4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/config.h>
8 #include <linux/module.h>
10 #include <linux/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/smp_lock.h>
13 #include <linux/notifier.h>
14 #include <linux/reboot.h>
15 #include <linux/prctl.h>
16 #include <linux/init.h>
17 #include <linux/highuid.h>
19 #include <linux/kernel.h>
20 #include <linux/kexec.h>
21 #include <linux/workqueue.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>
32 #include <linux/compat.h>
33 #include <linux/syscalls.h>
35 #include <asm/uaccess.h>
37 #include <asm/unistd.h>
39 #ifndef SET_UNALIGN_CTL
40 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
42 #ifndef GET_UNALIGN_CTL
43 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
46 # define SET_FPEMU_CTL(a,b) (-EINVAL)
49 # define GET_FPEMU_CTL(a,b) (-EINVAL)
52 # define SET_FPEXC_CTL(a,b) (-EINVAL)
55 # define GET_FPEXC_CTL(a,b) (-EINVAL)
59 * this is where the system-wide overflow UID and GID are defined, for
60 * architectures that now have 32-bit UID/GID but didn't in the past
63 int overflowuid = DEFAULT_OVERFLOWUID;
64 int overflowgid = DEFAULT_OVERFLOWGID;
67 EXPORT_SYMBOL(overflowuid);
68 EXPORT_SYMBOL(overflowgid);
72 * the same as above, but for filesystems which can only store a 16-bit
73 * UID and GID. as such, this is needed on all architectures
76 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
77 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
79 EXPORT_SYMBOL(fs_overflowuid);
80 EXPORT_SYMBOL(fs_overflowgid);
83 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
90 * Notifier list for kernel code which wants to be called
91 * at shutdown. This is used to stop any idling DMA operations
95 static struct notifier_block *reboot_notifier_list;
96 static DEFINE_RWLOCK(notifier_lock);
99 * notifier_chain_register - Add notifier to a notifier chain
100 * @list: Pointer to root list pointer
101 * @n: New entry in notifier chain
103 * Adds a notifier to a notifier chain.
105 * Currently always returns zero.
108 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
110 write_lock(¬ifier_lock);
113 if(n->priority > (*list)->priority)
115 list= &((*list)->next);
119 write_unlock(¬ifier_lock);
123 EXPORT_SYMBOL(notifier_chain_register);
126 * notifier_chain_unregister - Remove notifier from a notifier chain
127 * @nl: Pointer to root list pointer
128 * @n: New entry in notifier chain
130 * Removes a notifier from a notifier chain.
132 * Returns zero on success, or %-ENOENT on failure.
135 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
137 write_lock(¬ifier_lock);
143 write_unlock(¬ifier_lock);
148 write_unlock(¬ifier_lock);
152 EXPORT_SYMBOL(notifier_chain_unregister);
155 * notifier_call_chain - Call functions in a notifier chain
156 * @n: Pointer to root pointer of notifier chain
157 * @val: Value passed unmodified to notifier function
158 * @v: Pointer passed unmodified to notifier function
160 * Calls each function in a notifier chain in turn.
162 * If the return value of the notifier can be and'd
163 * with %NOTIFY_STOP_MASK, then notifier_call_chain
164 * will return immediately, with the return value of
165 * the notifier function which halted execution.
166 * Otherwise, the return value is the return value
167 * of the last notifier function called.
170 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
173 struct notifier_block *nb = *n;
177 ret=nb->notifier_call(nb,val,v);
178 if(ret&NOTIFY_STOP_MASK)
187 EXPORT_SYMBOL(notifier_call_chain);
190 * register_reboot_notifier - Register function to be called at reboot time
191 * @nb: Info about notifier function to be called
193 * Registers a function with the list of functions
194 * to be called at reboot time.
196 * Currently always returns zero, as notifier_chain_register
197 * always returns zero.
200 int register_reboot_notifier(struct notifier_block * nb)
202 return notifier_chain_register(&reboot_notifier_list, nb);
205 EXPORT_SYMBOL(register_reboot_notifier);
208 * unregister_reboot_notifier - Unregister previously registered reboot notifier
209 * @nb: Hook to be unregistered
211 * Unregisters a previously registered reboot
214 * Returns zero on success, or %-ENOENT on failure.
217 int unregister_reboot_notifier(struct notifier_block * nb)
219 return notifier_chain_unregister(&reboot_notifier_list, nb);
222 EXPORT_SYMBOL(unregister_reboot_notifier);
224 static int set_one_prio(struct task_struct *p, int niceval, int error)
228 if (p->uid != current->euid &&
229 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
233 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
237 no_nice = security_task_setnice(p, niceval);
244 set_user_nice(p, niceval);
249 asmlinkage long sys_setpriority(int which, int who, int niceval)
251 struct task_struct *g, *p;
252 struct user_struct *user;
255 if (which > 2 || which < 0)
258 /* normalize: avoid signed division (rounding problems) */
265 read_lock(&tasklist_lock);
270 p = find_task_by_pid(who);
272 error = set_one_prio(p, niceval, error);
276 who = process_group(current);
277 do_each_task_pid(who, PIDTYPE_PGID, p) {
278 error = set_one_prio(p, niceval, error);
279 } while_each_task_pid(who, PIDTYPE_PGID, p);
282 user = current->user;
286 if ((who != current->uid) && !(user = find_user(who)))
287 goto out_unlock; /* No processes for this user */
291 error = set_one_prio(p, niceval, error);
292 while_each_thread(g, p);
293 if (who != current->uid)
294 free_uid(user); /* For find_user() */
298 read_unlock(&tasklist_lock);
304 * Ugh. To avoid negative return values, "getpriority()" will
305 * not return the normal nice-value, but a negated value that
306 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
307 * to stay compatible.
309 asmlinkage long sys_getpriority(int which, int who)
311 struct task_struct *g, *p;
312 struct user_struct *user;
313 long niceval, retval = -ESRCH;
315 if (which > 2 || which < 0)
318 read_lock(&tasklist_lock);
323 p = find_task_by_pid(who);
325 niceval = 20 - task_nice(p);
326 if (niceval > retval)
332 who = process_group(current);
333 do_each_task_pid(who, PIDTYPE_PGID, p) {
334 niceval = 20 - task_nice(p);
335 if (niceval > retval)
337 } while_each_task_pid(who, PIDTYPE_PGID, p);
340 user = current->user;
344 if ((who != current->uid) && !(user = find_user(who)))
345 goto out_unlock; /* No processes for this user */
349 niceval = 20 - task_nice(p);
350 if (niceval > retval)
353 while_each_thread(g, p);
354 if (who != current->uid)
355 free_uid(user); /* for find_user() */
359 read_unlock(&tasklist_lock);
366 * Reboot system call: for obvious reasons only root may call it,
367 * and even root needs to set up some magic numbers in the registers
368 * so that some mistake won't make this reboot the whole machine.
369 * You can also set the meaning of the ctrl-alt-del-key here.
371 * reboot doesn't sync: do that yourself before calling this.
373 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
377 /* We only trust the superuser with rebooting the system. */
378 if (!capable(CAP_SYS_BOOT))
381 /* For safety, we require "magic" arguments. */
382 if (magic1 != LINUX_REBOOT_MAGIC1 ||
383 (magic2 != LINUX_REBOOT_MAGIC2 &&
384 magic2 != LINUX_REBOOT_MAGIC2A &&
385 magic2 != LINUX_REBOOT_MAGIC2B &&
386 magic2 != LINUX_REBOOT_MAGIC2C))
391 case LINUX_REBOOT_CMD_RESTART:
392 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
393 system_state = SYSTEM_RESTART;
394 device_suspend(PMSG_FREEZE);
396 printk(KERN_EMERG "Restarting system.\n");
397 machine_restart(NULL);
400 case LINUX_REBOOT_CMD_CAD_ON:
404 case LINUX_REBOOT_CMD_CAD_OFF:
408 case LINUX_REBOOT_CMD_HALT:
409 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
410 system_state = SYSTEM_HALT;
411 device_suspend(PMSG_SUSPEND);
413 printk(KERN_EMERG "System halted.\n");
419 case LINUX_REBOOT_CMD_POWER_OFF:
420 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
421 system_state = SYSTEM_POWER_OFF;
422 device_suspend(PMSG_SUSPEND);
424 printk(KERN_EMERG "Power down.\n");
430 case LINUX_REBOOT_CMD_RESTART2:
431 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
435 buffer[sizeof(buffer) - 1] = '\0';
437 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
438 system_state = SYSTEM_RESTART;
439 device_suspend(PMSG_FREEZE);
441 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
442 machine_restart(buffer);
446 case LINUX_REBOOT_CMD_KEXEC:
448 struct kimage *image;
449 image = xchg(&kexec_image, 0);
454 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
455 system_state = SYSTEM_RESTART;
456 device_suspend(PMSG_FREEZE);
458 printk(KERN_EMERG "Starting new kernel\n");
460 machine_kexec(image);
464 #ifdef CONFIG_SOFTWARE_SUSPEND
465 case LINUX_REBOOT_CMD_SW_SUSPEND:
467 int ret = software_suspend();
481 static void deferred_cad(void *dummy)
483 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
484 machine_restart(NULL);
488 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
489 * As it's called within an interrupt, it may NOT sync: the only choice
490 * is whether to reboot at once, or just ignore the ctrl-alt-del.
492 void ctrl_alt_del(void)
494 static DECLARE_WORK(cad_work, deferred_cad, NULL);
497 schedule_work(&cad_work);
499 kill_proc(cad_pid, SIGINT, 1);
504 * Unprivileged users may change the real gid to the effective gid
505 * or vice versa. (BSD-style)
507 * If you set the real gid at all, or set the effective gid to a value not
508 * equal to the real gid, then the saved gid is set to the new effective gid.
510 * This makes it possible for a setgid program to completely drop its
511 * privileges, which is often a useful assertion to make when you are doing
512 * a security audit over a program.
514 * The general idea is that a program which uses just setregid() will be
515 * 100% compatible with BSD. A program which uses just setgid() will be
516 * 100% compatible with POSIX with saved IDs.
518 * SMP: There are not races, the GIDs are checked only by filesystem
519 * operations (as far as semantic preservation is concerned).
521 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
523 int old_rgid = current->gid;
524 int old_egid = current->egid;
525 int new_rgid = old_rgid;
526 int new_egid = old_egid;
529 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
533 if (rgid != (gid_t) -1) {
534 if ((old_rgid == rgid) ||
535 (current->egid==rgid) ||
541 if (egid != (gid_t) -1) {
542 if ((old_rgid == egid) ||
543 (current->egid == egid) ||
544 (current->sgid == egid) ||
551 if (new_egid != old_egid)
553 current->mm->dumpable = suid_dumpable;
556 if (rgid != (gid_t) -1 ||
557 (egid != (gid_t) -1 && egid != old_rgid))
558 current->sgid = new_egid;
559 current->fsgid = new_egid;
560 current->egid = new_egid;
561 current->gid = new_rgid;
562 key_fsgid_changed(current);
567 * setgid() is implemented like SysV w/ SAVED_IDS
569 * SMP: Same implicit races as above.
571 asmlinkage long sys_setgid(gid_t gid)
573 int old_egid = current->egid;
576 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
580 if (capable(CAP_SETGID))
584 current->mm->dumpable = suid_dumpable;
587 current->gid = current->egid = current->sgid = current->fsgid = gid;
589 else if ((gid == current->gid) || (gid == current->sgid))
593 current->mm->dumpable = suid_dumpable;
596 current->egid = current->fsgid = gid;
601 key_fsgid_changed(current);
605 static int set_user(uid_t new_ruid, int dumpclear)
607 struct user_struct *new_user;
609 new_user = alloc_uid(new_ruid);
613 if (atomic_read(&new_user->processes) >=
614 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
615 new_user != &root_user) {
620 switch_uid(new_user);
624 current->mm->dumpable = suid_dumpable;
627 current->uid = new_ruid;
632 * Unprivileged users may change the real uid to the effective uid
633 * or vice versa. (BSD-style)
635 * If you set the real uid at all, or set the effective uid to a value not
636 * equal to the real uid, then the saved uid is set to the new effective uid.
638 * This makes it possible for a setuid program to completely drop its
639 * privileges, which is often a useful assertion to make when you are doing
640 * a security audit over a program.
642 * The general idea is that a program which uses just setreuid() will be
643 * 100% compatible with BSD. A program which uses just setuid() will be
644 * 100% compatible with POSIX with saved IDs.
646 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
648 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
651 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
655 new_ruid = old_ruid = current->uid;
656 new_euid = old_euid = current->euid;
657 old_suid = current->suid;
659 if (ruid != (uid_t) -1) {
661 if ((old_ruid != ruid) &&
662 (current->euid != ruid) &&
663 !capable(CAP_SETUID))
667 if (euid != (uid_t) -1) {
669 if ((old_ruid != euid) &&
670 (current->euid != euid) &&
671 (current->suid != euid) &&
672 !capable(CAP_SETUID))
676 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
679 if (new_euid != old_euid)
681 current->mm->dumpable = suid_dumpable;
684 current->fsuid = current->euid = new_euid;
685 if (ruid != (uid_t) -1 ||
686 (euid != (uid_t) -1 && euid != old_ruid))
687 current->suid = current->euid;
688 current->fsuid = current->euid;
690 key_fsuid_changed(current);
692 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
698 * setuid() is implemented like SysV with SAVED_IDS
700 * Note that SAVED_ID's is deficient in that a setuid root program
701 * like sendmail, for example, cannot set its uid to be a normal
702 * user and then switch back, because if you're root, setuid() sets
703 * the saved uid too. If you don't like this, blame the bright people
704 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
705 * will allow a root program to temporarily drop privileges and be able to
706 * regain them by swapping the real and effective uid.
708 asmlinkage long sys_setuid(uid_t uid)
710 int old_euid = current->euid;
711 int old_ruid, old_suid, new_ruid, new_suid;
714 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
718 old_ruid = new_ruid = current->uid;
719 old_suid = current->suid;
722 if (capable(CAP_SETUID)) {
723 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
726 } else if ((uid != current->uid) && (uid != new_suid))
731 current->mm->dumpable = suid_dumpable;
734 current->fsuid = current->euid = uid;
735 current->suid = new_suid;
737 key_fsuid_changed(current);
739 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
744 * This function implements a generic ability to update ruid, euid,
745 * and suid. This allows you to implement the 4.4 compatible seteuid().
747 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
749 int old_ruid = current->uid;
750 int old_euid = current->euid;
751 int old_suid = current->suid;
754 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
758 if (!capable(CAP_SETUID)) {
759 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
760 (ruid != current->euid) && (ruid != current->suid))
762 if ((euid != (uid_t) -1) && (euid != current->uid) &&
763 (euid != current->euid) && (euid != current->suid))
765 if ((suid != (uid_t) -1) && (suid != current->uid) &&
766 (suid != current->euid) && (suid != current->suid))
769 if (ruid != (uid_t) -1) {
770 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
773 if (euid != (uid_t) -1) {
774 if (euid != current->euid)
776 current->mm->dumpable = suid_dumpable;
779 current->euid = euid;
781 current->fsuid = current->euid;
782 if (suid != (uid_t) -1)
783 current->suid = suid;
785 key_fsuid_changed(current);
787 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
790 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
794 if (!(retval = put_user(current->uid, ruid)) &&
795 !(retval = put_user(current->euid, euid)))
796 retval = put_user(current->suid, suid);
802 * Same as above, but for rgid, egid, sgid.
804 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
808 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
812 if (!capable(CAP_SETGID)) {
813 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
814 (rgid != current->egid) && (rgid != current->sgid))
816 if ((egid != (gid_t) -1) && (egid != current->gid) &&
817 (egid != current->egid) && (egid != current->sgid))
819 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
820 (sgid != current->egid) && (sgid != current->sgid))
823 if (egid != (gid_t) -1) {
824 if (egid != current->egid)
826 current->mm->dumpable = suid_dumpable;
829 current->egid = egid;
831 current->fsgid = current->egid;
832 if (rgid != (gid_t) -1)
834 if (sgid != (gid_t) -1)
835 current->sgid = sgid;
837 key_fsgid_changed(current);
841 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
845 if (!(retval = put_user(current->gid, rgid)) &&
846 !(retval = put_user(current->egid, egid)))
847 retval = put_user(current->sgid, sgid);
854 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
855 * is used for "access()" and for the NFS daemon (letting nfsd stay at
856 * whatever uid it wants to). It normally shadows "euid", except when
857 * explicitly set by setfsuid() or for access..
859 asmlinkage long sys_setfsuid(uid_t uid)
863 old_fsuid = current->fsuid;
864 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
867 if (uid == current->uid || uid == current->euid ||
868 uid == current->suid || uid == current->fsuid ||
871 if (uid != old_fsuid)
873 current->mm->dumpable = suid_dumpable;
876 current->fsuid = uid;
879 key_fsuid_changed(current);
881 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
887 * Samma på svenska..
889 asmlinkage long sys_setfsgid(gid_t gid)
893 old_fsgid = current->fsgid;
894 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
897 if (gid == current->gid || gid == current->egid ||
898 gid == current->sgid || gid == current->fsgid ||
901 if (gid != old_fsgid)
903 current->mm->dumpable = suid_dumpable;
906 current->fsgid = gid;
907 key_fsgid_changed(current);
912 asmlinkage long sys_times(struct tms __user * tbuf)
915 * In the SMP world we might just be unlucky and have one of
916 * the times increment as we use it. Since the value is an
917 * atomically safe type this is just fine. Conceptually its
918 * as if the syscall took an instant longer to occur.
922 cputime_t utime, stime, cutime, cstime;
925 if (thread_group_empty(current)) {
927 * Single thread case without the use of any locks.
929 * We may race with release_task if two threads are
930 * executing. However, release task first adds up the
931 * counters (__exit_signal) before removing the task
932 * from the process tasklist (__unhash_process).
933 * __exit_signal also acquires and releases the
934 * siglock which results in the proper memory ordering
935 * so that the list modifications are always visible
936 * after the counters have been updated.
938 * If the counters have been updated by the second thread
939 * but the thread has not yet been removed from the list
940 * then the other branch will be executing which will
941 * block on tasklist_lock until the exit handling of the
942 * other task is finished.
944 * This also implies that the sighand->siglock cannot
945 * be held by another processor. So we can also
946 * skip acquiring that lock.
948 utime = cputime_add(current->signal->utime, current->utime);
949 stime = cputime_add(current->signal->utime, current->stime);
950 cutime = current->signal->cutime;
951 cstime = current->signal->cstime;
956 /* Process with multiple threads */
957 struct task_struct *tsk = current;
958 struct task_struct *t;
960 read_lock(&tasklist_lock);
961 utime = tsk->signal->utime;
962 stime = tsk->signal->stime;
965 utime = cputime_add(utime, t->utime);
966 stime = cputime_add(stime, t->stime);
971 * While we have tasklist_lock read-locked, no dying thread
972 * can be updating current->signal->[us]time. Instead,
973 * we got their counts included in the live thread loop.
974 * However, another thread can come in right now and
975 * do a wait call that updates current->signal->c[us]time.
976 * To make sure we always see that pair updated atomically,
977 * we take the siglock around fetching them.
979 spin_lock_irq(&tsk->sighand->siglock);
980 cutime = tsk->signal->cutime;
981 cstime = tsk->signal->cstime;
982 spin_unlock_irq(&tsk->sighand->siglock);
983 read_unlock(&tasklist_lock);
985 tmp.tms_utime = cputime_to_clock_t(utime);
986 tmp.tms_stime = cputime_to_clock_t(stime);
987 tmp.tms_cutime = cputime_to_clock_t(cutime);
988 tmp.tms_cstime = cputime_to_clock_t(cstime);
989 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
992 return (long) jiffies_64_to_clock_t(get_jiffies_64());
996 * This needs some heavy checking ...
997 * I just haven't the stomach for it. I also don't fully
998 * understand sessions/pgrp etc. Let somebody who does explain it.
1000 * OK, I think I have the protection semantics right.... this is really
1001 * only important on a multi-user system anyway, to make sure one user
1002 * can't send a signal to a process owned by another. -TYT, 12/12/91
1004 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1008 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1010 struct task_struct *p;
1020 /* From this point forward we keep holding onto the tasklist lock
1021 * so that our parent does not change from under us. -DaveM
1023 write_lock_irq(&tasklist_lock);
1026 p = find_task_by_pid(pid);
1031 if (!thread_group_leader(p))
1034 if (p->parent == current || p->real_parent == current) {
1036 if (p->signal->session != current->signal->session)
1048 if (p->signal->leader)
1052 struct task_struct *p;
1054 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1055 if (p->signal->session == current->signal->session)
1057 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1062 err = security_task_setpgid(p, pgid);
1066 if (process_group(p) != pgid) {
1067 detach_pid(p, PIDTYPE_PGID);
1068 p->signal->pgrp = pgid;
1069 attach_pid(p, PIDTYPE_PGID, pgid);
1074 /* All paths lead to here, thus we are safe. -DaveM */
1075 write_unlock_irq(&tasklist_lock);
1079 asmlinkage long sys_getpgid(pid_t pid)
1082 return process_group(current);
1085 struct task_struct *p;
1087 read_lock(&tasklist_lock);
1088 p = find_task_by_pid(pid);
1092 retval = security_task_getpgid(p);
1094 retval = process_group(p);
1096 read_unlock(&tasklist_lock);
1101 #ifdef __ARCH_WANT_SYS_GETPGRP
1103 asmlinkage long sys_getpgrp(void)
1105 /* SMP - assuming writes are word atomic this is fine */
1106 return process_group(current);
1111 asmlinkage long sys_getsid(pid_t pid)
1114 return current->signal->session;
1117 struct task_struct *p;
1119 read_lock(&tasklist_lock);
1120 p = find_task_by_pid(pid);
1124 retval = security_task_getsid(p);
1126 retval = p->signal->session;
1128 read_unlock(&tasklist_lock);
1133 asmlinkage long sys_setsid(void)
1138 if (!thread_group_leader(current))
1142 write_lock_irq(&tasklist_lock);
1144 pid = find_pid(PIDTYPE_PGID, current->pid);
1148 current->signal->leader = 1;
1149 __set_special_pids(current->pid, current->pid);
1150 current->signal->tty = NULL;
1151 current->signal->tty_old_pgrp = 0;
1152 err = process_group(current);
1154 write_unlock_irq(&tasklist_lock);
1160 * Supplementary group IDs
1163 /* init to 2 - one for init_task, one to ensure it is never freed */
1164 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1166 struct group_info *groups_alloc(int gidsetsize)
1168 struct group_info *group_info;
1172 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1173 /* Make sure we always allocate at least one indirect block pointer */
1174 nblocks = nblocks ? : 1;
1175 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1178 group_info->ngroups = gidsetsize;
1179 group_info->nblocks = nblocks;
1180 atomic_set(&group_info->usage, 1);
1182 if (gidsetsize <= NGROUPS_SMALL) {
1183 group_info->blocks[0] = group_info->small_block;
1185 for (i = 0; i < nblocks; i++) {
1187 b = (void *)__get_free_page(GFP_USER);
1189 goto out_undo_partial_alloc;
1190 group_info->blocks[i] = b;
1195 out_undo_partial_alloc:
1197 free_page((unsigned long)group_info->blocks[i]);
1203 EXPORT_SYMBOL(groups_alloc);
1205 void groups_free(struct group_info *group_info)
1207 if (group_info->blocks[0] != group_info->small_block) {
1209 for (i = 0; i < group_info->nblocks; i++)
1210 free_page((unsigned long)group_info->blocks[i]);
1215 EXPORT_SYMBOL(groups_free);
1217 /* export the group_info to a user-space array */
1218 static int groups_to_user(gid_t __user *grouplist,
1219 struct group_info *group_info)
1222 int count = group_info->ngroups;
1224 for (i = 0; i < group_info->nblocks; i++) {
1225 int cp_count = min(NGROUPS_PER_BLOCK, count);
1226 int off = i * NGROUPS_PER_BLOCK;
1227 int len = cp_count * sizeof(*grouplist);
1229 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1237 /* fill a group_info from a user-space array - it must be allocated already */
1238 static int groups_from_user(struct group_info *group_info,
1239 gid_t __user *grouplist)
1242 int count = group_info->ngroups;
1244 for (i = 0; i < group_info->nblocks; i++) {
1245 int cp_count = min(NGROUPS_PER_BLOCK, count);
1246 int off = i * NGROUPS_PER_BLOCK;
1247 int len = cp_count * sizeof(*grouplist);
1249 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1257 /* a simple Shell sort */
1258 static void groups_sort(struct group_info *group_info)
1260 int base, max, stride;
1261 int gidsetsize = group_info->ngroups;
1263 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1268 max = gidsetsize - stride;
1269 for (base = 0; base < max; base++) {
1271 int right = left + stride;
1272 gid_t tmp = GROUP_AT(group_info, right);
1274 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1275 GROUP_AT(group_info, right) =
1276 GROUP_AT(group_info, left);
1280 GROUP_AT(group_info, right) = tmp;
1286 /* a simple bsearch */
1287 int groups_search(struct group_info *group_info, gid_t grp)
1295 right = group_info->ngroups;
1296 while (left < right) {
1297 int mid = (left+right)/2;
1298 int cmp = grp - GROUP_AT(group_info, mid);
1309 /* validate and set current->group_info */
1310 int set_current_groups(struct group_info *group_info)
1313 struct group_info *old_info;
1315 retval = security_task_setgroups(group_info);
1319 groups_sort(group_info);
1320 get_group_info(group_info);
1323 old_info = current->group_info;
1324 current->group_info = group_info;
1325 task_unlock(current);
1327 put_group_info(old_info);
1332 EXPORT_SYMBOL(set_current_groups);
1334 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1339 * SMP: Nobody else can change our grouplist. Thus we are
1346 /* no need to grab task_lock here; it cannot change */
1347 get_group_info(current->group_info);
1348 i = current->group_info->ngroups;
1350 if (i > gidsetsize) {
1354 if (groups_to_user(grouplist, current->group_info)) {
1360 put_group_info(current->group_info);
1365 * SMP: Our groups are copy-on-write. We can set them safely
1366 * without another task interfering.
1369 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1371 struct group_info *group_info;
1374 if (!capable(CAP_SETGID))
1376 if ((unsigned)gidsetsize > NGROUPS_MAX)
1379 group_info = groups_alloc(gidsetsize);
1382 retval = groups_from_user(group_info, grouplist);
1384 put_group_info(group_info);
1388 retval = set_current_groups(group_info);
1389 put_group_info(group_info);
1395 * Check whether we're fsgid/egid or in the supplemental group..
1397 int in_group_p(gid_t grp)
1400 if (grp != current->fsgid) {
1401 get_group_info(current->group_info);
1402 retval = groups_search(current->group_info, grp);
1403 put_group_info(current->group_info);
1408 EXPORT_SYMBOL(in_group_p);
1410 int in_egroup_p(gid_t grp)
1413 if (grp != current->egid) {
1414 get_group_info(current->group_info);
1415 retval = groups_search(current->group_info, grp);
1416 put_group_info(current->group_info);
1421 EXPORT_SYMBOL(in_egroup_p);
1423 DECLARE_RWSEM(uts_sem);
1425 EXPORT_SYMBOL(uts_sem);
1427 asmlinkage long sys_newuname(struct new_utsname __user * name)
1431 down_read(&uts_sem);
1432 if (copy_to_user(name,&system_utsname,sizeof *name))
1438 asmlinkage long sys_sethostname(char __user *name, int len)
1441 char tmp[__NEW_UTS_LEN];
1443 if (!capable(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 memcpy(system_utsname.nodename, tmp, len);
1451 system_utsname.nodename[len] = 0;
1458 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1460 asmlinkage long sys_gethostname(char __user *name, int len)
1466 down_read(&uts_sem);
1467 i = 1 + strlen(system_utsname.nodename);
1471 if (copy_to_user(name, system_utsname.nodename, i))
1480 * Only setdomainname; getdomainname can be implemented by calling
1483 asmlinkage long sys_setdomainname(char __user *name, int len)
1486 char tmp[__NEW_UTS_LEN];
1488 if (!capable(CAP_SYS_ADMIN))
1490 if (len < 0 || len > __NEW_UTS_LEN)
1493 down_write(&uts_sem);
1495 if (!copy_from_user(tmp, name, len)) {
1496 memcpy(system_utsname.domainname, tmp, len);
1497 system_utsname.domainname[len] = 0;
1504 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1506 if (resource >= RLIM_NLIMITS)
1509 struct rlimit value;
1510 task_lock(current->group_leader);
1511 value = current->signal->rlim[resource];
1512 task_unlock(current->group_leader);
1513 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1517 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1520 * Back compatibility for getrlimit. Needed for some apps.
1523 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1526 if (resource >= RLIM_NLIMITS)
1529 task_lock(current->group_leader);
1530 x = current->signal->rlim[resource];
1531 task_unlock(current->group_leader);
1532 if(x.rlim_cur > 0x7FFFFFFF)
1533 x.rlim_cur = 0x7FFFFFFF;
1534 if(x.rlim_max > 0x7FFFFFFF)
1535 x.rlim_max = 0x7FFFFFFF;
1536 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1541 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1543 struct rlimit new_rlim, *old_rlim;
1546 if (resource >= RLIM_NLIMITS)
1548 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1550 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1552 old_rlim = current->signal->rlim + resource;
1553 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1554 !capable(CAP_SYS_RESOURCE))
1556 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1559 retval = security_task_setrlimit(resource, &new_rlim);
1563 task_lock(current->group_leader);
1564 *old_rlim = new_rlim;
1565 task_unlock(current->group_leader);
1567 if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1568 (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1569 new_rlim.rlim_cur <= cputime_to_secs(
1570 current->signal->it_prof_expires))) {
1571 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1572 read_lock(&tasklist_lock);
1573 spin_lock_irq(¤t->sighand->siglock);
1574 set_process_cpu_timer(current, CPUCLOCK_PROF,
1576 spin_unlock_irq(¤t->sighand->siglock);
1577 read_unlock(&tasklist_lock);
1584 * It would make sense to put struct rusage in the task_struct,
1585 * except that would make the task_struct be *really big*. After
1586 * task_struct gets moved into malloc'ed memory, it would
1587 * make sense to do this. It will make moving the rest of the information
1588 * a lot simpler! (Which we're not doing right now because we're not
1589 * measuring them yet).
1591 * This expects to be called with tasklist_lock read-locked or better,
1592 * and the siglock not locked. It may momentarily take the siglock.
1594 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1595 * races with threads incrementing their own counters. But since word
1596 * reads are atomic, we either get new values or old values and we don't
1597 * care which for the sums. We always take the siglock to protect reading
1598 * the c* fields from p->signal from races with exit.c updating those
1599 * fields when reaping, so a sample either gets all the additions of a
1600 * given child after it's reaped, or none so this sample is before reaping.
1603 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1605 struct task_struct *t;
1606 unsigned long flags;
1607 cputime_t utime, stime;
1609 memset((char *) r, 0, sizeof *r);
1611 if (unlikely(!p->signal))
1615 case RUSAGE_CHILDREN:
1616 spin_lock_irqsave(&p->sighand->siglock, flags);
1617 utime = p->signal->cutime;
1618 stime = p->signal->cstime;
1619 r->ru_nvcsw = p->signal->cnvcsw;
1620 r->ru_nivcsw = p->signal->cnivcsw;
1621 r->ru_minflt = p->signal->cmin_flt;
1622 r->ru_majflt = p->signal->cmaj_flt;
1623 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1624 cputime_to_timeval(utime, &r->ru_utime);
1625 cputime_to_timeval(stime, &r->ru_stime);
1628 spin_lock_irqsave(&p->sighand->siglock, flags);
1629 utime = stime = cputime_zero;
1632 spin_lock_irqsave(&p->sighand->siglock, flags);
1633 utime = p->signal->cutime;
1634 stime = p->signal->cstime;
1635 r->ru_nvcsw = p->signal->cnvcsw;
1636 r->ru_nivcsw = p->signal->cnivcsw;
1637 r->ru_minflt = p->signal->cmin_flt;
1638 r->ru_majflt = p->signal->cmaj_flt;
1640 utime = cputime_add(utime, p->signal->utime);
1641 stime = cputime_add(stime, p->signal->stime);
1642 r->ru_nvcsw += p->signal->nvcsw;
1643 r->ru_nivcsw += p->signal->nivcsw;
1644 r->ru_minflt += p->signal->min_flt;
1645 r->ru_majflt += p->signal->maj_flt;
1648 utime = cputime_add(utime, t->utime);
1649 stime = cputime_add(stime, t->stime);
1650 r->ru_nvcsw += t->nvcsw;
1651 r->ru_nivcsw += t->nivcsw;
1652 r->ru_minflt += t->min_flt;
1653 r->ru_majflt += t->maj_flt;
1656 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1657 cputime_to_timeval(utime, &r->ru_utime);
1658 cputime_to_timeval(stime, &r->ru_stime);
1665 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1668 read_lock(&tasklist_lock);
1669 k_getrusage(p, who, &r);
1670 read_unlock(&tasklist_lock);
1671 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1674 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1676 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1678 return getrusage(current, who, ru);
1681 asmlinkage long sys_umask(int mask)
1683 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1687 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1688 unsigned long arg4, unsigned long arg5)
1693 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1698 case PR_SET_PDEATHSIG:
1700 if (!valid_signal(sig)) {
1704 current->pdeath_signal = sig;
1706 case PR_GET_PDEATHSIG:
1707 error = put_user(current->pdeath_signal, (int __user *)arg2);
1709 case PR_GET_DUMPABLE:
1710 if (current->mm->dumpable)
1713 case PR_SET_DUMPABLE:
1714 if (arg2 < 0 || arg2 > 2) {
1718 current->mm->dumpable = arg2;
1721 case PR_SET_UNALIGN:
1722 error = SET_UNALIGN_CTL(current, arg2);
1724 case PR_GET_UNALIGN:
1725 error = GET_UNALIGN_CTL(current, arg2);
1728 error = SET_FPEMU_CTL(current, arg2);
1731 error = GET_FPEMU_CTL(current, arg2);
1734 error = SET_FPEXC_CTL(current, arg2);
1737 error = GET_FPEXC_CTL(current, arg2);
1740 error = PR_TIMING_STATISTICAL;
1743 if (arg2 == PR_TIMING_STATISTICAL)
1749 case PR_GET_KEEPCAPS:
1750 if (current->keep_capabilities)
1753 case PR_SET_KEEPCAPS:
1754 if (arg2 != 0 && arg2 != 1) {
1758 current->keep_capabilities = arg2;
1761 struct task_struct *me = current;
1762 unsigned char ncomm[sizeof(me->comm)];
1764 ncomm[sizeof(me->comm)-1] = 0;
1765 if (strncpy_from_user(ncomm, (char __user *)arg2,
1766 sizeof(me->comm)-1) < 0)
1768 set_task_comm(me, ncomm);
1772 struct task_struct *me = current;
1773 unsigned char tcomm[sizeof(me->comm)];
1775 get_task_comm(tcomm, me);
1776 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))