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/smp_lock.h>
12 #include <linux/notifier.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.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>
37 #include <linux/compat.h>
38 #include <linux/syscalls.h>
39 #include <linux/kprobes.h>
40 #include <linux/user_namespace.h>
42 #include <asm/uaccess.h>
44 #include <asm/unistd.h>
46 #ifndef SET_UNALIGN_CTL
47 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
49 #ifndef GET_UNALIGN_CTL
50 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
53 # define SET_FPEMU_CTL(a,b) (-EINVAL)
56 # define GET_FPEMU_CTL(a,b) (-EINVAL)
59 # define SET_FPEXC_CTL(a,b) (-EINVAL)
62 # define GET_FPEXC_CTL(a,b) (-EINVAL)
65 # define GET_ENDIAN(a,b) (-EINVAL)
68 # define SET_ENDIAN(a,b) (-EINVAL)
71 # define GET_TSC_CTL(a) (-EINVAL)
74 # define SET_TSC_CTL(a) (-EINVAL)
78 * this is where the system-wide overflow UID and GID are defined, for
79 * architectures that now have 32-bit UID/GID but didn't in the past
82 int overflowuid = DEFAULT_OVERFLOWUID;
83 int overflowgid = DEFAULT_OVERFLOWGID;
86 EXPORT_SYMBOL(overflowuid);
87 EXPORT_SYMBOL(overflowgid);
91 * the same as above, but for filesystems which can only store a 16-bit
92 * UID and GID. as such, this is needed on all architectures
95 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
96 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
98 EXPORT_SYMBOL(fs_overflowuid);
99 EXPORT_SYMBOL(fs_overflowgid);
102 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
107 EXPORT_SYMBOL(cad_pid);
110 * If set, this is used for preparing the system to power off.
113 void (*pm_power_off_prepare)(void);
116 * set the priority of a task
117 * - the caller must hold the RCU read lock
119 static int set_one_prio(struct task_struct *p, int niceval, int error)
121 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
124 if (pcred->uid != cred->euid &&
125 pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
129 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
133 no_nice = security_task_setnice(p, niceval);
140 set_user_nice(p, niceval);
145 asmlinkage long sys_setpriority(int which, int who, int niceval)
147 struct task_struct *g, *p;
148 struct user_struct *user;
149 const struct cred *cred = current_cred();
153 if (which > PRIO_USER || which < PRIO_PROCESS)
156 /* normalize: avoid signed division (rounding problems) */
163 read_lock(&tasklist_lock);
167 p = find_task_by_vpid(who);
171 error = set_one_prio(p, niceval, error);
175 pgrp = find_vpid(who);
177 pgrp = task_pgrp(current);
178 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
179 error = set_one_prio(p, niceval, error);
180 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
183 user = (struct user_struct *) cred->user;
186 else if ((who != cred->uid) &&
187 !(user = find_user(who)))
188 goto out_unlock; /* No processes for this user */
191 if (__task_cred(p)->uid == who)
192 error = set_one_prio(p, niceval, error);
193 while_each_thread(g, p);
194 if (who != cred->uid)
195 free_uid(user); /* For find_user() */
199 read_unlock(&tasklist_lock);
205 * Ugh. To avoid negative return values, "getpriority()" will
206 * not return the normal nice-value, but a negated value that
207 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
208 * to stay compatible.
210 asmlinkage long sys_getpriority(int which, int who)
212 struct task_struct *g, *p;
213 struct user_struct *user;
214 const struct cred *cred = current_cred();
215 long niceval, retval = -ESRCH;
218 if (which > PRIO_USER || which < PRIO_PROCESS)
221 read_lock(&tasklist_lock);
225 p = find_task_by_vpid(who);
229 niceval = 20 - task_nice(p);
230 if (niceval > retval)
236 pgrp = find_vpid(who);
238 pgrp = task_pgrp(current);
239 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
240 niceval = 20 - task_nice(p);
241 if (niceval > retval)
243 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
246 user = (struct user_struct *) cred->user;
249 else if ((who != cred->uid) &&
250 !(user = find_user(who)))
251 goto out_unlock; /* No processes for this user */
254 if (__task_cred(p)->uid == who) {
255 niceval = 20 - task_nice(p);
256 if (niceval > retval)
259 while_each_thread(g, p);
260 if (who != cred->uid)
261 free_uid(user); /* for find_user() */
265 read_unlock(&tasklist_lock);
271 * emergency_restart - reboot the system
273 * Without shutting down any hardware or taking any locks
274 * reboot the system. This is called when we know we are in
275 * trouble so this is our best effort to reboot. This is
276 * safe to call in interrupt context.
278 void emergency_restart(void)
280 machine_emergency_restart();
282 EXPORT_SYMBOL_GPL(emergency_restart);
284 void kernel_restart_prepare(char *cmd)
286 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
287 system_state = SYSTEM_RESTART;
293 * kernel_restart - reboot the system
294 * @cmd: pointer to buffer containing command to execute for restart
297 * Shutdown everything and perform a clean reboot.
298 * This is not safe to call in interrupt context.
300 void kernel_restart(char *cmd)
302 kernel_restart_prepare(cmd);
304 printk(KERN_EMERG "Restarting system.\n");
306 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
307 machine_restart(cmd);
309 EXPORT_SYMBOL_GPL(kernel_restart);
311 static void kernel_shutdown_prepare(enum system_states state)
313 blocking_notifier_call_chain(&reboot_notifier_list,
314 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
315 system_state = state;
319 * kernel_halt - halt the system
321 * Shutdown everything and perform a clean system halt.
323 void kernel_halt(void)
325 kernel_shutdown_prepare(SYSTEM_HALT);
327 printk(KERN_EMERG "System halted.\n");
331 EXPORT_SYMBOL_GPL(kernel_halt);
334 * kernel_power_off - power_off the system
336 * Shutdown everything and perform a clean system power_off.
338 void kernel_power_off(void)
340 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
341 if (pm_power_off_prepare)
342 pm_power_off_prepare();
343 disable_nonboot_cpus();
345 printk(KERN_EMERG "Power down.\n");
348 EXPORT_SYMBOL_GPL(kernel_power_off);
350 * Reboot system call: for obvious reasons only root may call it,
351 * and even root needs to set up some magic numbers in the registers
352 * so that some mistake won't make this reboot the whole machine.
353 * You can also set the meaning of the ctrl-alt-del-key here.
355 * reboot doesn't sync: do that yourself before calling this.
357 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
361 /* We only trust the superuser with rebooting the system. */
362 if (!capable(CAP_SYS_BOOT))
365 /* For safety, we require "magic" arguments. */
366 if (magic1 != LINUX_REBOOT_MAGIC1 ||
367 (magic2 != LINUX_REBOOT_MAGIC2 &&
368 magic2 != LINUX_REBOOT_MAGIC2A &&
369 magic2 != LINUX_REBOOT_MAGIC2B &&
370 magic2 != LINUX_REBOOT_MAGIC2C))
373 /* Instead of trying to make the power_off code look like
374 * halt when pm_power_off is not set do it the easy way.
376 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
377 cmd = LINUX_REBOOT_CMD_HALT;
381 case LINUX_REBOOT_CMD_RESTART:
382 kernel_restart(NULL);
385 case LINUX_REBOOT_CMD_CAD_ON:
389 case LINUX_REBOOT_CMD_CAD_OFF:
393 case LINUX_REBOOT_CMD_HALT:
399 case LINUX_REBOOT_CMD_POWER_OFF:
405 case LINUX_REBOOT_CMD_RESTART2:
406 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
410 buffer[sizeof(buffer) - 1] = '\0';
412 kernel_restart(buffer);
416 case LINUX_REBOOT_CMD_KEXEC:
419 ret = kernel_kexec();
425 #ifdef CONFIG_HIBERNATION
426 case LINUX_REBOOT_CMD_SW_SUSPEND:
428 int ret = hibernate();
442 static void deferred_cad(struct work_struct *dummy)
444 kernel_restart(NULL);
448 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
449 * As it's called within an interrupt, it may NOT sync: the only choice
450 * is whether to reboot at once, or just ignore the ctrl-alt-del.
452 void ctrl_alt_del(void)
454 static DECLARE_WORK(cad_work, deferred_cad);
457 schedule_work(&cad_work);
459 kill_cad_pid(SIGINT, 1);
463 * Unprivileged users may change the real gid to the effective gid
464 * or vice versa. (BSD-style)
466 * If you set the real gid at all, or set the effective gid to a value not
467 * equal to the real gid, then the saved gid is set to the new effective gid.
469 * This makes it possible for a setgid program to completely drop its
470 * privileges, which is often a useful assertion to make when you are doing
471 * a security audit over a program.
473 * The general idea is that a program which uses just setregid() will be
474 * 100% compatible with BSD. A program which uses just setgid() will be
475 * 100% compatible with POSIX with saved IDs.
477 * SMP: There are not races, the GIDs are checked only by filesystem
478 * operations (as far as semantic preservation is concerned).
480 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
482 const struct cred *old;
486 new = prepare_creds();
489 old = current_cred();
491 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
496 if (rgid != (gid_t) -1) {
497 if (old->gid == rgid ||
504 if (egid != (gid_t) -1) {
505 if (old->gid == egid ||
514 if (rgid != (gid_t) -1 ||
515 (egid != (gid_t) -1 && egid != old->gid))
516 new->sgid = new->egid;
517 new->fsgid = new->egid;
519 return commit_creds(new);
527 * setgid() is implemented like SysV w/ SAVED_IDS
529 * SMP: Same implicit races as above.
531 asmlinkage long sys_setgid(gid_t gid)
533 const struct cred *old;
537 new = prepare_creds();
540 old = current_cred();
542 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
547 if (capable(CAP_SETGID))
548 new->gid = new->egid = new->sgid = new->fsgid = gid;
549 else if (gid == old->gid || gid == old->sgid)
550 new->egid = new->fsgid = gid;
554 return commit_creds(new);
562 * change the user struct in a credentials set to match the new UID
564 static int set_user(struct cred *new)
566 struct user_struct *new_user;
568 new_user = alloc_uid(current_user_ns(), new->uid);
572 if (atomic_read(&new_user->processes) >=
573 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
574 new_user != INIT_USER) {
580 new->user = new_user;
585 * Unprivileged users may change the real uid to the effective uid
586 * or vice versa. (BSD-style)
588 * If you set the real uid at all, or set the effective uid to a value not
589 * equal to the real uid, then the saved uid is set to the new effective uid.
591 * This makes it possible for a setuid program to completely drop its
592 * privileges, which is often a useful assertion to make when you are doing
593 * a security audit over a program.
595 * The general idea is that a program which uses just setreuid() will be
596 * 100% compatible with BSD. A program which uses just setuid() will be
597 * 100% compatible with POSIX with saved IDs.
599 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
601 const struct cred *old;
605 new = prepare_creds();
608 old = current_cred();
610 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
615 if (ruid != (uid_t) -1) {
617 if (old->uid != ruid &&
619 !capable(CAP_SETUID))
623 if (euid != (uid_t) -1) {
625 if (old->uid != euid &&
628 !capable(CAP_SETUID))
633 if (new->uid != old->uid && set_user(new) < 0)
636 if (ruid != (uid_t) -1 ||
637 (euid != (uid_t) -1 && euid != old->uid))
638 new->suid = new->euid;
639 new->fsuid = new->euid;
641 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
645 return commit_creds(new);
653 * setuid() is implemented like SysV with SAVED_IDS
655 * Note that SAVED_ID's is deficient in that a setuid root program
656 * like sendmail, for example, cannot set its uid to be a normal
657 * user and then switch back, because if you're root, setuid() sets
658 * the saved uid too. If you don't like this, blame the bright people
659 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
660 * will allow a root program to temporarily drop privileges and be able to
661 * regain them by swapping the real and effective uid.
663 asmlinkage long sys_setuid(uid_t uid)
665 const struct cred *old;
669 new = prepare_creds();
672 old = current_cred();
674 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
679 if (capable(CAP_SETUID)) {
680 new->suid = new->uid = uid;
681 if (uid != old->uid && set_user(new) < 0) {
685 } else if (uid != old->uid && uid != new->suid) {
689 new->fsuid = new->euid = uid;
691 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
695 return commit_creds(new);
704 * This function implements a generic ability to update ruid, euid,
705 * and suid. This allows you to implement the 4.4 compatible seteuid().
707 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
709 const struct cred *old;
713 new = prepare_creds();
717 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
720 old = current_cred();
723 if (!capable(CAP_SETUID)) {
724 if (ruid != (uid_t) -1 && ruid != old->uid &&
725 ruid != old->euid && ruid != old->suid)
727 if (euid != (uid_t) -1 && euid != old->uid &&
728 euid != old->euid && euid != old->suid)
730 if (suid != (uid_t) -1 && suid != old->uid &&
731 suid != old->euid && suid != old->suid)
736 if (ruid != (uid_t) -1) {
738 if (ruid != old->uid && set_user(new) < 0)
741 if (euid != (uid_t) -1)
743 if (suid != (uid_t) -1)
745 new->fsuid = new->euid;
747 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
751 return commit_creds(new);
758 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
760 const struct cred *cred = current_cred();
763 if (!(retval = put_user(cred->uid, ruid)) &&
764 !(retval = put_user(cred->euid, euid)))
765 retval = put_user(cred->suid, suid);
771 * Same as above, but for rgid, egid, sgid.
773 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
775 const struct cred *old;
779 new = prepare_creds();
782 old = current_cred();
784 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
789 if (!capable(CAP_SETGID)) {
790 if (rgid != (gid_t) -1 && rgid != old->gid &&
791 rgid != old->egid && rgid != old->sgid)
793 if (egid != (gid_t) -1 && egid != old->gid &&
794 egid != old->egid && egid != old->sgid)
796 if (sgid != (gid_t) -1 && sgid != old->gid &&
797 sgid != old->egid && sgid != old->sgid)
801 if (rgid != (gid_t) -1)
803 if (egid != (gid_t) -1)
805 if (sgid != (gid_t) -1)
807 new->fsgid = new->egid;
809 return commit_creds(new);
816 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
818 const struct cred *cred = current_cred();
821 if (!(retval = put_user(cred->gid, rgid)) &&
822 !(retval = put_user(cred->egid, egid)))
823 retval = put_user(cred->sgid, sgid);
830 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
831 * is used for "access()" and for the NFS daemon (letting nfsd stay at
832 * whatever uid it wants to). It normally shadows "euid", except when
833 * explicitly set by setfsuid() or for access..
835 asmlinkage long sys_setfsuid(uid_t uid)
837 const struct cred *old;
841 new = prepare_creds();
843 return current_fsuid();
844 old = current_cred();
845 old_fsuid = old->fsuid;
847 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
850 if (uid == old->uid || uid == old->euid ||
851 uid == old->suid || uid == old->fsuid ||
852 capable(CAP_SETUID)) {
853 if (uid != old_fsuid) {
855 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
870 * Samma på svenska..
872 asmlinkage long sys_setfsgid(gid_t gid)
874 const struct cred *old;
878 new = prepare_creds();
880 return current_fsgid();
881 old = current_cred();
882 old_fsgid = old->fsgid;
884 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
887 if (gid == old->gid || gid == old->egid ||
888 gid == old->sgid || gid == old->fsgid ||
889 capable(CAP_SETGID)) {
890 if (gid != old_fsgid) {
905 void do_sys_times(struct tms *tms)
907 struct task_cputime cputime;
908 cputime_t cutime, cstime;
910 thread_group_cputime(current, &cputime);
911 spin_lock_irq(¤t->sighand->siglock);
912 cutime = current->signal->cutime;
913 cstime = current->signal->cstime;
914 spin_unlock_irq(¤t->sighand->siglock);
915 tms->tms_utime = cputime_to_clock_t(cputime.utime);
916 tms->tms_stime = cputime_to_clock_t(cputime.stime);
917 tms->tms_cutime = cputime_to_clock_t(cutime);
918 tms->tms_cstime = cputime_to_clock_t(cstime);
921 asmlinkage long sys_times(struct tms __user * tbuf)
927 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
930 return (long) jiffies_64_to_clock_t(get_jiffies_64());
934 * This needs some heavy checking ...
935 * I just haven't the stomach for it. I also don't fully
936 * understand sessions/pgrp etc. Let somebody who does explain it.
938 * OK, I think I have the protection semantics right.... this is really
939 * only important on a multi-user system anyway, to make sure one user
940 * can't send a signal to a process owned by another. -TYT, 12/12/91
942 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
945 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
947 struct task_struct *p;
948 struct task_struct *group_leader = current->group_leader;
953 pid = task_pid_vnr(group_leader);
959 /* From this point forward we keep holding onto the tasklist lock
960 * so that our parent does not change from under us. -DaveM
962 write_lock_irq(&tasklist_lock);
965 p = find_task_by_vpid(pid);
970 if (!thread_group_leader(p))
973 if (same_thread_group(p->real_parent, group_leader)) {
975 if (task_session(p) != task_session(group_leader))
982 if (p != group_leader)
987 if (p->signal->leader)
992 struct task_struct *g;
994 pgrp = find_vpid(pgid);
995 g = pid_task(pgrp, PIDTYPE_PGID);
996 if (!g || task_session(g) != task_session(group_leader))
1000 err = security_task_setpgid(p, pgid);
1004 if (task_pgrp(p) != pgrp) {
1005 change_pid(p, PIDTYPE_PGID, pgrp);
1006 set_task_pgrp(p, pid_nr(pgrp));
1011 /* All paths lead to here, thus we are safe. -DaveM */
1012 write_unlock_irq(&tasklist_lock);
1016 asmlinkage long sys_getpgid(pid_t pid)
1018 struct task_struct *p;
1024 grp = task_pgrp(current);
1027 p = find_task_by_vpid(pid);
1034 retval = security_task_getpgid(p);
1038 retval = pid_vnr(grp);
1044 #ifdef __ARCH_WANT_SYS_GETPGRP
1046 asmlinkage long sys_getpgrp(void)
1048 return sys_getpgid(0);
1053 asmlinkage long sys_getsid(pid_t pid)
1055 struct task_struct *p;
1061 sid = task_session(current);
1064 p = find_task_by_vpid(pid);
1067 sid = task_session(p);
1071 retval = security_task_getsid(p);
1075 retval = pid_vnr(sid);
1081 asmlinkage long sys_setsid(void)
1083 struct task_struct *group_leader = current->group_leader;
1084 struct pid *sid = task_pid(group_leader);
1085 pid_t session = pid_vnr(sid);
1088 write_lock_irq(&tasklist_lock);
1089 /* Fail if I am already a session leader */
1090 if (group_leader->signal->leader)
1093 /* Fail if a process group id already exists that equals the
1094 * proposed session id.
1096 if (pid_task(sid, PIDTYPE_PGID))
1099 group_leader->signal->leader = 1;
1100 __set_special_pids(sid);
1102 proc_clear_tty(group_leader);
1106 write_unlock_irq(&tasklist_lock);
1111 * Supplementary group IDs
1114 /* init to 2 - one for init_task, one to ensure it is never freed */
1115 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1117 struct group_info *groups_alloc(int gidsetsize)
1119 struct group_info *group_info;
1123 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1124 /* Make sure we always allocate at least one indirect block pointer */
1125 nblocks = nblocks ? : 1;
1126 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1129 group_info->ngroups = gidsetsize;
1130 group_info->nblocks = nblocks;
1131 atomic_set(&group_info->usage, 1);
1133 if (gidsetsize <= NGROUPS_SMALL)
1134 group_info->blocks[0] = group_info->small_block;
1136 for (i = 0; i < nblocks; i++) {
1138 b = (void *)__get_free_page(GFP_USER);
1140 goto out_undo_partial_alloc;
1141 group_info->blocks[i] = b;
1146 out_undo_partial_alloc:
1148 free_page((unsigned long)group_info->blocks[i]);
1154 EXPORT_SYMBOL(groups_alloc);
1156 void groups_free(struct group_info *group_info)
1158 if (group_info->blocks[0] != group_info->small_block) {
1160 for (i = 0; i < group_info->nblocks; i++)
1161 free_page((unsigned long)group_info->blocks[i]);
1166 EXPORT_SYMBOL(groups_free);
1168 /* export the group_info to a user-space array */
1169 static int groups_to_user(gid_t __user *grouplist,
1170 const struct group_info *group_info)
1173 unsigned int count = group_info->ngroups;
1175 for (i = 0; i < group_info->nblocks; i++) {
1176 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1177 unsigned int len = cp_count * sizeof(*grouplist);
1179 if (copy_to_user(grouplist, group_info->blocks[i], len))
1182 grouplist += NGROUPS_PER_BLOCK;
1188 /* fill a group_info from a user-space array - it must be allocated already */
1189 static int groups_from_user(struct group_info *group_info,
1190 gid_t __user *grouplist)
1193 unsigned int count = group_info->ngroups;
1195 for (i = 0; i < group_info->nblocks; i++) {
1196 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1197 unsigned int len = cp_count * sizeof(*grouplist);
1199 if (copy_from_user(group_info->blocks[i], grouplist, len))
1202 grouplist += NGROUPS_PER_BLOCK;
1208 /* a simple Shell sort */
1209 static void groups_sort(struct group_info *group_info)
1211 int base, max, stride;
1212 int gidsetsize = group_info->ngroups;
1214 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1219 max = gidsetsize - stride;
1220 for (base = 0; base < max; base++) {
1222 int right = left + stride;
1223 gid_t tmp = GROUP_AT(group_info, right);
1225 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1226 GROUP_AT(group_info, right) =
1227 GROUP_AT(group_info, left);
1231 GROUP_AT(group_info, right) = tmp;
1237 /* a simple bsearch */
1238 int groups_search(const struct group_info *group_info, gid_t grp)
1240 unsigned int left, right;
1246 right = group_info->ngroups;
1247 while (left < right) {
1248 unsigned int mid = (left+right)/2;
1249 int cmp = grp - GROUP_AT(group_info, mid);
1261 * set_groups - Change a group subscription in a set of credentials
1262 * @new: The newly prepared set of credentials to alter
1263 * @group_info: The group list to install
1265 * Validate a group subscription and, if valid, insert it into a set
1268 int set_groups(struct cred *new, struct group_info *group_info)
1272 retval = security_task_setgroups(group_info);
1276 put_group_info(new->group_info);
1277 groups_sort(group_info);
1278 get_group_info(group_info);
1279 new->group_info = group_info;
1283 EXPORT_SYMBOL(set_groups);
1286 * set_current_groups - Change current's group subscription
1287 * @group_info: The group list to impose
1289 * Validate a group subscription and, if valid, impose it upon current's task
1292 int set_current_groups(struct group_info *group_info)
1297 new = prepare_creds();
1301 ret = set_groups(new, group_info);
1307 return commit_creds(new);
1310 EXPORT_SYMBOL(set_current_groups);
1312 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1314 const struct cred *cred = current_cred();
1320 /* no need to grab task_lock here; it cannot change */
1321 i = cred->group_info->ngroups;
1323 if (i > gidsetsize) {
1327 if (groups_to_user(grouplist, cred->group_info)) {
1337 * SMP: Our groups are copy-on-write. We can set them safely
1338 * without another task interfering.
1341 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1343 struct group_info *group_info;
1346 if (!capable(CAP_SETGID))
1348 if ((unsigned)gidsetsize > NGROUPS_MAX)
1351 group_info = groups_alloc(gidsetsize);
1354 retval = groups_from_user(group_info, grouplist);
1356 put_group_info(group_info);
1360 retval = set_current_groups(group_info);
1361 put_group_info(group_info);
1367 * Check whether we're fsgid/egid or in the supplemental group..
1369 int in_group_p(gid_t grp)
1371 const struct cred *cred = current_cred();
1374 if (grp != cred->fsgid)
1375 retval = groups_search(cred->group_info, grp);
1379 EXPORT_SYMBOL(in_group_p);
1381 int in_egroup_p(gid_t grp)
1383 const struct cred *cred = current_cred();
1386 if (grp != cred->egid)
1387 retval = groups_search(cred->group_info, grp);
1391 EXPORT_SYMBOL(in_egroup_p);
1393 DECLARE_RWSEM(uts_sem);
1395 asmlinkage long sys_newuname(struct new_utsname __user * name)
1399 down_read(&uts_sem);
1400 if (copy_to_user(name, utsname(), sizeof *name))
1406 asmlinkage long sys_sethostname(char __user *name, int len)
1409 char tmp[__NEW_UTS_LEN];
1411 if (!capable(CAP_SYS_ADMIN))
1413 if (len < 0 || len > __NEW_UTS_LEN)
1415 down_write(&uts_sem);
1417 if (!copy_from_user(tmp, name, len)) {
1418 struct new_utsname *u = utsname();
1420 memcpy(u->nodename, tmp, len);
1421 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1428 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1430 asmlinkage long sys_gethostname(char __user *name, int len)
1433 struct new_utsname *u;
1437 down_read(&uts_sem);
1439 i = 1 + strlen(u->nodename);
1443 if (copy_to_user(name, u->nodename, i))
1452 * Only setdomainname; getdomainname can be implemented by calling
1455 asmlinkage long sys_setdomainname(char __user *name, int len)
1458 char tmp[__NEW_UTS_LEN];
1460 if (!capable(CAP_SYS_ADMIN))
1462 if (len < 0 || len > __NEW_UTS_LEN)
1465 down_write(&uts_sem);
1467 if (!copy_from_user(tmp, name, len)) {
1468 struct new_utsname *u = utsname();
1470 memcpy(u->domainname, tmp, len);
1471 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1478 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1480 if (resource >= RLIM_NLIMITS)
1483 struct rlimit value;
1484 task_lock(current->group_leader);
1485 value = current->signal->rlim[resource];
1486 task_unlock(current->group_leader);
1487 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1491 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1494 * Back compatibility for getrlimit. Needed for some apps.
1497 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1500 if (resource >= RLIM_NLIMITS)
1503 task_lock(current->group_leader);
1504 x = current->signal->rlim[resource];
1505 task_unlock(current->group_leader);
1506 if (x.rlim_cur > 0x7FFFFFFF)
1507 x.rlim_cur = 0x7FFFFFFF;
1508 if (x.rlim_max > 0x7FFFFFFF)
1509 x.rlim_max = 0x7FFFFFFF;
1510 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1515 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1517 struct rlimit new_rlim, *old_rlim;
1520 if (resource >= RLIM_NLIMITS)
1522 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1524 old_rlim = current->signal->rlim + resource;
1525 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1526 !capable(CAP_SYS_RESOURCE))
1529 if (resource == RLIMIT_NOFILE) {
1530 if (new_rlim.rlim_max == RLIM_INFINITY)
1531 new_rlim.rlim_max = sysctl_nr_open;
1532 if (new_rlim.rlim_cur == RLIM_INFINITY)
1533 new_rlim.rlim_cur = sysctl_nr_open;
1534 if (new_rlim.rlim_max > sysctl_nr_open)
1538 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1541 retval = security_task_setrlimit(resource, &new_rlim);
1545 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1547 * The caller is asking for an immediate RLIMIT_CPU
1548 * expiry. But we use the zero value to mean "it was
1549 * never set". So let's cheat and make it one second
1552 new_rlim.rlim_cur = 1;
1555 task_lock(current->group_leader);
1556 *old_rlim = new_rlim;
1557 task_unlock(current->group_leader);
1559 if (resource != RLIMIT_CPU)
1563 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1564 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1565 * very long-standing error, and fixing it now risks breakage of
1566 * applications, so we live with it
1568 if (new_rlim.rlim_cur == RLIM_INFINITY)
1571 update_rlimit_cpu(new_rlim.rlim_cur);
1577 * It would make sense to put struct rusage in the task_struct,
1578 * except that would make the task_struct be *really big*. After
1579 * task_struct gets moved into malloc'ed memory, it would
1580 * make sense to do this. It will make moving the rest of the information
1581 * a lot simpler! (Which we're not doing right now because we're not
1582 * measuring them yet).
1584 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1585 * races with threads incrementing their own counters. But since word
1586 * reads are atomic, we either get new values or old values and we don't
1587 * care which for the sums. We always take the siglock to protect reading
1588 * the c* fields from p->signal from races with exit.c updating those
1589 * fields when reaping, so a sample either gets all the additions of a
1590 * given child after it's reaped, or none so this sample is before reaping.
1593 * We need to take the siglock for CHILDEREN, SELF and BOTH
1594 * for the cases current multithreaded, non-current single threaded
1595 * non-current multithreaded. Thread traversal is now safe with
1597 * Strictly speaking, we donot need to take the siglock if we are current and
1598 * single threaded, as no one else can take our signal_struct away, no one
1599 * else can reap the children to update signal->c* counters, and no one else
1600 * can race with the signal-> fields. If we do not take any lock, the
1601 * signal-> fields could be read out of order while another thread was just
1602 * exiting. So we should place a read memory barrier when we avoid the lock.
1603 * On the writer side, write memory barrier is implied in __exit_signal
1604 * as __exit_signal releases the siglock spinlock after updating the signal->
1605 * fields. But we don't do this yet to keep things simple.
1609 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1611 r->ru_nvcsw += t->nvcsw;
1612 r->ru_nivcsw += t->nivcsw;
1613 r->ru_minflt += t->min_flt;
1614 r->ru_majflt += t->maj_flt;
1615 r->ru_inblock += task_io_get_inblock(t);
1616 r->ru_oublock += task_io_get_oublock(t);
1619 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1621 struct task_struct *t;
1622 unsigned long flags;
1623 cputime_t utime, stime;
1624 struct task_cputime cputime;
1626 memset((char *) r, 0, sizeof *r);
1627 utime = stime = cputime_zero;
1629 if (who == RUSAGE_THREAD) {
1630 accumulate_thread_rusage(p, r);
1634 if (!lock_task_sighand(p, &flags))
1639 case RUSAGE_CHILDREN:
1640 utime = p->signal->cutime;
1641 stime = p->signal->cstime;
1642 r->ru_nvcsw = p->signal->cnvcsw;
1643 r->ru_nivcsw = p->signal->cnivcsw;
1644 r->ru_minflt = p->signal->cmin_flt;
1645 r->ru_majflt = p->signal->cmaj_flt;
1646 r->ru_inblock = p->signal->cinblock;
1647 r->ru_oublock = p->signal->coublock;
1649 if (who == RUSAGE_CHILDREN)
1653 thread_group_cputime(p, &cputime);
1654 utime = cputime_add(utime, cputime.utime);
1655 stime = cputime_add(stime, cputime.stime);
1656 r->ru_nvcsw += p->signal->nvcsw;
1657 r->ru_nivcsw += p->signal->nivcsw;
1658 r->ru_minflt += p->signal->min_flt;
1659 r->ru_majflt += p->signal->maj_flt;
1660 r->ru_inblock += p->signal->inblock;
1661 r->ru_oublock += p->signal->oublock;
1664 accumulate_thread_rusage(t, r);
1672 unlock_task_sighand(p, &flags);
1675 cputime_to_timeval(utime, &r->ru_utime);
1676 cputime_to_timeval(stime, &r->ru_stime);
1679 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1682 k_getrusage(p, who, &r);
1683 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1686 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1688 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1689 who != RUSAGE_THREAD)
1691 return getrusage(current, who, ru);
1694 asmlinkage long sys_umask(int mask)
1696 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1700 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1701 unsigned long arg4, unsigned long arg5)
1703 struct task_struct *me = current;
1704 unsigned char comm[sizeof(me->comm)];
1707 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1708 if (error != -ENOSYS)
1713 case PR_SET_PDEATHSIG:
1714 if (!valid_signal(arg2)) {
1718 me->pdeath_signal = arg2;
1721 case PR_GET_PDEATHSIG:
1722 error = put_user(me->pdeath_signal, (int __user *)arg2);
1724 case PR_GET_DUMPABLE:
1725 error = get_dumpable(me->mm);
1727 case PR_SET_DUMPABLE:
1728 if (arg2 < 0 || arg2 > 1) {
1732 set_dumpable(me->mm, arg2);
1736 case PR_SET_UNALIGN:
1737 error = SET_UNALIGN_CTL(me, arg2);
1739 case PR_GET_UNALIGN:
1740 error = GET_UNALIGN_CTL(me, arg2);
1743 error = SET_FPEMU_CTL(me, arg2);
1746 error = GET_FPEMU_CTL(me, arg2);
1749 error = SET_FPEXC_CTL(me, arg2);
1752 error = GET_FPEXC_CTL(me, arg2);
1755 error = PR_TIMING_STATISTICAL;
1758 if (arg2 != PR_TIMING_STATISTICAL)
1765 comm[sizeof(me->comm)-1] = 0;
1766 if (strncpy_from_user(comm, (char __user *)arg2,
1767 sizeof(me->comm) - 1) < 0)
1769 set_task_comm(me, comm);
1772 get_task_comm(comm, me);
1773 if (copy_to_user((char __user *)arg2, comm,
1778 error = GET_ENDIAN(me, arg2);
1781 error = SET_ENDIAN(me, arg2);
1784 case PR_GET_SECCOMP:
1785 error = prctl_get_seccomp();
1787 case PR_SET_SECCOMP:
1788 error = prctl_set_seccomp(arg2);
1791 error = GET_TSC_CTL(arg2);
1794 error = SET_TSC_CTL(arg2);
1796 case PR_GET_TIMERSLACK:
1797 error = current->timer_slack_ns;
1799 case PR_SET_TIMERSLACK:
1801 current->timer_slack_ns =
1802 current->default_timer_slack_ns;
1804 current->timer_slack_ns = arg2;
1814 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
1815 struct getcpu_cache __user *unused)
1818 int cpu = raw_smp_processor_id();
1820 err |= put_user(cpu, cpup);
1822 err |= put_user(cpu_to_node(cpu), nodep);
1823 return err ? -EFAULT : 0;
1826 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1828 static void argv_cleanup(char **argv, char **envp)
1834 * orderly_poweroff - Trigger an orderly system poweroff
1835 * @force: force poweroff if command execution fails
1837 * This may be called from any context to trigger a system shutdown.
1838 * If the orderly shutdown fails, it will force an immediate shutdown.
1840 int orderly_poweroff(bool force)
1843 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1844 static char *envp[] = {
1846 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1850 struct subprocess_info *info;
1853 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1854 __func__, poweroff_cmd);
1858 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1864 call_usermodehelper_setcleanup(info, argv_cleanup);
1866 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1870 printk(KERN_WARNING "Failed to start orderly shutdown: "
1871 "forcing the issue\n");
1873 /* I guess this should try to kick off some daemon to
1874 sync and poweroff asap. Or not even bother syncing
1875 if we're doing an emergency shutdown? */
1882 EXPORT_SYMBOL_GPL(orderly_poweroff);