diff options
Diffstat (limited to '')
33 files changed, 11644 insertions, 0 deletions
diff --git a/lib/isc/unix/Makefile.in b/lib/isc/unix/Makefile.in new file mode 100644 index 0000000..4af7d6e --- /dev/null +++ b/lib/isc/unix/Makefile.in @@ -0,0 +1,48 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# SPDX-License-Identifier: MPL-2.0 +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, you can obtain one at https://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +CINCLUDES = -I${srcdir}/include \ + -I${srcdir}/../pthreads/include \ + -I../include \ + -I${srcdir}/../include \ + -I${srcdir}/.. \ + ${OPENSSL_CFLAGS} \ + ${JSON_C_CFLAGS} \ + ${LIBXML2_CFLAGS} + +CDEFINES = +CWARNINGS = + +# Alphabetically +OBJS = pk11_api.@O@ \ + dir.@O@ errno.@O@ errno2result.@O@ \ + file.@O@ fsaccess.@O@ interfaceiter.@O@ \ + meminfo.@O@ \ + net.@O@ os.@O@ resource.@O@ socket.@O@ stdio.@O@ stdtime.@O@ \ + syslog.@O@ time.@O@ + +# Alphabetically +SRCS = pk11_api.c \ + dir.c errno.c errno2result.c \ + file.c fsaccess.c interfaceiter.c meminfo.c \ + net.c os.c resource.c socket.c stdio.c stdtime.c \ + syslog.c time.c + +SUBDIRS = include +TARGETS = ${OBJS} + +@BIND9_MAKE_RULES@ + +interfaceiter.@O@: interfaceiter.c ifiter_getifaddrs.c diff --git a/lib/isc/unix/dir.c b/lib/isc/unix/dir.c new file mode 100644 index 0000000..b7eabe9 --- /dev/null +++ b/lib/isc/unix/dir.c @@ -0,0 +1,268 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <ctype.h> +#include <errno.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <unistd.h> + +#include <isc/dir.h> +#include <isc/magic.h> +#include <isc/netdb.h> +#include <isc/print.h> +#include <isc/string.h> +#include <isc/util.h> + +#include "errno2result.h" + +#define ISC_DIR_MAGIC ISC_MAGIC('D', 'I', 'R', '*') +#define VALID_DIR(dir) ISC_MAGIC_VALID(dir, ISC_DIR_MAGIC) + +void +isc_dir_init(isc_dir_t *dir) { + REQUIRE(dir != NULL); + + dir->entry.name[0] = '\0'; + dir->entry.length = 0; + + dir->handle = NULL; + + dir->magic = ISC_DIR_MAGIC; +} + +/*! + * \brief Allocate workspace and open directory stream. If either one fails, + * NULL will be returned. + */ +isc_result_t +isc_dir_open(isc_dir_t *dir, const char *dirname) { + char *p; + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(VALID_DIR(dir)); + REQUIRE(dirname != NULL); + + /* + * Copy directory name. Need to have enough space for the name, + * a possible path separator, the wildcard, and the final NUL. + */ + if (strlen(dirname) + 3 > sizeof(dir->dirname)) { + /* XXXDCL ? */ + return (ISC_R_NOSPACE); + } + strlcpy(dir->dirname, dirname, sizeof(dir->dirname)); + + /* + * Append path separator, if needed, and "*". + */ + p = dir->dirname + strlen(dir->dirname); + if (dir->dirname < p && *(p - 1) != '/') { + *p++ = '/'; + } + *p++ = '*'; + *p = '\0'; + + /* + * Open stream. + */ + dir->handle = opendir(dirname); + + if (dir->handle == NULL) { + return (isc__errno2result(errno)); + } + + return (result); +} + +/*! + * \brief Return previously retrieved file or get next one. + * + * Unix's dirent has + * separate open and read functions, but the Win32 and DOS interfaces open + * the dir stream and reads the first file in one operation. + */ +isc_result_t +isc_dir_read(isc_dir_t *dir) { + struct dirent *entry; + + REQUIRE(VALID_DIR(dir) && dir->handle != NULL); + + /* + * Fetch next file in directory. + */ + entry = readdir(dir->handle); + + if (entry == NULL) { + return (ISC_R_NOMORE); + } + + /* + * Make sure that the space for the name is long enough. + */ + if (sizeof(dir->entry.name) <= strlen(entry->d_name)) { + return (ISC_R_UNEXPECTED); + } + + strlcpy(dir->entry.name, entry->d_name, sizeof(dir->entry.name)); + + /* + * Some dirents have d_namlen, but it is not portable. + */ + dir->entry.length = strlen(entry->d_name); + + return (ISC_R_SUCCESS); +} + +/*! + * \brief Close directory stream. + */ +void +isc_dir_close(isc_dir_t *dir) { + REQUIRE(VALID_DIR(dir) && dir->handle != NULL); + + (void)closedir(dir->handle); + dir->handle = NULL; +} + +/*! + * \brief Reposition directory stream at start. + */ +isc_result_t +isc_dir_reset(isc_dir_t *dir) { + REQUIRE(VALID_DIR(dir) && dir->handle != NULL); + + rewinddir(dir->handle); + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_dir_chdir(const char *dirname) { + /*! + * \brief Change the current directory to 'dirname'. + */ + + REQUIRE(dirname != NULL); + + if (chdir(dirname) < 0) { + return (isc__errno2result(errno)); + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_dir_chroot(const char *dirname) { +#ifdef HAVE_CHROOT + void *tmp; +#endif /* ifdef HAVE_CHROOT */ + + REQUIRE(dirname != NULL); + +#ifdef HAVE_CHROOT + /* + * Try to use getservbyname and getprotobyname before chroot. + * If WKS records are used in a zone under chroot, Name Service Switch + * may fail to load library in chroot. + * Do not report errors if it fails, we do not need any result now. + */ + tmp = getprotobyname("udp"); + if (tmp != NULL) { + (void)getservbyname("domain", "udp"); + } + + if (chroot(dirname) < 0 || chdir("/") < 0) { + return (isc__errno2result(errno)); + } + + return (ISC_R_SUCCESS); +#else /* ifdef HAVE_CHROOT */ + return (ISC_R_NOTIMPLEMENTED); +#endif /* ifdef HAVE_CHROOT */ +} + +isc_result_t +isc_dir_createunique(char *templet) { + isc_result_t result; + char *x; + char *p; + int i; + int pid; + + REQUIRE(templet != NULL); + + /*! + * \brief mkdtemp is not portable, so this emulates it. + */ + + pid = getpid(); + + /* + * Replace trailing Xs with the process-id, zero-filled. + */ + for (x = templet + strlen(templet) - 1; *x == 'X' && x >= templet; + x--, pid /= 10) + { + *x = pid % 10 + '0'; + } + + x++; /* Set x to start of ex-Xs. */ + + do { + i = mkdir(templet, 0700); + if (i == 0 || errno != EEXIST) { + break; + } + + /* + * The BSD algorithm. + */ + p = x; + while (*p != '\0') { + if (isdigit((unsigned char)*p)) { + *p = 'a'; + } else if (*p != 'z') { + ++*p; + } else { + /* + * Reset character and move to next. + */ + *p++ = 'a'; + continue; + } + + break; + } + + if (*p == '\0') { + /* + * Tried all combinations. errno should already + * be EEXIST, but ensure it is anyway for + * isc__errno2result(). + */ + errno = EEXIST; + break; + } + } while (1); + + if (i == -1) { + result = isc__errno2result(errno); + } else { + result = ISC_R_SUCCESS; + } + + return (result); +} diff --git a/lib/isc/unix/errno.c b/lib/isc/unix/errno.c new file mode 100644 index 0000000..5875f3b --- /dev/null +++ b/lib/isc/unix/errno.c @@ -0,0 +1,24 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <isc/errno.h> +#include <isc/util.h> + +#include "errno2result.h" + +isc_result_t +isc_errno_toresult(int err) { + return (isc___errno2result(err, false, 0, 0)); +} diff --git a/lib/isc/unix/errno2result.c b/lib/isc/unix/errno2result.c new file mode 100644 index 0000000..6f752df --- /dev/null +++ b/lib/isc/unix/errno2result.c @@ -0,0 +1,131 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include "errno2result.h" +#include <stdbool.h> + +#include <isc/platform.h> +#include <isc/result.h> +#include <isc/strerr.h> +#include <isc/string.h> +#include <isc/util.h> + +/*% + * Convert a POSIX errno value into an isc_result_t. The + * list of supported errno values is not complete; new users + * of this function should add any expected errors that are + * not already there. + */ +isc_result_t +isc___errno2result(int posixerrno, bool dolog, const char *file, + unsigned int line) { + char strbuf[ISC_STRERRORSIZE]; + + switch (posixerrno) { + case ENOTDIR: + case ELOOP: + case EINVAL: /* XXX sometimes this is not for files */ + case ENAMETOOLONG: + case EBADF: + return (ISC_R_INVALIDFILE); + case ENOENT: + return (ISC_R_FILENOTFOUND); + case EACCES: + case EPERM: + return (ISC_R_NOPERM); + case EEXIST: + return (ISC_R_FILEEXISTS); + case EIO: + return (ISC_R_IOERROR); + case ENOMEM: + return (ISC_R_NOMEMORY); + case ENFILE: + case EMFILE: + return (ISC_R_TOOMANYOPENFILES); +#ifdef EDQUOT + case EDQUOT: + return (ISC_R_DISCQUOTA); +#endif /* ifdef EDQUOT */ + case ENOSPC: + return (ISC_R_DISCFULL); +#ifdef EOVERFLOW + case EOVERFLOW: + return (ISC_R_RANGE); +#endif /* ifdef EOVERFLOW */ + case EPIPE: +#ifdef ECONNRESET + case ECONNRESET: +#endif /* ifdef ECONNRESET */ +#ifdef ECONNABORTED + case ECONNABORTED: +#endif /* ifdef ECONNABORTED */ + return (ISC_R_CONNECTIONRESET); +#ifdef ENOTCONN + case ENOTCONN: + return (ISC_R_NOTCONNECTED); +#endif /* ifdef ENOTCONN */ +#ifdef ETIMEDOUT + case ETIMEDOUT: + return (ISC_R_TIMEDOUT); +#endif /* ifdef ETIMEDOUT */ +#ifdef ENOBUFS + case ENOBUFS: + return (ISC_R_NORESOURCES); +#endif /* ifdef ENOBUFS */ +#ifdef EAFNOSUPPORT + case EAFNOSUPPORT: + return (ISC_R_FAMILYNOSUPPORT); +#endif /* ifdef EAFNOSUPPORT */ +#ifdef ENETDOWN + case ENETDOWN: + return (ISC_R_NETDOWN); +#endif /* ifdef ENETDOWN */ +#ifdef EHOSTDOWN + case EHOSTDOWN: + return (ISC_R_HOSTDOWN); +#endif /* ifdef EHOSTDOWN */ +#ifdef ENETUNREACH + case ENETUNREACH: + return (ISC_R_NETUNREACH); +#endif /* ifdef ENETUNREACH */ +#ifdef EHOSTUNREACH + case EHOSTUNREACH: + return (ISC_R_HOSTUNREACH); +#endif /* ifdef EHOSTUNREACH */ +#ifdef EADDRINUSE + case EADDRINUSE: + return (ISC_R_ADDRINUSE); +#endif /* ifdef EADDRINUSE */ + case EADDRNOTAVAIL: + return (ISC_R_ADDRNOTAVAIL); + case ECONNREFUSED: + return (ISC_R_CONNREFUSED); + default: + if (dolog) { + strerror_r(posixerrno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(file, line, + "unable to convert errno " + "to isc_result: %d: %s", + posixerrno, strbuf); + } + /* + * XXXDCL would be nice if perhaps this function could + * return the system's error string, so the caller + * might have something more descriptive than "unexpected + * error" to log with. + */ + return (ISC_R_UNEXPECTED); + } +} diff --git a/lib/isc/unix/errno2result.h b/lib/isc/unix/errno2result.h new file mode 100644 index 0000000..05de1f2 --- /dev/null +++ b/lib/isc/unix/errno2result.h @@ -0,0 +1,37 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef UNIX_ERRNO2RESULT_H +#define UNIX_ERRNO2RESULT_H 1 + +/*! \file */ + +/* XXXDCL this should be moved to lib/isc/include/isc/errno2result.h. */ + +#include <errno.h> /* Provides errno. */ +#include <stdbool.h> + +#include <isc/lang.h> +#include <isc/types.h> + +ISC_LANG_BEGINDECLS + +#define isc__errno2result(x) isc___errno2result(x, true, __FILE__, __LINE__) + +isc_result_t +isc___errno2result(int posixerrno, bool dolog, const char *file, + unsigned int line); + +ISC_LANG_ENDDECLS + +#endif /* UNIX_ERRNO2RESULT_H */ diff --git a/lib/isc/unix/file.c b/lib/isc/unix/file.c new file mode 100644 index 0000000..0604f14 --- /dev/null +++ b/lib/isc/unix/file.c @@ -0,0 +1,840 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/* + * Portions Copyright (c) 1987, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +/*! \file */ + +#include <errno.h> +#include <fcntl.h> +#include <inttypes.h> +#include <limits.h> +#include <stdbool.h> +#include <stdlib.h> +#include <sys/stat.h> +#include <sys/time.h> +#include <time.h> /* Required for utimes on some platforms. */ +#include <unistd.h> /* Required for mkstemp on NetBSD. */ + +#ifdef HAVE_SYS_MMAN_H +#include <sys/mman.h> +#endif /* ifdef HAVE_SYS_MMAN_H */ + +#include <isc/dir.h> +#include <isc/file.h> +#include <isc/log.h> +#include <isc/md.h> +#include <isc/mem.h> +#include <isc/platform.h> +#include <isc/print.h> +#include <isc/random.h> +#include <isc/string.h> +#include <isc/time.h> +#include <isc/util.h> + +#include "errno2result.h" + +/* + * XXXDCL As the API for accessing file statistics undoubtedly gets expanded, + * it might be good to provide a mechanism that allows for the results + * of a previous stat() to be used again without having to do another stat, + * such as perl's mechanism of using "_" in place of a file name to indicate + * that the results of the last stat should be used. But then you get into + * annoying MP issues. BTW, Win32 has stat(). + */ +static isc_result_t +file_stats(const char *file, struct stat *stats) { + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(file != NULL); + REQUIRE(stats != NULL); + + if (stat(file, stats) != 0) { + result = isc__errno2result(errno); + } + + return (result); +} + +static isc_result_t +fd_stats(int fd, struct stat *stats) { + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(stats != NULL); + + if (fstat(fd, stats) != 0) { + result = isc__errno2result(errno); + } + + return (result); +} + +isc_result_t +isc_file_getsizefd(int fd, off_t *size) { + isc_result_t result; + struct stat stats; + + REQUIRE(size != NULL); + + result = fd_stats(fd, &stats); + + if (result == ISC_R_SUCCESS) { + *size = stats.st_size; + } + + return (result); +} + +isc_result_t +isc_file_mode(const char *file, mode_t *modep) { + isc_result_t result; + struct stat stats; + + REQUIRE(modep != NULL); + + result = file_stats(file, &stats); + if (result == ISC_R_SUCCESS) { + *modep = (stats.st_mode & 07777); + } + + return (result); +} + +isc_result_t +isc_file_getmodtime(const char *file, isc_time_t *modtime) { + isc_result_t result; + struct stat stats; + + REQUIRE(file != NULL); + REQUIRE(modtime != NULL); + + result = file_stats(file, &stats); + + if (result == ISC_R_SUCCESS) { +#if defined(HAVE_STAT_NSEC) + isc_time_set(modtime, stats.st_mtime, stats.st_mtim.tv_nsec); +#else /* if defined(HAVE_STAT_NSEC) */ + isc_time_set(modtime, stats.st_mtime, 0); +#endif /* if defined(HAVE_STAT_NSEC) */ + } + + return (result); +} + +isc_result_t +isc_file_getsize(const char *file, off_t *size) { + isc_result_t result; + struct stat stats; + + REQUIRE(file != NULL); + REQUIRE(size != NULL); + + result = file_stats(file, &stats); + + if (result == ISC_R_SUCCESS) { + *size = stats.st_size; + } + + return (result); +} + +isc_result_t +isc_file_settime(const char *file, isc_time_t *when) { + struct timeval times[2]; + + REQUIRE(file != NULL && when != NULL); + + /* + * tv_sec is at least a 32 bit quantity on all platforms we're + * dealing with, but it is signed on most (all?) of them, + * so we need to make sure the high bit isn't set. This unfortunately + * loses when either: + * * tv_sec becomes a signed 64 bit integer but long is 32 bits + * and isc_time_seconds > LONG_MAX, or + * * isc_time_seconds is changed to be > 32 bits but long is 32 bits + * and isc_time_seconds has at least 33 significant bits. + */ + times[0].tv_sec = times[1].tv_sec = (long)isc_time_seconds(when); + + /* + * Here is the real check for the high bit being set. + */ + if ((times[0].tv_sec & + (1ULL << (sizeof(times[0].tv_sec) * CHAR_BIT - 1))) != 0) + { + return (ISC_R_RANGE); + } + + /* + * isc_time_nanoseconds guarantees a value that divided by 1000 will + * fit into the minimum possible size tv_usec field. + */ + times[0].tv_usec = times[1].tv_usec = + (int32_t)(isc_time_nanoseconds(when) / 1000); + + if (utimes(file, times) < 0) { + return (isc__errno2result(errno)); + } + + return (ISC_R_SUCCESS); +} + +#undef TEMPLATE +#define TEMPLATE "tmp-XXXXXXXXXX" /*%< 14 characters. */ + +isc_result_t +isc_file_mktemplate(const char *path, char *buf, size_t buflen) { + return (isc_file_template(path, TEMPLATE, buf, buflen)); +} + +isc_result_t +isc_file_template(const char *path, const char *templet, char *buf, + size_t buflen) { + const char *s; + + REQUIRE(templet != NULL); + REQUIRE(buf != NULL); + + if (path == NULL) { + path = ""; + } + + s = strrchr(templet, '/'); + if (s != NULL) { + templet = s + 1; + } + + s = strrchr(path, '/'); + + if (s != NULL) { + size_t prefixlen = s - path + 1; + if ((prefixlen + strlen(templet) + 1) > buflen) { + return (ISC_R_NOSPACE); + } + + /* Copy 'prefixlen' bytes and NUL terminate. */ + strlcpy(buf, path, ISC_MIN(prefixlen + 1, buflen)); + strlcat(buf, templet, buflen); + } else { + if ((strlen(templet) + 1) > buflen) { + return (ISC_R_NOSPACE); + } + + strlcpy(buf, templet, buflen); + } + + return (ISC_R_SUCCESS); +} + +static const char alphnum[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuv" + "wxyz0123456789"; + +isc_result_t +isc_file_renameunique(const char *file, char *templet) { + char *x; + char *cp; + + REQUIRE(file != NULL); + REQUIRE(templet != NULL); + + cp = templet; + while (*cp != '\0') { + cp++; + } + if (cp == templet) { + return (ISC_R_FAILURE); + } + + x = cp--; + while (cp >= templet && *cp == 'X') { + *cp = alphnum[isc_random_uniform(sizeof(alphnum) - 1)]; + x = cp--; + } + while (link(file, templet) == -1) { + if (errno != EEXIST) { + return (isc__errno2result(errno)); + } + for (cp = x;;) { + const char *t; + if (*cp == '\0') { + return (ISC_R_FAILURE); + } + t = strchr(alphnum, *cp); + if (t == NULL || *++t == '\0') { + *cp++ = alphnum[0]; + } else { + *cp = *t; + break; + } + } + } + if (unlink(file) < 0) { + if (errno != ENOENT) { + return (isc__errno2result(errno)); + } + } + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_openunique(char *templet, FILE **fp) { + int mode = S_IWUSR | S_IRUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_openuniqueprivate(char *templet, FILE **fp) { + int mode = S_IWUSR | S_IRUSR; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_openuniquemode(char *templet, int mode, FILE **fp) { + int fd; + FILE *f; + isc_result_t result = ISC_R_SUCCESS; + char *x; + char *cp; + + REQUIRE(templet != NULL); + REQUIRE(fp != NULL && *fp == NULL); + + cp = templet; + while (*cp != '\0') { + cp++; + } + if (cp == templet) { + return (ISC_R_FAILURE); + } + + x = cp--; + while (cp >= templet && *cp == 'X') { + *cp = alphnum[isc_random_uniform(sizeof(alphnum) - 1)]; + x = cp--; + } + + while ((fd = open(templet, O_RDWR | O_CREAT | O_EXCL, mode)) == -1) { + if (errno != EEXIST) { + return (isc__errno2result(errno)); + } + for (cp = x;;) { + char *t; + if (*cp == '\0') { + return (ISC_R_FAILURE); + } + t = strchr(alphnum, *cp); + if (t == NULL || *++t == '\0') { + *cp++ = alphnum[0]; + } else { + *cp = *t; + break; + } + } + } + f = fdopen(fd, "w+"); + if (f == NULL) { + result = isc__errno2result(errno); + if (remove(templet) < 0) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_FILE, ISC_LOG_ERROR, + "remove '%s': failed", templet); + } + (void)close(fd); + } else { + *fp = f; + } + + return (result); +} + +isc_result_t +isc_file_bopenunique(char *templet, FILE **fp) { + int mode = S_IWUSR | S_IRUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_bopenuniqueprivate(char *templet, FILE **fp) { + int mode = S_IWUSR | S_IRUSR; + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_bopenuniquemode(char *templet, int mode, FILE **fp) { + return (isc_file_openuniquemode(templet, mode, fp)); +} + +isc_result_t +isc_file_remove(const char *filename) { + int r; + + REQUIRE(filename != NULL); + + r = unlink(filename); + if (r == 0) { + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} + +isc_result_t +isc_file_rename(const char *oldname, const char *newname) { + int r; + + REQUIRE(oldname != NULL); + REQUIRE(newname != NULL); + + r = rename(oldname, newname); + if (r == 0) { + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} + +bool +isc_file_exists(const char *pathname) { + struct stat stats; + + REQUIRE(pathname != NULL); + + return (file_stats(pathname, &stats) == ISC_R_SUCCESS); +} + +isc_result_t +isc_file_isplainfile(const char *filename) { + /* + * This function returns success if filename is a plain file. + */ + struct stat filestat; + memset(&filestat, 0, sizeof(struct stat)); + + if ((stat(filename, &filestat)) == -1) { + return (isc__errno2result(errno)); + } + + if (!S_ISREG(filestat.st_mode)) { + return (ISC_R_INVALIDFILE); + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_isplainfilefd(int fd) { + /* + * This function returns success if filename is a plain file. + */ + struct stat filestat; + memset(&filestat, 0, sizeof(struct stat)); + + if ((fstat(fd, &filestat)) == -1) { + return (isc__errno2result(errno)); + } + + if (!S_ISREG(filestat.st_mode)) { + return (ISC_R_INVALIDFILE); + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_isdirectory(const char *filename) { + /* + * This function returns success if filename exists and is a + * directory. + */ + struct stat filestat; + memset(&filestat, 0, sizeof(struct stat)); + + if ((stat(filename, &filestat)) == -1) { + return (isc__errno2result(errno)); + } + + if (!S_ISDIR(filestat.st_mode)) { + return (ISC_R_INVALIDFILE); + } + + return (ISC_R_SUCCESS); +} + +bool +isc_file_isabsolute(const char *filename) { + REQUIRE(filename != NULL); + return (filename[0] == '/'); +} + +bool +isc_file_iscurrentdir(const char *filename) { + REQUIRE(filename != NULL); + return (filename[0] == '.' && filename[1] == '\0'); +} + +bool +isc_file_ischdiridempotent(const char *filename) { + REQUIRE(filename != NULL); + if (isc_file_isabsolute(filename)) { + return (true); + } + if (isc_file_iscurrentdir(filename)) { + return (true); + } + return (false); +} + +const char * +isc_file_basename(const char *filename) { + const char *s; + + REQUIRE(filename != NULL); + + s = strrchr(filename, '/'); + if (s == NULL) { + return (filename); + } + + return (s + 1); +} + +isc_result_t +isc_file_progname(const char *filename, char *buf, size_t buflen) { + const char *base; + size_t len; + + REQUIRE(filename != NULL); + REQUIRE(buf != NULL); + + base = isc_file_basename(filename); + len = strlen(base) + 1; + + if (len > buflen) { + return (ISC_R_NOSPACE); + } + memmove(buf, base, len); + + return (ISC_R_SUCCESS); +} + +/* + * Put the absolute name of the current directory into 'dirname', which is + * a buffer of at least 'length' characters. End the string with the + * appropriate path separator, such that the final product could be + * concatenated with a relative pathname to make a valid pathname string. + */ +static isc_result_t +dir_current(char *dirname, size_t length) { + char *cwd; + isc_result_t result = ISC_R_SUCCESS; + + REQUIRE(dirname != NULL); + REQUIRE(length > 0U); + + cwd = getcwd(dirname, length); + + if (cwd == NULL) { + if (errno == ERANGE) { + result = ISC_R_NOSPACE; + } else { + result = isc__errno2result(errno); + } + } else { + if (strlen(dirname) + 1 == length) { + result = ISC_R_NOSPACE; + } else if (dirname[1] != '\0') { + strlcat(dirname, "/", length); + } + } + + return (result); +} + +isc_result_t +isc_file_absolutepath(const char *filename, char *path, size_t pathlen) { + isc_result_t result; + result = dir_current(path, pathlen); + if (result != ISC_R_SUCCESS) { + return (result); + } + if (strlen(path) + strlen(filename) + 1 > pathlen) { + return (ISC_R_NOSPACE); + } + strlcat(path, filename, pathlen); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_truncate(const char *filename, isc_offset_t size) { + isc_result_t result = ISC_R_SUCCESS; + + if (truncate(filename, size) < 0) { + result = isc__errno2result(errno); + } + return (result); +} + +isc_result_t +isc_file_safecreate(const char *filename, FILE **fp) { + isc_result_t result; + int flags; + struct stat sb; + FILE *f; + int fd; + + REQUIRE(filename != NULL); + REQUIRE(fp != NULL && *fp == NULL); + + result = file_stats(filename, &sb); + if (result == ISC_R_SUCCESS) { + if ((sb.st_mode & S_IFREG) == 0) { + return (ISC_R_INVALIDFILE); + } + flags = O_WRONLY | O_TRUNC; + } else if (result == ISC_R_FILENOTFOUND) { + flags = O_WRONLY | O_CREAT | O_EXCL; + } else { + return (result); + } + + fd = open(filename, flags, S_IRUSR | S_IWUSR); + if (fd == -1) { + return (isc__errno2result(errno)); + } + + f = fdopen(fd, "w"); + if (f == NULL) { + result = isc__errno2result(errno); + close(fd); + return (result); + } + + *fp = f; + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_splitpath(isc_mem_t *mctx, const char *path, char **dirname, + char const **bname) { + char *dir; + const char *file, *slash; + + if (path == NULL) { + return (ISC_R_INVALIDFILE); + } + + slash = strrchr(path, '/'); + + if (slash == path) { + file = ++slash; + dir = isc_mem_strdup(mctx, "/"); + } else if (slash != NULL) { + file = ++slash; + dir = isc_mem_allocate(mctx, slash - path); + strlcpy(dir, path, slash - path); + } else { + file = path; + dir = isc_mem_strdup(mctx, "."); + } + + if (dir == NULL) { + return (ISC_R_NOMEMORY); + } + + if (*file == '\0') { + isc_mem_free(mctx, dir); + return (ISC_R_INVALIDFILE); + } + + *dirname = dir; + *bname = file; + + return (ISC_R_SUCCESS); +} + +void * +isc_file_mmap(void *addr, size_t len, int prot, int flags, int fd, + off_t offset) { +#ifdef HAVE_MMAP + return (mmap(addr, len, prot, flags, fd, offset)); +#else /* ifdef HAVE_MMAP */ + void *buf; + ssize_t ret; + off_t end; + + UNUSED(addr); + UNUSED(prot); + UNUSED(flags); + + end = lseek(fd, 0, SEEK_END); + lseek(fd, offset, SEEK_SET); + if (end - offset < (off_t)len) { + len = end - offset; + } + + buf = malloc(len); + if (buf == NULL) { + return (NULL); + } + + ret = read(fd, buf, len); + if (ret != (ssize_t)len) { + free(buf); + buf = NULL; + } + + return (buf); +#endif /* ifdef HAVE_MMAP */ +} + +int +isc_file_munmap(void *addr, size_t len) { +#ifdef HAVE_MMAP + return (munmap(addr, len)); +#else /* ifdef HAVE_MMAP */ + UNUSED(len); + + free(addr); + return (0); +#endif /* ifdef HAVE_MMAP */ +} + +#define DISALLOW "\\/ABCDEFGHIJKLMNOPQRSTUVWXYZ" + +static isc_result_t +digest2hex(unsigned char *digest, unsigned int digestlen, char *hash, + size_t hashlen) { + unsigned int i; + int ret; + for (i = 0; i < digestlen; i++) { + size_t left = hashlen - i * 2; + ret = snprintf(hash + i * 2, left, "%02x", digest[i]); + if (ret < 0 || (size_t)ret >= left) { + return (ISC_R_NOSPACE); + } + } + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_file_sanitize(const char *dir, const char *base, const char *ext, + char *path, size_t length) { + char buf[PATH_MAX]; + unsigned char digest[ISC_MAX_MD_SIZE]; + unsigned int digestlen; + char hash[ISC_MAX_MD_SIZE * 2 + 1]; + size_t l = 0; + isc_result_t err; + + REQUIRE(base != NULL); + REQUIRE(path != NULL); + + l = strlen(base) + 1; + + /* + * allow room for a full sha256 hash (64 chars + * plus null terminator) + */ + if (l < 65U) { + l = 65; + } + + if (dir != NULL) { + l += strlen(dir) + 1; + } + if (ext != NULL) { + l += strlen(ext) + 1; + } + + if (l > length || l > (unsigned)PATH_MAX) { + return (ISC_R_NOSPACE); + } + + /* Check whether the full-length SHA256 hash filename exists */ + err = isc_md(ISC_MD_SHA256, (const unsigned char *)base, strlen(base), + digest, &digestlen); + if (err != ISC_R_SUCCESS) { + return (err); + } + + err = digest2hex(digest, digestlen, hash, sizeof(hash)); + if (err != ISC_R_SUCCESS) { + return (err); + } + + snprintf(buf, sizeof(buf), "%s%s%s%s%s", dir != NULL ? dir : "", + dir != NULL ? "/" : "", hash, ext != NULL ? "." : "", + ext != NULL ? ext : ""); + if (isc_file_exists(buf)) { + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); + } + + /* Check for a truncated SHA256 hash filename */ + hash[16] = '\0'; + snprintf(buf, sizeof(buf), "%s%s%s%s%s", dir != NULL ? dir : "", + dir != NULL ? "/" : "", hash, ext != NULL ? "." : "", + ext != NULL ? ext : ""); + if (isc_file_exists(buf)) { + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); + } + + /* + * If neither hash filename already exists, then we'll use + * the original base name if it has no disallowed characters, + * or the truncated hash name if it does. + */ + if (strpbrk(base, DISALLOW) != NULL) { + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); + } + + snprintf(buf, sizeof(buf), "%s%s%s%s%s", dir != NULL ? dir : "", + dir != NULL ? "/" : "", base, ext != NULL ? "." : "", + ext != NULL ? ext : ""); + strlcpy(path, buf, length); + return (ISC_R_SUCCESS); +} + +bool +isc_file_isdirwritable(const char *path) { + return (access(path, W_OK | X_OK) == 0); +} diff --git a/lib/isc/unix/fsaccess.c b/lib/isc/unix/fsaccess.c new file mode 100644 index 0000000..306cdfd --- /dev/null +++ b/lib/isc/unix/fsaccess.c @@ -0,0 +1,87 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <errno.h> +#include <stdbool.h> +#include <sys/stat.h> +#include <sys/types.h> + +#include "errno2result.h" + +/*! \file + * \brief + * The OS-independent part of the API is in lib/isc. + */ +#include "../fsaccess.c" + +isc_result_t +isc_fsaccess_set(const char *path, isc_fsaccess_t access) { + struct stat statb; + mode_t mode; + bool is_dir = false; + isc_fsaccess_t bits; + isc_result_t result; + + if (stat(path, &statb) != 0) { + return (isc__errno2result(errno)); + } + + if ((statb.st_mode & S_IFDIR) != 0) { + is_dir = true; + } else if ((statb.st_mode & S_IFREG) == 0) { + return (ISC_R_INVALIDFILE); + } + + result = check_bad_bits(access, is_dir); + if (result != ISC_R_SUCCESS) { + return (result); + } + + /* + * Done with checking bad bits. Set mode_t. + */ + mode = 0; + +#define SET_AND_CLEAR1(modebit) \ + if ((access & bits) != 0) { \ + mode |= modebit; \ + access &= ~bits; \ + } +#define SET_AND_CLEAR(user, group, other) \ + SET_AND_CLEAR1(user); \ + bits <<= STEP; \ + SET_AND_CLEAR1(group); \ + bits <<= STEP; \ + SET_AND_CLEAR1(other); + + bits = ISC_FSACCESS_READ | ISC_FSACCESS_LISTDIRECTORY; + + SET_AND_CLEAR(S_IRUSR, S_IRGRP, S_IROTH); + + bits = ISC_FSACCESS_WRITE | ISC_FSACCESS_CREATECHILD | + ISC_FSACCESS_DELETECHILD; + + SET_AND_CLEAR(S_IWUSR, S_IWGRP, S_IWOTH); + + bits = ISC_FSACCESS_EXECUTE | ISC_FSACCESS_ACCESSCHILD; + + SET_AND_CLEAR(S_IXUSR, S_IXGRP, S_IXOTH); + + INSIST(access == 0); + + if (chmod(path, mode) < 0) { + return (isc__errno2result(errno)); + } + + return (ISC_R_SUCCESS); +} diff --git a/lib/isc/unix/ifiter_getifaddrs.c b/lib/isc/unix/ifiter_getifaddrs.c new file mode 100644 index 0000000..9f8b31a --- /dev/null +++ b/lib/isc/unix/ifiter_getifaddrs.c @@ -0,0 +1,240 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file + * \brief + * Obtain the list of network interfaces using the getifaddrs(3) library. + */ + +#include <ifaddrs.h> +#include <stdbool.h> + +#include <isc/strerr.h> + +/*% Iterator Magic */ +#define IFITER_MAGIC ISC_MAGIC('I', 'F', 'I', 'G') +/*% Valid Iterator */ +#define VALID_IFITER(t) ISC_MAGIC_VALID(t, IFITER_MAGIC) + +#ifdef __linux +static bool seenv6 = false; +#endif /* ifdef __linux */ + +/*% Iterator structure */ +struct isc_interfaceiter { + unsigned int magic; /*%< Magic number. */ + isc_mem_t *mctx; + void *buf; /*%< (unused) */ + unsigned int bufsize; /*%< (always 0) */ + struct ifaddrs *ifaddrs; /*%< List of ifaddrs */ + struct ifaddrs *pos; /*%< Ptr to current ifaddr */ + isc_interface_t current; /*%< Current interface data. */ + isc_result_t result; /*%< Last result code. */ +#ifdef __linux + FILE *proc; + char entry[ISC_IF_INET6_SZ]; + isc_result_t valid; +#endif /* ifdef __linux */ +}; + +isc_result_t +isc_interfaceiter_create(isc_mem_t *mctx, isc_interfaceiter_t **iterp) { + isc_interfaceiter_t *iter; + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(mctx != NULL); + REQUIRE(iterp != NULL); + REQUIRE(*iterp == NULL); + + iter = isc_mem_get(mctx, sizeof(*iter)); + + iter->mctx = mctx; + iter->buf = NULL; + iter->bufsize = 0; + iter->ifaddrs = NULL; +#ifdef __linux + /* + * Only open "/proc/net/if_inet6" if we have never seen a IPv6 + * address returned by getifaddrs(). + */ + if (!seenv6) { + iter->proc = fopen("/proc/net/if_inet6", "r"); + } else { + iter->proc = NULL; + } + iter->valid = ISC_R_FAILURE; +#endif /* ifdef __linux */ + + if (getifaddrs(&iter->ifaddrs) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "getting interface addresses: getifaddrs: %s", + strbuf); + result = ISC_R_UNEXPECTED; + goto failure; + } + + /* + * A newly created iterator has an undefined position + * until isc_interfaceiter_first() is called. + */ + iter->pos = NULL; + iter->result = ISC_R_FAILURE; + + iter->magic = IFITER_MAGIC; + *iterp = iter; + return (ISC_R_SUCCESS); + +failure: +#ifdef __linux + if (iter->proc != NULL) { + fclose(iter->proc); + } +#endif /* ifdef __linux */ + if (iter->ifaddrs != NULL) { /* just in case */ + freeifaddrs(iter->ifaddrs); + } + isc_mem_put(mctx, iter, sizeof(*iter)); + return (result); +} + +/* + * Get information about the current interface to iter->current. + * If successful, return ISC_R_SUCCESS. + * If the interface has an unsupported address family, + * return ISC_R_IGNORE. + */ + +static isc_result_t +internal_current(isc_interfaceiter_t *iter) { + struct ifaddrs *ifa; + int family; + unsigned int namelen; + + REQUIRE(VALID_IFITER(iter)); + + ifa = iter->pos; + +#ifdef __linux + if (iter->pos == NULL) { + return (linux_if_inet6_current(iter)); + } +#endif /* ifdef __linux */ + + INSIST(ifa != NULL); + INSIST(ifa->ifa_name != NULL); + + if (ifa->ifa_addr == NULL) { + return (ISC_R_IGNORE); + } + + family = ifa->ifa_addr->sa_family; + if (family != AF_INET && family != AF_INET6) { + return (ISC_R_IGNORE); + } + +#ifdef __linux + if (family == AF_INET6) { + seenv6 = true; + } +#endif /* ifdef __linux */ + + memset(&iter->current, 0, sizeof(iter->current)); + + namelen = strlen(ifa->ifa_name); + if (namelen > sizeof(iter->current.name) - 1) { + namelen = sizeof(iter->current.name) - 1; + } + + memset(iter->current.name, 0, sizeof(iter->current.name)); + memmove(iter->current.name, ifa->ifa_name, namelen); + + iter->current.flags = 0; + + if ((ifa->ifa_flags & IFF_UP) != 0) { + iter->current.flags |= INTERFACE_F_UP; + } + + if ((ifa->ifa_flags & IFF_POINTOPOINT) != 0) { + iter->current.flags |= INTERFACE_F_POINTTOPOINT; + } + + if ((ifa->ifa_flags & IFF_LOOPBACK) != 0) { + iter->current.flags |= INTERFACE_F_LOOPBACK; + } + + iter->current.af = family; + + get_addr(family, &iter->current.address, ifa->ifa_addr, ifa->ifa_name); + + if (ifa->ifa_netmask != NULL) { + get_addr(family, &iter->current.netmask, ifa->ifa_netmask, + ifa->ifa_name); + } + + if (ifa->ifa_dstaddr != NULL && + (iter->current.flags & INTERFACE_F_POINTTOPOINT) != 0) + { + get_addr(family, &iter->current.dstaddress, ifa->ifa_dstaddr, + ifa->ifa_name); + } + + return (ISC_R_SUCCESS); +} + +/* + * Step the iterator to the next interface. Unlike + * isc_interfaceiter_next(), this may leave the iterator + * positioned on an interface that will ultimately + * be ignored. Return ISC_R_NOMORE if there are no more + * interfaces, otherwise ISC_R_SUCCESS. + */ +static isc_result_t +internal_next(isc_interfaceiter_t *iter) { + if (iter->pos != NULL) { + iter->pos = iter->pos->ifa_next; + } + if (iter->pos == NULL) { +#ifdef __linux + if (!seenv6) { + return (linux_if_inet6_next(iter)); + } +#endif /* ifdef __linux */ + return (ISC_R_NOMORE); + } + + return (ISC_R_SUCCESS); +} + +static void +internal_destroy(isc_interfaceiter_t *iter) { +#ifdef __linux + if (iter->proc != NULL) { + fclose(iter->proc); + } + iter->proc = NULL; +#endif /* ifdef __linux */ + if (iter->ifaddrs) { + freeifaddrs(iter->ifaddrs); + } + iter->ifaddrs = NULL; +} + +static void +internal_first(isc_interfaceiter_t *iter) { +#ifdef __linux + linux_if_inet6_first(iter); +#endif /* ifdef __linux */ + iter->pos = iter->ifaddrs; +} diff --git a/lib/isc/unix/include/.clang-format b/lib/isc/unix/include/.clang-format new file mode 120000 index 0000000..e919bba --- /dev/null +++ b/lib/isc/unix/include/.clang-format @@ -0,0 +1 @@ +../../../../.clang-format.headers
\ No newline at end of file diff --git a/lib/isc/unix/include/Makefile.in b/lib/isc/unix/include/Makefile.in new file mode 100644 index 0000000..60a0a67 --- /dev/null +++ b/lib/isc/unix/include/Makefile.in @@ -0,0 +1,19 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# SPDX-License-Identifier: MPL-2.0 +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, you can obtain one at https://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +SUBDIRS = isc +TARGETS = + +@BIND9_MAKE_RULES@ diff --git a/lib/isc/unix/include/isc/Makefile.in b/lib/isc/unix/include/isc/Makefile.in new file mode 100644 index 0000000..06f7806 --- /dev/null +++ b/lib/isc/unix/include/isc/Makefile.in @@ -0,0 +1,37 @@ +# Copyright (C) Internet Systems Consortium, Inc. ("ISC") +# +# SPDX-License-Identifier: MPL-2.0 +# +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, you can obtain one at https://mozilla.org/MPL/2.0/. +# +# See the COPYRIGHT file distributed with this work for additional +# information regarding copyright ownership. + +srcdir = @srcdir@ +VPATH = @srcdir@ +top_srcdir = @top_srcdir@ + +VERSION=@BIND9_VERSION@ + +HEADERS = align.h dir.h net.h netdb.h offset.h stat.h \ + stdatomic.h stdtime.h syslog.h time.h + +SUBDIRS = +TARGETS = + +@BIND9_MAKE_RULES@ + +installdirs: + $(SHELL) ${top_srcdir}/mkinstalldirs ${DESTDIR}${includedir}/isc + +install:: installdirs + for i in ${HEADERS}; do \ + ${INSTALL_DATA} $(srcdir)/$$i ${DESTDIR}${includedir}/isc || exit 1; \ + done + +uninstall:: + for i in ${HEADERS}; do \ + rm -f ${DESTDIR}${includedir}/isc/$$i || exit 1; \ + done diff --git a/lib/isc/unix/include/isc/align.h b/lib/isc/unix/include/isc/align.h new file mode 100644 index 0000000..7b72e9d --- /dev/null +++ b/lib/isc/unix/include/isc/align.h @@ -0,0 +1,20 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#pragma once + +#ifdef HAVE_STDALIGN_H +#include <stdalign.h> +#else /* ifdef HAVE_STDALIGN_H */ +#define alignas(x) __attribute__((__aligned__(x))) +#endif /* ifdef HAVE_STDALIGN_H */ diff --git a/lib/isc/unix/include/isc/dir.h b/lib/isc/unix/include/isc/dir.h new file mode 100644 index 0000000..89d1338 --- /dev/null +++ b/lib/isc/unix/include/isc/dir.h @@ -0,0 +1,75 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_DIR_H +#define ISC_DIR_H 1 + +/*! \file */ + +#include <dirent.h> + +#include <isc/lang.h> +#include <isc/platform.h> +#include <isc/result.h> + +#include <sys/types.h> /* Required on some systems. */ + +/*% Directory Entry */ +typedef struct isc_direntry { + char name[NAME_MAX]; + unsigned int length; +} isc_direntry_t; + +/*% Directory */ +typedef struct isc_dir { + unsigned int magic; + char dirname[PATH_MAX]; + isc_direntry_t entry; + DIR *handle; +} isc_dir_t; + +ISC_LANG_BEGINDECLS + +void +isc_dir_init(isc_dir_t *dir); + +isc_result_t +isc_dir_open(isc_dir_t *dir, const char *dirname); + +isc_result_t +isc_dir_read(isc_dir_t *dir); + +isc_result_t +isc_dir_reset(isc_dir_t *dir); + +void +isc_dir_close(isc_dir_t *dir); + +isc_result_t +isc_dir_chdir(const char *dirname); + +isc_result_t +isc_dir_chroot(const char *dirname); + +isc_result_t +isc_dir_createunique(char *templet); +/*!< + * Use a templet (such as from isc_file_mktemplate()) to create a uniquely + * named, empty directory. The templet string is modified in place. + * If result == ISC_R_SUCCESS, it is the name of the directory that was + * created. + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_DIR_H */ diff --git a/lib/isc/unix/include/isc/net.h b/lib/isc/unix/include/isc/net.h new file mode 100644 index 0000000..ead9c7f --- /dev/null +++ b/lib/isc/unix/include/isc/net.h @@ -0,0 +1,310 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_NET_H +#define ISC_NET_H 1 + +/***** +***** Module Info +*****/ + +/*! \file + * \brief + * Basic Networking Types + * + * This module is responsible for defining the following basic networking + * types: + * + *\li struct in_addr + *\li struct in6_addr + *\li struct in6_pktinfo + *\li struct sockaddr + *\li struct sockaddr_in + *\li struct sockaddr_in6 + *\li struct sockaddr_storage + *\li in_port_t + * + * It ensures that the AF_ and PF_ macros are defined. + * + * It declares ntoh[sl]() and hton[sl](). + * + * It declares inet_ntop(), and inet_pton(). + * + * It ensures that #INADDR_LOOPBACK, #INADDR_ANY, #IN6ADDR_ANY_INIT, + * IN6ADDR_V4MAPPED_INIT, in6addr_any, and in6addr_loopback are available. + * + * It ensures that IN_MULTICAST() is available to check for multicast + * addresses. + * + * MP: + *\li No impact. + * + * Reliability: + *\li No anticipated impact. + * + * Resources: + *\li N/A. + * + * Security: + *\li No anticipated impact. + * + * Standards: + *\li BSD Socket API + *\li RFC2553 + */ + +/*** + *** Imports. + ***/ +#include <inttypes.h> + +#include <isc/lang.h> +#include <isc/platform.h> +#include <isc/types.h> + +#include <arpa/inet.h> /* Contractual promise. */ +#include <net/if.h> +#include <netinet/in.h> /* Contractual promise. */ +#include <sys/socket.h> /* Contractual promise. */ +#include <sys/types.h> + +#ifndef IN6ADDR_LOOPBACK_INIT +#ifdef s6_addr +/*% IPv6 address loopback init */ +#define IN6ADDR_LOOPBACK_INIT \ + { \ + { \ + { \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1 \ + } \ + } \ + } +#else /* ifdef s6_addr */ +#define IN6ADDR_LOOPBACK_INIT \ + { \ + { \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1 \ + } \ + } +#endif /* ifdef s6_addr */ +#endif /* ifndef IN6ADDR_LOOPBACK_INIT */ + +#ifndef IN6ADDR_V4MAPPED_INIT +#ifdef s6_addr +/*% IPv6 v4mapped prefix init */ +#define IN6ADDR_V4MAPPED_INIT \ + { \ + { \ + { \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xff, 0xff, 0, \ + 0, 0, 0 \ + } \ + } \ + } +#else /* ifdef s6_addr */ +#define IN6ADDR_V4MAPPED_INIT \ + { \ + { \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xff, 0xff, 0, 0, 0, 0 \ + } \ + } +#endif /* ifdef s6_addr */ +#endif /* ifndef IN6ADDR_V4MAPPED_INIT */ + +#ifndef IN6_IS_ADDR_V4MAPPED +/*% Is IPv6 address V4 mapped? */ +#define IN6_IS_ADDR_V4MAPPED(x) \ + (memcmp((x)->s6_addr, in6addr_any.s6_addr, 10) == 0 && \ + (x)->s6_addr[10] == 0xff && (x)->s6_addr[11] == 0xff) +#endif /* ifndef IN6_IS_ADDR_V4MAPPED */ + +#ifndef IN6_IS_ADDR_V4COMPAT +/*% Is IPv6 address V4 compatible? */ +#define IN6_IS_ADDR_V4COMPAT(x) \ + (memcmp((x)->s6_addr, in6addr_any.s6_addr, 12) == 0 && \ + ((x)->s6_addr[12] != 0 || (x)->s6_addr[13] != 0 || \ + (x)->s6_addr[14] != 0 || \ + ((x)->s6_addr[15] != 0 && (x)->s6_addr[15] != 1))) +#endif /* ifndef IN6_IS_ADDR_V4COMPAT */ + +#ifndef IN6_IS_ADDR_MULTICAST +/*% Is IPv6 address multicast? */ +#define IN6_IS_ADDR_MULTICAST(a) ((a)->s6_addr[0] == 0xff) +#endif /* ifndef IN6_IS_ADDR_MULTICAST */ + +#ifndef IN6_IS_ADDR_LINKLOCAL +/*% Is IPv6 address linklocal? */ +#define IN6_IS_ADDR_LINKLOCAL(a) \ + (((a)->s6_addr[0] == 0xfe) && (((a)->s6_addr[1] & 0xc0) == 0x80)) +#endif /* ifndef IN6_IS_ADDR_LINKLOCAL */ + +#ifndef IN6_IS_ADDR_SITELOCAL +/*% is IPv6 address sitelocal? */ +#define IN6_IS_ADDR_SITELOCAL(a) \ + (((a)->s6_addr[0] == 0xfe) && (((a)->s6_addr[1] & 0xc0) == 0xc0)) +#endif /* ifndef IN6_IS_ADDR_SITELOCAL */ + +#ifndef IN6_IS_ADDR_LOOPBACK +/*% is IPv6 address loopback? */ +#define IN6_IS_ADDR_LOOPBACK(x) \ + (memcmp((x)->s6_addr, in6addr_loopback.s6_addr, 16) == 0) +#endif /* ifndef IN6_IS_ADDR_LOOPBACK */ + +#ifndef AF_INET6 +/*% IPv6 */ +#define AF_INET6 99 +#endif /* ifndef AF_INET6 */ + +#ifndef PF_INET6 +/*% IPv6 */ +#define PF_INET6 AF_INET6 +#endif /* ifndef PF_INET6 */ + +#ifndef INADDR_ANY +/*% inaddr any */ +#define INADDR_ANY 0x00000000UL +#endif /* ifndef INADDR_ANY */ + +#ifndef INADDR_LOOPBACK +/*% inaddr loopback */ +#define INADDR_LOOPBACK 0x7f000001UL +#endif /* ifndef INADDR_LOOPBACK */ + +#ifndef MSG_TRUNC +/*% + * If this system does not have MSG_TRUNC (as returned from recvmsg()) + * ISC_PLATFORM_RECVOVERFLOW will be defined. This will enable the MSG_TRUNC + * faking code in socket.c. + */ +#define ISC_PLATFORM_RECVOVERFLOW +#endif /* ifndef MSG_TRUNC */ + +/*% IP address. */ +#define ISC__IPADDR(x) ((uint32_t)htonl((uint32_t)(x))) + +/*% Is IP address multicast? */ +#define ISC_IPADDR_ISMULTICAST(i) \ + (((uint32_t)(i)&ISC__IPADDR(0xf0000000)) == ISC__IPADDR(0xe0000000)) + +#define ISC_IPADDR_ISEXPERIMENTAL(i) \ + (((uint32_t)(i)&ISC__IPADDR(0xf0000000)) == ISC__IPADDR(0xf0000000)) + +/*** + *** Functions. + ***/ + +ISC_LANG_BEGINDECLS + +isc_result_t +isc_net_probeipv4(void); +/*%< + * Check if the system's kernel supports IPv4. + * + * Returns: + * + *\li #ISC_R_SUCCESS IPv4 is supported. + *\li #ISC_R_NOTFOUND IPv4 is not supported. + *\li #ISC_R_DISABLED IPv4 is disabled. + *\li #ISC_R_UNEXPECTED + */ + +isc_result_t +isc_net_probeipv6(void); +/*%< + * Check if the system's kernel supports IPv6. + * + * Returns: + * + *\li #ISC_R_SUCCESS IPv6 is supported. + *\li #ISC_R_NOTFOUND IPv6 is not supported. + *\li #ISC_R_DISABLED IPv6 is disabled. + *\li #ISC_R_UNEXPECTED + */ + +isc_result_t +isc_net_probe_ipv6only(void); +/*%< + * Check if the system's kernel supports the IPV6_V6ONLY socket option. + * + * Returns: + * + *\li #ISC_R_SUCCESS the option is supported for both TCP and UDP. + *\li #ISC_R_NOTFOUND IPv6 itself or the option is not supported. + *\li #ISC_R_UNEXPECTED + */ + +isc_result_t +isc_net_probe_ipv6pktinfo(void); +/* + * Check if the system's kernel supports the IPV6_(RECV)PKTINFO socket option + * for UDP sockets. + * + * Returns: + * + * \li #ISC_R_SUCCESS the option is supported. + * \li #ISC_R_NOTFOUND IPv6 itself or the option is not supported. + * \li #ISC_R_UNEXPECTED + */ + +void +isc_net_disableipv4(void); + +void +isc_net_disableipv6(void); + +void +isc_net_enableipv4(void); + +void +isc_net_enableipv6(void); + +isc_result_t +isc_net_probeunix(void); +/* + * Returns whether UNIX domain sockets are supported. + */ + +#define ISC_NET_DSCPRECVV4 0x01 /* Can receive sent DSCP value IPv4 */ +#define ISC_NET_DSCPRECVV6 0x02 /* Can receive sent DSCP value IPv6 */ +#define ISC_NET_DSCPSETV4 0x04 /* Can set DSCP on socket IPv4 */ +#define ISC_NET_DSCPSETV6 0x08 /* Can set DSCP on socket IPv6 */ +#define ISC_NET_DSCPPKTV4 0x10 /* Can set DSCP on per packet IPv4 */ +#define ISC_NET_DSCPPKTV6 0x20 /* Can set DSCP on per packet IPv6 */ +#define ISC_NET_DSCPALL 0x3f /* All valid flags */ + +unsigned int +isc_net_probedscp(void); +/*%< + * Probe the level of DSCP support. + */ + +isc_result_t +isc_net_getudpportrange(int af, in_port_t *low, in_port_t *high); +/*%< + * Returns system's default range of ephemeral UDP ports, if defined. + * If the range is not available or unknown, ISC_NET_PORTRANGELOW and + * ISC_NET_PORTRANGEHIGH will be returned. + * + * Requires: + * + *\li 'low' and 'high' must be non NULL. + * + * Returns: + * + *\li *low and *high will be the ports specifying the low and high ends of + * the range. + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_NET_H */ diff --git a/lib/isc/unix/include/isc/netdb.h b/lib/isc/unix/include/isc/netdb.h new file mode 100644 index 0000000..dab845a --- /dev/null +++ b/lib/isc/unix/include/isc/netdb.h @@ -0,0 +1,51 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_NETDB_H +#define ISC_NETDB_H 1 + +/***** +***** Module Info +*****/ + +/*! \file + * \brief + * Portable netdb.h support. + * + * This module is responsible for defining the get<x>by<y> APIs. + * + * MP: + *\li No impact. + * + * Reliability: + *\li No anticipated impact. + * + * Resources: + *\li N/A. + * + * Security: + *\li No anticipated impact. + * + * Standards: + *\li BSD API + */ + +/*** + *** Imports. + ***/ + +#include <netdb.h> + +#include <isc/net.h> + +#endif /* ISC_NETDB_H */ diff --git a/lib/isc/unix/include/isc/offset.h b/lib/isc/unix/include/isc/offset.h new file mode 100644 index 0000000..fa24ae4 --- /dev/null +++ b/lib/isc/unix/include/isc/offset.h @@ -0,0 +1,28 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_OFFSET_H +#define ISC_OFFSET_H 1 + +/*! \file + * \brief + * File offsets are operating-system dependent. + */ +#include <limits.h> /* Required for CHAR_BIT. */ +#include <stddef.h> /* For Linux Standard Base. */ + +#include <sys/types.h> + +typedef off_t isc_offset_t; + +#endif /* ISC_OFFSET_H */ diff --git a/lib/isc/unix/include/isc/stat.h b/lib/isc/unix/include/isc/stat.h new file mode 100644 index 0000000..6ddddd7 --- /dev/null +++ b/lib/isc/unix/include/isc/stat.h @@ -0,0 +1,47 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_STAT_H +#define ISC_STAT_H 1 + +/***** +***** Module Info +*****/ + +/* + * Portable <sys/stat.h> support. + * + * This module is responsible for defining S_IS??? macros. + * + * MP: + * No impact. + * + * Reliability: + * No anticipated impact. + * + * Resources: + * N/A. + * + * Security: + * No anticipated impact. + * + */ + +/*** + *** Imports. + ***/ + +#include <sys/stat.h> +#include <sys/types.h> + +#endif /* ISC_STAT_H */ diff --git a/lib/isc/unix/include/isc/stdatomic.h b/lib/isc/unix/include/isc/stdatomic.h new file mode 100644 index 0000000..f071b6a --- /dev/null +++ b/lib/isc/unix/include/isc/stdatomic.h @@ -0,0 +1,224 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#pragma once + +#include <inttypes.h> +#include <stdbool.h> +#include <stddef.h> +#if HAVE_UCHAR_H +#include <uchar.h> +#endif /* HAVE_UCHAR_H */ + +#if !defined(__has_feature) +#define __has_feature(x) 0 +#endif /* if !defined(__has_feature) */ + +#if !defined(__has_extension) +#define __has_extension(x) __has_feature(x) +#endif /* if !defined(__has_extension) */ + +#if !defined(__GNUC_PREREQ__) +#if defined(__GNUC__) && defined(__GNUC_MINOR__) +#define __GNUC_PREREQ__(maj, min) \ + ((__GNUC__ << 16) + __GNUC_MINOR__ >= ((maj) << 16) + (min)) +#else /* if defined(__GNUC__) && defined(__GNUC_MINOR__) */ +#define __GNUC_PREREQ__(maj, min) 0 +#endif /* if defined(__GNUC__) && defined(__GNUC_MINOR__) */ +#endif /* if !defined(__GNUC_PREREQ__) */ + +#if !defined(__CLANG_ATOMICS) && !defined(__GNUC_ATOMICS) +#if __has_extension(c_atomic) || __has_extension(cxx_atomic) +#define __CLANG_ATOMICS +#elif __GNUC_PREREQ__(4, 7) +#define __GNUC_ATOMICS +#elif !defined(__GNUC__) +#error "isc/stdatomic.h does not support your compiler" +#endif /* if __has_extension(c_atomic) || __has_extension(cxx_atomic) */ +#endif /* if !defined(__CLANG_ATOMICS) && !defined(__GNUC_ATOMICS) */ + +#ifndef __ATOMIC_RELAXED +#define __ATOMIC_RELAXED 0 +#endif /* ifndef __ATOMIC_RELAXED */ +#ifndef __ATOMIC_CONSUME +#define __ATOMIC_CONSUME 1 +#endif /* ifndef __ATOMIC_CONSUME */ +#ifndef __ATOMIC_ACQUIRE +#define __ATOMIC_ACQUIRE 2 +#endif /* ifndef __ATOMIC_ACQUIRE */ +#ifndef __ATOMIC_RELEASE +#define __ATOMIC_RELEASE 3 +#endif /* ifndef __ATOMIC_RELEASE */ +#ifndef __ATOMIC_ACQ_REL +#define __ATOMIC_ACQ_REL 4 +#endif /* ifndef __ATOMIC_ACQ_REL */ +#ifndef __ATOMIC_SEQ_CST +#define __ATOMIC_SEQ_CST 5 +#endif /* ifndef __ATOMIC_SEQ_CST */ + +enum memory_order { + memory_order_relaxed = __ATOMIC_RELAXED, + memory_order_consume = __ATOMIC_CONSUME, + memory_order_acquire = __ATOMIC_ACQUIRE, + memory_order_release = __ATOMIC_RELEASE, + memory_order_acq_rel = __ATOMIC_ACQ_REL, + memory_order_seq_cst = __ATOMIC_SEQ_CST +}; + +typedef enum memory_order memory_order; + +#ifndef HAVE_UCHAR_H +typedef uint_least16_t char16_t; +typedef uint_least32_t char32_t; +#endif /* HAVE_UCHAR_H */ + +typedef bool atomic_bool; +typedef char atomic_char; +typedef signed char atomic_schar; +typedef unsigned char atomic_uchar; +typedef short atomic_short; +typedef unsigned short atomic_ushort; +typedef int atomic_int; +typedef unsigned int atomic_uint; +typedef long atomic_long; +typedef unsigned long atomic_ulong; +typedef long long atomic_llong; +typedef unsigned long long atomic_ullong; +typedef char16_t atomic_char16_t; +typedef char32_t atomic_char32_t; +typedef wchar_t atomic_wchar_t; +typedef int_least8_t atomic_int_least8_t; +typedef uint_least8_t atomic_uint_least8_t; +typedef int_least16_t atomic_int_least16_t; +typedef uint_least16_t atomic_uint_least16_t; +typedef int_least32_t atomic_int_least32_t; +typedef uint_least32_t atomic_uint_least32_t; +typedef int_least64_t atomic_int_least64_t; +typedef uint_least64_t atomic_uint_least64_t; +typedef int_fast8_t atomic_int_fast8_t; +typedef uint_fast8_t atomic_uint_fast8_t; +typedef int_fast16_t atomic_int_fast16_t; +typedef uint_fast16_t atomic_uint_fast16_t; +typedef int_fast32_t atomic_int_fast32_t; +typedef uint_fast32_t atomic_uint_fast32_t; +typedef int_fast64_t atomic_int_fast64_t; +typedef uint_fast64_t atomic_uint_fast64_t; +typedef intptr_t atomic_intptr_t; +typedef uintptr_t atomic_uintptr_t; +typedef size_t atomic_size_t; +typedef ptrdiff_t atomic_ptrdiff_t; +typedef intmax_t atomic_intmax_t; +typedef uintmax_t atomic_uintmax_t; + +#if defined(__CLANG_ATOMICS) /* __c11_atomic builtins */ +#define atomic_init(obj, desired) __c11_atomic_init(obj, desired) +#define atomic_load_explicit(obj, order) __c11_atomic_load(obj, order) +#define atomic_store_explicit(obj, desired, order) \ + __c11_atomic_store(obj, desired, order) +#define atomic_fetch_add_explicit(obj, arg, order) \ + __c11_atomic_fetch_add(obj, arg, order) +#define atomic_fetch_sub_explicit(obj, arg, order) \ + __c11_atomic_fetch_sub(obj, arg, order) +#define atomic_fetch_and_explicit(obj, arg, order) \ + __c11_atomic_fetch_and(obj, arg, order) +#define atomic_fetch_or_explicit(obj, arg, order) \ + __c11_atomic_fetch_or(obj, arg, order) +#define atomic_compare_exchange_strong_explicit(obj, expected, desired, succ, \ + fail) \ + __c11_atomic_compare_exchange_strong_explicit(obj, expected, desired, \ + succ, fail) +#define atomic_compare_exchange_weak_explicit(obj, expected, desired, succ, \ + fail) \ + __c11_atomic_compare_exchange_weak_explicit(obj, expected, desired, \ + succ, fail) +#define atomic_exchange_explicit(obj, desired, order) \ + __c11_atomic_exchange_explicit(obj, expected, order) +#elif defined(__GNUC_ATOMICS) /* __atomic builtins */ +#define atomic_init(obj, desired) (*obj = desired) +#define atomic_load_explicit(obj, order) __atomic_load_n(obj, order) +#define atomic_store_explicit(obj, desired, order) \ + __atomic_store_n(obj, desired, order) +#define atomic_fetch_add_explicit(obj, arg, order) \ + __atomic_fetch_add(obj, arg, order) +#define atomic_fetch_sub_explicit(obj, arg, order) \ + __atomic_fetch_sub(obj, arg, order) +#define atomic_fetch_and_explicit(obj, arg, order) \ + __atomic_fetch_and(obj, arg, order) +#define atomic_fetch_or_explicit(obj, arg, order) \ + __atomic_fetch_or(obj, arg, order) +#define atomic_compare_exchange_strong_explicit(obj, expected, desired, succ, \ + fail) \ + __atomic_compare_exchange_n(obj, expected, desired, 0, succ, fail) +#define atomic_compare_exchange_weak_explicit(obj, expected, desired, succ, \ + fail) \ + __atomic_compare_exchange_n(obj, expected, desired, 1, succ, fail) +#define atomic_exchange_explicit(obj, desired, order) \ + __atomic_exchange_n(obj, desired, order) +#else /* __sync builtins */ +#define atomic_init(obj, desired) (*obj = desired) +#define atomic_load_explicit(obj, order) __sync_fetch_and_add(obj, 0) +#define atomic_store_explicit(obj, desired, order) \ + do { \ + __sync_synchronize(); \ + *obj = desired; \ + __sync_synchronize(); \ + } while (0); +#define atomic_fetch_add_explicit(obj, arg, order) \ + __sync_fetch_and_add(obj, arg) +#define atomic_fetch_sub_explicit(obj, arg, order) \ + __sync_fetch_and_sub(obj, arg, order) +#define atomic_fetch_and_explicit(obj, arg, order) \ + __sync_fetch_and_and(obj, arg, order) +#define atomic_fetch_or_explicit(obj, arg, order) \ + __sync_fetch_and_or(obj, arg, order) +#define atomic_compare_exchange_strong_explicit(obj, expected, desired, succ, \ + fail) \ + ({ \ + __typeof__(obj) __v; \ + _Bool __r; \ + __v = (__typeof__(obj))__sync_val_compare_and_swap( \ + obj, *(expected), desired); \ + __r = ((__typeof__(obj))*(expected) == __v); \ + *(expected) = __v; \ + __r; \ + }) +#define atomic_compare_exchange_weak_explicit(obj, expected, desired, succ, \ + fail) \ + atomic_compare_exchange_strong_explicit(obj, expected, desired, succ, \ + fail) +#define atomic_exchange_explicit(obj, desired, order) \ + __sync_lock_test_and_set(obj, desired) + +#endif /* if defined(__CLANG_ATOMICS) */ + +#define atomic_load(obj) atomic_load_explicit(obj, memory_order_seq_cst) +#define atomic_store(obj, arg) \ + atomic_store_explicit(obj, arg, memory_order_seq_cst) +#define atomic_fetch_add(obj, arg) \ + atomic_fetch_add_explicit(obj, arg, memory_order_seq_cst) +#define atomic_fetch_sub(obj, arg) \ + atomic_fetch_sub_explicit(obj, arg, memory_order_seq_cst) +#define atomic_fetch_and(obj, arg) \ + atomic_fetch_and_explicit(obj, arg, memory_order_seq_cst) +#define atomic_fetch_or(obj, arg) \ + atomic_fetch_or_explicit(obj, arg, memory_order_seq_cst) +#define atomic_compare_exchange_strong(obj, expected, desired) \ + atomic_compare_exchange_strong_explicit(obj, expected, desired, \ + memory_order_seq_cst, \ + memory_order_seq_cst) +#define atomic_compare_exchange_weak(obj, expected, desired) \ + atomic_compare_exchange_weak_explicit(obj, expected, desired, \ + memory_order_seq_cst, \ + memory_order_seq_cst) +#define atomic_exchange(obj, desired) \ + atomic_exchange_explicit(obj, desired, memory_order_seq_cst) diff --git a/lib/isc/unix/include/isc/stdtime.h b/lib/isc/unix/include/isc/stdtime.h new file mode 100644 index 0000000..e38a2c6 --- /dev/null +++ b/lib/isc/unix/include/isc/stdtime.h @@ -0,0 +1,64 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_STDTIME_H +#define ISC_STDTIME_H 1 + +/*! \file */ + +#include <inttypes.h> +#include <stdlib.h> + +#include <isc/lang.h> + +/*% + * It's public information that 'isc_stdtime_t' is an unsigned integral type. + * Applications that want maximum portability should not assume anything + * about its size. + */ +typedef uint32_t isc_stdtime_t; + +ISC_LANG_BEGINDECLS +/* */ +void +isc_stdtime_get(isc_stdtime_t *t); +/*%< + * Set 't' to the number of seconds since 00:00:00 UTC, January 1, 1970. + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +void +isc_stdtime_tostring(isc_stdtime_t t, char *out, size_t outlen); +/* + * Convert 't' into a null-terminated string of the form + * "Wed Jun 30 21:49:08 1993". Store the string in the 'out' + * buffer. + * + * Requires: + * + * 't' is a valid time. + * 'out' is a valid pointer. + * 'outlen' is at least 26. + */ + +#define isc_stdtime_convert32(t, t32p) (*(t32p) = t) +/* + * Convert the standard time to its 32-bit version. + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_STDTIME_H */ diff --git a/lib/isc/unix/include/isc/syslog.h b/lib/isc/unix/include/isc/syslog.h new file mode 100644 index 0000000..f46dc1a --- /dev/null +++ b/lib/isc/unix/include/isc/syslog.h @@ -0,0 +1,41 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_SYSLOG_H +#define ISC_SYSLOG_H 1 + +/*! \file */ + +#include <isc/lang.h> +#include <isc/types.h> + +ISC_LANG_BEGINDECLS + +isc_result_t +isc_syslog_facilityfromstring(const char *str, int *facilityp); +/*%< + * Convert 'str' to the appropriate syslog facility constant. + * + * Requires: + * + *\li 'str' is not NULL + *\li 'facilityp' is not NULL + * + * Returns: + * \li #ISC_R_SUCCESS + * \li #ISC_R_NOTFOUND + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_SYSLOG_H */ diff --git a/lib/isc/unix/include/isc/time.h b/lib/isc/unix/include/isc/time.h new file mode 100644 index 0000000..ecec64d --- /dev/null +++ b/lib/isc/unix/include/isc/time.h @@ -0,0 +1,452 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_TIME_H +#define ISC_TIME_H 1 + +/*! \file */ + +#include <inttypes.h> +#include <stdbool.h> + +#include <isc/lang.h> +#include <isc/types.h> + +/*** + *** Intervals + ***/ + +/*! + * \brief + * The contents of this structure are private, and MUST NOT be accessed + * directly by callers. + * + * The contents are exposed only to allow callers to avoid dynamic allocation. + */ +struct isc_interval { + unsigned int seconds; + unsigned int nanoseconds; +}; + +extern const isc_interval_t *const isc_interval_zero; + +/* + * ISC_FORMATHTTPTIMESTAMP_SIZE needs to be 30 in C locale and potentially + * more for other locales to handle longer national abbreviations when + * expanding strftime's %a and %b. + */ +#define ISC_FORMATHTTPTIMESTAMP_SIZE 50 + +ISC_LANG_BEGINDECLS + +void +isc_interval_set(isc_interval_t *i, unsigned int seconds, + unsigned int nanoseconds); +/*%< + * Set 'i' to a value representing an interval of 'seconds' seconds and + * 'nanoseconds' nanoseconds, suitable for use in isc_time_add() and + * isc_time_subtract(). + * + * Requires: + * + *\li 't' is a valid pointer. + *\li nanoseconds < 1000000000. + */ + +bool +isc_interval_iszero(const isc_interval_t *i); +/*%< + * Returns true iff. 'i' is the zero interval. + * + * Requires: + * + *\li 'i' is a valid pointer. + */ + +/*** + *** Absolute Times + ***/ + +/*% + * The contents of this structure are private, and MUST NOT be accessed + * directly by callers. + * + * The contents are exposed only to allow callers to avoid dynamic allocation. + */ + +struct isc_time { + unsigned int seconds; + unsigned int nanoseconds; +}; + +extern const isc_time_t *const isc_time_epoch; + +void +isc_time_set(isc_time_t *t, unsigned int seconds, unsigned int nanoseconds); +/*%< + * Set 't' to a value which represents the given number of seconds and + * nanoseconds since 00:00:00 January 1, 1970, UTC. + * + * Notes: + *\li The Unix version of this call is equivalent to: + *\code + * isc_time_settoepoch(t); + * isc_interval_set(i, seconds, nanoseconds); + * isc_time_add(t, i, t); + *\endcode + * + * Requires: + *\li 't' is a valid pointer. + *\li nanoseconds < 1000000000. + */ + +void +isc_time_settoepoch(isc_time_t *t); +/*%< + * Set 't' to the time of the epoch. + * + * Notes: + *\li The date of the epoch is platform-dependent. + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +bool +isc_time_isepoch(const isc_time_t *t); +/*%< + * Returns true iff. 't' is the epoch ("time zero"). + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +isc_result_t +isc_time_now(isc_time_t *t); +/*%< + * Set 't' to the current absolute time. + * + * Requires: + * + *\li 't' is a valid pointer. + * + * Returns: + * + *\li Success + *\li Unexpected error + * Getting the time from the system failed. + *\li Out of range + * The time from the system is too large to be represented + * in the current definition of isc_time_t. + */ + +isc_result_t +isc_time_now_hires(isc_time_t *t); +/*%< + * Set 't' to the current absolute time. Uses higher resolution clocks + * recommended when microsecond accuracy is required. + * + * Requires: + * + *\li 't' is a valid pointer. + * + * Returns: + * + *\li Success + *\li Unexpected error + * Getting the time from the system failed. + *\li Out of range + * The time from the system is too large to be represented + * in the current definition of isc_time_t. + */ + +isc_result_t +isc_time_nowplusinterval(isc_time_t *t, const isc_interval_t *i); +/*%< + * Set *t to the current absolute time + i. + * + * Note: + *\li This call is equivalent to: + * + *\code + * isc_time_now(t); + * isc_time_add(t, i, t); + *\endcode + * + * Requires: + * + *\li 't' and 'i' are valid pointers. + * + * Returns: + * + *\li Success + *\li Unexpected error + * Getting the time from the system failed. + *\li Out of range + * The interval added to the time from the system is too large to + * be represented in the current definition of isc_time_t. + */ + +int +isc_time_compare(const isc_time_t *t1, const isc_time_t *t2); +/*%< + * Compare the times referenced by 't1' and 't2' + * + * Requires: + * + *\li 't1' and 't2' are valid pointers. + * + * Returns: + * + *\li -1 t1 < t2 (comparing times, not pointers) + *\li 0 t1 = t2 + *\li 1 t1 > t2 + */ + +isc_result_t +isc_time_add(const isc_time_t *t, const isc_interval_t *i, isc_time_t *result); +/*%< + * Add 'i' to 't', storing the result in 'result'. + * + * Requires: + * + *\li 't', 'i', and 'result' are valid pointers. + * + * Returns: + *\li Success + *\li Out of range + * The interval added to the time is too large to + * be represented in the current definition of isc_time_t. + */ + +isc_result_t +isc_time_subtract(const isc_time_t *t, const isc_interval_t *i, + isc_time_t *result); +/*%< + * Subtract 'i' from 't', storing the result in 'result'. + * + * Requires: + * + *\li 't', 'i', and 'result' are valid pointers. + * + * Returns: + *\li Success + *\li Out of range + * The interval is larger than the time since the epoch. + */ + +uint64_t +isc_time_microdiff(const isc_time_t *t1, const isc_time_t *t2); +/*%< + * Find the difference in microseconds between time t1 and time t2. + * t2 is the subtrahend of t1; ie, difference = t1 - t2. + * + * Requires: + * + *\li 't1' and 't2' are valid pointers. + * + * Returns: + *\li The difference of t1 - t2, or 0 if t1 <= t2. + */ + +uint32_t +isc_time_seconds(const isc_time_t *t); +/*%< + * Return the number of seconds since the epoch stored in a time structure. + * + * Requires: + * + *\li 't' is a valid pointer. + */ + +isc_result_t +isc_time_secondsastimet(const isc_time_t *t, time_t *secondsp); +/*%< + * Ensure the number of seconds in an isc_time_t is representable by a time_t. + * + * Notes: + *\li The number of seconds stored in an isc_time_t might be larger + * than the number of seconds a time_t is able to handle. Since + * time_t is mostly opaque according to the ANSI/ISO standard + * (essentially, all you can be sure of is that it is an arithmetic type, + * not even necessarily integral), it can be tricky to ensure that + * the isc_time_t is in the range a time_t can handle. Use this + * function in place of isc_time_seconds() any time you need to set a + * time_t from an isc_time_t. + * + * Requires: + *\li 't' is a valid pointer. + * + * Returns: + *\li Success + *\li Out of range + */ + +uint32_t +isc_time_nanoseconds(const isc_time_t *t); +/*%< + * Return the number of nanoseconds stored in a time structure. + * + * Notes: + *\li This is the number of nanoseconds in excess of the number + * of seconds since the epoch; it will always be less than one + * full second. + * + * Requires: + *\li 't' is a valid pointer. + * + * Ensures: + *\li The returned value is less than 1*10^9. + */ + +void +isc_time_formattimestamp(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using a format like "30-Aug-2000 04:06:47.997" and the local time zone. + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formathttptimestamp(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using a format like "Mon, 30 Aug 2000 04:06:47 GMT" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +isc_result_t +isc_time_parsehttptimestamp(char *input, isc_time_t *t); +/*%< + * Parse the time in 'input' into the isc_time_t pointed to by 't', + * expecting a format like "Mon, 30 Aug 2000 04:06:47 GMT" + * + * Requires: + *\li 'buf' and 't' are not NULL. + */ + +void +isc_time_formatISO8601L(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ss" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatISO8601Lms(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ss.sss" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatISO8601Lus(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ss.ssssss" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatISO8601(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ssZ" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatISO8601ms(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ss.sssZ" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatISO8601us(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the ISO8601 format: "yyyy-mm-ddThh:mm:ss.ssssssZ" + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +void +isc_time_formatshorttimestamp(const isc_time_t *t, char *buf, unsigned int len); +/*%< + * Format the time 't' into the buffer 'buf' of length 'len', + * using the format "yyyymmddhhmmsssss" useful for file timestamping. + * If the text does not fit in the buffer, the result is indeterminate, + * but is always guaranteed to be null terminated. + * + * Requires: + *\li 'len' > 0 + *\li 'buf' points to an array of at least len chars + * + */ + +ISC_LANG_ENDDECLS + +#endif /* ISC_TIME_H */ diff --git a/lib/isc/unix/interfaceiter.c b/lib/isc/unix/interfaceiter.c new file mode 100644 index 0000000..a1f6c9d --- /dev/null +++ b/lib/isc/unix/interfaceiter.c @@ -0,0 +1,301 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <sys/ioctl.h> +#include <sys/types.h> +#ifdef HAVE_SYS_SOCKIO_H +#include <sys/sockio.h> /* Required for ifiter_ioctl.c. */ +#endif /* ifdef HAVE_SYS_SOCKIO_H */ + +#include <errno.h> +#include <inttypes.h> +#include <stdio.h> +#include <stdlib.h> +#include <unistd.h> + +#include <isc/interfaceiter.h> +#include <isc/log.h> +#include <isc/magic.h> +#include <isc/mem.h> +#include <isc/net.h> +#include <isc/print.h> +#include <isc/result.h> +#include <isc/strerr.h> +#include <isc/string.h> +#include <isc/types.h> +#include <isc/util.h> + +/* Must follow <isc/net.h>. */ +#ifdef HAVE_NET_IF6_H +#include <net/if6.h> +#endif /* ifdef HAVE_NET_IF6_H */ +#include <net/if.h> + +/* Common utility functions */ + +/*% + * Extract the network address part from a "struct sockaddr". + * \brief + * The address family is given explicitly + * instead of using src->sa_family, because the latter does not work + * for copying a network mask obtained by SIOCGIFNETMASK (it does + * not have a valid address family). + */ + +static void +get_addr(unsigned int family, isc_netaddr_t *dst, struct sockaddr *src, + char *ifname) { + struct sockaddr_in6 *sa6; + +#if !defined(HAVE_IF_NAMETOINDEX) + UNUSED(ifname); +#endif /* if !defined(HAVE_IF_NAMETOINDEX) */ + + /* clear any remaining value for safety */ + memset(dst, 0, sizeof(*dst)); + + dst->family = family; + switch (family) { + case AF_INET: + memmove(&dst->type.in, &((struct sockaddr_in *)src)->sin_addr, + sizeof(struct in_addr)); + break; + case AF_INET6: + sa6 = (struct sockaddr_in6 *)src; + memmove(&dst->type.in6, &sa6->sin6_addr, + sizeof(struct in6_addr)); + if (sa6->sin6_scope_id != 0) { + isc_netaddr_setzone(dst, sa6->sin6_scope_id); + } else { + /* + * BSD variants embed scope zone IDs in the 128bit + * address as a kernel internal form. Unfortunately, + * the embedded IDs are not hidden from applications + * when getting access to them by sysctl or ioctl. + * We convert the internal format to the pure address + * part and the zone ID part. + * Since multicast addresses should not appear here + * and they cannot be distinguished from netmasks, + * we only consider unicast link-local addresses. + */ + if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) { + uint16_t zone16; + + memmove(&zone16, &sa6->sin6_addr.s6_addr[2], + sizeof(zone16)); + zone16 = ntohs(zone16); + if (zone16 != 0) { + /* the zone ID is embedded */ + isc_netaddr_setzone(dst, + (uint32_t)zone16); + dst->type.in6.s6_addr[2] = 0; + dst->type.in6.s6_addr[3] = 0; +#ifdef HAVE_IF_NAMETOINDEX + } else if (ifname != NULL) { + unsigned int zone; + + /* + * sin6_scope_id is still not provided, + * but the corresponding interface name + * is know. Use the interface ID as + * the link ID. + */ + zone = if_nametoindex(ifname); + if (zone != 0) { + isc_netaddr_setzone( + dst, (uint32_t)zone); + } +#endif /* ifdef HAVE_IF_NAMETOINDEX */ + } + } + } + break; + default: + UNREACHABLE(); + } +} + +/* + * Include system-dependent code. + */ + +#ifdef __linux +#define ISC_IF_INET6_SZ \ + sizeof("00000000000000000000000000000001 01 80 10 80 " \ + "XXXXXXloXXXXXXXX\n") +static isc_result_t +linux_if_inet6_next(isc_interfaceiter_t *); +static isc_result_t +linux_if_inet6_current(isc_interfaceiter_t *); +static void +linux_if_inet6_first(isc_interfaceiter_t *iter); +#endif /* ifdef __linux */ + +#include "ifiter_getifaddrs.c" + +#ifdef __linux +static void +linux_if_inet6_first(isc_interfaceiter_t *iter) { + if (iter->proc != NULL) { + rewind(iter->proc); + (void)linux_if_inet6_next(iter); + } else { + iter->valid = ISC_R_NOMORE; + } +} + +static isc_result_t +linux_if_inet6_next(isc_interfaceiter_t *iter) { + if (iter->proc != NULL && + fgets(iter->entry, sizeof(iter->entry), iter->proc) != NULL) + { + iter->valid = ISC_R_SUCCESS; + } else { + iter->valid = ISC_R_NOMORE; + } + return (iter->valid); +} + +static isc_result_t +linux_if_inet6_current(isc_interfaceiter_t *iter) { + char address[33]; + char name[IF_NAMESIZE + 1]; + struct in6_addr addr6; + unsigned int ifindex, prefix, flag3, flag4; + int res; + unsigned int i; + + if (iter->valid != ISC_R_SUCCESS) { + return (iter->valid); + } + if (iter->proc == NULL) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, ISC_LOG_ERROR, + "/proc/net/if_inet6:iter->proc == NULL"); + return (ISC_R_FAILURE); + } + + res = sscanf(iter->entry, "%32[a-f0-9] %x %x %x %x %16s\n", address, + &ifindex, &prefix, &flag3, &flag4, name); + if (res != 6) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, ISC_LOG_ERROR, + "/proc/net/if_inet6:sscanf() -> %d (expected 6)", + res); + return (ISC_R_FAILURE); + } + if (strlen(address) != 32) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_INTERFACE, ISC_LOG_ERROR, + "/proc/net/if_inet6:strlen(%s) != 32", address); + return (ISC_R_FAILURE); + } + for (i = 0; i < 16; i++) { + unsigned char byte; + static const char hex[] = "0123456789abcdef"; + byte = ((strchr(hex, address[i * 2]) - hex) << 4) | + (strchr(hex, address[i * 2 + 1]) - hex); + addr6.s6_addr[i] = byte; + } + iter->current.af = AF_INET6; + iter->current.flags = INTERFACE_F_UP; + isc_netaddr_fromin6(&iter->current.address, &addr6); + if (isc_netaddr_islinklocal(&iter->current.address)) { + isc_netaddr_setzone(&iter->current.address, (uint32_t)ifindex); + } + for (i = 0; i < 16; i++) { + if (prefix > 8) { + addr6.s6_addr[i] = 0xff; + prefix -= 8; + } else { + addr6.s6_addr[i] = (0xff << (8 - prefix)) & 0xff; + prefix = 0; + } + } + isc_netaddr_fromin6(&iter->current.netmask, &addr6); + strlcpy(iter->current.name, name, sizeof(iter->current.name)); + return (ISC_R_SUCCESS); +} +#endif /* ifdef __linux */ + +/* + * The remaining code is common to the sysctl and ioctl case. + */ + +isc_result_t +isc_interfaceiter_current(isc_interfaceiter_t *iter, isc_interface_t *ifdata) { + REQUIRE(iter->result == ISC_R_SUCCESS); + memmove(ifdata, &iter->current, sizeof(*ifdata)); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_interfaceiter_first(isc_interfaceiter_t *iter) { + isc_result_t result; + + REQUIRE(VALID_IFITER(iter)); + + internal_first(iter); + for (;;) { + result = internal_current(iter); + if (result != ISC_R_IGNORE) { + break; + } + result = internal_next(iter); + if (result != ISC_R_SUCCESS) { + break; + } + } + iter->result = result; + return (result); +} + +isc_result_t +isc_interfaceiter_next(isc_interfaceiter_t *iter) { + isc_result_t result; + + REQUIRE(VALID_IFITER(iter)); + REQUIRE(iter->result == ISC_R_SUCCESS); + + for (;;) { + result = internal_next(iter); + if (result != ISC_R_SUCCESS) { + break; + } + result = internal_current(iter); + if (result != ISC_R_IGNORE) { + break; + } + } + iter->result = result; + return (result); +} + +void +isc_interfaceiter_destroy(isc_interfaceiter_t **iterp) { + isc_interfaceiter_t *iter; + REQUIRE(iterp != NULL); + iter = *iterp; + *iterp = NULL; + REQUIRE(VALID_IFITER(iter)); + + internal_destroy(iter); + if (iter->buf != NULL) { + isc_mem_put(iter->mctx, iter->buf, iter->bufsize); + } + + iter->magic = 0; + isc_mem_put(iter->mctx, iter, sizeof(*iter)); +} diff --git a/lib/isc/unix/meminfo.c b/lib/isc/unix/meminfo.c new file mode 100644 index 0000000..612ccea --- /dev/null +++ b/lib/isc/unix/meminfo.c @@ -0,0 +1,50 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <inttypes.h> +#include <unistd.h> + +#include <isc/meminfo.h> +#if defined(HAVE_SYS_SYSCTL_H) && !defined(__linux__) +#include <sys/sysctl.h> +#endif /* if defined(HAVE_SYS_SYSCTL_H) && !defined(__linux__) */ + +uint64_t +isc_meminfo_totalphys(void) { +#if defined(CTL_HW) && (defined(HW_PHYSMEM64) || defined(HW_MEMSIZE)) + int mib[2]; + mib[0] = CTL_HW; +#if defined(HW_MEMSIZE) + mib[1] = HW_MEMSIZE; +#elif defined(HW_PHYSMEM64) + mib[1] = HW_PHYSMEM64; +#endif /* if defined(HW_MEMSIZE) */ + uint64_t size = 0; + size_t len = sizeof(size); + if (sysctl(mib, 2, &size, &len, NULL, 0) == 0) { + return (size); + } +#endif /* if defined(CTL_HW) && (defined(HW_PHYSMEM64) || defined(HW_MEMSIZE)) \ + * */ +#if defined(_SC_PHYS_PAGES) && defined(_SC_PAGESIZE) + long pages = sysconf(_SC_PHYS_PAGES); + long pagesize = sysconf(_SC_PAGESIZE); + + if (pages == -1 || pagesize == -1) { + return (0); + } + + return ((size_t)pages * pagesize); +#endif /* if defined(_SC_PHYS_PAGES) && defined(_SC_PAGESIZE) */ + return (0); +} diff --git a/lib/isc/unix/net.c b/lib/isc/unix/net.c new file mode 100644 index 0000000..1e43b4b --- /dev/null +++ b/lib/isc/unix/net.c @@ -0,0 +1,860 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <stdbool.h> +#include <sys/types.h> + +#if defined(HAVE_SYS_SYSCTL_H) && !defined(__linux__) +#if defined(HAVE_SYS_PARAM_H) +#include <sys/param.h> +#endif /* if defined(HAVE_SYS_PARAM_H) */ +#include <sys/sysctl.h> +#endif /* if defined(HAVE_SYS_SYSCTL_H) && !defined(__linux__) */ +#include <errno.h> +#include <fcntl.h> +#include <sys/uio.h> +#include <unistd.h> + +#include <isc/log.h> +#include <isc/net.h> +#include <isc/netdb.h> +#include <isc/once.h> +#include <isc/strerr.h> +#include <isc/string.h> +#include <isc/util.h> + +#ifndef socklen_t +#define socklen_t unsigned int +#endif /* ifndef socklen_t */ + +/*% + * Definitions about UDP port range specification. This is a total mess of + * portability variants: some use sysctl (but the sysctl names vary), some use + * system-specific interfaces, some have the same interface for IPv4 and IPv6, + * some separate them, etc... + */ + +/*% + * The last resort defaults: use all non well known port space + */ +#ifndef ISC_NET_PORTRANGELOW +#define ISC_NET_PORTRANGELOW 1024 +#endif /* ISC_NET_PORTRANGELOW */ +#ifndef ISC_NET_PORTRANGEHIGH +#define ISC_NET_PORTRANGEHIGH 65535 +#endif /* ISC_NET_PORTRANGEHIGH */ + +#ifdef HAVE_SYSCTLBYNAME + +/*% + * sysctl variants + */ +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__DragonFly__) +#define USE_SYSCTL_PORTRANGE +#define SYSCTL_V4PORTRANGE_LOW "net.inet.ip.portrange.hifirst" +#define SYSCTL_V4PORTRANGE_HIGH "net.inet.ip.portrange.hilast" +#define SYSCTL_V6PORTRANGE_LOW "net.inet.ip.portrange.hifirst" +#define SYSCTL_V6PORTRANGE_HIGH "net.inet.ip.portrange.hilast" +#endif /* if defined(__FreeBSD__) || defined(__APPLE__) || \ + * defined(__DragonFly__) */ + +#ifdef __NetBSD__ +#define USE_SYSCTL_PORTRANGE +#define SYSCTL_V4PORTRANGE_LOW "net.inet.ip.anonportmin" +#define SYSCTL_V4PORTRANGE_HIGH "net.inet.ip.anonportmax" +#define SYSCTL_V6PORTRANGE_LOW "net.inet6.ip6.anonportmin" +#define SYSCTL_V6PORTRANGE_HIGH "net.inet6.ip6.anonportmax" +#endif /* ifdef __NetBSD__ */ + +#else /* !HAVE_SYSCTLBYNAME */ + +#ifdef __OpenBSD__ +#define USE_SYSCTL_PORTRANGE +#define SYSCTL_V4PORTRANGE_LOW \ + { \ + CTL_NET, PF_INET, IPPROTO_IP, IPCTL_IPPORT_HIFIRSTAUTO \ + } +#define SYSCTL_V4PORTRANGE_HIGH \ + { \ + CTL_NET, PF_INET, IPPROTO_IP, IPCTL_IPPORT_HILASTAUTO \ + } +/* Same for IPv6 */ +#define SYSCTL_V6PORTRANGE_LOW SYSCTL_V4PORTRANGE_LOW +#define SYSCTL_V6PORTRANGE_HIGH SYSCTL_V4PORTRANGE_HIGH +#endif /* ifdef __OpenBSD__ */ + +#endif /* HAVE_SYSCTLBYNAME */ + +static isc_once_t once_ipv6only = ISC_ONCE_INIT; +#ifdef __notyet__ +static isc_once_t once_ipv6pktinfo = ISC_ONCE_INIT; +#endif /* ifdef __notyet__ */ + +#ifndef ISC_CMSG_IP_TOS +#ifdef __APPLE__ +#define ISC_CMSG_IP_TOS 0 /* As of 10.8.2. */ +#else /* ! __APPLE__ */ +#define ISC_CMSG_IP_TOS 1 +#endif /* ! __APPLE__ */ +#endif /* ! ISC_CMSG_IP_TOS */ + +static isc_once_t once = ISC_ONCE_INIT; +static isc_once_t once_dscp = ISC_ONCE_INIT; + +static isc_result_t ipv4_result = ISC_R_NOTFOUND; +static isc_result_t ipv6_result = ISC_R_NOTFOUND; +static isc_result_t unix_result = ISC_R_NOTFOUND; +static isc_result_t ipv6only_result = ISC_R_NOTFOUND; +static isc_result_t ipv6pktinfo_result = ISC_R_NOTFOUND; +static unsigned int dscp_result = 0; + +static isc_result_t +try_proto(int domain) { + int s; + isc_result_t result = ISC_R_SUCCESS; + char strbuf[ISC_STRERRORSIZE]; + + s = socket(domain, SOCK_STREAM, 0); + if (s == -1) { + switch (errno) { +#ifdef EAFNOSUPPORT + case EAFNOSUPPORT: +#endif /* ifdef EAFNOSUPPORT */ +#ifdef EPFNOSUPPORT + case EPFNOSUPPORT: +#endif /* ifdef EPFNOSUPPORT */ +#ifdef EPROTONOSUPPORT + case EPROTONOSUPPORT: +#endif /* ifdef EPROTONOSUPPORT */ +#ifdef EINVAL + case EINVAL: +#endif /* ifdef EINVAL */ + return (ISC_R_NOTFOUND); + default: + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "socket() failed: %s", strbuf); + return (ISC_R_UNEXPECTED); + } + } + + if (domain == PF_INET6) { + struct sockaddr_in6 sin6; + unsigned int len; + + /* + * Check to see if IPv6 is broken, as is common on Linux. + */ + len = sizeof(sin6); + if (getsockname(s, (struct sockaddr *)&sin6, (void *)&len) < 0) + { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "retrieving the address of an IPv6 " + "socket from the kernel failed."); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "IPv6 is not supported."); + result = ISC_R_NOTFOUND; + } else { + if (len == sizeof(struct sockaddr_in6)) { + result = ISC_R_SUCCESS; + } else { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_ERROR, + "IPv6 structures in kernel and " + "user space do not match."); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_ERROR, + "IPv6 is not supported."); + result = ISC_R_NOTFOUND; + } + } + } + + (void)close(s); + + return (result); +} + +static void +initialize_action(void) { + ipv4_result = try_proto(PF_INET); + ipv6_result = try_proto(PF_INET6); +#ifdef ISC_PLATFORM_HAVESYSUNH + unix_result = try_proto(PF_UNIX); +#endif /* ifdef ISC_PLATFORM_HAVESYSUNH */ +} + +static void +initialize(void) { + RUNTIME_CHECK(isc_once_do(&once, initialize_action) == ISC_R_SUCCESS); +} + +isc_result_t +isc_net_probeipv4(void) { + initialize(); + return (ipv4_result); +} + +isc_result_t +isc_net_probeipv6(void) { + initialize(); + return (ipv6_result); +} + +isc_result_t +isc_net_probeunix(void) { + initialize(); + return (unix_result); +} + +static void +try_ipv6only(void) { +#ifdef IPV6_V6ONLY + int s, on; + char strbuf[ISC_STRERRORSIZE]; +#endif /* ifdef IPV6_V6ONLY */ + isc_result_t result; + + result = isc_net_probeipv6(); + if (result != ISC_R_SUCCESS) { + ipv6only_result = result; + return; + } + +#ifndef IPV6_V6ONLY + ipv6only_result = ISC_R_NOTFOUND; + return; +#else /* ifndef IPV6_V6ONLY */ + /* check for TCP sockets */ + s = socket(PF_INET6, SOCK_STREAM, 0); + if (s == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "socket() failed: %s", + strbuf); + ipv6only_result = ISC_R_UNEXPECTED; + return; + } + + on = 1; + if (setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) < 0) { + ipv6only_result = ISC_R_NOTFOUND; + goto close; + } + + close(s); + + /* check for UDP sockets */ + s = socket(PF_INET6, SOCK_DGRAM, 0); + if (s == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "socket() failed: %s", + strbuf); + ipv6only_result = ISC_R_UNEXPECTED; + return; + } + + on = 1; + if (setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) < 0) { + ipv6only_result = ISC_R_NOTFOUND; + goto close; + } + + ipv6only_result = ISC_R_SUCCESS; + +close: + close(s); + return; +#endif /* IPV6_V6ONLY */ +} + +static void +initialize_ipv6only(void) { + RUNTIME_CHECK(isc_once_do(&once_ipv6only, try_ipv6only) == + ISC_R_SUCCESS); +} + +#ifdef __notyet__ +static void +try_ipv6pktinfo(void) { + int s, on; + char strbuf[ISC_STRERRORSIZE]; + isc_result_t result; + int optname; + + result = isc_net_probeipv6(); + if (result != ISC_R_SUCCESS) { + ipv6pktinfo_result = result; + return; + } + + /* we only use this for UDP sockets */ + s = socket(PF_INET6, SOCK_DGRAM, IPPROTO_UDP); + if (s == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "socket() failed: %s", + strbuf); + ipv6pktinfo_result = ISC_R_UNEXPECTED; + return; + } + +#ifdef IPV6_RECVPKTINFO + optname = IPV6_RECVPKTINFO; +#else /* ifdef IPV6_RECVPKTINFO */ + optname = IPV6_PKTINFO; +#endif /* ifdef IPV6_RECVPKTINFO */ + on = 1; + if (setsockopt(s, IPPROTO_IPV6, optname, &on, sizeof(on)) < 0) { + ipv6pktinfo_result = ISC_R_NOTFOUND; + goto close; + } + + ipv6pktinfo_result = ISC_R_SUCCESS; + +close: + close(s); + return; +} + +static void +initialize_ipv6pktinfo(void) { + RUNTIME_CHECK(isc_once_do(&once_ipv6pktinfo, try_ipv6pktinfo) == + ISC_R_SUCCESS); +} +#endif /* ifdef __notyet__ */ + +isc_result_t +isc_net_probe_ipv6only(void) { + initialize_ipv6only(); + return (ipv6only_result); +} + +isc_result_t +isc_net_probe_ipv6pktinfo(void) { +/* + * XXXWPK if pktinfo were supported then we could listen on :: for ipv6 and get + * the information about the destination address from pktinfo structure passed + * in recvmsg but this method is not portable and libuv doesn't support it - so + * we need to listen on all interfaces. + * We should verify that this doesn't impact performance (we already do it for + * ipv4) and either remove all the ipv6pktinfo detection code from above + * or think of fixing libuv. + */ +#ifdef __notyet__ + initialize_ipv6pktinfo(); +#endif /* ifdef __notyet__ */ + return (ipv6pktinfo_result); +} + +#if ISC_CMSG_IP_TOS || defined(IPV6_TCLASS) + +static socklen_t +cmsg_len(socklen_t len) { +#ifdef CMSG_LEN + return (CMSG_LEN(len)); +#else /* ifdef CMSG_LEN */ + socklen_t hdrlen; + + /* + * Cast NULL so that any pointer arithmetic performed by CMSG_DATA + * is correct. + */ + hdrlen = (socklen_t)CMSG_DATA(((struct cmsghdr *)NULL)); + return (hdrlen + len); +#endif /* ifdef CMSG_LEN */ +} + +static socklen_t +cmsg_space(socklen_t len) { +#ifdef CMSG_SPACE + return (CMSG_SPACE(len)); +#else /* ifdef CMSG_SPACE */ + struct msghdr msg; + struct cmsghdr *cmsgp; + /* + * XXX: The buffer length is an ad-hoc value, but should be enough + * in a practical sense. + */ + char dummybuf[sizeof(struct cmsghdr) + 1024]; + + memset(&msg, 0, sizeof(msg)); + msg.msg_control = dummybuf; + msg.msg_controllen = sizeof(dummybuf); + + cmsgp = (struct cmsghdr *)dummybuf; + cmsgp->cmsg_len = cmsg_len(len); + + cmsgp = CMSG_NXTHDR(&msg, cmsgp); + if (cmsgp != NULL) { + return ((char *)cmsgp - (char *)msg.msg_control); + } else { + return (0); + } +#endif /* ifdef CMSG_SPACE */ +} + +/* + * Make a fd non-blocking. + */ +static isc_result_t +make_nonblock(int fd) { + int ret; + int flags; + char strbuf[ISC_STRERRORSIZE]; +#ifdef USE_FIONBIO_IOCTL + int on = 1; + + ret = ioctl(fd, FIONBIO, (char *)&on); +#else /* ifdef USE_FIONBIO_IOCTL */ + flags = fcntl(fd, F_GETFL, 0); + flags |= PORT_NONBLOCK; + ret = fcntl(fd, F_SETFL, flags); +#endif /* ifdef USE_FIONBIO_IOCTL */ + + if (ret == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, +#ifdef USE_FIONBIO_IOCTL + "ioctl(%d, FIONBIO, &on): %s", fd, +#else /* ifdef USE_FIONBIO_IOCTL */ + "fcntl(%d, F_SETFL, %d): %s", fd, flags, +#endif /* ifdef USE_FIONBIO_IOCTL */ + strbuf); + + return (ISC_R_UNEXPECTED); + } + + return (ISC_R_SUCCESS); +} + +static bool +cmsgsend(int s, int level, int type, struct addrinfo *res) { + char strbuf[ISC_STRERRORSIZE]; + struct sockaddr_storage ss; + socklen_t len = sizeof(ss); + struct msghdr msg; + union { + struct cmsghdr h; + unsigned char b[256]; + } control; + struct cmsghdr *cmsgp; + int dscp = (46 << 2); /* Expedited forwarding. */ + struct iovec iovec; + char buf[1] = { 0 }; + isc_result_t result; + + if (bind(s, res->ai_addr, res->ai_addrlen) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "bind: %s", strbuf); + return (false); + } + + if (getsockname(s, (struct sockaddr *)&ss, &len) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "getsockname: %s", strbuf); + return (false); + } + + iovec.iov_base = buf; + iovec.iov_len = sizeof(buf); + + memset(&msg, 0, sizeof(msg)); + msg.msg_name = (struct sockaddr *)&ss; + msg.msg_namelen = len; + msg.msg_iov = &iovec; + msg.msg_iovlen = 1; + msg.msg_control = (void *)&control; + msg.msg_controllen = 0; + msg.msg_flags = 0; + + cmsgp = msg.msg_control; + + switch (type) { +#ifdef IP_TOS + case IP_TOS: + memset(cmsgp, 0, cmsg_space(sizeof(char))); + cmsgp->cmsg_level = level; + cmsgp->cmsg_type = type; + cmsgp->cmsg_len = cmsg_len(sizeof(char)); + *(unsigned char *)CMSG_DATA(cmsgp) = dscp; + msg.msg_controllen += cmsg_space(sizeof(char)); + break; +#endif /* ifdef IP_TOS */ +#ifdef IPV6_TCLASS + case IPV6_TCLASS: + memset(cmsgp, 0, cmsg_space(sizeof(dscp))); + cmsgp->cmsg_level = level; + cmsgp->cmsg_type = type; + cmsgp->cmsg_len = cmsg_len(sizeof(dscp)); + memmove(CMSG_DATA(cmsgp), &dscp, sizeof(dscp)); + msg.msg_controllen += cmsg_space(sizeof(dscp)); + break; +#endif /* ifdef IPV6_TCLASS */ + default: + UNREACHABLE(); + } + + if (sendmsg(s, &msg, 0) < 0) { + int debug = ISC_LOG_DEBUG(10); + const char *typestr; + switch (errno) { +#ifdef ENOPROTOOPT + case ENOPROTOOPT: +#endif /* ifdef ENOPROTOOPT */ +#ifdef EOPNOTSUPP + case EOPNOTSUPP: +#endif /* ifdef EOPNOTSUPP */ + case EINVAL: + case EPERM: + break; + default: + debug = ISC_LOG_NOTICE; + } + strerror_r(errno, strbuf, sizeof(strbuf)); + if (debug != ISC_LOG_NOTICE) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "sendmsg: %s", strbuf); + } else { + typestr = (type == IP_TOS) ? "IP_TOS" : "IPV6_TCLASS"; + UNEXPECTED_ERROR(__FILE__, __LINE__, + "probing " + "sendmsg() with %s=%02x failed: %s", + typestr, dscp, strbuf); + } + return (false); + } + + /* + * Make sure the message actually got sent. + */ + result = make_nonblock(s); + RUNTIME_CHECK(result == ISC_R_SUCCESS); + + iovec.iov_base = buf; + iovec.iov_len = sizeof(buf); + + memset(&msg, 0, sizeof(msg)); + msg.msg_name = (struct sockaddr *)&ss; + msg.msg_namelen = sizeof(ss); + msg.msg_iov = &iovec; + msg.msg_iovlen = 1; + msg.msg_control = NULL; + msg.msg_controllen = 0; + msg.msg_flags = 0; + + if (recvmsg(s, &msg, 0) < 0) { + return (false); + } + + return (true); +} +#endif /* if ISC_CMSG_IP_TOS || defined(IPV6_TCLASS) */ + +static void +try_dscp_v4(void) { +#ifdef IP_TOS + char strbuf[ISC_STRERRORSIZE]; + struct addrinfo hints, *res0; + int s, dscp = 0, n; +#ifdef IP_RECVTOS + int on = 1; +#endif /* IP_RECVTOS */ + + memset(&hints, 0, sizeof(hints)); + hints.ai_family = AF_INET; + hints.ai_socktype = SOCK_DGRAM; + hints.ai_protocol = IPPROTO_UDP; +#ifdef AI_NUMERICHOST + hints.ai_flags = AI_PASSIVE | AI_NUMERICHOST; +#else /* ifdef AI_NUMERICHOST */ + hints.ai_flags = AI_PASSIVE; +#endif /* ifdef AI_NUMERICHOST */ + + n = getaddrinfo("127.0.0.1", NULL, &hints, &res0); + if (n != 0 || res0 == NULL) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "getaddrinfo(127.0.0.1): %s", gai_strerror(n)); + return; + } + + s = socket(res0->ai_family, res0->ai_socktype, res0->ai_protocol); + + if (s == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "socket: %s", strbuf); + freeaddrinfo(res0); + return; + } + + if (setsockopt(s, IPPROTO_IP, IP_TOS, &dscp, sizeof(dscp)) == 0) { + dscp_result |= ISC_NET_DSCPSETV4; + } + +#ifdef IP_RECVTOS + on = 1; + if (setsockopt(s, IPPROTO_IP, IP_RECVTOS, &on, sizeof(on)) == 0) { + dscp_result |= ISC_NET_DSCPRECVV4; + } +#endif /* IP_RECVTOS */ + +#if ISC_CMSG_IP_TOS + if (cmsgsend(s, IPPROTO_IP, IP_TOS, res0)) { + dscp_result |= ISC_NET_DSCPPKTV4; + } +#endif /* ISC_CMSG_IP_TOS */ + + freeaddrinfo(res0); + close(s); + +#endif /* IP_TOS */ +} + +static void +try_dscp_v6(void) { +#ifdef IPV6_TCLASS + char strbuf[ISC_STRERRORSIZE]; + struct addrinfo hints, *res0; + int s, dscp = 0, n; +#if defined(IPV6_RECVTCLASS) + int on = 1; +#endif /* IPV6_RECVTCLASS */ + + memset(&hints, 0, sizeof(hints)); + hints.ai_family = AF_INET6; + hints.ai_socktype = SOCK_DGRAM; + hints.ai_protocol = IPPROTO_UDP; +#ifdef AI_NUMERICHOST + hints.ai_flags = AI_PASSIVE | AI_NUMERICHOST; +#else /* ifdef AI_NUMERICHOST */ + hints.ai_flags = AI_PASSIVE; +#endif /* ifdef AI_NUMERICHOST */ + + n = getaddrinfo("::1", NULL, &hints, &res0); + if (n != 0 || res0 == NULL) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "getaddrinfo(::1): %s", gai_strerror(n)); + return; + } + + s = socket(res0->ai_family, res0->ai_socktype, res0->ai_protocol); + if (s == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(10), + "socket: %s", strbuf); + freeaddrinfo(res0); + return; + } + if (setsockopt(s, IPPROTO_IPV6, IPV6_TCLASS, &dscp, sizeof(dscp)) == 0) + { + dscp_result |= ISC_NET_DSCPSETV6; + } + +#ifdef IPV6_RECVTCLASS + on = 1; + if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVTCLASS, &on, sizeof(on)) == 0) + { + dscp_result |= ISC_NET_DSCPRECVV6; + } +#endif /* IPV6_RECVTCLASS */ + + if (cmsgsend(s, IPPROTO_IPV6, IPV6_TCLASS, res0)) { + dscp_result |= ISC_NET_DSCPPKTV6; + } + + freeaddrinfo(res0); + close(s); + +#endif /* IPV6_TCLASS */ +} + +static void +try_dscp(void) { + try_dscp_v4(); + try_dscp_v6(); +} + +static void +initialize_dscp(void) { + RUNTIME_CHECK(isc_once_do(&once_dscp, try_dscp) == ISC_R_SUCCESS); +} + +unsigned int +isc_net_probedscp(void) { + initialize_dscp(); + return (dscp_result); +} + +#if defined(USE_SYSCTL_PORTRANGE) +#if defined(HAVE_SYSCTLBYNAME) +static isc_result_t +getudpportrange_sysctl(int af, in_port_t *low, in_port_t *high) { + int port_low, port_high; + size_t portlen; + const char *sysctlname_lowport, *sysctlname_hiport; + + if (af == AF_INET) { + sysctlname_lowport = SYSCTL_V4PORTRANGE_LOW; + sysctlname_hiport = SYSCTL_V4PORTRANGE_HIGH; + } else { + sysctlname_lowport = SYSCTL_V6PORTRANGE_LOW; + sysctlname_hiport = SYSCTL_V6PORTRANGE_HIGH; + } + portlen = sizeof(port_low); + if (sysctlbyname(sysctlname_lowport, &port_low, &portlen, NULL, 0) < 0) + { + return (ISC_R_FAILURE); + } + portlen = sizeof(port_high); + if (sysctlbyname(sysctlname_hiport, &port_high, &portlen, NULL, 0) < 0) + { + return (ISC_R_FAILURE); + } + if ((port_low & ~0xffff) != 0 || (port_high & ~0xffff) != 0) { + return (ISC_R_RANGE); + } + + *low = (in_port_t)port_low; + *high = (in_port_t)port_high; + + return (ISC_R_SUCCESS); +} +#else /* !HAVE_SYSCTLBYNAME */ +static isc_result_t +getudpportrange_sysctl(int af, in_port_t *low, in_port_t *high) { + int mib_lo4[4] = SYSCTL_V4PORTRANGE_LOW; + int mib_hi4[4] = SYSCTL_V4PORTRANGE_HIGH; + int mib_lo6[4] = SYSCTL_V6PORTRANGE_LOW; + int mib_hi6[4] = SYSCTL_V6PORTRANGE_HIGH; + int *mib_lo, *mib_hi, miblen; + int port_low, port_high; + size_t portlen; + + if (af == AF_INET) { + mib_lo = mib_lo4; + mib_hi = mib_hi4; + miblen = sizeof(mib_lo4) / sizeof(mib_lo4[0]); + } else { + mib_lo = mib_lo6; + mib_hi = mib_hi6; + miblen = sizeof(mib_lo6) / sizeof(mib_lo6[0]); + } + + portlen = sizeof(port_low); + if (sysctl(mib_lo, miblen, &port_low, &portlen, NULL, 0) < 0) { + return (ISC_R_FAILURE); + } + + portlen = sizeof(port_high); + if (sysctl(mib_hi, miblen, &port_high, &portlen, NULL, 0) < 0) { + return (ISC_R_FAILURE); + } + + if ((port_low & ~0xffff) != 0 || (port_high & ~0xffff) != 0) { + return (ISC_R_RANGE); + } + + *low = (in_port_t)port_low; + *high = (in_port_t)port_high; + + return (ISC_R_SUCCESS); +} +#endif /* HAVE_SYSCTLBYNAME */ +#endif /* USE_SYSCTL_PORTRANGE */ + +isc_result_t +isc_net_getudpportrange(int af, in_port_t *low, in_port_t *high) { + int result = ISC_R_FAILURE; +#if !defined(USE_SYSCTL_PORTRANGE) && defined(__linux) + FILE *fp; +#endif /* if !defined(USE_SYSCTL_PORTRANGE) && defined(__linux) */ + + REQUIRE(low != NULL && high != NULL); + +#if defined(USE_SYSCTL_PORTRANGE) + result = getudpportrange_sysctl(af, low, high); +#elif defined(__linux) + + UNUSED(af); + + /* + * Linux local ports are address family agnostic. + */ + fp = fopen("/proc/sys/net/ipv4/ip_local_port_range", "r"); + if (fp != NULL) { + int n; + unsigned int l, h; + + n = fscanf(fp, "%u %u", &l, &h); + if (n == 2 && (l & ~0xffff) == 0 && (h & ~0xffff) == 0) { + *low = l; + *high = h; + result = ISC_R_SUCCESS; + } + fclose(fp); + } +#else /* if defined(USE_SYSCTL_PORTRANGE) */ + UNUSED(af); +#endif /* if defined(USE_SYSCTL_PORTRANGE) */ + + if (result != ISC_R_SUCCESS) { + *low = ISC_NET_PORTRANGELOW; + *high = ISC_NET_PORTRANGEHIGH; + } + + return (ISC_R_SUCCESS); /* we currently never fail in this function */ +} + +void +isc_net_disableipv4(void) { + initialize(); + if (ipv4_result == ISC_R_SUCCESS) { + ipv4_result = ISC_R_DISABLED; + } +} + +void +isc_net_disableipv6(void) { + initialize(); + if (ipv6_result == ISC_R_SUCCESS) { + ipv6_result = ISC_R_DISABLED; + } +} + +void +isc_net_enableipv4(void) { + initialize(); + if (ipv4_result == ISC_R_DISABLED) { + ipv4_result = ISC_R_SUCCESS; + } +} + +void +isc_net_enableipv6(void) { + initialize(); + if (ipv6_result == ISC_R_DISABLED) { + ipv6_result = ISC_R_SUCCESS; + } +} diff --git a/lib/isc/unix/os.c b/lib/isc/unix/os.c new file mode 100644 index 0000000..5577587 --- /dev/null +++ b/lib/isc/unix/os.c @@ -0,0 +1,68 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <isc/os.h> + +#ifdef HAVE_SYSCONF + +#include <unistd.h> + +static long +sysconf_ncpus(void) { +#if defined(_SC_NPROCESSORS_ONLN) + return (sysconf((_SC_NPROCESSORS_ONLN))); +#elif defined(_SC_NPROC_ONLN) + return (sysconf((_SC_NPROC_ONLN))); +#else /* if defined(_SC_NPROCESSORS_ONLN) */ + return (0); +#endif /* if defined(_SC_NPROCESSORS_ONLN) */ +} +#endif /* HAVE_SYSCONF */ + +#if defined(HAVE_SYS_SYSCTL_H) && defined(HAVE_SYSCTLBYNAME) +#include <sys/param.h> /* for NetBSD */ +#include <sys/sysctl.h> +#include <sys/types.h> /* for FreeBSD */ + +static int +sysctl_ncpus(void) { + int ncpu, result; + size_t len; + + len = sizeof(ncpu); + result = sysctlbyname("hw.ncpu", &ncpu, &len, 0, 0); + if (result != -1) { + return (ncpu); + } + return (0); +} +#endif /* if defined(HAVE_SYS_SYSCTL_H) && defined(HAVE_SYSCTLBYNAME) */ + +unsigned int +isc_os_ncpus(void) { + long ncpus = 0; + +#if defined(HAVE_SYSCONF) + ncpus = sysconf_ncpus(); +#endif /* if defined(HAVE_SYSCONF) */ +#if defined(HAVE_SYS_SYSCTL_H) && defined(HAVE_SYSCTLBYNAME) + if (ncpus <= 0) { + ncpus = sysctl_ncpus(); + } +#endif /* if defined(HAVE_SYS_SYSCTL_H) && defined(HAVE_SYSCTLBYNAME) */ + if (ncpus <= 0) { + ncpus = 1; + } + + return ((unsigned int)ncpus); +} diff --git a/lib/isc/unix/pk11_api.c b/lib/isc/unix/pk11_api.c new file mode 100644 index 0000000..babd1a9 --- /dev/null +++ b/lib/isc/unix/pk11_api.c @@ -0,0 +1,708 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <dlfcn.h> +#include <string.h> + +#include <isc/log.h> +#include <isc/mem.h> +#include <isc/once.h> +#include <isc/print.h> +#include <isc/stdio.h> +#include <isc/thread.h> +#include <isc/util.h> + +#include <pkcs11/pkcs11.h> + +#define KEEP_PKCS11_NAMES +#include <pk11/internal.h> +#include <pk11/pk11.h> + +static void *hPK11 = NULL; +static char loaderrmsg[1024]; + +CK_RV +pkcs_C_Initialize(CK_VOID_PTR pReserved) { + CK_C_Initialize sym; + + if (hPK11 != NULL) { + return (CKR_CRYPTOKI_ALREADY_INITIALIZED); + } + + hPK11 = dlopen(pk11_get_lib_name(), RTLD_NOW); + + if (hPK11 == NULL) { + snprintf(loaderrmsg, sizeof(loaderrmsg), + "dlopen(\"%s\") failed: %s\n", pk11_get_lib_name(), + dlerror()); + return (CKR_LIBRARY_LOAD_FAILED); + } + sym = (CK_C_Initialize)dlsym(hPK11, "C_Initialize"); + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(pReserved)); +} + +char * +pk11_get_load_error_message(void) { + return (loaderrmsg); +} + +CK_RV +pkcs_C_Finalize(CK_VOID_PTR pReserved) { + CK_C_Finalize sym; + CK_RV rv; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + sym = (CK_C_Finalize)dlsym(hPK11, "C_Finalize"); + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + rv = (*sym)(pReserved); + if ((rv == CKR_OK) && (dlclose(hPK11) != 0)) { + return (CKR_LIBRARY_LOAD_FAILED); + } + hPK11 = NULL; + return (rv); +} + +CK_RV +pkcs_C_GetSlotList(CK_BBOOL tokenPresent, CK_SLOT_ID_PTR pSlotList, + CK_ULONG_PTR pulCount) { + static CK_C_GetSlotList sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetSlotList)dlsym(hPK11, "C_GetSlotList"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(tokenPresent, pSlotList, pulCount)); +} + +CK_RV +pkcs_C_GetTokenInfo(CK_SLOT_ID slotID, CK_TOKEN_INFO_PTR pInfo) { + static CK_C_GetTokenInfo sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetTokenInfo)dlsym(hPK11, "C_GetTokenInfo"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(slotID, pInfo)); +} + +CK_RV +pkcs_C_GetMechanismInfo(CK_SLOT_ID slotID, CK_MECHANISM_TYPE type, + CK_MECHANISM_INFO_PTR pInfo) { + static CK_C_GetMechanismInfo sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetMechanismInfo)dlsym(hPK11, "C_GetMechanismInfo"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(slotID, type, pInfo)); +} + +CK_RV +pkcs_C_OpenSession(CK_SLOT_ID slotID, CK_FLAGS flags, CK_VOID_PTR pApplication, + CK_RV (*Notify)(CK_SESSION_HANDLE hSession, + CK_NOTIFICATION event, + CK_VOID_PTR pApplication), + CK_SESSION_HANDLE_PTR phSession) { + static CK_C_OpenSession sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + hPK11 = dlopen(pk11_get_lib_name(), RTLD_NOW); + } + if (hPK11 == NULL) { + snprintf(loaderrmsg, sizeof(loaderrmsg), + "dlopen(\"%s\") failed: %s\n", pk11_get_lib_name(), + dlerror()); + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_OpenSession)dlsym(hPK11, "C_OpenSession"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(slotID, flags, pApplication, Notify, phSession)); +} + +CK_RV +pkcs_C_CloseSession(CK_SESSION_HANDLE hSession) { + static CK_C_CloseSession sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_CloseSession)dlsym(hPK11, "C_CloseSession"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession)); +} + +CK_RV +pkcs_C_Login(CK_SESSION_HANDLE hSession, CK_USER_TYPE userType, + CK_CHAR_PTR pPin, CK_ULONG usPinLen) { + static CK_C_Login sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Login)dlsym(hPK11, "C_Login"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, userType, pPin, usPinLen)); +} + +CK_RV +pkcs_C_Logout(CK_SESSION_HANDLE hSession) { + static CK_C_Logout sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Logout)dlsym(hPK11, "C_Logout"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession)); +} + +CK_RV +pkcs_C_CreateObject(CK_SESSION_HANDLE hSession, CK_ATTRIBUTE_PTR pTemplate, + CK_ULONG usCount, CK_OBJECT_HANDLE_PTR phObject) { + static CK_C_CreateObject sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_CreateObject)dlsym(hPK11, "C_CreateObject"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pTemplate, usCount, phObject)); +} + +CK_RV +pkcs_C_DestroyObject(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE hObject) { + static CK_C_DestroyObject sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DestroyObject)dlsym(hPK11, "C_DestroyObject"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, hObject)); +} + +CK_RV +pkcs_C_GetAttributeValue(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE hObject, + CK_ATTRIBUTE_PTR pTemplate, CK_ULONG usCount) { + static CK_C_GetAttributeValue sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GetAttributeValue)dlsym(hPK11, "C_" + "GetAttributeValue"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, hObject, pTemplate, usCount)); +} + +CK_RV +pkcs_C_SetAttributeValue(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE hObject, + CK_ATTRIBUTE_PTR pTemplate, CK_ULONG usCount) { + static CK_C_SetAttributeValue sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SetAttributeValue)dlsym(hPK11, "C_" + "SetAttributeValue"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, hObject, pTemplate, usCount)); +} + +CK_RV +pkcs_C_FindObjectsInit(CK_SESSION_HANDLE hSession, CK_ATTRIBUTE_PTR pTemplate, + CK_ULONG usCount) { + static CK_C_FindObjectsInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_FindObjectsInit)dlsym(hPK11, "C_FindObjectsInit"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pTemplate, usCount)); +} + +CK_RV +pkcs_C_FindObjects(CK_SESSION_HANDLE hSession, CK_OBJECT_HANDLE_PTR phObject, + CK_ULONG usMaxObjectCount, CK_ULONG_PTR pusObjectCount) { + static CK_C_FindObjects sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_FindObjects)dlsym(hPK11, "C_FindObjects"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, phObject, usMaxObjectCount, pusObjectCount)); +} + +CK_RV +pkcs_C_FindObjectsFinal(CK_SESSION_HANDLE hSession) { + static CK_C_FindObjectsFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_FindObjectsFinal)dlsym(hPK11, "C_FindObjectsFinal"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession)); +} + +CK_RV +pkcs_C_EncryptInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hKey) { + static CK_C_EncryptInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_EncryptInit)dlsym(hPK11, "C_EncryptInit"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism, hKey)); +} + +CK_RV +pkcs_C_Encrypt(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pData, + CK_ULONG ulDataLen, CK_BYTE_PTR pEncryptedData, + CK_ULONG_PTR pulEncryptedDataLen) { + static CK_C_Encrypt sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Encrypt)dlsym(hPK11, "C_Encrypt"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pData, ulDataLen, pEncryptedData, + pulEncryptedDataLen)); +} + +CK_RV +pkcs_C_DigestInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism) { + static CK_C_DigestInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DigestInit)dlsym(hPK11, "C_DigestInit"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism)); +} + +CK_RV +pkcs_C_DigestUpdate(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pPart, + CK_ULONG ulPartLen) { + static CK_C_DigestUpdate sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DigestUpdate)dlsym(hPK11, "C_DigestUpdate"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pPart, ulPartLen)); +} + +CK_RV +pkcs_C_DigestFinal(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pDigest, + CK_ULONG_PTR pulDigestLen) { + static CK_C_DigestFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DigestFinal)dlsym(hPK11, "C_DigestFinal"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pDigest, pulDigestLen)); +} + +CK_RV +pkcs_C_SignInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hKey) { + static CK_C_SignInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SignInit)dlsym(hPK11, "C_SignInit"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism, hKey)); +} + +CK_RV +pkcs_C_Sign(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pData, CK_ULONG ulDataLen, + CK_BYTE_PTR pSignature, CK_ULONG_PTR pulSignatureLen) { + static CK_C_Sign sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Sign)dlsym(hPK11, "C_Sign"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pData, ulDataLen, pSignature, + pulSignatureLen)); +} + +CK_RV +pkcs_C_SignUpdate(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pPart, + CK_ULONG ulPartLen) { + static CK_C_SignUpdate sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SignUpdate)dlsym(hPK11, "C_SignUpdate"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pPart, ulPartLen)); +} + +CK_RV +pkcs_C_SignFinal(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pSignature, + CK_ULONG_PTR pulSignatureLen) { + static CK_C_SignFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SignFinal)dlsym(hPK11, "C_SignFinal"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pSignature, pulSignatureLen)); +} + +CK_RV +pkcs_C_VerifyInit(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hKey) { + static CK_C_VerifyInit sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_VerifyInit)dlsym(hPK11, "C_VerifyInit"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism, hKey)); +} + +CK_RV +pkcs_C_Verify(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pData, CK_ULONG ulDataLen, + CK_BYTE_PTR pSignature, CK_ULONG ulSignatureLen) { + static CK_C_Verify sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_Verify)dlsym(hPK11, "C_Verify"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pData, ulDataLen, pSignature, ulSignatureLen)); +} + +CK_RV +pkcs_C_VerifyUpdate(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pPart, + CK_ULONG ulPartLen) { + static CK_C_VerifyUpdate sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_VerifyUpdate)dlsym(hPK11, "C_VerifyUpdate"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pPart, ulPartLen)); +} + +CK_RV +pkcs_C_VerifyFinal(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pSignature, + CK_ULONG ulSignatureLen) { + static CK_C_VerifyFinal sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_VerifyFinal)dlsym(hPK11, "C_VerifyFinal"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pSignature, ulSignatureLen)); +} + +CK_RV +pkcs_C_GenerateKey(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_ATTRIBUTE_PTR pTemplate, CK_ULONG ulCount, + CK_OBJECT_HANDLE_PTR phKey) { + static CK_C_GenerateKey sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GenerateKey)dlsym(hPK11, "C_GenerateKey"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism, pTemplate, ulCount, phKey)); +} + +CK_RV +pkcs_C_GenerateKeyPair(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_ATTRIBUTE_PTR pPublicKeyTemplate, + CK_ULONG usPublicKeyAttributeCount, + CK_ATTRIBUTE_PTR pPrivateKeyTemplate, + CK_ULONG usPrivateKeyAttributeCount, + CK_OBJECT_HANDLE_PTR phPrivateKey, + CK_OBJECT_HANDLE_PTR phPublicKey) { + static CK_C_GenerateKeyPair sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GenerateKeyPair)dlsym(hPK11, "C_GenerateKeyPair"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism, pPublicKeyTemplate, + usPublicKeyAttributeCount, pPrivateKeyTemplate, + usPrivateKeyAttributeCount, phPrivateKey, phPublicKey)); +} + +CK_RV +pkcs_C_DeriveKey(CK_SESSION_HANDLE hSession, CK_MECHANISM_PTR pMechanism, + CK_OBJECT_HANDLE hBaseKey, CK_ATTRIBUTE_PTR pTemplate, + CK_ULONG ulAttributeCount, CK_OBJECT_HANDLE_PTR phKey) { + static CK_C_DeriveKey sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_DeriveKey)dlsym(hPK11, "C_DeriveKey"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pMechanism, hBaseKey, pTemplate, + ulAttributeCount, phKey)); +} + +CK_RV +pkcs_C_SeedRandom(CK_SESSION_HANDLE hSession, CK_BYTE_PTR pSeed, + CK_ULONG ulSeedLen) { + static CK_C_SeedRandom sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_SeedRandom)dlsym(hPK11, "C_SeedRandom"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, pSeed, ulSeedLen)); +} + +CK_RV +pkcs_C_GenerateRandom(CK_SESSION_HANDLE hSession, CK_BYTE_PTR RandomData, + CK_ULONG ulRandomLen) { + static CK_C_GenerateRandom sym = NULL; + static void *pPK11 = NULL; + + if (hPK11 == NULL) { + return (CKR_LIBRARY_LOAD_FAILED); + } + if ((sym == NULL) || (hPK11 != pPK11)) { + pPK11 = hPK11; + sym = (CK_C_GenerateRandom)dlsym(hPK11, "C_GenerateRandom"); + } + if (sym == NULL) { + return (CKR_FUNCTION_NOT_SUPPORTED); + } + return ((*sym)(hSession, RandomData, ulRandomLen)); +} diff --git a/lib/isc/unix/resource.c b/lib/isc/unix/resource.c new file mode 100644 index 0000000..6020b39 --- /dev/null +++ b/lib/isc/unix/resource.c @@ -0,0 +1,219 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <inttypes.h> +#include <stdbool.h> +#include <sys/resource.h> +#include <sys/time.h> /* Required on some systems for <sys/resource.h>. */ +#include <sys/types.h> + +#include <isc/platform.h> +#include <isc/resource.h> +#include <isc/result.h> +#include <isc/util.h> + +#ifdef __linux__ +#include <linux/fs.h> /* To get the large NR_OPEN. */ +#endif /* ifdef __linux__ */ + +#include "errno2result.h" + +static isc_result_t +resource2rlim(isc_resource_t resource, int *rlim_resource) { + isc_result_t result = ISC_R_SUCCESS; + + switch (resource) { + case isc_resource_coresize: + *rlim_resource = RLIMIT_CORE; + break; + case isc_resource_cputime: + *rlim_resource = RLIMIT_CPU; + break; + case isc_resource_datasize: + *rlim_resource = RLIMIT_DATA; + break; + case isc_resource_filesize: + *rlim_resource = RLIMIT_FSIZE; + break; + case isc_resource_lockedmemory: +#ifdef RLIMIT_MEMLOCK + *rlim_resource = RLIMIT_MEMLOCK; +#else /* ifdef RLIMIT_MEMLOCK */ + result = ISC_R_NOTIMPLEMENTED; +#endif /* ifdef RLIMIT_MEMLOCK */ + break; + case isc_resource_openfiles: +#ifdef RLIMIT_NOFILE + *rlim_resource = RLIMIT_NOFILE; +#else /* ifdef RLIMIT_NOFILE */ + result = ISC_R_NOTIMPLEMENTED; +#endif /* ifdef RLIMIT_NOFILE */ + break; + case isc_resource_processes: +#ifdef RLIMIT_NPROC + *rlim_resource = RLIMIT_NPROC; +#else /* ifdef RLIMIT_NPROC */ + result = ISC_R_NOTIMPLEMENTED; +#endif /* ifdef RLIMIT_NPROC */ + break; + case isc_resource_residentsize: +#ifdef RLIMIT_RSS + *rlim_resource = RLIMIT_RSS; +#else /* ifdef RLIMIT_RSS */ + result = ISC_R_NOTIMPLEMENTED; +#endif /* ifdef RLIMIT_RSS */ + break; + case isc_resource_stacksize: + *rlim_resource = RLIMIT_STACK; + break; + default: + /* + * This test is not very robust if isc_resource_t + * changes, but generates a clear assertion message. + */ + REQUIRE(resource >= isc_resource_coresize && + resource <= isc_resource_stacksize); + + result = ISC_R_RANGE; + break; + } + + return (result); +} + +isc_result_t +isc_resource_setlimit(isc_resource_t resource, isc_resourcevalue_t value) { + struct rlimit rl; + rlim_t rlim_value; + int unixresult; + int unixresource; + isc_result_t result; + + result = resource2rlim(resource, &unixresource); + if (result != ISC_R_SUCCESS) { + return (result); + } + + if (value == ISC_RESOURCE_UNLIMITED) { + rlim_value = RLIM_INFINITY; + } else { + /* + * isc_resourcevalue_t was chosen as an unsigned 64 bit + * integer so that it could contain the maximum range of + * reasonable values. Unfortunately, this exceeds the typical + * range on Unix systems. Ensure the range of + * rlim_t is not overflowed. + */ + isc_resourcevalue_t rlim_max; + bool rlim_t_is_signed = (((double)(rlim_t)-1) < 0); + + if (rlim_t_is_signed) { + rlim_max = ~((rlim_t)1 << (sizeof(rlim_t) * 8 - 1)); + } else { + rlim_max = (rlim_t)-1; + } + + if (value > rlim_max) { + value = rlim_max; + } + + rlim_value = value; + } + + rl.rlim_cur = rl.rlim_max = rlim_value; + unixresult = setrlimit(unixresource, &rl); + + if (unixresult == 0) { + return (ISC_R_SUCCESS); + } + +#if defined(OPEN_MAX) && defined(__APPLE__) + /* + * The Darwin kernel doesn't accept RLIM_INFINITY for rlim_cur; the + * maximum possible value is OPEN_MAX. BIND8 used to use + * sysconf(_SC_OPEN_MAX) for such a case, but this value is much + * smaller than OPEN_MAX and is not really effective. + */ + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + rl.rlim_cur = OPEN_MAX; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) { + return (ISC_R_SUCCESS); + } + } +#elif defined(__linux__) +#ifndef NR_OPEN +#define NR_OPEN (1024 * 1024) +#endif /* ifndef NR_OPEN */ + + /* + * Some Linux kernels don't accept RLIM_INFINIT; the maximum + * possible value is the NR_OPEN defined in linux/fs.h. + */ + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + rl.rlim_cur = rl.rlim_max = NR_OPEN; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) { + return (ISC_R_SUCCESS); + } + } +#endif /* if defined(OPEN_MAX) && defined(__APPLE__) */ + if (resource == isc_resource_openfiles && rlim_value == RLIM_INFINITY) { + if (getrlimit(unixresource, &rl) == 0) { + rl.rlim_cur = rl.rlim_max; + unixresult = setrlimit(unixresource, &rl); + if (unixresult == 0) { + return (ISC_R_SUCCESS); + } + } + } + return (isc__errno2result(errno)); +} + +isc_result_t +isc_resource_getlimit(isc_resource_t resource, isc_resourcevalue_t *value) { + int unixresource; + struct rlimit rl; + isc_result_t result; + + result = resource2rlim(resource, &unixresource); + if (result != ISC_R_SUCCESS) { + return (result); + } + + if (getrlimit(unixresource, &rl) != 0) { + return (isc__errno2result(errno)); + } + + *value = rl.rlim_max; + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_resource_getcurlimit(isc_resource_t resource, isc_resourcevalue_t *value) { + int unixresource; + struct rlimit rl; + isc_result_t result; + + result = resource2rlim(resource, &unixresource); + if (result != ISC_R_SUCCESS) { + return (result); + } + + if (getrlimit(unixresource, &rl) != 0) { + return (isc__errno2result(errno)); + } + + *value = rl.rlim_cur; + return (ISC_R_SUCCESS); +} diff --git a/lib/isc/unix/socket.c b/lib/isc/unix/socket.c new file mode 100644 index 0000000..d6578ea --- /dev/null +++ b/lib/isc/unix/socket.c @@ -0,0 +1,5521 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <inttypes.h> +#include <stdbool.h> +#include <sys/param.h> +#include <sys/socket.h> +#include <sys/stat.h> +#include <sys/types.h> +#if defined(HAVE_SYS_SYSCTL_H) && !defined(__linux__) +#include <sys/sysctl.h> +#endif /* if defined(HAVE_SYS_SYSCTL_H) && !defined(__linux__) */ +#include <sys/time.h> +#include <sys/uio.h> + +#if defined(HAVE_LINUX_NETLINK_H) && defined(HAVE_LINUX_RTNETLINK_H) +#include <linux/netlink.h> +#include <linux/rtnetlink.h> +#endif /* if defined(HAVE_LINUX_NETLINK_H) && defined(HAVE_LINUX_RTNETLINK_H) \ + */ + +#include <errno.h> +#include <fcntl.h> +#include <stddef.h> +#include <stdlib.h> +#include <unistd.h> + +#include <isc/app.h> +#include <isc/buffer.h> +#include <isc/condition.h> +#include <isc/formatcheck.h> +#include <isc/list.h> +#include <isc/log.h> +#include <isc/mem.h> +#include <isc/mutex.h> +#include <isc/net.h> +#include <isc/once.h> +#include <isc/platform.h> +#include <isc/print.h> +#include <isc/refcount.h> +#include <isc/region.h> +#include <isc/resource.h> +#include <isc/socket.h> +#include <isc/stats.h> +#include <isc/strerr.h> +#include <isc/string.h> +#include <isc/task.h> +#include <isc/thread.h> +#include <isc/util.h> + +#ifdef ISC_PLATFORM_HAVESYSUNH +#include <sys/un.h> +#endif /* ifdef ISC_PLATFORM_HAVESYSUNH */ +#ifdef HAVE_KQUEUE +#include <sys/event.h> +#endif /* ifdef HAVE_KQUEUE */ +#ifdef HAVE_EPOLL_CREATE1 +#include <sys/epoll.h> +#endif /* ifdef HAVE_EPOLL_CREATE1 */ +#if defined(HAVE_SYS_DEVPOLL_H) +#include <sys/devpoll.h> +#elif defined(HAVE_DEVPOLL_H) +#include <devpoll.h> +#endif /* if defined(HAVE_SYS_DEVPOLL_H) */ + +#include <netinet/tcp.h> + +#include "errno2result.h" + +#ifdef ENABLE_TCP_FASTOPEN +#include <netinet/tcp.h> +#endif /* ifdef ENABLE_TCP_FASTOPEN */ + +#ifdef HAVE_JSON_C +#include <json_object.h> +#endif /* HAVE_JSON_C */ + +#ifdef HAVE_LIBXML2 +#include <libxml/xmlwriter.h> +#define ISC_XMLCHAR (const xmlChar *) +#endif /* HAVE_LIBXML2 */ + +/*% + * Choose the most preferable multiplex method. + */ +#if defined(HAVE_KQUEUE) +#define USE_KQUEUE +#elif defined(HAVE_EPOLL_CREATE1) +#define USE_EPOLL +#elif defined(HAVE_SYS_DEVPOLL_H) || defined(HAVE_DEVPOLL_H) +#define USE_DEVPOLL +typedef struct { + unsigned int want_read : 1, want_write : 1; +} pollinfo_t; +#else /* if defined(HAVE_KQUEUE) */ +#define USE_SELECT +#endif /* HAVE_KQUEUE */ + +/* + * Set by the -T dscp option on the command line. If set to a value + * other than -1, we check to make sure DSCP values match it, and + * assert if not. + */ +int isc_dscp_check_value = -1; + +/*% + * Maximum number of allowable open sockets. This is also the maximum + * allowable socket file descriptor. + * + * Care should be taken before modifying this value for select(): + * The API standard doesn't ensure select() accept more than (the system default + * of) FD_SETSIZE descriptors, and the default size should in fact be fine in + * the vast majority of cases. This constant should therefore be increased only + * when absolutely necessary and possible, i.e., the server is exhausting all + * available file descriptors (up to FD_SETSIZE) and the select() function + * and FD_xxx macros support larger values than FD_SETSIZE (which may not + * always by true, but we keep using some of them to ensure as much + * portability as possible). Note also that overall server performance + * may be rather worsened with a larger value of this constant due to + * inherent scalability problems of select(). + * + * As a special note, this value shouldn't have to be touched if + * this is a build for an authoritative only DNS server. + */ +#ifndef ISC_SOCKET_MAXSOCKETS +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) +#ifdef TUNE_LARGE +#define ISC_SOCKET_MAXSOCKETS 21000 +#else /* ifdef TUNE_LARGE */ +#define ISC_SOCKET_MAXSOCKETS 4096 +#endif /* TUNE_LARGE */ +#elif defined(USE_SELECT) +#define ISC_SOCKET_MAXSOCKETS FD_SETSIZE +#endif /* USE_KQUEUE... */ +#endif /* ISC_SOCKET_MAXSOCKETS */ + +#ifdef USE_SELECT +/*% + * Mac OS X needs a special definition to support larger values in select(). + * We always define this because a larger value can be specified run-time. + */ +#ifdef __APPLE__ +#define _DARWIN_UNLIMITED_SELECT +#endif /* __APPLE__ */ +#endif /* USE_SELECT */ + +#ifdef ISC_SOCKET_USE_POLLWATCH +/*% + * If this macro is defined, enable workaround for a Solaris /dev/poll kernel + * bug: DP_POLL ioctl could keep sleeping even if socket I/O is possible for + * some of the specified FD. The idea is based on the observation that it's + * likely for a busy server to keep receiving packets. It specifically works + * as follows: the socket watcher is first initialized with the state of + * "poll_idle". While it's in the idle state it keeps sleeping until a socket + * event occurs. When it wakes up for a socket I/O event, it moves to the + * poll_active state, and sets the poll timeout to a short period + * (ISC_SOCKET_POLLWATCH_TIMEOUT msec). If timeout occurs in this state, the + * watcher goes to the poll_checking state with the same timeout period. + * In this state, the watcher tries to detect whether this is a break + * during intermittent events or the kernel bug is triggered. If the next + * polling reports an event within the short period, the previous timeout is + * likely to be a kernel bug, and so the watcher goes back to the active state. + * Otherwise, it moves to the idle state again. + * + * It's not clear whether this is a thread-related bug, but since we've only + * seen this with threads, this workaround is used only when enabling threads. + */ + +typedef enum { poll_idle, poll_active, poll_checking } pollstate_t; + +#ifndef ISC_SOCKET_POLLWATCH_TIMEOUT +#define ISC_SOCKET_POLLWATCH_TIMEOUT 10 +#endif /* ISC_SOCKET_POLLWATCH_TIMEOUT */ +#endif /* ISC_SOCKET_USE_POLLWATCH */ + +/*% + * Per-FD lock buckets, we shuffle them around a bit as FDs come in herds. + */ +#define FDLOCK_BITS 10 +#define FDLOCK_COUNT (1 << FDLOCK_BITS) +#define FDLOCK_ID(fd) \ + (((fd) % (FDLOCK_COUNT) >> (FDLOCK_BITS / 2)) | \ + (((fd) << (FDLOCK_BITS / 2)) % (FDLOCK_COUNT))) + +/*% + * Maximum number of events communicated with the kernel. There should normally + * be no need for having a large number. + */ +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) +#ifndef ISC_SOCKET_MAXEVENTS +#ifdef TUNE_LARGE +#define ISC_SOCKET_MAXEVENTS 2048 +#else /* ifdef TUNE_LARGE */ +#define ISC_SOCKET_MAXEVENTS 64 +#endif /* TUNE_LARGE */ +#endif /* ifndef ISC_SOCKET_MAXEVENTS */ +#endif /* if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) \ + * */ + +/*% + * Some systems define the socket length argument as an int, some as size_t, + * some as socklen_t. This is here so it can be easily changed if needed. + */ +#ifndef socklen_t +#define socklen_t unsigned int +#endif /* ifndef socklen_t */ + +/*% + * Define what the possible "soft" errors can be. These are non-fatal returns + * of various network related functions, like recv() and so on. + * + * For some reason, BSDI (and perhaps others) will sometimes return <0 + * from recv() but will have errno==0. This is broken, but we have to + * work around it here. + */ +#define SOFT_ERROR(e) \ + ((e) == EAGAIN || (e) == EWOULDBLOCK || (e) == ENOBUFS || \ + (e) == EINTR || (e) == 0) + +#define DLVL(x) ISC_LOGCATEGORY_GENERAL, ISC_LOGMODULE_SOCKET, ISC_LOG_DEBUG(x) + +/*!< + * DLVL(90) -- Function entry/exit and other tracing. + * DLVL(70) -- Socket "correctness" -- including returning of events, etc. + * DLVL(60) -- Socket data send/receive + * DLVL(50) -- Event tracing, including receiving/sending completion events. + * DLVL(20) -- Socket creation/destruction. + */ +#define TRACE_LEVEL 90 +#define CORRECTNESS_LEVEL 70 +#define IOEVENT_LEVEL 60 +#define EVENT_LEVEL 50 +#define CREATION_LEVEL 20 + +#define TRACE DLVL(TRACE_LEVEL) +#define CORRECTNESS DLVL(CORRECTNESS_LEVEL) +#define IOEVENT DLVL(IOEVENT_LEVEL) +#define EVENT DLVL(EVENT_LEVEL) +#define CREATION DLVL(CREATION_LEVEL) + +typedef isc_event_t intev_t; + +#define SOCKET_MAGIC ISC_MAGIC('I', 'O', 'i', 'o') +#define VALID_SOCKET(s) ISC_MAGIC_VALID(s, SOCKET_MAGIC) + +/*! + * IPv6 control information. If the socket is an IPv6 socket we want + * to collect the destination address and interface so the client can + * set them on outgoing packets. + */ +#ifndef USE_CMSG +#define USE_CMSG 1 +#endif /* ifndef USE_CMSG */ + +/*% + * NetBSD and FreeBSD can timestamp packets. XXXMLG Should we have + * a setsockopt() like interface to request timestamps, and if the OS + * doesn't do it for us, call gettimeofday() on every UDP receive? + */ +#ifdef SO_TIMESTAMP +#ifndef USE_CMSG +#define USE_CMSG 1 +#endif /* ifndef USE_CMSG */ +#endif /* ifdef SO_TIMESTAMP */ + +#if defined(SO_RCVBUF) && defined(ISC_RECV_BUFFER_SIZE) +#define SET_RCVBUF +#endif + +#if defined(SO_SNDBUF) && defined(ISC_SEND_BUFFER_SIZE) +#define SET_SNDBUF +#endif + +/*% + * Instead of calculating the cmsgbuf lengths every time we take + * a rule of thumb approach - sizes are taken from x86_64 linux, + * multiplied by 2, everything should fit. Those sizes are not + * large enough to cause any concern. + */ +#if defined(USE_CMSG) +#define CMSG_SP_IN6PKT 40 +#else /* if defined(USE_CMSG) */ +#define CMSG_SP_IN6PKT 0 +#endif /* if defined(USE_CMSG) */ + +#if defined(USE_CMSG) && defined(SO_TIMESTAMP) +#define CMSG_SP_TIMESTAMP 32 +#else /* if defined(USE_CMSG) && defined(SO_TIMESTAMP) */ +#define CMSG_SP_TIMESTAMP 0 +#endif /* if defined(USE_CMSG) && defined(SO_TIMESTAMP) */ + +#if defined(USE_CMSG) && (defined(IPV6_TCLASS) || defined(IP_TOS)) +#define CMSG_SP_TCTOS 24 +#else /* if defined(USE_CMSG) && (defined(IPV6_TCLASS) || defined(IP_TOS)) */ +#define CMSG_SP_TCTOS 0 +#endif /* if defined(USE_CMSG) && (defined(IPV6_TCLASS) || defined(IP_TOS)) */ + +#define CMSG_SP_INT 24 + +/* Align cmsg buffers to be safe on SPARC etc. */ +#define RECVCMSGBUFLEN \ + ISC_ALIGN(2 * (CMSG_SP_IN6PKT + CMSG_SP_TIMESTAMP + CMSG_SP_TCTOS) + \ + 1, \ + sizeof(void *)) +#define SENDCMSGBUFLEN \ + ISC_ALIGN(2 * (CMSG_SP_IN6PKT + CMSG_SP_INT + CMSG_SP_TCTOS) + 1, \ + sizeof(void *)) + +/*% + * The number of times a send operation is repeated if the result is EINTR. + */ +#define NRETRIES 10 + +typedef struct isc__socketthread isc__socketthread_t; + +#define NEWCONNSOCK(ev) ((ev)->newsocket) + +struct isc_socket { + /* Not locked. */ + unsigned int magic; + isc_socketmgr_t *manager; + isc_mutex_t lock; + isc_sockettype_t type; + const isc_statscounter_t *statsindex; + isc_refcount_t references; + + /* Locked by socket lock. */ + ISC_LINK(isc_socket_t) link; + int fd; + int pf; + int threadid; + char name[16]; + void *tag; + + ISC_LIST(isc_socketevent_t) send_list; + ISC_LIST(isc_socketevent_t) recv_list; + ISC_LIST(isc_socket_newconnev_t) accept_list; + ISC_LIST(isc_socket_connev_t) connect_list; + + isc_sockaddr_t peer_address; /* remote address */ + + unsigned int listener : 1, /* listener socket */ + connected : 1, connecting : 1, /* connect pending + * */ + bound : 1, /* bound to local addr */ + dupped : 1, active : 1, /* currently active */ + pktdscp : 1; /* per packet dscp */ + +#ifdef ISC_PLATFORM_RECVOVERFLOW + unsigned char overflow; /* used for MSG_TRUNC fake */ +#endif /* ifdef ISC_PLATFORM_RECVOVERFLOW */ + + unsigned int dscp; +}; + +#define SOCKET_MANAGER_MAGIC ISC_MAGIC('I', 'O', 'm', 'g') +#define VALID_MANAGER(m) ISC_MAGIC_VALID(m, SOCKET_MANAGER_MAGIC) + +struct isc_socketmgr { + /* Not locked. */ + unsigned int magic; + isc_mem_t *mctx; + isc_mutex_t lock; + isc_stats_t *stats; + int nthreads; + isc__socketthread_t *threads; + unsigned int maxsocks; + /* Locked by manager lock. */ + ISC_LIST(isc_socket_t) socklist; + int reserved; /* unlocked */ + isc_condition_t shutdown_ok; + size_t maxudp; +}; + +struct isc__socketthread { + isc_socketmgr_t *manager; + int threadid; + isc_thread_t thread; + int pipe_fds[2]; + isc_mutex_t *fdlock; + /* Locked by fdlock. */ + isc_socket_t **fds; + int *fdstate; +#ifdef USE_KQUEUE + int kqueue_fd; + int nevents; + struct kevent *events; +#endif /* USE_KQUEUE */ +#ifdef USE_EPOLL + int epoll_fd; + int nevents; + struct epoll_event *events; + uint32_t *epoll_events; +#endif /* USE_EPOLL */ +#ifdef USE_DEVPOLL + int devpoll_fd; + isc_resourcevalue_t open_max; + unsigned int calls; + int nevents; + struct pollfd *events; + pollinfo_t *fdpollinfo; +#endif /* USE_DEVPOLL */ +#ifdef USE_SELECT + int fd_bufsize; + fd_set *read_fds; + fd_set *read_fds_copy; + fd_set *write_fds; + fd_set *write_fds_copy; + int maxfd; +#endif /* USE_SELECT */ +}; + +#define CLOSED 0 /* this one must be zero */ +#define MANAGED 1 +#define CLOSE_PENDING 2 + +/* + * send() and recv() iovec counts + */ +#define MAXSCATTERGATHER_SEND (ISC_SOCKET_MAXSCATTERGATHER) +#ifdef ISC_PLATFORM_RECVOVERFLOW +#define MAXSCATTERGATHER_RECV (ISC_SOCKET_MAXSCATTERGATHER + 1) +#else /* ifdef ISC_PLATFORM_RECVOVERFLOW */ +#define MAXSCATTERGATHER_RECV (ISC_SOCKET_MAXSCATTERGATHER) +#endif /* ifdef ISC_PLATFORM_RECVOVERFLOW */ + +static isc_result_t +socket_create(isc_socketmgr_t *manager0, int pf, isc_sockettype_t type, + isc_socket_t **socketp, isc_socket_t *dup_socket); +static void +send_recvdone_event(isc_socket_t *, isc_socketevent_t **); +static void +send_senddone_event(isc_socket_t *, isc_socketevent_t **); +static void +send_connectdone_event(isc_socket_t *, isc_socket_connev_t **); +static void +free_socket(isc_socket_t **); +static isc_result_t +allocate_socket(isc_socketmgr_t *, isc_sockettype_t, isc_socket_t **); +static void +destroy(isc_socket_t **); +static void +internal_accept(isc_socket_t *); +static void +internal_connect(isc_socket_t *); +static void +internal_recv(isc_socket_t *); +static void +internal_send(isc_socket_t *); +static void +process_cmsg(isc_socket_t *, struct msghdr *, isc_socketevent_t *); +static void +build_msghdr_send(isc_socket_t *, char *, isc_socketevent_t *, struct msghdr *, + struct iovec *, size_t *); +static void +build_msghdr_recv(isc_socket_t *, char *, isc_socketevent_t *, struct msghdr *, + struct iovec *, size_t *); +static bool +process_ctlfd(isc__socketthread_t *thread); +static void +setdscp(isc_socket_t *sock, isc_dscp_t dscp); + +#define SELECT_POKE_SHUTDOWN (-1) +#define SELECT_POKE_NOTHING (-2) +#define SELECT_POKE_READ (-3) +#define SELECT_POKE_ACCEPT (-3) /*%< Same as _READ */ +#define SELECT_POKE_WRITE (-4) +#define SELECT_POKE_CONNECT (-4) /*%< Same as _WRITE */ +#define SELECT_POKE_CLOSE (-5) + +/*% + * Shortcut index arrays to get access to statistics counters. + */ +enum { + STATID_OPEN = 0, + STATID_OPENFAIL = 1, + STATID_CLOSE = 2, + STATID_BINDFAIL = 3, + STATID_CONNECTFAIL = 4, + STATID_CONNECT = 5, + STATID_ACCEPTFAIL = 6, + STATID_ACCEPT = 7, + STATID_SENDFAIL = 8, + STATID_RECVFAIL = 9, + STATID_ACTIVE = 10 +}; +static const isc_statscounter_t udp4statsindex[] = { + isc_sockstatscounter_udp4open, + isc_sockstatscounter_udp4openfail, + isc_sockstatscounter_udp4close, + isc_sockstatscounter_udp4bindfail, + isc_sockstatscounter_udp4connectfail, + isc_sockstatscounter_udp4connect, + -1, + -1, + isc_sockstatscounter_udp4sendfail, + isc_sockstatscounter_udp4recvfail, + isc_sockstatscounter_udp4active +}; +static const isc_statscounter_t udp6statsindex[] = { + isc_sockstatscounter_udp6open, + isc_sockstatscounter_udp6openfail, + isc_sockstatscounter_udp6close, + isc_sockstatscounter_udp6bindfail, + isc_sockstatscounter_udp6connectfail, + isc_sockstatscounter_udp6connect, + -1, + -1, + isc_sockstatscounter_udp6sendfail, + isc_sockstatscounter_udp6recvfail, + isc_sockstatscounter_udp6active +}; +static const isc_statscounter_t tcp4statsindex[] = { + isc_sockstatscounter_tcp4open, isc_sockstatscounter_tcp4openfail, + isc_sockstatscounter_tcp4close, isc_sockstatscounter_tcp4bindfail, + isc_sockstatscounter_tcp4connectfail, isc_sockstatscounter_tcp4connect, + isc_sockstatscounter_tcp4acceptfail, isc_sockstatscounter_tcp4accept, + isc_sockstatscounter_tcp4sendfail, isc_sockstatscounter_tcp4recvfail, + isc_sockstatscounter_tcp4active +}; +static const isc_statscounter_t tcp6statsindex[] = { + isc_sockstatscounter_tcp6open, isc_sockstatscounter_tcp6openfail, + isc_sockstatscounter_tcp6close, isc_sockstatscounter_tcp6bindfail, + isc_sockstatscounter_tcp6connectfail, isc_sockstatscounter_tcp6connect, + isc_sockstatscounter_tcp6acceptfail, isc_sockstatscounter_tcp6accept, + isc_sockstatscounter_tcp6sendfail, isc_sockstatscounter_tcp6recvfail, + isc_sockstatscounter_tcp6active +}; +static const isc_statscounter_t unixstatsindex[] = { + isc_sockstatscounter_unixopen, isc_sockstatscounter_unixopenfail, + isc_sockstatscounter_unixclose, isc_sockstatscounter_unixbindfail, + isc_sockstatscounter_unixconnectfail, isc_sockstatscounter_unixconnect, + isc_sockstatscounter_unixacceptfail, isc_sockstatscounter_unixaccept, + isc_sockstatscounter_unixsendfail, isc_sockstatscounter_unixrecvfail, + isc_sockstatscounter_unixactive +}; +static const isc_statscounter_t rawstatsindex[] = { + isc_sockstatscounter_rawopen, + isc_sockstatscounter_rawopenfail, + isc_sockstatscounter_rawclose, + -1, + -1, + -1, + -1, + -1, + -1, + isc_sockstatscounter_rawrecvfail, + isc_sockstatscounter_rawactive +}; + +static int +gen_threadid(isc_socket_t *sock); + +static int +gen_threadid(isc_socket_t *sock) { + return (sock->fd % sock->manager->nthreads); +} + +static void +manager_log(isc_socketmgr_t *sockmgr, isc_logcategory_t *category, + isc_logmodule_t *module, int level, const char *fmt, ...) + ISC_FORMAT_PRINTF(5, 6); +static void +manager_log(isc_socketmgr_t *sockmgr, isc_logcategory_t *category, + isc_logmodule_t *module, int level, const char *fmt, ...) { + char msgbuf[2048]; + va_list ap; + + if (!isc_log_wouldlog(isc_lctx, level)) { + return; + } + + va_start(ap, fmt); + vsnprintf(msgbuf, sizeof(msgbuf), fmt, ap); + va_end(ap); + + isc_log_write(isc_lctx, category, module, level, "sockmgr %p: %s", + sockmgr, msgbuf); +} + +static void +thread_log(isc__socketthread_t *thread, isc_logcategory_t *category, + isc_logmodule_t *module, int level, const char *fmt, ...) + ISC_FORMAT_PRINTF(5, 6); +static void +thread_log(isc__socketthread_t *thread, isc_logcategory_t *category, + isc_logmodule_t *module, int level, const char *fmt, ...) { + char msgbuf[2048]; + va_list ap; + + if (!isc_log_wouldlog(isc_lctx, level)) { + return; + } + + va_start(ap, fmt); + vsnprintf(msgbuf, sizeof(msgbuf), fmt, ap); + va_end(ap); + + isc_log_write(isc_lctx, category, module, level, + "sockmgr %p thread %d: %s", thread->manager, + thread->threadid, msgbuf); +} + +static void +socket_log(isc_socket_t *sock, const isc_sockaddr_t *address, + isc_logcategory_t *category, isc_logmodule_t *module, int level, + const char *fmt, ...) ISC_FORMAT_PRINTF(6, 7); +static void +socket_log(isc_socket_t *sock, const isc_sockaddr_t *address, + isc_logcategory_t *category, isc_logmodule_t *module, int level, + const char *fmt, ...) { + char msgbuf[2048]; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + va_list ap; + + if (!isc_log_wouldlog(isc_lctx, level)) { + return; + } + + va_start(ap, fmt); + vsnprintf(msgbuf, sizeof(msgbuf), fmt, ap); + va_end(ap); + + if (address == NULL) { + isc_log_write(isc_lctx, category, module, level, + "socket %p: %s", sock, msgbuf); + } else { + isc_sockaddr_format(address, peerbuf, sizeof(peerbuf)); + isc_log_write(isc_lctx, category, module, level, + "socket %p %s: %s", sock, peerbuf, msgbuf); + } +} + +/*% + * Increment socket-related statistics counters. + */ +static void +inc_stats(isc_stats_t *stats, isc_statscounter_t counterid) { + REQUIRE(counterid != -1); + + if (stats != NULL) { + isc_stats_increment(stats, counterid); + } +} + +/*% + * Decrement socket-related statistics counters. + */ +static void +dec_stats(isc_stats_t *stats, isc_statscounter_t counterid) { + REQUIRE(counterid != -1); + + if (stats != NULL) { + isc_stats_decrement(stats, counterid); + } +} + +static isc_result_t +watch_fd(isc__socketthread_t *thread, int fd, int msg) { + isc_result_t result = ISC_R_SUCCESS; + +#ifdef USE_KQUEUE + struct kevent evchange; + + memset(&evchange, 0, sizeof(evchange)); + if (msg == SELECT_POKE_READ) { + evchange.filter = EVFILT_READ; + } else { + evchange.filter = EVFILT_WRITE; + } + evchange.flags = EV_ADD; + evchange.ident = fd; + if (kevent(thread->kqueue_fd, &evchange, 1, NULL, 0, NULL) != 0) { + result = isc__errno2result(errno); + } + + return (result); +#elif defined(USE_EPOLL) + struct epoll_event event; + uint32_t oldevents; + int ret; + int op; + + oldevents = thread->epoll_events[fd]; + if (msg == SELECT_POKE_READ) { + thread->epoll_events[fd] |= EPOLLIN; + } else { + thread->epoll_events[fd] |= EPOLLOUT; + } + + event.events = thread->epoll_events[fd]; + memset(&event.data, 0, sizeof(event.data)); + event.data.fd = fd; + + op = (oldevents == 0U) ? EPOLL_CTL_ADD : EPOLL_CTL_MOD; + if (thread->fds[fd] != NULL) { + LOCK(&thread->fds[fd]->lock); + } + ret = epoll_ctl(thread->epoll_fd, op, fd, &event); + if (thread->fds[fd] != NULL) { + UNLOCK(&thread->fds[fd]->lock); + } + if (ret == -1) { + if (errno == EEXIST) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "epoll_ctl(ADD/MOD) returned " + "EEXIST for fd %d", + fd); + } + result = isc__errno2result(errno); + } + + return (result); +#elif defined(USE_DEVPOLL) + struct pollfd pfd; + + memset(&pfd, 0, sizeof(pfd)); + if (msg == SELECT_POKE_READ) { + pfd.events = POLLIN; + } else { + pfd.events = POLLOUT; + } + pfd.fd = fd; + pfd.revents = 0; + if (write(thread->devpoll_fd, &pfd, sizeof(pfd)) == -1) { + result = isc__errno2result(errno); + } else { + if (msg == SELECT_POKE_READ) { + thread->fdpollinfo[fd].want_read = 1; + } else { + thread->fdpollinfo[fd].want_write = 1; + } + } + + return (result); +#elif defined(USE_SELECT) + LOCK(&thread->manager->lock); + if (msg == SELECT_POKE_READ) { + FD_SET(fd, thread->read_fds); + } + if (msg == SELECT_POKE_WRITE) { + FD_SET(fd, thread->write_fds); + } + UNLOCK(&thread->manager->lock); + + return (result); +#endif /* ifdef USE_KQUEUE */ +} + +static isc_result_t +unwatch_fd(isc__socketthread_t *thread, int fd, int msg) { + isc_result_t result = ISC_R_SUCCESS; + +#ifdef USE_KQUEUE + struct kevent evchange; + + memset(&evchange, 0, sizeof(evchange)); + if (msg == SELECT_POKE_READ) { + evchange.filter = EVFILT_READ; + } else { + evchange.filter = EVFILT_WRITE; + } + evchange.flags = EV_DELETE; + evchange.ident = fd; + if (kevent(thread->kqueue_fd, &evchange, 1, NULL, 0, NULL) != 0) { + result = isc__errno2result(errno); + } + + return (result); +#elif defined(USE_EPOLL) + struct epoll_event event; + int ret; + int op; + + if (msg == SELECT_POKE_READ) { + thread->epoll_events[fd] &= ~(EPOLLIN); + } else { + thread->epoll_events[fd] &= ~(EPOLLOUT); + } + + event.events = thread->epoll_events[fd]; + memset(&event.data, 0, sizeof(event.data)); + event.data.fd = fd; + + op = (event.events == 0U) ? EPOLL_CTL_DEL : EPOLL_CTL_MOD; + ret = epoll_ctl(thread->epoll_fd, op, fd, &event); + if (ret == -1 && errno != ENOENT) { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "epoll_ctl(DEL), %d: %s", + fd, strbuf); + result = ISC_R_UNEXPECTED; + } + return (result); +#elif defined(USE_DEVPOLL) + struct pollfd pfds[2]; + size_t writelen = sizeof(pfds[0]); + + memset(pfds, 0, sizeof(pfds)); + pfds[0].events = POLLREMOVE; + pfds[0].fd = fd; + + /* + * Canceling read or write polling via /dev/poll is tricky. Since it + * only provides a way of canceling per FD, we may need to re-poll the + * socket for the other operation. + */ + if (msg == SELECT_POKE_READ && thread->fdpollinfo[fd].want_write == 1) { + pfds[1].events = POLLOUT; + pfds[1].fd = fd; + writelen += sizeof(pfds[1]); + } + if (msg == SELECT_POKE_WRITE && thread->fdpollinfo[fd].want_read == 1) { + pfds[1].events = POLLIN; + pfds[1].fd = fd; + writelen += sizeof(pfds[1]); + } + + if (write(thread->devpoll_fd, pfds, writelen) == -1) { + result = isc__errno2result(errno); + } else { + if (msg == SELECT_POKE_READ) { + thread->fdpollinfo[fd].want_read = 0; + } else { + thread->fdpollinfo[fd].want_write = 0; + } + } + + return (result); +#elif defined(USE_SELECT) + LOCK(&thread->manager->lock); + if (msg == SELECT_POKE_READ) { + FD_CLR(fd, thread->read_fds); + } else if (msg == SELECT_POKE_WRITE) { + FD_CLR(fd, thread->write_fds); + } + UNLOCK(&thread->manager->lock); + + return (result); +#endif /* ifdef USE_KQUEUE */ +} + +/* + * A poke message was received, perform a proper watch/unwatch + * on a fd provided + */ +static void +wakeup_socket(isc__socketthread_t *thread, int fd, int msg) { + isc_result_t result; + int lockid = FDLOCK_ID(fd); + + /* + * This is a wakeup on a socket. If the socket is not in the + * process of being closed, start watching it for either reads + * or writes. + */ + + INSIST(fd >= 0 && fd < (int)thread->manager->maxsocks); + + if (msg == SELECT_POKE_CLOSE) { + LOCK(&thread->fdlock[lockid]); + INSIST(thread->fdstate[fd] == CLOSE_PENDING); + thread->fdstate[fd] = CLOSED; + (void)unwatch_fd(thread, fd, SELECT_POKE_READ); + (void)unwatch_fd(thread, fd, SELECT_POKE_WRITE); + (void)close(fd); + UNLOCK(&thread->fdlock[lockid]); + return; + } + + LOCK(&thread->fdlock[lockid]); + if (thread->fdstate[fd] == CLOSE_PENDING) { + /* + * We accept (and ignore) any error from unwatch_fd() as we are + * closing the socket, hoping it doesn't leave dangling state in + * the kernel. + * Note that unwatch_fd() must be called after releasing the + * fdlock; otherwise it could cause deadlock due to a lock order + * reversal. + */ + (void)unwatch_fd(thread, fd, SELECT_POKE_READ); + (void)unwatch_fd(thread, fd, SELECT_POKE_WRITE); + UNLOCK(&thread->fdlock[lockid]); + return; + } + if (thread->fdstate[fd] != MANAGED) { + UNLOCK(&thread->fdlock[lockid]); + return; + } + + /* + * Set requested bit. + */ + result = watch_fd(thread, fd, msg); + if (result != ISC_R_SUCCESS) { + /* + * XXXJT: what should we do? Ignoring the failure of watching + * a socket will make the application dysfunctional, but there + * seems to be no reasonable recovery process. + */ + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "failed to start watching FD (%d): %s", fd, + isc_result_totext(result)); + } + UNLOCK(&thread->fdlock[lockid]); +} + +/* + * Poke the select loop when there is something for us to do. + * The write is required (by POSIX) to complete. That is, we + * will not get partial writes. + */ +static void +select_poke(isc_socketmgr_t *mgr, int threadid, int fd, int msg) { + int cc; + int buf[2]; + char strbuf[ISC_STRERRORSIZE]; + + buf[0] = fd; + buf[1] = msg; + + do { + cc = write(mgr->threads[threadid].pipe_fds[1], buf, + sizeof(buf)); +#ifdef ENOSR + /* + * Treat ENOSR as EAGAIN but loop slowly as it is + * unlikely to clear fast. + */ + if (cc < 0 && errno == ENOSR) { + sleep(1); + errno = EAGAIN; + } +#endif /* ifdef ENOSR */ + } while (cc < 0 && SOFT_ERROR(errno)); + + if (cc < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + FATAL_ERROR(__FILE__, __LINE__, + "write() failed during watcher poke: %s", strbuf); + } + + INSIST(cc == sizeof(buf)); +} + +/* + * Read a message on the internal fd. + */ +static void +select_readmsg(isc__socketthread_t *thread, int *fd, int *msg) { + int buf[2]; + int cc; + char strbuf[ISC_STRERRORSIZE]; + + cc = read(thread->pipe_fds[0], buf, sizeof(buf)); + if (cc < 0) { + *msg = SELECT_POKE_NOTHING; + *fd = -1; /* Silence compiler. */ + if (SOFT_ERROR(errno)) { + return; + } + + strerror_r(errno, strbuf, sizeof(strbuf)); + FATAL_ERROR(__FILE__, __LINE__, + "read() failed during watcher poke: %s", strbuf); + } + INSIST(cc == sizeof(buf)); + + *fd = buf[0]; + *msg = buf[1]; +} + +/* + * Make a fd non-blocking. + */ +static isc_result_t +make_nonblock(int fd) { + int ret; + char strbuf[ISC_STRERRORSIZE]; +#ifdef USE_FIONBIO_IOCTL + int on = 1; +#else /* ifdef USE_FIONBIO_IOCTL */ + int flags; +#endif /* ifdef USE_FIONBIO_IOCTL */ + +#ifdef USE_FIONBIO_IOCTL + ret = ioctl(fd, FIONBIO, (char *)&on); +#else /* ifdef USE_FIONBIO_IOCTL */ + flags = fcntl(fd, F_GETFL, 0); + flags |= PORT_NONBLOCK; + ret = fcntl(fd, F_SETFL, flags); +#endif /* ifdef USE_FIONBIO_IOCTL */ + + if (ret == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, +#ifdef USE_FIONBIO_IOCTL + "ioctl(%d, FIONBIO, &on): %s", fd, +#else /* ifdef USE_FIONBIO_IOCTL */ + "fcntl(%d, F_SETFL, %d): %s", fd, flags, +#endif /* ifdef USE_FIONBIO_IOCTL */ + strbuf); + + return (ISC_R_UNEXPECTED); + } + + return (ISC_R_SUCCESS); +} + +#ifdef USE_CMSG +/* + * Not all OSes support advanced CMSG macros: CMSG_LEN and CMSG_SPACE. + * In order to ensure as much portability as possible, we provide wrapper + * functions of these macros. + * Note that cmsg_space() could run slow on OSes that do not have + * CMSG_SPACE. + */ +static socklen_t +cmsg_len(socklen_t len) { +#ifdef CMSG_LEN + return (CMSG_LEN(len)); +#else /* ifdef CMSG_LEN */ + socklen_t hdrlen; + + /* + * Cast NULL so that any pointer arithmetic performed by CMSG_DATA + * is correct. + */ + hdrlen = (socklen_t)CMSG_DATA(((struct cmsghdr *)NULL)); + return (hdrlen + len); +#endif /* ifdef CMSG_LEN */ +} + +static socklen_t +cmsg_space(socklen_t len) { +#ifdef CMSG_SPACE + return (CMSG_SPACE(len)); +#else /* ifdef CMSG_SPACE */ + struct msghdr msg; + struct cmsghdr *cmsgp; + /* + * XXX: The buffer length is an ad-hoc value, but should be enough + * in a practical sense. + */ + char dummybuf[sizeof(struct cmsghdr) + 1024]; + + memset(&msg, 0, sizeof(msg)); + msg.msg_control = dummybuf; + msg.msg_controllen = sizeof(dummybuf); + + cmsgp = (struct cmsghdr *)dummybuf; + cmsgp->cmsg_len = cmsg_len(len); + + cmsgp = CMSG_NXTHDR(&msg, cmsgp); + if (cmsgp != NULL) { + return ((char *)cmsgp - (char *)msg.msg_control); + } else { + return (0); + } +#endif /* ifdef CMSG_SPACE */ +} +#endif /* USE_CMSG */ + +/* + * Process control messages received on a socket. + */ +static void +process_cmsg(isc_socket_t *sock, struct msghdr *msg, isc_socketevent_t *dev) { +#ifdef USE_CMSG + struct cmsghdr *cmsgp; + struct in6_pktinfo *pktinfop; +#ifdef SO_TIMESTAMP + void *timevalp; +#endif /* ifdef SO_TIMESTAMP */ +#endif /* ifdef USE_CMSG */ + + /* + * sock is used only when ISC_NET_BSD44MSGHDR and USE_CMSG are defined. + * msg and dev are used only when ISC_NET_BSD44MSGHDR is defined. + * They are all here, outside of the CPP tests, because it is + * more consistent with the usual ISC coding style. + */ + UNUSED(sock); + UNUSED(msg); + UNUSED(dev); + +#ifdef MSG_TRUNC + if ((msg->msg_flags & MSG_TRUNC) != 0) { + dev->attributes |= ISC_SOCKEVENTATTR_TRUNC; + } +#endif /* ifdef MSG_TRUNC */ + +#ifdef MSG_CTRUNC + if ((msg->msg_flags & MSG_CTRUNC) != 0) { + dev->attributes |= ISC_SOCKEVENTATTR_CTRUNC; + } +#endif /* ifdef MSG_CTRUNC */ + +#ifndef USE_CMSG + return; +#else /* ifndef USE_CMSG */ + if (msg->msg_controllen == 0U || msg->msg_control == NULL) { + return; + } + +#ifdef SO_TIMESTAMP + timevalp = NULL; +#endif /* ifdef SO_TIMESTAMP */ + pktinfop = NULL; + + cmsgp = CMSG_FIRSTHDR(msg); + while (cmsgp != NULL) { + socket_log(sock, NULL, TRACE, "processing cmsg %p", cmsgp); + + if (cmsgp->cmsg_level == IPPROTO_IPV6 && + cmsgp->cmsg_type == IPV6_PKTINFO) + { + pktinfop = (struct in6_pktinfo *)CMSG_DATA(cmsgp); + memmove(&dev->pktinfo, pktinfop, + sizeof(struct in6_pktinfo)); + dev->attributes |= ISC_SOCKEVENTATTR_PKTINFO; + socket_log(sock, NULL, TRACE, + "interface received on ifindex %u", + dev->pktinfo.ipi6_ifindex); + if (IN6_IS_ADDR_MULTICAST(&pktinfop->ipi6_addr)) { + dev->attributes |= ISC_SOCKEVENTATTR_MULTICAST; + } + goto next; + } + +#ifdef SO_TIMESTAMP + if (cmsgp->cmsg_level == SOL_SOCKET && + cmsgp->cmsg_type == SCM_TIMESTAMP) + { + struct timeval tv; + timevalp = CMSG_DATA(cmsgp); + memmove(&tv, timevalp, sizeof(tv)); + dev->timestamp.seconds = tv.tv_sec; + dev->timestamp.nanoseconds = tv.tv_usec * 1000; + dev->attributes |= ISC_SOCKEVENTATTR_TIMESTAMP; + goto next; + } +#endif /* ifdef SO_TIMESTAMP */ + +#ifdef IPV6_TCLASS + if (cmsgp->cmsg_level == IPPROTO_IPV6 && + cmsgp->cmsg_type == IPV6_TCLASS) + { + dev->dscp = *(int *)CMSG_DATA(cmsgp); + dev->dscp >>= 2; + dev->attributes |= ISC_SOCKEVENTATTR_DSCP; + goto next; + } +#endif /* ifdef IPV6_TCLASS */ + +#ifdef IP_TOS + if (cmsgp->cmsg_level == IPPROTO_IP && + (cmsgp->cmsg_type == IP_TOS +#ifdef IP_RECVTOS + || cmsgp->cmsg_type == IP_RECVTOS +#endif /* ifdef IP_RECVTOS */ + )) + { + dev->dscp = (int)*(unsigned char *)CMSG_DATA(cmsgp); + dev->dscp >>= 2; + dev->attributes |= ISC_SOCKEVENTATTR_DSCP; + goto next; + } +#endif /* ifdef IP_TOS */ + next: + cmsgp = CMSG_NXTHDR(msg, cmsgp); + } +#endif /* USE_CMSG */ +} + +/* + * Construct an iov array and attach it to the msghdr passed in. This is + * the SEND constructor, which will use the used region of the buffer + * (if using a buffer list) or will use the internal region (if a single + * buffer I/O is requested). + * + * Nothing can be NULL, and the done event must list at least one buffer + * on the buffer linked list for this function to be meaningful. + * + * If write_countp != NULL, *write_countp will hold the number of bytes + * this transaction can send. + */ +static void +build_msghdr_send(isc_socket_t *sock, char *cmsgbuf, isc_socketevent_t *dev, + struct msghdr *msg, struct iovec *iov, size_t *write_countp) { + unsigned int iovcount; + size_t write_count; + struct cmsghdr *cmsgp; + + memset(msg, 0, sizeof(*msg)); + + if (!sock->connected) { + msg->msg_name = (void *)&dev->address.type.sa; + msg->msg_namelen = dev->address.length; + } else { + msg->msg_name = NULL; + msg->msg_namelen = 0; + } + + write_count = dev->region.length - dev->n; + iov[0].iov_base = (void *)(dev->region.base + dev->n); + iov[0].iov_len = write_count; + iovcount = 1; + + msg->msg_iov = iov; + msg->msg_iovlen = iovcount; + msg->msg_control = NULL; + msg->msg_controllen = 0; + msg->msg_flags = 0; +#if defined(USE_CMSG) + + if ((sock->type == isc_sockettype_udp) && + ((dev->attributes & ISC_SOCKEVENTATTR_PKTINFO) != 0)) + { + struct in6_pktinfo *pktinfop; + + socket_log(sock, NULL, TRACE, "sendto pktinfo data, ifindex %u", + dev->pktinfo.ipi6_ifindex); + + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen = cmsg_space(sizeof(struct in6_pktinfo)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp = (struct cmsghdr *)cmsgbuf; + cmsgp->cmsg_level = IPPROTO_IPV6; + cmsgp->cmsg_type = IPV6_PKTINFO; + cmsgp->cmsg_len = cmsg_len(sizeof(struct in6_pktinfo)); + pktinfop = (struct in6_pktinfo *)CMSG_DATA(cmsgp); + memmove(pktinfop, &dev->pktinfo, sizeof(struct in6_pktinfo)); + } + +#if defined(IPV6_USE_MIN_MTU) + if ((sock->type == isc_sockettype_udp) && (sock->pf == AF_INET6) && + ((dev->attributes & ISC_SOCKEVENTATTR_USEMINMTU) != 0)) + { + int use_min_mtu = 1; /* -1, 0, 1 */ + + cmsgp = (struct cmsghdr *)(cmsgbuf + msg->msg_controllen); + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen += cmsg_space(sizeof(use_min_mtu)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp->cmsg_level = IPPROTO_IPV6; + cmsgp->cmsg_type = IPV6_USE_MIN_MTU; + cmsgp->cmsg_len = cmsg_len(sizeof(use_min_mtu)); + memmove(CMSG_DATA(cmsgp), &use_min_mtu, sizeof(use_min_mtu)); + } +#endif /* if defined(IPV6_USE_MIN_MTU) */ + + if (isc_dscp_check_value > -1) { + if (sock->type == isc_sockettype_udp) { + INSIST((int)dev->dscp == isc_dscp_check_value); + } else if (sock->type == isc_sockettype_tcp) { + INSIST((int)sock->dscp == isc_dscp_check_value); + } + } + +#if defined(IP_TOS) || (defined(IPPROTO_IPV6) && defined(IPV6_TCLASS)) + if ((sock->type == isc_sockettype_udp) && + ((dev->attributes & ISC_SOCKEVENTATTR_DSCP) != 0)) + { + int dscp = (dev->dscp << 2) & 0xff; + + INSIST(dev->dscp < 0x40); + +#ifdef IP_TOS + if (sock->pf == AF_INET && sock->pktdscp) { + cmsgp = (struct cmsghdr *)(cmsgbuf + + msg->msg_controllen); + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen += cmsg_space(sizeof(dscp)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp->cmsg_level = IPPROTO_IP; + cmsgp->cmsg_type = IP_TOS; + cmsgp->cmsg_len = cmsg_len(sizeof(char)); + *(unsigned char *)CMSG_DATA(cmsgp) = dscp; + } else if (sock->pf == AF_INET && sock->dscp != dev->dscp) { + if (setsockopt(sock->fd, IPPROTO_IP, IP_TOS, + (void *)&dscp, sizeof(int)) < 0) + { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IP_TOS, %.02x)" + " failed: %s", + sock->fd, dscp >> 2, strbuf); + } else { + sock->dscp = dscp; + } + } +#endif /* ifdef IP_TOS */ +#if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS) + if (sock->pf == AF_INET6 && sock->pktdscp) { + cmsgp = (struct cmsghdr *)(cmsgbuf + + msg->msg_controllen); + msg->msg_control = (void *)cmsgbuf; + msg->msg_controllen += cmsg_space(sizeof(dscp)); + INSIST(msg->msg_controllen <= SENDCMSGBUFLEN); + + cmsgp->cmsg_level = IPPROTO_IPV6; + cmsgp->cmsg_type = IPV6_TCLASS; + cmsgp->cmsg_len = cmsg_len(sizeof(dscp)); + memmove(CMSG_DATA(cmsgp), &dscp, sizeof(dscp)); + } else if (sock->pf == AF_INET6 && sock->dscp != dev->dscp) { + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_TCLASS, + (void *)&dscp, sizeof(int)) < 0) + { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_TCLASS, " + "%.02x) failed: %s", + sock->fd, dscp >> 2, strbuf); + } else { + sock->dscp = dscp; + } + } +#endif /* if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS) */ + if (msg->msg_controllen != 0 && + msg->msg_controllen < SENDCMSGBUFLEN) + { + memset(cmsgbuf + msg->msg_controllen, 0, + SENDCMSGBUFLEN - msg->msg_controllen); + } + } +#endif /* if defined(IP_TOS) || (defined(IPPROTO_IPV6) && \ + * defined(IPV6_TCLASS)) \ + * */ +#endif /* USE_CMSG */ + + if (write_countp != NULL) { + *write_countp = write_count; + } +} + +/* + * Construct an iov array and attach it to the msghdr passed in. This is + * the RECV constructor, which will use the available region of the buffer + * (if using a buffer list) or will use the internal region (if a single + * buffer I/O is requested). + * + * Nothing can be NULL, and the done event must list at least one buffer + * on the buffer linked list for this function to be meaningful. + * + * If read_countp != NULL, *read_countp will hold the number of bytes + * this transaction can receive. + */ +static void +build_msghdr_recv(isc_socket_t *sock, char *cmsgbuf, isc_socketevent_t *dev, + struct msghdr *msg, struct iovec *iov, size_t *read_countp) { + unsigned int iovcount; + size_t read_count; + + memset(msg, 0, sizeof(struct msghdr)); + + if (sock->type == isc_sockettype_udp) { + memset(&dev->address, 0, sizeof(dev->address)); + msg->msg_name = (void *)&dev->address.type.sa; + msg->msg_namelen = sizeof(dev->address.type); + } else { /* TCP */ + msg->msg_name = NULL; + msg->msg_namelen = 0; + dev->address = sock->peer_address; + } + + read_count = dev->region.length - dev->n; + iov[0].iov_base = (void *)(dev->region.base + dev->n); + iov[0].iov_len = read_count; + iovcount = 1; + + /* + * If needed, set up to receive that one extra byte. + */ +#ifdef ISC_PLATFORM_RECVOVERFLOW + if (sock->type == isc_sockettype_udp) { + INSIST(iovcount < MAXSCATTERGATHER_RECV); + iov[iovcount].iov_base = (void *)(&sock->overflow); + iov[iovcount].iov_len = 1; + iovcount++; + } +#endif /* ifdef ISC_PLATFORM_RECVOVERFLOW */ + + msg->msg_iov = iov; + msg->msg_iovlen = iovcount; + +#if defined(USE_CMSG) + msg->msg_control = cmsgbuf; + msg->msg_controllen = RECVCMSGBUFLEN; +#else /* if defined(USE_CMSG) */ + msg->msg_control = NULL; + msg->msg_controllen = 0; +#endif /* USE_CMSG */ + msg->msg_flags = 0; + + if (read_countp != NULL) { + *read_countp = read_count; + } +} + +static void +set_dev_address(const isc_sockaddr_t *address, isc_socket_t *sock, + isc_socketevent_t *dev) { + if (sock->type == isc_sockettype_udp) { + if (address != NULL) { + dev->address = *address; + } else { + dev->address = sock->peer_address; + } + } else if (sock->type == isc_sockettype_tcp) { + INSIST(address == NULL); + dev->address = sock->peer_address; + } +} + +static void +destroy_socketevent(isc_event_t *event) { + isc_socketevent_t *ev = (isc_socketevent_t *)event; + + (ev->destroy)(event); +} + +static isc_socketevent_t * +allocate_socketevent(isc_mem_t *mctx, void *sender, isc_eventtype_t eventtype, + isc_taskaction_t action, void *arg) { + isc_socketevent_t *ev; + + ev = (isc_socketevent_t *)isc_event_allocate(mctx, sender, eventtype, + action, arg, sizeof(*ev)); + + ev->result = ISC_R_UNSET; + ISC_LINK_INIT(ev, ev_link); + ev->region.base = NULL; + ev->n = 0; + ev->offset = 0; + ev->attributes = 0; + ev->destroy = ev->ev_destroy; + ev->ev_destroy = destroy_socketevent; + ev->dscp = 0; + + return (ev); +} + +#if defined(ISC_SOCKET_DEBUG) +static void +dump_msg(struct msghdr *msg) { + unsigned int i; + + printf("MSGHDR %p\n", msg); + printf("\tname %p, namelen %ld\n", msg->msg_name, + (long)msg->msg_namelen); + printf("\tiov %p, iovlen %ld\n", msg->msg_iov, (long)msg->msg_iovlen); + for (i = 0; i < (unsigned int)msg->msg_iovlen; i++) { + printf("\t\t%u\tbase %p, len %ld\n", i, + msg->msg_iov[i].iov_base, (long)msg->msg_iov[i].iov_len); + } + printf("\tcontrol %p, controllen %ld\n", msg->msg_control, + (long)msg->msg_controllen); +} +#endif /* if defined(ISC_SOCKET_DEBUG) */ + +#define DOIO_SUCCESS 0 /* i/o ok, event sent */ +#define DOIO_SOFT 1 /* i/o ok, soft error, no event sent */ +#define DOIO_HARD 2 /* i/o error, event sent */ +#define DOIO_EOF 3 /* EOF, no event sent */ + +static int +doio_recv(isc_socket_t *sock, isc_socketevent_t *dev) { + int cc; + struct iovec iov[MAXSCATTERGATHER_RECV]; + size_t read_count; + struct msghdr msghdr; + int recv_errno; + char strbuf[ISC_STRERRORSIZE]; + char cmsgbuf[RECVCMSGBUFLEN] = { 0 }; + + build_msghdr_recv(sock, cmsgbuf, dev, &msghdr, iov, &read_count); + +#if defined(ISC_SOCKET_DEBUG) + dump_msg(&msghdr); +#endif /* if defined(ISC_SOCKET_DEBUG) */ + + cc = recvmsg(sock->fd, &msghdr, 0); + recv_errno = errno; + +#if defined(ISC_SOCKET_DEBUG) + dump_msg(&msghdr); +#endif /* if defined(ISC_SOCKET_DEBUG) */ + + if (cc < 0) { + if (SOFT_ERROR(recv_errno)) { + return (DOIO_SOFT); + } + + if (isc_log_wouldlog(isc_lctx, IOEVENT_LEVEL)) { + strerror_r(recv_errno, strbuf, sizeof(strbuf)); + socket_log(sock, NULL, IOEVENT, + "doio_recv: recvmsg(%d) %d bytes, err %d/%s", + sock->fd, cc, recv_errno, strbuf); + } + +#define SOFT_OR_HARD(_system, _isc) \ + if (recv_errno == _system) { \ + if (sock->connected) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_RECVFAIL]); \ + return (DOIO_HARD); \ + } \ + return (DOIO_SOFT); \ + } +#define ALWAYS_HARD(_system, _isc) \ + if (recv_errno == _system) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_RECVFAIL]); \ + return (DOIO_HARD); \ + } + + SOFT_OR_HARD(ECONNREFUSED, ISC_R_CONNREFUSED); + SOFT_OR_HARD(ENETUNREACH, ISC_R_NETUNREACH); + SOFT_OR_HARD(EHOSTUNREACH, ISC_R_HOSTUNREACH); + SOFT_OR_HARD(EHOSTDOWN, ISC_R_HOSTDOWN); + SOFT_OR_HARD(ENOBUFS, ISC_R_NORESOURCES); + /* + * Older operating systems may still return EPROTO in some + * situations, for example when receiving ICMP/ICMPv6 errors. + * A real life scenario is when ICMPv6 returns code 5 or 6. + * These codes are introduced in RFC 4443 from March 2006, + * and the document obsoletes RFC 1885. But unfortunately not + * all operating systems have caught up with the new standard + * (in 2020) and thus a generic protocol error is returned. + */ + SOFT_OR_HARD(EPROTO, ISC_R_HOSTUNREACH); + /* Should never get this one but it was seen. */ +#ifdef ENOPROTOOPT + SOFT_OR_HARD(ENOPROTOOPT, ISC_R_HOSTUNREACH); +#endif /* ifdef ENOPROTOOPT */ + SOFT_OR_HARD(EINVAL, ISC_R_HOSTUNREACH); + +#undef SOFT_OR_HARD +#undef ALWAYS_HARD + + dev->result = isc__errno2result(recv_errno); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_RECVFAIL]); + return (DOIO_HARD); + } + + /* + * On TCP and UNIX sockets, zero length reads indicate EOF, + * while on UDP sockets, zero length reads are perfectly valid, + * although strange. + */ + switch (sock->type) { + case isc_sockettype_tcp: + case isc_sockettype_unix: + if (cc == 0) { + return (DOIO_EOF); + } + break; + case isc_sockettype_udp: + case isc_sockettype_raw: + break; + default: + UNREACHABLE(); + } + + if (sock->type == isc_sockettype_udp) { + dev->address.length = msghdr.msg_namelen; + if (isc_sockaddr_getport(&dev->address) == 0) { + if (isc_log_wouldlog(isc_lctx, IOEVENT_LEVEL)) { + socket_log(sock, &dev->address, IOEVENT, + "dropping source port zero packet"); + } + return (DOIO_SOFT); + } + /* + * Simulate a firewall blocking UDP responses bigger than + * 'maxudp' bytes. + */ + if (sock->manager->maxudp != 0 && + cc > (int)sock->manager->maxudp) + { + return (DOIO_SOFT); + } + } + + socket_log(sock, &dev->address, IOEVENT, "packet received correctly"); + + /* + * Overflow bit detection. If we received MORE bytes than we should, + * this indicates an overflow situation. Set the flag in the + * dev entry and adjust how much we read by one. + */ +#ifdef ISC_PLATFORM_RECVOVERFLOW + if ((sock->type == isc_sockettype_udp) && ((size_t)cc > read_count)) { + dev->attributes |= ISC_SOCKEVENTATTR_TRUNC; + cc--; + } +#endif /* ifdef ISC_PLATFORM_RECVOVERFLOW */ + + /* + * If there are control messages attached, run through them and pull + * out the interesting bits. + */ + process_cmsg(sock, &msghdr, dev); + + /* + * update the buffers (if any) and the i/o count + */ + dev->n += cc; + + /* + * If we read less than we expected, update counters, + * and let the upper layer poke the descriptor. + */ + if (((size_t)cc != read_count) && (dev->n < dev->minimum)) { + return (DOIO_SOFT); + } + + /* + * Full reads are posted, or partials if partials are ok. + */ + dev->result = ISC_R_SUCCESS; + return (DOIO_SUCCESS); +} + +/* + * Returns: + * DOIO_SUCCESS The operation succeeded. dev->result contains + * ISC_R_SUCCESS. + * + * DOIO_HARD A hard or unexpected I/O error was encountered. + * dev->result contains the appropriate error. + * + * DOIO_SOFT A soft I/O error was encountered. No senddone + * event was sent. The operation should be retried. + * + * No other return values are possible. + */ +static int +doio_send(isc_socket_t *sock, isc_socketevent_t *dev) { + int cc; + struct iovec iov[MAXSCATTERGATHER_SEND]; + size_t write_count; + struct msghdr msghdr; + char addrbuf[ISC_SOCKADDR_FORMATSIZE]; + int attempts = 0; + int send_errno; + char strbuf[ISC_STRERRORSIZE]; + char cmsgbuf[SENDCMSGBUFLEN] = { 0 }; + + build_msghdr_send(sock, cmsgbuf, dev, &msghdr, iov, &write_count); + +resend: + if (sock->type == isc_sockettype_udp && sock->manager->maxudp != 0 && + write_count > sock->manager->maxudp) + { + cc = write_count; + } else { + cc = sendmsg(sock->fd, &msghdr, 0); + } + send_errno = errno; + + /* + * Check for error or block condition. + */ + if (cc < 0) { + if (send_errno == EINTR && ++attempts < NRETRIES) { + goto resend; + } + + if (SOFT_ERROR(send_errno)) { + if (errno == EWOULDBLOCK || errno == EAGAIN) { + dev->result = ISC_R_WOULDBLOCK; + } + return (DOIO_SOFT); + } + +#define SOFT_OR_HARD(_system, _isc) \ + if (send_errno == _system) { \ + if (sock->connected) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_SENDFAIL]); \ + return (DOIO_HARD); \ + } \ + return (DOIO_SOFT); \ + } +#define ALWAYS_HARD(_system, _isc) \ + if (send_errno == _system) { \ + dev->result = _isc; \ + inc_stats(sock->manager->stats, \ + sock->statsindex[STATID_SENDFAIL]); \ + return (DOIO_HARD); \ + } + + SOFT_OR_HARD(ECONNREFUSED, ISC_R_CONNREFUSED); + ALWAYS_HARD(EACCES, ISC_R_NOPERM); + ALWAYS_HARD(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL); + ALWAYS_HARD(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + ALWAYS_HARD(EHOSTUNREACH, ISC_R_HOSTUNREACH); +#ifdef EHOSTDOWN + ALWAYS_HARD(EHOSTDOWN, ISC_R_HOSTUNREACH); +#endif /* ifdef EHOSTDOWN */ + ALWAYS_HARD(ENETUNREACH, ISC_R_NETUNREACH); + SOFT_OR_HARD(ENOBUFS, ISC_R_NORESOURCES); + ALWAYS_HARD(EPERM, ISC_R_HOSTUNREACH); + ALWAYS_HARD(EPIPE, ISC_R_NOTCONNECTED); + ALWAYS_HARD(ECONNRESET, ISC_R_CONNECTIONRESET); + +#undef SOFT_OR_HARD +#undef ALWAYS_HARD + + /* + * The other error types depend on whether or not the + * socket is UDP or TCP. If it is UDP, some errors + * that we expect to be fatal under TCP are merely + * annoying, and are really soft errors. + * + * However, these soft errors are still returned as + * a status. + */ + isc_sockaddr_format(&dev->address, addrbuf, sizeof(addrbuf)); + strerror_r(send_errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "internal_send: %s: %s", + addrbuf, strbuf); + dev->result = isc__errno2result(send_errno); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_SENDFAIL]); + return (DOIO_HARD); + } + + if (cc == 0) { + inc_stats(sock->manager->stats, + sock->statsindex[STATID_SENDFAIL]); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "doio_send: send() returned 0"); + } + + /* + * If we write less than we expected, update counters, poke. + */ + dev->n += cc; + if ((size_t)cc != write_count) { + return (DOIO_SOFT); + } + + /* + * Exactly what we wanted to write. We're done with this + * entry. Post its completion event. + */ + dev->result = ISC_R_SUCCESS; + return (DOIO_SUCCESS); +} + +/* + * Kill. + * + * Caller must ensure that the socket is not locked and no external + * references exist. + */ +static void +socketclose(isc__socketthread_t *thread, isc_socket_t *sock, int fd) { + int lockid = FDLOCK_ID(fd); + /* + * No one has this socket open, so the watcher doesn't have to be + * poked, and the socket doesn't have to be locked. + */ + LOCK(&thread->fdlock[lockid]); + thread->fds[fd] = NULL; + thread->fdstate[fd] = CLOSE_PENDING; + UNLOCK(&thread->fdlock[lockid]); + select_poke(thread->manager, thread->threadid, fd, SELECT_POKE_CLOSE); + + inc_stats(thread->manager->stats, sock->statsindex[STATID_CLOSE]); + + LOCK(&sock->lock); + if (sock->active == 1) { + dec_stats(thread->manager->stats, + sock->statsindex[STATID_ACTIVE]); + sock->active = 0; + } + UNLOCK(&sock->lock); + + /* + * update manager->maxfd here (XXX: this should be implemented more + * efficiently) + */ +#ifdef USE_SELECT + LOCK(&thread->manager->lock); + if (thread->maxfd == fd) { + int i; + + thread->maxfd = 0; + for (i = fd - 1; i >= 0; i--) { + lockid = FDLOCK_ID(i); + + LOCK(&thread->fdlock[lockid]); + if (thread->fdstate[i] == MANAGED) { + thread->maxfd = i; + UNLOCK(&thread->fdlock[lockid]); + break; + } + UNLOCK(&thread->fdlock[lockid]); + } + if (thread->maxfd < thread->pipe_fds[0]) { + thread->maxfd = thread->pipe_fds[0]; + } + } + + UNLOCK(&thread->manager->lock); +#endif /* USE_SELECT */ +} + +static void +destroy(isc_socket_t **sockp) { + int fd = 0; + isc_socket_t *sock = *sockp; + isc_socketmgr_t *manager = sock->manager; + isc__socketthread_t *thread = NULL; + + socket_log(sock, NULL, CREATION, "destroying"); + + isc_refcount_destroy(&sock->references); + + LOCK(&sock->lock); + INSIST(ISC_LIST_EMPTY(sock->connect_list)); + INSIST(ISC_LIST_EMPTY(sock->accept_list)); + INSIST(ISC_LIST_EMPTY(sock->recv_list)); + INSIST(ISC_LIST_EMPTY(sock->send_list)); + INSIST(sock->fd >= -1 && sock->fd < (int)manager->maxsocks); + + if (sock->fd >= 0) { + fd = sock->fd; + thread = &manager->threads[sock->threadid]; + sock->fd = -1; + sock->threadid = -1; + } + UNLOCK(&sock->lock); + + if (fd > 0) { + socketclose(thread, sock, fd); + } + + LOCK(&manager->lock); + + ISC_LIST_UNLINK(manager->socklist, sock, link); + + if (ISC_LIST_EMPTY(manager->socklist)) { + SIGNAL(&manager->shutdown_ok); + } + + /* can't unlock manager as its memory context is still used */ + free_socket(sockp); + + UNLOCK(&manager->lock); +} + +static isc_result_t +allocate_socket(isc_socketmgr_t *manager, isc_sockettype_t type, + isc_socket_t **socketp) { + isc_socket_t *sock; + + sock = isc_mem_get(manager->mctx, sizeof(*sock)); + + sock->magic = 0; + isc_refcount_init(&sock->references, 0); + + sock->manager = manager; + sock->type = type; + sock->fd = -1; + sock->threadid = -1; + sock->dscp = 0; /* TOS/TCLASS is zero until set. */ + sock->dupped = 0; + sock->statsindex = NULL; + sock->active = 0; + + ISC_LINK_INIT(sock, link); + + memset(sock->name, 0, sizeof(sock->name)); + sock->tag = NULL; + + /* + * Set up list of readers and writers to be initially empty. + */ + ISC_LIST_INIT(sock->recv_list); + ISC_LIST_INIT(sock->send_list); + ISC_LIST_INIT(sock->accept_list); + ISC_LIST_INIT(sock->connect_list); + + sock->listener = 0; + sock->connected = 0; + sock->connecting = 0; + sock->bound = 0; + sock->pktdscp = 0; + + /* + * Initialize the lock. + */ + isc_mutex_init(&sock->lock); + + sock->magic = SOCKET_MAGIC; + *socketp = sock; + + return (ISC_R_SUCCESS); +} + +/* + * This event requires that the various lists be empty, that the reference + * count be 1, and that the magic number is valid. The other socket bits, + * like the lock, must be initialized as well. The fd associated must be + * marked as closed, by setting it to -1 on close, or this routine will + * also close the socket. + */ +static void +free_socket(isc_socket_t **socketp) { + isc_socket_t *sock = *socketp; + *socketp = NULL; + + INSIST(VALID_SOCKET(sock)); + isc_refcount_destroy(&sock->references); + LOCK(&sock->lock); + INSIST(!sock->connecting); + INSIST(ISC_LIST_EMPTY(sock->recv_list)); + INSIST(ISC_LIST_EMPTY(sock->send_list)); + INSIST(ISC_LIST_EMPTY(sock->accept_list)); + INSIST(ISC_LIST_EMPTY(sock->connect_list)); + INSIST(!ISC_LINK_LINKED(sock, link)); + UNLOCK(&sock->lock); + + sock->magic = 0; + + isc_mutex_destroy(&sock->lock); + + isc_mem_put(sock->manager->mctx, sock, sizeof(*sock)); +} + +#if defined(SET_RCVBUF) +static isc_once_t rcvbuf_once = ISC_ONCE_INIT; +static int rcvbuf = ISC_RECV_BUFFER_SIZE; + +static void +set_rcvbuf(void) { + int fd; + int max = rcvbuf, min; + socklen_t len; + + fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); + if (fd == -1) { + switch (errno) { + case EPROTONOSUPPORT: + case EPFNOSUPPORT: + case EAFNOSUPPORT: + /* + * Linux 2.2 (and maybe others) return EINVAL instead of + * EAFNOSUPPORT. + */ + case EINVAL: + fd = socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP); + break; + } + } + if (fd == -1) { + return; + } + + len = sizeof(min); + if (getsockopt(fd, SOL_SOCKET, SO_RCVBUF, (void *)&min, &len) == 0 && + min < rcvbuf) + { + again: + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, (void *)&rcvbuf, + sizeof(rcvbuf)) == -1) + { + if (errno == ENOBUFS && rcvbuf > min) { + max = rcvbuf - 1; + rcvbuf = (rcvbuf + min) / 2; + goto again; + } else { + rcvbuf = min; + goto cleanup; + } + } else { + min = rcvbuf; + } + if (min != max) { + rcvbuf = max; + goto again; + } + } +cleanup: + close(fd); +} +#endif /* ifdef SO_RCVBUF */ + +#if defined(SET_SNDBUF) +static isc_once_t sndbuf_once = ISC_ONCE_INIT; +static int sndbuf = ISC_SEND_BUFFER_SIZE; + +static void +set_sndbuf(void) { + int fd; + int max = sndbuf, min; + socklen_t len; + + fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); + if (fd == -1) { + switch (errno) { + case EPROTONOSUPPORT: + case EPFNOSUPPORT: + case EAFNOSUPPORT: + /* + * Linux 2.2 (and maybe others) return EINVAL instead of + * EAFNOSUPPORT. + */ + case EINVAL: + fd = socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP); + break; + } + } + if (fd == -1) { + return; + } + + len = sizeof(min); + if (getsockopt(fd, SOL_SOCKET, SO_SNDBUF, (void *)&min, &len) == 0 && + min < sndbuf) + { + again: + if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, (void *)&sndbuf, + sizeof(sndbuf)) == -1) + { + if (errno == ENOBUFS && sndbuf > min) { + max = sndbuf - 1; + sndbuf = (sndbuf + min) / 2; + goto again; + } else { + sndbuf = min; + goto cleanup; + } + } else { + min = sndbuf; + } + if (min != max) { + sndbuf = max; + goto again; + } + } +cleanup: + close(fd); +} +#endif /* ifdef SO_SNDBUF */ + +static void +use_min_mtu(isc_socket_t *sock) { +#if !defined(IPV6_USE_MIN_MTU) && !defined(IPV6_MTU) + UNUSED(sock); +#endif /* if !defined(IPV6_USE_MIN_MTU) && !defined(IPV6_MTU) */ +#ifdef IPV6_USE_MIN_MTU + /* use minimum MTU */ + if (sock->pf == AF_INET6) { + int on = 1; + (void)setsockopt(sock->fd, IPPROTO_IPV6, IPV6_USE_MIN_MTU, + (void *)&on, sizeof(on)); + } +#endif /* ifdef IPV6_USE_MIN_MTU */ +#if defined(IPV6_MTU) + /* + * Use minimum MTU on IPv6 sockets. + */ + if (sock->pf == AF_INET6) { + int mtu = 1280; + (void)setsockopt(sock->fd, IPPROTO_IPV6, IPV6_MTU, &mtu, + sizeof(mtu)); + } +#endif /* if defined(IPV6_MTU) */ +} + +static void +set_tcp_maxseg(isc_socket_t *sock, int size) { +#ifdef TCP_MAXSEG + if (sock->type == isc_sockettype_tcp) { + (void)setsockopt(sock->fd, IPPROTO_TCP, TCP_MAXSEG, + (void *)&size, sizeof(size)); + } +#endif /* ifdef TCP_MAXSEG */ +} + +static void +set_ip_disable_pmtud(isc_socket_t *sock) { + /* + * Disable Path MTU Discover on IP packets + */ + if (sock->pf == AF_INET6) { +#if defined(IPV6_DONTFRAG) + (void)setsockopt(sock->fd, IPPROTO_IPV6, IPV6_DONTFRAG, + &(int){ 0 }, sizeof(int)); +#endif +#if defined(IPV6_MTU_DISCOVER) && defined(IP_PMTUDISC_OMIT) + (void)setsockopt(sock->fd, IPPROTO_IPV6, IPV6_MTU_DISCOVER, + &(int){ IP_PMTUDISC_OMIT }, sizeof(int)); +#endif + } else if (sock->pf == AF_INET) { +#if defined(IP_DONTFRAG) + (void)setsockopt(sock->fd, IPPROTO_IP, IP_DONTFRAG, &(int){ 0 }, + sizeof(int)); +#endif +#if defined(IP_MTU_DISCOVER) && defined(IP_PMTUDISC_OMIT) + (void)setsockopt(sock->fd, IPPROTO_IP, IP_MTU_DISCOVER, + &(int){ IP_PMTUDISC_OMIT }, sizeof(int)); +#endif + } +} + +static isc_result_t +opensocket(isc_socketmgr_t *manager, isc_socket_t *sock, + isc_socket_t *dup_socket) { + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + const char *err = "socket"; + int tries = 0; +#if defined(USE_CMSG) || defined(SO_NOSIGPIPE) + int on = 1; +#endif /* if defined(USE_CMSG) || defined(SO_NOSIGPIPE) */ +#if defined(SET_RCVBUF) || defined(SET_SNDBUF) + socklen_t optlen; + int size = 0; +#endif + +again: + if (dup_socket == NULL) { + switch (sock->type) { + case isc_sockettype_udp: + sock->fd = socket(sock->pf, SOCK_DGRAM, IPPROTO_UDP); + break; + case isc_sockettype_tcp: + sock->fd = socket(sock->pf, SOCK_STREAM, IPPROTO_TCP); + break; + case isc_sockettype_unix: + sock->fd = socket(sock->pf, SOCK_STREAM, 0); + break; + case isc_sockettype_raw: + errno = EPFNOSUPPORT; + /* + * PF_ROUTE is a alias for PF_NETLINK on linux. + */ +#if defined(PF_ROUTE) + if (sock->fd == -1 && sock->pf == PF_ROUTE) { +#ifdef NETLINK_ROUTE + sock->fd = socket(sock->pf, SOCK_RAW, + NETLINK_ROUTE); +#else /* ifdef NETLINK_ROUTE */ + sock->fd = socket(sock->pf, SOCK_RAW, 0); +#endif /* ifdef NETLINK_ROUTE */ + if (sock->fd != -1) { +#ifdef NETLINK_ROUTE + struct sockaddr_nl sa; + int n; + + /* + * Do an implicit bind. + */ + memset(&sa, 0, sizeof(sa)); + sa.nl_family = AF_NETLINK; + sa.nl_groups = RTMGRP_IPV4_IFADDR | + RTMGRP_IPV6_IFADDR; + n = bind(sock->fd, + (struct sockaddr *)&sa, + sizeof(sa)); + if (n < 0) { + close(sock->fd); + sock->fd = -1; + } +#endif /* ifdef NETLINK_ROUTE */ + sock->bound = 1; + } + } +#endif /* if defined(PF_ROUTE) */ + break; + } + } else { + sock->fd = dup(dup_socket->fd); + sock->dupped = 1; + sock->bound = dup_socket->bound; + } + if (sock->fd == -1 && errno == EINTR && tries++ < 42) { + goto again; + } + +#ifdef F_DUPFD + /* + * Leave a space for stdio and TCP to work in. + */ + if (manager->reserved != 0 && sock->type == isc_sockettype_udp && + sock->fd >= 0 && sock->fd < manager->reserved) + { + int newfd, tmp; + newfd = fcntl(sock->fd, F_DUPFD, manager->reserved); + tmp = errno; + (void)close(sock->fd); + errno = tmp; + sock->fd = newfd; + err = "isc_socket_create: fcntl/reserved"; + } else if (sock->fd >= 0 && sock->fd < 20) { + int newfd, tmp; + newfd = fcntl(sock->fd, F_DUPFD, 20); + tmp = errno; + (void)close(sock->fd); + errno = tmp; + sock->fd = newfd; + err = "isc_socket_create: fcntl"; + } +#endif /* ifdef F_DUPFD */ + + if (sock->fd >= (int)manager->maxsocks) { + (void)close(sock->fd); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "socket: file descriptor exceeds limit (%d/%u)", + sock->fd, manager->maxsocks); + inc_stats(manager->stats, sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_NORESOURCES); + } + + if (sock->fd < 0) { + switch (errno) { + case EMFILE: + case ENFILE: + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "%s: %s", err, strbuf); + FALLTHROUGH; + case ENOBUFS: + inc_stats(manager->stats, + sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_NORESOURCES); + + case EPROTONOSUPPORT: + case EPFNOSUPPORT: + case EAFNOSUPPORT: + /* + * Linux 2.2 (and maybe others) return EINVAL instead of + * EAFNOSUPPORT. + */ + case EINVAL: + inc_stats(manager->stats, + sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_FAMILYNOSUPPORT); + + default: + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "%s() failed: %s", + err, strbuf); + inc_stats(manager->stats, + sock->statsindex[STATID_OPENFAIL]); + return (ISC_R_UNEXPECTED); + } + } + + if (dup_socket != NULL) { + goto setup_done; + } + + result = make_nonblock(sock->fd); + if (result != ISC_R_SUCCESS) { + (void)close(sock->fd); + inc_stats(manager->stats, sock->statsindex[STATID_OPENFAIL]); + return (result); + } + +#ifdef SO_NOSIGPIPE + if (setsockopt(sock->fd, SOL_SOCKET, SO_NOSIGPIPE, (void *)&on, + sizeof(on)) < 0) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_NOSIGPIPE) failed: %s", + sock->fd, strbuf); + /* Press on... */ + } +#endif /* ifdef SO_NOSIGPIPE */ + + /* + * Use minimum mtu if possible. + */ + if (sock->type == isc_sockettype_tcp && sock->pf == AF_INET6) { + use_min_mtu(sock); + set_tcp_maxseg(sock, 1280 - 20 - 40); /* 1280 - TCP - IPV6 */ + } + +#if defined(USE_CMSG) || defined(SET_RCVBUF) || defined(SET_SNDBUF) + if (sock->type == isc_sockettype_udp) { +#if defined(USE_CMSG) +#if defined(SO_TIMESTAMP) + if (setsockopt(sock->fd, SOL_SOCKET, SO_TIMESTAMP, (void *)&on, + sizeof(on)) < 0 && + errno != ENOPROTOOPT) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_TIMESTAMP) failed: " + "%s", + sock->fd, strbuf); + /* Press on... */ + } +#endif /* SO_TIMESTAMP */ + +#ifdef IPV6_RECVPKTINFO + /* RFC 3542 */ + if ((sock->pf == AF_INET6) && + (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_RECVPKTINFO, + (void *)&on, sizeof(on)) < 0)) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_RECVPKTINFO) " + "failed: %s", + sock->fd, strbuf); + } +#else /* ifdef IPV6_RECVPKTINFO */ + /* RFC 2292 */ + if ((sock->pf == AF_INET6) && + (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_PKTINFO, + (void *)&on, sizeof(on)) < 0)) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_PKTINFO) failed: " + "%s", + sock->fd, strbuf); + } +#endif /* IPV6_RECVPKTINFO */ +#endif /* defined(USE_CMSG) */ + +#if defined(SET_RCVBUF) + optlen = sizeof(size); + if (getsockopt(sock->fd, SOL_SOCKET, SO_RCVBUF, (void *)&size, + &optlen) == 0 && + size < rcvbuf) + { + RUNTIME_CHECK(isc_once_do(&rcvbuf_once, set_rcvbuf) == + ISC_R_SUCCESS); + if (setsockopt(sock->fd, SOL_SOCKET, SO_RCVBUF, + (void *)&rcvbuf, sizeof(rcvbuf)) == -1) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_RCVBUF, " + "%d) failed: %s", + sock->fd, rcvbuf, strbuf); + } + } +#endif /* if defined(SET_RCVBUF) */ + +#if defined(SET_SNDBUF) + optlen = sizeof(size); + if (getsockopt(sock->fd, SOL_SOCKET, SO_SNDBUF, (void *)&size, + &optlen) == 0 && + size < sndbuf) + { + RUNTIME_CHECK(isc_once_do(&sndbuf_once, set_sndbuf) == + ISC_R_SUCCESS); + if (setsockopt(sock->fd, SOL_SOCKET, SO_SNDBUF, + (void *)&sndbuf, sizeof(sndbuf)) == -1) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, SO_SNDBUF, " + "%d) failed: %s", + sock->fd, sndbuf, strbuf); + } + } +#endif /* if defined(SO_SNDBUF) */ + } +#ifdef IPV6_RECVTCLASS + if ((sock->pf == AF_INET6) && + (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_RECVTCLASS, (void *)&on, + sizeof(on)) < 0)) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_RECVTCLASS) " + "failed: %s", + sock->fd, strbuf); + } +#endif /* ifdef IPV6_RECVTCLASS */ +#ifdef IP_RECVTOS + if ((sock->pf == AF_INET) && + (setsockopt(sock->fd, IPPROTO_IP, IP_RECVTOS, (void *)&on, + sizeof(on)) < 0)) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IP_RECVTOS) " + "failed: %s", + sock->fd, strbuf); + } +#endif /* ifdef IP_RECVTOS */ +#endif /* defined(USE_CMSG) || defined(SET_RCVBUF) || defined(SET_SNDBUF) */ + + set_ip_disable_pmtud(sock); + +setup_done: + inc_stats(manager->stats, sock->statsindex[STATID_OPEN]); + if (sock->active == 0) { + inc_stats(manager->stats, sock->statsindex[STATID_ACTIVE]); + sock->active = 1; + } + + return (ISC_R_SUCCESS); +} + +/* + * Create a 'type' socket or duplicate an existing socket, managed + * by 'manager'. Events will be posted to 'task' and when dispatched + * 'action' will be called with 'arg' as the arg value. The new + * socket is returned in 'socketp'. + */ +static isc_result_t +socket_create(isc_socketmgr_t *manager, int pf, isc_sockettype_t type, + isc_socket_t **socketp, isc_socket_t *dup_socket) { + isc_socket_t *sock = NULL; + isc__socketthread_t *thread; + isc_result_t result; + int lockid; + + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(socketp != NULL && *socketp == NULL); + + result = allocate_socket(manager, type, &sock); + if (result != ISC_R_SUCCESS) { + return (result); + } + + switch (sock->type) { + case isc_sockettype_udp: + sock->statsindex = (pf == AF_INET) ? udp4statsindex + : udp6statsindex; +#define DCSPPKT(pf) ((pf == AF_INET) ? ISC_NET_DSCPPKTV4 : ISC_NET_DSCPPKTV6) + sock->pktdscp = (isc_net_probedscp() & DCSPPKT(pf)) != 0; + break; + case isc_sockettype_tcp: + sock->statsindex = (pf == AF_INET) ? tcp4statsindex + : tcp6statsindex; + break; + case isc_sockettype_unix: + sock->statsindex = unixstatsindex; + break; + case isc_sockettype_raw: + sock->statsindex = rawstatsindex; + break; + default: + UNREACHABLE(); + } + + sock->pf = pf; + + result = opensocket(manager, sock, dup_socket); + if (result != ISC_R_SUCCESS) { + free_socket(&sock); + return (result); + } + + if (sock->fd == -1) { + abort(); + } + sock->threadid = gen_threadid(sock); + isc_refcount_increment0(&sock->references); + thread = &manager->threads[sock->threadid]; + *socketp = sock; + + /* + * Note we don't have to lock the socket like we normally would because + * there are no external references to it yet. + */ + + lockid = FDLOCK_ID(sock->fd); + LOCK(&thread->fdlock[lockid]); + thread->fds[sock->fd] = sock; + thread->fdstate[sock->fd] = MANAGED; +#if defined(USE_EPOLL) + thread->epoll_events[sock->fd] = 0; +#endif /* if defined(USE_EPOLL) */ +#ifdef USE_DEVPOLL + INSIST(thread->fdpollinfo[sock->fd].want_read == 0 && + thread->fdpollinfo[sock->fd].want_write == 0); +#endif /* ifdef USE_DEVPOLL */ + UNLOCK(&thread->fdlock[lockid]); + + LOCK(&manager->lock); + ISC_LIST_APPEND(manager->socklist, sock, link); +#ifdef USE_SELECT + if (thread->maxfd < sock->fd) { + thread->maxfd = sock->fd; + } +#endif /* ifdef USE_SELECT */ + UNLOCK(&manager->lock); + + socket_log(sock, NULL, CREATION, + dup_socket != NULL ? "dupped" : "created"); + + return (ISC_R_SUCCESS); +} + +/*% + * Create a new 'type' socket managed by 'manager'. Events + * will be posted to 'task' and when dispatched 'action' will be + * called with 'arg' as the arg value. The new socket is returned + * in 'socketp'. + */ +isc_result_t +isc_socket_create(isc_socketmgr_t *manager0, int pf, isc_sockettype_t type, + isc_socket_t **socketp) { + return (socket_create(manager0, pf, type, socketp, NULL)); +} + +/*% + * Duplicate an existing socket. The new socket is returned + * in 'socketp'. + */ +isc_result_t +isc_socket_dup(isc_socket_t *sock, isc_socket_t **socketp) { + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(socketp != NULL && *socketp == NULL); + + return (socket_create(sock->manager, sock->pf, sock->type, socketp, + sock)); +} + +isc_result_t +isc_socket_open(isc_socket_t *sock) { + isc_result_t result; + isc__socketthread_t *thread; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + REQUIRE(isc_refcount_current(&sock->references) >= 1); + REQUIRE(sock->fd == -1); + REQUIRE(sock->threadid == -1); + + result = opensocket(sock->manager, sock, NULL); + + UNLOCK(&sock->lock); + + if (result != ISC_R_SUCCESS) { + sock->fd = -1; + } else { + sock->threadid = gen_threadid(sock); + thread = &sock->manager->threads[sock->threadid]; + int lockid = FDLOCK_ID(sock->fd); + + LOCK(&thread->fdlock[lockid]); + thread->fds[sock->fd] = sock; + thread->fdstate[sock->fd] = MANAGED; +#if defined(USE_EPOLL) + thread->epoll_events[sock->fd] = 0; +#endif /* if defined(USE_EPOLL) */ +#ifdef USE_DEVPOLL + INSIST(thread->fdpollinfo[sock->fd].want_read == 0 && + thread->fdpollinfo[sock->fd].want_write == 0); +#endif /* ifdef USE_DEVPOLL */ + UNLOCK(&thread->fdlock[lockid]); + +#ifdef USE_SELECT + LOCK(&sock->manager->lock); + if (thread->maxfd < sock->fd) { + thread->maxfd = sock->fd; + } + UNLOCK(&sock->manager->lock); +#endif /* ifdef USE_SELECT */ + } + + return (result); +} + +/* + * Attach to a socket. Caller must explicitly detach when it is done. + */ +void +isc_socket_attach(isc_socket_t *sock, isc_socket_t **socketp) { + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(socketp != NULL && *socketp == NULL); + + int old_refs = isc_refcount_increment(&sock->references); + REQUIRE(old_refs > 0); + + *socketp = sock; +} + +/* + * Dereference a socket. If this is the last reference to it, clean things + * up by destroying the socket. + */ +void +isc_socket_detach(isc_socket_t **socketp) { + isc_socket_t *sock; + + REQUIRE(socketp != NULL); + sock = *socketp; + REQUIRE(VALID_SOCKET(sock)); + if (isc_refcount_decrement(&sock->references) == 1) { + destroy(&sock); + } + + *socketp = NULL; +} + +isc_result_t +isc_socket_close(isc_socket_t *sock) { + int fd; + isc_socketmgr_t *manager; + isc__socketthread_t *thread; + fflush(stdout); + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + REQUIRE(sock->fd >= 0 && sock->fd < (int)sock->manager->maxsocks); + + INSIST(!sock->connecting); + INSIST(ISC_LIST_EMPTY(sock->recv_list)); + INSIST(ISC_LIST_EMPTY(sock->send_list)); + INSIST(ISC_LIST_EMPTY(sock->accept_list)); + INSIST(ISC_LIST_EMPTY(sock->connect_list)); + + manager = sock->manager; + thread = &manager->threads[sock->threadid]; + fd = sock->fd; + sock->fd = -1; + sock->threadid = -1; + + sock->dupped = 0; + memset(sock->name, 0, sizeof(sock->name)); + sock->tag = NULL; + sock->listener = 0; + sock->connected = 0; + sock->connecting = 0; + sock->bound = 0; + isc_sockaddr_any(&sock->peer_address); + + UNLOCK(&sock->lock); + + socketclose(thread, sock, fd); + + return (ISC_R_SUCCESS); +} + +/* + * Dequeue an item off the given socket's read queue, set the result code + * in the done event to the one provided, and send it to the task it was + * destined for. + * + * If the event to be sent is on a list, remove it before sending. If + * asked to, send and detach from the socket as well. + * + * Caller must have the socket locked if the event is attached to the socket. + */ +static void +send_recvdone_event(isc_socket_t *sock, isc_socketevent_t **dev) { + isc_task_t *task; + + task = (*dev)->ev_sender; + + (*dev)->ev_sender = sock; + + if (ISC_LINK_LINKED(*dev, ev_link)) { + ISC_LIST_DEQUEUE(sock->recv_list, *dev, ev_link); + } + + if (((*dev)->attributes & ISC_SOCKEVENTATTR_ATTACHED) != 0) { + isc_task_sendtoanddetach(&task, (isc_event_t **)dev, + sock->threadid); + } else { + isc_task_sendto(task, (isc_event_t **)dev, sock->threadid); + } +} + +/* + * See comments for send_recvdone_event() above. + * + * Caller must have the socket locked if the event is attached to the socket. + */ +static void +send_senddone_event(isc_socket_t *sock, isc_socketevent_t **dev) { + isc_task_t *task; + + INSIST(dev != NULL && *dev != NULL); + + task = (*dev)->ev_sender; + (*dev)->ev_sender = sock; + + if (ISC_LINK_LINKED(*dev, ev_link)) { + ISC_LIST_DEQUEUE(sock->send_list, *dev, ev_link); + } + + if (((*dev)->attributes & ISC_SOCKEVENTATTR_ATTACHED) != 0) { + isc_task_sendtoanddetach(&task, (isc_event_t **)dev, + sock->threadid); + } else { + isc_task_sendto(task, (isc_event_t **)dev, sock->threadid); + } +} + +/* + * See comments for send_recvdone_event() above. + * + * Caller must have the socket locked if the event is attached to the socket. + */ +static void +send_connectdone_event(isc_socket_t *sock, isc_socket_connev_t **dev) { + isc_task_t *task; + + INSIST(dev != NULL && *dev != NULL); + + task = (*dev)->ev_sender; + (*dev)->ev_sender = sock; + + if (ISC_LINK_LINKED(*dev, ev_link)) { + ISC_LIST_DEQUEUE(sock->connect_list, *dev, ev_link); + } + + isc_task_sendtoanddetach(&task, (isc_event_t **)dev, sock->threadid); +} + +/* + * Call accept() on a socket, to get the new file descriptor. The listen + * socket is used as a prototype to create a new isc_socket_t. The new + * socket has one outstanding reference. The task receiving the event + * will be detached from just after the event is delivered. + * + * On entry to this function, the event delivered is the internal + * readable event, and the first item on the accept_list should be + * the done event we want to send. If the list is empty, this is a no-op, + * so just unlock and return. + */ +static void +internal_accept(isc_socket_t *sock) { + isc_socketmgr_t *manager; + isc__socketthread_t *thread, *nthread; + isc_socket_newconnev_t *dev; + isc_task_t *task; + socklen_t addrlen; + int fd; + isc_result_t result = ISC_R_SUCCESS; + char strbuf[ISC_STRERRORSIZE]; + const char *err = "accept"; + + INSIST(VALID_SOCKET(sock)); + REQUIRE(sock->fd >= 0); + + socket_log(sock, NULL, TRACE, "internal_accept called, locked socket"); + + manager = sock->manager; + INSIST(VALID_MANAGER(manager)); + thread = &manager->threads[sock->threadid]; + + INSIST(sock->listener); + + /* + * Get the first item off the accept list. + * If it is empty, unlock the socket and return. + */ + dev = ISC_LIST_HEAD(sock->accept_list); + if (dev == NULL) { + unwatch_fd(thread, sock->fd, SELECT_POKE_ACCEPT); + UNLOCK(&sock->lock); + return; + } + + /* + * Try to accept the new connection. If the accept fails with + * EAGAIN or EINTR, simply poke the watcher to watch this socket + * again. Also ignore ECONNRESET, which has been reported to + * be spuriously returned on Linux 2.2.19 although it is not + * a documented error for accept(). ECONNABORTED has been + * reported for Solaris 8. The rest are thrown in not because + * we have seen them but because they are ignored by other + * daemons such as BIND 8 and Apache. + */ + + addrlen = sizeof(NEWCONNSOCK(dev)->peer_address.type); + memset(&NEWCONNSOCK(dev)->peer_address.type, 0, addrlen); + fd = accept(sock->fd, &NEWCONNSOCK(dev)->peer_address.type.sa, + (void *)&addrlen); + +#ifdef F_DUPFD + /* + * Leave a space for stdio to work in. + */ + if (fd >= 0 && fd < 20) { + int newfd, tmp; + newfd = fcntl(fd, F_DUPFD, 20); + tmp = errno; + (void)close(fd); + errno = tmp; + fd = newfd; + err = "accept/fcntl"; + } +#endif /* ifdef F_DUPFD */ + + if (fd < 0) { + if (SOFT_ERROR(errno)) { + goto soft_error; + } + switch (errno) { + case ENFILE: + case EMFILE: + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "%s: too many open file descriptors", + err); + goto soft_error; + + case ENOBUFS: + case ENOMEM: + case ECONNRESET: + case ECONNABORTED: + case EHOSTUNREACH: + case EHOSTDOWN: + case ENETUNREACH: + case ENETDOWN: + case ECONNREFUSED: +#ifdef EPROTO + case EPROTO: +#endif /* ifdef EPROTO */ +#ifdef ENONET + case ENONET: +#endif /* ifdef ENONET */ + goto soft_error; + default: + break; + } + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_accept: %s() failed: %s", err, + strbuf); + fd = -1; + result = ISC_R_UNEXPECTED; + } else { + if (addrlen == 0U) { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_accept(): " + "accept() failed to return " + "remote address"); + + (void)close(fd); + goto soft_error; + } else if (NEWCONNSOCK(dev)->peer_address.type.sa.sa_family != + sock->pf) + { + UNEXPECTED_ERROR( + __FILE__, __LINE__, + "internal_accept(): " + "accept() returned peer address " + "family %u (expected %u)", + NEWCONNSOCK(dev)->peer_address.type.sa.sa_family, + sock->pf); + (void)close(fd); + goto soft_error; + } else if (fd >= (int)manager->maxsocks) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "accept: file descriptor exceeds limit " + "(%d/%u)", + fd, manager->maxsocks); + (void)close(fd); + goto soft_error; + } + } + + if (fd != -1) { + NEWCONNSOCK(dev)->peer_address.length = addrlen; + NEWCONNSOCK(dev)->pf = sock->pf; + } + + /* + * Pull off the done event. + */ + ISC_LIST_UNLINK(sock->accept_list, dev, ev_link); + + /* + * Poke watcher if there are more pending accepts. + */ + if (ISC_LIST_EMPTY(sock->accept_list)) { + unwatch_fd(thread, sock->fd, SELECT_POKE_ACCEPT); + } + + if (fd != -1) { + result = make_nonblock(fd); + if (result != ISC_R_SUCCESS) { + (void)close(fd); + fd = -1; + } + } + + /* + * We need to unlock sock->lock now to be able to lock manager->lock + * without risking a deadlock with xmlstats. + */ + UNLOCK(&sock->lock); + + /* + * -1 means the new socket didn't happen. + */ + if (fd != -1) { + int lockid = FDLOCK_ID(fd); + + NEWCONNSOCK(dev)->fd = fd; + NEWCONNSOCK(dev)->threadid = gen_threadid(NEWCONNSOCK(dev)); + NEWCONNSOCK(dev)->bound = 1; + NEWCONNSOCK(dev)->connected = 1; + nthread = &manager->threads[NEWCONNSOCK(dev)->threadid]; + + /* + * We already hold a lock on one fdlock in accepting thread, + * we need to make sure that we don't double lock. + */ + bool same_bucket = (sock->threadid == + NEWCONNSOCK(dev)->threadid) && + (FDLOCK_ID(sock->fd) == lockid); + + /* + * Use minimum mtu if possible. + */ + use_min_mtu(NEWCONNSOCK(dev)); + set_tcp_maxseg(NEWCONNSOCK(dev), 1280 - 20 - 40); + + /* + * Ensure DSCP settings are inherited across accept. + */ + setdscp(NEWCONNSOCK(dev), sock->dscp); + + /* + * Save away the remote address + */ + dev->address = NEWCONNSOCK(dev)->peer_address; + + if (NEWCONNSOCK(dev)->active == 0) { + inc_stats(manager->stats, + NEWCONNSOCK(dev)->statsindex[STATID_ACTIVE]); + NEWCONNSOCK(dev)->active = 1; + } + + if (!same_bucket) { + LOCK(&nthread->fdlock[lockid]); + } + nthread->fds[fd] = NEWCONNSOCK(dev); + nthread->fdstate[fd] = MANAGED; +#if defined(USE_EPOLL) + nthread->epoll_events[fd] = 0; +#endif /* if defined(USE_EPOLL) */ + if (!same_bucket) { + UNLOCK(&nthread->fdlock[lockid]); + } + + LOCK(&manager->lock); + +#ifdef USE_SELECT + if (nthread->maxfd < fd) { + nthread->maxfd = fd; + } +#endif /* ifdef USE_SELECT */ + + socket_log(sock, &NEWCONNSOCK(dev)->peer_address, CREATION, + "accepted connection, new socket %p", + dev->newsocket); + + ISC_LIST_APPEND(manager->socklist, NEWCONNSOCK(dev), link); + + UNLOCK(&manager->lock); + + inc_stats(manager->stats, sock->statsindex[STATID_ACCEPT]); + } else { + inc_stats(manager->stats, sock->statsindex[STATID_ACCEPTFAIL]); + isc_refcount_decrementz(&NEWCONNSOCK(dev)->references); + free_socket((isc_socket_t **)&dev->newsocket); + } + + /* + * Fill in the done event details and send it off. + */ + dev->result = result; + task = dev->ev_sender; + dev->ev_sender = sock; + + isc_task_sendtoanddetach(&task, ISC_EVENT_PTR(&dev), sock->threadid); + return; + +soft_error: + watch_fd(thread, sock->fd, SELECT_POKE_ACCEPT); + UNLOCK(&sock->lock); + + inc_stats(manager->stats, sock->statsindex[STATID_ACCEPTFAIL]); + return; +} + +static void +internal_recv(isc_socket_t *sock) { + isc_socketevent_t *dev; + + INSIST(VALID_SOCKET(sock)); + REQUIRE(sock->fd >= 0); + + dev = ISC_LIST_HEAD(sock->recv_list); + if (dev == NULL) { + goto finish; + } + + socket_log(sock, NULL, IOEVENT, "internal_recv: event %p -> task %p", + dev, dev->ev_sender); + + /* + * Try to do as much I/O as possible on this socket. There are no + * limits here, currently. + */ + while (dev != NULL) { + switch (doio_recv(sock, dev)) { + case DOIO_SOFT: + goto finish; + + case DOIO_EOF: + /* + * read of 0 means the remote end was closed. + * Run through the event queue and dispatch all + * the events with an EOF result code. + */ + do { + dev->result = ISC_R_EOF; + send_recvdone_event(sock, &dev); + dev = ISC_LIST_HEAD(sock->recv_list); + } while (dev != NULL); + goto finish; + + case DOIO_SUCCESS: + case DOIO_HARD: + send_recvdone_event(sock, &dev); + break; + } + + dev = ISC_LIST_HEAD(sock->recv_list); + } + +finish: + if (ISC_LIST_EMPTY(sock->recv_list)) { + unwatch_fd(&sock->manager->threads[sock->threadid], sock->fd, + SELECT_POKE_READ); + } +} + +static void +internal_send(isc_socket_t *sock) { + isc_socketevent_t *dev; + + INSIST(VALID_SOCKET(sock)); + REQUIRE(sock->fd >= 0); + + dev = ISC_LIST_HEAD(sock->send_list); + if (dev == NULL) { + goto finish; + } + socket_log(sock, NULL, EVENT, "internal_send: event %p -> task %p", dev, + dev->ev_sender); + + /* + * Try to do as much I/O as possible on this socket. There are no + * limits here, currently. + */ + while (dev != NULL) { + switch (doio_send(sock, dev)) { + case DOIO_SOFT: + goto finish; + + case DOIO_HARD: + case DOIO_SUCCESS: + send_senddone_event(sock, &dev); + break; + } + + dev = ISC_LIST_HEAD(sock->send_list); + } + +finish: + if (ISC_LIST_EMPTY(sock->send_list)) { + unwatch_fd(&sock->manager->threads[sock->threadid], sock->fd, + SELECT_POKE_WRITE); + } +} + +/* + * Process read/writes on each fd here. Avoid locking + * and unlocking twice if both reads and writes are possible. + */ +static void +process_fd(isc__socketthread_t *thread, int fd, bool readable, bool writeable) { + isc_socket_t *sock; + int lockid = FDLOCK_ID(fd); + + /* + * If the socket is going to be closed, don't do more I/O. + */ + LOCK(&thread->fdlock[lockid]); + if (thread->fdstate[fd] == CLOSE_PENDING) { + UNLOCK(&thread->fdlock[lockid]); + + (void)unwatch_fd(thread, fd, SELECT_POKE_READ); + (void)unwatch_fd(thread, fd, SELECT_POKE_WRITE); + return; + } + + sock = thread->fds[fd]; + if (sock == NULL) { + UNLOCK(&thread->fdlock[lockid]); + return; + } + + LOCK(&sock->lock); + + if (sock->fd < 0) { + /* + * Sock is being closed - the final external reference + * is gone but it was not yet removed from event loop + * and fdstate[]/fds[] as destroy() is waiting on + * thread->fdlock[lockid] or sock->lock that we're holding. + * Just release the locks and bail. + */ + UNLOCK(&sock->lock); + UNLOCK(&thread->fdlock[lockid]); + return; + } + + REQUIRE(readable || writeable); + if (writeable) { + if (sock->connecting) { + internal_connect(sock); + } else { + internal_send(sock); + } + } + + if (readable) { + if (sock->listener) { + internal_accept(sock); /* unlocks sock */ + } else { + internal_recv(sock); + UNLOCK(&sock->lock); + } + } else { + UNLOCK(&sock->lock); + } + + UNLOCK(&thread->fdlock[lockid]); + + /* + * Socket destruction might be pending, it will resume + * after releasing fdlock and sock->lock. + */ +} + +/* + * process_fds is different for different event loops + * it takes the events from event loops and for each FD + * launches process_fd + */ +#ifdef USE_KQUEUE +static bool +process_fds(isc__socketthread_t *thread, struct kevent *events, int nevents) { + int i; + bool readable, writable; + bool done = false; + bool have_ctlevent = false; + if (nevents == thread->nevents) { + /* + * This is not an error, but something unexpected. If this + * happens, it may indicate the need for increasing + * ISC_SOCKET_MAXEVENTS. + */ + thread_log(thread, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_INFO, + "maximum number of FD events (%d) received", + nevents); + } + + for (i = 0; i < nevents; i++) { + REQUIRE(events[i].ident < thread->manager->maxsocks); + if (events[i].ident == (uintptr_t)thread->pipe_fds[0]) { + have_ctlevent = true; + continue; + } + readable = (events[i].filter == EVFILT_READ); + writable = (events[i].filter == EVFILT_WRITE); + process_fd(thread, events[i].ident, readable, writable); + } + + if (have_ctlevent) { + done = process_ctlfd(thread); + } + + return (done); +} +#elif defined(USE_EPOLL) +static bool +process_fds(isc__socketthread_t *thread, struct epoll_event *events, + int nevents) { + int i; + bool done = false; + bool have_ctlevent = false; + + if (nevents == thread->nevents) { + thread_log(thread, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_INFO, + "maximum number of FD events (%d) received", + nevents); + } + + for (i = 0; i < nevents; i++) { + REQUIRE(events[i].data.fd < (int)thread->manager->maxsocks); + if (events[i].data.fd == thread->pipe_fds[0]) { + have_ctlevent = true; + continue; + } + if ((events[i].events & EPOLLERR) != 0 || + (events[i].events & EPOLLHUP) != 0) + { + /* + * epoll does not set IN/OUT bits on an erroneous + * condition, so we need to try both anyway. This is a + * bit inefficient, but should be okay for such rare + * events. Note also that the read or write attempt + * won't block because we use non-blocking sockets. + */ + int fd = events[i].data.fd; + events[i].events |= thread->epoll_events[fd]; + } + process_fd(thread, events[i].data.fd, + (events[i].events & EPOLLIN) != 0, + (events[i].events & EPOLLOUT) != 0); + } + + if (have_ctlevent) { + done = process_ctlfd(thread); + } + + return (done); +} +#elif defined(USE_DEVPOLL) +static bool +process_fds(isc__socketthread_t *thread, struct pollfd *events, int nevents) { + int i; + bool done = false; + bool have_ctlevent = false; + + if (nevents == thread->nevents) { + thread_log(thread, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_INFO, + "maximum number of FD events (%d) received", + nevents); + } + + for (i = 0; i < nevents; i++) { + REQUIRE(events[i].fd < (int)thread->manager->maxsocks); + if (events[i].fd == thread->pipe_fds[0]) { + have_ctlevent = true; + continue; + } + process_fd(thread, events[i].fd, + (events[i].events & POLLIN) != 0, + (events[i].events & POLLOUT) != 0); + } + + if (have_ctlevent) { + done = process_ctlfd(thread); + } + + return (done); +} +#elif defined(USE_SELECT) +static void +process_fds(isc__socketthread_t *thread, int maxfd, fd_set *readfds, + fd_set *writefds) { + int i; + + REQUIRE(maxfd <= (int)thread->manager->maxsocks); + + for (i = 0; i < maxfd; i++) { + if (i == thread->pipe_fds[0] || i == thread->pipe_fds[1]) { + continue; + } + process_fd(thread, i, FD_ISSET(i, readfds), + FD_ISSET(i, writefds)); + } +} +#endif /* ifdef USE_KQUEUE */ + +static bool +process_ctlfd(isc__socketthread_t *thread) { + int msg, fd; + + for (;;) { + select_readmsg(thread, &fd, &msg); + + thread_log(thread, IOEVENT, + "watcher got message %d for socket %d", msg, fd); + + /* + * Nothing to read? + */ + if (msg == SELECT_POKE_NOTHING) { + break; + } + + /* + * Handle shutdown message. We really should + * jump out of this loop right away, but + * it doesn't matter if we have to do a little + * more work first. + */ + if (msg == SELECT_POKE_SHUTDOWN) { + return (true); + } + + /* + * This is a wakeup on a socket. Look + * at the event queue for both read and write, + * and decide if we need to watch on it now + * or not. + */ + wakeup_socket(thread, fd, msg); + } + + return (false); +} + +/* + * This is the thread that will loop forever, always in a select or poll + * call. + * + * When select returns something to do, do whatever's necessary and post + * an event to the task that was requesting the action. + */ +static isc_threadresult_t +netthread(void *uap) { + isc__socketthread_t *thread = uap; + isc_socketmgr_t *manager = thread->manager; + (void)manager; + bool done; + int cc; +#ifdef USE_KQUEUE + const char *fnname = "kevent()"; +#elif defined(USE_EPOLL) + const char *fnname = "epoll_wait()"; +#elif defined(USE_DEVPOLL) + isc_result_t result; + const char *fnname = "ioctl(DP_POLL)"; + struct dvpoll dvp; + int pass; +#if defined(ISC_SOCKET_USE_POLLWATCH) + pollstate_t pollstate = poll_idle; +#endif /* if defined(ISC_SOCKET_USE_POLLWATCH) */ +#elif defined(USE_SELECT) + const char *fnname = "select()"; + int maxfd; + int ctlfd; +#endif /* ifdef USE_KQUEUE */ + char strbuf[ISC_STRERRORSIZE]; + +#if defined(USE_SELECT) + /* + * Get the control fd here. This will never change. + */ + ctlfd = thread->pipe_fds[0]; +#endif /* if defined(USE_SELECT) */ + done = false; + while (!done) { + do { +#ifdef USE_KQUEUE + cc = kevent(thread->kqueue_fd, NULL, 0, thread->events, + thread->nevents, NULL); +#elif defined(USE_EPOLL) + cc = epoll_wait(thread->epoll_fd, thread->events, + thread->nevents, -1); +#elif defined(USE_DEVPOLL) + /* + * Re-probe every thousand calls. + */ + if (thread->calls++ > 1000U) { + result = isc_resource_getcurlimit( + isc_resource_openfiles, + &thread->open_max); + if (result != ISC_R_SUCCESS) { + thread->open_max = 64; + } + thread->calls = 0; + } + for (pass = 0; pass < 2; pass++) { + dvp.dp_fds = thread->events; + dvp.dp_nfds = thread->nevents; + if (dvp.dp_nfds >= thread->open_max) { + dvp.dp_nfds = thread->open_max - 1; + } +#ifndef ISC_SOCKET_USE_POLLWATCH + dvp.dp_timeout = -1; +#else /* ifndef ISC_SOCKET_USE_POLLWATCH */ + if (pollstate == poll_idle) { + dvp.dp_timeout = -1; + } else { + dvp.dp_timeout = + ISC_SOCKET_POLLWATCH_TIMEOUT; + } +#endif /* ISC_SOCKET_USE_POLLWATCH */ + cc = ioctl(thread->devpoll_fd, DP_POLL, &dvp); + if (cc == -1 && errno == EINVAL) { + /* + * {OPEN_MAX} may have dropped. Look + * up the current value and try again. + */ + result = isc_resource_getcurlimit( + isc_resource_openfiles, + &thread->open_max); + if (result != ISC_R_SUCCESS) { + thread->open_max = 64; + } + } else { + break; + } + } +#elif defined(USE_SELECT) + /* + * We will have only one thread anyway, we can lock + * manager lock and don't care + */ + LOCK(&manager->lock); + memmove(thread->read_fds_copy, thread->read_fds, + thread->fd_bufsize); + memmove(thread->write_fds_copy, thread->write_fds, + thread->fd_bufsize); + maxfd = thread->maxfd + 1; + UNLOCK(&manager->lock); + + cc = select(maxfd, thread->read_fds_copy, + thread->write_fds_copy, NULL, NULL); +#endif /* USE_KQUEUE */ + + if (cc < 0 && !SOFT_ERROR(errno)) { + strerror_r(errno, strbuf, sizeof(strbuf)); + FATAL_ERROR(__FILE__, __LINE__, "%s failed: %s", + fnname, strbuf); + } + +#if defined(USE_DEVPOLL) && defined(ISC_SOCKET_USE_POLLWATCH) + if (cc == 0) { + if (pollstate == poll_active) { + pollstate = poll_checking; + } else if (pollstate == poll_checking) { + pollstate = poll_idle; + } + } else if (cc > 0) { + if (pollstate == poll_checking) { + /* + * XXX: We'd like to use a more + * verbose log level as it's actually an + * unexpected event, but the kernel bug + * reportedly happens pretty frequently + * (and it can also be a false positive) + * so it would be just too noisy. + */ + thread_log(thread, + ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + ISC_LOG_DEBUG(1), + "unexpected POLL timeout"); + } + pollstate = poll_active; + } +#endif /* if defined(USE_DEVPOLL) && defined(ISC_SOCKET_USE_POLLWATCH) */ + } while (cc < 0); + +#if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) + done = process_fds(thread, thread->events, cc); +#elif defined(USE_SELECT) + process_fds(thread, maxfd, thread->read_fds_copy, + thread->write_fds_copy); + + /* + * Process reads on internal, control fd. + */ + if (FD_ISSET(ctlfd, thread->read_fds_copy)) { + done = process_ctlfd(thread); + } +#endif /* if defined(USE_KQUEUE) || defined(USE_EPOLL) || defined(USE_DEVPOLL) \ + * */ + } + + thread_log(thread, TRACE, "watcher exiting"); + return ((isc_threadresult_t)0); +} + +void +isc_socketmgr_setreserved(isc_socketmgr_t *manager, uint32_t reserved) { + REQUIRE(VALID_MANAGER(manager)); + + manager->reserved = reserved; +} + +void +isc_socketmgr_maxudp(isc_socketmgr_t *manager, unsigned int maxudp) { + REQUIRE(VALID_MANAGER(manager)); + + manager->maxudp = maxudp; +} + +/* + * Setup socket thread, thread->manager and thread->threadid must be filled. + */ + +static isc_result_t +setup_thread(isc__socketthread_t *thread) { + isc_result_t result = ISC_R_SUCCESS; + int i; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(thread != NULL); + REQUIRE(VALID_MANAGER(thread->manager)); + REQUIRE(thread->threadid >= 0 && + thread->threadid < thread->manager->nthreads); + + thread->fds = + isc_mem_get(thread->manager->mctx, + thread->manager->maxsocks * sizeof(isc_socket_t *)); + + memset(thread->fds, 0, + thread->manager->maxsocks * sizeof(isc_socket_t *)); + + thread->fdstate = isc_mem_get(thread->manager->mctx, + thread->manager->maxsocks * sizeof(int)); + + memset(thread->fdstate, 0, thread->manager->maxsocks * sizeof(int)); + + thread->fdlock = isc_mem_get(thread->manager->mctx, + FDLOCK_COUNT * sizeof(isc_mutex_t)); + + for (i = 0; i < FDLOCK_COUNT; i++) { + isc_mutex_init(&thread->fdlock[i]); + } + + if (pipe(thread->pipe_fds) != 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "pipe() failed: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } + RUNTIME_CHECK(make_nonblock(thread->pipe_fds[0]) == ISC_R_SUCCESS); + +#ifdef USE_KQUEUE + thread->nevents = ISC_SOCKET_MAXEVENTS; + thread->events = isc_mem_get(thread->manager->mctx, + sizeof(struct kevent) * thread->nevents); + + thread->kqueue_fd = kqueue(); + if (thread->kqueue_fd == -1) { + result = isc__errno2result(errno); + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "kqueue failed: %s", + strbuf); + isc_mem_put(thread->manager->mctx, thread->events, + sizeof(struct kevent) * thread->nevents); + return (result); + } + + result = watch_fd(thread, thread->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + close(thread->kqueue_fd); + isc_mem_put(thread->manager->mctx, thread->events, + sizeof(struct kevent) * thread->nevents); + } + return (result); + +#elif defined(USE_EPOLL) + thread->nevents = ISC_SOCKET_MAXEVENTS; + thread->epoll_events = + isc_mem_get(thread->manager->mctx, + (thread->manager->maxsocks * sizeof(uint32_t))); + + memset(thread->epoll_events, 0, + thread->manager->maxsocks * sizeof(uint32_t)); + + thread->events = + isc_mem_get(thread->manager->mctx, + sizeof(struct epoll_event) * thread->nevents); + + thread->epoll_fd = epoll_create(thread->nevents); + if (thread->epoll_fd == -1) { + result = isc__errno2result(errno); + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "epoll_create failed: %s", + strbuf); + return (result); + } + + result = watch_fd(thread, thread->pipe_fds[0], SELECT_POKE_READ); + return (result); + +#elif defined(USE_DEVPOLL) + thread->nevents = ISC_SOCKET_MAXEVENTS; + result = isc_resource_getcurlimit(isc_resource_openfiles, + &thread->open_max); + if (result != ISC_R_SUCCESS) { + thread->open_max = 64; + } + thread->calls = 0; + thread->events = isc_mem_get(thread->manager->mctx, + sizeof(struct pollfd) * thread->nevents); + + /* + * Note: fdpollinfo should be able to support all possible FDs, so + * it must have maxsocks entries (not nevents). + */ + thread->fdpollinfo = + isc_mem_get(thread->manager->mctx, + sizeof(pollinfo_t) * thread->manager->maxsocks); + memset(thread->fdpollinfo, 0, + sizeof(pollinfo_t) * thread->manager->maxsocks); + thread->devpoll_fd = open("/dev/poll", O_RDWR); + if (thread->devpoll_fd == -1) { + result = isc__errno2result(errno); + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "open(/dev/poll) failed: %s", strbuf); + isc_mem_put(thread->manager->mctx, thread->events, + sizeof(struct pollfd) * thread->nevents); + isc_mem_put(thread->manager->mctx, thread->fdpollinfo, + sizeof(pollinfo_t) * thread->manager->maxsocks); + return (result); + } + result = watch_fd(thread, thread->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + close(thread->devpoll_fd); + isc_mem_put(thread->manager->mctx, thread->events, + sizeof(struct pollfd) * thread->nevents); + isc_mem_put(thread->manager->mctx, thread->fdpollinfo, + sizeof(pollinfo_t) * thread->manager->maxsocks); + return (result); + } + + return (ISC_R_SUCCESS); +#elif defined(USE_SELECT) + UNUSED(result); + +#if ISC_SOCKET_MAXSOCKETS > FD_SETSIZE + /* + * Note: this code should also cover the case of MAXSOCKETS <= + * FD_SETSIZE, but we separate the cases to avoid possible portability + * issues regarding howmany() and the actual representation of fd_set. + */ + thread->fd_bufsize = howmany(manager->maxsocks, NFDBITS) * + sizeof(fd_mask); +#else /* if ISC_SOCKET_MAXSOCKETS > FD_SETSIZE */ + thread->fd_bufsize = sizeof(fd_set); +#endif /* if ISC_SOCKET_MAXSOCKETS > FD_SETSIZE */ + + thread->read_fds = isc_mem_get(thread->manager->mctx, + thread->fd_bufsize); + thread->read_fds_copy = isc_mem_get(thread->manager->mctx, + thread->fd_bufsize); + thread->write_fds = isc_mem_get(thread->manager->mctx, + thread->fd_bufsize); + thread->write_fds_copy = isc_mem_get(thread->manager->mctx, + thread->fd_bufsize); + memset(thread->read_fds, 0, thread->fd_bufsize); + memset(thread->write_fds, 0, thread->fd_bufsize); + + (void)watch_fd(thread, thread->pipe_fds[0], SELECT_POKE_READ); + thread->maxfd = thread->pipe_fds[0]; + + return (ISC_R_SUCCESS); +#endif /* USE_KQUEUE */ +} + +static void +cleanup_thread(isc_mem_t *mctx, isc__socketthread_t *thread) { + isc_result_t result; + int i; + + result = unwatch_fd(thread, thread->pipe_fds[0], SELECT_POKE_READ); + if (result != ISC_R_SUCCESS) { + UNEXPECTED_ERROR(__FILE__, __LINE__, "epoll_ctl(DEL) failed"); + } +#ifdef USE_KQUEUE + close(thread->kqueue_fd); + isc_mem_put(mctx, thread->events, + sizeof(struct kevent) * thread->nevents); +#elif defined(USE_EPOLL) + close(thread->epoll_fd); + + isc_mem_put(mctx, thread->events, + sizeof(struct epoll_event) * thread->nevents); +#elif defined(USE_DEVPOLL) + close(thread->devpoll_fd); + isc_mem_put(mctx, thread->events, + sizeof(struct pollfd) * thread->nevents); + isc_mem_put(mctx, thread->fdpollinfo, + sizeof(pollinfo_t) * thread->manager->maxsocks); +#elif defined(USE_SELECT) + if (thread->read_fds != NULL) { + isc_mem_put(mctx, thread->read_fds, thread->fd_bufsize); + } + if (thread->read_fds_copy != NULL) { + isc_mem_put(mctx, thread->read_fds_copy, thread->fd_bufsize); + } + if (thread->write_fds != NULL) { + isc_mem_put(mctx, thread->write_fds, thread->fd_bufsize); + } + if (thread->write_fds_copy != NULL) { + isc_mem_put(mctx, thread->write_fds_copy, thread->fd_bufsize); + } +#endif /* USE_KQUEUE */ + for (i = 0; i < (int)thread->manager->maxsocks; i++) { + if (thread->fdstate[i] == CLOSE_PENDING) { + /* no need to lock */ + (void)close(i); + } + } + +#if defined(USE_EPOLL) + isc_mem_put(thread->manager->mctx, thread->epoll_events, + thread->manager->maxsocks * sizeof(uint32_t)); +#endif /* if defined(USE_EPOLL) */ + isc_mem_put(thread->manager->mctx, thread->fds, + thread->manager->maxsocks * sizeof(isc_socket_t *)); + isc_mem_put(thread->manager->mctx, thread->fdstate, + thread->manager->maxsocks * sizeof(int)); + + for (i = 0; i < FDLOCK_COUNT; i++) { + isc_mutex_destroy(&thread->fdlock[i]); + } + isc_mem_put(thread->manager->mctx, thread->fdlock, + FDLOCK_COUNT * sizeof(isc_mutex_t)); +} + +isc_result_t +isc_socketmgr_create(isc_mem_t *mctx, isc_socketmgr_t **managerp) { + return (isc_socketmgr_create2(mctx, managerp, 0, 1)); +} + +isc_result_t +isc_socketmgr_create2(isc_mem_t *mctx, isc_socketmgr_t **managerp, + unsigned int maxsocks, int nthreads) { + int i; + isc_socketmgr_t *manager; + + REQUIRE(managerp != NULL && *managerp == NULL); + + if (maxsocks == 0) { + maxsocks = ISC_SOCKET_MAXSOCKETS; + } + + manager = isc_mem_get(mctx, sizeof(*manager)); + + /* zero-clear so that necessary cleanup on failure will be easy */ + memset(manager, 0, sizeof(*manager)); + manager->maxsocks = maxsocks; + manager->reserved = 0; + manager->maxudp = 0; + manager->nthreads = nthreads; + manager->stats = NULL; + + manager->magic = SOCKET_MANAGER_MAGIC; + manager->mctx = NULL; + ISC_LIST_INIT(manager->socklist); + isc_mutex_init(&manager->lock); + isc_condition_init(&manager->shutdown_ok); + + /* + * Start up the select/poll thread. + */ + manager->threads = isc_mem_get(mctx, sizeof(isc__socketthread_t) * + manager->nthreads); + isc_mem_attach(mctx, &manager->mctx); + + for (i = 0; i < manager->nthreads; i++) { + manager->threads[i].manager = manager; + manager->threads[i].threadid = i; + setup_thread(&manager->threads[i]); + isc_thread_create(netthread, &manager->threads[i], + &manager->threads[i].thread); + char tname[1024]; + sprintf(tname, "isc-socket-%d", i); + isc_thread_setname(manager->threads[i].thread, tname); + } + + *managerp = manager; + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_socketmgr_getmaxsockets(isc_socketmgr_t *manager, unsigned int *nsockp) { + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(nsockp != NULL); + + *nsockp = manager->maxsocks; + + return (ISC_R_SUCCESS); +} + +void +isc_socketmgr_setstats(isc_socketmgr_t *manager, isc_stats_t *stats) { + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(ISC_LIST_EMPTY(manager->socklist)); + REQUIRE(manager->stats == NULL); + REQUIRE(isc_stats_ncounters(stats) == isc_sockstatscounter_max); + + isc_stats_attach(stats, &manager->stats); +} + +void +isc_socketmgr_destroy(isc_socketmgr_t **managerp) { + isc_socketmgr_t *manager; + + /* + * Destroy a socket manager. + */ + + REQUIRE(managerp != NULL); + manager = *managerp; + REQUIRE(VALID_MANAGER(manager)); + + LOCK(&manager->lock); + + /* + * Wait for all sockets to be destroyed. + */ + while (!ISC_LIST_EMPTY(manager->socklist)) { + manager_log(manager, CREATION, "sockets exist"); + WAIT(&manager->shutdown_ok, &manager->lock); + } + + UNLOCK(&manager->lock); + + /* + * Here, poke our select/poll thread. Do this by closing the write + * half of the pipe, which will send EOF to the read half. + * This is currently a no-op in the non-threaded case. + */ + for (int i = 0; i < manager->nthreads; i++) { + select_poke(manager, i, 0, SELECT_POKE_SHUTDOWN); + } + + /* + * Wait for thread to exit. + */ + for (int i = 0; i < manager->nthreads; i++) { + isc_thread_join(manager->threads[i].thread, NULL); + cleanup_thread(manager->mctx, &manager->threads[i]); + } + /* + * Clean up. + */ + isc_mem_put(manager->mctx, manager->threads, + sizeof(isc__socketthread_t) * manager->nthreads); + (void)isc_condition_destroy(&manager->shutdown_ok); + + if (manager->stats != NULL) { + isc_stats_detach(&manager->stats); + } + isc_mutex_destroy(&manager->lock); + manager->magic = 0; + isc_mem_putanddetach(&manager->mctx, manager, sizeof(*manager)); + + *managerp = NULL; +} + +static isc_result_t +socket_recv(isc_socket_t *sock, isc_socketevent_t *dev, isc_task_t *task, + unsigned int flags) { + int io_state; + bool have_lock = false; + isc_task_t *ntask = NULL; + isc_result_t result = ISC_R_SUCCESS; + + dev->ev_sender = task; + + if (sock->type == isc_sockettype_udp) { + io_state = doio_recv(sock, dev); + } else { + LOCK(&sock->lock); + have_lock = true; + + if (ISC_LIST_EMPTY(sock->recv_list)) { + io_state = doio_recv(sock, dev); + } else { + io_state = DOIO_SOFT; + } + } + + switch (io_state) { + case DOIO_SOFT: + /* + * We couldn't read all or part of the request right now, so + * queue it. + * + * Attach to socket and to task + */ + isc_task_attach(task, &ntask); + dev->attributes |= ISC_SOCKEVENTATTR_ATTACHED; + + if (!have_lock) { + LOCK(&sock->lock); + have_lock = true; + } + + /* + * Enqueue the request. If the socket was previously not being + * watched, poke the watcher to start paying attention to it. + */ + bool do_poke = ISC_LIST_EMPTY(sock->recv_list); + ISC_LIST_ENQUEUE(sock->recv_list, dev, ev_link); + if (do_poke) { + select_poke(sock->manager, sock->threadid, sock->fd, + SELECT_POKE_READ); + } + + socket_log(sock, NULL, EVENT, + "socket_recv: event %p -> task %p", dev, ntask); + + if ((flags & ISC_SOCKFLAG_IMMEDIATE) != 0) { + result = ISC_R_INPROGRESS; + } + break; + + case DOIO_EOF: + dev->result = ISC_R_EOF; + FALLTHROUGH; + + case DOIO_HARD: + case DOIO_SUCCESS: + if ((flags & ISC_SOCKFLAG_IMMEDIATE) == 0) { + send_recvdone_event(sock, &dev); + } + break; + } + + if (have_lock) { + UNLOCK(&sock->lock); + } + + return (result); +} + +isc_result_t +isc_socket_recv(isc_socket_t *sock, isc_region_t *region, unsigned int minimum, + isc_task_t *task, isc_taskaction_t action, void *arg) { + isc_socketevent_t *dev; + isc_socketmgr_t *manager; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + INSIST(sock->bound); + + dev = allocate_socketevent(manager->mctx, sock, ISC_SOCKEVENT_RECVDONE, + action, arg); + if (dev == NULL) { + return (ISC_R_NOMEMORY); + } + + return (isc_socket_recv2(sock, region, minimum, task, dev, 0)); +} + +isc_result_t +isc_socket_recv2(isc_socket_t *sock, isc_region_t *region, unsigned int minimum, + isc_task_t *task, isc_socketevent_t *event, + unsigned int flags) { + event->ev_sender = sock; + event->result = ISC_R_UNSET; + event->region = *region; + event->n = 0; + event->offset = 0; + event->attributes = 0; + + /* + * UDP sockets are always partial read. + */ + if (sock->type == isc_sockettype_udp) { + event->minimum = 1; + } else { + if (minimum == 0) { + event->minimum = region->length; + } else { + event->minimum = minimum; + } + } + + return (socket_recv(sock, event, task, flags)); +} + +static isc_result_t +socket_send(isc_socket_t *sock, isc_socketevent_t *dev, isc_task_t *task, + const isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + unsigned int flags) { + int io_state; + bool have_lock = false; + isc_task_t *ntask = NULL; + isc_result_t result = ISC_R_SUCCESS; + + dev->ev_sender = task; + + set_dev_address(address, sock, dev); + if (pktinfo != NULL) { + dev->attributes |= ISC_SOCKEVENTATTR_PKTINFO; + dev->pktinfo = *pktinfo; + + if (!isc_sockaddr_issitelocal(&dev->address) && + !isc_sockaddr_islinklocal(&dev->address)) + { + socket_log(sock, NULL, TRACE, + "pktinfo structure provided, ifindex %u " + "(set to 0)", + pktinfo->ipi6_ifindex); + + /* + * Set the pktinfo index to 0 here, to let the + * kernel decide what interface it should send on. + */ + dev->pktinfo.ipi6_ifindex = 0; + } + } + + if (sock->type == isc_sockettype_udp) { + io_state = doio_send(sock, dev); + } else { + LOCK(&sock->lock); + have_lock = true; + + if (ISC_LIST_EMPTY(sock->send_list)) { + io_state = doio_send(sock, dev); + } else { + io_state = DOIO_SOFT; + } + } + + switch (io_state) { + case DOIO_SOFT: + /* + * We couldn't send all or part of the request right now, so + * queue it unless ISC_SOCKFLAG_NORETRY is set. + */ + if ((flags & ISC_SOCKFLAG_NORETRY) == 0) { + isc_task_attach(task, &ntask); + dev->attributes |= ISC_SOCKEVENTATTR_ATTACHED; + + if (!have_lock) { + LOCK(&sock->lock); + have_lock = true; + } + + /* + * Enqueue the request. If the socket was previously + * not being watched, poke the watcher to start + * paying attention to it. + */ + bool do_poke = ISC_LIST_EMPTY(sock->send_list); + ISC_LIST_ENQUEUE(sock->send_list, dev, ev_link); + if (do_poke) { + select_poke(sock->manager, sock->threadid, + sock->fd, SELECT_POKE_WRITE); + } + socket_log(sock, NULL, EVENT, + "socket_send: event %p -> task %p", dev, + ntask); + + if ((flags & ISC_SOCKFLAG_IMMEDIATE) != 0) { + result = ISC_R_INPROGRESS; + } + break; + } + + FALLTHROUGH; + + case DOIO_HARD: + case DOIO_SUCCESS: + if (!have_lock) { + LOCK(&sock->lock); + have_lock = true; + } + if ((flags & ISC_SOCKFLAG_IMMEDIATE) == 0) { + send_senddone_event(sock, &dev); + } + break; + } + + if (have_lock) { + UNLOCK(&sock->lock); + } + + return (result); +} + +isc_result_t +isc_socket_send(isc_socket_t *sock, isc_region_t *region, isc_task_t *task, + isc_taskaction_t action, void *arg) { + /* + * REQUIRE() checking is performed in isc_socket_sendto(). + */ + return (isc_socket_sendto(sock, region, task, action, arg, NULL, NULL)); +} + +isc_result_t +isc_socket_sendto(isc_socket_t *sock, isc_region_t *region, isc_task_t *task, + isc_taskaction_t action, void *arg, + const isc_sockaddr_t *address, struct in6_pktinfo *pktinfo) { + isc_socketevent_t *dev; + isc_socketmgr_t *manager; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(region != NULL); + REQUIRE(task != NULL); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + INSIST(sock->bound); + + dev = allocate_socketevent(manager->mctx, sock, ISC_SOCKEVENT_SENDDONE, + action, arg); + if (dev == NULL) { + return (ISC_R_NOMEMORY); + } + + dev->region = *region; + + return (socket_send(sock, dev, task, address, pktinfo, 0)); +} + +isc_result_t +isc_socket_sendto2(isc_socket_t *sock, isc_region_t *region, isc_task_t *task, + const isc_sockaddr_t *address, struct in6_pktinfo *pktinfo, + isc_socketevent_t *event, unsigned int flags) { + REQUIRE(VALID_SOCKET(sock)); + REQUIRE((flags & ~(ISC_SOCKFLAG_IMMEDIATE | ISC_SOCKFLAG_NORETRY)) == + 0); + if ((flags & ISC_SOCKFLAG_NORETRY) != 0) { + REQUIRE(sock->type == isc_sockettype_udp); + } + event->ev_sender = sock; + event->result = ISC_R_UNSET; + event->region = *region; + event->n = 0; + event->offset = 0; + event->attributes &= ~ISC_SOCKEVENTATTR_ATTACHED; + + return (socket_send(sock, event, task, address, pktinfo, flags)); +} + +void +isc_socket_cleanunix(const isc_sockaddr_t *sockaddr, bool active) { +#ifdef ISC_PLATFORM_HAVESYSUNH + int s; + struct stat sb; + char strbuf[ISC_STRERRORSIZE]; + + if (sockaddr->type.sa.sa_family != AF_UNIX) { + return; + } + +#ifndef S_ISSOCK +#if defined(S_IFMT) && defined(S_IFSOCK) +#define S_ISSOCK(mode) ((mode & S_IFMT) == S_IFSOCK) +#elif defined(_S_IFMT) && defined(S_IFSOCK) +#define S_ISSOCK(mode) ((mode & _S_IFMT) == S_IFSOCK) +#endif /* if defined(S_IFMT) && defined(S_IFSOCK) */ +#endif /* ifndef S_ISSOCK */ + +#ifndef S_ISFIFO +#if defined(S_IFMT) && defined(S_IFIFO) +#define S_ISFIFO(mode) ((mode & S_IFMT) == S_IFIFO) +#elif defined(_S_IFMT) && defined(S_IFIFO) +#define S_ISFIFO(mode) ((mode & _S_IFMT) == S_IFIFO) +#endif /* if defined(S_IFMT) && defined(S_IFIFO) */ +#endif /* ifndef S_ISFIFO */ + +#if !defined(S_ISFIFO) && !defined(S_ISSOCK) +/* cppcheck-suppress preprocessorErrorDirective */ +#error \ + You need to define S_ISFIFO and S_ISSOCK as appropriate for your platform. See <sys/stat.h>. +#endif /* if !defined(S_ISFIFO) && !defined(S_ISSOCK) */ + +#ifndef S_ISFIFO +#define S_ISFIFO(mode) 0 +#endif /* ifndef S_ISFIFO */ + +#ifndef S_ISSOCK +#define S_ISSOCK(mode) 0 +#endif /* ifndef S_ISSOCK */ + + if (stat(sockaddr->type.sunix.sun_path, &sb) < 0) { + switch (errno) { + case ENOENT: + if (active) { /* We exited cleanly last time */ + break; + } + FALLTHROUGH; + default: + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + active ? ISC_LOG_ERROR : ISC_LOG_WARNING, + "isc_socket_cleanunix: stat(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + return; + } + } else { + if (!(S_ISSOCK(sb.st_mode) || S_ISFIFO(sb.st_mode))) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, + active ? ISC_LOG_ERROR : ISC_LOG_WARNING, + "isc_socket_cleanunix: %s: not a socket", + sockaddr->type.sunix.sun_path); + return; + } + } + + if (active) { + if (unlink(sockaddr->type.sunix.sun_path) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_cleanunix: unlink(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + } + return; + } + + s = socket(AF_UNIX, SOCK_STREAM, 0); + if (s < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: socket(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + return; + } + + if (connect(s, (const struct sockaddr *)&sockaddr->type.sunix, + sizeof(sockaddr->type.sunix)) < 0) + { + switch (errno) { + case ECONNREFUSED: + case ECONNRESET: + if (unlink(sockaddr->type.sunix.sun_path) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write( + isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: " + "unlink(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + } + break; + default: + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_WARNING, + "isc_socket_cleanunix: connect(%s): %s", + sockaddr->type.sunix.sun_path, strbuf); + break; + } + } + close(s); +#else /* ifdef ISC_PLATFORM_HAVESYSUNH */ + UNUSED(sockaddr); + UNUSED(active); +#endif /* ifdef ISC_PLATFORM_HAVESYSUNH */ +} + +isc_result_t +isc_socket_permunix(const isc_sockaddr_t *sockaddr, uint32_t perm, + uint32_t owner, uint32_t group) { +#ifdef ISC_PLATFORM_HAVESYSUNH + isc_result_t result = ISC_R_SUCCESS; + char strbuf[ISC_STRERRORSIZE]; + char path[sizeof(sockaddr->type.sunix.sun_path)]; +#ifdef NEED_SECURE_DIRECTORY + char *slash; +#endif /* ifdef NEED_SECURE_DIRECTORY */ + + REQUIRE(sockaddr->type.sa.sa_family == AF_UNIX); + INSIST(strlen(sockaddr->type.sunix.sun_path) < sizeof(path)); + strlcpy(path, sockaddr->type.sunix.sun_path, sizeof(path)); + +#ifdef NEED_SECURE_DIRECTORY + slash = strrchr(path, '/'); + if (slash != NULL) { + if (slash != path) { + *slash = '\0'; + } else { + strlcpy(path, "/", sizeof(path)); + } + } else { + strlcpy(path, ".", sizeof(path)); + } +#endif /* ifdef NEED_SECURE_DIRECTORY */ + + if (chmod(path, perm) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_permunix: chmod(%s, %d): %s", path, + perm, strbuf); + result = ISC_R_FAILURE; + } + if (chown(path, owner, group) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR, + "isc_socket_permunix: chown(%s, %d, %d): %s", + path, owner, group, strbuf); + result = ISC_R_FAILURE; + } + return (result); +#else /* ifdef ISC_PLATFORM_HAVESYSUNH */ + UNUSED(sockaddr); + UNUSED(perm); + UNUSED(owner); + UNUSED(group); + return (ISC_R_NOTIMPLEMENTED); +#endif /* ifdef ISC_PLATFORM_HAVESYSUNH */ +} + +isc_result_t +isc_socket_bind(isc_socket_t *sock, const isc_sockaddr_t *sockaddr, + isc_socket_options_t options) { + char strbuf[ISC_STRERRORSIZE]; + int on = 1; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + INSIST(!sock->bound); + INSIST(!sock->dupped); + + if (sock->pf != sockaddr->type.sa.sa_family) { + UNLOCK(&sock->lock); + return (ISC_R_FAMILYMISMATCH); + } + + /* + * Only set SO_REUSEADDR when we want a specific port. + */ +#ifdef AF_UNIX + if (sock->pf == AF_UNIX) { + goto bind_socket; + } +#endif /* ifdef AF_UNIX */ + if ((options & ISC_SOCKET_REUSEADDRESS) != 0 && + isc_sockaddr_getport(sockaddr) != (in_port_t)0) + { + if (setsockopt(sock->fd, SOL_SOCKET, SO_REUSEADDR, (void *)&on, + sizeof(on)) < 0) + { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d) failed", sock->fd); + } +#if defined(__FreeBSD_kernel__) && defined(SO_REUSEPORT_LB) + if (setsockopt(sock->fd, SOL_SOCKET, SO_REUSEPORT_LB, + (void *)&on, sizeof(on)) < 0) + { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d) failed", sock->fd); + } +#elif defined(__linux__) && defined(SO_REUSEPORT) + if (setsockopt(sock->fd, SOL_SOCKET, SO_REUSEPORT, (void *)&on, + sizeof(on)) < 0) + { + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d) failed", sock->fd); + } +#endif /* if defined(__FreeBSD_kernel__) && defined(SO_REUSEPORT_LB) */ + /* Press on... */ + } +#ifdef AF_UNIX +bind_socket: +#endif /* ifdef AF_UNIX */ + if (bind(sock->fd, &sockaddr->type.sa, sockaddr->length) < 0) { + inc_stats(sock->manager->stats, + sock->statsindex[STATID_BINDFAIL]); + + UNLOCK(&sock->lock); + switch (errno) { + case EACCES: + return (ISC_R_NOPERM); + case EADDRNOTAVAIL: + return (ISC_R_ADDRNOTAVAIL); + case EADDRINUSE: + return (ISC_R_ADDRINUSE); + case EINVAL: + return (ISC_R_BOUND); + default: + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "bind: %s", + strbuf); + return (ISC_R_UNEXPECTED); + } + } + + socket_log(sock, sockaddr, TRACE, "bound"); + sock->bound = 1; + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +/* + * Enable this only for specific OS versions, and only when they have repaired + * their problems with it. Until then, this is is broken and needs to be + * disabled by default. See RT22589 for details. + */ +#undef ENABLE_ACCEPTFILTER + +isc_result_t +isc_socket_filter(isc_socket_t *sock, const char *filter) { +#if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) + char strbuf[ISC_STRERRORSIZE]; + struct accept_filter_arg afa; +#else /* if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) */ + UNUSED(sock); + UNUSED(filter); +#endif /* if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) */ + + REQUIRE(VALID_SOCKET(sock)); + +#if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) + bzero(&afa, sizeof(afa)); + strlcpy(afa.af_name, filter, sizeof(afa.af_name)); + if (setsockopt(sock->fd, SOL_SOCKET, SO_ACCEPTFILTER, &afa, + sizeof(afa)) == -1) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + socket_log(sock, NULL, CREATION, + "setsockopt(SO_ACCEPTFILTER): %s", strbuf); + return (ISC_R_FAILURE); + } + return (ISC_R_SUCCESS); +#else /* if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) */ + return (ISC_R_NOTIMPLEMENTED); +#endif /* if defined(SO_ACCEPTFILTER) && defined(ENABLE_ACCEPTFILTER) */ +} + +/* + * Try enabling TCP Fast Open for a given socket if the OS supports it. + */ +static void +set_tcp_fastopen(isc_socket_t *sock, unsigned int backlog) { +#if defined(ENABLE_TCP_FASTOPEN) && defined(TCP_FASTOPEN) + char strbuf[ISC_STRERRORSIZE]; + +/* + * FreeBSD, as of versions 10.3 and 11.0, defines TCP_FASTOPEN while also + * shipping a default kernel without TFO support, so we special-case it by + * performing an additional runtime check for TFO support using sysctl to + * prevent setsockopt() errors from being logged. + */ +#if defined(__FreeBSD__) && defined(HAVE_SYSCTLBYNAME) +#define SYSCTL_TFO "net.inet.tcp.fastopen.enabled" + unsigned int enabled; + size_t enabledlen = sizeof(enabled); + static bool tfo_notice_logged = false; + + if (sysctlbyname(SYSCTL_TFO, &enabled, &enabledlen, NULL, 0) < 0) { + /* + * This kernel does not support TCP Fast Open. There is + * nothing more we can do. + */ + return; + } else if (enabled == 0) { + /* + * This kernel does support TCP Fast Open, but it is disabled + * by sysctl. Notify the user, but do not nag. + */ + if (!tfo_notice_logged) { + isc_log_write(isc_lctx, ISC_LOGCATEGORY_GENERAL, + ISC_LOGMODULE_SOCKET, ISC_LOG_NOTICE, + "TCP_FASTOPEN support is disabled by " + "sysctl (" SYSCTL_TFO " = 0)"); + tfo_notice_logged = true; + } + return; + } +#endif /* if defined(__FreeBSD__) && defined(HAVE_SYSCTLBYNAME) */ + +#ifdef __APPLE__ + backlog = 1; +#else /* ifdef __APPLE__ */ + backlog = backlog / 2; + if (backlog == 0) { + backlog = 1; + } +#endif /* ifdef __APPLE__ */ + if (setsockopt(sock->fd, IPPROTO_TCP, TCP_FASTOPEN, (void *)&backlog, + sizeof(backlog)) < 0) + { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, TCP_FASTOPEN) failed with %s", + sock->fd, strbuf); + /* TCP_FASTOPEN is experimental so ignore failures */ + } +#else /* if defined(ENABLE_TCP_FASTOPEN) && defined(TCP_FASTOPEN) */ + UNUSED(sock); + UNUSED(backlog); +#endif /* if defined(ENABLE_TCP_FASTOPEN) && defined(TCP_FASTOPEN) */ +} + +/* + * Set up to listen on a given socket. We do this by creating an internal + * event that will be dispatched when the socket has read activity. The + * watcher will send the internal event to the task when there is a new + * connection. + * + * Unlike in read, we don't preallocate a done event here. Every time there + * is a new connection we'll have to allocate a new one anyway, so we might + * as well keep things simple rather than having to track them. + */ +isc_result_t +isc_socket_listen(isc_socket_t *sock, unsigned int backlog) { + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + + REQUIRE(!sock->listener); + REQUIRE(sock->bound); + REQUIRE(sock->type == isc_sockettype_tcp || + sock->type == isc_sockettype_unix); + + if (backlog == 0) { + backlog = SOMAXCONN; + } + + if (listen(sock->fd, (int)backlog) < 0) { + UNLOCK(&sock->lock); + strerror_r(errno, strbuf, sizeof(strbuf)); + + UNEXPECTED_ERROR(__FILE__, __LINE__, "listen: %s", strbuf); + + return (ISC_R_UNEXPECTED); + } + + set_tcp_fastopen(sock, backlog); + + sock->listener = 1; + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +/* + * This should try to do aggressive accept() XXXMLG + */ +isc_result_t +isc_socket_accept(isc_socket_t *sock, isc_task_t *task, isc_taskaction_t action, + void *arg) { + isc_socket_newconnev_t *dev; + isc_socketmgr_t *manager; + isc_task_t *ntask = NULL; + isc_socket_t *nsock; + isc_result_t result; + bool do_poke = false; + + REQUIRE(VALID_SOCKET(sock)); + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + + LOCK(&sock->lock); + + REQUIRE(sock->listener); + + /* + * Sender field is overloaded here with the task we will be sending + * this event to. Just before the actual event is delivered the + * actual ev_sender will be touched up to be the socket. + */ + dev = (isc_socket_newconnev_t *)isc_event_allocate( + manager->mctx, task, ISC_SOCKEVENT_NEWCONN, action, arg, + sizeof(*dev)); + ISC_LINK_INIT(dev, ev_link); + + result = allocate_socket(manager, sock->type, &nsock); + if (result != ISC_R_SUCCESS) { + isc_event_free(ISC_EVENT_PTR(&dev)); + UNLOCK(&sock->lock); + return (result); + } + + /* + * Attach to socket and to task. + */ + isc_task_attach(task, &ntask); + if (isc_task_exiting(ntask)) { + free_socket(&nsock); + isc_task_detach(&ntask); + isc_event_free(ISC_EVENT_PTR(&dev)); + UNLOCK(&sock->lock); + return (ISC_R_SHUTTINGDOWN); + } + isc_refcount_increment0(&nsock->references); + nsock->statsindex = sock->statsindex; + + dev->ev_sender = ntask; + dev->newsocket = nsock; + + /* + * Poke watcher here. We still have the socket locked, so there + * is no race condition. We will keep the lock for such a short + * bit of time waking it up now or later won't matter all that much. + */ + do_poke = ISC_LIST_EMPTY(sock->accept_list); + ISC_LIST_ENQUEUE(sock->accept_list, dev, ev_link); + if (do_poke) { + select_poke(manager, sock->threadid, sock->fd, + SELECT_POKE_ACCEPT); + } + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_socket_connect(isc_socket_t *sock, const isc_sockaddr_t *addr, + isc_task_t *task, isc_taskaction_t action, void *arg) { + isc_socket_connev_t *dev; + isc_task_t *ntask = NULL; + isc_socketmgr_t *manager; + int cc; + char strbuf[ISC_STRERRORSIZE]; + char addrbuf[ISC_SOCKADDR_FORMATSIZE]; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(addr != NULL); + REQUIRE(task != NULL); + REQUIRE(action != NULL); + + manager = sock->manager; + REQUIRE(VALID_MANAGER(manager)); + REQUIRE(addr != NULL); + + if (isc_sockaddr_ismulticast(addr)) { + return (ISC_R_MULTICAST); + } + + LOCK(&sock->lock); + + dev = (isc_socket_connev_t *)isc_event_allocate( + manager->mctx, sock, ISC_SOCKEVENT_CONNECT, action, arg, + sizeof(*dev)); + ISC_LINK_INIT(dev, ev_link); + + if (sock->connecting) { + INSIST(isc_sockaddr_equal(&sock->peer_address, addr)); + goto queue; + } + + if (sock->connected) { + INSIST(isc_sockaddr_equal(&sock->peer_address, addr)); + dev->result = ISC_R_SUCCESS; + isc_task_sendto(task, ISC_EVENT_PTR(&dev), sock->threadid); + + UNLOCK(&sock->lock); + + return (ISC_R_SUCCESS); + } + + /* + * Try to do the connect right away, as there can be only one + * outstanding, and it might happen to complete. + */ + sock->peer_address = *addr; + cc = connect(sock->fd, &addr->type.sa, addr->length); + if (cc < 0) { + /* + * The socket is nonblocking and the connection cannot be + * completed immediately. It is possible to select(2) or + * poll(2) for completion by selecting the socket for writing. + * After select(2) indicates writability, use getsockopt(2) to + * read the SO_ERROR option at level SOL_SOCKET to determine + * whether connect() completed successfully (SO_ERROR is zero) + * or unsuccessfully (SO_ERROR is one of the usual error codes + * listed here, explaining the reason for the failure). + */ + if (sock->type == isc_sockettype_udp && errno == EINPROGRESS) { + cc = 0; + goto success; + } + if (SOFT_ERROR(errno) || errno == EINPROGRESS) { + goto queue; + } + + switch (errno) { +#define ERROR_MATCH(a, b) \ + case a: \ + dev->result = b; \ + goto err_exit; + ERROR_MATCH(EACCES, ISC_R_NOPERM); + ERROR_MATCH(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(ECONNREFUSED, ISC_R_CONNREFUSED); + ERROR_MATCH(EHOSTUNREACH, ISC_R_HOSTUNREACH); +#ifdef EHOSTDOWN + ERROR_MATCH(EHOSTDOWN, ISC_R_HOSTUNREACH); +#endif /* ifdef EHOSTDOWN */ + ERROR_MATCH(ENETUNREACH, ISC_R_NETUNREACH); + ERROR_MATCH(ENOBUFS, ISC_R_NORESOURCES); + ERROR_MATCH(EPERM, ISC_R_HOSTUNREACH); + ERROR_MATCH(EPIPE, ISC_R_NOTCONNECTED); + ERROR_MATCH(ETIMEDOUT, ISC_R_TIMEDOUT); + ERROR_MATCH(ECONNRESET, ISC_R_CONNECTIONRESET); +#undef ERROR_MATCH + } + + sock->connected = 0; + + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_sockaddr_format(addr, addrbuf, sizeof(addrbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "connect(%s) %d/%s", + addrbuf, errno, strbuf); + + UNLOCK(&sock->lock); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECTFAIL]); + isc_event_free(ISC_EVENT_PTR(&dev)); + return (ISC_R_UNEXPECTED); + + err_exit: + sock->connected = 0; + isc_task_sendto(task, ISC_EVENT_PTR(&dev), sock->threadid); + + UNLOCK(&sock->lock); + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECTFAIL]); + return (ISC_R_SUCCESS); + } + + /* + * If connect completed, fire off the done event. + */ +success: + if (cc == 0) { + sock->connected = 1; + sock->bound = 1; + dev->result = ISC_R_SUCCESS; + isc_task_sendto(task, ISC_EVENT_PTR(&dev), sock->threadid); + + UNLOCK(&sock->lock); + + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECT]); + + return (ISC_R_SUCCESS); + } + +queue: + + /* + * Attach to task. + */ + isc_task_attach(task, &ntask); + + dev->ev_sender = ntask; + + /* + * Poke watcher here. We still have the socket locked, so there + * is no race condition. We will keep the lock for such a short + * bit of time waking it up now or later won't matter all that much. + */ + bool do_poke = ISC_LIST_EMPTY(sock->connect_list); + ISC_LIST_ENQUEUE(sock->connect_list, dev, ev_link); + if (do_poke && !sock->connecting) { + sock->connecting = 1; + select_poke(manager, sock->threadid, sock->fd, + SELECT_POKE_CONNECT); + } + + UNLOCK(&sock->lock); + return (ISC_R_SUCCESS); +} + +/* + * Called when a socket with a pending connect() finishes. + */ +static void +internal_connect(isc_socket_t *sock) { + isc_socket_connev_t *dev; + int cc; + isc_result_t result; + socklen_t optlen; + char strbuf[ISC_STRERRORSIZE]; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + + INSIST(VALID_SOCKET(sock)); + REQUIRE(sock->fd >= 0); + + /* + * Get the first item off the connect list. + * If it is empty, unlock the socket and return. + */ + dev = ISC_LIST_HEAD(sock->connect_list); + if (dev == NULL) { + INSIST(!sock->connecting); + goto finish; + } + + INSIST(sock->connecting); + sock->connecting = 0; + + /* + * Get any possible error status here. + */ + optlen = sizeof(cc); + if (getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void *)&cc, + (void *)&optlen) != 0) + { + cc = errno; + } else { + errno = cc; + } + + if (errno != 0) { + /* + * If the error is EAGAIN, just re-select on this + * fd and pretend nothing strange happened. + */ + if (SOFT_ERROR(errno) || errno == EINPROGRESS) { + sock->connecting = 1; + return; + } + + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECTFAIL]); + + /* + * Translate other errors into ISC_R_* flavors. + */ + switch (errno) { +#define ERROR_MATCH(a, b) \ + case a: \ + result = b; \ + break; + ERROR_MATCH(EACCES, ISC_R_NOPERM); + ERROR_MATCH(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL); + ERROR_MATCH(ECONNREFUSED, ISC_R_CONNREFUSED); + ERROR_MATCH(EHOSTUNREACH, ISC_R_HOSTUNREACH); +#ifdef EHOSTDOWN + ERROR_MATCH(EHOSTDOWN, ISC_R_HOSTUNREACH); +#endif /* ifdef EHOSTDOWN */ + ERROR_MATCH(ENETUNREACH, ISC_R_NETUNREACH); + ERROR_MATCH(ENOBUFS, ISC_R_NORESOURCES); + ERROR_MATCH(EPERM, ISC_R_HOSTUNREACH); + ERROR_MATCH(EPIPE, ISC_R_NOTCONNECTED); + ERROR_MATCH(ETIMEDOUT, ISC_R_TIMEDOUT); + ERROR_MATCH(ECONNRESET, ISC_R_CONNECTIONRESET); +#undef ERROR_MATCH + default: + result = ISC_R_UNEXPECTED; + isc_sockaddr_format(&sock->peer_address, peerbuf, + sizeof(peerbuf)); + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "internal_connect: connect(%s) %s", + peerbuf, strbuf); + } + } else { + inc_stats(sock->manager->stats, + sock->statsindex[STATID_CONNECT]); + result = ISC_R_SUCCESS; + sock->connected = 1; + sock->bound = 1; + } + + do { + dev->result = result; + send_connectdone_event(sock, &dev); + dev = ISC_LIST_HEAD(sock->connect_list); + } while (dev != NULL); + +finish: + unwatch_fd(&sock->manager->threads[sock->threadid], sock->fd, + SELECT_POKE_CONNECT); +} + +isc_result_t +isc_socket_getpeername(isc_socket_t *sock, isc_sockaddr_t *addressp) { + isc_result_t result; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(addressp != NULL); + + LOCK(&sock->lock); + + if (sock->connected) { + *addressp = sock->peer_address; + result = ISC_R_SUCCESS; + } else { + result = ISC_R_NOTCONNECTED; + } + + UNLOCK(&sock->lock); + + return (result); +} + +isc_result_t +isc_socket_getsockname(isc_socket_t *sock, isc_sockaddr_t *addressp) { + socklen_t len; + isc_result_t result; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(addressp != NULL); + + LOCK(&sock->lock); + + if (!sock->bound) { + result = ISC_R_NOTBOUND; + goto out; + } + + result = ISC_R_SUCCESS; + + len = sizeof(addressp->type); + if (getsockname(sock->fd, &addressp->type.sa, (void *)&len) < 0) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "getsockname: %s", strbuf); + result = ISC_R_UNEXPECTED; + goto out; + } + addressp->length = (unsigned int)len; + +out: + UNLOCK(&sock->lock); + + return (result); +} + +/* + * Run through the list of events on this socket, and cancel the ones + * queued for task "task" of type "how". "how" is a bitmask. + */ +void +isc_socket_cancel(isc_socket_t *sock, isc_task_t *task, unsigned int how) { + REQUIRE(VALID_SOCKET(sock)); + + /* + * Quick exit if there is nothing to do. Don't even bother locking + * in this case. + */ + if (how == 0) { + return; + } + + LOCK(&sock->lock); + + /* + * All of these do the same thing, more or less. + * Each will: + * o If the internal event is marked as "posted" try to + * remove it from the task's queue. If this fails, mark it + * as canceled instead, and let the task clean it up later. + * o For each I/O request for that task of that type, post + * its done event with status of "ISC_R_CANCELED". + * o Reset any state needed. + */ + if (((how & ISC_SOCKCANCEL_RECV) != 0) && + !ISC_LIST_EMPTY(sock->recv_list)) + { + isc_socketevent_t *dev; + isc_socketevent_t *next; + isc_task_t *current_task; + + dev = ISC_LIST_HEAD(sock->recv_list); + + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + dev->result = ISC_R_CANCELED; + send_recvdone_event(sock, &dev); + } + dev = next; + } + } + + if (((how & ISC_SOCKCANCEL_SEND) != 0) && + !ISC_LIST_EMPTY(sock->send_list)) + { + isc_socketevent_t *dev; + isc_socketevent_t *next; + isc_task_t *current_task; + + dev = ISC_LIST_HEAD(sock->send_list); + + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + dev->result = ISC_R_CANCELED; + send_senddone_event(sock, &dev); + } + dev = next; + } + } + + if (((how & ISC_SOCKCANCEL_ACCEPT) != 0) && + !ISC_LIST_EMPTY(sock->accept_list)) + { + isc_socket_newconnev_t *dev; + isc_socket_newconnev_t *next; + isc_task_t *current_task; + + dev = ISC_LIST_HEAD(sock->accept_list); + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + ISC_LIST_UNLINK(sock->accept_list, dev, + ev_link); + + isc_refcount_decrementz( + &NEWCONNSOCK(dev)->references); + free_socket((isc_socket_t **)&dev->newsocket); + + dev->result = ISC_R_CANCELED; + dev->ev_sender = sock; + isc_task_sendtoanddetach(¤t_task, + ISC_EVENT_PTR(&dev), + sock->threadid); + } + + dev = next; + } + } + + if (((how & ISC_SOCKCANCEL_CONNECT) != 0) && + !ISC_LIST_EMPTY(sock->connect_list)) + { + isc_socket_connev_t *dev; + isc_socket_connev_t *next; + isc_task_t *current_task; + + INSIST(sock->connecting); + sock->connecting = 0; + + dev = ISC_LIST_HEAD(sock->connect_list); + + while (dev != NULL) { + current_task = dev->ev_sender; + next = ISC_LIST_NEXT(dev, ev_link); + + if ((task == NULL) || (task == current_task)) { + dev->result = ISC_R_CANCELED; + send_connectdone_event(sock, &dev); + } + dev = next; + } + } + + UNLOCK(&sock->lock); +} + +isc_sockettype_t +isc_socket_gettype(isc_socket_t *sock) { + REQUIRE(VALID_SOCKET(sock)); + + return (sock->type); +} + +void +isc_socket_ipv6only(isc_socket_t *sock, bool yes) { +#if defined(IPV6_V6ONLY) && !defined(__OpenBSD__) + int onoff = yes ? 1 : 0; +#else /* if defined(IPV6_V6ONLY) */ + UNUSED(yes); + UNUSED(sock); +#endif /* if defined(IPV6_V6ONLY) */ + + REQUIRE(VALID_SOCKET(sock)); + INSIST(!sock->dupped); + +#if defined(IPV6_V6ONLY) && !defined(__OpenBSD__) + if (sock->pf == AF_INET6) { + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_V6ONLY, + (void *)&onoff, sizeof(int)) < 0) + { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_V6ONLY) failed: " + "%s", + sock->fd, strbuf); + } + } +#endif /* ifdef IPV6_V6ONLY */ +} + +static void +setdscp(isc_socket_t *sock, isc_dscp_t dscp) { +#if defined(IP_TOS) || defined(IPV6_TCLASS) + int value = dscp << 2; +#endif /* if defined(IP_TOS) || defined(IPV6_TCLASS) */ + + sock->dscp = dscp; + +#ifdef IP_TOS + if (sock->pf == AF_INET) { + if (setsockopt(sock->fd, IPPROTO_IP, IP_TOS, (void *)&value, + sizeof(value)) < 0) + { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IP_TOS, %.02x) " + "failed: %s", + sock->fd, value >> 2, strbuf); + } + } +#endif /* ifdef IP_TOS */ +#ifdef IPV6_TCLASS + if (sock->pf == AF_INET6) { + if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_TCLASS, + (void *)&value, sizeof(value)) < 0) + { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, + "setsockopt(%d, IPV6_TCLASS, %.02x) " + "failed: %s", + sock->fd, dscp >> 2, strbuf); + } + } +#endif /* ifdef IPV6_TCLASS */ +} + +void +isc_socket_dscp(isc_socket_t *sock, isc_dscp_t dscp) { + REQUIRE(VALID_SOCKET(sock)); + REQUIRE(dscp < 0x40); + +#if !defined(IP_TOS) && !defined(IPV6_TCLASS) + UNUSED(dscp); +#else /* if !defined(IP_TOS) && !defined(IPV6_TCLASS) */ + if (dscp < 0) { + return; + } + + /* The DSCP value must not be changed once it has been set. */ + if (isc_dscp_check_value != -1) { + INSIST(dscp == isc_dscp_check_value); + } +#endif /* if !defined(IP_TOS) && !defined(IPV6_TCLASS) */ + +#ifdef notyet + REQUIRE(!sock->dupped); +#endif /* ifdef notyet */ + + setdscp(sock, dscp); +} + +isc_socketevent_t * +isc_socket_socketevent(isc_mem_t *mctx, void *sender, isc_eventtype_t eventtype, + isc_taskaction_t action, void *arg) { + return (allocate_socketevent(mctx, sender, eventtype, action, arg)); +} + +void +isc_socket_setname(isc_socket_t *sock, const char *name, void *tag) { + /* + * Name 'sock'. + */ + + REQUIRE(VALID_SOCKET(sock)); + + LOCK(&sock->lock); + strlcpy(sock->name, name, sizeof(sock->name)); + sock->tag = tag; + UNLOCK(&sock->lock); +} + +const char * +isc_socket_getname(isc_socket_t *sock) { + return (sock->name); +} + +void * +isc_socket_gettag(isc_socket_t *sock) { + return (sock->tag); +} + +int +isc_socket_getfd(isc_socket_t *sock) { + return ((short)sock->fd); +} + +static isc_once_t hasreuseport_once = ISC_ONCE_INIT; +static bool hasreuseport = false; + +static void +init_hasreuseport() { +/* + * SO_REUSEPORT works very differently on *BSD and on Linux (because why not). + * We only want to use it on Linux, if it's available. On BSD we want to dup() + * sockets instead of re-binding them. + */ +#if (defined(SO_REUSEPORT) && defined(__linux__)) || \ + (defined(SO_REUSEPORT_LB) && defined(__FreeBSD_kernel__)) + int sock, yes = 1; + sock = socket(AF_INET, SOCK_DGRAM, 0); + if (sock < 0) { + sock = socket(AF_INET6, SOCK_DGRAM, 0); + if (sock < 0) { + return; + } + } + if (setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (void *)&yes, + sizeof(yes)) < 0) + { + close(sock); + return; +#if defined(__FreeBSD_kernel__) + } else if (setsockopt(sock, SOL_SOCKET, SO_REUSEPORT_LB, (void *)&yes, + sizeof(yes)) < 0) +#else /* if defined(__FreeBSD_kernel__) */ + } else if (setsockopt(sock, SOL_SOCKET, SO_REUSEPORT, (void *)&yes, + sizeof(yes)) < 0) +#endif /* if defined(__FreeBSD_kernel__) */ + { + close(sock); + return; + } + hasreuseport = true; + close(sock); +#endif /* if (defined(SO_REUSEPORT) && defined(__linux__)) || \ + * (defined(SO_REUSEPORT_LB) && defined(__FreeBSD_kernel__)) */ +} + +bool +isc_socket_hasreuseport() { + RUNTIME_CHECK(isc_once_do(&hasreuseport_once, init_hasreuseport) == + ISC_R_SUCCESS); + return (hasreuseport); +} + +#if defined(HAVE_LIBXML2) || defined(HAVE_JSON_C) +static const char * +_socktype(isc_sockettype_t type) { + switch (type) { + case isc_sockettype_udp: + return ("udp"); + case isc_sockettype_tcp: + return ("tcp"); + case isc_sockettype_unix: + return ("unix"); + default: + return ("not-initialized"); + } +} +#endif /* if defined(HAVE_LIBXML2) || defined(HAVE_JSON_C) */ + +#ifdef HAVE_LIBXML2 +#define TRY0(a) \ + do { \ + xmlrc = (a); \ + if (xmlrc < 0) \ + goto error; \ + } while (0) +int +isc_socketmgr_renderxml(isc_socketmgr_t *mgr, void *writer0) { + isc_socket_t *sock = NULL; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + isc_sockaddr_t addr; + socklen_t len; + int xmlrc; + xmlTextWriterPtr writer = (xmlTextWriterPtr)writer0; + + LOCK(&mgr->lock); + + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "sockets")); + sock = ISC_LIST_HEAD(mgr->socklist); + while (sock != NULL) { + LOCK(&sock->lock); + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "socket")); + + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "id")); + TRY0(xmlTextWriterWriteFormatString(writer, "%p", sock)); + TRY0(xmlTextWriterEndElement(writer)); + + if (sock->name[0] != 0) { + TRY0(xmlTextWriterStartElement(writer, + ISC_XMLCHAR "name")); + TRY0(xmlTextWriterWriteFormatString(writer, "%s", + sock->name)); + TRY0(xmlTextWriterEndElement(writer)); /* name */ + } + + TRY0(xmlTextWriterStartElement(writer, + ISC_XMLCHAR "references")); + TRY0(xmlTextWriterWriteFormatString( + writer, "%d", + (int)isc_refcount_current(&sock->references))); + TRY0(xmlTextWriterEndElement(writer)); + + TRY0(xmlTextWriterWriteElement( + writer, ISC_XMLCHAR "type", + ISC_XMLCHAR _socktype(sock->type))); + + if (sock->connected) { + isc_sockaddr_format(&sock->peer_address, peerbuf, + sizeof(peerbuf)); + TRY0(xmlTextWriterWriteElement( + writer, ISC_XMLCHAR "peer-address", + ISC_XMLCHAR peerbuf)); + } + + len = sizeof(addr); + if (getsockname(sock->fd, &addr.type.sa, (void *)&len) == 0) { + isc_sockaddr_format(&addr, peerbuf, sizeof(peerbuf)); + TRY0(xmlTextWriterWriteElement( + writer, ISC_XMLCHAR "local-address", + ISC_XMLCHAR peerbuf)); + } + + TRY0(xmlTextWriterStartElement(writer, ISC_XMLCHAR "states")); + if (sock->listener) { + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "listener")); + } + if (sock->connected) { + TRY0(xmlTextWriterWriteElement( + writer, ISC_XMLCHAR "state", + ISC_XMLCHAR "connected")); + } + if (sock->connecting) { + TRY0(xmlTextWriterWriteElement( + writer, ISC_XMLCHAR "state", + ISC_XMLCHAR "connecting")); + } + if (sock->bound) { + TRY0(xmlTextWriterWriteElement(writer, + ISC_XMLCHAR "state", + ISC_XMLCHAR "bound")); + } + + TRY0(xmlTextWriterEndElement(writer)); /* states */ + + TRY0(xmlTextWriterEndElement(writer)); /* socket */ + + UNLOCK(&sock->lock); + sock = ISC_LIST_NEXT(sock, link); + } + TRY0(xmlTextWriterEndElement(writer)); /* sockets */ + +error: + if (sock != NULL) { + UNLOCK(&sock->lock); + } + + UNLOCK(&mgr->lock); + + return (xmlrc); +} +#endif /* HAVE_LIBXML2 */ + +#ifdef HAVE_JSON_C +#define CHECKMEM(m) \ + do { \ + if (m == NULL) { \ + result = ISC_R_NOMEMORY; \ + goto error; \ + } \ + } while (0) + +isc_result_t +isc_socketmgr_renderjson(isc_socketmgr_t *mgr, void *stats0) { + isc_result_t result = ISC_R_SUCCESS; + isc_socket_t *sock = NULL; + char peerbuf[ISC_SOCKADDR_FORMATSIZE]; + isc_sockaddr_t addr; + socklen_t len; + json_object *obj, *array = json_object_new_array(); + json_object *stats = (json_object *)stats0; + + CHECKMEM(array); + + LOCK(&mgr->lock); + + sock = ISC_LIST_HEAD(mgr->socklist); + while (sock != NULL) { + json_object *states, *entry = json_object_new_object(); + char buf[255]; + + CHECKMEM(entry); + json_object_array_add(array, entry); + + LOCK(&sock->lock); + + snprintf(buf, sizeof(buf), "%p", sock); + obj = json_object_new_string(buf); + CHECKMEM(obj); + json_object_object_add(entry, "id", obj); + + if (sock->name[0] != 0) { + obj = json_object_new_string(sock->name); + CHECKMEM(obj); + json_object_object_add(entry, "name", obj); + } + + obj = json_object_new_int( + (int)isc_refcount_current(&sock->references)); + CHECKMEM(obj); + json_object_object_add(entry, "references", obj); + + obj = json_object_new_string(_socktype(sock->type)); + CHECKMEM(obj); + json_object_object_add(entry, "type", obj); + + if (sock->connected) { + isc_sockaddr_format(&sock->peer_address, peerbuf, + sizeof(peerbuf)); + obj = json_object_new_string(peerbuf); + CHECKMEM(obj); + json_object_object_add(entry, "peer-address", obj); + } + + len = sizeof(addr); + if (getsockname(sock->fd, &addr.type.sa, (void *)&len) == 0) { + isc_sockaddr_format(&addr, peerbuf, sizeof(peerbuf)); + obj = json_object_new_string(peerbuf); + CHECKMEM(obj); + json_object_object_add(entry, "local-address", obj); + } + + states = json_object_new_array(); + CHECKMEM(states); + json_object_object_add(entry, "states", states); + + if (sock->listener) { + obj = json_object_new_string("listener"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->connected) { + obj = json_object_new_string("connected"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->connecting) { + obj = json_object_new_string("connecting"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + if (sock->bound) { + obj = json_object_new_string("bound"); + CHECKMEM(obj); + json_object_array_add(states, obj); + } + + UNLOCK(&sock->lock); + sock = ISC_LIST_NEXT(sock, link); + } + + json_object_object_add(stats, "sockets", array); + array = NULL; + result = ISC_R_SUCCESS; + +error: + if (array != NULL) { + json_object_put(array); + } + + if (sock != NULL) { + UNLOCK(&sock->lock); + } + + UNLOCK(&mgr->lock); + + return (result); +} +#endif /* HAVE_JSON_C */ diff --git a/lib/isc/unix/socket_p.h b/lib/isc/unix/socket_p.h new file mode 100644 index 0000000..3323cab --- /dev/null +++ b/lib/isc/unix/socket_p.h @@ -0,0 +1,27 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#ifndef ISC_SOCKET_P_H +#define ISC_SOCKET_P_H + +/*! \file */ + +#include <sys/time.h> + +typedef struct isc_socketwait isc_socketwait_t; +int +isc__socketmgr_waitevents(isc_socketmgr_t *, struct timeval *, + isc_socketwait_t **); +isc_result_t +isc__socketmgr_dispatch(isc_socketmgr_t *, isc_socketwait_t *); +#endif /* ISC_SOCKET_P_H */ diff --git a/lib/isc/unix/stdio.c b/lib/isc/unix/stdio.c new file mode 100644 index 0000000..12ab678 --- /dev/null +++ b/lib/isc/unix/stdio.c @@ -0,0 +1,152 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +#include <errno.h> +#include <unistd.h> + +#include <isc/stat.h> +#include <isc/stdio.h> +#include <isc/util.h> + +#include "errno2result.h" + +isc_result_t +isc_stdio_open(const char *filename, const char *mode, FILE **fp) { + FILE *f; + + f = fopen(filename, mode); + if (f == NULL) { + return (isc__errno2result(errno)); + } + *fp = f; + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_stdio_close(FILE *f) { + int r; + + r = fclose(f); + if (r == 0) { + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} + +isc_result_t +isc_stdio_seek(FILE *f, off_t offset, int whence) { + int r; + + r = fseeko(f, offset, whence); + if (r == 0) { + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} + +isc_result_t +isc_stdio_tell(FILE *f, off_t *offsetp) { + off_t r; + + REQUIRE(offsetp != NULL); + + r = ftello(f); + if (r >= 0) { + *offsetp = r; + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} + +isc_result_t +isc_stdio_read(void *ptr, size_t size, size_t nmemb, FILE *f, size_t *nret) { + isc_result_t result = ISC_R_SUCCESS; + size_t r; + + clearerr(f); + r = fread(ptr, size, nmemb, f); + if (r != nmemb) { + if (feof(f)) { + result = ISC_R_EOF; + } else { + result = isc__errno2result(errno); + } + } + if (nret != NULL) { + *nret = r; + } + return (result); +} + +isc_result_t +isc_stdio_write(const void *ptr, size_t size, size_t nmemb, FILE *f, + size_t *nret) { + isc_result_t result = ISC_R_SUCCESS; + size_t r; + + clearerr(f); + r = fwrite(ptr, size, nmemb, f); + if (r != nmemb) { + result = isc__errno2result(errno); + } + if (nret != NULL) { + *nret = r; + } + return (result); +} + +isc_result_t +isc_stdio_flush(FILE *f) { + int r; + + r = fflush(f); + if (r == 0) { + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} + +/* + * OpenBSD has deprecated ENOTSUP in favor of EOPNOTSUPP. + */ +#if defined(EOPNOTSUPP) && !defined(ENOTSUP) +#define ENOTSUP EOPNOTSUPP +#endif /* if defined(EOPNOTSUPP) && !defined(ENOTSUP) */ + +isc_result_t +isc_stdio_sync(FILE *f) { + struct stat buf; + int r; + + if (fstat(fileno(f), &buf) != 0) { + return (isc__errno2result(errno)); + } + + /* + * Only call fsync() on regular files. + */ + if ((buf.st_mode & S_IFMT) != S_IFREG) { + return (ISC_R_SUCCESS); + } + + r = fsync(fileno(f)); + if (r == 0) { + return (ISC_R_SUCCESS); + } else { + return (isc__errno2result(errno)); + } +} diff --git a/lib/isc/unix/stdtime.c b/lib/isc/unix/stdtime.c new file mode 100644 index 0000000..ada19d0 --- /dev/null +++ b/lib/isc/unix/stdtime.c @@ -0,0 +1,68 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <errno.h> +#include <stdbool.h> +#include <stddef.h> /* NULL */ +#include <stdlib.h> /* NULL */ +#include <syslog.h> +#include <time.h> + +#include <isc/stdtime.h> +#include <isc/strerr.h> +#include <isc/util.h> + +#define NS_PER_S 1000000000 /*%< Nanoseconds per second. */ + +#if defined(CLOCK_REALTIME_COARSE) +#define CLOCKSOURCE CLOCK_REALTIME_COARSE +#elif defined(CLOCK_REALTIME_FAST) +#define CLOCKSOURCE CLOCK_REALTIME_FAST +#else /* if defined(CLOCK_REALTIME_COARSE) */ +#define CLOCKSOURCE CLOCK_REALTIME +#endif /* if defined(CLOCK_REALTIME_COARSE) */ + +void +isc_stdtime_get(isc_stdtime_t *t) { + REQUIRE(t != NULL); + + struct timespec ts; + + if (clock_gettime(CLOCKSOURCE, &ts) == -1) { + char strbuf[ISC_STRERRORSIZE]; + strerror_r(errno, strbuf, sizeof(strbuf)); + isc_error_fatal(__FILE__, __LINE__, "clock_gettime failed: %s", + strbuf); + } + + REQUIRE(ts.tv_sec > 0 && ts.tv_nsec >= 0 && ts.tv_nsec < NS_PER_S); + + *t = (isc_stdtime_t)ts.tv_sec; +} + +void +isc_stdtime_tostring(isc_stdtime_t t, char *out, size_t outlen) { + time_t when; + + REQUIRE(out != NULL); + REQUIRE(outlen >= 26); + + UNUSED(outlen); + + /* time_t and isc_stdtime_t might be different sizes */ + when = t; + INSIST((ctime_r(&when, out) != NULL)); + *(out + strlen(out) - 1) = '\0'; +} diff --git a/lib/isc/unix/syslog.c b/lib/isc/unix/syslog.c new file mode 100644 index 0000000..81b99ca --- /dev/null +++ b/lib/isc/unix/syslog.c @@ -0,0 +1,73 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <stdlib.h> +#include <syslog.h> + +#include <isc/result.h> +#include <isc/string.h> +#include <isc/syslog.h> +#include <isc/util.h> + +static struct dsn_c_pvt_sfnt { + int val; + const char *strval; +} facilities[] = { { LOG_KERN, "kern" }, + { LOG_USER, "user" }, + { LOG_MAIL, "mail" }, + { LOG_DAEMON, "daemon" }, + { LOG_AUTH, "auth" }, + { LOG_SYSLOG, "syslog" }, + { LOG_LPR, "lpr" }, +#ifdef LOG_NEWS + { LOG_NEWS, "news" }, +#endif /* ifdef LOG_NEWS */ +#ifdef LOG_UUCP + { LOG_UUCP, "uucp" }, +#endif /* ifdef LOG_UUCP */ +#ifdef LOG_CRON + { LOG_CRON, "cron" }, +#endif /* ifdef LOG_CRON */ +#ifdef LOG_AUTHPRIV + { LOG_AUTHPRIV, "authpriv" }, +#endif /* ifdef LOG_AUTHPRIV */ +#ifdef LOG_FTP + { LOG_FTP, "ftp" }, +#endif /* ifdef LOG_FTP */ + { LOG_LOCAL0, "local0" }, + { LOG_LOCAL1, "local1" }, + { LOG_LOCAL2, "local2" }, + { LOG_LOCAL3, "local3" }, + { LOG_LOCAL4, "local4" }, + { LOG_LOCAL5, "local5" }, + { LOG_LOCAL6, "local6" }, + { LOG_LOCAL7, "local7" }, + { 0, NULL } }; + +isc_result_t +isc_syslog_facilityfromstring(const char *str, int *facilityp) { + int i; + + REQUIRE(str != NULL); + REQUIRE(facilityp != NULL); + + for (i = 0; facilities[i].strval != NULL; i++) { + if (strcasecmp(facilities[i].strval, str) == 0) { + *facilityp = facilities[i].val; + return (ISC_R_SUCCESS); + } + } + return (ISC_R_NOTFOUND); +} diff --git a/lib/isc/unix/time.c b/lib/isc/unix/time.c new file mode 100644 index 0000000..02291e4 --- /dev/null +++ b/lib/isc/unix/time.c @@ -0,0 +1,553 @@ +/* + * Copyright (C) Internet Systems Consortium, Inc. ("ISC") + * + * SPDX-License-Identifier: MPL-2.0 + * + * This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, you can obtain one at https://mozilla.org/MPL/2.0/. + * + * See the COPYRIGHT file distributed with this work for additional + * information regarding copyright ownership. + */ + +/*! \file */ + +#include <errno.h> +#include <inttypes.h> +#include <limits.h> +#include <stdbool.h> +#include <stdlib.h> +#include <sys/time.h> /* Required for struct timeval on some platforms. */ +#include <syslog.h> +#include <time.h> + +#include <isc/log.h> +#include <isc/platform.h> +#include <isc/print.h> +#include <isc/strerr.h> +#include <isc/string.h> +#include <isc/time.h> +#include <isc/tm.h> +#include <isc/util.h> + +#define NS_PER_S 1000000000 /*%< Nanoseconds per second. */ +#define NS_PER_US 1000 /*%< Nanoseconds per microsecond. */ +#define NS_PER_MS 1000000 /*%< Nanoseconds per millisecond. */ + +#if defined(CLOCK_REALTIME) +#define CLOCKSOURCE_HIRES CLOCK_REALTIME +#endif /* #if defined(CLOCK_REALTIME) */ + +#if defined(CLOCK_REALTIME_COARSE) +#define CLOCKSOURCE CLOCK_REALTIME_COARSE +#elif defined(CLOCK_REALTIME_FAST) +#define CLOCKSOURCE CLOCK_REALTIME_FAST +#else /* if defined(CLOCK_REALTIME_COARSE) */ +#define CLOCKSOURCE CLOCK_REALTIME +#endif /* if defined(CLOCK_REALTIME_COARSE) */ + +#if !defined(CLOCKSOURCE_HIRES) +#define CLOCKSOURCE_HIRES CLOCKSOURCE +#endif /* #ifndef CLOCKSOURCE_HIRES */ + +/*% + *** Intervals + ***/ + +#if !defined(UNIT_TESTING) +static const isc_interval_t zero_interval = { 0, 0 }; +const isc_interval_t *const isc_interval_zero = &zero_interval; +#endif + +void +isc_interval_set(isc_interval_t *i, unsigned int seconds, + unsigned int nanoseconds) { + REQUIRE(i != NULL); + REQUIRE(nanoseconds < NS_PER_S); + + i->seconds = seconds; + i->nanoseconds = nanoseconds; +} + +bool +isc_interval_iszero(const isc_interval_t *i) { + REQUIRE(i != NULL); + INSIST(i->nanoseconds < NS_PER_S); + + if (i->seconds == 0 && i->nanoseconds == 0) { + return (true); + } + + return (false); +} + +/*** + *** Absolute Times + ***/ + +#if !defined(UNIT_TESTING) +static const isc_time_t epoch = { 0, 0 }; +const isc_time_t *const isc_time_epoch = &epoch; +#endif + +void +isc_time_set(isc_time_t *t, unsigned int seconds, unsigned int nanoseconds) { + REQUIRE(t != NULL); + REQUIRE(nanoseconds < NS_PER_S); + + t->seconds = seconds; + t->nanoseconds = nanoseconds; +} + +void +isc_time_settoepoch(isc_time_t *t) { + REQUIRE(t != NULL); + + t->seconds = 0; + t->nanoseconds = 0; +} + +bool +isc_time_isepoch(const isc_time_t *t) { + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + + if (t->seconds == 0 && t->nanoseconds == 0) { + return (true); + } + + return (false); +} + +static isc_result_t +time_now(isc_time_t *t, clockid_t clock) { + struct timespec ts; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(t != NULL); + + if (clock_gettime(clock, &ts) == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "%s", strbuf); + return (ISC_R_UNEXPECTED); + } + + if (ts.tv_sec < 0 || ts.tv_nsec < 0 || ts.tv_nsec >= NS_PER_S) { + return (ISC_R_UNEXPECTED); + } + + /* + * Ensure the tv_sec value fits in t->seconds. + */ + if (sizeof(ts.tv_sec) > sizeof(t->seconds) && + ((ts.tv_sec | (unsigned int)-1) ^ (unsigned int)-1) != 0U) + { + return (ISC_R_RANGE); + } + + t->seconds = ts.tv_sec; + t->nanoseconds = ts.tv_nsec; + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_time_now_hires(isc_time_t *t) { + return time_now(t, CLOCKSOURCE_HIRES); +} + +isc_result_t +isc_time_now(isc_time_t *t) { + return time_now(t, CLOCKSOURCE); +} + +isc_result_t +isc_time_nowplusinterval(isc_time_t *t, const isc_interval_t *i) { + struct timespec ts; + char strbuf[ISC_STRERRORSIZE]; + + REQUIRE(t != NULL); + REQUIRE(i != NULL); + INSIST(i->nanoseconds < NS_PER_S); + + if (clock_gettime(CLOCKSOURCE, &ts) == -1) { + strerror_r(errno, strbuf, sizeof(strbuf)); + UNEXPECTED_ERROR(__FILE__, __LINE__, "%s", strbuf); + return (ISC_R_UNEXPECTED); + } + + if (ts.tv_sec < 0 || ts.tv_nsec < 0 || ts.tv_nsec >= NS_PER_S) { + return (ISC_R_UNEXPECTED); + } + + /* + * Ensure the resulting seconds value fits in the size of an + * unsigned int. (It is written this way as a slight optimization; + * note that even if both values == INT_MAX, then when added + * and getting another 1 added below the result is UINT_MAX.) + */ + if ((ts.tv_sec > INT_MAX || i->seconds > INT_MAX) && + ((long long)ts.tv_sec + i->seconds > UINT_MAX)) + { + return (ISC_R_RANGE); + } + + t->seconds = ts.tv_sec + i->seconds; + t->nanoseconds = ts.tv_nsec + i->nanoseconds; + if (t->nanoseconds >= NS_PER_S) { + t->seconds++; + t->nanoseconds -= NS_PER_S; + } + + return (ISC_R_SUCCESS); +} + +int +isc_time_compare(const isc_time_t *t1, const isc_time_t *t2) { + REQUIRE(t1 != NULL && t2 != NULL); + INSIST(t1->nanoseconds < NS_PER_S && t2->nanoseconds < NS_PER_S); + + if (t1->seconds < t2->seconds) { + return (-1); + } + if (t1->seconds > t2->seconds) { + return (1); + } + if (t1->nanoseconds < t2->nanoseconds) { + return (-1); + } + if (t1->nanoseconds > t2->nanoseconds) { + return (1); + } + return (0); +} + +isc_result_t +isc_time_add(const isc_time_t *t, const isc_interval_t *i, isc_time_t *result) { + REQUIRE(t != NULL && i != NULL && result != NULL); + REQUIRE(t->nanoseconds < NS_PER_S && i->nanoseconds < NS_PER_S); + + /* Seconds */ + if (t->seconds > UINT_MAX - i->seconds) { + return (ISC_R_RANGE); + } + result->seconds = t->seconds + i->seconds; + + /* Nanoseconds */ + result->nanoseconds = t->nanoseconds + i->nanoseconds; + if (result->nanoseconds >= NS_PER_S) { + if (result->seconds == UINT_MAX) { + return (ISC_R_RANGE); + } + result->nanoseconds -= NS_PER_S; + result->seconds++; + } + + return (ISC_R_SUCCESS); +} + +isc_result_t +isc_time_subtract(const isc_time_t *t, const isc_interval_t *i, + isc_time_t *result) { + REQUIRE(t != NULL && i != NULL && result != NULL); + REQUIRE(t->nanoseconds < NS_PER_S && i->nanoseconds < NS_PER_S); + + /* Seconds */ + if (t->seconds < i->seconds) { + return (ISC_R_RANGE); + } + result->seconds = t->seconds - i->seconds; + + /* Nanoseconds */ + if (t->nanoseconds >= i->nanoseconds) { + result->nanoseconds = t->nanoseconds - i->nanoseconds; + } else { + if (result->seconds == 0) { + return (ISC_R_RANGE); + } + result->seconds--; + result->nanoseconds = NS_PER_S + t->nanoseconds - + i->nanoseconds; + } + + return (ISC_R_SUCCESS); +} + +uint64_t +isc_time_microdiff(const isc_time_t *t1, const isc_time_t *t2) { + uint64_t i1, i2, i3; + + REQUIRE(t1 != NULL && t2 != NULL); + INSIST(t1->nanoseconds < NS_PER_S && t2->nanoseconds < NS_PER_S); + + i1 = (uint64_t)t1->seconds * NS_PER_S + t1->nanoseconds; + i2 = (uint64_t)t2->seconds * NS_PER_S + t2->nanoseconds; + + if (i1 <= i2) { + return (0); + } + + i3 = i1 - i2; + + /* + * Convert to microseconds. + */ + i3 /= NS_PER_US; + + return (i3); +} + +uint32_t +isc_time_seconds(const isc_time_t *t) { + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + + return ((uint32_t)t->seconds); +} + +isc_result_t +isc_time_secondsastimet(const isc_time_t *t, time_t *secondsp) { + time_t seconds; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + + /* + * Ensure that the number of seconds represented by t->seconds + * can be represented by a time_t. Since t->seconds is an + * unsigned int and since time_t is mostly opaque, this is + * trickier than it seems. (This standardized opaqueness of + * time_t is *very* frustrating; time_t is not even limited to + * being an integral type.) + * + * The mission, then, is to avoid generating any kind of warning + * about "signed versus unsigned" while trying to determine if + * the unsigned int t->seconds is out range for tv_sec, + * which is pretty much only true if time_t is a signed integer + * of the same size as the return value of isc_time_seconds. + * + * If the paradox in the if clause below is true, t->seconds is + * out of range for time_t. + */ + seconds = (time_t)t->seconds; + + INSIST(sizeof(unsigned int) == sizeof(uint32_t)); + INSIST(sizeof(time_t) >= sizeof(uint32_t)); + + if (t->seconds > (~0U >> 1) && seconds <= (time_t)(~0U >> 1)) { + return (ISC_R_RANGE); + } + + *secondsp = seconds; + + return (ISC_R_SUCCESS); +} + +uint32_t +isc_time_nanoseconds(const isc_time_t *t) { + REQUIRE(t != NULL); + + ENSURE(t->nanoseconds < NS_PER_S); + + return ((uint32_t)t->nanoseconds); +} + +void +isc_time_formattimestamp(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%d-%b-%Y %X", localtime_r(&now, &tm)); + INSIST(flen < len); + if (flen != 0) { + snprintf(buf + flen, len - flen, ".%03u", + t->nanoseconds / NS_PER_MS); + } else { + strlcpy(buf, "99-Bad-9999 99:99:99.999", len); + } +} + +void +isc_time_formathttptimestamp(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + /* + * 5 spaces, 1 comma, 3 GMT, 2 %d, 4 %Y, 8 %H:%M:%S, 3+ %a, 3+ + * %b (29+) + */ + now = (time_t)t->seconds; + flen = strftime(buf, len, "%a, %d %b %Y %H:%M:%S GMT", + gmtime_r(&now, &tm)); + INSIST(flen < len); +} + +isc_result_t +isc_time_parsehttptimestamp(char *buf, isc_time_t *t) { + struct tm t_tm; + time_t when; + char *p; + + REQUIRE(buf != NULL); + REQUIRE(t != NULL); + + p = isc_tm_strptime(buf, "%a, %d %b %Y %H:%M:%S", &t_tm); + if (p == NULL) { + return (ISC_R_UNEXPECTED); + } + when = isc_tm_timegm(&t_tm); + if (when == -1) { + return (ISC_R_UNEXPECTED); + } + isc_time_set(t, when, 0); + return (ISC_R_SUCCESS); +} + +void +isc_time_formatISO8601L(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%S", localtime_r(&now, &tm)); + INSIST(flen < len); +} + +void +isc_time_formatISO8601Lms(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%S", localtime_r(&now, &tm)); + INSIST(flen < len); + if (flen > 0U && len - flen >= 6) { + snprintf(buf + flen, len - flen, ".%03u", + t->nanoseconds / NS_PER_MS); + } +} + +void +isc_time_formatISO8601Lus(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%S", localtime_r(&now, &tm)); + INSIST(flen < len); + if (flen > 0U && len - flen >= 6) { + snprintf(buf + flen, len - flen, ".%06u", + t->nanoseconds / NS_PER_US); + } +} + +void +isc_time_formatISO8601(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime_r(&now, &tm)); + INSIST(flen < len); +} + +void +isc_time_formatISO8601ms(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime_r(&now, &tm)); + INSIST(flen < len); + if (flen > 0U && len - flen >= 5) { + flen -= 1; /* rewind one character (Z) */ + snprintf(buf + flen, len - flen, ".%03uZ", + t->nanoseconds / NS_PER_MS); + } +} + +void +isc_time_formatISO8601us(const isc_time_t *t, char *buf, unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y-%m-%dT%H:%M:%SZ", gmtime_r(&now, &tm)); + INSIST(flen < len); + if (flen > 0U && len - flen >= 5) { + flen -= 1; /* rewind one character (Z) */ + snprintf(buf + flen, len - flen, ".%06uZ", + t->nanoseconds / NS_PER_US); + } +} + +void +isc_time_formatshorttimestamp(const isc_time_t *t, char *buf, + unsigned int len) { + time_t now; + unsigned int flen; + struct tm tm; + + REQUIRE(t != NULL); + INSIST(t->nanoseconds < NS_PER_S); + REQUIRE(buf != NULL); + REQUIRE(len > 0); + + now = (time_t)t->seconds; + flen = strftime(buf, len, "%Y%m%d%H%M%S", gmtime_r(&now, &tm)); + INSIST(flen < len); + if (flen > 0U && len - flen >= 5) { + snprintf(buf + flen, len - flen, "%03u", + t->nanoseconds / NS_PER_MS); + } +} |