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>
31 #include <linux/cn_proc.h>
33 #include <linux/compat.h>
34 #include <linux/syscalls.h>
35 #include <linux/kprobes.h>
37 #include <asm/uaccess.h>
39 #include <asm/unistd.h>
41 #ifndef SET_UNALIGN_CTL
42 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
44 #ifndef GET_UNALIGN_CTL
45 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
48 # define SET_FPEMU_CTL(a,b) (-EINVAL)
51 # define GET_FPEMU_CTL(a,b) (-EINVAL)
54 # define SET_FPEXC_CTL(a,b) (-EINVAL)
57 # define GET_FPEXC_CTL(a,b) (-EINVAL)
61 * this is where the system-wide overflow UID and GID are defined, for
62 * architectures that now have 32-bit UID/GID but didn't in the past
65 int overflowuid = DEFAULT_OVERFLOWUID;
66 int overflowgid = DEFAULT_OVERFLOWGID;
69 EXPORT_SYMBOL(overflowuid);
70 EXPORT_SYMBOL(overflowgid);
74 * the same as above, but for filesystems which can only store a 16-bit
75 * UID and GID. as such, this is needed on all architectures
78 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
79 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
81 EXPORT_SYMBOL(fs_overflowuid);
82 EXPORT_SYMBOL(fs_overflowgid);
85 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
92 * Notifier list for kernel code which wants to be called
93 * at shutdown. This is used to stop any idling DMA operations
97 static struct notifier_block *reboot_notifier_list;
98 static DEFINE_RWLOCK(notifier_lock);
101 * notifier_chain_register - Add notifier to a notifier chain
102 * @list: Pointer to root list pointer
103 * @n: New entry in notifier chain
105 * Adds a notifier to a notifier chain.
107 * Currently always returns zero.
110 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
112 write_lock(¬ifier_lock);
115 if(n->priority > (*list)->priority)
117 list= &((*list)->next);
121 write_unlock(¬ifier_lock);
125 EXPORT_SYMBOL(notifier_chain_register);
128 * notifier_chain_unregister - Remove notifier from a notifier chain
129 * @nl: Pointer to root list pointer
130 * @n: New entry in notifier chain
132 * Removes a notifier from a notifier chain.
134 * Returns zero on success, or %-ENOENT on failure.
137 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
139 write_lock(¬ifier_lock);
145 write_unlock(¬ifier_lock);
150 write_unlock(¬ifier_lock);
154 EXPORT_SYMBOL(notifier_chain_unregister);
157 * notifier_call_chain - Call functions in a notifier chain
158 * @n: Pointer to root pointer of notifier chain
159 * @val: Value passed unmodified to notifier function
160 * @v: Pointer passed unmodified to notifier function
162 * Calls each function in a notifier chain in turn.
164 * If the return value of the notifier can be and'd
165 * with %NOTIFY_STOP_MASK, then notifier_call_chain
166 * will return immediately, with the return value of
167 * the notifier function which halted execution.
168 * Otherwise, the return value is the return value
169 * of the last notifier function called.
172 int __kprobes notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
175 struct notifier_block *nb = *n;
179 ret=nb->notifier_call(nb,val,v);
180 if(ret&NOTIFY_STOP_MASK)
189 EXPORT_SYMBOL(notifier_call_chain);
192 * register_reboot_notifier - Register function to be called at reboot time
193 * @nb: Info about notifier function to be called
195 * Registers a function with the list of functions
196 * to be called at reboot time.
198 * Currently always returns zero, as notifier_chain_register
199 * always returns zero.
202 int register_reboot_notifier(struct notifier_block * nb)
204 return notifier_chain_register(&reboot_notifier_list, nb);
207 EXPORT_SYMBOL(register_reboot_notifier);
210 * unregister_reboot_notifier - Unregister previously registered reboot notifier
211 * @nb: Hook to be unregistered
213 * Unregisters a previously registered reboot
216 * Returns zero on success, or %-ENOENT on failure.
219 int unregister_reboot_notifier(struct notifier_block * nb)
221 return notifier_chain_unregister(&reboot_notifier_list, nb);
224 EXPORT_SYMBOL(unregister_reboot_notifier);
226 #ifndef CONFIG_SECURITY
229 if (cap_raised(current->cap_effective, cap)) {
230 current->flags |= PF_SUPERPRIV;
235 EXPORT_SYMBOL(capable);
238 static int set_one_prio(struct task_struct *p, int niceval, int error)
242 if (p->uid != current->euid &&
243 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
247 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
251 no_nice = security_task_setnice(p, niceval);
258 set_user_nice(p, niceval);
263 asmlinkage long sys_setpriority(int which, int who, int niceval)
265 struct task_struct *g, *p;
266 struct user_struct *user;
269 if (which > 2 || which < 0)
272 /* normalize: avoid signed division (rounding problems) */
279 read_lock(&tasklist_lock);
284 p = find_task_by_pid(who);
286 error = set_one_prio(p, niceval, error);
290 who = process_group(current);
291 do_each_task_pid(who, PIDTYPE_PGID, p) {
292 error = set_one_prio(p, niceval, error);
293 } while_each_task_pid(who, PIDTYPE_PGID, p);
296 user = current->user;
300 if ((who != current->uid) && !(user = find_user(who)))
301 goto out_unlock; /* No processes for this user */
305 error = set_one_prio(p, niceval, error);
306 while_each_thread(g, p);
307 if (who != current->uid)
308 free_uid(user); /* For find_user() */
312 read_unlock(&tasklist_lock);
318 * Ugh. To avoid negative return values, "getpriority()" will
319 * not return the normal nice-value, but a negated value that
320 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
321 * to stay compatible.
323 asmlinkage long sys_getpriority(int which, int who)
325 struct task_struct *g, *p;
326 struct user_struct *user;
327 long niceval, retval = -ESRCH;
329 if (which > 2 || which < 0)
332 read_lock(&tasklist_lock);
337 p = find_task_by_pid(who);
339 niceval = 20 - task_nice(p);
340 if (niceval > retval)
346 who = process_group(current);
347 do_each_task_pid(who, PIDTYPE_PGID, p) {
348 niceval = 20 - task_nice(p);
349 if (niceval > retval)
351 } while_each_task_pid(who, PIDTYPE_PGID, p);
354 user = current->user;
358 if ((who != current->uid) && !(user = find_user(who)))
359 goto out_unlock; /* No processes for this user */
363 niceval = 20 - task_nice(p);
364 if (niceval > retval)
367 while_each_thread(g, p);
368 if (who != current->uid)
369 free_uid(user); /* for find_user() */
373 read_unlock(&tasklist_lock);
379 * emergency_restart - reboot the system
381 * Without shutting down any hardware or taking any locks
382 * reboot the system. This is called when we know we are in
383 * trouble so this is our best effort to reboot. This is
384 * safe to call in interrupt context.
386 void emergency_restart(void)
388 machine_emergency_restart();
390 EXPORT_SYMBOL_GPL(emergency_restart);
392 void kernel_restart_prepare(char *cmd)
394 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
395 system_state = SYSTEM_RESTART;
400 * kernel_restart - reboot the system
401 * @cmd: pointer to buffer containing command to execute for restart
404 * Shutdown everything and perform a clean reboot.
405 * This is not safe to call in interrupt context.
407 void kernel_restart(char *cmd)
409 kernel_restart_prepare(cmd);
411 printk(KERN_EMERG "Restarting system.\n");
413 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
416 machine_restart(cmd);
418 EXPORT_SYMBOL_GPL(kernel_restart);
421 * kernel_kexec - reboot the system
423 * Move into place and start executing a preloaded standalone
424 * executable. If nothing was preloaded return an error.
426 void kernel_kexec(void)
429 struct kimage *image;
430 image = xchg(&kexec_image, 0);
434 kernel_restart_prepare(NULL);
435 printk(KERN_EMERG "Starting new kernel\n");
437 machine_kexec(image);
440 EXPORT_SYMBOL_GPL(kernel_kexec);
443 * kernel_halt - halt the system
445 * Shutdown everything and perform a clean system halt.
447 void kernel_halt_prepare(void)
449 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
450 system_state = SYSTEM_HALT;
453 void kernel_halt(void)
455 kernel_halt_prepare();
456 printk(KERN_EMERG "System halted.\n");
459 EXPORT_SYMBOL_GPL(kernel_halt);
462 * kernel_power_off - power_off the system
464 * Shutdown everything and perform a clean system power_off.
466 void kernel_power_off_prepare(void)
468 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
469 system_state = SYSTEM_POWER_OFF;
472 void kernel_power_off(void)
474 kernel_power_off_prepare();
475 printk(KERN_EMERG "Power down.\n");
478 EXPORT_SYMBOL_GPL(kernel_power_off);
481 * Reboot system call: for obvious reasons only root may call it,
482 * and even root needs to set up some magic numbers in the registers
483 * so that some mistake won't make this reboot the whole machine.
484 * You can also set the meaning of the ctrl-alt-del-key here.
486 * reboot doesn't sync: do that yourself before calling this.
488 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
492 /* We only trust the superuser with rebooting the system. */
493 if (!capable(CAP_SYS_BOOT))
496 /* For safety, we require "magic" arguments. */
497 if (magic1 != LINUX_REBOOT_MAGIC1 ||
498 (magic2 != LINUX_REBOOT_MAGIC2 &&
499 magic2 != LINUX_REBOOT_MAGIC2A &&
500 magic2 != LINUX_REBOOT_MAGIC2B &&
501 magic2 != LINUX_REBOOT_MAGIC2C))
504 /* Instead of trying to make the power_off code look like
505 * halt when pm_power_off is not set do it the easy way.
507 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
508 cmd = LINUX_REBOOT_CMD_HALT;
512 case LINUX_REBOOT_CMD_RESTART:
513 kernel_restart(NULL);
516 case LINUX_REBOOT_CMD_CAD_ON:
520 case LINUX_REBOOT_CMD_CAD_OFF:
524 case LINUX_REBOOT_CMD_HALT:
530 case LINUX_REBOOT_CMD_POWER_OFF:
536 case LINUX_REBOOT_CMD_RESTART2:
537 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
541 buffer[sizeof(buffer) - 1] = '\0';
543 kernel_restart(buffer);
546 case LINUX_REBOOT_CMD_KEXEC:
551 #ifdef CONFIG_SOFTWARE_SUSPEND
552 case LINUX_REBOOT_CMD_SW_SUSPEND:
554 int ret = software_suspend();
568 static void deferred_cad(void *dummy)
570 kernel_restart(NULL);
574 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
575 * As it's called within an interrupt, it may NOT sync: the only choice
576 * is whether to reboot at once, or just ignore the ctrl-alt-del.
578 void ctrl_alt_del(void)
580 static DECLARE_WORK(cad_work, deferred_cad, NULL);
583 schedule_work(&cad_work);
585 kill_proc(cad_pid, SIGINT, 1);
590 * Unprivileged users may change the real gid to the effective gid
591 * or vice versa. (BSD-style)
593 * If you set the real gid at all, or set the effective gid to a value not
594 * equal to the real gid, then the saved gid is set to the new effective gid.
596 * This makes it possible for a setgid program to completely drop its
597 * privileges, which is often a useful assertion to make when you are doing
598 * a security audit over a program.
600 * The general idea is that a program which uses just setregid() will be
601 * 100% compatible with BSD. A program which uses just setgid() will be
602 * 100% compatible with POSIX with saved IDs.
604 * SMP: There are not races, the GIDs are checked only by filesystem
605 * operations (as far as semantic preservation is concerned).
607 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
609 int old_rgid = current->gid;
610 int old_egid = current->egid;
611 int new_rgid = old_rgid;
612 int new_egid = old_egid;
615 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
619 if (rgid != (gid_t) -1) {
620 if ((old_rgid == rgid) ||
621 (current->egid==rgid) ||
627 if (egid != (gid_t) -1) {
628 if ((old_rgid == egid) ||
629 (current->egid == egid) ||
630 (current->sgid == egid) ||
637 if (new_egid != old_egid)
639 current->mm->dumpable = suid_dumpable;
642 if (rgid != (gid_t) -1 ||
643 (egid != (gid_t) -1 && egid != old_rgid))
644 current->sgid = new_egid;
645 current->fsgid = new_egid;
646 current->egid = new_egid;
647 current->gid = new_rgid;
648 key_fsgid_changed(current);
649 proc_id_connector(current, PROC_EVENT_GID);
654 * setgid() is implemented like SysV w/ SAVED_IDS
656 * SMP: Same implicit races as above.
658 asmlinkage long sys_setgid(gid_t gid)
660 int old_egid = current->egid;
663 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
667 if (capable(CAP_SETGID))
671 current->mm->dumpable = suid_dumpable;
674 current->gid = current->egid = current->sgid = current->fsgid = gid;
676 else if ((gid == current->gid) || (gid == current->sgid))
680 current->mm->dumpable = suid_dumpable;
683 current->egid = current->fsgid = gid;
688 key_fsgid_changed(current);
689 proc_id_connector(current, PROC_EVENT_GID);
693 static int set_user(uid_t new_ruid, int dumpclear)
695 struct user_struct *new_user;
697 new_user = alloc_uid(new_ruid);
701 if (atomic_read(&new_user->processes) >=
702 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
703 new_user != &root_user) {
708 switch_uid(new_user);
712 current->mm->dumpable = suid_dumpable;
715 current->uid = new_ruid;
720 * Unprivileged users may change the real uid to the effective uid
721 * or vice versa. (BSD-style)
723 * If you set the real uid at all, or set the effective uid to a value not
724 * equal to the real uid, then the saved uid is set to the new effective uid.
726 * This makes it possible for a setuid program to completely drop its
727 * privileges, which is often a useful assertion to make when you are doing
728 * a security audit over a program.
730 * The general idea is that a program which uses just setreuid() will be
731 * 100% compatible with BSD. A program which uses just setuid() will be
732 * 100% compatible with POSIX with saved IDs.
734 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
736 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
739 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
743 new_ruid = old_ruid = current->uid;
744 new_euid = old_euid = current->euid;
745 old_suid = current->suid;
747 if (ruid != (uid_t) -1) {
749 if ((old_ruid != ruid) &&
750 (current->euid != ruid) &&
751 !capable(CAP_SETUID))
755 if (euid != (uid_t) -1) {
757 if ((old_ruid != euid) &&
758 (current->euid != euid) &&
759 (current->suid != euid) &&
760 !capable(CAP_SETUID))
764 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
767 if (new_euid != old_euid)
769 current->mm->dumpable = suid_dumpable;
772 current->fsuid = current->euid = new_euid;
773 if (ruid != (uid_t) -1 ||
774 (euid != (uid_t) -1 && euid != old_ruid))
775 current->suid = current->euid;
776 current->fsuid = current->euid;
778 key_fsuid_changed(current);
779 proc_id_connector(current, PROC_EVENT_UID);
781 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
787 * setuid() is implemented like SysV with SAVED_IDS
789 * Note that SAVED_ID's is deficient in that a setuid root program
790 * like sendmail, for example, cannot set its uid to be a normal
791 * user and then switch back, because if you're root, setuid() sets
792 * the saved uid too. If you don't like this, blame the bright people
793 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
794 * will allow a root program to temporarily drop privileges and be able to
795 * regain them by swapping the real and effective uid.
797 asmlinkage long sys_setuid(uid_t uid)
799 int old_euid = current->euid;
800 int old_ruid, old_suid, new_ruid, new_suid;
803 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
807 old_ruid = new_ruid = current->uid;
808 old_suid = current->suid;
811 if (capable(CAP_SETUID)) {
812 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
815 } else if ((uid != current->uid) && (uid != new_suid))
820 current->mm->dumpable = suid_dumpable;
823 current->fsuid = current->euid = uid;
824 current->suid = new_suid;
826 key_fsuid_changed(current);
827 proc_id_connector(current, PROC_EVENT_UID);
829 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
834 * This function implements a generic ability to update ruid, euid,
835 * and suid. This allows you to implement the 4.4 compatible seteuid().
837 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
839 int old_ruid = current->uid;
840 int old_euid = current->euid;
841 int old_suid = current->suid;
844 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
848 if (!capable(CAP_SETUID)) {
849 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
850 (ruid != current->euid) && (ruid != current->suid))
852 if ((euid != (uid_t) -1) && (euid != current->uid) &&
853 (euid != current->euid) && (euid != current->suid))
855 if ((suid != (uid_t) -1) && (suid != current->uid) &&
856 (suid != current->euid) && (suid != current->suid))
859 if (ruid != (uid_t) -1) {
860 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
863 if (euid != (uid_t) -1) {
864 if (euid != current->euid)
866 current->mm->dumpable = suid_dumpable;
869 current->euid = euid;
871 current->fsuid = current->euid;
872 if (suid != (uid_t) -1)
873 current->suid = suid;
875 key_fsuid_changed(current);
876 proc_id_connector(current, PROC_EVENT_UID);
878 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
881 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
885 if (!(retval = put_user(current->uid, ruid)) &&
886 !(retval = put_user(current->euid, euid)))
887 retval = put_user(current->suid, suid);
893 * Same as above, but for rgid, egid, sgid.
895 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
899 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
903 if (!capable(CAP_SETGID)) {
904 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
905 (rgid != current->egid) && (rgid != current->sgid))
907 if ((egid != (gid_t) -1) && (egid != current->gid) &&
908 (egid != current->egid) && (egid != current->sgid))
910 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
911 (sgid != current->egid) && (sgid != current->sgid))
914 if (egid != (gid_t) -1) {
915 if (egid != current->egid)
917 current->mm->dumpable = suid_dumpable;
920 current->egid = egid;
922 current->fsgid = current->egid;
923 if (rgid != (gid_t) -1)
925 if (sgid != (gid_t) -1)
926 current->sgid = sgid;
928 key_fsgid_changed(current);
929 proc_id_connector(current, PROC_EVENT_GID);
933 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
937 if (!(retval = put_user(current->gid, rgid)) &&
938 !(retval = put_user(current->egid, egid)))
939 retval = put_user(current->sgid, sgid);
946 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
947 * is used for "access()" and for the NFS daemon (letting nfsd stay at
948 * whatever uid it wants to). It normally shadows "euid", except when
949 * explicitly set by setfsuid() or for access..
951 asmlinkage long sys_setfsuid(uid_t uid)
955 old_fsuid = current->fsuid;
956 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
959 if (uid == current->uid || uid == current->euid ||
960 uid == current->suid || uid == current->fsuid ||
963 if (uid != old_fsuid)
965 current->mm->dumpable = suid_dumpable;
968 current->fsuid = uid;
971 key_fsuid_changed(current);
972 proc_id_connector(current, PROC_EVENT_UID);
974 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
980 * Samma på svenska..
982 asmlinkage long sys_setfsgid(gid_t gid)
986 old_fsgid = current->fsgid;
987 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
990 if (gid == current->gid || gid == current->egid ||
991 gid == current->sgid || gid == current->fsgid ||
994 if (gid != old_fsgid)
996 current->mm->dumpable = suid_dumpable;
999 current->fsgid = gid;
1000 key_fsgid_changed(current);
1001 proc_id_connector(current, PROC_EVENT_GID);
1006 asmlinkage long sys_times(struct tms __user * tbuf)
1009 * In the SMP world we might just be unlucky and have one of
1010 * the times increment as we use it. Since the value is an
1011 * atomically safe type this is just fine. Conceptually its
1012 * as if the syscall took an instant longer to occur.
1016 cputime_t utime, stime, cutime, cstime;
1019 if (thread_group_empty(current)) {
1021 * Single thread case without the use of any locks.
1023 * We may race with release_task if two threads are
1024 * executing. However, release task first adds up the
1025 * counters (__exit_signal) before removing the task
1026 * from the process tasklist (__unhash_process).
1027 * __exit_signal also acquires and releases the
1028 * siglock which results in the proper memory ordering
1029 * so that the list modifications are always visible
1030 * after the counters have been updated.
1032 * If the counters have been updated by the second thread
1033 * but the thread has not yet been removed from the list
1034 * then the other branch will be executing which will
1035 * block on tasklist_lock until the exit handling of the
1036 * other task is finished.
1038 * This also implies that the sighand->siglock cannot
1039 * be held by another processor. So we can also
1040 * skip acquiring that lock.
1042 utime = cputime_add(current->signal->utime, current->utime);
1043 stime = cputime_add(current->signal->utime, current->stime);
1044 cutime = current->signal->cutime;
1045 cstime = current->signal->cstime;
1050 /* Process with multiple threads */
1051 struct task_struct *tsk = current;
1052 struct task_struct *t;
1054 read_lock(&tasklist_lock);
1055 utime = tsk->signal->utime;
1056 stime = tsk->signal->stime;
1059 utime = cputime_add(utime, t->utime);
1060 stime = cputime_add(stime, t->stime);
1065 * While we have tasklist_lock read-locked, no dying thread
1066 * can be updating current->signal->[us]time. Instead,
1067 * we got their counts included in the live thread loop.
1068 * However, another thread can come in right now and
1069 * do a wait call that updates current->signal->c[us]time.
1070 * To make sure we always see that pair updated atomically,
1071 * we take the siglock around fetching them.
1073 spin_lock_irq(&tsk->sighand->siglock);
1074 cutime = tsk->signal->cutime;
1075 cstime = tsk->signal->cstime;
1076 spin_unlock_irq(&tsk->sighand->siglock);
1077 read_unlock(&tasklist_lock);
1079 tmp.tms_utime = cputime_to_clock_t(utime);
1080 tmp.tms_stime = cputime_to_clock_t(stime);
1081 tmp.tms_cutime = cputime_to_clock_t(cutime);
1082 tmp.tms_cstime = cputime_to_clock_t(cstime);
1083 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1086 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1090 * This needs some heavy checking ...
1091 * I just haven't the stomach for it. I also don't fully
1092 * understand sessions/pgrp etc. Let somebody who does explain it.
1094 * OK, I think I have the protection semantics right.... this is really
1095 * only important on a multi-user system anyway, to make sure one user
1096 * can't send a signal to a process owned by another. -TYT, 12/12/91
1098 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1102 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1104 struct task_struct *p;
1105 struct task_struct *group_leader = current->group_leader;
1109 pid = group_leader->pid;
1115 /* From this point forward we keep holding onto the tasklist lock
1116 * so that our parent does not change from under us. -DaveM
1118 write_lock_irq(&tasklist_lock);
1121 p = find_task_by_pid(pid);
1126 if (!thread_group_leader(p))
1129 if (p->real_parent == group_leader) {
1131 if (p->signal->session != group_leader->signal->session)
1138 if (p != group_leader)
1143 if (p->signal->leader)
1147 struct task_struct *p;
1149 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1150 if (p->signal->session == group_leader->signal->session)
1152 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1157 err = security_task_setpgid(p, pgid);
1161 if (process_group(p) != pgid) {
1162 detach_pid(p, PIDTYPE_PGID);
1163 p->signal->pgrp = pgid;
1164 attach_pid(p, PIDTYPE_PGID, pgid);
1169 /* All paths lead to here, thus we are safe. -DaveM */
1170 write_unlock_irq(&tasklist_lock);
1174 asmlinkage long sys_getpgid(pid_t pid)
1177 return process_group(current);
1180 struct task_struct *p;
1182 read_lock(&tasklist_lock);
1183 p = find_task_by_pid(pid);
1187 retval = security_task_getpgid(p);
1189 retval = process_group(p);
1191 read_unlock(&tasklist_lock);
1196 #ifdef __ARCH_WANT_SYS_GETPGRP
1198 asmlinkage long sys_getpgrp(void)
1200 /* SMP - assuming writes are word atomic this is fine */
1201 return process_group(current);
1206 asmlinkage long sys_getsid(pid_t pid)
1209 return current->signal->session;
1212 struct task_struct *p;
1214 read_lock(&tasklist_lock);
1215 p = find_task_by_pid(pid);
1219 retval = security_task_getsid(p);
1221 retval = p->signal->session;
1223 read_unlock(&tasklist_lock);
1228 asmlinkage long sys_setsid(void)
1230 struct task_struct *group_leader = current->group_leader;
1235 write_lock_irq(&tasklist_lock);
1237 pid = find_pid(PIDTYPE_PGID, group_leader->pid);
1241 group_leader->signal->leader = 1;
1242 __set_special_pids(group_leader->pid, group_leader->pid);
1243 group_leader->signal->tty = NULL;
1244 group_leader->signal->tty_old_pgrp = 0;
1245 err = process_group(group_leader);
1247 write_unlock_irq(&tasklist_lock);
1253 * Supplementary group IDs
1256 /* init to 2 - one for init_task, one to ensure it is never freed */
1257 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1259 struct group_info *groups_alloc(int gidsetsize)
1261 struct group_info *group_info;
1265 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1266 /* Make sure we always allocate at least one indirect block pointer */
1267 nblocks = nblocks ? : 1;
1268 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1271 group_info->ngroups = gidsetsize;
1272 group_info->nblocks = nblocks;
1273 atomic_set(&group_info->usage, 1);
1275 if (gidsetsize <= NGROUPS_SMALL) {
1276 group_info->blocks[0] = group_info->small_block;
1278 for (i = 0; i < nblocks; i++) {
1280 b = (void *)__get_free_page(GFP_USER);
1282 goto out_undo_partial_alloc;
1283 group_info->blocks[i] = b;
1288 out_undo_partial_alloc:
1290 free_page((unsigned long)group_info->blocks[i]);
1296 EXPORT_SYMBOL(groups_alloc);
1298 void groups_free(struct group_info *group_info)
1300 if (group_info->blocks[0] != group_info->small_block) {
1302 for (i = 0; i < group_info->nblocks; i++)
1303 free_page((unsigned long)group_info->blocks[i]);
1308 EXPORT_SYMBOL(groups_free);
1310 /* export the group_info to a user-space array */
1311 static int groups_to_user(gid_t __user *grouplist,
1312 struct group_info *group_info)
1315 int count = group_info->ngroups;
1317 for (i = 0; i < group_info->nblocks; i++) {
1318 int cp_count = min(NGROUPS_PER_BLOCK, count);
1319 int off = i * NGROUPS_PER_BLOCK;
1320 int len = cp_count * sizeof(*grouplist);
1322 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1330 /* fill a group_info from a user-space array - it must be allocated already */
1331 static int groups_from_user(struct group_info *group_info,
1332 gid_t __user *grouplist)
1335 int count = group_info->ngroups;
1337 for (i = 0; i < group_info->nblocks; i++) {
1338 int cp_count = min(NGROUPS_PER_BLOCK, count);
1339 int off = i * NGROUPS_PER_BLOCK;
1340 int len = cp_count * sizeof(*grouplist);
1342 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1350 /* a simple Shell sort */
1351 static void groups_sort(struct group_info *group_info)
1353 int base, max, stride;
1354 int gidsetsize = group_info->ngroups;
1356 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1361 max = gidsetsize - stride;
1362 for (base = 0; base < max; base++) {
1364 int right = left + stride;
1365 gid_t tmp = GROUP_AT(group_info, right);
1367 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1368 GROUP_AT(group_info, right) =
1369 GROUP_AT(group_info, left);
1373 GROUP_AT(group_info, right) = tmp;
1379 /* a simple bsearch */
1380 int groups_search(struct group_info *group_info, gid_t grp)
1388 right = group_info->ngroups;
1389 while (left < right) {
1390 int mid = (left+right)/2;
1391 int cmp = grp - GROUP_AT(group_info, mid);
1402 /* validate and set current->group_info */
1403 int set_current_groups(struct group_info *group_info)
1406 struct group_info *old_info;
1408 retval = security_task_setgroups(group_info);
1412 groups_sort(group_info);
1413 get_group_info(group_info);
1416 old_info = current->group_info;
1417 current->group_info = group_info;
1418 task_unlock(current);
1420 put_group_info(old_info);
1425 EXPORT_SYMBOL(set_current_groups);
1427 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1432 * SMP: Nobody else can change our grouplist. Thus we are
1439 /* no need to grab task_lock here; it cannot change */
1440 get_group_info(current->group_info);
1441 i = current->group_info->ngroups;
1443 if (i > gidsetsize) {
1447 if (groups_to_user(grouplist, current->group_info)) {
1453 put_group_info(current->group_info);
1458 * SMP: Our groups are copy-on-write. We can set them safely
1459 * without another task interfering.
1462 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1464 struct group_info *group_info;
1467 if (!capable(CAP_SETGID))
1469 if ((unsigned)gidsetsize > NGROUPS_MAX)
1472 group_info = groups_alloc(gidsetsize);
1475 retval = groups_from_user(group_info, grouplist);
1477 put_group_info(group_info);
1481 retval = set_current_groups(group_info);
1482 put_group_info(group_info);
1488 * Check whether we're fsgid/egid or in the supplemental group..
1490 int in_group_p(gid_t grp)
1493 if (grp != current->fsgid) {
1494 get_group_info(current->group_info);
1495 retval = groups_search(current->group_info, grp);
1496 put_group_info(current->group_info);
1501 EXPORT_SYMBOL(in_group_p);
1503 int in_egroup_p(gid_t grp)
1506 if (grp != current->egid) {
1507 get_group_info(current->group_info);
1508 retval = groups_search(current->group_info, grp);
1509 put_group_info(current->group_info);
1514 EXPORT_SYMBOL(in_egroup_p);
1516 DECLARE_RWSEM(uts_sem);
1518 EXPORT_SYMBOL(uts_sem);
1520 asmlinkage long sys_newuname(struct new_utsname __user * name)
1524 down_read(&uts_sem);
1525 if (copy_to_user(name,&system_utsname,sizeof *name))
1531 asmlinkage long sys_sethostname(char __user *name, int len)
1534 char tmp[__NEW_UTS_LEN];
1536 if (!capable(CAP_SYS_ADMIN))
1538 if (len < 0 || len > __NEW_UTS_LEN)
1540 down_write(&uts_sem);
1542 if (!copy_from_user(tmp, name, len)) {
1543 memcpy(system_utsname.nodename, tmp, len);
1544 system_utsname.nodename[len] = 0;
1551 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1553 asmlinkage long sys_gethostname(char __user *name, int len)
1559 down_read(&uts_sem);
1560 i = 1 + strlen(system_utsname.nodename);
1564 if (copy_to_user(name, system_utsname.nodename, i))
1573 * Only setdomainname; getdomainname can be implemented by calling
1576 asmlinkage long sys_setdomainname(char __user *name, int len)
1579 char tmp[__NEW_UTS_LEN];
1581 if (!capable(CAP_SYS_ADMIN))
1583 if (len < 0 || len > __NEW_UTS_LEN)
1586 down_write(&uts_sem);
1588 if (!copy_from_user(tmp, name, len)) {
1589 memcpy(system_utsname.domainname, tmp, len);
1590 system_utsname.domainname[len] = 0;
1597 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1599 if (resource >= RLIM_NLIMITS)
1602 struct rlimit value;
1603 task_lock(current->group_leader);
1604 value = current->signal->rlim[resource];
1605 task_unlock(current->group_leader);
1606 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1610 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1613 * Back compatibility for getrlimit. Needed for some apps.
1616 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1619 if (resource >= RLIM_NLIMITS)
1622 task_lock(current->group_leader);
1623 x = current->signal->rlim[resource];
1624 task_unlock(current->group_leader);
1625 if(x.rlim_cur > 0x7FFFFFFF)
1626 x.rlim_cur = 0x7FFFFFFF;
1627 if(x.rlim_max > 0x7FFFFFFF)
1628 x.rlim_max = 0x7FFFFFFF;
1629 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1634 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1636 struct rlimit new_rlim, *old_rlim;
1639 if (resource >= RLIM_NLIMITS)
1641 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1643 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1645 old_rlim = current->signal->rlim + resource;
1646 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1647 !capable(CAP_SYS_RESOURCE))
1649 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1652 retval = security_task_setrlimit(resource, &new_rlim);
1656 task_lock(current->group_leader);
1657 *old_rlim = new_rlim;
1658 task_unlock(current->group_leader);
1660 if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1661 (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1662 new_rlim.rlim_cur <= cputime_to_secs(
1663 current->signal->it_prof_expires))) {
1664 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1665 read_lock(&tasklist_lock);
1666 spin_lock_irq(¤t->sighand->siglock);
1667 set_process_cpu_timer(current, CPUCLOCK_PROF,
1669 spin_unlock_irq(¤t->sighand->siglock);
1670 read_unlock(&tasklist_lock);
1677 * It would make sense to put struct rusage in the task_struct,
1678 * except that would make the task_struct be *really big*. After
1679 * task_struct gets moved into malloc'ed memory, it would
1680 * make sense to do this. It will make moving the rest of the information
1681 * a lot simpler! (Which we're not doing right now because we're not
1682 * measuring them yet).
1684 * This expects to be called with tasklist_lock read-locked or better,
1685 * and the siglock not locked. It may momentarily take the siglock.
1687 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1688 * races with threads incrementing their own counters. But since word
1689 * reads are atomic, we either get new values or old values and we don't
1690 * care which for the sums. We always take the siglock to protect reading
1691 * the c* fields from p->signal from races with exit.c updating those
1692 * fields when reaping, so a sample either gets all the additions of a
1693 * given child after it's reaped, or none so this sample is before reaping.
1696 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1698 struct task_struct *t;
1699 unsigned long flags;
1700 cputime_t utime, stime;
1702 memset((char *) r, 0, sizeof *r);
1704 if (unlikely(!p->signal))
1707 utime = stime = cputime_zero;
1711 case RUSAGE_CHILDREN:
1712 spin_lock_irqsave(&p->sighand->siglock, flags);
1713 utime = p->signal->cutime;
1714 stime = p->signal->cstime;
1715 r->ru_nvcsw = p->signal->cnvcsw;
1716 r->ru_nivcsw = p->signal->cnivcsw;
1717 r->ru_minflt = p->signal->cmin_flt;
1718 r->ru_majflt = p->signal->cmaj_flt;
1719 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1721 if (who == RUSAGE_CHILDREN)
1725 utime = cputime_add(utime, p->signal->utime);
1726 stime = cputime_add(stime, p->signal->stime);
1727 r->ru_nvcsw += p->signal->nvcsw;
1728 r->ru_nivcsw += p->signal->nivcsw;
1729 r->ru_minflt += p->signal->min_flt;
1730 r->ru_majflt += p->signal->maj_flt;
1733 utime = cputime_add(utime, t->utime);
1734 stime = cputime_add(stime, t->stime);
1735 r->ru_nvcsw += t->nvcsw;
1736 r->ru_nivcsw += t->nivcsw;
1737 r->ru_minflt += t->min_flt;
1738 r->ru_majflt += t->maj_flt;
1747 cputime_to_timeval(utime, &r->ru_utime);
1748 cputime_to_timeval(stime, &r->ru_stime);
1751 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1754 read_lock(&tasklist_lock);
1755 k_getrusage(p, who, &r);
1756 read_unlock(&tasklist_lock);
1757 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1760 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1762 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1764 return getrusage(current, who, ru);
1767 asmlinkage long sys_umask(int mask)
1769 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1773 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1774 unsigned long arg4, unsigned long arg5)
1778 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1783 case PR_SET_PDEATHSIG:
1784 if (!valid_signal(arg2)) {
1788 current->pdeath_signal = arg2;
1790 case PR_GET_PDEATHSIG:
1791 error = put_user(current->pdeath_signal, (int __user *)arg2);
1793 case PR_GET_DUMPABLE:
1794 error = current->mm->dumpable;
1796 case PR_SET_DUMPABLE:
1797 if (arg2 < 0 || arg2 > 2) {
1801 current->mm->dumpable = arg2;
1804 case PR_SET_UNALIGN:
1805 error = SET_UNALIGN_CTL(current, arg2);
1807 case PR_GET_UNALIGN:
1808 error = GET_UNALIGN_CTL(current, arg2);
1811 error = SET_FPEMU_CTL(current, arg2);
1814 error = GET_FPEMU_CTL(current, arg2);
1817 error = SET_FPEXC_CTL(current, arg2);
1820 error = GET_FPEXC_CTL(current, arg2);
1823 error = PR_TIMING_STATISTICAL;
1826 if (arg2 == PR_TIMING_STATISTICAL)
1832 case PR_GET_KEEPCAPS:
1833 if (current->keep_capabilities)
1836 case PR_SET_KEEPCAPS:
1837 if (arg2 != 0 && arg2 != 1) {
1841 current->keep_capabilities = arg2;
1844 struct task_struct *me = current;
1845 unsigned char ncomm[sizeof(me->comm)];
1847 ncomm[sizeof(me->comm)-1] = 0;
1848 if (strncpy_from_user(ncomm, (char __user *)arg2,
1849 sizeof(me->comm)-1) < 0)
1851 set_task_comm(me, ncomm);
1855 struct task_struct *me = current;
1856 unsigned char tcomm[sizeof(me->comm)];
1858 get_task_comm(tcomm, me);
1859 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))