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/security.h>
25 #include <linux/dcookies.h>
26 #include <linux/suspend.h>
27 #include <linux/ckrm.h>
28 #include <linux/vs_base.h>
29 #include <linux/vs_cvirt.h>
30 #include <linux/tty.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 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
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);
223 static int set_one_prio(struct task_struct *p, int niceval, int error)
227 if (p->uid != current->euid &&
228 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
232 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
233 if (vx_flags(VXF_IGNEG_NICE, 0))
239 no_nice = security_task_setnice(p, niceval);
246 set_user_nice(p, niceval);
251 asmlinkage long sys_setpriority(int which, int who, int niceval)
253 struct task_struct *g, *p;
254 struct user_struct *user;
257 if (which > 2 || which < 0)
260 /* normalize: avoid signed division (rounding problems) */
267 read_lock(&tasklist_lock);
272 p = find_task_by_pid(who);
274 error = set_one_prio(p, niceval, error);
278 who = process_group(current);
279 do_each_task_pid(who, PIDTYPE_PGID, p) {
280 error = set_one_prio(p, niceval, error);
281 } while_each_task_pid(who, PIDTYPE_PGID, p);
284 user = current->user;
288 if ((who != current->uid) &&
289 !(user = find_user(vx_current_xid(), who)))
290 goto out_unlock; /* No processes for this user */
294 error = set_one_prio(p, niceval, error);
295 while_each_thread(g, p);
296 if (who != current->uid)
297 free_uid(user); /* For find_user() */
301 read_unlock(&tasklist_lock);
307 * Ugh. To avoid negative return values, "getpriority()" will
308 * not return the normal nice-value, but a negated value that
309 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
310 * to stay compatible.
312 asmlinkage long sys_getpriority(int which, int who)
314 struct task_struct *g, *p;
315 struct user_struct *user;
316 long niceval, retval = -ESRCH;
318 if (which > 2 || which < 0)
321 read_lock(&tasklist_lock);
326 p = find_task_by_pid(who);
328 niceval = 20 - task_nice(p);
329 if (niceval > retval)
335 who = process_group(current);
336 do_each_task_pid(who, PIDTYPE_PGID, p) {
337 niceval = 20 - task_nice(p);
338 if (niceval > retval)
340 } while_each_task_pid(who, PIDTYPE_PGID, p);
343 user = current->user;
347 if ((who != current->uid) &&
348 !(user = find_user(vx_current_xid(), who)))
349 goto out_unlock; /* No processes for this user */
353 niceval = 20 - task_nice(p);
354 if (niceval > retval)
357 while_each_thread(g, p);
358 if (who != current->uid)
359 free_uid(user); /* for find_user() */
363 read_unlock(&tasklist_lock);
368 long vs_reboot(unsigned int, void *);
371 * Reboot system call: for obvious reasons only root may call it,
372 * and even root needs to set up some magic numbers in the registers
373 * so that some mistake won't make this reboot the whole machine.
374 * You can also set the meaning of the ctrl-alt-del-key here.
376 * reboot doesn't sync: do that yourself before calling this.
378 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
382 /* We only trust the superuser with rebooting the system. */
383 if (!capable(CAP_SYS_BOOT))
386 /* For safety, we require "magic" arguments. */
387 if (magic1 != LINUX_REBOOT_MAGIC1 ||
388 (magic2 != LINUX_REBOOT_MAGIC2 &&
389 magic2 != LINUX_REBOOT_MAGIC2A &&
390 magic2 != LINUX_REBOOT_MAGIC2B &&
391 magic2 != LINUX_REBOOT_MAGIC2C))
394 if (!vx_check(0, VX_ADMIN|VX_WATCH))
395 return vs_reboot(cmd, arg);
399 case LINUX_REBOOT_CMD_RESTART:
400 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
401 system_state = SYSTEM_RESTART;
403 printk(KERN_EMERG "Restarting system.\n");
404 machine_restart(NULL);
407 case LINUX_REBOOT_CMD_CAD_ON:
411 case LINUX_REBOOT_CMD_CAD_OFF:
415 case LINUX_REBOOT_CMD_HALT:
416 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
417 system_state = SYSTEM_HALT;
419 printk(KERN_EMERG "System halted.\n");
425 case LINUX_REBOOT_CMD_POWER_OFF:
426 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
427 system_state = SYSTEM_POWER_OFF;
429 printk(KERN_EMERG "Power down.\n");
435 case LINUX_REBOOT_CMD_RESTART2:
436 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
440 buffer[sizeof(buffer) - 1] = '\0';
442 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
443 system_state = SYSTEM_RESTART;
445 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
446 machine_restart(buffer);
450 case LINUX_REBOOT_CMD_KEXEC:
452 struct kimage *image;
453 image = xchg(&kexec_image, 0);
458 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
459 system_state = SYSTEM_RESTART;
461 system_state = SYSTEM_BOOTING;
462 printk(KERN_EMERG "Starting new kernel\n");
464 machine_kexec(image);
469 #ifdef CONFIG_SOFTWARE_SUSPEND
470 case LINUX_REBOOT_CMD_SW_SUSPEND:
472 int ret = software_suspend();
486 static void deferred_cad(void *dummy)
488 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
489 machine_restart(NULL);
493 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
494 * As it's called within an interrupt, it may NOT sync: the only choice
495 * is whether to reboot at once, or just ignore the ctrl-alt-del.
497 void ctrl_alt_del(void)
499 static DECLARE_WORK(cad_work, deferred_cad, NULL);
502 schedule_work(&cad_work);
504 kill_proc(cad_pid, SIGINT, 1);
509 * Unprivileged users may change the real gid to the effective gid
510 * or vice versa. (BSD-style)
512 * If you set the real gid at all, or set the effective gid to a value not
513 * equal to the real gid, then the saved gid is set to the new effective gid.
515 * This makes it possible for a setgid program to completely drop its
516 * privileges, which is often a useful assertion to make when you are doing
517 * a security audit over a program.
519 * The general idea is that a program which uses just setregid() will be
520 * 100% compatible with BSD. A program which uses just setgid() will be
521 * 100% compatible with POSIX with saved IDs.
523 * SMP: There are not races, the GIDs are checked only by filesystem
524 * operations (as far as semantic preservation is concerned).
526 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
528 int old_rgid = current->gid;
529 int old_egid = current->egid;
530 int new_rgid = old_rgid;
531 int new_egid = old_egid;
534 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
538 if (rgid != (gid_t) -1) {
539 if ((old_rgid == rgid) ||
540 (current->egid==rgid) ||
546 if (egid != (gid_t) -1) {
547 if ((old_rgid == egid) ||
548 (current->egid == egid) ||
549 (current->sgid == egid) ||
556 if (new_egid != old_egid)
558 current->mm->dumpable = suid_dumpable;
561 if (rgid != (gid_t) -1 ||
562 (egid != (gid_t) -1 && egid != old_rgid))
563 current->sgid = new_egid;
564 current->fsgid = new_egid;
565 current->egid = new_egid;
566 current->gid = new_rgid;
570 key_fsgid_changed(current);
575 * setgid() is implemented like SysV w/ SAVED_IDS
577 * SMP: Same implicit races as above.
579 asmlinkage long sys_setgid(gid_t gid)
581 int old_egid = current->egid;
584 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
588 if (capable(CAP_SETGID))
592 current->mm->dumpable = suid_dumpable;
595 current->gid = current->egid = current->sgid = current->fsgid = gid;
597 else if ((gid == current->gid) || (gid == current->sgid))
601 current->mm->dumpable = suid_dumpable;
604 current->egid = current->fsgid = gid;
611 key_fsgid_changed(current);
615 static int set_user(uid_t new_ruid, int dumpclear)
617 struct user_struct *new_user;
619 new_user = alloc_uid(vx_current_xid(), new_ruid);
623 if (atomic_read(&new_user->processes) >=
624 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
625 new_user != &root_user) {
630 switch_uid(new_user);
634 current->mm->dumpable = suid_dumpable;
637 current->uid = new_ruid;
642 * Unprivileged users may change the real uid to the effective uid
643 * or vice versa. (BSD-style)
645 * If you set the real uid at all, or set the effective uid to a value not
646 * equal to the real uid, then the saved uid is set to the new effective uid.
648 * This makes it possible for a setuid program to completely drop its
649 * privileges, which is often a useful assertion to make when you are doing
650 * a security audit over a program.
652 * The general idea is that a program which uses just setreuid() will be
653 * 100% compatible with BSD. A program which uses just setuid() will be
654 * 100% compatible with POSIX with saved IDs.
656 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
658 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
661 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
665 new_ruid = old_ruid = current->uid;
666 new_euid = old_euid = current->euid;
667 old_suid = current->suid;
669 if (ruid != (uid_t) -1) {
671 if ((old_ruid != ruid) &&
672 (current->euid != ruid) &&
673 !capable(CAP_SETUID))
677 if (euid != (uid_t) -1) {
679 if ((old_ruid != euid) &&
680 (current->euid != euid) &&
681 (current->suid != euid) &&
682 !capable(CAP_SETUID))
686 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
689 if (new_euid != old_euid)
691 current->mm->dumpable = suid_dumpable;
694 current->fsuid = current->euid = new_euid;
695 if (ruid != (uid_t) -1 ||
696 (euid != (uid_t) -1 && euid != old_ruid))
697 current->suid = current->euid;
698 current->fsuid = current->euid;
702 key_fsuid_changed(current);
704 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
710 * setuid() is implemented like SysV with SAVED_IDS
712 * Note that SAVED_ID's is deficient in that a setuid root program
713 * like sendmail, for example, cannot set its uid to be a normal
714 * user and then switch back, because if you're root, setuid() sets
715 * the saved uid too. If you don't like this, blame the bright people
716 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
717 * will allow a root program to temporarily drop privileges and be able to
718 * regain them by swapping the real and effective uid.
720 asmlinkage long sys_setuid(uid_t uid)
722 int old_euid = current->euid;
723 int old_ruid, old_suid, new_ruid, new_suid;
726 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
730 old_ruid = new_ruid = current->uid;
731 old_suid = current->suid;
734 if (capable(CAP_SETUID)) {
735 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
738 } else if ((uid != current->uid) && (uid != new_suid))
743 current->mm->dumpable = suid_dumpable;
746 current->fsuid = current->euid = uid;
747 current->suid = new_suid;
751 key_fsuid_changed(current);
753 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
758 * This function implements a generic ability to update ruid, euid,
759 * and suid. This allows you to implement the 4.4 compatible seteuid().
761 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
763 int old_ruid = current->uid;
764 int old_euid = current->euid;
765 int old_suid = current->suid;
768 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
772 if (!capable(CAP_SETUID)) {
773 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
774 (ruid != current->euid) && (ruid != current->suid))
776 if ((euid != (uid_t) -1) && (euid != current->uid) &&
777 (euid != current->euid) && (euid != current->suid))
779 if ((suid != (uid_t) -1) && (suid != current->uid) &&
780 (suid != current->euid) && (suid != current->suid))
783 if (ruid != (uid_t) -1) {
784 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
787 if (euid != (uid_t) -1) {
788 if (euid != current->euid)
790 current->mm->dumpable = suid_dumpable;
793 current->euid = euid;
795 current->fsuid = current->euid;
796 if (suid != (uid_t) -1)
797 current->suid = suid;
801 key_fsuid_changed(current);
803 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
806 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
810 if (!(retval = put_user(current->uid, ruid)) &&
811 !(retval = put_user(current->euid, euid)))
812 retval = put_user(current->suid, suid);
818 * Same as above, but for rgid, egid, sgid.
820 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
824 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
828 if (!capable(CAP_SETGID)) {
829 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
830 (rgid != current->egid) && (rgid != current->sgid))
832 if ((egid != (gid_t) -1) && (egid != current->gid) &&
833 (egid != current->egid) && (egid != current->sgid))
835 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
836 (sgid != current->egid) && (sgid != current->sgid))
839 if (egid != (gid_t) -1) {
840 if (egid != current->egid)
842 current->mm->dumpable = suid_dumpable;
845 current->egid = egid;
847 current->fsgid = current->egid;
848 if (rgid != (gid_t) -1)
850 if (sgid != (gid_t) -1)
851 current->sgid = sgid;
856 key_fsgid_changed(current);
860 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
864 if (!(retval = put_user(current->gid, rgid)) &&
865 !(retval = put_user(current->egid, egid)))
866 retval = put_user(current->sgid, sgid);
873 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
874 * is used for "access()" and for the NFS daemon (letting nfsd stay at
875 * whatever uid it wants to). It normally shadows "euid", except when
876 * explicitly set by setfsuid() or for access..
878 asmlinkage long sys_setfsuid(uid_t uid)
882 old_fsuid = current->fsuid;
883 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
886 if (uid == current->uid || uid == current->euid ||
887 uid == current->suid || uid == current->fsuid ||
890 if (uid != old_fsuid)
892 current->mm->dumpable = suid_dumpable;
895 current->fsuid = uid;
898 key_fsuid_changed(current);
900 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
906 * Samma på svenska..
908 asmlinkage long sys_setfsgid(gid_t gid)
912 old_fsgid = current->fsgid;
913 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
916 if (gid == current->gid || gid == current->egid ||
917 gid == current->sgid || gid == current->fsgid ||
920 if (gid != old_fsgid)
922 current->mm->dumpable = suid_dumpable;
925 current->fsgid = gid;
926 key_fsgid_changed(current);
931 asmlinkage long sys_times(struct tms __user * tbuf)
934 * In the SMP world we might just be unlucky and have one of
935 * the times increment as we use it. Since the value is an
936 * atomically safe type this is just fine. Conceptually its
937 * as if the syscall took an instant longer to occur.
941 struct task_struct *tsk = current;
942 struct task_struct *t;
943 unsigned long utime, stime, cutime, cstime;
945 read_lock(&tasklist_lock);
946 utime = tsk->signal->utime;
947 stime = tsk->signal->stime;
956 * While we have tasklist_lock read-locked, no dying thread
957 * can be updating current->signal->[us]time. Instead,
958 * we got their counts included in the live thread loop.
959 * However, another thread can come in right now and
960 * do a wait call that updates current->signal->c[us]time.
961 * To make sure we always see that pair updated atomically,
962 * we take the siglock around fetching them.
964 spin_lock_irq(&tsk->sighand->siglock);
965 cutime = tsk->signal->cutime;
966 cstime = tsk->signal->cstime;
967 spin_unlock_irq(&tsk->sighand->siglock);
968 read_unlock(&tasklist_lock);
970 tmp.tms_utime = jiffies_to_clock_t(utime);
971 tmp.tms_stime = jiffies_to_clock_t(stime);
972 tmp.tms_cutime = jiffies_to_clock_t(cutime);
973 tmp.tms_cstime = jiffies_to_clock_t(cstime);
974 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
977 return (long) jiffies_64_to_clock_t(get_jiffies_64());
981 * This needs some heavy checking ...
982 * I just haven't the stomach for it. I also don't fully
983 * understand sessions/pgrp etc. Let somebody who does explain it.
985 * OK, I think I have the protection semantics right.... this is really
986 * only important on a multi-user system anyway, to make sure one user
987 * can't send a signal to a process owned by another. -TYT, 12/12/91
989 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
993 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
995 struct task_struct *p;
1000 pid = vx_map_pid(current->pid);
1006 rpgid = vx_rmap_pid(pgid);
1008 /* From this point forward we keep holding onto the tasklist lock
1009 * so that our parent does not change from under us. -DaveM
1011 write_lock_irq(&tasklist_lock);
1014 p = find_task_by_pid(pid);
1019 if (!thread_group_leader(p))
1022 if (p->parent == current || p->real_parent == current) {
1024 if (p->signal->session != current->signal->session)
1036 if (p->signal->leader)
1040 struct task_struct *p;
1042 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1043 if (p->signal->session == current->signal->session)
1045 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1050 err = security_task_setpgid(p, rpgid);
1054 if (process_group(p) != rpgid) {
1055 detach_pid(p, PIDTYPE_PGID);
1056 p->signal->pgrp = rpgid;
1057 attach_pid(p, PIDTYPE_PGID, rpgid);
1062 /* All paths lead to here, thus we are safe. -DaveM */
1063 write_unlock_irq(&tasklist_lock);
1067 asmlinkage long sys_getpgid(pid_t pid)
1070 return vx_rmap_pid(process_group(current));
1073 struct task_struct *p;
1075 read_lock(&tasklist_lock);
1076 p = find_task_by_pid(pid);
1080 retval = security_task_getpgid(p);
1082 retval = vx_rmap_pid(process_group(p));
1084 read_unlock(&tasklist_lock);
1089 #ifdef __ARCH_WANT_SYS_GETPGRP
1091 asmlinkage long sys_getpgrp(void)
1093 /* SMP - assuming writes are word atomic this is fine */
1094 return process_group(current);
1099 asmlinkage long sys_getsid(pid_t pid)
1102 return current->signal->session;
1105 struct task_struct *p;
1107 read_lock(&tasklist_lock);
1108 p = find_task_by_pid(pid);
1112 retval = security_task_getsid(p);
1114 retval = p->signal->session;
1116 read_unlock(&tasklist_lock);
1121 asmlinkage long sys_setsid(void)
1126 if (!thread_group_leader(current))
1130 write_lock_irq(&tasklist_lock);
1132 pid = find_pid(PIDTYPE_PGID, current->pid);
1136 current->signal->leader = 1;
1137 __set_special_pids(current->pid, current->pid);
1138 current->signal->tty = NULL;
1139 current->signal->tty_old_pgrp = 0;
1140 err = process_group(current);
1142 write_unlock_irq(&tasklist_lock);
1148 * Supplementary group IDs
1151 /* init to 2 - one for init_task, one to ensure it is never freed */
1152 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1154 struct group_info *groups_alloc(int gidsetsize)
1156 struct group_info *group_info;
1160 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1161 /* Make sure we always allocate at least one indirect block pointer */
1162 nblocks = nblocks ? : 1;
1163 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1166 group_info->ngroups = gidsetsize;
1167 group_info->nblocks = nblocks;
1168 atomic_set(&group_info->usage, 1);
1170 if (gidsetsize <= NGROUPS_SMALL) {
1171 group_info->blocks[0] = group_info->small_block;
1173 for (i = 0; i < nblocks; i++) {
1175 b = (void *)__get_free_page(GFP_USER);
1177 goto out_undo_partial_alloc;
1178 group_info->blocks[i] = b;
1183 out_undo_partial_alloc:
1185 free_page((unsigned long)group_info->blocks[i]);
1191 EXPORT_SYMBOL(groups_alloc);
1193 void groups_free(struct group_info *group_info)
1195 if (group_info->blocks[0] != group_info->small_block) {
1197 for (i = 0; i < group_info->nblocks; i++)
1198 free_page((unsigned long)group_info->blocks[i]);
1203 EXPORT_SYMBOL(groups_free);
1205 /* export the group_info to a user-space array */
1206 static int groups_to_user(gid_t __user *grouplist,
1207 struct group_info *group_info)
1210 int count = group_info->ngroups;
1212 for (i = 0; i < group_info->nblocks; i++) {
1213 int cp_count = min(NGROUPS_PER_BLOCK, count);
1214 int off = i * NGROUPS_PER_BLOCK;
1215 int len = cp_count * sizeof(*grouplist);
1217 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1225 /* fill a group_info from a user-space array - it must be allocated already */
1226 static int groups_from_user(struct group_info *group_info,
1227 gid_t __user *grouplist)
1230 int count = group_info->ngroups;
1232 for (i = 0; i < group_info->nblocks; i++) {
1233 int cp_count = min(NGROUPS_PER_BLOCK, count);
1234 int off = i * NGROUPS_PER_BLOCK;
1235 int len = cp_count * sizeof(*grouplist);
1237 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1245 /* a simple shell-metzner sort */
1246 static void groups_sort(struct group_info *group_info)
1248 int base, max, stride;
1249 int gidsetsize = group_info->ngroups;
1251 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1256 max = gidsetsize - stride;
1257 for (base = 0; base < max; base++) {
1259 int right = left + stride;
1260 gid_t tmp = GROUP_AT(group_info, right);
1262 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1263 GROUP_AT(group_info, right) =
1264 GROUP_AT(group_info, left);
1268 GROUP_AT(group_info, right) = tmp;
1274 /* a simple bsearch */
1275 static int groups_search(struct group_info *group_info, gid_t grp)
1283 right = group_info->ngroups;
1284 while (left < right) {
1285 int mid = (left+right)/2;
1286 int cmp = grp - GROUP_AT(group_info, mid);
1297 /* validate and set current->group_info */
1298 int set_current_groups(struct group_info *group_info)
1301 struct group_info *old_info;
1303 retval = security_task_setgroups(group_info);
1307 groups_sort(group_info);
1308 get_group_info(group_info);
1311 old_info = current->group_info;
1312 current->group_info = group_info;
1313 task_unlock(current);
1315 put_group_info(old_info);
1320 EXPORT_SYMBOL(set_current_groups);
1322 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1327 * SMP: Nobody else can change our grouplist. Thus we are
1334 /* no need to grab task_lock here; it cannot change */
1335 get_group_info(current->group_info);
1336 i = current->group_info->ngroups;
1338 if (i > gidsetsize) {
1342 if (groups_to_user(grouplist, current->group_info)) {
1348 put_group_info(current->group_info);
1353 * SMP: Our groups are copy-on-write. We can set them safely
1354 * without another task interfering.
1357 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1359 struct group_info *group_info;
1362 if (!capable(CAP_SETGID))
1364 if ((unsigned)gidsetsize > NGROUPS_MAX)
1367 group_info = groups_alloc(gidsetsize);
1370 retval = groups_from_user(group_info, grouplist);
1372 put_group_info(group_info);
1376 retval = set_current_groups(group_info);
1377 put_group_info(group_info);
1383 * Check whether we're fsgid/egid or in the supplemental group..
1385 int in_group_p(gid_t grp)
1388 if (grp != current->fsgid) {
1389 get_group_info(current->group_info);
1390 retval = groups_search(current->group_info, grp);
1391 put_group_info(current->group_info);
1396 EXPORT_SYMBOL(in_group_p);
1398 int in_egroup_p(gid_t grp)
1401 if (grp != current->egid) {
1402 get_group_info(current->group_info);
1403 retval = groups_search(current->group_info, grp);
1404 put_group_info(current->group_info);
1409 EXPORT_SYMBOL(in_egroup_p);
1411 DECLARE_RWSEM(uts_sem);
1413 EXPORT_SYMBOL(uts_sem);
1415 asmlinkage long sys_newuname(struct new_utsname __user * name)
1419 down_read(&uts_sem);
1420 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1426 asmlinkage long sys_sethostname(char __user *name, int len)
1429 char tmp[__NEW_UTS_LEN];
1431 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1433 if (len < 0 || len > __NEW_UTS_LEN)
1435 down_write(&uts_sem);
1437 if (!copy_from_user(tmp, name, len)) {
1438 char *ptr = vx_new_uts(nodename);
1440 memcpy(ptr, tmp, len);
1448 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1450 asmlinkage long sys_gethostname(char __user *name, int len)
1457 down_read(&uts_sem);
1458 ptr = vx_new_uts(nodename);
1459 i = 1 + strlen(ptr);
1463 if (copy_to_user(name, ptr, i))
1472 * Only setdomainname; getdomainname can be implemented by calling
1475 asmlinkage long sys_setdomainname(char __user *name, int len)
1478 char tmp[__NEW_UTS_LEN];
1480 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1482 if (len < 0 || len > __NEW_UTS_LEN)
1485 down_write(&uts_sem);
1487 if (!copy_from_user(tmp, name, len)) {
1488 char *ptr = vx_new_uts(domainname);
1490 memcpy(ptr, tmp, len);
1498 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1500 if (resource >= RLIM_NLIMITS)
1503 struct rlimit value;
1504 task_lock(current->group_leader);
1505 value = current->signal->rlim[resource];
1506 task_unlock(current->group_leader);
1507 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1511 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1514 * Back compatibility for getrlimit. Needed for some apps.
1517 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1520 if (resource >= RLIM_NLIMITS)
1523 task_lock(current->group_leader);
1524 x = current->signal->rlim[resource];
1525 task_unlock(current->group_leader);
1526 if(x.rlim_cur > 0x7FFFFFFF)
1527 x.rlim_cur = 0x7FFFFFFF;
1528 if(x.rlim_max > 0x7FFFFFFF)
1529 x.rlim_max = 0x7FFFFFFF;
1530 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1535 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1537 struct rlimit new_rlim, *old_rlim;
1540 if (resource >= RLIM_NLIMITS)
1542 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1544 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1546 old_rlim = current->signal->rlim + resource;
1547 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1548 !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1550 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1553 retval = security_task_setrlimit(resource, &new_rlim);
1557 task_lock(current->group_leader);
1558 *old_rlim = new_rlim;
1559 task_unlock(current->group_leader);
1564 * It would make sense to put struct rusage in the task_struct,
1565 * except that would make the task_struct be *really big*. After
1566 * task_struct gets moved into malloc'ed memory, it would
1567 * make sense to do this. It will make moving the rest of the information
1568 * a lot simpler! (Which we're not doing right now because we're not
1569 * measuring them yet).
1571 * This expects to be called with tasklist_lock read-locked or better,
1572 * and the siglock not locked. It may momentarily take the siglock.
1574 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1575 * races with threads incrementing their own counters. But since word
1576 * reads are atomic, we either get new values or old values and we don't
1577 * care which for the sums. We always take the siglock to protect reading
1578 * the c* fields from p->signal from races with exit.c updating those
1579 * fields when reaping, so a sample either gets all the additions of a
1580 * given child after it's reaped, or none so this sample is before reaping.
1583 void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1585 struct task_struct *t;
1586 unsigned long flags;
1587 unsigned long utime, stime;
1589 memset((char *) r, 0, sizeof *r);
1591 if (unlikely(!p->signal))
1595 case RUSAGE_CHILDREN:
1596 spin_lock_irqsave(&p->sighand->siglock, flags);
1597 utime = p->signal->cutime;
1598 stime = p->signal->cstime;
1599 r->ru_nvcsw = p->signal->cnvcsw;
1600 r->ru_nivcsw = p->signal->cnivcsw;
1601 r->ru_minflt = p->signal->cmin_flt;
1602 r->ru_majflt = p->signal->cmaj_flt;
1603 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1604 jiffies_to_timeval(utime, &r->ru_utime);
1605 jiffies_to_timeval(stime, &r->ru_stime);
1608 spin_lock_irqsave(&p->sighand->siglock, flags);
1612 spin_lock_irqsave(&p->sighand->siglock, flags);
1613 utime = p->signal->cutime;
1614 stime = p->signal->cstime;
1615 r->ru_nvcsw = p->signal->cnvcsw;
1616 r->ru_nivcsw = p->signal->cnivcsw;
1617 r->ru_minflt = p->signal->cmin_flt;
1618 r->ru_majflt = p->signal->cmaj_flt;
1620 utime += p->signal->utime;
1621 stime += p->signal->stime;
1622 r->ru_nvcsw += p->signal->nvcsw;
1623 r->ru_nivcsw += p->signal->nivcsw;
1624 r->ru_minflt += p->signal->min_flt;
1625 r->ru_majflt += p->signal->maj_flt;
1630 r->ru_nvcsw += t->nvcsw;
1631 r->ru_nivcsw += t->nivcsw;
1632 r->ru_minflt += t->min_flt;
1633 r->ru_majflt += t->maj_flt;
1636 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1637 jiffies_to_timeval(utime, &r->ru_utime);
1638 jiffies_to_timeval(stime, &r->ru_stime);
1645 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1648 read_lock(&tasklist_lock);
1649 k_getrusage(p, who, &r);
1650 read_unlock(&tasklist_lock);
1651 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1654 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1656 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1658 return getrusage(current, who, ru);
1661 asmlinkage long sys_umask(int mask)
1663 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1667 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1668 unsigned long arg4, unsigned long arg5)
1673 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1678 case PR_SET_PDEATHSIG:
1680 if (sig < 0 || sig > _NSIG) {
1684 current->pdeath_signal = sig;
1686 case PR_GET_PDEATHSIG:
1687 error = put_user(current->pdeath_signal, (int __user *)arg2);
1689 case PR_GET_DUMPABLE:
1690 if (current->mm->dumpable)
1693 case PR_SET_DUMPABLE:
1694 if (arg2 < 0 && arg2 > 2) {
1698 current->mm->dumpable = arg2;
1701 case PR_SET_UNALIGN:
1702 error = SET_UNALIGN_CTL(current, arg2);
1704 case PR_GET_UNALIGN:
1705 error = GET_UNALIGN_CTL(current, arg2);
1708 error = SET_FPEMU_CTL(current, arg2);
1711 error = GET_FPEMU_CTL(current, arg2);
1714 error = SET_FPEXC_CTL(current, arg2);
1717 error = GET_FPEXC_CTL(current, arg2);
1720 error = PR_TIMING_STATISTICAL;
1723 if (arg2 == PR_TIMING_STATISTICAL)
1729 case PR_GET_KEEPCAPS:
1730 if (current->keep_capabilities)
1733 case PR_SET_KEEPCAPS:
1734 if (arg2 != 0 && arg2 != 1) {
1738 current->keep_capabilities = arg2;
1741 struct task_struct *me = current;
1742 unsigned char ncomm[sizeof(me->comm)];
1744 ncomm[sizeof(me->comm)-1] = 0;
1745 if (strncpy_from_user(ncomm, (char __user *)arg2,
1746 sizeof(me->comm)-1) < 0)
1748 set_task_comm(me, ncomm);