4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/module.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/notifier.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.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/gfp.h>
40 #include <linux/syscore_ops.h>
41 #include <linux/version.h>
42 #include <linux/ctype.h>
44 #include <linux/compat.h>
45 #include <linux/syscalls.h>
46 #include <linux/kprobes.h>
47 #include <linux/user_namespace.h>
49 #include <linux/kmsg_dump.h>
50 /* Move somewhere else to avoid recompiling? */
51 #include <generated/utsrelease.h>
53 #include <asm/uaccess.h>
55 #include <asm/unistd.h>
57 #ifndef SET_UNALIGN_CTL
58 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
60 #ifndef GET_UNALIGN_CTL
61 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
64 # define SET_FPEMU_CTL(a,b) (-EINVAL)
67 # define GET_FPEMU_CTL(a,b) (-EINVAL)
70 # define SET_FPEXC_CTL(a,b) (-EINVAL)
73 # define GET_FPEXC_CTL(a,b) (-EINVAL)
76 # define GET_ENDIAN(a,b) (-EINVAL)
79 # define SET_ENDIAN(a,b) (-EINVAL)
82 # define GET_TSC_CTL(a) (-EINVAL)
85 # define SET_TSC_CTL(a) (-EINVAL)
89 * this is where the system-wide overflow UID and GID are defined, for
90 * architectures that now have 32-bit UID/GID but didn't in the past
93 int overflowuid = DEFAULT_OVERFLOWUID;
94 int overflowgid = DEFAULT_OVERFLOWGID;
97 EXPORT_SYMBOL(overflowuid);
98 EXPORT_SYMBOL(overflowgid);
102 * the same as above, but for filesystems which can only store a 16-bit
103 * UID and GID. as such, this is needed on all architectures
106 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
107 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
109 EXPORT_SYMBOL(fs_overflowuid);
110 EXPORT_SYMBOL(fs_overflowgid);
113 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
118 EXPORT_SYMBOL(cad_pid);
121 * If set, this is used for preparing the system to power off.
124 void (*pm_power_off_prepare)(void);
127 * Returns true if current's euid is same as p's uid or euid,
128 * or has CAP_SYS_NICE to p's user_ns.
130 * Called with rcu_read_lock, creds are safe
132 static bool set_one_prio_perm(struct task_struct *p)
134 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
136 if (pcred->user->user_ns == cred->user->user_ns &&
137 (pcred->uid == cred->euid ||
138 pcred->euid == cred->euid))
140 if (ns_capable(pcred->user->user_ns, CAP_SYS_NICE))
146 * set the priority of a task
147 * - the caller must hold the RCU read lock
149 static int set_one_prio(struct task_struct *p, int niceval, int error)
153 if (!set_one_prio_perm(p)) {
157 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
161 no_nice = security_task_setnice(p, niceval);
168 set_user_nice(p, niceval);
173 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
175 struct task_struct *g, *p;
176 struct user_struct *user;
177 const struct cred *cred = current_cred();
181 if (which > PRIO_USER || which < PRIO_PROCESS)
184 /* normalize: avoid signed division (rounding problems) */
192 read_lock(&tasklist_lock);
196 p = find_task_by_vpid(who);
200 error = set_one_prio(p, niceval, error);
204 pgrp = find_vpid(who);
206 pgrp = task_pgrp(current);
207 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
208 error = set_one_prio(p, niceval, error);
209 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
212 user = (struct user_struct *) cred->user;
215 else if ((who != cred->uid) &&
216 !(user = find_user(who)))
217 goto out_unlock; /* No processes for this user */
219 do_each_thread(g, p) {
220 if (__task_cred(p)->uid == who)
221 error = set_one_prio(p, niceval, error);
222 } while_each_thread(g, p);
223 if (who != cred->uid)
224 free_uid(user); /* For find_user() */
228 read_unlock(&tasklist_lock);
235 * Ugh. To avoid negative return values, "getpriority()" will
236 * not return the normal nice-value, but a negated value that
237 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
238 * to stay compatible.
240 SYSCALL_DEFINE2(getpriority, int, which, int, who)
242 struct task_struct *g, *p;
243 struct user_struct *user;
244 const struct cred *cred = current_cred();
245 long niceval, retval = -ESRCH;
248 if (which > PRIO_USER || which < PRIO_PROCESS)
252 read_lock(&tasklist_lock);
256 p = find_task_by_vpid(who);
260 niceval = 20 - task_nice(p);
261 if (niceval > retval)
267 pgrp = find_vpid(who);
269 pgrp = task_pgrp(current);
270 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
271 niceval = 20 - task_nice(p);
272 if (niceval > retval)
274 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
277 user = (struct user_struct *) cred->user;
280 else if ((who != cred->uid) &&
281 !(user = find_user(who)))
282 goto out_unlock; /* No processes for this user */
284 do_each_thread(g, p) {
285 if (__task_cred(p)->uid == who) {
286 niceval = 20 - task_nice(p);
287 if (niceval > retval)
290 } while_each_thread(g, p);
291 if (who != cred->uid)
292 free_uid(user); /* for find_user() */
296 read_unlock(&tasklist_lock);
303 * emergency_restart - reboot the system
305 * Without shutting down any hardware or taking any locks
306 * reboot the system. This is called when we know we are in
307 * trouble so this is our best effort to reboot. This is
308 * safe to call in interrupt context.
310 void emergency_restart(void)
312 kmsg_dump(KMSG_DUMP_EMERG);
313 machine_emergency_restart();
315 EXPORT_SYMBOL_GPL(emergency_restart);
317 void kernel_restart_prepare(char *cmd)
319 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
320 system_state = SYSTEM_RESTART;
321 usermodehelper_disable();
327 * kernel_restart - reboot the system
328 * @cmd: pointer to buffer containing command to execute for restart
331 * Shutdown everything and perform a clean reboot.
332 * This is not safe to call in interrupt context.
334 void kernel_restart(char *cmd)
336 kernel_restart_prepare(cmd);
337 disable_nonboot_cpus();
339 printk(KERN_EMERG "Restarting system.\n");
341 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
342 kmsg_dump(KMSG_DUMP_RESTART);
343 machine_restart(cmd);
345 EXPORT_SYMBOL_GPL(kernel_restart);
347 static void kernel_shutdown_prepare(enum system_states state)
349 blocking_notifier_call_chain(&reboot_notifier_list,
350 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
351 system_state = state;
352 usermodehelper_disable();
356 * kernel_halt - halt the system
358 * Shutdown everything and perform a clean system halt.
360 void kernel_halt(void)
362 kernel_shutdown_prepare(SYSTEM_HALT);
364 printk(KERN_EMERG "System halted.\n");
365 kmsg_dump(KMSG_DUMP_HALT);
369 EXPORT_SYMBOL_GPL(kernel_halt);
372 * kernel_power_off - power_off the system
374 * Shutdown everything and perform a clean system power_off.
376 void kernel_power_off(void)
378 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
379 if (pm_power_off_prepare)
380 pm_power_off_prepare();
381 disable_nonboot_cpus();
383 printk(KERN_EMERG "Power down.\n");
384 kmsg_dump(KMSG_DUMP_POWEROFF);
387 EXPORT_SYMBOL_GPL(kernel_power_off);
389 static DEFINE_MUTEX(reboot_mutex);
392 * Reboot system call: for obvious reasons only root may call it,
393 * and even root needs to set up some magic numbers in the registers
394 * so that some mistake won't make this reboot the whole machine.
395 * You can also set the meaning of the ctrl-alt-del-key here.
397 * reboot doesn't sync: do that yourself before calling this.
399 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
405 /* We only trust the superuser with rebooting the system. */
406 if (!capable(CAP_SYS_BOOT))
409 /* For safety, we require "magic" arguments. */
410 if (magic1 != LINUX_REBOOT_MAGIC1 ||
411 (magic2 != LINUX_REBOOT_MAGIC2 &&
412 magic2 != LINUX_REBOOT_MAGIC2A &&
413 magic2 != LINUX_REBOOT_MAGIC2B &&
414 magic2 != LINUX_REBOOT_MAGIC2C))
417 /* Instead of trying to make the power_off code look like
418 * halt when pm_power_off is not set do it the easy way.
420 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
421 cmd = LINUX_REBOOT_CMD_HALT;
423 mutex_lock(&reboot_mutex);
425 case LINUX_REBOOT_CMD_RESTART:
426 kernel_restart(NULL);
429 case LINUX_REBOOT_CMD_CAD_ON:
433 case LINUX_REBOOT_CMD_CAD_OFF:
437 case LINUX_REBOOT_CMD_HALT:
440 panic("cannot halt");
442 case LINUX_REBOOT_CMD_POWER_OFF:
447 case LINUX_REBOOT_CMD_RESTART2:
448 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
452 buffer[sizeof(buffer) - 1] = '\0';
454 kernel_restart(buffer);
458 case LINUX_REBOOT_CMD_KEXEC:
459 ret = kernel_kexec();
463 #ifdef CONFIG_HIBERNATION
464 case LINUX_REBOOT_CMD_SW_SUSPEND:
473 mutex_unlock(&reboot_mutex);
477 static void deferred_cad(struct work_struct *dummy)
479 kernel_restart(NULL);
483 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
484 * As it's called within an interrupt, it may NOT sync: the only choice
485 * is whether to reboot at once, or just ignore the ctrl-alt-del.
487 void ctrl_alt_del(void)
489 static DECLARE_WORK(cad_work, deferred_cad);
492 schedule_work(&cad_work);
494 kill_cad_pid(SIGINT, 1);
498 * Unprivileged users may change the real gid to the effective gid
499 * or vice versa. (BSD-style)
501 * If you set the real gid at all, or set the effective gid to a value not
502 * equal to the real gid, then the saved gid is set to the new effective gid.
504 * This makes it possible for a setgid program to completely drop its
505 * privileges, which is often a useful assertion to make when you are doing
506 * a security audit over a program.
508 * The general idea is that a program which uses just setregid() will be
509 * 100% compatible with BSD. A program which uses just setgid() will be
510 * 100% compatible with POSIX with saved IDs.
512 * SMP: There are not races, the GIDs are checked only by filesystem
513 * operations (as far as semantic preservation is concerned).
515 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
517 const struct cred *old;
521 new = prepare_creds();
524 old = current_cred();
527 if (rgid != (gid_t) -1) {
528 if (old->gid == rgid ||
530 nsown_capable(CAP_SETGID))
535 if (egid != (gid_t) -1) {
536 if (old->gid == egid ||
539 nsown_capable(CAP_SETGID))
545 if (rgid != (gid_t) -1 ||
546 (egid != (gid_t) -1 && egid != old->gid))
547 new->sgid = new->egid;
548 new->fsgid = new->egid;
550 return commit_creds(new);
558 * setgid() is implemented like SysV w/ SAVED_IDS
560 * SMP: Same implicit races as above.
562 SYSCALL_DEFINE1(setgid, gid_t, gid)
564 const struct cred *old;
568 new = prepare_creds();
571 old = current_cred();
574 if (nsown_capable(CAP_SETGID))
575 new->gid = new->egid = new->sgid = new->fsgid = gid;
576 else if (gid == old->gid || gid == old->sgid)
577 new->egid = new->fsgid = gid;
581 return commit_creds(new);
589 * change the user struct in a credentials set to match the new UID
591 static int set_user(struct cred *new)
593 struct user_struct *new_user;
595 new_user = alloc_uid(current_user_ns(), new->uid);
599 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
600 new_user != INIT_USER) {
606 new->user = new_user;
611 * Unprivileged users may change the real uid to the effective uid
612 * or vice versa. (BSD-style)
614 * If you set the real uid at all, or set the effective uid to a value not
615 * equal to the real uid, then the saved uid is set to the new effective uid.
617 * This makes it possible for a setuid program to completely drop its
618 * privileges, which is often a useful assertion to make when you are doing
619 * a security audit over a program.
621 * The general idea is that a program which uses just setreuid() will be
622 * 100% compatible with BSD. A program which uses just setuid() will be
623 * 100% compatible with POSIX with saved IDs.
625 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
627 const struct cred *old;
631 new = prepare_creds();
634 old = current_cred();
637 if (ruid != (uid_t) -1) {
639 if (old->uid != ruid &&
641 !nsown_capable(CAP_SETUID))
645 if (euid != (uid_t) -1) {
647 if (old->uid != euid &&
650 !nsown_capable(CAP_SETUID))
654 if (new->uid != old->uid) {
655 retval = set_user(new);
659 if (ruid != (uid_t) -1 ||
660 (euid != (uid_t) -1 && euid != old->uid))
661 new->suid = new->euid;
662 new->fsuid = new->euid;
664 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
668 return commit_creds(new);
676 * setuid() is implemented like SysV with SAVED_IDS
678 * Note that SAVED_ID's is deficient in that a setuid root program
679 * like sendmail, for example, cannot set its uid to be a normal
680 * user and then switch back, because if you're root, setuid() sets
681 * the saved uid too. If you don't like this, blame the bright people
682 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
683 * will allow a root program to temporarily drop privileges and be able to
684 * regain them by swapping the real and effective uid.
686 SYSCALL_DEFINE1(setuid, uid_t, uid)
688 const struct cred *old;
692 new = prepare_creds();
695 old = current_cred();
698 if (nsown_capable(CAP_SETUID)) {
699 new->suid = new->uid = uid;
700 if (uid != old->uid) {
701 retval = set_user(new);
705 } else if (uid != old->uid && uid != new->suid) {
709 new->fsuid = new->euid = uid;
711 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
715 return commit_creds(new);
724 * This function implements a generic ability to update ruid, euid,
725 * and suid. This allows you to implement the 4.4 compatible seteuid().
727 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
729 const struct cred *old;
733 new = prepare_creds();
737 old = current_cred();
740 if (!nsown_capable(CAP_SETUID)) {
741 if (ruid != (uid_t) -1 && ruid != old->uid &&
742 ruid != old->euid && ruid != old->suid)
744 if (euid != (uid_t) -1 && euid != old->uid &&
745 euid != old->euid && euid != old->suid)
747 if (suid != (uid_t) -1 && suid != old->uid &&
748 suid != old->euid && suid != old->suid)
752 if (ruid != (uid_t) -1) {
754 if (ruid != old->uid) {
755 retval = set_user(new);
760 if (euid != (uid_t) -1)
762 if (suid != (uid_t) -1)
764 new->fsuid = new->euid;
766 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
770 return commit_creds(new);
777 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
779 const struct cred *cred = current_cred();
782 if (!(retval = put_user(cred->uid, ruid)) &&
783 !(retval = put_user(cred->euid, euid)))
784 retval = put_user(cred->suid, suid);
790 * Same as above, but for rgid, egid, sgid.
792 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
794 const struct cred *old;
798 new = prepare_creds();
801 old = current_cred();
804 if (!nsown_capable(CAP_SETGID)) {
805 if (rgid != (gid_t) -1 && rgid != old->gid &&
806 rgid != old->egid && rgid != old->sgid)
808 if (egid != (gid_t) -1 && egid != old->gid &&
809 egid != old->egid && egid != old->sgid)
811 if (sgid != (gid_t) -1 && sgid != old->gid &&
812 sgid != old->egid && sgid != old->sgid)
816 if (rgid != (gid_t) -1)
818 if (egid != (gid_t) -1)
820 if (sgid != (gid_t) -1)
822 new->fsgid = new->egid;
824 return commit_creds(new);
831 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
833 const struct cred *cred = current_cred();
836 if (!(retval = put_user(cred->gid, rgid)) &&
837 !(retval = put_user(cred->egid, egid)))
838 retval = put_user(cred->sgid, sgid);
845 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
846 * is used for "access()" and for the NFS daemon (letting nfsd stay at
847 * whatever uid it wants to). It normally shadows "euid", except when
848 * explicitly set by setfsuid() or for access..
850 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
852 const struct cred *old;
856 new = prepare_creds();
858 return current_fsuid();
859 old = current_cred();
860 old_fsuid = old->fsuid;
862 if (uid == old->uid || uid == old->euid ||
863 uid == old->suid || uid == old->fsuid ||
864 nsown_capable(CAP_SETUID)) {
865 if (uid != old_fsuid) {
867 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
881 * Samma på svenska..
883 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
885 const struct cred *old;
889 new = prepare_creds();
891 return current_fsgid();
892 old = current_cred();
893 old_fsgid = old->fsgid;
895 if (gid == old->gid || gid == old->egid ||
896 gid == old->sgid || gid == old->fsgid ||
897 nsown_capable(CAP_SETGID)) {
898 if (gid != old_fsgid) {
912 void do_sys_times(struct tms *tms)
914 cputime_t tgutime, tgstime, cutime, cstime;
916 spin_lock_irq(¤t->sighand->siglock);
917 thread_group_times(current, &tgutime, &tgstime);
918 cutime = current->signal->cutime;
919 cstime = current->signal->cstime;
920 spin_unlock_irq(¤t->sighand->siglock);
921 tms->tms_utime = cputime_to_clock_t(tgutime);
922 tms->tms_stime = cputime_to_clock_t(tgstime);
923 tms->tms_cutime = cputime_to_clock_t(cutime);
924 tms->tms_cstime = cputime_to_clock_t(cstime);
927 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
933 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
936 force_successful_syscall_return();
937 return (long) jiffies_64_to_clock_t(get_jiffies_64());
941 * This needs some heavy checking ...
942 * I just haven't the stomach for it. I also don't fully
943 * understand sessions/pgrp etc. Let somebody who does explain it.
945 * OK, I think I have the protection semantics right.... this is really
946 * only important on a multi-user system anyway, to make sure one user
947 * can't send a signal to a process owned by another. -TYT, 12/12/91
949 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
952 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
954 struct task_struct *p;
955 struct task_struct *group_leader = current->group_leader;
960 pid = task_pid_vnr(group_leader);
967 /* From this point forward we keep holding onto the tasklist lock
968 * so that our parent does not change from under us. -DaveM
970 write_lock_irq(&tasklist_lock);
973 p = find_task_by_vpid(pid);
978 if (!thread_group_leader(p))
981 if (same_thread_group(p->real_parent, group_leader)) {
983 if (task_session(p) != task_session(group_leader))
990 if (p != group_leader)
995 if (p->signal->leader)
1000 struct task_struct *g;
1002 pgrp = find_vpid(pgid);
1003 g = pid_task(pgrp, PIDTYPE_PGID);
1004 if (!g || task_session(g) != task_session(group_leader))
1008 err = security_task_setpgid(p, pgid);
1012 if (task_pgrp(p) != pgrp)
1013 change_pid(p, PIDTYPE_PGID, pgrp);
1017 /* All paths lead to here, thus we are safe. -DaveM */
1018 write_unlock_irq(&tasklist_lock);
1023 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1025 struct task_struct *p;
1031 grp = task_pgrp(current);
1034 p = find_task_by_vpid(pid);
1041 retval = security_task_getpgid(p);
1045 retval = pid_vnr(grp);
1051 #ifdef __ARCH_WANT_SYS_GETPGRP
1053 SYSCALL_DEFINE0(getpgrp)
1055 return sys_getpgid(0);
1060 SYSCALL_DEFINE1(getsid, pid_t, pid)
1062 struct task_struct *p;
1068 sid = task_session(current);
1071 p = find_task_by_vpid(pid);
1074 sid = task_session(p);
1078 retval = security_task_getsid(p);
1082 retval = pid_vnr(sid);
1088 SYSCALL_DEFINE0(setsid)
1090 struct task_struct *group_leader = current->group_leader;
1091 struct pid *sid = task_pid(group_leader);
1092 pid_t session = pid_vnr(sid);
1095 write_lock_irq(&tasklist_lock);
1096 /* Fail if I am already a session leader */
1097 if (group_leader->signal->leader)
1100 /* Fail if a process group id already exists that equals the
1101 * proposed session id.
1103 if (pid_task(sid, PIDTYPE_PGID))
1106 group_leader->signal->leader = 1;
1107 __set_special_pids(sid);
1109 proc_clear_tty(group_leader);
1113 write_unlock_irq(&tasklist_lock);
1115 proc_sid_connector(group_leader);
1116 sched_autogroup_create_attach(group_leader);
1121 DECLARE_RWSEM(uts_sem);
1123 #ifdef COMPAT_UTS_MACHINE
1124 #define override_architecture(name) \
1125 (personality(current->personality) == PER_LINUX32 && \
1126 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1127 sizeof(COMPAT_UTS_MACHINE)))
1129 #define override_architecture(name) 0
1133 * Work around broken programs that cannot handle "Linux 3.0".
1134 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1136 static int override_release(char __user *release, size_t len)
1140 if (current->personality & UNAME26) {
1141 const char *rest = UTS_RELEASE;
1142 char buf[65] = { 0 };
1148 if (*rest == '.' && ++ndots >= 3)
1150 if (!isdigit(*rest) && *rest != '.')
1154 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1155 copy = clamp_t(size_t, len, 1, sizeof(buf));
1156 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1157 ret = copy_to_user(release, buf, copy + 1);
1162 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1166 down_read(&uts_sem);
1167 if (copy_to_user(name, utsname(), sizeof *name))
1171 if (!errno && override_release(name->release, sizeof(name->release)))
1173 if (!errno && override_architecture(name))
1178 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1182 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1189 down_read(&uts_sem);
1190 if (copy_to_user(name, utsname(), sizeof(*name)))
1194 if (!error && override_release(name->release, sizeof(name->release)))
1196 if (!error && override_architecture(name))
1201 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1207 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1210 down_read(&uts_sem);
1211 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1213 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1214 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1216 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1217 error |= __copy_to_user(&name->release, &utsname()->release,
1219 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1220 error |= __copy_to_user(&name->version, &utsname()->version,
1222 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1223 error |= __copy_to_user(&name->machine, &utsname()->machine,
1225 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1228 if (!error && override_architecture(name))
1230 if (!error && override_release(name->release, sizeof(name->release)))
1232 return error ? -EFAULT : 0;
1236 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1239 char tmp[__NEW_UTS_LEN];
1241 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1244 if (len < 0 || len > __NEW_UTS_LEN)
1246 down_write(&uts_sem);
1248 if (!copy_from_user(tmp, name, len)) {
1249 struct new_utsname *u = utsname();
1251 memcpy(u->nodename, tmp, len);
1252 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1259 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1261 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1264 struct new_utsname *u;
1268 down_read(&uts_sem);
1270 i = 1 + strlen(u->nodename);
1274 if (copy_to_user(name, u->nodename, i))
1283 * Only setdomainname; getdomainname can be implemented by calling
1286 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1289 char tmp[__NEW_UTS_LEN];
1291 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1293 if (len < 0 || len > __NEW_UTS_LEN)
1296 down_write(&uts_sem);
1298 if (!copy_from_user(tmp, name, len)) {
1299 struct new_utsname *u = utsname();
1301 memcpy(u->domainname, tmp, len);
1302 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1309 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1311 struct rlimit value;
1314 ret = do_prlimit(current, resource, NULL, &value);
1316 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1321 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1324 * Back compatibility for getrlimit. Needed for some apps.
1327 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1328 struct rlimit __user *, rlim)
1331 if (resource >= RLIM_NLIMITS)
1334 task_lock(current->group_leader);
1335 x = current->signal->rlim[resource];
1336 task_unlock(current->group_leader);
1337 if (x.rlim_cur > 0x7FFFFFFF)
1338 x.rlim_cur = 0x7FFFFFFF;
1339 if (x.rlim_max > 0x7FFFFFFF)
1340 x.rlim_max = 0x7FFFFFFF;
1341 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1346 static inline bool rlim64_is_infinity(__u64 rlim64)
1348 #if BITS_PER_LONG < 64
1349 return rlim64 >= ULONG_MAX;
1351 return rlim64 == RLIM64_INFINITY;
1355 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1357 if (rlim->rlim_cur == RLIM_INFINITY)
1358 rlim64->rlim_cur = RLIM64_INFINITY;
1360 rlim64->rlim_cur = rlim->rlim_cur;
1361 if (rlim->rlim_max == RLIM_INFINITY)
1362 rlim64->rlim_max = RLIM64_INFINITY;
1364 rlim64->rlim_max = rlim->rlim_max;
1367 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1369 if (rlim64_is_infinity(rlim64->rlim_cur))
1370 rlim->rlim_cur = RLIM_INFINITY;
1372 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1373 if (rlim64_is_infinity(rlim64->rlim_max))
1374 rlim->rlim_max = RLIM_INFINITY;
1376 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1379 /* make sure you are allowed to change @tsk limits before calling this */
1380 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1381 struct rlimit *new_rlim, struct rlimit *old_rlim)
1383 struct rlimit *rlim;
1386 if (resource >= RLIM_NLIMITS)
1389 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1391 if (resource == RLIMIT_NOFILE &&
1392 new_rlim->rlim_max > sysctl_nr_open)
1396 /* protect tsk->signal and tsk->sighand from disappearing */
1397 read_lock(&tasklist_lock);
1398 if (!tsk->sighand) {
1403 rlim = tsk->signal->rlim + resource;
1404 task_lock(tsk->group_leader);
1406 /* Keep the capable check against init_user_ns until
1407 cgroups can contain all limits */
1408 if (new_rlim->rlim_max > rlim->rlim_max &&
1409 !capable(CAP_SYS_RESOURCE))
1412 retval = security_task_setrlimit(tsk->group_leader,
1413 resource, new_rlim);
1414 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1416 * The caller is asking for an immediate RLIMIT_CPU
1417 * expiry. But we use the zero value to mean "it was
1418 * never set". So let's cheat and make it one second
1421 new_rlim->rlim_cur = 1;
1430 task_unlock(tsk->group_leader);
1433 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1434 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1435 * very long-standing error, and fixing it now risks breakage of
1436 * applications, so we live with it
1438 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1439 new_rlim->rlim_cur != RLIM_INFINITY)
1440 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1442 read_unlock(&tasklist_lock);
1446 /* rcu lock must be held */
1447 static int check_prlimit_permission(struct task_struct *task)
1449 const struct cred *cred = current_cred(), *tcred;
1451 if (current == task)
1454 tcred = __task_cred(task);
1455 if (cred->user->user_ns == tcred->user->user_ns &&
1456 (cred->uid == tcred->euid &&
1457 cred->uid == tcred->suid &&
1458 cred->uid == tcred->uid &&
1459 cred->gid == tcred->egid &&
1460 cred->gid == tcred->sgid &&
1461 cred->gid == tcred->gid))
1463 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1469 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1470 const struct rlimit64 __user *, new_rlim,
1471 struct rlimit64 __user *, old_rlim)
1473 struct rlimit64 old64, new64;
1474 struct rlimit old, new;
1475 struct task_struct *tsk;
1479 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1481 rlim64_to_rlim(&new64, &new);
1485 tsk = pid ? find_task_by_vpid(pid) : current;
1490 ret = check_prlimit_permission(tsk);
1495 get_task_struct(tsk);
1498 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1499 old_rlim ? &old : NULL);
1501 if (!ret && old_rlim) {
1502 rlim_to_rlim64(&old, &old64);
1503 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1507 put_task_struct(tsk);
1511 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1513 struct rlimit new_rlim;
1515 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1517 return do_prlimit(current, resource, &new_rlim, NULL);
1521 * It would make sense to put struct rusage in the task_struct,
1522 * except that would make the task_struct be *really big*. After
1523 * task_struct gets moved into malloc'ed memory, it would
1524 * make sense to do this. It will make moving the rest of the information
1525 * a lot simpler! (Which we're not doing right now because we're not
1526 * measuring them yet).
1528 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1529 * races with threads incrementing their own counters. But since word
1530 * reads are atomic, we either get new values or old values and we don't
1531 * care which for the sums. We always take the siglock to protect reading
1532 * the c* fields from p->signal from races with exit.c updating those
1533 * fields when reaping, so a sample either gets all the additions of a
1534 * given child after it's reaped, or none so this sample is before reaping.
1537 * We need to take the siglock for CHILDEREN, SELF and BOTH
1538 * for the cases current multithreaded, non-current single threaded
1539 * non-current multithreaded. Thread traversal is now safe with
1541 * Strictly speaking, we donot need to take the siglock if we are current and
1542 * single threaded, as no one else can take our signal_struct away, no one
1543 * else can reap the children to update signal->c* counters, and no one else
1544 * can race with the signal-> fields. If we do not take any lock, the
1545 * signal-> fields could be read out of order while another thread was just
1546 * exiting. So we should place a read memory barrier when we avoid the lock.
1547 * On the writer side, write memory barrier is implied in __exit_signal
1548 * as __exit_signal releases the siglock spinlock after updating the signal->
1549 * fields. But we don't do this yet to keep things simple.
1553 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1555 r->ru_nvcsw += t->nvcsw;
1556 r->ru_nivcsw += t->nivcsw;
1557 r->ru_minflt += t->min_flt;
1558 r->ru_majflt += t->maj_flt;
1559 r->ru_inblock += task_io_get_inblock(t);
1560 r->ru_oublock += task_io_get_oublock(t);
1563 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1565 struct task_struct *t;
1566 unsigned long flags;
1567 cputime_t tgutime, tgstime, utime, stime;
1568 unsigned long maxrss = 0;
1570 memset((char *) r, 0, sizeof *r);
1571 utime = stime = cputime_zero;
1573 if (who == RUSAGE_THREAD) {
1574 task_times(current, &utime, &stime);
1575 accumulate_thread_rusage(p, r);
1576 maxrss = p->signal->maxrss;
1580 if (!lock_task_sighand(p, &flags))
1585 case RUSAGE_CHILDREN:
1586 utime = p->signal->cutime;
1587 stime = p->signal->cstime;
1588 r->ru_nvcsw = p->signal->cnvcsw;
1589 r->ru_nivcsw = p->signal->cnivcsw;
1590 r->ru_minflt = p->signal->cmin_flt;
1591 r->ru_majflt = p->signal->cmaj_flt;
1592 r->ru_inblock = p->signal->cinblock;
1593 r->ru_oublock = p->signal->coublock;
1594 maxrss = p->signal->cmaxrss;
1596 if (who == RUSAGE_CHILDREN)
1600 thread_group_times(p, &tgutime, &tgstime);
1601 utime = cputime_add(utime, tgutime);
1602 stime = cputime_add(stime, tgstime);
1603 r->ru_nvcsw += p->signal->nvcsw;
1604 r->ru_nivcsw += p->signal->nivcsw;
1605 r->ru_minflt += p->signal->min_flt;
1606 r->ru_majflt += p->signal->maj_flt;
1607 r->ru_inblock += p->signal->inblock;
1608 r->ru_oublock += p->signal->oublock;
1609 if (maxrss < p->signal->maxrss)
1610 maxrss = p->signal->maxrss;
1613 accumulate_thread_rusage(t, r);
1621 unlock_task_sighand(p, &flags);
1624 cputime_to_timeval(utime, &r->ru_utime);
1625 cputime_to_timeval(stime, &r->ru_stime);
1627 if (who != RUSAGE_CHILDREN) {
1628 struct mm_struct *mm = get_task_mm(p);
1630 setmax_mm_hiwater_rss(&maxrss, mm);
1634 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1637 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1640 k_getrusage(p, who, &r);
1641 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1644 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1646 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1647 who != RUSAGE_THREAD)
1649 return getrusage(current, who, ru);
1652 SYSCALL_DEFINE1(umask, int, mask)
1654 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1658 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1659 unsigned long, arg4, unsigned long, arg5)
1661 struct task_struct *me = current;
1662 unsigned char comm[sizeof(me->comm)];
1665 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1666 if (error != -ENOSYS)
1671 case PR_SET_PDEATHSIG:
1672 if (!valid_signal(arg2)) {
1676 me->pdeath_signal = arg2;
1679 case PR_GET_PDEATHSIG:
1680 error = put_user(me->pdeath_signal, (int __user *)arg2);
1682 case PR_GET_DUMPABLE:
1683 error = get_dumpable(me->mm);
1685 case PR_SET_DUMPABLE:
1686 if (arg2 < 0 || arg2 > 1) {
1690 set_dumpable(me->mm, arg2);
1694 case PR_SET_UNALIGN:
1695 error = SET_UNALIGN_CTL(me, arg2);
1697 case PR_GET_UNALIGN:
1698 error = GET_UNALIGN_CTL(me, arg2);
1701 error = SET_FPEMU_CTL(me, arg2);
1704 error = GET_FPEMU_CTL(me, arg2);
1707 error = SET_FPEXC_CTL(me, arg2);
1710 error = GET_FPEXC_CTL(me, arg2);
1713 error = PR_TIMING_STATISTICAL;
1716 if (arg2 != PR_TIMING_STATISTICAL)
1723 comm[sizeof(me->comm)-1] = 0;
1724 if (strncpy_from_user(comm, (char __user *)arg2,
1725 sizeof(me->comm) - 1) < 0)
1727 set_task_comm(me, comm);
1730 get_task_comm(comm, me);
1731 if (copy_to_user((char __user *)arg2, comm,
1736 error = GET_ENDIAN(me, arg2);
1739 error = SET_ENDIAN(me, arg2);
1742 case PR_GET_SECCOMP:
1743 error = prctl_get_seccomp();
1745 case PR_SET_SECCOMP:
1746 error = prctl_set_seccomp(arg2);
1749 error = GET_TSC_CTL(arg2);
1752 error = SET_TSC_CTL(arg2);
1754 case PR_TASK_PERF_EVENTS_DISABLE:
1755 error = perf_event_task_disable();
1757 case PR_TASK_PERF_EVENTS_ENABLE:
1758 error = perf_event_task_enable();
1760 case PR_GET_TIMERSLACK:
1761 error = current->timer_slack_ns;
1763 case PR_SET_TIMERSLACK:
1765 current->timer_slack_ns =
1766 current->default_timer_slack_ns;
1768 current->timer_slack_ns = arg2;
1775 case PR_MCE_KILL_CLEAR:
1778 current->flags &= ~PF_MCE_PROCESS;
1780 case PR_MCE_KILL_SET:
1781 current->flags |= PF_MCE_PROCESS;
1782 if (arg3 == PR_MCE_KILL_EARLY)
1783 current->flags |= PF_MCE_EARLY;
1784 else if (arg3 == PR_MCE_KILL_LATE)
1785 current->flags &= ~PF_MCE_EARLY;
1786 else if (arg3 == PR_MCE_KILL_DEFAULT)
1788 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1797 case PR_MCE_KILL_GET:
1798 if (arg2 | arg3 | arg4 | arg5)
1800 if (current->flags & PF_MCE_PROCESS)
1801 error = (current->flags & PF_MCE_EARLY) ?
1802 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1804 error = PR_MCE_KILL_DEFAULT;
1813 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1814 struct getcpu_cache __user *, unused)
1817 int cpu = raw_smp_processor_id();
1819 err |= put_user(cpu, cpup);
1821 err |= put_user(cpu_to_node(cpu), nodep);
1822 return err ? -EFAULT : 0;
1825 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1827 static void argv_cleanup(struct subprocess_info *info)
1829 argv_free(info->argv);
1833 * orderly_poweroff - Trigger an orderly system poweroff
1834 * @force: force poweroff if command execution fails
1836 * This may be called from any context to trigger a system shutdown.
1837 * If the orderly shutdown fails, it will force an immediate shutdown.
1839 int orderly_poweroff(bool force)
1842 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1843 static char *envp[] = {
1845 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1849 struct subprocess_info *info;
1852 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1853 __func__, poweroff_cmd);
1857 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1863 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
1865 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1869 printk(KERN_WARNING "Failed to start orderly shutdown: "
1870 "forcing the issue\n");
1872 /* I guess this should try to kick off some daemon to
1873 sync and poweroff asap. Or not even bother syncing
1874 if we're doing an emergency shutdown? */
1881 EXPORT_SYMBOL_GPL(orderly_poweroff);