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);
364 void kernel_restart(char *cmd)
366 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
367 system_state = SYSTEM_RESTART;
368 device_suspend(PMSG_FREEZE);
371 printk(KERN_EMERG "Restarting system.\n");
373 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
376 machine_restart(cmd);
378 EXPORT_SYMBOL_GPL(kernel_restart);
380 void kernel_kexec(void)
383 struct kimage *image;
384 image = xchg(&kexec_image, 0);
388 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
389 system_state = SYSTEM_RESTART;
390 device_suspend(PMSG_FREEZE);
392 printk(KERN_EMERG "Starting new kernel\n");
394 machine_kexec(image);
397 EXPORT_SYMBOL_GPL(kernel_kexec);
399 void kernel_halt(void)
401 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
402 system_state = SYSTEM_HALT;
403 device_suspend(PMSG_SUSPEND);
405 printk(KERN_EMERG "System halted.\n");
408 EXPORT_SYMBOL_GPL(kernel_halt);
410 void kernel_power_off(void)
412 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
413 system_state = SYSTEM_POWER_OFF;
414 device_suspend(PMSG_SUSPEND);
416 printk(KERN_EMERG "Power down.\n");
419 EXPORT_SYMBOL_GPL(kernel_power_off);
422 * Reboot system call: for obvious reasons only root may call it,
423 * and even root needs to set up some magic numbers in the registers
424 * so that some mistake won't make this reboot the whole machine.
425 * You can also set the meaning of the ctrl-alt-del-key here.
427 * reboot doesn't sync: do that yourself before calling this.
429 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
433 /* We only trust the superuser with rebooting the system. */
434 if (!capable(CAP_SYS_BOOT))
437 /* For safety, we require "magic" arguments. */
438 if (magic1 != LINUX_REBOOT_MAGIC1 ||
439 (magic2 != LINUX_REBOOT_MAGIC2 &&
440 magic2 != LINUX_REBOOT_MAGIC2A &&
441 magic2 != LINUX_REBOOT_MAGIC2B &&
442 magic2 != LINUX_REBOOT_MAGIC2C))
447 case LINUX_REBOOT_CMD_RESTART:
448 kernel_restart(NULL);
451 case LINUX_REBOOT_CMD_CAD_ON:
455 case LINUX_REBOOT_CMD_CAD_OFF:
459 case LINUX_REBOOT_CMD_HALT:
465 case LINUX_REBOOT_CMD_POWER_OFF:
471 case LINUX_REBOOT_CMD_RESTART2:
472 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
476 buffer[sizeof(buffer) - 1] = '\0';
478 kernel_restart(buffer);
481 case LINUX_REBOOT_CMD_KEXEC:
486 #ifdef CONFIG_SOFTWARE_SUSPEND
487 case LINUX_REBOOT_CMD_SW_SUSPEND:
489 int ret = software_suspend();
503 static void deferred_cad(void *dummy)
505 kernel_restart(NULL);
509 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
510 * As it's called within an interrupt, it may NOT sync: the only choice
511 * is whether to reboot at once, or just ignore the ctrl-alt-del.
513 void ctrl_alt_del(void)
515 static DECLARE_WORK(cad_work, deferred_cad, NULL);
518 schedule_work(&cad_work);
520 kill_proc(cad_pid, SIGINT, 1);
525 * Unprivileged users may change the real gid to the effective gid
526 * or vice versa. (BSD-style)
528 * If you set the real gid at all, or set the effective gid to a value not
529 * equal to the real gid, then the saved gid is set to the new effective gid.
531 * This makes it possible for a setgid program to completely drop its
532 * privileges, which is often a useful assertion to make when you are doing
533 * a security audit over a program.
535 * The general idea is that a program which uses just setregid() will be
536 * 100% compatible with BSD. A program which uses just setgid() will be
537 * 100% compatible with POSIX with saved IDs.
539 * SMP: There are not races, the GIDs are checked only by filesystem
540 * operations (as far as semantic preservation is concerned).
542 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
544 int old_rgid = current->gid;
545 int old_egid = current->egid;
546 int new_rgid = old_rgid;
547 int new_egid = old_egid;
550 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
554 if (rgid != (gid_t) -1) {
555 if ((old_rgid == rgid) ||
556 (current->egid==rgid) ||
562 if (egid != (gid_t) -1) {
563 if ((old_rgid == egid) ||
564 (current->egid == egid) ||
565 (current->sgid == egid) ||
572 if (new_egid != old_egid)
574 current->mm->dumpable = suid_dumpable;
577 if (rgid != (gid_t) -1 ||
578 (egid != (gid_t) -1 && egid != old_rgid))
579 current->sgid = new_egid;
580 current->fsgid = new_egid;
581 current->egid = new_egid;
582 current->gid = new_rgid;
583 key_fsgid_changed(current);
588 * setgid() is implemented like SysV w/ SAVED_IDS
590 * SMP: Same implicit races as above.
592 asmlinkage long sys_setgid(gid_t gid)
594 int old_egid = current->egid;
597 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
601 if (capable(CAP_SETGID))
605 current->mm->dumpable = suid_dumpable;
608 current->gid = current->egid = current->sgid = current->fsgid = gid;
610 else if ((gid == current->gid) || (gid == current->sgid))
614 current->mm->dumpable = suid_dumpable;
617 current->egid = current->fsgid = gid;
622 key_fsgid_changed(current);
626 static int set_user(uid_t new_ruid, int dumpclear)
628 struct user_struct *new_user;
630 new_user = alloc_uid(new_ruid);
634 if (atomic_read(&new_user->processes) >=
635 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
636 new_user != &root_user) {
641 switch_uid(new_user);
645 current->mm->dumpable = suid_dumpable;
648 current->uid = new_ruid;
653 * Unprivileged users may change the real uid to the effective uid
654 * or vice versa. (BSD-style)
656 * If you set the real uid at all, or set the effective uid to a value not
657 * equal to the real uid, then the saved uid is set to the new effective uid.
659 * This makes it possible for a setuid program to completely drop its
660 * privileges, which is often a useful assertion to make when you are doing
661 * a security audit over a program.
663 * The general idea is that a program which uses just setreuid() will be
664 * 100% compatible with BSD. A program which uses just setuid() will be
665 * 100% compatible with POSIX with saved IDs.
667 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
669 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
672 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
676 new_ruid = old_ruid = current->uid;
677 new_euid = old_euid = current->euid;
678 old_suid = current->suid;
680 if (ruid != (uid_t) -1) {
682 if ((old_ruid != ruid) &&
683 (current->euid != ruid) &&
684 !capable(CAP_SETUID))
688 if (euid != (uid_t) -1) {
690 if ((old_ruid != euid) &&
691 (current->euid != euid) &&
692 (current->suid != euid) &&
693 !capable(CAP_SETUID))
697 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
700 if (new_euid != old_euid)
702 current->mm->dumpable = suid_dumpable;
705 current->fsuid = current->euid = new_euid;
706 if (ruid != (uid_t) -1 ||
707 (euid != (uid_t) -1 && euid != old_ruid))
708 current->suid = current->euid;
709 current->fsuid = current->euid;
711 key_fsuid_changed(current);
713 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
719 * setuid() is implemented like SysV with SAVED_IDS
721 * Note that SAVED_ID's is deficient in that a setuid root program
722 * like sendmail, for example, cannot set its uid to be a normal
723 * user and then switch back, because if you're root, setuid() sets
724 * the saved uid too. If you don't like this, blame the bright people
725 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
726 * will allow a root program to temporarily drop privileges and be able to
727 * regain them by swapping the real and effective uid.
729 asmlinkage long sys_setuid(uid_t uid)
731 int old_euid = current->euid;
732 int old_ruid, old_suid, new_ruid, new_suid;
735 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
739 old_ruid = new_ruid = current->uid;
740 old_suid = current->suid;
743 if (capable(CAP_SETUID)) {
744 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
747 } else if ((uid != current->uid) && (uid != new_suid))
752 current->mm->dumpable = suid_dumpable;
755 current->fsuid = current->euid = uid;
756 current->suid = new_suid;
758 key_fsuid_changed(current);
760 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
765 * This function implements a generic ability to update ruid, euid,
766 * and suid. This allows you to implement the 4.4 compatible seteuid().
768 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
770 int old_ruid = current->uid;
771 int old_euid = current->euid;
772 int old_suid = current->suid;
775 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
779 if (!capable(CAP_SETUID)) {
780 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
781 (ruid != current->euid) && (ruid != current->suid))
783 if ((euid != (uid_t) -1) && (euid != current->uid) &&
784 (euid != current->euid) && (euid != current->suid))
786 if ((suid != (uid_t) -1) && (suid != current->uid) &&
787 (suid != current->euid) && (suid != current->suid))
790 if (ruid != (uid_t) -1) {
791 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
794 if (euid != (uid_t) -1) {
795 if (euid != current->euid)
797 current->mm->dumpable = suid_dumpable;
800 current->euid = euid;
802 current->fsuid = current->euid;
803 if (suid != (uid_t) -1)
804 current->suid = suid;
806 key_fsuid_changed(current);
808 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
811 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
815 if (!(retval = put_user(current->uid, ruid)) &&
816 !(retval = put_user(current->euid, euid)))
817 retval = put_user(current->suid, suid);
823 * Same as above, but for rgid, egid, sgid.
825 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
829 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
833 if (!capable(CAP_SETGID)) {
834 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
835 (rgid != current->egid) && (rgid != current->sgid))
837 if ((egid != (gid_t) -1) && (egid != current->gid) &&
838 (egid != current->egid) && (egid != current->sgid))
840 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
841 (sgid != current->egid) && (sgid != current->sgid))
844 if (egid != (gid_t) -1) {
845 if (egid != current->egid)
847 current->mm->dumpable = suid_dumpable;
850 current->egid = egid;
852 current->fsgid = current->egid;
853 if (rgid != (gid_t) -1)
855 if (sgid != (gid_t) -1)
856 current->sgid = sgid;
858 key_fsgid_changed(current);
862 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
866 if (!(retval = put_user(current->gid, rgid)) &&
867 !(retval = put_user(current->egid, egid)))
868 retval = put_user(current->sgid, sgid);
875 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
876 * is used for "access()" and for the NFS daemon (letting nfsd stay at
877 * whatever uid it wants to). It normally shadows "euid", except when
878 * explicitly set by setfsuid() or for access..
880 asmlinkage long sys_setfsuid(uid_t uid)
884 old_fsuid = current->fsuid;
885 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
888 if (uid == current->uid || uid == current->euid ||
889 uid == current->suid || uid == current->fsuid ||
892 if (uid != old_fsuid)
894 current->mm->dumpable = suid_dumpable;
897 current->fsuid = uid;
900 key_fsuid_changed(current);
902 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
908 * Samma på svenska..
910 asmlinkage long sys_setfsgid(gid_t gid)
914 old_fsgid = current->fsgid;
915 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
918 if (gid == current->gid || gid == current->egid ||
919 gid == current->sgid || gid == current->fsgid ||
922 if (gid != old_fsgid)
924 current->mm->dumpable = suid_dumpable;
927 current->fsgid = gid;
928 key_fsgid_changed(current);
933 asmlinkage long sys_times(struct tms __user * tbuf)
936 * In the SMP world we might just be unlucky and have one of
937 * the times increment as we use it. Since the value is an
938 * atomically safe type this is just fine. Conceptually its
939 * as if the syscall took an instant longer to occur.
943 cputime_t utime, stime, cutime, cstime;
946 if (thread_group_empty(current)) {
948 * Single thread case without the use of any locks.
950 * We may race with release_task if two threads are
951 * executing. However, release task first adds up the
952 * counters (__exit_signal) before removing the task
953 * from the process tasklist (__unhash_process).
954 * __exit_signal also acquires and releases the
955 * siglock which results in the proper memory ordering
956 * so that the list modifications are always visible
957 * after the counters have been updated.
959 * If the counters have been updated by the second thread
960 * but the thread has not yet been removed from the list
961 * then the other branch will be executing which will
962 * block on tasklist_lock until the exit handling of the
963 * other task is finished.
965 * This also implies that the sighand->siglock cannot
966 * be held by another processor. So we can also
967 * skip acquiring that lock.
969 utime = cputime_add(current->signal->utime, current->utime);
970 stime = cputime_add(current->signal->utime, current->stime);
971 cutime = current->signal->cutime;
972 cstime = current->signal->cstime;
977 /* Process with multiple threads */
978 struct task_struct *tsk = current;
979 struct task_struct *t;
981 read_lock(&tasklist_lock);
982 utime = tsk->signal->utime;
983 stime = tsk->signal->stime;
986 utime = cputime_add(utime, t->utime);
987 stime = cputime_add(stime, t->stime);
992 * While we have tasklist_lock read-locked, no dying thread
993 * can be updating current->signal->[us]time. Instead,
994 * we got their counts included in the live thread loop.
995 * However, another thread can come in right now and
996 * do a wait call that updates current->signal->c[us]time.
997 * To make sure we always see that pair updated atomically,
998 * we take the siglock around fetching them.
1000 spin_lock_irq(&tsk->sighand->siglock);
1001 cutime = tsk->signal->cutime;
1002 cstime = tsk->signal->cstime;
1003 spin_unlock_irq(&tsk->sighand->siglock);
1004 read_unlock(&tasklist_lock);
1006 tmp.tms_utime = cputime_to_clock_t(utime);
1007 tmp.tms_stime = cputime_to_clock_t(stime);
1008 tmp.tms_cutime = cputime_to_clock_t(cutime);
1009 tmp.tms_cstime = cputime_to_clock_t(cstime);
1010 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1013 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1017 * This needs some heavy checking ...
1018 * I just haven't the stomach for it. I also don't fully
1019 * understand sessions/pgrp etc. Let somebody who does explain it.
1021 * OK, I think I have the protection semantics right.... this is really
1022 * only important on a multi-user system anyway, to make sure one user
1023 * can't send a signal to a process owned by another. -TYT, 12/12/91
1025 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1029 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1031 struct task_struct *p;
1041 /* From this point forward we keep holding onto the tasklist lock
1042 * so that our parent does not change from under us. -DaveM
1044 write_lock_irq(&tasklist_lock);
1047 p = find_task_by_pid(pid);
1052 if (!thread_group_leader(p))
1055 if (p->parent == current || p->real_parent == current) {
1057 if (p->signal->session != current->signal->session)
1069 if (p->signal->leader)
1073 struct task_struct *p;
1075 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1076 if (p->signal->session == current->signal->session)
1078 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1083 err = security_task_setpgid(p, pgid);
1087 if (process_group(p) != pgid) {
1088 detach_pid(p, PIDTYPE_PGID);
1089 p->signal->pgrp = pgid;
1090 attach_pid(p, PIDTYPE_PGID, pgid);
1095 /* All paths lead to here, thus we are safe. -DaveM */
1096 write_unlock_irq(&tasklist_lock);
1100 asmlinkage long sys_getpgid(pid_t pid)
1103 return process_group(current);
1106 struct task_struct *p;
1108 read_lock(&tasklist_lock);
1109 p = find_task_by_pid(pid);
1113 retval = security_task_getpgid(p);
1115 retval = process_group(p);
1117 read_unlock(&tasklist_lock);
1122 #ifdef __ARCH_WANT_SYS_GETPGRP
1124 asmlinkage long sys_getpgrp(void)
1126 /* SMP - assuming writes are word atomic this is fine */
1127 return process_group(current);
1132 asmlinkage long sys_getsid(pid_t pid)
1135 return current->signal->session;
1138 struct task_struct *p;
1140 read_lock(&tasklist_lock);
1141 p = find_task_by_pid(pid);
1145 retval = security_task_getsid(p);
1147 retval = p->signal->session;
1149 read_unlock(&tasklist_lock);
1154 asmlinkage long sys_setsid(void)
1159 if (!thread_group_leader(current))
1163 write_lock_irq(&tasklist_lock);
1165 pid = find_pid(PIDTYPE_PGID, current->pid);
1169 current->signal->leader = 1;
1170 __set_special_pids(current->pid, current->pid);
1171 current->signal->tty = NULL;
1172 current->signal->tty_old_pgrp = 0;
1173 err = process_group(current);
1175 write_unlock_irq(&tasklist_lock);
1181 * Supplementary group IDs
1184 /* init to 2 - one for init_task, one to ensure it is never freed */
1185 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1187 struct group_info *groups_alloc(int gidsetsize)
1189 struct group_info *group_info;
1193 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1194 /* Make sure we always allocate at least one indirect block pointer */
1195 nblocks = nblocks ? : 1;
1196 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1199 group_info->ngroups = gidsetsize;
1200 group_info->nblocks = nblocks;
1201 atomic_set(&group_info->usage, 1);
1203 if (gidsetsize <= NGROUPS_SMALL) {
1204 group_info->blocks[0] = group_info->small_block;
1206 for (i = 0; i < nblocks; i++) {
1208 b = (void *)__get_free_page(GFP_USER);
1210 goto out_undo_partial_alloc;
1211 group_info->blocks[i] = b;
1216 out_undo_partial_alloc:
1218 free_page((unsigned long)group_info->blocks[i]);
1224 EXPORT_SYMBOL(groups_alloc);
1226 void groups_free(struct group_info *group_info)
1228 if (group_info->blocks[0] != group_info->small_block) {
1230 for (i = 0; i < group_info->nblocks; i++)
1231 free_page((unsigned long)group_info->blocks[i]);
1236 EXPORT_SYMBOL(groups_free);
1238 /* export the group_info to a user-space array */
1239 static int groups_to_user(gid_t __user *grouplist,
1240 struct group_info *group_info)
1243 int count = group_info->ngroups;
1245 for (i = 0; i < group_info->nblocks; i++) {
1246 int cp_count = min(NGROUPS_PER_BLOCK, count);
1247 int off = i * NGROUPS_PER_BLOCK;
1248 int len = cp_count * sizeof(*grouplist);
1250 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1258 /* fill a group_info from a user-space array - it must be allocated already */
1259 static int groups_from_user(struct group_info *group_info,
1260 gid_t __user *grouplist)
1263 int count = group_info->ngroups;
1265 for (i = 0; i < group_info->nblocks; i++) {
1266 int cp_count = min(NGROUPS_PER_BLOCK, count);
1267 int off = i * NGROUPS_PER_BLOCK;
1268 int len = cp_count * sizeof(*grouplist);
1270 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1278 /* a simple Shell sort */
1279 static void groups_sort(struct group_info *group_info)
1281 int base, max, stride;
1282 int gidsetsize = group_info->ngroups;
1284 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1289 max = gidsetsize - stride;
1290 for (base = 0; base < max; base++) {
1292 int right = left + stride;
1293 gid_t tmp = GROUP_AT(group_info, right);
1295 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1296 GROUP_AT(group_info, right) =
1297 GROUP_AT(group_info, left);
1301 GROUP_AT(group_info, right) = tmp;
1307 /* a simple bsearch */
1308 int groups_search(struct group_info *group_info, gid_t grp)
1316 right = group_info->ngroups;
1317 while (left < right) {
1318 int mid = (left+right)/2;
1319 int cmp = grp - GROUP_AT(group_info, mid);
1330 /* validate and set current->group_info */
1331 int set_current_groups(struct group_info *group_info)
1334 struct group_info *old_info;
1336 retval = security_task_setgroups(group_info);
1340 groups_sort(group_info);
1341 get_group_info(group_info);
1344 old_info = current->group_info;
1345 current->group_info = group_info;
1346 task_unlock(current);
1348 put_group_info(old_info);
1353 EXPORT_SYMBOL(set_current_groups);
1355 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1360 * SMP: Nobody else can change our grouplist. Thus we are
1367 /* no need to grab task_lock here; it cannot change */
1368 get_group_info(current->group_info);
1369 i = current->group_info->ngroups;
1371 if (i > gidsetsize) {
1375 if (groups_to_user(grouplist, current->group_info)) {
1381 put_group_info(current->group_info);
1386 * SMP: Our groups are copy-on-write. We can set them safely
1387 * without another task interfering.
1390 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1392 struct group_info *group_info;
1395 if (!capable(CAP_SETGID))
1397 if ((unsigned)gidsetsize > NGROUPS_MAX)
1400 group_info = groups_alloc(gidsetsize);
1403 retval = groups_from_user(group_info, grouplist);
1405 put_group_info(group_info);
1409 retval = set_current_groups(group_info);
1410 put_group_info(group_info);
1416 * Check whether we're fsgid/egid or in the supplemental group..
1418 int in_group_p(gid_t grp)
1421 if (grp != current->fsgid) {
1422 get_group_info(current->group_info);
1423 retval = groups_search(current->group_info, grp);
1424 put_group_info(current->group_info);
1429 EXPORT_SYMBOL(in_group_p);
1431 int in_egroup_p(gid_t grp)
1434 if (grp != current->egid) {
1435 get_group_info(current->group_info);
1436 retval = groups_search(current->group_info, grp);
1437 put_group_info(current->group_info);
1442 EXPORT_SYMBOL(in_egroup_p);
1444 DECLARE_RWSEM(uts_sem);
1446 EXPORT_SYMBOL(uts_sem);
1448 asmlinkage long sys_newuname(struct new_utsname __user * name)
1452 down_read(&uts_sem);
1453 if (copy_to_user(name,&system_utsname,sizeof *name))
1459 asmlinkage long sys_sethostname(char __user *name, int len)
1462 char tmp[__NEW_UTS_LEN];
1464 if (!capable(CAP_SYS_ADMIN))
1466 if (len < 0 || len > __NEW_UTS_LEN)
1468 down_write(&uts_sem);
1470 if (!copy_from_user(tmp, name, len)) {
1471 memcpy(system_utsname.nodename, tmp, len);
1472 system_utsname.nodename[len] = 0;
1479 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1481 asmlinkage long sys_gethostname(char __user *name, int len)
1487 down_read(&uts_sem);
1488 i = 1 + strlen(system_utsname.nodename);
1492 if (copy_to_user(name, system_utsname.nodename, i))
1501 * Only setdomainname; getdomainname can be implemented by calling
1504 asmlinkage long sys_setdomainname(char __user *name, int len)
1507 char tmp[__NEW_UTS_LEN];
1509 if (!capable(CAP_SYS_ADMIN))
1511 if (len < 0 || len > __NEW_UTS_LEN)
1514 down_write(&uts_sem);
1516 if (!copy_from_user(tmp, name, len)) {
1517 memcpy(system_utsname.domainname, tmp, len);
1518 system_utsname.domainname[len] = 0;
1525 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1527 if (resource >= RLIM_NLIMITS)
1530 struct rlimit value;
1531 task_lock(current->group_leader);
1532 value = current->signal->rlim[resource];
1533 task_unlock(current->group_leader);
1534 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1538 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1541 * Back compatibility for getrlimit. Needed for some apps.
1544 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1547 if (resource >= RLIM_NLIMITS)
1550 task_lock(current->group_leader);
1551 x = current->signal->rlim[resource];
1552 task_unlock(current->group_leader);
1553 if(x.rlim_cur > 0x7FFFFFFF)
1554 x.rlim_cur = 0x7FFFFFFF;
1555 if(x.rlim_max > 0x7FFFFFFF)
1556 x.rlim_max = 0x7FFFFFFF;
1557 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1562 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1564 struct rlimit new_rlim, *old_rlim;
1567 if (resource >= RLIM_NLIMITS)
1569 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1571 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1573 old_rlim = current->signal->rlim + resource;
1574 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1575 !capable(CAP_SYS_RESOURCE))
1577 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1580 retval = security_task_setrlimit(resource, &new_rlim);
1584 task_lock(current->group_leader);
1585 *old_rlim = new_rlim;
1586 task_unlock(current->group_leader);
1588 if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1589 (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1590 new_rlim.rlim_cur <= cputime_to_secs(
1591 current->signal->it_prof_expires))) {
1592 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1593 read_lock(&tasklist_lock);
1594 spin_lock_irq(¤t->sighand->siglock);
1595 set_process_cpu_timer(current, CPUCLOCK_PROF,
1597 spin_unlock_irq(¤t->sighand->siglock);
1598 read_unlock(&tasklist_lock);
1605 * It would make sense to put struct rusage in the task_struct,
1606 * except that would make the task_struct be *really big*. After
1607 * task_struct gets moved into malloc'ed memory, it would
1608 * make sense to do this. It will make moving the rest of the information
1609 * a lot simpler! (Which we're not doing right now because we're not
1610 * measuring them yet).
1612 * This expects to be called with tasklist_lock read-locked or better,
1613 * and the siglock not locked. It may momentarily take the siglock.
1615 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1616 * races with threads incrementing their own counters. But since word
1617 * reads are atomic, we either get new values or old values and we don't
1618 * care which for the sums. We always take the siglock to protect reading
1619 * the c* fields from p->signal from races with exit.c updating those
1620 * fields when reaping, so a sample either gets all the additions of a
1621 * given child after it's reaped, or none so this sample is before reaping.
1624 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1626 struct task_struct *t;
1627 unsigned long flags;
1628 cputime_t utime, stime;
1630 memset((char *) r, 0, sizeof *r);
1632 if (unlikely(!p->signal))
1636 case RUSAGE_CHILDREN:
1637 spin_lock_irqsave(&p->sighand->siglock, flags);
1638 utime = p->signal->cutime;
1639 stime = p->signal->cstime;
1640 r->ru_nvcsw = p->signal->cnvcsw;
1641 r->ru_nivcsw = p->signal->cnivcsw;
1642 r->ru_minflt = p->signal->cmin_flt;
1643 r->ru_majflt = p->signal->cmaj_flt;
1644 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1645 cputime_to_timeval(utime, &r->ru_utime);
1646 cputime_to_timeval(stime, &r->ru_stime);
1649 spin_lock_irqsave(&p->sighand->siglock, flags);
1650 utime = stime = cputime_zero;
1653 spin_lock_irqsave(&p->sighand->siglock, flags);
1654 utime = p->signal->cutime;
1655 stime = p->signal->cstime;
1656 r->ru_nvcsw = p->signal->cnvcsw;
1657 r->ru_nivcsw = p->signal->cnivcsw;
1658 r->ru_minflt = p->signal->cmin_flt;
1659 r->ru_majflt = p->signal->cmaj_flt;
1661 utime = cputime_add(utime, p->signal->utime);
1662 stime = cputime_add(stime, p->signal->stime);
1663 r->ru_nvcsw += p->signal->nvcsw;
1664 r->ru_nivcsw += p->signal->nivcsw;
1665 r->ru_minflt += p->signal->min_flt;
1666 r->ru_majflt += p->signal->maj_flt;
1669 utime = cputime_add(utime, t->utime);
1670 stime = cputime_add(stime, t->stime);
1671 r->ru_nvcsw += t->nvcsw;
1672 r->ru_nivcsw += t->nivcsw;
1673 r->ru_minflt += t->min_flt;
1674 r->ru_majflt += t->maj_flt;
1677 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1678 cputime_to_timeval(utime, &r->ru_utime);
1679 cputime_to_timeval(stime, &r->ru_stime);
1686 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1689 read_lock(&tasklist_lock);
1690 k_getrusage(p, who, &r);
1691 read_unlock(&tasklist_lock);
1692 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1695 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1697 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1699 return getrusage(current, who, ru);
1702 asmlinkage long sys_umask(int mask)
1704 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1708 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1709 unsigned long arg4, unsigned long arg5)
1714 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1719 case PR_SET_PDEATHSIG:
1721 if (!valid_signal(sig)) {
1725 current->pdeath_signal = sig;
1727 case PR_GET_PDEATHSIG:
1728 error = put_user(current->pdeath_signal, (int __user *)arg2);
1730 case PR_GET_DUMPABLE:
1731 if (current->mm->dumpable)
1734 case PR_SET_DUMPABLE:
1735 if (arg2 < 0 || arg2 > 2) {
1739 current->mm->dumpable = arg2;
1742 case PR_SET_UNALIGN:
1743 error = SET_UNALIGN_CTL(current, arg2);
1745 case PR_GET_UNALIGN:
1746 error = GET_UNALIGN_CTL(current, arg2);
1749 error = SET_FPEMU_CTL(current, arg2);
1752 error = GET_FPEMU_CTL(current, arg2);
1755 error = SET_FPEXC_CTL(current, arg2);
1758 error = GET_FPEXC_CTL(current, arg2);
1761 error = PR_TIMING_STATISTICAL;
1764 if (arg2 == PR_TIMING_STATISTICAL)
1770 case PR_GET_KEEPCAPS:
1771 if (current->keep_capabilities)
1774 case PR_SET_KEEPCAPS:
1775 if (arg2 != 0 && arg2 != 1) {
1779 current->keep_capabilities = arg2;
1782 struct task_struct *me = current;
1783 unsigned char ncomm[sizeof(me->comm)];
1785 ncomm[sizeof(me->comm)-1] = 0;
1786 if (strncpy_from_user(ncomm, (char __user *)arg2,
1787 sizeof(me->comm)-1) < 0)
1789 set_task_comm(me, ncomm);
1793 struct task_struct *me = current;
1794 unsigned char tcomm[sizeof(me->comm)];
1796 get_task_comm(tcomm, me);
1797 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))