summaryrefslogtreecommitdiffstats
path: root/common/exechelp-posix.c
diff options
context:
space:
mode:
Diffstat (limited to 'common/exechelp-posix.c')
-rw-r--r--common/exechelp-posix.c895
1 files changed, 895 insertions, 0 deletions
diff --git a/common/exechelp-posix.c b/common/exechelp-posix.c
new file mode 100644
index 0000000..0782152
--- /dev/null
+++ b/common/exechelp-posix.c
@@ -0,0 +1,895 @@
+/* exechelp.c - Fork and exec helpers for POSIX
+ * Copyright (C) 2004, 2007, 2008, 2009,
+ * 2010 Free Software Foundation, Inc.
+ *
+ * This file is part of GnuPG.
+ *
+ * This file is free software; you can redistribute it and/or modify
+ * it under the terms of either
+ *
+ * - the GNU Lesser General Public License as published by the Free
+ * Software Foundation; either version 3 of the License, or (at
+ * your option) any later version.
+ *
+ * or
+ *
+ * - the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at
+ * your option) any later version.
+ *
+ * or both in parallel, as here.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <config.h>
+
+#if defined(HAVE_W32_SYSTEM) || defined (HAVE_W32CE_SYSTEM)
+#error This code is only used on POSIX
+#endif
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <errno.h>
+#include <assert.h>
+#ifdef HAVE_SIGNAL_H
+# include <signal.h>
+#endif
+#include <unistd.h>
+#include <fcntl.h>
+
+#ifdef WITHOUT_NPTH /* Give the Makefile a chance to build without Pth. */
+#undef HAVE_NPTH
+#undef USE_NPTH
+#endif
+
+#ifdef HAVE_NPTH
+#include <npth.h>
+#endif
+#include <sys/wait.h>
+
+#ifdef HAVE_GETRLIMIT
+#include <sys/time.h>
+#include <sys/resource.h>
+#endif /*HAVE_GETRLIMIT*/
+
+#ifdef HAVE_STAT
+# include <sys/stat.h>
+#endif
+
+#if __linux__
+# include <sys/types.h>
+# include <dirent.h>
+#endif /*__linux__ */
+
+#include "util.h"
+#include "i18n.h"
+#include "sysutils.h"
+#include "exechelp.h"
+
+
+/* Helper */
+static inline gpg_error_t
+my_error_from_syserror (void)
+{
+ return gpg_err_make (default_errsource, gpg_err_code_from_syserror ());
+}
+
+static inline gpg_error_t
+my_error (int errcode)
+{
+ return gpg_err_make (default_errsource, errcode);
+}
+
+
+/* Return the maximum number of currently allowed open file
+ descriptors. Only useful on POSIX systems but returns a value on
+ other systems too. */
+int
+get_max_fds (void)
+{
+ int max_fds = -1;
+#ifdef HAVE_GETRLIMIT
+ struct rlimit rl;
+
+ /* Under Linux we can figure out the highest used file descriptor by
+ * reading /proc/PID/fd. This is in the common cases much fast than
+ * for example doing 4096 close calls where almost all of them will
+ * fail. On a system with a limit of 4096 files and only 8 files
+ * open with the highest number being 10, we speedup close_all_fds
+ * from 125ms to 0.4ms including readdir.
+ *
+ * Another option would be to close the file descriptors as returned
+ * from reading that directory - however then we need to snapshot
+ * that list before starting to close them. */
+#ifdef __linux__
+ {
+ DIR *dir = NULL;
+ struct dirent *dir_entry;
+ const char *s;
+ int x;
+
+ dir = opendir ("/proc/self/fd");
+ if (dir)
+ {
+ while ((dir_entry = readdir (dir)))
+ {
+ s = dir_entry->d_name;
+ if ( *s < '0' || *s > '9')
+ continue;
+ x = atoi (s);
+ if (x > max_fds)
+ max_fds = x;
+ }
+ closedir (dir);
+ }
+ if (max_fds != -1)
+ return max_fds + 1;
+ }
+#endif /* __linux__ */
+
+
+# ifdef RLIMIT_NOFILE
+ if (!getrlimit (RLIMIT_NOFILE, &rl))
+ max_fds = rl.rlim_max;
+# endif
+
+# ifdef RLIMIT_OFILE
+ if (max_fds == -1 && !getrlimit (RLIMIT_OFILE, &rl))
+ max_fds = rl.rlim_max;
+
+# endif
+#endif /*HAVE_GETRLIMIT*/
+
+#ifdef _SC_OPEN_MAX
+ if (max_fds == -1)
+ {
+ long int scres = sysconf (_SC_OPEN_MAX);
+ if (scres >= 0)
+ max_fds = scres;
+ }
+#endif
+
+#ifdef _POSIX_OPEN_MAX
+ if (max_fds == -1)
+ max_fds = _POSIX_OPEN_MAX;
+#endif
+
+#ifdef OPEN_MAX
+ if (max_fds == -1)
+ max_fds = OPEN_MAX;
+#endif
+
+ if (max_fds == -1)
+ max_fds = 256; /* Arbitrary limit. */
+
+ /* AIX returns INT32_MAX instead of a proper value. We assume that
+ this is always an error and use an arbitrary limit. */
+#ifdef INT32_MAX
+ if (max_fds == INT32_MAX)
+ max_fds = 256;
+#endif
+
+ return max_fds;
+}
+
+
+/* Close all file descriptors starting with descriptor FIRST. If
+ EXCEPT is not NULL, it is expected to be a list of file descriptors
+ which shall not be closed. This list shall be sorted in ascending
+ order with the end marked by -1. */
+void
+close_all_fds (int first, int *except)
+{
+ int max_fd = get_max_fds ();
+ int fd, i, except_start;
+
+ if (except)
+ {
+ except_start = 0;
+ for (fd=first; fd < max_fd; fd++)
+ {
+ for (i=except_start; except[i] != -1; i++)
+ {
+ if (except[i] == fd)
+ {
+ /* If we found the descriptor in the exception list
+ we can start the next compare run at the next
+ index because the exception list is ordered. */
+ except_start = i + 1;
+ break;
+ }
+ }
+ if (except[i] == -1)
+ close (fd);
+ }
+ }
+ else
+ {
+ for (fd=first; fd < max_fd; fd++)
+ close (fd);
+ }
+
+ gpg_err_set_errno (0);
+}
+
+
+/* Returns an array with all currently open file descriptors. The end
+ of the array is marked by -1. The caller needs to release this
+ array using the *standard free* and not with xfree. This allow the
+ use of this function right at startup even before libgcrypt has
+ been initialized. Returns NULL on error and sets ERRNO
+ accordingly. */
+int *
+get_all_open_fds (void)
+{
+ int *array;
+ size_t narray;
+ int fd, max_fd, idx;
+#ifndef HAVE_STAT
+ array = calloc (1, sizeof *array);
+ if (array)
+ array[0] = -1;
+#else /*HAVE_STAT*/
+ struct stat statbuf;
+
+ max_fd = get_max_fds ();
+ narray = 32; /* If you change this change also t-exechelp.c. */
+ array = calloc (narray, sizeof *array);
+ if (!array)
+ return NULL;
+
+ /* Note: The list we return is ordered. */
+ for (idx=0, fd=0; fd < max_fd; fd++)
+ if (!(fstat (fd, &statbuf) == -1 && errno == EBADF))
+ {
+ if (idx+1 >= narray)
+ {
+ int *tmp;
+
+ narray += (narray < 256)? 32:256;
+ tmp = realloc (array, narray * sizeof *array);
+ if (!tmp)
+ {
+ free (array);
+ return NULL;
+ }
+ array = tmp;
+ }
+ array[idx++] = fd;
+ }
+ array[idx] = -1;
+#endif /*HAVE_STAT*/
+ return array;
+}
+
+
+/* The exec core used right after the fork. This will never return. */
+static void
+do_exec (const char *pgmname, const char *argv[],
+ int fd_in, int fd_out, int fd_err,
+ int *except, void (*preexec)(void) )
+{
+ char **arg_list;
+ int i, j;
+ int fds[3];
+
+ fds[0] = fd_in;
+ fds[1] = fd_out;
+ fds[2] = fd_err;
+
+ /* Create the command line argument array. */
+ i = 0;
+ if (argv)
+ while (argv[i])
+ i++;
+ arg_list = xcalloc (i+2, sizeof *arg_list);
+ arg_list[0] = strrchr (pgmname, '/');
+ if (arg_list[0])
+ arg_list[0]++;
+ else
+ arg_list[0] = xstrdup (pgmname);
+ if (argv)
+ for (i=0,j=1; argv[i]; i++, j++)
+ arg_list[j] = (char*)argv[i];
+
+ /* Assign /dev/null to unused FDs. */
+ for (i=0; i <= 2; i++)
+ {
+ if (fds[i] == -1 )
+ {
+ fds[i] = open ("/dev/null", i? O_WRONLY : O_RDONLY);
+ if (fds[i] == -1)
+ log_fatal ("failed to open '%s': %s\n",
+ "/dev/null", strerror (errno));
+ }
+ }
+
+ /* Connect the standard files. */
+ for (i=0; i <= 2; i++)
+ {
+ if (fds[i] != i && dup2 (fds[i], i) == -1)
+ log_fatal ("dup2 std%s failed: %s\n",
+ i==0?"in":i==1?"out":"err", strerror (errno));
+ }
+
+ /* Close all other files. */
+ close_all_fds (3, except);
+
+ if (preexec)
+ preexec ();
+ execv (pgmname, arg_list);
+ /* No way to print anything, as we have closed all streams. */
+ _exit (127);
+}
+
+
+static gpg_error_t
+do_create_pipe (int filedes[2])
+{
+ gpg_error_t err = 0;
+
+ if (pipe (filedes) == -1)
+ {
+ err = my_error_from_syserror ();
+ filedes[0] = filedes[1] = -1;
+ }
+
+ return err;
+}
+
+
+static gpg_error_t
+create_pipe_and_estream (int filedes[2], estream_t *r_fp,
+ int outbound, int nonblock)
+{
+ gpg_error_t err;
+
+ if (pipe (filedes) == -1)
+ {
+ err = my_error_from_syserror ();
+ log_error (_("error creating a pipe: %s\n"), gpg_strerror (err));
+ filedes[0] = filedes[1] = -1;
+ *r_fp = NULL;
+ return err;
+ }
+
+ if (!outbound)
+ *r_fp = es_fdopen (filedes[0], nonblock? "r,nonblock" : "r");
+ else
+ *r_fp = es_fdopen (filedes[1], nonblock? "w,nonblock" : "w");
+ if (!*r_fp)
+ {
+ err = my_error_from_syserror ();
+ log_error (_("error creating a stream for a pipe: %s\n"),
+ gpg_strerror (err));
+ close (filedes[0]);
+ close (filedes[1]);
+ filedes[0] = filedes[1] = -1;
+ return err;
+ }
+ return 0;
+}
+
+
+/* Portable function to create a pipe. Under Windows the write end is
+ inheritable. If R_FP is not NULL, an estream is created for the
+ read end and stored at R_FP. */
+gpg_error_t
+gnupg_create_inbound_pipe (int filedes[2], estream_t *r_fp, int nonblock)
+{
+ if (r_fp)
+ return create_pipe_and_estream (filedes, r_fp, 0, nonblock);
+ else
+ return do_create_pipe (filedes);
+}
+
+
+/* Portable function to create a pipe. Under Windows the read end is
+ inheritable. If R_FP is not NULL, an estream is created for the
+ write end and stored at R_FP. */
+gpg_error_t
+gnupg_create_outbound_pipe (int filedes[2], estream_t *r_fp, int nonblock)
+{
+ if (r_fp)
+ return create_pipe_and_estream (filedes, r_fp, 1, nonblock);
+ else
+ return do_create_pipe (filedes);
+}
+
+
+/* Portable function to create a pipe. Under Windows both ends are
+ inheritable. */
+gpg_error_t
+gnupg_create_pipe (int filedes[2])
+{
+ return do_create_pipe (filedes);
+}
+
+
+/* Fork and exec the PGMNAME, see exechelp.h for details. */
+gpg_error_t
+gnupg_spawn_process (const char *pgmname, const char *argv[],
+ int *except, void (*preexec)(void), unsigned int flags,
+ estream_t *r_infp,
+ estream_t *r_outfp,
+ estream_t *r_errfp,
+ pid_t *pid)
+{
+ gpg_error_t err;
+ int inpipe[2] = {-1, -1};
+ int outpipe[2] = {-1, -1};
+ int errpipe[2] = {-1, -1};
+ estream_t infp = NULL;
+ estream_t outfp = NULL;
+ estream_t errfp = NULL;
+ int nonblock = !!(flags & GNUPG_SPAWN_NONBLOCK);
+
+ if (r_infp)
+ *r_infp = NULL;
+ if (r_outfp)
+ *r_outfp = NULL;
+ if (r_errfp)
+ *r_errfp = NULL;
+ *pid = (pid_t)(-1); /* Always required. */
+
+ if (r_infp)
+ {
+ err = create_pipe_and_estream (inpipe, &infp, 1, nonblock);
+ if (err)
+ return err;
+ }
+
+ if (r_outfp)
+ {
+ err = create_pipe_and_estream (outpipe, &outfp, 0, nonblock);
+ if (err)
+ {
+ if (infp)
+ es_fclose (infp);
+ else if (inpipe[1] != -1)
+ close (inpipe[1]);
+ if (inpipe[0] != -1)
+ close (inpipe[0]);
+
+ return err;
+ }
+ }
+
+ if (r_errfp)
+ {
+ err = create_pipe_and_estream (errpipe, &errfp, 0, nonblock);
+ if (err)
+ {
+ if (infp)
+ es_fclose (infp);
+ else if (inpipe[1] != -1)
+ close (inpipe[1]);
+ if (inpipe[0] != -1)
+ close (inpipe[0]);
+
+ if (outfp)
+ es_fclose (outfp);
+ else if (outpipe[0] != -1)
+ close (outpipe[0]);
+ if (outpipe[1] != -1)
+ close (outpipe[1]);
+
+ return err;
+ }
+ }
+
+
+ *pid = fork ();
+ if (*pid == (pid_t)(-1))
+ {
+ err = my_error_from_syserror ();
+ log_error (_("error forking process: %s\n"), gpg_strerror (err));
+
+ if (infp)
+ es_fclose (infp);
+ else if (inpipe[1] != -1)
+ close (inpipe[1]);
+ if (inpipe[0] != -1)
+ close (inpipe[0]);
+
+ if (outfp)
+ es_fclose (outfp);
+ else if (outpipe[0] != -1)
+ close (outpipe[0]);
+ if (outpipe[1] != -1)
+ close (outpipe[1]);
+
+ if (errfp)
+ es_fclose (errfp);
+ else if (errpipe[0] != -1)
+ close (errpipe[0]);
+ if (errpipe[1] != -1)
+ close (errpipe[1]);
+ return err;
+ }
+
+ if (!*pid)
+ {
+ /* This is the child. */
+ gcry_control (GCRYCTL_TERM_SECMEM);
+ es_fclose (infp);
+ es_fclose (outfp);
+ es_fclose (errfp);
+ do_exec (pgmname, argv, inpipe[0], outpipe[1], errpipe[1],
+ except, preexec);
+ /*NOTREACHED*/
+ }
+
+ /* This is the parent. */
+ if (inpipe[0] != -1)
+ close (inpipe[0]);
+ if (outpipe[1] != -1)
+ close (outpipe[1]);
+ if (errpipe[1] != -1)
+ close (errpipe[1]);
+
+ if (r_infp)
+ *r_infp = infp;
+ if (r_outfp)
+ *r_outfp = outfp;
+ if (r_errfp)
+ *r_errfp = errfp;
+
+ return 0;
+}
+
+
+
+/* Simplified version of gnupg_spawn_process. This function forks and
+ then execs PGMNAME, while connecting INFD to stdin, OUTFD to stdout
+ and ERRFD to stderr (any of them may be -1 to connect them to
+ /dev/null). The arguments for the process are expected in the NULL
+ terminated array ARGV. The program name itself should not be
+ included there. Calling gnupg_wait_process is required.
+
+ Returns 0 on success or an error code. */
+gpg_error_t
+gnupg_spawn_process_fd (const char *pgmname, const char *argv[],
+ int infd, int outfd, int errfd, pid_t *pid)
+{
+ gpg_error_t err;
+
+ *pid = fork ();
+ if (*pid == (pid_t)(-1))
+ {
+ err = my_error_from_syserror ();
+ log_error (_("error forking process: %s\n"), strerror (errno));
+ return err;
+ }
+
+ if (!*pid)
+ {
+ gcry_control (GCRYCTL_TERM_SECMEM);
+ /* Run child. */
+ do_exec (pgmname, argv, infd, outfd, errfd, NULL, NULL);
+ /*NOTREACHED*/
+ }
+
+ return 0;
+}
+
+
+
+
+/* Waiting for child processes.
+
+ waitpid(2) may return information about terminated children that we
+ did not yet request, and there is no portable way to wait for a
+ specific set of children.
+
+ As a workaround, we store the results of children for later use.
+
+ XXX: This assumes that PIDs are not reused too quickly. */
+
+struct terminated_child
+{
+ pid_t pid;
+ int exitcode;
+ struct terminated_child *next;
+};
+
+struct terminated_child *terminated_children;
+
+
+static gpg_error_t
+store_result (pid_t pid, int exitcode)
+{
+ struct terminated_child *c;
+
+ c = xtrymalloc (sizeof *c);
+ if (c == NULL)
+ return gpg_err_code_from_syserror ();
+
+ c->pid = pid;
+ c->exitcode = exitcode;
+ c->next = terminated_children;
+ terminated_children = c;
+
+ return 0;
+}
+
+
+static int
+get_result (pid_t pid, int *r_exitcode)
+{
+ struct terminated_child *c, **prevp;
+
+ for (prevp = &terminated_children, c = terminated_children;
+ c;
+ prevp = &c->next, c = c->next)
+ if (c->pid == pid)
+ {
+ *prevp = c->next;
+ *r_exitcode = c->exitcode;
+ xfree (c);
+ return 1;
+ }
+
+ return 0;
+}
+
+
+/* See exechelp.h for a description. */
+gpg_error_t
+gnupg_wait_process (const char *pgmname, pid_t pid, int hang, int *r_exitcode)
+{
+ gpg_err_code_t ec;
+ int i, status;
+
+ if (r_exitcode)
+ *r_exitcode = -1;
+
+ if (pid == (pid_t)(-1))
+ return gpg_error (GPG_ERR_INV_VALUE);
+
+#ifdef USE_NPTH
+ i = npth_waitpid (pid, &status, hang? 0:WNOHANG);
+#else
+ while ((i=waitpid (pid, &status, hang? 0:WNOHANG)) == (pid_t)(-1)
+ && errno == EINTR);
+#endif
+
+ if (i == (pid_t)(-1))
+ {
+ ec = gpg_err_code_from_errno (errno);
+ log_error (_("waiting for process %d to terminate failed: %s\n"),
+ (int)pid, strerror (errno));
+ }
+ else if (!i)
+ {
+ ec = GPG_ERR_TIMEOUT; /* Still running. */
+ }
+ else if (WIFEXITED (status) && WEXITSTATUS (status) == 127)
+ {
+ log_error (_("error running '%s': probably not installed\n"), pgmname);
+ ec = GPG_ERR_CONFIGURATION;
+ }
+ else if (WIFEXITED (status) && WEXITSTATUS (status))
+ {
+ if (!r_exitcode)
+ log_error (_("error running '%s': exit status %d\n"), pgmname,
+ WEXITSTATUS (status));
+ else
+ *r_exitcode = WEXITSTATUS (status);
+ ec = GPG_ERR_GENERAL;
+ }
+ else if (!WIFEXITED (status))
+ {
+ log_error (_("error running '%s': terminated\n"), pgmname);
+ ec = GPG_ERR_GENERAL;
+ }
+ else
+ {
+ if (r_exitcode)
+ *r_exitcode = 0;
+ ec = 0;
+ }
+
+ return gpg_err_make (GPG_ERR_SOURCE_DEFAULT, ec);
+}
+
+/* See exechelp.h for a description. */
+gpg_error_t
+gnupg_wait_processes (const char **pgmnames, pid_t *pids, size_t count,
+ int hang, int *r_exitcodes)
+{
+ gpg_err_code_t ec = 0;
+ size_t i, left;
+ int *dummy = NULL;
+
+ if (r_exitcodes == NULL)
+ {
+ dummy = r_exitcodes = xtrymalloc (sizeof *r_exitcodes * count);
+ if (dummy == NULL)
+ return gpg_err_code_from_syserror ();
+ }
+
+ for (i = 0, left = count; i < count; i++)
+ {
+ int status = -1;
+
+ if (pids[i] == (pid_t)(-1))
+ return my_error (GPG_ERR_INV_VALUE);
+
+ /* See if there was a previously stored result for this pid. */
+ if (get_result (pids[i], &status))
+ left -= 1;
+
+ r_exitcodes[i] = status;
+ }
+
+ while (left > 0)
+ {
+ pid_t pid;
+ int status;
+
+#ifdef USE_NPTH
+ pid = npth_waitpid (-1, &status, hang ? 0 : WNOHANG);
+#else
+ while ((pid = waitpid (-1, &status, hang ? 0 : WNOHANG)) == (pid_t)(-1)
+ && errno == EINTR);
+#endif
+
+ if (pid == (pid_t)(-1))
+ {
+ ec = gpg_err_code_from_errno (errno);
+ log_error (_("waiting for processes to terminate failed: %s\n"),
+ strerror (errno));
+ break;
+ }
+ else if (!pid)
+ {
+ ec = GPG_ERR_TIMEOUT; /* Still running. */
+ break;
+ }
+ else
+ {
+ for (i = 0; i < count; i++)
+ if (pid == pids[i])
+ break;
+
+ if (i == count)
+ {
+ /* No match, store this result. */
+ ec = store_result (pid, status);
+ if (ec)
+ break;
+ continue;
+ }
+
+ /* Process PIDS[i] died. */
+ if (r_exitcodes[i] != (pid_t) -1)
+ {
+ log_error ("PID %d was reused", pid);
+ ec = GPG_ERR_GENERAL;
+ break;
+ }
+
+ left -= 1;
+ r_exitcodes[i] = status;
+ }
+ }
+
+ for (i = 0; i < count; i++)
+ {
+ if (r_exitcodes[i] == -1)
+ continue;
+
+ if (WIFEXITED (r_exitcodes[i]) && WEXITSTATUS (r_exitcodes[i]) == 127)
+ {
+ log_error (_("error running '%s': probably not installed\n"),
+ pgmnames[i]);
+ ec = GPG_ERR_CONFIGURATION;
+ }
+ else if (WIFEXITED (r_exitcodes[i]) && WEXITSTATUS (r_exitcodes[i]))
+ {
+ if (dummy)
+ log_error (_("error running '%s': exit status %d\n"),
+ pgmnames[i], WEXITSTATUS (r_exitcodes[i]));
+ else
+ r_exitcodes[i] = WEXITSTATUS (r_exitcodes[i]);
+ ec = GPG_ERR_GENERAL;
+ }
+ else if (!WIFEXITED (r_exitcodes[i]))
+ {
+ log_error (_("error running '%s': terminated\n"), pgmnames[i]);
+ ec = GPG_ERR_GENERAL;
+ }
+ }
+
+ xfree (dummy);
+ return gpg_err_make (GPG_ERR_SOURCE_DEFAULT, ec);
+}
+
+
+
+void
+gnupg_release_process (pid_t pid)
+{
+ (void)pid;
+}
+
+
+/* Spawn a new process and immediately detach from it. The name of
+ the program to exec is PGMNAME and its arguments are in ARGV (the
+ programname is automatically passed as first argument).
+ Environment strings in ENVP are set. An error is returned if
+ pgmname is not executable; to make this work it is necessary to
+ provide an absolute file name. All standard file descriptors are
+ connected to /dev/null. */
+gpg_error_t
+gnupg_spawn_process_detached (const char *pgmname, const char *argv[],
+ const char *envp[] )
+{
+ gpg_err_code_t ec;
+ pid_t pid;
+ int i;
+
+ if (getuid() != geteuid())
+ return my_error (GPG_ERR_BUG);
+
+ if ((ec = gnupg_access (pgmname, X_OK)))
+ return gpg_err_make (default_errsource, ec);
+
+ pid = fork ();
+ if (pid == (pid_t)(-1))
+ {
+ log_error (_("error forking process: %s\n"), strerror (errno));
+ return my_error_from_syserror ();
+ }
+ if (!pid)
+ {
+ pid_t pid2;
+
+ gcry_control (GCRYCTL_TERM_SECMEM);
+ if (setsid() == -1 || chdir ("/"))
+ _exit (1);
+
+ pid2 = fork (); /* Double fork to let init take over the new child. */
+ if (pid2 == (pid_t)(-1))
+ _exit (1);
+ if (pid2)
+ _exit (0); /* Let the parent exit immediately. */
+
+ if (envp)
+ for (i=0; envp[i]; i++)
+ putenv (xstrdup (envp[i]));
+
+ do_exec (pgmname, argv, -1, -1, -1, NULL, NULL);
+
+ /*NOTREACHED*/
+ }
+
+ if (waitpid (pid, NULL, 0) == -1)
+ log_error ("waitpid failed in gnupg_spawn_process_detached: %s",
+ strerror (errno));
+
+ return 0;
+}
+
+
+/* Kill a process; that is send an appropriate signal to the process.
+ gnupg_wait_process must be called to actually remove the process
+ from the system. An invalid PID is ignored. */
+void
+gnupg_kill_process (pid_t pid)
+{
+ if (pid != (pid_t)(-1))
+ {
+ kill (pid, SIGTERM);
+ }
+}