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/kernel.h>
21 #include <linux/kexec.h>
22 #include <linux/workqueue.h>
23 #include <linux/device.h>
24 #include <linux/key.h>
25 #include <linux/times.h>
26 #include <linux/posix-timers.h>
27 #include <linux/security.h>
28 #include <linux/dcookies.h>
29 #include <linux/suspend.h>
30 #include <linux/tty.h>
31 #include <linux/signal.h>
33 #include <linux/compat.h>
34 #include <linux/syscalls.h>
35 #include <linux/vs_cvirt.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 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 static int set_one_prio(struct task_struct *p, int niceval, int error)
230 if (p->uid != current->euid &&
231 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
235 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
236 if (vx_flags(VXF_IGNEG_NICE, 0))
242 no_nice = security_task_setnice(p, niceval);
249 set_user_nice(p, niceval);
254 asmlinkage long sys_setpriority(int which, int who, int niceval)
256 struct task_struct *g, *p;
257 struct user_struct *user;
260 if (which > 2 || which < 0)
263 /* normalize: avoid signed division (rounding problems) */
270 read_lock(&tasklist_lock);
275 p = find_task_by_pid(who);
277 error = set_one_prio(p, niceval, error);
281 who = process_group(current);
282 do_each_task_pid(who, PIDTYPE_PGID, p) {
283 error = set_one_prio(p, niceval, error);
284 } while_each_task_pid(who, PIDTYPE_PGID, p);
287 user = current->user;
291 if ((who != current->uid) &&
292 !(user = find_user(vx_current_xid(), who)))
293 goto out_unlock; /* No processes for this user */
297 error = set_one_prio(p, niceval, error);
298 while_each_thread(g, p);
299 if (who != current->uid)
300 free_uid(user); /* For find_user() */
304 read_unlock(&tasklist_lock);
310 * Ugh. To avoid negative return values, "getpriority()" will
311 * not return the normal nice-value, but a negated value that
312 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
313 * to stay compatible.
315 asmlinkage long sys_getpriority(int which, int who)
317 struct task_struct *g, *p;
318 struct user_struct *user;
319 long niceval, retval = -ESRCH;
321 if (which > 2 || which < 0)
324 read_lock(&tasklist_lock);
329 p = find_task_by_pid(who);
331 niceval = 20 - task_nice(p);
332 if (niceval > retval)
338 who = process_group(current);
339 do_each_task_pid(who, PIDTYPE_PGID, p) {
340 niceval = 20 - task_nice(p);
341 if (niceval > retval)
343 } while_each_task_pid(who, PIDTYPE_PGID, p);
346 user = current->user;
350 if ((who != current->uid) &&
351 !(user = find_user(vx_current_xid(), who)))
352 goto out_unlock; /* No processes for this user */
356 niceval = 20 - task_nice(p);
357 if (niceval > retval)
360 while_each_thread(g, p);
361 if (who != current->uid)
362 free_uid(user); /* for find_user() */
366 read_unlock(&tasklist_lock);
368 key_fsgid_changed(current);
372 long vs_reboot(unsigned int, void *);
375 * Reboot system call: for obvious reasons only root may call it,
376 * and even root needs to set up some magic numbers in the registers
377 * so that some mistake won't make this reboot the whole machine.
378 * You can also set the meaning of the ctrl-alt-del-key here.
380 * reboot doesn't sync: do that yourself before calling this.
382 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
386 /* We only trust the superuser with rebooting the system. */
387 if (!capable(CAP_SYS_BOOT))
390 /* For safety, we require "magic" arguments. */
391 if (magic1 != LINUX_REBOOT_MAGIC1 ||
392 (magic2 != LINUX_REBOOT_MAGIC2 &&
393 magic2 != LINUX_REBOOT_MAGIC2A &&
394 magic2 != LINUX_REBOOT_MAGIC2B &&
395 magic2 != LINUX_REBOOT_MAGIC2C))
398 if (!vx_check(0, VX_ADMIN|VX_WATCH))
399 return vs_reboot(cmd, arg);
403 case LINUX_REBOOT_CMD_RESTART:
404 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
405 system_state = SYSTEM_RESTART;
407 printk(KERN_EMERG "Restarting system.\n");
408 machine_restart(NULL);
411 case LINUX_REBOOT_CMD_CAD_ON:
415 case LINUX_REBOOT_CMD_CAD_OFF:
419 case LINUX_REBOOT_CMD_HALT:
420 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
421 system_state = SYSTEM_HALT;
423 printk(KERN_EMERG "System halted.\n");
429 case LINUX_REBOOT_CMD_POWER_OFF:
430 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
431 system_state = SYSTEM_POWER_OFF;
433 printk(KERN_EMERG "Power down.\n");
439 case LINUX_REBOOT_CMD_RESTART2:
440 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
444 buffer[sizeof(buffer) - 1] = '\0';
446 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
447 system_state = SYSTEM_RESTART;
449 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
450 machine_restart(buffer);
454 case LINUX_REBOOT_CMD_KEXEC:
456 struct kimage *image;
457 image = xchg(&kexec_image, 0);
462 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
463 system_state = SYSTEM_RESTART;
465 system_state = SYSTEM_BOOTING;
466 printk(KERN_EMERG "Starting new kernel\n");
468 machine_kexec(image);
473 #ifdef CONFIG_SOFTWARE_SUSPEND
474 case LINUX_REBOOT_CMD_SW_SUSPEND:
476 int ret = software_suspend();
490 static void deferred_cad(void *dummy)
492 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
493 machine_restart(NULL);
497 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
498 * As it's called within an interrupt, it may NOT sync: the only choice
499 * is whether to reboot at once, or just ignore the ctrl-alt-del.
501 void ctrl_alt_del(void)
503 static DECLARE_WORK(cad_work, deferred_cad, NULL);
506 schedule_work(&cad_work);
508 kill_proc(cad_pid, SIGINT, 1);
514 * Unprivileged users may change the real gid to the effective gid
515 * or vice versa. (BSD-style)
517 * If you set the real gid at all, or set the effective gid to a value not
518 * equal to the real gid, then the saved gid is set to the new effective gid.
520 * This makes it possible for a setgid program to completely drop its
521 * privileges, which is often a useful assertion to make when you are doing
522 * a security audit over a program.
524 * The general idea is that a program which uses just setregid() will be
525 * 100% compatible with BSD. A program which uses just setgid() will be
526 * 100% compatible with POSIX with saved IDs.
528 * SMP: There are not races, the GIDs are checked only by filesystem
529 * operations (as far as semantic preservation is concerned).
531 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
533 int old_rgid = current->gid;
534 int old_egid = current->egid;
535 int new_rgid = old_rgid;
536 int new_egid = old_egid;
539 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
543 if (rgid != (gid_t) -1) {
544 if ((old_rgid == rgid) ||
545 (current->egid==rgid) ||
551 if (egid != (gid_t) -1) {
552 if ((old_rgid == egid) ||
553 (current->egid == egid) ||
554 (current->sgid == egid) ||
561 if (new_egid != old_egid)
563 current->mm->dumpable = 0;
566 if (rgid != (gid_t) -1 ||
567 (egid != (gid_t) -1 && egid != old_rgid))
568 current->sgid = new_egid;
569 current->fsgid = new_egid;
570 current->egid = new_egid;
571 current->gid = new_rgid;
573 key_fsgid_changed(current);
578 * setgid() is implemented like SysV w/ SAVED_IDS
580 * SMP: Same implicit races as above.
582 asmlinkage long sys_setgid(gid_t gid)
584 int old_egid = current->egid;
587 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
591 if (capable(CAP_SETGID))
595 current->mm->dumpable=0;
598 current->gid = current->egid = current->sgid = current->fsgid = gid;
600 else if ((gid == current->gid) || (gid == current->sgid))
604 current->mm->dumpable=0;
607 current->egid = current->fsgid = gid;
612 key_fsgid_changed(current);
617 static int set_user(uid_t new_ruid, int dumpclear)
619 struct user_struct *new_user;
621 new_user = alloc_uid(vx_current_xid(), new_ruid);
625 if (atomic_read(&new_user->processes) >=
626 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
627 new_user != &root_user) {
632 switch_uid(new_user);
636 current->mm->dumpable = 0;
639 current->uid = new_ruid;
644 * Unprivileged users may change the real uid to the effective uid
645 * or vice versa. (BSD-style)
647 * If you set the real uid at all, or set the effective uid to a value not
648 * equal to the real uid, then the saved uid is set to the new effective uid.
650 * This makes it possible for a setuid program to completely drop its
651 * privileges, which is often a useful assertion to make when you are doing
652 * a security audit over a program.
654 * The general idea is that a program which uses just setreuid() will be
655 * 100% compatible with BSD. A program which uses just setuid() will be
656 * 100% compatible with POSIX with saved IDs.
658 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
660 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
663 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
667 new_ruid = old_ruid = current->uid;
668 new_euid = old_euid = current->euid;
669 old_suid = current->suid;
671 if (ruid != (uid_t) -1) {
673 if ((old_ruid != ruid) &&
674 (current->euid != ruid) &&
675 !capable(CAP_SETUID))
679 if (euid != (uid_t) -1) {
681 if ((old_ruid != euid) &&
682 (current->euid != euid) &&
683 (current->suid != euid) &&
684 !capable(CAP_SETUID))
688 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
691 if (new_euid != old_euid)
693 current->mm->dumpable=0;
696 current->fsuid = current->euid = new_euid;
697 if (ruid != (uid_t) -1 ||
698 (euid != (uid_t) -1 && euid != old_ruid))
699 current->suid = current->euid;
700 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 = 0;
746 current->fsuid = current->euid = uid;
747 current->suid = new_suid;
749 key_fsuid_changed(current);
751 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
756 * This function implements a generic ability to update ruid, euid,
757 * and suid. This allows you to implement the 4.4 compatible seteuid().
759 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
761 int old_ruid = current->uid;
762 int old_euid = current->euid;
763 int old_suid = current->suid;
766 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
770 if (!capable(CAP_SETUID)) {
771 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
772 (ruid != current->euid) && (ruid != current->suid))
774 if ((euid != (uid_t) -1) && (euid != current->uid) &&
775 (euid != current->euid) && (euid != current->suid))
777 if ((suid != (uid_t) -1) && (suid != current->uid) &&
778 (suid != current->euid) && (suid != current->suid))
781 if (ruid != (uid_t) -1) {
782 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
785 if (euid != (uid_t) -1) {
786 if (euid != current->euid)
788 current->mm->dumpable = 0;
791 current->euid = euid;
793 current->fsuid = current->euid;
794 if (suid != (uid_t) -1)
795 current->suid = suid;
797 key_fsuid_changed(current);
799 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
802 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
806 if (!(retval = put_user(current->uid, ruid)) &&
807 !(retval = put_user(current->euid, euid)))
808 retval = put_user(current->suid, suid);
814 * Same as above, but for rgid, egid, sgid.
816 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
820 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
824 if (!capable(CAP_SETGID)) {
825 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
826 (rgid != current->egid) && (rgid != current->sgid))
828 if ((egid != (gid_t) -1) && (egid != current->gid) &&
829 (egid != current->egid) && (egid != current->sgid))
831 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
832 (sgid != current->egid) && (sgid != current->sgid))
835 if (egid != (gid_t) -1) {
836 if (egid != current->egid)
838 current->mm->dumpable = 0;
841 current->egid = egid;
843 current->fsgid = current->egid;
844 if (rgid != (gid_t) -1)
846 if (sgid != (gid_t) -1)
847 current->sgid = sgid;
849 key_fsgid_changed(current);
853 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
857 if (!(retval = put_user(current->gid, rgid)) &&
858 !(retval = put_user(current->egid, egid)))
859 retval = put_user(current->sgid, sgid);
866 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
867 * is used for "access()" and for the NFS daemon (letting nfsd stay at
868 * whatever uid it wants to). It normally shadows "euid", except when
869 * explicitly set by setfsuid() or for access..
871 asmlinkage long sys_setfsuid(uid_t uid)
875 old_fsuid = current->fsuid;
876 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
879 if (uid == current->uid || uid == current->euid ||
880 uid == current->suid || uid == current->fsuid ||
883 if (uid != old_fsuid)
885 current->mm->dumpable = 0;
888 current->fsuid = uid;
891 key_fsuid_changed(current);
893 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
899 * Samma på svenska..
901 asmlinkage long sys_setfsgid(gid_t gid)
905 old_fsgid = current->fsgid;
906 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
909 if (gid == current->gid || gid == current->egid ||
910 gid == current->sgid || gid == current->fsgid ||
913 if (gid != old_fsgid)
915 current->mm->dumpable = 0;
918 current->fsgid = gid;
919 key_fsgid_changed(current);
924 asmlinkage long sys_times(struct tms __user * tbuf)
927 * In the SMP world we might just be unlucky and have one of
928 * the times increment as we use it. Since the value is an
929 * atomically safe type this is just fine. Conceptually its
930 * as if the syscall took an instant longer to occur.
934 struct task_struct *tsk = current;
935 struct task_struct *t;
936 cputime_t utime, stime, cutime, cstime;
938 read_lock(&tasklist_lock);
939 utime = tsk->signal->utime;
940 stime = tsk->signal->stime;
943 utime = cputime_add(utime, t->utime);
944 stime = cputime_add(stime, t->stime);
949 * While we have tasklist_lock read-locked, no dying thread
950 * can be updating current->signal->[us]time. Instead,
951 * we got their counts included in the live thread loop.
952 * However, another thread can come in right now and
953 * do a wait call that updates current->signal->c[us]time.
954 * To make sure we always see that pair updated atomically,
955 * we take the siglock around fetching them.
957 spin_lock_irq(&tsk->sighand->siglock);
958 cutime = tsk->signal->cutime;
959 cstime = tsk->signal->cstime;
960 spin_unlock_irq(&tsk->sighand->siglock);
961 read_unlock(&tasklist_lock);
963 tmp.tms_utime = cputime_to_clock_t(utime);
964 tmp.tms_stime = cputime_to_clock_t(stime);
965 tmp.tms_cutime = cputime_to_clock_t(cutime);
966 tmp.tms_cstime = cputime_to_clock_t(cstime);
967 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
970 return (long) jiffies_64_to_clock_t(get_jiffies_64());
974 * This needs some heavy checking ...
975 * I just haven't the stomach for it. I also don't fully
976 * understand sessions/pgrp etc. Let somebody who does explain it.
978 * OK, I think I have the protection semantics right.... this is really
979 * only important on a multi-user system anyway, to make sure one user
980 * can't send a signal to a process owned by another. -TYT, 12/12/91
982 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
986 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
988 struct task_struct *p;
993 pid = vx_map_pid(current->pid);
999 rpgid = vx_rmap_pid(pgid);
1001 /* From this point forward we keep holding onto the tasklist lock
1002 * so that our parent does not change from under us. -DaveM
1004 write_lock_irq(&tasklist_lock);
1007 p = find_task_by_pid(pid);
1012 if (!thread_group_leader(p))
1015 if (p->parent == current || p->real_parent == current) {
1017 if (p->signal->session != current->signal->session)
1029 if (p->signal->leader)
1033 struct task_struct *p;
1035 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1036 if (p->signal->session == current->signal->session)
1038 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1043 err = security_task_setpgid(p, rpgid);
1047 if (process_group(p) != rpgid) {
1048 detach_pid(p, PIDTYPE_PGID);
1049 p->signal->pgrp = rpgid;
1050 attach_pid(p, PIDTYPE_PGID, rpgid);
1055 /* All paths lead to here, thus we are safe. -DaveM */
1056 write_unlock_irq(&tasklist_lock);
1060 asmlinkage long sys_getpgid(pid_t pid)
1063 return vx_rmap_pid(process_group(current));
1066 struct task_struct *p;
1068 read_lock(&tasklist_lock);
1069 p = find_task_by_pid(pid);
1073 retval = security_task_getpgid(p);
1075 retval = vx_rmap_pid(process_group(p));
1077 read_unlock(&tasklist_lock);
1082 #ifdef __ARCH_WANT_SYS_GETPGRP
1084 asmlinkage long sys_getpgrp(void)
1086 /* SMP - assuming writes are word atomic this is fine */
1087 return process_group(current);
1092 asmlinkage long sys_getsid(pid_t pid)
1095 return current->signal->session;
1098 struct task_struct *p;
1100 read_lock(&tasklist_lock);
1101 p = find_task_by_pid(pid);
1105 retval = security_task_getsid(p);
1107 retval = p->signal->session;
1109 read_unlock(&tasklist_lock);
1114 asmlinkage long sys_setsid(void)
1119 if (!thread_group_leader(current))
1123 write_lock_irq(&tasklist_lock);
1125 pid = find_pid(PIDTYPE_PGID, current->pid);
1129 current->signal->leader = 1;
1130 __set_special_pids(current->pid, current->pid);
1131 current->signal->tty = NULL;
1132 current->signal->tty_old_pgrp = 0;
1133 err = process_group(current);
1135 write_unlock_irq(&tasklist_lock);
1141 * Supplementary group IDs
1144 /* init to 2 - one for init_task, one to ensure it is never freed */
1145 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1147 struct group_info *groups_alloc(int gidsetsize)
1149 struct group_info *group_info;
1153 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1154 /* Make sure we always allocate at least one indirect block pointer */
1155 nblocks = nblocks ? : 1;
1156 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1159 group_info->ngroups = gidsetsize;
1160 group_info->nblocks = nblocks;
1161 atomic_set(&group_info->usage, 1);
1163 if (gidsetsize <= NGROUPS_SMALL) {
1164 group_info->blocks[0] = group_info->small_block;
1166 for (i = 0; i < nblocks; i++) {
1168 b = (void *)__get_free_page(GFP_USER);
1170 goto out_undo_partial_alloc;
1171 group_info->blocks[i] = b;
1176 out_undo_partial_alloc:
1178 free_page((unsigned long)group_info->blocks[i]);
1184 EXPORT_SYMBOL(groups_alloc);
1186 void groups_free(struct group_info *group_info)
1188 if (group_info->blocks[0] != group_info->small_block) {
1190 for (i = 0; i < group_info->nblocks; i++)
1191 free_page((unsigned long)group_info->blocks[i]);
1196 EXPORT_SYMBOL(groups_free);
1198 /* export the group_info to a user-space array */
1199 static int groups_to_user(gid_t __user *grouplist,
1200 struct group_info *group_info)
1203 int count = group_info->ngroups;
1205 for (i = 0; i < group_info->nblocks; i++) {
1206 int cp_count = min(NGROUPS_PER_BLOCK, count);
1207 int off = i * NGROUPS_PER_BLOCK;
1208 int len = cp_count * sizeof(*grouplist);
1210 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1218 /* fill a group_info from a user-space array - it must be allocated already */
1219 static int groups_from_user(struct group_info *group_info,
1220 gid_t __user *grouplist)
1223 int count = group_info->ngroups;
1225 for (i = 0; i < group_info->nblocks; i++) {
1226 int cp_count = min(NGROUPS_PER_BLOCK, count);
1227 int off = i * NGROUPS_PER_BLOCK;
1228 int len = cp_count * sizeof(*grouplist);
1230 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1238 /* a simple Shell sort */
1239 static void groups_sort(struct group_info *group_info)
1241 int base, max, stride;
1242 int gidsetsize = group_info->ngroups;
1244 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1249 max = gidsetsize - stride;
1250 for (base = 0; base < max; base++) {
1252 int right = left + stride;
1253 gid_t tmp = GROUP_AT(group_info, right);
1255 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1256 GROUP_AT(group_info, right) =
1257 GROUP_AT(group_info, left);
1261 GROUP_AT(group_info, right) = tmp;
1267 /* a simple bsearch */
1268 static int groups_search(struct group_info *group_info, gid_t grp)
1276 right = group_info->ngroups;
1277 while (left < right) {
1278 int mid = (left+right)/2;
1279 int cmp = grp - GROUP_AT(group_info, mid);
1290 /* validate and set current->group_info */
1291 int set_current_groups(struct group_info *group_info)
1294 struct group_info *old_info;
1296 retval = security_task_setgroups(group_info);
1300 groups_sort(group_info);
1301 get_group_info(group_info);
1304 old_info = current->group_info;
1305 current->group_info = group_info;
1306 task_unlock(current);
1308 put_group_info(old_info);
1313 EXPORT_SYMBOL(set_current_groups);
1315 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1320 * SMP: Nobody else can change our grouplist. Thus we are
1327 /* no need to grab task_lock here; it cannot change */
1328 get_group_info(current->group_info);
1329 i = current->group_info->ngroups;
1331 if (i > gidsetsize) {
1335 if (groups_to_user(grouplist, current->group_info)) {
1341 put_group_info(current->group_info);
1346 * SMP: Our groups are copy-on-write. We can set them safely
1347 * without another task interfering.
1350 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1352 struct group_info *group_info;
1355 if (!capable(CAP_SETGID))
1357 if ((unsigned)gidsetsize > NGROUPS_MAX)
1360 group_info = groups_alloc(gidsetsize);
1363 retval = groups_from_user(group_info, grouplist);
1365 put_group_info(group_info);
1369 retval = set_current_groups(group_info);
1370 put_group_info(group_info);
1376 * Check whether we're fsgid/egid or in the supplemental group..
1378 int in_group_p(gid_t grp)
1381 if (grp != current->fsgid) {
1382 get_group_info(current->group_info);
1383 retval = groups_search(current->group_info, grp);
1384 put_group_info(current->group_info);
1389 EXPORT_SYMBOL(in_group_p);
1391 int in_egroup_p(gid_t grp)
1394 if (grp != current->egid) {
1395 get_group_info(current->group_info);
1396 retval = groups_search(current->group_info, grp);
1397 put_group_info(current->group_info);
1402 EXPORT_SYMBOL(in_egroup_p);
1404 DECLARE_RWSEM(uts_sem);
1406 EXPORT_SYMBOL(uts_sem);
1408 asmlinkage long sys_newuname(struct new_utsname __user * name)
1412 down_read(&uts_sem);
1413 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1419 asmlinkage long sys_sethostname(char __user *name, int len)
1422 char tmp[__NEW_UTS_LEN];
1424 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1426 if (len < 0 || len > __NEW_UTS_LEN)
1428 down_write(&uts_sem);
1430 if (!copy_from_user(tmp, name, len)) {
1431 char *ptr = vx_new_uts(nodename);
1433 memcpy(ptr, tmp, len);
1441 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1443 asmlinkage long sys_gethostname(char __user *name, int len)
1450 down_read(&uts_sem);
1451 ptr = vx_new_uts(nodename);
1452 i = 1 + strlen(ptr);
1456 if (copy_to_user(name, ptr, i))
1465 * Only setdomainname; getdomainname can be implemented by calling
1468 asmlinkage long sys_setdomainname(char __user *name, int len)
1471 char tmp[__NEW_UTS_LEN];
1473 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1475 if (len < 0 || len > __NEW_UTS_LEN)
1478 down_write(&uts_sem);
1480 if (!copy_from_user(tmp, name, len)) {
1481 char *ptr = vx_new_uts(domainname);
1483 memcpy(ptr, tmp, len);
1491 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1493 if (resource >= RLIM_NLIMITS)
1496 struct rlimit value;
1497 task_lock(current->group_leader);
1498 value = current->signal->rlim[resource];
1499 task_unlock(current->group_leader);
1500 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1504 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1507 * Back compatibility for getrlimit. Needed for some apps.
1510 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1513 if (resource >= RLIM_NLIMITS)
1516 task_lock(current->group_leader);
1517 x = current->signal->rlim[resource];
1518 task_unlock(current->group_leader);
1519 if(x.rlim_cur > 0x7FFFFFFF)
1520 x.rlim_cur = 0x7FFFFFFF;
1521 if(x.rlim_max > 0x7FFFFFFF)
1522 x.rlim_max = 0x7FFFFFFF;
1523 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1528 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1530 struct rlimit new_rlim, *old_rlim;
1533 if (resource >= RLIM_NLIMITS)
1535 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1537 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1539 old_rlim = current->signal->rlim + resource;
1540 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1541 !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1543 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1546 retval = security_task_setrlimit(resource, &new_rlim);
1550 task_lock(current->group_leader);
1551 *old_rlim = new_rlim;
1552 task_unlock(current->group_leader);
1554 if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1555 (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1556 new_rlim.rlim_cur <= cputime_to_secs(
1557 current->signal->it_prof_expires))) {
1558 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1559 read_lock(&tasklist_lock);
1560 spin_lock_irq(¤t->sighand->siglock);
1561 set_process_cpu_timer(current, CPUCLOCK_PROF,
1563 spin_unlock_irq(¤t->sighand->siglock);
1564 read_unlock(&tasklist_lock);
1571 * It would make sense to put struct rusage in the task_struct,
1572 * except that would make the task_struct be *really big*. After
1573 * task_struct gets moved into malloc'ed memory, it would
1574 * make sense to do this. It will make moving the rest of the information
1575 * a lot simpler! (Which we're not doing right now because we're not
1576 * measuring them yet).
1578 * This expects to be called with tasklist_lock read-locked or better,
1579 * and the siglock not locked. It may momentarily take the siglock.
1581 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1582 * races with threads incrementing their own counters. But since word
1583 * reads are atomic, we either get new values or old values and we don't
1584 * care which for the sums. We always take the siglock to protect reading
1585 * the c* fields from p->signal from races with exit.c updating those
1586 * fields when reaping, so a sample either gets all the additions of a
1587 * given child after it's reaped, or none so this sample is before reaping.
1590 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1592 struct task_struct *t;
1593 unsigned long flags;
1594 cputime_t utime, stime;
1596 memset((char *) r, 0, sizeof *r);
1598 if (unlikely(!p->signal))
1602 case RUSAGE_CHILDREN:
1603 spin_lock_irqsave(&p->sighand->siglock, flags);
1604 utime = p->signal->cutime;
1605 stime = p->signal->cstime;
1606 r->ru_nvcsw = p->signal->cnvcsw;
1607 r->ru_nivcsw = p->signal->cnivcsw;
1608 r->ru_minflt = p->signal->cmin_flt;
1609 r->ru_majflt = p->signal->cmaj_flt;
1610 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1611 cputime_to_timeval(utime, &r->ru_utime);
1612 cputime_to_timeval(stime, &r->ru_stime);
1615 spin_lock_irqsave(&p->sighand->siglock, flags);
1616 utime = stime = cputime_zero;
1619 spin_lock_irqsave(&p->sighand->siglock, flags);
1620 utime = p->signal->cutime;
1621 stime = p->signal->cstime;
1622 r->ru_nvcsw = p->signal->cnvcsw;
1623 r->ru_nivcsw = p->signal->cnivcsw;
1624 r->ru_minflt = p->signal->cmin_flt;
1625 r->ru_majflt = p->signal->cmaj_flt;
1627 utime = cputime_add(utime, p->signal->utime);
1628 stime = cputime_add(stime, p->signal->stime);
1629 r->ru_nvcsw += p->signal->nvcsw;
1630 r->ru_nivcsw += p->signal->nivcsw;
1631 r->ru_minflt += p->signal->min_flt;
1632 r->ru_majflt += p->signal->maj_flt;
1635 utime = cputime_add(utime, t->utime);
1636 stime = cputime_add(stime, t->stime);
1637 r->ru_nvcsw += t->nvcsw;
1638 r->ru_nivcsw += t->nivcsw;
1639 r->ru_minflt += t->min_flt;
1640 r->ru_majflt += t->maj_flt;
1643 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1644 cputime_to_timeval(utime, &r->ru_utime);
1645 cputime_to_timeval(stime, &r->ru_stime);
1652 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1655 read_lock(&tasklist_lock);
1656 k_getrusage(p, who, &r);
1657 read_unlock(&tasklist_lock);
1658 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1661 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1663 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1665 return getrusage(current, who, ru);
1668 asmlinkage long sys_umask(int mask)
1670 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1674 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1675 unsigned long arg4, unsigned long arg5)
1680 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1685 case PR_SET_PDEATHSIG:
1687 if (!valid_signal(sig)) {
1691 current->pdeath_signal = sig;
1693 case PR_GET_PDEATHSIG:
1694 error = put_user(current->pdeath_signal, (int __user *)arg2);
1696 case PR_GET_DUMPABLE:
1697 if (current->mm->dumpable)
1700 case PR_SET_DUMPABLE:
1701 if (arg2 != 0 && arg2 != 1) {
1705 current->mm->dumpable = arg2;
1708 case PR_SET_UNALIGN:
1709 error = SET_UNALIGN_CTL(current, arg2);
1711 case PR_GET_UNALIGN:
1712 error = GET_UNALIGN_CTL(current, arg2);
1715 error = SET_FPEMU_CTL(current, arg2);
1718 error = GET_FPEMU_CTL(current, arg2);
1721 error = SET_FPEXC_CTL(current, arg2);
1724 error = GET_FPEXC_CTL(current, arg2);
1727 error = PR_TIMING_STATISTICAL;
1730 if (arg2 == PR_TIMING_STATISTICAL)
1736 case PR_GET_KEEPCAPS:
1737 if (current->keep_capabilities)
1740 case PR_SET_KEEPCAPS:
1741 if (arg2 != 0 && arg2 != 1) {
1745 current->keep_capabilities = arg2;
1748 struct task_struct *me = current;
1749 unsigned char ncomm[sizeof(me->comm)];
1751 ncomm[sizeof(me->comm)-1] = 0;
1752 if (strncpy_from_user(ncomm, (char __user *)arg2,
1753 sizeof(me->comm)-1) < 0)
1755 set_task_comm(me, ncomm);
1759 struct task_struct *me = current;
1760 unsigned char tcomm[sizeof(me->comm)];
1762 get_task_comm(tcomm, me);
1763 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))