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/kmod.h>
15 #include <linux/reboot.h>
16 #include <linux/prctl.h>
17 #include <linux/init.h>
18 #include <linux/highuid.h>
20 #include <linux/workqueue.h>
21 #include <linux/device.h>
22 #include <linux/key.h>
23 #include <linux/times.h>
24 #include <linux/posix-timers.h>
25 #include <linux/security.h>
26 #include <linux/dcookies.h>
27 #include <linux/suspend.h>
28 #include <linux/tty.h>
29 #include <linux/signal.h>
31 #include <linux/compat.h>
32 #include <linux/syscalls.h>
33 #include <linux/vs_cvirt.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)) {
234 if (vx_flags(VXF_IGNEG_NICE, 0))
240 no_nice = security_task_setnice(p, niceval);
247 set_user_nice(p, niceval);
252 asmlinkage long sys_setpriority(int which, int who, int niceval)
254 struct task_struct *g, *p;
255 struct user_struct *user;
258 if (which > 2 || which < 0)
261 /* normalize: avoid signed division (rounding problems) */
268 read_lock(&tasklist_lock);
273 p = find_task_by_pid(who);
275 error = set_one_prio(p, niceval, error);
279 who = process_group(current);
280 do_each_task_pid(who, PIDTYPE_PGID, p) {
281 error = set_one_prio(p, niceval, error);
282 } while_each_task_pid(who, PIDTYPE_PGID, p);
285 user = current->user;
289 if ((who != current->uid) &&
290 !(user = find_user(vx_current_xid(), who)))
291 goto out_unlock; /* No processes for this user */
295 error = set_one_prio(p, niceval, error);
296 while_each_thread(g, p);
297 if (who != current->uid)
298 free_uid(user); /* For find_user() */
302 read_unlock(&tasklist_lock);
308 * Ugh. To avoid negative return values, "getpriority()" will
309 * not return the normal nice-value, but a negated value that
310 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
311 * to stay compatible.
313 asmlinkage long sys_getpriority(int which, int who)
315 struct task_struct *g, *p;
316 struct user_struct *user;
317 long niceval, retval = -ESRCH;
319 if (which > 2 || which < 0)
322 read_lock(&tasklist_lock);
327 p = find_task_by_pid(who);
329 niceval = 20 - task_nice(p);
330 if (niceval > retval)
336 who = process_group(current);
337 do_each_task_pid(who, PIDTYPE_PGID, p) {
338 niceval = 20 - task_nice(p);
339 if (niceval > retval)
341 } while_each_task_pid(who, PIDTYPE_PGID, p);
344 user = current->user;
348 if ((who != current->uid) &&
349 !(user = find_user(vx_current_xid(), who)))
350 goto out_unlock; /* No processes for this user */
354 niceval = 20 - task_nice(p);
355 if (niceval > retval)
358 while_each_thread(g, p);
359 if (who != current->uid)
360 free_uid(user); /* for find_user() */
364 read_unlock(&tasklist_lock);
369 long vs_reboot(unsigned int, void *);
372 * Reboot system call: for obvious reasons only root may call it,
373 * and even root needs to set up some magic numbers in the registers
374 * so that some mistake won't make this reboot the whole machine.
375 * You can also set the meaning of the ctrl-alt-del-key here.
377 * reboot doesn't sync: do that yourself before calling this.
379 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
383 /* We only trust the superuser with rebooting the system. */
384 if (!capable(CAP_SYS_BOOT))
387 /* For safety, we require "magic" arguments. */
388 if (magic1 != LINUX_REBOOT_MAGIC1 ||
389 (magic2 != LINUX_REBOOT_MAGIC2 &&
390 magic2 != LINUX_REBOOT_MAGIC2A &&
391 magic2 != LINUX_REBOOT_MAGIC2B &&
392 magic2 != LINUX_REBOOT_MAGIC2C))
395 if (!vx_check(0, VX_ADMIN|VX_WATCH))
396 return vs_reboot(cmd, arg);
400 case LINUX_REBOOT_CMD_RESTART:
401 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
402 system_state = SYSTEM_RESTART;
404 printk(KERN_EMERG "Restarting system.\n");
405 machine_restart(NULL);
408 case LINUX_REBOOT_CMD_CAD_ON:
412 case LINUX_REBOOT_CMD_CAD_OFF:
416 case LINUX_REBOOT_CMD_HALT:
417 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
418 system_state = SYSTEM_HALT;
420 printk(KERN_EMERG "System halted.\n");
426 case LINUX_REBOOT_CMD_POWER_OFF:
427 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
428 system_state = SYSTEM_POWER_OFF;
430 printk(KERN_EMERG "Power down.\n");
436 case LINUX_REBOOT_CMD_RESTART2:
437 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
441 buffer[sizeof(buffer) - 1] = '\0';
443 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
444 system_state = SYSTEM_RESTART;
446 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
447 machine_restart(buffer);
450 #ifdef CONFIG_SOFTWARE_SUSPEND
451 case LINUX_REBOOT_CMD_SW_SUSPEND:
453 int ret = software_suspend();
467 static void deferred_cad(void *dummy)
469 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
470 machine_restart(NULL);
474 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
475 * As it's called within an interrupt, it may NOT sync: the only choice
476 * is whether to reboot at once, or just ignore the ctrl-alt-del.
478 void ctrl_alt_del(void)
480 static DECLARE_WORK(cad_work, deferred_cad, NULL);
483 schedule_work(&cad_work);
485 kill_proc(cad_pid, SIGINT, 1);
490 * Unprivileged users may change the real gid to the effective gid
491 * or vice versa. (BSD-style)
493 * If you set the real gid at all, or set the effective gid to a value not
494 * equal to the real gid, then the saved gid is set to the new effective gid.
496 * This makes it possible for a setgid program to completely drop its
497 * privileges, which is often a useful assertion to make when you are doing
498 * a security audit over a program.
500 * The general idea is that a program which uses just setregid() will be
501 * 100% compatible with BSD. A program which uses just setgid() will be
502 * 100% compatible with POSIX with saved IDs.
504 * SMP: There are not races, the GIDs are checked only by filesystem
505 * operations (as far as semantic preservation is concerned).
507 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
509 int old_rgid = current->gid;
510 int old_egid = current->egid;
511 int new_rgid = old_rgid;
512 int new_egid = old_egid;
515 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
519 if (rgid != (gid_t) -1) {
520 if ((old_rgid == rgid) ||
521 (current->egid==rgid) ||
527 if (egid != (gid_t) -1) {
528 if ((old_rgid == egid) ||
529 (current->egid == egid) ||
530 (current->sgid == egid) ||
537 if (new_egid != old_egid)
539 current->mm->dumpable = 0;
542 if (rgid != (gid_t) -1 ||
543 (egid != (gid_t) -1 && egid != old_rgid))
544 current->sgid = new_egid;
545 current->fsgid = new_egid;
546 current->egid = new_egid;
547 current->gid = new_rgid;
548 key_fsgid_changed(current);
553 * setgid() is implemented like SysV w/ SAVED_IDS
555 * SMP: Same implicit races as above.
557 asmlinkage long sys_setgid(gid_t gid)
559 int old_egid = current->egid;
562 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
566 if (capable(CAP_SETGID))
570 current->mm->dumpable=0;
573 current->gid = current->egid = current->sgid = current->fsgid = gid;
575 else if ((gid == current->gid) || (gid == current->sgid))
579 current->mm->dumpable=0;
582 current->egid = current->fsgid = gid;
587 key_fsgid_changed(current);
591 static int set_user(uid_t new_ruid, int dumpclear)
593 struct user_struct *new_user;
595 new_user = alloc_uid(vx_current_xid(), new_ruid);
599 if (atomic_read(&new_user->processes) >=
600 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
601 new_user != &root_user) {
606 switch_uid(new_user);
610 current->mm->dumpable = 0;
613 current->uid = new_ruid;
618 * Unprivileged users may change the real uid to the effective uid
619 * or vice versa. (BSD-style)
621 * If you set the real uid at all, or set the effective uid to a value not
622 * equal to the real uid, then the saved uid is set to the new effective uid.
624 * This makes it possible for a setuid program to completely drop its
625 * privileges, which is often a useful assertion to make when you are doing
626 * a security audit over a program.
628 * The general idea is that a program which uses just setreuid() will be
629 * 100% compatible with BSD. A program which uses just setuid() will be
630 * 100% compatible with POSIX with saved IDs.
632 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
634 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
637 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
641 new_ruid = old_ruid = current->uid;
642 new_euid = old_euid = current->euid;
643 old_suid = current->suid;
645 if (ruid != (uid_t) -1) {
647 if ((old_ruid != ruid) &&
648 (current->euid != ruid) &&
649 !capable(CAP_SETUID))
653 if (euid != (uid_t) -1) {
655 if ((old_ruid != euid) &&
656 (current->euid != euid) &&
657 (current->suid != euid) &&
658 !capable(CAP_SETUID))
662 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
665 if (new_euid != old_euid)
667 current->mm->dumpable=0;
670 current->fsuid = current->euid = new_euid;
671 if (ruid != (uid_t) -1 ||
672 (euid != (uid_t) -1 && euid != old_ruid))
673 current->suid = current->euid;
674 current->fsuid = current->euid;
676 key_fsuid_changed(current);
678 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
684 * setuid() is implemented like SysV with SAVED_IDS
686 * Note that SAVED_ID's is deficient in that a setuid root program
687 * like sendmail, for example, cannot set its uid to be a normal
688 * user and then switch back, because if you're root, setuid() sets
689 * the saved uid too. If you don't like this, blame the bright people
690 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
691 * will allow a root program to temporarily drop privileges and be able to
692 * regain them by swapping the real and effective uid.
694 asmlinkage long sys_setuid(uid_t uid)
696 int old_euid = current->euid;
697 int old_ruid, old_suid, new_ruid, new_suid;
700 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
704 old_ruid = new_ruid = current->uid;
705 old_suid = current->suid;
708 if (capable(CAP_SETUID)) {
709 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
712 } else if ((uid != current->uid) && (uid != new_suid))
717 current->mm->dumpable = 0;
720 current->fsuid = current->euid = uid;
721 current->suid = new_suid;
723 key_fsuid_changed(current);
725 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
730 * This function implements a generic ability to update ruid, euid,
731 * and suid. This allows you to implement the 4.4 compatible seteuid().
733 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
735 int old_ruid = current->uid;
736 int old_euid = current->euid;
737 int old_suid = current->suid;
740 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
744 if (!capable(CAP_SETUID)) {
745 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
746 (ruid != current->euid) && (ruid != current->suid))
748 if ((euid != (uid_t) -1) && (euid != current->uid) &&
749 (euid != current->euid) && (euid != current->suid))
751 if ((suid != (uid_t) -1) && (suid != current->uid) &&
752 (suid != current->euid) && (suid != current->suid))
755 if (ruid != (uid_t) -1) {
756 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
759 if (euid != (uid_t) -1) {
760 if (euid != current->euid)
762 current->mm->dumpable = 0;
765 current->euid = euid;
767 current->fsuid = current->euid;
768 if (suid != (uid_t) -1)
769 current->suid = suid;
771 key_fsuid_changed(current);
773 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
776 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
780 if (!(retval = put_user(current->uid, ruid)) &&
781 !(retval = put_user(current->euid, euid)))
782 retval = put_user(current->suid, suid);
788 * Same as above, but for rgid, egid, sgid.
790 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
794 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
798 if (!capable(CAP_SETGID)) {
799 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
800 (rgid != current->egid) && (rgid != current->sgid))
802 if ((egid != (gid_t) -1) && (egid != current->gid) &&
803 (egid != current->egid) && (egid != current->sgid))
805 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
806 (sgid != current->egid) && (sgid != current->sgid))
809 if (egid != (gid_t) -1) {
810 if (egid != current->egid)
812 current->mm->dumpable = 0;
815 current->egid = egid;
817 current->fsgid = current->egid;
818 if (rgid != (gid_t) -1)
820 if (sgid != (gid_t) -1)
821 current->sgid = sgid;
823 key_fsgid_changed(current);
827 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
831 if (!(retval = put_user(current->gid, rgid)) &&
832 !(retval = put_user(current->egid, egid)))
833 retval = put_user(current->sgid, sgid);
840 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
841 * is used for "access()" and for the NFS daemon (letting nfsd stay at
842 * whatever uid it wants to). It normally shadows "euid", except when
843 * explicitly set by setfsuid() or for access..
845 asmlinkage long sys_setfsuid(uid_t uid)
849 old_fsuid = current->fsuid;
850 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
853 if (uid == current->uid || uid == current->euid ||
854 uid == current->suid || uid == current->fsuid ||
857 if (uid != old_fsuid)
859 current->mm->dumpable = 0;
862 current->fsuid = uid;
865 key_fsuid_changed(current);
867 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
873 * Samma på svenska..
875 asmlinkage long sys_setfsgid(gid_t gid)
879 old_fsgid = current->fsgid;
880 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
883 if (gid == current->gid || gid == current->egid ||
884 gid == current->sgid || gid == current->fsgid ||
887 if (gid != old_fsgid)
889 current->mm->dumpable = 0;
892 current->fsgid = gid;
893 key_fsgid_changed(current);
898 asmlinkage long sys_times(struct tms __user * tbuf)
901 * In the SMP world we might just be unlucky and have one of
902 * the times increment as we use it. Since the value is an
903 * atomically safe type this is just fine. Conceptually its
904 * as if the syscall took an instant longer to occur.
908 struct task_struct *tsk = current;
909 struct task_struct *t;
910 cputime_t utime, stime, cutime, cstime;
912 read_lock(&tasklist_lock);
913 utime = tsk->signal->utime;
914 stime = tsk->signal->stime;
917 utime = cputime_add(utime, t->utime);
918 stime = cputime_add(stime, t->stime);
923 * While we have tasklist_lock read-locked, no dying thread
924 * can be updating current->signal->[us]time. Instead,
925 * we got their counts included in the live thread loop.
926 * However, another thread can come in right now and
927 * do a wait call that updates current->signal->c[us]time.
928 * To make sure we always see that pair updated atomically,
929 * we take the siglock around fetching them.
931 spin_lock_irq(&tsk->sighand->siglock);
932 cutime = tsk->signal->cutime;
933 cstime = tsk->signal->cstime;
934 spin_unlock_irq(&tsk->sighand->siglock);
935 read_unlock(&tasklist_lock);
937 tmp.tms_utime = cputime_to_clock_t(utime);
938 tmp.tms_stime = cputime_to_clock_t(stime);
939 tmp.tms_cutime = cputime_to_clock_t(cutime);
940 tmp.tms_cstime = cputime_to_clock_t(cstime);
941 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
944 return (long) jiffies_64_to_clock_t(get_jiffies_64());
948 * This needs some heavy checking ...
949 * I just haven't the stomach for it. I also don't fully
950 * understand sessions/pgrp etc. Let somebody who does explain it.
952 * OK, I think I have the protection semantics right.... this is really
953 * only important on a multi-user system anyway, to make sure one user
954 * can't send a signal to a process owned by another. -TYT, 12/12/91
956 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
960 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
962 struct task_struct *p;
967 pid = vx_map_pid(current->pid);
973 rpgid = vx_rmap_pid(pgid);
975 /* From this point forward we keep holding onto the tasklist lock
976 * so that our parent does not change from under us. -DaveM
978 write_lock_irq(&tasklist_lock);
981 p = find_task_by_pid(pid);
986 if (!thread_group_leader(p))
989 if (p->parent == current || p->real_parent == current) {
991 if (p->signal->session != current->signal->session)
1003 if (p->signal->leader)
1007 struct task_struct *p;
1009 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1010 if (p->signal->session == current->signal->session)
1012 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1017 err = security_task_setpgid(p, rpgid);
1021 if (process_group(p) != rpgid) {
1022 detach_pid(p, PIDTYPE_PGID);
1023 p->signal->pgrp = rpgid;
1024 attach_pid(p, PIDTYPE_PGID, rpgid);
1029 /* All paths lead to here, thus we are safe. -DaveM */
1030 write_unlock_irq(&tasklist_lock);
1034 asmlinkage long sys_getpgid(pid_t pid)
1037 return vx_rmap_pid(process_group(current));
1040 struct task_struct *p;
1042 read_lock(&tasklist_lock);
1043 p = find_task_by_pid(pid);
1047 retval = security_task_getpgid(p);
1049 retval = vx_rmap_pid(process_group(p));
1051 read_unlock(&tasklist_lock);
1056 #ifdef __ARCH_WANT_SYS_GETPGRP
1058 asmlinkage long sys_getpgrp(void)
1060 /* SMP - assuming writes are word atomic this is fine */
1061 return process_group(current);
1066 asmlinkage long sys_getsid(pid_t pid)
1069 return current->signal->session;
1072 struct task_struct *p;
1074 read_lock(&tasklist_lock);
1075 p = find_task_by_pid(pid);
1079 retval = security_task_getsid(p);
1081 retval = p->signal->session;
1083 read_unlock(&tasklist_lock);
1088 asmlinkage long sys_setsid(void)
1093 if (!thread_group_leader(current))
1097 write_lock_irq(&tasklist_lock);
1099 pid = find_pid(PIDTYPE_PGID, current->pid);
1103 current->signal->leader = 1;
1104 __set_special_pids(current->pid, current->pid);
1105 current->signal->tty = NULL;
1106 current->signal->tty_old_pgrp = 0;
1107 err = process_group(current);
1109 write_unlock_irq(&tasklist_lock);
1115 * Supplementary group IDs
1118 /* init to 2 - one for init_task, one to ensure it is never freed */
1119 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1121 struct group_info *groups_alloc(int gidsetsize)
1123 struct group_info *group_info;
1127 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1128 /* Make sure we always allocate at least one indirect block pointer */
1129 nblocks = nblocks ? : 1;
1130 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1133 group_info->ngroups = gidsetsize;
1134 group_info->nblocks = nblocks;
1135 atomic_set(&group_info->usage, 1);
1137 if (gidsetsize <= NGROUPS_SMALL) {
1138 group_info->blocks[0] = group_info->small_block;
1140 for (i = 0; i < nblocks; i++) {
1142 b = (void *)__get_free_page(GFP_USER);
1144 goto out_undo_partial_alloc;
1145 group_info->blocks[i] = b;
1150 out_undo_partial_alloc:
1152 free_page((unsigned long)group_info->blocks[i]);
1158 EXPORT_SYMBOL(groups_alloc);
1160 void groups_free(struct group_info *group_info)
1162 if (group_info->blocks[0] != group_info->small_block) {
1164 for (i = 0; i < group_info->nblocks; i++)
1165 free_page((unsigned long)group_info->blocks[i]);
1170 EXPORT_SYMBOL(groups_free);
1172 /* export the group_info to a user-space array */
1173 static int groups_to_user(gid_t __user *grouplist,
1174 struct group_info *group_info)
1177 int count = group_info->ngroups;
1179 for (i = 0; i < group_info->nblocks; i++) {
1180 int cp_count = min(NGROUPS_PER_BLOCK, count);
1181 int off = i * NGROUPS_PER_BLOCK;
1182 int len = cp_count * sizeof(*grouplist);
1184 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1192 /* fill a group_info from a user-space array - it must be allocated already */
1193 static int groups_from_user(struct group_info *group_info,
1194 gid_t __user *grouplist)
1197 int count = group_info->ngroups;
1199 for (i = 0; i < group_info->nblocks; i++) {
1200 int cp_count = min(NGROUPS_PER_BLOCK, count);
1201 int off = i * NGROUPS_PER_BLOCK;
1202 int len = cp_count * sizeof(*grouplist);
1204 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1212 /* a simple Shell sort */
1213 static void groups_sort(struct group_info *group_info)
1215 int base, max, stride;
1216 int gidsetsize = group_info->ngroups;
1218 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1223 max = gidsetsize - stride;
1224 for (base = 0; base < max; base++) {
1226 int right = left + stride;
1227 gid_t tmp = GROUP_AT(group_info, right);
1229 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1230 GROUP_AT(group_info, right) =
1231 GROUP_AT(group_info, left);
1235 GROUP_AT(group_info, right) = tmp;
1241 /* a simple bsearch */
1242 static int groups_search(struct group_info *group_info, gid_t grp)
1250 right = group_info->ngroups;
1251 while (left < right) {
1252 int mid = (left+right)/2;
1253 int cmp = grp - GROUP_AT(group_info, mid);
1264 /* validate and set current->group_info */
1265 int set_current_groups(struct group_info *group_info)
1268 struct group_info *old_info;
1270 retval = security_task_setgroups(group_info);
1274 groups_sort(group_info);
1275 get_group_info(group_info);
1278 old_info = current->group_info;
1279 current->group_info = group_info;
1280 task_unlock(current);
1282 put_group_info(old_info);
1287 EXPORT_SYMBOL(set_current_groups);
1289 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1294 * SMP: Nobody else can change our grouplist. Thus we are
1301 /* no need to grab task_lock here; it cannot change */
1302 get_group_info(current->group_info);
1303 i = current->group_info->ngroups;
1305 if (i > gidsetsize) {
1309 if (groups_to_user(grouplist, current->group_info)) {
1315 put_group_info(current->group_info);
1320 * SMP: Our groups are copy-on-write. We can set them safely
1321 * without another task interfering.
1324 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1326 struct group_info *group_info;
1329 if (!capable(CAP_SETGID))
1331 if ((unsigned)gidsetsize > NGROUPS_MAX)
1334 group_info = groups_alloc(gidsetsize);
1337 retval = groups_from_user(group_info, grouplist);
1339 put_group_info(group_info);
1343 retval = set_current_groups(group_info);
1344 put_group_info(group_info);
1350 * Check whether we're fsgid/egid or in the supplemental group..
1352 int in_group_p(gid_t grp)
1355 if (grp != current->fsgid) {
1356 get_group_info(current->group_info);
1357 retval = groups_search(current->group_info, grp);
1358 put_group_info(current->group_info);
1363 EXPORT_SYMBOL(in_group_p);
1365 int in_egroup_p(gid_t grp)
1368 if (grp != current->egid) {
1369 get_group_info(current->group_info);
1370 retval = groups_search(current->group_info, grp);
1371 put_group_info(current->group_info);
1376 EXPORT_SYMBOL(in_egroup_p);
1378 DECLARE_RWSEM(uts_sem);
1380 EXPORT_SYMBOL(uts_sem);
1382 asmlinkage long sys_newuname(struct new_utsname __user * name)
1386 down_read(&uts_sem);
1387 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1393 asmlinkage long sys_sethostname(char __user *name, int len)
1396 char tmp[__NEW_UTS_LEN];
1398 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1400 if (len < 0 || len > __NEW_UTS_LEN)
1402 down_write(&uts_sem);
1404 if (!copy_from_user(tmp, name, len)) {
1405 char *ptr = vx_new_uts(nodename);
1407 memcpy(ptr, tmp, len);
1415 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1417 asmlinkage long sys_gethostname(char __user *name, int len)
1424 down_read(&uts_sem);
1425 ptr = vx_new_uts(nodename);
1426 i = 1 + strlen(ptr);
1430 if (copy_to_user(name, ptr, i))
1439 * Only setdomainname; getdomainname can be implemented by calling
1442 asmlinkage long sys_setdomainname(char __user *name, int len)
1445 char tmp[__NEW_UTS_LEN];
1447 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1449 if (len < 0 || len > __NEW_UTS_LEN)
1452 down_write(&uts_sem);
1454 if (!copy_from_user(tmp, name, len)) {
1455 char *ptr = vx_new_uts(domainname);
1457 memcpy(ptr, tmp, len);
1465 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1467 if (resource >= RLIM_NLIMITS)
1470 struct rlimit value;
1471 task_lock(current->group_leader);
1472 value = current->signal->rlim[resource];
1473 task_unlock(current->group_leader);
1474 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1478 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1481 * Back compatibility for getrlimit. Needed for some apps.
1484 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1487 if (resource >= RLIM_NLIMITS)
1490 task_lock(current->group_leader);
1491 x = current->signal->rlim[resource];
1492 task_unlock(current->group_leader);
1493 if(x.rlim_cur > 0x7FFFFFFF)
1494 x.rlim_cur = 0x7FFFFFFF;
1495 if(x.rlim_max > 0x7FFFFFFF)
1496 x.rlim_max = 0x7FFFFFFF;
1497 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1502 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1504 struct rlimit new_rlim, *old_rlim;
1507 if (resource >= RLIM_NLIMITS)
1509 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1511 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1513 old_rlim = current->signal->rlim + resource;
1514 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1515 !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1517 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1520 retval = security_task_setrlimit(resource, &new_rlim);
1524 task_lock(current->group_leader);
1525 *old_rlim = new_rlim;
1526 task_unlock(current->group_leader);
1528 if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1529 (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1530 new_rlim.rlim_cur <= cputime_to_secs(
1531 current->signal->it_prof_expires))) {
1532 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1533 read_lock(&tasklist_lock);
1534 spin_lock_irq(¤t->sighand->siglock);
1535 set_process_cpu_timer(current, CPUCLOCK_PROF,
1537 spin_unlock_irq(¤t->sighand->siglock);
1538 read_unlock(&tasklist_lock);
1545 * It would make sense to put struct rusage in the task_struct,
1546 * except that would make the task_struct be *really big*. After
1547 * task_struct gets moved into malloc'ed memory, it would
1548 * make sense to do this. It will make moving the rest of the information
1549 * a lot simpler! (Which we're not doing right now because we're not
1550 * measuring them yet).
1552 * This expects to be called with tasklist_lock read-locked or better,
1553 * and the siglock not locked. It may momentarily take the siglock.
1555 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1556 * races with threads incrementing their own counters. But since word
1557 * reads are atomic, we either get new values or old values and we don't
1558 * care which for the sums. We always take the siglock to protect reading
1559 * the c* fields from p->signal from races with exit.c updating those
1560 * fields when reaping, so a sample either gets all the additions of a
1561 * given child after it's reaped, or none so this sample is before reaping.
1564 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1566 struct task_struct *t;
1567 unsigned long flags;
1568 cputime_t utime, stime;
1570 memset((char *) r, 0, sizeof *r);
1572 if (unlikely(!p->signal))
1576 case RUSAGE_CHILDREN:
1577 spin_lock_irqsave(&p->sighand->siglock, flags);
1578 utime = p->signal->cutime;
1579 stime = p->signal->cstime;
1580 r->ru_nvcsw = p->signal->cnvcsw;
1581 r->ru_nivcsw = p->signal->cnivcsw;
1582 r->ru_minflt = p->signal->cmin_flt;
1583 r->ru_majflt = p->signal->cmaj_flt;
1584 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1585 cputime_to_timeval(utime, &r->ru_utime);
1586 cputime_to_timeval(stime, &r->ru_stime);
1589 spin_lock_irqsave(&p->sighand->siglock, flags);
1590 utime = stime = cputime_zero;
1593 spin_lock_irqsave(&p->sighand->siglock, flags);
1594 utime = p->signal->cutime;
1595 stime = p->signal->cstime;
1596 r->ru_nvcsw = p->signal->cnvcsw;
1597 r->ru_nivcsw = p->signal->cnivcsw;
1598 r->ru_minflt = p->signal->cmin_flt;
1599 r->ru_majflt = p->signal->cmaj_flt;
1601 utime = cputime_add(utime, p->signal->utime);
1602 stime = cputime_add(stime, p->signal->stime);
1603 r->ru_nvcsw += p->signal->nvcsw;
1604 r->ru_nivcsw += p->signal->nivcsw;
1605 r->ru_minflt += p->signal->min_flt;
1606 r->ru_majflt += p->signal->maj_flt;
1609 utime = cputime_add(utime, t->utime);
1610 stime = cputime_add(stime, t->stime);
1611 r->ru_nvcsw += t->nvcsw;
1612 r->ru_nivcsw += t->nivcsw;
1613 r->ru_minflt += t->min_flt;
1614 r->ru_majflt += t->maj_flt;
1617 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1618 cputime_to_timeval(utime, &r->ru_utime);
1619 cputime_to_timeval(stime, &r->ru_stime);
1626 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1629 read_lock(&tasklist_lock);
1630 k_getrusage(p, who, &r);
1631 read_unlock(&tasklist_lock);
1632 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1635 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1637 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1639 return getrusage(current, who, ru);
1642 asmlinkage long sys_umask(int mask)
1644 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1648 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1649 unsigned long arg4, unsigned long arg5)
1654 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1659 case PR_SET_PDEATHSIG:
1661 if (!valid_signal(sig)) {
1665 current->pdeath_signal = sig;
1667 case PR_GET_PDEATHSIG:
1668 error = put_user(current->pdeath_signal, (int __user *)arg2);
1670 case PR_GET_DUMPABLE:
1671 if (current->mm->dumpable)
1674 case PR_SET_DUMPABLE:
1675 if (arg2 != 0 && arg2 != 1) {
1679 current->mm->dumpable = arg2;
1682 case PR_SET_UNALIGN:
1683 error = SET_UNALIGN_CTL(current, arg2);
1685 case PR_GET_UNALIGN:
1686 error = GET_UNALIGN_CTL(current, arg2);
1689 error = SET_FPEMU_CTL(current, arg2);
1692 error = GET_FPEMU_CTL(current, arg2);
1695 error = SET_FPEXC_CTL(current, arg2);
1698 error = GET_FPEXC_CTL(current, arg2);
1701 error = PR_TIMING_STATISTICAL;
1704 if (arg2 == PR_TIMING_STATISTICAL)
1710 case PR_GET_KEEPCAPS:
1711 if (current->keep_capabilities)
1714 case PR_SET_KEEPCAPS:
1715 if (arg2 != 0 && arg2 != 1) {
1719 current->keep_capabilities = arg2;
1722 struct task_struct *me = current;
1723 unsigned char ncomm[sizeof(me->comm)];
1725 ncomm[sizeof(me->comm)-1] = 0;
1726 if (strncpy_from_user(ncomm, (char __user *)arg2,
1727 sizeof(me->comm)-1) < 0)
1729 set_task_comm(me, ncomm);
1733 struct task_struct *me = current;
1734 unsigned char tcomm[sizeof(me->comm)];
1736 get_task_comm(tcomm, me);
1737 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))