1 /* SPDX-License-Identifier: LGPL-2.1+ */
3 This file is part of systemd.
5 Copyright 2010 Lennart Poettering
7 systemd is free software; you can redistribute it and/or modify it
8 under the terms of the GNU Lesser General Public License as published by
9 the Free Software Foundation; either version 2.1 of the License, or
10 (at your option) any later version.
12 systemd is distributed in the hope that it will be useful, but
13 WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public License
18 along with systemd; If not, see <http://www.gnu.org/licenses/>.
30 //#include <sys/mman.h>
31 #include <sys/prctl.h>
32 #include <sys/statfs.h>
33 #include <sys/sysmacros.h>
34 //#include <sys/types.h>
37 #include "alloc-util.h"
38 //#include "btrfs-util.h"
40 #include "cgroup-util.h"
42 //#include "device-nodes.h"
43 #include "dirent-util.h"
46 //#include "format-util.h"
48 #include "hostname-util.h"
51 //#include "missing.h"
52 #include "parse-util.h"
53 //#include "path-util.h"
54 #include "process-util.h"
55 #include "procfs-util.h"
57 #include "signal-util.h"
58 #include "stat-util.h"
59 #include "string-util.h"
61 #include "time-util.h"
62 #include "umask-util.h"
63 #include "user-util.h"
68 char **saved_argv = NULL;
69 static int saved_in_initrd = -1;
71 size_t page_size(void) {
72 static thread_local size_t pgsz = 0;
75 if (_likely_(pgsz > 0))
78 r = sysconf(_SC_PAGESIZE);
85 #if 0 /// UNNEEDED by elogind
86 bool plymouth_running(void) {
87 return access("/run/plymouth/pid", F_OK) >= 0;
91 bool display_is_local(const char *display) {
100 int socket_from_display(const char *display, char **path) {
107 if (!display_is_local(display))
110 k = strspn(display+1, "0123456789");
112 f = new(char, STRLEN("/tmp/.X11-unix/X") + k + 1);
116 c = stpcpy(f, "/tmp/.X11-unix/X");
117 memcpy(c, display+1, k);
125 #if 0 /// UNNEEDED by elogind
126 bool kexec_loaded(void) {
127 _cleanup_free_ char *s = NULL;
129 if (read_one_line_file("/sys/kernel/kexec_loaded", &s) < 0)
135 int prot_from_flags(int flags) {
137 switch (flags & O_ACCMODE) {
146 return PROT_READ|PROT_WRITE;
154 bool in_initrd(void) {
157 if (saved_in_initrd >= 0)
158 return saved_in_initrd;
160 /* We make two checks here:
162 * 1. the flag file /etc/initrd-release must exist
163 * 2. the root file system must be a memory file system
165 * The second check is extra paranoia, since misdetecting an
166 * initrd can have bad consequences due the initrd
167 * emptying when transititioning to the main systemd.
170 saved_in_initrd = access("/etc/initrd-release", F_OK) >= 0 &&
171 statfs("/", &s) >= 0 &&
174 return saved_in_initrd;
177 #if 0 /// UNNEEDED by elogind
178 void in_initrd_force(bool value) {
179 saved_in_initrd = value;
182 /* hey glibc, APIs with callbacks without a user pointer are so useless */
183 void *xbsearch_r(const void *key, const void *base, size_t nmemb, size_t size,
184 int (*compar) (const void *, const void *, void *), void *arg) {
189 assert(!size_multiply_overflow(nmemb, size));
195 p = (const uint8_t*) base + idx * size;
196 comparison = compar(key, p, arg);
199 else if (comparison > 0)
208 int on_ac_power(void) {
209 bool found_offline = false, found_online = false;
210 _cleanup_closedir_ DIR *d = NULL;
213 d = opendir("/sys/class/power_supply");
215 return errno == ENOENT ? true : -errno;
217 FOREACH_DIRENT(de, d, return -errno) {
218 _cleanup_close_ int fd = -1, device = -1;
222 device = openat(dirfd(d), de->d_name, O_DIRECTORY|O_RDONLY|O_CLOEXEC|O_NOCTTY);
224 if (IN_SET(errno, ENOENT, ENOTDIR))
230 fd = openat(device, "type", O_RDONLY|O_CLOEXEC|O_NOCTTY);
238 n = read(fd, contents, sizeof(contents));
242 if (n != 6 || memcmp(contents, "Mains\n", 6))
246 fd = openat(device, "online", O_RDONLY|O_CLOEXEC|O_NOCTTY);
254 n = read(fd, contents, sizeof(contents));
258 if (n != 2 || contents[1] != '\n')
261 if (contents[0] == '1') {
264 } else if (contents[0] == '0')
265 found_offline = true;
270 return found_online || !found_offline;
273 int container_get_leader(const char *machine, pid_t *pid) {
274 _cleanup_free_ char *s = NULL, *class = NULL;
282 if (!machine_name_is_valid(machine))
285 p = strjoina("/run/systemd/machines/", machine);
286 r = parse_env_file(p, NEWLINE, "LEADER", &s, "CLASS", &class, NULL);
294 if (!streq_ptr(class, "container"))
297 r = parse_pid(s, &leader);
307 int namespace_open(pid_t pid, int *pidns_fd, int *mntns_fd, int *netns_fd, int *userns_fd, int *root_fd) {
308 _cleanup_close_ int pidnsfd = -1, mntnsfd = -1, netnsfd = -1, usernsfd = -1;
316 mntns = procfs_file_alloca(pid, "ns/mnt");
317 mntnsfd = open(mntns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
325 pidns = procfs_file_alloca(pid, "ns/pid");
326 pidnsfd = open(pidns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
334 netns = procfs_file_alloca(pid, "ns/net");
335 netnsfd = open(netns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
343 userns = procfs_file_alloca(pid, "ns/user");
344 usernsfd = open(userns, O_RDONLY|O_NOCTTY|O_CLOEXEC);
345 if (usernsfd < 0 && errno != ENOENT)
352 root = procfs_file_alloca(pid, "root");
353 rfd = open(root, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
368 *userns_fd = usernsfd;
373 pidnsfd = mntnsfd = netnsfd = usernsfd = -1;
378 int namespace_enter(int pidns_fd, int mntns_fd, int netns_fd, int userns_fd, int root_fd) {
379 if (userns_fd >= 0) {
380 /* Can't setns to your own userns, since then you could
381 * escalate from non-root to root in your own namespace, so
382 * check if namespaces equal before attempting to enter. */
383 _cleanup_free_ char *userns_fd_path = NULL;
385 if (asprintf(&userns_fd_path, "/proc/self/fd/%d", userns_fd) < 0)
388 r = files_same(userns_fd_path, "/proc/self/ns/user", 0);
396 if (setns(pidns_fd, CLONE_NEWPID) < 0)
400 if (setns(mntns_fd, CLONE_NEWNS) < 0)
404 if (setns(netns_fd, CLONE_NEWNET) < 0)
408 if (setns(userns_fd, CLONE_NEWUSER) < 0)
412 if (fchdir(root_fd) < 0)
419 return reset_uid_gid();
422 uint64_t physical_memory(void) {
423 _cleanup_free_ char *root = NULL, *value = NULL;
428 /* We return this as uint64_t in case we are running as 32bit process on a 64bit kernel with huge amounts of
431 * In order to support containers nicely that have a configured memory limit we'll take the minimum of the
432 * physically reported amount of memory and the limit configured for the root cgroup, if there is any. */
434 sc = sysconf(_SC_PHYS_PAGES);
438 mem = (uint64_t) sc * (uint64_t) ps;
440 if (cg_get_root_path(&root) < 0)
443 if (cg_get_attribute("memory", root, "memory.limit_in_bytes", &value))
446 if (safe_atou64(value, &lim) < 0)
449 /* Make sure the limit is a multiple of our own page size */
453 return MIN(mem, lim);
456 uint64_t physical_memory_scale(uint64_t v, uint64_t max) {
457 uint64_t p, m, ps, r;
461 /* Returns the physical memory size, multiplied by v divided by max. Returns UINT64_MAX on overflow. On success
462 * the result is a multiple of the page size (rounds down). */
467 p = physical_memory() / ps;
483 uint64_t system_tasks_max(void) {
485 uint64_t a = TASKS_MAX, b = TASKS_MAX;
486 _cleanup_free_ char *root = NULL;
488 /* Determine the maximum number of tasks that may run on this system. We check three sources to determine this
491 * a) the maximum tasks value the kernel allows on this architecture
492 * b) the cgroups pids_max attribute for the system
493 * c) the kernel's configured maximum PID value
495 * And then pick the smallest of the three */
497 (void) procfs_tasks_get_limit(&a);
499 if (cg_get_root_path(&root) >= 0) {
500 _cleanup_free_ char *value = NULL;
502 if (cg_get_attribute("pids", root, "pids.max", &value) >= 0)
503 (void) safe_atou64(value, &b);
506 return MIN3(TASKS_MAX,
507 a <= 0 ? TASKS_MAX : a,
508 b <= 0 ? TASKS_MAX : b);
511 uint64_t system_tasks_max_scale(uint64_t v, uint64_t max) {
516 /* Multiply the system's task value by the fraction v/max. Hence, if max==100 this calculates percentages
517 * relative to the system's maximum number of tasks. Returns UINT64_MAX on overflow. */
519 t = system_tasks_max();
523 if (m / t != v) /* overflow? */
530 puts(PACKAGE_STRING "\n"
535 #if 0 /// UNNEEDED by elogind
536 /* This is a direct translation of str_verscmp from boot.c */
537 static bool is_digit(int c) {
538 return c >= '0' && c <= '9';
541 static int c_order(int c) {
542 if (c == 0 || is_digit(c))
545 if ((c >= 'a') && (c <= 'z'))
551 int str_verscmp(const char *s1, const char *s2) {
552 const char *os1, *os2;
563 while ((*s1 && !is_digit(*s1)) || (*s2 && !is_digit(*s2))) {
566 order = c_order(*s1) - c_order(*s2);
579 while (is_digit(*s1) && is_digit(*s2)) {
595 return strcmp(os1, os2);
598 /* Turn off core dumps but only if we're running outside of a container. */
599 void disable_coredumps(void) {
602 if (detect_container() > 0)
605 r = write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
607 log_debug_errno(r, "Failed to turn off coredumps, ignoring: %m");