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>
67 #ifdef CONFIG_ARCH_ROCKCHIP
68 #include <asm/system_misc.h>
71 #ifndef SET_UNALIGN_CTL
72 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
74 #ifndef GET_UNALIGN_CTL
75 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
78 # define SET_FPEMU_CTL(a,b) (-EINVAL)
81 # define GET_FPEMU_CTL(a,b) (-EINVAL)
84 # define SET_FPEXC_CTL(a,b) (-EINVAL)
87 # define GET_FPEXC_CTL(a,b) (-EINVAL)
90 # define GET_ENDIAN(a,b) (-EINVAL)
93 # define SET_ENDIAN(a,b) (-EINVAL)
96 # define GET_TSC_CTL(a) (-EINVAL)
99 # define SET_TSC_CTL(a) (-EINVAL)
103 * this is where the system-wide overflow UID and GID are defined, for
104 * architectures that now have 32-bit UID/GID but didn't in the past
107 int overflowuid = DEFAULT_OVERFLOWUID;
108 int overflowgid = DEFAULT_OVERFLOWGID;
110 EXPORT_SYMBOL(overflowuid);
111 EXPORT_SYMBOL(overflowgid);
114 * the same as above, but for filesystems which can only store a 16-bit
115 * UID and GID. as such, this is needed on all architectures
118 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
119 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
121 EXPORT_SYMBOL(fs_overflowuid);
122 EXPORT_SYMBOL(fs_overflowgid);
125 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
130 EXPORT_SYMBOL(cad_pid);
133 * If set, this is used for preparing the system to power off.
136 void (*pm_power_off_prepare)(void);
139 * Returns true if current's euid is same as p's uid or euid,
140 * or has CAP_SYS_NICE to p's user_ns.
142 * Called with rcu_read_lock, creds are safe
144 static bool set_one_prio_perm(struct task_struct *p)
146 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
148 if (uid_eq(pcred->uid, cred->euid) ||
149 uid_eq(pcred->euid, cred->euid))
151 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
157 * set the priority of a task
158 * - the caller must hold the RCU read lock
160 static int set_one_prio(struct task_struct *p, int niceval, int error)
164 if (!set_one_prio_perm(p)) {
168 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
172 no_nice = security_task_setnice(p, niceval);
179 set_user_nice(p, niceval);
184 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
186 struct task_struct *g, *p;
187 struct user_struct *user;
188 const struct cred *cred = current_cred();
193 if (which > PRIO_USER || which < PRIO_PROCESS)
196 /* normalize: avoid signed division (rounding problems) */
204 read_lock(&tasklist_lock);
208 p = find_task_by_vpid(who);
212 error = set_one_prio(p, niceval, error);
216 pgrp = find_vpid(who);
218 pgrp = task_pgrp(current);
219 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
220 error = set_one_prio(p, niceval, error);
221 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
224 uid = make_kuid(cred->user_ns, who);
228 else if (!uid_eq(uid, cred->uid) &&
229 !(user = find_user(uid)))
230 goto out_unlock; /* No processes for this user */
232 do_each_thread(g, p) {
233 if (uid_eq(task_uid(p), uid))
234 error = set_one_prio(p, niceval, error);
235 } while_each_thread(g, p);
236 if (!uid_eq(uid, cred->uid))
237 free_uid(user); /* For find_user() */
241 read_unlock(&tasklist_lock);
248 * Ugh. To avoid negative return values, "getpriority()" will
249 * not return the normal nice-value, but a negated value that
250 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
251 * to stay compatible.
253 SYSCALL_DEFINE2(getpriority, int, which, int, who)
255 struct task_struct *g, *p;
256 struct user_struct *user;
257 const struct cred *cred = current_cred();
258 long niceval, retval = -ESRCH;
262 if (which > PRIO_USER || which < PRIO_PROCESS)
266 read_lock(&tasklist_lock);
270 p = find_task_by_vpid(who);
274 niceval = 20 - task_nice(p);
275 if (niceval > retval)
281 pgrp = find_vpid(who);
283 pgrp = task_pgrp(current);
284 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
285 niceval = 20 - task_nice(p);
286 if (niceval > retval)
288 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
291 uid = make_kuid(cred->user_ns, who);
295 else if (!uid_eq(uid, cred->uid) &&
296 !(user = find_user(uid)))
297 goto out_unlock; /* No processes for this user */
299 do_each_thread(g, p) {
300 if (uid_eq(task_uid(p), uid)) {
301 niceval = 20 - task_nice(p);
302 if (niceval > retval)
305 } while_each_thread(g, p);
306 if (!uid_eq(uid, cred->uid))
307 free_uid(user); /* for find_user() */
311 read_unlock(&tasklist_lock);
318 * emergency_restart - reboot the system
320 * Without shutting down any hardware or taking any locks
321 * reboot the system. This is called when we know we are in
322 * trouble so this is our best effort to reboot. This is
323 * safe to call in interrupt context.
325 void emergency_restart(void)
327 kmsg_dump(KMSG_DUMP_EMERG);
328 machine_emergency_restart();
330 EXPORT_SYMBOL_GPL(emergency_restart);
332 void kernel_restart_prepare(char *cmd)
334 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
335 system_state = SYSTEM_RESTART;
336 usermodehelper_disable();
341 * register_reboot_notifier - Register function to be called at reboot time
342 * @nb: Info about notifier function to be called
344 * Registers a function with the list of functions
345 * to be called at reboot time.
347 * Currently always returns zero, as blocking_notifier_chain_register()
348 * always returns zero.
350 int register_reboot_notifier(struct notifier_block *nb)
352 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
354 EXPORT_SYMBOL(register_reboot_notifier);
357 * unregister_reboot_notifier - Unregister previously registered reboot notifier
358 * @nb: Hook to be unregistered
360 * Unregisters a previously registered reboot
363 * Returns zero on success, or %-ENOENT on failure.
365 int unregister_reboot_notifier(struct notifier_block *nb)
367 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
369 EXPORT_SYMBOL(unregister_reboot_notifier);
371 /* Add backwards compatibility for stable trees. */
372 #ifndef PF_NO_SETAFFINITY
373 #define PF_NO_SETAFFINITY PF_THREAD_BOUND
376 static void migrate_to_reboot_cpu(void)
378 /* The boot cpu is always logical cpu 0 */
381 cpu_hotplug_disable();
383 /* Make certain the cpu I'm about to reboot on is online */
384 if (!cpu_online(cpu))
385 cpu = cpumask_first(cpu_online_mask);
387 /* Prevent races with other tasks migrating this task */
388 current->flags |= PF_NO_SETAFFINITY;
390 /* Make certain I only run on the appropriate processor */
391 set_cpus_allowed_ptr(current, cpumask_of(cpu));
395 * kernel_restart - reboot the system
396 * @cmd: pointer to buffer containing command to execute for restart
399 * Shutdown everything and perform a clean reboot.
400 * This is not safe to call in interrupt context.
402 void kernel_restart(char *cmd)
404 kernel_restart_prepare(cmd);
405 migrate_to_reboot_cpu();
408 printk(KERN_EMERG "Restarting system.\n");
410 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
411 kmsg_dump(KMSG_DUMP_RESTART);
412 machine_restart(cmd);
414 EXPORT_SYMBOL_GPL(kernel_restart);
416 static void kernel_shutdown_prepare(enum system_states state)
418 blocking_notifier_call_chain(&reboot_notifier_list,
419 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
420 system_state = state;
421 usermodehelper_disable();
425 * kernel_halt - halt the system
427 * Shutdown everything and perform a clean system halt.
429 void kernel_halt(void)
431 kernel_shutdown_prepare(SYSTEM_HALT);
432 migrate_to_reboot_cpu();
434 printk(KERN_EMERG "System halted.\n");
435 kmsg_dump(KMSG_DUMP_HALT);
439 EXPORT_SYMBOL_GPL(kernel_halt);
442 * kernel_power_off - power_off the system
444 * Shutdown everything and perform a clean system power_off.
446 void kernel_power_off(void)
448 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
449 if (pm_power_off_prepare)
450 pm_power_off_prepare();
451 migrate_to_reboot_cpu();
453 printk(KERN_EMERG "Power down.\n");
454 kmsg_dump(KMSG_DUMP_POWEROFF);
456 #ifdef CONFIG_ARCH_ROCKCHIP
457 arm_pm_restart('h', "charge");
460 EXPORT_SYMBOL_GPL(kernel_power_off);
462 static DEFINE_MUTEX(reboot_mutex);
465 * Reboot system call: for obvious reasons only root may call it,
466 * and even root needs to set up some magic numbers in the registers
467 * so that some mistake won't make this reboot the whole machine.
468 * You can also set the meaning of the ctrl-alt-del-key here.
470 * reboot doesn't sync: do that yourself before calling this.
472 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
475 struct pid_namespace *pid_ns = task_active_pid_ns(current);
479 /* We only trust the superuser with rebooting the system. */
480 if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
483 /* For safety, we require "magic" arguments. */
484 if (magic1 != LINUX_REBOOT_MAGIC1 ||
485 (magic2 != LINUX_REBOOT_MAGIC2 &&
486 magic2 != LINUX_REBOOT_MAGIC2A &&
487 magic2 != LINUX_REBOOT_MAGIC2B &&
488 magic2 != LINUX_REBOOT_MAGIC2C))
492 * If pid namespaces are enabled and the current task is in a child
493 * pid_namespace, the command is handled by reboot_pid_ns() which will
496 ret = reboot_pid_ns(pid_ns, cmd);
500 /* Instead of trying to make the power_off code look like
501 * halt when pm_power_off is not set do it the easy way.
503 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
504 cmd = LINUX_REBOOT_CMD_HALT;
506 mutex_lock(&reboot_mutex);
508 case LINUX_REBOOT_CMD_RESTART:
509 kernel_restart(NULL);
512 case LINUX_REBOOT_CMD_CAD_ON:
516 case LINUX_REBOOT_CMD_CAD_OFF:
520 case LINUX_REBOOT_CMD_HALT:
523 panic("cannot halt");
525 case LINUX_REBOOT_CMD_POWER_OFF:
530 case LINUX_REBOOT_CMD_RESTART2:
531 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
535 buffer[sizeof(buffer) - 1] = '\0';
537 kernel_restart(buffer);
541 case LINUX_REBOOT_CMD_KEXEC:
542 ret = kernel_kexec();
546 #ifdef CONFIG_HIBERNATION
547 case LINUX_REBOOT_CMD_SW_SUSPEND:
556 mutex_unlock(&reboot_mutex);
560 static void deferred_cad(struct work_struct *dummy)
562 kernel_restart(NULL);
566 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
567 * As it's called within an interrupt, it may NOT sync: the only choice
568 * is whether to reboot at once, or just ignore the ctrl-alt-del.
570 void ctrl_alt_del(void)
572 static DECLARE_WORK(cad_work, deferred_cad);
575 schedule_work(&cad_work);
577 kill_cad_pid(SIGINT, 1);
581 * Unprivileged users may change the real gid to the effective gid
582 * or vice versa. (BSD-style)
584 * If you set the real gid at all, or set the effective gid to a value not
585 * equal to the real gid, then the saved gid is set to the new effective gid.
587 * This makes it possible for a setgid program to completely drop its
588 * privileges, which is often a useful assertion to make when you are doing
589 * a security audit over a program.
591 * The general idea is that a program which uses just setregid() will be
592 * 100% compatible with BSD. A program which uses just setgid() will be
593 * 100% compatible with POSIX with saved IDs.
595 * SMP: There are not races, the GIDs are checked only by filesystem
596 * operations (as far as semantic preservation is concerned).
598 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
600 struct user_namespace *ns = current_user_ns();
601 const struct cred *old;
606 krgid = make_kgid(ns, rgid);
607 kegid = make_kgid(ns, egid);
609 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
611 if ((egid != (gid_t) -1) && !gid_valid(kegid))
614 new = prepare_creds();
617 old = current_cred();
620 if (rgid != (gid_t) -1) {
621 if (gid_eq(old->gid, krgid) ||
622 gid_eq(old->egid, krgid) ||
623 nsown_capable(CAP_SETGID))
628 if (egid != (gid_t) -1) {
629 if (gid_eq(old->gid, kegid) ||
630 gid_eq(old->egid, kegid) ||
631 gid_eq(old->sgid, kegid) ||
632 nsown_capable(CAP_SETGID))
638 if (rgid != (gid_t) -1 ||
639 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
640 new->sgid = new->egid;
641 new->fsgid = new->egid;
643 return commit_creds(new);
651 * setgid() is implemented like SysV w/ SAVED_IDS
653 * SMP: Same implicit races as above.
655 SYSCALL_DEFINE1(setgid, gid_t, gid)
657 struct user_namespace *ns = current_user_ns();
658 const struct cred *old;
663 kgid = make_kgid(ns, gid);
664 if (!gid_valid(kgid))
667 new = prepare_creds();
670 old = current_cred();
673 if (nsown_capable(CAP_SETGID))
674 new->gid = new->egid = new->sgid = new->fsgid = kgid;
675 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
676 new->egid = new->fsgid = kgid;
680 return commit_creds(new);
688 * change the user struct in a credentials set to match the new UID
690 static int set_user(struct cred *new)
692 struct user_struct *new_user;
694 new_user = alloc_uid(new->uid);
699 * We don't fail in case of NPROC limit excess here because too many
700 * poorly written programs don't check set*uid() return code, assuming
701 * it never fails if called by root. We may still enforce NPROC limit
702 * for programs doing set*uid()+execve() by harmlessly deferring the
703 * failure to the execve() stage.
705 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
706 new_user != INIT_USER)
707 current->flags |= PF_NPROC_EXCEEDED;
709 current->flags &= ~PF_NPROC_EXCEEDED;
712 new->user = new_user;
717 * Unprivileged users may change the real uid to the effective uid
718 * or vice versa. (BSD-style)
720 * If you set the real uid at all, or set the effective uid to a value not
721 * equal to the real uid, then the saved uid is set to the new effective uid.
723 * This makes it possible for a setuid program to completely drop its
724 * privileges, which is often a useful assertion to make when you are doing
725 * a security audit over a program.
727 * The general idea is that a program which uses just setreuid() will be
728 * 100% compatible with BSD. A program which uses just setuid() will be
729 * 100% compatible with POSIX with saved IDs.
731 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
733 struct user_namespace *ns = current_user_ns();
734 const struct cred *old;
739 kruid = make_kuid(ns, ruid);
740 keuid = make_kuid(ns, euid);
742 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
744 if ((euid != (uid_t) -1) && !uid_valid(keuid))
747 new = prepare_creds();
750 old = current_cred();
753 if (ruid != (uid_t) -1) {
755 if (!uid_eq(old->uid, kruid) &&
756 !uid_eq(old->euid, kruid) &&
757 !nsown_capable(CAP_SETUID))
761 if (euid != (uid_t) -1) {
763 if (!uid_eq(old->uid, keuid) &&
764 !uid_eq(old->euid, keuid) &&
765 !uid_eq(old->suid, keuid) &&
766 !nsown_capable(CAP_SETUID))
770 if (!uid_eq(new->uid, old->uid)) {
771 retval = set_user(new);
775 if (ruid != (uid_t) -1 ||
776 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
777 new->suid = new->euid;
778 new->fsuid = new->euid;
780 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
784 return commit_creds(new);
792 * setuid() is implemented like SysV with SAVED_IDS
794 * Note that SAVED_ID's is deficient in that a setuid root program
795 * like sendmail, for example, cannot set its uid to be a normal
796 * user and then switch back, because if you're root, setuid() sets
797 * the saved uid too. If you don't like this, blame the bright people
798 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
799 * will allow a root program to temporarily drop privileges and be able to
800 * regain them by swapping the real and effective uid.
802 SYSCALL_DEFINE1(setuid, uid_t, uid)
804 struct user_namespace *ns = current_user_ns();
805 const struct cred *old;
810 kuid = make_kuid(ns, uid);
811 if (!uid_valid(kuid))
814 new = prepare_creds();
817 old = current_cred();
820 if (nsown_capable(CAP_SETUID)) {
821 new->suid = new->uid = kuid;
822 if (!uid_eq(kuid, old->uid)) {
823 retval = set_user(new);
827 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
831 new->fsuid = new->euid = kuid;
833 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
837 return commit_creds(new);
846 * This function implements a generic ability to update ruid, euid,
847 * and suid. This allows you to implement the 4.4 compatible seteuid().
849 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
851 struct user_namespace *ns = current_user_ns();
852 const struct cred *old;
855 kuid_t kruid, keuid, ksuid;
857 kruid = make_kuid(ns, ruid);
858 keuid = make_kuid(ns, euid);
859 ksuid = make_kuid(ns, suid);
861 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
864 if ((euid != (uid_t) -1) && !uid_valid(keuid))
867 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
870 new = prepare_creds();
874 old = current_cred();
877 if (!nsown_capable(CAP_SETUID)) {
878 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
879 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
881 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
882 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
884 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
885 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
889 if (ruid != (uid_t) -1) {
891 if (!uid_eq(kruid, old->uid)) {
892 retval = set_user(new);
897 if (euid != (uid_t) -1)
899 if (suid != (uid_t) -1)
901 new->fsuid = new->euid;
903 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
907 return commit_creds(new);
914 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
916 const struct cred *cred = current_cred();
918 uid_t ruid, euid, suid;
920 ruid = from_kuid_munged(cred->user_ns, cred->uid);
921 euid = from_kuid_munged(cred->user_ns, cred->euid);
922 suid = from_kuid_munged(cred->user_ns, cred->suid);
924 if (!(retval = put_user(ruid, ruidp)) &&
925 !(retval = put_user(euid, euidp)))
926 retval = put_user(suid, suidp);
932 * Same as above, but for rgid, egid, sgid.
934 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
936 struct user_namespace *ns = current_user_ns();
937 const struct cred *old;
940 kgid_t krgid, kegid, ksgid;
942 krgid = make_kgid(ns, rgid);
943 kegid = make_kgid(ns, egid);
944 ksgid = make_kgid(ns, sgid);
946 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
948 if ((egid != (gid_t) -1) && !gid_valid(kegid))
950 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
953 new = prepare_creds();
956 old = current_cred();
959 if (!nsown_capable(CAP_SETGID)) {
960 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
961 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
963 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
964 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
966 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
967 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
971 if (rgid != (gid_t) -1)
973 if (egid != (gid_t) -1)
975 if (sgid != (gid_t) -1)
977 new->fsgid = new->egid;
979 return commit_creds(new);
986 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
988 const struct cred *cred = current_cred();
990 gid_t rgid, egid, sgid;
992 rgid = from_kgid_munged(cred->user_ns, cred->gid);
993 egid = from_kgid_munged(cred->user_ns, cred->egid);
994 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
996 if (!(retval = put_user(rgid, rgidp)) &&
997 !(retval = put_user(egid, egidp)))
998 retval = put_user(sgid, sgidp);
1005 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1006 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1007 * whatever uid it wants to). It normally shadows "euid", except when
1008 * explicitly set by setfsuid() or for access..
1010 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
1012 const struct cred *old;
1017 old = current_cred();
1018 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
1020 kuid = make_kuid(old->user_ns, uid);
1021 if (!uid_valid(kuid))
1024 new = prepare_creds();
1028 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
1029 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
1030 nsown_capable(CAP_SETUID)) {
1031 if (!uid_eq(kuid, old->fsuid)) {
1033 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
1047 * Samma på svenska..
1049 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
1051 const struct cred *old;
1056 old = current_cred();
1057 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1059 kgid = make_kgid(old->user_ns, gid);
1060 if (!gid_valid(kgid))
1063 new = prepare_creds();
1067 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1068 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
1069 nsown_capable(CAP_SETGID)) {
1070 if (!gid_eq(kgid, old->fsgid)) {
1085 * sys_getpid - return the thread group id of the current process
1087 * Note, despite the name, this returns the tgid not the pid. The tgid and
1088 * the pid are identical unless CLONE_THREAD was specified on clone() in
1089 * which case the tgid is the same in all threads of the same group.
1091 * This is SMP safe as current->tgid does not change.
1093 SYSCALL_DEFINE0(getpid)
1095 return task_tgid_vnr(current);
1098 /* Thread ID - the internal kernel "pid" */
1099 SYSCALL_DEFINE0(gettid)
1101 return task_pid_vnr(current);
1105 * Accessing ->real_parent is not SMP-safe, it could
1106 * change from under us. However, we can use a stale
1107 * value of ->real_parent under rcu_read_lock(), see
1108 * release_task()->call_rcu(delayed_put_task_struct).
1110 SYSCALL_DEFINE0(getppid)
1115 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1121 SYSCALL_DEFINE0(getuid)
1123 /* Only we change this so SMP safe */
1124 return from_kuid_munged(current_user_ns(), current_uid());
1127 SYSCALL_DEFINE0(geteuid)
1129 /* Only we change this so SMP safe */
1130 return from_kuid_munged(current_user_ns(), current_euid());
1133 SYSCALL_DEFINE0(getgid)
1135 /* Only we change this so SMP safe */
1136 return from_kgid_munged(current_user_ns(), current_gid());
1139 SYSCALL_DEFINE0(getegid)
1141 /* Only we change this so SMP safe */
1142 return from_kgid_munged(current_user_ns(), current_egid());
1145 void do_sys_times(struct tms *tms)
1147 cputime_t tgutime, tgstime, cutime, cstime;
1149 spin_lock_irq(¤t->sighand->siglock);
1150 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1151 cutime = current->signal->cutime;
1152 cstime = current->signal->cstime;
1153 spin_unlock_irq(¤t->sighand->siglock);
1154 tms->tms_utime = cputime_to_clock_t(tgutime);
1155 tms->tms_stime = cputime_to_clock_t(tgstime);
1156 tms->tms_cutime = cputime_to_clock_t(cutime);
1157 tms->tms_cstime = cputime_to_clock_t(cstime);
1160 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1166 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1169 force_successful_syscall_return();
1170 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1174 * This needs some heavy checking ...
1175 * I just haven't the stomach for it. I also don't fully
1176 * understand sessions/pgrp etc. Let somebody who does explain it.
1178 * OK, I think I have the protection semantics right.... this is really
1179 * only important on a multi-user system anyway, to make sure one user
1180 * can't send a signal to a process owned by another. -TYT, 12/12/91
1182 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1185 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1187 struct task_struct *p;
1188 struct task_struct *group_leader = current->group_leader;
1193 pid = task_pid_vnr(group_leader);
1200 /* From this point forward we keep holding onto the tasklist lock
1201 * so that our parent does not change from under us. -DaveM
1203 write_lock_irq(&tasklist_lock);
1206 p = find_task_by_vpid(pid);
1211 if (!thread_group_leader(p))
1214 if (same_thread_group(p->real_parent, group_leader)) {
1216 if (task_session(p) != task_session(group_leader))
1223 if (p != group_leader)
1228 if (p->signal->leader)
1233 struct task_struct *g;
1235 pgrp = find_vpid(pgid);
1236 g = pid_task(pgrp, PIDTYPE_PGID);
1237 if (!g || task_session(g) != task_session(group_leader))
1241 err = security_task_setpgid(p, pgid);
1245 if (task_pgrp(p) != pgrp)
1246 change_pid(p, PIDTYPE_PGID, pgrp);
1250 /* All paths lead to here, thus we are safe. -DaveM */
1251 write_unlock_irq(&tasklist_lock);
1256 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1258 struct task_struct *p;
1264 grp = task_pgrp(current);
1267 p = find_task_by_vpid(pid);
1274 retval = security_task_getpgid(p);
1278 retval = pid_vnr(grp);
1284 #ifdef __ARCH_WANT_SYS_GETPGRP
1286 SYSCALL_DEFINE0(getpgrp)
1288 return sys_getpgid(0);
1293 SYSCALL_DEFINE1(getsid, pid_t, pid)
1295 struct task_struct *p;
1301 sid = task_session(current);
1304 p = find_task_by_vpid(pid);
1307 sid = task_session(p);
1311 retval = security_task_getsid(p);
1315 retval = pid_vnr(sid);
1321 SYSCALL_DEFINE0(setsid)
1323 struct task_struct *group_leader = current->group_leader;
1324 struct pid *sid = task_pid(group_leader);
1325 pid_t session = pid_vnr(sid);
1328 write_lock_irq(&tasklist_lock);
1329 /* Fail if I am already a session leader */
1330 if (group_leader->signal->leader)
1333 /* Fail if a process group id already exists that equals the
1334 * proposed session id.
1336 if (pid_task(sid, PIDTYPE_PGID))
1339 group_leader->signal->leader = 1;
1340 __set_special_pids(sid);
1342 proc_clear_tty(group_leader);
1346 write_unlock_irq(&tasklist_lock);
1348 proc_sid_connector(group_leader);
1349 sched_autogroup_create_attach(group_leader);
1354 DECLARE_RWSEM(uts_sem);
1356 #ifdef COMPAT_UTS_MACHINE
1357 #define override_architecture(name) \
1358 (personality(current->personality) == PER_LINUX32 && \
1359 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1360 sizeof(COMPAT_UTS_MACHINE)))
1362 #define override_architecture(name) 0
1366 * Work around broken programs that cannot handle "Linux 3.0".
1367 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1369 static int override_release(char __user *release, size_t len)
1373 if (current->personality & UNAME26) {
1374 const char *rest = UTS_RELEASE;
1375 char buf[65] = { 0 };
1381 if (*rest == '.' && ++ndots >= 3)
1383 if (!isdigit(*rest) && *rest != '.')
1387 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1388 copy = clamp_t(size_t, len, 1, sizeof(buf));
1389 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1390 ret = copy_to_user(release, buf, copy + 1);
1395 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1399 down_read(&uts_sem);
1400 if (copy_to_user(name, utsname(), sizeof *name))
1404 if (!errno && override_release(name->release, sizeof(name->release)))
1406 if (!errno && override_architecture(name))
1411 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1415 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1422 down_read(&uts_sem);
1423 if (copy_to_user(name, utsname(), sizeof(*name)))
1427 if (!error && override_release(name->release, sizeof(name->release)))
1429 if (!error && override_architecture(name))
1434 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1440 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1443 down_read(&uts_sem);
1444 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1446 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1447 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1449 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1450 error |= __copy_to_user(&name->release, &utsname()->release,
1452 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1453 error |= __copy_to_user(&name->version, &utsname()->version,
1455 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1456 error |= __copy_to_user(&name->machine, &utsname()->machine,
1458 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1461 if (!error && override_architecture(name))
1463 if (!error && override_release(name->release, sizeof(name->release)))
1465 return error ? -EFAULT : 0;
1469 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1472 char tmp[__NEW_UTS_LEN];
1474 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1477 if (len < 0 || len > __NEW_UTS_LEN)
1479 down_write(&uts_sem);
1481 if (!copy_from_user(tmp, name, len)) {
1482 struct new_utsname *u = utsname();
1484 memcpy(u->nodename, tmp, len);
1485 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1487 uts_proc_notify(UTS_PROC_HOSTNAME);
1493 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1495 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1498 struct new_utsname *u;
1502 down_read(&uts_sem);
1504 i = 1 + strlen(u->nodename);
1508 if (copy_to_user(name, u->nodename, i))
1517 * Only setdomainname; getdomainname can be implemented by calling
1520 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1523 char tmp[__NEW_UTS_LEN];
1525 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1527 if (len < 0 || len > __NEW_UTS_LEN)
1530 down_write(&uts_sem);
1532 if (!copy_from_user(tmp, name, len)) {
1533 struct new_utsname *u = utsname();
1535 memcpy(u->domainname, tmp, len);
1536 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1538 uts_proc_notify(UTS_PROC_DOMAINNAME);
1544 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1546 struct rlimit value;
1549 ret = do_prlimit(current, resource, NULL, &value);
1551 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1556 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1559 * Back compatibility for getrlimit. Needed for some apps.
1562 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1563 struct rlimit __user *, rlim)
1566 if (resource >= RLIM_NLIMITS)
1569 task_lock(current->group_leader);
1570 x = current->signal->rlim[resource];
1571 task_unlock(current->group_leader);
1572 if (x.rlim_cur > 0x7FFFFFFF)
1573 x.rlim_cur = 0x7FFFFFFF;
1574 if (x.rlim_max > 0x7FFFFFFF)
1575 x.rlim_max = 0x7FFFFFFF;
1576 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1581 static inline bool rlim64_is_infinity(__u64 rlim64)
1583 #if BITS_PER_LONG < 64
1584 return rlim64 >= ULONG_MAX;
1586 return rlim64 == RLIM64_INFINITY;
1590 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1592 if (rlim->rlim_cur == RLIM_INFINITY)
1593 rlim64->rlim_cur = RLIM64_INFINITY;
1595 rlim64->rlim_cur = rlim->rlim_cur;
1596 if (rlim->rlim_max == RLIM_INFINITY)
1597 rlim64->rlim_max = RLIM64_INFINITY;
1599 rlim64->rlim_max = rlim->rlim_max;
1602 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1604 if (rlim64_is_infinity(rlim64->rlim_cur))
1605 rlim->rlim_cur = RLIM_INFINITY;
1607 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1608 if (rlim64_is_infinity(rlim64->rlim_max))
1609 rlim->rlim_max = RLIM_INFINITY;
1611 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1614 /* make sure you are allowed to change @tsk limits before calling this */
1615 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1616 struct rlimit *new_rlim, struct rlimit *old_rlim)
1618 struct rlimit *rlim;
1621 if (resource >= RLIM_NLIMITS)
1624 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1626 if (resource == RLIMIT_NOFILE &&
1627 new_rlim->rlim_max > sysctl_nr_open)
1631 /* protect tsk->signal and tsk->sighand from disappearing */
1632 read_lock(&tasklist_lock);
1633 if (!tsk->sighand) {
1638 rlim = tsk->signal->rlim + resource;
1639 task_lock(tsk->group_leader);
1641 /* Keep the capable check against init_user_ns until
1642 cgroups can contain all limits */
1643 if (new_rlim->rlim_max > rlim->rlim_max &&
1644 !capable(CAP_SYS_RESOURCE))
1647 retval = security_task_setrlimit(tsk->group_leader,
1648 resource, new_rlim);
1649 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1651 * The caller is asking for an immediate RLIMIT_CPU
1652 * expiry. But we use the zero value to mean "it was
1653 * never set". So let's cheat and make it one second
1656 new_rlim->rlim_cur = 1;
1665 task_unlock(tsk->group_leader);
1668 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1669 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1670 * very long-standing error, and fixing it now risks breakage of
1671 * applications, so we live with it
1673 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1674 new_rlim->rlim_cur != RLIM_INFINITY)
1675 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1677 read_unlock(&tasklist_lock);
1681 /* rcu lock must be held */
1682 static int check_prlimit_permission(struct task_struct *task)
1684 const struct cred *cred = current_cred(), *tcred;
1686 if (current == task)
1689 tcred = __task_cred(task);
1690 if (uid_eq(cred->uid, tcred->euid) &&
1691 uid_eq(cred->uid, tcred->suid) &&
1692 uid_eq(cred->uid, tcred->uid) &&
1693 gid_eq(cred->gid, tcred->egid) &&
1694 gid_eq(cred->gid, tcred->sgid) &&
1695 gid_eq(cred->gid, tcred->gid))
1697 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1703 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1704 const struct rlimit64 __user *, new_rlim,
1705 struct rlimit64 __user *, old_rlim)
1707 struct rlimit64 old64, new64;
1708 struct rlimit old, new;
1709 struct task_struct *tsk;
1713 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1715 rlim64_to_rlim(&new64, &new);
1719 tsk = pid ? find_task_by_vpid(pid) : current;
1724 ret = check_prlimit_permission(tsk);
1729 get_task_struct(tsk);
1732 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1733 old_rlim ? &old : NULL);
1735 if (!ret && old_rlim) {
1736 rlim_to_rlim64(&old, &old64);
1737 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1741 put_task_struct(tsk);
1745 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1747 struct rlimit new_rlim;
1749 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1751 return do_prlimit(current, resource, &new_rlim, NULL);
1755 * It would make sense to put struct rusage in the task_struct,
1756 * except that would make the task_struct be *really big*. After
1757 * task_struct gets moved into malloc'ed memory, it would
1758 * make sense to do this. It will make moving the rest of the information
1759 * a lot simpler! (Which we're not doing right now because we're not
1760 * measuring them yet).
1762 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1763 * races with threads incrementing their own counters. But since word
1764 * reads are atomic, we either get new values or old values and we don't
1765 * care which for the sums. We always take the siglock to protect reading
1766 * the c* fields from p->signal from races with exit.c updating those
1767 * fields when reaping, so a sample either gets all the additions of a
1768 * given child after it's reaped, or none so this sample is before reaping.
1771 * We need to take the siglock for CHILDEREN, SELF and BOTH
1772 * for the cases current multithreaded, non-current single threaded
1773 * non-current multithreaded. Thread traversal is now safe with
1775 * Strictly speaking, we donot need to take the siglock if we are current and
1776 * single threaded, as no one else can take our signal_struct away, no one
1777 * else can reap the children to update signal->c* counters, and no one else
1778 * can race with the signal-> fields. If we do not take any lock, the
1779 * signal-> fields could be read out of order while another thread was just
1780 * exiting. So we should place a read memory barrier when we avoid the lock.
1781 * On the writer side, write memory barrier is implied in __exit_signal
1782 * as __exit_signal releases the siglock spinlock after updating the signal->
1783 * fields. But we don't do this yet to keep things simple.
1787 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1789 r->ru_nvcsw += t->nvcsw;
1790 r->ru_nivcsw += t->nivcsw;
1791 r->ru_minflt += t->min_flt;
1792 r->ru_majflt += t->maj_flt;
1793 r->ru_inblock += task_io_get_inblock(t);
1794 r->ru_oublock += task_io_get_oublock(t);
1797 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1799 struct task_struct *t;
1800 unsigned long flags;
1801 cputime_t tgutime, tgstime, utime, stime;
1802 unsigned long maxrss = 0;
1804 memset((char *) r, 0, sizeof *r);
1807 if (who == RUSAGE_THREAD) {
1808 task_cputime_adjusted(current, &utime, &stime);
1809 accumulate_thread_rusage(p, r);
1810 maxrss = p->signal->maxrss;
1814 if (!lock_task_sighand(p, &flags))
1819 case RUSAGE_CHILDREN:
1820 utime = p->signal->cutime;
1821 stime = p->signal->cstime;
1822 r->ru_nvcsw = p->signal->cnvcsw;
1823 r->ru_nivcsw = p->signal->cnivcsw;
1824 r->ru_minflt = p->signal->cmin_flt;
1825 r->ru_majflt = p->signal->cmaj_flt;
1826 r->ru_inblock = p->signal->cinblock;
1827 r->ru_oublock = p->signal->coublock;
1828 maxrss = p->signal->cmaxrss;
1830 if (who == RUSAGE_CHILDREN)
1834 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1837 r->ru_nvcsw += p->signal->nvcsw;
1838 r->ru_nivcsw += p->signal->nivcsw;
1839 r->ru_minflt += p->signal->min_flt;
1840 r->ru_majflt += p->signal->maj_flt;
1841 r->ru_inblock += p->signal->inblock;
1842 r->ru_oublock += p->signal->oublock;
1843 if (maxrss < p->signal->maxrss)
1844 maxrss = p->signal->maxrss;
1847 accumulate_thread_rusage(t, r);
1855 unlock_task_sighand(p, &flags);
1858 cputime_to_timeval(utime, &r->ru_utime);
1859 cputime_to_timeval(stime, &r->ru_stime);
1861 if (who != RUSAGE_CHILDREN) {
1862 struct mm_struct *mm = get_task_mm(p);
1864 setmax_mm_hiwater_rss(&maxrss, mm);
1868 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1871 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1874 k_getrusage(p, who, &r);
1875 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1878 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1880 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1881 who != RUSAGE_THREAD)
1883 return getrusage(current, who, ru);
1886 #ifdef CONFIG_COMPAT
1887 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1891 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1892 who != RUSAGE_THREAD)
1895 k_getrusage(current, who, &r);
1896 return put_compat_rusage(&r, ru);
1900 SYSCALL_DEFINE1(umask, int, mask)
1902 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1906 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1909 struct inode *inode;
1916 inode = file_inode(exe.file);
1919 * Because the original mm->exe_file points to executable file, make
1920 * sure that this one is executable as well, to avoid breaking an
1924 if (!S_ISREG(inode->i_mode) ||
1925 exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1928 err = inode_permission(inode, MAY_EXEC);
1932 down_write(&mm->mmap_sem);
1935 * Forbid mm->exe_file change if old file still mapped.
1939 struct vm_area_struct *vma;
1941 for (vma = mm->mmap; vma; vma = vma->vm_next)
1943 path_equal(&vma->vm_file->f_path,
1944 &mm->exe_file->f_path))
1949 * The symlink can be changed only once, just to disallow arbitrary
1950 * transitions malicious software might bring in. This means one
1951 * could make a snapshot over all processes running and monitor
1952 * /proc/pid/exe changes to notice unusual activity if needed.
1955 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1959 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
1961 up_write(&mm->mmap_sem);
1968 static int prctl_set_mm(int opt, unsigned long addr,
1969 unsigned long arg4, unsigned long arg5)
1971 unsigned long rlim = rlimit(RLIMIT_DATA);
1972 struct mm_struct *mm = current->mm;
1973 struct vm_area_struct *vma;
1976 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
1979 if (!capable(CAP_SYS_RESOURCE))
1982 if (opt == PR_SET_MM_EXE_FILE)
1983 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1985 if (addr >= TASK_SIZE || addr < mmap_min_addr)
1990 down_read(&mm->mmap_sem);
1991 vma = find_vma(mm, addr);
1994 case PR_SET_MM_START_CODE:
1995 mm->start_code = addr;
1997 case PR_SET_MM_END_CODE:
1998 mm->end_code = addr;
2000 case PR_SET_MM_START_DATA:
2001 mm->start_data = addr;
2003 case PR_SET_MM_END_DATA:
2004 mm->end_data = addr;
2007 case PR_SET_MM_START_BRK:
2008 if (addr <= mm->end_data)
2011 if (rlim < RLIM_INFINITY &&
2013 (mm->end_data - mm->start_data) > rlim)
2016 mm->start_brk = addr;
2020 if (addr <= mm->end_data)
2023 if (rlim < RLIM_INFINITY &&
2024 (addr - mm->start_brk) +
2025 (mm->end_data - mm->start_data) > rlim)
2032 * If command line arguments and environment
2033 * are placed somewhere else on stack, we can
2034 * set them up here, ARG_START/END to setup
2035 * command line argumets and ENV_START/END
2038 case PR_SET_MM_START_STACK:
2039 case PR_SET_MM_ARG_START:
2040 case PR_SET_MM_ARG_END:
2041 case PR_SET_MM_ENV_START:
2042 case PR_SET_MM_ENV_END:
2047 if (opt == PR_SET_MM_START_STACK)
2048 mm->start_stack = addr;
2049 else if (opt == PR_SET_MM_ARG_START)
2050 mm->arg_start = addr;
2051 else if (opt == PR_SET_MM_ARG_END)
2053 else if (opt == PR_SET_MM_ENV_START)
2054 mm->env_start = addr;
2055 else if (opt == PR_SET_MM_ENV_END)
2060 * This doesn't move auxiliary vector itself
2061 * since it's pinned to mm_struct, but allow
2062 * to fill vector with new values. It's up
2063 * to a caller to provide sane values here
2064 * otherwise user space tools which use this
2065 * vector might be unhappy.
2067 case PR_SET_MM_AUXV: {
2068 unsigned long user_auxv[AT_VECTOR_SIZE];
2070 if (arg4 > sizeof(user_auxv))
2072 up_read(&mm->mmap_sem);
2074 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
2077 /* Make sure the last entry is always AT_NULL */
2078 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2079 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2081 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2084 memcpy(mm->saved_auxv, user_auxv, arg4);
2085 task_unlock(current);
2095 up_read(&mm->mmap_sem);
2099 #ifdef CONFIG_CHECKPOINT_RESTORE
2100 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2102 return put_user(me->clear_child_tid, tid_addr);
2105 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2112 static int prctl_update_vma_anon_name(struct vm_area_struct *vma,
2113 struct vm_area_struct **prev,
2114 unsigned long start, unsigned long end,
2115 const char __user *name_addr)
2117 struct mm_struct * mm = vma->vm_mm;
2121 if (name_addr == vma_get_anon_name(vma)) {
2126 pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
2127 *prev = vma_merge(mm, *prev, start, end, vma->vm_flags, vma->anon_vma,
2128 vma->vm_file, pgoff, vma_policy(vma),
2137 if (start != vma->vm_start) {
2138 error = split_vma(mm, vma, start, 1);
2143 if (end != vma->vm_end) {
2144 error = split_vma(mm, vma, end, 0);
2151 vma->shared.anon_name = name_addr;
2154 if (error == -ENOMEM)
2159 static int prctl_set_vma_anon_name(unsigned long start, unsigned long end,
2163 struct vm_area_struct * vma, *prev;
2164 int unmapped_error = 0;
2165 int error = -EINVAL;
2168 * If the interval [start,end) covers some unmapped address
2169 * ranges, just ignore them, but return -ENOMEM at the end.
2170 * - this matches the handling in madvise.
2172 vma = find_vma_prev(current->mm, start, &prev);
2173 if (vma && start > vma->vm_start)
2177 /* Still start < end. */
2182 /* Here start < (end|vma->vm_end). */
2183 if (start < vma->vm_start) {
2184 unmapped_error = -ENOMEM;
2185 start = vma->vm_start;
2190 /* Here vma->vm_start <= start < (end|vma->vm_end) */
2195 /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */
2196 error = prctl_update_vma_anon_name(vma, &prev, start, tmp,
2197 (const char __user *)arg);
2201 if (prev && start < prev->vm_end)
2202 start = prev->vm_end;
2203 error = unmapped_error;
2207 vma = prev->vm_next;
2208 else /* madvise_remove dropped mmap_sem */
2209 vma = find_vma(current->mm, start);
2213 static int prctl_set_vma(unsigned long opt, unsigned long start,
2214 unsigned long len_in, unsigned long arg)
2216 struct mm_struct *mm = current->mm;
2221 if (start & ~PAGE_MASK)
2223 len = (len_in + ~PAGE_MASK) & PAGE_MASK;
2225 /* Check to see whether len was rounded up from small -ve to zero */
2236 down_write(&mm->mmap_sem);
2239 case PR_SET_VMA_ANON_NAME:
2240 error = prctl_set_vma_anon_name(start, end, arg);
2246 up_write(&mm->mmap_sem);
2250 #else /* CONFIG_MMU */
2251 static int prctl_set_vma(unsigned long opt, unsigned long start,
2252 unsigned long len_in, unsigned long arg)
2258 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2259 unsigned long, arg4, unsigned long, arg5)
2261 struct task_struct *me = current;
2262 struct task_struct *tsk;
2263 unsigned char comm[sizeof(me->comm)];
2266 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2267 if (error != -ENOSYS)
2272 case PR_SET_PDEATHSIG:
2273 if (!valid_signal(arg2)) {
2277 me->pdeath_signal = arg2;
2279 case PR_GET_PDEATHSIG:
2280 error = put_user(me->pdeath_signal, (int __user *)arg2);
2282 case PR_GET_DUMPABLE:
2283 error = get_dumpable(me->mm);
2285 case PR_SET_DUMPABLE:
2286 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2290 set_dumpable(me->mm, arg2);
2293 case PR_SET_UNALIGN:
2294 error = SET_UNALIGN_CTL(me, arg2);
2296 case PR_GET_UNALIGN:
2297 error = GET_UNALIGN_CTL(me, arg2);
2300 error = SET_FPEMU_CTL(me, arg2);
2303 error = GET_FPEMU_CTL(me, arg2);
2306 error = SET_FPEXC_CTL(me, arg2);
2309 error = GET_FPEXC_CTL(me, arg2);
2312 error = PR_TIMING_STATISTICAL;
2315 if (arg2 != PR_TIMING_STATISTICAL)
2319 comm[sizeof(me->comm) - 1] = 0;
2320 if (strncpy_from_user(comm, (char __user *)arg2,
2321 sizeof(me->comm) - 1) < 0)
2323 set_task_comm(me, comm);
2324 proc_comm_connector(me);
2327 get_task_comm(comm, me);
2328 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2332 error = GET_ENDIAN(me, arg2);
2335 error = SET_ENDIAN(me, arg2);
2337 case PR_GET_SECCOMP:
2338 error = prctl_get_seccomp();
2340 case PR_SET_SECCOMP:
2341 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2344 error = GET_TSC_CTL(arg2);
2347 error = SET_TSC_CTL(arg2);
2349 case PR_TASK_PERF_EVENTS_DISABLE:
2350 error = perf_event_task_disable();
2352 case PR_TASK_PERF_EVENTS_ENABLE:
2353 error = perf_event_task_enable();
2355 case PR_GET_TIMERSLACK:
2356 error = current->timer_slack_ns;
2358 case PR_SET_TIMERSLACK:
2360 current->timer_slack_ns =
2361 current->default_timer_slack_ns;
2363 current->timer_slack_ns = arg2;
2369 case PR_MCE_KILL_CLEAR:
2372 current->flags &= ~PF_MCE_PROCESS;
2374 case PR_MCE_KILL_SET:
2375 current->flags |= PF_MCE_PROCESS;
2376 if (arg3 == PR_MCE_KILL_EARLY)
2377 current->flags |= PF_MCE_EARLY;
2378 else if (arg3 == PR_MCE_KILL_LATE)
2379 current->flags &= ~PF_MCE_EARLY;
2380 else if (arg3 == PR_MCE_KILL_DEFAULT)
2382 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2390 case PR_MCE_KILL_GET:
2391 if (arg2 | arg3 | arg4 | arg5)
2393 if (current->flags & PF_MCE_PROCESS)
2394 error = (current->flags & PF_MCE_EARLY) ?
2395 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2397 error = PR_MCE_KILL_DEFAULT;
2400 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2402 case PR_GET_TID_ADDRESS:
2403 error = prctl_get_tid_address(me, (int __user **)arg2);
2405 case PR_SET_TIMERSLACK_PID:
2406 if (task_pid_vnr(current) != (pid_t)arg3 &&
2407 !capable(CAP_SYS_NICE))
2410 tsk = find_task_by_vpid((pid_t)arg3);
2415 get_task_struct(tsk);
2418 tsk->timer_slack_ns =
2419 tsk->default_timer_slack_ns;
2421 tsk->timer_slack_ns = arg2;
2422 put_task_struct(tsk);
2425 case PR_SET_CHILD_SUBREAPER:
2426 me->signal->is_child_subreaper = !!arg2;
2428 case PR_GET_CHILD_SUBREAPER:
2429 error = put_user(me->signal->is_child_subreaper,
2430 (int __user *)arg2);
2432 case PR_SET_NO_NEW_PRIVS:
2433 if (arg2 != 1 || arg3 || arg4 || arg5)
2436 task_set_no_new_privs(current);
2438 case PR_GET_NO_NEW_PRIVS:
2439 if (arg2 || arg3 || arg4 || arg5)
2441 return task_no_new_privs(current) ? 1 : 0;
2443 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2452 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2453 struct getcpu_cache __user *, unused)
2456 int cpu = raw_smp_processor_id();
2458 err |= put_user(cpu, cpup);
2460 err |= put_user(cpu_to_node(cpu), nodep);
2461 return err ? -EFAULT : 0;
2464 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2466 static int __orderly_poweroff(bool force)
2469 static char *envp[] = {
2471 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2476 argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL);
2478 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
2481 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2482 __func__, poweroff_cmd);
2487 printk(KERN_WARNING "Failed to start orderly shutdown: "
2488 "forcing the issue\n");
2490 * I guess this should try to kick off some daemon to sync and
2491 * poweroff asap. Or not even bother syncing if we're doing an
2492 * emergency shutdown?
2501 static bool poweroff_force;
2503 static void poweroff_work_func(struct work_struct *work)
2505 __orderly_poweroff(poweroff_force);
2508 static DECLARE_WORK(poweroff_work, poweroff_work_func);
2511 * orderly_poweroff - Trigger an orderly system poweroff
2512 * @force: force poweroff if command execution fails
2514 * This may be called from any context to trigger a system shutdown.
2515 * If the orderly shutdown fails, it will force an immediate shutdown.
2517 int orderly_poweroff(bool force)
2519 if (force) /* do not override the pending "true" */
2520 poweroff_force = true;
2521 schedule_work(&poweroff_work);
2524 EXPORT_SYMBOL_GPL(orderly_poweroff);
2527 * do_sysinfo - fill in sysinfo struct
2528 * @info: pointer to buffer to fill
2530 static int do_sysinfo(struct sysinfo *info)
2532 unsigned long mem_total, sav_total;
2533 unsigned int mem_unit, bitcount;
2536 memset(info, 0, sizeof(struct sysinfo));
2539 monotonic_to_bootbased(&tp);
2540 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2542 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2544 info->procs = nr_threads;
2550 * If the sum of all the available memory (i.e. ram + swap)
2551 * is less than can be stored in a 32 bit unsigned long then
2552 * we can be binary compatible with 2.2.x kernels. If not,
2553 * well, in that case 2.2.x was broken anyways...
2555 * -Erik Andersen <andersee@debian.org>
2558 mem_total = info->totalram + info->totalswap;
2559 if (mem_total < info->totalram || mem_total < info->totalswap)
2562 mem_unit = info->mem_unit;
2563 while (mem_unit > 1) {
2566 sav_total = mem_total;
2568 if (mem_total < sav_total)
2573 * If mem_total did not overflow, multiply all memory values by
2574 * info->mem_unit and set it to 1. This leaves things compatible
2575 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2580 info->totalram <<= bitcount;
2581 info->freeram <<= bitcount;
2582 info->sharedram <<= bitcount;
2583 info->bufferram <<= bitcount;
2584 info->totalswap <<= bitcount;
2585 info->freeswap <<= bitcount;
2586 info->totalhigh <<= bitcount;
2587 info->freehigh <<= bitcount;
2593 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2599 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2605 #ifdef CONFIG_COMPAT
2606 struct compat_sysinfo {
2620 char _f[20-2*sizeof(u32)-sizeof(int)];
2623 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2629 /* Check to see if any memory value is too large for 32-bit and scale
2632 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2635 while (s.mem_unit < PAGE_SIZE) {
2640 s.totalram >>= bitcount;
2641 s.freeram >>= bitcount;
2642 s.sharedram >>= bitcount;
2643 s.bufferram >>= bitcount;
2644 s.totalswap >>= bitcount;
2645 s.freeswap >>= bitcount;
2646 s.totalhigh >>= bitcount;
2647 s.freehigh >>= bitcount;
2650 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2651 __put_user(s.uptime, &info->uptime) ||
2652 __put_user(s.loads[0], &info->loads[0]) ||
2653 __put_user(s.loads[1], &info->loads[1]) ||
2654 __put_user(s.loads[2], &info->loads[2]) ||
2655 __put_user(s.totalram, &info->totalram) ||
2656 __put_user(s.freeram, &info->freeram) ||
2657 __put_user(s.sharedram, &info->sharedram) ||
2658 __put_user(s.bufferram, &info->bufferram) ||
2659 __put_user(s.totalswap, &info->totalswap) ||
2660 __put_user(s.freeswap, &info->freeswap) ||
2661 __put_user(s.procs, &info->procs) ||
2662 __put_user(s.totalhigh, &info->totalhigh) ||
2663 __put_user(s.freehigh, &info->freehigh) ||
2664 __put_user(s.mem_unit, &info->mem_unit))
2669 #endif /* CONFIG_COMPAT */