2 This file is part of systemd.
4 Copyright 2010 Lennart Poettering
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
23 #include <linux/oom.h>
30 #include <sys/personality.h>
31 #include <sys/prctl.h>
32 #include <sys/types.h>
36 #ifdef HAVE_VALGRIND_VALGRIND_H
37 #include <valgrind/valgrind.h>
40 #include "alloc-util.h"
41 //#include "architecture.h"
50 #include "process-util.h"
51 #include "signal-util.h"
52 //#include "stat-util.h"
53 #include "string-table.h"
54 #include "string-util.h"
55 #include "user-util.h"
58 int get_process_state(pid_t pid) {
62 _cleanup_free_ char *line = NULL;
66 p = procfs_file_alloca(pid, "stat");
68 r = read_one_line_file(p, &line);
74 p = strrchr(line, ')');
80 if (sscanf(p, " %c", &state) != 1)
83 return (unsigned char) state;
86 int get_process_comm(pid_t pid, char **name) {
93 p = procfs_file_alloca(pid, "comm");
95 r = read_one_line_file(p, name);
102 int get_process_cmdline(pid_t pid, size_t max_length, bool comm_fallback, char **line) {
103 _cleanup_fclose_ FILE *f = NULL;
105 char *k, *ans = NULL;
112 /* Retrieves a process' command line. Replaces unprintable characters while doing so by whitespace (coalescing
113 * multiple sequential ones into one). If max_length is != 0 will return a string of the specified size at most
114 * (the trailing NUL byte does count towards the length here!), abbreviated with a "..." ellipsis. If
115 * comm_fallback is true and the process has no command line set (the case for kernel threads), or has a
116 * command line that resolves to the empty string will return the "comm" name of the process instead.
118 * Returns -ESRCH if the process doesn't exist, and -ENOENT if the process has no command line (and
119 * comm_fallback is false). Returns 0 and sets *line otherwise. */
121 p = procfs_file_alloca(pid, "cmdline");
130 if (max_length == 1) {
132 /* If there's only room for one byte, return the empty string */
140 } else if (max_length == 0) {
141 size_t len = 0, allocated = 0;
143 while ((c = getc(f)) != EOF) {
145 if (!GREEDY_REALLOC(ans, allocated, len+3)) {
167 bool dotdotdot = false;
170 ans = new(char, max_length);
176 while ((c = getc(f)) != EOF) {
203 if (max_length <= 4) {
207 k = ans + max_length - 4;
210 /* Eat up final spaces */
211 while (k > ans && isspace(k[-1])) {
217 strncpy(k, "...", left-1);
223 /* Kernel threads have no argv[] */
225 _cleanup_free_ char *t = NULL;
233 h = get_process_comm(pid, &t);
238 ans = strjoin("[", t, "]");
244 if (l + 3 <= max_length)
245 ans = strjoin("[", t, "]");
246 else if (max_length <= 6) {
248 ans = new(char, max_length);
252 memcpy(ans, "[...]", max_length-1);
253 ans[max_length-1] = 0;
257 t[max_length - 6] = 0;
259 /* Chop off final spaces */
261 while (e > t && isspace(e[-1]))
265 ans = strjoin("[", t, "...]");
276 #if 0 /// UNNEEDED by elogind
277 int rename_process(const char name[]) {
278 static size_t mm_size = 0;
279 static char *mm = NULL;
280 bool truncated = false;
283 /* This is a like a poor man's setproctitle(). It changes the comm field, argv[0], and also the glibc's
284 * internally used name of the process. For the first one a limit of 16 chars applies; to the second one in
285 * many cases one of 10 (i.e. length of "/sbin/init") — however if we have CAP_SYS_RESOURCES it is unbounded;
286 * to the third one 7 (i.e. the length of "systemd". If you pass a longer string it will likely be
289 * Returns 0 if a name was set but truncated, > 0 if it was set but not truncated. */
292 return -EINVAL; /* let's not confuse users unnecessarily with an empty name */
296 /* First step, change the comm field. */
297 (void) prctl(PR_SET_NAME, name);
298 if (l > 15) /* Linux process names can be 15 chars at max */
301 /* Second step, change glibc's ID of the process name. */
302 if (program_invocation_name) {
305 k = strlen(program_invocation_name);
306 strncpy(program_invocation_name, name, k);
311 /* Third step, completely replace the argv[] array the kernel maintains for us. This requires privileges, but
312 * has the advantage that the argv[] array is exactly what we want it to be, and not filled up with zeros at
313 * the end. This is the best option for changing /proc/self/cmdline. */
318 /* Let's not bother with this if we don't have euid == 0. Strictly speaking if people do weird stuff
319 * with capabilities this could work even for euid != 0, but our own code generally doesn't do that,
320 * hence let's use this as quick bypass check, to avoid calling mmap() if PR_SET_MM_ARG_START fails
321 * with EPERM later on anyway. After all geteuid() is dead cheap to call, but mmap() is not. */
322 if (geteuid() != 0) {
323 log_debug("Skipping PR_SET_MM_ARG_START, as we don't have privileges.");
327 nn_size = PAGE_ALIGN(l+1);
328 nn = mmap(NULL, nn_size, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
329 if (nn == MAP_FAILED) {
330 log_debug_errno(errno, "mmap() failed: %m");
334 strncpy(nn, name, nn_size);
336 /* Now, let's tell the kernel about this new memory */
337 if (prctl(PR_SET_MM, PR_SET_MM_ARG_START, (unsigned long) nn, 0, 0) < 0) {
338 log_debug_errno(errno, "PR_SET_MM_ARG_START failed, proceeding without: %m");
339 (void) munmap(nn, nn_size);
343 /* And update the end pointer to the new end, too. If this fails, we don't really know what to do, it's
344 * pretty unlikely that we can rollback, hence we'll just accept the failure, and continue. */
345 if (prctl(PR_SET_MM, PR_SET_MM_ARG_END, (unsigned long) nn + l + 1, 0, 0) < 0)
346 log_debug_errno(errno, "PR_SET_MM_ARG_END failed, proceeding without: %m");
349 (void) munmap(mm, mm_size);
354 strncpy(mm, name, mm_size);
357 /* Fourth step: in all cases we'll also update the original argv[], so that our own code gets it right too if
358 * it still looks here */
360 if (saved_argc > 0) {
366 k = strlen(saved_argv[0]);
367 strncpy(saved_argv[0], name, k);
372 for (i = 1; i < saved_argc; i++) {
376 memzero(saved_argv[i], strlen(saved_argv[i]));
384 int is_kernel_thread(pid_t pid) {
391 if (pid == 0 || pid == 1) /* pid 1, and we ourselves certainly aren't a kernel thread */
396 p = procfs_file_alloca(pid, "cmdline");
404 count = fread(&c, 1, 1, f);
408 /* Kernel threads have an empty cmdline */
411 return eof ? 1 : -errno;
416 #if 0 /// UNNEEDED by elogind
417 int get_process_capeff(pid_t pid, char **capeff) {
424 p = procfs_file_alloca(pid, "status");
426 r = get_proc_field(p, "CapEff", WHITESPACE, capeff);
434 static int get_process_link_contents(const char *proc_file, char **name) {
440 r = readlink_malloc(proc_file, name);
449 int get_process_exe(pid_t pid, char **name) {
456 p = procfs_file_alloca(pid, "exe");
457 r = get_process_link_contents(p, name);
461 d = endswith(*name, " (deleted)");
468 #if 0 /// UNNEEDED by elogind
469 static int get_process_id(pid_t pid, const char *field, uid_t *uid) {
470 _cleanup_fclose_ FILE *f = NULL;
477 p = procfs_file_alloca(pid, "status");
485 FOREACH_LINE(line, f, return -errno) {
490 if (startswith(l, field)) {
492 l += strspn(l, WHITESPACE);
494 l[strcspn(l, WHITESPACE)] = 0;
496 return parse_uid(l, uid);
503 int get_process_uid(pid_t pid, uid_t *uid) {
504 return get_process_id(pid, "Uid:", uid);
507 int get_process_gid(pid_t pid, gid_t *gid) {
508 assert_cc(sizeof(uid_t) == sizeof(gid_t));
509 return get_process_id(pid, "Gid:", gid);
512 int get_process_cwd(pid_t pid, char **cwd) {
517 p = procfs_file_alloca(pid, "cwd");
519 return get_process_link_contents(p, cwd);
522 int get_process_root(pid_t pid, char **root) {
527 p = procfs_file_alloca(pid, "root");
529 return get_process_link_contents(p, root);
532 int get_process_environ(pid_t pid, char **env) {
533 _cleanup_fclose_ FILE *f = NULL;
534 _cleanup_free_ char *outcome = NULL;
537 size_t allocated = 0, sz = 0;
542 p = procfs_file_alloca(pid, "environ");
551 while ((c = fgetc(f)) != EOF) {
552 if (!GREEDY_REALLOC(outcome, allocated, sz + 5))
556 outcome[sz++] = '\n';
558 sz += cescape_char(c, outcome + sz);
562 outcome = strdup("");
574 int get_process_ppid(pid_t pid, pid_t *_ppid) {
576 _cleanup_free_ char *line = NULL;
588 p = procfs_file_alloca(pid, "stat");
589 r = read_one_line_file(p, &line);
595 /* Let's skip the pid and comm fields. The latter is enclosed
596 * in () but does not escape any () in its value, so let's
597 * skip over it manually */
599 p = strrchr(line, ')');
611 if ((long unsigned) (pid_t) ppid != ppid)
614 *_ppid = (pid_t) ppid;
620 int wait_for_terminate(pid_t pid, siginfo_t *status) {
631 if (waitid(P_PID, pid, status, WEXITED) < 0) {
636 return negative_errno();
645 * < 0 : wait_for_terminate() failed to get the state of the
646 * process, the process was terminated by a signal, or
647 * failed for an unknown reason.
648 * >=0 : The process terminated normally, and its exit code is
651 * That is, success is indicated by a return value of zero, and an
652 * error is indicated by a non-zero value.
654 * A warning is emitted if the process terminates abnormally,
655 * and also if it returns non-zero unless check_exit_code is true.
657 int wait_for_terminate_and_warn(const char *name, pid_t pid, bool check_exit_code) {
664 r = wait_for_terminate(pid, &status);
666 return log_warning_errno(r, "Failed to wait for %s: %m", name);
668 if (status.si_code == CLD_EXITED) {
669 if (status.si_status != 0)
670 log_full(check_exit_code ? LOG_WARNING : LOG_DEBUG,
671 "%s failed with error code %i.", name, status.si_status);
673 log_debug("%s succeeded.", name);
675 return status.si_status;
676 } else if (status.si_code == CLD_KILLED ||
677 status.si_code == CLD_DUMPED) {
679 log_warning("%s terminated by signal %s.", name, signal_to_string(status.si_status));
683 log_warning("%s failed due to unknown reason.", name);
687 #if 0 /// UNNEEDED by elogind
688 void sigkill_wait(pid_t pid) {
691 if (kill(pid, SIGKILL) > 0)
692 (void) wait_for_terminate(pid, NULL);
695 void sigkill_waitp(pid_t *pid) {
704 int kill_and_sigcont(pid_t pid, int sig) {
707 r = kill(pid, sig) < 0 ? -errno : 0;
709 /* If this worked, also send SIGCONT, unless we already just sent a SIGCONT, or SIGKILL was sent which isn't
710 * affected by a process being suspended anyway. */
711 if (r >= 0 && !IN_SET(sig, SIGCONT, SIGKILL))
712 (void) kill(pid, SIGCONT);
718 int getenv_for_pid(pid_t pid, const char *field, char **_value) {
719 _cleanup_fclose_ FILE *f = NULL;
730 path = procfs_file_alloca(pid, "environ");
732 f = fopen(path, "re");
746 for (i = 0; i < sizeof(line)-1; i++) {
750 if (_unlikely_(c == EOF)) {
760 if (strneq(line, field, l) && line[l] == '=') {
761 value = strdup(line + l + 1);
775 bool pid_is_unwaited(pid_t pid) {
776 /* Checks whether a PID is still valid at all, including a zombie */
781 if (pid <= 1) /* If we or PID 1 would be dead and have been waited for, this code would not be running */
784 if (kill(pid, 0) >= 0)
787 return errno != ESRCH;
790 bool pid_is_alive(pid_t pid) {
793 /* Checks whether a PID is still valid and not a zombie */
798 if (pid <= 1) /* If we or PID 1 would be a zombie, this code would not be running */
801 r = get_process_state(pid);
802 if (r == -ESRCH || r == 'Z')
808 #if 0 /// UNNEEDED by elogind
809 int pid_from_same_root_fs(pid_t pid) {
815 root = procfs_file_alloca(pid, "root");
817 return files_same(root, "/proc/1/root", 0);
821 bool is_main_thread(void) {
822 static thread_local int cached = 0;
824 if (_unlikely_(cached == 0))
825 cached = getpid() == gettid() ? 1 : -1;
830 #if 0 /// UNNEEDED by elogind
831 noreturn void freeze(void) {
835 /* Make sure nobody waits for us on a socket anymore */
836 close_all_fds(NULL, 0);
844 bool oom_score_adjust_is_valid(int oa) {
845 return oa >= OOM_SCORE_ADJ_MIN && oa <= OOM_SCORE_ADJ_MAX;
848 unsigned long personality_from_string(const char *p) {
852 return PERSONALITY_INVALID;
854 /* Parse a personality specifier. We use our own identifiers that indicate specific ABIs, rather than just
855 * hints regarding the register size, since we want to keep things open for multiple locally supported ABIs for
856 * the same register size. */
858 architecture = architecture_from_string(p);
859 if (architecture < 0)
860 return PERSONALITY_INVALID;
862 if (architecture == native_architecture())
864 #ifdef SECONDARY_ARCHITECTURE
865 if (architecture == SECONDARY_ARCHITECTURE)
869 return PERSONALITY_INVALID;
872 const char* personality_to_string(unsigned long p) {
873 int architecture = _ARCHITECTURE_INVALID;
876 architecture = native_architecture();
877 #ifdef SECONDARY_ARCHITECTURE
878 else if (p == PER_LINUX32)
879 architecture = SECONDARY_ARCHITECTURE;
882 if (architecture < 0)
885 return architecture_to_string(architecture);
888 void valgrind_summary_hack(void) {
889 #ifdef HAVE_VALGRIND_VALGRIND_H
890 if (getpid() == 1 && RUNNING_ON_VALGRIND) {
892 pid = raw_clone(SIGCHLD);
894 log_emergency_errno(errno, "Failed to fork off valgrind helper: %m");
898 log_info("Spawned valgrind helper as PID "PID_FMT".", pid);
899 (void) wait_for_terminate(pid, NULL);
905 int pid_compare_func(const void *a, const void *b) {
906 const pid_t *p = a, *q = b;
908 /* Suitable for usage in qsort() */
917 int ioprio_parse_priority(const char *s, int *ret) {
923 r = safe_atoi(s, &i);
927 if (!ioprio_priority_is_valid(i))
934 static const char *const ioprio_class_table[] = {
935 [IOPRIO_CLASS_NONE] = "none",
936 [IOPRIO_CLASS_RT] = "realtime",
937 [IOPRIO_CLASS_BE] = "best-effort",
938 [IOPRIO_CLASS_IDLE] = "idle"
941 DEFINE_STRING_TABLE_LOOKUP_WITH_FALLBACK(ioprio_class, int, INT_MAX);
943 static const char *const sigchld_code_table[] = {
944 [CLD_EXITED] = "exited",
945 [CLD_KILLED] = "killed",
946 [CLD_DUMPED] = "dumped",
947 [CLD_TRAPPED] = "trapped",
948 [CLD_STOPPED] = "stopped",
949 [CLD_CONTINUED] = "continued",
952 DEFINE_STRING_TABLE_LOOKUP(sigchld_code, int);
954 static const char* const sched_policy_table[] = {
955 [SCHED_OTHER] = "other",
956 [SCHED_BATCH] = "batch",
957 [SCHED_IDLE] = "idle",
958 [SCHED_FIFO] = "fifo",
962 DEFINE_STRING_TABLE_LOOKUP_WITH_FALLBACK(sched_policy, int, INT_MAX);