4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/export.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/reboot.h>
12 #include <linux/prctl.h>
13 #include <linux/highuid.h>
15 #include <linux/kmod.h>
16 #include <linux/perf_event.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/kexec.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/binfmts.h>
52 #include <linux/sched.h>
53 #include <linux/rcupdate.h>
54 #include <linux/uidgid.h>
55 #include <linux/cred.h>
57 #include <linux/kmsg_dump.h>
58 /* Move somewhere else to avoid recompiling? */
59 #include <generated/utsrelease.h>
61 #include <asm/uaccess.h>
63 #include <asm/unistd.h>
65 #ifndef SET_UNALIGN_CTL
66 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
68 #ifndef GET_UNALIGN_CTL
69 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
72 # define SET_FPEMU_CTL(a,b) (-EINVAL)
75 # define GET_FPEMU_CTL(a,b) (-EINVAL)
78 # define SET_FPEXC_CTL(a,b) (-EINVAL)
81 # define GET_FPEXC_CTL(a,b) (-EINVAL)
84 # define GET_ENDIAN(a,b) (-EINVAL)
87 # define SET_ENDIAN(a,b) (-EINVAL)
90 # define GET_TSC_CTL(a) (-EINVAL)
93 # define SET_TSC_CTL(a) (-EINVAL)
97 * this is where the system-wide overflow UID and GID are defined, for
98 * architectures that now have 32-bit UID/GID but didn't in the past
101 int overflowuid = DEFAULT_OVERFLOWUID;
102 int overflowgid = DEFAULT_OVERFLOWGID;
104 EXPORT_SYMBOL(overflowuid);
105 EXPORT_SYMBOL(overflowgid);
108 * the same as above, but for filesystems which can only store a 16-bit
109 * UID and GID. as such, this is needed on all architectures
112 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
113 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
115 EXPORT_SYMBOL(fs_overflowuid);
116 EXPORT_SYMBOL(fs_overflowgid);
119 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
124 EXPORT_SYMBOL(cad_pid);
127 * If set, this is used for preparing the system to power off.
130 void (*pm_power_off_prepare)(void);
133 * Returns true if current's euid is same as p's uid or euid,
134 * or has CAP_SYS_NICE to p's user_ns.
136 * Called with rcu_read_lock, creds are safe
138 static bool set_one_prio_perm(struct task_struct *p)
140 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
142 if (uid_eq(pcred->uid, cred->euid) ||
143 uid_eq(pcred->euid, cred->euid))
145 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
151 * set the priority of a task
152 * - the caller must hold the RCU read lock
154 static int set_one_prio(struct task_struct *p, int niceval, int error)
158 if (!set_one_prio_perm(p)) {
162 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
166 no_nice = security_task_setnice(p, niceval);
173 set_user_nice(p, niceval);
178 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
180 struct task_struct *g, *p;
181 struct user_struct *user;
182 const struct cred *cred = current_cred();
187 if (which > PRIO_USER || which < PRIO_PROCESS)
190 /* normalize: avoid signed division (rounding problems) */
198 read_lock(&tasklist_lock);
202 p = find_task_by_vpid(who);
206 error = set_one_prio(p, niceval, error);
210 pgrp = find_vpid(who);
212 pgrp = task_pgrp(current);
213 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
214 error = set_one_prio(p, niceval, error);
215 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
218 uid = make_kuid(cred->user_ns, who);
222 else if (!uid_eq(uid, cred->uid) &&
223 !(user = find_user(uid)))
224 goto out_unlock; /* No processes for this user */
226 do_each_thread(g, p) {
227 if (uid_eq(task_uid(p), uid))
228 error = set_one_prio(p, niceval, error);
229 } while_each_thread(g, p);
230 if (!uid_eq(uid, cred->uid))
231 free_uid(user); /* For find_user() */
235 read_unlock(&tasklist_lock);
242 * Ugh. To avoid negative return values, "getpriority()" will
243 * not return the normal nice-value, but a negated value that
244 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
245 * to stay compatible.
247 SYSCALL_DEFINE2(getpriority, int, which, int, who)
249 struct task_struct *g, *p;
250 struct user_struct *user;
251 const struct cred *cred = current_cred();
252 long niceval, retval = -ESRCH;
256 if (which > PRIO_USER || which < PRIO_PROCESS)
260 read_lock(&tasklist_lock);
264 p = find_task_by_vpid(who);
268 niceval = 20 - task_nice(p);
269 if (niceval > retval)
275 pgrp = find_vpid(who);
277 pgrp = task_pgrp(current);
278 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
279 niceval = 20 - task_nice(p);
280 if (niceval > retval)
282 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
285 uid = make_kuid(cred->user_ns, who);
289 else if (!uid_eq(uid, cred->uid) &&
290 !(user = find_user(uid)))
291 goto out_unlock; /* No processes for this user */
293 do_each_thread(g, p) {
294 if (uid_eq(task_uid(p), uid)) {
295 niceval = 20 - task_nice(p);
296 if (niceval > retval)
299 } while_each_thread(g, p);
300 if (!uid_eq(uid, cred->uid))
301 free_uid(user); /* for find_user() */
305 read_unlock(&tasklist_lock);
312 * emergency_restart - reboot the system
314 * Without shutting down any hardware or taking any locks
315 * reboot the system. This is called when we know we are in
316 * trouble so this is our best effort to reboot. This is
317 * safe to call in interrupt context.
319 void emergency_restart(void)
321 kmsg_dump(KMSG_DUMP_EMERG);
322 machine_emergency_restart();
324 EXPORT_SYMBOL_GPL(emergency_restart);
326 void kernel_restart_prepare(char *cmd)
328 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
329 system_state = SYSTEM_RESTART;
330 usermodehelper_disable();
335 * register_reboot_notifier - Register function to be called at reboot time
336 * @nb: Info about notifier function to be called
338 * Registers a function with the list of functions
339 * to be called at reboot time.
341 * Currently always returns zero, as blocking_notifier_chain_register()
342 * always returns zero.
344 int register_reboot_notifier(struct notifier_block *nb)
346 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
348 EXPORT_SYMBOL(register_reboot_notifier);
351 * unregister_reboot_notifier - Unregister previously registered reboot notifier
352 * @nb: Hook to be unregistered
354 * Unregisters a previously registered reboot
357 * Returns zero on success, or %-ENOENT on failure.
359 int unregister_reboot_notifier(struct notifier_block *nb)
361 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
363 EXPORT_SYMBOL(unregister_reboot_notifier);
365 static void migrate_to_reboot_cpu(void)
367 /* The boot cpu is always logical cpu 0 */
370 cpu_hotplug_disable();
372 /* Make certain the cpu I'm about to reboot on is online */
373 if (!cpu_online(cpu))
374 cpu = cpumask_first(cpu_online_mask);
376 /* Prevent races with other tasks migrating this task */
377 current->flags |= PF_NO_SETAFFINITY;
379 /* Make certain I only run on the appropriate processor */
380 set_cpus_allowed_ptr(current, cpumask_of(cpu));
384 * kernel_restart - reboot the system
385 * @cmd: pointer to buffer containing command to execute for restart
388 * Shutdown everything and perform a clean reboot.
389 * This is not safe to call in interrupt context.
391 void kernel_restart(char *cmd)
393 kernel_restart_prepare(cmd);
394 migrate_to_reboot_cpu();
397 printk(KERN_EMERG "Restarting system.\n");
399 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
400 kmsg_dump(KMSG_DUMP_RESTART);
401 machine_restart(cmd);
403 EXPORT_SYMBOL_GPL(kernel_restart);
405 static void kernel_shutdown_prepare(enum system_states state)
407 blocking_notifier_call_chain(&reboot_notifier_list,
408 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
409 system_state = state;
410 usermodehelper_disable();
414 * kernel_halt - halt the system
416 * Shutdown everything and perform a clean system halt.
418 void kernel_halt(void)
420 kernel_shutdown_prepare(SYSTEM_HALT);
421 migrate_to_reboot_cpu();
423 printk(KERN_EMERG "System halted.\n");
424 kmsg_dump(KMSG_DUMP_HALT);
428 EXPORT_SYMBOL_GPL(kernel_halt);
431 * kernel_power_off - power_off the system
433 * Shutdown everything and perform a clean system power_off.
435 void kernel_power_off(void)
437 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
438 if (pm_power_off_prepare)
439 pm_power_off_prepare();
440 migrate_to_reboot_cpu();
442 printk(KERN_EMERG "Power down.\n");
443 kmsg_dump(KMSG_DUMP_POWEROFF);
446 EXPORT_SYMBOL_GPL(kernel_power_off);
448 static DEFINE_MUTEX(reboot_mutex);
451 * Reboot system call: for obvious reasons only root may call it,
452 * and even root needs to set up some magic numbers in the registers
453 * so that some mistake won't make this reboot the whole machine.
454 * You can also set the meaning of the ctrl-alt-del-key here.
456 * reboot doesn't sync: do that yourself before calling this.
458 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
461 struct pid_namespace *pid_ns = task_active_pid_ns(current);
465 /* We only trust the superuser with rebooting the system. */
466 if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
469 /* For safety, we require "magic" arguments. */
470 if (magic1 != LINUX_REBOOT_MAGIC1 ||
471 (magic2 != LINUX_REBOOT_MAGIC2 &&
472 magic2 != LINUX_REBOOT_MAGIC2A &&
473 magic2 != LINUX_REBOOT_MAGIC2B &&
474 magic2 != LINUX_REBOOT_MAGIC2C))
478 * If pid namespaces are enabled and the current task is in a child
479 * pid_namespace, the command is handled by reboot_pid_ns() which will
482 ret = reboot_pid_ns(pid_ns, cmd);
486 /* Instead of trying to make the power_off code look like
487 * halt when pm_power_off is not set do it the easy way.
489 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
490 cmd = LINUX_REBOOT_CMD_HALT;
492 mutex_lock(&reboot_mutex);
494 case LINUX_REBOOT_CMD_RESTART:
495 kernel_restart(NULL);
498 case LINUX_REBOOT_CMD_CAD_ON:
502 case LINUX_REBOOT_CMD_CAD_OFF:
506 case LINUX_REBOOT_CMD_HALT:
509 panic("cannot halt.\n");
511 case LINUX_REBOOT_CMD_POWER_OFF:
516 case LINUX_REBOOT_CMD_RESTART2:
517 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
521 buffer[sizeof(buffer) - 1] = '\0';
523 kernel_restart(buffer);
527 case LINUX_REBOOT_CMD_KEXEC:
528 ret = kernel_kexec();
532 #ifdef CONFIG_HIBERNATION
533 case LINUX_REBOOT_CMD_SW_SUSPEND:
542 mutex_unlock(&reboot_mutex);
546 static void deferred_cad(struct work_struct *dummy)
548 kernel_restart(NULL);
552 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
553 * As it's called within an interrupt, it may NOT sync: the only choice
554 * is whether to reboot at once, or just ignore the ctrl-alt-del.
556 void ctrl_alt_del(void)
558 static DECLARE_WORK(cad_work, deferred_cad);
561 schedule_work(&cad_work);
563 kill_cad_pid(SIGINT, 1);
567 * Unprivileged users may change the real gid to the effective gid
568 * or vice versa. (BSD-style)
570 * If you set the real gid at all, or set the effective gid to a value not
571 * equal to the real gid, then the saved gid is set to the new effective gid.
573 * This makes it possible for a setgid program to completely drop its
574 * privileges, which is often a useful assertion to make when you are doing
575 * a security audit over a program.
577 * The general idea is that a program which uses just setregid() will be
578 * 100% compatible with BSD. A program which uses just setgid() will be
579 * 100% compatible with POSIX with saved IDs.
581 * SMP: There are not races, the GIDs are checked only by filesystem
582 * operations (as far as semantic preservation is concerned).
584 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
586 struct user_namespace *ns = current_user_ns();
587 const struct cred *old;
592 krgid = make_kgid(ns, rgid);
593 kegid = make_kgid(ns, egid);
595 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
597 if ((egid != (gid_t) -1) && !gid_valid(kegid))
600 new = prepare_creds();
603 old = current_cred();
606 if (rgid != (gid_t) -1) {
607 if (gid_eq(old->gid, krgid) ||
608 gid_eq(old->egid, krgid) ||
609 nsown_capable(CAP_SETGID))
614 if (egid != (gid_t) -1) {
615 if (gid_eq(old->gid, kegid) ||
616 gid_eq(old->egid, kegid) ||
617 gid_eq(old->sgid, kegid) ||
618 nsown_capable(CAP_SETGID))
624 if (rgid != (gid_t) -1 ||
625 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
626 new->sgid = new->egid;
627 new->fsgid = new->egid;
629 return commit_creds(new);
637 * setgid() is implemented like SysV w/ SAVED_IDS
639 * SMP: Same implicit races as above.
641 SYSCALL_DEFINE1(setgid, gid_t, gid)
643 struct user_namespace *ns = current_user_ns();
644 const struct cred *old;
649 kgid = make_kgid(ns, gid);
650 if (!gid_valid(kgid))
653 new = prepare_creds();
656 old = current_cred();
659 if (nsown_capable(CAP_SETGID))
660 new->gid = new->egid = new->sgid = new->fsgid = kgid;
661 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
662 new->egid = new->fsgid = kgid;
666 return commit_creds(new);
674 * change the user struct in a credentials set to match the new UID
676 static int set_user(struct cred *new)
678 struct user_struct *new_user;
680 new_user = alloc_uid(new->uid);
685 * We don't fail in case of NPROC limit excess here because too many
686 * poorly written programs don't check set*uid() return code, assuming
687 * it never fails if called by root. We may still enforce NPROC limit
688 * for programs doing set*uid()+execve() by harmlessly deferring the
689 * failure to the execve() stage.
691 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
692 new_user != INIT_USER)
693 current->flags |= PF_NPROC_EXCEEDED;
695 current->flags &= ~PF_NPROC_EXCEEDED;
698 new->user = new_user;
703 * Unprivileged users may change the real uid to the effective uid
704 * or vice versa. (BSD-style)
706 * If you set the real uid at all, or set the effective uid to a value not
707 * equal to the real uid, then the saved uid is set to the new effective uid.
709 * This makes it possible for a setuid program to completely drop its
710 * privileges, which is often a useful assertion to make when you are doing
711 * a security audit over a program.
713 * The general idea is that a program which uses just setreuid() will be
714 * 100% compatible with BSD. A program which uses just setuid() will be
715 * 100% compatible with POSIX with saved IDs.
717 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
719 struct user_namespace *ns = current_user_ns();
720 const struct cred *old;
725 kruid = make_kuid(ns, ruid);
726 keuid = make_kuid(ns, euid);
728 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
730 if ((euid != (uid_t) -1) && !uid_valid(keuid))
733 new = prepare_creds();
736 old = current_cred();
739 if (ruid != (uid_t) -1) {
741 if (!uid_eq(old->uid, kruid) &&
742 !uid_eq(old->euid, kruid) &&
743 !nsown_capable(CAP_SETUID))
747 if (euid != (uid_t) -1) {
749 if (!uid_eq(old->uid, keuid) &&
750 !uid_eq(old->euid, keuid) &&
751 !uid_eq(old->suid, keuid) &&
752 !nsown_capable(CAP_SETUID))
756 if (!uid_eq(new->uid, old->uid)) {
757 retval = set_user(new);
761 if (ruid != (uid_t) -1 ||
762 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
763 new->suid = new->euid;
764 new->fsuid = new->euid;
766 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
770 return commit_creds(new);
778 * setuid() is implemented like SysV with SAVED_IDS
780 * Note that SAVED_ID's is deficient in that a setuid root program
781 * like sendmail, for example, cannot set its uid to be a normal
782 * user and then switch back, because if you're root, setuid() sets
783 * the saved uid too. If you don't like this, blame the bright people
784 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
785 * will allow a root program to temporarily drop privileges and be able to
786 * regain them by swapping the real and effective uid.
788 SYSCALL_DEFINE1(setuid, uid_t, uid)
790 struct user_namespace *ns = current_user_ns();
791 const struct cred *old;
796 kuid = make_kuid(ns, uid);
797 if (!uid_valid(kuid))
800 new = prepare_creds();
803 old = current_cred();
806 if (nsown_capable(CAP_SETUID)) {
807 new->suid = new->uid = kuid;
808 if (!uid_eq(kuid, old->uid)) {
809 retval = set_user(new);
813 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
817 new->fsuid = new->euid = kuid;
819 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
823 return commit_creds(new);
832 * This function implements a generic ability to update ruid, euid,
833 * and suid. This allows you to implement the 4.4 compatible seteuid().
835 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
837 struct user_namespace *ns = current_user_ns();
838 const struct cred *old;
841 kuid_t kruid, keuid, ksuid;
843 kruid = make_kuid(ns, ruid);
844 keuid = make_kuid(ns, euid);
845 ksuid = make_kuid(ns, suid);
847 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
850 if ((euid != (uid_t) -1) && !uid_valid(keuid))
853 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
856 new = prepare_creds();
860 old = current_cred();
863 if (!nsown_capable(CAP_SETUID)) {
864 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
865 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
867 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
868 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
870 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
871 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
875 if (ruid != (uid_t) -1) {
877 if (!uid_eq(kruid, old->uid)) {
878 retval = set_user(new);
883 if (euid != (uid_t) -1)
885 if (suid != (uid_t) -1)
887 new->fsuid = new->euid;
889 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
893 return commit_creds(new);
900 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
902 const struct cred *cred = current_cred();
904 uid_t ruid, euid, suid;
906 ruid = from_kuid_munged(cred->user_ns, cred->uid);
907 euid = from_kuid_munged(cred->user_ns, cred->euid);
908 suid = from_kuid_munged(cred->user_ns, cred->suid);
910 if (!(retval = put_user(ruid, ruidp)) &&
911 !(retval = put_user(euid, euidp)))
912 retval = put_user(suid, suidp);
918 * Same as above, but for rgid, egid, sgid.
920 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
922 struct user_namespace *ns = current_user_ns();
923 const struct cred *old;
926 kgid_t krgid, kegid, ksgid;
928 krgid = make_kgid(ns, rgid);
929 kegid = make_kgid(ns, egid);
930 ksgid = make_kgid(ns, sgid);
932 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
934 if ((egid != (gid_t) -1) && !gid_valid(kegid))
936 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
939 new = prepare_creds();
942 old = current_cred();
945 if (!nsown_capable(CAP_SETGID)) {
946 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
947 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
949 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
950 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
952 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
953 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
957 if (rgid != (gid_t) -1)
959 if (egid != (gid_t) -1)
961 if (sgid != (gid_t) -1)
963 new->fsgid = new->egid;
965 return commit_creds(new);
972 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
974 const struct cred *cred = current_cred();
976 gid_t rgid, egid, sgid;
978 rgid = from_kgid_munged(cred->user_ns, cred->gid);
979 egid = from_kgid_munged(cred->user_ns, cred->egid);
980 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
982 if (!(retval = put_user(rgid, rgidp)) &&
983 !(retval = put_user(egid, egidp)))
984 retval = put_user(sgid, sgidp);
991 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
992 * is used for "access()" and for the NFS daemon (letting nfsd stay at
993 * whatever uid it wants to). It normally shadows "euid", except when
994 * explicitly set by setfsuid() or for access..
996 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
998 const struct cred *old;
1003 old = current_cred();
1004 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
1006 kuid = make_kuid(old->user_ns, uid);
1007 if (!uid_valid(kuid))
1010 new = prepare_creds();
1014 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
1015 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
1016 nsown_capable(CAP_SETUID)) {
1017 if (!uid_eq(kuid, old->fsuid)) {
1019 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
1033 * Samma på svenska..
1035 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
1037 const struct cred *old;
1042 old = current_cred();
1043 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1045 kgid = make_kgid(old->user_ns, gid);
1046 if (!gid_valid(kgid))
1049 new = prepare_creds();
1053 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1054 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
1055 nsown_capable(CAP_SETGID)) {
1056 if (!gid_eq(kgid, old->fsgid)) {
1071 * sys_getpid - return the thread group id of the current process
1073 * Note, despite the name, this returns the tgid not the pid. The tgid and
1074 * the pid are identical unless CLONE_THREAD was specified on clone() in
1075 * which case the tgid is the same in all threads of the same group.
1077 * This is SMP safe as current->tgid does not change.
1079 SYSCALL_DEFINE0(getpid)
1081 return task_tgid_vnr(current);
1084 /* Thread ID - the internal kernel "pid" */
1085 SYSCALL_DEFINE0(gettid)
1087 return task_pid_vnr(current);
1091 * Accessing ->real_parent is not SMP-safe, it could
1092 * change from under us. However, we can use a stale
1093 * value of ->real_parent under rcu_read_lock(), see
1094 * release_task()->call_rcu(delayed_put_task_struct).
1096 SYSCALL_DEFINE0(getppid)
1101 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1107 SYSCALL_DEFINE0(getuid)
1109 /* Only we change this so SMP safe */
1110 return from_kuid_munged(current_user_ns(), current_uid());
1113 SYSCALL_DEFINE0(geteuid)
1115 /* Only we change this so SMP safe */
1116 return from_kuid_munged(current_user_ns(), current_euid());
1119 SYSCALL_DEFINE0(getgid)
1121 /* Only we change this so SMP safe */
1122 return from_kgid_munged(current_user_ns(), current_gid());
1125 SYSCALL_DEFINE0(getegid)
1127 /* Only we change this so SMP safe */
1128 return from_kgid_munged(current_user_ns(), current_egid());
1131 void do_sys_times(struct tms *tms)
1133 cputime_t tgutime, tgstime, cutime, cstime;
1135 spin_lock_irq(¤t->sighand->siglock);
1136 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1137 cutime = current->signal->cutime;
1138 cstime = current->signal->cstime;
1139 spin_unlock_irq(¤t->sighand->siglock);
1140 tms->tms_utime = cputime_to_clock_t(tgutime);
1141 tms->tms_stime = cputime_to_clock_t(tgstime);
1142 tms->tms_cutime = cputime_to_clock_t(cutime);
1143 tms->tms_cstime = cputime_to_clock_t(cstime);
1146 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1152 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1155 force_successful_syscall_return();
1156 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1160 * This needs some heavy checking ...
1161 * I just haven't the stomach for it. I also don't fully
1162 * understand sessions/pgrp etc. Let somebody who does explain it.
1164 * OK, I think I have the protection semantics right.... this is really
1165 * only important on a multi-user system anyway, to make sure one user
1166 * can't send a signal to a process owned by another. -TYT, 12/12/91
1168 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1171 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1173 struct task_struct *p;
1174 struct task_struct *group_leader = current->group_leader;
1179 pid = task_pid_vnr(group_leader);
1186 /* From this point forward we keep holding onto the tasklist lock
1187 * so that our parent does not change from under us. -DaveM
1189 write_lock_irq(&tasklist_lock);
1192 p = find_task_by_vpid(pid);
1197 if (!thread_group_leader(p))
1200 if (same_thread_group(p->real_parent, group_leader)) {
1202 if (task_session(p) != task_session(group_leader))
1209 if (p != group_leader)
1214 if (p->signal->leader)
1219 struct task_struct *g;
1221 pgrp = find_vpid(pgid);
1222 g = pid_task(pgrp, PIDTYPE_PGID);
1223 if (!g || task_session(g) != task_session(group_leader))
1227 err = security_task_setpgid(p, pgid);
1231 if (task_pgrp(p) != pgrp)
1232 change_pid(p, PIDTYPE_PGID, pgrp);
1236 /* All paths lead to here, thus we are safe. -DaveM */
1237 write_unlock_irq(&tasklist_lock);
1242 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1244 struct task_struct *p;
1250 grp = task_pgrp(current);
1253 p = find_task_by_vpid(pid);
1260 retval = security_task_getpgid(p);
1264 retval = pid_vnr(grp);
1270 #ifdef __ARCH_WANT_SYS_GETPGRP
1272 SYSCALL_DEFINE0(getpgrp)
1274 return sys_getpgid(0);
1279 SYSCALL_DEFINE1(getsid, pid_t, pid)
1281 struct task_struct *p;
1287 sid = task_session(current);
1290 p = find_task_by_vpid(pid);
1293 sid = task_session(p);
1297 retval = security_task_getsid(p);
1301 retval = pid_vnr(sid);
1307 static void set_special_pids(struct pid *pid)
1309 struct task_struct *curr = current->group_leader;
1311 if (task_session(curr) != pid)
1312 change_pid(curr, PIDTYPE_SID, pid);
1314 if (task_pgrp(curr) != pid)
1315 change_pid(curr, PIDTYPE_PGID, pid);
1318 SYSCALL_DEFINE0(setsid)
1320 struct task_struct *group_leader = current->group_leader;
1321 struct pid *sid = task_pid(group_leader);
1322 pid_t session = pid_vnr(sid);
1325 write_lock_irq(&tasklist_lock);
1326 /* Fail if I am already a session leader */
1327 if (group_leader->signal->leader)
1330 /* Fail if a process group id already exists that equals the
1331 * proposed session id.
1333 if (pid_task(sid, PIDTYPE_PGID))
1336 group_leader->signal->leader = 1;
1337 set_special_pids(sid);
1339 proc_clear_tty(group_leader);
1343 write_unlock_irq(&tasklist_lock);
1345 proc_sid_connector(group_leader);
1346 sched_autogroup_create_attach(group_leader);
1351 DECLARE_RWSEM(uts_sem);
1353 #ifdef COMPAT_UTS_MACHINE
1354 #define override_architecture(name) \
1355 (personality(current->personality) == PER_LINUX32 && \
1356 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1357 sizeof(COMPAT_UTS_MACHINE)))
1359 #define override_architecture(name) 0
1363 * Work around broken programs that cannot handle "Linux 3.0".
1364 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1366 static int override_release(char __user *release, size_t len)
1370 if (current->personality & UNAME26) {
1371 const char *rest = UTS_RELEASE;
1372 char buf[65] = { 0 };
1378 if (*rest == '.' && ++ndots >= 3)
1380 if (!isdigit(*rest) && *rest != '.')
1384 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1385 copy = clamp_t(size_t, len, 1, sizeof(buf));
1386 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1387 ret = copy_to_user(release, buf, copy + 1);
1392 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1396 down_read(&uts_sem);
1397 if (copy_to_user(name, utsname(), sizeof *name))
1401 if (!errno && override_release(name->release, sizeof(name->release)))
1403 if (!errno && override_architecture(name))
1408 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1412 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1419 down_read(&uts_sem);
1420 if (copy_to_user(name, utsname(), sizeof(*name)))
1424 if (!error && override_release(name->release, sizeof(name->release)))
1426 if (!error && override_architecture(name))
1431 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1437 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1440 down_read(&uts_sem);
1441 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1443 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1444 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1446 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1447 error |= __copy_to_user(&name->release, &utsname()->release,
1449 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1450 error |= __copy_to_user(&name->version, &utsname()->version,
1452 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1453 error |= __copy_to_user(&name->machine, &utsname()->machine,
1455 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1458 if (!error && override_architecture(name))
1460 if (!error && override_release(name->release, sizeof(name->release)))
1462 return error ? -EFAULT : 0;
1466 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1469 char tmp[__NEW_UTS_LEN];
1471 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1474 if (len < 0 || len > __NEW_UTS_LEN)
1476 down_write(&uts_sem);
1478 if (!copy_from_user(tmp, name, len)) {
1479 struct new_utsname *u = utsname();
1481 memcpy(u->nodename, tmp, len);
1482 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1484 uts_proc_notify(UTS_PROC_HOSTNAME);
1490 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1492 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1495 struct new_utsname *u;
1499 down_read(&uts_sem);
1501 i = 1 + strlen(u->nodename);
1505 if (copy_to_user(name, u->nodename, i))
1514 * Only setdomainname; getdomainname can be implemented by calling
1517 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1520 char tmp[__NEW_UTS_LEN];
1522 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1524 if (len < 0 || len > __NEW_UTS_LEN)
1527 down_write(&uts_sem);
1529 if (!copy_from_user(tmp, name, len)) {
1530 struct new_utsname *u = utsname();
1532 memcpy(u->domainname, tmp, len);
1533 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1535 uts_proc_notify(UTS_PROC_DOMAINNAME);
1541 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1543 struct rlimit value;
1546 ret = do_prlimit(current, resource, NULL, &value);
1548 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1553 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1556 * Back compatibility for getrlimit. Needed for some apps.
1559 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1560 struct rlimit __user *, rlim)
1563 if (resource >= RLIM_NLIMITS)
1566 task_lock(current->group_leader);
1567 x = current->signal->rlim[resource];
1568 task_unlock(current->group_leader);
1569 if (x.rlim_cur > 0x7FFFFFFF)
1570 x.rlim_cur = 0x7FFFFFFF;
1571 if (x.rlim_max > 0x7FFFFFFF)
1572 x.rlim_max = 0x7FFFFFFF;
1573 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1578 static inline bool rlim64_is_infinity(__u64 rlim64)
1580 #if BITS_PER_LONG < 64
1581 return rlim64 >= ULONG_MAX;
1583 return rlim64 == RLIM64_INFINITY;
1587 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1589 if (rlim->rlim_cur == RLIM_INFINITY)
1590 rlim64->rlim_cur = RLIM64_INFINITY;
1592 rlim64->rlim_cur = rlim->rlim_cur;
1593 if (rlim->rlim_max == RLIM_INFINITY)
1594 rlim64->rlim_max = RLIM64_INFINITY;
1596 rlim64->rlim_max = rlim->rlim_max;
1599 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1601 if (rlim64_is_infinity(rlim64->rlim_cur))
1602 rlim->rlim_cur = RLIM_INFINITY;
1604 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1605 if (rlim64_is_infinity(rlim64->rlim_max))
1606 rlim->rlim_max = RLIM_INFINITY;
1608 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1611 /* make sure you are allowed to change @tsk limits before calling this */
1612 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1613 struct rlimit *new_rlim, struct rlimit *old_rlim)
1615 struct rlimit *rlim;
1618 if (resource >= RLIM_NLIMITS)
1621 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1623 if (resource == RLIMIT_NOFILE &&
1624 new_rlim->rlim_max > sysctl_nr_open)
1628 /* protect tsk->signal and tsk->sighand from disappearing */
1629 read_lock(&tasklist_lock);
1630 if (!tsk->sighand) {
1635 rlim = tsk->signal->rlim + resource;
1636 task_lock(tsk->group_leader);
1638 /* Keep the capable check against init_user_ns until
1639 cgroups can contain all limits */
1640 if (new_rlim->rlim_max > rlim->rlim_max &&
1641 !capable(CAP_SYS_RESOURCE))
1644 retval = security_task_setrlimit(tsk->group_leader,
1645 resource, new_rlim);
1646 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1648 * The caller is asking for an immediate RLIMIT_CPU
1649 * expiry. But we use the zero value to mean "it was
1650 * never set". So let's cheat and make it one second
1653 new_rlim->rlim_cur = 1;
1662 task_unlock(tsk->group_leader);
1665 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1666 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1667 * very long-standing error, and fixing it now risks breakage of
1668 * applications, so we live with it
1670 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1671 new_rlim->rlim_cur != RLIM_INFINITY)
1672 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1674 read_unlock(&tasklist_lock);
1678 /* rcu lock must be held */
1679 static int check_prlimit_permission(struct task_struct *task)
1681 const struct cred *cred = current_cred(), *tcred;
1683 if (current == task)
1686 tcred = __task_cred(task);
1687 if (uid_eq(cred->uid, tcred->euid) &&
1688 uid_eq(cred->uid, tcred->suid) &&
1689 uid_eq(cred->uid, tcred->uid) &&
1690 gid_eq(cred->gid, tcred->egid) &&
1691 gid_eq(cred->gid, tcred->sgid) &&
1692 gid_eq(cred->gid, tcred->gid))
1694 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1700 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1701 const struct rlimit64 __user *, new_rlim,
1702 struct rlimit64 __user *, old_rlim)
1704 struct rlimit64 old64, new64;
1705 struct rlimit old, new;
1706 struct task_struct *tsk;
1710 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1712 rlim64_to_rlim(&new64, &new);
1716 tsk = pid ? find_task_by_vpid(pid) : current;
1721 ret = check_prlimit_permission(tsk);
1726 get_task_struct(tsk);
1729 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1730 old_rlim ? &old : NULL);
1732 if (!ret && old_rlim) {
1733 rlim_to_rlim64(&old, &old64);
1734 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1738 put_task_struct(tsk);
1742 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1744 struct rlimit new_rlim;
1746 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1748 return do_prlimit(current, resource, &new_rlim, NULL);
1752 * It would make sense to put struct rusage in the task_struct,
1753 * except that would make the task_struct be *really big*. After
1754 * task_struct gets moved into malloc'ed memory, it would
1755 * make sense to do this. It will make moving the rest of the information
1756 * a lot simpler! (Which we're not doing right now because we're not
1757 * measuring them yet).
1759 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1760 * races with threads incrementing their own counters. But since word
1761 * reads are atomic, we either get new values or old values and we don't
1762 * care which for the sums. We always take the siglock to protect reading
1763 * the c* fields from p->signal from races with exit.c updating those
1764 * fields when reaping, so a sample either gets all the additions of a
1765 * given child after it's reaped, or none so this sample is before reaping.
1768 * We need to take the siglock for CHILDEREN, SELF and BOTH
1769 * for the cases current multithreaded, non-current single threaded
1770 * non-current multithreaded. Thread traversal is now safe with
1772 * Strictly speaking, we donot need to take the siglock if we are current and
1773 * single threaded, as no one else can take our signal_struct away, no one
1774 * else can reap the children to update signal->c* counters, and no one else
1775 * can race with the signal-> fields. If we do not take any lock, the
1776 * signal-> fields could be read out of order while another thread was just
1777 * exiting. So we should place a read memory barrier when we avoid the lock.
1778 * On the writer side, write memory barrier is implied in __exit_signal
1779 * as __exit_signal releases the siglock spinlock after updating the signal->
1780 * fields. But we don't do this yet to keep things simple.
1784 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1786 r->ru_nvcsw += t->nvcsw;
1787 r->ru_nivcsw += t->nivcsw;
1788 r->ru_minflt += t->min_flt;
1789 r->ru_majflt += t->maj_flt;
1790 r->ru_inblock += task_io_get_inblock(t);
1791 r->ru_oublock += task_io_get_oublock(t);
1794 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1796 struct task_struct *t;
1797 unsigned long flags;
1798 cputime_t tgutime, tgstime, utime, stime;
1799 unsigned long maxrss = 0;
1801 memset((char *) r, 0, sizeof *r);
1804 if (who == RUSAGE_THREAD) {
1805 task_cputime_adjusted(current, &utime, &stime);
1806 accumulate_thread_rusage(p, r);
1807 maxrss = p->signal->maxrss;
1811 if (!lock_task_sighand(p, &flags))
1816 case RUSAGE_CHILDREN:
1817 utime = p->signal->cutime;
1818 stime = p->signal->cstime;
1819 r->ru_nvcsw = p->signal->cnvcsw;
1820 r->ru_nivcsw = p->signal->cnivcsw;
1821 r->ru_minflt = p->signal->cmin_flt;
1822 r->ru_majflt = p->signal->cmaj_flt;
1823 r->ru_inblock = p->signal->cinblock;
1824 r->ru_oublock = p->signal->coublock;
1825 maxrss = p->signal->cmaxrss;
1827 if (who == RUSAGE_CHILDREN)
1831 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1834 r->ru_nvcsw += p->signal->nvcsw;
1835 r->ru_nivcsw += p->signal->nivcsw;
1836 r->ru_minflt += p->signal->min_flt;
1837 r->ru_majflt += p->signal->maj_flt;
1838 r->ru_inblock += p->signal->inblock;
1839 r->ru_oublock += p->signal->oublock;
1840 if (maxrss < p->signal->maxrss)
1841 maxrss = p->signal->maxrss;
1844 accumulate_thread_rusage(t, r);
1852 unlock_task_sighand(p, &flags);
1855 cputime_to_timeval(utime, &r->ru_utime);
1856 cputime_to_timeval(stime, &r->ru_stime);
1858 if (who != RUSAGE_CHILDREN) {
1859 struct mm_struct *mm = get_task_mm(p);
1861 setmax_mm_hiwater_rss(&maxrss, mm);
1865 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1868 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1871 k_getrusage(p, who, &r);
1872 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1875 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1877 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1878 who != RUSAGE_THREAD)
1880 return getrusage(current, who, ru);
1883 #ifdef CONFIG_COMPAT
1884 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1888 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1889 who != RUSAGE_THREAD)
1892 k_getrusage(current, who, &r);
1893 return put_compat_rusage(&r, ru);
1897 SYSCALL_DEFINE1(umask, int, mask)
1899 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1903 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1906 struct inode *inode;
1913 inode = file_inode(exe.file);
1916 * Because the original mm->exe_file points to executable file, make
1917 * sure that this one is executable as well, to avoid breaking an
1921 if (!S_ISREG(inode->i_mode) ||
1922 exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1925 err = inode_permission(inode, MAY_EXEC);
1929 down_write(&mm->mmap_sem);
1932 * Forbid mm->exe_file change if old file still mapped.
1936 struct vm_area_struct *vma;
1938 for (vma = mm->mmap; vma; vma = vma->vm_next)
1940 path_equal(&vma->vm_file->f_path,
1941 &mm->exe_file->f_path))
1946 * The symlink can be changed only once, just to disallow arbitrary
1947 * transitions malicious software might bring in. This means one
1948 * could make a snapshot over all processes running and monitor
1949 * /proc/pid/exe changes to notice unusual activity if needed.
1952 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1956 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
1958 up_write(&mm->mmap_sem);
1965 static int prctl_set_mm(int opt, unsigned long addr,
1966 unsigned long arg4, unsigned long arg5)
1968 unsigned long rlim = rlimit(RLIMIT_DATA);
1969 struct mm_struct *mm = current->mm;
1970 struct vm_area_struct *vma;
1973 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
1976 if (!capable(CAP_SYS_RESOURCE))
1979 if (opt == PR_SET_MM_EXE_FILE)
1980 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1982 if (addr >= TASK_SIZE || addr < mmap_min_addr)
1987 down_read(&mm->mmap_sem);
1988 vma = find_vma(mm, addr);
1991 case PR_SET_MM_START_CODE:
1992 mm->start_code = addr;
1994 case PR_SET_MM_END_CODE:
1995 mm->end_code = addr;
1997 case PR_SET_MM_START_DATA:
1998 mm->start_data = addr;
2000 case PR_SET_MM_END_DATA:
2001 mm->end_data = addr;
2004 case PR_SET_MM_START_BRK:
2005 if (addr <= mm->end_data)
2008 if (rlim < RLIM_INFINITY &&
2010 (mm->end_data - mm->start_data) > rlim)
2013 mm->start_brk = addr;
2017 if (addr <= mm->end_data)
2020 if (rlim < RLIM_INFINITY &&
2021 (addr - mm->start_brk) +
2022 (mm->end_data - mm->start_data) > rlim)
2029 * If command line arguments and environment
2030 * are placed somewhere else on stack, we can
2031 * set them up here, ARG_START/END to setup
2032 * command line argumets and ENV_START/END
2035 case PR_SET_MM_START_STACK:
2036 case PR_SET_MM_ARG_START:
2037 case PR_SET_MM_ARG_END:
2038 case PR_SET_MM_ENV_START:
2039 case PR_SET_MM_ENV_END:
2044 if (opt == PR_SET_MM_START_STACK)
2045 mm->start_stack = addr;
2046 else if (opt == PR_SET_MM_ARG_START)
2047 mm->arg_start = addr;
2048 else if (opt == PR_SET_MM_ARG_END)
2050 else if (opt == PR_SET_MM_ENV_START)
2051 mm->env_start = addr;
2052 else if (opt == PR_SET_MM_ENV_END)
2057 * This doesn't move auxiliary vector itself
2058 * since it's pinned to mm_struct, but allow
2059 * to fill vector with new values. It's up
2060 * to a caller to provide sane values here
2061 * otherwise user space tools which use this
2062 * vector might be unhappy.
2064 case PR_SET_MM_AUXV: {
2065 unsigned long user_auxv[AT_VECTOR_SIZE];
2067 if (arg4 > sizeof(user_auxv))
2069 up_read(&mm->mmap_sem);
2071 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
2074 /* Make sure the last entry is always AT_NULL */
2075 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2076 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2078 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2081 memcpy(mm->saved_auxv, user_auxv, arg4);
2082 task_unlock(current);
2092 up_read(&mm->mmap_sem);
2096 #ifdef CONFIG_CHECKPOINT_RESTORE
2097 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2099 return put_user(me->clear_child_tid, tid_addr);
2102 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2108 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2109 unsigned long, arg4, unsigned long, arg5)
2111 struct task_struct *me = current;
2112 unsigned char comm[sizeof(me->comm)];
2115 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2116 if (error != -ENOSYS)
2121 case PR_SET_PDEATHSIG:
2122 if (!valid_signal(arg2)) {
2126 me->pdeath_signal = arg2;
2128 case PR_GET_PDEATHSIG:
2129 error = put_user(me->pdeath_signal, (int __user *)arg2);
2131 case PR_GET_DUMPABLE:
2132 error = get_dumpable(me->mm);
2134 case PR_SET_DUMPABLE:
2135 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2139 set_dumpable(me->mm, arg2);
2142 case PR_SET_UNALIGN:
2143 error = SET_UNALIGN_CTL(me, arg2);
2145 case PR_GET_UNALIGN:
2146 error = GET_UNALIGN_CTL(me, arg2);
2149 error = SET_FPEMU_CTL(me, arg2);
2152 error = GET_FPEMU_CTL(me, arg2);
2155 error = SET_FPEXC_CTL(me, arg2);
2158 error = GET_FPEXC_CTL(me, arg2);
2161 error = PR_TIMING_STATISTICAL;
2164 if (arg2 != PR_TIMING_STATISTICAL)
2168 comm[sizeof(me->comm) - 1] = 0;
2169 if (strncpy_from_user(comm, (char __user *)arg2,
2170 sizeof(me->comm) - 1) < 0)
2172 set_task_comm(me, comm);
2173 proc_comm_connector(me);
2176 get_task_comm(comm, me);
2177 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2181 error = GET_ENDIAN(me, arg2);
2184 error = SET_ENDIAN(me, arg2);
2186 case PR_GET_SECCOMP:
2187 error = prctl_get_seccomp();
2189 case PR_SET_SECCOMP:
2190 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2193 error = GET_TSC_CTL(arg2);
2196 error = SET_TSC_CTL(arg2);
2198 case PR_TASK_PERF_EVENTS_DISABLE:
2199 error = perf_event_task_disable();
2201 case PR_TASK_PERF_EVENTS_ENABLE:
2202 error = perf_event_task_enable();
2204 case PR_GET_TIMERSLACK:
2205 error = current->timer_slack_ns;
2207 case PR_SET_TIMERSLACK:
2209 current->timer_slack_ns =
2210 current->default_timer_slack_ns;
2212 current->timer_slack_ns = arg2;
2218 case PR_MCE_KILL_CLEAR:
2221 current->flags &= ~PF_MCE_PROCESS;
2223 case PR_MCE_KILL_SET:
2224 current->flags |= PF_MCE_PROCESS;
2225 if (arg3 == PR_MCE_KILL_EARLY)
2226 current->flags |= PF_MCE_EARLY;
2227 else if (arg3 == PR_MCE_KILL_LATE)
2228 current->flags &= ~PF_MCE_EARLY;
2229 else if (arg3 == PR_MCE_KILL_DEFAULT)
2231 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2239 case PR_MCE_KILL_GET:
2240 if (arg2 | arg3 | arg4 | arg5)
2242 if (current->flags & PF_MCE_PROCESS)
2243 error = (current->flags & PF_MCE_EARLY) ?
2244 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2246 error = PR_MCE_KILL_DEFAULT;
2249 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2251 case PR_GET_TID_ADDRESS:
2252 error = prctl_get_tid_address(me, (int __user **)arg2);
2254 case PR_SET_CHILD_SUBREAPER:
2255 me->signal->is_child_subreaper = !!arg2;
2257 case PR_GET_CHILD_SUBREAPER:
2258 error = put_user(me->signal->is_child_subreaper,
2259 (int __user *)arg2);
2261 case PR_SET_NO_NEW_PRIVS:
2262 if (arg2 != 1 || arg3 || arg4 || arg5)
2265 current->no_new_privs = 1;
2267 case PR_GET_NO_NEW_PRIVS:
2268 if (arg2 || arg3 || arg4 || arg5)
2270 return current->no_new_privs ? 1 : 0;
2278 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2279 struct getcpu_cache __user *, unused)
2282 int cpu = raw_smp_processor_id();
2284 err |= put_user(cpu, cpup);
2286 err |= put_user(cpu_to_node(cpu), nodep);
2287 return err ? -EFAULT : 0;
2290 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2292 static int __orderly_poweroff(bool force)
2295 static char *envp[] = {
2297 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2302 argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL);
2304 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
2307 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2308 __func__, poweroff_cmd);
2313 printk(KERN_WARNING "Failed to start orderly shutdown: "
2314 "forcing the issue\n");
2316 * I guess this should try to kick off some daemon to sync and
2317 * poweroff asap. Or not even bother syncing if we're doing an
2318 * emergency shutdown?
2327 static bool poweroff_force;
2329 static void poweroff_work_func(struct work_struct *work)
2331 __orderly_poweroff(poweroff_force);
2334 static DECLARE_WORK(poweroff_work, poweroff_work_func);
2337 * orderly_poweroff - Trigger an orderly system poweroff
2338 * @force: force poweroff if command execution fails
2340 * This may be called from any context to trigger a system shutdown.
2341 * If the orderly shutdown fails, it will force an immediate shutdown.
2343 int orderly_poweroff(bool force)
2345 if (force) /* do not override the pending "true" */
2346 poweroff_force = true;
2347 schedule_work(&poweroff_work);
2350 EXPORT_SYMBOL_GPL(orderly_poweroff);
2353 * do_sysinfo - fill in sysinfo struct
2354 * @info: pointer to buffer to fill
2356 static int do_sysinfo(struct sysinfo *info)
2358 unsigned long mem_total, sav_total;
2359 unsigned int mem_unit, bitcount;
2362 memset(info, 0, sizeof(struct sysinfo));
2364 get_monotonic_boottime(&tp);
2365 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2367 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2369 info->procs = nr_threads;
2375 * If the sum of all the available memory (i.e. ram + swap)
2376 * is less than can be stored in a 32 bit unsigned long then
2377 * we can be binary compatible with 2.2.x kernels. If not,
2378 * well, in that case 2.2.x was broken anyways...
2380 * -Erik Andersen <andersee@debian.org>
2383 mem_total = info->totalram + info->totalswap;
2384 if (mem_total < info->totalram || mem_total < info->totalswap)
2387 mem_unit = info->mem_unit;
2388 while (mem_unit > 1) {
2391 sav_total = mem_total;
2393 if (mem_total < sav_total)
2398 * If mem_total did not overflow, multiply all memory values by
2399 * info->mem_unit and set it to 1. This leaves things compatible
2400 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2405 info->totalram <<= bitcount;
2406 info->freeram <<= bitcount;
2407 info->sharedram <<= bitcount;
2408 info->bufferram <<= bitcount;
2409 info->totalswap <<= bitcount;
2410 info->freeswap <<= bitcount;
2411 info->totalhigh <<= bitcount;
2412 info->freehigh <<= bitcount;
2418 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2424 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2430 #ifdef CONFIG_COMPAT
2431 struct compat_sysinfo {
2445 char _f[20-2*sizeof(u32)-sizeof(int)];
2448 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2454 /* Check to see if any memory value is too large for 32-bit and scale
2457 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2460 while (s.mem_unit < PAGE_SIZE) {
2465 s.totalram >>= bitcount;
2466 s.freeram >>= bitcount;
2467 s.sharedram >>= bitcount;
2468 s.bufferram >>= bitcount;
2469 s.totalswap >>= bitcount;
2470 s.freeswap >>= bitcount;
2471 s.totalhigh >>= bitcount;
2472 s.freehigh >>= bitcount;
2475 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2476 __put_user(s.uptime, &info->uptime) ||
2477 __put_user(s.loads[0], &info->loads[0]) ||
2478 __put_user(s.loads[1], &info->loads[1]) ||
2479 __put_user(s.loads[2], &info->loads[2]) ||
2480 __put_user(s.totalram, &info->totalram) ||
2481 __put_user(s.freeram, &info->freeram) ||
2482 __put_user(s.sharedram, &info->sharedram) ||
2483 __put_user(s.bufferram, &info->bufferram) ||
2484 __put_user(s.totalswap, &info->totalswap) ||
2485 __put_user(s.freeswap, &info->freeswap) ||
2486 __put_user(s.procs, &info->procs) ||
2487 __put_user(s.totalhigh, &info->totalhigh) ||
2488 __put_user(s.freehigh, &info->freehigh) ||
2489 __put_user(s.mem_unit, &info->mem_unit))
2494 #endif /* CONFIG_COMPAT */