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/mempolicy.h>
47 #include <linux/sched.h>
49 #include <linux/compat.h>
50 #include <linux/syscalls.h>
51 #include <linux/kprobes.h>
52 #include <linux/user_namespace.h>
53 #include <linux/binfmts.h>
55 #include <linux/sched.h>
56 #include <linux/rcupdate.h>
57 #include <linux/uidgid.h>
58 #include <linux/cred.h>
60 #include <linux/kmsg_dump.h>
61 /* Move somewhere else to avoid recompiling? */
62 #include <generated/utsrelease.h>
64 #include <asm/uaccess.h>
66 #include <asm/unistd.h>
68 #ifndef SET_UNALIGN_CTL
69 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
71 #ifndef GET_UNALIGN_CTL
72 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
75 # define SET_FPEMU_CTL(a,b) (-EINVAL)
78 # define GET_FPEMU_CTL(a,b) (-EINVAL)
81 # define SET_FPEXC_CTL(a,b) (-EINVAL)
84 # define GET_FPEXC_CTL(a,b) (-EINVAL)
87 # define GET_ENDIAN(a,b) (-EINVAL)
90 # define SET_ENDIAN(a,b) (-EINVAL)
93 # define GET_TSC_CTL(a) (-EINVAL)
96 # define SET_TSC_CTL(a) (-EINVAL)
100 * this is where the system-wide overflow UID and GID are defined, for
101 * architectures that now have 32-bit UID/GID but didn't in the past
104 int overflowuid = DEFAULT_OVERFLOWUID;
105 int overflowgid = DEFAULT_OVERFLOWGID;
107 EXPORT_SYMBOL(overflowuid);
108 EXPORT_SYMBOL(overflowgid);
111 * the same as above, but for filesystems which can only store a 16-bit
112 * UID and GID. as such, this is needed on all architectures
115 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
116 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
118 EXPORT_SYMBOL(fs_overflowuid);
119 EXPORT_SYMBOL(fs_overflowgid);
122 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
127 EXPORT_SYMBOL(cad_pid);
130 * If set, this is used for preparing the system to power off.
133 void (*pm_power_off_prepare)(void);
136 * Returns true if current's euid is same as p's uid or euid,
137 * or has CAP_SYS_NICE to p's user_ns.
139 * Called with rcu_read_lock, creds are safe
141 static bool set_one_prio_perm(struct task_struct *p)
143 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
145 if (uid_eq(pcred->uid, cred->euid) ||
146 uid_eq(pcred->euid, cred->euid))
148 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
154 * set the priority of a task
155 * - the caller must hold the RCU read lock
157 static int set_one_prio(struct task_struct *p, int niceval, int error)
161 if (!set_one_prio_perm(p)) {
165 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
169 no_nice = security_task_setnice(p, niceval);
176 set_user_nice(p, niceval);
181 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
183 struct task_struct *g, *p;
184 struct user_struct *user;
185 const struct cred *cred = current_cred();
190 if (which > PRIO_USER || which < PRIO_PROCESS)
193 /* normalize: avoid signed division (rounding problems) */
201 read_lock(&tasklist_lock);
205 p = find_task_by_vpid(who);
209 error = set_one_prio(p, niceval, error);
213 pgrp = find_vpid(who);
215 pgrp = task_pgrp(current);
216 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
217 error = set_one_prio(p, niceval, error);
218 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
221 uid = make_kuid(cred->user_ns, who);
225 else if (!uid_eq(uid, cred->uid) &&
226 !(user = find_user(uid)))
227 goto out_unlock; /* No processes for this user */
229 do_each_thread(g, p) {
230 if (uid_eq(task_uid(p), uid))
231 error = set_one_prio(p, niceval, error);
232 } while_each_thread(g, p);
233 if (!uid_eq(uid, cred->uid))
234 free_uid(user); /* For find_user() */
238 read_unlock(&tasklist_lock);
245 * Ugh. To avoid negative return values, "getpriority()" will
246 * not return the normal nice-value, but a negated value that
247 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
248 * to stay compatible.
250 SYSCALL_DEFINE2(getpriority, int, which, int, who)
252 struct task_struct *g, *p;
253 struct user_struct *user;
254 const struct cred *cred = current_cred();
255 long niceval, retval = -ESRCH;
259 if (which > PRIO_USER || which < PRIO_PROCESS)
263 read_lock(&tasklist_lock);
267 p = find_task_by_vpid(who);
271 niceval = 20 - task_nice(p);
272 if (niceval > retval)
278 pgrp = find_vpid(who);
280 pgrp = task_pgrp(current);
281 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
282 niceval = 20 - task_nice(p);
283 if (niceval > retval)
285 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
288 uid = make_kuid(cred->user_ns, who);
292 else if (!uid_eq(uid, cred->uid) &&
293 !(user = find_user(uid)))
294 goto out_unlock; /* No processes for this user */
296 do_each_thread(g, p) {
297 if (uid_eq(task_uid(p), uid)) {
298 niceval = 20 - task_nice(p);
299 if (niceval > retval)
302 } while_each_thread(g, p);
303 if (!uid_eq(uid, cred->uid))
304 free_uid(user); /* for find_user() */
308 read_unlock(&tasklist_lock);
315 * emergency_restart - reboot the system
317 * Without shutting down any hardware or taking any locks
318 * reboot the system. This is called when we know we are in
319 * trouble so this is our best effort to reboot. This is
320 * safe to call in interrupt context.
322 void emergency_restart(void)
324 kmsg_dump(KMSG_DUMP_EMERG);
325 machine_emergency_restart();
327 EXPORT_SYMBOL_GPL(emergency_restart);
329 void kernel_restart_prepare(char *cmd)
331 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
332 system_state = SYSTEM_RESTART;
333 usermodehelper_disable();
338 * register_reboot_notifier - Register function to be called at reboot time
339 * @nb: Info about notifier function to be called
341 * Registers a function with the list of functions
342 * to be called at reboot time.
344 * Currently always returns zero, as blocking_notifier_chain_register()
345 * always returns zero.
347 int register_reboot_notifier(struct notifier_block *nb)
349 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
351 EXPORT_SYMBOL(register_reboot_notifier);
354 * unregister_reboot_notifier - Unregister previously registered reboot notifier
355 * @nb: Hook to be unregistered
357 * Unregisters a previously registered reboot
360 * Returns zero on success, or %-ENOENT on failure.
362 int unregister_reboot_notifier(struct notifier_block *nb)
364 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
366 EXPORT_SYMBOL(unregister_reboot_notifier);
368 /* Add backwards compatibility for stable trees. */
369 #ifndef PF_NO_SETAFFINITY
370 #define PF_NO_SETAFFINITY PF_THREAD_BOUND
373 static void migrate_to_reboot_cpu(void)
375 /* The boot cpu is always logical cpu 0 */
378 cpu_hotplug_disable();
380 /* Make certain the cpu I'm about to reboot on is online */
381 if (!cpu_online(cpu))
382 cpu = cpumask_first(cpu_online_mask);
384 /* Prevent races with other tasks migrating this task */
385 current->flags |= PF_NO_SETAFFINITY;
387 /* Make certain I only run on the appropriate processor */
388 set_cpus_allowed_ptr(current, cpumask_of(cpu));
392 * kernel_restart - reboot the system
393 * @cmd: pointer to buffer containing command to execute for restart
396 * Shutdown everything and perform a clean reboot.
397 * This is not safe to call in interrupt context.
399 void kernel_restart(char *cmd)
401 kernel_restart_prepare(cmd);
402 migrate_to_reboot_cpu();
405 printk(KERN_EMERG "Restarting system.\n");
407 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
408 kmsg_dump(KMSG_DUMP_RESTART);
410 machine_restart(cmd);
412 EXPORT_SYMBOL_GPL(kernel_restart);
414 static void kernel_shutdown_prepare(enum system_states state)
416 blocking_notifier_call_chain(&reboot_notifier_list,
417 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
418 system_state = state;
419 usermodehelper_disable();
423 * kernel_halt - halt the system
425 * Shutdown everything and perform a clean system halt.
427 void kernel_halt(void)
429 kernel_shutdown_prepare(SYSTEM_HALT);
430 migrate_to_reboot_cpu();
432 printk(KERN_EMERG "System halted.\n");
433 kmsg_dump(KMSG_DUMP_HALT);
437 EXPORT_SYMBOL_GPL(kernel_halt);
440 * kernel_power_off - power_off the system
442 * Shutdown everything and perform a clean system power_off.
444 void kernel_power_off(void)
446 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
447 if (pm_power_off_prepare)
448 pm_power_off_prepare();
449 migrate_to_reboot_cpu();
451 printk(KERN_EMERG "Power down.\n");
452 kmsg_dump(KMSG_DUMP_POWEROFF);
455 EXPORT_SYMBOL_GPL(kernel_power_off);
457 static DEFINE_MUTEX(reboot_mutex);
460 * Reboot system call: for obvious reasons only root may call it,
461 * and even root needs to set up some magic numbers in the registers
462 * so that some mistake won't make this reboot the whole machine.
463 * You can also set the meaning of the ctrl-alt-del-key here.
465 * reboot doesn't sync: do that yourself before calling this.
467 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
470 struct pid_namespace *pid_ns = task_active_pid_ns(current);
474 /* We only trust the superuser with rebooting the system. */
475 if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
478 /* For safety, we require "magic" arguments. */
479 if (magic1 != LINUX_REBOOT_MAGIC1 ||
480 (magic2 != LINUX_REBOOT_MAGIC2 &&
481 magic2 != LINUX_REBOOT_MAGIC2A &&
482 magic2 != LINUX_REBOOT_MAGIC2B &&
483 magic2 != LINUX_REBOOT_MAGIC2C))
487 * If pid namespaces are enabled and the current task is in a child
488 * pid_namespace, the command is handled by reboot_pid_ns() which will
491 ret = reboot_pid_ns(pid_ns, cmd);
495 /* Instead of trying to make the power_off code look like
496 * halt when pm_power_off is not set do it the easy way.
498 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
499 cmd = LINUX_REBOOT_CMD_HALT;
501 mutex_lock(&reboot_mutex);
503 case LINUX_REBOOT_CMD_RESTART:
504 kernel_restart(NULL);
507 case LINUX_REBOOT_CMD_CAD_ON:
511 case LINUX_REBOOT_CMD_CAD_OFF:
515 case LINUX_REBOOT_CMD_HALT:
518 panic("cannot halt");
520 case LINUX_REBOOT_CMD_POWER_OFF:
525 case LINUX_REBOOT_CMD_RESTART2:
526 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
530 buffer[sizeof(buffer) - 1] = '\0';
532 kernel_restart(buffer);
536 case LINUX_REBOOT_CMD_KEXEC:
537 ret = kernel_kexec();
541 #ifdef CONFIG_HIBERNATION
542 case LINUX_REBOOT_CMD_SW_SUSPEND:
551 mutex_unlock(&reboot_mutex);
555 static void deferred_cad(struct work_struct *dummy)
557 kernel_restart(NULL);
561 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
562 * As it's called within an interrupt, it may NOT sync: the only choice
563 * is whether to reboot at once, or just ignore the ctrl-alt-del.
565 void ctrl_alt_del(void)
567 static DECLARE_WORK(cad_work, deferred_cad);
570 schedule_work(&cad_work);
572 kill_cad_pid(SIGINT, 1);
576 * Unprivileged users may change the real gid to the effective gid
577 * or vice versa. (BSD-style)
579 * If you set the real gid at all, or set the effective gid to a value not
580 * equal to the real gid, then the saved gid is set to the new effective gid.
582 * This makes it possible for a setgid program to completely drop its
583 * privileges, which is often a useful assertion to make when you are doing
584 * a security audit over a program.
586 * The general idea is that a program which uses just setregid() will be
587 * 100% compatible with BSD. A program which uses just setgid() will be
588 * 100% compatible with POSIX with saved IDs.
590 * SMP: There are not races, the GIDs are checked only by filesystem
591 * operations (as far as semantic preservation is concerned).
593 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
595 struct user_namespace *ns = current_user_ns();
596 const struct cred *old;
601 krgid = make_kgid(ns, rgid);
602 kegid = make_kgid(ns, egid);
604 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
606 if ((egid != (gid_t) -1) && !gid_valid(kegid))
609 new = prepare_creds();
612 old = current_cred();
615 if (rgid != (gid_t) -1) {
616 if (gid_eq(old->gid, krgid) ||
617 gid_eq(old->egid, krgid) ||
618 nsown_capable(CAP_SETGID))
623 if (egid != (gid_t) -1) {
624 if (gid_eq(old->gid, kegid) ||
625 gid_eq(old->egid, kegid) ||
626 gid_eq(old->sgid, kegid) ||
627 nsown_capable(CAP_SETGID))
633 if (rgid != (gid_t) -1 ||
634 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
635 new->sgid = new->egid;
636 new->fsgid = new->egid;
638 return commit_creds(new);
646 * setgid() is implemented like SysV w/ SAVED_IDS
648 * SMP: Same implicit races as above.
650 SYSCALL_DEFINE1(setgid, gid_t, gid)
652 struct user_namespace *ns = current_user_ns();
653 const struct cred *old;
658 kgid = make_kgid(ns, gid);
659 if (!gid_valid(kgid))
662 new = prepare_creds();
665 old = current_cred();
668 if (nsown_capable(CAP_SETGID))
669 new->gid = new->egid = new->sgid = new->fsgid = kgid;
670 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
671 new->egid = new->fsgid = kgid;
675 return commit_creds(new);
683 * change the user struct in a credentials set to match the new UID
685 static int set_user(struct cred *new)
687 struct user_struct *new_user;
689 new_user = alloc_uid(new->uid);
694 * We don't fail in case of NPROC limit excess here because too many
695 * poorly written programs don't check set*uid() return code, assuming
696 * it never fails if called by root. We may still enforce NPROC limit
697 * for programs doing set*uid()+execve() by harmlessly deferring the
698 * failure to the execve() stage.
700 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
701 new_user != INIT_USER)
702 current->flags |= PF_NPROC_EXCEEDED;
704 current->flags &= ~PF_NPROC_EXCEEDED;
707 new->user = new_user;
712 * Unprivileged users may change the real uid to the effective uid
713 * or vice versa. (BSD-style)
715 * If you set the real uid at all, or set the effective uid to a value not
716 * equal to the real uid, then the saved uid is set to the new effective uid.
718 * This makes it possible for a setuid program to completely drop its
719 * privileges, which is often a useful assertion to make when you are doing
720 * a security audit over a program.
722 * The general idea is that a program which uses just setreuid() will be
723 * 100% compatible with BSD. A program which uses just setuid() will be
724 * 100% compatible with POSIX with saved IDs.
726 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
728 struct user_namespace *ns = current_user_ns();
729 const struct cred *old;
734 kruid = make_kuid(ns, ruid);
735 keuid = make_kuid(ns, euid);
737 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
739 if ((euid != (uid_t) -1) && !uid_valid(keuid))
742 new = prepare_creds();
745 old = current_cred();
748 if (ruid != (uid_t) -1) {
750 if (!uid_eq(old->uid, kruid) &&
751 !uid_eq(old->euid, kruid) &&
752 !nsown_capable(CAP_SETUID))
756 if (euid != (uid_t) -1) {
758 if (!uid_eq(old->uid, keuid) &&
759 !uid_eq(old->euid, keuid) &&
760 !uid_eq(old->suid, keuid) &&
761 !nsown_capable(CAP_SETUID))
765 if (!uid_eq(new->uid, old->uid)) {
766 retval = set_user(new);
770 if (ruid != (uid_t) -1 ||
771 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
772 new->suid = new->euid;
773 new->fsuid = new->euid;
775 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
779 return commit_creds(new);
787 * setuid() is implemented like SysV with SAVED_IDS
789 * Note that SAVED_ID's is deficient in that a setuid root program
790 * like sendmail, for example, cannot set its uid to be a normal
791 * user and then switch back, because if you're root, setuid() sets
792 * the saved uid too. If you don't like this, blame the bright people
793 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
794 * will allow a root program to temporarily drop privileges and be able to
795 * regain them by swapping the real and effective uid.
797 SYSCALL_DEFINE1(setuid, uid_t, uid)
799 struct user_namespace *ns = current_user_ns();
800 const struct cred *old;
805 kuid = make_kuid(ns, uid);
806 if (!uid_valid(kuid))
809 new = prepare_creds();
812 old = current_cred();
815 if (nsown_capable(CAP_SETUID)) {
816 new->suid = new->uid = kuid;
817 if (!uid_eq(kuid, old->uid)) {
818 retval = set_user(new);
822 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
826 new->fsuid = new->euid = kuid;
828 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
832 return commit_creds(new);
841 * This function implements a generic ability to update ruid, euid,
842 * and suid. This allows you to implement the 4.4 compatible seteuid().
844 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
846 struct user_namespace *ns = current_user_ns();
847 const struct cred *old;
850 kuid_t kruid, keuid, ksuid;
852 kruid = make_kuid(ns, ruid);
853 keuid = make_kuid(ns, euid);
854 ksuid = make_kuid(ns, suid);
856 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
859 if ((euid != (uid_t) -1) && !uid_valid(keuid))
862 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
865 new = prepare_creds();
869 old = current_cred();
872 if (!nsown_capable(CAP_SETUID)) {
873 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
874 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
876 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
877 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
879 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
880 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
884 if (ruid != (uid_t) -1) {
886 if (!uid_eq(kruid, old->uid)) {
887 retval = set_user(new);
892 if (euid != (uid_t) -1)
894 if (suid != (uid_t) -1)
896 new->fsuid = new->euid;
898 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
902 return commit_creds(new);
909 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
911 const struct cred *cred = current_cred();
913 uid_t ruid, euid, suid;
915 ruid = from_kuid_munged(cred->user_ns, cred->uid);
916 euid = from_kuid_munged(cred->user_ns, cred->euid);
917 suid = from_kuid_munged(cred->user_ns, cred->suid);
919 if (!(retval = put_user(ruid, ruidp)) &&
920 !(retval = put_user(euid, euidp)))
921 retval = put_user(suid, suidp);
927 * Same as above, but for rgid, egid, sgid.
929 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
931 struct user_namespace *ns = current_user_ns();
932 const struct cred *old;
935 kgid_t krgid, kegid, ksgid;
937 krgid = make_kgid(ns, rgid);
938 kegid = make_kgid(ns, egid);
939 ksgid = make_kgid(ns, sgid);
941 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
943 if ((egid != (gid_t) -1) && !gid_valid(kegid))
945 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
948 new = prepare_creds();
951 old = current_cred();
954 if (!nsown_capable(CAP_SETGID)) {
955 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
956 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
958 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
959 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
961 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
962 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
966 if (rgid != (gid_t) -1)
968 if (egid != (gid_t) -1)
970 if (sgid != (gid_t) -1)
972 new->fsgid = new->egid;
974 return commit_creds(new);
981 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
983 const struct cred *cred = current_cred();
985 gid_t rgid, egid, sgid;
987 rgid = from_kgid_munged(cred->user_ns, cred->gid);
988 egid = from_kgid_munged(cred->user_ns, cred->egid);
989 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
991 if (!(retval = put_user(rgid, rgidp)) &&
992 !(retval = put_user(egid, egidp)))
993 retval = put_user(sgid, sgidp);
1000 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1001 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1002 * whatever uid it wants to). It normally shadows "euid", except when
1003 * explicitly set by setfsuid() or for access..
1005 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
1007 const struct cred *old;
1012 old = current_cred();
1013 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
1015 kuid = make_kuid(old->user_ns, uid);
1016 if (!uid_valid(kuid))
1019 new = prepare_creds();
1023 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
1024 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
1025 nsown_capable(CAP_SETUID)) {
1026 if (!uid_eq(kuid, old->fsuid)) {
1028 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
1042 * Samma på svenska..
1044 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
1046 const struct cred *old;
1051 old = current_cred();
1052 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1054 kgid = make_kgid(old->user_ns, gid);
1055 if (!gid_valid(kgid))
1058 new = prepare_creds();
1062 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1063 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
1064 nsown_capable(CAP_SETGID)) {
1065 if (!gid_eq(kgid, old->fsgid)) {
1080 * sys_getpid - return the thread group id of the current process
1082 * Note, despite the name, this returns the tgid not the pid. The tgid and
1083 * the pid are identical unless CLONE_THREAD was specified on clone() in
1084 * which case the tgid is the same in all threads of the same group.
1086 * This is SMP safe as current->tgid does not change.
1088 SYSCALL_DEFINE0(getpid)
1090 return task_tgid_vnr(current);
1093 /* Thread ID - the internal kernel "pid" */
1094 SYSCALL_DEFINE0(gettid)
1096 return task_pid_vnr(current);
1100 * Accessing ->real_parent is not SMP-safe, it could
1101 * change from under us. However, we can use a stale
1102 * value of ->real_parent under rcu_read_lock(), see
1103 * release_task()->call_rcu(delayed_put_task_struct).
1105 SYSCALL_DEFINE0(getppid)
1110 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1116 SYSCALL_DEFINE0(getuid)
1118 /* Only we change this so SMP safe */
1119 return from_kuid_munged(current_user_ns(), current_uid());
1122 SYSCALL_DEFINE0(geteuid)
1124 /* Only we change this so SMP safe */
1125 return from_kuid_munged(current_user_ns(), current_euid());
1128 SYSCALL_DEFINE0(getgid)
1130 /* Only we change this so SMP safe */
1131 return from_kgid_munged(current_user_ns(), current_gid());
1134 SYSCALL_DEFINE0(getegid)
1136 /* Only we change this so SMP safe */
1137 return from_kgid_munged(current_user_ns(), current_egid());
1140 void do_sys_times(struct tms *tms)
1142 cputime_t tgutime, tgstime, cutime, cstime;
1144 spin_lock_irq(¤t->sighand->siglock);
1145 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1146 cutime = current->signal->cutime;
1147 cstime = current->signal->cstime;
1148 spin_unlock_irq(¤t->sighand->siglock);
1149 tms->tms_utime = cputime_to_clock_t(tgutime);
1150 tms->tms_stime = cputime_to_clock_t(tgstime);
1151 tms->tms_cutime = cputime_to_clock_t(cutime);
1152 tms->tms_cstime = cputime_to_clock_t(cstime);
1155 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1161 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1164 force_successful_syscall_return();
1165 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1169 * This needs some heavy checking ...
1170 * I just haven't the stomach for it. I also don't fully
1171 * understand sessions/pgrp etc. Let somebody who does explain it.
1173 * OK, I think I have the protection semantics right.... this is really
1174 * only important on a multi-user system anyway, to make sure one user
1175 * can't send a signal to a process owned by another. -TYT, 12/12/91
1177 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1180 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1182 struct task_struct *p;
1183 struct task_struct *group_leader = current->group_leader;
1188 pid = task_pid_vnr(group_leader);
1195 /* From this point forward we keep holding onto the tasklist lock
1196 * so that our parent does not change from under us. -DaveM
1198 write_lock_irq(&tasklist_lock);
1201 p = find_task_by_vpid(pid);
1206 if (!thread_group_leader(p))
1209 if (same_thread_group(p->real_parent, group_leader)) {
1211 if (task_session(p) != task_session(group_leader))
1218 if (p != group_leader)
1223 if (p->signal->leader)
1228 struct task_struct *g;
1230 pgrp = find_vpid(pgid);
1231 g = pid_task(pgrp, PIDTYPE_PGID);
1232 if (!g || task_session(g) != task_session(group_leader))
1236 err = security_task_setpgid(p, pgid);
1240 if (task_pgrp(p) != pgrp)
1241 change_pid(p, PIDTYPE_PGID, pgrp);
1245 /* All paths lead to here, thus we are safe. -DaveM */
1246 write_unlock_irq(&tasklist_lock);
1251 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1253 struct task_struct *p;
1259 grp = task_pgrp(current);
1262 p = find_task_by_vpid(pid);
1269 retval = security_task_getpgid(p);
1273 retval = pid_vnr(grp);
1279 #ifdef __ARCH_WANT_SYS_GETPGRP
1281 SYSCALL_DEFINE0(getpgrp)
1283 return sys_getpgid(0);
1288 SYSCALL_DEFINE1(getsid, pid_t, pid)
1290 struct task_struct *p;
1296 sid = task_session(current);
1299 p = find_task_by_vpid(pid);
1302 sid = task_session(p);
1306 retval = security_task_getsid(p);
1310 retval = pid_vnr(sid);
1316 SYSCALL_DEFINE0(setsid)
1318 struct task_struct *group_leader = current->group_leader;
1319 struct pid *sid = task_pid(group_leader);
1320 pid_t session = pid_vnr(sid);
1323 write_lock_irq(&tasklist_lock);
1324 /* Fail if I am already a session leader */
1325 if (group_leader->signal->leader)
1328 /* Fail if a process group id already exists that equals the
1329 * proposed session id.
1331 if (pid_task(sid, PIDTYPE_PGID))
1334 group_leader->signal->leader = 1;
1335 __set_special_pids(sid);
1337 proc_clear_tty(group_leader);
1341 write_unlock_irq(&tasklist_lock);
1343 proc_sid_connector(group_leader);
1344 sched_autogroup_create_attach(group_leader);
1349 DECLARE_RWSEM(uts_sem);
1351 #ifdef COMPAT_UTS_MACHINE
1352 #define override_architecture(name) \
1353 (personality(current->personality) == PER_LINUX32 && \
1354 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1355 sizeof(COMPAT_UTS_MACHINE)))
1357 #define override_architecture(name) 0
1361 * Work around broken programs that cannot handle "Linux 3.0".
1362 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1364 static int override_release(char __user *release, size_t len)
1368 if (current->personality & UNAME26) {
1369 const char *rest = UTS_RELEASE;
1370 char buf[65] = { 0 };
1376 if (*rest == '.' && ++ndots >= 3)
1378 if (!isdigit(*rest) && *rest != '.')
1382 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1383 copy = clamp_t(size_t, len, 1, sizeof(buf));
1384 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1385 ret = copy_to_user(release, buf, copy + 1);
1390 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1394 down_read(&uts_sem);
1395 if (copy_to_user(name, utsname(), sizeof *name))
1399 if (!errno && override_release(name->release, sizeof(name->release)))
1401 if (!errno && override_architecture(name))
1406 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1410 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1417 down_read(&uts_sem);
1418 if (copy_to_user(name, utsname(), sizeof(*name)))
1422 if (!error && override_release(name->release, sizeof(name->release)))
1424 if (!error && override_architecture(name))
1429 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1435 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1438 down_read(&uts_sem);
1439 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1441 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1442 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1444 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1445 error |= __copy_to_user(&name->release, &utsname()->release,
1447 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1448 error |= __copy_to_user(&name->version, &utsname()->version,
1450 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1451 error |= __copy_to_user(&name->machine, &utsname()->machine,
1453 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1456 if (!error && override_architecture(name))
1458 if (!error && override_release(name->release, sizeof(name->release)))
1460 return error ? -EFAULT : 0;
1464 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1467 char tmp[__NEW_UTS_LEN];
1469 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1472 if (len < 0 || len > __NEW_UTS_LEN)
1474 down_write(&uts_sem);
1476 if (!copy_from_user(tmp, name, len)) {
1477 struct new_utsname *u = utsname();
1479 memcpy(u->nodename, tmp, len);
1480 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1482 uts_proc_notify(UTS_PROC_HOSTNAME);
1488 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1490 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1493 struct new_utsname *u;
1497 down_read(&uts_sem);
1499 i = 1 + strlen(u->nodename);
1503 if (copy_to_user(name, u->nodename, i))
1512 * Only setdomainname; getdomainname can be implemented by calling
1515 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1518 char tmp[__NEW_UTS_LEN];
1520 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1522 if (len < 0 || len > __NEW_UTS_LEN)
1525 down_write(&uts_sem);
1527 if (!copy_from_user(tmp, name, len)) {
1528 struct new_utsname *u = utsname();
1530 memcpy(u->domainname, tmp, len);
1531 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1533 uts_proc_notify(UTS_PROC_DOMAINNAME);
1539 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1541 struct rlimit value;
1544 ret = do_prlimit(current, resource, NULL, &value);
1546 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1551 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1554 * Back compatibility for getrlimit. Needed for some apps.
1557 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1558 struct rlimit __user *, rlim)
1561 if (resource >= RLIM_NLIMITS)
1564 task_lock(current->group_leader);
1565 x = current->signal->rlim[resource];
1566 task_unlock(current->group_leader);
1567 if (x.rlim_cur > 0x7FFFFFFF)
1568 x.rlim_cur = 0x7FFFFFFF;
1569 if (x.rlim_max > 0x7FFFFFFF)
1570 x.rlim_max = 0x7FFFFFFF;
1571 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1576 static inline bool rlim64_is_infinity(__u64 rlim64)
1578 #if BITS_PER_LONG < 64
1579 return rlim64 >= ULONG_MAX;
1581 return rlim64 == RLIM64_INFINITY;
1585 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1587 if (rlim->rlim_cur == RLIM_INFINITY)
1588 rlim64->rlim_cur = RLIM64_INFINITY;
1590 rlim64->rlim_cur = rlim->rlim_cur;
1591 if (rlim->rlim_max == RLIM_INFINITY)
1592 rlim64->rlim_max = RLIM64_INFINITY;
1594 rlim64->rlim_max = rlim->rlim_max;
1597 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1599 if (rlim64_is_infinity(rlim64->rlim_cur))
1600 rlim->rlim_cur = RLIM_INFINITY;
1602 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1603 if (rlim64_is_infinity(rlim64->rlim_max))
1604 rlim->rlim_max = RLIM_INFINITY;
1606 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1609 /* make sure you are allowed to change @tsk limits before calling this */
1610 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1611 struct rlimit *new_rlim, struct rlimit *old_rlim)
1613 struct rlimit *rlim;
1616 if (resource >= RLIM_NLIMITS)
1619 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1621 if (resource == RLIMIT_NOFILE &&
1622 new_rlim->rlim_max > sysctl_nr_open)
1626 /* protect tsk->signal and tsk->sighand from disappearing */
1627 read_lock(&tasklist_lock);
1628 if (!tsk->sighand) {
1633 rlim = tsk->signal->rlim + resource;
1634 task_lock(tsk->group_leader);
1636 /* Keep the capable check against init_user_ns until
1637 cgroups can contain all limits */
1638 if (new_rlim->rlim_max > rlim->rlim_max &&
1639 !capable(CAP_SYS_RESOURCE))
1642 retval = security_task_setrlimit(tsk->group_leader,
1643 resource, new_rlim);
1644 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1646 * The caller is asking for an immediate RLIMIT_CPU
1647 * expiry. But we use the zero value to mean "it was
1648 * never set". So let's cheat and make it one second
1651 new_rlim->rlim_cur = 1;
1660 task_unlock(tsk->group_leader);
1663 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1664 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1665 * very long-standing error, and fixing it now risks breakage of
1666 * applications, so we live with it
1668 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1669 new_rlim->rlim_cur != RLIM_INFINITY)
1670 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1672 read_unlock(&tasklist_lock);
1676 /* rcu lock must be held */
1677 static int check_prlimit_permission(struct task_struct *task)
1679 const struct cred *cred = current_cred(), *tcred;
1681 if (current == task)
1684 tcred = __task_cred(task);
1685 if (uid_eq(cred->uid, tcred->euid) &&
1686 uid_eq(cred->uid, tcred->suid) &&
1687 uid_eq(cred->uid, tcred->uid) &&
1688 gid_eq(cred->gid, tcred->egid) &&
1689 gid_eq(cred->gid, tcred->sgid) &&
1690 gid_eq(cred->gid, tcred->gid))
1692 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1698 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1699 const struct rlimit64 __user *, new_rlim,
1700 struct rlimit64 __user *, old_rlim)
1702 struct rlimit64 old64, new64;
1703 struct rlimit old, new;
1704 struct task_struct *tsk;
1708 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1710 rlim64_to_rlim(&new64, &new);
1714 tsk = pid ? find_task_by_vpid(pid) : current;
1719 ret = check_prlimit_permission(tsk);
1724 get_task_struct(tsk);
1727 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1728 old_rlim ? &old : NULL);
1730 if (!ret && old_rlim) {
1731 rlim_to_rlim64(&old, &old64);
1732 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1736 put_task_struct(tsk);
1740 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1742 struct rlimit new_rlim;
1744 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1746 return do_prlimit(current, resource, &new_rlim, NULL);
1750 * It would make sense to put struct rusage in the task_struct,
1751 * except that would make the task_struct be *really big*. After
1752 * task_struct gets moved into malloc'ed memory, it would
1753 * make sense to do this. It will make moving the rest of the information
1754 * a lot simpler! (Which we're not doing right now because we're not
1755 * measuring them yet).
1757 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1758 * races with threads incrementing their own counters. But since word
1759 * reads are atomic, we either get new values or old values and we don't
1760 * care which for the sums. We always take the siglock to protect reading
1761 * the c* fields from p->signal from races with exit.c updating those
1762 * fields when reaping, so a sample either gets all the additions of a
1763 * given child after it's reaped, or none so this sample is before reaping.
1766 * We need to take the siglock for CHILDEREN, SELF and BOTH
1767 * for the cases current multithreaded, non-current single threaded
1768 * non-current multithreaded. Thread traversal is now safe with
1770 * Strictly speaking, we donot need to take the siglock if we are current and
1771 * single threaded, as no one else can take our signal_struct away, no one
1772 * else can reap the children to update signal->c* counters, and no one else
1773 * can race with the signal-> fields. If we do not take any lock, the
1774 * signal-> fields could be read out of order while another thread was just
1775 * exiting. So we should place a read memory barrier when we avoid the lock.
1776 * On the writer side, write memory barrier is implied in __exit_signal
1777 * as __exit_signal releases the siglock spinlock after updating the signal->
1778 * fields. But we don't do this yet to keep things simple.
1782 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1784 r->ru_nvcsw += t->nvcsw;
1785 r->ru_nivcsw += t->nivcsw;
1786 r->ru_minflt += t->min_flt;
1787 r->ru_majflt += t->maj_flt;
1788 r->ru_inblock += task_io_get_inblock(t);
1789 r->ru_oublock += task_io_get_oublock(t);
1792 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1794 struct task_struct *t;
1795 unsigned long flags;
1796 cputime_t tgutime, tgstime, utime, stime;
1797 unsigned long maxrss = 0;
1799 memset((char *) r, 0, sizeof *r);
1802 if (who == RUSAGE_THREAD) {
1803 task_cputime_adjusted(current, &utime, &stime);
1804 accumulate_thread_rusage(p, r);
1805 maxrss = p->signal->maxrss;
1809 if (!lock_task_sighand(p, &flags))
1814 case RUSAGE_CHILDREN:
1815 utime = p->signal->cutime;
1816 stime = p->signal->cstime;
1817 r->ru_nvcsw = p->signal->cnvcsw;
1818 r->ru_nivcsw = p->signal->cnivcsw;
1819 r->ru_minflt = p->signal->cmin_flt;
1820 r->ru_majflt = p->signal->cmaj_flt;
1821 r->ru_inblock = p->signal->cinblock;
1822 r->ru_oublock = p->signal->coublock;
1823 maxrss = p->signal->cmaxrss;
1825 if (who == RUSAGE_CHILDREN)
1829 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1832 r->ru_nvcsw += p->signal->nvcsw;
1833 r->ru_nivcsw += p->signal->nivcsw;
1834 r->ru_minflt += p->signal->min_flt;
1835 r->ru_majflt += p->signal->maj_flt;
1836 r->ru_inblock += p->signal->inblock;
1837 r->ru_oublock += p->signal->oublock;
1838 if (maxrss < p->signal->maxrss)
1839 maxrss = p->signal->maxrss;
1842 accumulate_thread_rusage(t, r);
1850 unlock_task_sighand(p, &flags);
1853 cputime_to_timeval(utime, &r->ru_utime);
1854 cputime_to_timeval(stime, &r->ru_stime);
1856 if (who != RUSAGE_CHILDREN) {
1857 struct mm_struct *mm = get_task_mm(p);
1859 setmax_mm_hiwater_rss(&maxrss, mm);
1863 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1866 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1869 k_getrusage(p, who, &r);
1870 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1873 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1875 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1876 who != RUSAGE_THREAD)
1878 return getrusage(current, who, ru);
1881 #ifdef CONFIG_COMPAT
1882 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1886 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1887 who != RUSAGE_THREAD)
1890 k_getrusage(current, who, &r);
1891 return put_compat_rusage(&r, ru);
1895 SYSCALL_DEFINE1(umask, int, mask)
1897 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1901 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1904 struct inode *inode;
1911 inode = file_inode(exe.file);
1914 * Because the original mm->exe_file points to executable file, make
1915 * sure that this one is executable as well, to avoid breaking an
1919 if (!S_ISREG(inode->i_mode) ||
1920 exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1923 err = inode_permission(inode, MAY_EXEC);
1927 down_write(&mm->mmap_sem);
1930 * Forbid mm->exe_file change if old file still mapped.
1934 struct vm_area_struct *vma;
1936 for (vma = mm->mmap; vma; vma = vma->vm_next)
1938 path_equal(&vma->vm_file->f_path,
1939 &mm->exe_file->f_path))
1944 * The symlink can be changed only once, just to disallow arbitrary
1945 * transitions malicious software might bring in. This means one
1946 * could make a snapshot over all processes running and monitor
1947 * /proc/pid/exe changes to notice unusual activity if needed.
1950 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1954 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
1956 up_write(&mm->mmap_sem);
1963 static int prctl_set_mm(int opt, unsigned long addr,
1964 unsigned long arg4, unsigned long arg5)
1966 unsigned long rlim = rlimit(RLIMIT_DATA);
1967 struct mm_struct *mm = current->mm;
1968 struct vm_area_struct *vma;
1971 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
1974 if (!capable(CAP_SYS_RESOURCE))
1977 if (opt == PR_SET_MM_EXE_FILE)
1978 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1980 if (addr >= TASK_SIZE || addr < mmap_min_addr)
1985 down_read(&mm->mmap_sem);
1986 vma = find_vma(mm, addr);
1989 case PR_SET_MM_START_CODE:
1990 mm->start_code = addr;
1992 case PR_SET_MM_END_CODE:
1993 mm->end_code = addr;
1995 case PR_SET_MM_START_DATA:
1996 mm->start_data = addr;
1998 case PR_SET_MM_END_DATA:
1999 mm->end_data = addr;
2002 case PR_SET_MM_START_BRK:
2003 if (addr <= mm->end_data)
2006 if (rlim < RLIM_INFINITY &&
2008 (mm->end_data - mm->start_data) > rlim)
2011 mm->start_brk = addr;
2015 if (addr <= mm->end_data)
2018 if (rlim < RLIM_INFINITY &&
2019 (addr - mm->start_brk) +
2020 (mm->end_data - mm->start_data) > rlim)
2027 * If command line arguments and environment
2028 * are placed somewhere else on stack, we can
2029 * set them up here, ARG_START/END to setup
2030 * command line argumets and ENV_START/END
2033 case PR_SET_MM_START_STACK:
2034 case PR_SET_MM_ARG_START:
2035 case PR_SET_MM_ARG_END:
2036 case PR_SET_MM_ENV_START:
2037 case PR_SET_MM_ENV_END:
2042 if (opt == PR_SET_MM_START_STACK)
2043 mm->start_stack = addr;
2044 else if (opt == PR_SET_MM_ARG_START)
2045 mm->arg_start = addr;
2046 else if (opt == PR_SET_MM_ARG_END)
2048 else if (opt == PR_SET_MM_ENV_START)
2049 mm->env_start = addr;
2050 else if (opt == PR_SET_MM_ENV_END)
2055 * This doesn't move auxiliary vector itself
2056 * since it's pinned to mm_struct, but allow
2057 * to fill vector with new values. It's up
2058 * to a caller to provide sane values here
2059 * otherwise user space tools which use this
2060 * vector might be unhappy.
2062 case PR_SET_MM_AUXV: {
2063 unsigned long user_auxv[AT_VECTOR_SIZE];
2065 if (arg4 > sizeof(user_auxv))
2067 up_read(&mm->mmap_sem);
2069 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
2072 /* Make sure the last entry is always AT_NULL */
2073 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2074 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2076 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2079 memcpy(mm->saved_auxv, user_auxv, arg4);
2080 task_unlock(current);
2090 up_read(&mm->mmap_sem);
2094 #ifdef CONFIG_CHECKPOINT_RESTORE
2095 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2097 return put_user(me->clear_child_tid, tid_addr);
2100 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2107 static int prctl_update_vma_anon_name(struct vm_area_struct *vma,
2108 struct vm_area_struct **prev,
2109 unsigned long start, unsigned long end,
2110 const char __user *name_addr)
2112 struct mm_struct * mm = vma->vm_mm;
2116 if (name_addr == vma_get_anon_name(vma)) {
2121 pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
2122 *prev = vma_merge(mm, *prev, start, end, vma->vm_flags, vma->anon_vma,
2123 vma->vm_file, pgoff, vma_policy(vma),
2132 if (start != vma->vm_start) {
2133 error = split_vma(mm, vma, start, 1);
2138 if (end != vma->vm_end) {
2139 error = split_vma(mm, vma, end, 0);
2146 vma->shared.anon_name = name_addr;
2149 if (error == -ENOMEM)
2154 static int prctl_set_vma_anon_name(unsigned long start, unsigned long end,
2158 struct vm_area_struct * vma, *prev;
2159 int unmapped_error = 0;
2160 int error = -EINVAL;
2163 * If the interval [start,end) covers some unmapped address
2164 * ranges, just ignore them, but return -ENOMEM at the end.
2165 * - this matches the handling in madvise.
2167 vma = find_vma_prev(current->mm, start, &prev);
2168 if (vma && start > vma->vm_start)
2172 /* Still start < end. */
2177 /* Here start < (end|vma->vm_end). */
2178 if (start < vma->vm_start) {
2179 unmapped_error = -ENOMEM;
2180 start = vma->vm_start;
2185 /* Here vma->vm_start <= start < (end|vma->vm_end) */
2190 /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */
2191 error = prctl_update_vma_anon_name(vma, &prev, start, tmp,
2192 (const char __user *)arg);
2196 if (prev && start < prev->vm_end)
2197 start = prev->vm_end;
2198 error = unmapped_error;
2202 vma = prev->vm_next;
2203 else /* madvise_remove dropped mmap_sem */
2204 vma = find_vma(current->mm, start);
2208 static int prctl_set_vma(unsigned long opt, unsigned long start,
2209 unsigned long len_in, unsigned long arg)
2211 struct mm_struct *mm = current->mm;
2216 if (start & ~PAGE_MASK)
2218 len = (len_in + ~PAGE_MASK) & PAGE_MASK;
2220 /* Check to see whether len was rounded up from small -ve to zero */
2231 down_write(&mm->mmap_sem);
2234 case PR_SET_VMA_ANON_NAME:
2235 error = prctl_set_vma_anon_name(start, end, arg);
2241 up_write(&mm->mmap_sem);
2245 #else /* CONFIG_MMU */
2246 static int prctl_set_vma(unsigned long opt, unsigned long start,
2247 unsigned long len_in, unsigned long arg)
2253 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2254 unsigned long, arg4, unsigned long, arg5)
2256 struct task_struct *me = current;
2257 struct task_struct *tsk;
2258 unsigned char comm[sizeof(me->comm)];
2261 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2262 if (error != -ENOSYS)
2267 case PR_SET_PDEATHSIG:
2268 if (!valid_signal(arg2)) {
2272 me->pdeath_signal = arg2;
2274 case PR_GET_PDEATHSIG:
2275 error = put_user(me->pdeath_signal, (int __user *)arg2);
2277 case PR_GET_DUMPABLE:
2278 error = get_dumpable(me->mm);
2280 case PR_SET_DUMPABLE:
2281 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2285 set_dumpable(me->mm, arg2);
2288 case PR_SET_UNALIGN:
2289 error = SET_UNALIGN_CTL(me, arg2);
2291 case PR_GET_UNALIGN:
2292 error = GET_UNALIGN_CTL(me, arg2);
2295 error = SET_FPEMU_CTL(me, arg2);
2298 error = GET_FPEMU_CTL(me, arg2);
2301 error = SET_FPEXC_CTL(me, arg2);
2304 error = GET_FPEXC_CTL(me, arg2);
2307 error = PR_TIMING_STATISTICAL;
2310 if (arg2 != PR_TIMING_STATISTICAL)
2314 comm[sizeof(me->comm) - 1] = 0;
2315 if (strncpy_from_user(comm, (char __user *)arg2,
2316 sizeof(me->comm) - 1) < 0)
2318 set_task_comm(me, comm);
2319 proc_comm_connector(me);
2322 get_task_comm(comm, me);
2323 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2327 error = GET_ENDIAN(me, arg2);
2330 error = SET_ENDIAN(me, arg2);
2332 case PR_GET_SECCOMP:
2333 error = prctl_get_seccomp();
2335 case PR_SET_SECCOMP:
2336 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2339 error = GET_TSC_CTL(arg2);
2342 error = SET_TSC_CTL(arg2);
2344 case PR_TASK_PERF_EVENTS_DISABLE:
2345 error = perf_event_task_disable();
2347 case PR_TASK_PERF_EVENTS_ENABLE:
2348 error = perf_event_task_enable();
2350 case PR_GET_TIMERSLACK:
2351 error = current->timer_slack_ns;
2353 case PR_SET_TIMERSLACK:
2355 current->timer_slack_ns =
2356 current->default_timer_slack_ns;
2358 current->timer_slack_ns = arg2;
2364 case PR_MCE_KILL_CLEAR:
2367 current->flags &= ~PF_MCE_PROCESS;
2369 case PR_MCE_KILL_SET:
2370 current->flags |= PF_MCE_PROCESS;
2371 if (arg3 == PR_MCE_KILL_EARLY)
2372 current->flags |= PF_MCE_EARLY;
2373 else if (arg3 == PR_MCE_KILL_LATE)
2374 current->flags &= ~PF_MCE_EARLY;
2375 else if (arg3 == PR_MCE_KILL_DEFAULT)
2377 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2381 case PR_SET_TIMERSLACK_PID:
2382 if (current->pid != (pid_t)arg3 &&
2383 !capable(CAP_SYS_NICE))
2386 tsk = find_task_by_pid_ns((pid_t)arg3, &init_pid_ns);
2391 get_task_struct(tsk);
2394 tsk->timer_slack_ns =
2395 tsk->default_timer_slack_ns;
2397 tsk->timer_slack_ns = arg2;
2398 put_task_struct(tsk);
2405 case PR_MCE_KILL_GET:
2406 if (arg2 | arg3 | arg4 | arg5)
2408 if (current->flags & PF_MCE_PROCESS)
2409 error = (current->flags & PF_MCE_EARLY) ?
2410 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2412 error = PR_MCE_KILL_DEFAULT;
2415 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2417 case PR_GET_TID_ADDRESS:
2418 error = prctl_get_tid_address(me, (int __user **)arg2);
2420 case PR_SET_CHILD_SUBREAPER:
2421 me->signal->is_child_subreaper = !!arg2;
2423 case PR_GET_CHILD_SUBREAPER:
2424 error = put_user(me->signal->is_child_subreaper,
2425 (int __user *)arg2);
2427 case PR_SET_NO_NEW_PRIVS:
2428 if (arg2 != 1 || arg3 || arg4 || arg5)
2431 task_set_no_new_privs(current);
2433 case PR_GET_NO_NEW_PRIVS:
2434 if (arg2 || arg3 || arg4 || arg5)
2436 return task_no_new_privs(current) ? 1 : 0;
2438 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2447 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2448 struct getcpu_cache __user *, unused)
2451 int cpu = raw_smp_processor_id();
2453 err |= put_user(cpu, cpup);
2455 err |= put_user(cpu_to_node(cpu), nodep);
2456 return err ? -EFAULT : 0;
2459 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2461 static int __orderly_poweroff(bool force)
2464 static char *envp[] = {
2466 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2471 argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL);
2473 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
2476 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2477 __func__, poweroff_cmd);
2482 printk(KERN_WARNING "Failed to start orderly shutdown: "
2483 "forcing the issue\n");
2485 * I guess this should try to kick off some daemon to sync and
2486 * poweroff asap. Or not even bother syncing if we're doing an
2487 * emergency shutdown?
2496 static bool poweroff_force;
2498 static void poweroff_work_func(struct work_struct *work)
2500 __orderly_poweroff(poweroff_force);
2503 static DECLARE_WORK(poweroff_work, poweroff_work_func);
2506 * orderly_poweroff - Trigger an orderly system poweroff
2507 * @force: force poweroff if command execution fails
2509 * This may be called from any context to trigger a system shutdown.
2510 * If the orderly shutdown fails, it will force an immediate shutdown.
2512 int orderly_poweroff(bool force)
2514 if (force) /* do not override the pending "true" */
2515 poweroff_force = true;
2516 schedule_work(&poweroff_work);
2519 EXPORT_SYMBOL_GPL(orderly_poweroff);
2522 * do_sysinfo - fill in sysinfo struct
2523 * @info: pointer to buffer to fill
2525 static int do_sysinfo(struct sysinfo *info)
2527 unsigned long mem_total, sav_total;
2528 unsigned int mem_unit, bitcount;
2531 memset(info, 0, sizeof(struct sysinfo));
2534 monotonic_to_bootbased(&tp);
2535 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2537 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2539 info->procs = nr_threads;
2545 * If the sum of all the available memory (i.e. ram + swap)
2546 * is less than can be stored in a 32 bit unsigned long then
2547 * we can be binary compatible with 2.2.x kernels. If not,
2548 * well, in that case 2.2.x was broken anyways...
2550 * -Erik Andersen <andersee@debian.org>
2553 mem_total = info->totalram + info->totalswap;
2554 if (mem_total < info->totalram || mem_total < info->totalswap)
2557 mem_unit = info->mem_unit;
2558 while (mem_unit > 1) {
2561 sav_total = mem_total;
2563 if (mem_total < sav_total)
2568 * If mem_total did not overflow, multiply all memory values by
2569 * info->mem_unit and set it to 1. This leaves things compatible
2570 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2575 info->totalram <<= bitcount;
2576 info->freeram <<= bitcount;
2577 info->sharedram <<= bitcount;
2578 info->bufferram <<= bitcount;
2579 info->totalswap <<= bitcount;
2580 info->freeswap <<= bitcount;
2581 info->totalhigh <<= bitcount;
2582 info->freehigh <<= bitcount;
2588 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2594 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2600 #ifdef CONFIG_COMPAT
2601 struct compat_sysinfo {
2615 char _f[20-2*sizeof(u32)-sizeof(int)];
2618 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2624 /* Check to see if any memory value is too large for 32-bit and scale
2627 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2630 while (s.mem_unit < PAGE_SIZE) {
2635 s.totalram >>= bitcount;
2636 s.freeram >>= bitcount;
2637 s.sharedram >>= bitcount;
2638 s.bufferram >>= bitcount;
2639 s.totalswap >>= bitcount;
2640 s.freeswap >>= bitcount;
2641 s.totalhigh >>= bitcount;
2642 s.freehigh >>= bitcount;
2645 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2646 __put_user(s.uptime, &info->uptime) ||
2647 __put_user(s.loads[0], &info->loads[0]) ||
2648 __put_user(s.loads[1], &info->loads[1]) ||
2649 __put_user(s.loads[2], &info->loads[2]) ||
2650 __put_user(s.totalram, &info->totalram) ||
2651 __put_user(s.freeram, &info->freeram) ||
2652 __put_user(s.sharedram, &info->sharedram) ||
2653 __put_user(s.bufferram, &info->bufferram) ||
2654 __put_user(s.totalswap, &info->totalswap) ||
2655 __put_user(s.freeswap, &info->freeswap) ||
2656 __put_user(s.procs, &info->procs) ||
2657 __put_user(s.totalhigh, &info->totalhigh) ||
2658 __put_user(s.freehigh, &info->freehigh) ||
2659 __put_user(s.mem_unit, &info->mem_unit))
2664 #endif /* CONFIG_COMPAT */