + } + + off_t offCurrent = lseek(RTFileToNative(hFile), (off_t)offSeek, aSeekRecode[uMethod]); + if (offCurrent != ~0) + { + if (poffActual) + *poffActual = (uint64_t)offCurrent; + return VINF_SUCCESS; + } + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileRead(RTFILE hFile, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + if (cbToRead <= 0) + { + if (pcbRead) + *pcbRead = 0; + return VINF_SUCCESS; + } + + /* + * Attempt read. + */ + ssize_t cbRead = read(RTFileToNative(hFile), pvBuf, cbToRead); + if (cbRead >= 0) + { + if (pcbRead) + /* caller can handle partial read. */ + *pcbRead = cbRead; + else + { + /* Caller expects all to be read. */ + while ((ssize_t)cbToRead > cbRead) + { + ssize_t cbReadPart = read(RTFileToNative(hFile), (char*)pvBuf + cbRead, cbToRead - cbRead); + if (cbReadPart <= 0) + { + if (cbReadPart == 0) + return VERR_EOF; + return RTErrConvertFromErrno(errno); + } + cbRead += cbReadPart; + } + } + return VINF_SUCCESS; + } + + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileWrite(RTFILE hFile, const void *pvBuf, size_t cbToWrite, size_t *pcbWritten) +{ + if (cbToWrite <= 0) + return VINF_SUCCESS; + + /* + * Attempt write. + */ + ssize_t cbWritten = write(RTFileToNative(hFile), pvBuf, cbToWrite); + if (cbWritten >= 0) + { + if (pcbWritten) + /* caller can handle partial write. */ + *pcbWritten = cbWritten; + else + { + /* Caller expects all to be write. */ + while ((ssize_t)cbToWrite > cbWritten) + { + ssize_t cbWrittenPart = write(RTFileToNative(hFile), (const char *)pvBuf + cbWritten, cbToWrite - cbWritten); + if (cbWrittenPart <= 0) + return cbWrittenPart < 0 ? RTErrConvertFromErrno(errno) : VERR_TRY_AGAIN; + cbWritten += cbWrittenPart; + } + } + return VINF_SUCCESS; + } + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileSetSize(RTFILE hFile, uint64_t cbSize) +{ + /* + * Validate offset. + */ + if ( sizeof(off_t) < sizeof(cbSize) + && (cbSize >> 32) != 0) + { + AssertMsgFailed(("64-bit filesize not supported! cbSize=%lld\n", cbSize)); + return VERR_NOT_SUPPORTED; + } + +#if defined(_MSC_VER) || (defined(RT_OS_OS2) && (!defined(__INNOTEK_LIBC__) || __INNOTEK_LIBC__ < 0x006)) + if (chsize(RTFileToNative(hFile), (off_t)cbSize) == 0) +#else + /* This relies on a non-standard feature of FreeBSD, Linux, and OS/2 + * LIBC v0.6 and higher. (SuS doesn't define ftruncate() and size bigger + * than the file.) + */ + if (ftruncate(RTFileToNative(hFile), (off_t)cbSize) == 0) +#endif + return VINF_SUCCESS; + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileQuerySize(RTFILE hFile, uint64_t *pcbSize) +{ + /* + * Ask fstat() first. + */ + struct stat st; + if (!fstat(RTFileToNative(hFile), &st)) + { + *pcbSize = st.st_size; + if ( st.st_size != 0 +#if defined(RT_OS_SOLARIS) || defined(RT_OS_DARWIN) + || (!S_ISBLK(st.st_mode) && !S_ISCHR(st.st_mode)) +#elif defined(RT_OS_FREEBSD) || defined(RT_OS_NETBSD) || defined(RT_OS_DARWIN) + || !S_ISCHR(st.st_mode) +#else + || !S_ISBLK(st.st_mode) +#endif + ) + return VINF_SUCCESS; + + /* + * It could be a block device. Try determin the size by I/O control + * query or seek. + */ +#ifdef RT_OS_DARWIN + uint64_t cBlocks; + if (!ioctl(RTFileToNative(hFile), DKIOCGETBLOCKCOUNT, &cBlocks)) + { + uint32_t cbBlock; + if (!ioctl(RTFileToNative(hFile), DKIOCGETBLOCKSIZE, &cbBlock)) + { + *pcbSize = cBlocks * cbBlock; + return VINF_SUCCESS; + } + } + + /* Always fail block devices. Character devices doesn't all need to be + /dev/rdisk* nodes, they should return ENOTTY but /dev/null returns ENODEV + and we include EINVAL just in case. */ + if (!S_ISBLK(st.st_mode) && (errno == ENOTTY || errno == ENODEV || errno == EINVAL)) + return VINF_SUCCESS; + +#elif defined(RT_OS_SOLARIS) + struct dk_minfo MediaInfo; + if (!ioctl(RTFileToNative(hFile), DKIOCGMEDIAINFO, &MediaInfo)) + { + *pcbSize = MediaInfo.dki_capacity * MediaInfo.dki_lbsize; + return VINF_SUCCESS; + } + /* might not be a block device. */ + if (errno == EINVAL || errno == ENOTTY) + return VINF_SUCCESS; + +#elif defined(RT_OS_FREEBSD) + off_t cbMedia = 0; + if (!ioctl(RTFileToNative(hFile), DIOCGMEDIASIZE, &cbMedia)) + { + *pcbSize = cbMedia; + return VINF_SUCCESS; + } + /* might not be a block device. */ + if (errno == EINVAL || errno == ENOTTY) + return VINF_SUCCESS; + +#else + /* PORTME! Avoid this path when possible. */ + uint64_t offSaved = UINT64_MAX; + int rc = RTFileSeek(hFile, 0, RTFILE_SEEK_CURRENT, &offSaved); + if (RT_SUCCESS(rc)) + { + rc = RTFileSeek(hFile, 0, RTFILE_SEEK_END, pcbSize); + int rc2 = RTFileSeek(hFile, offSaved, RTFILE_SEEK_BEGIN, NULL); + if (RT_SUCCESS(rc)) + return rc2; + } +#endif + } + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileQueryMaxSizeEx(RTFILE hFile, PRTFOFF pcbMax) +{ + /* + * Save the current location + */ + uint64_t offOld = UINT64_MAX; + int rc = RTFileSeek(hFile, 0, RTFILE_SEEK_CURRENT, &offOld); + if (RT_FAILURE(rc)) + return rc; + + uint64_t offLow = 0; + uint64_t offHigh = INT64_MAX; /* we don't need bigger files */ + /** @todo Unfortunately this does not work for certain file system types, + * for instance cifs mounts. Even worse, statvfs.f_fsid returns 0 for such + * file systems. */ + + /* + * Quickly guess the order of magnitude for offHigh and offLow. + */ + { + uint64_t offHighPrev = offHigh; + while (offHigh >= INT32_MAX) + { + rc = RTFileSeek(hFile, offHigh, RTFILE_SEEK_BEGIN, NULL); + if (RT_SUCCESS(rc)) + { + offLow = offHigh; + offHigh = offHighPrev; + break; + } + else + { + offHighPrev = offHigh; + offHigh >>= 8; + } + } + } + + /* + * Sanity: if the seek to the initial offHigh (INT64_MAX) works, then + * this algorithm cannot possibly work. Declare defeat. + */ + if (offLow == offHigh) + { + rc = RTFileSeek(hFile, offOld, RTFILE_SEEK_BEGIN, NULL); + if (RT_SUCCESS(rc)) + rc = VERR_NOT_IMPLEMENTED; + + return rc; + } + + /* + * Perform a binary search for the max file size. + */ + while (offLow <= offHigh) + { + uint64_t offMid = offLow + (offHigh - offLow) / 2; + rc = RTFileSeek(hFile, offMid, RTFILE_SEEK_BEGIN, NULL); + if (RT_FAILURE(rc)) + offHigh = offMid - 1; + else + offLow = offMid + 1; + } + + if (pcbMax) + *pcbMax = RT_MIN(offLow, offHigh); + return RTFileSeek(hFile, offOld, RTFILE_SEEK_BEGIN, NULL); +} + + +RTR3DECL(bool) RTFileIsValid(RTFILE hFile) +{ + if (hFile != NIL_RTFILE) + { + int fFlags = fcntl(RTFileToNative(hFile), F_GETFD); + if (fFlags >= 0) + return true; + } + return false; +} + + +RTR3DECL(int) RTFileFlush(RTFILE hFile) +{ + if (!fsync(RTFileToNative(hFile))) + return VINF_SUCCESS; + /* Ignore EINVAL here as that's what returned for pseudo ttys + and other odd handles. */ + if (errno == EINVAL) + return VINF_NOT_SUPPORTED; + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileIoCtl(RTFILE hFile, unsigned long ulRequest, void *pvData, unsigned cbData, int *piRet) +{ + NOREF(cbData); + int rc = ioctl(RTFileToNative(hFile), ulRequest, pvData); + if (piRet) + *piRet = rc; + return rc >= 0 ? VINF_SUCCESS : RTErrConvertFromErrno(errno); +} + + +RTR3DECL(int) RTFileSetMode(RTFILE hFile, RTFMODE fMode) +{ + /* + * Normalize the mode and call the API. + */ + fMode = rtFsModeNormalize(fMode, NULL, 0, RTFS_TYPE_FILE); + if (!rtFsModeIsValid(fMode)) + return VERR_INVALID_PARAMETER; + + if (fchmod(RTFileToNative(hFile), fMode & RTFS_UNIX_MASK)) + { + int rc = RTErrConvertFromErrno(errno); + Log(("RTFileSetMode(%RTfile,%RTfmode): returns %Rrc\n", hFile, fMode, rc)); + return rc; + } + return VINF_SUCCESS; +} + + +RTDECL(int) RTFileSetOwner(RTFILE hFile, uint32_t uid, uint32_t gid) +{ + uid_t uidNative = uid != NIL_RTUID ? (uid_t)uid : (uid_t)-1; + AssertReturn(uid == uidNative, VERR_INVALID_PARAMETER); + gid_t gidNative = gid != NIL_RTGID ? (gid_t)gid : (gid_t)-1; + AssertReturn(gid == gidNative, VERR_INVALID_PARAMETER); + + if (fchown(RTFileToNative(hFile), uidNative, gidNative)) + return RTErrConvertFromErrno(errno); + return VINF_SUCCESS; +} + + +RTR3DECL(int) RTFileRename(const char *pszSrc, const char *pszDst, unsigned fRename) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszSrc, VERR_INVALID_POINTER); + AssertPtrReturn(pszDst, VERR_INVALID_POINTER); + AssertMsgReturn(*pszSrc, ("%p\n", pszSrc), VERR_INVALID_PARAMETER); + AssertMsgReturn(*pszDst, ("%p\n", pszDst), VERR_INVALID_PARAMETER); + AssertMsgReturn(!(fRename & ~RTPATHRENAME_FLAGS_REPLACE), ("%#x\n", fRename), VERR_INVALID_PARAMETER); + + /* + * Take common cause with RTPathRename. + */ + int rc = rtPathPosixRename(pszSrc, pszDst, fRename, RTFS_TYPE_FILE); + + LogFlow(("RTDirRename(%p:{%s}, %p:{%s}, %#x): returns %Rrc\n", + pszSrc, pszSrc, pszDst, pszDst, fRename, rc)); + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/fileio-sg-at-posix.cpp b/src/VBox/Runtime/r3/posix/fileio-sg-at-posix.cpp new file mode 100644 index 00000000..5fdea41a --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fileio-sg-at-posix.cpp @@ -0,0 +1,298 @@ +/* $Id: fileio-sg-at-posix.cpp $ */ +/** @file + * IPRT - File I/O, RTFileSgReadAt & RTFileSgWriteAt, posixy. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +/* + * Determin whether we've got preadv and pwritev. + */ +#include <iprt/cdefs.h> +#ifdef RT_OS_LINUX +/* Linux has these since glibc 2.10 and Linux 2.6.30: */ +# include <features.h> +# ifdef __GLIBC_PREREQ +# if __GLIBC_PREREQ(2,10) +# define HAVE_PREADV_AND_PWRITEV 1 +#else +# endif +# endif + +#elif defined(RT_OS_FREEBSD) +/* FreeBSD has these since 6.0: */ +# include <osreldate.h> +# ifdef __FreeBSD_version +# if __FreeBSD_version >= 600000 +# define HAVE_PREADV_AND_PWRITEV 1 +# endif +# endif + +#endif + +#ifndef HAVE_PREADV_AND_PWRITEV + +# include "../../generic/fileio-sg-at-generic.cpp" + +#else /* HAVE_PREADV_AND_PWRITEV - rest of the file */ + +# include <errno.h> +# include <sys/types.h> +# include <sys/uio.h> +# include <unistd.h> +# include <limits.h> +# if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) || defined(RT_OS_NETBSD) || defined(RT_OS_OPENBSD) +# include <sys/syslimits.h> +# endif + +# include "internal/iprt.h" +# include <iprt/file.h> + +# include <iprt/assert.h> +# include <iprt/err.h> +# include <iprt/log.h> + +# ifndef UIO_MAXIOV +# ifdef IOV_MAX +# define UIO_MAXIOV IOV_MAX +# else +# error "UIO_MAXIOV and IOV_MAX are undefined" +# endif +# endif + + +/* These assumptions simplifies things a lot here. */ +AssertCompileMembersSameSizeAndOffset(struct iovec, iov_base, RTSGSEG, pvSeg); +AssertCompileMembersSameSizeAndOffset(struct iovec, iov_len, RTSGSEG, cbSeg); + + +RTDECL(int) RTFileSgReadAt(RTFILE hFile, RTFOFF off, PRTSGBUF pSgBuf, size_t cbToRead, size_t *pcbRead) +{ + /* + * Make sure we set pcbRead. + */ + if (pcbRead) + *pcbRead = 0; + + /* + * Special case: Zero read == seek. + */ + if (cbToRead == 0) + return RTFileSeek(hFile, off, RTFILE_SEEK_BEGIN, NULL); + + /* + * We can use the segment array directly if we're at the start of the + * current S/G segment and cbToRead matches the remainder exactly. + */ + size_t cbTotalRead = 0; + + size_t const cbSgBufLeft = RTSgBufCalcLengthLeft(pSgBuf); + AssertMsgReturn(cbSgBufLeft >= cbToRead, ("%#zx vs %#zx\n", cbSgBufLeft, cbToRead), VERR_INVALID_PARAMETER); + + if (cbToRead == cbSgBufLeft) + while (RTSgBufIsAtStartOfSegment(pSgBuf)) + { + size_t const cSegsLeft = pSgBuf->cSegs - pSgBuf->idxSeg; + ssize_t cbThisRead = preadv(RTFileToNative(hFile), (const struct iovec *)&pSgBuf->paSegs[pSgBuf->idxSeg], + RT_MIN(cSegsLeft, UIO_MAXIOV), off); + if (cbThisRead >= 0) + { + AssertStmt((size_t)cbThisRead <= cbToRead, cbThisRead = cbToRead); + + RTSgBufAdvance(pSgBuf, cbThisRead); + cbTotalRead += cbThisRead; + cbToRead -= cbThisRead; + if (cbToRead == 0) + { + if (pcbRead) + *pcbRead = cbTotalRead; + return VINF_SUCCESS; + } + + if ( pcbRead + && ( cSegsLeft <= UIO_MAXIOV + || cbThisRead == 0 /* typically EOF */ )) + { + *pcbRead = cbTotalRead; + return VINF_SUCCESS; + } + if (cbThisRead == 0) + return VERR_EOF; + + off += cbThisRead; + } + else if (cbTotalRead > 0 && pcbRead) + { + *pcbRead = cbTotalRead; + return VINF_SUCCESS; + } + else + return RTErrConvertFromErrno(errno); + } + + /* + * Unaligned start or not reading the whole buffer. For reasons of + * simplicity, we work the input segment by segment like the generic code. + */ + int rc = VINF_SUCCESS; + while (cbToRead > 0) + { + size_t cbSeg; + void *pvSeg = RTSgBufGetCurrentSegment(pSgBuf, cbToRead, &cbSeg); + size_t cbThisRead = cbSeg; + rc = RTFileReadAt(hFile, off, pvSeg, cbSeg, pcbRead ? &cbThisRead : NULL); + if (RT_SUCCESS(rc)) + { + RTSgBufAdvance(pSgBuf, cbThisRead); + cbTotalRead += cbThisRead; + } + else + break; + if ((size_t)cbThisRead < cbSeg) + { + AssertStmt(pcbRead, rc = VERR_INTERNAL_ERROR_2); + break; + } + + Assert(cbSeg == cbThisRead); + cbToRead -= cbSeg; + off += cbSeg; + } + if (pcbRead) + *pcbRead = cbTotalRead; + return rc; +} + + +RTDECL(int) RTFileSgWriteAt(RTFILE hFile, RTFOFF off, PRTSGBUF pSgBuf, size_t cbToWrite, size_t *pcbWritten) +{ + /* + * Make sure we set pcbWritten. + */ + if (pcbWritten) + *pcbWritten = 0; + + /* + * Special case: Zero write == seek. + */ + if (cbToWrite == 0) + return RTFileSeek(hFile, off, RTFILE_SEEK_BEGIN, NULL); + + /* + * We can use the segment array directly if we're at the start of the + * current S/G segment and cbToWrite matches the remainder exactly. + */ + size_t cbTotalWritten = 0; + + size_t const cbSgBufLeft = RTSgBufCalcLengthLeft(pSgBuf); + AssertMsgReturn(cbSgBufLeft >= cbToWrite, ("%#zx vs %#zx\n", cbSgBufLeft, cbToWrite), VERR_INVALID_PARAMETER); + + if (cbToWrite == cbSgBufLeft) + while (RTSgBufIsAtStartOfSegment(pSgBuf)) + { + size_t const cSegsLeft = pSgBuf->cSegs - pSgBuf->idxSeg; + ssize_t cbThisWritten = pwritev(RTFileToNative(hFile), (const struct iovec *)&pSgBuf->paSegs[pSgBuf->idxSeg], + RT_MIN(cSegsLeft, UIO_MAXIOV), off); + if (cbThisWritten >= 0) + { + AssertStmt((size_t)cbThisWritten <= cbToWrite, cbThisWritten = cbToWrite); + + RTSgBufAdvance(pSgBuf, cbThisWritten); + cbTotalWritten += cbThisWritten; + cbToWrite -= cbThisWritten; + if (cbToWrite == 0) + { + if (pcbWritten) + *pcbWritten = cbTotalWritten; + return VINF_SUCCESS; + } + + if ( pcbWritten + && ( cSegsLeft <= UIO_MAXIOV + || cbThisWritten == 0 /* non-file, full buffer/whatever */ )) + { + *pcbWritten = cbTotalWritten; + return VINF_SUCCESS; + } + if (cbThisWritten == 0) + return VERR_TRY_AGAIN; + + off += cbThisWritten; + } + else if (cbTotalWritten > 0 && pcbWritten) + { + *pcbWritten = cbTotalWritten; + return VINF_SUCCESS; + } + else + return RTErrConvertFromErrno(errno); + } + + /* + * Unaligned start or not writing the whole buffer. For reasons of + * simplicity, we work the input segment by segment like the generic code. + */ + int rc = VINF_SUCCESS; + while (cbToWrite > 0) + { + size_t cbSeg; + void *pvSeg = RTSgBufGetCurrentSegment(pSgBuf, cbToWrite, &cbSeg); + size_t cbThisWritten = cbSeg; + rc = RTFileWriteAt(hFile, off, pvSeg, cbSeg, pcbWritten ? &cbThisWritten : NULL); + if (RT_SUCCESS(rc)) + { + RTSgBufAdvance(pSgBuf, cbThisWritten); + cbTotalWritten += cbThisWritten; + } + else + break; + if ((size_t)cbThisWritten < cbSeg) + { + AssertStmt(pcbWritten, rc = VERR_INTERNAL_ERROR_2); + break; + } + + Assert(cbSeg == cbThisWritten); + cbToWrite -= cbSeg; + off += cbSeg; + } + if (pcbWritten) + *pcbWritten = cbTotalWritten; + return rc; +} + +#endif /* HAVE_PREADV_AND_PWRITEV */ + diff --git a/src/VBox/Runtime/r3/posix/fileio-sg-posix.cpp b/src/VBox/Runtime/r3/posix/fileio-sg-posix.cpp new file mode 100644 index 00000000..c8c95786 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fileio-sg-posix.cpp @@ -0,0 +1,260 @@ +/* $Id: fileio-sg-posix.cpp $ */ +/** @file + * IPRT - File I/O, RTFileSgRead & RTFileSgWrite, posixy. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/cdefs.h> +#include <errno.h> +#include <sys/types.h> +#include <sys/uio.h> +#include <unistd.h> +#include <limits.h> +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) || defined(RT_OS_NETBSD) || defined(RT_OS_OPENBSD) +# include <sys/syslimits.h> +#endif + +#include "internal/iprt.h" +#include <iprt/file.h> + +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/log.h> + +#ifndef UIO_MAXIOV +# ifdef IOV_MAX +# define UIO_MAXIOV IOV_MAX +# else +# error "UIO_MAXIOV and IOV_MAX are undefined" +# endif +#endif + + +/* These assumptions simplifies things a lot here. */ +AssertCompileMembersSameSizeAndOffset(struct iovec, iov_base, RTSGSEG, pvSeg); +AssertCompileMembersSameSizeAndOffset(struct iovec, iov_len, RTSGSEG, cbSeg); + + +RTDECL(int) RTFileSgRead(RTFILE hFile, PRTSGBUF pSgBuf, size_t cbToRead, size_t *pcbRead) +{ + /* + * Make sure we set pcbRead. + */ + if (pcbRead) + *pcbRead = 0; + + /* + * Special case: Zero read == nop. + */ + if (cbToRead == 0) + return VINF_SUCCESS; + + /* + * We can use the segment array directly if we're at the start of the + * current S/G segment and cbToRead matches the remainder exactly. + */ + size_t cbTotalRead = 0; + + size_t const cbSgBufLeft = RTSgBufCalcLengthLeft(pSgBuf); + AssertMsgReturn(cbSgBufLeft >= cbToRead, ("%#zx vs %#zx\n", cbSgBufLeft, cbToRead), VERR_INVALID_PARAMETER); + + if (cbToRead == cbSgBufLeft) + while (RTSgBufIsAtStartOfSegment(pSgBuf)) + { + size_t const cSegsLeft = pSgBuf->cSegs - pSgBuf->idxSeg; + ssize_t cbThisRead = readv(RTFileToNative(hFile), (const struct iovec *)&pSgBuf->paSegs[pSgBuf->idxSeg], + RT_MIN(cSegsLeft, UIO_MAXIOV)); + if (cbThisRead >= 0) + { + AssertStmt((size_t)cbThisRead <= cbToRead, cbThisRead = cbToRead); + + RTSgBufAdvance(pSgBuf, cbThisRead); + cbTotalRead += cbThisRead; + cbToRead -= cbThisRead; + if (cbToRead == 0) + { + if (pcbRead) + *pcbRead = cbTotalRead; + return VINF_SUCCESS; + } + + if ( pcbRead + && ( cSegsLeft <= UIO_MAXIOV + || cbThisRead == 0 /* typically EOF */ )) + { + *pcbRead = cbTotalRead; + return VINF_SUCCESS; + } + if (cbThisRead == 0) + return VERR_EOF; + } + else if (cbTotalRead > 0 && pcbRead) + { + *pcbRead = cbTotalRead; + return VINF_SUCCESS; + } + else + return RTErrConvertFromErrno(errno); + } + + /* + * Unaligned start or not reading the whole buffer. For reasons of + * simplicity, we work the input segment by segment like the generic code. + */ + int rc = VINF_SUCCESS; + while (cbToRead > 0) + { + size_t cbSeg; + void *pvSeg = RTSgBufGetCurrentSegment(pSgBuf, cbToRead, &cbSeg); + size_t cbThisRead = cbSeg; + rc = RTFileRead(hFile, pvSeg, cbSeg, pcbRead ? &cbThisRead : NULL); + if (RT_SUCCESS(rc)) + { + RTSgBufAdvance(pSgBuf, cbThisRead); + cbTotalRead += cbThisRead; + } + else + break; + if ((size_t)cbThisRead < cbSeg) + { + AssertStmt(pcbRead, rc = VERR_INTERNAL_ERROR_2); + break; + } + + Assert(cbSeg == cbThisRead); + cbToRead -= cbSeg; + } + if (pcbRead) + *pcbRead = cbTotalRead; + return rc; +} + + +RTDECL(int) RTFileSgWrite(RTFILE hFile, PRTSGBUF pSgBuf, size_t cbToWrite, size_t *pcbWritten) +{ + /* + * Make sure we set pcbWritten. + */ + if (pcbWritten) + *pcbWritten = 0; + + /* + * Special case: Zero write == nop. + */ + if (cbToWrite == 0) + return VINF_SUCCESS; + + /* + * We can use the segment array directly if we're at the start of the + * current S/G segment and cbToWrite matches the remainder exactly. + */ + size_t cbTotalWritten = 0; + + size_t const cbSgBufLeft = RTSgBufCalcLengthLeft(pSgBuf); + AssertMsgReturn(cbSgBufLeft >= cbToWrite, ("%#zx vs %#zx\n", cbSgBufLeft, cbToWrite), VERR_INVALID_PARAMETER); + + if (cbToWrite == cbSgBufLeft) + while (RTSgBufIsAtStartOfSegment(pSgBuf)) + { + size_t const cSegsLeft = pSgBuf->cSegs - pSgBuf->idxSeg; + ssize_t cbThisWritten = writev(RTFileToNative(hFile), (const struct iovec *)&pSgBuf->paSegs[pSgBuf->idxSeg], + RT_MIN(cSegsLeft, UIO_MAXIOV)); + if (cbThisWritten >= 0) + { + AssertStmt((size_t)cbThisWritten <= cbToWrite, cbThisWritten = cbToWrite); + + RTSgBufAdvance(pSgBuf, cbThisWritten); + cbTotalWritten += cbThisWritten; + cbToWrite -= cbThisWritten; + if (cbToWrite == 0) + { + if (pcbWritten) + *pcbWritten = cbTotalWritten; + return VINF_SUCCESS; + } + + if ( pcbWritten + && ( cSegsLeft <= UIO_MAXIOV + || cbThisWritten == 0 /* non-file, full buffer/whatever */ )) + { + *pcbWritten = cbTotalWritten; + return VINF_SUCCESS; + } + if (cbThisWritten == 0) + return VERR_TRY_AGAIN; + } + else if (cbTotalWritten > 0 && pcbWritten) + { + *pcbWritten = cbTotalWritten; + return VINF_SUCCESS; + } + else + return RTErrConvertFromErrno(errno); + } + + /* + * Unaligned start or not writing the whole buffer. For reasons of + * simplicity, we work the input segment by segment like the generic code. + */ + int rc = VINF_SUCCESS; + while (cbToWrite > 0) + { + size_t cbSeg; + void *pvSeg = RTSgBufGetCurrentSegment(pSgBuf, cbToWrite, &cbSeg); + size_t cbThisWritten = cbSeg; + rc = RTFileWrite(hFile, pvSeg, cbSeg, pcbWritten ? &cbThisWritten : NULL); + if (RT_SUCCESS(rc)) + { + RTSgBufAdvance(pSgBuf, cbThisWritten); + cbTotalWritten += cbThisWritten; + } + else + break; + if ((size_t)cbThisWritten < cbSeg) + { + AssertStmt(pcbWritten, rc = VERR_INTERNAL_ERROR_2); + break; + } + + Assert(cbSeg == cbThisWritten); + cbToWrite -= cbSeg; + } + if (pcbWritten) + *pcbWritten = cbTotalWritten; + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/fileio2-posix.cpp b/src/VBox/Runtime/r3/posix/fileio2-posix.cpp new file mode 100644 index 00000000..4879f51e --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fileio2-posix.cpp @@ -0,0 +1,210 @@ +/* $Id: fileio2-posix.cpp $ */ +/** @file + * IPRT - File I/O, POSIX, Part 2. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_FILE + +#include <errno.h> +#include <sys/stat.h> +#include <sys/types.h> +#ifdef _MSC_VER +# include <io.h> +# include <stdio.h> +#else +# include <unistd.h> +# include <sys/time.h> +#endif +#ifdef RT_OS_LINUX +# include <sys/file.h> +#endif +#if defined(RT_OS_OS2) && (!defined(__INNOTEK_LIBC__) || __INNOTEK_LIBC__ < 0x006) +# include <io.h> +#endif + +#ifdef RT_OS_SOLARIS +# define futimes(filedes, timeval) futimesat(filedes, NULL, timeval) +#endif + +#ifdef RT_OS_HAIKU +# define USE_FUTIMENS +#endif + +#include <iprt/file.h> +#include <iprt/path.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/errcore.h> +#include <iprt/log.h> +#include "internal/file.h" +#include "internal/fs.h" +#include "internal/path.h" + + + +RTR3DECL(int) RTFileQueryInfo(RTFILE hFile, PRTFSOBJINFO pObjInfo, RTFSOBJATTRADD enmAdditionalAttribs) +{ + /* + * Validate input. + */ + AssertReturn(hFile != NIL_RTFILE, VERR_INVALID_PARAMETER); + AssertPtrReturn(pObjInfo, VERR_INVALID_PARAMETER); + if ( enmAdditionalAttribs < RTFSOBJATTRADD_NOTHING + || enmAdditionalAttribs > RTFSOBJATTRADD_LAST) + { + AssertMsgFailed(("Invalid enmAdditionalAttribs=%p\n", enmAdditionalAttribs)); + return VERR_INVALID_PARAMETER; + } + + /* + * Query file info. + */ + struct stat Stat; + if (fstat(RTFileToNative(hFile), &Stat)) + { + int rc = RTErrConvertFromErrno(errno); + Log(("RTFileQueryInfo(%RTfile,,%d): returns %Rrc\n", hFile, enmAdditionalAttribs, rc)); + return rc; + } + + /* + * Setup the returned data. + */ + rtFsConvertStatToObjInfo(pObjInfo, &Stat, NULL, 0); + + /* + * Requested attributes (we cannot provide anything actually). + */ + switch (enmAdditionalAttribs) + { + case RTFSOBJATTRADD_NOTHING: + case RTFSOBJATTRADD_UNIX: + /* done */ + break; + + case RTFSOBJATTRADD_UNIX_OWNER: + rtFsObjInfoAttrSetUnixOwner(pObjInfo, Stat.st_uid); + break; + + case RTFSOBJATTRADD_UNIX_GROUP: + rtFsObjInfoAttrSetUnixGroup(pObjInfo, Stat.st_gid); + break; + + case RTFSOBJATTRADD_EASIZE: + pObjInfo->Attr.enmAdditional = RTFSOBJATTRADD_EASIZE; + pObjInfo->Attr.u.EASize.cb = 0; + break; + + default: + AssertMsgFailed(("Impossible!\n")); + return VERR_INTERNAL_ERROR; + } + + LogFlow(("RTFileQueryInfo(%RTfile,,%d): returns VINF_SUCCESS\n", hFile, enmAdditionalAttribs)); + return VINF_SUCCESS; +} + + +RTR3DECL(int) RTFileSetTimes(RTFILE hFile, PCRTTIMESPEC pAccessTime, PCRTTIMESPEC pModificationTime, + PCRTTIMESPEC pChangeTime, PCRTTIMESPEC pBirthTime) +{ + NOREF(pChangeTime); NOREF(pBirthTime); + + /* + * We can only set AccessTime and ModificationTime, so if neither + * are specified we can return immediately. + */ + if (!pAccessTime && !pModificationTime) + return VINF_SUCCESS; + +#ifdef USE_FUTIMENS + struct timespec aTimespecs[2]; + if (pAccessTime && pModificationTime) + { + memcpy(&aTimespecs[0], pAccessTime, sizeof(struct timespec)); + memcpy(&aTimespecs[1], pModificationTime, sizeof(struct timespec)); + } + else + { + RTFSOBJINFO ObjInfo; + int rc = RTFileQueryInfo(hFile, &ObjInfo, RTFSOBJATTRADD_UNIX); + if (RT_FAILURE(rc)) + return rc; + memcpy(&aTimespecs[0], pAccessTime ? pAccessTime : &ObjInfo.AccessTime, sizeof(struct timespec)); + memcpy(&aTimespecs[1], pModificationTime ? pModificationTime : &ObjInfo.ModificationTime, sizeof(struct timespec)); + } + + if (futimens(RTFileToNative(hFile), aTimespecs)) + { + int rc = RTErrConvertFromErrno(errno); + Log(("RTFileSetTimes(%RTfile,%p,%p,,): returns %Rrc\n", hFile, pAccessTime, pModificationTime, rc)); + return rc; + } +#else + /* + * Convert the input to timeval, getting the missing one if necessary, + * and call the API which does the change. + */ + struct timeval aTimevals[2]; + if (pAccessTime && pModificationTime) + { + RTTimeSpecGetTimeval(pAccessTime, &aTimevals[0]); + RTTimeSpecGetTimeval(pModificationTime, &aTimevals[1]); + } + else + { + RTFSOBJINFO ObjInfo; + int rc = RTFileQueryInfo(hFile, &ObjInfo, RTFSOBJATTRADD_UNIX); + if (RT_FAILURE(rc)) + return rc; + RTTimeSpecGetTimeval(pAccessTime ? pAccessTime : &ObjInfo.AccessTime, &aTimevals[0]); + RTTimeSpecGetTimeval(pModificationTime ? pModificationTime : &ObjInfo.ModificationTime, &aTimevals[1]); + } + + /* XXX this falls back to utimes("/proc/self/fd/...",...) for older kernels/glibcs and this + * will not work for hardened builds where this directory is owned by root.root and mode 0500 */ + if (futimes(RTFileToNative(hFile), aTimevals)) + { + int rc = RTErrConvertFromErrno(errno); + Log(("RTFileSetTimes(%RTfile,%p,%p,,): returns %Rrc\n", hFile, pAccessTime, pModificationTime, rc)); + return rc; + } +#endif + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/filelock-posix.cpp b/src/VBox/Runtime/r3/posix/filelock-posix.cpp new file mode 100644 index 00000000..3a0e230f --- /dev/null +++ b/src/VBox/Runtime/r3/posix/filelock-posix.cpp @@ -0,0 +1,148 @@ +/* $Id: filelock-posix.cpp $ */ +/** @file + * IPRT - File Locking, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_FILE + +#include <errno.h> +#include <sys/types.h> +#include <sys/ioctl.h> +#include <fcntl.h> +#include <unistd.h> +#include <sys/time.h> + +#include <iprt/file.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include "internal/file.h" +#include "internal/fs.h" + + + + +RTR3DECL(int) RTFileLock(RTFILE hFile, unsigned fLock, int64_t offLock, uint64_t cbLock) +{ + Assert(offLock >= 0); + + /* Check arguments. */ + if (fLock & ~RTFILE_LOCK_MASK) + { + AssertMsgFailed(("Invalid fLock=%08X\n", fLock)); + return VERR_INVALID_PARAMETER; + } + + /* + * Validate offset. + */ + if ( sizeof(off_t) < sizeof(cbLock) + && ( (offLock >> 32) != 0 + || (cbLock >> 32) != 0 + || ((offLock + cbLock) >> 32) != 0)) + { + AssertMsgFailed(("64-bit file i/o not supported! offLock=%lld cbLock=%lld\n", offLock, cbLock)); + return VERR_NOT_SUPPORTED; + } + + /* Prepare flock structure. */ + struct flock fl; + Assert(RTFILE_LOCK_WRITE); + fl.l_type = (fLock & RTFILE_LOCK_WRITE) ? F_WRLCK : F_RDLCK; + fl.l_whence = SEEK_SET; + fl.l_start = (off_t)offLock; + fl.l_len = (off_t)cbLock; + fl.l_pid = 0; + + Assert(RTFILE_LOCK_WAIT); + if (fcntl(RTFileToNative(hFile), (fLock & RTFILE_LOCK_WAIT) ? F_SETLKW : F_SETLK, &fl) >= 0) + return VINF_SUCCESS; + + int iErr = errno; + if ( iErr == EAGAIN + || iErr == EACCES) + return VERR_FILE_LOCK_VIOLATION; + + return RTErrConvertFromErrno(iErr); +} + + +RTR3DECL(int) RTFileChangeLock(RTFILE hFile, unsigned fLock, int64_t offLock, uint64_t cbLock) +{ + /** @todo We never returns VERR_FILE_NOT_LOCKED for now. */ + return RTFileLock(hFile, fLock, offLock, cbLock); +} + + +RTR3DECL(int) RTFileUnlock(RTFILE hFile, int64_t offLock, uint64_t cbLock) +{ + Assert(offLock >= 0); + + /* + * Validate offset. + */ + if ( sizeof(off_t) < sizeof(cbLock) + && ( (offLock >> 32) != 0 + || (cbLock >> 32) != 0 + || ((offLock + cbLock) >> 32) != 0)) + { + AssertMsgFailed(("64-bit file i/o not supported! offLock=%lld cbLock=%lld\n", offLock, cbLock)); + return VERR_NOT_SUPPORTED; + } + + /* Prepare flock structure. */ + struct flock fl; + fl.l_type = F_UNLCK; + fl.l_whence = SEEK_SET; + fl.l_start = (off_t)offLock; + fl.l_len = (off_t)cbLock; + fl.l_pid = 0; + + if (fcntl(RTFileToNative(hFile), F_SETLK, &fl) >= 0) + return VINF_SUCCESS; + + /** @todo check error codes for non existing lock. */ + int iErr = errno; + if ( iErr == EAGAIN + || iErr == EACCES) + return VERR_FILE_LOCK_VIOLATION; + + return RTErrConvertFromErrno(iErr); +} + diff --git a/src/VBox/Runtime/r3/posix/fs-posix.cpp b/src/VBox/Runtime/r3/posix/fs-posix.cpp new file mode 100644 index 00000000..04b3289a --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fs-posix.cpp @@ -0,0 +1,346 @@ +/* $Id: fs-posix.cpp $ */ +/** @file + * IPRT - File System, Linux. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_FS +#include <sys/statvfs.h> +#include <errno.h> +#include <stdio.h> +#ifdef RT_OS_LINUX +# include <mntent.h> +#endif +#if defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) +# include <sys/mount.h> +#endif + +#include <iprt/fs.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include <iprt/errcore.h> +#include <iprt/log.h> +#include <iprt/string.h> +#include "internal/fs.h" +#include "internal/path.h" + + + +RTR3DECL(int) RTFsQuerySizes(const char *pszFsPath, RTFOFF *pcbTotal, RTFOFF *pcbFree, + uint32_t *pcbBlock, uint32_t *pcbSector) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszFsPath, VERR_INVALID_POINTER); + AssertReturn(*pszFsPath != '\0', VERR_INVALID_PARAMETER); + + /* + * Convert the path and query the information. + */ + char const *pszNativeFsPath; + int rc = rtPathToNative(&pszNativeFsPath, pszFsPath, NULL); + if (RT_SUCCESS(rc)) + { + /** @todo I'm not quite sure if statvfs was properly specified by SuS, I have to check my own + * implementation and FreeBSD before this can eventually be promoted to posix. */ + struct statvfs StatVFS; + RT_ZERO(StatVFS); + if (!statvfs(pszNativeFsPath, &StatVFS)) + { + /* + * Calc the returned values. + */ + if (pcbTotal) + *pcbTotal = (RTFOFF)StatVFS.f_blocks * StatVFS.f_frsize; + if (pcbFree) + *pcbFree = (RTFOFF)StatVFS.f_bavail * StatVFS.f_frsize; + if (pcbBlock) + *pcbBlock = StatVFS.f_frsize; + /* no idea how to get the sector... */ + if (pcbSector) + *pcbSector = 512; + } + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativeFsPath, pszFsPath); + } + + LogFlow(("RTFsQuerySizes(%p:{%s}, %p:{%RTfoff}, %p:{%RTfoff}, %p:{%RX32}, %p:{%RX32}): returns %Rrc\n", + pszFsPath, pszFsPath, pcbTotal, pcbTotal ? *pcbTotal : 0, pcbFree, pcbFree ? *pcbFree : 0, + pcbBlock, pcbBlock ? *pcbBlock : 0, pcbSector, pcbSector ? *pcbSector : 0, rc)); + return rc; +} + + +RTR3DECL(int) RTFsQuerySerial(const char *pszFsPath, uint32_t *pu32Serial) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszFsPath, VERR_INVALID_POINTER); + AssertReturn(*pszFsPath != '\0', VERR_INVALID_PARAMETER); + AssertPtrReturn(pu32Serial, VERR_INVALID_POINTER); + + /* + * Convert the path and query the stats. + * We're simply return the device id. + */ + char const *pszNativeFsPath; + int rc = rtPathToNative(&pszNativeFsPath, pszFsPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat Stat; + if (!stat(pszNativeFsPath, &Stat)) + { + if (pu32Serial) + *pu32Serial = (uint32_t)Stat.st_dev; + } + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativeFsPath, pszFsPath); + } + LogFlow(("RTFsQuerySerial(%p:{%s}, %p:{%RX32}: returns %Rrc\n", + pszFsPath, pszFsPath, pu32Serial, pu32Serial ? *pu32Serial : 0, rc)); + return rc; +} + + +RTR3DECL(int) RTFsQueryProperties(const char *pszFsPath, PRTFSPROPERTIES pProperties) +{ + /* + * Validate. + */ + AssertPtrReturn(pszFsPath, VERR_INVALID_POINTER); + AssertReturn(*pszFsPath != '\0', VERR_INVALID_PARAMETER); + AssertPtrReturn(pProperties, VERR_INVALID_POINTER); + + /* + * Convert the path and query the information. + */ + char const *pszNativeFsPath; + int rc = rtPathToNative(&pszNativeFsPath, pszFsPath, NULL); + if (RT_SUCCESS(rc)) + { + struct statvfs StatVFS; + RT_ZERO(StatVFS); + if (!statvfs(pszNativeFsPath, &StatVFS)) + { + /* + * Calc/fake the returned values. + */ + pProperties->cbMaxComponent = StatVFS.f_namemax; +#if defined(RT_OS_OS2) || defined(RT_OS_WINDOWS) + pProperties->fCaseSensitive = false; +#else + pProperties->fCaseSensitive = true; +#endif + pProperties->fCompressed = false; + pProperties->fFileCompression = false; + pProperties->fReadOnly = !!(StatVFS.f_flag & ST_RDONLY); + pProperties->fRemote = false; + pProperties->fSupportsUnicode = true; + } + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativeFsPath, pszFsPath); + } + + LogFlow(("RTFsQueryProperties(%p:{%s}, %p:{.cbMaxComponent=%u, .fReadOnly=%RTbool}): returns %Rrc\n", + pszFsPath, pszFsPath, pProperties, pProperties->cbMaxComponent, pProperties->fReadOnly, rc)); + return rc; +} + + +RTR3DECL(bool) RTFsIsCaseSensitive(const char *pszFsPath) +{ + RT_NOREF_PV(pszFsPath); +#if defined(RT_OS_OS2) || defined(RT_OS_WINDOWS) + return false; +#else + return true; +#endif +} + + +RTR3DECL(int) RTFsQueryType(const char *pszFsPath, PRTFSTYPE penmType) +{ + *penmType = RTFSTYPE_UNKNOWN; + + /* + * Validate input. + */ + AssertPtrReturn(pszFsPath, VERR_INVALID_POINTER); + AssertReturn(*pszFsPath, VERR_INVALID_PARAMETER); + + /* + * Convert the path and query the stats. + * We're simply return the device id. + */ + char const *pszNativeFsPath; + int rc = rtPathToNative(&pszNativeFsPath, pszFsPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat Stat; + if (!stat(pszNativeFsPath, &Stat)) + { +#if defined(RT_OS_LINUX) + FILE *mounted = setmntent("/proc/mounts", "r"); + if (!mounted) + mounted = setmntent("/etc/mtab", "r"); + if (mounted) + { + char szBuf[1024]; + struct stat mntStat; + struct mntent mntEnt; + while (getmntent_r(mounted, &mntEnt, szBuf, sizeof(szBuf))) + { + if (!stat(mntEnt.mnt_dir, &mntStat)) + { + if (mntStat.st_dev == Stat.st_dev) + { + if (!strcmp("ext4", mntEnt.mnt_type)) + *penmType = RTFSTYPE_EXT4; + else if (!strcmp("ext3", mntEnt.mnt_type)) + *penmType = RTFSTYPE_EXT3; + else if (!strcmp("ext2", mntEnt.mnt_type)) + *penmType = RTFSTYPE_EXT2; + else if (!strcmp("jfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_JFS; + else if (!strcmp("xfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_XFS; + else if (!strcmp("btrfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_BTRFS; + else if ( !strcmp("vfat", mntEnt.mnt_type) + || !strcmp("msdos", mntEnt.mnt_type)) + *penmType = RTFSTYPE_FAT; + else if (!strcmp("ntfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_NTFS; + else if (!strcmp("hpfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_HPFS; + else if (!strcmp("ufs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_UFS; + else if (!strcmp("tmpfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_TMPFS; + else if (!strcmp("hfsplus", mntEnt.mnt_type)) + *penmType = RTFSTYPE_HFS; + else if (!strcmp("udf", mntEnt.mnt_type)) + *penmType = RTFSTYPE_UDF; + else if (!strcmp("iso9660", mntEnt.mnt_type)) + *penmType = RTFSTYPE_ISO9660; + else if (!strcmp("smbfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_SMBFS; + else if (!strcmp("cifs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_CIFS; + else if (!strcmp("nfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_NFS; + else if (!strcmp("nfs4", mntEnt.mnt_type)) + *penmType = RTFSTYPE_NFS; + else if (!strcmp("ocfs2", mntEnt.mnt_type)) + *penmType = RTFSTYPE_OCFS2; + else if (!strcmp("sysfs", mntEnt.mnt_type)) + *penmType = RTFSTYPE_SYSFS; + else if (!strcmp("proc", mntEnt.mnt_type)) + *penmType = RTFSTYPE_PROC; + else if ( !strcmp("fuse", mntEnt.mnt_type) + || !strncmp("fuse.", mntEnt.mnt_type, 5) + || !strcmp("fuseblk", mntEnt.mnt_type)) + *penmType = RTFSTYPE_FUSE; + else + { + /* sometimes there are more than one entry for the same partition */ + continue; + } + break; + } + } + } + endmntent(mounted); + } + +#elif defined(RT_OS_SOLARIS) + /* + * Home directories are normally loopback mounted in Solaris 11 (st_fstype=="lofs") + * so statvfs(2) is needed to get the underlying file system information. + */ + struct statvfs statvfsBuf; + if (!statvfs(pszNativeFsPath, &statvfsBuf)) + { + if (!strcmp("zfs", statvfsBuf.f_basetype)) + *penmType = RTFSTYPE_ZFS; + else if (!strcmp("ufs", statvfsBuf.f_basetype)) + *penmType = RTFSTYPE_UFS; + else if (!strcmp("nfs", statvfsBuf.f_basetype)) + *penmType = RTFSTYPE_NFS; + } + +#elif defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) + struct statfs statfsBuf; + if (!statfs(pszNativeFsPath, &statfsBuf)) + { + if (!strcmp("hfs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_HFS; + else if (!strcmp("apfs", statfsBuf.f_fstypename)) /** @todo verify apfs signature. */ + *penmType = RTFSTYPE_APFS; + else if ( !strcmp("fat", statfsBuf.f_fstypename) + || !strcmp("msdos", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_FAT; + else if (!strcmp("ntfs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_NTFS; + else if (!strcmp("autofs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_AUTOFS; + else if (!strcmp("devfs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_DEVFS; + else if (!strcmp("nfs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_NFS; + else if (!strcmp("ufs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_UFS; + else if (!strcmp("zfs", statfsBuf.f_fstypename)) + *penmType = RTFSTYPE_ZFS; + } + else + rc = RTErrConvertFromErrno(errno); +#endif + } + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativeFsPath, pszFsPath); + } + + return rc; +} diff --git a/src/VBox/Runtime/r3/posix/fs2-posix.cpp b/src/VBox/Runtime/r3/posix/fs2-posix.cpp new file mode 100644 index 00000000..adbdecfe --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fs2-posix.cpp @@ -0,0 +1,165 @@ +/* $Id: fs2-posix.cpp $ */ +/** @file + * IPRT - File System Helpers, POSIX, Part 2. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define RTTIME_INCL_TIMESPEC +#include <sys/time.h> +#include <sys/param.h> +#ifndef DEV_BSIZE +# include <sys/stat.h> +# if defined(RT_OS_HAIKU) && !defined(S_BLKSIZE) +# define S_BLKSIZE 512 +# endif +# define DEV_BSIZE S_BLKSIZE /** @todo bird: add DEV_BSIZE to sys/param.h on OS/2. */ +#endif + +#include <iprt/fs.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include <iprt/time.h> +#include "internal/fs.h" + + +/** + * Internal worker function which setups RTFSOBJINFO based on a UNIX stat struct. + * + * @param pObjInfo The file system object info structure to setup. + * @param pStat The stat structure to use. + * @param pszName The filename which this applies to (exe/hidden check). + * @param cbName The length of that filename. (optional, set 0) + */ +void rtFsConvertStatToObjInfo(PRTFSOBJINFO pObjInfo, const struct stat *pStat, const char *pszName, unsigned cbName) +{ + pObjInfo->cbObject = pStat->st_size; + pObjInfo->cbAllocated = pStat->st_blocks * DEV_BSIZE; + +#ifdef HAVE_STAT_NSEC + RTTimeSpecAddNano(RTTimeSpecSetSeconds(&pObjInfo->AccessTime, pStat->st_atime), pStat->st_atimensec); + RTTimeSpecAddNano(RTTimeSpecSetSeconds(&pObjInfo->ModificationTime, pStat->st_mtime), pStat->st_mtimensec); + RTTimeSpecAddNano(RTTimeSpecSetSeconds(&pObjInfo->ChangeTime, pStat->st_ctime), pStat->st_ctimensec); +# ifdef HAVE_STAT_BIRTHTIME + RTTimeSpecAddNano(RTTimeSpecSetSeconds(&pObjInfo->BirthTime, pStat->st_birthtime), pStat->st_birthtimensec); +# endif + +#elif defined(HAVE_STAT_TIMESPEC_BRIEF) + RTTimeSpecSetTimespec(&pObjInfo->AccessTime, &pStat->st_atim); + RTTimeSpecSetTimespec(&pObjInfo->ModificationTime, &pStat->st_mtim); + RTTimeSpecSetTimespec(&pObjInfo->ChangeTime, &pStat->st_ctim); +# ifdef HAVE_STAT_BIRTHTIME + RTTimeSpecSetTimespec(&pObjInfo->BirthTime, &pStat->st_birthtim); +# endif + +#elif defined(HAVE_STAT_TIMESPEC) + RTTimeSpecSetTimespec(&pObjInfo->AccessTime, pStat->st_atimespec); + RTTimeSpecSetTimespec(&pObjInfo->ModificationTime, pStat->st_mtimespec); + RTTimeSpecSetTimespec(&pObjInfo->ChangeTime, pStat->st_ctimespec); +# ifdef HAVE_STAT_BIRTHTIME + RTTimeSpecSetTimespec(&pObjInfo->BirthTime, pStat->st_birthtimespec); +# endif + +#else /* just the normal stuff */ + RTTimeSpecSetSeconds(&pObjInfo->AccessTime, pStat->st_atime); + RTTimeSpecSetSeconds(&pObjInfo->ModificationTime, pStat->st_mtime); + RTTimeSpecSetSeconds(&pObjInfo->ChangeTime, pStat->st_ctime); +# ifdef HAVE_STAT_BIRTHTIME + RTTimeSpecSetSeconds(&pObjInfo->BirthTime, pStat->st_birthtime); +# endif +#endif +#ifndef HAVE_STAT_BIRTHTIME + pObjInfo->BirthTime = pObjInfo->ChangeTime; +#endif + + /* the file mode */ + RTFMODE fMode = pStat->st_mode & RTFS_UNIX_MASK; + Assert(RTFS_UNIX_ISUID == S_ISUID); + Assert(RTFS_UNIX_ISGID == S_ISGID); +#ifdef S_ISTXT + Assert(RTFS_UNIX_ISTXT == S_ISTXT); +#elif defined(S_ISVTX) + Assert(RTFS_UNIX_ISTXT == S_ISVTX); +#else +#error "S_ISVTX / S_ISTXT isn't defined" +#endif + Assert(RTFS_UNIX_IRWXU == S_IRWXU); + Assert(RTFS_UNIX_IRUSR == S_IRUSR); + Assert(RTFS_UNIX_IWUSR == S_IWUSR); + Assert(RTFS_UNIX_IXUSR == S_IXUSR); + Assert(RTFS_UNIX_IRWXG == S_IRWXG); + Assert(RTFS_UNIX_IRGRP == S_IRGRP); + Assert(RTFS_UNIX_IWGRP == S_IWGRP); + Assert(RTFS_UNIX_IXGRP == S_IXGRP); + Assert(RTFS_UNIX_IRWXO == S_IRWXO); + Assert(RTFS_UNIX_IROTH == S_IROTH); + Assert(RTFS_UNIX_IWOTH == S_IWOTH); + Assert(RTFS_UNIX_IXOTH == S_IXOTH); + Assert(RTFS_TYPE_FIFO == S_IFIFO); + Assert(RTFS_TYPE_DEV_CHAR == S_IFCHR); + Assert(RTFS_TYPE_DIRECTORY == S_IFDIR); + Assert(RTFS_TYPE_DEV_BLOCK == S_IFBLK); + Assert(RTFS_TYPE_FILE == S_IFREG); + Assert(RTFS_TYPE_SYMLINK == S_IFLNK); + Assert(RTFS_TYPE_SOCKET == S_IFSOCK); +#ifdef S_IFWHT + Assert(RTFS_TYPE_WHITEOUT == S_IFWHT); +#endif + Assert(RTFS_TYPE_MASK == S_IFMT); + + pObjInfo->Attr.fMode = rtFsModeFromUnix(fMode, pszName, cbName, 0); + + /* additional unix attribs */ + pObjInfo->Attr.enmAdditional = RTFSOBJATTRADD_UNIX; + pObjInfo->Attr.u.Unix.uid = pStat->st_uid; + pObjInfo->Attr.u.Unix.gid = pStat->st_gid; + pObjInfo->Attr.u.Unix.cHardlinks = pStat->st_nlink; + pObjInfo->Attr.u.Unix.INodeIdDevice = pStat->st_dev; + pObjInfo->Attr.u.Unix.INodeId = pStat->st_ino; +#ifdef HAVE_STAT_FLAGS + pObjInfo->Attr.u.Unix.fFlags = pStat->st_flags; +#else + pObjInfo->Attr.u.Unix.fFlags = 0; +#endif +#ifdef HAVE_STAT_GEN + pObjInfo->Attr.u.Unix.GenerationId = pStat->st_gen; +#else + pObjInfo->Attr.u.Unix.GenerationId = 0; +#endif + pObjInfo->Attr.u.Unix.Device = pStat->st_rdev; +} + diff --git a/src/VBox/Runtime/r3/posix/fs3-posix.cpp b/src/VBox/Runtime/r3/posix/fs3-posix.cpp new file mode 100644 index 00000000..368ae8e3 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/fs3-posix.cpp @@ -0,0 +1,94 @@ +/* $Id: fs3-posix.cpp $ */ +/** @file + * IPRT - File System Helpers, POSIX, Part 3. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/fs.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include <iprt/string.h> +#include "internal/fs.h" + +#include <sys/time.h> +#include <grp.h> +#include <pwd.h> + + +/** + * Set user-owner additional attributes. + * + * @param pObjInfo The object info to fill add attrs for. + * @param uid The user id. + */ +void rtFsObjInfoAttrSetUnixOwner(PRTFSOBJINFO pObjInfo, RTUID uid) +{ + pObjInfo->Attr.enmAdditional = RTFSOBJATTRADD_UNIX_OWNER; + pObjInfo->Attr.u.UnixOwner.uid = uid; + pObjInfo->Attr.u.UnixOwner.szName[0] = '\0'; + + char achBuf[_4K]; + struct passwd Pwd; + struct passwd *pPwd; + int rc = getpwuid_r(uid, &Pwd, achBuf, sizeof(achBuf), &pPwd); + if (!rc && pPwd) + RTStrCopy(pObjInfo->Attr.u.UnixOwner.szName, sizeof(pObjInfo->Attr.u.UnixOwner.szName), pPwd->pw_name); +} + + +/** + * Set user-group additional attributes. + * + * @param pObjInfo The object info to fill add attrs for. + * @param gid The group id. + */ +void rtFsObjInfoAttrSetUnixGroup(PRTFSOBJINFO pObjInfo, RTUID gid) +{ + pObjInfo->Attr.enmAdditional = RTFSOBJATTRADD_UNIX_GROUP; + pObjInfo->Attr.u.UnixGroup.gid = gid; + pObjInfo->Attr.u.UnixGroup.szName[0] = '\0'; + + char achBuf[_4K]; + struct group Grp; + struct group *pGrp; + + int rc = getgrgid_r(gid, &Grp, achBuf, sizeof(achBuf), &pGrp); + if (!rc && pGrp) + RTStrCopy(pObjInfo->Attr.u.UnixGroup.szName, sizeof(pObjInfo->Attr.u.UnixGroup.szName), pGrp->gr_name); +} + diff --git a/src/VBox/Runtime/r3/posix/ldrNative-posix.cpp b/src/VBox/Runtime/r3/posix/ldrNative-posix.cpp new file mode 100644 index 00000000..9b33a0ff --- /dev/null +++ b/src/VBox/Runtime/r3/posix/ldrNative-posix.cpp @@ -0,0 +1,207 @@ +/* $Id: ldrNative-posix.cpp $ */ +/** @file + * IPRT - Binary Image Loader, POSIX native. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_LDR +#include <dlfcn.h> + +#include <iprt/ldr.h> +#include <iprt/assert.h> +#include <iprt/path.h> +#include <iprt/alloca.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include "internal/ldr.h" + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +#if defined(RT_OS_OS2) || defined(RT_OS_WINDOWS) +static const char g_szSuff[] = ".DLL"; +#elif defined(RT_OS_L4) +static const char g_szSuff[] = ".s.so"; +#elif defined(RT_OS_DARWIN) +static const char g_szSuff[] = ".dylib"; +#else +static const char g_szSuff[] = ".so"; +#endif + + +DECLHIDDEN(int) rtldrNativeLoad(const char *pszFilename, uintptr_t *phHandle, uint32_t fFlags, PRTERRINFO pErrInfo) +{ + /* + * Do we need to add an extension? + */ + if (!RTPathHasSuffix(pszFilename) && !(fFlags & RTLDRLOAD_FLAGS_NO_SUFFIX)) + { + size_t cch = strlen(pszFilename); + char *psz = (char *)alloca(cch + sizeof(g_szSuff)); + if (!psz) + return RTErrInfoSet(pErrInfo, VERR_NO_MEMORY, "alloca failed"); + memcpy(psz, pszFilename, cch); + memcpy(psz + cch, g_szSuff, sizeof(g_szSuff)); + pszFilename = psz; + } + + /* + * Attempt load. + */ + int fFlagsNative = RTLD_NOW; + if (fFlags & RTLDRLOAD_FLAGS_GLOBAL) + fFlagsNative |= RTLD_GLOBAL; + else + fFlagsNative |= RTLD_LOCAL; + void *pvMod = dlopen(pszFilename, fFlagsNative); + if (pvMod) + { + *phHandle = (uintptr_t)pvMod; + return VINF_SUCCESS; + } + + const char *pszDlError = dlerror(); + RTErrInfoSet(pErrInfo, VERR_FILE_NOT_FOUND, pszDlError); + LogRel(("rtldrNativeLoad: dlopen('%s', RTLD_NOW | RTLD_LOCAL) failed: %s\n", pszFilename, pszDlError)); + return VERR_FILE_NOT_FOUND; +} + + +DECLCALLBACK(int) rtldrNativeGetSymbol(PRTLDRMODINTERNAL pMod, const char *pszSymbol, void **ppvValue) +{ + PRTLDRMODNATIVE pModNative = (PRTLDRMODNATIVE)pMod; +#ifdef RT_OS_OS2 + /* Prefix the symbol with an underscore (assuming __cdecl/gcc-default). */ + size_t cch = strlen(pszSymbol); + char *psz = (char *)alloca(cch + 2); + psz[0] = '_'; + memcpy(psz + 1, pszSymbol, cch + 1); + pszSymbol = psz; +#endif + *ppvValue = dlsym((void *)pModNative->hNative, pszSymbol); + if (*ppvValue) + return VINF_SUCCESS; + return VERR_SYMBOL_NOT_FOUND; +} + + +DECLCALLBACK(int) rtldrNativeClose(PRTLDRMODINTERNAL pMod) +{ + PRTLDRMODNATIVE pModNative = (PRTLDRMODNATIVE)pMod; +#ifdef __SANITIZE_ADDRESS__ + /* If we are compiled with enabled address sanitizer (gcc/llvm), don't + * unload the module to prevent <unknown module> in the stack trace */ + pModNative->fFlags |= RTLDRLOAD_FLAGS_NO_UNLOAD; +#endif + if ( (pModNative->fFlags & RTLDRLOAD_FLAGS_NO_UNLOAD) + || !dlclose((void *)pModNative->hNative)) + { + pModNative->hNative = (uintptr_t)0; + return VINF_SUCCESS; + } + Log(("rtldrNativeFree: dlclose(%p) failed: %s\n", pModNative->hNative, dlerror())); + return VERR_GENERAL_FAILURE; +} + + +DECLHIDDEN(int) rtldrNativeLoadSystem(const char *pszFilename, const char *pszExt, uint32_t fFlags, PRTLDRMOD phLdrMod) +{ + /* + * For the present we ASSUME that we can trust dlopen to load what we want + * when not specifying a path. There seems to be very little we can do to + * restrict the places dlopen will search for library without doing + * auditing (linux) or something like that. + */ + Assert(strchr(pszFilename, '/') == NULL); + + uint32_t const fFlags2 = fFlags & ~(RTLDRLOAD_FLAGS_SO_VER_BEGIN_MASK | RTLDRLOAD_FLAGS_SO_VER_END_MASK); + + /* + * If no suffix is given and we haven't got any RTLDRLOAD_FLAGS_SO_VER_ range to work + * with, we can call RTLdrLoadEx directly. + */ + if (!pszExt) + { +#if !defined(RT_OS_DARWIN) && !defined(RT_OS_OS2) && !defined(RT_OS_WINDOWS) + if ( (fFlags & RTLDRLOAD_FLAGS_SO_VER_BEGIN_MASK) >> RTLDRLOAD_FLAGS_SO_VER_BEGIN_SHIFT + == (fFlags & RTLDRLOAD_FLAGS_SO_VER_END_MASK) >> RTLDRLOAD_FLAGS_SO_VER_END_SHIFT) +#endif + return RTLdrLoadEx(pszFilename, phLdrMod, fFlags2, NULL); + pszExt = ""; + } + + /* + * Combine filename and suffix and then do the loading. + */ + size_t const cchFilename = strlen(pszFilename); + size_t const cchSuffix = strlen(pszExt); + char *pszTmp = (char *)alloca(cchFilename + cchSuffix + 16 + 1); + memcpy(pszTmp, pszFilename, cchFilename); + memcpy(&pszTmp[cchFilename], pszExt, cchSuffix); + pszTmp[cchFilename + cchSuffix] = '\0'; + + int rc = RTLdrLoadEx(pszTmp, phLdrMod, fFlags2, NULL); + +#if !defined(RT_OS_DARWIN) && !defined(RT_OS_OS2) && !defined(RT_OS_WINDOWS) + /* + * If no version was given after the .so and do .so.MAJOR search according + * to the range in the fFlags. + */ + if (RT_FAILURE(rc) && !(fFlags & RTLDRLOAD_FLAGS_NO_SUFFIX)) + { + const char *pszActualSuff = RTPathSuffix(pszTmp); + if (pszActualSuff && strcmp(pszActualSuff, ".so") == 0) + { + int32_t const iBegin = (fFlags & RTLDRLOAD_FLAGS_SO_VER_BEGIN_MASK) >> RTLDRLOAD_FLAGS_SO_VER_BEGIN_SHIFT; + int32_t const iEnd = (fFlags & RTLDRLOAD_FLAGS_SO_VER_END_MASK) >> RTLDRLOAD_FLAGS_SO_VER_END_SHIFT; + int32_t const iIncr = iBegin <= iEnd ? 1 : -1; + for (int32_t iMajorVer = iBegin; iMajorVer != iEnd; iMajorVer += iIncr) + { + RTStrPrintf(&pszTmp[cchFilename + cchSuffix], 16 + 1, ".%d", iMajorVer); + rc = RTLdrLoadEx(pszTmp, phLdrMod, fFlags2, NULL); + if (RT_SUCCESS(rc)) + break; + } + } + } +#endif + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/localipc-posix.cpp b/src/VBox/Runtime/r3/posix/localipc-posix.cpp new file mode 100644 index 00000000..131d78cf --- /dev/null +++ b/src/VBox/Runtime/r3/posix/localipc-posix.cpp @@ -0,0 +1,1172 @@ +/* $Id: localipc-posix.cpp $ */ +/** @file + * IPRT - Local IPC Server & Client, Posix. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_LOCALIPC +#include "internal/iprt.h" +#include <iprt/localipc.h> + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/ctype.h> +#include <iprt/critsect.h> +#include <iprt/err.h> +#include <iprt/mem.h> +#include <iprt/log.h> +#include <iprt/poll.h> +#include <iprt/socket.h> +#include <iprt/string.h> +#include <iprt/time.h> +#include <iprt/path.h> + +#include <sys/types.h> +#include <sys/socket.h> +#include <sys/un.h> +#ifndef RT_OS_OS2 +# include <sys/poll.h> +#endif +#include <errno.h> +#include <fcntl.h> +#include <signal.h> +#include <unistd.h> +#include <sys/stat.h> +#ifndef AF_LOCAL +# define AF_LOCAL AF_UNIX +#endif + +#include "internal/magics.h" +#include "internal/path.h" +#include "internal/socket.h" + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** + * Local IPC service instance, POSIX. + */ +typedef struct RTLOCALIPCSERVERINT +{ + /** The magic (RTLOCALIPCSERVER_MAGIC). */ + uint32_t u32Magic; + /** The creation flags. */ + uint32_t fFlags; + /** Critical section protecting the structure. */ + RTCRITSECT CritSect; + /** The number of references to the instance. */ + uint32_t volatile cRefs; + /** Indicates that there is a pending cancel request. */ + bool volatile fCancelled; + /** The server socket. */ + RTSOCKET hSocket; + /** Thread currently listening for clients. */ + RTTHREAD hListenThread; + /** The name we bound the server to (native charset encoding). */ + struct sockaddr_un Name; +} RTLOCALIPCSERVERINT; +/** Pointer to a local IPC server instance (POSIX). */ +typedef RTLOCALIPCSERVERINT *PRTLOCALIPCSERVERINT; + + +/** + * Local IPC session instance, POSIX. + */ +typedef struct RTLOCALIPCSESSIONINT +{ + /** The magic (RTLOCALIPCSESSION_MAGIC). */ + uint32_t u32Magic; + /** Critical section protecting the structure. */ + RTCRITSECT CritSect; + /** The number of references to the instance. */ + uint32_t volatile cRefs; + /** Indicates that there is a pending cancel request. */ + bool volatile fCancelled; + /** Set if this is the server side, clear if the client. */ + bool fServerSide; + /** The client socket. */ + RTSOCKET hSocket; + /** Thread currently doing read related activites. */ + RTTHREAD hWriteThread; + /** Thread currently doing write related activies. */ + RTTHREAD hReadThread; +} RTLOCALIPCSESSIONINT; +/** Pointer to a local IPC session instance (Windows). */ +typedef RTLOCALIPCSESSIONINT *PRTLOCALIPCSESSIONINT; + + +/** Local IPC name prefix for portable names. */ +#define RTLOCALIPC_POSIX_NAME_PREFIX "/tmp/.iprt-localipc-" + + +/** + * Validates the user specified name. + * + * @returns IPRT status code. + * @param pszName The name to validate. + * @param fNative Whether it's a native name or a portable name. + */ +static int rtLocalIpcPosixValidateName(const char *pszName, bool fNative) +{ + AssertPtrReturn(pszName, VERR_INVALID_POINTER); + AssertReturn(*pszName, VERR_INVALID_NAME); + + if (!fNative) + { + for (;;) + { + char ch = *pszName++; + if (!ch) + break; + AssertReturn(!RT_C_IS_CNTRL(ch), VERR_INVALID_NAME); + AssertReturn((unsigned)ch < 0x80, VERR_INVALID_NAME); + AssertReturn(ch != '\\', VERR_INVALID_NAME); + AssertReturn(ch != '/', VERR_INVALID_NAME); + } + } + else + { + int rc = RTStrValidateEncoding(pszName); + AssertRCReturn(rc, rc); + } + + return VINF_SUCCESS; +} + + +/** + * Constructs a local (unix) domain socket name. + * + * @returns IPRT status code. + * @param pAddr The address structure to construct the name in. + * @param pcbAddr Where to return the address size. + * @param pszName The user specified name (valid). + * @param fNative Whether it's a native name or a portable name. + */ +static int rtLocalIpcPosixConstructName(struct sockaddr_un *pAddr, uint8_t *pcbAddr, const char *pszName, bool fNative) +{ + const char *pszNativeName; + int rc = rtPathToNative(&pszNativeName, pszName, NULL /*pszBasePath not support*/); + if (RT_SUCCESS(rc)) + { + size_t cchNativeName = strlen(pszNativeName); + size_t cbFull = !fNative ? cchNativeName + sizeof(RTLOCALIPC_POSIX_NAME_PREFIX) : cchNativeName + 1; + if (cbFull <= sizeof(pAddr->sun_path)) + { + RT_ZERO(*pAddr); +#ifdef RT_OS_OS2 /* Size must be exactly right on OS/2. */ + *pcbAddr = sizeof(*pAddr); +#else + *pcbAddr = RT_UOFFSETOF(struct sockaddr_un, sun_path) + (uint8_t)cbFull; +#endif +#ifdef HAVE_SUN_LEN_MEMBER + pAddr->sun_len = *pcbAddr; +#endif + pAddr->sun_family = AF_LOCAL; + + if (!fNative) + { + memcpy(pAddr->sun_path, RTLOCALIPC_POSIX_NAME_PREFIX, sizeof(RTLOCALIPC_POSIX_NAME_PREFIX) - 1); + memcpy(&pAddr->sun_path[sizeof(RTLOCALIPC_POSIX_NAME_PREFIX) - 1], pszNativeName, cchNativeName + 1); + } + else + memcpy(pAddr->sun_path, pszNativeName, cchNativeName + 1); + } + else + rc = VERR_FILENAME_TOO_LONG; + rtPathFreeNative(pszNativeName, pszName); + } + return rc; +} + + + +RTDECL(int) RTLocalIpcServerCreate(PRTLOCALIPCSERVER phServer, const char *pszName, uint32_t fFlags) +{ + /* + * Parameter validation. + */ + AssertPtrReturn(phServer, VERR_INVALID_POINTER); + *phServer = NIL_RTLOCALIPCSERVER; + AssertReturn(!(fFlags & ~RTLOCALIPC_FLAGS_VALID_MASK), VERR_INVALID_FLAGS); + int rc = rtLocalIpcPosixValidateName(pszName, RT_BOOL(fFlags & RTLOCALIPC_FLAGS_NATIVE_NAME)); + if (RT_SUCCESS(rc)) + { + /* + * Allocate memory for the instance and initialize it. + */ + PRTLOCALIPCSERVERINT pThis = (PRTLOCALIPCSERVERINT)RTMemAllocZ(sizeof(*pThis)); + if (pThis) + { + pThis->u32Magic = RTLOCALIPCSERVER_MAGIC; + pThis->fFlags = fFlags; + pThis->cRefs = 1; + pThis->fCancelled = false; + pThis->hListenThread = NIL_RTTHREAD; + rc = RTCritSectInit(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + /* + * Create the local (unix) socket and bind to it. + */ + rc = rtSocketCreate(&pThis->hSocket, AF_LOCAL, SOCK_STREAM, 0 /*iProtocol*/, false /*fInheritable*/); + if (RT_SUCCESS(rc)) + { + signal(SIGPIPE, SIG_IGN); /* Required on solaris, at least. */ + + uint8_t cbAddr; + rc = rtLocalIpcPosixConstructName(&pThis->Name, &cbAddr, pszName, + RT_BOOL(fFlags & RTLOCALIPC_FLAGS_NATIVE_NAME)); + if (RT_SUCCESS(rc)) + { + rc = rtSocketBindRawAddr(pThis->hSocket, &pThis->Name, cbAddr); + if (rc == VERR_NET_ADDRESS_IN_USE) + { + unlink(pThis->Name.sun_path); + rc = rtSocketBindRawAddr(pThis->hSocket, &pThis->Name, cbAddr); + } + if (RT_SUCCESS(rc)) + { + rc = rtSocketListen(pThis->hSocket, 16); + if (RT_SUCCESS(rc)) + { + LogFlow(("RTLocalIpcServerCreate: Created %p (%s)\n", pThis, pThis->Name.sun_path)); + *phServer = pThis; + return VINF_SUCCESS; + } + unlink(pThis->Name.sun_path); + } + } + RTSocketRelease(pThis->hSocket); + } + RTCritSectDelete(&pThis->CritSect); + } + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + } + Log(("RTLocalIpcServerCreate: failed, rc=%Rrc\n", rc)); + return rc; +} + + +RTDECL(int) RTLocalIpcServerGrantGroupAccess(RTLOCALIPCSERVER hServer, RTGID gid) +{ + PRTLOCALIPCSERVERINT pThis = (PRTLOCALIPCSERVERINT)hServer; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSERVER_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(pThis->Name.sun_path[0] != '\0', VERR_INVALID_STATE); + + if (chown(pThis->Name.sun_path, (uid_t)-1, gid) == 0) + { + if (chmod(pThis->Name.sun_path, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP) == 0) + { + LogRel2(("RTLocalIpcServerGrantGroupAccess: IPC socket %s access has been granted to group %RTgid\n", + pThis->Name.sun_path, gid)); + return VINF_SUCCESS; + } + LogRel(("RTLocalIpcServerGrantGroupAccess: cannot grant IPC socket %s write permission to group %RTgid: errno=%d\n", + pThis->Name.sun_path, gid, errno)); + } + else + LogRel(("RTLocalIpcServerGrantGroupAccess: cannot change IPC socket %s group ownership to %RTgid: errno=%d\n", + pThis->Name.sun_path, gid, errno)); + return RTErrConvertFromErrno(errno); +} + + +RTDECL(int) RTLocalIpcServerSetAccessMode(RTLOCALIPCSERVER hServer, RTFMODE fMode) +{ + PRTLOCALIPCSERVERINT pThis = (PRTLOCALIPCSERVERINT)hServer; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSERVER_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(pThis->Name.sun_path[0] != '\0', VERR_INVALID_STATE); + + if (chmod(pThis->Name.sun_path, fMode & RTFS_UNIX_ALL_ACCESS_PERMS) == 0) + return VINF_SUCCESS; + + return RTErrConvertFromErrno(errno); +} + + +/** + * Retains a reference to the server instance. + * + * @returns + * @param pThis The server instance. + */ +DECLINLINE(void) rtLocalIpcServerRetain(PRTLOCALIPCSERVERINT pThis) +{ + uint32_t cRefs = ASMAtomicIncU32(&pThis->cRefs); + Assert(cRefs < UINT32_MAX / 2 && cRefs); RT_NOREF_PV(cRefs); +} + + +/** + * Server instance destructor. + * + * @returns VINF_OBJECT_DESTROYED + * @param pThis The server instance. + */ +static int rtLocalIpcServerDtor(PRTLOCALIPCSERVERINT pThis) +{ + pThis->u32Magic = ~RTLOCALIPCSERVER_MAGIC; + if (RTSocketRelease(pThis->hSocket) == 0) + Log(("rtLocalIpcServerDtor: Released socket\n")); + else + Log(("rtLocalIpcServerDtor: Socket still has references (impossible?)\n")); + RTCritSectDelete(&pThis->CritSect); + unlink(pThis->Name.sun_path); + RTMemFree(pThis); + return VINF_OBJECT_DESTROYED; +} + + +/** + * Releases a reference to the server instance. + * + * @returns VINF_SUCCESS if only release, VINF_OBJECT_DESTROYED if destroyed. + * @param pThis The server instance. + */ +DECLINLINE(int) rtLocalIpcServerRelease(PRTLOCALIPCSERVERINT pThis) +{ + uint32_t cRefs = ASMAtomicDecU32(&pThis->cRefs); + Assert(cRefs < UINT32_MAX / 2); + if (!cRefs) + return rtLocalIpcServerDtor(pThis); + return VINF_SUCCESS; +} + + +/** + * The core of RTLocalIpcServerCancel, used by both the destroy and cancel APIs. + * + * @returns IPRT status code + * @param pThis The server instance. + */ +static int rtLocalIpcServerCancel(PRTLOCALIPCSERVERINT pThis) +{ + RTCritSectEnter(&pThis->CritSect); + pThis->fCancelled = true; + Log(("rtLocalIpcServerCancel:\n")); + if (pThis->hListenThread != NIL_RTTHREAD) + RTThreadPoke(pThis->hListenThread); + RTCritSectLeave(&pThis->CritSect); + return VINF_SUCCESS; +} + + + +RTDECL(int) RTLocalIpcServerDestroy(RTLOCALIPCSERVER hServer) +{ + /* + * Validate input. + */ + if (hServer == NIL_RTLOCALIPCSERVER) + return VINF_SUCCESS; + PRTLOCALIPCSERVERINT pThis = (PRTLOCALIPCSERVERINT)hServer; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSERVER_MAGIC, VERR_INVALID_HANDLE); + + /* + * Invalidate the server, releasing the caller's reference to the instance + * data and making sure any other thread in the listen API will wake up. + */ + AssertReturn(ASMAtomicCmpXchgU32(&pThis->u32Magic, ~RTLOCALIPCSERVER_MAGIC, RTLOCALIPCSERVER_MAGIC), VERR_WRONG_ORDER); + + rtLocalIpcServerCancel(pThis); + return rtLocalIpcServerRelease(pThis); +} + + +RTDECL(int) RTLocalIpcServerCancel(RTLOCALIPCSERVER hServer) +{ + /* + * Validate input. + */ + PRTLOCALIPCSERVERINT pThis = (PRTLOCALIPCSERVERINT)hServer; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSERVER_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the job. + */ + rtLocalIpcServerRetain(pThis); + rtLocalIpcServerCancel(pThis); + rtLocalIpcServerRelease(pThis); + return VINF_SUCCESS; +} + + +RTDECL(int) RTLocalIpcServerListen(RTLOCALIPCSERVER hServer, PRTLOCALIPCSESSION phClientSession) +{ + /* + * Validate input. + */ + PRTLOCALIPCSERVERINT pThis = (PRTLOCALIPCSERVERINT)hServer; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSERVER_MAGIC, VERR_INVALID_HANDLE); + + /* + * Begin listening. + */ + rtLocalIpcServerRetain(pThis); + int rc = RTCritSectEnter(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + if (pThis->hListenThread == NIL_RTTHREAD) + { + pThis->hListenThread = RTThreadSelf(); + + /* + * The listening retry loop. + */ + for (;;) + { + if (!pThis->fCancelled) + { + rc = RTCritSectLeave(&pThis->CritSect); + AssertRCBreak(rc); + + struct sockaddr_un Addr; + size_t cbAddr = sizeof(Addr); + RTSOCKET hClient; + Log(("RTLocalIpcServerListen: Calling rtSocketAccept...\n")); + rc = rtSocketAccept(pThis->hSocket, &hClient, (struct sockaddr *)&Addr, &cbAddr); + Log(("RTLocalIpcServerListen: rtSocketAccept returns %Rrc.\n", rc)); + + int rc2 = RTCritSectEnter(&pThis->CritSect); + AssertRCBreakStmt(rc2, rc = RT_SUCCESS(rc) ? rc2 : rc); + + if (RT_SUCCESS(rc)) + { + /* + * Create a client session. + */ + PRTLOCALIPCSESSIONINT pSession = (PRTLOCALIPCSESSIONINT)RTMemAllocZ(sizeof(*pSession)); + if (pSession) + { + pSession->u32Magic = RTLOCALIPCSESSION_MAGIC; + pSession->cRefs = 1; + pSession->fCancelled = false; + pSession->fServerSide = true; + pSession->hSocket = hClient; + pSession->hReadThread = NIL_RTTHREAD; + pSession->hWriteThread = NIL_RTTHREAD; + rc = RTCritSectInit(&pSession->CritSect); + if (RT_SUCCESS(rc)) + { + Log(("RTLocalIpcServerListen: Returning new client session: %p\n", pSession)); + *phClientSession = pSession; + break; + } + + RTMemFree(pSession); + } + else + rc = VERR_NO_MEMORY; + } + else if ( rc != VERR_INTERRUPTED + && rc != VERR_TRY_AGAIN) + break; + } + else + { + rc = VERR_CANCELLED; + break; + } + } + + pThis->hListenThread = NIL_RTTHREAD; + } + else + { + AssertFailed(); + rc = VERR_RESOURCE_BUSY; + } + int rc2 = RTCritSectLeave(&pThis->CritSect); + AssertStmt(RT_SUCCESS(rc2), rc = RT_SUCCESS(rc) ? rc2 : rc); + } + rtLocalIpcServerRelease(pThis); + + Log(("RTLocalIpcServerListen: returns %Rrc\n", rc)); + return rc; +} + + +RTDECL(int) RTLocalIpcSessionConnect(PRTLOCALIPCSESSION phSession, const char *pszName, uint32_t fFlags) +{ + /* + * Parameter validation. + */ + AssertPtrReturn(phSession, VERR_INVALID_POINTER); + *phSession = NIL_RTLOCALIPCSESSION; + + AssertReturn(!(fFlags & ~RTLOCALIPC_C_FLAGS_VALID_MASK), VERR_INVALID_FLAGS); + + int rc = rtLocalIpcPosixValidateName(pszName, RT_BOOL(fFlags & RTLOCALIPC_C_FLAGS_NATIVE_NAME)); + if (RT_SUCCESS(rc)) + { + /* + * Allocate memory for the instance and initialize it. + */ + PRTLOCALIPCSESSIONINT pThis = (PRTLOCALIPCSESSIONINT)RTMemAllocZ(sizeof(*pThis)); + if (pThis) + { + pThis->u32Magic = RTLOCALIPCSESSION_MAGIC; + pThis->cRefs = 1; + pThis->fCancelled = false; + pThis->fServerSide = false; + pThis->hSocket = NIL_RTSOCKET; + pThis->hReadThread = NIL_RTTHREAD; + pThis->hWriteThread = NIL_RTTHREAD; + rc = RTCritSectInit(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + /* + * Create the local (unix) socket and try connect to the server. + */ + rc = rtSocketCreate(&pThis->hSocket, AF_LOCAL, SOCK_STREAM, 0 /*iProtocol*/, false /*fInheritable*/); + if (RT_SUCCESS(rc)) + { + signal(SIGPIPE, SIG_IGN); /* Required on solaris, at least. */ + + struct sockaddr_un Addr; + uint8_t cbAddr; + rc = rtLocalIpcPosixConstructName(&Addr, &cbAddr, pszName, RT_BOOL(fFlags & RTLOCALIPC_C_FLAGS_NATIVE_NAME)); + if (RT_SUCCESS(rc)) + { + rc = rtSocketConnectRaw(pThis->hSocket, &Addr, cbAddr); + if (RT_SUCCESS(rc)) + { + *phSession = pThis; + Log(("RTLocalIpcSessionConnect: Returns new session %p\n", pThis)); + return VINF_SUCCESS; + } + } + RTSocketRelease(pThis->hSocket); + } + RTCritSectDelete(&pThis->CritSect); + } + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + } + Log(("RTLocalIpcSessionConnect: returns %Rrc\n", rc)); + return rc; +} + + +/** + * Retains a reference to the session instance. + * + * @param pThis The server instance. + */ +DECLINLINE(void) rtLocalIpcSessionRetain(PRTLOCALIPCSESSIONINT pThis) +{ + uint32_t cRefs = ASMAtomicIncU32(&pThis->cRefs); + Assert(cRefs < UINT32_MAX / 2 && cRefs); RT_NOREF_PV(cRefs); +} + + +RTDECL(uint32_t) RTLocalIpcSessionRetain(RTLOCALIPCSESSION hSession) +{ + PRTLOCALIPCSESSIONINT pThis = (PRTLOCALIPCSESSIONINT)hSession; + AssertPtrReturn(pThis, UINT32_MAX); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, UINT32_MAX); + + uint32_t cRefs = ASMAtomicIncU32(&pThis->cRefs); + Assert(cRefs < UINT32_MAX / 2 && cRefs); + return cRefs; +} + + +/** + * Session instance destructor. + * + * @returns VINF_OBJECT_DESTROYED + * @param pThis The server instance. + */ +static int rtLocalIpcSessionDtor(PRTLOCALIPCSESSIONINT pThis) +{ + pThis->u32Magic = ~RTLOCALIPCSESSION_MAGIC; + if (RTSocketRelease(pThis->hSocket) == 0) + Log(("rtLocalIpcSessionDtor: Released socket\n")); + else + Log(("rtLocalIpcSessionDtor: Socket still has references (impossible?)\n")); + RTCritSectDelete(&pThis->CritSect); + RTMemFree(pThis); + return VINF_OBJECT_DESTROYED; +} + + +/** + * Releases a reference to the session instance. + * + * @returns VINF_SUCCESS or VINF_OBJECT_DESTROYED as appropriate. + * @param pThis The session instance. + */ +DECLINLINE(int) rtLocalIpcSessionRelease(PRTLOCALIPCSESSIONINT pThis) +{ + uint32_t cRefs = ASMAtomicDecU32(&pThis->cRefs); + Assert(cRefs < UINT32_MAX / 2); + if (!cRefs) + return rtLocalIpcSessionDtor(pThis); + Log(("rtLocalIpcSessionRelease: %u refs left\n", cRefs)); + return VINF_SUCCESS; +} + + +RTDECL(uint32_t) RTLocalIpcSessionRelease(RTLOCALIPCSESSION hSession) +{ + if (hSession == NIL_RTLOCALIPCSESSION) + return 0; + + PRTLOCALIPCSESSIONINT pThis = (PRTLOCALIPCSESSIONINT)hSession; + AssertPtrReturn(pThis, UINT32_MAX); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, UINT32_MAX); + + uint32_t cRefs = ASMAtomicDecU32(&pThis->cRefs); + Assert(cRefs < UINT32_MAX / 2); + if (cRefs) + Log(("RTLocalIpcSessionRelease: %u refs left\n", cRefs)); + else + rtLocalIpcSessionDtor(pThis); + return cRefs; +} + + +/** + * The core of RTLocalIpcSessionCancel, used by both the destroy and cancel APIs. + * + * @returns IPRT status code + * @param pThis The session instance. + */ +static int rtLocalIpcSessionCancel(PRTLOCALIPCSESSIONINT pThis) +{ + RTCritSectEnter(&pThis->CritSect); + pThis->fCancelled = true; + Log(("rtLocalIpcSessionCancel:\n")); + if (pThis->hReadThread != NIL_RTTHREAD) + RTThreadPoke(pThis->hReadThread); + if (pThis->hWriteThread != NIL_RTTHREAD) + RTThreadPoke(pThis->hWriteThread); + RTCritSectLeave(&pThis->CritSect); + return VINF_SUCCESS; +} + + +RTDECL(int) RTLocalIpcSessionClose(RTLOCALIPCSESSION hSession) +{ + /* + * Validate input. + */ + if (hSession == NIL_RTLOCALIPCSESSION) + return VINF_SUCCESS; + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * Invalidate the session, releasing the caller's reference to the instance + * data and making sure any other thread in the listen API will wake up. + */ + Log(("RTLocalIpcSessionClose:\n")); + + rtLocalIpcSessionCancel(pThis); + return rtLocalIpcSessionRelease(pThis); +} + + +RTDECL(int) RTLocalIpcSessionCancel(RTLOCALIPCSESSION hSession) +{ + /* + * Validate input. + */ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the job. + */ + rtLocalIpcSessionRetain(pThis); + rtLocalIpcSessionCancel(pThis); + rtLocalIpcSessionRelease(pThis); + return VINF_SUCCESS; +} + + +/** + * Checks if the socket has has a HUP condition after reading zero bytes. + * + * @returns true if HUP, false if no. + * @param pThis The IPC session handle. + */ +static bool rtLocalIpcPosixHasHup(PRTLOCALIPCSESSIONINT pThis) +{ + int fdNative = RTSocketToNative(pThis->hSocket); + +#if !defined(RT_OS_OS2) && !defined(RT_OS_SOLARIS) + struct pollfd PollFd; + RT_ZERO(PollFd); + PollFd.fd = fdNative; + PollFd.events = POLLHUP | POLLERR; + if (poll(&PollFd, 1, 0) <= 0) + return false; + if (!(PollFd.revents & (POLLHUP | POLLERR))) + return false; +#else /* RT_OS_OS2 || RT_OS_SOLARIS */ + /* + * OS/2: No native poll, do zero byte send to check for EPIPE. + * Solaris: We don't get POLLHUP. + */ + uint8_t bDummy; + ssize_t rcSend = send(fdNative, &bDummy, 0, 0); + if (rcSend >= 0 || (errno != EPIPE && errno != ECONNRESET)) + return false; +#endif /* RT_OS_OS2 || RT_OS_SOLARIS */ + + /* + * We've established EPIPE. Now make sure there aren't any last bytes to + * read that came in between the recv made by the caller and the disconnect. + */ + uint8_t bPeek; + ssize_t rcRecv = recv(fdNative, &bPeek, 1, MSG_DONTWAIT | MSG_PEEK); + return rcRecv <= 0; +} + + +RTDECL(int) RTLocalIpcSessionRead(RTLOCALIPCSESSION hSession, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + /* + * Validate input. + */ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the job. + */ + rtLocalIpcSessionRetain(pThis); + + int rc = RTCritSectEnter(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + if (pThis->hReadThread == NIL_RTTHREAD) + { + pThis->hReadThread = RTThreadSelf(); + + for (;;) + { + if (!pThis->fCancelled) + { + rc = RTCritSectLeave(&pThis->CritSect); + AssertRCBreak(rc); + + rc = RTSocketRead(pThis->hSocket, pvBuf, cbToRead, pcbRead); + + /* Detect broken pipe. */ + if (rc == VINF_SUCCESS) + { + if (!pcbRead || *pcbRead) + { /* likely */ } + else if (rtLocalIpcPosixHasHup(pThis)) + rc = VERR_BROKEN_PIPE; + } + else if (rc == VERR_NET_CONNECTION_RESET_BY_PEER || rc == VERR_NET_SHUTDOWN) + rc = VERR_BROKEN_PIPE; + + int rc2 = RTCritSectEnter(&pThis->CritSect); + AssertRCBreakStmt(rc2, rc = RT_SUCCESS(rc) ? rc2 : rc); + + if ( rc == VERR_INTERRUPTED + || rc == VERR_TRY_AGAIN) + continue; + } + else + rc = VERR_CANCELLED; + break; + } + + pThis->hReadThread = NIL_RTTHREAD; + } + int rc2 = RTCritSectLeave(&pThis->CritSect); + AssertStmt(RT_SUCCESS(rc2), rc = RT_SUCCESS(rc) ? rc2 : rc); + } + + rtLocalIpcSessionRelease(pThis); + return rc; +} + + +RTDECL(int) RTLocalIpcSessionReadNB(RTLOCALIPCSESSION hSession, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + /* + * Validate input. + */ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the job. + */ + rtLocalIpcSessionRetain(pThis); + + int rc = RTCritSectEnter(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + if (pThis->hReadThread == NIL_RTTHREAD) + { + pThis->hReadThread = RTThreadSelf(); /* not really required, but whatever. */ + + for (;;) + { + if (!pThis->fCancelled) + { + rc = RTSocketReadNB(pThis->hSocket, pvBuf, cbToRead, pcbRead); + + /* Detect broken pipe. */ + if (rc == VINF_SUCCESS) + { + if (!pcbRead || *pcbRead) + { /* likely */ } + else if (rtLocalIpcPosixHasHup(pThis)) + rc = VERR_BROKEN_PIPE; + } + else if (rc == VERR_NET_CONNECTION_RESET_BY_PEER || rc == VERR_NET_SHUTDOWN) + rc = VERR_BROKEN_PIPE; + + if (rc == VERR_INTERRUPTED) + continue; + } + else + rc = VERR_CANCELLED; + break; + } + + pThis->hReadThread = NIL_RTTHREAD; + } + int rc2 = RTCritSectLeave(&pThis->CritSect); + AssertStmt(RT_SUCCESS(rc2), rc = RT_SUCCESS(rc) ? rc2 : rc); + } + + rtLocalIpcSessionRelease(pThis); + return rc; +} + + +RTDECL(int) RTLocalIpcSessionWrite(RTLOCALIPCSESSION hSession, const void *pvBuf, size_t cbToWrite) +{ + /* + * Validate input. + */ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the job. + */ + rtLocalIpcSessionRetain(pThis); + + int rc = RTCritSectEnter(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + if (pThis->hWriteThread == NIL_RTTHREAD) + { + pThis->hWriteThread = RTThreadSelf(); + + for (;;) + { + if (!pThis->fCancelled) + { + rc = RTCritSectLeave(&pThis->CritSect); + AssertRCBreak(rc); + + rc = RTSocketWrite(pThis->hSocket, pvBuf, cbToWrite); + + int rc2 = RTCritSectEnter(&pThis->CritSect); + AssertRCBreakStmt(rc2, rc = RT_SUCCESS(rc) ? rc2 : rc); + + if ( rc == VERR_INTERRUPTED + || rc == VERR_TRY_AGAIN) + continue; + } + else + rc = VERR_CANCELLED; + break; + } + + pThis->hWriteThread = NIL_RTTHREAD; + } + int rc2 = RTCritSectLeave(&pThis->CritSect); + AssertStmt(RT_SUCCESS(rc2), rc = RT_SUCCESS(rc) ? rc2 : rc); + } + + rtLocalIpcSessionRelease(pThis); + return rc; +} + + +RTDECL(int) RTLocalIpcSessionFlush(RTLOCALIPCSESSION hSession) +{ + /* + * Validate input. + */ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * This is a no-op because apparently write doesn't return until the + * result is read. At least that's what the reply to a 2003-04-08 LKML + * posting title "fsync() on unix domain sockets?" indicates. + * + * For conformity, make sure there isn't any active writes concurrent to this call. + */ + rtLocalIpcSessionRetain(pThis); + + int rc = RTCritSectEnter(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + if (pThis->hWriteThread == NIL_RTTHREAD) + rc = RTCritSectLeave(&pThis->CritSect); + else + { + rc = RTCritSectLeave(&pThis->CritSect); + if (RT_SUCCESS(rc)) + rc = VERR_RESOURCE_BUSY; + } + } + + rtLocalIpcSessionRelease(pThis); + return rc; +} + + +RTDECL(int) RTLocalIpcSessionWaitForData(RTLOCALIPCSESSION hSession, uint32_t cMillies) +{ + /* + * Validate input. + */ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the job. + */ + rtLocalIpcSessionRetain(pThis); + + int rc = RTCritSectEnter(&pThis->CritSect); + if (RT_SUCCESS(rc)) + { + if (pThis->hReadThread == NIL_RTTHREAD) + { + pThis->hReadThread = RTThreadSelf(); + uint64_t const msStart = RTTimeMilliTS(); + RTMSINTERVAL const cMsOriginalTimeout = cMillies; + + for (;;) + { + if (!pThis->fCancelled) + { + rc = RTCritSectLeave(&pThis->CritSect); + AssertRCBreak(rc); + + uint32_t fEvents = 0; +#ifdef RT_OS_OS2 + /* This doesn't give us any error condition on hangup, so use HUP check. */ + Log(("RTLocalIpcSessionWaitForData: Calling RTSocketSelectOneEx...\n")); + rc = RTSocketSelectOneEx(pThis->hSocket, RTPOLL_EVT_READ | RTPOLL_EVT_ERROR, &fEvents, cMillies); + Log(("RTLocalIpcSessionWaitForData: RTSocketSelectOneEx returns %Rrc, fEvents=%#x\n", rc, fEvents)); + if (RT_SUCCESS(rc) && fEvents == RTPOLL_EVT_READ && rtLocalIpcPosixHasHup(pThis)) + rc = VERR_BROKEN_PIPE; +#else +/** @todo RTSocketPoll? */ + /* POLLHUP will be set on hangup. */ + struct pollfd PollFd; + RT_ZERO(PollFd); + PollFd.fd = RTSocketToNative(pThis->hSocket); + PollFd.events = POLLHUP | POLLERR | POLLIN; + Log(("RTLocalIpcSessionWaitForData: Calling poll...\n")); + int cFds = poll(&PollFd, 1, cMillies == RT_INDEFINITE_WAIT ? -1 : (int)cMillies); + if (cFds >= 1) + { + /* Linux & Darwin sets both POLLIN and POLLHUP when the pipe is + broken and but no more data to read. Google hints at NetBSD + returning more sane values (POLLIN till no more data, then + POLLHUP). Solairs OTOH, doesn't ever seem to return POLLHUP. */ + fEvents = RTPOLL_EVT_READ; + if ( (PollFd.revents & (POLLHUP | POLLERR)) + && !(PollFd.revents & POLLIN)) + fEvents = RTPOLL_EVT_ERROR; +# if defined(RT_OS_SOLARIS) + else if (PollFd.revents & POLLIN) +# else + else if ((PollFd.revents & (POLLIN | POLLHUP)) == (POLLIN | POLLHUP)) +# endif + { + /* Check if there is actually data available. */ + uint8_t bPeek; + ssize_t rcRecv = recv(PollFd.fd, &bPeek, 1, MSG_DONTWAIT | MSG_PEEK); + if (rcRecv <= 0) + fEvents = RTPOLL_EVT_ERROR; + } + rc = VINF_SUCCESS; + } + else if (rc == 0) + rc = VERR_TIMEOUT; + else + rc = RTErrConvertFromErrno(errno); + Log(("RTLocalIpcSessionWaitForData: poll returns %u (rc=%d), revents=%#x\n", cFds, rc, PollFd.revents)); +#endif + + int rc2 = RTCritSectEnter(&pThis->CritSect); + AssertRCBreakStmt(rc2, rc = RT_SUCCESS(rc) ? rc2 : rc); + + if (RT_SUCCESS(rc)) + { + if (pThis->fCancelled) + rc = VERR_CANCELLED; + else if (fEvents & RTPOLL_EVT_ERROR) + rc = VERR_BROKEN_PIPE; + } + else if ( rc == VERR_INTERRUPTED + || rc == VERR_TRY_AGAIN) + { + /* Recalc cMillies. */ + if (cMsOriginalTimeout != RT_INDEFINITE_WAIT) + { + uint64_t cMsElapsed = RTTimeMilliTS() - msStart; + cMillies = cMsElapsed >= cMsOriginalTimeout ? 0 : cMsOriginalTimeout - (RTMSINTERVAL)cMsElapsed; + } + continue; + } + } + else + rc = VERR_CANCELLED; + break; + } + + pThis->hReadThread = NIL_RTTHREAD; + } + int rc2 = RTCritSectLeave(&pThis->CritSect); + AssertStmt(RT_SUCCESS(rc2), rc = RT_SUCCESS(rc) ? rc2 : rc); + } + + rtLocalIpcSessionRelease(pThis); + return rc; +} + + +/** + * Get IPC session socket peer credentials. + * + * @returns IPRT status code. + * @param hSession IPC session handle. + * @param pProcess Where to return the remote peer's PID (can be NULL). + * @param pUid Where to return the remote peer's UID (can be NULL). + * @param pGid Where to return the remote peer's GID (can be NULL). + */ +static int rtLocalIpcSessionQueryUcred(RTLOCALIPCSESSION hSession, PRTPROCESS pProcess, PRTUID pUid, PRTGID pGid) +{ + PRTLOCALIPCSESSIONINT pThis = hSession; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTLOCALIPCSESSION_MAGIC, VERR_INVALID_HANDLE); + +#if defined(RT_OS_LINUX) + struct ucred PeerCred = { (pid_t)NIL_RTPROCESS, (uid_t)NIL_RTUID, (gid_t)NIL_RTGID }; + socklen_t cbPeerCred = sizeof(PeerCred); + + rtLocalIpcSessionRetain(pThis); + + int rc = RTCritSectEnter(&pThis->CritSect);; + if (RT_SUCCESS(rc)) + { + if (getsockopt(RTSocketToNative(pThis->hSocket), SOL_SOCKET, SO_PEERCRED, &PeerCred, &cbPeerCred) >= 0) + { + if (pProcess) + *pProcess = PeerCred.pid; + if (pUid) + *pUid = PeerCred.uid; + if (pGid) + *pGid = PeerCred.gid; + rc = VINF_SUCCESS; + } + else + rc = RTErrConvertFromErrno(errno); + + int rc2 = RTCritSectLeave(&pThis->CritSect); + AssertStmt(RT_SUCCESS(rc2), rc = RT_SUCCESS(rc) ? rc2 : rc); + } + + rtLocalIpcSessionRelease(pThis); + + return rc; + +#else + /** @todo Implement on other platforms too (mostly platform specific this). + * Solaris: getpeerucred? Darwin: LOCALPEERCRED or getpeereid? */ + RT_NOREF(pProcess, pUid, pGid); + return VERR_NOT_SUPPORTED; +#endif +} + + +RTDECL(int) RTLocalIpcSessionQueryProcess(RTLOCALIPCSESSION hSession, PRTPROCESS pProcess) +{ + return rtLocalIpcSessionQueryUcred(hSession, pProcess, NULL, NULL); +} + + +RTDECL(int) RTLocalIpcSessionQueryUserId(RTLOCALIPCSESSION hSession, PRTUID pUid) +{ + return rtLocalIpcSessionQueryUcred(hSession, NULL, pUid, NULL); +} + +RTDECL(int) RTLocalIpcSessionQueryGroupId(RTLOCALIPCSESSION hSession, PRTGID pGid) +{ + return rtLocalIpcSessionQueryUcred(hSession, NULL, NULL, pGid); +} + diff --git a/src/VBox/Runtime/r3/posix/path-posix.cpp b/src/VBox/Runtime/r3/posix/path-posix.cpp new file mode 100644 index 00000000..04462315 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/path-posix.cpp @@ -0,0 +1,418 @@ +/* $Id: path-posix.cpp $ */ +/** @file + * IPRT - Path Manipulation, POSIX, Part 1. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_PATH +#include <stdlib.h> +#include <limits.h> +#include <errno.h> +#include <unistd.h> +#include <sys/stat.h> +#include <sys/time.h> +#include <stdio.h> +#include <sys/types.h> +#include <pwd.h> + +#include <iprt/path.h> +#include <iprt/env.h> +#include <iprt/assert.h> +#include <iprt/mem.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include "internal/path.h" +#include "internal/process.h" +#include "internal/fs.h" + + + +RTDECL(int) RTPathReal(const char *pszPath, char *pszRealPath, size_t cchRealPath) +{ + /* + * Convert input. + */ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + /* + * On POSIX platforms the API doesn't take a length parameter, which makes it + * a little bit more work. + */ + char szTmpPath[PATH_MAX + 1]; + const char *psz = realpath(pszNativePath, szTmpPath); + if (psz) + rc = rtPathFromNativeCopy(pszRealPath, cchRealPath, szTmpPath, NULL); + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTPathReal(%p:{%s}, %p:{%s}, %u): returns %Rrc\n", pszPath, pszPath, + pszRealPath, RT_SUCCESS(rc) ? pszRealPath : "<failed>", cchRealPath, rc)); + return rc; +} + + +RTR3DECL(int) RTPathSetMode(const char *pszPath, RTFMODE fMode) +{ + AssertPtrReturn(pszPath, VERR_INVALID_POINTER); + AssertReturn(*pszPath, VERR_INVALID_PARAMETER); + + int rc; + fMode = rtFsModeNormalize(fMode, pszPath, 0, 0); + if (rtFsModeIsValidPermissions(fMode)) + { + char const *pszNativePath; + rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + if (chmod(pszNativePath, fMode & RTFS_UNIX_MASK) != 0) + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativePath, pszPath); + } + } + else + { + AssertMsgFailed(("Invalid file mode! %RTfmode\n", fMode)); + rc = VERR_INVALID_FMODE; + } + return rc; +} + + +/** + * Checks if two files are the one and same file. + */ +static bool rtPathSame(const char *pszNativeSrc, const char *pszNativeDst) +{ + struct stat SrcStat; + if (lstat(pszNativeSrc, &SrcStat)) + return false; + struct stat DstStat; + if (lstat(pszNativeDst, &DstStat)) + return false; + Assert(SrcStat.st_dev && DstStat.st_dev); + Assert(SrcStat.st_ino && DstStat.st_ino); + if ( SrcStat.st_dev == DstStat.st_dev + && SrcStat.st_ino == DstStat.st_ino + && (SrcStat.st_mode & S_IFMT) == (DstStat.st_mode & S_IFMT)) + return true; + return false; +} + + +/** + * Worker for RTPathRename, RTDirRename, RTFileRename. + * + * @returns IPRT status code. + * @param pszSrc The source path. + * @param pszDst The destination path. + * @param fRename The rename flags. + * @param fFileType The filetype. We use the RTFMODE filetypes here. If it's 0, + * anything goes. If it's RTFS_TYPE_DIRECTORY we'll check that the + * source is a directory. If Its RTFS_TYPE_FILE we'll check that it's + * not a directory (we are NOT checking whether it's a file). + */ +DECLHIDDEN(int) rtPathPosixRename(const char *pszSrc, const char *pszDst, unsigned fRename, RTFMODE fFileType) +{ + /* + * Convert the paths. + */ + char const *pszNativeSrc; + int rc = rtPathToNative(&pszNativeSrc, pszSrc, NULL); + if (RT_SUCCESS(rc)) + { + char const *pszNativeDst; + rc = rtPathToNative(&pszNativeDst, pszDst, NULL); + if (RT_SUCCESS(rc)) + { + /* + * Check that the source exists and that any types that's specified matches. + * We have to check this first to avoid getting errnous VERR_ALREADY_EXISTS + * errors from the next step. + * + * There are race conditions here (perhaps unlikely ones, but still), but I'm + * afraid there is little with can do to fix that. + */ + struct stat SrcStat; + if (lstat(pszNativeSrc, &SrcStat)) + rc = RTErrConvertFromErrno(errno); + else if (!fFileType) + rc = VINF_SUCCESS; + else if (RTFS_IS_DIRECTORY(fFileType)) + rc = S_ISDIR(SrcStat.st_mode) ? VINF_SUCCESS : VERR_NOT_A_DIRECTORY; + else + rc = S_ISDIR(SrcStat.st_mode) ? VERR_IS_A_DIRECTORY : VINF_SUCCESS; + if (RT_SUCCESS(rc)) + { + bool fSameFile = false; + + /* + * Check if the target exists, rename is rather destructive. + * We'll have to make sure we don't overwrite the source! + * Another race condition btw. + */ + struct stat DstStat; + if (lstat(pszNativeDst, &DstStat)) + rc = errno == ENOENT ? VINF_SUCCESS : RTErrConvertFromErrno(errno); + else + { + Assert(SrcStat.st_dev && DstStat.st_dev); + Assert(SrcStat.st_ino && DstStat.st_ino); + if ( SrcStat.st_dev == DstStat.st_dev + && SrcStat.st_ino == DstStat.st_ino + && (SrcStat.st_mode & S_IFMT) == (DstStat.st_mode & S_IFMT)) + { + /* + * It's likely that we're talking about the same file here. + * We should probably check paths or whatever, but for now this'll have to be enough. + */ + fSameFile = true; + } + if (fSameFile) + rc = VINF_SUCCESS; + else if (S_ISDIR(DstStat.st_mode) || !(fRename & RTPATHRENAME_FLAGS_REPLACE)) + rc = VERR_ALREADY_EXISTS; + else + rc = VINF_SUCCESS; + + } + if (RT_SUCCESS(rc)) + { + if (!rename(pszNativeSrc, pszNativeDst)) + rc = VINF_SUCCESS; + else if ( (fRename & RTPATHRENAME_FLAGS_REPLACE) + && (errno == ENOTDIR || errno == EEXIST)) + { + /* + * Check that the destination isn't a directory. + * Yet another race condition. + */ + if (rtPathSame(pszNativeSrc, pszNativeDst)) + { + rc = VINF_SUCCESS; + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): appears to be the same file... (errno=%d)\n", + pszSrc, pszDst, fRename, fFileType, errno)); + } + else + { + if (lstat(pszNativeDst, &DstStat)) + rc = errno != ENOENT ? RTErrConvertFromErrno(errno) : VINF_SUCCESS; + else if (S_ISDIR(DstStat.st_mode)) + rc = VERR_ALREADY_EXISTS; + else + rc = VINF_SUCCESS; + if (RT_SUCCESS(rc)) + { + if (!unlink(pszNativeDst)) + { + if (!rename(pszNativeSrc, pszNativeDst)) + rc = VINF_SUCCESS; + else + { + rc = RTErrConvertFromErrno(errno); + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): rename failed rc=%Rrc errno=%d\n", + pszSrc, pszDst, fRename, fFileType, rc, errno)); + } + } + else + { + rc = RTErrConvertFromErrno(errno); + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): failed to unlink dst rc=%Rrc errno=%d\n", + pszSrc, pszDst, fRename, fFileType, rc, errno)); + } + } + else + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): dst !dir check failed rc=%Rrc\n", + pszSrc, pszDst, fRename, fFileType, rc)); + } + } + else + { + rc = RTErrConvertFromErrno(errno); + if (errno == ENOTDIR) + rc = VERR_ALREADY_EXISTS; /* unless somebody is racing us, this is the right interpretation */ + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): rename failed rc=%Rrc errno=%d\n", + pszSrc, pszDst, fRename, fFileType, rc, errno)); + } + } + else + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): destination check failed rc=%Rrc errno=%d\n", + pszSrc, pszDst, fRename, fFileType, rc, errno)); + } + else + Log(("rtPathRename('%s', '%s', %#x ,%RTfmode): source type check failed rc=%Rrc errno=%d\n", + pszSrc, pszDst, fRename, fFileType, rc, errno)); + + rtPathFreeNative(pszNativeDst, pszDst); + } + rtPathFreeNative(pszNativeSrc, pszSrc); + } + return rc; +} + + +RTR3DECL(int) RTPathRename(const char *pszSrc, const char *pszDst, unsigned fRename) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszSrc, VERR_INVALID_POINTER); + AssertPtrReturn(pszDst, VERR_INVALID_POINTER); + AssertMsgReturn(*pszSrc, ("%p\n", pszSrc), VERR_INVALID_PARAMETER); + AssertMsgReturn(*pszDst, ("%p\n", pszDst), VERR_INVALID_PARAMETER); + AssertMsgReturn(!(fRename & ~RTPATHRENAME_FLAGS_REPLACE), ("%#x\n", fRename), VERR_INVALID_PARAMETER); + + /* + * Hand it to the worker. + */ + int rc = rtPathPosixRename(pszSrc, pszDst, fRename, 0); + + Log(("RTPathRename(%p:{%s}, %p:{%s}, %#x): returns %Rrc\n", pszSrc, pszSrc, pszDst, pszDst, fRename, rc)); + return rc; +} + + +RTR3DECL(int) RTPathUnlink(const char *pszPath, uint32_t fUnlink) +{ + RT_NOREF_PV(pszPath); RT_NOREF_PV(fUnlink); + return VERR_NOT_IMPLEMENTED; +} + + +RTDECL(bool) RTPathExists(const char *pszPath) +{ + return RTPathExistsEx(pszPath, RTPATH_F_FOLLOW_LINK); +} + + +RTDECL(bool) RTPathExistsEx(const char *pszPath, uint32_t fFlags) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszPath, false); + AssertReturn(*pszPath, false); + Assert(RTPATH_F_IS_VALID(fFlags, 0)); + + /* + * Convert the path and check if it exists using stat(). + */ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat Stat; + if (fFlags & RTPATH_F_FOLLOW_LINK) + rc = stat(pszNativePath, &Stat); + else + rc = lstat(pszNativePath, &Stat); + if (!rc) + rc = VINF_SUCCESS; + else + rc = VERR_GENERAL_FAILURE; + rtPathFreeNative(pszNativePath, pszPath); + } + return RT_SUCCESS(rc); +} + + +RTDECL(int) RTPathGetCurrent(char *pszPath, size_t cchPath) +{ + /* + * Try with a reasonably sized buffer first. + */ + char szNativeCurDir[RTPATH_MAX]; + if (getcwd(szNativeCurDir, sizeof(szNativeCurDir)) != NULL) + return rtPathFromNativeCopy(pszPath, cchPath, szNativeCurDir, NULL); + + /* + * Retry a few times with really big buffers if we failed because CWD is unreasonably long. + */ + int iErr = errno; + if (iErr != ERANGE) + return RTErrConvertFromErrno(iErr); + + size_t cbNativeTmp = RTPATH_BIG_MAX; + for (;;) + { + char *pszNativeTmp = (char *)RTMemTmpAlloc(cbNativeTmp); + if (!pszNativeTmp) + return VERR_NO_TMP_MEMORY; + if (getcwd(pszNativeTmp, cbNativeTmp) != NULL) + { + int rc = rtPathFromNativeCopy(pszPath, cchPath, pszNativeTmp, NULL); + RTMemTmpFree(pszNativeTmp); + return rc; + } + iErr = errno; + RTMemTmpFree(pszNativeTmp); + if (iErr != ERANGE) + return RTErrConvertFromErrno(iErr); + + cbNativeTmp += RTPATH_BIG_MAX; + if (cbNativeTmp > RTPATH_BIG_MAX * 4) + return VERR_FILENAME_TOO_LONG; + } +} + + +RTDECL(int) RTPathSetCurrent(const char *pszPath) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszPath, VERR_INVALID_POINTER); + AssertReturn(*pszPath, VERR_INVALID_PARAMETER); + + /* + * Change the directory. + */ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + if (chdir(pszNativePath)) + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativePath, pszPath); + } + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/path2-posix.cpp b/src/VBox/Runtime/r3/posix/path2-posix.cpp new file mode 100644 index 00000000..9af9512f --- /dev/null +++ b/src/VBox/Runtime/r3/posix/path2-posix.cpp @@ -0,0 +1,316 @@ +/* $Id: path2-posix.cpp $ */ +/** @file + * IPRT - Path Manipulation, POSIX, Part 2 - RTPathQueryInfo. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_PATH +#include <stdlib.h> +#include <limits.h> +#include <errno.h> +#include <unistd.h> +#include <sys/stat.h> +#include <sys/time.h> +#include <stdio.h> +#include <sys/types.h> + +#include <iprt/path.h> +#include <iprt/env.h> +#include <iprt/assert.h> +#include <iprt/string.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include "internal/path.h" +#include "internal/process.h" +#include "internal/fs.h" + + +RTR3DECL(int) RTPathQueryInfo(const char *pszPath, PRTFSOBJINFO pObjInfo, RTFSOBJATTRADD enmAdditionalAttribs) +{ + return RTPathQueryInfoEx(pszPath, pObjInfo, enmAdditionalAttribs, RTPATH_F_ON_LINK); +} + + +RTR3DECL(int) RTPathQueryInfoEx(const char *pszPath, PRTFSOBJINFO pObjInfo, RTFSOBJATTRADD enmAdditionalAttribs, uint32_t fFlags) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszPath, VERR_INVALID_POINTER); + AssertReturn(*pszPath, VERR_INVALID_PARAMETER); + AssertPtrReturn(pObjInfo, VERR_INVALID_POINTER); + AssertMsgReturn( enmAdditionalAttribs >= RTFSOBJATTRADD_NOTHING + && enmAdditionalAttribs <= RTFSOBJATTRADD_LAST, + ("Invalid enmAdditionalAttribs=%p\n", enmAdditionalAttribs), + VERR_INVALID_PARAMETER); + AssertMsgReturn(RTPATH_F_IS_VALID(fFlags, 0), ("%#x\n", fFlags), VERR_INVALID_PARAMETER); + + /* + * Convert the filename. + */ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + struct stat Stat; + if (fFlags & RTPATH_F_FOLLOW_LINK) + rc = stat(pszNativePath, &Stat); + else + rc = lstat(pszNativePath, &Stat); /** @todo how doesn't have lstat again? */ + if (!rc) + { + rtFsConvertStatToObjInfo(pObjInfo, &Stat, pszPath, 0); + switch (enmAdditionalAttribs) + { + case RTFSOBJATTRADD_NOTHING: + case RTFSOBJATTRADD_UNIX: + Assert(pObjInfo->Attr.enmAdditional == RTFSOBJATTRADD_UNIX); + break; + + case RTFSOBJATTRADD_UNIX_OWNER: + rtFsObjInfoAttrSetUnixOwner(pObjInfo, Stat.st_uid); + break; + + case RTFSOBJATTRADD_UNIX_GROUP: + rtFsObjInfoAttrSetUnixGroup(pObjInfo, Stat.st_gid); + break; + + case RTFSOBJATTRADD_EASIZE: + /** @todo Use SGI extended attribute interface to query EA info. */ + pObjInfo->Attr.enmAdditional = RTFSOBJATTRADD_EASIZE; + pObjInfo->Attr.u.EASize.cb = 0; + break; + + default: + AssertMsgFailed(("Impossible!\n")); + return VERR_INTERNAL_ERROR; + } + } + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTPathQueryInfoEx(%p:{%s}, pObjInfo=%p, %d): returns %Rrc\n", + pszPath, pszPath, pObjInfo, enmAdditionalAttribs, rc)); + return rc; +} + + +RTR3DECL(int) RTPathSetTimes(const char *pszPath, PCRTTIMESPEC pAccessTime, PCRTTIMESPEC pModificationTime, + PCRTTIMESPEC pChangeTime, PCRTTIMESPEC pBirthTime) +{ + return RTPathSetTimesEx(pszPath, pAccessTime, pModificationTime, pChangeTime, pBirthTime, RTPATH_F_ON_LINK); +} + + +RTR3DECL(int) RTPathSetTimesEx(const char *pszPath, PCRTTIMESPEC pAccessTime, PCRTTIMESPEC pModificationTime, + PCRTTIMESPEC pChangeTime, PCRTTIMESPEC pBirthTime, uint32_t fFlags) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszPath, VERR_INVALID_POINTER); + AssertReturn(*pszPath, VERR_INVALID_PARAMETER); + AssertPtrNullReturn(pAccessTime, VERR_INVALID_POINTER); + AssertPtrNullReturn(pModificationTime, VERR_INVALID_POINTER); + AssertPtrNullReturn(pChangeTime, VERR_INVALID_POINTER); + AssertPtrNullReturn(pBirthTime, VERR_INVALID_POINTER); + AssertMsgReturn(RTPATH_F_IS_VALID(fFlags, 0), ("%#x\n", fFlags), VERR_INVALID_PARAMETER); + + /* + * Convert the paths. + */ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + RTFSOBJINFO ObjInfo; + + /* + * If it's a no-op, we'll only verify the existance of the file. + */ + if (!pAccessTime && !pModificationTime) + rc = RTPathQueryInfoEx(pszPath, &ObjInfo, RTFSOBJATTRADD_NOTHING, fFlags); + else + { + /* + * Convert the input to timeval, getting the missing one if necessary, + * and call the API which does the change. + */ + struct timeval aTimevals[2]; + if (pAccessTime && pModificationTime) + { + RTTimeSpecGetTimeval(pAccessTime, &aTimevals[0]); + RTTimeSpecGetTimeval(pModificationTime, &aTimevals[1]); + } + else + { + rc = RTPathQueryInfoEx(pszPath, &ObjInfo, RTFSOBJATTRADD_UNIX, fFlags); + if (RT_SUCCESS(rc)) + { + RTTimeSpecGetTimeval(pAccessTime ? pAccessTime : &ObjInfo.AccessTime, &aTimevals[0]); + RTTimeSpecGetTimeval(pModificationTime ? pModificationTime : &ObjInfo.ModificationTime, &aTimevals[1]); + } + else + Log(("RTPathSetTimes('%s',%p,%p,,): RTPathQueryInfo failed with %Rrc\n", + pszPath, pAccessTime, pModificationTime, rc)); + } + if (RT_SUCCESS(rc)) + { + if (fFlags & RTPATH_F_FOLLOW_LINK) + { + if (utimes(pszNativePath, aTimevals)) + rc = RTErrConvertFromErrno(errno); + } +#if (defined(RT_OS_DARWIN) && MAC_OS_X_VERSION_MIN_REQUIRED >= 1050) \ + || defined(RT_OS_FREEBSD) \ + || defined(RT_OS_LINUX) \ + || defined(RT_OS_OS2) /** @todo who really has lutimes? */ + else + { + if (lutimes(pszNativePath, aTimevals)) + { + /* If lutimes is not supported (e.g. linux < 2.6.22), try fall back on utimes: */ + if (errno != ENOSYS) + rc = RTErrConvertFromErrno(errno); + else + { + if (pAccessTime && pModificationTime) + rc = RTPathQueryInfoEx(pszPath, &ObjInfo, RTFSOBJATTRADD_UNIX, fFlags); + if (RT_SUCCESS(rc) && !RTFS_IS_SYMLINK(ObjInfo.Attr.fMode)) + { + if (utimes(pszNativePath, aTimevals)) + rc = RTErrConvertFromErrno(errno); + } + else + rc = VERR_NOT_SUPPORTED; + } + } + } +#else + else + { + if (pAccessTime && pModificationTime) + rc = RTPathQueryInfoEx(pszPath, &ObjInfo, RTFSOBJATTRADD_UNIX, fFlags); + if (RT_SUCCESS(rc) && RTFS_IS_SYMLINK(ObjInfo.Attr.fMode)) + rc = VERR_NS_SYMLINK_SET_TIME; + else if (RT_SUCCESS(rc)) + { + if (utimes(pszNativePath, aTimevals)) + rc = RTErrConvertFromErrno(errno); + } + } +#endif + if (RT_FAILURE(rc)) + Log(("RTPathSetTimes('%s',%p,%p,,): failed with %Rrc and errno=%d\n", + pszPath, pAccessTime, pModificationTime, rc, errno)); + } + } + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTPathSetTimes(%p:{%s}, %p:{%RDtimespec}, %p:{%RDtimespec}, %p:{%RDtimespec}, %p:{%RDtimespec}): return %Rrc\n", + pszPath, pszPath, pAccessTime, pAccessTime, pModificationTime, pModificationTime, + pChangeTime, pChangeTime, pBirthTime, pBirthTime, rc)); + return rc; +} + + +RTR3DECL(int) RTPathSetOwner(const char *pszPath, uint32_t uid, uint32_t gid) +{ + return RTPathSetOwnerEx(pszPath, uid, gid, RTPATH_F_ON_LINK); +} + + +RTR3DECL(int) RTPathSetOwnerEx(const char *pszPath, uint32_t uid, uint32_t gid, uint32_t fFlags) +{ + /* + * Validate input. + */ + AssertPtrReturn(pszPath, VERR_INVALID_POINTER); + AssertReturn(*pszPath, VERR_INVALID_PARAMETER); + AssertMsgReturn(RTPATH_F_IS_VALID(fFlags, 0), ("%#x\n", fFlags), VERR_INVALID_PARAMETER); + uid_t uidNative = uid != NIL_RTUID ? (uid_t)uid : (uid_t)-1; + AssertReturn(uid == uidNative, VERR_INVALID_PARAMETER); + gid_t gidNative = gid != NIL_RTGID ? (gid_t)gid : (uid_t)-1; + AssertReturn(gid == gidNative, VERR_INVALID_PARAMETER); + + /* + * Convert the path. + */ + char const *pszNativePath; + int rc = rtPathToNative(&pszNativePath, pszPath, NULL); + if (RT_SUCCESS(rc)) + { + if (fFlags & RTPATH_F_FOLLOW_LINK) + { + if (chown(pszNativePath, uidNative, gidNative)) + rc = RTErrConvertFromErrno(errno); + } +#if 1 + else + { + if (lchown(pszNativePath, uidNative, gidNative)) + rc = RTErrConvertFromErrno(errno); + } +#else + else + { + RTFSOBJINFO ObjInfo; + rc = RTPathQueryInfoEx(pszPath, &ObjInfo, RTFSOBJATTRADD_UNIX, fFlags); + if (RT_SUCCESS(rc) && RTFS_IS_SYMLINK(ObjInfo.Attr.fMode)) + rc = VERR_NS_SYMLINK_CHANGE_OWNER; + else if (RT_SUCCESS(rc)) + { + if (lchown(pszNativePath, uidNative, gidNative)) + rc = RTErrConvertFromErrno(errno); + } + } +#endif + if (RT_FAILURE(rc)) + Log(("RTPathSetOwnerEx('%s',%d,%d): failed with %Rrc and errno=%d\n", + pszPath, uid, gid, rc, errno)); + + rtPathFreeNative(pszNativePath, pszPath); + } + + LogFlow(("RTPathSetOwnerEx(%p:{%s}, uid=%d, gid=%d): return %Rrc\n", + pszPath, pszPath, uid, gid, rc)); + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/pathhost-posix.cpp b/src/VBox/Runtime/r3/posix/pathhost-posix.cpp new file mode 100644 index 00000000..b0c4fe46 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/pathhost-posix.cpp @@ -0,0 +1,294 @@ +/* $Id: pathhost-posix.cpp $ */ +/** @file + * IPRT - Path Conversions, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_PATH +#include "internal/iprt.h" +#include "internal/path.h" +#include "internal/string.h" +#include "internal/thread.h" + +#include <iprt/env.h> +#include <iprt/err.h> +#include <iprt/string.h> +#include <iprt/once.h> + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +/** Initialize once object. */ +static RTONCE g_OnceInitPathConv = RTONCE_INITIALIZER; +/** If set, then we can pass UTF-8 thru directly. */ +static bool g_fPassthruUtf8 = false; +/** The UTF-8 to FS iconv cache entry. */ +static RTSTRICONV g_enmUtf8ToFsIdx = RTSTRICONV_UTF8_TO_LOCALE; +/** The FS to UTF-8 iconv cache entry. */ +static RTSTRICONV g_enmFsToUtf8Idx = RTSTRICONV_LOCALE_TO_UTF8; +/** The codeset we're using. */ +static char g_szFsCodeset[32]; + + +/** + * Do a case insensitive compare where the 2nd string is known and can be case + * folded when writing the code. + * + * @returns see strcmp. + * @param pszStr1 The string to compare against pszLower and + * pszUpper. + * @param pszUpper The upper case edition of the 2nd string. + * @param pszLower The lower case edition of the 2nd string. + */ +static int rtPathStrICmp(const char *pszStr1, const char *pszUpper, const char *pszLower) +{ + Assert(strlen(pszLower) == strlen(pszUpper)); + for (;;) + { + char ch1 = *pszStr1++; + char ch2Upper = *pszUpper++; + char ch2Lower = *pszLower++; + if ( ch1 != ch2Upper + && ch1 != ch2Lower) + return ch1 < ch2Upper ? -1 : 1; + if (!ch1) + return 0; + } +} + +/** + * Is the specified codeset something we can treat as UTF-8. + * + * @returns true if we can do UTF-8 passthru, false if not. + * @param pszCodeset The codeset in question. + */ +static bool rtPathConvInitIsUtf8(const char *pszCodeset) +{ + /* Paranoia. */ + if (!pszCodeset) + return false; + + /* + * Avoid RTStrICmp at this point. + */ + static struct + { + const char *pszUpper; + const char *pszLower; + } const s_aUtf8Compatible[] = + { + /* The default locale. */ + { "C" , "c" }, + { "POSIX" , "posix" }, + /* 7-bit ASCII. */ + { "ANSI_X3.4-1968" , "ansi_x3.4-1968" }, + { "ANSI_X3.4-1986" , "ansi_x3.4-1986" }, + { "US-ASCII" , "us-ascii" }, + { "ISO646-US" , "iso646-us" }, + { "ISO_646.IRV:1991" , "iso_646.irv:1991" }, + { "ISO-IR-6" , "iso-ir-6" }, + { "IBM367" , "ibm367" }, + /* UTF-8 */ + { "UTF-8" , "utf-8" }, + { "UTF8" , "utf8" }, + { "ISO-10646/UTF-8" , "iso-10646/utf-8" }, + { "ISO-10646/UTF8" , "iso-10646/utf8" } + }; + + for (size_t i = 0; i < RT_ELEMENTS(s_aUtf8Compatible); i++) + if (!rtPathStrICmp(pszCodeset, s_aUtf8Compatible[i].pszUpper, s_aUtf8Compatible[i].pszLower)) + return true; + + return false; +} + + +/** + * Init once for the path conversion code. + * + * @returns IPRT status code. + * @param pvUser1 Unused. + * @param pvUser2 Unused. + */ +static DECLCALLBACK(int32_t) rtPathConvInitOnce(void *pvUser) +{ + /* + * Read the environment variable, no mercy on misconfigs here except that + * empty values are quietly ignored. (We use a temp buffer for stripping.) + */ + char *pszEnvValue = NULL; + char szEnvValue[sizeof(g_szFsCodeset)]; + int rc = RTEnvGetEx(RTENV_DEFAULT, RTPATH_CODESET_ENV_VAR, szEnvValue, sizeof(szEnvValue), NULL); + if (rc != VERR_ENV_VAR_NOT_FOUND && RT_FAILURE(rc)) + return rc; + if (RT_SUCCESS(rc)) + pszEnvValue = RTStrStrip(szEnvValue); + + if (pszEnvValue && *pszEnvValue) + { + g_fPassthruUtf8 = rtPathConvInitIsUtf8(pszEnvValue); + g_enmFsToUtf8Idx = RTSTRICONV_FS_TO_UTF8; + g_enmUtf8ToFsIdx = RTSTRICONV_UTF8_TO_FS; + strcpy(g_szFsCodeset, pszEnvValue); + } + else + { + const char *pszCodeset = rtStrGetLocaleCodeset(); + size_t cchCodeset = pszCodeset ? strlen(pszCodeset) : sizeof(g_szFsCodeset); + if (cchCodeset >= sizeof(g_szFsCodeset)) + /* This shouldn't happen, but we'll manage. */ + g_szFsCodeset[0] = '\0'; + else + { + memcpy(g_szFsCodeset, pszCodeset, cchCodeset + 1); + pszCodeset = g_szFsCodeset; + } + g_fPassthruUtf8 = rtPathConvInitIsUtf8(pszCodeset); + g_enmFsToUtf8Idx = RTSTRICONV_LOCALE_TO_UTF8; + g_enmUtf8ToFsIdx = RTSTRICONV_UTF8_TO_LOCALE; + } + + NOREF(pvUser); + return VINF_SUCCESS; +} + + +int rtPathToNative(char const **ppszNativePath, const char *pszPath, const char *pszBasePath) +{ + *ppszNativePath = NULL; + + int rc = RTOnce(&g_OnceInitPathConv, rtPathConvInitOnce, NULL); + if (RT_SUCCESS(rc)) + { + if (g_fPassthruUtf8 || !*pszPath) + *ppszNativePath = pszPath; + else + rc = rtStrConvert(pszPath, strlen(pszPath), "UTF-8", + (char **)ppszNativePath, 0, g_szFsCodeset, + 2, g_enmUtf8ToFsIdx); + } + NOREF(pszBasePath); /* We don't query the FS for codeset preferences. */ + return rc; +} + + +void rtPathFreeNative(char const *pszNativePath, const char *pszPath) +{ + if ( pszNativePath != pszPath + && pszNativePath) + RTStrFree((char *)pszNativePath); +} + + +int rtPathFromNative(const char **ppszPath, const char *pszNativePath, const char *pszBasePath) +{ + *ppszPath = NULL; + + int rc = RTOnce(&g_OnceInitPathConv, rtPathConvInitOnce, NULL); + if (RT_SUCCESS(rc)) + { + if (g_fPassthruUtf8 || !*pszNativePath) + { + size_t cCpsIgnored; + size_t cchNativePath; + rc = rtUtf8Length(pszNativePath, RTSTR_MAX, &cCpsIgnored, &cchNativePath); + if (RT_SUCCESS(rc)) + { + char *pszPath; + *ppszPath = pszPath = RTStrAlloc(cchNativePath + 1); + if (pszPath) + memcpy(pszPath, pszNativePath, cchNativePath + 1); + else + rc = VERR_NO_STR_MEMORY; + } + } + else + rc = rtStrConvert(pszNativePath, strlen(pszNativePath), g_szFsCodeset, + (char **)ppszPath, 0, "UTF-8", + 2, g_enmFsToUtf8Idx); + } + NOREF(pszBasePath); /* We don't query the FS for codeset preferences. */ + return rc; +} + + +void rtPathFreeIprt(const char *pszPath, const char *pszNativePath) +{ + if ( pszPath != pszNativePath + && pszPath) + RTStrFree((char *)pszPath); +} + + +int rtPathFromNativeCopy(char *pszPath, size_t cbPath, const char *pszNativePath, const char *pszBasePath) +{ + int rc = RTOnce(&g_OnceInitPathConv, rtPathConvInitOnce, NULL); + if (RT_SUCCESS(rc)) + { + if (g_fPassthruUtf8 || !*pszNativePath) + rc = RTStrCopy(pszPath, cbPath, pszNativePath); + else if (cbPath) + rc = rtStrConvert(pszNativePath, strlen(pszNativePath), g_szFsCodeset, + &pszPath, cbPath, "UTF-8", + 2, g_enmFsToUtf8Idx); + else + rc = VERR_BUFFER_OVERFLOW; + } + + NOREF(pszBasePath); /* We don't query the FS for codeset preferences. */ + return rc; +} + + +int rtPathFromNativeDup(char **ppszPath, const char *pszNativePath, const char *pszBasePath) +{ + int rc = RTOnce(&g_OnceInitPathConv, rtPathConvInitOnce, NULL); + if (RT_SUCCESS(rc)) + { + if (g_fPassthruUtf8 || !*pszNativePath) + rc = RTStrDupEx(ppszPath, pszNativePath); + else + rc = rtStrConvert(pszNativePath, strlen(pszNativePath), g_szFsCodeset, + ppszPath, 0, "UTF-8", + 2, g_enmFsToUtf8Idx); + } + + NOREF(pszBasePath); /* We don't query the FS for codeset preferences. */ + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/pipe-posix.cpp b/src/VBox/Runtime/r3/posix/pipe-posix.cpp new file mode 100644 index 00000000..dc673bd3 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/pipe-posix.cpp @@ -0,0 +1,754 @@ +/* $Id: pipe-posix.cpp $ */ +/** @file + * IPRT - Anonymous Pipes, POSIX Implementation. + */ + +/* + * Copyright (C) 2010-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/pipe.h> +#include "internal/iprt.h" + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/mem.h> +#include <iprt/poll.h> +#include <iprt/string.h> +#include <iprt/thread.h> +#include "internal/magics.h" + +#include <errno.h> +#include <fcntl.h> +#include <limits.h> +#include <unistd.h> +#include <sys/ioctl.h> +#include <sys/poll.h> +#include <sys/stat.h> +#include <signal.h> +#ifdef RT_OS_LINUX +# include <sys/syscall.h> +#endif +#ifdef RT_OS_SOLARIS +# include <sys/filio.h> +#endif + +#include "internal/pipe.h" + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +typedef struct RTPIPEINTERNAL +{ + /** Magic value (RTPIPE_MAGIC). */ + uint32_t u32Magic; + /** The file descriptor. */ + int fd; + /** Set if this is the read end, clear if it's the write end. */ + bool fRead; + /** RTPipeFromNative: Leave it open on RTPipeClose. */ + bool fLeaveOpen; + /** Atomically operated state variable. + * + * - Bits 0 thru 29 - Users of the new mode. + * - Bit 30 - The pipe mode, set indicates blocking. + * - Bit 31 - Set when we're switching the mode. + */ + uint32_t volatile u32State; +} RTPIPEINTERNAL; + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +/** @name RTPIPEINTERNAL::u32State defines + * @{ */ +#define RTPIPE_POSIX_BLOCKING UINT32_C(0x40000000) +#define RTPIPE_POSIX_SWITCHING UINT32_C(0x80000000) +#define RTPIPE_POSIX_SWITCHING_BIT 31 +#define RTPIPE_POSIX_USERS_MASK UINT32_C(0x3fffffff) +/** @} */ + + + +/** + * Wrapper for calling pipe2() or pipe(). + * + * When using pipe2() the returned handles are marked close-on-exec and does + * not risk racing process creation calls on other threads. + * + * @returns See pipe(). + * @param paFds See pipe(). + * @param piNewPipeSyscall Where to cache which call we should used. -1 if + * pipe(), 1 if pipe2(), 0 if not yet decided. + */ +static int my_pipe_wrapper(int *paFds, int *piNewPipeSyscall) +{ + if (*piNewPipeSyscall >= 0) + { +#if defined(RT_OS_LINUX) && defined(__NR_pipe2) && defined(O_CLOEXEC) + long rc = syscall(__NR_pipe2, paFds, O_CLOEXEC); + if (rc >= 0) + { + if (*piNewPipeSyscall == 0) + *piNewPipeSyscall = 1; + return (int)rc; + } +#endif + *piNewPipeSyscall = -1; + } + + return pipe(paFds); +} + + +RTDECL(int) RTPipeCreate(PRTPIPE phPipeRead, PRTPIPE phPipeWrite, uint32_t fFlags) +{ + AssertPtrReturn(phPipeRead, VERR_INVALID_POINTER); + AssertPtrReturn(phPipeWrite, VERR_INVALID_POINTER); + AssertReturn(!(fFlags & ~RTPIPE_C_VALID_MASK), VERR_INVALID_PARAMETER); + + /* + * Create the pipe and clear/set the close-on-exec flag as required. + */ + int aFds[2] = {-1, -1}; + static int s_iNewPipeSyscall = 0; + if (my_pipe_wrapper(aFds, &s_iNewPipeSyscall)) + return RTErrConvertFromErrno(errno); + + int rc = VINF_SUCCESS; + if (s_iNewPipeSyscall > 0) + { + /* created with close-on-exec set. */ + if (fFlags & RTPIPE_C_INHERIT_READ) + { + if (fcntl(aFds[0], F_SETFD, 0)) + rc = RTErrConvertFromErrno(errno); + } + + if (fFlags & RTPIPE_C_INHERIT_WRITE) + { + if (fcntl(aFds[1], F_SETFD, 0)) + rc = RTErrConvertFromErrno(errno); + } + } + else + { + /* created with close-on-exec cleared. */ + if (!(fFlags & RTPIPE_C_INHERIT_READ)) + { + if (fcntl(aFds[0], F_SETFD, FD_CLOEXEC)) + rc = RTErrConvertFromErrno(errno); + } + + if (!(fFlags & RTPIPE_C_INHERIT_WRITE)) + { + if (fcntl(aFds[1], F_SETFD, FD_CLOEXEC)) + rc = RTErrConvertFromErrno(errno); + } + } + + if (RT_SUCCESS(rc)) + { + /* + * Create the two handles. + */ + RTPIPEINTERNAL *pThisR = (RTPIPEINTERNAL *)RTMemAlloc(sizeof(RTPIPEINTERNAL)); + if (pThisR) + { + RTPIPEINTERNAL *pThisW = (RTPIPEINTERNAL *)RTMemAlloc(sizeof(RTPIPEINTERNAL)); + if (pThisW) + { + pThisR->u32Magic = RTPIPE_MAGIC; + pThisW->u32Magic = RTPIPE_MAGIC; + pThisR->fd = aFds[0]; + pThisW->fd = aFds[1]; + pThisR->fRead = true; + pThisW->fRead = false; + pThisR->fLeaveOpen = false; + pThisW->fLeaveOpen = false; + pThisR->u32State = RTPIPE_POSIX_BLOCKING; + pThisW->u32State = RTPIPE_POSIX_BLOCKING; + + *phPipeRead = pThisR; + *phPipeWrite = pThisW; + + /* + * Before we leave, make sure to shut up SIGPIPE. + */ + signal(SIGPIPE, SIG_IGN); + return VINF_SUCCESS; + } + + RTMemFree(pThisR); + rc = VERR_NO_MEMORY; + } + else + rc = VERR_NO_MEMORY; + } + + close(aFds[0]); + close(aFds[1]); + return rc; +} + + +RTDECL(int) RTPipeCloseEx(RTPIPE hPipe, bool fLeaveOpen) +{ + RTPIPEINTERNAL *pThis = hPipe; + if (pThis == NIL_RTPIPE) + return VINF_SUCCESS; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the cleanup. + */ + AssertReturn(ASMAtomicCmpXchgU32(&pThis->u32Magic, ~RTPIPE_MAGIC, RTPIPE_MAGIC), VERR_INVALID_HANDLE); + + int fd = pThis->fd; + pThis->fd = -1; + if (!fLeaveOpen && !pThis->fLeaveOpen) + close(fd); + + if (ASMAtomicReadU32(&pThis->u32State) & RTPIPE_POSIX_USERS_MASK) + { + AssertFailed(); + RTThreadSleep(1); + } + + RTMemFree(pThis); + + return VINF_SUCCESS; +} + + +RTDECL(int) RTPipeClose(RTPIPE hPipe) +{ + return RTPipeCloseEx(hPipe, false /*fLeaveOpen*/); +} + + +RTDECL(int) RTPipeFromNative(PRTPIPE phPipe, RTHCINTPTR hNativePipe, uint32_t fFlags) +{ + AssertPtrReturn(phPipe, VERR_INVALID_POINTER); + AssertReturn(!(fFlags & ~RTPIPE_N_VALID_MASK_FN), VERR_INVALID_PARAMETER); + AssertReturn(!!(fFlags & RTPIPE_N_READ) != !!(fFlags & RTPIPE_N_WRITE), VERR_INVALID_PARAMETER); + + /* + * Get and validate the pipe handle info. + */ + int hNative = (int)hNativePipe; + struct stat st; + AssertReturn(fstat(hNative, &st) == 0, RTErrConvertFromErrno(errno)); + AssertMsgReturn(S_ISFIFO(st.st_mode) || S_ISSOCK(st.st_mode), ("%#x (%o)\n", st.st_mode, st.st_mode), VERR_INVALID_HANDLE); + + int fFd = fcntl(hNative, F_GETFL, 0); + AssertReturn(fFd != -1, VERR_INVALID_HANDLE); + AssertMsgReturn( (fFd & O_ACCMODE) == (fFlags & RTPIPE_N_READ ? O_RDONLY : O_WRONLY) + || (fFd & O_ACCMODE) == O_RDWR /* Solaris creates bi-directional pipes. */ + , ("%#x\n", fFd), VERR_INVALID_HANDLE); + + /* + * Create the handle. + */ + RTPIPEINTERNAL *pThis = (RTPIPEINTERNAL *)RTMemAlloc(sizeof(RTPIPEINTERNAL)); + if (!pThis) + return VERR_NO_MEMORY; + + pThis->u32Magic = RTPIPE_MAGIC; + pThis->fd = hNative; + pThis->fRead = RT_BOOL(fFlags & RTPIPE_N_READ); + pThis->fLeaveOpen = RT_BOOL(fFlags & RTPIPE_N_LEAVE_OPEN); + pThis->u32State = fFd & O_NONBLOCK ? 0 : RTPIPE_POSIX_BLOCKING; + + /* + * Fix up inheritability and shut up SIGPIPE and we're done. + */ + if (fcntl(hNative, F_SETFD, fFlags & RTPIPE_N_INHERIT ? 0 : FD_CLOEXEC) == 0) + { + signal(SIGPIPE, SIG_IGN); + *phPipe = pThis; + return VINF_SUCCESS; + } + + int rc = RTErrConvertFromErrno(errno); + RTMemFree(pThis); + return rc; +} + + +RTDECL(RTHCINTPTR) RTPipeToNative(RTPIPE hPipe) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, -1); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, -1); + + return pThis->fd; +} + + +/** + * Prepare blocking mode. + * + * @returns VINF_SUCCESS + * @retval VERR_WRONG_ORDER + * @retval VERR_INTERNAL_ERROR_4 + * + * @param pThis The pipe handle. + */ +static int rtPipeTryBlocking(RTPIPEINTERNAL *pThis) +{ + /* + * Update the state. + */ + for (;;) + { + uint32_t u32State = ASMAtomicReadU32(&pThis->u32State); + uint32_t const u32StateOld = u32State; + uint32_t const cUsers = (u32State & RTPIPE_POSIX_USERS_MASK); + + if (u32State & RTPIPE_POSIX_BLOCKING) + { + AssertReturn(cUsers < RTPIPE_POSIX_USERS_MASK / 2, VERR_INTERNAL_ERROR_4); + u32State &= ~RTPIPE_POSIX_USERS_MASK; + u32State |= cUsers + 1; + if (ASMAtomicCmpXchgU32(&pThis->u32State, u32State, u32StateOld)) + { + if (u32State & RTPIPE_POSIX_SWITCHING) + break; + return VINF_SUCCESS; + } + } + else if (cUsers == 0) + { + u32State = 1 | RTPIPE_POSIX_SWITCHING | RTPIPE_POSIX_BLOCKING; + if (ASMAtomicCmpXchgU32(&pThis->u32State, u32State, u32StateOld)) + break; + } + else + return VERR_WRONG_ORDER; + ASMNopPause(); + } + + /* + * Do the switching. + */ + int fFlags = fcntl(pThis->fd, F_GETFL, 0); + if (fFlags != -1) + { + if ( !(fFlags & O_NONBLOCK) + || fcntl(pThis->fd, F_SETFL, fFlags & ~O_NONBLOCK) != -1) + { + ASMAtomicBitClear(&pThis->u32State, RTPIPE_POSIX_SWITCHING_BIT); + return VINF_SUCCESS; + } + } + + ASMAtomicDecU32(&pThis->u32State); + return RTErrConvertFromErrno(errno); +} + + +/** + * Prepare non-blocking mode. + * + * @returns VINF_SUCCESS + * @retval VERR_WRONG_ORDER + * @retval VERR_INTERNAL_ERROR_4 + * + * @param pThis The pipe handle. + */ +static int rtPipeTryNonBlocking(RTPIPEINTERNAL *pThis) +{ + /* + * Update the state. + */ + for (;;) + { + uint32_t u32State = ASMAtomicReadU32(&pThis->u32State); + uint32_t const u32StateOld = u32State; + uint32_t const cUsers = (u32State & RTPIPE_POSIX_USERS_MASK); + + if (!(u32State & RTPIPE_POSIX_BLOCKING)) + { + AssertReturn(cUsers < RTPIPE_POSIX_USERS_MASK / 2, VERR_INTERNAL_ERROR_4); + u32State &= ~RTPIPE_POSIX_USERS_MASK; + u32State |= cUsers + 1; + if (ASMAtomicCmpXchgU32(&pThis->u32State, u32State, u32StateOld)) + { + if (u32State & RTPIPE_POSIX_SWITCHING) + break; + return VINF_SUCCESS; + } + } + else if (cUsers == 0) + { + u32State = 1 | RTPIPE_POSIX_SWITCHING; + if (ASMAtomicCmpXchgU32(&pThis->u32State, u32State, u32StateOld)) + break; + } + else + return VERR_WRONG_ORDER; + ASMNopPause(); + } + + /* + * Do the switching. + */ + int fFlags = fcntl(pThis->fd, F_GETFL, 0); + if (fFlags != -1) + { + if ( (fFlags & O_NONBLOCK) + || fcntl(pThis->fd, F_SETFL, fFlags | O_NONBLOCK) != -1) + { + ASMAtomicBitClear(&pThis->u32State, RTPIPE_POSIX_SWITCHING_BIT); + return VINF_SUCCESS; + } + } + + ASMAtomicDecU32(&pThis->u32State); + return RTErrConvertFromErrno(errno); +} + + +/** + * Checks if the read pipe has a HUP condition. + * + * @returns true if HUP, false if no. + * @param pThis The pipe handle (read). + */ +static bool rtPipePosixHasHup(RTPIPEINTERNAL *pThis) +{ + Assert(pThis->fRead); + + struct pollfd PollFd; + RT_ZERO(PollFd); + PollFd.fd = pThis->fd; + PollFd.events = POLLHUP; + return poll(&PollFd, 1, 0) >= 1 + && (PollFd.revents & POLLHUP); +} + + +RTDECL(int) RTPipeRead(RTPIPE hPipe, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(pThis->fRead, VERR_ACCESS_DENIED); + AssertPtr(pcbRead); + AssertPtr(pvBuf); + + int rc = rtPipeTryNonBlocking(pThis); + if (RT_SUCCESS(rc)) + { + ssize_t cbRead = read(pThis->fd, pvBuf, RT_MIN(cbToRead, SSIZE_MAX)); + if (cbRead >= 0) + { + if (cbRead || !cbToRead || !rtPipePosixHasHup(pThis)) + *pcbRead = cbRead; + else + rc = VERR_BROKEN_PIPE; + } + else if (errno == EAGAIN) + { + *pcbRead = 0; + rc = VINF_TRY_AGAIN; + } + else + rc = RTErrConvertFromErrno(errno); + + ASMAtomicDecU32(&pThis->u32State); + } + return rc; +} + + +RTDECL(int) RTPipeReadBlocking(RTPIPE hPipe, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(pThis->fRead, VERR_ACCESS_DENIED); + AssertPtr(pvBuf); + + int rc = rtPipeTryBlocking(pThis); + if (RT_SUCCESS(rc)) + { + size_t cbTotalRead = 0; + while (cbToRead > 0) + { + ssize_t cbRead = read(pThis->fd, pvBuf, RT_MIN(cbToRead, SSIZE_MAX)); + if (cbRead < 0) + { + rc = RTErrConvertFromErrno(errno); + break; + } + if (!cbRead && rtPipePosixHasHup(pThis)) + { + rc = VERR_BROKEN_PIPE; + break; + } + + /* advance */ + pvBuf = (char *)pvBuf + cbRead; + cbTotalRead += cbRead; + cbToRead -= cbRead; + } + + if (pcbRead) + { + *pcbRead = cbTotalRead; + if ( RT_FAILURE(rc) + && cbTotalRead + && rc != VERR_INVALID_POINTER) + rc = VINF_SUCCESS; + } + + ASMAtomicDecU32(&pThis->u32State); + } + return rc; +} + + +RTDECL(int) RTPipeWrite(RTPIPE hPipe, const void *pvBuf, size_t cbToWrite, size_t *pcbWritten) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(!pThis->fRead, VERR_ACCESS_DENIED); + AssertPtr(pcbWritten); + AssertPtr(pvBuf); + + int rc = rtPipeTryNonBlocking(pThis); + if (RT_SUCCESS(rc)) + { + if (cbToWrite) + { + ssize_t cbWritten = write(pThis->fd, pvBuf, RT_MIN(cbToWrite, SSIZE_MAX)); + if (cbWritten >= 0) + *pcbWritten = cbWritten; + else if (errno == EAGAIN) + { + *pcbWritten = 0; + rc = VINF_TRY_AGAIN; + } + else + rc = RTErrConvertFromErrno(errno); + } + else + *pcbWritten = 0; + + ASMAtomicDecU32(&pThis->u32State); + } + return rc; +} + + +RTDECL(int) RTPipeWriteBlocking(RTPIPE hPipe, const void *pvBuf, size_t cbToWrite, size_t *pcbWritten) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(!pThis->fRead, VERR_ACCESS_DENIED); + AssertPtr(pvBuf); + AssertPtrNull(pcbWritten); + + int rc = rtPipeTryBlocking(pThis); + if (RT_SUCCESS(rc)) + { + size_t cbTotalWritten = 0; + while (cbToWrite > 0) + { + ssize_t cbWritten = write(pThis->fd, pvBuf, RT_MIN(cbToWrite, SSIZE_MAX)); + if (cbWritten < 0) + { + rc = RTErrConvertFromErrno(errno); + break; + } + + /* advance */ + pvBuf = (char const *)pvBuf + cbWritten; + cbTotalWritten += cbWritten; + cbToWrite -= cbWritten; + } + + if (pcbWritten) + { + *pcbWritten = cbTotalWritten; + if ( RT_FAILURE(rc) + && cbTotalWritten + && rc != VERR_INVALID_POINTER) + rc = VINF_SUCCESS; + } + + ASMAtomicDecU32(&pThis->u32State); + } + return rc; +} + + +RTDECL(int) RTPipeFlush(RTPIPE hPipe) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(!pThis->fRead, VERR_ACCESS_DENIED); + + if (fsync(pThis->fd)) + { + if (errno == EINVAL || errno == ENOTSUP) + return VERR_NOT_SUPPORTED; + return RTErrConvertFromErrno(errno); + } + return VINF_SUCCESS; +} + + +RTDECL(int) RTPipeSelectOne(RTPIPE hPipe, RTMSINTERVAL cMillies) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + + struct pollfd PollFd; + RT_ZERO(PollFd); + PollFd.fd = pThis->fd; + PollFd.events = POLLHUP | POLLERR; + if (pThis->fRead) + PollFd.events |= POLLIN | POLLPRI; + else + PollFd.events |= POLLOUT; + + int timeout; + if ( cMillies == RT_INDEFINITE_WAIT + || cMillies >= INT_MAX /* lazy bird */) + timeout = -1; + else + timeout = cMillies; + + int rc = poll(&PollFd, 1, timeout); + if (rc == -1) + return RTErrConvertFromErrno(errno); + return rc > 0 ? VINF_SUCCESS : VERR_TIMEOUT; +} + + +RTDECL(int) RTPipeQueryReadable(RTPIPE hPipe, size_t *pcbReadable) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(pThis->fRead, VERR_PIPE_NOT_READ); + AssertPtrReturn(pcbReadable, VERR_INVALID_POINTER); + + int cb = 0; + int rc = ioctl(pThis->fd, FIONREAD, &cb); + if (rc != -1) + { + AssertStmt(cb >= 0, cb = 0); + *pcbReadable = cb; + return VINF_SUCCESS; + } + + rc = errno; + if (rc == ENOTTY) + rc = VERR_NOT_SUPPORTED; + else + rc = RTErrConvertFromErrno(rc); + return rc; +} + + +RTDECL(int) RTPipeQueryInfo(RTPIPE hPipe, PRTFSOBJINFO pObjInfo, RTFSOBJATTRADD enmAddAttr) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, 0); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, 0); + + rtPipeFakeQueryInfo(pObjInfo, enmAddAttr, pThis->fRead); + + if (pThis->fRead) + { + int cb = 0; + int rc = ioctl(pThis->fd, FIONREAD, &cb); + if (rc >= 0) + pObjInfo->cbObject = cb; + } +#ifdef FIONSPACE + else + { + int cb = 0; + int rc = ioctl(pThis->fd, FIONSPACE, &cb); + if (rc >= 0) + pObjInfo->cbObject = cb; + } +#endif + + /** @todo Check this out on linux, solaris and darwin... (Currently going by a + * FreeBSD manpage.) */ + struct stat St; + if (fstat(pThis->fd, &St)) + { + pObjInfo->cbAllocated = St.st_blksize; + if ( enmAddAttr == RTFSOBJATTRADD_NOTHING + || enmAddAttr == RTFSOBJATTRADD_UNIX) + { + pObjInfo->Attr.enmAdditional = RTFSOBJATTRADD_UNIX; + pObjInfo->Attr.u.Unix.INodeId = St.st_ino; + pObjInfo->Attr.u.Unix.INodeIdDevice = St.st_dev; + } + } + /** @todo error handling? */ + + return VINF_SUCCESS; +} + + +int rtPipePollGetHandle(RTPIPE hPipe, uint32_t fEvents, PRTHCINTPTR phNative) +{ + RTPIPEINTERNAL *pThis = hPipe; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTPIPE_MAGIC, VERR_INVALID_HANDLE); + + AssertReturn(!(fEvents & RTPOLL_EVT_READ) || pThis->fRead, VERR_INVALID_PARAMETER); + AssertReturn(!(fEvents & RTPOLL_EVT_WRITE) || !pThis->fRead, VERR_INVALID_PARAMETER); + + *phNative = pThis->fd; + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/process-creation-posix.cpp b/src/VBox/Runtime/r3/posix/process-creation-posix.cpp new file mode 100644 index 00000000..6cf7be73 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/process-creation-posix.cpp @@ -0,0 +1,2408 @@ +/* $Id: process-creation-posix.cpp $ */ +/** @file + * IPRT - Process Creation, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_PROCESS +#include <iprt/cdefs.h> +#ifdef RT_OS_LINUX +# define IPRT_WITH_DYNAMIC_CRYPT_R +#endif +#if (defined(RT_OS_LINUX) || defined(RT_OS_OS2)) && !defined(_GNU_SOURCE) +# define _GNU_SOURCE +#endif +#if defined(RT_OS_LINUX) && !defined(_XOPEN_SOURCE) +# define _XOPEN_SOURCE 700 /* for newlocale */ +#endif + +#ifdef RT_OS_OS2 +# define crypt unistd_crypt +# define setkey unistd_setkey +# define encrypt unistd_encrypt +# include <unistd.h> +# undef crypt +# undef setkey +# undef encrypt +#else +# include <unistd.h> +#endif +#include <stdlib.h> +#include <errno.h> +#include <langinfo.h> +#include <locale.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <sys/wait.h> +#include <fcntl.h> +#include <signal.h> +#include <grp.h> +#include <pwd.h> +#if defined(RT_OS_LINUX) || defined(RT_OS_OS2) || defined(RT_OS_SOLARIS) +# include <crypt.h> +#endif +#if defined(RT_OS_LINUX) || defined(RT_OS_SOLARIS) +# include <shadow.h> +#endif +#if defined(RT_OS_DARWIN) +# include <xlocale.h> /* for newlocale() */ +#endif + +#if defined(RT_OS_LINUX) || defined(RT_OS_OS2) +/* While Solaris has posix_spawn() of course we don't want to use it as + * we need to have the child in a different process contract, no matter + * whether it is started detached or not. */ +# define HAVE_POSIX_SPAWN 1 +#endif +#if defined(RT_OS_DARWIN) && defined(MAC_OS_X_VERSION_MIN_REQUIRED) +# if MAC_OS_X_VERSION_MIN_REQUIRED >= 1050 +# define HAVE_POSIX_SPAWN 1 +# endif +#endif +#ifdef HAVE_POSIX_SPAWN +# include <spawn.h> +#endif + +#if !defined(IPRT_USE_PAM) \ + && !defined(IPRT_WITHOUT_PAM) \ + && ( defined(RT_OS_DARWIN) || defined(RT_OS_FREEBSD) || defined(RT_OS_LINUX) || defined(RT_OS_NETBSD) || defined(RT_OS_OPENBSD) || defined(RT_OS_SOLARIS) ) +# define IPRT_USE_PAM +#endif +#ifdef IPRT_USE_PAM +# include <security/pam_appl.h> +# include <stdlib.h> +# include <dlfcn.h> +# include <iprt/asm.h> +#endif + +#ifdef RT_OS_SOLARIS +# include <limits.h> +# include <sys/ctfs.h> +# include <sys/contract/process.h> +# include <libcontract.h> +#endif + +#ifndef RT_OS_SOLARIS +# include <paths.h> +#else +# define _PATH_MAILDIR "/var/mail" +# define _PATH_DEFPATH "/usr/bin:/bin" +# define _PATH_STDPATH "/sbin:/usr/sbin:/bin:/usr/bin" +#endif +#ifndef _PATH_BSHELL +# define _PATH_BSHELL "/bin/sh" +#endif + + +#include <iprt/process.h> +#include "internal/iprt.h" + +#include <iprt/alloca.h> +#include <iprt/assert.h> +#include <iprt/ctype.h> +#include <iprt/env.h> +#include <iprt/err.h> +#include <iprt/file.h> +#if defined(IPRT_WITH_DYNAMIC_CRYPT_R) || defined(IPRT_USE_PAM) +# include <iprt/ldr.h> +#endif +#include <iprt/log.h> +#include <iprt/path.h> +#include <iprt/pipe.h> +#include <iprt/socket.h> +#include <iprt/string.h> +#include <iprt/mem.h> +#include "internal/process.h" +#include "internal/path.h" +#include "internal/string.h" + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +#ifdef IPRT_USE_PAM +/* + * The PAM library names and version ranges to try. + */ +# ifdef RT_OS_DARWIN +# include <mach-o/dyld.h> +/** @node libpam.2.dylib was introduced with 10.6.x (OpenPAM); we use + * libpam.dylib as that's a symlink to the latest and greatest. */ +# define IPRT_LIBPAM_FILE_1 "libpam.dylib" +# define IPRT_LIBPAM_FILE_1_FIRST_VER 0 +# define IPRT_LIBPAM_FILE_1_END_VER 0 +# define IPRT_LIBPAM_FILE_2 "libpam.2.dylib" +# define IPRT_LIBPAM_FILE_2_FIRST_VER 0 +# define IPRT_LIBPAM_FILE_2_END_VER 0 +# define IPRT_LIBPAM_FILE_3 "libpam.1.dylib" +# define IPRT_LIBPAM_FILE_3_FIRST_VER 0 +# define IPRT_LIBPAM_FILE_3_END_VER 0 +# elif RT_OS_LINUX +# define IPRT_LIBPAM_FILE_1 "libpam.so.0" +# define IPRT_LIBPAM_FILE_1_FIRST_VER 0 +# define IPRT_LIBPAM_FILE_1_END_VER 0 +# define IPRT_LIBPAM_FILE_2 "libpam.so" +# define IPRT_LIBPAM_FILE_2_FIRST_VER 16 +# define IPRT_LIBPAM_FILE_2_END_VER 1 +# else +# define IPRT_LIBPAM_FILE_1 "libpam.so" +# define IPRT_LIBPAM_FILE_1_FIRST_VER 16 +# define IPRT_LIBPAM_FILE_1_END_VER 0 +# endif +#endif + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +#ifdef IPRT_USE_PAM +/** For passing info between rtCheckCredentials and rtPamConv. */ +typedef struct RTPROCPAMARGS +{ + const char *pszUser; + const char *pszPassword; +} RTPROCPAMARGS; +/** Pointer to rtPamConv argument package. */ +typedef RTPROCPAMARGS *PRTPROCPAMARGS; +#endif + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +/** Environment dump marker used with CSH. */ +static const char g_szEnvMarkerBegin[] = "IPRT_EnvEnvEnv_Begin_EnvEnvEnv"; +/** Environment dump marker used with CSH. */ +static const char g_szEnvMarkerEnd[] = "IPRT_EnvEnvEnv_End_EnvEnvEnv"; + + +/********************************************************************************************************************************* +* Internal Functions * +*********************************************************************************************************************************/ +static int rtProcPosixCreateInner(const char *pszExec, const char * const *papszArgs, RTENV hEnv, RTENV hEnvToUse, + uint32_t fFlags, const char *pszAsUser, uid_t uid, gid_t gid, + unsigned cRedirFds, int *paRedirFds, PRTPROCESS phProcess); + + +#ifdef IPRT_USE_PAM +/** + * Worker for rtCheckCredentials that feeds password and maybe username to PAM. + * + * @returns PAM status. + * @param cMessages Number of messages. + * @param papMessages Message vector. + * @param ppaResponses Where to put our responses. + * @param pvAppData Pointer to RTPROCPAMARGS. + */ +#if defined(RT_OS_SOLARIS) +static int rtPamConv(int cMessages, struct pam_message **papMessages, struct pam_response **ppaResponses, void *pvAppData) +#else +static int rtPamConv(int cMessages, const struct pam_message **papMessages, struct pam_response **ppaResponses, void *pvAppData) +#endif +{ + LogFlow(("rtPamConv: cMessages=%d\n", cMessages)); + PRTPROCPAMARGS pArgs = (PRTPROCPAMARGS)pvAppData; + AssertPtrReturn(pArgs, PAM_CONV_ERR); + + struct pam_response *paResponses = (struct pam_response *)calloc(cMessages, sizeof(paResponses[0])); + AssertReturn(paResponses, PAM_CONV_ERR); + for (int i = 0; i < cMessages; i++) + { + LogFlow(("rtPamConv: #%d: msg_style=%d msg=%s\n", i, papMessages[i]->msg_style, papMessages[i]->msg)); + + paResponses[i].resp_retcode = 0; + if (papMessages[i]->msg_style == PAM_PROMPT_ECHO_OFF) + paResponses[i].resp = strdup(pArgs->pszPassword); + else if (papMessages[i]->msg_style == PAM_PROMPT_ECHO_ON) + paResponses[i].resp = strdup(pArgs->pszUser); + else + { + paResponses[i].resp = NULL; + continue; + } + if (paResponses[i].resp == NULL) + { + while (i-- > 0) + free(paResponses[i].resp); + free(paResponses); + LogFlow(("rtPamConv: out of memory\n")); + return PAM_CONV_ERR; + } + } + + *ppaResponses = paResponses; + return PAM_SUCCESS; +} + + +/** + * Common PAM driver for rtCheckCredentials and the case where pszAsUser is NULL + * but RTPROC_FLAGS_PROFILE is set. + * + * @returns IPRT status code. + * @param pszPamService The PAM service to use for the run. + * @param pszUser The user. + * @param pszPassword The password. + * @param ppapszEnv Where to return PAM environment variables, NULL is + * fine if no variables to return. Call + * rtProcPosixFreePamEnv to free. Optional, so NULL + * can be passed in. + * @param pfMayFallBack Where to return whether a fallback to crypt is + * acceptable or if the failure result is due to + * authentication failing. Optional. + */ +static int rtProcPosixAuthenticateUsingPam(const char *pszPamService, const char *pszUser, const char *pszPassword, + char ***ppapszEnv, bool *pfMayFallBack) +{ + if (pfMayFallBack) + *pfMayFallBack = true; + + /* + * Dynamically load pam the first time we go thru here. + */ + static int (*s_pfnPamStart)(const char *, const char *, struct pam_conv *, pam_handle_t **); + static int (*s_pfnPamAuthenticate)(pam_handle_t *, int); + static int (*s_pfnPamAcctMgmt)(pam_handle_t *, int); + static int (*s_pfnPamSetItem)(pam_handle_t *, int, const void *); + static int (*s_pfnPamSetCred)(pam_handle_t *, int); + static char ** (*s_pfnPamGetEnvList)(pam_handle_t *); + static int (*s_pfnPamOpenSession)(pam_handle_t *, int); + static int (*s_pfnPamCloseSession)(pam_handle_t *, int); + static int (*s_pfnPamEnd)(pam_handle_t *, int); + if ( s_pfnPamStart == NULL + || s_pfnPamAuthenticate == NULL + || s_pfnPamAcctMgmt == NULL + || s_pfnPamSetItem == NULL + || s_pfnPamEnd == NULL) + { + RTLDRMOD hModPam = NIL_RTLDRMOD; + const char *pszLast; + int rc = RTLdrLoadSystemEx(pszLast = IPRT_LIBPAM_FILE_1, RTLDRLOAD_FLAGS_GLOBAL | RTLDRLOAD_FLAGS_NO_UNLOAD + | RTLDRLOAD_FLAGS_SO_VER_RANGE(IPRT_LIBPAM_FILE_1_FIRST_VER, IPRT_LIBPAM_FILE_1_END_VER), + &hModPam); +# ifdef IPRT_LIBPAM_FILE_2 + if (RT_FAILURE(rc)) + rc = RTLdrLoadSystemEx(pszLast = IPRT_LIBPAM_FILE_2, RTLDRLOAD_FLAGS_GLOBAL | RTLDRLOAD_FLAGS_NO_UNLOAD + | RTLDRLOAD_FLAGS_SO_VER_RANGE(IPRT_LIBPAM_FILE_2_FIRST_VER, IPRT_LIBPAM_FILE_2_END_VER), + &hModPam); +# endif +# ifdef IPRT_LIBPAM_FILE_3 + if (RT_FAILURE(rc)) + rc = RTLdrLoadSystemEx(pszLast = IPRT_LIBPAM_FILE_3, RTLDRLOAD_FLAGS_GLOBAL | RTLDRLOAD_FLAGS_NO_UNLOAD + | RTLDRLOAD_FLAGS_SO_VER_RANGE(IPRT_LIBPAM_FILE_3_FIRST_VER, IPRT_LIBPAM_FILE_3_END_VER), + &hModPam); +# endif + if (RT_FAILURE(rc)) + { + LogRelMax(10, ("failed to load %s: %Rrc\n", pszLast, rc)); + return VERR_AUTHENTICATION_FAILURE; + } + + *(uintptr_t *)&s_pfnPamStart = (uintptr_t)RTLdrGetFunction(hModPam, "pam_start"); + *(uintptr_t *)&s_pfnPamAuthenticate = (uintptr_t)RTLdrGetFunction(hModPam, "pam_authenticate"); + *(uintptr_t *)&s_pfnPamAcctMgmt = (uintptr_t)RTLdrGetFunction(hModPam, "pam_acct_mgmt"); + *(uintptr_t *)&s_pfnPamSetItem = (uintptr_t)RTLdrGetFunction(hModPam, "pam_set_item"); + *(uintptr_t *)&s_pfnPamSetCred = (uintptr_t)RTLdrGetFunction(hModPam, "pam_setcred"); + *(uintptr_t *)&s_pfnPamGetEnvList = (uintptr_t)RTLdrGetFunction(hModPam, "pam_getenvlist"); + *(uintptr_t *)&s_pfnPamOpenSession = (uintptr_t)RTLdrGetFunction(hModPam, "pam_open_session"); + *(uintptr_t *)&s_pfnPamCloseSession = (uintptr_t)RTLdrGetFunction(hModPam, "pam_close_session"); + *(uintptr_t *)&s_pfnPamEnd = (uintptr_t)RTLdrGetFunction(hModPam, "pam_end"); + ASMCompilerBarrier(); + + RTLdrClose(hModPam); + + if ( s_pfnPamStart == NULL + || s_pfnPamAuthenticate == NULL + || s_pfnPamAcctMgmt == NULL + || s_pfnPamSetItem == NULL + || s_pfnPamEnd == NULL) + { + LogRelMax(10, ("failed to resolve symbols: %p %p %p %p %p\n", + s_pfnPamStart, s_pfnPamAuthenticate, s_pfnPamAcctMgmt, s_pfnPamSetItem, s_pfnPamEnd)); + return VERR_AUTHENTICATION_FAILURE; + } + } + +# define pam_start s_pfnPamStart +# define pam_authenticate s_pfnPamAuthenticate +# define pam_acct_mgmt s_pfnPamAcctMgmt +# define pam_set_item s_pfnPamSetItem +# define pam_setcred s_pfnPamSetCred +# define pam_getenvlist s_pfnPamGetEnvList +# define pam_open_session s_pfnPamOpenSession +# define pam_close_session s_pfnPamCloseSession +# define pam_end s_pfnPamEnd + + /* + * Do the PAM stuff. + */ + pam_handle_t *hPam = NULL; + RTPROCPAMARGS PamConvArgs = { pszUser, pszPassword }; + struct pam_conv PamConversation; + RT_ZERO(PamConversation); + PamConversation.appdata_ptr = &PamConvArgs; + PamConversation.conv = rtPamConv; + int rc = pam_start(pszPamService, pszUser, &PamConversation, &hPam); + if (rc == PAM_SUCCESS) + { + rc = pam_set_item(hPam, PAM_RUSER, pszUser); + LogRel2(("rtProcPosixAuthenticateUsingPam(%s): pam_setitem/PAM_RUSER: %s\n", pszPamService, pszUser)); + if (rc == PAM_SUCCESS) + { + /* + * Secure TTY fun ahead (for pam_securetty). + * + * We need to set PAM_TTY (if available) to make PAM stacks work which + * require a secure TTY via pam_securetty (Debian 10 + 11, for example). This + * is typically an issue when launching as 'root'. See @bugref{10225}. + * + * Note! We only can try (or better: guess) to a certain amount, as it really + * depends on the distribution or Administrator which has set up the + * system which (and how) things are allowed (see /etc/securetty). + * + * Note! We don't acctually try or guess anything about the distro like + * suggested by the above note, we just try determine the TTY of + * the _parent_ process and hope for the best. (bird) + */ + char szTTY[64]; + int rc2 = RTEnvGetEx(RTENV_DEFAULT, "DISPLAY", szTTY, sizeof(szTTY), NULL); + if (RT_FAILURE(rc2)) + { + /* Virtual terminal hint given? */ + static char const s_szPrefix[] = "tty"; + memcpy(szTTY, s_szPrefix, sizeof(s_szPrefix)); + rc2 = RTEnvGetEx(RTENV_DEFAULT, "XDG_VTNR", &szTTY[sizeof(s_szPrefix) - 1], sizeof(s_szPrefix) - 1, NULL); + } + + /** @todo Should we - distinguished from the login service - also set the hostname as PAM_TTY? + * The pam_access and pam_systemd talk about this. Similarly, SSH and cron use "ssh" and "cron" for PAM_TTY + * (see PAM_TTY_KLUDGE). */ +#ifdef IPRT_WITH_PAM_TTY_KLUDGE + if (RT_FAILURE(rc2)) + if (!RTStrICmp(pszPamService, "access")) /* Access management needed? */ + { + int err = gethostname(szTTY, sizeof(szTTY)); + if (err == 0) + rc2 = VINF_SUCCESS; + } +#endif + /* As a last resort, try stdin's TTY name instead (if any). */ + if (RT_FAILURE(rc2)) + { + rc2 = ttyname_r(0 /*stdin*/, szTTY, sizeof(szTTY)); + if (rc2 != 0) + rc2 = RTErrConvertFromErrno(rc2); + } + + LogRel2(("rtProcPosixAuthenticateUsingPam(%s): pam_setitem/PAM_TTY: %s, rc2=%Rrc\n", pszPamService, szTTY, rc2)); + if (szTTY[0] == '\0') + LogRel2(("rtProcPosixAuthenticateUsingPam(%s): Hint: Looks like running as a non-interactive user (no TTY/PTY).\n" + "Authentication requiring a secure terminal might fail.\n", pszPamService)); + + if ( RT_SUCCESS(rc2) + && szTTY[0] != '\0') /* Only try using PAM_TTY if we have something to set. */ + rc = pam_set_item(hPam, PAM_TTY, szTTY); + + if (rc == PAM_SUCCESS) + { + /* From this point on we don't allow falling back to other auth methods. */ + if (pfMayFallBack) + *pfMayFallBack = false; + + rc = pam_authenticate(hPam, 0); + if (rc == PAM_SUCCESS) + { + rc = pam_acct_mgmt(hPam, 0); + if ( rc == PAM_SUCCESS + || rc == PAM_AUTHINFO_UNAVAIL /*??*/) + { + if ( ppapszEnv + && s_pfnPamGetEnvList + && s_pfnPamSetCred) + { + /* pam_env.so creates the environment when pam_setcred is called,. */ + int rcSetCred = pam_setcred(hPam, PAM_ESTABLISH_CRED | PAM_SILENT); + /** @todo check pam_setcred status code? */ + + /* Unless it does it during session opening (Ubuntu 21.10). This + unfortunately means we might mount user dir and other crap: */ + /** @todo do session handling properly */ + int rcOpenSession = PAM_ABORT; + if ( s_pfnPamOpenSession + && s_pfnPamCloseSession) + rcOpenSession = pam_open_session(hPam, PAM_SILENT); + + *ppapszEnv = pam_getenvlist(hPam); + LogFlowFunc(("pam_getenvlist -> %p ([0]=%p); rcSetCred=%d rcOpenSession=%d\n", + *ppapszEnv, *ppapszEnv ? **ppapszEnv : NULL, rcSetCred, rcOpenSession)); RT_NOREF(rcSetCred); + + if (rcOpenSession == PAM_SUCCESS) + pam_close_session(hPam, PAM_SILENT); + pam_setcred(hPam, PAM_DELETE_CRED); + } + + pam_end(hPam, PAM_SUCCESS); + LogFlowFunc(("pam auth (for %s) successful\n", pszPamService)); + return VINF_SUCCESS; + } + LogFunc(("pam_acct_mgmt -> %d\n", rc)); + } + else + LogFunc(("pam_authenticate -> %d\n", rc)); + } + else + LogFunc(("pam_setitem/PAM_TTY -> %d\n", rc)); + } + else + LogFunc(("pam_set_item/PAM_RUSER -> %d\n", rc)); + pam_end(hPam, rc); + } + else + LogFunc(("pam_start(%s) -> %d\n", pszPamService, rc)); + + LogRel2(("rtProcPosixAuthenticateUsingPam(%s): Failed authenticating user '%s' with %d\n", pszPamService, pszUser, rc)); + return VERR_AUTHENTICATION_FAILURE; +} + + +/** + * Checks if the given service file is present in any of the pam.d directories. + */ +static bool rtProcPosixPamServiceExists(const char *pszService) +{ + char szPath[256]; + + /* PAM_CONFIG_D: */ + int rc = RTPathJoin(szPath, sizeof(szPath), "/etc/pam.d/", pszService); AssertRC(rc); + if (RTFileExists(szPath)) + return true; + + /* PAM_CONFIG_DIST_D: */ + rc = RTPathJoin(szPath, sizeof(szPath), "/usr/lib/pam.d/", pszService); AssertRC(rc); + if (RTFileExists(szPath)) + return true; + + /* No support for PAM_CONFIG_DIST2_D. */ + return false; +} + +#endif /* IPRT_USE_PAM */ + + +#if defined(IPRT_WITH_DYNAMIC_CRYPT_R) +/** Pointer to crypt_r(). */ +typedef char *(*PFNCRYPTR)(const char *, const char *, struct crypt_data *); + +/** + * Wrapper for resolving and calling crypt_r dynamically. + * + * The reason for this is that fedora 30+ wants to use libxcrypt rather than the + * glibc libcrypt. The two libraries has different crypt_data sizes and layout, + * so we allocate a 256KB data block to be on the safe size (caller does this). + */ +static char *rtProcDynamicCryptR(const char *pszKey, const char *pszSalt, struct crypt_data *pData) +{ + static PFNCRYPTR volatile s_pfnCryptR = NULL; + PFNCRYPTR pfnCryptR = s_pfnCryptR; + if (pfnCryptR) + return pfnCryptR(pszKey, pszSalt, pData); + + pfnCryptR = (PFNCRYPTR)(uintptr_t)RTLdrGetSystemSymbolEx("libcrypt.so", "crypt_r", RTLDRLOAD_FLAGS_SO_VER_RANGE(1, 6)); + if (!pfnCryptR) + pfnCryptR = (PFNCRYPTR)(uintptr_t)RTLdrGetSystemSymbolEx("libxcrypt.so", "crypt_r", RTLDRLOAD_FLAGS_SO_VER_RANGE(1, 32)); + if (pfnCryptR) + { + s_pfnCryptR = pfnCryptR; + return pfnCryptR(pszKey, pszSalt, pData); + } + + LogRel(("IPRT/RTProc: Unable to locate crypt_r!\n")); + return NULL; +} +#endif /* IPRT_WITH_DYNAMIC_CRYPT_R */ + + +/** Free the environment list returned by rtCheckCredentials. */ +static void rtProcPosixFreePamEnv(char **papszEnv) +{ + if (papszEnv) + { + for (size_t i = 0; papszEnv[i] != NULL; i++) + free(papszEnv[i]); + free(papszEnv); + } +} + + +/** + * Check the credentials and return the gid/uid of user. + * + * @param pszUser The username. + * @param pszPasswd The password to authenticate with. + * @param gid Where to store the GID of the user. + * @param uid Where to store the UID of the user. + * @param ppapszEnv Where to return PAM environment variables, NULL is fine + * if no variables to return. Call rtProcPosixFreePamEnv to + * free. Optional, so NULL can be passed in. + * @returns IPRT status code + */ +static int rtCheckCredentials(const char *pszUser, const char *pszPasswd, gid_t *pGid, uid_t *pUid, char ***ppapszEnv) +{ + Log(("rtCheckCredentials: pszUser=%s\n", pszUser)); + int rc; + + if (ppapszEnv) + *ppapszEnv = NULL; + + /* + * Resolve user to UID and GID. + */ + char achBuf[_4K]; + struct passwd Pw; + struct passwd *pPw; + if (getpwnam_r(pszUser, &Pw, achBuf, sizeof(achBuf), &pPw) != 0) + return VERR_AUTHENTICATION_FAILURE; + if (!pPw) + return VERR_AUTHENTICATION_FAILURE; + + *pUid = pPw->pw_uid; + *pGid = pPw->pw_gid; + +#ifdef IPRT_USE_PAM + /* + * Try authenticate using PAM, and falling back on crypto if allowed. + */ + const char *pszService = "iprt-as-user"; + if (!rtProcPosixPamServiceExists("iprt-as-user")) +# ifdef IPRT_PAM_NATIVE_SERVICE_NAME_AS_USER + pszService = IPRT_PAM_NATIVE_SERVICE_NAME_AS_USER; +# else + pszService = "login"; +# endif + bool fMayFallBack = false; + rc = rtProcPosixAuthenticateUsingPam(pszService, pszUser, pszPasswd, ppapszEnv, &fMayFallBack); + if (RT_SUCCESS(rc) || !fMayFallBack) + { + RTMemWipeThoroughly(achBuf, sizeof(achBuf), 3); + return rc; + } +#endif + +#if !defined(IPRT_USE_PAM) || defined(RT_OS_LINUX) || defined(RT_OS_SOLARIS) || defined(RT_OS_OS2) +# if defined(RT_OS_LINUX) || defined(RT_OS_SOLARIS) + /* + * Ditto for /etc/shadow and replace pw_passwd from above if we can access it: + * + * Note! On FreeBSD and OS/2 the root user will open /etc/shadow above, so + * this getspnam_r step is not necessary. + */ + struct spwd ShwPwd; + char achBuf2[_4K]; +# if defined(RT_OS_LINUX) + struct spwd *pShwPwd = NULL; + if (getspnam_r(pszUser, &ShwPwd, achBuf2, sizeof(achBuf2), &pShwPwd) != 0) + pShwPwd = NULL; +# else + struct spwd *pShwPwd = getspnam_r(pszUser, &ShwPwd, achBuf2, sizeof(achBuf2)); +# endif + if (pShwPwd != NULL) + pPw->pw_passwd = pShwPwd->sp_pwdp; +# endif + + /* + * Encrypt the passed in password and see if it matches. + */ +# if defined(RT_OS_LINUX) + /* Default fCorrect=true if no password specified. In that case, pPw->pw_passwd + must be NULL (no password set for this user). Fail if a password is specified + but the user does not have one assigned. */ + rc = !pszPasswd || !*pszPasswd ? VINF_SUCCESS : VERR_AUTHENTICATION_FAILURE; + if (pPw->pw_passwd && *pPw->pw_passwd) +# endif + { +# if defined(RT_OS_LINUX) || defined(RT_OS_OS2) +# ifdef IPRT_WITH_DYNAMIC_CRYPT_R + size_t const cbCryptData = RT_MAX(sizeof(struct crypt_data) * 2, _256K); +# else + size_t const cbCryptData = sizeof(struct crypt_data); +# endif + struct crypt_data *pCryptData = (struct crypt_data *)RTMemTmpAllocZ(cbCryptData); + if (pCryptData) + { +# ifdef IPRT_WITH_DYNAMIC_CRYPT_R + char *pszEncPasswd = rtProcDynamicCryptR(pszPasswd, pPw->pw_passwd, pCryptData); +# else + char *pszEncPasswd = crypt_r(pszPasswd, pPw->pw_passwd, pCryptData); +# endif + rc = pszEncPasswd && !strcmp(pszEncPasswd, pPw->pw_passwd) ? VINF_SUCCESS : VERR_AUTHENTICATION_FAILURE; + RTMemWipeThoroughly(pCryptData, cbCryptData, 3); + RTMemTmpFree(pCryptData); + } + else + rc = VERR_NO_TMP_MEMORY; +# else + char *pszEncPasswd = crypt(pszPasswd, pPw->pw_passwd); + rc = strcmp(pszEncPasswd, pPw->pw_passwd) == 0 ? VINF_SUCCESS : VERR_AUTHENTICATION_FAILURE; +# endif + } + + /* + * Return GID and UID on success. Always wipe stack buffers. + */ + if (RT_SUCCESS(rc)) + { + *pGid = pPw->pw_gid; + *pUid = pPw->pw_uid; + } +# if defined(RT_OS_LINUX) || defined(RT_OS_SOLARIS) + RTMemWipeThoroughly(achBuf2, sizeof(achBuf2), 3); +# endif +#endif + RTMemWipeThoroughly(achBuf, sizeof(achBuf), 3); + return rc; +} + +#ifdef RT_OS_SOLARIS + +/** @todo the error reporting of the Solaris process contract code could be + * a lot better, but essentially it is not meant to run into errors after + * the debugging phase. */ +static int rtSolarisContractPreFork(void) +{ + int templateFd = open64(CTFS_ROOT "/process/template", O_RDWR); + if (templateFd < 0) + return -1; + + /* Set template parameters and event sets. */ + if (ct_pr_tmpl_set_param(templateFd, CT_PR_PGRPONLY)) + { + close(templateFd); + return -1; + } + if (ct_pr_tmpl_set_fatal(templateFd, CT_PR_EV_HWERR)) + { + close(templateFd); + return -1; + } + if (ct_tmpl_set_critical(templateFd, 0)) + { + close(templateFd); + return -1; + } + if (ct_tmpl_set_informative(templateFd, CT_PR_EV_HWERR)) + { + close(templateFd); + return -1; + } + + /* Make this the active template for the process. */ + if (ct_tmpl_activate(templateFd)) + { + close(templateFd); + return -1; + } + + return templateFd; +} + +static void rtSolarisContractPostForkChild(int templateFd) +{ + if (templateFd == -1) + return; + + /* Clear the active template. */ + ct_tmpl_clear(templateFd); + close(templateFd); +} + +static void rtSolarisContractPostForkParent(int templateFd, pid_t pid) +{ + if (templateFd == -1) + return; + + /* Clear the active template. */ + int cleared = ct_tmpl_clear(templateFd); + close(templateFd); + + /* If the clearing failed or the fork failed there's nothing more to do. */ + if (cleared || pid <= 0) + return; + + /* Look up the contract which was created by this thread. */ + int statFd = open64(CTFS_ROOT "/process/latest", O_RDONLY); + if (statFd == -1) + return; + ct_stathdl_t statHdl; + if (ct_status_read(statFd, CTD_COMMON, &statHdl)) + { + close(statFd); + return; + } + ctid_t ctId = ct_status_get_id(statHdl); + ct_status_free(statHdl); + close(statFd); + if (ctId < 0) + return; + + /* Abandon this contract we just created. */ + char ctlPath[PATH_MAX]; + size_t len = snprintf(ctlPath, sizeof(ctlPath), + CTFS_ROOT "/process/%ld/ctl", (long)ctId); + if (len >= sizeof(ctlPath)) + return; + int ctlFd = open64(ctlPath, O_WRONLY); + if (statFd == -1) + return; + if (ct_ctl_abandon(ctlFd) < 0) + { + close(ctlFd); + return; + } + close(ctlFd); +} + +#endif /* RT_OS_SOLARIS */ + + +RTR3DECL(int) RTProcCreate(const char *pszExec, const char * const *papszArgs, RTENV Env, unsigned fFlags, PRTPROCESS pProcess) +{ + return RTProcCreateEx(pszExec, papszArgs, Env, fFlags, + NULL, NULL, NULL, /* standard handles */ + NULL /*pszAsUser*/, NULL /* pszPassword*/, NULL /*pvExtraData*/, + pProcess); +} + + +/** + * Adjust the profile environment after forking the child process and changing + * the UID. + * + * @returns IRPT status code. + * @param hEnvToUse The environment we're going to use with execve. + * @param fFlags The process creation flags. + * @param hEnv The environment passed in by the user. + */ +static int rtProcPosixAdjustProfileEnvFromChild(RTENV hEnvToUse, uint32_t fFlags, RTENV hEnv) +{ + int rc = VINF_SUCCESS; +#ifdef RT_OS_DARWIN + if ( RT_SUCCESS(rc) + && (!(fFlags & RTPROC_FLAGS_ENV_CHANGE_RECORD) || RTEnvExistEx(hEnv, "TMPDIR")) ) + { + char szValue[RTPATH_MAX]; + size_t cbNeeded = confstr(_CS_DARWIN_USER_TEMP_DIR, szValue, sizeof(szValue)); + if (cbNeeded > 0 && cbNeeded < sizeof(szValue)) + { + char *pszTmp; + rc = RTStrCurrentCPToUtf8(&pszTmp, szValue); + if (RT_SUCCESS(rc)) + { + rc = RTEnvSetEx(hEnvToUse, "TMPDIR", pszTmp); + RTStrFree(pszTmp); + } + } + else + rc = VERR_BUFFER_OVERFLOW; + } +#else + RT_NOREF_PV(hEnvToUse); RT_NOREF_PV(fFlags); RT_NOREF_PV(hEnv); +#endif + return rc; +} + + +/** + * Undos quoting and escape sequences and looks for stop characters. + * + * @returns Where to continue scanning in @a pszString. This points to the + * next character after the stop character, but for the zero terminator + * it points to the terminator character. + * @param pszString The string to undo quoting and escaping for. + * This is both input and output as the work is + * done in place. + * @param pfStoppedOnEqual Where to return whether we stopped work on a + * plain equal characater or not. If this is NULL, + * then the equal character is not a stop + * character, then only newline and the zero + * terminator are. + */ +static char *rtProcPosixProfileEnvUnquoteAndUnescapeString(char *pszString, bool *pfStoppedOnEqual) +{ + if (pfStoppedOnEqual) + *pfStoppedOnEqual = false; + + enum { kPlain, kSingleQ, kDoubleQ } enmState = kPlain; + char *pszDst = pszString; + for (;;) + { + char ch = *pszString++; + switch (ch) + { + default: + *pszDst++ = ch; + break; + + case '\\': + { + char ch2; + if ( enmState == kSingleQ + || (ch2 = *pszString) == '\0' + || (enmState == kDoubleQ && strchr("\\$`\"\n", ch2) == NULL) ) + *pszDst++ = ch; + else + { + *pszDst++ = ch2; + pszString++; + } + break; + } + + case '"': + if (enmState == kSingleQ) + *pszDst++ = ch; + else + enmState = enmState == kPlain ? kDoubleQ : kPlain; + break; + + case '\'': + if (enmState == kDoubleQ) + *pszDst++ = ch; + else + enmState = enmState == kPlain ? kSingleQ : kPlain; + break; + + case '\n': + if (enmState == kPlain) + { + *pszDst = '\0'; + return pszString; + } + *pszDst++ = ch; + break; + + case '=': + if (enmState == kPlain && pfStoppedOnEqual) + { + *pszDst = '\0'; + *pfStoppedOnEqual = true; + return pszString; + } + *pszDst++ = ch; + break; + + case '\0': + Assert(enmState == kPlain); + *pszDst = '\0'; + return pszString - 1; + } + } +} + + +/** + * Worker for rtProcPosixProfileEnvRunAndHarvest that parses the environment + * dump and loads it into hEnvToUse. + * + * @note This isn't entirely correct should any of the profile setup scripts + * unset any of the environment variables in the basic initial + * enviornment, but since that's unlikely and it's very convenient to + * have something half sensible as a basis if don't don't grok the dump + * entirely and would skip central stuff like PATH or HOME. + * + * @returns IPRT status code. + * @retval -VERR_PARSE_ERROR (positive, e.g. warning) if we run into trouble. + * @retval -VERR_INVALID_UTF8_ENCODING (positive, e.g. warning) if there are + * invalid UTF-8 in the environment. This isn't unlikely if the + * profile doesn't use UTF-8. This is unfortunately not something we + * can guess to accurately up front, so we don't do any guessing and + * hope everyone is sensible and use UTF-8. + * + * @param hEnvToUse The basic environment to extend with what we manage + * to parse here. + * @param pszEnvDump The environment dump to parse. Nominally in Bourne + * shell 'export -p' format. + * @param fWithMarkers Whether there are markers around the dump (C shell, + * tmux) or not. + */ +static int rtProcPosixProfileEnvHarvest(RTENV hEnvToUse, char *pszEnvDump, bool fWithMarkers) +{ + LogRel3(("**** pszEnvDump start ****\n%s**** pszEnvDump end ****\n", pszEnvDump)); + if (!LogIs3Enabled()) + LogFunc(("**** pszEnvDump start ****\n%s**** pszEnvDump end ****\n", pszEnvDump)); + + /* + * Clip dump at markers if we're using them (C shell). + */ + if (fWithMarkers) + { + char *pszStart = strstr(pszEnvDump, g_szEnvMarkerBegin); + AssertReturn(pszStart, -VERR_PARSE_ERROR); + pszStart += sizeof(g_szEnvMarkerBegin) - 1; + if (*pszStart == '\n') + pszStart++; + pszEnvDump = pszStart; + + char *pszEnd = strstr(pszStart, g_szEnvMarkerEnd); + AssertReturn(pszEnd, -VERR_PARSE_ERROR); + *pszEnd = '\0'; + } + + /* + * Since we're using /bin/sh -c "export -p" for all the dumping, we should + * always get lines on the format: + * export VAR1="Value 1" + * export VAR2=Value2 + * + * However, just in case something goes wrong, like bash doesn't think it + * needs to be posixly correct, try deal with the alternative where + * "declare -x " replaces the "export". + */ + const char *pszPrefix; + if ( strncmp(pszEnvDump, RT_STR_TUPLE("export")) == 0 + && RT_C_IS_BLANK(pszEnvDump[6])) + pszPrefix = "export "; + else if ( strncmp(pszEnvDump, RT_STR_TUPLE("declare")) == 0 + && RT_C_IS_BLANK(pszEnvDump[7]) + && pszEnvDump[8] == '-') + pszPrefix = "declare -x "; /* We only need to care about the non-array, non-function lines. */ + else + AssertFailedReturn(-VERR_PARSE_ERROR); + size_t const cchPrefix = strlen(pszPrefix); + + /* + * Process the lines, ignoring stuff which we don't grok. + * The shell should quote problematic characters. Bash double quotes stuff + * by default, whereas almquist's shell does it as needed and only the value + * side. + */ + int rc = VINF_SUCCESS; + while (pszEnvDump && *pszEnvDump != '\0') + { + /* + * Skip the prefixing command. + */ + if ( cchPrefix == 0 + || strncmp(pszEnvDump, pszPrefix, cchPrefix) == 0) + { + pszEnvDump += cchPrefix; + while (RT_C_IS_BLANK(*pszEnvDump)) + pszEnvDump++; + } + else + { + /* Oops, must find our bearings for some reason... */ + pszEnvDump = strchr(pszEnvDump, '\n'); + rc = -VERR_PARSE_ERROR; + continue; + } + + /* + * Parse out the variable name using typical bourne shell escaping + * and quoting rules. + */ + /** @todo We should throw away lines that aren't propertly quoted, now we + * just continue and use what we found. */ + const char *pszVar = pszEnvDump; + bool fStoppedOnPlainEqual = false; + pszEnvDump = rtProcPosixProfileEnvUnquoteAndUnescapeString(pszEnvDump, &fStoppedOnPlainEqual); + const char *pszValue = pszEnvDump; + if (fStoppedOnPlainEqual) + pszEnvDump = rtProcPosixProfileEnvUnquoteAndUnescapeString(pszEnvDump, NULL /*pfStoppedOnPlainEqual*/); + else + pszValue = ""; + + /* + * Add them if valid UTF-8, otherwise we simply drop them for now. + * The whole codeset stuff goes seriously wonky here as the environment + * we're harvesting probably contains it's own LC_CTYPE or LANG variables, + * so ignore the problem for now. + */ + if ( RTStrIsValidEncoding(pszVar) + && RTStrIsValidEncoding(pszValue)) + { + int rc2 = RTEnvSetEx(hEnvToUse, pszVar, pszValue); + AssertRCReturn(rc2, rc2); + } + else if (rc == VINF_SUCCESS) + rc = -VERR_INVALID_UTF8_ENCODING; + } + + return rc; +} + + +/** + * Runs the user's shell in login mode with some environment dumping logic and + * harvests the dump, putting it into hEnvToUse. + * + * This is a bit hairy, esp. with regards to codesets. + * + * @returns IPRT status code. Not all error statuses will be returned and the + * caller should just continue with whatever is in hEnvToUse. + * + * @param hEnvToUse On input this is the basic user environment, on success + * in is fleshed out with stuff from the login shell dump. + * @param pszAsUser The user name for the profile. + * @param uid The UID corrsponding to @a pszAsUser, ~0 if current user. + * @param gid The GID corrsponding to @a pszAsUser, ~0 if current user. + * @param pszShell The login shell. This is a writable string to avoid + * needing to make a copy of it when examining the path + * part, instead we make a temporary change to it which is + * always reverted before returning. + */ +static int rtProcPosixProfileEnvRunAndHarvest(RTENV hEnvToUse, const char *pszAsUser, uid_t uid, gid_t gid, char *pszShell) +{ + LogFlowFunc(("pszAsUser=%s uid=%u gid=%u pszShell=%s; hEnvToUse contains %u variables on entry\n", + pszAsUser, uid, gid, pszShell, RTEnvCountEx(hEnvToUse) )); + + /* + * The three standard handles should be pointed to /dev/null, the 3rd handle + * is used to dump the environment. + */ + RTPIPE hPipeR, hPipeW; + int rc = RTPipeCreate(&hPipeR, &hPipeW, 0); + if (RT_SUCCESS(rc)) + { + RTFILE hFileNull; + rc = RTFileOpenBitBucket(&hFileNull, RTFILE_O_READWRITE); + if (RT_SUCCESS(rc)) + { + int aRedirFds[4]; + aRedirFds[0] = aRedirFds[1] = aRedirFds[2] = RTFileToNative(hFileNull); + aRedirFds[3] = RTPipeToNative(hPipeW); + + /* + * Allocate a buffer for receiving the environment dump. + * + * This is fixed sized for simplicity and safety (creative user script + * shouldn't be allowed to exhaust our memory or such, after all we're + * most likely running with root privileges in this code path). + */ + size_t const cbEnvDump = _64K; + char * const pszEnvDump = (char *)RTMemTmpAllocZ(cbEnvDump); + if (pszEnvDump) + { + /* + * Our default approach is using /bin/sh: + */ + const char *pszExec = _PATH_BSHELL; + const char *apszArgs[8]; + apszArgs[0] = "-sh"; /* First arg must start with a dash for login shells. */ + apszArgs[1] = "-c"; + apszArgs[2] = "POSIXLY_CORRECT=1;export -p >&3"; + apszArgs[3] = NULL; + + /* + * But see if we can trust the shell to be a real usable shell. + * This would be great as different shell typically has different profile setup + * files and we'll endup with the wrong enviornment if we use a different shell. + */ + char szDashShell[32]; + char szExportArg[128]; + bool fWithMarkers = false; + const char *pszShellNm = RTPathFilename(pszShell); + if ( pszShellNm + && access(pszShellNm, X_OK)) + { + /* + * First the check that it's a known bin directory: + */ + size_t const cchShellPath = pszShellNm - pszShell; + char const chSaved = pszShell[cchShellPath - 1]; + pszShell[cchShellPath - 1] = '\0'; + if ( RTPathCompare(pszShell, "/bin") == 0 + || RTPathCompare(pszShell, "/usr/bin") == 0 + || RTPathCompare(pszShell, "/usr/local/bin") == 0) + { + /* + * Then see if we recognize the shell name. + */ + RTStrCopy(&szDashShell[1], sizeof(szDashShell) - 1, pszShellNm); + szDashShell[0] = '-'; + if ( strcmp(pszShellNm, "bash") == 0 + || strcmp(pszShellNm, "ksh") == 0 + || strcmp(pszShellNm, "ksh93") == 0 + || strcmp(pszShellNm, "zsh") == 0 + || strcmp(pszShellNm, "fish") == 0) + { + pszExec = pszShell; + apszArgs[0] = szDashShell; + + /* Use /bin/sh for doing the environment dumping so we get the same kind + of output from everyone and can limit our parsing + testing efforts. */ + RTStrPrintf(szExportArg, sizeof(szExportArg), + "%s -c 'POSIXLY_CORRECT=1;export -p >&3'", _PATH_BSHELL); + apszArgs[2] = szExportArg; + } + /* C shell is very annoying in that it closes fd 3 without regard to what + we might have put there, so we must use stdout here but with markers so + we can find the dump. + Seems tmux have similar issues as it doesn't work above, but works fine here. */ + else if ( strcmp(pszShellNm, "csh") == 0 + || strcmp(pszShellNm, "tcsh") == 0 + || strcmp(pszShellNm, "tmux") == 0) + { + pszExec = pszShell; + apszArgs[0] = szDashShell; + + fWithMarkers = true; + size_t cch = RTStrPrintf(szExportArg, sizeof(szExportArg), + "%s -c 'set -e;POSIXLY_CORRECT=1;echo %s;export -p;echo %s'", + _PATH_BSHELL, g_szEnvMarkerBegin, g_szEnvMarkerEnd); + Assert(cch < sizeof(szExportArg) - 1); RT_NOREF(cch); + apszArgs[2] = szExportArg; + + aRedirFds[1] = aRedirFds[3]; + aRedirFds[3] = -1; + } + } + pszShell[cchShellPath - 1] = chSaved; + } + + /* + * Create the process and wait for the output. + */ + LogFunc(("Executing '%s': '%s', '%s', '%s'\n", pszExec, apszArgs[0], apszArgs[1], apszArgs[2])); + RTPROCESS hProcess = NIL_RTPROCESS; + rc = rtProcPosixCreateInner(pszExec, apszArgs, hEnvToUse, hEnvToUse, 0 /*fFlags*/, + pszAsUser, uid, gid, RT_ELEMENTS(aRedirFds), aRedirFds, &hProcess); + if (RT_SUCCESS(rc)) + { + RTPipeClose(hPipeW); + hPipeW = NIL_RTPIPE; + + size_t offEnvDump = 0; + uint64_t const msStart = RTTimeMilliTS(); + for (;;) + { + size_t cbRead = 0; + if (offEnvDump < cbEnvDump - 1) + { + rc = RTPipeRead(hPipeR, &pszEnvDump[offEnvDump], cbEnvDump - 1 - offEnvDump, &cbRead); + if (RT_SUCCESS(rc)) + offEnvDump += cbRead; + else + { + LogFlowFunc(("Breaking out of read loop: %Rrc\n", rc)); + if (rc == VERR_BROKEN_PIPE) + rc = VINF_SUCCESS; + break; + } + pszEnvDump[offEnvDump] = '\0'; + } + else + { + LogFunc(("Too much data in environment dump, dropping it\n")); + rc = VERR_TOO_MUCH_DATA; + break; + } + + /* Do the timout check. */ + uint64_t const cMsElapsed = RTTimeMilliTS() - msStart; + if (cMsElapsed >= RT_MS_15SEC) + { + LogFunc(("Timed out after %RU64 ms\n", cMsElapsed)); + rc = VERR_TIMEOUT; + break; + } + + /* If we got no data in above wait for more to become ready. */ + if (!cbRead) + RTPipeSelectOne(hPipeR, RT_MS_15SEC - cMsElapsed); + } + + /* + * Kill the process and wait for it to avoid leaving zombies behind. + */ + /** @todo do we check the exit code? */ + int rc2 = RTProcWait(hProcess, RTPROCWAIT_FLAGS_NOBLOCK, NULL); + if (RT_SUCCESS(rc2)) + LogFlowFunc(("First RTProcWait succeeded\n")); + else + { + LogFunc(("First RTProcWait failed (%Rrc), terminating and doing a blocking wait\n", rc2)); + RTProcTerminate(hProcess); + RTProcWait(hProcess, RTPROCWAIT_FLAGS_BLOCK, NULL); + } + + /* + * Parse the result. + */ + if (RT_SUCCESS(rc)) + rc = rtProcPosixProfileEnvHarvest(hEnvToUse, pszEnvDump, fWithMarkers); + else + { + LogFunc(("Ignoring rc=%Rrc from the pipe read loop and continues with basic environment\n", rc)); + rc = -rc; + } + } + else + LogFunc(("Failed to create process '%s': %Rrc\n", pszExec, rc)); + RTMemTmpFree(pszEnvDump); + } + else + { + LogFunc(("Failed to allocate %#zx bytes for the dump\n", cbEnvDump)); + rc = VERR_NO_TMP_MEMORY; + } + RTFileClose(hFileNull); + } + else + LogFunc(("Failed to open /dev/null: %Rrc\n", rc)); + RTPipeClose(hPipeR); + RTPipeClose(hPipeW); + } + else + LogFunc(("Failed to create pipe: %Rrc\n", rc)); + LogFlowFunc(("returns %Rrc (hEnvToUse contains %u variables now)\n", rc, RTEnvCountEx(hEnvToUse))); + return rc; +} + + +/** + * Create an environment for the given user. + * + * This starts by creating a very basic environment and then tries to do it + * properly by running the user's shell in login mode with some environment + * dumping attached. The latter may fail and we'll ignore that for now and move + * ahead with the very basic environment. + * + * @returns IPRT status code. + * @param phEnvToUse Where to return the created environment. + * @param pszAsUser The user name for the profile. NULL if the current + * user. + * @param uid The UID corrsponding to @a pszAsUser, ~0 if NULL. + * @param gid The GID corrsponding to @a pszAsUser, ~0 if NULL. + * @param fFlags RTPROC_FLAGS_XXX + * @param papszPamEnv Array of environment variables returned by PAM, if + * it was used for authentication and produced anything. + * Otherwise NULL. + */ +static int rtProcPosixCreateProfileEnv(PRTENV phEnvToUse, const char *pszAsUser, uid_t uid, gid_t gid, + uint32_t fFlags, char **papszPamEnv) +{ + /* + * Get the passwd entry for the user. + */ + struct passwd Pwd; + struct passwd *pPwd = NULL; + char achBuf[_4K]; + int rc; + errno = 0; + if (pszAsUser) + rc = getpwnam_r(pszAsUser, &Pwd, achBuf, sizeof(achBuf), &pPwd); + else + rc = getpwuid_r(getuid(), &Pwd, achBuf, sizeof(achBuf), &pPwd); + if (rc == 0 && pPwd) + { + /* + * Convert stuff to UTF-8 since the environment is UTF-8. + */ + char *pszDir; + rc = RTStrCurrentCPToUtf8(&pszDir, pPwd->pw_dir); + if (RT_SUCCESS(rc)) + { +#if 0 /* Enable and modify this to test shells other that your login shell. */ + pPwd->pw_shell = (char *)"/bin/tmux"; +#endif + char *pszShell; + rc = RTStrCurrentCPToUtf8(&pszShell, pPwd->pw_shell); + if (RT_SUCCESS(rc)) + { + char *pszAsUserFree = NULL; + if (!pszAsUser) + { + rc = RTStrCurrentCPToUtf8(&pszAsUserFree, pPwd->pw_name); + if (RT_SUCCESS(rc)) + pszAsUser = pszAsUserFree; + } + if (RT_SUCCESS(rc)) + { + /* + * Create and populate the environment. + */ + rc = RTEnvCreate(phEnvToUse); + if (RT_SUCCESS(rc)) + { + RTENV hEnvToUse = *phEnvToUse; + rc = RTEnvSetEx(hEnvToUse, "HOME", pszDir); + if (RT_SUCCESS(rc)) + rc = RTEnvSetEx(hEnvToUse, "SHELL", pszShell); + if (RT_SUCCESS(rc)) + rc = RTEnvSetEx(hEnvToUse, "USER", pszAsUser); + if (RT_SUCCESS(rc)) + rc = RTEnvSetEx(hEnvToUse, "LOGNAME", pszAsUser); + if (RT_SUCCESS(rc)) + rc = RTEnvSetEx(hEnvToUse, "PATH", pPwd->pw_uid == 0 ? _PATH_STDPATH : _PATH_DEFPATH); + char szTmpPath[RTPATH_MAX]; + if (RT_SUCCESS(rc)) + { + RTStrPrintf(szTmpPath, sizeof(szTmpPath), "%s/%s", _PATH_MAILDIR, pszAsUser); + rc = RTEnvSetEx(hEnvToUse, "MAIL", szTmpPath); + } +#ifdef RT_OS_DARWIN + if (RT_SUCCESS(rc)) + { + /* TMPDIR is some unique per user directory under /var/folders on darwin, + so get the one for the current user. If we're launching the process as + a different user, rtProcPosixAdjustProfileEnvFromChild will update it + again for the actual child process user (provided we set it here). See + https://opensource.apple.com/source/Libc/Libc-997.1.1/darwin/_dirhelper.c + for the implementation of this query. */ + size_t cbNeeded = confstr(_CS_DARWIN_USER_TEMP_DIR, szTmpPath, sizeof(szTmpPath)); + if (cbNeeded > 0 && cbNeeded < sizeof(szTmpPath)) + { + char *pszTmp; + rc = RTStrCurrentCPToUtf8(&pszTmp, szTmpPath); + if (RT_SUCCESS(rc)) + { + rc = RTEnvSetEx(hEnvToUse, "TMPDIR", pszTmp); + RTStrFree(pszTmp); + } + } + else + rc = VERR_BUFFER_OVERFLOW; + } +#endif + /* + * Add everything from the PAM environment. + */ + if (RT_SUCCESS(rc) && papszPamEnv != NULL) + for (size_t i = 0; papszPamEnv[i] != NULL && RT_SUCCESS(rc); i++) + { + char *pszEnvVar; + rc = RTStrCurrentCPToUtf8(&pszEnvVar, papszPamEnv[i]); + if (RT_SUCCESS(rc)) + { + char *pszValue = strchr(pszEnvVar, '='); + if (pszValue) + *pszValue++ = '\0'; + rc = RTEnvSetEx(hEnvToUse, pszEnvVar, pszValue ? pszValue : ""); + RTStrFree(pszEnvVar); + } + /* Ignore conversion issue, though LogRel them. */ + else if (rc != VERR_NO_STR_MEMORY && rc != VERR_NO_MEMORY) + { + LogRelMax(256, ("RTStrCurrentCPToUtf8(,%.*Rhxs) -> %Rrc\n", strlen(pszEnvVar), pszEnvVar, rc)); + rc = -rc; + } + } + if (RT_SUCCESS(rc)) + { + /* + * Now comes the fun part where we need to try run a shell in login mode + * and harvest its final environment to get the proper environment for + * the user. We ignore some failures here so buggy login scrips and + * other weird stuff won't trip us up too badly. + */ + if (!(fFlags & RTPROC_FLAGS_ONLY_BASIC_PROFILE)) + rc = rtProcPosixProfileEnvRunAndHarvest(hEnvToUse, pszAsUser, uid, gid, pszShell); + } + + if (RT_FAILURE(rc)) + RTEnvDestroy(hEnvToUse); + } + RTStrFree(pszAsUserFree); + } + RTStrFree(pszShell); + } + RTStrFree(pszDir); + } + } + else + rc = errno ? RTErrConvertFromErrno(errno) : VERR_ACCESS_DENIED; + return rc; +} + + +/** + * Converts the arguments to the child's LC_CTYPE charset if necessary. + * + * @returns IPRT status code. + * @param papszArgs The arguments (UTF-8). + * @param hEnvToUse The child process environment. + * @param ppapszArgs Where to return the converted arguments. The array + * entries must be freed by RTStrFree and the array itself + * by RTMemFree. + */ +static int rtProcPosixConvertArgv(const char * const *papszArgs, RTENV hEnvToUse, char ***ppapszArgs) +{ + *ppapszArgs = (char **)papszArgs; + + /* + * The first thing we need to do here is to try guess the codeset of the + * child process and check if it's UTF-8 or not. + */ + const char *pszEncoding; + char szEncoding[512]; + if (hEnvToUse == RTENV_DEFAULT) + { + /* Same environment as us, assume setlocale is up to date: */ + pszEncoding = rtStrGetLocaleCodeset(); + } + else + { + /* + * LC_ALL overrides everything else. The LC_* environment variables are often set + * to the empty string so move on the next variable if that is the case (that's + * what setlocale in glibc does). + */ + const char *pszVar; + int rc = RTEnvGetEx(hEnvToUse, pszVar = "LC_ALL", szEncoding, sizeof(szEncoding), NULL); + if (rc == VERR_ENV_VAR_NOT_FOUND || (RT_SUCCESS(rc) && szEncoding[0] == '\0')) + rc = RTEnvGetEx(hEnvToUse, pszVar = "LC_CTYPE", szEncoding, sizeof(szEncoding), NULL); + if (rc == VERR_ENV_VAR_NOT_FOUND || (RT_SUCCESS(rc) && szEncoding[0] == '\0')) + rc = RTEnvGetEx(hEnvToUse, pszVar = "LANG", szEncoding, sizeof(szEncoding), NULL); + if (RT_SUCCESS(rc) && szEncoding[0] != '\0') + { + /* + * LC_ALL can contain a composite locale consisting of the locales of each of the + * categories in two different formats depending on the OS. On Solaris, macOS, and + * *BSD composite locale names use slash ('/') as the separator and the following + * order for the categories: + * LC_CTYPE/LC_NUMERIC/LC_TIME/LC_COLLATE/LC_MONETARY/LC_MESSAGES + * e.g.: + * en_US.UTF-8/POSIX/el_GR.UTF-8/el_CY.UTF-8/en_GB.UTF-8/es_ES.UTF-8 + * + * On Solaris there is also a leading slash. + * + * On Linux and OS/2 the composite locale format is made up of key-value pairs + * of category names and locales of the form 'name=value' with each element + * separated by a semicolon in the same order as above with following additional + * categories included as well: + * LC_PAPER/LC_NAME/LC_ADDRESS/LC_TELEPHONE/LC_MEASUREMENT/LC_IDENTIFICATION + * e.g. + * LC_CTYPE=fr_BE;LC_NUMERIC=fr_BE@euro;LC_TIME=fr_BE.utf8;LC_COLLATE=fr_CA;\ + * LC_MONETARY=fr_CA.utf8;LC_MESSAGES=fr_CH;LC_PAPER=fr_CH.utf8;LC_NAME=fr_FR;\ + * LC_ADDRESS=fr_FR.utf8;LC_TELEPHONE=fr_LU;LC_MEASUREMENT=fr_LU@euro;\ + * LC_IDENTIFICATION=fr_LU.utf8 + */ + char *pszEncodingStart = szEncoding; +#if !defined(RT_OS_LINUX) && !defined(RT_OS_OS2) + if (*pszEncodingStart == '/') + pszEncodingStart++; + char *pszSlash = strchr(pszEncodingStart, '/'); + if (pszSlash) + *pszSlash = '\0'; /* This ASSUMES the first one is LC_CTYPE! */ +#else + char *pszCType = strstr(pszEncodingStart, "LC_CTYPE="); + if (pszCType) + { + pszEncodingStart = pszCType + sizeof("LC_CTYPE=") - 1; + + char *pszSemiColon = strchr(pszEncodingStart, ';'); + if (pszSemiColon) + *pszSemiColon = '\0'; + } +#endif + + /* + * Use newlocale and nl_langinfo_l to determine the default codeset for the locale + * specified in the child's environment. These routines have been around since + * ancient days on Linux and for quite a long time on macOS, Solaris, and *BSD but + * to ensure their availability check that LC_CTYPE_MASK is defined. + * + * Note! The macOS nl_langinfo(3)/nl_langinfo_l(3) routines return a pointer to an + * empty string for "short" locale names like en_NZ, it_IT, el_GR, etc. so use + * UTF-8 in those cases as it is the default for short name locales on macOS + * (see also rtStrGetLocaleCodeset). + */ +#ifdef LC_CTYPE_MASK + locale_t hLocale = newlocale(LC_CTYPE_MASK, pszEncodingStart, (locale_t)0); + if (hLocale != (locale_t)0) + { + const char *pszCodeset = nl_langinfo_l(CODESET, hLocale); + Log2Func(("nl_langinfo_l(CODESET, %s=%s) -> %s\n", pszVar, pszEncodingStart, pszCodeset)); + if (!pszCodeset || *pszCodeset == '\0') +# ifdef RT_OS_DARWIN + pszEncoding = "UTF-8"; +# else + pszEncoding = "ASCII"; +# endif + else + { + rc = RTStrCopy(szEncoding, sizeof(szEncoding), pszCodeset); + AssertRC(rc); /* cannot possibly overflow */ + } + + freelocale(hLocale); + pszEncoding = szEncoding; + } + else +#endif + { + /* If there is something that ought to be a character set encoding, try use it: */ + const char *pszDot = strchr(pszEncodingStart, '.'); + if (pszDot) + pszDot = RTStrStripL(pszDot + 1); + if (pszDot && *pszDot != '\0') + { + pszEncoding = pszDot; + Log2Func(("%s=%s -> %s (simple)\n", pszVar, szEncoding, pszEncoding)); + } + else + { + /* This is mostly wrong, but I cannot think of anything better now: */ + pszEncoding = rtStrGetLocaleCodeset(); + LogFunc(("No newlocale or it failed (on '%s=%s', errno=%d), falling back on %s that we're using...\n", + pszVar, pszEncodingStart, errno, pszEncoding)); + } + } + RT_NOREF_PV(pszVar); + } + else +#ifdef RT_OS_DARWIN /* @bugref{10153}: Darwin defaults to UTF-8. */ + pszEncoding = "UTF-8"; +#else + pszEncoding = "ASCII"; +#endif + } + + /* + * Do nothing if it's UTF-8. + */ + if (rtStrIsCodesetUtf8(pszEncoding)) + { + LogFlowFunc(("No conversion needed (%s)\n", pszEncoding)); + return VINF_SUCCESS; + } + + + /* + * Do the conversion. + */ + size_t cArgs = 0; + while (papszArgs[cArgs] != NULL) + cArgs++; + LogFunc(("Converting #%u arguments to %s...\n", cArgs, pszEncoding)); + + char **papszArgsConverted = (char **)RTMemAllocZ(sizeof(papszArgsConverted[0]) * (cArgs + 2)); + AssertReturn(papszArgsConverted, VERR_NO_MEMORY); + + void *pvConversionCache = NULL; + rtStrLocalCacheInit(&pvConversionCache); + for (size_t i = 0; i < cArgs; i++) + { + int rc = rtStrLocalCacheConvert(papszArgs[i], strlen(papszArgs[i]), "UTF-8", + &papszArgsConverted[i], 0, pszEncoding, &pvConversionCache); + if (RT_SUCCESS(rc) && rc != VWRN_NO_TRANSLATION) + { /* likely */ } + else + { + LogRelMax(100, ("Failed to convert argument #%u '%s' to '%s': %Rrc\n", i, papszArgs[i], pszEncoding, rc)); + while (i-- > 0) + RTStrFree(papszArgsConverted[i]); + RTMemFree(papszArgsConverted); + rtStrLocalCacheDelete(&pvConversionCache); + return rc == VWRN_NO_TRANSLATION || rc == VERR_NO_TRANSLATION ? VERR_PROC_NO_ARG_TRANSLATION : rc; + } + } + + rtStrLocalCacheDelete(&pvConversionCache); + *ppapszArgs = papszArgsConverted; + return VINF_SUCCESS; +} + + +/** + * The result structure for rtPathFindExec/RTPathTraverseList. + * @todo move to common path code? + */ +typedef struct RTPATHINTSEARCH +{ + /** For EACCES or EPERM errors that we continued on. + * @note Must be initialized to VINF_SUCCESS. */ + int rcSticky; + /** Buffer containing the filename. */ + char szFound[RTPATH_MAX]; +} RTPATHINTSEARCH; +/** Pointer to a rtPathFindExec/RTPathTraverseList result. */ +typedef RTPATHINTSEARCH *PRTPATHINTSEARCH; + + +/** + * RTPathTraverseList callback used by RTProcCreateEx to locate the executable. + */ +static DECLCALLBACK(int) rtPathFindExec(char const *pchPath, size_t cchPath, void *pvUser1, void *pvUser2) +{ + const char *pszExec = (const char *)pvUser1; + PRTPATHINTSEARCH pResult = (PRTPATHINTSEARCH)pvUser2; + int rc = RTPathJoinEx(pResult->szFound, sizeof(pResult->szFound), pchPath, cchPath, pszExec, RTSTR_MAX, + RTPATH_STR_F_STYLE_HOST); + if (RT_SUCCESS(rc)) + { + const char *pszNativeExec = NULL; + rc = rtPathToNative(&pszNativeExec, pResult->szFound, NULL); + if (RT_SUCCESS(rc)) + { + if (!access(pszNativeExec, X_OK)) + rc = VINF_SUCCESS; + else + { + if ( errno == EACCES + || errno == EPERM) + pResult->rcSticky = RTErrConvertFromErrno(errno); + rc = VERR_TRY_AGAIN; + } + rtPathFreeNative(pszNativeExec, pResult->szFound); + } + else + AssertRCStmt(rc, rc = VERR_TRY_AGAIN /* don't stop on this, whatever it is */); + } + return rc; +} + + +RTR3DECL(int) RTProcCreateEx(const char *pszExec, const char * const *papszArgs, RTENV hEnv, uint32_t fFlags, + PCRTHANDLE phStdIn, PCRTHANDLE phStdOut, PCRTHANDLE phStdErr, const char *pszAsUser, + const char *pszPassword, void *pvExtraData, PRTPROCESS phProcess) +{ + int rc; + LogFlow(("RTProcCreateEx: pszExec=%s pszAsUser=%s fFlags=%#x phStdIn=%p phStdOut=%p phStdErr=%p\n", + pszExec, pszAsUser, fFlags, phStdIn, phStdOut, phStdErr)); + + /* + * Input validation + */ + AssertPtrReturn(pszExec, VERR_INVALID_POINTER); + AssertReturn(*pszExec, VERR_INVALID_PARAMETER); + AssertReturn(!(fFlags & ~RTPROC_FLAGS_VALID_MASK), VERR_INVALID_PARAMETER); + AssertReturn(!(fFlags & RTPROC_FLAGS_DETACHED) || !phProcess, VERR_INVALID_PARAMETER); + AssertReturn(hEnv != NIL_RTENV, VERR_INVALID_PARAMETER); + AssertPtrReturn(papszArgs, VERR_INVALID_PARAMETER); + AssertPtrNullReturn(pszAsUser, VERR_INVALID_POINTER); + AssertReturn(!pszAsUser || *pszAsUser, VERR_INVALID_PARAMETER); + AssertReturn(!pszPassword || pszAsUser, VERR_INVALID_PARAMETER); + AssertPtrNullReturn(pszPassword, VERR_INVALID_POINTER); +#if defined(RT_OS_OS2) + if (fFlags & RTPROC_FLAGS_DETACHED) + return VERR_PROC_DETACH_NOT_SUPPORTED; +#endif + AssertReturn(pvExtraData == NULL || (fFlags & RTPROC_FLAGS_DESIRED_SESSION_ID), VERR_INVALID_PARAMETER); + + /* + * Get the file descriptors for the handles we've been passed. + */ + PCRTHANDLE paHandles[3] = { phStdIn, phStdOut, phStdErr }; + int aStdFds[3] = { -1, -1, -1 }; + for (int i = 0; i < 3; i++) + { + if (paHandles[i]) + { + AssertPtrReturn(paHandles[i], VERR_INVALID_POINTER); + switch (paHandles[i]->enmType) + { + case RTHANDLETYPE_FILE: + aStdFds[i] = paHandles[i]->u.hFile != NIL_RTFILE + ? (int)RTFileToNative(paHandles[i]->u.hFile) + : -2 /* close it */; + break; + + case RTHANDLETYPE_PIPE: + aStdFds[i] = paHandles[i]->u.hPipe != NIL_RTPIPE + ? (int)RTPipeToNative(paHandles[i]->u.hPipe) + : -2 /* close it */; + break; + + case RTHANDLETYPE_SOCKET: + aStdFds[i] = paHandles[i]->u.hSocket != NIL_RTSOCKET + ? (int)RTSocketToNative(paHandles[i]->u.hSocket) + : -2 /* close it */; + break; + + default: + AssertMsgFailedReturn(("%d: %d\n", i, paHandles[i]->enmType), VERR_INVALID_PARAMETER); + } + /** @todo check the close-on-execness of these handles? */ + } + } + + for (int i = 0; i < 3; i++) + if (aStdFds[i] == i) + aStdFds[i] = -1; + LogFlowFunc(("aStdFds={%d, %d, %d}\n", aStdFds[0], aStdFds[1], aStdFds[2])); + + for (int i = 0; i < 3; i++) + AssertMsgReturn(aStdFds[i] < 0 || aStdFds[i] > i, + ("%i := %i not possible because we're lazy\n", i, aStdFds[i]), + VERR_NOT_SUPPORTED); + + /* + * Validate the credentials if a user is specified. + */ + bool const fNeedLoginEnv = (fFlags & RTPROC_FLAGS_PROFILE) + && ((fFlags & RTPROC_FLAGS_ENV_CHANGE_RECORD) || hEnv == RTENV_DEFAULT); + uid_t uid = ~(uid_t)0; + gid_t gid = ~(gid_t)0; + char **papszPamEnv = NULL; + if (pszAsUser) + { + rc = rtCheckCredentials(pszAsUser, pszPassword, &gid, &uid, fNeedLoginEnv ? &papszPamEnv : NULL); + if (RT_FAILURE(rc)) + return rc; + } +#ifdef IPRT_USE_PAM + /* + * User unchanged, but if PROFILE is request we must try get the PAM + * environmnet variables. + * + * For this to work, we'll need a special PAM service profile which doesn't + * actually do any authentication, only concerns itself with the enviornment + * setup. gdm-launch-environment is such one, and we use it if we haven't + * got an IPRT specific one there. + */ + else if (fNeedLoginEnv) + { + const char *pszService; + if (rtProcPosixPamServiceExists("iprt-environment")) + pszService = "iprt-environment"; +# ifdef IPRT_PAM_NATIVE_SERVICE_NAME_ENVIRONMENT + else if (rtProcPosixPamServiceExists(IPRT_PAM_NATIVE_SERVICE_NAME_ENVIRONMENT)) + pszService = IPRT_PAM_NATIVE_SERVICE_NAME_ENVIRONMENT; +# endif + else if (rtProcPosixPamServiceExists("gdm-launch-environment")) + pszService = "gdm-launch-environment"; + else + pszService = NULL; + if (pszService) + { + char szLoginName[512]; + rc = getlogin_r(szLoginName, sizeof(szLoginName)); + if (rc == 0) + rc = rtProcPosixAuthenticateUsingPam(pszService, szLoginName, "xxx", &papszPamEnv, NULL); + } + } +#endif + + /* + * Create the child environment if either RTPROC_FLAGS_PROFILE or + * RTPROC_FLAGS_ENV_CHANGE_RECORD are in effect. + */ + RTENV hEnvToUse = hEnv; + if ( (fFlags & (RTPROC_FLAGS_ENV_CHANGE_RECORD | RTPROC_FLAGS_PROFILE)) + && ( (fFlags & RTPROC_FLAGS_ENV_CHANGE_RECORD) + || hEnv == RTENV_DEFAULT) ) + { + if (fFlags & RTPROC_FLAGS_PROFILE) + rc = rtProcPosixCreateProfileEnv(&hEnvToUse, pszAsUser, uid, gid, fFlags, papszPamEnv); + else + rc = RTEnvClone(&hEnvToUse, RTENV_DEFAULT); + rtProcPosixFreePamEnv(papszPamEnv); + papszPamEnv = NULL; + if (RT_FAILURE(rc)) + return rc; + + if ((fFlags & RTPROC_FLAGS_ENV_CHANGE_RECORD) && hEnv != RTENV_DEFAULT) + { + rc = RTEnvApplyChanges(hEnvToUse, hEnv); + if (RT_FAILURE(rc)) + { + RTEnvDestroy(hEnvToUse); + return rc; + } + } + } + Assert(papszPamEnv == NULL); + + /* + * Check for execute access to the file, searching the PATH if needed. + */ + const char *pszNativeExec = NULL; + rc = rtPathToNative(&pszNativeExec, pszExec, NULL); + if (RT_SUCCESS(rc)) + { + if (access(pszNativeExec, X_OK) == 0) + rc = VINF_SUCCESS; + else + { + rc = errno; + rtPathFreeNative(pszNativeExec, pszExec); + + if ( !(fFlags & RTPROC_FLAGS_SEARCH_PATH) + || rc != ENOENT + || RTPathHavePath(pszExec) ) + rc = RTErrConvertFromErrno(rc); + else + { + /* Search the PATH for it: */ + char *pszPath = RTEnvDupEx(hEnvToUse, "PATH"); + if (pszPath) + { + PRTPATHINTSEARCH pResult = (PRTPATHINTSEARCH)alloca(sizeof(*pResult)); + pResult->rcSticky = VINF_SUCCESS; + rc = RTPathTraverseList(pszPath, ':', rtPathFindExec, (void *)pszExec, pResult); + RTStrFree(pszPath); + if (RT_SUCCESS(rc)) + { + /* Found it. Now, convert to native path: */ + pszExec = pResult->szFound; + rc = rtPathToNative(&pszNativeExec, pszExec, NULL); + } + else + rc = rc != VERR_END_OF_STRING ? rc + : pResult->rcSticky == VINF_SUCCESS ? VERR_FILE_NOT_FOUND : pResult->rcSticky; + } + else + rc = VERR_NO_STR_MEMORY; + } + } + if (RT_SUCCESS(rc)) + { + /* + * Convert arguments to child codeset if necessary. + */ + char **papszArgsConverted = (char **)papszArgs; + if (!(fFlags & RTPROC_FLAGS_UTF8_ARGV)) + rc = rtProcPosixConvertArgv(papszArgs, hEnvToUse, &papszArgsConverted); + if (RT_SUCCESS(rc)) + { + /* + * The rest of the process creation is reused internally by rtProcPosixCreateProfileEnv. + */ + rc = rtProcPosixCreateInner(pszNativeExec, papszArgsConverted, hEnv, hEnvToUse, fFlags, pszAsUser, uid, gid, + RT_ELEMENTS(aStdFds), aStdFds, phProcess); + + } + + /* Free the translated argv copy, if needed. */ + if (papszArgsConverted != (char **)papszArgs) + { + for (size_t i = 0; papszArgsConverted[i] != NULL; i++) + RTStrFree(papszArgsConverted[i]); + RTMemFree(papszArgsConverted); + } + rtPathFreeNative(pszNativeExec, pszExec); + } + } + if (hEnvToUse != hEnv) + RTEnvDestroy(hEnvToUse); + return rc; +} + + +/** + * The inner 2nd half of RTProcCreateEx. + * + * This is also used by rtProcPosixCreateProfileEnv(). + * + * @returns IPRT status code. + * @param pszNativeExec The executable to run (absolute path, X_OK). + * Native path. + * @param papszArgs The arguments. Caller has done codeset conversions. + * @param hEnv The original enviornment request, needed for + * adjustments if starting as different user. + * @param hEnvToUse The environment we should use. + * @param fFlags The process creation flags, RTPROC_FLAGS_XXX. + * @param pszAsUser The user to start the process as, if requested. + * @param uid The UID corrsponding to @a pszAsUser, ~0 if NULL. + * @param gid The GID corrsponding to @a pszAsUser, ~0 if NULL. + * @param cRedirFds Number of redirection file descriptors. + * @param paRedirFds Pointer to redirection file descriptors. Entries + * containing -1 are not modified (inherit from parent), + * -2 indicates that the descriptor should be closed in the + * child. + * @param phProcess Where to return the process ID on success. + */ +static int rtProcPosixCreateInner(const char *pszNativeExec, const char * const *papszArgs, RTENV hEnv, RTENV hEnvToUse, + uint32_t fFlags, const char *pszAsUser, uid_t uid, gid_t gid, + unsigned cRedirFds, int *paRedirFds, PRTPROCESS phProcess) +{ + /* + * Get the environment block. + */ + const char * const *papszEnv = RTEnvGetExecEnvP(hEnvToUse); + AssertPtrReturn(papszEnv, VERR_INVALID_HANDLE); + + /* + * Optimize the redirections. + */ + while (cRedirFds > 0 && paRedirFds[cRedirFds - 1] == -1) + cRedirFds--; + + /* + * Child PID. + */ + pid_t pid = -1; + + /* + * Take care of detaching the process. + * + * HACK ALERT! Put the process into a new process group with pgid = pid + * to make sure it differs from that of the parent process to ensure that + * the IPRT waitpid call doesn't race anyone (read XPCOM) doing group wide + * waits. setsid() includes the setpgid() functionality. + * 2010-10-11 XPCOM no longer waits for anything, but it cannot hurt. + */ +#ifndef RT_OS_OS2 + if (fFlags & RTPROC_FLAGS_DETACHED) + { +# ifdef RT_OS_SOLARIS + int templateFd = -1; + if (!(fFlags & RTPROC_FLAGS_SAME_CONTRACT)) + { + templateFd = rtSolarisContractPreFork(); + if (templateFd == -1) + return VERR_OPEN_FAILED; + } +# endif /* RT_OS_SOLARIS */ + pid = fork(); + if (!pid) + { +# ifdef RT_OS_SOLARIS + if (!(fFlags & RTPROC_FLAGS_SAME_CONTRACT)) + rtSolarisContractPostForkChild(templateFd); +# endif + setsid(); /* see comment above */ + + pid = -1; + /* Child falls through to the actual spawn code below. */ + } + else + { +# ifdef RT_OS_SOLARIS + if (!(fFlags & RTPROC_FLAGS_SAME_CONTRACT)) + rtSolarisContractPostForkParent(templateFd, pid); +# endif + if (pid > 0) + { + /* Must wait for the temporary process to avoid a zombie. */ + int status = 0; + pid_t pidChild = 0; + + /* Restart if we get interrupted. */ + do + { + pidChild = waitpid(pid, &status, 0); + } while ( pidChild == -1 + && errno == EINTR); + + /* Assume that something wasn't found. No detailed info. */ + if (status) + return VERR_PROCESS_NOT_FOUND; + if (phProcess) + *phProcess = 0; + return VINF_SUCCESS; + } + return RTErrConvertFromErrno(errno); + } + } +#endif + + /* + * Spawn the child. + * + * Any spawn code MUST not execute any atexit functions if it is for a + * detached process. It would lead to running the atexit functions which + * make only sense for the parent. libORBit e.g. gets confused by multiple + * execution. Remember, there was only a fork() so far, and until exec() + * is successfully run there is nothing which would prevent doing anything + * silly with the (duplicated) file descriptors. + */ + int rc; +#ifdef HAVE_POSIX_SPAWN + /** @todo OS/2: implement DETACHED (BACKGROUND stuff), see VbglR3Daemonize. */ + if ( uid == ~(uid_t)0 + && gid == ~(gid_t)0) + { + /* Spawn attributes. */ + posix_spawnattr_t Attr; + rc = posix_spawnattr_init(&Attr); + if (!rc) + { + /* Indicate that process group and signal mask are to be changed, + and that the child should use default signal actions. */ + rc = posix_spawnattr_setflags(&Attr, POSIX_SPAWN_SETPGROUP | POSIX_SPAWN_SETSIGMASK | POSIX_SPAWN_SETSIGDEF); + Assert(rc == 0); + + /* The child starts in its own process group. */ + if (!rc) + { + rc = posix_spawnattr_setpgroup(&Attr, 0 /* pg == child pid */); + Assert(rc == 0); + } + + /* Unmask all signals. */ + if (!rc) + { + sigset_t SigMask; + sigemptyset(&SigMask); + rc = posix_spawnattr_setsigmask(&Attr, &SigMask); Assert(rc == 0); + } + + /* File changes. */ + posix_spawn_file_actions_t FileActions; + posix_spawn_file_actions_t *pFileActions = NULL; + if (!rc && cRedirFds > 0) + { + rc = posix_spawn_file_actions_init(&FileActions); + if (!rc) + { + pFileActions = &FileActions; + for (unsigned i = 0; i < cRedirFds; i++) + { + int fd = paRedirFds[i]; + if (fd == -2) + rc = posix_spawn_file_actions_addclose(&FileActions, i); + else if (fd >= 0 && fd != (int)i) + { + rc = posix_spawn_file_actions_adddup2(&FileActions, fd, i); + if (!rc) + { + for (unsigned j = i + 1; j < cRedirFds; j++) + if (paRedirFds[j] == fd) + { + fd = -1; + break; + } + if (fd >= 0) + rc = posix_spawn_file_actions_addclose(&FileActions, fd); + } + } + if (rc) + break; + } + } + } + + if (!rc) + rc = posix_spawn(&pid, pszNativeExec, pFileActions, &Attr, (char * const *)papszArgs, + (char * const *)papszEnv); + + /* cleanup */ + int rc2 = posix_spawnattr_destroy(&Attr); Assert(rc2 == 0); NOREF(rc2); + if (pFileActions) + { + rc2 = posix_spawn_file_actions_destroy(pFileActions); + Assert(rc2 == 0); + } + + /* return on success.*/ + if (!rc) + { + /* For a detached process this happens in the temp process, so + * it's not worth doing anything as this process must exit. */ + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(0); + if (phProcess) + *phProcess = pid; + return VINF_SUCCESS; + } + } + /* For a detached process this happens in the temp process, so + * it's not worth doing anything as this process must exit. */ + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(124); + } + else +#endif + { +#ifdef RT_OS_SOLARIS + int templateFd = -1; + if (!(fFlags & RTPROC_FLAGS_SAME_CONTRACT)) + { + templateFd = rtSolarisContractPreFork(); + if (templateFd == -1) + return VERR_OPEN_FAILED; + } +#endif /* RT_OS_SOLARIS */ + pid = fork(); + if (!pid) + { +#ifdef RT_OS_SOLARIS + if (!(fFlags & RTPROC_FLAGS_SAME_CONTRACT)) + rtSolarisContractPostForkChild(templateFd); +#endif /* RT_OS_SOLARIS */ + if (!(fFlags & RTPROC_FLAGS_DETACHED)) + setpgid(0, 0); /* see comment above */ + + /* + * Change group and user if requested. + */ +#if 1 /** @todo This needs more work, see suplib/hardening. */ + if (pszAsUser) + { + int ret = initgroups(pszAsUser, gid); + if (ret) + { + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(126); + else + exit(126); + } + } + if (gid != ~(gid_t)0) + { + if (setgid(gid)) + { + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(126); + else + exit(126); + } + } + + if (uid != ~(uid_t)0) + { + if (setuid(uid)) + { + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(126); + else + exit(126); + } + } +#endif + + /* + * Some final profile environment tweaks, if running as user. + */ + if ( (fFlags & RTPROC_FLAGS_PROFILE) + && pszAsUser + && ( (fFlags & RTPROC_FLAGS_ENV_CHANGE_RECORD) + || hEnv == RTENV_DEFAULT) ) + { + rc = rtProcPosixAdjustProfileEnvFromChild(hEnvToUse, fFlags, hEnv); + papszEnv = RTEnvGetExecEnvP(hEnvToUse); + if (RT_FAILURE(rc) || !papszEnv) + { + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(126); + else + exit(126); + } + } + + /* + * Unset the signal mask. + */ + sigset_t SigMask; + sigemptyset(&SigMask); + rc = sigprocmask(SIG_SETMASK, &SigMask, NULL); + Assert(rc == 0); + + /* + * Apply changes to the standard file descriptor and stuff. + */ + for (unsigned i = 0; i < cRedirFds; i++) + { + int fd = paRedirFds[i]; + if (fd == -2) + close(i); + else if (fd >= 0) + { + int rc2 = dup2(fd, i); + if (rc2 != (int)i) + { + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(125); + else + exit(125); + } + for (unsigned j = i + 1; j < cRedirFds; j++) + if (paRedirFds[j] == fd) + { + fd = -1; + break; + } + if (fd >= 0) + close(fd); + } + } + + /* + * Finally, execute the requested program. + */ + rc = execve(pszNativeExec, (char * const *)papszArgs, (char * const *)papszEnv); + if (errno == ENOEXEC) + { + /* This can happen when trying to start a shell script without the magic #!/bin/sh */ + RTAssertMsg2Weak("Cannot execute this binary format!\n"); + } + else + RTAssertMsg2Weak("execve returns %d errno=%d (%s)\n", rc, errno, pszNativeExec); + RTAssertReleasePanic(); + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(127); + else + exit(127); + } +#ifdef RT_OS_SOLARIS + if (!(fFlags & RTPROC_FLAGS_SAME_CONTRACT)) + rtSolarisContractPostForkParent(templateFd, pid); +#endif /* RT_OS_SOLARIS */ + if (pid > 0) + { + /* For a detached process this happens in the temp process, so + * it's not worth doing anything as this process must exit. */ + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(0); + if (phProcess) + *phProcess = pid; + return VINF_SUCCESS; + } + /* For a detached process this happens in the temp process, so + * it's not worth doing anything as this process must exit. */ + if (fFlags & RTPROC_FLAGS_DETACHED) + _Exit(124); + return RTErrConvertFromErrno(errno); + } + + return VERR_NOT_IMPLEMENTED; +} + + +RTR3DECL(int) RTProcDaemonizeUsingFork(bool fNoChDir, bool fNoClose, const char *pszPidfile) +{ + /* + * Fork the child process in a new session and quit the parent. + * + * - fork once and create a new session (setsid). This will detach us + * from the controlling tty meaning that we won't receive the SIGHUP + * (or any other signal) sent to that session. + * - The SIGHUP signal is ignored because the session/parent may throw + * us one before we get to the setsid. + * - When the parent exit(0) we will become an orphan and re-parented to + * the init process. + * - Because of the sometimes unexpected semantics of assigning the + * controlling tty automagically when a session leader first opens a tty, + * we will fork() once more to get rid of the session leadership role. + */ + + /* We start off by opening the pidfile, so that we can fail straight away + * if it already exists. */ + int fdPidfile = -1; + if (pszPidfile != NULL) + { + /* @note the exclusive create is not guaranteed on all file + * systems (e.g. NFSv2) */ + if ((fdPidfile = open(pszPidfile, O_RDWR | O_CREAT | O_EXCL, 0644)) == -1) + return RTErrConvertFromErrno(errno); + } + + /* Ignore SIGHUP straight away. */ + struct sigaction OldSigAct; + struct sigaction SigAct; + memset(&SigAct, 0, sizeof(SigAct)); + SigAct.sa_handler = SIG_IGN; + int rcSigAct = sigaction(SIGHUP, &SigAct, &OldSigAct); + + /* First fork, to become independent process. */ + pid_t pid = fork(); + if (pid == -1) + { + if (fdPidfile != -1) + close(fdPidfile); + return RTErrConvertFromErrno(errno); + } + if (pid != 0) + { + /* Parent exits, no longer necessary. The child gets reparented + * to the init process. */ + exit(0); + } + + /* Create new session, fix up the standard file descriptors and the + * current working directory. */ + /** @todo r=klaus the webservice uses this function and assumes that the + * contract id of the daemon is the same as that of the original process. + * Whenever this code is changed this must still remain possible. */ + pid_t newpgid = setsid(); + int SavedErrno = errno; + if (rcSigAct != -1) + sigaction(SIGHUP, &OldSigAct, NULL); + if (newpgid == -1) + { + if (fdPidfile != -1) + close(fdPidfile); + return RTErrConvertFromErrno(SavedErrno); + } + + if (!fNoClose) + { + /* Open stdin(0), stdout(1) and stderr(2) as /dev/null. */ + int fd = open("/dev/null", O_RDWR); + if (fd == -1) /* paranoia */ + { + close(STDIN_FILENO); + close(STDOUT_FILENO); + close(STDERR_FILENO); + fd = open("/dev/null", O_RDWR); + } + if (fd != -1) + { + dup2(fd, STDIN_FILENO); + dup2(fd, STDOUT_FILENO); + dup2(fd, STDERR_FILENO); + if (fd > 2) + close(fd); + } + } + + if (!fNoChDir) + { + int rcIgnored = chdir("/"); + NOREF(rcIgnored); + } + + /* Second fork to lose session leader status. */ + pid = fork(); + if (pid == -1) + { + if (fdPidfile != -1) + close(fdPidfile); + return RTErrConvertFromErrno(errno); + } + + if (pid != 0) + { + /* Write the pid file, this is done in the parent, before exiting. */ + if (fdPidfile != -1) + { + char szBuf[256]; + size_t cbPid = RTStrPrintf(szBuf, sizeof(szBuf), "%d\n", pid); + ssize_t cbIgnored = write(fdPidfile, szBuf, cbPid); NOREF(cbIgnored); + close(fdPidfile); + } + exit(0); + } + + if (fdPidfile != -1) + close(fdPidfile); + + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/process-posix.cpp b/src/VBox/Runtime/r3/posix/process-posix.cpp new file mode 100644 index 00000000..b0074f1c --- /dev/null +++ b/src/VBox/Runtime/r3/posix/process-posix.cpp @@ -0,0 +1,279 @@ +/* $Id: process-posix.cpp $ */ +/** @file + * IPRT - Process, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_PROCESS +#include <unistd.h> +#include <stdlib.h> +#include <errno.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <sys/wait.h> +#include <signal.h> +#include <pwd.h> + +#include <iprt/process.h> +#include "internal/iprt.h" + +#include <iprt/assert.h> +#include <iprt/env.h> +#include <iprt/err.h> +#include <iprt/file.h> +#include <iprt/pipe.h> +#include <iprt/socket.h> +#include <iprt/string.h> +#include <iprt/mem.h> +#include "internal/process.h" + + +RTR3DECL(int) RTProcWait(RTPROCESS Process, unsigned fFlags, PRTPROCSTATUS pProcStatus) +{ + int rc; + do rc = RTProcWaitNoResume(Process, fFlags, pProcStatus); + while (rc == VERR_INTERRUPTED); + return rc; +} + + +RTR3DECL(int) RTProcWaitNoResume(RTPROCESS Process, unsigned fFlags, PRTPROCSTATUS pProcStatus) +{ + /* + * Validate input. + */ + if (Process <= 0) + { + AssertMsgFailed(("Invalid Process=%d\n", Process)); + return VERR_INVALID_PARAMETER; + } + if (fFlags & ~(RTPROCWAIT_FLAGS_NOBLOCK | RTPROCWAIT_FLAGS_BLOCK)) + { + AssertMsgFailed(("Invalid flags %#x\n", fFlags)); + return VERR_INVALID_PARAMETER; + } + + /* + * Perform the wait. + */ + int iStatus = 0; + int rc = waitpid(Process, &iStatus, fFlags & RTPROCWAIT_FLAGS_NOBLOCK ? WNOHANG : 0); + if (rc > 0) + { + /* + * Fill in the status structure. + */ + if (pProcStatus) + { + if (WIFEXITED(iStatus)) + { + pProcStatus->enmReason = RTPROCEXITREASON_NORMAL; + pProcStatus->iStatus = WEXITSTATUS(iStatus); + } + else if (WIFSIGNALED(iStatus)) + { + pProcStatus->enmReason = RTPROCEXITREASON_SIGNAL; + pProcStatus->iStatus = WTERMSIG(iStatus); + } + else + { + Assert(!WIFSTOPPED(iStatus)); + pProcStatus->enmReason = RTPROCEXITREASON_ABEND; + pProcStatus->iStatus = iStatus; + } + } + return VINF_SUCCESS; + } + + /* + * Child running? + */ + if (!rc) + { + Assert(fFlags & RTPROCWAIT_FLAGS_NOBLOCK); + return VERR_PROCESS_RUNNING; + } + + /* + * Figure out which error to return. + */ + int iErr = errno; + if (iErr == ECHILD) + return VERR_PROCESS_NOT_FOUND; + return RTErrConvertFromErrno(iErr); +} + + +RTR3DECL(int) RTProcTerminate(RTPROCESS Process) +{ + if (Process == NIL_RTPROCESS) + return VINF_SUCCESS; + + if (!kill(Process, SIGKILL)) + return VINF_SUCCESS; + return RTErrConvertFromErrno(errno); +} + + +RTR3DECL(uint64_t) RTProcGetAffinityMask(void) +{ + /// @todo + return 1; +} + + +RTR3DECL(int) RTProcQueryParent(RTPROCESS hProcess, PRTPROCESS phParent) +{ + if (hProcess == RTProcSelf()) + { + *phParent = getppid(); + return VINF_SUCCESS; + } + return VERR_NOT_SUPPORTED; +} + + +RTR3DECL(int) RTProcQueryUsername(RTPROCESS hProcess, char *pszUser, size_t cbUser, size_t *pcbUser) +{ + AssertReturn( (pszUser && cbUser > 0) + || (!pszUser && !cbUser), VERR_INVALID_PARAMETER); + AssertReturn(pcbUser || pszUser, VERR_INVALID_PARAMETER); + + int rc; + if ( hProcess == NIL_RTPROCESS + || hProcess == RTProcSelf()) + { + /* + * Figure a good buffer estimate. + */ + int32_t cbPwdMax = sysconf(_SC_GETPW_R_SIZE_MAX); + if (cbPwdMax <= _1K) + cbPwdMax = _1K; + else + AssertStmt(cbPwdMax <= 32*_1M, cbPwdMax = 32*_1M); + char *pchBuf = (char *)RTMemTmpAllocZ(cbPwdMax); + if (pchBuf) + { + /* + * Get the password file entry. + */ + struct passwd Pwd; + struct passwd *pPwd = NULL; + rc = getpwuid_r(geteuid(), &Pwd, pchBuf, cbPwdMax, &pPwd); + if (!rc) + { + /* + * Convert the name to UTF-8, assuming that we're getting it in the local codeset. + */ + /** @todo This isn't exactly optimal... the current codeset/page conversion + * stuff never was. Should optimize that for UTF-8 and ASCII one day. + * And also optimize for avoiding heap. */ + char *pszTmp = NULL; + rc = RTStrCurrentCPToUtf8(&pszTmp, pPwd->pw_name); + if (RT_SUCCESS(rc)) + { + size_t cbTmp = strlen(pszTmp) + 1; + if (pcbUser) + *pcbUser = cbTmp; + if (cbTmp <= cbUser) + { + memcpy(pszUser, pszTmp, cbTmp); + rc = VINF_SUCCESS; + } + else + rc = VERR_BUFFER_OVERFLOW; + RTStrFree(pszTmp); + } + } + else + rc = RTErrConvertFromErrno(rc); + RTMemFree(pchBuf); + } + else + rc = VERR_NO_TMP_MEMORY; + } + else + rc = VERR_NOT_SUPPORTED; + return rc; +} + + +RTR3DECL(int) RTProcQueryUsernameA(RTPROCESS hProcess, char **ppszUser) +{ + AssertPtrReturn(ppszUser, VERR_INVALID_POINTER); + + int rc; + if ( hProcess == NIL_RTPROCESS + || hProcess == RTProcSelf()) + { + /* + * Figure a good buffer estimate. + */ + int32_t cbPwdMax = sysconf(_SC_GETPW_R_SIZE_MAX); + if (cbPwdMax <= _1K) + cbPwdMax = _1K; + else + AssertStmt(cbPwdMax <= 32*_1M, cbPwdMax = 32*_1M); + char *pchBuf = (char *)RTMemTmpAllocZ(cbPwdMax); + if (pchBuf) + { + /* + * Get the password file entry. + */ + struct passwd Pwd; + struct passwd *pPwd = NULL; + rc = getpwuid_r(geteuid(), &Pwd, pchBuf, cbPwdMax, &pPwd); + if (!rc) + { + /* + * Convert the name to UTF-8, assuming that we're getting it in the local codeset. + */ + rc = RTStrCurrentCPToUtf8(ppszUser, pPwd->pw_name); + } + else + rc = RTErrConvertFromErrno(rc); + RTMemFree(pchBuf); + } + else + rc = VERR_NO_TMP_MEMORY; + } + else + rc = VERR_NOT_SUPPORTED; + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/rand-posix.cpp b/src/VBox/Runtime/r3/posix/rand-posix.cpp new file mode 100644 index 00000000..191e3db8 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/rand-posix.cpp @@ -0,0 +1,148 @@ +/* $Id: rand-posix.cpp $ */ +/** @file + * IPRT - Random Numbers and Byte Streams, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <errno.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <sys/ioctl.h> +#include <fcntl.h> +#ifdef _MSC_VER +# include <io.h> +# include <stdio.h> +#else +# include <unistd.h> +# include <sys/time.h> +#endif + +#include <iprt/rand.h> +#include <iprt/mem.h> +#include <iprt/errcore.h> +#include <iprt/assert.h> +#include "internal/rand.h" +#include "internal/magics.h" + + + +/** @copydoc RTRANDINT::pfnGetBytes */ +static DECLCALLBACK(void) rtRandAdvPosixGetBytes(PRTRANDINT pThis, uint8_t *pb, size_t cb) +{ + ssize_t cbRead = read(pThis->u.File.hFile, pb, cb); + if ((size_t)cbRead != cb) + { + /* S10 has been observed returning 1040 bytes at the time from /dev/urandom. + Which means we need to do than 256 rounds to reach 668171 bytes if + that's what demanded by the caller (like tstRTMemWipe.cpp). */ + ssize_t cTries = RT_MAX(256, cb / 64); + do + { + if (cbRead > 0) + { + cb -= cbRead; + pb += cbRead; + } + cbRead = read(pThis->u.File.hFile, pb, cb); + } while ( (size_t)cbRead != cb + && cTries-- > 0); + AssertReleaseMsg((size_t)cbRead == cb, ("%zu != %zu, cTries=%zd errno=%d\n", cbRead, cb, cTries, errno)); + } +} + + +/** @copydoc RTRANDINT::pfnDestroy */ +static DECLCALLBACK(int) rtRandAdvPosixDestroy(PRTRANDINT pThis) +{ + pThis->u32Magic = ~RTRANDINT_MAGIC; + int fd = pThis->u.File.hFile; + pThis->u.File.hFile = -1; + RTMemFree(pThis); + close(fd); + return VINF_SUCCESS; +} + + +static int rtRandAdvPosixCreateSystem(PRTRAND phRand, const char *pszDev) RT_NO_THROW_DEF +{ + /* + * Try open it first and then setup the handle structure. + */ + int fd = open(pszDev, O_RDONLY); + if (fd < 0) + return RTErrConvertFromErrno(errno); + int rc; + if (fcntl(fd, F_SETFD, FD_CLOEXEC) != -1) + { + PRTRANDINT pThis = (PRTRANDINT)RTMemAlloc(sizeof(*pThis)); + if (pThis) + { + pThis->u32Magic = RTRANDINT_MAGIC; + pThis->pfnGetBytes = rtRandAdvPosixGetBytes; + pThis->pfnGetU32 = rtRandAdvSynthesizeU32FromBytes; + pThis->pfnGetU64 = rtRandAdvSynthesizeU64FromBytes; + pThis->pfnSeed = rtRandAdvStubSeed; + pThis->pfnSaveState = rtRandAdvStubSaveState; + pThis->pfnRestoreState = rtRandAdvStubRestoreState; + pThis->pfnDestroy = rtRandAdvPosixDestroy; + pThis->u.File.hFile = fd; + + *phRand = pThis; + return VINF_SUCCESS; + } + + /* bail out */ + rc = VERR_NO_MEMORY; + } + else + rc = RTErrConvertFromErrno(errno); + close(fd); + return rc; +} + + +RTDECL(int) RTRandAdvCreateSystemFaster(PRTRAND phRand) RT_NO_THROW_DEF +{ + return rtRandAdvPosixCreateSystem(phRand, "/dev/urandom"); +} + + +RTDECL(int) RTRandAdvCreateSystemTruer(PRTRAND phRand) RT_NO_THROW_DEF +{ + return rtRandAdvPosixCreateSystem(phRand, "/dev/random"); +} + diff --git a/src/VBox/Runtime/r3/posix/rtmempage-exec-mmap-heap-posix.cpp b/src/VBox/Runtime/r3/posix/rtmempage-exec-mmap-heap-posix.cpp new file mode 100644 index 00000000..2bb60178 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/rtmempage-exec-mmap-heap-posix.cpp @@ -0,0 +1,797 @@ +/* $Id: rtmempage-exec-mmap-heap-posix.cpp $ */ +/** @file + * IPRT - RTMemPage*, POSIX with heap. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include "internal/iprt.h" +#include <iprt/mem.h> + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/avl.h> +#include <iprt/critsect.h> +#include <iprt/errcore.h> +#include <iprt/once.h> +#include <iprt/param.h> +#include <iprt/string.h> +#include "internal/mem.h" +#include "../alloc-ef.h" + +#include <stdlib.h> +#include <errno.h> +#include <sys/mman.h> +#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON) +# define MAP_ANONYMOUS MAP_ANON +#endif + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +/** Threshold at which to we switch to simply calling mmap. */ +#define RTMEMPAGEPOSIX_MMAP_THRESHOLD _128K +/** The size of a heap block (power of two) - in bytes. */ +#define RTMEMPAGEPOSIX_BLOCK_SIZE _2M +AssertCompile(RTMEMPAGEPOSIX_BLOCK_SIZE == (RTMEMPAGEPOSIX_BLOCK_SIZE / PAGE_SIZE) * PAGE_SIZE); +/** The number of pages per heap block. */ +#define RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT (RTMEMPAGEPOSIX_BLOCK_SIZE / PAGE_SIZE) + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** Pointer to a page heap block. */ +typedef struct RTHEAPPAGEBLOCK *PRTHEAPPAGEBLOCK; + +/** + * A simple page heap. + */ +typedef struct RTHEAPPAGE +{ + /** Magic number (RTHEAPPAGE_MAGIC). */ + uint32_t u32Magic; + /** The number of pages in the heap (in BlockTree). */ + uint32_t cHeapPages; + /** The number of currently free pages. */ + uint32_t cFreePages; + /** Number of successful calls. */ + uint32_t cAllocCalls; + /** Number of successful free calls. */ + uint32_t cFreeCalls; + /** The free call number at which we last tried to minimize the heap. */ + uint32_t uLastMinimizeCall; + /** Tree of heap blocks. */ + AVLRPVTREE BlockTree; + /** Allocation hint no 1 (last freed). */ + PRTHEAPPAGEBLOCK pHint1; + /** Allocation hint no 2 (last alloc). */ + PRTHEAPPAGEBLOCK pHint2; + /** Critical section protecting the heap. */ + RTCRITSECT CritSect; + /** Set if the memory must allocated with execute access. */ + bool fExec; +} RTHEAPPAGE; +#define RTHEAPPAGE_MAGIC UINT32_C(0xfeedface) +/** Pointer to a page heap. */ +typedef RTHEAPPAGE *PRTHEAPPAGE; + + +/** + * Describes a page heap block. + */ +typedef struct RTHEAPPAGEBLOCK +{ + /** The AVL tree node core (void pointer range). */ + AVLRPVNODECORE Core; + /** Allocation bitmap. Set bits marks allocated pages. */ + uint32_t bmAlloc[RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT / 32]; + /** Allocation boundrary bitmap. Set bits marks the start of + * allocations. */ + uint32_t bmFirst[RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT / 32]; + /** The number of free pages. */ + uint32_t cFreePages; + /** Pointer back to the heap. */ + PRTHEAPPAGE pHeap; +} RTHEAPPAGEBLOCK; + + +/** + * Argument package for rtHeapPageAllocCallback. + */ +typedef struct RTHEAPPAGEALLOCARGS +{ + /** The number of pages to allocate. */ + size_t cPages; + /** Non-null on success. */ + void *pvAlloc; + /** RTMEMPAGEALLOC_F_XXX. */ + uint32_t fFlags; +} RTHEAPPAGEALLOCARGS; + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +/** Initialize once structure. */ +static RTONCE g_MemPagePosixInitOnce = RTONCE_INITIALIZER; +/** The page heap. */ +static RTHEAPPAGE g_MemPagePosixHeap; +/** The exec page heap. */ +static RTHEAPPAGE g_MemExecPosixHeap; + + +#ifdef RT_OS_OS2 +/* + * A quick mmap/munmap mockup for avoid duplicating lots of good code. + */ +# define INCL_BASE +# include <os2.h> +# undef MAP_PRIVATE +# define MAP_PRIVATE 0 +# undef MAP_ANONYMOUS +# define MAP_ANONYMOUS 0 +# undef MAP_FAILED +# define MAP_FAILED (void *)-1 +# undef mmap +# define mmap iprt_mmap +# undef munmap +# define munmap iprt_munmap + +static void *mmap(void *pvWhere, size_t cb, int fProt, int fFlags, int fd, off_t off) +{ + NOREF(pvWhere); NOREF(fd); NOREF(off); + void *pv = NULL; + ULONG fAlloc = OBJ_ANY | PAG_COMMIT; + if (fProt & PROT_EXEC) + fAlloc |= PAG_EXECUTE; + if (fProt & PROT_READ) + fAlloc |= PAG_READ; + if (fProt & PROT_WRITE) + fAlloc |= PAG_WRITE; + APIRET rc = DosAllocMem(&pv, cb, fAlloc); + if (rc == NO_ERROR) + return pv; + errno = ENOMEM; + return MAP_FAILED; +} + +static int munmap(void *pv, size_t cb) +{ + APIRET rc = DosFreeMem(pv); + if (rc == NO_ERROR) + return 0; + errno = EINVAL; + return -1; +} + +#endif + +/** + * Initializes the heap. + * + * @returns IPRT status code. + * @param pHeap The page heap to initialize. + * @param fExec Whether the heap memory should be marked as + * executable or not. + */ +int RTHeapPageInit(PRTHEAPPAGE pHeap, bool fExec) +{ + int rc = RTCritSectInitEx(&pHeap->CritSect, + RTCRITSECT_FLAGS_NO_LOCK_VAL | RTCRITSECT_FLAGS_NO_NESTING | RTCRITSECT_FLAGS_BOOTSTRAP_HACK, + NIL_RTLOCKVALCLASS, RTLOCKVAL_SUB_CLASS_NONE, NULL); + if (RT_SUCCESS(rc)) + { + pHeap->cHeapPages = 0; + pHeap->cFreePages = 0; + pHeap->cAllocCalls = 0; + pHeap->cFreeCalls = 0; + pHeap->uLastMinimizeCall = 0; + pHeap->BlockTree = NULL; + pHeap->fExec = fExec; + pHeap->u32Magic = RTHEAPPAGE_MAGIC; + } + return rc; +} + + +/** + * Deletes the heap and all the memory it tracks. + * + * @returns IPRT status code. + * @param pHeap The page heap to delete. + */ +int RTHeapPageDelete(PRTHEAPPAGE pHeap) +{ + NOREF(pHeap); + return VERR_NOT_IMPLEMENTED; +} + + +/** + * Applies flags to an allocation. + * + * @param pv The allocation. + * @param cb The size of the allocation (page aligned). + * @param fFlags RTMEMPAGEALLOC_F_XXX. + */ +DECLINLINE(void) rtMemPagePosixApplyFlags(void *pv, size_t cb, uint32_t fFlags) +{ +#ifndef RT_OS_OS2 + if (fFlags & RTMEMPAGEALLOC_F_ADVISE_LOCKED) + { + int rc = mlock(pv, cb); +# ifndef RT_OS_SOLARIS /* mlock(3C) on Solaris requires the priv_lock_memory privilege */ + AssertMsg(rc == 0, ("mlock %p LB %#zx -> %d errno=%d\n", pv, cb, rc, errno)); +# endif + NOREF(rc); + } + +# ifdef MADV_DONTDUMP + if (fFlags & RTMEMPAGEALLOC_F_ADVISE_NO_DUMP) + { + int rc = madvise(pv, cb, MADV_DONTDUMP); + AssertMsg(rc == 0, ("madvice %p LB %#zx MADV_DONTDUMP -> %d errno=%d\n", pv, cb, rc, errno)); + NOREF(rc); + } +# endif +#endif + + if (fFlags & RTMEMPAGEALLOC_F_ZERO) + RT_BZERO(pv, cb); +} + + +/** + * Avoids some gotos in rtHeapPageAllocFromBlock. + * + * @returns VINF_SUCCESS. + * @param pBlock The block. + * @param iPage The page to start allocating at. + * @param cPages The number of pages. + * @param fFlags RTMEMPAGEALLOC_F_XXX. + * @param ppv Where to return the allocation address. + */ +DECLINLINE(int) rtHeapPageAllocFromBlockSuccess(PRTHEAPPAGEBLOCK pBlock, uint32_t iPage, size_t cPages, uint32_t fFlags, void **ppv) +{ + PRTHEAPPAGE pHeap = pBlock->pHeap; + + ASMBitSet(&pBlock->bmFirst[0], iPage); + pBlock->cFreePages -= cPages; + pHeap->cFreePages -= cPages; + if (!pHeap->pHint2 || pHeap->pHint2->cFreePages < pBlock->cFreePages) + pHeap->pHint2 = pBlock; + pHeap->cAllocCalls++; + + void *pv = (uint8_t *)pBlock->Core.Key + (iPage << PAGE_SHIFT); + *ppv = pv; + + if (fFlags) + rtMemPagePosixApplyFlags(pv, cPages << PAGE_SHIFT, fFlags); + + return VINF_SUCCESS; +} + + +/** + * Checks if a page range is free in the specified block. + * + * @returns @c true if the range is free, @c false if not. + * @param pBlock The block. + * @param iFirst The first page to check. + * @param cPages The number of pages to check. + */ +DECLINLINE(bool) rtHeapPageIsPageRangeFree(PRTHEAPPAGEBLOCK pBlock, uint32_t iFirst, uint32_t cPages) +{ + uint32_t i = iFirst + cPages; + while (i-- > iFirst) + { + if (ASMBitTest(&pBlock->bmAlloc[0], i)) + return false; + Assert(!ASMBitTest(&pBlock->bmFirst[0], i)); + } + return true; +} + + +/** + * Tries to allocate a chunk of pages from a heap block. + * + * @retval VINF_SUCCESS on success. + * @retval VERR_NO_MEMORY if the allocation failed. + * @param pBlock The block to allocate from. + * @param cPages The size of the allocation. + * @param fFlags RTMEMPAGEALLOC_F_XXX. + * @param ppv Where to return the allocation address on success. + */ +DECLINLINE(int) rtHeapPageAllocFromBlock(PRTHEAPPAGEBLOCK pBlock, size_t cPages, uint32_t fFlags, void **ppv) +{ + if (pBlock->cFreePages >= cPages) + { + int iPage = ASMBitFirstClear(&pBlock->bmAlloc[0], RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT); + Assert(iPage >= 0); + + /* special case: single page. */ + if (cPages == 1) + { + ASMBitSet(&pBlock->bmAlloc[0], iPage); + return rtHeapPageAllocFromBlockSuccess(pBlock, iPage, cPages, fFlags, ppv); + } + + while ( iPage >= 0 + && (unsigned)iPage <= RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT - cPages) + { + if (rtHeapPageIsPageRangeFree(pBlock, iPage + 1, cPages - 1)) + { + ASMBitSetRange(&pBlock->bmAlloc[0], iPage, iPage + cPages); + return rtHeapPageAllocFromBlockSuccess(pBlock, iPage, cPages, fFlags, ppv); + } + + /* next */ + iPage = ASMBitNextSet(&pBlock->bmAlloc[0], RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT, iPage); + if (iPage < 0 || iPage >= RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT - 1) + break; + iPage = ASMBitNextClear(&pBlock->bmAlloc[0], RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT, iPage); + } + } + + return VERR_NO_MEMORY; +} + + +/** + * RTAvlrPVDoWithAll callback. + * + * @returns 0 to continue the enum, non-zero to quit it. + * @param pNode The node. + * @param pvUser The user argument. + */ +static DECLCALLBACK(int) rtHeapPageAllocCallback(PAVLRPVNODECORE pNode, void *pvUser) +{ + PRTHEAPPAGEBLOCK pBlock = RT_FROM_MEMBER(pNode, RTHEAPPAGEBLOCK, Core); + RTHEAPPAGEALLOCARGS *pArgs = (RTHEAPPAGEALLOCARGS *)pvUser; + int rc = rtHeapPageAllocFromBlock(pBlock, pArgs->cPages, pArgs->fFlags, &pArgs->pvAlloc); + return RT_SUCCESS(rc) ? 1 : 0; +} + + +/** + * Worker for RTHeapPageAlloc. + * + * @returns IPRT status code + * @param pHeap The heap - locked. + * @param cPages The page count. + * @param pszTag The tag. + * @param fFlags RTMEMPAGEALLOC_F_XXX. + * @param ppv Where to return the address of the allocation + * on success. + */ +static int rtHeapPageAllocLocked(PRTHEAPPAGE pHeap, size_t cPages, const char *pszTag, uint32_t fFlags, void **ppv) +{ + int rc; + NOREF(pszTag); + + /* + * Use the hints first. + */ + if (pHeap->pHint1) + { + rc = rtHeapPageAllocFromBlock(pHeap->pHint1, cPages, fFlags, ppv); + if (rc != VERR_NO_MEMORY) + return rc; + } + if (pHeap->pHint2) + { + rc = rtHeapPageAllocFromBlock(pHeap->pHint2, cPages, fFlags, ppv); + if (rc != VERR_NO_MEMORY) + return rc; + } + + /* + * Search the heap for a block with enough free space. + * + * N.B. This search algorithm is not optimal at all. What (hopefully) saves + * it are the two hints above. + */ + if (pHeap->cFreePages >= cPages) + { + RTHEAPPAGEALLOCARGS Args; + Args.cPages = cPages; + Args.pvAlloc = NULL; + Args.fFlags = fFlags; + RTAvlrPVDoWithAll(&pHeap->BlockTree, true /*fFromLeft*/, rtHeapPageAllocCallback, &Args); + if (Args.pvAlloc) + { + *ppv = Args.pvAlloc; + return VINF_SUCCESS; + } + } + + /* + * Didn't find anytyhing, so expand the heap with a new block. + */ + RTCritSectLeave(&pHeap->CritSect); + void *pvPages; + pvPages = mmap(NULL, RTMEMPAGEPOSIX_BLOCK_SIZE, + PROT_READ | PROT_WRITE | (pHeap->fExec ? PROT_EXEC : 0), + MAP_PRIVATE | MAP_ANONYMOUS, + -1, 0); + if (pvPages == MAP_FAILED) + { + RTCritSectEnter(&pHeap->CritSect); + return RTErrConvertFromErrno(errno); + + } + /** @todo Eliminate this rtMemBaseAlloc dependency! */ + PRTHEAPPAGEBLOCK pBlock; +#ifdef RTALLOC_REPLACE_MALLOC + if (g_pfnOrgMalloc) + pBlock = (PRTHEAPPAGEBLOCK)g_pfnOrgMalloc(sizeof(*pBlock)); + else +#endif + pBlock = (PRTHEAPPAGEBLOCK)rtMemBaseAlloc(sizeof(*pBlock)); + if (!pBlock) + { + munmap(pvPages, RTMEMPAGEPOSIX_BLOCK_SIZE); + RTCritSectEnter(&pHeap->CritSect); + return VERR_NO_MEMORY; + } + + RT_ZERO(*pBlock); + pBlock->Core.Key = pvPages; + pBlock->Core.KeyLast = (uint8_t *)pvPages + RTMEMPAGEPOSIX_BLOCK_SIZE - 1; + pBlock->cFreePages = RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT; + pBlock->pHeap = pHeap; + + RTCritSectEnter(&pHeap->CritSect); + + bool fRc = RTAvlrPVInsert(&pHeap->BlockTree, &pBlock->Core); Assert(fRc); NOREF(fRc); + pHeap->cFreePages += RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT; + pHeap->cHeapPages += RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT; + + /* + * Grab memory from the new block (cannot fail). + */ + rc = rtHeapPageAllocFromBlock(pBlock, cPages, fFlags, ppv); + Assert(rc == VINF_SUCCESS); + + return rc; +} + + +/** + * Allocates one or more pages off the heap. + * + * @returns IPRT status code. + * @param pHeap The page heap. + * @param cPages The number of pages to allocate. + * @param pszTag The allocation tag. + * @param fFlags RTMEMPAGEALLOC_F_XXX. + * @param ppv Where to return the pointer to the pages. + */ +int RTHeapPageAlloc(PRTHEAPPAGE pHeap, size_t cPages, const char *pszTag, uint32_t fFlags, void **ppv) +{ + /* + * Validate input. + */ + AssertPtr(ppv); + *ppv = NULL; + AssertPtrReturn(pHeap, VERR_INVALID_HANDLE); + AssertReturn(pHeap->u32Magic == RTHEAPPAGE_MAGIC, VERR_INVALID_HANDLE); + AssertMsgReturn(cPages < RTMEMPAGEPOSIX_BLOCK_SIZE, ("%#zx\n", cPages), VERR_OUT_OF_RANGE); + + /* + * Grab the lock and call a worker with many returns. + */ + int rc = RTCritSectEnter(&pHeap->CritSect); + if (RT_SUCCESS(rc)) + { + rc = rtHeapPageAllocLocked(pHeap, cPages, pszTag, fFlags, ppv); + RTCritSectLeave(&pHeap->CritSect); + } + + return rc; +} + + +/** + * RTAvlrPVDoWithAll callback. + * + * @returns 0 to continue the enum, non-zero to quit it. + * @param pNode The node. + * @param pvUser Pointer to a block pointer variable. For returning + * the address of the block to be freed. + */ +static DECLCALLBACK(int) rtHeapPageFindUnusedBlockCallback(PAVLRPVNODECORE pNode, void *pvUser) +{ + PRTHEAPPAGEBLOCK pBlock = RT_FROM_MEMBER(pNode, RTHEAPPAGEBLOCK, Core); + if (pBlock->cFreePages == RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT) + { + *(PRTHEAPPAGEBLOCK *)pvUser = pBlock; + return 1; + } + return 0; +} + + +/** + * Allocates one or more pages off the heap. + * + * @returns IPRT status code. + * @param pHeap The page heap. + * @param pv Pointer to what RTHeapPageAlloc returned. + * @param cPages The number of pages that was allocated. + */ +int RTHeapPageFree(PRTHEAPPAGE pHeap, void *pv, size_t cPages) +{ + /* + * Validate input. + */ + if (!pv) + return VINF_SUCCESS; + AssertPtrReturn(pHeap, VERR_INVALID_HANDLE); + AssertReturn(pHeap->u32Magic == RTHEAPPAGE_MAGIC, VERR_INVALID_HANDLE); + + /* + * Grab the lock and look up the page. + */ + int rc = RTCritSectEnter(&pHeap->CritSect); + if (RT_SUCCESS(rc)) + { + PRTHEAPPAGEBLOCK pBlock = (PRTHEAPPAGEBLOCK)RTAvlrPVRangeGet(&pHeap->BlockTree, pv); + if (pBlock) + { + /* + * Validate the specified address range. + */ + uint32_t const iPage = (uint32_t)(((uintptr_t)pv - (uintptr_t)pBlock->Core.Key) >> PAGE_SHIFT); + /* Check the range is within the block. */ + bool fOk = iPage + cPages <= RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT; + /* Check that it's the start of an allocation. */ + fOk = fOk && ASMBitTest(&pBlock->bmFirst[0], iPage); + /* Check that the range ends at an allocation boundrary. */ + fOk = fOk && ( iPage + cPages == RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT + || ASMBitTest(&pBlock->bmFirst[0], iPage + cPages) + || !ASMBitTest(&pBlock->bmAlloc[0], iPage + cPages)); + /* Check the other pages. */ + uint32_t const iLastPage = iPage + cPages - 1; + for (uint32_t i = iPage + 1; i < iLastPage && fOk; i++) + fOk = ASMBitTest(&pBlock->bmAlloc[0], i) + && !ASMBitTest(&pBlock->bmFirst[0], i); + if (fOk) + { + /* + * Free the memory. + */ + ASMBitClearRange(&pBlock->bmAlloc[0], iPage, iPage + cPages); + ASMBitClear(&pBlock->bmFirst[0], iPage); + pBlock->cFreePages += cPages; + pHeap->cFreePages += cPages; + pHeap->cFreeCalls++; + if (!pHeap->pHint1 || pHeap->pHint1->cFreePages < pBlock->cFreePages) + pHeap->pHint1 = pBlock; + + /** @todo Add bitmaps for tracking madvice and mlock so we can undo those. */ + + /* + * Shrink the heap. Not very efficient because of the AVL tree. + */ + if ( pHeap->cFreePages >= RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT * 3 + && pHeap->cFreePages >= pHeap->cHeapPages / 2 /* 50% free */ + && pHeap->cFreeCalls - pHeap->uLastMinimizeCall > RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT + ) + { + uint32_t cFreePageTarget = pHeap->cHeapPages / 4; /* 25% free */ + while (pHeap->cFreePages > cFreePageTarget) + { + pHeap->uLastMinimizeCall = pHeap->cFreeCalls; + + pBlock = NULL; + RTAvlrPVDoWithAll(&pHeap->BlockTree, false /*fFromLeft*/, + rtHeapPageFindUnusedBlockCallback, &pBlock); + if (!pBlock) + break; + + void *pv2 = RTAvlrPVRemove(&pHeap->BlockTree, pBlock->Core.Key); Assert(pv2); NOREF(pv2); + pHeap->cHeapPages -= RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT; + pHeap->cFreePages -= RTMEMPAGEPOSIX_BLOCK_PAGE_COUNT; + pHeap->pHint1 = NULL; + pHeap->pHint2 = NULL; + RTCritSectLeave(&pHeap->CritSect); + + munmap(pBlock->Core.Key, RTMEMPAGEPOSIX_BLOCK_SIZE); + pBlock->Core.Key = pBlock->Core.KeyLast = NULL; + pBlock->cFreePages = 0; +#ifdef RTALLOC_REPLACE_MALLOC + if (g_pfnOrgFree) + g_pfnOrgFree(pBlock); + else +#endif + rtMemBaseFree(pBlock); + + RTCritSectEnter(&pHeap->CritSect); + } + } + } + else + rc = VERR_INVALID_POINTER; + } + else + rc = VERR_INVALID_POINTER; + + RTCritSectLeave(&pHeap->CritSect); + } + + return rc; +} + + +/** + * Initializes the heap. + * + * @returns IPRT status code + * @param pvUser Unused. + */ +static DECLCALLBACK(int) rtMemPagePosixInitOnce(void *pvUser) +{ + NOREF(pvUser); + int rc = RTHeapPageInit(&g_MemPagePosixHeap, false /*fExec*/); + if (RT_SUCCESS(rc)) + { + rc = RTHeapPageInit(&g_MemExecPosixHeap, true /*fExec*/); + if (RT_SUCCESS(rc)) + return rc; + RTHeapPageDelete(&g_MemPagePosixHeap); + } + return rc; +} + + +/** + * Allocates memory from the specified heap. + * + * @returns Address of the allocated memory. + * @param cb The number of bytes to allocate. + * @param pszTag The tag. + * @param fFlags RTMEMPAGEALLOC_F_XXX. + * @param pHeap The heap to use. + */ +static void *rtMemPagePosixAlloc(size_t cb, const char *pszTag, uint32_t fFlags, PRTHEAPPAGE pHeap) +{ + /* + * Validate & adjust the input. + */ + Assert(cb > 0); + NOREF(pszTag); + cb = RT_ALIGN_Z(cb, PAGE_SIZE); + + /* + * If the allocation is relatively large, we use mmap/munmap directly. + */ + void *pv; + if (cb >= RTMEMPAGEPOSIX_MMAP_THRESHOLD) + { + + pv = mmap(NULL, cb, + PROT_READ | PROT_WRITE | (pHeap == &g_MemExecPosixHeap ? PROT_EXEC : 0), + MAP_PRIVATE | MAP_ANONYMOUS, + -1, 0); + if (pv != MAP_FAILED) + { + AssertPtr(pv); + + if (fFlags) + rtMemPagePosixApplyFlags(pv, cb, fFlags); + } + else + pv = NULL; + } + else + { + int rc = RTOnce(&g_MemPagePosixInitOnce, rtMemPagePosixInitOnce, NULL); + if (RT_SUCCESS(rc)) + rc = RTHeapPageAlloc(pHeap, cb >> PAGE_SHIFT, pszTag, fFlags, &pv); + if (RT_FAILURE(rc)) + pv = NULL; + } + + return pv; +} + + +/** + * Free memory allocated by rtMemPagePosixAlloc. + * + * @param pv The address of the memory to free. + * @param cb The size. + * @param pHeap The heap. + */ +static void rtMemPagePosixFree(void *pv, size_t cb, PRTHEAPPAGE pHeap) +{ + /* + * Validate & adjust the input. + */ + if (!pv) + return; + AssertPtr(pv); + Assert(cb > 0); + Assert(!((uintptr_t)pv & PAGE_OFFSET_MASK)); + cb = RT_ALIGN_Z(cb, PAGE_SIZE); + + /* + * If the allocation is relatively large, we use mmap/munmap directly. + */ + if (cb >= RTMEMPAGEPOSIX_MMAP_THRESHOLD) + { + int rc = munmap(pv, cb); + AssertMsg(rc == 0, ("rc=%d pv=%p cb=%#zx\n", rc, pv, cb)); NOREF(rc); + } + else + { + int rc = RTHeapPageFree(pHeap, pv, cb >> PAGE_SHIFT); + AssertRC(rc); + } +} + + + + + +RTDECL(void *) RTMemPageAllocTag(size_t cb, const char *pszTag) RT_NO_THROW_DEF +{ + return rtMemPagePosixAlloc(cb, pszTag, 0, &g_MemPagePosixHeap); +} + + +RTDECL(void *) RTMemPageAllocZTag(size_t cb, const char *pszTag) RT_NO_THROW_DEF +{ + return rtMemPagePosixAlloc(cb, pszTag, RTMEMPAGEALLOC_F_ZERO, &g_MemPagePosixHeap); +} + + +RTDECL(void *) RTMemPageAllocExTag(size_t cb, uint32_t fFlags, const char *pszTag) RT_NO_THROW_DEF +{ + AssertReturn(!(fFlags & ~RTMEMPAGEALLOC_F_VALID_MASK), NULL); + return rtMemPagePosixAlloc(cb, pszTag, fFlags, &g_MemPagePosixHeap); +} + + +RTDECL(void) RTMemPageFree(void *pv, size_t cb) RT_NO_THROW_DEF +{ + return rtMemPagePosixFree(pv, cb, &g_MemPagePosixHeap); +} + diff --git a/src/VBox/Runtime/r3/posix/rtmempage-exec-mmap-posix.cpp b/src/VBox/Runtime/r3/posix/rtmempage-exec-mmap-posix.cpp new file mode 100644 index 00000000..abe285a2 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/rtmempage-exec-mmap-posix.cpp @@ -0,0 +1,182 @@ +/* $Id: rtmempage-exec-mmap-posix.cpp $ */ +/** @file + * IPRT - RTMemPage*, POSIX with mmap only. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include "internal/iprt.h" +#include <iprt/mem.h> + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/errcore.h> +#include <iprt/param.h> +#include <iprt/string.h> + +#include <stdlib.h> +#include <errno.h> +#include <sys/mman.h> +#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON) +# define MAP_ANONYMOUS MAP_ANON +#endif + + +/** + * Applies flags to an allocation. + * + * @param pv The allocation. + * @param cb The size of the allocation (page aligned). + * @param fFlags RTMEMPAGEALLOC_F_XXX. + */ +DECLINLINE(void) rtMemPagePosixApplyFlags(void *pv, size_t cb, uint32_t fFlags) +{ +#ifndef RT_OS_OS2 + if (fFlags & RTMEMPAGEALLOC_F_ADVISE_LOCKED) + { + int rc = mlock(pv, cb); +# ifndef RT_OS_SOLARIS /* mlock(3C) on Solaris requires the priv_lock_memory privilege */ + AssertMsg(rc == 0, ("mlock %p LB %#zx -> %d errno=%d\n", pv, cb, rc, errno)); +# endif + NOREF(rc); + } + +# ifdef MADV_DONTDUMP + if (fFlags & RTMEMPAGEALLOC_F_ADVISE_NO_DUMP) + { + int rc = madvise(pv, cb, MADV_DONTDUMP); + AssertMsg(rc == 0, ("madvice %p LB %#zx MADV_DONTDUMP -> %d errno=%d\n", pv, cb, rc, errno)); + NOREF(rc); + } +# endif +#endif + + if (fFlags & RTMEMPAGEALLOC_F_ZERO) + RT_BZERO(pv, cb); +} + + +/** + * Allocates memory from the specified heap. + * + * @returns Address of the allocated memory. + * @param cb The number of bytes to allocate. + * @param pszTag The tag. + * @param fFlags RTMEMPAGEALLOC_F_XXX. + * @param fProtExec PROT_EXEC or 0. + */ +static void *rtMemPagePosixAlloc(size_t cb, const char *pszTag, uint32_t fFlags, int fProtExec) +{ + /* + * Validate & adjust the input. + */ + Assert(cb > 0); + NOREF(pszTag); + cb = RT_ALIGN_Z(cb, PAGE_SIZE); + + /* + * Do the allocation. + */ + void *pv = mmap(NULL, cb, + PROT_READ | PROT_WRITE | fProtExec, + MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); + if (pv != MAP_FAILED) + { + AssertPtr(pv); + + if (fFlags) + rtMemPagePosixApplyFlags(pv, cb, fFlags); + } + else + pv = NULL; + + return pv; +} + + +/** + * Free memory allocated by rtMemPagePosixAlloc. + * + * @param pv The address of the memory to free. + * @param cb The size. + */ +static void rtMemPagePosixFree(void *pv, size_t cb) +{ + /* + * Validate & adjust the input. + */ + if (!pv) + return; + AssertPtr(pv); + Assert(cb > 0); + Assert(!((uintptr_t)pv & PAGE_OFFSET_MASK)); + cb = RT_ALIGN_Z(cb, PAGE_SIZE); + + /* + * Free the memory. + */ + int rc = munmap(pv, cb); + AssertMsg(rc == 0, ("rc=%d pv=%p cb=%#zx\n", rc, pv, cb)); NOREF(rc); +} + + + + + +RTDECL(void *) RTMemPageAllocTag(size_t cb, const char *pszTag) RT_NO_THROW_DEF +{ + return rtMemPagePosixAlloc(cb, pszTag, 0, 0); +} + + +RTDECL(void *) RTMemPageAllocZTag(size_t cb, const char *pszTag) RT_NO_THROW_DEF +{ + return rtMemPagePosixAlloc(cb, pszTag, RTMEMPAGEALLOC_F_ZERO, 0); +} + + +RTDECL(void *) RTMemPageAllocExTag(size_t cb, uint32_t fFlags, const char *pszTag) RT_NO_THROW_DEF +{ + AssertReturn(!(fFlags & ~RTMEMPAGEALLOC_F_VALID_MASK), NULL); + return rtMemPagePosixAlloc(cb, pszTag, fFlags, 0); +} + + +RTDECL(void) RTMemPageFree(void *pv, size_t cb) RT_NO_THROW_DEF +{ + return rtMemPagePosixFree(pv, cb); +} + diff --git a/src/VBox/Runtime/r3/posix/sched-posix.cpp b/src/VBox/Runtime/r3/posix/sched-posix.cpp new file mode 100644 index 00000000..b30b9c67 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/sched-posix.cpp @@ -0,0 +1,849 @@ +/* $Id: sched-posix.cpp $ */ +/** @file + * IPRT - Scheduling, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + +/* + * !WARNING! + * + * When talking about lowering and raising priority, we do *NOT* refer to + * the common direction priority values takes on unix systems (lower means + * higher). So, when we raise the priority of a linux thread the nice + * value will decrease, and when we lower the priority the nice value + * will increase. Confusing, right? + * + * !WARNING! + */ + + + +/** @def THREAD_LOGGING + * Be very careful with enabling this, it may cause deadlocks when combined + * with the 'thread' logging prefix. + */ +#ifdef DOXYGEN_RUNNING +#define THREAD_LOGGING +#endif + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_THREAD +#include <errno.h> +#include <pthread.h> +#include <sched.h> +#include <unistd.h> +#include <sys/resource.h> + +#include <iprt/thread.h> +#include <iprt/process.h> +#include <iprt/semaphore.h> +#include <iprt/string.h> +#include <iprt/assert.h> +#include <iprt/log.h> +#include <iprt/err.h> +#include "internal/sched.h" +#include "internal/thread.h" + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ + +/** Array scheduler attributes corresponding to each of the thread types. */ +typedef struct PROCPRIORITYTYPE +{ + /** For sanity include the array index. */ + RTTHREADTYPE enmType; + /** The thread priority or nice delta - depends on which priority type. */ + int iPriority; +} PROCPRIORITYTYPE; + + +/** + * Configuration of one priority. + */ +typedef struct +{ + /** The priority. */ + RTPROCPRIORITY enmPriority; + /** The name of this priority. */ + const char *pszName; + /** The process nice value. */ + int iNice; + /** The delta applied to the iPriority value. */ + int iDelta; + /** Array scheduler attributes corresponding to each of the thread types. */ + const PROCPRIORITYTYPE *paTypes; +} PROCPRIORITY; + + +/** + * Saved priority settings + */ +typedef struct +{ + /** Process priority. */ + int iPriority; + /** Process level. */ + struct sched_param SchedParam; + /** Process level. */ + int iPolicy; + /** pthread level. */ + struct sched_param PthreadSchedParam; + /** pthread level. */ + int iPthreadPolicy; +} SAVEDPRIORITY, *PSAVEDPRIORITY; + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +/** + * Thread level priorities based on a 0..31 priority range + * as specified as the minimum for SCHED_RR/FIFO. FreeBSD + * seems to be using this (needs more research to be + * certain). + */ +static const PROCPRIORITYTYPE g_aTypesThread[RTTHREADTYPE_END] = +{ + { RTTHREADTYPE_INVALID, -999999999 }, + { RTTHREADTYPE_INFREQUENT_POLLER, 5 }, + { RTTHREADTYPE_MAIN_HEAVY_WORKER, 12 }, + { RTTHREADTYPE_EMULATION, 14 }, + { RTTHREADTYPE_DEFAULT, 15 }, + { RTTHREADTYPE_GUI, 16 }, + { RTTHREADTYPE_MAIN_WORKER, 18 }, + { RTTHREADTYPE_VRDP_IO, 24 }, + { RTTHREADTYPE_DEBUGGER, 28 }, + { RTTHREADTYPE_MSG_PUMP, 29 }, + { RTTHREADTYPE_IO, 30 }, + { RTTHREADTYPE_TIMER, 31 } +}; + +static const PROCPRIORITYTYPE g_aTypesThreadFlat[RTTHREADTYPE_END] = +{ + { RTTHREADTYPE_INVALID, ~0 }, + { RTTHREADTYPE_INFREQUENT_POLLER, 15 }, + { RTTHREADTYPE_MAIN_HEAVY_WORKER, 15 }, + { RTTHREADTYPE_EMULATION, 15 }, + { RTTHREADTYPE_DEFAULT, 15 }, + { RTTHREADTYPE_GUI, 15 }, + { RTTHREADTYPE_MAIN_WORKER, 15 }, + { RTTHREADTYPE_VRDP_IO, 15 }, + { RTTHREADTYPE_DEBUGGER, 15 }, + { RTTHREADTYPE_MSG_PUMP, 15 }, + { RTTHREADTYPE_IO, 15 }, + { RTTHREADTYPE_TIMER, 15 } +}; + +/** + * Process and thread level priority, full access at thread level. + */ +static const PROCPRIORITY g_aProcessAndThread[] = +{ + { RTPROCPRIORITY_FLAT, "Flat", 0, 0, g_aTypesThreadFlat }, + { RTPROCPRIORITY_LOW, "Low", 9, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 11, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 15, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 17, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 19, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 7, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 5, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 3, 0, g_aTypesThread }, + { RTPROCPRIORITY_LOW, "Low", 1, 0, g_aTypesThread }, + { RTPROCPRIORITY_NORMAL, "Normal", 0, 0, g_aTypesThread }, + { RTPROCPRIORITY_NORMAL, "Normal", 0, 0, g_aTypesThreadFlat }, + { RTPROCPRIORITY_HIGH, "High", -9, 0, g_aTypesThread }, + { RTPROCPRIORITY_HIGH, "High", -7, 0, g_aTypesThread }, + { RTPROCPRIORITY_HIGH, "High", -5, 0, g_aTypesThread }, + { RTPROCPRIORITY_HIGH, "High", -3, 0, g_aTypesThread }, + { RTPROCPRIORITY_HIGH, "High", -1, 0, g_aTypesThread }, + { RTPROCPRIORITY_HIGH, "High", -9, 0, g_aTypesThreadFlat }, + { RTPROCPRIORITY_HIGH, "High", -1, 0, g_aTypesThreadFlat } +}; + +/** + * Deltas for a process in which we are not restricted + * to only be lowering the priority. + */ +static const PROCPRIORITYTYPE g_aTypesUnixFree[RTTHREADTYPE_END] = +{ + { RTTHREADTYPE_INVALID, -999999999 }, + { RTTHREADTYPE_INFREQUENT_POLLER, +3 }, + { RTTHREADTYPE_MAIN_HEAVY_WORKER, +2 }, + { RTTHREADTYPE_EMULATION, +1 }, + { RTTHREADTYPE_DEFAULT, 0 }, + { RTTHREADTYPE_GUI, 0 }, + { RTTHREADTYPE_MAIN_WORKER, 0 }, + { RTTHREADTYPE_VRDP_IO, -1 }, + { RTTHREADTYPE_DEBUGGER, -1 }, + { RTTHREADTYPE_MSG_PUMP, -2 }, + { RTTHREADTYPE_IO, -3 }, + { RTTHREADTYPE_TIMER, -4 } +}; + +/** + * Deltas for a process in which we are restricted + * to only be lowering the priority. + */ +static const PROCPRIORITYTYPE g_aTypesUnixRestricted[RTTHREADTYPE_END] = +{ + { RTTHREADTYPE_INVALID, -999999999 }, + { RTTHREADTYPE_INFREQUENT_POLLER, +3 }, + { RTTHREADTYPE_MAIN_HEAVY_WORKER, +2 }, + { RTTHREADTYPE_EMULATION, +1 }, + { RTTHREADTYPE_DEFAULT, 0 }, + { RTTHREADTYPE_GUI, 0 }, + { RTTHREADTYPE_MAIN_WORKER, 0 }, + { RTTHREADTYPE_VRDP_IO, 0 }, + { RTTHREADTYPE_DEBUGGER, 0 }, + { RTTHREADTYPE_MSG_PUMP, 0 }, + { RTTHREADTYPE_IO, 0 }, + { RTTHREADTYPE_TIMER, 0 } +}; + +/** + * Deltas for a process in which we are restricted + * to only be lowering the priority. + */ +static const PROCPRIORITYTYPE g_aTypesUnixFlat[RTTHREADTYPE_END] = +{ + { RTTHREADTYPE_INVALID, -999999999 }, + { RTTHREADTYPE_INFREQUENT_POLLER, 0 }, + { RTTHREADTYPE_MAIN_HEAVY_WORKER, 0 }, + { RTTHREADTYPE_EMULATION, 0 }, + { RTTHREADTYPE_DEFAULT, 0 }, + { RTTHREADTYPE_GUI, 0 }, + { RTTHREADTYPE_MAIN_WORKER, 0 }, + { RTTHREADTYPE_VRDP_IO, 0 }, + { RTTHREADTYPE_DEBUGGER, 0 }, + { RTTHREADTYPE_MSG_PUMP, 0 }, + { RTTHREADTYPE_IO, 0 }, + { RTTHREADTYPE_TIMER, 0 } +}; + +/** + * Process and thread level priority, full access at thread level. + */ +static const PROCPRIORITY g_aUnixConfigs[] = +{ + { RTPROCPRIORITY_FLAT, "Flat", 0, 0, g_aTypesUnixFlat }, + { RTPROCPRIORITY_LOW, "Low", 9, 9, g_aTypesUnixFree }, + { RTPROCPRIORITY_LOW, "Low", 9, 9, g_aTypesUnixFlat }, + { RTPROCPRIORITY_LOW, "Low", 15, 15, g_aTypesUnixFree }, + { RTPROCPRIORITY_LOW, "Low", 15, 15, g_aTypesUnixFlat }, + { RTPROCPRIORITY_LOW, "Low", 17, 17, g_aTypesUnixFree }, + { RTPROCPRIORITY_LOW, "Low", 17, 17, g_aTypesUnixFlat }, + { RTPROCPRIORITY_LOW, "Low", 19, 19, g_aTypesUnixFlat }, + { RTPROCPRIORITY_LOW, "Low", 9, 9, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_LOW, "Low", 15, 15, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_LOW, "Low", 17, 17, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_NORMAL, "Normal", 0, 0, g_aTypesUnixFree }, + { RTPROCPRIORITY_NORMAL, "Normal", 0, 0, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_NORMAL, "Normal", 0, 0, g_aTypesUnixFlat }, + { RTPROCPRIORITY_HIGH, "High", -9, -9, g_aTypesUnixFree }, + { RTPROCPRIORITY_HIGH, "High", -7, -7, g_aTypesUnixFree }, + { RTPROCPRIORITY_HIGH, "High", -5, -5, g_aTypesUnixFree }, + { RTPROCPRIORITY_HIGH, "High", -3, -3, g_aTypesUnixFree }, + { RTPROCPRIORITY_HIGH, "High", -1, -1, g_aTypesUnixFree }, + { RTPROCPRIORITY_HIGH, "High", -9, -9, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_HIGH, "High", -7, -7, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_HIGH, "High", -5, -5, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_HIGH, "High", -3, -3, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_HIGH, "High", -1, -1, g_aTypesUnixRestricted }, + { RTPROCPRIORITY_HIGH, "High", -9, -9, g_aTypesUnixFlat }, + { RTPROCPRIORITY_HIGH, "High", -7, -7, g_aTypesUnixFlat }, + { RTPROCPRIORITY_HIGH, "High", -5, -5, g_aTypesUnixFlat }, + { RTPROCPRIORITY_HIGH, "High", -3, -3, g_aTypesUnixFlat }, + { RTPROCPRIORITY_HIGH, "High", -1, -1, g_aTypesUnixFlat } +}; + +/** + * The dynamic default priority configuration. + * + * This will be recalulated at runtime depending on what the + * system allow us to do and what the current priority is. + */ +static PROCPRIORITY g_aDefaultPriority = +{ + RTPROCPRIORITY_LOW, "Default", 0, 0, g_aTypesUnixRestricted +}; + +/** Pointer to the current priority configuration. */ +static const PROCPRIORITY *g_pProcessPriority = &g_aDefaultPriority; + + +/** Set to what kind of scheduling priority support the host + * OS seems to be offering. Determined at runtime. + */ +static enum +{ + OSPRIOSUP_UNDETERMINED = 0, + /** An excellent combination of process and thread level + * I.e. setpriority() works on process level, one have to be supervisor + * to raise priority as is the custom in unix. While pthread_setschedparam() + * works on thread level and we can raise the priority just like we want. + * + * I think this is what FreeBSD offers. (It is certainly analogous to what + * NT offers if you wondered.) Linux on the other hand doesn't provide this + * for processes with SCHED_OTHER policy, and I'm not sure if we want to + * play around with using the real-time SCHED_RR and SCHED_FIFO which would + * require special privileges anyway. + */ + OSPRIOSUP_PROCESS_AND_THREAD_LEVEL, + /** A rough thread level priority only. + * setpriority() is the only real game in town, and it works on thread level. + */ + OSPRIOSUP_THREAD_LEVEL +} volatile g_enmOsPrioSup = OSPRIOSUP_UNDETERMINED; + +/** Set if we figure we have nice capability, meaning we can use setpriority + * to raise the priority. */ +static bool g_fCanNice = false; + + +/********************************************************************************************************************************* +* Internal Functions * +*********************************************************************************************************************************/ + + +/** + * Saves all the scheduling attributes we can think of. + */ +static void rtSchedNativeSave(PSAVEDPRIORITY pSave) +{ + memset(pSave, 0xff, sizeof(*pSave)); + + errno = 0; + pSave->iPriority = getpriority(PRIO_PROCESS, 0 /* current process */); + Assert(errno == 0); + + errno = 0; + sched_getparam(0 /* current process */, &pSave->SchedParam); + Assert(errno == 0); + + errno = 0; + pSave->iPolicy = sched_getscheduler(0 /* current process */); + Assert(errno == 0); + + int rc = pthread_getschedparam(pthread_self(), &pSave->iPthreadPolicy, &pSave->PthreadSchedParam); + Assert(rc == 0); NOREF(rc); +} + + +/** + * Restores scheduling attributes. + * Most of this won't work right, but anyway... + */ +static void rtSchedNativeRestore(PSAVEDPRIORITY pSave) +{ + setpriority(PRIO_PROCESS, 0, pSave->iPriority); + sched_setscheduler(0, pSave->iPolicy, &pSave->SchedParam); + sched_setparam(0, &pSave->SchedParam); + pthread_setschedparam(pthread_self(), pSave->iPthreadPolicy, &pSave->PthreadSchedParam); +} + + +/** + * Starts a worker thread and wait for it to complete. + * We cannot use RTThreadCreate since we're already owner of the RW lock. + */ +static int rtSchedCreateThread(void *(*pfnThread)(void *pvArg), void *pvArg) +{ + /* + * Setup thread attributes. + */ + pthread_attr_t ThreadAttr; + int rc = pthread_attr_init(&ThreadAttr); + if (!rc) + { + rc = pthread_attr_setdetachstate(&ThreadAttr, PTHREAD_CREATE_JOINABLE); + if (!rc) + { + rc = pthread_attr_setstacksize(&ThreadAttr, 128*1024); + if (!rc) + { + /* + * Create the thread. + */ + pthread_t Thread; + rc = pthread_create(&Thread, &ThreadAttr, pfnThread, pvArg); + if (!rc) + { + pthread_attr_destroy(&ThreadAttr); + /* + * Wait for the thread to finish. + */ + void *pvRet = (void *)-1; + do + { + rc = pthread_join(Thread, &pvRet); + } while (rc == EINTR); + if (rc) + return RTErrConvertFromErrno(rc); + return (int)(uintptr_t)pvRet; + } + } + } + pthread_attr_destroy(&ThreadAttr); + } + return RTErrConvertFromErrno(rc); +} + + +static void rtSchedDumpPriority(void) +{ +#ifdef THREAD_LOGGING + Log(("Priority: g_fCanNice=%d g_enmOsPrioSup=%d\n", g_fCanNice, g_enmOsPrioSup)); + Log(("Priority: enmPriority=%d \"%s\" iNice=%d iDelta=%d\n", + g_pProcessPriority->enmPriority, + g_pProcessPriority->pszName, + g_pProcessPriority->iNice, + g_pProcessPriority->iDelta)); + Log(("Priority: %2d INFREQUENT_POLLER = %d\n", RTTHREADTYPE_INFREQUENT_POLLER, g_pProcessPriority->paTypes[RTTHREADTYPE_INFREQUENT_POLLER].iPriority)); + Log(("Priority: %2d MAIN_HEAVY_WORKER = %d\n", RTTHREADTYPE_MAIN_HEAVY_WORKER, g_pProcessPriority->paTypes[RTTHREADTYPE_MAIN_HEAVY_WORKER].iPriority)); + Log(("Priority: %2d EMULATION = %d\n", RTTHREADTYPE_EMULATION , g_pProcessPriority->paTypes[RTTHREADTYPE_EMULATION ].iPriority)); + Log(("Priority: %2d DEFAULT = %d\n", RTTHREADTYPE_DEFAULT , g_pProcessPriority->paTypes[RTTHREADTYPE_DEFAULT ].iPriority)); + Log(("Priority: %2d GUI = %d\n", RTTHREADTYPE_GUI , g_pProcessPriority->paTypes[RTTHREADTYPE_GUI ].iPriority)); + Log(("Priority: %2d MAIN_WORKER = %d\n", RTTHREADTYPE_MAIN_WORKER , g_pProcessPriority->paTypes[RTTHREADTYPE_MAIN_WORKER ].iPriority)); + Log(("Priority: %2d VRDP_IO = %d\n", RTTHREADTYPE_VRDP_IO , g_pProcessPriority->paTypes[RTTHREADTYPE_VRDP_IO ].iPriority)); + Log(("Priority: %2d DEBUGGER = %d\n", RTTHREADTYPE_DEBUGGER , g_pProcessPriority->paTypes[RTTHREADTYPE_DEBUGGER ].iPriority)); + Log(("Priority: %2d MSG_PUMP = %d\n", RTTHREADTYPE_MSG_PUMP , g_pProcessPriority->paTypes[RTTHREADTYPE_MSG_PUMP ].iPriority)); + Log(("Priority: %2d IO = %d\n", RTTHREADTYPE_IO , g_pProcessPriority->paTypes[RTTHREADTYPE_IO ].iPriority)); + Log(("Priority: %2d TIMER = %d\n", RTTHREADTYPE_TIMER , g_pProcessPriority->paTypes[RTTHREADTYPE_TIMER ].iPriority)); +#endif +} + + +/** + * The prober thread. + * We don't want to mess with the priority of the calling thread. + * + * @remark This is pretty presumptive stuff, but if it works on Linux and + * FreeBSD it does what I want. + */ +static void *rtSchedNativeProberThread(void *pvUser) +{ + SAVEDPRIORITY SavedPriority; + rtSchedNativeSave(&SavedPriority); + + /* + * Let's first try and see what we get on a thread level. + */ + int iMax = sched_get_priority_max(SavedPriority.iPthreadPolicy); + int iMin = sched_get_priority_min(SavedPriority.iPthreadPolicy); + if (iMax - iMin >= 32) + { + pthread_t Self = pthread_self(); + int i = iMin; + while (i <= iMax) + { + struct sched_param SchedParam = SavedPriority.PthreadSchedParam; + SchedParam.sched_priority = i; + if (pthread_setschedparam(Self, SavedPriority.iPthreadPolicy, &SchedParam)) + break; + i++; + } + if (i == iMax) + g_enmOsPrioSup = OSPRIOSUP_PROCESS_AND_THREAD_LEVEL; + } + + /* + * Ok, we didn't have the good stuff, so let's fall back on the unix stuff. + */ + if (g_enmOsPrioSup == OSPRIOSUP_UNDETERMINED) + g_enmOsPrioSup = OSPRIOSUP_THREAD_LEVEL; + + /* + * Check if we can get higher priority (typically only root can do this). + * (Won't work right if our priority is -19 to start with, but what the heck.) + * + * We assume that the unix priority is -19 to 19. I know there are defines + * for this, but I don't remember which and if I'm awake enough to make sense + * of them from any SuS spec. + */ + int iStart = getpriority(PRIO_PROCESS, 0); + int i = iStart; + while (i-- > -19) + { + if (setpriority(PRIO_PROCESS, 0, i)) + break; + } + if (getpriority(PRIO_PROCESS, 0) != iStart) + g_fCanNice = true; + else + g_fCanNice = false; + + /* done */ + rtSchedNativeRestore(&SavedPriority); + RT_NOREF(pvUser); + return (void *)VINF_SUCCESS; +} + + +/** + * Calculate the scheduling properties for all the threads in the default + * process priority, assuming the current thread have the type enmType. + * + * @returns iprt status code. + * @param enmType The thread type to be assumed for the current thread. + */ +DECLHIDDEN(int) rtSchedNativeCalcDefaultPriority(RTTHREADTYPE enmType) +{ + Assert(enmType > RTTHREADTYPE_INVALID && enmType < RTTHREADTYPE_END); + + /* + * First figure out what's supported by the OS. + */ + if (g_enmOsPrioSup == OSPRIOSUP_UNDETERMINED) + { + int iPriority = getpriority(PRIO_PROCESS, 0); + int rc = rtSchedCreateThread(rtSchedNativeProberThread, NULL); + if (RT_FAILURE(rc)) + return rc; + if (g_enmOsPrioSup == OSPRIOSUP_UNDETERMINED) + g_enmOsPrioSup = OSPRIOSUP_THREAD_LEVEL; + Assert(getpriority(PRIO_PROCESS, 0) == iPriority); NOREF(iPriority); + } + + /* + * Now let's see what we can do... + */ + int iPriority = getpriority(PRIO_PROCESS, 0); + switch (g_enmOsPrioSup) + { + case OSPRIOSUP_PROCESS_AND_THREAD_LEVEL: + { + g_aDefaultPriority.iNice = iPriority; + g_aDefaultPriority.iDelta = 0; + g_aDefaultPriority.paTypes = g_aTypesThread; + Assert(enmType == g_aDefaultPriority.paTypes[enmType].enmType); + break; + } + + case OSPRIOSUP_THREAD_LEVEL: + { + if (g_fCanNice) + g_aDefaultPriority.paTypes = g_aTypesUnixFree; + else + g_aDefaultPriority.paTypes = g_aTypesUnixRestricted; + Assert(enmType == g_aDefaultPriority.paTypes[enmType].enmType); + g_aDefaultPriority.iNice = iPriority - g_aDefaultPriority.paTypes[enmType].iPriority; + g_aDefaultPriority.iDelta = g_aDefaultPriority.iNice; + break; + } + + default: + AssertFailed(); + break; + } + rtSchedDumpPriority(); + return VINF_SUCCESS; +} + + +/** + * The validator thread. + * We don't want to mess with the priority of the calling thread. + * + * @remark This is pretty presumptive stuff, but if it works on Linux and + * FreeBSD it does what I want. + */ +static void *rtSchedNativeValidatorThread(void *pvUser) +{ + const PROCPRIORITY *pCfg = (const PROCPRIORITY *)pvUser; + SAVEDPRIORITY SavedPriority; + rtSchedNativeSave(&SavedPriority); + + int rc = VINF_SUCCESS; + switch (g_enmOsPrioSup) + { + /* + * Try set the specified process priority and then try + * out all the thread priorities which are used. + */ + case OSPRIOSUP_PROCESS_AND_THREAD_LEVEL: + { + if (!setpriority(PRIO_PROCESS, 0, pCfg->iNice)) + { + int iMin = sched_get_priority_min(SavedPriority.iPolicy); + pthread_t Self = pthread_self(); + for (int i = RTTHREADTYPE_INVALID + 1; i < RTTHREADTYPE_END; i++) + { + struct sched_param SchedParam = SavedPriority.PthreadSchedParam; + SchedParam.sched_priority = pCfg->paTypes[i].iPriority + + pCfg->iDelta + iMin; + rc = pthread_setschedparam(Self, SavedPriority.iPthreadPolicy, &SchedParam); + if (rc) + { + rc = RTErrConvertFromErrno(rc); + break; + } + } + } + else + rc = RTErrConvertFromErrno(errno); + break; + } + + /* + * Try out the priorities from the top and down. + */ + case OSPRIOSUP_THREAD_LEVEL: + { + int i = RTTHREADTYPE_END; + while (--i > RTTHREADTYPE_INVALID) + { + int iPriority = pCfg->paTypes[i].iPriority + pCfg->iDelta; + if (setpriority(PRIO_PROCESS, 0, iPriority)) + { + rc = RTErrConvertFromErrno(errno); + break; + } + } + break; + } + + default: + AssertFailed(); + break; + } + + /* done */ + rtSchedNativeRestore(&SavedPriority); + return (void *)(intptr_t)rc; +} + + +DECLHIDDEN(int) rtProcNativeSetPriority(RTPROCPRIORITY enmPriority) +{ + Assert(enmPriority > RTPROCPRIORITY_INVALID && enmPriority < RTPROCPRIORITY_LAST); + +#ifdef RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY + /* + * Make sure the proxy creation thread is started so we don't 'lose' our + * initial priority if it's lowered. + */ + rtThreadPosixPriorityProxyStart(); +#endif + + /* + * Nothing to validate for the default priority (assuming no external renice). + */ + int rc = VINF_SUCCESS; + if (enmPriority == RTPROCPRIORITY_DEFAULT) + g_pProcessPriority = &g_aDefaultPriority; + else + { + /* + * Select the array to search. + */ + const PROCPRIORITY *pa; + unsigned c; + switch (g_enmOsPrioSup) + { + case OSPRIOSUP_PROCESS_AND_THREAD_LEVEL: + pa = g_aProcessAndThread; + c = RT_ELEMENTS(g_aProcessAndThread); + break; + case OSPRIOSUP_THREAD_LEVEL: + pa = g_aUnixConfigs; + c = RT_ELEMENTS(g_aUnixConfigs); + break; + default: + pa = NULL; + c = 0; + break; + } + + /* + * Search the array. + */ + rc = VERR_FILE_NOT_FOUND; + unsigned i; + for (i = 0; i < c; i++) + { + if (pa[i].enmPriority == enmPriority) + { + /* + * Validate it. + */ + int iPriority = getpriority(PRIO_PROCESS, 0); + int rc3 = rtSchedCreateThread(rtSchedNativeValidatorThread, (void *)&pa[i]); + Assert(getpriority(PRIO_PROCESS, 0) == iPriority); NOREF(iPriority); + if (RT_SUCCESS(rc)) + rc = rc3; + if (RT_SUCCESS(rc)) + break; + } + } + + /* + * Did we get lucky? + * If so update process priority and globals. + */ + if (RT_SUCCESS(rc)) + { + switch (g_enmOsPrioSup) + { + case OSPRIOSUP_PROCESS_AND_THREAD_LEVEL: + if (setpriority(PRIO_PROCESS, 0, pa[i].iNice)) + { + rc = RTErrConvertFromErrno(errno); + AssertMsgFailed(("setpriority(,,%d) -> errno=%d rc=%Rrc\n", pa[i].iNice, errno, rc)); + } + break; + + default: + break; + } + + if (RT_SUCCESS(rc)) + g_pProcessPriority = &pa[i]; + } + } + +#ifdef THREAD_LOGGING + LogFlow(("rtProcNativeSetPriority: returns %Rrc enmPriority=%d\n", rc, enmPriority)); + rtSchedDumpPriority(); +#endif + return rc; +} + + +/** + * Worker for rtThreadNativeSetPriority/OSPRIOSUP_PROCESS_AND_THREAD_LEVEL + * that's either called on the priority proxy thread or directly if no proxy. + */ +static DECLCALLBACK(int) rtThreadPosixSetPriorityOnProcAndThrdCallback(PRTTHREADINT pThread, RTTHREADTYPE enmType) +{ + struct sched_param SchedParam = {-9999999}; + int iPolicy = -7777777; + int rc = pthread_getschedparam((pthread_t)pThread->Core.Key, &iPolicy, &SchedParam); + if (!rc) + { + SchedParam.sched_priority = g_pProcessPriority->paTypes[enmType].iPriority + + g_pProcessPriority->iDelta + + sched_get_priority_min(iPolicy); + + rc = pthread_setschedparam((pthread_t)pThread->Core.Key, iPolicy, &SchedParam); + if (!rc) + { +#ifdef THREAD_LOGGING + Log(("rtThreadNativeSetPriority: Thread=%p enmType=%d iPolicy=%d sched_priority=%d pid=%d\n", + pThread->Core.Key, enmType, iPolicy, SchedParam.sched_priority, getpid())); +#endif + return VINF_SUCCESS; + } + } + + int rcNative = rc; + rc = RTErrConvertFromErrno(rc); + AssertMsgFailed(("pthread_[gs]etschedparam(%p, %d, {%d}) -> rcNative=%d rc=%Rrc\n", + (void *)pThread->Core.Key, iPolicy, SchedParam.sched_priority, rcNative, rc)); NOREF(rcNative); + return rc; +} + + +DECLHIDDEN(int) rtThreadNativeSetPriority(PRTTHREADINT pThread, RTTHREADTYPE enmType) +{ + Assert(enmType > RTTHREADTYPE_INVALID && enmType < RTTHREADTYPE_END); + Assert(enmType == g_pProcessPriority->paTypes[enmType].enmType); + + int rc = VINF_SUCCESS; + switch (g_enmOsPrioSup) + { + case OSPRIOSUP_PROCESS_AND_THREAD_LEVEL: + { +#ifdef RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY + if (rtThreadPosixPriorityProxyStart()) + rc = rtThreadPosixPriorityProxyCall(pThread, (PFNRT)rtThreadPosixSetPriorityOnProcAndThrdCallback, + 2, pThread, enmType); + else +#endif + rc = rtThreadPosixSetPriorityOnProcAndThrdCallback(pThread, enmType); + break; + } + + case OSPRIOSUP_THREAD_LEVEL: + { + /* No cross platform way of getting the 'who' parameter value for + arbitrary threads, so this is restricted to the calling thread only. */ + AssertReturn((pthread_t)pThread->Core.Key == pthread_self(), VERR_NOT_SUPPORTED); + + int iPriority = g_pProcessPriority->paTypes[enmType].iPriority + g_pProcessPriority->iDelta; + if (!setpriority(PRIO_PROCESS, 0, iPriority)) + { + AssertMsg(iPriority == getpriority(PRIO_PROCESS, 0), ("iPriority=%d getpriority()=%d\n", iPriority, getpriority(PRIO_PROCESS, 0))); +#ifdef THREAD_LOGGING + Log(("rtThreadNativeSetPriority: Thread=%p enmType=%d iPriority=%d pid=%d\n", pThread->Core.Key, enmType, iPriority, getpid())); +#endif + } + else + { +#if 0 + rc = RTErrConvertFromErrno(errno); + AssertMsgFailed(("setpriority(,, %d) -> errno=%d rc=%Rrc\n", iPriority, errno, rc)); +#else + /** @todo + * Just keep quiet about failures now - we'll fail here because we're not + * allowed to raise our own priority. This is a problem when starting the + * threads with higher priority from EMT (i.e. most threads it starts). + * This is apparently inherited from the parent in some cases and not + * in other cases. I guess this would come down to which kind of pthread + * implementation is actually in use, and how many sensible patches which + * are installed. + * I need to find a system where this problem shows up in order to come up + * with a proper fix. There's an pthread_create attribute for not inheriting + * scheduler stuff I think... + */ + rc = VINF_SUCCESS; +#endif + } + break; + } + + /* + * Any thread created before we determine the default config, remains unchanged! + * The prober thread above is one of those. + */ + default: + break; + } + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/semevent-posix.cpp b/src/VBox/Runtime/r3/posix/semevent-posix.cpp new file mode 100644 index 00000000..96d01bac --- /dev/null +++ b/src/VBox/Runtime/r3/posix/semevent-posix.cpp @@ -0,0 +1,654 @@ +/* $Id: semevent-posix.cpp $ */ +/** @file + * IPRT - Event Semaphore, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/semaphore.h> +#include "internal/iprt.h" + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/mem.h> +#include <iprt/lockvalidator.h> +#include <iprt/time.h> + +#include "internal/mem.h" +#include "internal/strict.h" + +#include <errno.h> +#include <pthread.h> +#include <unistd.h> +#include <sys/time.h> +#include <sched.h> + +#include "semwait.h" + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ + +/** Internal representation of the POSIX implementation of an Event semaphore. + * The POSIX implementation uses a mutex and a condition variable to implement + * the automatic reset event semaphore semantics. + */ +struct RTSEMEVENTINTERNAL +{ + /** pthread condition. */ + pthread_cond_t Cond; + /** pthread mutex which protects the condition and the event state. */ + pthread_mutex_t Mutex; + /** The state of the semaphore. + * This is operated while owning mutex and using atomic updating. */ + volatile uint32_t u32State; + /** Number of waiters. */ + volatile uint32_t cWaiters; +#ifdef RTSEMEVENT_STRICT + /** Signallers. */ + RTLOCKVALRECSHRD Signallers; + /** Indicates that lock validation should be performed. */ + bool volatile fEverHadSignallers; +#endif + /** The creation flags. */ + uint32_t fFlags; + /** Set if we're using the monotonic clock. */ + bool fMonotonicClock; +}; + +/** The values of the u32State variable in a RTSEMEVENTINTERNAL. + * @{ */ +/** The object isn't initialized. */ +#define EVENT_STATE_UNINITIALIZED 0 +/** The semaphore is signaled. */ +#define EVENT_STATE_SIGNALED 0xff00ff00 +/** The semaphore is not signaled. */ +#define EVENT_STATE_NOT_SIGNALED 0x00ff00ff +/** @} */ + + +RTDECL(int) RTSemEventCreate(PRTSEMEVENT phEventSem) +{ + return RTSemEventCreateEx(phEventSem, 0 /*fFlags*/, NIL_RTLOCKVALCLASS, NULL); +} + + +RTDECL(int) RTSemEventCreateEx(PRTSEMEVENT phEventSem, uint32_t fFlags, RTLOCKVALCLASS hClass, const char *pszNameFmt, ...) +{ + AssertReturn(!(fFlags & ~(RTSEMEVENT_FLAGS_NO_LOCK_VAL | RTSEMEVENT_FLAGS_BOOTSTRAP_HACK)), VERR_INVALID_PARAMETER); + Assert(!(fFlags & RTSEMEVENT_FLAGS_BOOTSTRAP_HACK) || (fFlags & RTSEMEVENT_FLAGS_NO_LOCK_VAL)); + + /* + * Allocate semaphore handle. + */ + int rc; + struct RTSEMEVENTINTERNAL *pThis; + if (!(fFlags & RTSEMEVENT_FLAGS_BOOTSTRAP_HACK)) + pThis = (struct RTSEMEVENTINTERNAL *)RTMemAlloc(sizeof(*pThis)); + else + pThis = (struct RTSEMEVENTINTERNAL *)rtMemBaseAlloc(sizeof(*pThis)); + if (pThis) + { + /* + * Create the condition variable. + */ + pthread_condattr_t CondAttr; + rc = pthread_condattr_init(&CondAttr); + if (!rc) + { +#if defined(CLOCK_MONOTONIC) && defined(IPRT_HAVE_PTHREAD_CONDATTR_SETCLOCK) + /* ASSUMES RTTimeSystemNanoTS() == RTTimeNanoTS() == clock_gettime(CLOCK_MONOTONIC). */ + rc = pthread_condattr_setclock(&CondAttr, CLOCK_MONOTONIC); + pThis->fMonotonicClock = rc == 0; +#else + pThis->fMonotonicClock = false; +#endif + rc = pthread_cond_init(&pThis->Cond, &CondAttr); + if (!rc) + { + /* + * Create the semaphore. + */ + rc = pthread_mutex_init(&pThis->Mutex, NULL); + if (!rc) + { + pthread_condattr_destroy(&CondAttr); + + ASMAtomicWriteU32(&pThis->u32State, EVENT_STATE_NOT_SIGNALED); + ASMAtomicWriteU32(&pThis->cWaiters, 0); + pThis->fFlags = fFlags; +#ifdef RTSEMEVENT_STRICT + if (!pszNameFmt) + { + static uint32_t volatile s_iSemEventAnon = 0; + RTLockValidatorRecSharedInit(&pThis->Signallers, hClass, RTLOCKVAL_SUB_CLASS_ANY, pThis, + true /*fSignaller*/, !(fFlags & RTSEMEVENT_FLAGS_NO_LOCK_VAL), + "RTSemEvent-%u", ASMAtomicIncU32(&s_iSemEventAnon) - 1); + } + else + { + va_list va; + va_start(va, pszNameFmt); + RTLockValidatorRecSharedInitV(&pThis->Signallers, hClass, RTLOCKVAL_SUB_CLASS_ANY, pThis, + true /*fSignaller*/, !(fFlags & RTSEMEVENT_FLAGS_NO_LOCK_VAL), + pszNameFmt, va); + va_end(va); + } + pThis->fEverHadSignallers = false; +#else + RT_NOREF_PV(hClass); RT_NOREF_PV(pszNameFmt); +#endif + + *phEventSem = pThis; + return VINF_SUCCESS; + } + pthread_cond_destroy(&pThis->Cond); + } + pthread_condattr_destroy(&CondAttr); + } + + rc = RTErrConvertFromErrno(rc); + if (!(fFlags & RTSEMEVENT_FLAGS_BOOTSTRAP_HACK)) + RTMemFree(pThis); + else + rtMemBaseFree(pThis); + } + else + rc = VERR_NO_MEMORY; + + return rc; +} + + +RTDECL(int) RTSemEventDestroy(RTSEMEVENT hEventSem) +{ + /* + * Validate handle. + */ + struct RTSEMEVENTINTERNAL *pThis = hEventSem; + if (pThis == NIL_RTSEMEVENT) + return VINF_SUCCESS; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENT_STATE_NOT_SIGNALED || u32 == EVENT_STATE_SIGNALED, VERR_INVALID_HANDLE); + + /* + * Abort all waiters forcing them to return failure. + */ + int rc; + for (int i = 30; i > 0; i--) + { + ASMAtomicWriteU32(&pThis->u32State, EVENT_STATE_UNINITIALIZED); + rc = pthread_cond_destroy(&pThis->Cond); + if (rc != EBUSY) + break; + pthread_cond_broadcast(&pThis->Cond); + usleep(1000); + } + if (rc) + { + AssertMsgFailed(("Failed to destroy event sem %p, rc=%d.\n", pThis, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Destroy the semaphore + * If it's busy we'll wait a bit to give the threads a chance to be scheduled. + */ + for (int i = 30; i > 0; i--) + { + rc = pthread_mutex_destroy(&pThis->Mutex); + if (rc != EBUSY) + break; + usleep(1000); + } + if (rc) + { + AssertMsgFailed(("Failed to destroy event sem %p, rc=%d. (mutex)\n", pThis, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Free the semaphore memory and be gone. + */ +#ifdef RTSEMEVENT_STRICT + RTLockValidatorRecSharedDelete(&pThis->Signallers); +#endif + if (!(pThis->fFlags & RTSEMEVENT_FLAGS_BOOTSTRAP_HACK)) + RTMemFree(pThis); + else + rtMemBaseFree(pThis); + return VINF_SUCCESS; +} + + +RTDECL(int) RTSemEventSignal(RTSEMEVENT hEventSem) +{ + /* + * Validate input. + */ + struct RTSEMEVENTINTERNAL *pThis = hEventSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENT_STATE_NOT_SIGNALED || u32 == EVENT_STATE_SIGNALED, VERR_INVALID_HANDLE); + +#ifdef RTSEMEVENT_STRICT + if (pThis->fEverHadSignallers) + { + int rc9 = RTLockValidatorRecSharedCheckSignaller(&pThis->Signallers, NIL_RTTHREAD); + if (RT_FAILURE(rc9)) + return rc9; + } +#endif + + /* + * Lock the mutex semaphore. + */ + int rc = pthread_mutex_lock(&pThis->Mutex); + if (rc) + { + AssertMsgFailed(("Failed to lock event sem %p, rc=%d.\n", hEventSem, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Check the state. + */ + if (pThis->u32State == EVENT_STATE_NOT_SIGNALED) + { + ASMAtomicWriteU32(&pThis->u32State, EVENT_STATE_SIGNALED); + rc = pthread_cond_signal(&pThis->Cond); + AssertMsg(!rc, ("Failed to signal event sem %p, rc=%d.\n", hEventSem, rc)); + } + else if (pThis->u32State == EVENT_STATE_SIGNALED) + { + rc = pthread_cond_signal(&pThis->Cond); /* give'm another kick... */ + AssertMsg(!rc, ("Failed to signal event sem %p, rc=%d. (2)\n", hEventSem, rc)); + } + else + rc = VERR_SEM_DESTROYED; + + /* + * Release the mutex and return. + */ + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event sem %p, rc=%d.\n", hEventSem, rc)); + if (rc) + return RTErrConvertFromErrno(rc); + if (rc2) + return RTErrConvertFromErrno(rc2); + + return VINF_SUCCESS; +} + + +/** + * Handle polling (timeout already expired at the time of the call). + * + * @returns VINF_SUCCESS, VERR_TIMEOUT, VERR_SEM_DESTROYED. + * @param pThis The semaphore. + */ +DECLINLINE(int) rtSemEventPosixWaitPoll(struct RTSEMEVENTINTERNAL *pThis) +{ + int rc = pthread_mutex_lock(&pThis->Mutex); + AssertMsgReturn(!rc, ("Failed to lock event sem %p, rc=%d.\n", pThis, rc), RTErrConvertFromErrno(rc)); + + uint32_t u32OldState; + bool fSuccess = ASMAtomicCmpXchgExU32(&pThis->u32State, EVENT_STATE_NOT_SIGNALED, EVENT_STATE_SIGNALED, &u32OldState); + + rc = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc, ("Failed to unlock event sem %p, rc=%d.\n", pThis, rc)); NOREF(rc); + + return fSuccess + ? VINF_SUCCESS + : u32OldState != EVENT_STATE_UNINITIALIZED + ? VERR_TIMEOUT + : VERR_SEM_DESTROYED; +} + + +/** + * Performs an indefinite wait on the event. + */ +static int rtSemEventPosixWaitIndefinite(struct RTSEMEVENTINTERNAL *pThis, uint32_t fFlags, PCRTLOCKVALSRCPOS pSrcPos) +{ + RT_NOREF_PV(pSrcPos); + + /* for fairness, yield before going to sleep. */ + if ( ASMAtomicIncU32(&pThis->cWaiters) > 1 + && pThis->u32State == EVENT_STATE_SIGNALED) + sched_yield(); + + /* take mutex */ + int rc = pthread_mutex_lock(&pThis->Mutex); + if (rc) + { + ASMAtomicDecU32(&pThis->cWaiters); + AssertMsgFailed(("Failed to lock event sem %p, rc=%d.\n", pThis, rc)); + return RTErrConvertFromErrno(rc); + } + + for (;;) + { + /* check state. */ + if (pThis->u32State == EVENT_STATE_SIGNALED) + { + ASMAtomicWriteU32(&pThis->u32State, EVENT_STATE_NOT_SIGNALED); + ASMAtomicDecU32(&pThis->cWaiters); + rc = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc, ("Failed to unlock event sem %p, rc=%d.\n", pThis, rc)); NOREF(rc); + return VINF_SUCCESS; + } + if (pThis->u32State == EVENT_STATE_UNINITIALIZED) + { + rc = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc, ("Failed to unlock event sem %p, rc=%d.\n", pThis, rc)); NOREF(rc); + return VERR_SEM_DESTROYED; + } + + /* wait */ +#ifdef RTSEMEVENT_STRICT + RTTHREAD hThreadSelf = !(pThis->fFlags & RTSEMEVENT_FLAGS_BOOTSTRAP_HACK) + ? RTThreadSelfAutoAdopt() + : RTThreadSelf(); + if (pThis->fEverHadSignallers) + { + rc = RTLockValidatorRecSharedCheckBlocking(&pThis->Signallers, hThreadSelf, pSrcPos, false, + RT_INDEFINITE_WAIT, RTTHREADSTATE_EVENT, true); + if (RT_FAILURE(rc)) + { + ASMAtomicDecU32(&pThis->cWaiters); + pthread_mutex_unlock(&pThis->Mutex); + return rc; + } + } +#else + RTTHREAD hThreadSelf = RTThreadSelf(); +#endif + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_EVENT, true); + RT_NOREF_PV(fFlags); /** @todo interruptible wait is not implementable... */ + rc = pthread_cond_wait(&pThis->Cond, &pThis->Mutex); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_EVENT); + if (rc) + { + AssertMsgFailed(("Failed to wait on event sem %p, rc=%d.\n", pThis, rc)); + ASMAtomicDecU32(&pThis->cWaiters); + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event sem %p, rc=%d.\n", pThis, rc2)); NOREF(rc2); + return RTErrConvertFromErrno(rc); + } + } +} + + +/** + * Performs an timed wait on the event. + */ +static int rtSemEventPosixWaitTimed(struct RTSEMEVENTINTERNAL *pThis, uint32_t fFlags, uint64_t uTimeout, + PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Convert the timeout specification to absolute and relative deadlines, + * divierting polling and infinite waits to the appropriate workers. + */ + struct timespec AbsDeadline = { 0, 0 }; + uint64_t const cNsRelativeDeadline = rtSemPosixCalcDeadline(fFlags, uTimeout, pThis->fMonotonicClock, &AbsDeadline); + if (cNsRelativeDeadline == 0) + return rtSemEventPosixWaitPoll(pThis); + if (cNsRelativeDeadline == UINT64_MAX) + return rtSemEventPosixWaitIndefinite(pThis, fFlags, pSrcPos); + + /* + * Now to the business of waiting... + */ + + /* for fairness, yield before going to sleep. */ + if (ASMAtomicIncU32(&pThis->cWaiters) > 1) + sched_yield(); + + /* take mutex */ + int rc = pthread_mutex_lock(&pThis->Mutex); + if (rc) + { + ASMAtomicDecU32(&pThis->cWaiters); + AssertMsg(rc == ETIMEDOUT, ("Failed to lock event sem %p, rc=%d.\n", pThis, rc)); + return RTErrConvertFromErrno(rc); + } + + for (;;) + { + /* check state. */ + uint32_t const u32State = pThis->u32State; + if (u32State != EVENT_STATE_NOT_SIGNALED) + { + if (u32State == EVENT_STATE_SIGNALED) + { + ASMAtomicWriteU32(&pThis->u32State, EVENT_STATE_NOT_SIGNALED); + ASMAtomicDecU32(&pThis->cWaiters); + rc = VINF_SUCCESS; + } + else + { + Assert(u32State == EVENT_STATE_UNINITIALIZED); + rc = VERR_SEM_DESTROYED; + } + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event sem %p, rc2=%d.\n", pThis, rc2)); RT_NOREF(rc2); + return rc; + } + + /* wait */ +#ifdef RTSEMEVENT_STRICT + RTTHREAD hThreadSelf = !(pThis->fFlags & RTSEMEVENT_FLAGS_BOOTSTRAP_HACK) + ? RTThreadSelfAutoAdopt() + : RTThreadSelf(); + if (pThis->fEverHadSignallers) + { + rc = RTLockValidatorRecSharedCheckBlocking(&pThis->Signallers, hThreadSelf, pSrcPos, false, + (cNsRelativeDeadline + RT_NS_1MS - 1) / RT_NS_1MS, + RTTHREADSTATE_EVENT, true); + if (RT_FAILURE(rc)) + { + ASMAtomicDecU32(&pThis->cWaiters); + pthread_mutex_unlock(&pThis->Mutex); + return rc; + } + } +#else + RTTHREAD hThreadSelf = RTThreadSelf(); +#endif + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_EVENT, true); + rc = pthread_cond_timedwait(&pThis->Cond, &pThis->Mutex, &AbsDeadline); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_EVENT); + + /* According to SuS this function shall not return EINTR, but linux man page might have said differently at some point... */ + if ( rc != 0 + && ( rc != EINTR + || !(fFlags & RTSEMWAIT_FLAGS_NORESUME))) + { + AssertMsg(rc == ETIMEDOUT, ("Failed to wait on event sem %p, rc=%d.\n", pThis, rc)); + ASMAtomicDecU32(&pThis->cWaiters); + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event sem %p, rc2=%d.\n", pThis, rc2)); NOREF(rc2); + return RTErrConvertFromErrno(rc); + } + } /* for (;;) */ +} + + +/** + * Internal wait worker function. + */ +DECLINLINE(int) rtSemEventPosixWait(RTSEMEVENT hEventSem, uint32_t fFlags, uint64_t uTimeout, PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Validate input. + */ + struct RTSEMEVENTINTERNAL *pThis = hEventSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENT_STATE_NOT_SIGNALED || u32 == EVENT_STATE_SIGNALED, VERR_INVALID_HANDLE); + AssertReturn(RTSEMWAIT_FLAGS_ARE_VALID(fFlags), VERR_INVALID_PARAMETER); + + /* + * Timed or indefinite wait? + */ + if (fFlags & RTSEMWAIT_FLAGS_INDEFINITE) + return rtSemEventPosixWaitIndefinite(pThis, fFlags, pSrcPos); + return rtSemEventPosixWaitTimed(hEventSem, fFlags, uTimeout, pSrcPos); +} + + +RTDECL(int) RTSemEventWait(RTSEMEVENT hEventSem, RTMSINTERVAL cMillies) +{ + int rc; +#ifndef RTSEMEVENT_STRICT + if (cMillies == RT_INDEFINITE_WAIT) + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_RESUME | RTSEMWAIT_FLAGS_INDEFINITE, 0, NULL); + else + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_RESUME | RTSEMWAIT_FLAGS_RELATIVE | RTSEMWAIT_FLAGS_MILLISECS, + cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + if (cMillies == RT_INDEFINITE_WAIT) + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_RESUME | RTSEMWAIT_FLAGS_INDEFINITE, 0, &SrcPos); + else + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_RESUME | RTSEMWAIT_FLAGS_RELATIVE | RTSEMWAIT_FLAGS_MILLISECS, + cMillies, &SrcPos); +#endif + Assert(rc != VERR_INTERRUPTED); + return rc; +} + + +RTDECL(int) RTSemEventWaitNoResume(RTSEMEVENT hEventSem, RTMSINTERVAL cMillies) +{ + int rc; +#ifndef RTSEMEVENT_STRICT + if (cMillies == RT_INDEFINITE_WAIT) + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_NORESUME | RTSEMWAIT_FLAGS_INDEFINITE, 0, NULL); + else + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_NORESUME | RTSEMWAIT_FLAGS_RELATIVE | RTSEMWAIT_FLAGS_MILLISECS, + cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + if (cMillies == RT_INDEFINITE_WAIT) + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_NORESUME | RTSEMWAIT_FLAGS_INDEFINITE, 0, &SrcPos); + else + rc = rtSemEventPosixWait(hEventSem, RTSEMWAIT_FLAGS_NORESUME | RTSEMWAIT_FLAGS_RELATIVE | RTSEMWAIT_FLAGS_MILLISECS, + cMillies, &SrcPos); +#endif + Assert(rc != VERR_INTERRUPTED); + return rc; +} + + +RTDECL(int) RTSemEventWaitEx(RTSEMEVENT hEventSem, uint32_t fFlags, uint64_t uTimeout) +{ +#ifndef RTSEMEVENT_STRICT + return rtSemEventPosixWait(hEventSem, fFlags, uTimeout, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemEventPosixWait(hEventSem, fFlags, uTimeout, &SrcPos); +#endif +} + + +RTDECL(int) RTSemEventWaitExDebug(RTSEMEVENT hEventSem, uint32_t fFlags, uint64_t uTimeout, + RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemEventPosixWait(hEventSem, fFlags, uTimeout, &SrcPos); +} + + +RTDECL(uint32_t) RTSemEventGetResolution(void) +{ + /** @todo we have 1ns parameter resolution, but we need to check each host + * what the actual resolution might be once the parameter makes it to the + * kernel and is processed there. */ + return 1; +} + + +RTDECL(void) RTSemEventSetSignaller(RTSEMEVENT hEventSem, RTTHREAD hThread) +{ +#ifdef RTSEMEVENT_STRICT + struct RTSEMEVENTINTERNAL *pThis = hEventSem; + AssertPtrReturnVoid(pThis); + uint32_t u32 = pThis->u32State; + AssertReturnVoid(u32 == EVENT_STATE_NOT_SIGNALED || u32 == EVENT_STATE_SIGNALED); + + ASMAtomicWriteBool(&pThis->fEverHadSignallers, true); + RTLockValidatorRecSharedResetOwner(&pThis->Signallers, hThread, NULL); +#else + RT_NOREF_PV(hEventSem); RT_NOREF_PV(hThread); +#endif +} + + +RTDECL(void) RTSemEventAddSignaller(RTSEMEVENT hEventSem, RTTHREAD hThread) +{ +#ifdef RTSEMEVENT_STRICT + struct RTSEMEVENTINTERNAL *pThis = hEventSem; + AssertPtrReturnVoid(pThis); + uint32_t u32 = pThis->u32State; + AssertReturnVoid(u32 == EVENT_STATE_NOT_SIGNALED || u32 == EVENT_STATE_SIGNALED); + + ASMAtomicWriteBool(&pThis->fEverHadSignallers, true); + RTLockValidatorRecSharedAddOwner(&pThis->Signallers, hThread, NULL); +#else + RT_NOREF_PV(hEventSem); RT_NOREF_PV(hThread); +#endif +} + + +RTDECL(void) RTSemEventRemoveSignaller(RTSEMEVENT hEventSem, RTTHREAD hThread) +{ +#ifdef RTSEMEVENT_STRICT + struct RTSEMEVENTINTERNAL *pThis = hEventSem; + AssertPtrReturnVoid(pThis); + uint32_t u32 = pThis->u32State; + AssertReturnVoid(u32 == EVENT_STATE_NOT_SIGNALED || u32 == EVENT_STATE_SIGNALED); + + RTLockValidatorRecSharedRemoveOwner(&pThis->Signallers, hThread); +#else + RT_NOREF_PV(hEventSem); RT_NOREF_PV(hThread); +#endif +} + diff --git a/src/VBox/Runtime/r3/posix/semeventmulti-posix.cpp b/src/VBox/Runtime/r3/posix/semeventmulti-posix.cpp new file mode 100644 index 00000000..52171b4a --- /dev/null +++ b/src/VBox/Runtime/r3/posix/semeventmulti-posix.cpp @@ -0,0 +1,613 @@ +/* $Id: semeventmulti-posix.cpp $ */ +/** @file + * IPRT - Multiple Release Event Semaphore, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/semaphore.h> +#include "internal/iprt.h" + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/lockvalidator.h> +#include <iprt/mem.h> +#include <iprt/time.h> + +#include "internal/strict.h" + +#include <errno.h> +#include <pthread.h> +#include <unistd.h> +#include <sys/time.h> + +#include "semwait.h" + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** Posix internal representation of a Mutex Multi semaphore. + * The POSIX implementation uses a mutex and a condition variable to implement + * the automatic reset event semaphore semantics. */ +struct RTSEMEVENTMULTIINTERNAL +{ + /** pthread condition. */ + pthread_cond_t Cond; + /** pthread mutex which protects the condition and the event state. */ + pthread_mutex_t Mutex; + /** The state of the semaphore. + * This is operated while owning mutex and using atomic updating. */ + volatile uint32_t u32State; + /** Number of waiters. */ + volatile uint32_t cWaiters; +#ifdef RTSEMEVENTMULTI_STRICT + /** Signallers. */ + RTLOCKVALRECSHRD Signallers; + /** Indicates that lock validation should be performed. */ + bool volatile fEverHadSignallers; +#endif + /** Set if we're using the monotonic clock. */ + bool fMonotonicClock; +}; + +/** The values of the u32State variable in RTSEMEVENTMULTIINTERNAL. + * @{ */ +/** The object isn't initialized. */ +#define EVENTMULTI_STATE_UNINITIALIZED 0 +/** The semaphore is signaled. */ +#define EVENTMULTI_STATE_SIGNALED 0xff00ff00 +/** The semaphore is not signaled. */ +#define EVENTMULTI_STATE_NOT_SIGNALED 0x00ff00ff +/** @} */ + + + +RTDECL(int) RTSemEventMultiCreate(PRTSEMEVENTMULTI phEventMultiSem) +{ + return RTSemEventMultiCreateEx(phEventMultiSem, 0 /*fFlags*/, NIL_RTLOCKVALCLASS, NULL); +} + + +RTDECL(int) RTSemEventMultiCreateEx(PRTSEMEVENTMULTI phEventMultiSem, uint32_t fFlags, RTLOCKVALCLASS hClass, + const char *pszNameFmt, ...) +{ + AssertReturn(!(fFlags & ~RTSEMEVENTMULTI_FLAGS_NO_LOCK_VAL), VERR_INVALID_PARAMETER); + + /* + * Allocate semaphore handle. + */ + int rc; + struct RTSEMEVENTMULTIINTERNAL *pThis = (struct RTSEMEVENTMULTIINTERNAL *)RTMemAlloc(sizeof(struct RTSEMEVENTMULTIINTERNAL)); + if (pThis) + { + /* + * Create the condition variable. + */ + pthread_condattr_t CondAttr; + rc = pthread_condattr_init(&CondAttr); + if (!rc) + { +#if defined(CLOCK_MONOTONIC) && defined(IPRT_HAVE_PTHREAD_CONDATTR_SETCLOCK) + /* ASSUMES RTTimeSystemNanoTS() == RTTimeNanoTS() == clock_gettime(CLOCK_MONOTONIC). */ + rc = pthread_condattr_setclock(&CondAttr, CLOCK_MONOTONIC); + pThis->fMonotonicClock = rc == 0; +#else + pThis->fMonotonicClock = false; +#endif + rc = pthread_cond_init(&pThis->Cond, &CondAttr); + if (!rc) + { + /* + * Create the semaphore. + */ + rc = pthread_mutex_init(&pThis->Mutex, NULL); + if (!rc) + { + pthread_condattr_destroy(&CondAttr); + + ASMAtomicWriteU32(&pThis->u32State, EVENTMULTI_STATE_NOT_SIGNALED); + ASMAtomicWriteU32(&pThis->cWaiters, 0); +#ifdef RTSEMEVENTMULTI_STRICT + if (!pszNameFmt) + { + static uint32_t volatile s_iSemEventMultiAnon = 0; + RTLockValidatorRecSharedInit(&pThis->Signallers, hClass, RTLOCKVAL_SUB_CLASS_ANY, pThis, + true /*fSignaller*/, !(fFlags & RTSEMEVENTMULTI_FLAGS_NO_LOCK_VAL), + "RTSemEventMulti-%u", ASMAtomicIncU32(&s_iSemEventMultiAnon) - 1); + } + else + { + va_list va; + va_start(va, pszNameFmt); + RTLockValidatorRecSharedInitV(&pThis->Signallers, hClass, RTLOCKVAL_SUB_CLASS_ANY, pThis, + true /*fSignaller*/, !(fFlags & RTSEMEVENTMULTI_FLAGS_NO_LOCK_VAL), + pszNameFmt, va); + va_end(va); + } + pThis->fEverHadSignallers = false; +#else + RT_NOREF_PV(hClass); RT_NOREF_PV(pszNameFmt); +#endif + + *phEventMultiSem = pThis; + return VINF_SUCCESS; + } + + pthread_cond_destroy(&pThis->Cond); + } + pthread_condattr_destroy(&CondAttr); + } + rc = RTErrConvertFromErrno(rc); + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + + return rc; + +} + + +RTDECL(int) RTSemEventMultiDestroy(RTSEMEVENTMULTI hEventMultiSem) +{ + /* + * Validate handle. + */ + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + if (pThis == NIL_RTSEMEVENTMULTI) + return VINF_SUCCESS; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED, VERR_INVALID_HANDLE); + + /* + * Abort all waiters forcing them to return failure. + */ + int rc; + for (int i = 30; i > 0; i--) + { + ASMAtomicXchgU32(&pThis->u32State, EVENTMULTI_STATE_UNINITIALIZED); + rc = pthread_cond_destroy(&pThis->Cond); + if (rc != EBUSY) + break; + pthread_cond_broadcast(&pThis->Cond); + usleep(1000); + } + if (rc) + { + AssertMsgFailed(("Failed to destroy event sem %p, rc=%d.\n", hEventMultiSem, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Destroy the semaphore + * If it's busy we'll wait a bit to give the threads a chance to be scheduled. + */ + for (int i = 30; i > 0; i--) + { + rc = pthread_mutex_destroy(&pThis->Mutex); + if (rc != EBUSY) + break; + usleep(1000); + } + if (rc) + { + AssertMsgFailed(("Failed to destroy event sem %p, rc=%d. (mutex)\n", hEventMultiSem, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Free the semaphore memory and be gone. + */ +#ifdef RTSEMEVENTMULTI_STRICT + RTLockValidatorRecSharedDelete(&pThis->Signallers); +#endif + RTMemFree(pThis); + return VINF_SUCCESS; +} + + +RTDECL(int) RTSemEventMultiSignal(RTSEMEVENTMULTI hEventMultiSem) +{ + /* + * Validate input. + */ + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED, VERR_INVALID_HANDLE); + +#ifdef RTSEMEVENTMULTI_STRICT + if (pThis->fEverHadSignallers) + { + int rc9 = RTLockValidatorRecSharedCheckSignaller(&pThis->Signallers, NIL_RTTHREAD); + if (RT_FAILURE(rc9)) + return rc9; + } +#endif + + /* + * Lock the mutex semaphore. + */ + int rc = pthread_mutex_lock(&pThis->Mutex); + if (rc) + { + AssertMsgFailed(("Failed to lock event sem %p, rc=%d.\n", hEventMultiSem, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Check the state. + */ + if (pThis->u32State == EVENTMULTI_STATE_NOT_SIGNALED) + { + ASMAtomicXchgU32(&pThis->u32State, EVENTMULTI_STATE_SIGNALED); + rc = pthread_cond_broadcast(&pThis->Cond); + AssertMsg(!rc, ("Failed to signal event sem %p, rc=%d.\n", hEventMultiSem, rc)); + } + else if (pThis->u32State == EVENTMULTI_STATE_SIGNALED) + { + rc = pthread_cond_broadcast(&pThis->Cond); /* give'm another kick... */ + AssertMsg(!rc, ("Failed to signal event sem %p, rc=%d. (2)\n", hEventMultiSem, rc)); + } + else + rc = VERR_SEM_DESTROYED; + + /* + * Release the mutex and return. + */ + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event sem %p, rc=%d.\n", hEventMultiSem, rc)); + if (rc) + return RTErrConvertFromErrno(rc); + if (rc2) + return RTErrConvertFromErrno(rc2); + + return VINF_SUCCESS; +} + + +RTDECL(int) RTSemEventMultiReset(RTSEMEVENTMULTI hEventMultiSem) +{ + /* + * Validate input. + */ + int rc = VINF_SUCCESS; + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED, VERR_INVALID_HANDLE); + + /* + * Lock the mutex semaphore. + */ + int rcPosix = pthread_mutex_lock(&pThis->Mutex); + if (RT_UNLIKELY(rcPosix)) + { + AssertMsgFailed(("Failed to lock event multi sem %p, rc=%d.\n", hEventMultiSem, rcPosix)); + return RTErrConvertFromErrno(rcPosix); + } + + /* + * Check the state. + */ + if (pThis->u32State == EVENTMULTI_STATE_SIGNALED) + ASMAtomicXchgU32(&pThis->u32State, EVENTMULTI_STATE_NOT_SIGNALED); + else if (pThis->u32State != EVENTMULTI_STATE_NOT_SIGNALED) + rc = VERR_SEM_DESTROYED; + + /* + * Release the mutex and return. + */ + rcPosix = pthread_mutex_unlock(&pThis->Mutex); + if (RT_UNLIKELY(rcPosix)) + { + AssertMsgFailed(("Failed to unlock event multi sem %p, rc=%d.\n", hEventMultiSem, rcPosix)); + return RTErrConvertFromErrno(rcPosix); + } + + return rc; +} + + +/** + * Handle polling (timeout already expired at the time of the call). + * + * @returns VINF_SUCCESS, VERR_TIMEOUT, VERR_SEM_DESTROYED. + * @param pThis The semaphore. + */ +DECLINLINE(int) rtSemEventMultiPosixWaitPoll(struct RTSEMEVENTMULTIINTERNAL *pThis) +{ + int rc = pthread_mutex_lock(&pThis->Mutex); + AssertMsgReturn(!rc, ("Failed to lock event multi sem %p, rc=%d.\n", pThis, rc), RTErrConvertFromErrno(rc)); + + uint32_t const u32State = pThis->u32State; + + rc = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc, ("Failed to unlock event multi sem %p, rc=%d.\n", pThis, rc)); NOREF(rc); + + return u32State == EVENTMULTI_STATE_SIGNALED + ? VINF_SUCCESS + : u32State != EVENTMULTI_STATE_UNINITIALIZED + ? VERR_TIMEOUT + : VERR_SEM_DESTROYED; +} + + + +/** + * Implements the indefinite wait. + * + * @returns See RTSemEventMultiWaitEx. + * @param pThis The semaphore. + * @param fFlags See RTSemEventMultiWaitEx. + * @param pSrcPos The source position, can be NULL. + */ +static int rtSemEventMultiPosixWaitIndefinite(struct RTSEMEVENTMULTIINTERNAL *pThis, uint32_t fFlags, PCRTLOCKVALSRCPOS pSrcPos) +{ + /* take mutex */ + int rc = pthread_mutex_lock(&pThis->Mutex); + AssertMsgReturn(!rc, ("Failed to lock event multi sem %p, rc=%d.\n", pThis, rc), RTErrConvertFromErrno(rc)); + ASMAtomicIncU32(&pThis->cWaiters); + + for (;;) + { + /* check state. */ + uint32_t const u32State = pThis->u32State; + if (u32State != EVENTMULTI_STATE_NOT_SIGNALED) + { + ASMAtomicDecU32(&pThis->cWaiters); + rc = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc, ("Failed to unlock event multi sem %p, rc=%d.\n", pThis, rc)); + return u32State == EVENTMULTI_STATE_SIGNALED + ? VINF_SUCCESS + : VERR_SEM_DESTROYED; + } + + /* wait */ +#ifdef RTSEMEVENTMULTI_STRICT + RTTHREAD hThreadSelf = RTThreadSelfAutoAdopt(); + if (pThis->fEverHadSignallers) + { + rc = RTLockValidatorRecSharedCheckBlocking(&pThis->Signallers, hThreadSelf, pSrcPos, false, + RT_INDEFINITE_WAIT, RTTHREADSTATE_EVENT_MULTI, true); + if (RT_FAILURE(rc)) + { + ASMAtomicDecU32(&pThis->cWaiters); + pthread_mutex_unlock(&pThis->Mutex); + return rc; + } + } +#else + RTTHREAD hThreadSelf = RTThreadSelf(); + RT_NOREF_PV(pSrcPos); +#endif + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_EVENT_MULTI, true); + /** @todo interruptible wait is not implementable... */ NOREF(fFlags); + rc = pthread_cond_wait(&pThis->Cond, &pThis->Mutex); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_EVENT_MULTI); + if (RT_UNLIKELY(rc)) + { + AssertMsgFailed(("Failed to wait on event multi sem %p, rc=%d.\n", pThis, rc)); + ASMAtomicDecU32(&pThis->cWaiters); + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event multi sem %p, rc=%d.\n", pThis, rc2)); NOREF(rc2); + return RTErrConvertFromErrno(rc); + } + } +} + + +/** + * Implements the timed wait. + * + * @returns See RTSemEventMultiWaitEx + * @param pThis The semaphore. + * @param fFlags See RTSemEventMultiWaitEx. + * @param uTimeout See RTSemEventMultiWaitEx. + * @param pSrcPos The source position, can be NULL. + */ +static int rtSemEventMultiPosixWaitTimed(struct RTSEMEVENTMULTIINTERNAL *pThis, uint32_t fFlags, uint64_t uTimeout, + PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Convert the timeout specification to absolute and relative deadlines, + * divierting polling and infinite waits to the appropriate workers. + */ + struct timespec AbsDeadline = { 0, 0 }; + uint64_t const cNsRelativeDeadline = rtSemPosixCalcDeadline(fFlags, uTimeout, pThis->fMonotonicClock, &AbsDeadline); + if (cNsRelativeDeadline == 0) + return rtSemEventMultiPosixWaitPoll(pThis); + if (cNsRelativeDeadline == UINT64_MAX) + return rtSemEventMultiPosixWaitIndefinite(pThis, fFlags, pSrcPos); + + /* + * To business! + */ + /* take mutex */ + int rc = pthread_mutex_lock(&pThis->Mutex); + AssertMsgReturn(rc == 0, ("rc=%d pThis=%p\n", rc, pThis), RTErrConvertFromErrno(rc)); NOREF(rc); + ASMAtomicIncU32(&pThis->cWaiters); + + for (;;) + { + /* check state. */ + uint32_t const u32State = pThis->u32State; + if (u32State != EVENTMULTI_STATE_NOT_SIGNALED) + { + ASMAtomicDecU32(&pThis->cWaiters); + rc = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc, ("Failed to unlock event multi sem %p, rc=%d.\n", pThis, rc)); + return u32State == EVENTMULTI_STATE_SIGNALED + ? VINF_SUCCESS + : VERR_SEM_DESTROYED; + } + + /* wait */ +#ifdef RTSEMEVENTMULTI_STRICT + RTTHREAD hThreadSelf = RTThreadSelfAutoAdopt(); + if (pThis->fEverHadSignallers) + { + rc = RTLockValidatorRecSharedCheckBlocking(&pThis->Signallers, hThreadSelf, pSrcPos, false, + (uTimeout + RT_NS_1MS - 1)/ RT_NS_1MS, RTTHREADSTATE_EVENT_MULTI, true); + if (RT_FAILURE(rc)) + { + ASMAtomicDecU32(&pThis->cWaiters); + pthread_mutex_unlock(&pThis->Mutex); + return rc; + } + } +#else + RTTHREAD hThreadSelf = RTThreadSelf(); +#endif + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_EVENT_MULTI, true); + rc = pthread_cond_timedwait(&pThis->Cond, &pThis->Mutex, &AbsDeadline); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_EVENT_MULTI); + + /* According to SuS this function shall not return EINTR, but linux man page might have said differently at some point... */ + if ( rc != 0 + && ( rc != EINTR + || (fFlags & RTSEMWAIT_FLAGS_NORESUME)) ) + { + AssertMsg(rc == ETIMEDOUT, ("Failed to wait on event multi sem %p, rc=%d.\n", pThis, rc)); + ASMAtomicDecU32(&pThis->cWaiters); + int rc2 = pthread_mutex_unlock(&pThis->Mutex); + AssertMsg(!rc2, ("Failed to unlock event multi sem %p, rc=%d.\n", pThis, rc2)); NOREF(rc2); + return RTErrConvertFromErrno(rc); + } + } +} + + +DECLINLINE(int) rtSemEventMultiPosixWait(RTSEMEVENTMULTI hEventMultiSem, uint32_t fFlags, uint64_t uTimeout, + PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Validate input. + */ + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + uint32_t u32 = pThis->u32State; + AssertReturn(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED, VERR_INVALID_HANDLE); + AssertReturn(RTSEMWAIT_FLAGS_ARE_VALID(fFlags), VERR_INVALID_PARAMETER); + + /* + * Optimize the case where the event is signalled. + */ + if (ASMAtomicUoReadU32(&pThis->u32State) == EVENTMULTI_STATE_SIGNALED) + { + int rc = rtSemEventMultiPosixWaitPoll(pThis); + if (RT_LIKELY(rc != VERR_TIMEOUT)) + return rc; + } + + /* + * Indefinite or timed wait? + */ + if (fFlags & RTSEMWAIT_FLAGS_INDEFINITE) + return rtSemEventMultiPosixWaitIndefinite(pThis, fFlags, pSrcPos); + return rtSemEventMultiPosixWaitTimed(pThis, fFlags, uTimeout, pSrcPos); +} + + +#undef RTSemEventMultiWaitEx +RTDECL(int) RTSemEventMultiWaitEx(RTSEMEVENTMULTI hEventMultiSem, uint32_t fFlags, uint64_t uTimeout) +{ +#ifndef RTSEMEVENT_STRICT + return rtSemEventMultiPosixWait(hEventMultiSem, fFlags, uTimeout, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemEventMultiPosixWait(hEventMultiSem, fFlags, uTimeout, &SrcPos); +#endif +} + + +RTDECL(int) RTSemEventMultiWaitExDebug(RTSEMEVENTMULTI hEventMultiSem, uint32_t fFlags, uint64_t uTimeout, + RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemEventMultiPosixWait(hEventMultiSem, fFlags, uTimeout, &SrcPos); +} + + +RTDECL(void) RTSemEventMultiSetSignaller(RTSEMEVENTMULTI hEventMultiSem, RTTHREAD hThread) +{ +#ifdef RTSEMEVENTMULTI_STRICT + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + AssertPtrReturnVoid(pThis); + uint32_t u32 = pThis->u32State; + AssertReturnVoid(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED); + + ASMAtomicWriteBool(&pThis->fEverHadSignallers, true); + RTLockValidatorRecSharedResetOwner(&pThis->Signallers, hThread, NULL); +#else + RT_NOREF_PV(hEventMultiSem); RT_NOREF_PV(hThread); +#endif +} + + +RTDECL(void) RTSemEventMultiAddSignaller(RTSEMEVENTMULTI hEventMultiSem, RTTHREAD hThread) +{ +#ifdef RTSEMEVENTMULTI_STRICT + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + AssertPtrReturnVoid(pThis); + uint32_t u32 = pThis->u32State; + AssertReturnVoid(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED); + + ASMAtomicWriteBool(&pThis->fEverHadSignallers, true); + RTLockValidatorRecSharedAddOwner(&pThis->Signallers, hThread, NULL); +#else + RT_NOREF_PV(hEventMultiSem); RT_NOREF_PV(hThread); +#endif +} + + +RTDECL(void) RTSemEventMultiRemoveSignaller(RTSEMEVENTMULTI hEventMultiSem, RTTHREAD hThread) +{ +#ifdef RTSEMEVENTMULTI_STRICT + struct RTSEMEVENTMULTIINTERNAL *pThis = hEventMultiSem; + AssertPtrReturnVoid(pThis); + uint32_t u32 = pThis->u32State; + AssertReturnVoid(u32 == EVENTMULTI_STATE_NOT_SIGNALED || u32 == EVENTMULTI_STATE_SIGNALED); + + RTLockValidatorRecSharedRemoveOwner(&pThis->Signallers, hThread); +#else + RT_NOREF_PV(hEventMultiSem); RT_NOREF_PV(hThread); +#endif +} + diff --git a/src/VBox/Runtime/r3/posix/semmutex-posix.cpp b/src/VBox/Runtime/r3/posix/semmutex-posix.cpp new file mode 100644 index 00000000..f0df48ef --- /dev/null +++ b/src/VBox/Runtime/r3/posix/semmutex-posix.cpp @@ -0,0 +1,467 @@ +/* $Id: semmutex-posix.cpp $ */ +/** @file + * IPRT - Mutex Semaphore, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/semaphore.h> +#include "internal/iprt.h" + +#include <iprt/alloc.h> +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/lockvalidator.h> +#include <iprt/thread.h> +#include "internal/magics.h" +#include "internal/strict.h" + +#include <errno.h> +#include <pthread.h> +#include <unistd.h> +#include <sys/time.h> + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** Posix internal representation of a Mutex semaphore. */ +struct RTSEMMUTEXINTERNAL +{ + /** pthread mutex. */ + pthread_mutex_t Mutex; + /** The owner of the mutex. */ + volatile pthread_t Owner; + /** Nesting count. */ + volatile uint32_t cNesting; + /** Magic value (RTSEMMUTEX_MAGIC). */ + uint32_t u32Magic; +#ifdef RTSEMMUTEX_STRICT + /** Lock validator record associated with this mutex. */ + RTLOCKVALRECEXCL ValidatorRec; +#endif +}; + +#if defined(RT_OS_DARWIN) || defined(RT_OS_NETBSD) +/** + * This function is a crude approximation of pthread_mutex_timedlock. + */ +int rtSemFallbackPthreadMutexTimedlock(pthread_mutex_t *mutex, RTMSINTERVAL cMillies) +{ + struct timespec ts; + int rc; + + rc = pthread_mutex_trylock(mutex); + if (rc != EBUSY) + return rc; + + ts.tv_sec = cMillies / 1000; + ts.tv_nsec = (cMillies % 1000) * 1000000; + + while (ts.tv_sec > 0 || ts.tv_nsec > 0) + { + struct timespec delta, remaining; + + if (ts.tv_sec > 0) + { + delta.tv_sec = 1; + delta.tv_nsec = 0; + ts.tv_sec--; + } + else + { + delta.tv_sec = 0; + delta.tv_nsec = ts.tv_nsec; + ts.tv_nsec = 0; + } + + nanosleep(&delta, &remaining); + + rc = pthread_mutex_trylock(mutex); + if (rc != EBUSY) + return rc; + + if (RT_UNLIKELY(remaining.tv_nsec > 0 || remaining.tv_sec > 0)) + { + ts.tv_sec += remaining.tv_sec; + ts.tv_nsec += remaining.tv_nsec; + if (ts.tv_nsec >= 1000000000) + { + ts.tv_nsec -= 1000000000; + ts.tv_sec++; + } + } + } + + return ETIMEDOUT; +} +#endif + + +#undef RTSemMutexCreate +RTDECL(int) RTSemMutexCreate(PRTSEMMUTEX phMutexSem) +{ + return RTSemMutexCreateEx(phMutexSem, 0 /*fFlags*/, NIL_RTLOCKVALCLASS, RTLOCKVAL_SUB_CLASS_NONE, NULL); +} + + +RTDECL(int) RTSemMutexCreateEx(PRTSEMMUTEX phMutexSem, uint32_t fFlags, + RTLOCKVALCLASS hClass, uint32_t uSubClass, const char *pszNameFmt, ...) +{ + AssertReturn(!(fFlags & ~RTSEMMUTEX_FLAGS_NO_LOCK_VAL), VERR_INVALID_PARAMETER); + + /* + * Allocate semaphore handle. + */ + int rc; + struct RTSEMMUTEXINTERNAL *pThis = (struct RTSEMMUTEXINTERNAL *)RTMemAlloc(sizeof(struct RTSEMMUTEXINTERNAL)); + if (pThis) + { + /* + * Create the semaphore. + */ + rc = pthread_mutex_init(&pThis->Mutex, NULL); + if (!rc) + { + pThis->Owner = (pthread_t)-1; + pThis->cNesting = 0; + pThis->u32Magic = RTSEMMUTEX_MAGIC; +#ifdef RTSEMMUTEX_STRICT + if (!pszNameFmt) + { + static uint32_t volatile s_iMutexAnon = 0; + RTLockValidatorRecExclInit(&pThis->ValidatorRec, hClass, uSubClass, pThis, + !(fFlags & RTSEMMUTEX_FLAGS_NO_LOCK_VAL), + "RTSemMutex-%u", ASMAtomicIncU32(&s_iMutexAnon) - 1); + } + else + { + va_list va; + va_start(va, pszNameFmt); + RTLockValidatorRecExclInitV(&pThis->ValidatorRec, hClass, uSubClass, pThis, + !(fFlags & RTSEMMUTEX_FLAGS_NO_LOCK_VAL), pszNameFmt, va); + va_end(va); + } +#else + RT_NOREF_PV(hClass); RT_NOREF_PV(uSubClass); RT_NOREF_PV(pszNameFmt); +#endif + + *phMutexSem = pThis; + return VINF_SUCCESS; + } + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + + return rc; +} + + +RTDECL(int) RTSemMutexDestroy(RTSEMMUTEX hMutexSem) +{ + /* + * Validate input. + */ + if (hMutexSem == NIL_RTSEMMUTEX) + return VINF_SUCCESS; + struct RTSEMMUTEXINTERNAL *pThis = hMutexSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTSEMMUTEX_MAGIC, VERR_INVALID_HANDLE); + + /* + * Try destroy it. + */ + int rc = pthread_mutex_destroy(&pThis->Mutex); + if (rc) + { + AssertMsgFailed(("Failed to destroy mutex sem %p, rc=%d.\n", hMutexSem, rc)); + return RTErrConvertFromErrno(rc); + } + + /* + * Free the memory and be gone. + */ + ASMAtomicWriteU32(&pThis->u32Magic, RTSEMMUTEX_MAGIC_DEAD); + pThis->Owner = (pthread_t)-1; + pThis->cNesting = UINT32_MAX; +#ifdef RTSEMMUTEX_STRICT + RTLockValidatorRecExclDelete(&pThis->ValidatorRec); +#endif + RTMemTmpFree(pThis); + + return VINF_SUCCESS; +} + + +RTDECL(uint32_t) RTSemMutexSetSubClass(RTSEMMUTEX hMutexSem, uint32_t uSubClass) +{ +#ifdef RTSEMMUTEX_STRICT + /* + * Validate. + */ + RTSEMMUTEXINTERNAL *pThis = hMutexSem; + AssertPtrReturn(pThis, RTLOCKVAL_SUB_CLASS_INVALID); + AssertReturn(pThis->u32Magic == RTSEMMUTEX_MAGIC, RTLOCKVAL_SUB_CLASS_INVALID); + + return RTLockValidatorRecExclSetSubClass(&pThis->ValidatorRec, uSubClass); +#else + RT_NOREF_PV(hMutexSem); RT_NOREF_PV(uSubClass); + return RTLOCKVAL_SUB_CLASS_INVALID; +#endif +} + + +DECL_FORCE_INLINE(int) rtSemMutexRequest(RTSEMMUTEX hMutexSem, RTMSINTERVAL cMillies, PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Validate input. + */ + struct RTSEMMUTEXINTERNAL *pThis = hMutexSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTSEMMUTEX_MAGIC, VERR_INVALID_HANDLE); + + /* + * Check if nested request. + */ + pthread_t Self = pthread_self(); + if ( pThis->Owner == Self + && pThis->cNesting > 0) + { +#ifdef RTSEMMUTEX_STRICT + int rc9 = RTLockValidatorRecExclRecursion(&pThis->ValidatorRec, pSrcPos); + if (RT_FAILURE(rc9)) + return rc9; +#endif + ASMAtomicIncU32(&pThis->cNesting); + return VINF_SUCCESS; + } + + /* + * Lock it. + */ + RTTHREAD hThreadSelf = NIL_RTTHREAD; + if (cMillies != 0) + { +#ifdef RTSEMMUTEX_STRICT + hThreadSelf = RTThreadSelfAutoAdopt(); + int rc9 = RTLockValidatorRecExclCheckOrderAndBlocking(&pThis->ValidatorRec, hThreadSelf, pSrcPos, true, + cMillies, RTTHREADSTATE_MUTEX, true); + if (RT_FAILURE(rc9)) + return rc9; +#else + hThreadSelf = RTThreadSelf(); + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_MUTEX, true); + RT_NOREF_PV(pSrcPos); +#endif + } + + if (cMillies == RT_INDEFINITE_WAIT) + { + /* take mutex */ + int rc = pthread_mutex_lock(&pThis->Mutex); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_MUTEX); + if (rc) + { + AssertMsgFailed(("Failed to lock mutex sem %p, rc=%d.\n", hMutexSem, rc)); NOREF(rc); + return RTErrConvertFromErrno(rc); + } + } + else + { + int rc; +#if !defined(RT_OS_DARWIN) && !defined(RT_OS_NETBSD) + struct timespec ts = {0,0}; +# if defined(RT_OS_HAIKU) + struct timeval tv = {0,0}; + gettimeofday(&tv, NULL); + ts.tv_sec = tv.tv_sec; + ts.tv_nsec = tv.tv_usec * 1000; +# else + clock_gettime(CLOCK_REALTIME, &ts); +# endif + if (cMillies != 0) + { + ts.tv_nsec += (cMillies % 1000) * 1000000; + ts.tv_sec += cMillies / 1000; + if (ts.tv_nsec >= 1000000000) + { + ts.tv_nsec -= 1000000000; + ts.tv_sec++; + } + } + + /* take mutex */ + rc = pthread_mutex_timedlock(&pThis->Mutex, &ts); +#else + /* + * When there's no pthread_mutex_timedlock() use a crude sleep + * and retry approximation. Since the sleep interval is + * relative, we don't need to convert to the absolute time + * here only to convert back to relative in the fallback + * function. + */ + rc = rtSemFallbackPthreadMutexTimedlock(&pThis->Mutex, cMillies); +#endif + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_MUTEX); + if (rc) + { + AssertMsg(rc == ETIMEDOUT, ("Failed to lock mutex sem %p, rc=%d.\n", hMutexSem, rc)); NOREF(rc); + return RTErrConvertFromErrno(rc); + } + } + + /* + * Set the owner and nesting. + */ + pThis->Owner = Self; + ASMAtomicWriteU32(&pThis->cNesting, 1); +#ifdef RTSEMMUTEX_STRICT + RTLockValidatorRecExclSetOwner(&pThis->ValidatorRec, hThreadSelf, pSrcPos, true); +#endif + + return VINF_SUCCESS; +} + + +#undef RTSemMutexRequest +RTDECL(int) RTSemMutexRequest(RTSEMMUTEX hMutexSem, RTMSINTERVAL cMillies) +{ +#ifndef RTSEMMUTEX_STRICT + return rtSemMutexRequest(hMutexSem, cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemMutexRequest(hMutexSem, cMillies, &SrcPos); +#endif +} + + +RTDECL(int) RTSemMutexRequestDebug(RTSEMMUTEX hMutexSem, RTMSINTERVAL cMillies, RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemMutexRequest(hMutexSem, cMillies, &SrcPos); +} + + +#undef RTSemMutexRequestNoResume +RTDECL(int) RTSemMutexRequestNoResume(RTSEMMUTEX hMutexSem, RTMSINTERVAL cMillies) +{ + /* (EINTR isn't returned by the wait functions we're using.) */ +#ifndef RTSEMMUTEX_STRICT + return rtSemMutexRequest(hMutexSem, cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemMutexRequest(hMutexSem, cMillies, &SrcPos); +#endif +} + + +RTDECL(int) RTSemMutexRequestNoResumeDebug(RTSEMMUTEX hMutexSem, RTMSINTERVAL cMillies, RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemMutexRequest(hMutexSem, cMillies, &SrcPos); +} + + +RTDECL(int) RTSemMutexRelease(RTSEMMUTEX hMutexSem) +{ + /* + * Validate input. + */ + struct RTSEMMUTEXINTERNAL *pThis = hMutexSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertReturn(pThis->u32Magic == RTSEMMUTEX_MAGIC, VERR_INVALID_HANDLE); + +#ifdef RTSEMMUTEX_STRICT + int rc9 = RTLockValidatorRecExclReleaseOwner(&pThis->ValidatorRec, pThis->cNesting == 1); + if (RT_FAILURE(rc9)) + return rc9; +#endif + + /* + * Check if nested. + */ + pthread_t Self = pthread_self(); + if (RT_UNLIKELY( pThis->Owner != Self + || pThis->cNesting == 0)) + { + AssertMsgFailed(("Not owner of mutex %p!! Self=%08x Owner=%08x cNesting=%d\n", + pThis, Self, pThis->Owner, pThis->cNesting)); + return VERR_NOT_OWNER; + } + + /* + * If nested we'll just pop a nesting. + */ + if (pThis->cNesting > 1) + { + ASMAtomicDecU32(&pThis->cNesting); + return VINF_SUCCESS; + } + + /* + * Clear the state. (cNesting == 1) + */ + pThis->Owner = (pthread_t)-1; + ASMAtomicWriteU32(&pThis->cNesting, 0); + + /* + * Unlock mutex semaphore. + */ + int rc = pthread_mutex_unlock(&pThis->Mutex); + if (RT_UNLIKELY(rc)) + { + AssertMsgFailed(("Failed to unlock mutex sem %p, rc=%d.\n", hMutexSem, rc)); NOREF(rc); + return RTErrConvertFromErrno(rc); + } + + return VINF_SUCCESS; +} + + +RTDECL(bool) RTSemMutexIsOwned(RTSEMMUTEX hMutexSem) +{ + /* + * Validate. + */ + RTSEMMUTEXINTERNAL *pThis = hMutexSem; + AssertPtrReturn(pThis, false); + AssertReturn(pThis->u32Magic == RTSEMMUTEX_MAGIC, false); + + return pThis->Owner != (pthread_t)-1; +} + diff --git a/src/VBox/Runtime/r3/posix/semrw-posix.cpp b/src/VBox/Runtime/r3/posix/semrw-posix.cpp new file mode 100644 index 00000000..ab657566 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/semrw-posix.cpp @@ -0,0 +1,741 @@ +/* $Id: semrw-posix.cpp $ */ +/** @file + * IPRT - Read-Write Semaphore, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/semaphore.h> +#include "internal/iprt.h" + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/lockvalidator.h> +#include <iprt/mem.h> +#include <iprt/thread.h> + +#include <errno.h> +#include <pthread.h> +#include <unistd.h> +#include <sys/time.h> + +#include "internal/magics.h" +#include "internal/strict.h" + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +/** @todo move this to r3/posix/something.h. */ +#ifdef RT_OS_SOLARIS +# define ATOMIC_GET_PTHREAD_T(ppvVar, pThread) ASMAtomicReadSize(ppvVar, pThread) +# define ATOMIC_SET_PTHREAD_T(ppvVar, pThread) ASMAtomicWriteSize(ppvVar, pThread) +#else +AssertCompileSize(pthread_t, sizeof(void *)); +# define ATOMIC_GET_PTHREAD_T(ppvVar, pThread) do { *(pThread) = (pthread_t)ASMAtomicReadPtr((void * volatile *)ppvVar); } while (0) +# define ATOMIC_SET_PTHREAD_T(ppvVar, pThread) ASMAtomicWritePtr((void * volatile *)ppvVar, (void *)pThread) +#endif + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** Posix internal representation of a read-write semaphore. */ +struct RTSEMRWINTERNAL +{ + /** The usual magic. (RTSEMRW_MAGIC) */ + uint32_t u32Magic; + /** The number of readers. + * (For preventing screwing up the lock on linux). */ + uint32_t volatile cReaders; + /** Number of write recursions. */ + uint32_t cWrites; + /** Number of read recursions by the writer. */ + uint32_t cWriterReads; + /** The write owner of the lock. */ + volatile pthread_t Writer; + /** pthread rwlock. */ + pthread_rwlock_t RWLock; +#ifdef RTSEMRW_STRICT + /** The validator record for the writer. */ + RTLOCKVALRECEXCL ValidatorWrite; + /** The validator record for the readers. */ + RTLOCKVALRECSHRD ValidatorRead; +#endif +}; + + + +#undef RTSemRWCreate +RTDECL(int) RTSemRWCreate(PRTSEMRW phRWSem) +{ + return RTSemRWCreateEx(phRWSem, 0 /*fFlags*/, NIL_RTLOCKVALCLASS, RTLOCKVAL_SUB_CLASS_NONE, "RTSemRW"); +} + + +RTDECL(int) RTSemRWCreateEx(PRTSEMRW phRWSem, uint32_t fFlags, + RTLOCKVALCLASS hClass, uint32_t uSubClass, const char *pszNameFmt, ...) +{ + AssertReturn(!(fFlags & ~RTSEMRW_FLAGS_NO_LOCK_VAL), VERR_INVALID_PARAMETER); + + /* + * Allocate handle. + */ + int rc; + struct RTSEMRWINTERNAL *pThis = (struct RTSEMRWINTERNAL *)RTMemAlloc(sizeof(struct RTSEMRWINTERNAL)); + if (pThis) + { + /* + * Create the rwlock. + */ + rc = pthread_rwlock_init(&pThis->RWLock, NULL); + if (!rc) + { + pThis->u32Magic = RTSEMRW_MAGIC; + pThis->cReaders = 0; + pThis->cWrites = 0; + pThis->cWriterReads = 0; + pThis->Writer = (pthread_t)-1; +#ifdef RTSEMRW_STRICT + bool const fLVEnabled = !(fFlags & RTSEMRW_FLAGS_NO_LOCK_VAL); + if (!pszNameFmt) + { + static uint32_t volatile s_iSemRWAnon = 0; + uint32_t i = ASMAtomicIncU32(&s_iSemRWAnon) - 1; + RTLockValidatorRecExclInit(&pThis->ValidatorWrite, hClass, uSubClass, pThis, + fLVEnabled, "RTSemRW-%u", i); + RTLockValidatorRecSharedInit(&pThis->ValidatorRead, hClass, uSubClass, pThis, + false /*fSignaller*/, fLVEnabled, "RTSemRW-%u", i); + } + else + { + va_list va; + va_start(va, pszNameFmt); + RTLockValidatorRecExclInitV(&pThis->ValidatorWrite, hClass, uSubClass, pThis, + fLVEnabled, pszNameFmt, va); + va_end(va); + va_start(va, pszNameFmt); + RTLockValidatorRecSharedInitV(&pThis->ValidatorRead, hClass, uSubClass, pThis, + false /*fSignaller*/, fLVEnabled, pszNameFmt, va); + va_end(va); + } + RTLockValidatorRecMakeSiblings(&pThis->ValidatorWrite.Core, &pThis->ValidatorRead.Core); +#else + RT_NOREF_PV(hClass); RT_NOREF_PV(uSubClass); RT_NOREF_PV(pszNameFmt); +#endif + *phRWSem = pThis; + return VINF_SUCCESS; + } + + rc = RTErrConvertFromErrno(rc); + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + + return rc; +} + + +RTDECL(int) RTSemRWDestroy(RTSEMRW hRWSem) +{ + /* + * Validate input, nil handle is fine. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + if (pThis == NIL_RTSEMRW) + return VINF_SUCCESS; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + VERR_INVALID_HANDLE); + Assert(pThis->Writer == (pthread_t)-1); + Assert(!pThis->cReaders); + Assert(!pThis->cWrites); + Assert(!pThis->cWriterReads); + + /* + * Try destroy it. + */ + AssertReturn(ASMAtomicCmpXchgU32(&pThis->u32Magic, ~RTSEMRW_MAGIC, RTSEMRW_MAGIC), VERR_INVALID_HANDLE); + int rc = pthread_rwlock_destroy(&pThis->RWLock); + if (!rc) + { +#ifdef RTSEMRW_STRICT + RTLockValidatorRecSharedDelete(&pThis->ValidatorRead); + RTLockValidatorRecExclDelete(&pThis->ValidatorWrite); +#endif + RTMemFree(pThis); + rc = VINF_SUCCESS; + } + else + { + ASMAtomicWriteU32(&pThis->u32Magic, RTSEMRW_MAGIC); + AssertMsgFailed(("Failed to destroy read-write sem %p, rc=%d.\n", hRWSem, rc)); + rc = RTErrConvertFromErrno(rc); + } + + return rc; +} + + +RTDECL(uint32_t) RTSemRWSetSubClass(RTSEMRW hRWSem, uint32_t uSubClass) +{ +#ifdef RTSEMRW_STRICT + /* + * Validate handle. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, RTLOCKVAL_SUB_CLASS_INVALID); + AssertReturn(pThis->u32Magic == RTSEMRW_MAGIC, RTLOCKVAL_SUB_CLASS_INVALID); + + RTLockValidatorRecSharedSetSubClass(&pThis->ValidatorRead, uSubClass); + return RTLockValidatorRecExclSetSubClass(&pThis->ValidatorWrite, uSubClass); +#else + RT_NOREF_PV(hRWSem); RT_NOREF_PV(uSubClass); + return RTLOCKVAL_SUB_CLASS_INVALID; +#endif +} + + +DECL_FORCE_INLINE(int) rtSemRWRequestRead(RTSEMRW hRWSem, RTMSINTERVAL cMillies, PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + VERR_INVALID_HANDLE); + + /* + * Check if it's the writer (implement write+read recursion). + */ + pthread_t Self = pthread_self(); + pthread_t Writer; + ATOMIC_GET_PTHREAD_T(&pThis->Writer, &Writer); + if (Writer == Self) + { +#ifdef RTSEMRW_STRICT + int rc9 = RTLockValidatorRecExclRecursionMixed(&pThis->ValidatorWrite, &pThis->ValidatorRead.Core, pSrcPos); + if (RT_FAILURE(rc9)) + return rc9; +#endif + Assert(pThis->cWriterReads < INT32_MAX); + pThis->cWriterReads++; + return VINF_SUCCESS; + } + + /* + * Try lock it. + */ + RTTHREAD hThreadSelf = NIL_RTTHREAD; + if (cMillies > 0) + { +#ifdef RTSEMRW_STRICT + hThreadSelf = RTThreadSelfAutoAdopt(); + int rc9 = RTLockValidatorRecSharedCheckOrderAndBlocking(&pThis->ValidatorRead, hThreadSelf, pSrcPos, true, + cMillies, RTTHREADSTATE_RW_READ, true); + if (RT_FAILURE(rc9)) + return rc9; +#else + hThreadSelf = RTThreadSelf(); + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_RW_READ, true); + RT_NOREF_PV(pSrcPos); +#endif + } + + if (cMillies == RT_INDEFINITE_WAIT) + { + /* take rwlock */ + int rc = pthread_rwlock_rdlock(&pThis->RWLock); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_RW_READ); + if (rc) + { + AssertMsgFailed(("Failed read lock read-write sem %p, rc=%d.\n", hRWSem, rc)); + return RTErrConvertFromErrno(rc); + } + } + else + { +#ifdef RT_OS_DARWIN + AssertMsgFailed(("Not implemented on Darwin yet because of incomplete pthreads API.")); + return VERR_NOT_IMPLEMENTED; + +#else /* !RT_OS_DARWIN */ + /* + * Get current time and calc end of wait time. + */ + struct timespec ts = {0,0}; + clock_gettime(CLOCK_REALTIME, &ts); + if (cMillies != 0) + { + ts.tv_nsec += (cMillies % 1000) * 1000000; + ts.tv_sec += cMillies / 1000; + if (ts.tv_nsec >= 1000000000) + { + ts.tv_nsec -= 1000000000; + ts.tv_sec++; + } + } + + /* take rwlock */ + int rc = pthread_rwlock_timedrdlock(&pThis->RWLock, &ts); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_RW_READ); + if (rc) + { + AssertMsg(rc == ETIMEDOUT, ("Failed read lock read-write sem %p, rc=%d.\n", hRWSem, rc)); + return RTErrConvertFromErrno(rc); + } +#endif /* !RT_OS_DARWIN */ + } + + ASMAtomicIncU32(&pThis->cReaders); +#ifdef RTSEMRW_STRICT + RTLockValidatorRecSharedAddOwner(&pThis->ValidatorRead, hThreadSelf, pSrcPos); +#endif + return VINF_SUCCESS; +} + + +#undef RTSemRWRequestRead +RTDECL(int) RTSemRWRequestRead(RTSEMRW hRWSem, RTMSINTERVAL cMillies) +{ +#ifndef RTSEMRW_STRICT + return rtSemRWRequestRead(hRWSem, cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemRWRequestRead(hRWSem, cMillies, &SrcPos); +#endif +} + + +RTDECL(int) RTSemRWRequestReadDebug(RTSEMRW hRWSem, RTMSINTERVAL cMillies, RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemRWRequestRead(hRWSem, cMillies, &SrcPos); +} + + +#undef RTSemRWRequestReadNoResume +RTDECL(int) RTSemRWRequestReadNoResume(RTSEMRW hRWSem, RTMSINTERVAL cMillies) +{ + /* EINTR isn't returned by the wait functions we're using. */ +#ifndef RTSEMRW_STRICT + return rtSemRWRequestRead(hRWSem, cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemRWRequestRead(hRWSem, cMillies, &SrcPos); +#endif +} + + +RTDECL(int) RTSemRWRequestReadNoResumeDebug(RTSEMRW hRWSem, RTMSINTERVAL cMillies, RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemRWRequestRead(hRWSem, cMillies, &SrcPos); +} + + +RTDECL(int) RTSemRWReleaseRead(RTSEMRW hRWSem) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + VERR_INVALID_HANDLE); + + /* + * Check if it's the writer. + */ + pthread_t Self = pthread_self(); + pthread_t Writer; + ATOMIC_GET_PTHREAD_T(&pThis->Writer, &Writer); + if (Writer == Self) + { + AssertMsgReturn(pThis->cWriterReads > 0, ("pThis=%p\n", pThis), VERR_NOT_OWNER); +#ifdef RTSEMRW_STRICT + int rc9 = RTLockValidatorRecExclUnwindMixed(&pThis->ValidatorWrite, &pThis->ValidatorRead.Core); + if (RT_FAILURE(rc9)) + return rc9; +#endif + pThis->cWriterReads--; + return VINF_SUCCESS; + } + + /* + * Try unlock it. + */ +#ifdef RTSEMRW_STRICT + int rc9 = RTLockValidatorRecSharedCheckAndRelease(&pThis->ValidatorRead, RTThreadSelf()); + if (RT_FAILURE(rc9)) + return rc9; +#endif +#ifdef RT_OS_LINUX /* glibc (at least 2.8) may screw up when unlocking a lock we don't own. */ + if (ASMAtomicReadU32(&pThis->cReaders) == 0) + { + AssertMsgFailed(("Not owner of %p\n", pThis)); + return VERR_NOT_OWNER; + } +#endif + ASMAtomicDecU32(&pThis->cReaders); + int rc = pthread_rwlock_unlock(&pThis->RWLock); + if (rc) + { + ASMAtomicIncU32(&pThis->cReaders); + AssertMsgFailed(("Failed read unlock read-write sem %p, rc=%d.\n", hRWSem, rc)); + return RTErrConvertFromErrno(rc); + } + return VINF_SUCCESS; +} + + +DECL_FORCE_INLINE(int) rtSemRWRequestWrite(RTSEMRW hRWSem, RTMSINTERVAL cMillies, PCRTLOCKVALSRCPOS pSrcPos) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + VERR_INVALID_HANDLE); + + /* + * Recursion? + */ + pthread_t Self = pthread_self(); + pthread_t Writer; + ATOMIC_GET_PTHREAD_T(&pThis->Writer, &Writer); + if (Writer == Self) + { +#ifdef RTSEMRW_STRICT + int rc9 = RTLockValidatorRecExclRecursion(&pThis->ValidatorWrite, pSrcPos); + if (RT_FAILURE(rc9)) + return rc9; +#endif + Assert(pThis->cWrites < INT32_MAX); + pThis->cWrites++; + return VINF_SUCCESS; + } + + /* + * Try lock it. + */ + RTTHREAD hThreadSelf = NIL_RTTHREAD; + if (cMillies) + { +#ifdef RTSEMRW_STRICT + hThreadSelf = RTThreadSelfAutoAdopt(); + int rc9 = RTLockValidatorRecExclCheckOrderAndBlocking(&pThis->ValidatorWrite, hThreadSelf, pSrcPos, true, + cMillies, RTTHREADSTATE_RW_WRITE, true); + if (RT_FAILURE(rc9)) + return rc9; +#else + hThreadSelf = RTThreadSelf(); + RTThreadBlocking(hThreadSelf, RTTHREADSTATE_RW_WRITE, true); + RT_NOREF_PV(pSrcPos); +#endif + } + + if (cMillies == RT_INDEFINITE_WAIT) + { + /* take rwlock */ + int rc = pthread_rwlock_wrlock(&pThis->RWLock); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_RW_WRITE); + if (rc) + { + AssertMsgFailed(("Failed write lock read-write sem %p, rc=%d.\n", hRWSem, rc)); + return RTErrConvertFromErrno(rc); + } + } + else + { +#ifdef RT_OS_DARWIN + AssertMsgFailed(("Not implemented on Darwin yet because of incomplete pthreads API.")); + return VERR_NOT_IMPLEMENTED; +#else /* !RT_OS_DARWIN */ + /* + * Get current time and calc end of wait time. + */ + struct timespec ts = {0,0}; + clock_gettime(CLOCK_REALTIME, &ts); + if (cMillies != 0) + { + ts.tv_nsec += (cMillies % 1000) * 1000000; + ts.tv_sec += cMillies / 1000; + if (ts.tv_nsec >= 1000000000) + { + ts.tv_nsec -= 1000000000; + ts.tv_sec++; + } + } + + /* take rwlock */ + int rc = pthread_rwlock_timedwrlock(&pThis->RWLock, &ts); + RTThreadUnblocked(hThreadSelf, RTTHREADSTATE_RW_WRITE); + if (rc) + { + AssertMsg(rc == ETIMEDOUT, ("Failed read lock read-write sem %p, rc=%d.\n", hRWSem, rc)); + return RTErrConvertFromErrno(rc); + } +#endif /* !RT_OS_DARWIN */ + } + + ATOMIC_SET_PTHREAD_T(&pThis->Writer, Self); + pThis->cWrites = 1; + Assert(!pThis->cReaders); +#ifdef RTSEMRW_STRICT + RTLockValidatorRecExclSetOwner(&pThis->ValidatorWrite, hThreadSelf, pSrcPos, true); +#endif + return VINF_SUCCESS; +} + + +#undef RTSemRWRequestWrite +RTDECL(int) RTSemRWRequestWrite(RTSEMRW hRWSem, RTMSINTERVAL cMillies) +{ +#ifndef RTSEMRW_STRICT + return rtSemRWRequestWrite(hRWSem, cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemRWRequestWrite(hRWSem, cMillies, &SrcPos); +#endif +} + + +RTDECL(int) RTSemRWRequestWriteDebug(RTSEMRW hRWSem, RTMSINTERVAL cMillies, RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemRWRequestWrite(hRWSem, cMillies, &SrcPos); +} + + +#undef RTSemRWRequestWriteNoResume +RTDECL(int) RTSemRWRequestWriteNoResume(RTSEMRW hRWSem, RTMSINTERVAL cMillies) +{ + /* EINTR isn't returned by the wait functions we're using. */ +#ifndef RTSEMRW_STRICT + return rtSemRWRequestWrite(hRWSem, cMillies, NULL); +#else + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_NORMAL_API(); + return rtSemRWRequestWrite(hRWSem, cMillies, &SrcPos); +#endif +} + + +RTDECL(int) RTSemRWRequestWriteNoResumeDebug(RTSEMRW hRWSem, RTMSINTERVAL cMillies, RTHCUINTPTR uId, RT_SRC_POS_DECL) +{ + /* EINTR isn't returned by the wait functions we're using. */ + RTLOCKVALSRCPOS SrcPos = RTLOCKVALSRCPOS_INIT_DEBUG_API(); + return rtSemRWRequestWrite(hRWSem, cMillies, &SrcPos); +} + + +RTDECL(int) RTSemRWReleaseWrite(RTSEMRW hRWSem) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, VERR_INVALID_HANDLE); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + VERR_INVALID_HANDLE); + + /* + * Verify ownership and implement recursion. + */ + pthread_t Self = pthread_self(); + pthread_t Writer; + ATOMIC_GET_PTHREAD_T(&pThis->Writer, &Writer); + AssertMsgReturn(Writer == Self, ("pThis=%p\n", pThis), VERR_NOT_OWNER); + AssertReturn(pThis->cWriterReads == 0 || pThis->cWrites > 1, VERR_WRONG_ORDER); + + if (pThis->cWrites > 1) + { +#ifdef RTSEMRW_STRICT + int rc9 = RTLockValidatorRecExclUnwind(&pThis->ValidatorWrite); + if (RT_FAILURE(rc9)) + return rc9; +#endif + pThis->cWrites--; + return VINF_SUCCESS; + } + + /* + * Try unlock it. + */ +#ifdef RTSEMRW_STRICT + int rc9 = RTLockValidatorRecExclReleaseOwner(&pThis->ValidatorWrite, true); + if (RT_FAILURE(rc9)) + return rc9; +#endif + + pThis->cWrites--; + ATOMIC_SET_PTHREAD_T(&pThis->Writer, (pthread_t)-1); + int rc = pthread_rwlock_unlock(&pThis->RWLock); + if (rc) + { + AssertMsgFailed(("Failed write unlock read-write sem %p, rc=%d.\n", hRWSem, rc)); + return RTErrConvertFromErrno(rc); + } + + return VINF_SUCCESS; +} + + +RTDECL(bool) RTSemRWIsWriteOwner(RTSEMRW hRWSem) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, false); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + false); + + /* + * Check ownership. + */ + pthread_t Self = pthread_self(); + pthread_t Writer; + ATOMIC_GET_PTHREAD_T(&pThis->Writer, &Writer); + return Writer == Self; +} + + +RTDECL(bool) RTSemRWIsReadOwner(RTSEMRW hRWSem, bool fWannaHear) +{ + /* + * Validate handle. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, false); + AssertReturn(pThis->u32Magic == RTSEMRW_MAGIC, false); + + /* + * Check write ownership. The writer is also a valid reader. + */ + pthread_t Self = pthread_self(); + pthread_t Writer; + ATOMIC_GET_PTHREAD_T(&pThis->Writer, &Writer); + if (Writer == Self) + return true; + if (Writer != (pthread_t)-1) + return false; + + /* + * If there are no readers, we cannot be one of them, can we? + */ + if (ASMAtomicReadU32(&pThis->cReaders) == 0) + return false; + +#ifdef RTSEMRW_STRICT + /* + * Ask the lock validator. + */ + NOREF(fWannaHear); + return RTLockValidatorRecSharedIsOwner(&pThis->ValidatorRead, NIL_RTTHREAD); +#else + /* + * Just tell the caller what he want to hear. + */ + return fWannaHear; +#endif +} +RT_EXPORT_SYMBOL(RTSemRWIsReadOwner); + + +RTDECL(uint32_t) RTSemRWGetWriteRecursion(RTSEMRW hRWSem) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, 0); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + 0); + + /* + * Return the requested data. + */ + return pThis->cWrites; +} + + +RTDECL(uint32_t) RTSemRWGetWriterReadRecursion(RTSEMRW hRWSem) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, 0); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + 0); + + /* + * Return the requested data. + */ + return pThis->cWriterReads; +} + + +RTDECL(uint32_t) RTSemRWGetReadCount(RTSEMRW hRWSem) +{ + /* + * Validate input. + */ + struct RTSEMRWINTERNAL *pThis = hRWSem; + AssertPtrReturn(pThis, 0); + AssertMsgReturn(pThis->u32Magic == RTSEMRW_MAGIC, + ("pThis=%p u32Magic=%#x\n", pThis, pThis->u32Magic), + 0); + + /* + * Return the requested data. + */ + return pThis->cReaders; +} + diff --git a/src/VBox/Runtime/r3/posix/semwait.h b/src/VBox/Runtime/r3/posix/semwait.h new file mode 100644 index 00000000..ade8c4a7 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/semwait.h @@ -0,0 +1,162 @@ +/* $Id: semwait.h $ */ +/** @file + * IPRT - Common semaphore wait code. + */ + +/* + * Copyright (C) 2021-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + +#ifndef IPRT_INCLUDED_SRC_r3_posix_semwait_h +#define IPRT_INCLUDED_SRC_r3_posix_semwait_h +#ifndef RT_WITHOUT_PRAGMA_ONCE +# pragma once +#endif + + +/** @def IPRT_HAVE_PTHREAD_CONDATTR_SETCLOCK + * Set if the platform implements pthread_condattr_setclock(). + * Enables the use of the monotonic clock for waiting on condition variables. */ +#ifndef IPRT_HAVE_PTHREAD_CONDATTR_SETCLOCK +/* Linux detection */ +# if defined(RT_OS_LINUX) && defined(__USE_XOPEN2K) +# include <features.h> +# if __GLIBC_PREREQ(2,6) /** @todo figure the exact version where this was added */ +# define IPRT_HAVE_PTHREAD_CONDATTR_SETCLOCK +# endif +# endif +/** @todo check other platforms */ +#endif + + +/** + * Converts a extended wait timeout specification to an absolute timespec and a + * relative nanosecond count. + * + * @note This does not check for RTSEMWAIT_FLAGS_INDEFINITE, caller should've + * done that already. + * + * @returns The relative wait in nanoseconds. 0 for a poll call, UINT64_MAX for + * an effectively indefinite wait. + * @param fFlags RTSEMWAIT_FLAGS_XXX. + * @param fMonotonicClock Whether the timeout is in monotonic (true) or real + * (false) time. + * @param uTimeout The timeout. + * @param pAbsDeadline Where to return the absolute deadline. + */ +DECLINLINE(uint64_t) rtSemPosixCalcDeadline(uint32_t fFlags, uint64_t uTimeout, bool fMonotonicClock, + struct timespec *pAbsDeadline) +{ + Assert(!(fFlags & RTSEMWAIT_FLAGS_INDEFINITE)); + + /* + * Convert uTimeout to a relative value in nanoseconds. + */ + if (fFlags & RTSEMWAIT_FLAGS_MILLISECS) + { + if (uTimeout < UINT64_MAX / RT_NS_1MS) + uTimeout = uTimeout * RT_NS_1MS; + else + return UINT64_MAX; + } + else if (uTimeout == UINT64_MAX) /* unofficial way of indicating an indefinite wait */ + return UINT64_MAX; + + /* + * Make uTimeout relative and check for polling (zero timeout) calls. + */ + uint64_t uAbsTimeout = uTimeout; + if (fFlags & RTSEMWAIT_FLAGS_ABSOLUTE) + { + uint64_t const u64Now = RTTimeSystemNanoTS(); + if (uTimeout > u64Now) + uTimeout -= u64Now; + else + return 0; + } + else if (uTimeout == 0) + return 0; + + /* + * Calculate the deadline according to the clock we're using. + */ + if (!fMonotonicClock) + { +#if defined(RT_OS_DARWIN) || defined(RT_OS_HAIKU) + struct timeval tv = {0,0}; + gettimeofday(&tv, NULL); + pAbsDeadline->tv_sec = tv.tv_sec; + pAbsDeadline->tv_nsec = tv.tv_usec * 1000; +#else + clock_gettime(CLOCK_REALTIME, pAbsDeadline); +#endif + struct timespec TsAdd; + TsAdd.tv_nsec = uTimeout % RT_NS_1SEC; + TsAdd.tv_sec = uTimeout / RT_NS_1SEC; + + /* Check for 32-bit tv_sec overflows: */ + if ( sizeof(pAbsDeadline->tv_sec) < sizeof(uint64_t) + && ( uTimeout >= (uint64_t)RT_NS_1SEC * UINT32_MAX + || (uint64_t)pAbsDeadline->tv_sec + pAbsDeadline->tv_sec >= UINT32_MAX) ) + return UINT64_MAX; + + pAbsDeadline->tv_sec += TsAdd.tv_sec; + pAbsDeadline->tv_nsec += TsAdd.tv_nsec; + if ((uint32_t)pAbsDeadline->tv_nsec >= RT_NS_1SEC) + { + pAbsDeadline->tv_nsec -= RT_NS_1SEC; + pAbsDeadline->tv_sec++; + } + } + else + { + /* ASSUMES RTTimeSystemNanoTS() == RTTimeNanoTS() == clock_gettime(CLOCK_MONOTONIC). */ + if (fFlags & RTSEMWAIT_FLAGS_RELATIVE) + { + uint64_t const nsNow = RTTimeSystemNanoTS(); + uAbsTimeout += nsNow; + if (uAbsTimeout < nsNow) + return UINT64_MAX; + } + + /* Check for 32-bit tv_sec overflows: */ + if ( sizeof(pAbsDeadline->tv_sec) < sizeof(uint64_t) + && uAbsTimeout >= (uint64_t)RT_NS_1SEC * UINT32_MAX) + return UINT64_MAX; + + pAbsDeadline->tv_nsec = uAbsTimeout % RT_NS_1SEC; + pAbsDeadline->tv_sec = uAbsTimeout / RT_NS_1SEC; + } + + return uTimeout; +} + +#endif /* !IPRT_INCLUDED_SRC_r3_posix_semwait_h */ + diff --git a/src/VBox/Runtime/r3/posix/serialport-posix.cpp b/src/VBox/Runtime/r3/posix/serialport-posix.cpp new file mode 100644 index 00000000..0afbb676 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/serialport-posix.cpp @@ -0,0 +1,1269 @@ +/* $Id: serialport-posix.cpp $ */ +/** @file + * IPRT - Serial Port API, POSIX Implementation. + */ + +/* + * Copyright (C) 2017-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/serialport.h> +#include "internal/iprt.h" + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/cdefs.h> +#include <iprt/err.h> +#include <iprt/mem.h> +#include <iprt/string.h> +#include <iprt/thread.h> +#include <iprt/time.h> +#include "internal/magics.h" + +#include <errno.h> +#ifdef RT_OS_SOLARIS +# include <sys/termios.h> +#else +# include <termios.h> +#endif +#include <sys/types.h> +#include <fcntl.h> +#include <string.h> +#include <unistd.h> +#ifdef RT_OS_DARWIN +# include <sys/poll.h> +#else +# include <sys/poll.h> +#endif +#include <sys/ioctl.h> +#include <pthread.h> + +#ifdef RT_OS_LINUX +/* + * TIOCM_LOOP is not defined in the above header files for some reason but in asm/termios.h. + * But inclusion of this file however leads to compilation errors because of redefinition of some + * structs. That's why it is defined here until a better solution is found. + */ +# ifndef TIOCM_LOOP +# define TIOCM_LOOP 0x8000 +# endif +/* For linux custom baudrate code we also need serial_struct */ +# include <linux/serial.h> +#endif /* linux */ + +/** Define fallback if not supported. */ +#if !defined(CMSPAR) +# define CMSPAR 0 +#endif + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ + +/** + * Internal serial port state. + */ +typedef struct RTSERIALPORTINTERNAL +{ + /** Magic value (RTSERIALPORT_MAGIC). */ + uint32_t u32Magic; + /** Flags given while opening the serial port. */ + uint32_t fOpenFlags; + /** The file descriptor of the serial port. */ + int iFd; + /** The status line monitor thread if enabled. */ + RTTHREAD hMonThrd; + /** Flag whether the monitoring thread should shutdown. */ + volatile bool fMonThrdShutdown; + /** Reading end of wakeup pipe. */ + int iFdPipeR; + /** Writing end of wakeup pipe. */ + int iFdPipeW; + /** Event pending mask. */ + volatile uint32_t fEvtsPending; + /** Flag whether we are in blocking or non blocking mode. */ + bool fBlocking; + /** The current active config (we assume no one changes this behind our back). */ + struct termios PortCfg; + /** Flag whether a custom baud rate was chosen (for hosts supporting this.). */ + bool fBaudrateCust; + /** The custom baud rate. */ + uint32_t uBaudRateCust; +} RTSERIALPORTINTERNAL; +/** Pointer to the internal serial port state. */ +typedef RTSERIALPORTINTERNAL *PRTSERIALPORTINTERNAL; + + +/** + * Baud rate conversion table descriptor. + */ +typedef struct RTSERIALPORTBRATECONVDESC +{ + /** The platform independent baud rate used by the RTSerialPort* API. */ + uint32_t uBaudRateCfg; + /** The speed identifier used in the termios structure. */ + speed_t iSpeedTermios; +} RTSERIALPORTBRATECONVDESC; +/** Pointer to a baud rate converions table descriptor. */ +typedef RTSERIALPORTBRATECONVDESC *PRTSERIALPORTBRATECONVDESC; +/** Pointer to a const baud rate conversion table descriptor. */ +typedef const RTSERIALPORTBRATECONVDESC *PCRTSERIALPORTBRATECONVDESC; + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ + +/** The event poller was woken up due to an external interrupt. */ +#define RTSERIALPORT_WAKEUP_PIPE_REASON_INTERRUPT 0x0 +/** The event poller was woken up due to a change in the monitored status lines. */ +#define RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_CHANGED 0x1 +/** The monitor thread encoutnered repeating errors querying the status lines and terminated. */ +#define RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_MONITOR_FAILED 0x2 + + +/********************************************************************************************************************************* +* Global variables * +*********************************************************************************************************************************/ + +/** The baud rate conversion table. */ +static const RTSERIALPORTBRATECONVDESC s_rtSerialPortBaudrateConv[] = +{ + { 50, B50 }, + { 75, B75 }, + { 110, B110 }, + { 134, B134 }, + { 150, B150 }, + { 200, B200 }, + { 300, B300 }, + { 600, B600 }, + { 1200, B1200 }, + { 1800, B1800 }, + { 2400, B2400 }, + { 4800, B4800 }, + { 9600, B9600 }, + { 19200, B19200 }, + { 38400, B38400 }, + { 57600, B57600 }, + { 115200, B115200 } +}; + + + +/********************************************************************************************************************************* +* Internal Functions * +*********************************************************************************************************************************/ + +/** + * Converts the given termios speed identifier to the baud rate used in the API. + * + * @returns Baud rate or 0 if not a standard baud rate + */ +DECLINLINE(uint32_t) rtSerialPortGetBaudrateFromTermiosSpeed(speed_t enmSpeed) +{ + for (unsigned i = 0; i < RT_ELEMENTS(s_rtSerialPortBaudrateConv); i++) + { + if (s_rtSerialPortBaudrateConv[i].iSpeedTermios == enmSpeed) + return s_rtSerialPortBaudrateConv[i].uBaudRateCfg; + } + + return 0; +} + + +/** + * Converts the given baud rate to proper termios speed identifier. + * + * @returns Speed identifier if available or B0 if no matching speed for the baud rate + * could be found. + * @param uBaudRate The baud rate to convert. + * @param pfBaudrateCust Where to store the flag whether a custom baudrate was selected. + */ +DECLINLINE(speed_t) rtSerialPortGetTermiosSpeedFromBaudrate(uint32_t uBaudRate, bool *pfBaudrateCust) +{ + *pfBaudrateCust = false; + + for (unsigned i = 0; i < RT_ELEMENTS(s_rtSerialPortBaudrateConv); i++) + { + if (s_rtSerialPortBaudrateConv[i].uBaudRateCfg == uBaudRate) + return s_rtSerialPortBaudrateConv[i].iSpeedTermios; + } + +#ifdef RT_OS_LINUX + *pfBaudrateCust = true; + return B38400; +#else + return B0; +#endif +} + + +/** + * Tries to set the default config on the given serial port. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + */ +static int rtSerialPortSetDefaultCfg(PRTSERIALPORTINTERNAL pThis) +{ + pThis->fBaudrateCust = false; + pThis->uBaudRateCust = 0; + pThis->PortCfg.c_iflag = INPCK; /* Input parity checking. */ + cfsetispeed(&pThis->PortCfg, B9600); + cfsetospeed(&pThis->PortCfg, B9600); + pThis->PortCfg.c_cflag |= CS8 | CLOCAL; /* 8 data bits, ignore modem control lines. */ + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_READ) + pThis->PortCfg.c_cflag |= CREAD; /* Enable receiver. */ + + /* Set to raw input mode. */ + pThis->PortCfg.c_lflag &= ~(ICANON | ECHO | ECHOE | ECHONL | ECHOK | ISIG | IEXTEN); + pThis->PortCfg.c_cc[VMIN] = 0; /* Achieve non-blocking behavior. */ + pThis->PortCfg.c_cc[VTIME] = 0; + + int rc = VINF_SUCCESS; + int rcPsx = tcflush(pThis->iFd, TCIOFLUSH); + if (!rcPsx) + { + rcPsx = tcsetattr(pThis->iFd, TCSANOW, &pThis->PortCfg); + if (rcPsx == -1) + rc = RTErrConvertFromErrno(errno); + + if (RT_SUCCESS(rc)) + { +#ifdef RT_OS_LINUX + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_ENABLE_LOOPBACK) + { + int fTiocmSet = TIOCM_LOOP; + rcPsx = ioctl(pThis->iFd, TIOCMBIS, &fTiocmSet); + if (rcPsx == -1) + rc = RTErrConvertFromErrno(errno); + } + else + { + /* Make sure it is clear. */ + int fTiocmClear = TIOCM_LOOP; + rcPsx = ioctl(pThis->iFd, TIOCMBIC, &fTiocmClear); + if (rcPsx == -1 && errno != EINVAL) /* Pseudo terminals don't support loopback mode so ignore an error here. */ + rc = RTErrConvertFromErrno(errno); + } +#else + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_ENABLE_LOOPBACK) + return VERR_NOT_SUPPORTED; +#endif + } + } + else + rc = RTErrConvertFromErrno(errno); + + return rc; +} + + +/** + * Converts the given serial port config to the appropriate termios counterpart. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + * @param pCfg Pointer to the serial port config descriptor. + * @param pTermios Pointer to the termios structure to fill. + * @param pfBaudrateCust Where to store the flag whether a custom baudrate was selected. + * @param pErrInfo Additional error to be set when the conversion fails. + */ +static int rtSerialPortCfg2Termios(PRTSERIALPORTINTERNAL pThis, PCRTSERIALPORTCFG pCfg, struct termios *pTermios, + bool *pfBaudrateCust, PRTERRINFO pErrInfo) +{ + RT_NOREF(pErrInfo); /** @todo Make use of the error info. */ + speed_t enmSpeed = rtSerialPortGetTermiosSpeedFromBaudrate(pCfg->uBaudRate, pfBaudrateCust); + if (enmSpeed != B0) + { + tcflag_t const fCFlagMask = (CS5 | CS6 | CS7 | CS8 | CSTOPB | PARENB | PARODD | CMSPAR); + tcflag_t fCFlagNew = CLOCAL; + + switch (pCfg->enmDataBitCount) + { + case RTSERIALPORTDATABITS_5BITS: + fCFlagNew |= CS5; + break; + case RTSERIALPORTDATABITS_6BITS: + fCFlagNew |= CS6; + break; + case RTSERIALPORTDATABITS_7BITS: + fCFlagNew |= CS7; + break; + case RTSERIALPORTDATABITS_8BITS: + fCFlagNew |= CS8; + break; + default: + AssertFailed(); + return VERR_INVALID_PARAMETER; + } + + switch (pCfg->enmParity) + { + case RTSERIALPORTPARITY_NONE: + break; + case RTSERIALPORTPARITY_EVEN: + fCFlagNew |= PARENB; + break; + case RTSERIALPORTPARITY_ODD: + fCFlagNew |= PARENB | PARODD; + break; +#if CMSPAR != 0 + case RTSERIALPORTPARITY_MARK: + fCFlagNew |= PARENB | CMSPAR | PARODD; + break; + case RTSERIALPORTPARITY_SPACE: + fCFlagNew |= PARENB | CMSPAR; + break; +#else + case RTSERIALPORTPARITY_MARK: + case RTSERIALPORTPARITY_SPACE: + return VERR_NOT_SUPPORTED; +#endif + default: + AssertFailed(); + return VERR_INVALID_PARAMETER; + } + + switch (pCfg->enmStopBitCount) + { + case RTSERIALPORTSTOPBITS_ONE: + break; + case RTSERIALPORTSTOPBITS_ONEPOINTFIVE: + if (pCfg->enmDataBitCount == RTSERIALPORTDATABITS_5BITS) + fCFlagNew |= CSTOPB; + else + return VERR_NOT_SUPPORTED; + break; + case RTSERIALPORTSTOPBITS_TWO: + if (pCfg->enmDataBitCount != RTSERIALPORTDATABITS_5BITS) + fCFlagNew |= CSTOPB; + else + return VERR_NOT_SUPPORTED; + break; + default: + AssertFailed(); + return VERR_INVALID_PARAMETER; + } + + /* Assign new flags. */ + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_READ) + pTermios->c_cflag |= CREAD; /* Enable receiver. */ + pTermios->c_cflag = (pTermios->c_cflag & ~fCFlagMask) | fCFlagNew; + pTermios->c_lflag &= ~(ICANON | ECHO | ECHOE | ECHONL | ECHOK | ISIG | IEXTEN); + pTermios->c_iflag = INPCK; /* Input parity checking. */ + pTermios->c_cc[VMIN] = 0; /* Achieve non-blocking behavior. */ + pTermios->c_cc[VTIME] = 0; + cfsetispeed(pTermios, enmSpeed); + cfsetospeed(pTermios, enmSpeed); + } + else + return VERR_SERIALPORT_INVALID_BAUDRATE; + + return VINF_SUCCESS; +} + + +/** + * Converts the given termios structure to an appropriate serial port config. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + * @param pTermios The termios structure to convert. + * @param pCfg The serial port config to fill in. + */ +static int rtSerialPortTermios2Cfg(PRTSERIALPORTINTERNAL pThis, struct termios *pTermios, PRTSERIALPORTCFG pCfg) +{ + int rc = VINF_SUCCESS; + bool f5DataBits = false; + speed_t enmSpeedIn = cfgetispeed(pTermios); + Assert(enmSpeedIn == cfgetospeed(pTermios)); /* Should always be the same. */ + + if (!pThis->fBaudrateCust) + { + pCfg->uBaudRate = rtSerialPortGetBaudrateFromTermiosSpeed(enmSpeedIn); + if (!pCfg->uBaudRate) + rc = VERR_SERIALPORT_INVALID_BAUDRATE; + } + else + pCfg->uBaudRate = pThis->uBaudRateCust; + + switch (pTermios->c_cflag & CSIZE) + { + case CS5: + pCfg->enmDataBitCount = RTSERIALPORTDATABITS_5BITS; + f5DataBits = true; + break; + case CS6: + pCfg->enmDataBitCount = RTSERIALPORTDATABITS_6BITS; + break; + case CS7: + pCfg->enmDataBitCount = RTSERIALPORTDATABITS_7BITS; + break; + case CS8: + pCfg->enmDataBitCount = RTSERIALPORTDATABITS_8BITS; + break; + default: + AssertFailed(); /* Should not happen. */ + pCfg->enmDataBitCount = RTSERIALPORTDATABITS_INVALID; + rc = RT_FAILURE(rc) ? rc : VERR_INVALID_PARAMETER; + } + + /* Convert parity. */ + if (pTermios->c_cflag & PARENB) + { + /* + * CMSPAR is not supported on all systems, especially OS X. As configuring + * mark/space parity there is not supported and we start from a known config + * when opening the serial port it is not required to check for this here. + */ +#if CMSPAR == 0 + bool fCmsParSet = RT_BOOL(pTermios->c_cflag & CMSPAR); +#else + bool fCmsParSet = false; +#endif + if (pTermios->c_cflag & PARODD) + pCfg->enmParity = fCmsParSet ? RTSERIALPORTPARITY_MARK : RTSERIALPORTPARITY_ODD; + else + pCfg->enmParity = fCmsParSet ? RTSERIALPORTPARITY_SPACE: RTSERIALPORTPARITY_EVEN; + } + else + pCfg->enmParity = RTSERIALPORTPARITY_NONE; + + /* + * 1.5 stop bits are used with a data count of 5 bits when a UART derived from the 8250 + * is used. + */ + if (pTermios->c_cflag & CSTOPB) + pCfg->enmStopBitCount = f5DataBits ? RTSERIALPORTSTOPBITS_ONEPOINTFIVE : RTSERIALPORTSTOPBITS_TWO; + else + pCfg->enmStopBitCount = RTSERIALPORTSTOPBITS_ONE; + + return rc; +} + + +/** + * Wakes up any thread polling for a serial port event with the given reason. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + * @param bWakeupReason The wakeup reason to pass to the event poller. + */ +DECLINLINE(int) rtSerialPortWakeupEvtPoller(PRTSERIALPORTINTERNAL pThis, uint8_t bWakeupReason) +{ + int rcPsx = write(pThis->iFdPipeW, &bWakeupReason, 1); + if (rcPsx != 1) + return RTErrConvertFromErrno(errno); + + return VINF_SUCCESS; +} + + +/** + * The status line monitor thread worker. + * + * @returns IPRT status code. + * @param ThreadSelf Thread handle to this thread. + * @param pvUser User argument. + */ +static DECLCALLBACK(int) rtSerialPortStsLineMonitorThrd(RTTHREAD hThreadSelf, void *pvUser) +{ + RT_NOREF(hThreadSelf); + PRTSERIALPORTINTERNAL pThis = (PRTSERIALPORTINTERNAL)pvUser; + unsigned long const fStsLinesChk = TIOCM_CAR | TIOCM_RNG | TIOCM_DSR | TIOCM_CTS; + int rc = VINF_SUCCESS; + uint32_t fStsLinesOld = 0; + uint32_t cStsLineGetErrors = 0; +#ifdef RT_OS_LINUX + bool fPoll = false; +#endif + + RTThreadUserSignal(hThreadSelf); + + int rcPsx = ioctl(pThis->iFd, TIOCMGET, &fStsLinesOld); + if (rcPsx == -1) + { + ASMAtomicXchgBool(&pThis->fMonThrdShutdown, true); + return RTErrConvertFromErrno(errno); + } + + while ( !pThis->fMonThrdShutdown + && RT_SUCCESS(rc)) + { +# ifdef RT_OS_LINUX + /* + * Wait for status line change. + * + * XXX In Linux, if a thread calls tcsetattr while the monitor thread is + * waiting in ioctl for a modem status change then 8250.c wrongly disables + * modem irqs and so the monitor thread never gets released. The workaround + * is to send a signal after each tcsetattr. + * + * TIOCMIWAIT doesn't work for the DSR line with TIOCM_DSR set + * (see http://lxr.linux.no/#linux+v4.7/drivers/usb/class/cdc-acm.c#L949) + * However as it is possible to query the line state we will not just clear + * the TIOCM_DSR bit from the lines to check but resort to the polling + * approach just like on other hosts. + */ + if (!fPoll) + { + rcPsx = ioctl(pThis->iFd, TIOCMIWAIT, fStsLinesChk); + if (!rcPsx) + { + rc = rtSerialPortWakeupEvtPoller(pThis, RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_CHANGED); + if (RT_FAILURE(rc)) + break; + } + else if (rcPsx == -1 && errno != EINTR) + fPoll = true; + } + else +#endif + { + uint32_t fStsLines = 0; + rcPsx = ioctl(pThis->iFd, TIOCMGET, &fStsLines); + if (!rcPsx) + { + cStsLineGetErrors = 0; /* Reset the error counter once we had one successful query. */ + + if (((fStsLines ^ fStsLinesOld) & fStsLinesChk)) + { + rc = rtSerialPortWakeupEvtPoller(pThis, RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_CHANGED); + if (RT_FAILURE(rc)) + break; + + fStsLinesOld = fStsLines; + } + else /* No change, sleep for a bit. */ + RTThreadSleep(100 /*ms*/); + } + else if (rcPsx == -1 && errno != EINTR) + { + /* + * If querying the status line fails too often we have to shut down the + * thread and notify the user of the serial port. + */ + if (cStsLineGetErrors++ >= 10) + { + rc = RTErrConvertFromErrno(errno); + rtSerialPortWakeupEvtPoller(pThis, RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_MONITOR_FAILED); + break; + } + + RTThreadSleep(100 /*ms*/); + } + } + } + + ASMAtomicXchgBool(&pThis->fMonThrdShutdown, true); + return rc; +} + + +/** + * Creates the status line monitoring thread. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + */ +static int rtSerialPortMonitorThreadCreate(PRTSERIALPORTINTERNAL pThis) +{ + int rc = VINF_SUCCESS; + + /* + * Check whether querying the status lines is supported at all, pseudo terminals + * don't support it so an error returned in that case. + */ + uint32_t fStsLines = 0; + int rcPsx = ioctl(pThis->iFd, TIOCMGET, &fStsLines); + if (!rcPsx) + { + pThis->fMonThrdShutdown = false; + rc = RTThreadCreate(&pThis->hMonThrd, rtSerialPortStsLineMonitorThrd, pThis, 0 /*cbStack*/, + RTTHREADTYPE_IO, RTTHREADFLAGS_WAITABLE, "IPRT-SerPortMon"); + if (RT_SUCCESS(rc)) + { + /* Wait for the thread to start up. */ + rc = RTThreadUserWait(pThis->hMonThrd, 20*RT_MS_1SEC); + if ( rc == VERR_TIMEOUT + || pThis->fMonThrdShutdown) + { + /* Startup failed, try to reap the thread. */ + int rcThrd; + rc = RTThreadWait(pThis->hMonThrd, 20*RT_MS_1SEC, &rcThrd); + if (RT_SUCCESS(rc)) + rc = rcThrd; + else + rc = VERR_INTERNAL_ERROR; + /* The thread is lost otherwise. */ + } + } + } + else if (errno == ENOTTY || errno == EINVAL) + rc = VERR_NOT_SUPPORTED; + else + rc = RTErrConvertFromErrno(errno); + + return rc; +} + + +/** + * Shuts down the status line monitor thread. + * + * @param pThis The internal serial port instance data. + */ +static void rtSerialPortMonitorThreadShutdown(PRTSERIALPORTINTERNAL pThis) +{ + bool fShutDown = ASMAtomicXchgBool(&pThis->fMonThrdShutdown, true); + if (!fShutDown) + { + int rc = RTThreadPoke(pThis->hMonThrd); + AssertRC(rc); + } + + int rcThrd = VINF_SUCCESS; + int rc = RTThreadWait(pThis->hMonThrd, 20*RT_MS_1SEC, &rcThrd); + AssertRC(rc); + AssertRC(rcThrd); +} + + +/** + * The slow path of rtSerialPortSwitchBlockingMode that does the actual switching. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + * @param fBlocking The desired mode of operation. + * @remarks Do not call directly. + */ +static int rtSerialPortSwitchBlockingModeSlow(PRTSERIALPORTINTERNAL pThis, bool fBlocking) +{ + int fFlags = fcntl(pThis->iFd, F_GETFL, 0); + if (fFlags == -1) + return RTErrConvertFromErrno(errno); + + if (fBlocking) + fFlags &= ~O_NONBLOCK; + else + fFlags |= O_NONBLOCK; + if (fcntl(pThis->iFd, F_SETFL, fFlags) == -1) + return RTErrConvertFromErrno(errno); + + pThis->fBlocking = fBlocking; + return VINF_SUCCESS; +} + + +/** + * Switches the serial port to the desired blocking mode if necessary. + * + * @returns IPRT status code. + * @param pThis The internal serial port instance data. + * @param fBlocking The desired mode of operation. + */ +DECLINLINE(int) rtSerialPortSwitchBlockingMode(PRTSERIALPORTINTERNAL pThis, bool fBlocking) +{ + if (pThis->fBlocking != fBlocking) + return rtSerialPortSwitchBlockingModeSlow(pThis, fBlocking); + return VINF_SUCCESS; +} + + +RTDECL(int) RTSerialPortOpen(PRTSERIALPORT phSerialPort, const char *pszPortAddress, uint32_t fFlags) +{ + AssertPtrReturn(phSerialPort, VERR_INVALID_POINTER); + AssertPtrReturn(pszPortAddress, VERR_INVALID_POINTER); + AssertReturn(*pszPortAddress != '\0', VERR_INVALID_PARAMETER); + AssertReturn(!(fFlags & ~RTSERIALPORT_OPEN_F_VALID_MASK), VERR_INVALID_PARAMETER); + AssertReturn((fFlags & RTSERIALPORT_OPEN_F_READ) || (fFlags & RTSERIALPORT_OPEN_F_WRITE), + VERR_INVALID_PARAMETER); + + int rc = VINF_SUCCESS; + PRTSERIALPORTINTERNAL pThis = (PRTSERIALPORTINTERNAL)RTMemAllocZ(sizeof(*pThis)); + if (pThis) + { + int fPsxFlags = O_NOCTTY | O_NONBLOCK; + + if ((fFlags & RTSERIALPORT_OPEN_F_READ) && !(fFlags & RTSERIALPORT_OPEN_F_WRITE)) + fPsxFlags |= O_RDONLY; + else if (!(fFlags & RTSERIALPORT_OPEN_F_READ) && (fFlags & RTSERIALPORT_OPEN_F_WRITE)) + fPsxFlags |= O_WRONLY; + else + fPsxFlags |= O_RDWR; + + pThis->u32Magic = RTSERIALPORT_MAGIC; + pThis->fOpenFlags = fFlags; + pThis->fEvtsPending = 0; + pThis->iFd = open(pszPortAddress, fPsxFlags); + pThis->fBlocking = false; + if (pThis->iFd != -1) + { + /* Create wakeup pipe for the event API. */ + int aPipeFds[2]; + int rcPsx = pipe(&aPipeFds[0]); + if (!rcPsx) + { + /* Make the pipes close on exec. */ + pThis->iFdPipeR = aPipeFds[0]; + pThis->iFdPipeW = aPipeFds[1]; + + if (fcntl(pThis->iFdPipeR, F_SETFD, FD_CLOEXEC)) + rc = RTErrConvertFromErrno(errno); + + if ( RT_SUCCESS(rc) + && fcntl(pThis->iFdPipeW, F_SETFD, FD_CLOEXEC)) + rc = RTErrConvertFromErrno(errno); + + if (RT_SUCCESS(rc)) + { + rc = rtSerialPortSetDefaultCfg(pThis); + if ( RT_SUCCESS(rc) + && (fFlags & RTSERIALPORT_OPEN_F_SUPPORT_STATUS_LINE_MONITORING)) + rc = rtSerialPortMonitorThreadCreate(pThis); + + if (RT_SUCCESS(rc)) + { + *phSerialPort = pThis; + return VINF_SUCCESS; + } + } + + close(pThis->iFdPipeR); + close(pThis->iFdPipeW); + } + else + rc = RTErrConvertFromErrno(errno); + + close(pThis->iFd); + } + else + rc = RTErrConvertFromErrno(errno); + + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + + return rc; +} + + +RTDECL(int) RTSerialPortClose(RTSERIALPORT hSerialPort) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + if (pThis == NIL_RTSERIALPORT) + return VINF_SUCCESS; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + + /* + * Do the cleanup. + */ + AssertReturn(ASMAtomicCmpXchgU32(&pThis->u32Magic, RTSERIALPORT_MAGIC_DEAD, RTSERIALPORT_MAGIC), VERR_INVALID_HANDLE); + + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_SUPPORT_STATUS_LINE_MONITORING) + rtSerialPortMonitorThreadShutdown(pThis); + + close(pThis->iFd); + close(pThis->iFdPipeR); + close(pThis->iFdPipeW); + RTMemFree(pThis); + return VINF_SUCCESS; +} + + +RTDECL(RTHCINTPTR) RTSerialPortToNative(RTSERIALPORT hSerialPort) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, -1); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, -1); + + return pThis->iFd; +} + + +RTDECL(int) RTSerialPortRead(RTSERIALPORT hSerialPort, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pvBuf, VERR_INVALID_POINTER); + AssertReturn(cbToRead > 0, VERR_INVALID_PARAMETER); + + int rc = rtSerialPortSwitchBlockingMode(pThis, true); + if (RT_SUCCESS(rc)) + { + /* + * Attempt read. + */ + ssize_t cbRead = read(pThis->iFd, pvBuf, cbToRead); + if (cbRead > 0) + { + if (pcbRead) + /* caller can handle partial read. */ + *pcbRead = cbRead; + else + { + /* Caller expects all to be read. */ + while ((ssize_t)cbToRead > cbRead) + { + ssize_t cbReadPart = read(pThis->iFd, (uint8_t *)pvBuf + cbRead, cbToRead - cbRead); + if (cbReadPart < 0) + return RTErrConvertFromErrno(errno); + else if (cbReadPart == 0) + return VERR_DEV_IO_ERROR; + + cbRead += cbReadPart; + } + } + } + else if (cbRead == 0) + rc = VERR_DEV_IO_ERROR; + else + rc = RTErrConvertFromErrno(errno); + } + + return rc; +} + + +RTDECL(int) RTSerialPortReadNB(RTSERIALPORT hSerialPort, void *pvBuf, size_t cbToRead, size_t *pcbRead) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pvBuf, VERR_INVALID_POINTER); + AssertReturn(cbToRead > 0, VERR_INVALID_PARAMETER); + AssertPtrReturn(pcbRead, VERR_INVALID_POINTER); + + *pcbRead = 0; + + int rc = rtSerialPortSwitchBlockingMode(pThis, false); + if (RT_SUCCESS(rc)) + { + ssize_t cbThisRead = read(pThis->iFd, pvBuf, cbToRead); + if (cbThisRead > 0) + { + /* + * The read data needs to be scanned for the BREAK condition marker encoded in the data stream, + * if break detection was enabled during open. + */ + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_DETECT_BREAK_CONDITION) + { /** @todo */ } + + *pcbRead = cbThisRead; + } + else if (cbThisRead == 0) + rc = VERR_DEV_IO_ERROR; + else if ( errno == EAGAIN +# ifdef EWOULDBLOCK +# if EWOULDBLOCK != EAGAIN + || errno == EWOULDBLOCK +# endif +# endif + ) + rc = VINF_TRY_AGAIN; + else + rc = RTErrConvertFromErrno(errno); + } + + return rc; +} + + +RTDECL(int) RTSerialPortWrite(RTSERIALPORT hSerialPort, const void *pvBuf, size_t cbToWrite, size_t *pcbWritten) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pvBuf, VERR_INVALID_POINTER); + AssertReturn(cbToWrite > 0, VERR_INVALID_PARAMETER); + + int rc = rtSerialPortSwitchBlockingMode(pThis, true); + if (RT_SUCCESS(rc)) + { + /* + * Attempt write. + */ + ssize_t cbWritten = write(pThis->iFd, pvBuf, cbToWrite); + if (cbWritten > 0) + { + if (pcbWritten) + /* caller can handle partial write. */ + *pcbWritten = cbWritten; + else + { + /* Caller expects all to be written. */ + while ((ssize_t)cbToWrite > cbWritten) + { + ssize_t cbWrittenPart = write(pThis->iFd, (const uint8_t *)pvBuf + cbWritten, cbToWrite - cbWritten); + if (cbWrittenPart < 0) + return RTErrConvertFromErrno(errno); + else if (cbWrittenPart == 0) + return VERR_DEV_IO_ERROR; + cbWritten += cbWrittenPart; + } + } + } + else if (cbWritten == 0) + rc = VERR_DEV_IO_ERROR; + else + rc = RTErrConvertFromErrno(errno); + } + + return rc; +} + + +RTDECL(int) RTSerialPortWriteNB(RTSERIALPORT hSerialPort, const void *pvBuf, size_t cbToWrite, size_t *pcbWritten) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pvBuf, VERR_INVALID_POINTER); + AssertReturn(cbToWrite > 0, VERR_INVALID_PARAMETER); + AssertPtrReturn(pcbWritten, VERR_INVALID_POINTER); + + *pcbWritten = 0; + + int rc = rtSerialPortSwitchBlockingMode(pThis, false); + if (RT_SUCCESS(rc)) + { + ssize_t cbThisWrite = write(pThis->iFd, pvBuf, cbToWrite); + if (cbThisWrite > 0) + *pcbWritten = cbThisWrite; + else if (cbThisWrite == 0) + rc = VERR_DEV_IO_ERROR; + else if ( errno == EAGAIN +# ifdef EWOULDBLOCK +# if EWOULDBLOCK != EAGAIN + || errno == EWOULDBLOCK +# endif +# endif + ) + rc = VINF_TRY_AGAIN; + else + rc = RTErrConvertFromErrno(errno); + } + + return rc; +} + + +RTDECL(int) RTSerialPortCfgQueryCurrent(RTSERIALPORT hSerialPort, PRTSERIALPORTCFG pCfg) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + + return rtSerialPortTermios2Cfg(pThis, &pThis->PortCfg, pCfg); +} + + +RTDECL(int) RTSerialPortCfgSet(RTSERIALPORT hSerialPort, PCRTSERIALPORTCFG pCfg, PRTERRINFO pErrInfo) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + + struct termios PortCfgNew; RT_ZERO(PortCfgNew); + bool fBaudrateCust = false; + int rc = rtSerialPortCfg2Termios(pThis, pCfg, &PortCfgNew, &fBaudrateCust, pErrInfo); + if (RT_SUCCESS(rc)) + { + int rcPsx = tcflush(pThis->iFd, TCIOFLUSH); + if (!rcPsx) + { +#ifdef RT_OS_LINUX + if (fBaudrateCust) + { + struct serial_struct SerLnx; + rcPsx = ioctl(pThis->iFd, TIOCGSERIAL, &SerLnx); + if (!rcPsx) + { + SerLnx.custom_divisor = SerLnx.baud_base / pCfg->uBaudRate; + if (!SerLnx.custom_divisor) + SerLnx.custom_divisor = 1; + SerLnx.flags &= ~ASYNC_SPD_MASK; + SerLnx.flags |= ASYNC_SPD_CUST; + rcPsx = ioctl(pThis->iFd, TIOCSSERIAL, &SerLnx); + } + } +#else /* !RT_OS_LINUX */ + /* Hosts not supporting custom baud rates should already fail in rtSerialPortCfg2Termios(). */ + AssertMsgFailed(("Should not get here!\n")); +#endif /* !RT_OS_LINUX */ + pThis->fBaudrateCust = fBaudrateCust; + pThis->uBaudRateCust = pCfg->uBaudRate; + + if (!rcPsx) + rcPsx = tcsetattr(pThis->iFd, TCSANOW, &PortCfgNew); + if (rcPsx == -1) + rc = RTErrConvertFromErrno(errno); + else + memcpy(&pThis->PortCfg, &PortCfgNew, sizeof(struct termios)); + +#ifdef RT_OS_LINUX + /* + * XXX In Linux, if a thread calls tcsetattr while the monitor thread is + * waiting in ioctl for a modem status change then 8250.c wrongly disables + * modem irqs and so the monitor thread never gets released. The workaround + * is to send a signal after each tcsetattr. + */ + if (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_SUPPORT_STATUS_LINE_MONITORING) + RTThreadPoke(pThis->hMonThrd); +#endif + } + else + rc = RTErrConvertFromErrno(errno); + } + + return rc; +} + + +RTDECL(int) RTSerialPortEvtPoll(RTSERIALPORT hSerialPort, uint32_t fEvtMask, uint32_t *pfEvtsRecv, + RTMSINTERVAL msTimeout) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(!(fEvtMask & ~RTSERIALPORT_EVT_F_VALID_MASK), VERR_INVALID_PARAMETER); + AssertPtrReturn(pfEvtsRecv, VERR_INVALID_POINTER); + + *pfEvtsRecv = 0; + + fEvtMask |= RTSERIALPORT_EVT_F_STATUS_LINE_MONITOR_FAILED; /* This will be reported always, no matter what the caller wants. */ + + /* Return early if there are events pending from previous calls which weren't fetched yet. */ + for (;;) + { + uint32_t fEvtsPending = ASMAtomicReadU32(&pThis->fEvtsPending); + if (fEvtsPending & fEvtMask) + { + *pfEvtsRecv = fEvtsPending & fEvtMask; + /* Write back, repeat the whole procedure if someone else raced us. */ + if (ASMAtomicCmpXchgU32(&pThis->fEvtsPending, fEvtsPending & ~fEvtMask, fEvtsPending)) + return VINF_SUCCESS; + } + else + break; + } + + int rc = rtSerialPortSwitchBlockingMode(pThis, false); + if (RT_SUCCESS(rc)) + { + struct pollfd aPollFds[2]; RT_ZERO(aPollFds); + aPollFds[0].fd = pThis->iFd; + aPollFds[0].events = POLLERR | POLLHUP; + aPollFds[0].revents = 0; + if ( (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_READ) + && (fEvtMask & RTSERIALPORT_EVT_F_DATA_RX)) + aPollFds[0].events |= POLLIN; + if ( (pThis->fOpenFlags & RTSERIALPORT_OPEN_F_WRITE) + && (fEvtMask & RTSERIALPORT_EVT_F_DATA_TX)) + aPollFds[0].events |= POLLOUT; + + aPollFds[1].fd = pThis->iFdPipeR; + aPollFds[1].events = POLLIN | POLLERR | POLLHUP; + aPollFds[1].revents = 0; + + int rcPsx = 0; + int msTimeoutLeft = msTimeout == RT_INDEFINITE_WAIT ? -1 : (int)msTimeout; + while (msTimeoutLeft != 0) + { + uint64_t tsPollStart = RTTimeMilliTS(); + + rcPsx = poll(&aPollFds[0], RT_ELEMENTS(aPollFds), msTimeoutLeft); + if (rcPsx != -1 || errno != EINTR) + break; + /* Restart when getting interrupted. */ + if (msTimeoutLeft > -1) + { + uint64_t tsPollEnd = RTTimeMilliTS(); + uint64_t tsPollSpan = tsPollEnd - tsPollStart; + msTimeoutLeft -= RT_MIN(tsPollSpan, (uint32_t)msTimeoutLeft); + } + } + + uint32_t fEvtsPending = 0; + if (rcPsx < 0 && errno != EINTR) + rc = RTErrConvertFromErrno(errno); + else if (rcPsx > 0) + { + if (aPollFds[0].revents != 0) + { + if (aPollFds[0].revents & POLLERR) + rc = VERR_DEV_IO_ERROR; + else + { + fEvtsPending |= (aPollFds[0].revents & POLLIN) ? RTSERIALPORT_EVT_F_DATA_RX : 0; + fEvtsPending |= (aPollFds[0].revents & POLLOUT) ? RTSERIALPORT_EVT_F_DATA_TX : 0; + /** @todo BREAK condition detection. */ + } + } + + if (aPollFds[1].revents != 0) + { + AssertReturn(!(aPollFds[1].revents & (POLLHUP | POLLERR | POLLNVAL)), VERR_INTERNAL_ERROR); + Assert(aPollFds[1].revents & POLLIN); + + uint8_t bWakeupReason = 0; + ssize_t cbRead = read(pThis->iFdPipeR, &bWakeupReason, 1); + if (cbRead == 1) + { + switch (bWakeupReason) + { + case RTSERIALPORT_WAKEUP_PIPE_REASON_INTERRUPT: + rc = VERR_INTERRUPTED; + break; + case RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_CHANGED: + fEvtsPending |= RTSERIALPORT_EVT_F_STATUS_LINE_CHANGED; + break; + case RTSERIALPORT_WAKEUP_PIPE_REASON_STS_LINE_MONITOR_FAILED: + fEvtsPending |= RTSERIALPORT_EVT_F_STATUS_LINE_MONITOR_FAILED; + break; + default: + AssertFailed(); + rc = VERR_INTERNAL_ERROR; + } + } + else + rc = VERR_INTERNAL_ERROR; + } + } + else + rc = VERR_TIMEOUT; + + *pfEvtsRecv = fEvtsPending & fEvtMask; + fEvtsPending &= ~fEvtMask; + ASMAtomicOrU32(&pThis->fEvtsPending, fEvtsPending); + } + + return rc; +} + + +RTDECL(int) RTSerialPortEvtPollInterrupt(RTSERIALPORT hSerialPort) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + + return rtSerialPortWakeupEvtPoller(pThis, RTSERIALPORT_WAKEUP_PIPE_REASON_INTERRUPT); +} + + +RTDECL(int) RTSerialPortChgBreakCondition(RTSERIALPORT hSerialPort, bool fSet) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + + int rc = VINF_SUCCESS; + int rcPsx = ioctl(pThis->iFd, fSet ? TIOCSBRK : TIOCCBRK); + if (rcPsx == -1) + rc = RTErrConvertFromErrno(errno); + + return rc; +} + + +RTDECL(int) RTSerialPortChgStatusLines(RTSERIALPORT hSerialPort, uint32_t fClear, uint32_t fSet) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + + int rc = VINF_SUCCESS; + int fTiocmSet = 0; + int fTiocmClear = 0; + + if (fClear & RTSERIALPORT_CHG_STS_LINES_F_RTS) + fTiocmClear |= TIOCM_RTS; + if (fClear & RTSERIALPORT_CHG_STS_LINES_F_DTR) + fTiocmClear |= TIOCM_DTR; + + if (fSet & RTSERIALPORT_CHG_STS_LINES_F_RTS) + fTiocmSet |= TIOCM_RTS; + if (fSet & RTSERIALPORT_CHG_STS_LINES_F_DTR) + fTiocmSet |= TIOCM_DTR; + + int rcPsx = ioctl(pThis->iFd, TIOCMBIS, &fTiocmSet); + if (!rcPsx) + { + rcPsx = ioctl(pThis->iFd, TIOCMBIC, &fTiocmClear); + if (rcPsx == -1) + rc = RTErrConvertFromErrno(errno); + } + return rc; +} + + +RTDECL(int) RTSerialPortQueryStatusLines(RTSERIALPORT hSerialPort, uint32_t *pfStsLines) +{ + PRTSERIALPORTINTERNAL pThis = hSerialPort; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSERIALPORT_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pfStsLines, VERR_INVALID_POINTER); + + *pfStsLines = 0; + + int rc = VINF_SUCCESS; + int fStsLines = 0; + int rcPsx = ioctl(pThis->iFd, TIOCMGET, &fStsLines); + if (!rcPsx) + { + /* This resets the status line event pending flag. */ + for (;;) + { + uint32_t fEvtsPending = ASMAtomicReadU32(&pThis->fEvtsPending); + if (ASMAtomicCmpXchgU32(&pThis->fEvtsPending, fEvtsPending & ~RTSERIALPORT_EVT_F_STATUS_LINE_CHANGED, fEvtsPending)) + break; + } + + *pfStsLines |= (fStsLines & TIOCM_CAR) ? RTSERIALPORT_STS_LINE_DCD : 0; + *pfStsLines |= (fStsLines & TIOCM_RNG) ? RTSERIALPORT_STS_LINE_RI : 0; + *pfStsLines |= (fStsLines & TIOCM_DSR) ? RTSERIALPORT_STS_LINE_DSR : 0; + *pfStsLines |= (fStsLines & TIOCM_CTS) ? RTSERIALPORT_STS_LINE_CTS : 0; + } + else + rc = RTErrConvertFromErrno(errno); + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/shmem-posix.cpp b/src/VBox/Runtime/r3/posix/shmem-posix.cpp new file mode 100644 index 00000000..5e21f26b --- /dev/null +++ b/src/VBox/Runtime/r3/posix/shmem-posix.cpp @@ -0,0 +1,419 @@ +/* $Id: shmem-posix.cpp $ */ +/** @file + * IPRT - Named shared memory object, POSIX Implementation. + */ + +/* + * Copyright (C) 2018-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/shmem.h> +#include "internal/iprt.h" + +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/cdefs.h> +#include <iprt/err.h> +#include <iprt/mem.h> +#include <iprt/string.h> +#include "internal/magics.h" + +#include <errno.h> +#include <sys/mman.h> +#include <sys/stat.h> +#include <fcntl.h> +#include <unistd.h> +#include <limits.h> + +/* Workaround on systems which do not provide this. */ +#ifndef NAME_MAX +# define NAME_MAX 255 +#endif + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ + +/** + * Shared memory object mapping descriptor. + */ +typedef struct RTSHMEMMAPPINGDESC +{ + /** Number of references held to this mapping, 0 if the descriptor is free. */ + volatile uint32_t cMappings; + /** Pointer to the region mapping. */ + void *pvMapping; + /** Start offset */ + size_t offRegion; + /** Size of the region. */ + size_t cbRegion; + /** Access flags for this region .*/ + uint32_t fFlags; +} RTSHMEMMAPPINGDESC; +/** Pointer to a shared memory object mapping descriptor. */ +typedef RTSHMEMMAPPINGDESC *PRTSHMEMMAPPINGDESC; +/** Pointer to a constant shared memory object mapping descriptor. */ +typedef const RTSHMEMMAPPINGDESC *PCRTSHMEMMAPPINGDESC; + + +/** + * Internal shared memory object state. + */ +typedef struct RTSHMEMINT +{ + /** Magic value (RTSHMEM_MAGIC). */ + uint32_t u32Magic; + /** File descriptor for the underlying shared memory object. */ + int iFdShm; + /** Pointer to the shared memory object name. */ + char *pszName; + /** Flag whether this instance created the named shared memory object. */ + bool fCreate; + /** Overall number of mappings active for this shared memory object. */ + volatile uint32_t cMappings; + /** Maximum number of mapping descriptors allocated. */ + uint32_t cMappingDescsMax; + /** Number of mapping descriptors used. */ + volatile uint32_t cMappingDescsUsed; + /** Array of mapping descriptors - variable in size. */ + RTSHMEMMAPPINGDESC aMappingDescs[1]; +} RTSHMEMINT; +/** Pointer to the internal shared memory object state. */ +typedef RTSHMEMINT *PRTSHMEMINT; + + + +/** + * Returns a mapping descriptor matching the given region properties or NULL if none was found. + * + * @returns Pointer to the matching mapping descriptor or NULL if not found. + * @param pThis Pointer to the shared memory object instance. + * @param offRegion Offset into the shared memory object to start mapping at. + * @param cbRegion Size of the region to map. + * @param fFlags Desired properties of the mapped region, combination of RTSHMEM_MAP_F_* defines. + */ +DECLINLINE(PRTSHMEMMAPPINGDESC) rtShMemMappingDescFindByProp(PRTSHMEMINT pThis, size_t offRegion, size_t cbRegion, uint32_t fFlags) +{ + for (uint32_t i = 0; i < pThis->cMappingDescsMax; i++) + { + if ( pThis->aMappingDescs[i].offRegion == offRegion + && pThis->aMappingDescs[i].cbRegion == cbRegion + && pThis->aMappingDescs[i].fFlags == fFlags) + return &pThis->aMappingDescs[i]; + } + + return NULL; +} + + +RTDECL(int) RTShMemOpen(PRTSHMEM phShMem, const char *pszName, uint32_t fFlags, size_t cbMax, uint32_t cMappingsHint) +{ + AssertPtrReturn(phShMem, VERR_INVALID_PARAMETER); + AssertPtrReturn(pszName, VERR_INVALID_PARAMETER); + AssertReturn(!(fFlags & ~RTSHMEM_O_F_VALID_MASK), VERR_INVALID_PARAMETER); + AssertReturn(cMappingsHint < 64, VERR_OUT_OF_RANGE); + + size_t cchName = strlen(pszName); + AssertReturn(cchName, VERR_INVALID_PARAMETER); + AssertReturn(cchName < NAME_MAX - 1, VERR_INVALID_PARAMETER); /* account for the / we add later on. */ + cMappingsHint = cMappingsHint == 0 ? 5 : cMappingsHint; + int rc = VINF_SUCCESS; + PRTSHMEMINT pThis = (PRTSHMEMINT)RTMemAllocZ(RT_UOFFSETOF_DYN(RTSHMEMINT, aMappingDescs[cMappingsHint]) + cchName + 2); /* '/' + terminator. */ + if (RT_LIKELY(pThis)) + { + pThis->u32Magic = RTSHMEM_MAGIC; + pThis->pszName = (char *)&pThis->aMappingDescs[cMappingsHint]; + /*pThis->fCreate = false; */ + /*pThis->cMappings = 0; */ + pThis->cMappingDescsMax = cMappingsHint; + /*pThis->cMappingDescsUsed = 0; */ + pThis->pszName[0] = '/'; + memcpy(&pThis->pszName[1], pszName, cchName); + int fShmFlags = 0; + if (fFlags & RTSHMEM_O_F_CREATE) + { + fShmFlags |= O_CREAT; + pThis->fCreate = true; + } + if ((fFlags & RTSHMEM_O_F_CREATE_EXCL) == RTSHMEM_O_F_CREATE_EXCL) + fShmFlags |= O_EXCL; + if ( (fFlags & RTSHMEM_O_F_READWRITE) == RTSHMEM_O_F_READWRITE + || (fFlags & RTSHMEM_O_F_WRITE)) + fShmFlags |= O_RDWR; + else + fShmFlags |= O_RDONLY; + if (fFlags & RTSHMEM_O_F_TRUNCATE) + fShmFlags |= O_TRUNC; + pThis->iFdShm = shm_open(pThis->pszName, fShmFlags , 0600); + if (pThis->iFdShm > 0) + { + if (cbMax) + rc = RTShMemSetSize(pThis, cbMax); + if (RT_SUCCESS(rc)) + { + *phShMem = pThis; + return VINF_SUCCESS; + } + + close(pThis->iFdShm); + } + else + rc = RTErrConvertFromErrno(errno); + + RTMemFree(pThis); + } + else + rc = VERR_NO_MEMORY; + + return rc; +} + + +RTDECL(int) RTShMemClose(RTSHMEM hShMem) +{ + PRTSHMEMINT pThis = hShMem; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSHMEM_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(!pThis->cMappings, VERR_INVALID_STATE); + + int rc = VINF_SUCCESS; + if (!close(pThis->iFdShm)) + { + if (pThis->fCreate) + shm_unlink(pThis->pszName); /* Ignore any error here. */ + pThis->u32Magic = RTSHMEM_MAGIC_DEAD; + RTMemFree(pThis); + } + else + rc = RTErrConvertFromErrno(errno); + + return rc; +} + + +RTDECL(int) RTShMemDelete(const char *pszName) +{ + AssertPtrReturn(pszName, VERR_INVALID_POINTER); + + size_t cchName = strlen(pszName); + AssertReturn(cchName, VERR_INVALID_PARAMETER); + AssertReturn(cchName < NAME_MAX - 1, VERR_INVALID_PARAMETER); /* account for the / we add later on. */ + char *psz = NULL; + + int rc = RTStrAllocEx(&psz, cchName + 2); /* '/' + terminator */ + if (RT_SUCCESS(rc)) + { + psz[0] = '/'; + memcpy(&psz[1], pszName, cchName + 1); + if (shm_unlink(psz)) + rc = RTErrConvertFromErrno(errno); + RTStrFree(psz); + } + + return rc; +} + + +RTDECL(uint32_t) RTShMemRefCount(RTSHMEM hShMem) +{ + PRTSHMEMINT pThis = hShMem; + AssertPtrReturn(pThis, 0); + AssertReturn(pThis->u32Magic == RTSHMEM_MAGIC, 0); + + return pThis->cMappings; +} + + +RTDECL(int) RTShMemSetSize(RTSHMEM hShMem, size_t cbMem) +{ + PRTSHMEMINT pThis = hShMem; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSHMEM_MAGIC, VERR_INVALID_HANDLE); + AssertReturn(!pThis->cMappings, VERR_INVALID_STATE); + + int rc = VINF_SUCCESS; + if (ftruncate(pThis->iFdShm, (off_t)cbMem)) + rc = RTErrConvertFromErrno(errno); + + return rc; +} + + +RTDECL(int) RTShMemQuerySize(RTSHMEM hShMem, size_t *pcbMem) +{ + PRTSHMEMINT pThis = hShMem; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSHMEM_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pcbMem, VERR_INVALID_PARAMETER); + + struct stat st; + if (!fstat(pThis->iFdShm, &st)) + { + *pcbMem = st.st_size; + return VINF_SUCCESS; + } + return RTErrConvertFromErrno(errno); +} + + +RTDECL(int) RTShMemMapRegion(RTSHMEM hShMem, size_t offRegion, size_t cbRegion, uint32_t fFlags, void **ppv) +{ + PRTSHMEMINT pThis = hShMem; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSHMEM_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(ppv, VERR_INVALID_PARAMETER); + AssertReturn(!(fFlags & ~RTSHMEM_MAP_F_VALID_MASK), VERR_INVALID_PARAMETER); + + /* Try to find a mapping with compatible parameters first. */ + PRTSHMEMMAPPINGDESC pMappingDesc = NULL; + for (uint32_t iTry = 0; iTry < 10; iTry++) + { + pMappingDesc = rtShMemMappingDescFindByProp(pThis, offRegion, cbRegion, fFlags); + if (!pMappingDesc) + break; + + /* Increase the mapping count and check that the region is still accessible by us. */ + if ( ASMAtomicIncU32(&pMappingDesc->cMappings) > 1 + && pMappingDesc->offRegion == offRegion + && pMappingDesc->cbRegion == cbRegion + && pMappingDesc->fFlags == fFlags) + break; + /* Mapping was freed inbetween, next round. */ + } + + int rc = VINF_SUCCESS; + if (!pMappingDesc) + { + /* Find an empty region descriptor and map the region. */ + for (uint32_t i = 0; i < pThis->cMappingDescsMax && !pMappingDesc; i++) + { + if (!pThis->aMappingDescs[i].cMappings) + { + pMappingDesc = &pThis->aMappingDescs[i]; + + /* Try to grab this one. */ + if (ASMAtomicIncU32(&pMappingDesc->cMappings) == 1) + break; + + /* Somebody raced us, drop reference and continue. */ + ASMAtomicDecU32(&pMappingDesc->cMappings); + pMappingDesc = NULL; + } + } + + if (RT_LIKELY(pMappingDesc)) + { + /* Try to map it. */ + int fMmapFlags = 0; + int fProt = 0; + if (fFlags & RTSHMEM_MAP_F_READ) + fProt |= PROT_READ; + if (fFlags & RTSHMEM_MAP_F_WRITE) + fProt |= PROT_WRITE; + if (fFlags & RTSHMEM_MAP_F_EXEC) + fProt |= PROT_EXEC; + if (fFlags & RTSHMEM_MAP_F_COW) + fMmapFlags |= MAP_PRIVATE; + else + fMmapFlags |= MAP_SHARED; + + void *pv = mmap(NULL, cbRegion, fProt, fMmapFlags, pThis->iFdShm, (off_t)offRegion); + if (pv != MAP_FAILED) + { + pMappingDesc->pvMapping = pv; + pMappingDesc->offRegion = offRegion; + pMappingDesc->cbRegion = cbRegion; + pMappingDesc->fFlags = fFlags; + } + else + { + rc = RTErrConvertFromErrno(errno); + ASMAtomicDecU32(&pMappingDesc->cMappings); + } + } + else + rc = VERR_SHMEM_MAXIMUM_MAPPINGS_REACHED; + } + + if (RT_SUCCESS(rc)) + { + *ppv = pMappingDesc->pvMapping; + ASMAtomicIncU32(&pThis->cMappings); + } + + return rc; +} + + +RTDECL(int) RTShMemUnmapRegion(RTSHMEM hShMem, void *pv) +{ + PRTSHMEMINT pThis = hShMem; + AssertPtrReturn(pThis, VERR_INVALID_PARAMETER); + AssertReturn(pThis->u32Magic == RTSHMEM_MAGIC, VERR_INVALID_HANDLE); + AssertPtrReturn(pv, VERR_INVALID_PARAMETER); + + /* Find the mapping descriptor by the given region address. */ + PRTSHMEMMAPPINGDESC pMappingDesc = NULL; + for (uint32_t i = 0; i < pThis->cMappingDescsMax && !pMappingDesc; i++) + { + if (pThis->aMappingDescs[i].pvMapping == pv) + { + pMappingDesc = &pThis->aMappingDescs[i]; + break; + } + } + + AssertPtrReturn(pMappingDesc, VERR_INVALID_PARAMETER); + + int rc = VINF_SUCCESS; + size_t cbRegion = pMappingDesc->cMappings; + if (!ASMAtomicDecU32(&pMappingDesc->cMappings)) + { + /* Last mapping of this region was unmapped, so do the real unmapping now. */ + if (munmap(pv, cbRegion)) + { + ASMAtomicIncU32(&pMappingDesc->cMappings); + rc = RTErrConvertFromErrno(errno); + } + else + { + ASMAtomicDecU32(&pThis->cMappingDescsUsed); + ASMAtomicDecU32(&pThis->cMappings); + } + } + + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/symlink-posix.cpp b/src/VBox/Runtime/r3/posix/symlink-posix.cpp new file mode 100644 index 00000000..e43f5b36 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/symlink-posix.cpp @@ -0,0 +1,247 @@ +/* $Id: symlink-posix.cpp $ */ +/** @file + * IPRT - Symbolic Links, POSIX. + */ + +/* + * Copyright (C) 2010-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_SYMLINK + +#include <errno.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <iprt/symlink.h> + +#include <iprt/assert.h> +#include <iprt/err.h> +#include <iprt/log.h> +#include <iprt/mem.h> +#include <iprt/string.h> +#include "internal/path.h" + + + +RTDECL(bool) RTSymlinkExists(const char *pszSymlink) +{ + bool fRc = false; + char const *pszNativeSymlink; + int rc = rtPathToNative(&pszNativeSymlink, pszSymlink, NULL); + if (RT_SUCCESS(rc)) + { + struct stat s; + fRc = !lstat(pszNativeSymlink, &s) + && S_ISLNK(s.st_mode); + + rtPathFreeNative(pszNativeSymlink, pszSymlink); + } + + LogFlow(("RTSymlinkExists(%p={%s}): returns %RTbool\n", pszSymlink, pszSymlink, fRc)); + return fRc; +} + + +RTDECL(bool) RTSymlinkIsDangling(const char *pszSymlink) +{ + bool fRc = false; + char const *pszNativeSymlink; + int rc = rtPathToNative(&pszNativeSymlink, pszSymlink, NULL); + if (RT_SUCCESS(rc)) + { + struct stat s; + fRc = !lstat(pszNativeSymlink, &s) + && S_ISLNK(s.st_mode); + if (fRc) + { + errno = 0; + fRc = stat(pszNativeSymlink, &s) != 0 + && ( errno == ENOENT + || errno == ENOTDIR + || errno == ELOOP); + } + + rtPathFreeNative(pszNativeSymlink, pszSymlink); + } + + LogFlow(("RTSymlinkIsDangling(%p={%s}): returns %RTbool\n", pszSymlink, pszSymlink, fRc)); + return fRc; +} + + +RTDECL(int) RTSymlinkCreate(const char *pszSymlink, const char *pszTarget, RTSYMLINKTYPE enmType, uint32_t fCreate) +{ + RT_NOREF_PV(fCreate); + + /* + * Validate the input. + */ + AssertReturn(enmType > RTSYMLINKTYPE_INVALID && enmType < RTSYMLINKTYPE_END, VERR_INVALID_PARAMETER); + AssertPtrReturn(pszSymlink, VERR_INVALID_POINTER); + AssertPtrReturn(pszTarget, VERR_INVALID_POINTER); + + /* + * Convert the paths. + */ + char const *pszNativeSymlink; + int rc = rtPathToNative(&pszNativeSymlink, pszSymlink, NULL); + if (RT_SUCCESS(rc)) + { + const char *pszNativeTarget; + rc = rtPathToNative(&pszNativeTarget, pszTarget, NULL); + if (RT_SUCCESS(rc)) + { + /* + * Create the link. + */ + if (symlink(pszNativeTarget, pszNativeSymlink) == 0) + rc = VINF_SUCCESS; + else + rc = RTErrConvertFromErrno(errno); + + rtPathFreeNative(pszNativeTarget, pszTarget); + } + rtPathFreeNative(pszNativeSymlink, pszSymlink); + } + + LogFlow(("RTSymlinkCreate(%p={%s}, %p={%s}, %d, %#x): returns %Rrc\n", pszSymlink, pszSymlink, pszTarget, pszTarget, enmType, fCreate, rc)); + return rc; +} + + +RTDECL(int) RTSymlinkDelete(const char *pszSymlink, uint32_t fDelete) +{ + RT_NOREF_PV(fDelete); + + char const *pszNativeSymlink; + int rc = rtPathToNative(&pszNativeSymlink, pszSymlink, NULL); + if (RT_SUCCESS(rc)) + { + struct stat s; + if (!lstat(pszNativeSymlink, &s)) + { + if (S_ISLNK(s.st_mode)) + { + if (unlink(pszNativeSymlink) == 0) + rc = VINF_SUCCESS; + else + rc = RTErrConvertFromErrno(errno); + } + else + rc = VERR_NOT_SYMLINK; + } + else + rc = RTErrConvertFromErrno(errno); + rtPathFreeNative(pszNativeSymlink, pszSymlink); + } + + LogFlow(("RTSymlinkDelete(%p={%s}, #%x): returns %Rrc\n", pszSymlink, pszSymlink, fDelete, rc)); + return rc; +} + + +RTDECL(int) RTSymlinkRead(const char *pszSymlink, char *pszTarget, size_t cbTarget, uint32_t fRead) +{ + RT_NOREF_PV(fRead); + + char *pszMyTarget; + int rc = RTSymlinkReadA(pszSymlink, &pszMyTarget); + if (RT_SUCCESS(rc)) + { + rc = RTStrCopy(pszTarget, cbTarget, pszMyTarget); + RTStrFree(pszMyTarget); + } + LogFlow(("RTSymlinkRead(%p={%s}): returns %Rrc\n", pszSymlink, pszSymlink, rc)); + return rc; +} + + +RTDECL(int) RTSymlinkReadA(const char *pszSymlink, char **ppszTarget) +{ + AssertPtr(ppszTarget); + char const *pszNativeSymlink; + int rc = rtPathToNative(&pszNativeSymlink, pszSymlink, NULL); + if (RT_SUCCESS(rc)) + { + /* Guess the initial buffer size. */ + ssize_t cbBuf; + struct stat s; + if (!lstat(pszNativeSymlink, &s)) + cbBuf = RT_MAX(RT_ALIGN_Z(s.st_size, 64), 64); + else + cbBuf = 1024; + + /* Read loop that grows the buffer. */ + char *pszBuf = NULL; + for (;;) + { + RTMemTmpFree(pszBuf); + pszBuf = (char *)RTMemTmpAlloc(cbBuf); + if (pszBuf) + { + ssize_t cbReturned = readlink(pszNativeSymlink, pszBuf, cbBuf); + if (cbReturned >= cbBuf) + { + /* Increase the buffer size and try again */ + cbBuf *= 2; + continue; + } + + if (cbReturned > 0) + { + pszBuf[cbReturned] = '\0'; + rc = rtPathFromNativeDup(ppszTarget, pszBuf, pszSymlink); + } + else if (errno == EINVAL) + rc = VERR_NOT_SYMLINK; + else + rc = RTErrConvertFromErrno(errno); + } + else + rc = VERR_NO_TMP_MEMORY; + break; + } /* for loop */ + + RTMemTmpFree(pszBuf); + rtPathFreeNative(pszNativeSymlink, pszSymlink); + } + + if (RT_SUCCESS(rc)) + LogFlow(("RTSymlinkReadA(%p={%s},%p): returns %Rrc *ppszTarget=%p:{%s}\n", pszSymlink, pszSymlink, ppszTarget, rc, *ppszTarget, *ppszTarget)); + else + LogFlow(("RTSymlinkReadA(%p={%s},%p): returns %Rrc\n", pszSymlink, pszSymlink, ppszTarget, rc)); + return rc; +} + diff --git a/src/VBox/Runtime/r3/posix/thread-posix.cpp b/src/VBox/Runtime/r3/posix/thread-posix.cpp new file mode 100644 index 00000000..cd621ce0 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/thread-posix.cpp @@ -0,0 +1,780 @@ +/* $Id: thread-posix.cpp $ */ +/** @file + * IPRT - Threads, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_THREAD +#include <errno.h> +#include <limits.h> +#include <pthread.h> +#include <signal.h> +#include <stdlib.h> +#if defined(RT_OS_LINUX) +# include <unistd.h> +# include <sys/syscall.h> +#endif +#if defined(RT_OS_SOLARIS) +# include <sched.h> +# include <sys/resource.h> +#endif +#if defined(RT_OS_DARWIN) +# include <mach/thread_act.h> +# include <mach/thread_info.h> +# include <mach/host_info.h> +# include <mach/mach_init.h> +# include <mach/mach_host.h> +#endif +#if defined(RT_OS_DARWIN) /*|| defined(RT_OS_FREEBSD) - later */ \ + || (defined(RT_OS_LINUX) && !defined(IN_RT_STATIC) /* static + dlsym = trouble */) \ + || defined(IPRT_MAY_HAVE_PTHREAD_SET_NAME_NP) +# define IPRT_MAY_HAVE_PTHREAD_SET_NAME_NP +# include <dlfcn.h> +#endif +#if defined(RT_OS_HAIKU) +# include <OS.h> +#endif +#if defined(RT_OS_DARWIN) +# define sigprocmask pthread_sigmask /* On xnu sigprocmask works on the process, not the calling thread as elsewhere. */ +#endif + +#include <iprt/thread.h> +#include <iprt/log.h> +#include <iprt/assert.h> +#include <iprt/asm.h> +#include <iprt/err.h> +#include <iprt/initterm.h> +#include <iprt/string.h> +#include <iprt/semaphore.h> +#include <iprt/list.h> +#include <iprt/once.h> +#include <iprt/critsect.h> +#include <iprt/req.h> +#include "internal/thread.h" + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +/*#ifndef IN_GUEST - shouldn't need to exclude this now with the non-obtrusive init option. */ +/** Includes RTThreadPoke. */ +# define RTTHREAD_POSIX_WITH_POKE +/*#endif*/ + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +/** The pthread key in which we store the pointer to our own PRTTHREAD structure. + * @note There is a defined NIL value here, nor can we really assume this is an + * integer. However, zero is a valid key on Linux, so we get into trouble + * if we accidentally use it uninitialized. + * + * So, we ASSUME it's a integer value and the valid range is in approx 0 + * to PTHREAD_KEYS_MAX. Solaris has at least one negative value (-1) + * defined. Thus, we go for 16 MAX values below zero and keep our fingers + * cross that it will always be an invalid key value everywhere... + * + * See also NIL_RTTLS, which is -1. + */ +static pthread_key_t g_SelfKey = (pthread_key_t)(-PTHREAD_KEYS_MAX * 16); +#ifdef RTTHREAD_POSIX_WITH_POKE +/** The signal we use for poking threads. + * This is set to -1 if no available signal was found. */ +static int volatile g_iSigPokeThread = -1; +#endif + +#ifdef IPRT_MAY_HAVE_PTHREAD_SET_NAME_NP +# if defined(RT_OS_DARWIN) +/** + * The Mac OS X (10.6 and later) variant of pthread_setname_np. + * + * @returns errno.h + * @param pszName The new thread name. + */ +typedef int (*PFNPTHREADSETNAME)(const char *pszName); +# else +/** + * The variant of pthread_setname_np most other unix-like systems implement. + * + * @returns errno.h + * @param hThread The thread. + * @param pszName The new thread name. + */ +typedef int (*PFNPTHREADSETNAME)(pthread_t hThread, const char *pszName); +# endif + +/** Pointer to pthread_setname_np if found. */ +static PFNPTHREADSETNAME g_pfnThreadSetName = NULL; +#endif /* IPRT_MAY_HAVE_PTHREAD_SET_NAME_NP */ + +#ifdef RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY +/** Atomic indicator of whether the priority proxy thread has been (attempted) started. + * + * The priority proxy thread is started under these circumstances: + * - RTThreadCreate + * - RTThreadSetType + * - RTProcSetPriority + * + * Which means that we'll be single threaded when this is modified. + * + * Speical values: + * - VERR_TRY_AGAIN: Not yet started. + * - VERR_WRONG_ORDER: Starting. + * - VINF_SUCCESS: Started successfully. + * - VERR_PROCESS_NOT_FOUND: Stopping or stopped + * - Other error status if failed to start. + * + * @note We could potentially optimize this by only start it when we lower the + * priority of ourselves, the process, or a newly created thread. But + * that would means we would need to take multi-threading into account, so + * let's not do that for now. + */ +static int32_t volatile g_rcPriorityProxyThreadStart = VERR_TRY_AGAIN; +/** The IPRT thread handle for the priority proxy. */ +static RTTHREAD g_hRTThreadPosixPriorityProxyThread = NIL_RTTHREAD; +/** The priority proxy queue. */ +static RTREQQUEUE g_hRTThreadPosixPriorityProxyQueue = NIL_RTREQQUEUE; +#endif /* RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY */ + + +/********************************************************************************************************************************* +* Internal Functions * +*********************************************************************************************************************************/ +static void *rtThreadNativeMain(void *pvArgs); +static void rtThreadKeyDestruct(void *pvValue); +#ifdef RTTHREAD_POSIX_WITH_POKE +static void rtThreadPosixPokeSignal(int iSignal); +#endif + + +#ifdef RTTHREAD_POSIX_WITH_POKE +/** + * Try register the dummy signal handler for RTThreadPoke. + */ +static void rtThreadPosixSelectPokeSignal(void) +{ + /* + * Note! Avoid SIGRTMIN thru SIGRTMIN+2 because of LinuxThreads. + */ +# if !defined(RT_OS_LINUX) && !defined(RT_OS_SOLARIS) /* glibc defines SIGRTMAX to __libc_current_sigrtmax() and Solaris libc defines it relying on _sysconf(), causing compiler to deploy serialization here. */ + static +# endif + const int s_aiSigCandidates[] = + { +# ifdef SIGRTMAX + SIGRTMAX-3, + SIGRTMAX-2, + SIGRTMAX-1, +# endif +# ifndef RT_OS_SOLARIS + SIGUSR2, +# endif + SIGWINCH + }; + + g_iSigPokeThread = -1; + if (!RTR3InitIsUnobtrusive()) + { + for (unsigned iSig = 0; iSig < RT_ELEMENTS(s_aiSigCandidates); iSig++) + { + struct sigaction SigActOld; + if (!sigaction(s_aiSigCandidates[iSig], NULL, &SigActOld)) + { + if ( SigActOld.sa_handler == SIG_DFL + || SigActOld.sa_handler == rtThreadPosixPokeSignal) + { + struct sigaction SigAct; + RT_ZERO(SigAct); + SigAct.sa_handler = rtThreadPosixPokeSignal; + SigAct.sa_flags = 0; /* no SA_RESTART! */ + sigfillset(&SigAct.sa_mask); + + /* ASSUMES no sigaction race... (lazy bird) */ + if (!sigaction(s_aiSigCandidates[iSig], &SigAct, NULL)) + { + g_iSigPokeThread = s_aiSigCandidates[iSig]; + break; + } + AssertMsgFailed(("rc=%Rrc errno=%d\n", RTErrConvertFromErrno(errno), errno)); + } + } + else + AssertMsgFailed(("rc=%Rrc errno=%d\n", RTErrConvertFromErrno(errno), errno)); + } + } +} +#endif /* RTTHREAD_POSIX_WITH_POKE */ + + +DECLHIDDEN(int) rtThreadNativeInit(void) +{ + /* + * Allocate the TLS (key in posix terms) where we store the pointer to + * a threads RTTHREADINT structure. + */ + int rc = pthread_key_create(&g_SelfKey, rtThreadKeyDestruct); + if (rc) + return VERR_NO_TLS_FOR_SELF; + +#ifdef RTTHREAD_POSIX_WITH_POKE + rtThreadPosixSelectPokeSignal(); +#endif + +#ifdef IPRT_MAY_HAVE_PTHREAD_SET_NAME_NP + if (RT_SUCCESS(rc)) + g_pfnThreadSetName = (PFNPTHREADSETNAME)(uintptr_t)dlsym(RTLD_DEFAULT, "pthread_setname_np"); +#endif + return rc; +} + +static void rtThreadPosixBlockSignals(PRTTHREADINT pThread) +{ + /* + * Mask all signals, including the poke one, if requested. + */ + if ( pThread + && (pThread->fFlags & RTTHREADFLAGS_NO_SIGNALS)) + { + sigset_t SigSet; + sigfillset(&SigSet); + sigdelset(&SigSet, SIGILL); /* On the m1 we end up spinning on UDF ... */ + sigdelset(&SigSet, SIGTRAP); /* ... and BRK instruction if these signals are masked. */ + sigdelset(&SigSet, SIGFPE); /* Just adding the rest here to be on the safe side. */ + sigdelset(&SigSet, SIGBUS); + sigdelset(&SigSet, SIGSEGV); + int rc = sigprocmask(SIG_BLOCK, &SigSet, NULL); + AssertMsg(rc == 0, ("rc=%Rrc errno=%d\n", RTErrConvertFromErrno(errno), errno)); RT_NOREF(rc); + } + /* + * Block SIGALRM - required for timer-posix.cpp. + * This is done to limit harm done by OSes which doesn't do special SIGALRM scheduling. + * It will not help much if someone creates threads directly using pthread_create. :/ + */ + else if (!RTR3InitIsUnobtrusive()) + { + sigset_t SigSet; + sigemptyset(&SigSet); + sigaddset(&SigSet, SIGALRM); + sigprocmask(SIG_BLOCK, &SigSet, NULL); + } + +#ifdef RTTHREAD_POSIX_WITH_POKE + /* + * bird 2020-10-28: Not entirely sure why we do this, but it makes sure the signal works + * on the new thread. Probably some pre-NPTL linux reasons. + */ + if (g_iSigPokeThread != -1) + { +# if 1 /* siginterrupt() is typically implemented as two sigaction calls, this should be faster and w/o deprecations: */ + struct sigaction SigActOld; + RT_ZERO(SigActOld); + + struct sigaction SigAct; + RT_ZERO(SigAct); + SigAct.sa_handler = rtThreadPosixPokeSignal; + SigAct.sa_flags = 0; /* no SA_RESTART! */ + sigfillset(&SigAct.sa_mask); + + int rc = sigaction(g_iSigPokeThread, &SigAct, &SigActOld); + AssertMsg(rc == 0, ("rc=%Rrc errno=%d\n", RTErrConvertFromErrno(errno), errno)); RT_NOREF(rc); + AssertMsg(rc || SigActOld.sa_handler == rtThreadPosixPokeSignal, ("%p\n", SigActOld.sa_handler)); +# else + siginterrupt(g_iSigPokeThread, 1); +# endif + } +#endif +} + +DECLHIDDEN(void) rtThreadNativeReInitObtrusive(void) +{ +#ifdef RTTHREAD_POSIX_WITH_POKE + Assert(!RTR3InitIsUnobtrusive()); + rtThreadPosixSelectPokeSignal(); +#endif + rtThreadPosixBlockSignals(NULL); +} + + +/** + * Destructor called when a thread terminates. + * @param pvValue The key value. PRTTHREAD in our case. + */ +static void rtThreadKeyDestruct(void *pvValue) +{ + /* + * Deal with alien threads. + */ + PRTTHREADINT pThread = (PRTTHREADINT)pvValue; + if (pThread->fIntFlags & RTTHREADINT_FLAGS_ALIEN) + { + pthread_setspecific(g_SelfKey, pThread); + rtThreadTerminate(pThread, 0); + pthread_setspecific(g_SelfKey, NULL); + } +} + + +#ifdef RTTHREAD_POSIX_WITH_POKE +/** + * Dummy signal handler for the poke signal. + * + * @param iSignal The signal number. + */ +static void rtThreadPosixPokeSignal(int iSignal) +{ + Assert(iSignal == g_iSigPokeThread); + NOREF(iSignal); +} +#endif + + +/** + * Adopts a thread, this is called immediately after allocating the + * thread structure. + * + * @param pThread Pointer to the thread structure. + */ +DECLHIDDEN(int) rtThreadNativeAdopt(PRTTHREADINT pThread) +{ + rtThreadPosixBlockSignals(pThread); + + int rc = pthread_setspecific(g_SelfKey, pThread); + if (!rc) + return VINF_SUCCESS; + return VERR_FAILED_TO_SET_SELF_TLS; +} + + +DECLHIDDEN(void) rtThreadNativeDestroy(PRTTHREADINT pThread) +{ + if (pThread == (PRTTHREADINT)pthread_getspecific(g_SelfKey)) + pthread_setspecific(g_SelfKey, NULL); +} + + +/** + * Wrapper which unpacks the params and calls thread function. + */ +static void *rtThreadNativeMain(void *pvArgs) +{ + PRTTHREADINT pThread = (PRTTHREADINT)pvArgs; + pthread_t Self = pthread_self(); +#if !defined(RT_OS_SOLARIS) /* On Solaris sizeof(pthread_t) = 4 and sizeof(NIL_RTNATIVETHREAD) = 8 */ + Assert((uintptr_t)Self != NIL_RTNATIVETHREAD); +#endif + Assert(Self == (pthread_t)(RTNATIVETHREAD)Self); + +#if defined(RT_OS_LINUX) + /* + * Set the TID. + */ + pThread->tid = syscall(__NR_gettid); + ASMMemoryFence(); +#endif + + rtThreadPosixBlockSignals(pThread); + + /* + * Set the TLS entry and, if possible, the thread name. + */ + int rc = pthread_setspecific(g_SelfKey, pThread); + AssertReleaseMsg(!rc, ("failed to set self TLS. rc=%d thread '%s'\n", rc, pThread->szName)); + +#ifdef IPRT_MAY_HAVE_PTHREAD_SET_NAME_NP + if (g_pfnThreadSetName) +# ifdef RT_OS_DARWIN + g_pfnThreadSetName(pThread->szName); +# else + g_pfnThreadSetName(Self, pThread->szName); +# endif +#endif + + /* + * Call common main. + */ + rc = rtThreadMain(pThread, (uintptr_t)Self, &pThread->szName[0]); + + pthread_setspecific(g_SelfKey, NULL); + pthread_exit((void *)(intptr_t)rc); + return (void *)(intptr_t)rc; +} + +#ifdef RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY + +/** + * @callback_method_impl{FNRTTHREAD, + * Priority proxy thread that services g_hRTThreadPosixPriorityProxyQueue.} + */ +static DECLCALLBACK(int) rtThreadPosixPriorityProxyThread(PRTTHREADINT, void *) +{ + for (;;) + { + RTREQQUEUE hReqQueue = g_hRTThreadPosixPriorityProxyQueue; + if (hReqQueue != NIL_RTREQQUEUE) + RTReqQueueProcess(hReqQueue, RT_INDEFINITE_WAIT); + else + break; + + int32_t rc = ASMAtomicUoReadS32(&g_rcPriorityProxyThreadStart); + if (rc != VINF_SUCCESS && rc != VERR_WRONG_ORDER) + break; + } + + return VINF_SUCCESS; +} + + +/** + * Just returns a non-success status codes to force the thread to re-evaluate + * the global shutdown variable. + */ +static DECLCALLBACK(int) rtThreadPosixPriorityProxyStopper(void) +{ + return VERR_CANCELLED; +} + + +/** + * An atexit() callback that stops the proxy creation/priority thread. + */ +static void rtThreadStopProxyThread(void) +{ + /* + * Signal to the thread that it's time to shut down. + */ + int32_t rc = ASMAtomicXchgS32(&g_rcPriorityProxyThreadStart, VERR_PROCESS_NOT_FOUND); + if (RT_SUCCESS(rc)) + { + /* + * Grab the associated handles. + */ + RTTHREAD hThread = g_hRTThreadPosixPriorityProxyThread; + RTREQQUEUE hQueue = g_hRTThreadPosixPriorityProxyQueue; + g_hRTThreadPosixPriorityProxyQueue = NIL_RTREQQUEUE; + g_hRTThreadPosixPriorityProxyThread = NIL_RTTHREAD; + ASMCompilerBarrier(); /* paranoia */ + + AssertReturnVoid(hThread != NIL_RTTHREAD); + AssertReturnVoid(hQueue != NIL_RTREQQUEUE); + + /* + * Kick the thread so it gets out of any pending RTReqQueueProcess call ASAP. + */ + rc = RTReqQueueCallEx(hQueue, NULL, 0 /*cMillies*/, RTREQFLAGS_IPRT_STATUS | RTREQFLAGS_NO_WAIT, + (PFNRT)rtThreadPosixPriorityProxyStopper, 0); + + /* + * Wait for the thread to complete. + */ + rc = RTThreadWait(hThread, RT_SUCCESS(rc) ? RT_MS_1SEC * 5 : 32, NULL); + if (RT_SUCCESS(rc)) + RTReqQueueDestroy(hQueue); + /* else: just leak the stuff, we're exitting, so nobody cares... */ + } +} + + +/** + * Ensure that the proxy priority proxy thread has been started. + * + * Since we will always start a proxy thread when asked to create a thread, + * there is no need for serialization here. + * + * @retval true if started + * @retval false if it failed to start (caller must handle this scenario). + */ +DECLHIDDEN(bool) rtThreadPosixPriorityProxyStart(void) +{ + /* + * Read the result. + */ + int rc = ASMAtomicUoReadS32(&g_rcPriorityProxyThreadStart); + if (rc != VERR_TRY_AGAIN) + return RT_SUCCESS(rc); + + /* If this triggers then there is a very unexpected race somewhere. It + should be harmless though. */ + AssertReturn(ASMAtomicCmpXchgS32(&g_rcPriorityProxyThreadStart, VERR_WRONG_ORDER, VERR_TRY_AGAIN), false); + + /* + * Not yet started, so do that. + */ + rc = RTReqQueueCreate(&g_hRTThreadPosixPriorityProxyQueue); + if (RT_SUCCESS(rc)) + { + rc = RTThreadCreate(&g_hRTThreadPosixPriorityProxyThread, rtThreadPosixPriorityProxyThread, NULL, 0 /*cbStack*/, + RTTHREADTYPE_DEFAULT, RTTHREADFLAGS_WAITABLE, "RTThrdPP"); + if (RT_SUCCESS(rc)) + { + ASMAtomicWriteS32(&g_rcPriorityProxyThreadStart, VINF_SUCCESS); + + atexit(rtThreadStopProxyThread); + return true; + } + RTReqQueueCreate(&g_hRTThreadPosixPriorityProxyQueue); + } + ASMAtomicWriteS32(&g_rcPriorityProxyThreadStart, rc != VERR_WRONG_ORDER ? rc : VERR_PROCESS_NOT_FOUND); + return false; +} + + +/** + * Calls @a pfnFunction from the priority proxy thread. + * + * Caller must have called rtThreadPosixStartProxy() to check that the priority + * proxy thread is running. + * + * @returns + * @param pTargetThread The target thread, NULL if not applicable. This is + * so we can skip calls pertaining to the priority + * proxy thread itself. + * @param pfnFunction The function to call. Must return IPRT status code. + * @param cArgs Number of arguments (see also RTReqQueueCall). + * @param ... Arguments (see also RTReqQueueCall). + */ +DECLHIDDEN(int) rtThreadPosixPriorityProxyCall(PRTTHREADINT pTargetThread, PFNRT pfnFunction, int cArgs, ...) +{ + int rc; + if ( !pTargetThread + || pTargetThread->pfnThread != rtThreadPosixPriorityProxyThread) + { + va_list va; + va_start(va, cArgs); + PRTREQ pReq; + rc = RTReqQueueCallV(g_hRTThreadPosixPriorityProxyQueue, &pReq, RT_INDEFINITE_WAIT, RTREQFLAGS_IPRT_STATUS, + pfnFunction, cArgs, va); + va_end(va); + RTReqRelease(pReq); + } + else + rc = VINF_SUCCESS; + return rc; +} + +#endif /* !RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY */ + +/** + * Worker for rtThreadNativeCreate that's either called on the priority proxy + * thread or directly on the calling thread depending on the proxy state. + */ +static DECLCALLBACK(int) rtThreadNativeInternalCreate(PRTTHREADINT pThread, PRTNATIVETHREAD pNativeThread) +{ + /* + * Set the default stack size. + */ + if (!pThread->cbStack) + pThread->cbStack = 512*1024; + +#ifdef RT_OS_LINUX + pThread->tid = -1; +#endif + + /* + * Setup thread attributes. + */ + pthread_attr_t ThreadAttr; + int rc = pthread_attr_init(&ThreadAttr); + if (!rc) + { + rc = pthread_attr_setdetachstate(&ThreadAttr, PTHREAD_CREATE_DETACHED); + if (!rc) + { + rc = pthread_attr_setstacksize(&ThreadAttr, pThread->cbStack); + if (!rc) + { + /* + * Create the thread. + */ + pthread_t ThreadId; + rc = pthread_create(&ThreadId, &ThreadAttr, rtThreadNativeMain, pThread); + if (!rc) + { + pthread_attr_destroy(&ThreadAttr); + *pNativeThread = (uintptr_t)ThreadId; + return VINF_SUCCESS; + } + } + } + pthread_attr_destroy(&ThreadAttr); + } + return RTErrConvertFromErrno(rc); +} + + +DECLHIDDEN(int) rtThreadNativeCreate(PRTTHREADINT pThread, PRTNATIVETHREAD pNativeThread) +{ +#ifdef RTTHREAD_POSIX_WITH_CREATE_PRIORITY_PROXY + /* + * If we have a priority proxy thread, use it. Make sure to ignore the + * staring of the proxy thread itself. + */ + if ( pThread->pfnThread != rtThreadPosixPriorityProxyThread + && rtThreadPosixPriorityProxyStart()) + { + PRTREQ pReq; + int rc = RTReqQueueCall(g_hRTThreadPosixPriorityProxyQueue, &pReq, RT_INDEFINITE_WAIT, + (PFNRT)rtThreadNativeInternalCreate, 2, pThread, pNativeThread); + RTReqRelease(pReq); + return rc; + } + + /* + * Fall back on creating it directly without regard to priority proxying. + */ +#endif + return rtThreadNativeInternalCreate(pThread, pNativeThread); +} + + +RTDECL(RTTHREAD) RTThreadSelf(void) +{ + /** @todo import alien threads? */ +#if defined(RT_OS_DARWIN) + /* On darwin, there seems to be input checking with pthread_getspecific. + So, we must prevent using g_SelfKey before rtThreadNativeInit has run, + otherwise we might crash or starting working with total garbage pointer + values here (see _os_tsd_get_direct in znu/libsyscall/os/tsd.h). + + Now, since the init value is a "negative" one, we just have to check + that it's positive or zero before calling the API. */ + if (RT_LIKELY((intptr_t)g_SelfKey >= 0)) + return (PRTTHREADINT)pthread_getspecific(g_SelfKey); + return NIL_RTTHREAD; +#else + return (PRTTHREADINT)pthread_getspecific(g_SelfKey); +#endif +} + + +#ifdef RTTHREAD_POSIX_WITH_POKE + +RTDECL(int) RTThreadPoke(RTTHREAD hThread) +{ + AssertReturn(hThread != RTThreadSelf(), VERR_INVALID_PARAMETER); + PRTTHREADINT pThread = rtThreadGet(hThread); + AssertReturn(pThread, VERR_INVALID_HANDLE); + + int rc; + if (g_iSigPokeThread != -1) + { + rc = pthread_kill((pthread_t)(uintptr_t)pThread->Core.Key, g_iSigPokeThread); + rc = RTErrConvertFromErrno(rc); + } + else + rc = VERR_NOT_SUPPORTED; + + rtThreadRelease(pThread); + return rc; +} + + +RTDECL(int) RTThreadControlPokeSignal(RTTHREAD hThread, bool fEnable) +{ + AssertReturn(hThread == RTThreadSelf() && hThread != NIL_RTTHREAD, VERR_INVALID_PARAMETER); + int rc; + if (g_iSigPokeThread != -1) + { + sigset_t SigSet; + sigemptyset(&SigSet); + sigaddset(&SigSet, g_iSigPokeThread); + + int rc2 = sigprocmask(fEnable ? SIG_UNBLOCK : SIG_BLOCK, &SigSet, NULL); + if (rc2 == 0) + rc = VINF_SUCCESS; + else + { + rc = RTErrConvertFromErrno(errno); + AssertMsgFailed(("rc=%Rrc errno=%d (rc2=%d)\n", rc, errno, rc2)); + } + } + else + rc = VERR_NOT_SUPPORTED; + return rc; +} + + +#endif + +/** @todo move this into platform specific files. */ +RTR3DECL(int) RTThreadGetExecutionTimeMilli(uint64_t *pKernelTime, uint64_t *pUserTime) +{ +#if defined(RT_OS_SOLARIS) + struct rusage ts; + int rc = getrusage(RUSAGE_LWP, &ts); + if (rc) + return RTErrConvertFromErrno(rc); + + *pKernelTime = ts.ru_stime.tv_sec * 1000 + ts.ru_stime.tv_usec / 1000; + *pUserTime = ts.ru_utime.tv_sec * 1000 + ts.ru_utime.tv_usec / 1000; + return VINF_SUCCESS; + +#elif defined(RT_OS_LINUX) || defined(RT_OS_FREEBSD) + /* on Linux, getrusage(RUSAGE_THREAD, ...) is available since 2.6.26 */ + struct timespec ts; + int rc = clock_gettime(CLOCK_THREAD_CPUTIME_ID, &ts); + if (rc) + return RTErrConvertFromErrno(rc); + + *pKernelTime = 0; + *pUserTime = (uint64_t)ts.tv_sec * 1000 + ts.tv_nsec / 1000000; + return VINF_SUCCESS; + +#elif defined(RT_OS_DARWIN) + thread_basic_info ThreadInfo; + mach_msg_type_number_t Count = THREAD_BASIC_INFO_COUNT; + kern_return_t krc = thread_info(mach_thread_self(), THREAD_BASIC_INFO, (thread_info_t)&ThreadInfo, &Count); + AssertReturn(krc == KERN_SUCCESS, RTErrConvertFromDarwinKern(krc)); + + *pKernelTime = ThreadInfo.system_time.seconds * 1000 + ThreadInfo.system_time.microseconds / 1000; + *pUserTime = ThreadInfo.user_time.seconds * 1000 + ThreadInfo.user_time.microseconds / 1000; + + return VINF_SUCCESS; +#elif defined(RT_OS_HAIKU) + thread_info ThreadInfo; + status_t status = get_thread_info(find_thread(NULL), &ThreadInfo); + AssertReturn(status == B_OK, RTErrConvertFromErrno(status)); + + *pKernelTime = ThreadInfo.kernel_time / 1000; + *pUserTime = ThreadInfo.user_time / 1000; + + return VINF_SUCCESS; +#else + return VERR_NOT_IMPLEMENTED; +#endif +} + diff --git a/src/VBox/Runtime/r3/posix/thread2-posix.cpp b/src/VBox/Runtime/r3/posix/thread2-posix.cpp new file mode 100644 index 00000000..6eaee22e --- /dev/null +++ b/src/VBox/Runtime/r3/posix/thread2-posix.cpp @@ -0,0 +1,133 @@ +/* $Id: thread2-posix.cpp $ */ +/** @file + * IPRT - Threads part 2, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_THREAD +#include <errno.h> +#include <pthread.h> +#include <unistd.h> +#include <sched.h> + +#include <iprt/thread.h> +#include <iprt/log.h> +#include <iprt/asm.h> +#if defined(RT_ARCH_AMD64) || defined(RT_ARCH_X86) +# include <iprt/asm-amd64-x86.h> +#endif +#include <iprt/errcore.h> +#include "internal/thread.h" + + +RTDECL(RTNATIVETHREAD) RTThreadNativeSelf(void) +{ + return (RTNATIVETHREAD)pthread_self(); +} + + +RTDECL(int) RTThreadSleep(RTMSINTERVAL cMillies) +{ + LogFlow(("RTThreadSleep: cMillies=%d\n", cMillies)); + if (!cMillies) + { + if (!sched_yield()) + { + LogFlow(("RTThreadSleep: returning %Rrc (cMillies=%d)\n", VINF_SUCCESS, cMillies)); + return VINF_SUCCESS; + } + } + else + { + struct timespec ts; + struct timespec tsrem = {0,0}; + + ts.tv_nsec = (cMillies % 1000) * 1000000; + ts.tv_sec = cMillies / 1000; + if (!nanosleep(&ts, &tsrem)) + { + LogFlow(("RTThreadSleep: returning %Rrc (cMillies=%d)\n", VINF_SUCCESS, cMillies)); + return VINF_SUCCESS; + } + } + + int rc = RTErrConvertFromErrno(errno); + LogFlow(("RTThreadSleep: returning %Rrc (cMillies=%d)\n", rc, cMillies)); + return rc; +} + + +RTDECL(int) RTThreadSleepNoLog(RTMSINTERVAL cMillies) +{ + if (!cMillies) + { + if (!sched_yield()) + return VINF_SUCCESS; + } + else + { + struct timespec ts; + struct timespec tsrem = {0,0}; + + ts.tv_nsec = (cMillies % 1000) * 1000000; + ts.tv_sec = cMillies / 1000; + if (!nanosleep(&ts, &tsrem)) + return VINF_SUCCESS; + } + + return RTErrConvertFromErrno(errno); +} + + +RTDECL(bool) RTThreadYield(void) +{ +#if defined(RT_ARCH_AMD64) || defined(RT_ARCH_X86) + uint64_t u64TS = ASMReadTSC(); +#endif + + sched_yield(); + +#if defined(RT_ARCH_AMD64) || defined(RT_ARCH_X86) + u64TS = ASMReadTSC() - u64TS; + bool fRc = u64TS > 1500; + LogFlow(("RTThreadYield: returning %d (%llu ticks)\n", fRc, u64TS)); +#else + bool fRc = true; /* PORTME: Add heuristics for determining whether the cpus was yielded. */ +#endif + return fRc; +} + diff --git a/src/VBox/Runtime/r3/posix/time-posix.cpp b/src/VBox/Runtime/r3/posix/time-posix.cpp new file mode 100644 index 00000000..3c028b73 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/time-posix.cpp @@ -0,0 +1,99 @@ +/* $Id: time-posix.cpp $ */ +/** @file + * IPRT - Time, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_TIME +#define RTTIME_INCL_TIMEVAL +#include <sys/time.h> +#include <time.h> + +#include <iprt/time.h> +#include "internal/time.h" + + +DECLINLINE(uint64_t) rtTimeGetSystemNanoTS(void) +{ +#if defined(CLOCK_MONOTONIC) && !defined(RT_OS_L4) && !defined(RT_OS_OS2) + /* check monotonic clock first. */ + static bool s_fMonoClock = true; + if (s_fMonoClock) + { + struct timespec ts; + if (!clock_gettime(CLOCK_MONOTONIC, &ts)) + return (uint64_t)ts.tv_sec * RT_NS_1SEC_64 + + ts.tv_nsec; + s_fMonoClock = false; + } +#endif + + /* fallback to gettimeofday(). */ + struct timeval tv; + gettimeofday(&tv, NULL); + return (uint64_t)tv.tv_sec * RT_NS_1SEC_64 + + (uint64_t)(tv.tv_usec * RT_NS_1US); +} + + +/** + * Gets the current nanosecond timestamp. + * + * This differs from RTTimeNanoTS in that it will use system APIs and not do any + * resolution or performance optimizations. + * + * @returns nanosecond timestamp. + */ +RTDECL(uint64_t) RTTimeSystemNanoTS(void) +{ + return rtTimeGetSystemNanoTS(); +} + + +/** + * Gets the current millisecond timestamp. + * + * This differs from RTTimeNanoTS in that it will use system APIs and not do any + * resolution or performance optimizations. + * + * @returns millisecond timestamp. + */ +RTDECL(uint64_t) RTTimeSystemMilliTS(void) +{ + return rtTimeGetSystemNanoTS() / RT_NS_1MS; +} + diff --git a/src/VBox/Runtime/r3/posix/timelocal-posix.cpp b/src/VBox/Runtime/r3/posix/timelocal-posix.cpp new file mode 100644 index 00000000..40e2fc16 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/timelocal-posix.cpp @@ -0,0 +1,215 @@ +/* $Id $ */ +/** @file + * IPRT - Local Time, Posix. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_TIME +#define RTTIME_INCL_TIMEVAL +#include <iprt/types.h> +#include <iprt/assert.h> + +#include <sys/time.h> +#include <time.h> + +#include <iprt/time.h> + + +/** + * This tries to find the UTC offset for a given timespec. + * + * It does probably not take into account changes in daylight + * saving over the years or similar stuff. + * + * @returns UTC offset in nanoseconds. + * @param pTime The time. + * @param fCurrentTime Whether the input is current time or not. + * This is for avoid infinit recursion on errors in the fallback path. + */ +static int64_t rtTimeLocalUTCOffset(PCRTTIMESPEC pTime, bool fCurrentTime) +{ + RTTIMESPEC Fallback; + + /* + * Convert to time_t. + */ + int64_t i64UnixTime = RTTimeSpecGetSeconds(pTime); + time_t UnixTime = i64UnixTime; + if (UnixTime != i64UnixTime) + return fCurrentTime ? 0 : rtTimeLocalUTCOffset(RTTimeNow(&Fallback), true); + + /* + * Explode it as both local and UTC time. + */ + struct tm TmLocal; + if ( !localtime_r(&UnixTime, &TmLocal) + || !TmLocal.tm_year) + return fCurrentTime ? 0 : rtTimeLocalUTCOffset(RTTimeNow(&Fallback), true); + struct tm TmUtc; + if (!gmtime_r(&UnixTime, &TmUtc)) + return fCurrentTime ? 0 : rtTimeLocalUTCOffset(RTTimeNow(&Fallback), true); + + /* + * Calc the difference (if any). + * We ASSUME that the difference is less that 24 hours. + */ + if ( TmLocal.tm_hour == TmUtc.tm_hour + && TmLocal.tm_min == TmUtc.tm_min + && TmLocal.tm_sec == TmUtc.tm_sec + && TmLocal.tm_mday == TmUtc.tm_mday) + return 0; + + int cLocalSecs = TmLocal.tm_hour * 3600 + + TmLocal.tm_min * 60 + + TmLocal.tm_sec; + int cUtcSecs = TmUtc.tm_hour * 3600 + + TmUtc.tm_min * 60 + + TmUtc.tm_sec; + if (TmLocal.tm_mday != TmUtc.tm_mday) + { + /* + * Must add 24 hours to the value that is ahead of the other. + * + * To determine which is ahead was busted for a long long time (bugref:9078), + * so here are some examples and two different approaches. + * + * TmLocal TmUtc => Add 24:00 to => Diff + * 2007-04-02 01:00 2007-04-01 23:00 => TmLocal => +02:00 + * 2007-04-01 01:00 2007-03-31 23:00 => TmLocal => +02:00 + * 2007-03-31 01:00 2007-03-30 23:00 => TmLocal => +02:00 + * + * 2007-04-01 01:00 2007-04-02 23:00 => TmUtc => -02:00 + * 2007-03-31 23:00 2007-04-01 01:00 => TmUtc => -02:00 + * 2007-03-30 23:00 2007-03-31 01:00 => TmUtc => -02:00 + * + */ +#if 0 + /* Using day of month turned out to be a little complicated. */ + if ( ( TmLocal.tm_mday > TmUtc.tm_mday + && (TmUtc.tm_mday != 1 || TmLocal.tm_mday < 28) ) + || (TmLocal.tm_mday == 1 && TmUtc.tm_mday >= 28) ) + { + cLocalSecs += 24*60*60; + Assert( TmLocal.tm_yday - TmUtc.tm_yday == 1 + || (TmLocal.tm_yday == 0 && TmUtc.tm_yday >= 364 && TmLocal.tm_year == TmUtc.tm_year + 1)); + } + else + { + cUtcSecs += 24*60*60; + Assert( TmUtc.tm_yday - TmLocal.tm_yday == 1 + || (TmUtc.tm_yday == 0 && TmLocal.tm_yday >= 364 && TmUtc.tm_year == TmLocal.tm_year + 1)); + } +#else + /* Using day of year and year is simpler. */ + if ( ( TmLocal.tm_year == TmUtc.tm_year + && TmLocal.tm_yday > TmUtc.tm_yday) + || TmLocal.tm_year > TmUtc.tm_year) + { + cLocalSecs += 24*60*60; + Assert( TmLocal.tm_yday - TmUtc.tm_yday == 1 + || (TmLocal.tm_yday == 0 && TmUtc.tm_yday >= 364 && TmLocal.tm_year == TmUtc.tm_year + 1)); + } + else + { + cUtcSecs += 24*60*60; + Assert( TmUtc.tm_yday - TmLocal.tm_yday == 1 + || (TmUtc.tm_yday == 0 && TmLocal.tm_yday >= 364 && TmUtc.tm_year == TmLocal.tm_year + 1)); + } +#endif + } + + return (cLocalSecs - cUtcSecs) * INT64_C(1000000000); +} + + +/** + * Gets the current delta between UTC and local time. + * + * @code + * RTTIMESPEC LocalTime; + * RTTimeSpecAddNano(RTTimeNow(&LocalTime), RTTimeLocalDeltaNano()); + * @endcode + * + * @returns Returns the nanosecond delta between UTC and local time. + */ +RTDECL(int64_t) RTTimeLocalDeltaNano(void) +{ + RTTIMESPEC Time; + return rtTimeLocalUTCOffset(RTTimeNow(&Time), true /* current time, skip fallback */); +} + + +/** + * Gets the delta between UTC and local time at the given time. + * + * @code + * RTTIMESPEC LocalTime; + * RTTimeNow(&LocalTime); + * RTTimeSpecAddNano(&LocalTime, RTTimeLocalDeltaNanoFor(&LocalTime)); + * @endcode + * + * @param pTimeSpec The time spec giving the time to get the delta for. + * @returns Returns the nanosecond delta between UTC and local time. + */ +RTDECL(int64_t) RTTimeLocalDeltaNanoFor(PCRTTIMESPEC pTimeSpec) +{ + AssertPtr(pTimeSpec); + return rtTimeLocalUTCOffset(pTimeSpec, false /* current time, skip fallback */); +} + + +/** + * Explodes a time spec to the localized timezone. + * + * @returns pTime. + * @param pTime Where to store the exploded time. + * @param pTimeSpec The time spec to exploded. (UTC) + */ +RTDECL(PRTTIME) RTTimeLocalExplode(PRTTIME pTime, PCRTTIMESPEC pTimeSpec) +{ + RTTIMESPEC LocalTime = *pTimeSpec; + int64_t cNsUtcOffset = rtTimeLocalUTCOffset(&LocalTime, true /* current time, skip fallback */); + RTTimeSpecAddNano(&LocalTime, cNsUtcOffset); + pTime = RTTimeExplode(pTime, &LocalTime); + if (pTime) + { + pTime->fFlags = (pTime->fFlags & ~RTTIME_FLAGS_TYPE_MASK) | RTTIME_FLAGS_TYPE_LOCAL; + pTime->offUTC = cNsUtcOffset / RT_NS_1MIN; + } + return pTime; +} + diff --git a/src/VBox/Runtime/r3/posix/timer-posix.cpp b/src/VBox/Runtime/r3/posix/timer-posix.cpp new file mode 100644 index 00000000..59c905b9 --- /dev/null +++ b/src/VBox/Runtime/r3/posix/timer-posix.cpp @@ -0,0 +1,847 @@ +/* $Id: timer-posix.cpp $ */ +/** @file + * IPRT - Timer, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +/** Enables the use of POSIX RT timers. */ +#ifndef RT_OS_SOLARIS /* Solaris 10 doesn't have SIGEV_THREAD */ +# define IPRT_WITH_POSIX_TIMERS +#endif /* !RT_OS_SOLARIS */ + +/** @def RT_TIMER_SIGNAL + * The signal number that the timers use. + * We currently use SIGALRM for both setitimer and posix real time timers + * out of simplicity, but we might want change this later for the posix ones. */ +#ifdef IPRT_WITH_POSIX_TIMERS +# define RT_TIMER_SIGNAL SIGALRM +#else +# define RT_TIMER_SIGNAL SIGALRM +#endif + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_TIMER +#include <iprt/timer.h> +#include <iprt/alloc.h> +#include <iprt/assert.h> +#include <iprt/thread.h> +#include <iprt/log.h> +#include <iprt/asm.h> +#include <iprt/semaphore.h> +#include <iprt/string.h> +#include <iprt/once.h> +#include <iprt/err.h> +#include <iprt/initterm.h> +#include <iprt/critsect.h> +#include "internal/magics.h" + +#include <unistd.h> +#include <sys/fcntl.h> +#include <sys/ioctl.h> +#ifdef RT_OS_LINUX +# include <linux/rtc.h> +#endif +#include <sys/time.h> +#include <signal.h> +#include <errno.h> +#include <pthread.h> +#if defined(RT_OS_DARWIN) +# define sigprocmask pthread_sigmask /* On xnu sigprocmask works on the process, not the calling thread as elsewhere. */ +#endif + + +/********************************************************************************************************************************* +* Global Variables * +*********************************************************************************************************************************/ +#ifdef IPRT_WITH_POSIX_TIMERS +/** Init the critsect on first call. */ +static RTONCE g_TimerOnce = RTONCE_INITIALIZER; +/** Global critsect that serializes timer creation and destruction. + * This is lazily created on the first RTTimerCreateEx call and will not be + * freed up (I'm afraid). */ +static RTCRITSECT g_TimerCritSect; +/** + * Global counter of RTTimer instances. The signal thread is + * started when it changes from 0 to 1. The signal thread + * terminates when it becomes 0 again. + */ +static uint32_t volatile g_cTimerInstances; +/** The signal handling thread. */ +static RTTHREAD g_TimerThread; +#endif /* IPRT_WITH_POSIX_TIMERS */ + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** + * The internal representation of a timer handle. + */ +typedef struct RTTIMER +{ + /** Magic. + * This is RTTIMER_MAGIC, but changes to something else before the timer + * is destroyed to indicate clearly that thread should exit. */ + uint32_t volatile u32Magic; + /** Flag indicating the timer is suspended. */ + uint8_t volatile fSuspended; + /** Flag indicating that the timer has been destroyed. */ + uint8_t volatile fDestroyed; +#ifndef IPRT_WITH_POSIX_TIMERS /** @todo We have to take the signals on a dedicated timer thread as + * we (might) have code assuming that signals doesn't screw around + * on existing threads. (It would be sufficient to have one thread + * per signal of course since the signal will be masked while it's + * running, however, it may just cause more complications than its + * worth - sigwait/sigwaitinfo work atomically anyway...) + * Also, must block the signal in the thread main procedure too. */ + /** The timer thread. */ + RTTHREAD Thread; + /** Event semaphore on which the thread is blocked. */ + RTSEMEVENT Event; +#endif /* !IPRT_WITH_POSIX_TIMERS */ + /** User argument. */ + void *pvUser; + /** Callback. */ + PFNRTTIMER pfnTimer; + /** The timer interval. 0 if one-shot. */ + uint64_t u64NanoInterval; +#ifndef IPRT_WITH_POSIX_TIMERS + /** The first shot interval. 0 if ASAP. */ + uint64_t volatile u64NanoFirst; +#endif /* !IPRT_WITH_POSIX_TIMERS */ + /** The current timer tick. */ + uint64_t volatile iTick; +#ifndef IPRT_WITH_POSIX_TIMERS + /** The error/status of the timer. + * Initially -1, set to 0 when the timer have been successfully started, and + * to errno on failure in starting the timer. */ + int volatile iError; +#else /* IPRT_WITH_POSIX_TIMERS */ + timer_t NativeTimer; +#endif /* IPRT_WITH_POSIX_TIMERS */ + +} RTTIMER; + + + +#ifdef IPRT_WITH_POSIX_TIMERS + +/** + * RTOnce callback that initializes the critical section. + * + * @returns RTCritSectInit return code. + * @param pvUser NULL, ignored. + * + */ +static DECLCALLBACK(int) rtTimerOnce(void *pvUser) +{ + NOREF(pvUser); + return RTCritSectInit(&g_TimerCritSect); +} +#endif + + +/** + * Signal handler which ignore everything it gets. + * + * @param iSignal The signal number. + */ +static void rttimerSignalIgnore(int iSignal) +{ + //AssertBreakpoint(); + NOREF(iSignal); +} + + +/** + * RT_TIMER_SIGNAL wait thread. + */ +static DECLCALLBACK(int) rttimerThread(RTTHREAD hThreadSelf, void *pvArg) +{ + NOREF(hThreadSelf); NOREF(pvArg); +#ifndef IPRT_WITH_POSIX_TIMERS + PRTTIMER pTimer = (PRTTIMER)pvArg; + RTTIMER Timer = *pTimer; + Assert(pTimer->u32Magic == RTTIMER_MAGIC); +#endif /* !IPRT_WITH_POSIX_TIMERS */ + + /* + * Install signal handler. + */ + struct sigaction SigAct; + memset(&SigAct, 0, sizeof(SigAct)); + SigAct.sa_flags = SA_RESTART; + sigemptyset(&SigAct.sa_mask); + SigAct.sa_handler = rttimerSignalIgnore; + if (sigaction(RT_TIMER_SIGNAL, &SigAct, NULL)) + { + SigAct.sa_flags &= ~SA_RESTART; + if (sigaction(RT_TIMER_SIGNAL, &SigAct, NULL)) + AssertMsgFailed(("sigaction failed, errno=%d\n", errno)); + } + + /* + * Mask most signals except those which might be used by the pthread implementation (linux). + */ + sigset_t SigSet; + sigfillset(&SigSet); + sigdelset(&SigSet, SIGTERM); + sigdelset(&SigSet, SIGHUP); + sigdelset(&SigSet, SIGINT); + sigdelset(&SigSet, SIGABRT); + sigdelset(&SigSet, SIGKILL); +#ifdef SIGRTMIN + for (int iSig = SIGRTMIN; iSig < SIGRTMAX; iSig++) + sigdelset(&SigSet, iSig); +#endif + if (sigprocmask(SIG_SETMASK, &SigSet, NULL)) + { +#ifdef IPRT_WITH_POSIX_TIMERS + int rc = RTErrConvertFromErrno(errno); +#else + int rc = pTimer->iError = RTErrConvertFromErrno(errno); +#endif + AssertMsgFailed(("sigprocmask -> errno=%d\n", errno)); + return rc; + } + + /* + * The work loop. + */ + RTThreadUserSignal(hThreadSelf); + +#ifndef IPRT_WITH_POSIX_TIMERS + while ( !pTimer->fDestroyed + && pTimer->u32Magic == RTTIMER_MAGIC) + { + /* + * Wait for a start or destroy event. + */ + if (pTimer->fSuspended) + { + int rc = RTSemEventWait(pTimer->Event, RT_INDEFINITE_WAIT); + if (RT_FAILURE(rc) && rc != VERR_INTERRUPTED) + { + AssertRC(rc); + if (pTimer->fDestroyed) + continue; + RTThreadSleep(1000); /* Don't cause trouble! */ + } + if ( pTimer->fSuspended + || pTimer->fDestroyed) + continue; + } + + /* + * Start the timer. + * + * For some SunOS (/SysV?) threading compatibility Linux will only + * deliver the RT_TIMER_SIGNAL to the thread calling setitimer(). Therefore + * we have to call it here. + * + * It turns out this might not always be the case, see RT_TIMER_SIGNAL killing + * processes on RH 2.4.21. + */ + struct itimerval TimerVal; + if (pTimer->u64NanoFirst) + { + uint64_t u64 = RT_MAX(1000, pTimer->u64NanoFirst); + TimerVal.it_value.tv_sec = u64 / 1000000000; + TimerVal.it_value.tv_usec = (u64 % 1000000000) / 1000; + } + else + { + TimerVal.it_value.tv_sec = 0; + TimerVal.it_value.tv_usec = 10; + } + if (pTimer->u64NanoInterval) + { + uint64_t u64 = RT_MAX(1000, pTimer->u64NanoInterval); + TimerVal.it_interval.tv_sec = u64 / 1000000000; + TimerVal.it_interval.tv_usec = (u64 % 1000000000) / 1000; + } + else + { + TimerVal.it_interval.tv_sec = 0; + TimerVal.it_interval.tv_usec = 0; + } + + if (setitimer(ITIMER_REAL, &TimerVal, NULL)) + { + ASMAtomicXchgU8(&pTimer->fSuspended, true); + pTimer->iError = RTErrConvertFromErrno(errno); + RTThreadUserSignal(hThreadSelf); + continue; /* back to suspended mode. */ + } + pTimer->iError = 0; + RTThreadUserSignal(hThreadSelf); + + /* + * Timer Service Loop. + */ + sigemptyset(&SigSet); + sigaddset(&SigSet, RT_TIMER_SIGNAL); + do + { + siginfo_t SigInfo; + RT_ZERO(SigInfo); +#ifdef RT_OS_DARWIN + if (RT_LIKELY(sigwait(&SigSet, &SigInfo.si_signo) >= 0)) + { +#else + if (RT_LIKELY(sigwaitinfo(&SigSet, &SigInfo) >= 0)) + { + if (RT_LIKELY(SigInfo.si_signo == RT_TIMER_SIGNAL)) +#endif + { + if (RT_UNLIKELY( pTimer->fSuspended + || pTimer->fDestroyed + || pTimer->u32Magic != RTTIMER_MAGIC)) + break; + + pTimer->pfnTimer(pTimer, pTimer->pvUser, ++pTimer->iTick); + + /* auto suspend one-shot timers. */ + if (RT_UNLIKELY(!pTimer->u64NanoInterval)) + { + ASMAtomicWriteU8(&pTimer->fSuspended, true); + break; + } + } + } + else if (errno != EINTR) + AssertMsgFailed(("sigwaitinfo -> errno=%d\n", errno)); + } while (RT_LIKELY( !pTimer->fSuspended + && !pTimer->fDestroyed + && pTimer->u32Magic == RTTIMER_MAGIC)); + + /* + * Disable the timer. + */ + struct itimerval TimerVal2 = {{0,0}, {0,0}}; + if (setitimer(ITIMER_REAL, &TimerVal2, NULL)) + AssertMsgFailed(("setitimer(ITIMER_REAL,&{0}, NULL) failed, errno=%d\n", errno)); + + /* + * ACK any pending suspend request. + */ + if (!pTimer->fDestroyed) + { + pTimer->iError = 0; + RTThreadUserSignal(hThreadSelf); + } + } + + /* + * Exit. + */ + pTimer->iError = 0; + RTThreadUserSignal(hThreadSelf); + +#else /* IPRT_WITH_POSIX_TIMERS */ + + sigemptyset(&SigSet); + sigaddset(&SigSet, RT_TIMER_SIGNAL); + while (g_cTimerInstances) + { + siginfo_t SigInfo; + RT_ZERO(SigInfo); + if (RT_LIKELY(sigwaitinfo(&SigSet, &SigInfo) >= 0)) + { + LogFlow(("rttimerThread: signo=%d pTimer=%p\n", SigInfo.si_signo, SigInfo.si_value.sival_ptr)); + if (RT_LIKELY( SigInfo.si_signo == RT_TIMER_SIGNAL + && SigInfo.si_code == SI_TIMER)) /* The SI_TIMER check is *essential* because of the pthread_kill. */ + { + PRTTIMER pTimer = (PRTTIMER)SigInfo.si_value.sival_ptr; + AssertPtr(pTimer); + if (RT_UNLIKELY( !RT_VALID_PTR(pTimer) + || ASMAtomicUoReadU8(&pTimer->fSuspended) + || ASMAtomicUoReadU8(&pTimer->fDestroyed) + || pTimer->u32Magic != RTTIMER_MAGIC)) + continue; + + pTimer->pfnTimer(pTimer, pTimer->pvUser, ++pTimer->iTick); + + /* auto suspend one-shot timers. */ + if (RT_UNLIKELY(!pTimer->u64NanoInterval)) + ASMAtomicWriteU8(&pTimer->fSuspended, true); + } + } + } +#endif /* IPRT_WITH_POSIX_TIMERS */ + + return VINF_SUCCESS; +} + + +RTDECL(int) RTTimerCreateEx(PRTTIMER *ppTimer, uint64_t u64NanoInterval, uint32_t fFlags, PFNRTTIMER pfnTimer, void *pvUser) +{ + /* + * We don't support the fancy MP features. + */ + if (fFlags & RTTIMER_FLAGS_CPU_SPECIFIC) + return VERR_NOT_SUPPORTED; + + /* + * We need the signal masks to be set correctly, which they won't be in + * unobtrusive mode. + */ + if (RTR3InitIsUnobtrusive()) + return VERR_NOT_SUPPORTED; + +#ifndef IPRT_WITH_POSIX_TIMERS + /* + * Check if timer is busy. + */ + struct itimerval TimerVal; + if (getitimer(ITIMER_REAL, &TimerVal)) + { + AssertMsgFailed(("getitimer() -> errno=%d\n", errno)); + return VERR_NOT_IMPLEMENTED; + } + if ( TimerVal.it_value.tv_usec + || TimerVal.it_value.tv_sec + || TimerVal.it_interval.tv_usec + || TimerVal.it_interval.tv_sec) + { + AssertMsgFailed(("A timer is running. System limit is one timer per process!\n")); + return VERR_TIMER_BUSY; + } +#endif /* !IPRT_WITH_POSIX_TIMERS */ + + /* + * Block RT_TIMER_SIGNAL from calling thread. + */ + sigset_t SigSet; + sigemptyset(&SigSet); + sigaddset(&SigSet, RT_TIMER_SIGNAL); + sigprocmask(SIG_BLOCK, &SigSet, NULL); + +#ifndef IPRT_WITH_POSIX_TIMERS /** @todo combine more of the setitimer/timer_create code. setitimer could also use the global thread. */ + /** @todo Move this RTC hack else where... */ + static bool fDoneRTC; + if (!fDoneRTC) + { + fDoneRTC = true; + /* check resolution. */ + TimerVal.it_interval.tv_sec = 0; + TimerVal.it_interval.tv_usec = 1000; + TimerVal.it_value = TimerVal.it_interval; + if ( setitimer(ITIMER_REAL, &TimerVal, NULL) + || getitimer(ITIMER_REAL, &TimerVal) + || TimerVal.it_interval.tv_usec > 1000) + { + /* + * Try open /dev/rtc to set the irq rate to 1024 and + * turn periodic + */ + Log(("RTTimerCreate: interval={%ld,%ld} trying to adjust /dev/rtc!\n", TimerVal.it_interval.tv_sec, TimerVal.it_interval.tv_usec)); +# ifdef RT_OS_LINUX + int fh = open("/dev/rtc", O_RDONLY); + if (fh >= 0) + { + if ( ioctl(fh, RTC_IRQP_SET, 1024) < 0 + || ioctl(fh, RTC_PIE_ON, 0) < 0) + Log(("RTTimerCreate: couldn't configure rtc! errno=%d\n", errno)); + ioctl(fh, F_SETFL, O_ASYNC); + ioctl(fh, F_SETOWN, getpid()); + /* not so sure if closing it is a good idea... */ + //close(fh); + } + else + Log(("RTTimerCreate: couldn't configure rtc! open failed with errno=%d\n", errno)); +# endif + } + /* disable it */ + TimerVal.it_interval.tv_sec = 0; + TimerVal.it_interval.tv_usec = 0; + TimerVal.it_value = TimerVal.it_interval; + setitimer(ITIMER_REAL, &TimerVal, NULL); + } + + /* + * Create a new timer. + */ + int rc; + PRTTIMER pTimer = (PRTTIMER)RTMemAlloc(sizeof(*pTimer)); + if (pTimer) + { + pTimer->u32Magic = RTTIMER_MAGIC; + pTimer->fSuspended = true; + pTimer->fDestroyed = false; + pTimer->Thread = NIL_RTTHREAD; + pTimer->Event = NIL_RTSEMEVENT; + pTimer->pfnTimer = pfnTimer; + pTimer->pvUser = pvUser; + pTimer->u64NanoInterval = u64NanoInterval; + pTimer->u64NanoFirst = 0; + pTimer->iTick = 0; + pTimer->iError = 0; + rc = RTSemEventCreate(&pTimer->Event); + AssertRC(rc); + if (RT_SUCCESS(rc)) + { + rc = RTThreadCreate(&pTimer->Thread, rttimerThread, pTimer, 0, RTTHREADTYPE_TIMER, RTTHREADFLAGS_WAITABLE, "Timer"); + AssertRC(rc); + if (RT_SUCCESS(rc)) + { + /* + * Wait for the timer thread to initialize it self. + * This might take a little while... + */ + rc = RTThreadUserWait(pTimer->Thread, 45*1000); + AssertRC(rc); + if (RT_SUCCESS(rc)) + { + rc = RTThreadUserReset(pTimer->Thread); AssertRC(rc); + rc = pTimer->iError; + AssertRC(rc); + if (RT_SUCCESS(rc)) + { + RTThreadYield(); /* <-- Horrible hack to make tstTimer work. (linux 2.6.12) */ + *ppTimer = pTimer; + return VINF_SUCCESS; + } + } + + /* bail out */ + ASMAtomicXchgU8(&pTimer->fDestroyed, true); + ASMAtomicXchgU32(&pTimer->u32Magic, ~RTTIMER_MAGIC); + RTThreadWait(pTimer->Thread, 45*1000, NULL); + } + RTSemEventDestroy(pTimer->Event); + pTimer->Event = NIL_RTSEMEVENT; + } + RTMemFree(pTimer); + } + else + rc = VERR_NO_MEMORY; + +#else /* IPRT_WITH_POSIX_TIMERS */ + + /* + * Do the global init first. + */ + int rc = RTOnce(&g_TimerOnce, rtTimerOnce, NULL); + if (RT_FAILURE(rc)) + return rc; + + /* + * Create a new timer structure. + */ + LogFlow(("RTTimerCreateEx: u64NanoInterval=%llu fFlags=%lu\n", u64NanoInterval, fFlags)); + PRTTIMER pTimer = (PRTTIMER)RTMemAlloc(sizeof(*pTimer)); + if (pTimer) + { + /* Initialize timer structure. */ + pTimer->u32Magic = RTTIMER_MAGIC; + pTimer->fSuspended = true; + pTimer->fDestroyed = false; + pTimer->pfnTimer = pfnTimer; + pTimer->pvUser = pvUser; + pTimer->u64NanoInterval = u64NanoInterval; + pTimer->iTick = 0; + + /* + * Create a timer that deliver RT_TIMER_SIGNAL upon timer expiration. + */ + struct sigevent SigEvt; + SigEvt.sigev_notify = SIGEV_SIGNAL; + SigEvt.sigev_signo = RT_TIMER_SIGNAL; + SigEvt.sigev_value.sival_ptr = pTimer; /* sigev_value gets copied to siginfo. */ + int err = timer_create(CLOCK_REALTIME, &SigEvt, &pTimer->NativeTimer); + if (!err) + { + /* + * Increment the timer count, do this behind the critsect to avoid races. + */ + RTCritSectEnter(&g_TimerCritSect); + + if (ASMAtomicIncU32(&g_cTimerInstances) != 1) + { + Assert(g_cTimerInstances > 1); + RTCritSectLeave(&g_TimerCritSect); + + LogFlow(("RTTimerCreateEx: rc=%Rrc pTimer=%p (thread already running)\n", rc, pTimer)); + *ppTimer = pTimer; + return VINF_SUCCESS; + } + + /* + * Create the signal handling thread. It will wait for the signal + * and execute the timer functions. + */ + rc = RTThreadCreate(&g_TimerThread, rttimerThread, NULL, 0, RTTHREADTYPE_TIMER, RTTHREADFLAGS_WAITABLE, "Timer"); + if (RT_SUCCESS(rc)) + { + rc = RTThreadUserWait(g_TimerThread, 45*1000); /* this better not fail... */ + if (RT_SUCCESS(rc)) + { + RTCritSectLeave(&g_TimerCritSect); + + LogFlow(("RTTimerCreateEx: rc=%Rrc pTimer=%p (thread already running)\n", rc, pTimer)); + *ppTimer = pTimer; + return VINF_SUCCESS; + } + /* darn, what do we do here? */ + } + + /* bail out */ + ASMAtomicDecU32(&g_cTimerInstances); + Assert(!g_cTimerInstances); + + RTCritSectLeave(&g_TimerCritSect); + + timer_delete(pTimer->NativeTimer); + } + else + { + rc = RTErrConvertFromErrno(err); + Log(("RTTimerCreateEx: err=%d (%Rrc)\n", err, rc)); + } + + RTMemFree(pTimer); + } + else + rc = VERR_NO_MEMORY; + +#endif /* IPRT_WITH_POSIX_TIMERS */ + return rc; +} + + +RTR3DECL(int) RTTimerDestroy(PRTTIMER pTimer) +{ + LogFlow(("RTTimerDestroy: pTimer=%p\n", pTimer)); + + /* + * Validate input. + */ + /* NULL is ok. */ + if (!pTimer) + return VINF_SUCCESS; + int rc = VINF_SUCCESS; + AssertPtrReturn(pTimer, VERR_INVALID_POINTER); + AssertReturn(pTimer->u32Magic == RTTIMER_MAGIC, VERR_INVALID_MAGIC); +#ifdef IPRT_WITH_POSIX_TIMERS + AssertReturn(g_TimerThread != RTThreadSelf(), VERR_INTERNAL_ERROR); +#else + AssertReturn(pTimer->Thread != RTThreadSelf(), VERR_INTERNAL_ERROR); +#endif + + /* + * Mark the semaphore as destroyed. + */ + ASMAtomicWriteU8(&pTimer->fDestroyed, true); + ASMAtomicWriteU32(&pTimer->u32Magic, ~RTTIMER_MAGIC); + +#ifdef IPRT_WITH_POSIX_TIMERS + /* + * Suspend the timer if it's running. + */ + if (!pTimer->fSuspended) + { + struct itimerspec TimerSpec; + TimerSpec.it_value.tv_sec = 0; + TimerSpec.it_value.tv_nsec = 0; + TimerSpec.it_interval.tv_sec = 0; + TimerSpec.it_interval.tv_nsec = 0; + int err = timer_settime(pTimer->NativeTimer, 0, &TimerSpec, NULL); NOREF(err); + AssertMsg(!err, ("%d / %d\n", err, errno)); + } +#endif + + /* + * Poke the thread and wait for it to finish. + * This is only done for the last timer when using posix timers. + */ +#ifdef IPRT_WITH_POSIX_TIMERS + RTTHREAD Thread = NIL_RTTHREAD; + RTCritSectEnter(&g_TimerCritSect); + if (ASMAtomicDecU32(&g_cTimerInstances) == 0) + { + Thread = g_TimerThread; + g_TimerThread = NIL_RTTHREAD; + } + RTCritSectLeave(&g_TimerCritSect); +#else /* IPRT_WITH_POSIX_TIMERS */ + RTTHREAD Thread = pTimer->Thread; + rc = RTSemEventSignal(pTimer->Event); + AssertRC(rc); +#endif /* IPRT_WITH_POSIX_TIMERS */ + if (Thread != NIL_RTTHREAD) + { + /* Signal it so it gets out of the sigwait if it's stuck there... */ + pthread_kill((pthread_t)RTThreadGetNative(Thread), RT_TIMER_SIGNAL); + + /* + * Wait for the thread to complete. + */ + rc = RTThreadWait(Thread, 30 * 1000, NULL); + AssertRC(rc); + } + + + /* + * Free up the resources associated with the timer. + */ +#ifdef IPRT_WITH_POSIX_TIMERS + timer_delete(pTimer->NativeTimer); +#else + RTSemEventDestroy(pTimer->Event); + pTimer->Event = NIL_RTSEMEVENT; +#endif /* !IPRT_WITH_POSIX_TIMERS */ + if (RT_SUCCESS(rc)) + RTMemFree(pTimer); + return rc; +} + + +RTDECL(int) RTTimerStart(PRTTIMER pTimer, uint64_t u64First) +{ + /* + * Validate input. + */ + AssertPtrReturn(pTimer, VERR_INVALID_POINTER); + AssertReturn(pTimer->u32Magic == RTTIMER_MAGIC, VERR_INVALID_MAGIC); +#ifndef IPRT_WITH_POSIX_TIMERS + AssertReturn(pTimer->Thread != RTThreadSelf(), VERR_INTERNAL_ERROR); +#endif + + /* + * Already running? + */ + if (!ASMAtomicXchgU8(&pTimer->fSuspended, false)) + return VERR_TIMER_ACTIVE; + LogFlow(("RTTimerStart: pTimer=%p u64First=%llu u64NanoInterval=%llu\n", pTimer, u64First, pTimer->u64NanoInterval)); + +#ifndef IPRT_WITH_POSIX_TIMERS + /* + * Tell the thread to start servicing the timer. + * Wait for it to ACK the request to avoid reset races. + */ + RTThreadUserReset(pTimer->Thread); + ASMAtomicUoWriteU64(&pTimer->u64NanoFirst, u64First); + ASMAtomicUoWriteU64(&pTimer->iTick, 0); + ASMAtomicWriteU8(&pTimer->fSuspended, false); + int rc = RTSemEventSignal(pTimer->Event); + if (RT_SUCCESS(rc)) + { + rc = RTThreadUserWait(pTimer->Thread, 45*1000); + AssertRC(rc); + RTThreadUserReset(pTimer->Thread); + } + else + AssertRC(rc); + +#else /* IPRT_WITH_POSIX_TIMERS */ + /* + * Start the timer. + */ + struct itimerspec TimerSpec; + TimerSpec.it_value.tv_sec = u64First / 1000000000; /* nanosec => sec */ + TimerSpec.it_value.tv_nsec = u64First ? u64First % 1000000000 : 10; /* 0 means disable, replace it with 10. */ + TimerSpec.it_interval.tv_sec = pTimer->u64NanoInterval / 1000000000; + TimerSpec.it_interval.tv_nsec = pTimer->u64NanoInterval % 1000000000; + int err = timer_settime(pTimer->NativeTimer, 0, &TimerSpec, NULL); + int rc = err == 0 ? VINF_SUCCESS : RTErrConvertFromErrno(errno); +#endif /* IPRT_WITH_POSIX_TIMERS */ + + if (RT_FAILURE(rc)) + ASMAtomicXchgU8(&pTimer->fSuspended, false); + return rc; +} + + +RTDECL(int) RTTimerStop(PRTTIMER pTimer) +{ + /* + * Validate input. + */ + AssertPtrReturn(pTimer, VERR_INVALID_POINTER); + AssertReturn(pTimer->u32Magic == RTTIMER_MAGIC, VERR_INVALID_MAGIC); + + /* + * Already running? + */ + if (ASMAtomicXchgU8(&pTimer->fSuspended, true)) + return VERR_TIMER_SUSPENDED; + LogFlow(("RTTimerStop: pTimer=%p\n", pTimer)); + +#ifndef IPRT_WITH_POSIX_TIMERS + /* + * Tell the thread to stop servicing the timer. + */ + RTThreadUserReset(pTimer->Thread); + ASMAtomicXchgU8(&pTimer->fSuspended, true); + int rc = VINF_SUCCESS; + if (RTThreadSelf() != pTimer->Thread) + { + pthread_kill((pthread_t)RTThreadGetNative(pTimer->Thread), RT_TIMER_SIGNAL); + rc = RTThreadUserWait(pTimer->Thread, 45*1000); + AssertRC(rc); + RTThreadUserReset(pTimer->Thread); + } + +#else /* IPRT_WITH_POSIX_TIMERS */ + /* + * Stop the timer. + */ + struct itimerspec TimerSpec; + TimerSpec.it_value.tv_sec = 0; + TimerSpec.it_value.tv_nsec = 0; + TimerSpec.it_interval.tv_sec = 0; + TimerSpec.it_interval.tv_nsec = 0; + int err = timer_settime(pTimer->NativeTimer, 0, &TimerSpec, NULL); + int rc = err == 0 ? VINF_SUCCESS : RTErrConvertFromErrno(errno); +#endif /* IPRT_WITH_POSIX_TIMERS */ + + return rc; +} + + +RTDECL(int) RTTimerChangeInterval(PRTTIMER pTimer, uint64_t u64NanoInterval) +{ + AssertPtrReturn(pTimer, VERR_INVALID_POINTER); + AssertReturn(pTimer->u32Magic == RTTIMER_MAGIC, VERR_INVALID_MAGIC); + NOREF(u64NanoInterval); + return VERR_NOT_SUPPORTED; +} + diff --git a/src/VBox/Runtime/r3/posix/tls-posix.cpp b/src/VBox/Runtime/r3/posix/tls-posix.cpp new file mode 100644 index 00000000..22ef28ab --- /dev/null +++ b/src/VBox/Runtime/r3/posix/tls-posix.cpp @@ -0,0 +1,119 @@ +/* $Id: tls-posix.cpp $ */ +/** @file + * IPRT - Thread Local Storage (TLS), POSIX. + */ + +/* + * Copyright (C) 2008-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP RTLOGGROUP_THREAD +#include <errno.h> +#include <pthread.h> + +#include <iprt/thread.h> +#include <iprt/log.h> +#include <iprt/assert.h> +#include <iprt/errcore.h> + + +AssertCompile(sizeof(pthread_key_t) <= sizeof(RTTLS)); + + +RTR3DECL(RTTLS) RTTlsAlloc(void) +{ + pthread_key_t iTls = (pthread_key_t)NIL_RTTLS; + int rc = pthread_key_create(&iTls, NULL); + if (!rc) + { + Assert(iTls != (pthread_key_t)NIL_RTTLS); + return iTls; + } + return NIL_RTTLS; +} + + +RTR3DECL(int) RTTlsAllocEx(PRTTLS piTls, PFNRTTLSDTOR pfnDestructor) +{ + pthread_key_t iTls = (pthread_key_t)NIL_RTTLS; +#if defined(__GNUC__) && defined(RT_ARCH_X86) + int rc = pthread_key_create(&iTls, (void (*)(void*))pfnDestructor); +#else + int rc = pthread_key_create(&iTls, pfnDestructor); +#endif + if (!rc) + { + *piTls = iTls; + Assert((pthread_key_t)*piTls == iTls); + Assert(*piTls != NIL_RTTLS); + return VINF_SUCCESS; + } + return RTErrConvertFromErrno(rc); +} + + +RTR3DECL(int) RTTlsFree(RTTLS iTls) +{ + if (iTls == NIL_RTTLS) + return VINF_SUCCESS; + int rc = pthread_key_delete(iTls); + if (!rc) + return VINF_SUCCESS; + return RTErrConvertFromErrno(rc); +} + + +RTR3DECL(void *) RTTlsGet(RTTLS iTls) +{ + return pthread_getspecific(iTls); +} + + +RTR3DECL(int) RTTlsGetEx(RTTLS iTls, void **ppvValue) +{ + if (RT_UNLIKELY(iTls == NIL_RTTLS)) + return VERR_INVALID_PARAMETER; + *ppvValue = pthread_getspecific(iTls); + return VINF_SUCCESS; +} + + +RTR3DECL(int) RTTlsSet(RTTLS iTls, void *pvValue) +{ + int rc = pthread_setspecific(iTls, pvValue); + if (RT_UNLIKELY(rc != 0)) + return RTErrConvertFromErrno(rc); + return VINF_SUCCESS; +} + diff --git a/src/VBox/Runtime/r3/posix/utf8-posix.cpp b/src/VBox/Runtime/r3/posix/utf8-posix.cpp new file mode 100644 index 00000000..5f4c947e --- /dev/null +++ b/src/VBox/Runtime/r3/posix/utf8-posix.cpp @@ -0,0 +1,709 @@ +/* $Id: utf8-posix.cpp $ */ +/** @file + * IPRT - UTF-8 helpers, POSIX. + */ + +/* + * Copyright (C) 2006-2023 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#include <iprt/string.h> +#include "internal/iprt.h" + +#include <iprt/alloc.h> +#include <iprt/assert.h> +#include <iprt/ctype.h> +#include <iprt/err.h> +#include <iprt/string.h> + +#include <errno.h> +#include <locale.h> +#ifdef RT_OS_DARWIN +# include <stdlib.h> +#endif + +/* iconv prototype changed with 165+ (thanks to PSARC/2010/160 Bugster 7037400) */ +#if defined(RT_OS_SOLARIS) +# if !defined(_XPG6) +# define IPRT_XPG6_TMP_DEF +# define _XPG6 +# endif +# if defined(__USE_LEGACY_PROTOTYPES__) +# define IPRT_LEGACY_PROTO_TMP_DEF +# undef __USE_LEGACY_PROTOTYPES__ +# endif +#endif /* RT_OS_SOLARIS */ + +# include <iconv.h> + +#if defined(RT_OS_SOLARIS) +# if defined(IPRT_XPG6_TMP_DEF) +# undef _XPG6 +# undef IPRT_XPG6_TMP_DEF +# endif +# if defined(IPRT_LEGACY_PROTO_TMP_DEF) +# define __USE_LEGACY_PROTOTYPES__ +# undef IPRT_LEGACY_PROTO_TMP_DEF +# endif +#endif /* RT_OS_SOLARIS */ + +#include <wctype.h> + +#include <langinfo.h> + +#include "internal/alignmentchecks.h" +#include "internal/string.h" +#ifdef RT_WITH_ICONV_CACHE +# include "internal/thread.h" +AssertCompile(sizeof(iconv_t) <= sizeof(void *)); +#endif + + +/* There are different opinions about the constness of the input buffer. */ +#if defined(RT_OS_LINUX) || defined(RT_OS_HAIKU) || defined(RT_OS_SOLARIS) \ + || (defined(RT_OS_DARWIN) && defined(_DARWIN_FEATURE_UNIX_CONFORMANCE)) +# define NON_CONST_ICONV_INPUT +#endif +#ifdef RT_OS_FREEBSD +# include <sys/param.h> +# if __FreeBSD_version >= 1002000 /* Changed around 10.2.2 (https://svnweb.freebsd.org/base?view=revision&revision=281550) */ +# define NON_CONST_ICONV_INPUT +# else +# error __FreeBSD_version__ +# endif +#endif +#ifdef RT_OS_NETBSD +/* iconv constness was changed on 2019-10-24, shortly after 9.99.17 */ +# include <sys/param.h> +# if __NetBSD_Prereq__(9,99,18) +# define NON_CONST_ICONV_INPUT +# endif +#endif + + +/** + * Gets the codeset of the current locale (LC_CTYPE). + * + * @returns Pointer to read-only string with the codeset name. + */ +DECLHIDDEN(const char *) rtStrGetLocaleCodeset(void) +{ +#ifdef RT_OS_DARWIN + /* + * @bugref{10153}: If no locale specified in the environment (typically the + * case when launched via Finder, LaunchPad or similar) default to UTF-8. + */ + static int8_t volatile s_fIsUtf8 = -1; + int8_t fIsUtf8 = s_fIsUtf8; + if (fIsUtf8) + { + if (fIsUtf8 == true) + return "UTF-8"; + + /* Initialize: */ + fIsUtf8 = true; + static const char * const s_papszVariables[] = { "LC_ALL", "LC_CTYPE", "LANG" }; + for (size_t i = 0; i < RT_ELEMENTS(s_papszVariables); i++) + { + const char *pszValue = getenv(s_papszVariables[i]); + if (pszValue && *pszValue) + { + fIsUtf8 = false; + break; + } + } + s_fIsUtf8 = fIsUtf8; + if (fIsUtf8 == true) + return "UTF-8"; + } +#endif + return nl_langinfo(CODESET); +} + + +/** + * Checks if the codeset specified by current locale (LC_CTYPE) is UTF-8. + * + * @returns true if UTF-8, false if not. + */ +DECLHIDDEN(bool) rtStrIsLocaleCodesetUtf8(void) +{ + return rtStrIsCodesetUtf8(rtStrGetLocaleCodeset()); +} + + +/** + * Checks if @a pszCodeset specified UTF-8. + * + * @returns true if UTF-8, false if not. + * @param pszCodeset Codeset to test. + */ +DECLHIDDEN(bool) rtStrIsCodesetUtf8(const char *pszCodeset) +{ + if (pszCodeset) + { + /* Skip leading spaces just in case: */ + while (RT_C_IS_SPACE(*pszCodeset)) + pszCodeset++; + + /* If prefixed by 'ISO-10646/' skip that (iconv access this, dunno about + LC_CTYPE et al., but play it safe): */ + if ( strncmp(pszCodeset, RT_STR_TUPLE("ISO-10646/")) == 0 + || strncmp(pszCodeset, RT_STR_TUPLE("iso-10646/")) == 0) + pszCodeset += sizeof("ISO-10646/") - 1; + + /* Match 'utf': */ + if ( (pszCodeset[0] == 'u' || pszCodeset[0] == 'U') + && (pszCodeset[1] == 't' || pszCodeset[1] == 'T') + && (pszCodeset[2] == 'f' || pszCodeset[2] == 'F')) + { + pszCodeset += 3; + + /* Treat the dash as optional: */ + if (*pszCodeset == '-') + pszCodeset++; + + /* Match '8': */ + if (*pszCodeset == '8') + { + do + pszCodeset++; + while (RT_C_IS_SPACE(*pszCodeset)); + + /* We ignore modifiers here (e.g. "[be_BY.]utf8@latin"). */ + if (!*pszCodeset || *pszCodeset == '@') + return true; + } + } + } + return false; +} + + + +#ifdef RT_WITH_ICONV_CACHE + +/** + * Initializes the iconv handle cache associated with a thread. + * + * @param pThread The thread in question. + */ +DECLHIDDEN(void) rtStrIconvCacheInit(PRTTHREADINT pThread) +{ + for (size_t i = 0; i < RT_ELEMENTS(pThread->ahIconvs); i++) + pThread->ahIconvs[i] = (iconv_t)-1; +} + +/** + * Destroys the iconv handle cache associated with a thread. + * + * @param pThread The thread in question. + */ +DECLHIDDEN(void) rtStrIconvCacheDestroy(PRTTHREADINT pThread) +{ + for (size_t i = 0; i < RT_ELEMENTS(pThread->ahIconvs); i++) + { + iconv_t hIconv = (iconv_t)pThread->ahIconvs[i]; + pThread->ahIconvs[i] = (iconv_t)-1; + if (hIconv != (iconv_t)-1) + iconv_close(hIconv); + } +} + + +/** + * Converts a string from one charset to another. + * + * @returns iprt status code. + * @param pvInput Pointer to intput string. + * @param cbInput Size (in bytes) of input string. Excludes any terminators. + * @param pszInputCS Codeset of the input string. + * @param ppvOutput Pointer to pointer to output buffer if cbOutput > 0. + * If cbOutput is 0 this is where the pointer to the allocated + * buffer is stored. + * @param cbOutput Size of the passed in buffer. + * @param pszOutputCS Codeset of the input string. + * @param cFactor Input vs. output size factor. + * @param phIconv Pointer to the cache entry. + */ +static int rtstrConvertCached(const void *pvInput, size_t cbInput, const char *pszInputCS, + void **ppvOutput, size_t cbOutput, const char *pszOutputCS, + unsigned cFactor, iconv_t *phIconv) +{ + /* + * Allocate buffer + */ + bool fUcs2Term; + void *pvOutput; + size_t cbOutput2; + if (!cbOutput) + { + cbOutput2 = cbInput * cFactor; + pvOutput = RTMemTmpAlloc(cbOutput2 + sizeof(RTUTF16)); + if (!pvOutput) + return VERR_NO_TMP_MEMORY; + fUcs2Term = true; + } + else + { + pvOutput = *ppvOutput; + fUcs2Term = !strcmp(pszOutputCS, "UCS-2") + || !strcmp(pszOutputCS, "UTF-16") + || !strcmp(pszOutputCS, "ucs-2") + || !strcmp(pszOutputCS, "utf-16"); + cbOutput2 = cbOutput - (fUcs2Term ? sizeof(RTUTF16) : 1); + if (cbOutput2 > cbOutput) + return VERR_BUFFER_OVERFLOW; + } + + /* + * Use a loop here to retry with bigger buffers. + */ + for (unsigned cTries = 10; cTries > 0; cTries--) + { + /* + * Create conversion object if necessary. + */ + iconv_t hIconv = (iconv_t)*phIconv; + if (hIconv == (iconv_t)-1) + { +#if defined(RT_OS_SOLARIS) || defined(RT_OS_NETBSD) || /* @bugref{10153}: Default to UTF-8: */ defined(RT_OS_DARWIN) + /* Some systems don't grok empty codeset strings, so help them find the current codeset. */ + if (!*pszInputCS) + pszInputCS = rtStrGetLocaleCodeset(); + if (!*pszOutputCS) + pszOutputCS = rtStrGetLocaleCodeset(); +#endif + IPRT_ALIGNMENT_CHECKS_DISABLE(); /* glibc causes trouble */ + *phIconv = hIconv = iconv_open(pszOutputCS, pszInputCS); + IPRT_ALIGNMENT_CHECKS_ENABLE(); + } + if (hIconv != (iconv_t)-1) + { + /* + * Do the conversion. + */ + size_t cbInLeft = cbInput; + size_t cbOutLeft = cbOutput2; + const void *pvInputLeft = pvInput; + void *pvOutputLeft = pvOutput; + size_t cchNonRev; +#ifdef NON_CONST_ICONV_INPUT + cchNonRev = iconv(hIconv, (char **)&pvInputLeft, &cbInLeft, (char **)&pvOutputLeft, &cbOutLeft); +#else + cchNonRev = iconv(hIconv, (const char **)&pvInputLeft, &cbInLeft, (char **)&pvOutputLeft, &cbOutLeft); +#endif + if (cchNonRev != (size_t)-1) + { + if (!cbInLeft) + { + /* + * We're done, just add the terminator and return. + * (Two terminators to support UCS-2 output, too.) + */ + ((char *)pvOutputLeft)[0] = '\0'; + if (fUcs2Term) + ((char *)pvOutputLeft)[1] = '\0'; + *ppvOutput = pvOutput; + if (cchNonRev == 0) + return VINF_SUCCESS; + return VWRN_NO_TRANSLATION; + } + errno = E2BIG; + } + + /* + * If we failed because of output buffer space we'll + * increase the output buffer size and retry. + */ + if (errno == E2BIG) + { + if (!cbOutput) + { + RTMemTmpFree(pvOutput); + cbOutput2 *= 2; + pvOutput = RTMemTmpAlloc(cbOutput2 + sizeof(RTUTF16)); + if (!pvOutput) + return VERR_NO_TMP_MEMORY; + continue; + } + return VERR_BUFFER_OVERFLOW; + } + + /* + * Close the handle on all other errors to make sure we won't carry + * any bad state with us. + */ + *phIconv = (iconv_t)-1; + iconv_close(hIconv); + } + break; + } + + /* failure */ + if (!cbOutput) + RTMemTmpFree(pvOutput); + return VERR_NO_TRANSLATION; +} + +#endif /* RT_WITH_ICONV_CACHE */ + +/** + * Converts a string from one charset to another without using the handle cache. + * + * @returns IPRT status code. + * + * @param pvInput Pointer to intput string. + * @param cbInput Size (in bytes) of input string. Excludes any terminators. + * @param pszInputCS Codeset of the input string. + * @param ppvOutput Pointer to pointer to output buffer if cbOutput > 0. + * If cbOutput is 0 this is where the pointer to the allocated + * buffer is stored. + * @param cbOutput Size of the passed in buffer. + * @param pszOutputCS Codeset of the input string. + * @param cFactor Input vs. output size factor. + */ +static int rtStrConvertUncached(const void *pvInput, size_t cbInput, const char *pszInputCS, + void **ppvOutput, size_t cbOutput, const char *pszOutputCS, + unsigned cFactor) +{ + /* + * Allocate buffer + */ + bool fUcs2Term; + void *pvOutput; + size_t cbOutput2; + if (!cbOutput) + { + cbOutput2 = cbInput * cFactor; + pvOutput = RTMemTmpAlloc(cbOutput2 + sizeof(RTUTF16)); + if (!pvOutput) + return VERR_NO_TMP_MEMORY; + fUcs2Term = true; + } + else + { + pvOutput = *ppvOutput; + fUcs2Term = !strcmp(pszOutputCS, "UCS-2"); + cbOutput2 = cbOutput - (fUcs2Term ? sizeof(RTUTF16) : 1); + if (cbOutput2 > cbOutput) + return VERR_BUFFER_OVERFLOW; + } + + /* + * Use a loop here to retry with bigger buffers. + */ + for (unsigned cTries = 10; cTries > 0; cTries--) + { + /* + * Create conversion object. + */ +#if defined(RT_OS_SOLARIS) || defined(RT_OS_NETBSD) || /* @bugref{10153}: Default to UTF-8: */ defined(RT_OS_DARWIN) + /* Some systems don't grok empty codeset strings, so help them find the current codeset. */ + if (!*pszInputCS) + pszInputCS = rtStrGetLocaleCodeset(); + if (!*pszOutputCS) + pszOutputCS = rtStrGetLocaleCodeset(); +#endif + IPRT_ALIGNMENT_CHECKS_DISABLE(); /* glibc causes trouble */ + iconv_t icHandle = iconv_open(pszOutputCS, pszInputCS); + IPRT_ALIGNMENT_CHECKS_ENABLE(); + if (icHandle != (iconv_t)-1) + { + /* + * Do the conversion. + */ + size_t cbInLeft = cbInput; + size_t cbOutLeft = cbOutput2; + const void *pvInputLeft = pvInput; + void *pvOutputLeft = pvOutput; + size_t cchNonRev; +#ifdef NON_CONST_ICONV_INPUT + cchNonRev = iconv(icHandle, (char **)&pvInputLeft, &cbInLeft, (char **)&pvOutputLeft, &cbOutLeft); +#else + cchNonRev = iconv(icHandle, (const char **)&pvInputLeft, &cbInLeft, (char **)&pvOutputLeft, &cbOutLeft); +#endif + if (cchNonRev != (size_t)-1) + { + if (!cbInLeft) + { + /* + * We're done, just add the terminator and return. + * (Two terminators to support UCS-2 output, too.) + */ + iconv_close(icHandle); + ((char *)pvOutputLeft)[0] = '\0'; + if (fUcs2Term) + ((char *)pvOutputLeft)[1] = '\0'; + *ppvOutput = pvOutput; + if (cchNonRev == 0) + return VINF_SUCCESS; + return VWRN_NO_TRANSLATION; + } + errno = E2BIG; + } + iconv_close(icHandle); + + /* + * If we failed because of output buffer space we'll + * increase the output buffer size and retry. + */ + if (errno == E2BIG) + { + if (!cbOutput) + { + RTMemTmpFree(pvOutput); + cbOutput2 *= 2; + pvOutput = RTMemTmpAlloc(cbOutput2 + sizeof(RTUTF16)); + if (!pvOutput) + return VERR_NO_TMP_MEMORY; + continue; + } + return VERR_BUFFER_OVERFLOW; + } + } + break; + } + + /* failure */ + if (!cbOutput) + RTMemTmpFree(pvOutput); + return VERR_NO_TRANSLATION; +} + + +/** + * Wrapper that selects rtStrConvertCached or rtStrConvertUncached. + * + * @returns IPRT status code. + * + * @param pszInput Pointer to intput string. + * @param cchInput Size (in bytes) of input string. Excludes any + * terminators. + * @param pszInputCS Codeset of the input string. + * @param ppszOutput Pointer to pointer to output buffer if cbOutput > 0. + * If cbOutput is 0 this is where the pointer to the + * allocated buffer is stored. + * @param cbOutput Size of the passed in buffer. + * @param pszOutputCS Codeset of the input string. + * @param cFactor Input vs. output size factor. + * @param enmCacheIdx The iconv cache index. + */ +DECLINLINE(int) rtStrConvertWrapper(const char *pchInput, size_t cchInput, const char *pszInputCS, + char **ppszOutput, size_t cbOutput, const char *pszOutputCS, + unsigned cFactor, RTSTRICONV enmCacheIdx) +{ +#ifdef RT_WITH_ICONV_CACHE + RTTHREAD hSelf = RTThreadSelf(); + if (hSelf != NIL_RTTHREAD) + { + PRTTHREADINT pThread = rtThreadGet(hSelf); + if (pThread) + { + if ((pThread->fIntFlags & (RTTHREADINT_FLAGS_ALIEN | RTTHREADINT_FLAGS_MAIN)) != RTTHREADINT_FLAGS_ALIEN) + { + int rc = rtstrConvertCached(pchInput, cchInput, pszInputCS, + (void **)ppszOutput, cbOutput, pszOutputCS, + cFactor, (iconv_t *)&pThread->ahIconvs[enmCacheIdx]); + rtThreadRelease(pThread); + return rc; + } + rtThreadRelease(pThread); + } + } +#endif + return rtStrConvertUncached(pchInput, cchInput, pszInputCS, + (void **)ppszOutput, cbOutput, pszOutputCS, + cFactor); +} + + +/** + * Internal API for use by the path conversion code. + * + * @returns IPRT status code. + * + * @param pszInput Pointer to intput string. + * @param cchInput Size (in bytes) of input string. Excludes any + * terminators. + * @param pszInputCS Codeset of the input string. + * @param ppszOutput Pointer to pointer to output buffer if cbOutput > 0. + * If cbOutput is 0 this is where the pointer to the + * allocated buffer is stored. + * @param cbOutput Size of the passed in buffer. + * @param pszOutputCS Codeset of the input string. + * @param cFactor Input vs. output size factor. + * @param enmCacheIdx The iconv cache index. + */ +DECLHIDDEN(int) rtStrConvert(const char *pchInput, size_t cchInput, const char *pszInputCS, + char **ppszOutput, size_t cbOutput, const char *pszOutputCS, + unsigned cFactor, RTSTRICONV enmCacheIdx) +{ + Assert(enmCacheIdx >= 0 && enmCacheIdx < RTSTRICONV_END); + return rtStrConvertWrapper(pchInput, cchInput, pszInputCS, + ppszOutput, cbOutput, pszOutputCS, + cFactor, enmCacheIdx); +} + + +/** + * Initializes a local conversion cache for use with rtStrLocalCacheConvert. + * + * Call rtStrLocalCacheDelete when done. + */ +DECLHIDDEN(void) rtStrLocalCacheInit(void **ppvTmpCache) +{ + *ppvTmpCache = (iconv_t)-1; +} + + +/** + * Cleans up a local conversion cache. + */ +DECLHIDDEN(void) rtStrLocalCacheDelete(void **ppvTmpCache) +{ +#ifdef RT_WITH_ICONV_CACHE + iconv_t icHandle = (iconv_t)*ppvTmpCache; + if (icHandle != (iconv_t)-1) + iconv_close(icHandle); +#endif + *ppvTmpCache = (iconv_t)-1; +} + + +/** + * Internal API for use by the process creation conversion code. + * + * @returns IPRT status code. + * + * @param pszInput Pointer to intput string. + * @param cchInput Size (in bytes) of input string. Excludes any + * terminators. + * @param pszInputCS Codeset of the input string. + * @param ppszOutput Pointer to pointer to output buffer if cbOutput > 0. + * If cbOutput is 0 this is where the pointer to the + * allocated buffer is stored. + * @param cbOutput Size of the passed in buffer. + * @param pszOutputCS Codeset of the input string. + * @param ppvTmpCache Pointer to local temporary cache. Must be + * initialized by calling rtStrLocalCacheInit and + * cleaned up afterwards by rtStrLocalCacheDelete. + * Optional. + */ +DECLHIDDEN(int) rtStrLocalCacheConvert(const char *pchInput, size_t cchInput, const char *pszInputCS, + char **ppszOutput, size_t cbOutput, const char *pszOutputCS, + void **ppvTmpCache) +{ +#ifdef RT_WITH_ICONV_CACHE + if (ppvTmpCache) + return rtstrConvertCached(pchInput, cchInput, pszInputCS, (void **)ppszOutput, cbOutput, pszOutputCS, + 1 /*cFactor*/, (iconv_t *)ppvTmpCache); +#else + RT_NOREF(ppvTmpCache); +#endif + + return rtStrConvertUncached(pchInput, cchInput, pszInputCS, (void **)ppszOutput, cbOutput, pszOutputCS, 1 /*cFactor*/); +} + + +RTR3DECL(int) RTStrUtf8ToCurrentCPTag(char **ppszString, const char *pszString, const char *pszTag) +{ + Assert(ppszString); + Assert(pszString); + *ppszString = NULL; + + /* + * Assume result string length is not longer than UTF-8 string. + */ + size_t cch = strlen(pszString); + if (cch <= 0) + { + /* zero length string passed. */ + *ppszString = (char *)RTMemTmpAllocZTag(sizeof(char), pszTag); + if (*ppszString) + return VINF_SUCCESS; + return VERR_NO_TMP_MEMORY; + } + return rtStrConvertWrapper(pszString, cch, "UTF-8", ppszString, 0, "", 1, RTSTRICONV_UTF8_TO_LOCALE); +} + + +RTR3DECL(int) RTStrUtf8ToCurrentCPExTag(char **ppszString, const char *pszString, size_t cchString, const char *pszTag) +{ + Assert(ppszString); + Assert(pszString); + *ppszString = NULL; + + /* + * Assume result string length is not longer than UTF-8 string. + */ + cchString = RTStrNLen(pszString, cchString); + if (cchString < 1) + { + /* zero length string passed. */ + *ppszString = (char *)RTMemTmpAllocZTag(sizeof(char), pszTag); + if (*ppszString) + return VINF_SUCCESS; + return VERR_NO_TMP_MEMORY; + } + return rtStrConvertWrapper(pszString, cchString, "UTF-8", ppszString, 0, "", 1, RTSTRICONV_UTF8_TO_LOCALE); +} + + +RTR3DECL(int) RTStrCurrentCPToUtf8Tag(char **ppszString, const char *pszString, const char *pszTag) +{ + Assert(ppszString); + Assert(pszString); + *ppszString = NULL; + + /* + * Attempt with UTF-8 length of 2x the native length. + */ + size_t cch = strlen(pszString); + if (cch <= 0) + { + /* zero length string passed. */ + *ppszString = (char *)RTMemTmpAllocZTag(sizeof(char), pszTag); + if (*ppszString) + return VINF_SUCCESS; + return VERR_NO_TMP_MEMORY; + } + return rtStrConvertWrapper(pszString, cch, "", ppszString, 0, "UTF-8", 2, RTSTRICONV_LOCALE_TO_UTF8); +} + + +RTR3DECL(int) RTStrConsoleCPToUtf8Tag(char **ppszString, const char *pszString, const char *pszTag) +{ + return RTStrCurrentCPToUtf8Tag(ppszString, pszString, pszTag); +} |