4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/export.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/reboot.h>
12 #include <linux/prctl.h>
13 #include <linux/highuid.h>
15 #include <linux/kmod.h>
16 #include <linux/perf_event.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/kexec.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/binfmts.h>
52 #include <linux/sched.h>
53 #include <linux/rcupdate.h>
54 #include <linux/uidgid.h>
55 #include <linux/cred.h>
57 #include <linux/kmsg_dump.h>
58 /* Move somewhere else to avoid recompiling? */
59 #include <generated/utsrelease.h>
61 #include <asm/uaccess.h>
63 #include <asm/unistd.h>
65 #ifndef SET_UNALIGN_CTL
66 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
68 #ifndef GET_UNALIGN_CTL
69 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
72 # define SET_FPEMU_CTL(a,b) (-EINVAL)
75 # define GET_FPEMU_CTL(a,b) (-EINVAL)
78 # define SET_FPEXC_CTL(a,b) (-EINVAL)
81 # define GET_FPEXC_CTL(a,b) (-EINVAL)
84 # define GET_ENDIAN(a,b) (-EINVAL)
87 # define SET_ENDIAN(a,b) (-EINVAL)
90 # define GET_TSC_CTL(a) (-EINVAL)
93 # define SET_TSC_CTL(a) (-EINVAL)
97 * this is where the system-wide overflow UID and GID are defined, for
98 * architectures that now have 32-bit UID/GID but didn't in the past
101 int overflowuid = DEFAULT_OVERFLOWUID;
102 int overflowgid = DEFAULT_OVERFLOWGID;
104 EXPORT_SYMBOL(overflowuid);
105 EXPORT_SYMBOL(overflowgid);
108 * the same as above, but for filesystems which can only store a 16-bit
109 * UID and GID. as such, this is needed on all architectures
112 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
113 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
115 EXPORT_SYMBOL(fs_overflowuid);
116 EXPORT_SYMBOL(fs_overflowgid);
119 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
124 EXPORT_SYMBOL(cad_pid);
127 * If set, this is used for preparing the system to power off.
130 void (*pm_power_off_prepare)(void);
133 * Returns true if current's euid is same as p's uid or euid,
134 * or has CAP_SYS_NICE to p's user_ns.
136 * Called with rcu_read_lock, creds are safe
138 static bool set_one_prio_perm(struct task_struct *p)
140 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
142 if (uid_eq(pcred->uid, cred->euid) ||
143 uid_eq(pcred->euid, cred->euid))
145 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
151 * set the priority of a task
152 * - the caller must hold the RCU read lock
154 static int set_one_prio(struct task_struct *p, int niceval, int error)
158 if (!set_one_prio_perm(p)) {
162 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
166 no_nice = security_task_setnice(p, niceval);
173 set_user_nice(p, niceval);
178 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
180 struct task_struct *g, *p;
181 struct user_struct *user;
182 const struct cred *cred = current_cred();
187 if (which > PRIO_USER || which < PRIO_PROCESS)
190 /* normalize: avoid signed division (rounding problems) */
198 read_lock(&tasklist_lock);
202 p = find_task_by_vpid(who);
206 error = set_one_prio(p, niceval, error);
210 pgrp = find_vpid(who);
212 pgrp = task_pgrp(current);
213 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
214 error = set_one_prio(p, niceval, error);
215 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
218 uid = make_kuid(cred->user_ns, who);
222 else if (!uid_eq(uid, cred->uid) &&
223 !(user = find_user(uid)))
224 goto out_unlock; /* No processes for this user */
226 do_each_thread(g, p) {
227 if (uid_eq(task_uid(p), uid))
228 error = set_one_prio(p, niceval, error);
229 } while_each_thread(g, p);
230 if (!uid_eq(uid, cred->uid))
231 free_uid(user); /* For find_user() */
235 read_unlock(&tasklist_lock);
242 * Ugh. To avoid negative return values, "getpriority()" will
243 * not return the normal nice-value, but a negated value that
244 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
245 * to stay compatible.
247 SYSCALL_DEFINE2(getpriority, int, which, int, who)
249 struct task_struct *g, *p;
250 struct user_struct *user;
251 const struct cred *cred = current_cred();
252 long niceval, retval = -ESRCH;
256 if (which > PRIO_USER || which < PRIO_PROCESS)
260 read_lock(&tasklist_lock);
264 p = find_task_by_vpid(who);
268 niceval = 20 - task_nice(p);
269 if (niceval > retval)
275 pgrp = find_vpid(who);
277 pgrp = task_pgrp(current);
278 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
279 niceval = 20 - task_nice(p);
280 if (niceval > retval)
282 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
285 uid = make_kuid(cred->user_ns, who);
289 else if (!uid_eq(uid, cred->uid) &&
290 !(user = find_user(uid)))
291 goto out_unlock; /* No processes for this user */
293 do_each_thread(g, p) {
294 if (uid_eq(task_uid(p), uid)) {
295 niceval = 20 - task_nice(p);
296 if (niceval > retval)
299 } while_each_thread(g, p);
300 if (!uid_eq(uid, cred->uid))
301 free_uid(user); /* for find_user() */
305 read_unlock(&tasklist_lock);
312 * emergency_restart - reboot the system
314 * Without shutting down any hardware or taking any locks
315 * reboot the system. This is called when we know we are in
316 * trouble so this is our best effort to reboot. This is
317 * safe to call in interrupt context.
319 void emergency_restart(void)
321 kmsg_dump(KMSG_DUMP_EMERG);
322 machine_emergency_restart();
324 EXPORT_SYMBOL_GPL(emergency_restart);
326 void kernel_restart_prepare(char *cmd)
328 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
329 system_state = SYSTEM_RESTART;
330 usermodehelper_disable();
335 * register_reboot_notifier - Register function to be called at reboot time
336 * @nb: Info about notifier function to be called
338 * Registers a function with the list of functions
339 * to be called at reboot time.
341 * Currently always returns zero, as blocking_notifier_chain_register()
342 * always returns zero.
344 int register_reboot_notifier(struct notifier_block *nb)
346 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
348 EXPORT_SYMBOL(register_reboot_notifier);
351 * unregister_reboot_notifier - Unregister previously registered reboot notifier
352 * @nb: Hook to be unregistered
354 * Unregisters a previously registered reboot
357 * Returns zero on success, or %-ENOENT on failure.
359 int unregister_reboot_notifier(struct notifier_block *nb)
361 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
363 EXPORT_SYMBOL(unregister_reboot_notifier);
366 * kernel_restart - reboot the system
367 * @cmd: pointer to buffer containing command to execute for restart
370 * Shutdown everything and perform a clean reboot.
371 * This is not safe to call in interrupt context.
373 void kernel_restart(char *cmd)
375 kernel_restart_prepare(cmd);
376 disable_nonboot_cpus();
379 printk(KERN_EMERG "Restarting system.\n");
381 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
382 kmsg_dump(KMSG_DUMP_RESTART);
383 machine_restart(cmd);
385 EXPORT_SYMBOL_GPL(kernel_restart);
387 static void kernel_shutdown_prepare(enum system_states state)
389 blocking_notifier_call_chain(&reboot_notifier_list,
390 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
391 system_state = state;
392 usermodehelper_disable();
396 * kernel_halt - halt the system
398 * Shutdown everything and perform a clean system halt.
400 void kernel_halt(void)
402 kernel_shutdown_prepare(SYSTEM_HALT);
403 disable_nonboot_cpus();
405 printk(KERN_EMERG "System halted.\n");
406 kmsg_dump(KMSG_DUMP_HALT);
410 EXPORT_SYMBOL_GPL(kernel_halt);
413 * kernel_power_off - power_off the system
415 * Shutdown everything and perform a clean system power_off.
417 void kernel_power_off(void)
419 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
420 if (pm_power_off_prepare)
421 pm_power_off_prepare();
422 disable_nonboot_cpus();
424 printk(KERN_EMERG "Power down.\n");
425 kmsg_dump(KMSG_DUMP_POWEROFF);
428 EXPORT_SYMBOL_GPL(kernel_power_off);
430 static DEFINE_MUTEX(reboot_mutex);
433 * Reboot system call: for obvious reasons only root may call it,
434 * and even root needs to set up some magic numbers in the registers
435 * so that some mistake won't make this reboot the whole machine.
436 * You can also set the meaning of the ctrl-alt-del-key here.
438 * reboot doesn't sync: do that yourself before calling this.
440 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
443 struct pid_namespace *pid_ns = task_active_pid_ns(current);
447 /* We only trust the superuser with rebooting the system. */
448 if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
451 /* For safety, we require "magic" arguments. */
452 if (magic1 != LINUX_REBOOT_MAGIC1 ||
453 (magic2 != LINUX_REBOOT_MAGIC2 &&
454 magic2 != LINUX_REBOOT_MAGIC2A &&
455 magic2 != LINUX_REBOOT_MAGIC2B &&
456 magic2 != LINUX_REBOOT_MAGIC2C))
460 * If pid namespaces are enabled and the current task is in a child
461 * pid_namespace, the command is handled by reboot_pid_ns() which will
464 ret = reboot_pid_ns(pid_ns, cmd);
468 /* Instead of trying to make the power_off code look like
469 * halt when pm_power_off is not set do it the easy way.
471 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
472 cmd = LINUX_REBOOT_CMD_HALT;
474 mutex_lock(&reboot_mutex);
476 case LINUX_REBOOT_CMD_RESTART:
477 kernel_restart(NULL);
480 case LINUX_REBOOT_CMD_CAD_ON:
484 case LINUX_REBOOT_CMD_CAD_OFF:
488 case LINUX_REBOOT_CMD_HALT:
491 panic("cannot halt");
493 case LINUX_REBOOT_CMD_POWER_OFF:
498 case LINUX_REBOOT_CMD_RESTART2:
499 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
503 buffer[sizeof(buffer) - 1] = '\0';
505 kernel_restart(buffer);
509 case LINUX_REBOOT_CMD_KEXEC:
510 ret = kernel_kexec();
514 #ifdef CONFIG_HIBERNATION
515 case LINUX_REBOOT_CMD_SW_SUSPEND:
524 mutex_unlock(&reboot_mutex);
528 static void deferred_cad(struct work_struct *dummy)
530 kernel_restart(NULL);
534 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
535 * As it's called within an interrupt, it may NOT sync: the only choice
536 * is whether to reboot at once, or just ignore the ctrl-alt-del.
538 void ctrl_alt_del(void)
540 static DECLARE_WORK(cad_work, deferred_cad);
543 schedule_work(&cad_work);
545 kill_cad_pid(SIGINT, 1);
549 * Unprivileged users may change the real gid to the effective gid
550 * or vice versa. (BSD-style)
552 * If you set the real gid at all, or set the effective gid to a value not
553 * equal to the real gid, then the saved gid is set to the new effective gid.
555 * This makes it possible for a setgid program to completely drop its
556 * privileges, which is often a useful assertion to make when you are doing
557 * a security audit over a program.
559 * The general idea is that a program which uses just setregid() will be
560 * 100% compatible with BSD. A program which uses just setgid() will be
561 * 100% compatible with POSIX with saved IDs.
563 * SMP: There are not races, the GIDs are checked only by filesystem
564 * operations (as far as semantic preservation is concerned).
566 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
568 struct user_namespace *ns = current_user_ns();
569 const struct cred *old;
574 krgid = make_kgid(ns, rgid);
575 kegid = make_kgid(ns, egid);
577 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
579 if ((egid != (gid_t) -1) && !gid_valid(kegid))
582 new = prepare_creds();
585 old = current_cred();
588 if (rgid != (gid_t) -1) {
589 if (gid_eq(old->gid, krgid) ||
590 gid_eq(old->egid, krgid) ||
591 nsown_capable(CAP_SETGID))
596 if (egid != (gid_t) -1) {
597 if (gid_eq(old->gid, kegid) ||
598 gid_eq(old->egid, kegid) ||
599 gid_eq(old->sgid, kegid) ||
600 nsown_capable(CAP_SETGID))
606 if (rgid != (gid_t) -1 ||
607 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
608 new->sgid = new->egid;
609 new->fsgid = new->egid;
611 return commit_creds(new);
619 * setgid() is implemented like SysV w/ SAVED_IDS
621 * SMP: Same implicit races as above.
623 SYSCALL_DEFINE1(setgid, gid_t, gid)
625 struct user_namespace *ns = current_user_ns();
626 const struct cred *old;
631 kgid = make_kgid(ns, gid);
632 if (!gid_valid(kgid))
635 new = prepare_creds();
638 old = current_cred();
641 if (nsown_capable(CAP_SETGID))
642 new->gid = new->egid = new->sgid = new->fsgid = kgid;
643 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
644 new->egid = new->fsgid = kgid;
648 return commit_creds(new);
656 * change the user struct in a credentials set to match the new UID
658 static int set_user(struct cred *new)
660 struct user_struct *new_user;
662 new_user = alloc_uid(new->uid);
667 * We don't fail in case of NPROC limit excess here because too many
668 * poorly written programs don't check set*uid() return code, assuming
669 * it never fails if called by root. We may still enforce NPROC limit
670 * for programs doing set*uid()+execve() by harmlessly deferring the
671 * failure to the execve() stage.
673 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
674 new_user != INIT_USER)
675 current->flags |= PF_NPROC_EXCEEDED;
677 current->flags &= ~PF_NPROC_EXCEEDED;
680 new->user = new_user;
685 * Unprivileged users may change the real uid to the effective uid
686 * or vice versa. (BSD-style)
688 * If you set the real uid at all, or set the effective uid to a value not
689 * equal to the real uid, then the saved uid is set to the new effective uid.
691 * This makes it possible for a setuid program to completely drop its
692 * privileges, which is often a useful assertion to make when you are doing
693 * a security audit over a program.
695 * The general idea is that a program which uses just setreuid() will be
696 * 100% compatible with BSD. A program which uses just setuid() will be
697 * 100% compatible with POSIX with saved IDs.
699 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
701 struct user_namespace *ns = current_user_ns();
702 const struct cred *old;
707 kruid = make_kuid(ns, ruid);
708 keuid = make_kuid(ns, euid);
710 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
712 if ((euid != (uid_t) -1) && !uid_valid(keuid))
715 new = prepare_creds();
718 old = current_cred();
721 if (ruid != (uid_t) -1) {
723 if (!uid_eq(old->uid, kruid) &&
724 !uid_eq(old->euid, kruid) &&
725 !nsown_capable(CAP_SETUID))
729 if (euid != (uid_t) -1) {
731 if (!uid_eq(old->uid, keuid) &&
732 !uid_eq(old->euid, keuid) &&
733 !uid_eq(old->suid, keuid) &&
734 !nsown_capable(CAP_SETUID))
738 if (!uid_eq(new->uid, old->uid)) {
739 retval = set_user(new);
743 if (ruid != (uid_t) -1 ||
744 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
745 new->suid = new->euid;
746 new->fsuid = new->euid;
748 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
752 return commit_creds(new);
760 * setuid() is implemented like SysV with SAVED_IDS
762 * Note that SAVED_ID's is deficient in that a setuid root program
763 * like sendmail, for example, cannot set its uid to be a normal
764 * user and then switch back, because if you're root, setuid() sets
765 * the saved uid too. If you don't like this, blame the bright people
766 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
767 * will allow a root program to temporarily drop privileges and be able to
768 * regain them by swapping the real and effective uid.
770 SYSCALL_DEFINE1(setuid, uid_t, uid)
772 struct user_namespace *ns = current_user_ns();
773 const struct cred *old;
778 kuid = make_kuid(ns, uid);
779 if (!uid_valid(kuid))
782 new = prepare_creds();
785 old = current_cred();
788 if (nsown_capable(CAP_SETUID)) {
789 new->suid = new->uid = kuid;
790 if (!uid_eq(kuid, old->uid)) {
791 retval = set_user(new);
795 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
799 new->fsuid = new->euid = kuid;
801 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
805 return commit_creds(new);
814 * This function implements a generic ability to update ruid, euid,
815 * and suid. This allows you to implement the 4.4 compatible seteuid().
817 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
819 struct user_namespace *ns = current_user_ns();
820 const struct cred *old;
823 kuid_t kruid, keuid, ksuid;
825 kruid = make_kuid(ns, ruid);
826 keuid = make_kuid(ns, euid);
827 ksuid = make_kuid(ns, suid);
829 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
832 if ((euid != (uid_t) -1) && !uid_valid(keuid))
835 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
838 new = prepare_creds();
842 old = current_cred();
845 if (!nsown_capable(CAP_SETUID)) {
846 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
847 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
849 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
850 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
852 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
853 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
857 if (ruid != (uid_t) -1) {
859 if (!uid_eq(kruid, old->uid)) {
860 retval = set_user(new);
865 if (euid != (uid_t) -1)
867 if (suid != (uid_t) -1)
869 new->fsuid = new->euid;
871 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
875 return commit_creds(new);
882 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
884 const struct cred *cred = current_cred();
886 uid_t ruid, euid, suid;
888 ruid = from_kuid_munged(cred->user_ns, cred->uid);
889 euid = from_kuid_munged(cred->user_ns, cred->euid);
890 suid = from_kuid_munged(cred->user_ns, cred->suid);
892 if (!(retval = put_user(ruid, ruidp)) &&
893 !(retval = put_user(euid, euidp)))
894 retval = put_user(suid, suidp);
900 * Same as above, but for rgid, egid, sgid.
902 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
904 struct user_namespace *ns = current_user_ns();
905 const struct cred *old;
908 kgid_t krgid, kegid, ksgid;
910 krgid = make_kgid(ns, rgid);
911 kegid = make_kgid(ns, egid);
912 ksgid = make_kgid(ns, sgid);
914 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
916 if ((egid != (gid_t) -1) && !gid_valid(kegid))
918 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
921 new = prepare_creds();
924 old = current_cred();
927 if (!nsown_capable(CAP_SETGID)) {
928 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
929 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
931 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
932 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
934 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
935 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
939 if (rgid != (gid_t) -1)
941 if (egid != (gid_t) -1)
943 if (sgid != (gid_t) -1)
945 new->fsgid = new->egid;
947 return commit_creds(new);
954 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
956 const struct cred *cred = current_cred();
958 gid_t rgid, egid, sgid;
960 rgid = from_kgid_munged(cred->user_ns, cred->gid);
961 egid = from_kgid_munged(cred->user_ns, cred->egid);
962 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
964 if (!(retval = put_user(rgid, rgidp)) &&
965 !(retval = put_user(egid, egidp)))
966 retval = put_user(sgid, sgidp);
973 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
974 * is used for "access()" and for the NFS daemon (letting nfsd stay at
975 * whatever uid it wants to). It normally shadows "euid", except when
976 * explicitly set by setfsuid() or for access..
978 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
980 const struct cred *old;
985 old = current_cred();
986 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
988 kuid = make_kuid(old->user_ns, uid);
989 if (!uid_valid(kuid))
992 new = prepare_creds();
996 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
997 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
998 nsown_capable(CAP_SETUID)) {
999 if (!uid_eq(kuid, old->fsuid)) {
1001 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
1015 * Samma på svenska..
1017 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
1019 const struct cred *old;
1024 old = current_cred();
1025 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1027 kgid = make_kgid(old->user_ns, gid);
1028 if (!gid_valid(kgid))
1031 new = prepare_creds();
1035 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1036 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
1037 nsown_capable(CAP_SETGID)) {
1038 if (!gid_eq(kgid, old->fsgid)) {
1053 * sys_getpid - return the thread group id of the current process
1055 * Note, despite the name, this returns the tgid not the pid. The tgid and
1056 * the pid are identical unless CLONE_THREAD was specified on clone() in
1057 * which case the tgid is the same in all threads of the same group.
1059 * This is SMP safe as current->tgid does not change.
1061 SYSCALL_DEFINE0(getpid)
1063 return task_tgid_vnr(current);
1066 /* Thread ID - the internal kernel "pid" */
1067 SYSCALL_DEFINE0(gettid)
1069 return task_pid_vnr(current);
1073 * Accessing ->real_parent is not SMP-safe, it could
1074 * change from under us. However, we can use a stale
1075 * value of ->real_parent under rcu_read_lock(), see
1076 * release_task()->call_rcu(delayed_put_task_struct).
1078 SYSCALL_DEFINE0(getppid)
1083 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1089 SYSCALL_DEFINE0(getuid)
1091 /* Only we change this so SMP safe */
1092 return from_kuid_munged(current_user_ns(), current_uid());
1095 SYSCALL_DEFINE0(geteuid)
1097 /* Only we change this so SMP safe */
1098 return from_kuid_munged(current_user_ns(), current_euid());
1101 SYSCALL_DEFINE0(getgid)
1103 /* Only we change this so SMP safe */
1104 return from_kgid_munged(current_user_ns(), current_gid());
1107 SYSCALL_DEFINE0(getegid)
1109 /* Only we change this so SMP safe */
1110 return from_kgid_munged(current_user_ns(), current_egid());
1113 void do_sys_times(struct tms *tms)
1115 cputime_t tgutime, tgstime, cutime, cstime;
1117 spin_lock_irq(¤t->sighand->siglock);
1118 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1119 cutime = current->signal->cutime;
1120 cstime = current->signal->cstime;
1121 spin_unlock_irq(¤t->sighand->siglock);
1122 tms->tms_utime = cputime_to_clock_t(tgutime);
1123 tms->tms_stime = cputime_to_clock_t(tgstime);
1124 tms->tms_cutime = cputime_to_clock_t(cutime);
1125 tms->tms_cstime = cputime_to_clock_t(cstime);
1128 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1134 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1137 force_successful_syscall_return();
1138 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1142 * This needs some heavy checking ...
1143 * I just haven't the stomach for it. I also don't fully
1144 * understand sessions/pgrp etc. Let somebody who does explain it.
1146 * OK, I think I have the protection semantics right.... this is really
1147 * only important on a multi-user system anyway, to make sure one user
1148 * can't send a signal to a process owned by another. -TYT, 12/12/91
1150 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1153 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1155 struct task_struct *p;
1156 struct task_struct *group_leader = current->group_leader;
1161 pid = task_pid_vnr(group_leader);
1168 /* From this point forward we keep holding onto the tasklist lock
1169 * so that our parent does not change from under us. -DaveM
1171 write_lock_irq(&tasklist_lock);
1174 p = find_task_by_vpid(pid);
1179 if (!thread_group_leader(p))
1182 if (same_thread_group(p->real_parent, group_leader)) {
1184 if (task_session(p) != task_session(group_leader))
1191 if (p != group_leader)
1196 if (p->signal->leader)
1201 struct task_struct *g;
1203 pgrp = find_vpid(pgid);
1204 g = pid_task(pgrp, PIDTYPE_PGID);
1205 if (!g || task_session(g) != task_session(group_leader))
1209 err = security_task_setpgid(p, pgid);
1213 if (task_pgrp(p) != pgrp)
1214 change_pid(p, PIDTYPE_PGID, pgrp);
1218 /* All paths lead to here, thus we are safe. -DaveM */
1219 write_unlock_irq(&tasklist_lock);
1224 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1226 struct task_struct *p;
1232 grp = task_pgrp(current);
1235 p = find_task_by_vpid(pid);
1242 retval = security_task_getpgid(p);
1246 retval = pid_vnr(grp);
1252 #ifdef __ARCH_WANT_SYS_GETPGRP
1254 SYSCALL_DEFINE0(getpgrp)
1256 return sys_getpgid(0);
1261 SYSCALL_DEFINE1(getsid, pid_t, pid)
1263 struct task_struct *p;
1269 sid = task_session(current);
1272 p = find_task_by_vpid(pid);
1275 sid = task_session(p);
1279 retval = security_task_getsid(p);
1283 retval = pid_vnr(sid);
1289 SYSCALL_DEFINE0(setsid)
1291 struct task_struct *group_leader = current->group_leader;
1292 struct pid *sid = task_pid(group_leader);
1293 pid_t session = pid_vnr(sid);
1296 write_lock_irq(&tasklist_lock);
1297 /* Fail if I am already a session leader */
1298 if (group_leader->signal->leader)
1301 /* Fail if a process group id already exists that equals the
1302 * proposed session id.
1304 if (pid_task(sid, PIDTYPE_PGID))
1307 group_leader->signal->leader = 1;
1308 __set_special_pids(sid);
1310 proc_clear_tty(group_leader);
1314 write_unlock_irq(&tasklist_lock);
1316 proc_sid_connector(group_leader);
1317 sched_autogroup_create_attach(group_leader);
1322 DECLARE_RWSEM(uts_sem);
1324 #ifdef COMPAT_UTS_MACHINE
1325 #define override_architecture(name) \
1326 (personality(current->personality) == PER_LINUX32 && \
1327 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1328 sizeof(COMPAT_UTS_MACHINE)))
1330 #define override_architecture(name) 0
1334 * Work around broken programs that cannot handle "Linux 3.0".
1335 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1337 static int override_release(char __user *release, size_t len)
1341 if (current->personality & UNAME26) {
1342 const char *rest = UTS_RELEASE;
1343 char buf[65] = { 0 };
1349 if (*rest == '.' && ++ndots >= 3)
1351 if (!isdigit(*rest) && *rest != '.')
1355 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1356 copy = clamp_t(size_t, len, 1, sizeof(buf));
1357 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1358 ret = copy_to_user(release, buf, copy + 1);
1363 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1367 down_read(&uts_sem);
1368 if (copy_to_user(name, utsname(), sizeof *name))
1372 if (!errno && override_release(name->release, sizeof(name->release)))
1374 if (!errno && override_architecture(name))
1379 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1383 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1390 down_read(&uts_sem);
1391 if (copy_to_user(name, utsname(), sizeof(*name)))
1395 if (!error && override_release(name->release, sizeof(name->release)))
1397 if (!error && override_architecture(name))
1402 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1408 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1411 down_read(&uts_sem);
1412 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1414 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1415 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1417 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1418 error |= __copy_to_user(&name->release, &utsname()->release,
1420 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1421 error |= __copy_to_user(&name->version, &utsname()->version,
1423 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1424 error |= __copy_to_user(&name->machine, &utsname()->machine,
1426 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1429 if (!error && override_architecture(name))
1431 if (!error && override_release(name->release, sizeof(name->release)))
1433 return error ? -EFAULT : 0;
1437 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1440 char tmp[__NEW_UTS_LEN];
1442 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1445 if (len < 0 || len > __NEW_UTS_LEN)
1447 down_write(&uts_sem);
1449 if (!copy_from_user(tmp, name, len)) {
1450 struct new_utsname *u = utsname();
1452 memcpy(u->nodename, tmp, len);
1453 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1455 uts_proc_notify(UTS_PROC_HOSTNAME);
1461 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1463 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1466 struct new_utsname *u;
1470 down_read(&uts_sem);
1472 i = 1 + strlen(u->nodename);
1476 if (copy_to_user(name, u->nodename, i))
1485 * Only setdomainname; getdomainname can be implemented by calling
1488 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1491 char tmp[__NEW_UTS_LEN];
1493 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1495 if (len < 0 || len > __NEW_UTS_LEN)
1498 down_write(&uts_sem);
1500 if (!copy_from_user(tmp, name, len)) {
1501 struct new_utsname *u = utsname();
1503 memcpy(u->domainname, tmp, len);
1504 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1506 uts_proc_notify(UTS_PROC_DOMAINNAME);
1512 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1514 struct rlimit value;
1517 ret = do_prlimit(current, resource, NULL, &value);
1519 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1524 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1527 * Back compatibility for getrlimit. Needed for some apps.
1530 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1531 struct rlimit __user *, rlim)
1534 if (resource >= RLIM_NLIMITS)
1537 task_lock(current->group_leader);
1538 x = current->signal->rlim[resource];
1539 task_unlock(current->group_leader);
1540 if (x.rlim_cur > 0x7FFFFFFF)
1541 x.rlim_cur = 0x7FFFFFFF;
1542 if (x.rlim_max > 0x7FFFFFFF)
1543 x.rlim_max = 0x7FFFFFFF;
1544 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1549 static inline bool rlim64_is_infinity(__u64 rlim64)
1551 #if BITS_PER_LONG < 64
1552 return rlim64 >= ULONG_MAX;
1554 return rlim64 == RLIM64_INFINITY;
1558 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1560 if (rlim->rlim_cur == RLIM_INFINITY)
1561 rlim64->rlim_cur = RLIM64_INFINITY;
1563 rlim64->rlim_cur = rlim->rlim_cur;
1564 if (rlim->rlim_max == RLIM_INFINITY)
1565 rlim64->rlim_max = RLIM64_INFINITY;
1567 rlim64->rlim_max = rlim->rlim_max;
1570 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1572 if (rlim64_is_infinity(rlim64->rlim_cur))
1573 rlim->rlim_cur = RLIM_INFINITY;
1575 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1576 if (rlim64_is_infinity(rlim64->rlim_max))
1577 rlim->rlim_max = RLIM_INFINITY;
1579 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1582 /* make sure you are allowed to change @tsk limits before calling this */
1583 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1584 struct rlimit *new_rlim, struct rlimit *old_rlim)
1586 struct rlimit *rlim;
1589 if (resource >= RLIM_NLIMITS)
1592 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1594 if (resource == RLIMIT_NOFILE &&
1595 new_rlim->rlim_max > sysctl_nr_open)
1599 /* protect tsk->signal and tsk->sighand from disappearing */
1600 read_lock(&tasklist_lock);
1601 if (!tsk->sighand) {
1606 rlim = tsk->signal->rlim + resource;
1607 task_lock(tsk->group_leader);
1609 /* Keep the capable check against init_user_ns until
1610 cgroups can contain all limits */
1611 if (new_rlim->rlim_max > rlim->rlim_max &&
1612 !capable(CAP_SYS_RESOURCE))
1615 retval = security_task_setrlimit(tsk->group_leader,
1616 resource, new_rlim);
1617 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1619 * The caller is asking for an immediate RLIMIT_CPU
1620 * expiry. But we use the zero value to mean "it was
1621 * never set". So let's cheat and make it one second
1624 new_rlim->rlim_cur = 1;
1633 task_unlock(tsk->group_leader);
1636 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1637 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1638 * very long-standing error, and fixing it now risks breakage of
1639 * applications, so we live with it
1641 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1642 new_rlim->rlim_cur != RLIM_INFINITY)
1643 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1645 read_unlock(&tasklist_lock);
1649 /* rcu lock must be held */
1650 static int check_prlimit_permission(struct task_struct *task)
1652 const struct cred *cred = current_cred(), *tcred;
1654 if (current == task)
1657 tcred = __task_cred(task);
1658 if (uid_eq(cred->uid, tcred->euid) &&
1659 uid_eq(cred->uid, tcred->suid) &&
1660 uid_eq(cred->uid, tcred->uid) &&
1661 gid_eq(cred->gid, tcred->egid) &&
1662 gid_eq(cred->gid, tcred->sgid) &&
1663 gid_eq(cred->gid, tcred->gid))
1665 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1671 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1672 const struct rlimit64 __user *, new_rlim,
1673 struct rlimit64 __user *, old_rlim)
1675 struct rlimit64 old64, new64;
1676 struct rlimit old, new;
1677 struct task_struct *tsk;
1681 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1683 rlim64_to_rlim(&new64, &new);
1687 tsk = pid ? find_task_by_vpid(pid) : current;
1692 ret = check_prlimit_permission(tsk);
1697 get_task_struct(tsk);
1700 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1701 old_rlim ? &old : NULL);
1703 if (!ret && old_rlim) {
1704 rlim_to_rlim64(&old, &old64);
1705 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1709 put_task_struct(tsk);
1713 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1715 struct rlimit new_rlim;
1717 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1719 return do_prlimit(current, resource, &new_rlim, NULL);
1723 * It would make sense to put struct rusage in the task_struct,
1724 * except that would make the task_struct be *really big*. After
1725 * task_struct gets moved into malloc'ed memory, it would
1726 * make sense to do this. It will make moving the rest of the information
1727 * a lot simpler! (Which we're not doing right now because we're not
1728 * measuring them yet).
1730 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1731 * races with threads incrementing their own counters. But since word
1732 * reads are atomic, we either get new values or old values and we don't
1733 * care which for the sums. We always take the siglock to protect reading
1734 * the c* fields from p->signal from races with exit.c updating those
1735 * fields when reaping, so a sample either gets all the additions of a
1736 * given child after it's reaped, or none so this sample is before reaping.
1739 * We need to take the siglock for CHILDEREN, SELF and BOTH
1740 * for the cases current multithreaded, non-current single threaded
1741 * non-current multithreaded. Thread traversal is now safe with
1743 * Strictly speaking, we donot need to take the siglock if we are current and
1744 * single threaded, as no one else can take our signal_struct away, no one
1745 * else can reap the children to update signal->c* counters, and no one else
1746 * can race with the signal-> fields. If we do not take any lock, the
1747 * signal-> fields could be read out of order while another thread was just
1748 * exiting. So we should place a read memory barrier when we avoid the lock.
1749 * On the writer side, write memory barrier is implied in __exit_signal
1750 * as __exit_signal releases the siglock spinlock after updating the signal->
1751 * fields. But we don't do this yet to keep things simple.
1755 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1757 r->ru_nvcsw += t->nvcsw;
1758 r->ru_nivcsw += t->nivcsw;
1759 r->ru_minflt += t->min_flt;
1760 r->ru_majflt += t->maj_flt;
1761 r->ru_inblock += task_io_get_inblock(t);
1762 r->ru_oublock += task_io_get_oublock(t);
1765 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1767 struct task_struct *t;
1768 unsigned long flags;
1769 cputime_t tgutime, tgstime, utime, stime;
1770 unsigned long maxrss = 0;
1772 memset((char *) r, 0, sizeof *r);
1775 if (who == RUSAGE_THREAD) {
1776 task_cputime_adjusted(current, &utime, &stime);
1777 accumulate_thread_rusage(p, r);
1778 maxrss = p->signal->maxrss;
1782 if (!lock_task_sighand(p, &flags))
1787 case RUSAGE_CHILDREN:
1788 utime = p->signal->cutime;
1789 stime = p->signal->cstime;
1790 r->ru_nvcsw = p->signal->cnvcsw;
1791 r->ru_nivcsw = p->signal->cnivcsw;
1792 r->ru_minflt = p->signal->cmin_flt;
1793 r->ru_majflt = p->signal->cmaj_flt;
1794 r->ru_inblock = p->signal->cinblock;
1795 r->ru_oublock = p->signal->coublock;
1796 maxrss = p->signal->cmaxrss;
1798 if (who == RUSAGE_CHILDREN)
1802 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1805 r->ru_nvcsw += p->signal->nvcsw;
1806 r->ru_nivcsw += p->signal->nivcsw;
1807 r->ru_minflt += p->signal->min_flt;
1808 r->ru_majflt += p->signal->maj_flt;
1809 r->ru_inblock += p->signal->inblock;
1810 r->ru_oublock += p->signal->oublock;
1811 if (maxrss < p->signal->maxrss)
1812 maxrss = p->signal->maxrss;
1815 accumulate_thread_rusage(t, r);
1823 unlock_task_sighand(p, &flags);
1826 cputime_to_timeval(utime, &r->ru_utime);
1827 cputime_to_timeval(stime, &r->ru_stime);
1829 if (who != RUSAGE_CHILDREN) {
1830 struct mm_struct *mm = get_task_mm(p);
1832 setmax_mm_hiwater_rss(&maxrss, mm);
1836 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1839 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1842 k_getrusage(p, who, &r);
1843 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1846 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1848 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1849 who != RUSAGE_THREAD)
1851 return getrusage(current, who, ru);
1854 #ifdef CONFIG_COMPAT
1855 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1859 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1860 who != RUSAGE_THREAD)
1863 k_getrusage(current, who, &r);
1864 return put_compat_rusage(&r, ru);
1868 SYSCALL_DEFINE1(umask, int, mask)
1870 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1874 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1877 struct inode *inode;
1884 inode = file_inode(exe.file);
1887 * Because the original mm->exe_file points to executable file, make
1888 * sure that this one is executable as well, to avoid breaking an
1892 if (!S_ISREG(inode->i_mode) ||
1893 exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1896 err = inode_permission(inode, MAY_EXEC);
1900 down_write(&mm->mmap_sem);
1903 * Forbid mm->exe_file change if old file still mapped.
1907 struct vm_area_struct *vma;
1909 for (vma = mm->mmap; vma; vma = vma->vm_next)
1911 path_equal(&vma->vm_file->f_path,
1912 &mm->exe_file->f_path))
1917 * The symlink can be changed only once, just to disallow arbitrary
1918 * transitions malicious software might bring in. This means one
1919 * could make a snapshot over all processes running and monitor
1920 * /proc/pid/exe changes to notice unusual activity if needed.
1923 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1927 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
1929 up_write(&mm->mmap_sem);
1936 static int prctl_set_mm(int opt, unsigned long addr,
1937 unsigned long arg4, unsigned long arg5)
1939 unsigned long rlim = rlimit(RLIMIT_DATA);
1940 struct mm_struct *mm = current->mm;
1941 struct vm_area_struct *vma;
1944 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
1947 if (!capable(CAP_SYS_RESOURCE))
1950 if (opt == PR_SET_MM_EXE_FILE)
1951 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1953 if (addr >= TASK_SIZE || addr < mmap_min_addr)
1958 down_read(&mm->mmap_sem);
1959 vma = find_vma(mm, addr);
1962 case PR_SET_MM_START_CODE:
1963 mm->start_code = addr;
1965 case PR_SET_MM_END_CODE:
1966 mm->end_code = addr;
1968 case PR_SET_MM_START_DATA:
1969 mm->start_data = addr;
1971 case PR_SET_MM_END_DATA:
1972 mm->end_data = addr;
1975 case PR_SET_MM_START_BRK:
1976 if (addr <= mm->end_data)
1979 if (rlim < RLIM_INFINITY &&
1981 (mm->end_data - mm->start_data) > rlim)
1984 mm->start_brk = addr;
1988 if (addr <= mm->end_data)
1991 if (rlim < RLIM_INFINITY &&
1992 (addr - mm->start_brk) +
1993 (mm->end_data - mm->start_data) > rlim)
2000 * If command line arguments and environment
2001 * are placed somewhere else on stack, we can
2002 * set them up here, ARG_START/END to setup
2003 * command line argumets and ENV_START/END
2006 case PR_SET_MM_START_STACK:
2007 case PR_SET_MM_ARG_START:
2008 case PR_SET_MM_ARG_END:
2009 case PR_SET_MM_ENV_START:
2010 case PR_SET_MM_ENV_END:
2015 if (opt == PR_SET_MM_START_STACK)
2016 mm->start_stack = addr;
2017 else if (opt == PR_SET_MM_ARG_START)
2018 mm->arg_start = addr;
2019 else if (opt == PR_SET_MM_ARG_END)
2021 else if (opt == PR_SET_MM_ENV_START)
2022 mm->env_start = addr;
2023 else if (opt == PR_SET_MM_ENV_END)
2028 * This doesn't move auxiliary vector itself
2029 * since it's pinned to mm_struct, but allow
2030 * to fill vector with new values. It's up
2031 * to a caller to provide sane values here
2032 * otherwise user space tools which use this
2033 * vector might be unhappy.
2035 case PR_SET_MM_AUXV: {
2036 unsigned long user_auxv[AT_VECTOR_SIZE];
2038 if (arg4 > sizeof(user_auxv))
2040 up_read(&mm->mmap_sem);
2042 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
2045 /* Make sure the last entry is always AT_NULL */
2046 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2047 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2049 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2052 memcpy(mm->saved_auxv, user_auxv, arg4);
2053 task_unlock(current);
2063 up_read(&mm->mmap_sem);
2067 #ifdef CONFIG_CHECKPOINT_RESTORE
2068 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2070 return put_user(me->clear_child_tid, tid_addr);
2073 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2079 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2080 unsigned long, arg4, unsigned long, arg5)
2082 struct task_struct *me = current;
2083 unsigned char comm[sizeof(me->comm)];
2086 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2087 if (error != -ENOSYS)
2092 case PR_SET_PDEATHSIG:
2093 if (!valid_signal(arg2)) {
2097 me->pdeath_signal = arg2;
2099 case PR_GET_PDEATHSIG:
2100 error = put_user(me->pdeath_signal, (int __user *)arg2);
2102 case PR_GET_DUMPABLE:
2103 error = get_dumpable(me->mm);
2105 case PR_SET_DUMPABLE:
2106 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2110 set_dumpable(me->mm, arg2);
2113 case PR_SET_UNALIGN:
2114 error = SET_UNALIGN_CTL(me, arg2);
2116 case PR_GET_UNALIGN:
2117 error = GET_UNALIGN_CTL(me, arg2);
2120 error = SET_FPEMU_CTL(me, arg2);
2123 error = GET_FPEMU_CTL(me, arg2);
2126 error = SET_FPEXC_CTL(me, arg2);
2129 error = GET_FPEXC_CTL(me, arg2);
2132 error = PR_TIMING_STATISTICAL;
2135 if (arg2 != PR_TIMING_STATISTICAL)
2139 comm[sizeof(me->comm) - 1] = 0;
2140 if (strncpy_from_user(comm, (char __user *)arg2,
2141 sizeof(me->comm) - 1) < 0)
2143 set_task_comm(me, comm);
2144 proc_comm_connector(me);
2147 get_task_comm(comm, me);
2148 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2152 error = GET_ENDIAN(me, arg2);
2155 error = SET_ENDIAN(me, arg2);
2157 case PR_GET_SECCOMP:
2158 error = prctl_get_seccomp();
2160 case PR_SET_SECCOMP:
2161 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2164 error = GET_TSC_CTL(arg2);
2167 error = SET_TSC_CTL(arg2);
2169 case PR_TASK_PERF_EVENTS_DISABLE:
2170 error = perf_event_task_disable();
2172 case PR_TASK_PERF_EVENTS_ENABLE:
2173 error = perf_event_task_enable();
2175 case PR_GET_TIMERSLACK:
2176 error = current->timer_slack_ns;
2178 case PR_SET_TIMERSLACK:
2180 current->timer_slack_ns =
2181 current->default_timer_slack_ns;
2183 current->timer_slack_ns = arg2;
2189 case PR_MCE_KILL_CLEAR:
2192 current->flags &= ~PF_MCE_PROCESS;
2194 case PR_MCE_KILL_SET:
2195 current->flags |= PF_MCE_PROCESS;
2196 if (arg3 == PR_MCE_KILL_EARLY)
2197 current->flags |= PF_MCE_EARLY;
2198 else if (arg3 == PR_MCE_KILL_LATE)
2199 current->flags &= ~PF_MCE_EARLY;
2200 else if (arg3 == PR_MCE_KILL_DEFAULT)
2202 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2210 case PR_MCE_KILL_GET:
2211 if (arg2 | arg3 | arg4 | arg5)
2213 if (current->flags & PF_MCE_PROCESS)
2214 error = (current->flags & PF_MCE_EARLY) ?
2215 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2217 error = PR_MCE_KILL_DEFAULT;
2220 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2222 case PR_GET_TID_ADDRESS:
2223 error = prctl_get_tid_address(me, (int __user **)arg2);
2225 case PR_SET_CHILD_SUBREAPER:
2226 me->signal->is_child_subreaper = !!arg2;
2228 case PR_GET_CHILD_SUBREAPER:
2229 error = put_user(me->signal->is_child_subreaper,
2230 (int __user *)arg2);
2232 case PR_SET_NO_NEW_PRIVS:
2233 if (arg2 != 1 || arg3 || arg4 || arg5)
2236 current->no_new_privs = 1;
2238 case PR_GET_NO_NEW_PRIVS:
2239 if (arg2 || arg3 || arg4 || arg5)
2241 return current->no_new_privs ? 1 : 0;
2249 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2250 struct getcpu_cache __user *, unused)
2253 int cpu = raw_smp_processor_id();
2255 err |= put_user(cpu, cpup);
2257 err |= put_user(cpu_to_node(cpu), nodep);
2258 return err ? -EFAULT : 0;
2261 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2263 static int __orderly_poweroff(bool force)
2266 static char *envp[] = {
2268 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2273 argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL);
2275 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
2278 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2279 __func__, poweroff_cmd);
2284 printk(KERN_WARNING "Failed to start orderly shutdown: "
2285 "forcing the issue\n");
2287 * I guess this should try to kick off some daemon to sync and
2288 * poweroff asap. Or not even bother syncing if we're doing an
2289 * emergency shutdown?
2298 static bool poweroff_force;
2300 static void poweroff_work_func(struct work_struct *work)
2302 __orderly_poweroff(poweroff_force);
2305 static DECLARE_WORK(poweroff_work, poweroff_work_func);
2308 * orderly_poweroff - Trigger an orderly system poweroff
2309 * @force: force poweroff if command execution fails
2311 * This may be called from any context to trigger a system shutdown.
2312 * If the orderly shutdown fails, it will force an immediate shutdown.
2314 int orderly_poweroff(bool force)
2316 if (force) /* do not override the pending "true" */
2317 poweroff_force = true;
2318 schedule_work(&poweroff_work);
2321 EXPORT_SYMBOL_GPL(orderly_poweroff);
2324 * do_sysinfo - fill in sysinfo struct
2325 * @info: pointer to buffer to fill
2327 static int do_sysinfo(struct sysinfo *info)
2329 unsigned long mem_total, sav_total;
2330 unsigned int mem_unit, bitcount;
2333 memset(info, 0, sizeof(struct sysinfo));
2336 monotonic_to_bootbased(&tp);
2337 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2339 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2341 info->procs = nr_threads;
2347 * If the sum of all the available memory (i.e. ram + swap)
2348 * is less than can be stored in a 32 bit unsigned long then
2349 * we can be binary compatible with 2.2.x kernels. If not,
2350 * well, in that case 2.2.x was broken anyways...
2352 * -Erik Andersen <andersee@debian.org>
2355 mem_total = info->totalram + info->totalswap;
2356 if (mem_total < info->totalram || mem_total < info->totalswap)
2359 mem_unit = info->mem_unit;
2360 while (mem_unit > 1) {
2363 sav_total = mem_total;
2365 if (mem_total < sav_total)
2370 * If mem_total did not overflow, multiply all memory values by
2371 * info->mem_unit and set it to 1. This leaves things compatible
2372 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2377 info->totalram <<= bitcount;
2378 info->freeram <<= bitcount;
2379 info->sharedram <<= bitcount;
2380 info->bufferram <<= bitcount;
2381 info->totalswap <<= bitcount;
2382 info->freeswap <<= bitcount;
2383 info->totalhigh <<= bitcount;
2384 info->freehigh <<= bitcount;
2390 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2396 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2402 #ifdef CONFIG_COMPAT
2403 struct compat_sysinfo {
2417 char _f[20-2*sizeof(u32)-sizeof(int)];
2420 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2426 /* Check to see if any memory value is too large for 32-bit and scale
2429 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2432 while (s.mem_unit < PAGE_SIZE) {
2437 s.totalram >>= bitcount;
2438 s.freeram >>= bitcount;
2439 s.sharedram >>= bitcount;
2440 s.bufferram >>= bitcount;
2441 s.totalswap >>= bitcount;
2442 s.freeswap >>= bitcount;
2443 s.totalhigh >>= bitcount;
2444 s.freehigh >>= bitcount;
2447 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2448 __put_user(s.uptime, &info->uptime) ||
2449 __put_user(s.loads[0], &info->loads[0]) ||
2450 __put_user(s.loads[1], &info->loads[1]) ||
2451 __put_user(s.loads[2], &info->loads[2]) ||
2452 __put_user(s.totalram, &info->totalram) ||
2453 __put_user(s.freeram, &info->freeram) ||
2454 __put_user(s.sharedram, &info->sharedram) ||
2455 __put_user(s.bufferram, &info->bufferram) ||
2456 __put_user(s.totalswap, &info->totalswap) ||
2457 __put_user(s.freeswap, &info->freeswap) ||
2458 __put_user(s.procs, &info->procs) ||
2459 __put_user(s.totalhigh, &info->totalhigh) ||
2460 __put_user(s.freehigh, &info->freehigh) ||
2461 __put_user(s.mem_unit, &info->mem_unit))
2466 #endif /* CONFIG_COMPAT */