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_events.h>
28 #include <linux/tty.h>
29 #include <linux/vs_cvirt.h>
30 #include <linux/compat.h>
31 #include <linux/syscalls.h>
33 #include <asm/uaccess.h>
35 #include <asm/unistd.h>
37 #ifndef SET_UNALIGN_CTL
38 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
40 #ifndef GET_UNALIGN_CTL
41 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
44 # define SET_FPEMU_CTL(a,b) (-EINVAL)
47 # define GET_FPEMU_CTL(a,b) (-EINVAL)
50 # define SET_FPEXC_CTL(a,b) (-EINVAL)
53 # define GET_FPEXC_CTL(a,b) (-EINVAL)
57 * this is where the system-wide overflow UID and GID are defined, for
58 * architectures that now have 32-bit UID/GID but didn't in the past
61 int overflowuid = DEFAULT_OVERFLOWUID;
62 int overflowgid = DEFAULT_OVERFLOWGID;
65 EXPORT_SYMBOL(overflowuid);
66 EXPORT_SYMBOL(overflowgid);
70 * the same as above, but for filesystems which can only store a 16-bit
71 * UID and GID. as such, this is needed on all architectures
74 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
75 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
77 EXPORT_SYMBOL(fs_overflowuid);
78 EXPORT_SYMBOL(fs_overflowgid);
81 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
88 * Notifier list for kernel code which wants to be called
89 * at shutdown. This is used to stop any idling DMA operations
93 static struct notifier_block *reboot_notifier_list;
94 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
97 * notifier_chain_register - Add notifier to a notifier chain
98 * @list: Pointer to root list pointer
99 * @n: New entry in notifier chain
101 * Adds a notifier to a notifier chain.
103 * Currently always returns zero.
106 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
108 write_lock(¬ifier_lock);
111 if(n->priority > (*list)->priority)
113 list= &((*list)->next);
117 write_unlock(¬ifier_lock);
121 EXPORT_SYMBOL(notifier_chain_register);
124 * notifier_chain_unregister - Remove notifier from a notifier chain
125 * @nl: Pointer to root list pointer
126 * @n: New entry in notifier chain
128 * Removes a notifier from a notifier chain.
130 * Returns zero on success, or %-ENOENT on failure.
133 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
135 write_lock(¬ifier_lock);
141 write_unlock(¬ifier_lock);
146 write_unlock(¬ifier_lock);
150 EXPORT_SYMBOL(notifier_chain_unregister);
153 * notifier_call_chain - Call functions in a notifier chain
154 * @n: Pointer to root pointer of notifier chain
155 * @val: Value passed unmodified to notifier function
156 * @v: Pointer passed unmodified to notifier function
158 * Calls each function in a notifier chain in turn.
160 * If the return value of the notifier can be and'd
161 * with %NOTIFY_STOP_MASK, then notifier_call_chain
162 * will return immediately, with the return value of
163 * the notifier function which halted execution.
164 * Otherwise, the return value is the return value
165 * of the last notifier function called.
168 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
171 struct notifier_block *nb = *n;
175 ret=nb->notifier_call(nb,val,v);
176 if(ret&NOTIFY_STOP_MASK)
185 EXPORT_SYMBOL(notifier_call_chain);
188 * register_reboot_notifier - Register function to be called at reboot time
189 * @nb: Info about notifier function to be called
191 * Registers a function with the list of functions
192 * to be called at reboot time.
194 * Currently always returns zero, as notifier_chain_register
195 * always returns zero.
198 int register_reboot_notifier(struct notifier_block * nb)
200 return notifier_chain_register(&reboot_notifier_list, nb);
203 EXPORT_SYMBOL(register_reboot_notifier);
206 * unregister_reboot_notifier - Unregister previously registered reboot notifier
207 * @nb: Hook to be unregistered
209 * Unregisters a previously registered reboot
212 * Returns zero on success, or %-ENOENT on failure.
215 int unregister_reboot_notifier(struct notifier_block * nb)
217 return notifier_chain_unregister(&reboot_notifier_list, nb);
220 EXPORT_SYMBOL(unregister_reboot_notifier);
221 static int set_one_prio(struct task_struct *p, int niceval, int error)
225 if (p->uid != current->euid &&
226 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
230 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
231 if (vx_flags(VXF_IGNEG_NICE, 0))
237 no_nice = security_task_setnice(p, niceval);
244 set_user_nice(p, niceval);
249 asmlinkage long sys_setpriority(int which, int who, int niceval)
251 struct task_struct *g, *p;
252 struct user_struct *user;
255 if (which > 2 || which < 0)
258 /* normalize: avoid signed division (rounding problems) */
265 read_lock(&tasklist_lock);
270 p = find_task_by_pid(who);
272 error = set_one_prio(p, niceval, error);
276 who = process_group(current);
277 do_each_task_pid(who, PIDTYPE_PGID, p) {
278 error = set_one_prio(p, niceval, error);
279 } while_each_task_pid(who, PIDTYPE_PGID, p);
282 user = current->user;
286 if ((who != current->uid) &&
287 !(user = find_user(vx_current_xid(), who)))
288 goto out_unlock; /* No processes for this user */
292 error = set_one_prio(p, niceval, error);
293 while_each_thread(g, p);
294 if (who != current->uid)
295 free_uid(user); /* For find_user() */
299 read_unlock(&tasklist_lock);
305 * Ugh. To avoid negative return values, "getpriority()" will
306 * not return the normal nice-value, but a negated value that
307 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
308 * to stay compatible.
310 asmlinkage long sys_getpriority(int which, int who)
312 struct task_struct *g, *p;
313 struct user_struct *user;
314 long niceval, retval = -ESRCH;
316 if (which > 2 || which < 0)
319 read_lock(&tasklist_lock);
324 p = find_task_by_pid(who);
326 niceval = 20 - task_nice(p);
327 if (niceval > retval)
333 who = process_group(current);
334 do_each_task_pid(who, PIDTYPE_PGID, p) {
335 niceval = 20 - task_nice(p);
336 if (niceval > retval)
338 } while_each_task_pid(who, PIDTYPE_PGID, p);
341 user = current->user;
345 if ((who != current->uid) &&
346 !(user = find_user(vx_current_xid(), who)))
347 goto out_unlock; /* No processes for this user */
351 niceval = 20 - task_nice(p);
352 if (niceval > retval)
355 while_each_thread(g, p);
356 if (who != current->uid)
357 free_uid(user); /* for find_user() */
361 read_unlock(&tasklist_lock);
363 key_fsgid_changed(current);
367 long vs_reboot(unsigned int, void *);
370 * Reboot system call: for obvious reasons only root may call it,
371 * and even root needs to set up some magic numbers in the registers
372 * so that some mistake won't make this reboot the whole machine.
373 * You can also set the meaning of the ctrl-alt-del-key here.
375 * reboot doesn't sync: do that yourself before calling this.
377 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
381 /* We only trust the superuser with rebooting the system. */
382 if (!capable(CAP_SYS_BOOT))
385 /* For safety, we require "magic" arguments. */
386 if (magic1 != LINUX_REBOOT_MAGIC1 ||
387 (magic2 != LINUX_REBOOT_MAGIC2 &&
388 magic2 != LINUX_REBOOT_MAGIC2A &&
389 magic2 != LINUX_REBOOT_MAGIC2B &&
390 magic2 != LINUX_REBOOT_MAGIC2C))
393 if (!vx_check(0, VX_ADMIN|VX_WATCH))
394 return vs_reboot(cmd, arg);
398 case LINUX_REBOOT_CMD_RESTART:
399 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
400 system_state = SYSTEM_RESTART;
402 printk(KERN_EMERG "Restarting system.\n");
403 machine_restart(NULL);
406 case LINUX_REBOOT_CMD_CAD_ON:
410 case LINUX_REBOOT_CMD_CAD_OFF:
414 case LINUX_REBOOT_CMD_HALT:
415 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
416 system_state = SYSTEM_HALT;
418 printk(KERN_EMERG "System halted.\n");
424 case LINUX_REBOOT_CMD_POWER_OFF:
425 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
426 system_state = SYSTEM_POWER_OFF;
428 printk(KERN_EMERG "Power down.\n");
434 case LINUX_REBOOT_CMD_RESTART2:
435 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
439 buffer[sizeof(buffer) - 1] = '\0';
441 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
442 system_state = SYSTEM_RESTART;
444 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
445 machine_restart(buffer);
449 case LINUX_REBOOT_CMD_KEXEC:
451 struct kimage *image;
452 image = xchg(&kexec_image, 0);
457 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
458 system_state = SYSTEM_RESTART;
460 system_state = SYSTEM_BOOTING;
461 printk(KERN_EMERG "Starting new kernel\n");
463 machine_kexec(image);
468 #ifdef CONFIG_SOFTWARE_SUSPEND
469 case LINUX_REBOOT_CMD_SW_SUSPEND:
471 int ret = software_suspend();
485 static void deferred_cad(void *dummy)
487 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
488 machine_restart(NULL);
492 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
493 * As it's called within an interrupt, it may NOT sync: the only choice
494 * is whether to reboot at once, or just ignore the ctrl-alt-del.
496 void ctrl_alt_del(void)
498 static DECLARE_WORK(cad_work, deferred_cad, NULL);
501 schedule_work(&cad_work);
503 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;
568 key_fsgid_changed(current);
574 * setgid() is implemented like SysV w/ SAVED_IDS
576 * SMP: Same implicit races as above.
578 asmlinkage long sys_setgid(gid_t gid)
580 int old_egid = current->egid;
583 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
587 if (capable(CAP_SETGID))
591 current->mm->dumpable = suid_dumpable;
594 current->gid = current->egid = current->sgid = current->fsgid = gid;
596 else if ((gid == current->gid) || (gid == current->sgid))
600 current->mm->dumpable = suid_dumpable;
603 current->egid = current->fsgid = gid;
608 key_fsgid_changed(current);
614 static int set_user(uid_t new_ruid, int dumpclear)
616 struct user_struct *new_user;
618 new_user = alloc_uid(vx_current_xid(), new_ruid);
622 if (atomic_read(&new_user->processes) >=
623 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
624 new_user != &root_user) {
629 switch_uid(new_user);
633 current->mm->dumpable = suid_dumpable;
636 current->uid = new_ruid;
641 * Unprivileged users may change the real uid to the effective uid
642 * or vice versa. (BSD-style)
644 * If you set the real uid at all, or set the effective uid to a value not
645 * equal to the real uid, then the saved uid is set to the new effective uid.
647 * This makes it possible for a setuid program to completely drop its
648 * privileges, which is often a useful assertion to make when you are doing
649 * a security audit over a program.
651 * The general idea is that a program which uses just setreuid() will be
652 * 100% compatible with BSD. A program which uses just setuid() will be
653 * 100% compatible with POSIX with saved IDs.
655 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
657 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
660 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
664 new_ruid = old_ruid = current->uid;
665 new_euid = old_euid = current->euid;
666 old_suid = current->suid;
668 if (ruid != (uid_t) -1) {
670 if ((old_ruid != ruid) &&
671 (current->euid != ruid) &&
672 !capable(CAP_SETUID))
676 if (euid != (uid_t) -1) {
678 if ((old_ruid != euid) &&
679 (current->euid != euid) &&
680 (current->suid != euid) &&
681 !capable(CAP_SETUID))
685 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
688 if (new_euid != old_euid)
690 current->mm->dumpable = suid_dumpable;
693 current->fsuid = current->euid = new_euid;
694 if (ruid != (uid_t) -1 ||
695 (euid != (uid_t) -1 && euid != old_ruid))
696 current->suid = current->euid;
697 current->fsuid = current->euid;
699 key_fsuid_changed(current);
702 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
708 * setuid() is implemented like SysV with SAVED_IDS
710 * Note that SAVED_ID's is deficient in that a setuid root program
711 * like sendmail, for example, cannot set its uid to be a normal
712 * user and then switch back, because if you're root, setuid() sets
713 * the saved uid too. If you don't like this, blame the bright people
714 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
715 * will allow a root program to temporarily drop privileges and be able to
716 * regain them by swapping the real and effective uid.
718 asmlinkage long sys_setuid(uid_t uid)
720 int old_euid = current->euid;
721 int old_ruid, old_suid, new_ruid, new_suid;
724 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
728 old_ruid = new_ruid = current->uid;
729 old_suid = current->suid;
732 if (capable(CAP_SETUID)) {
733 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
736 } else if ((uid != current->uid) && (uid != new_suid))
741 current->mm->dumpable = suid_dumpable;
744 current->fsuid = current->euid = uid;
745 current->suid = new_suid;
747 key_fsuid_changed(current);
750 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
755 * This function implements a generic ability to update ruid, euid,
756 * and suid. This allows you to implement the 4.4 compatible seteuid().
758 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
760 int old_ruid = current->uid;
761 int old_euid = current->euid;
762 int old_suid = current->suid;
765 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
769 if (!capable(CAP_SETUID)) {
770 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
771 (ruid != current->euid) && (ruid != current->suid))
773 if ((euid != (uid_t) -1) && (euid != current->uid) &&
774 (euid != current->euid) && (euid != current->suid))
776 if ((suid != (uid_t) -1) && (suid != current->uid) &&
777 (suid != current->euid) && (suid != current->suid))
780 if (ruid != (uid_t) -1) {
781 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
784 if (euid != (uid_t) -1) {
785 if (euid != current->euid)
787 current->mm->dumpable = suid_dumpable;
790 current->euid = euid;
792 current->fsuid = current->euid;
793 if (suid != (uid_t) -1)
794 current->suid = suid;
796 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 = suid_dumpable;
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);
854 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
858 if (!(retval = put_user(current->gid, rgid)) &&
859 !(retval = put_user(current->egid, egid)))
860 retval = put_user(current->sgid, sgid);
867 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
868 * is used for "access()" and for the NFS daemon (letting nfsd stay at
869 * whatever uid it wants to). It normally shadows "euid", except when
870 * explicitly set by setfsuid() or for access..
872 asmlinkage long sys_setfsuid(uid_t uid)
876 old_fsuid = current->fsuid;
877 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
880 if (uid == current->uid || uid == current->euid ||
881 uid == current->suid || uid == current->fsuid ||
884 if (uid != old_fsuid)
886 current->mm->dumpable = suid_dumpable;
889 current->fsuid = uid;
892 key_fsuid_changed(current);
894 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
900 * Samma på svenska..
902 asmlinkage long sys_setfsgid(gid_t gid)
906 old_fsgid = current->fsgid;
907 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
910 if (gid == current->gid || gid == current->egid ||
911 gid == current->sgid || gid == current->fsgid ||
914 if (gid != old_fsgid)
916 current->mm->dumpable = suid_dumpable;
919 current->fsgid = gid;
920 key_fsgid_changed(current);
925 asmlinkage long sys_times(struct tms __user * tbuf)
928 * In the SMP world we might just be unlucky and have one of
929 * the times increment as we use it. Since the value is an
930 * atomically safe type this is just fine. Conceptually its
931 * as if the syscall took an instant longer to occur.
935 struct task_struct *tsk = current;
936 struct task_struct *t;
937 unsigned long utime, stime, cutime, cstime;
939 read_lock(&tasklist_lock);
940 utime = tsk->signal->utime;
941 stime = tsk->signal->stime;
950 * While we have tasklist_lock read-locked, no dying thread
951 * can be updating current->signal->[us]time. Instead,
952 * we got their counts included in the live thread loop.
953 * However, another thread can come in right now and
954 * do a wait call that updates current->signal->c[us]time.
955 * To make sure we always see that pair updated atomically,
956 * we take the siglock around fetching them.
958 spin_lock_irq(&tsk->sighand->siglock);
959 cutime = tsk->signal->cutime;
960 cstime = tsk->signal->cstime;
961 spin_unlock_irq(&tsk->sighand->siglock);
962 read_unlock(&tasklist_lock);
964 tmp.tms_utime = jiffies_to_clock_t(utime);
965 tmp.tms_stime = jiffies_to_clock_t(stime);
966 tmp.tms_cutime = jiffies_to_clock_t(cutime);
967 tmp.tms_cstime = jiffies_to_clock_t(cstime);
968 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
971 return (long) jiffies_64_to_clock_t(get_jiffies_64());
975 * This needs some heavy checking ...
976 * I just haven't the stomach for it. I also don't fully
977 * understand sessions/pgrp etc. Let somebody who does explain it.
979 * OK, I think I have the protection semantics right.... this is really
980 * only important on a multi-user system anyway, to make sure one user
981 * can't send a signal to a process owned by another. -TYT, 12/12/91
983 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
987 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
989 struct task_struct *p;
994 pid = vx_map_pid(current->pid);
1000 rpgid = vx_rmap_pid(pgid);
1002 /* From this point forward we keep holding onto the tasklist lock
1003 * so that our parent does not change from under us. -DaveM
1005 write_lock_irq(&tasklist_lock);
1008 p = find_task_by_pid(pid);
1013 if (!thread_group_leader(p))
1016 if (p->parent == current || p->real_parent == current) {
1018 if (p->signal->session != current->signal->session)
1030 if (p->signal->leader)
1034 struct task_struct *p;
1036 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1037 if (p->signal->session == current->signal->session)
1039 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1044 err = security_task_setpgid(p, rpgid);
1048 if (process_group(p) != rpgid) {
1049 detach_pid(p, PIDTYPE_PGID);
1050 p->signal->pgrp = rpgid;
1051 attach_pid(p, PIDTYPE_PGID, rpgid);
1056 /* All paths lead to here, thus we are safe. -DaveM */
1057 write_unlock_irq(&tasklist_lock);
1061 asmlinkage long sys_getpgid(pid_t pid)
1064 return vx_rmap_pid(process_group(current));
1067 struct task_struct *p;
1069 read_lock(&tasklist_lock);
1070 p = find_task_by_pid(pid);
1074 retval = security_task_getpgid(p);
1076 retval = vx_rmap_pid(process_group(p));
1078 read_unlock(&tasklist_lock);
1083 #ifdef __ARCH_WANT_SYS_GETPGRP
1085 asmlinkage long sys_getpgrp(void)
1087 /* SMP - assuming writes are word atomic this is fine */
1088 return process_group(current);
1093 asmlinkage long sys_getsid(pid_t pid)
1096 return current->signal->session;
1099 struct task_struct *p;
1101 read_lock(&tasklist_lock);
1102 p = find_task_by_pid(pid);
1106 retval = security_task_getsid(p);
1108 retval = p->signal->session;
1110 read_unlock(&tasklist_lock);
1115 asmlinkage long sys_setsid(void)
1120 if (!thread_group_leader(current))
1124 write_lock_irq(&tasklist_lock);
1126 pid = find_pid(PIDTYPE_PGID, current->pid);
1130 current->signal->leader = 1;
1131 __set_special_pids(current->pid, current->pid);
1132 current->signal->tty = NULL;
1133 current->signal->tty_old_pgrp = 0;
1134 err = process_group(current);
1136 write_unlock_irq(&tasklist_lock);
1142 * Supplementary group IDs
1145 /* init to 2 - one for init_task, one to ensure it is never freed */
1146 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1148 struct group_info *groups_alloc(int gidsetsize)
1150 struct group_info *group_info;
1154 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1155 /* Make sure we always allocate at least one indirect block pointer */
1156 nblocks = nblocks ? : 1;
1157 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1160 group_info->ngroups = gidsetsize;
1161 group_info->nblocks = nblocks;
1162 atomic_set(&group_info->usage, 1);
1164 if (gidsetsize <= NGROUPS_SMALL) {
1165 group_info->blocks[0] = group_info->small_block;
1167 for (i = 0; i < nblocks; i++) {
1169 b = (void *)__get_free_page(GFP_USER);
1171 goto out_undo_partial_alloc;
1172 group_info->blocks[i] = b;
1177 out_undo_partial_alloc:
1179 free_page((unsigned long)group_info->blocks[i]);
1185 EXPORT_SYMBOL(groups_alloc);
1187 void groups_free(struct group_info *group_info)
1189 if (group_info->blocks[0] != group_info->small_block) {
1191 for (i = 0; i < group_info->nblocks; i++)
1192 free_page((unsigned long)group_info->blocks[i]);
1197 EXPORT_SYMBOL(groups_free);
1199 /* export the group_info to a user-space array */
1200 static int groups_to_user(gid_t __user *grouplist,
1201 struct group_info *group_info)
1204 int count = group_info->ngroups;
1206 for (i = 0; i < group_info->nblocks; i++) {
1207 int cp_count = min(NGROUPS_PER_BLOCK, count);
1208 int off = i * NGROUPS_PER_BLOCK;
1209 int len = cp_count * sizeof(*grouplist);
1211 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1219 /* fill a group_info from a user-space array - it must be allocated already */
1220 static int groups_from_user(struct group_info *group_info,
1221 gid_t __user *grouplist)
1224 int count = group_info->ngroups;
1226 for (i = 0; i < group_info->nblocks; i++) {
1227 int cp_count = min(NGROUPS_PER_BLOCK, count);
1228 int off = i * NGROUPS_PER_BLOCK;
1229 int len = cp_count * sizeof(*grouplist);
1231 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1239 /* a simple shell-metzner sort */
1240 static void groups_sort(struct group_info *group_info)
1242 int base, max, stride;
1243 int gidsetsize = group_info->ngroups;
1245 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1250 max = gidsetsize - stride;
1251 for (base = 0; base < max; base++) {
1253 int right = left + stride;
1254 gid_t tmp = GROUP_AT(group_info, right);
1256 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1257 GROUP_AT(group_info, right) =
1258 GROUP_AT(group_info, left);
1262 GROUP_AT(group_info, right) = tmp;
1268 /* a simple bsearch */
1269 static int groups_search(struct group_info *group_info, gid_t grp)
1277 right = group_info->ngroups;
1278 while (left < right) {
1279 int mid = (left+right)/2;
1280 int cmp = grp - GROUP_AT(group_info, mid);
1291 /* validate and set current->group_info */
1292 int set_current_groups(struct group_info *group_info)
1295 struct group_info *old_info;
1297 retval = security_task_setgroups(group_info);
1301 groups_sort(group_info);
1302 get_group_info(group_info);
1305 old_info = current->group_info;
1306 current->group_info = group_info;
1307 task_unlock(current);
1309 put_group_info(old_info);
1314 EXPORT_SYMBOL(set_current_groups);
1316 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1321 * SMP: Nobody else can change our grouplist. Thus we are
1328 /* no need to grab task_lock here; it cannot change */
1329 get_group_info(current->group_info);
1330 i = current->group_info->ngroups;
1332 if (i > gidsetsize) {
1336 if (groups_to_user(grouplist, current->group_info)) {
1342 put_group_info(current->group_info);
1347 * SMP: Our groups are copy-on-write. We can set them safely
1348 * without another task interfering.
1351 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1353 struct group_info *group_info;
1356 if (!capable(CAP_SETGID))
1358 if ((unsigned)gidsetsize > NGROUPS_MAX)
1361 group_info = groups_alloc(gidsetsize);
1364 retval = groups_from_user(group_info, grouplist);
1366 put_group_info(group_info);
1370 retval = set_current_groups(group_info);
1371 put_group_info(group_info);
1377 * Check whether we're fsgid/egid or in the supplemental group..
1379 int in_group_p(gid_t grp)
1382 if (grp != current->fsgid) {
1383 get_group_info(current->group_info);
1384 retval = groups_search(current->group_info, grp);
1385 put_group_info(current->group_info);
1390 EXPORT_SYMBOL(in_group_p);
1392 int in_egroup_p(gid_t grp)
1395 if (grp != current->egid) {
1396 get_group_info(current->group_info);
1397 retval = groups_search(current->group_info, grp);
1398 put_group_info(current->group_info);
1403 EXPORT_SYMBOL(in_egroup_p);
1405 DECLARE_RWSEM(uts_sem);
1407 EXPORT_SYMBOL(uts_sem);
1409 asmlinkage long sys_newuname(struct new_utsname __user * name)
1413 down_read(&uts_sem);
1414 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1420 asmlinkage long sys_sethostname(char __user *name, int len)
1423 char tmp[__NEW_UTS_LEN];
1425 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1427 if (len < 0 || len > __NEW_UTS_LEN)
1429 down_write(&uts_sem);
1431 if (!copy_from_user(tmp, name, len)) {
1432 char *ptr = vx_new_uts(nodename);
1434 memcpy(ptr, tmp, len);
1442 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1444 asmlinkage long sys_gethostname(char __user *name, int len)
1451 down_read(&uts_sem);
1452 ptr = vx_new_uts(nodename);
1453 i = 1 + strlen(ptr);
1457 if (copy_to_user(name, ptr, i))
1466 * Only setdomainname; getdomainname can be implemented by calling
1469 asmlinkage long sys_setdomainname(char __user *name, int len)
1472 char tmp[__NEW_UTS_LEN];
1474 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1476 if (len < 0 || len > __NEW_UTS_LEN)
1479 down_write(&uts_sem);
1481 if (!copy_from_user(tmp, name, len)) {
1482 char *ptr = vx_new_uts(domainname);
1484 memcpy(ptr, tmp, len);
1492 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1494 if (resource >= RLIM_NLIMITS)
1497 struct rlimit value;
1498 task_lock(current->group_leader);
1499 value = current->signal->rlim[resource];
1500 task_unlock(current->group_leader);
1501 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1505 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1508 * Back compatibility for getrlimit. Needed for some apps.
1511 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1514 if (resource >= RLIM_NLIMITS)
1517 task_lock(current->group_leader);
1518 x = current->signal->rlim[resource];
1519 task_unlock(current->group_leader);
1520 if(x.rlim_cur > 0x7FFFFFFF)
1521 x.rlim_cur = 0x7FFFFFFF;
1522 if(x.rlim_max > 0x7FFFFFFF)
1523 x.rlim_max = 0x7FFFFFFF;
1524 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1529 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1531 struct rlimit new_rlim, *old_rlim;
1534 if (resource >= RLIM_NLIMITS)
1536 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1538 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1540 old_rlim = current->signal->rlim + resource;
1541 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1542 !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1544 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1547 retval = security_task_setrlimit(resource, &new_rlim);
1551 task_lock(current->group_leader);
1552 *old_rlim = new_rlim;
1553 task_unlock(current->group_leader);
1558 * It would make sense to put struct rusage in the task_struct,
1559 * except that would make the task_struct be *really big*. After
1560 * task_struct gets moved into malloc'ed memory, it would
1561 * make sense to do this. It will make moving the rest of the information
1562 * a lot simpler! (Which we're not doing right now because we're not
1563 * measuring them yet).
1565 * This expects to be called with tasklist_lock read-locked or better,
1566 * and the siglock not locked. It may momentarily take the siglock.
1568 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1569 * races with threads incrementing their own counters. But since word
1570 * reads are atomic, we either get new values or old values and we don't
1571 * care which for the sums. We always take the siglock to protect reading
1572 * the c* fields from p->signal from races with exit.c updating those
1573 * fields when reaping, so a sample either gets all the additions of a
1574 * given child after it's reaped, or none so this sample is before reaping.
1577 void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1579 struct task_struct *t;
1580 unsigned long flags;
1581 unsigned long utime, stime;
1583 memset((char *) r, 0, sizeof *r);
1585 if (unlikely(!p->signal))
1589 case RUSAGE_CHILDREN:
1590 spin_lock_irqsave(&p->sighand->siglock, flags);
1591 utime = p->signal->cutime;
1592 stime = p->signal->cstime;
1593 r->ru_nvcsw = p->signal->cnvcsw;
1594 r->ru_nivcsw = p->signal->cnivcsw;
1595 r->ru_minflt = p->signal->cmin_flt;
1596 r->ru_majflt = p->signal->cmaj_flt;
1597 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1598 jiffies_to_timeval(utime, &r->ru_utime);
1599 jiffies_to_timeval(stime, &r->ru_stime);
1602 spin_lock_irqsave(&p->sighand->siglock, flags);
1606 spin_lock_irqsave(&p->sighand->siglock, flags);
1607 utime = p->signal->cutime;
1608 stime = p->signal->cstime;
1609 r->ru_nvcsw = p->signal->cnvcsw;
1610 r->ru_nivcsw = p->signal->cnivcsw;
1611 r->ru_minflt = p->signal->cmin_flt;
1612 r->ru_majflt = p->signal->cmaj_flt;
1614 utime += p->signal->utime;
1615 stime += p->signal->stime;
1616 r->ru_nvcsw += p->signal->nvcsw;
1617 r->ru_nivcsw += p->signal->nivcsw;
1618 r->ru_minflt += p->signal->min_flt;
1619 r->ru_majflt += p->signal->maj_flt;
1624 r->ru_nvcsw += t->nvcsw;
1625 r->ru_nivcsw += t->nivcsw;
1626 r->ru_minflt += t->min_flt;
1627 r->ru_majflt += t->maj_flt;
1630 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1631 jiffies_to_timeval(utime, &r->ru_utime);
1632 jiffies_to_timeval(stime, &r->ru_stime);
1639 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1642 read_lock(&tasklist_lock);
1643 k_getrusage(p, who, &r);
1644 read_unlock(&tasklist_lock);
1645 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1648 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1650 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1652 return getrusage(current, who, ru);
1655 asmlinkage long sys_umask(int mask)
1657 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1661 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1662 unsigned long arg4, unsigned long arg5)
1667 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1672 case PR_SET_PDEATHSIG:
1674 if (sig < 0 || sig > _NSIG) {
1678 current->pdeath_signal = sig;
1680 case PR_GET_PDEATHSIG:
1681 error = put_user(current->pdeath_signal, (int __user *)arg2);
1683 case PR_GET_DUMPABLE:
1684 if (current->mm->dumpable)
1687 case PR_SET_DUMPABLE:
1688 if (arg2 < 0 && arg2 > 2) {
1692 current->mm->dumpable = arg2;
1695 case PR_SET_UNALIGN:
1696 error = SET_UNALIGN_CTL(current, arg2);
1698 case PR_GET_UNALIGN:
1699 error = GET_UNALIGN_CTL(current, arg2);
1702 error = SET_FPEMU_CTL(current, arg2);
1705 error = GET_FPEMU_CTL(current, arg2);
1708 error = SET_FPEXC_CTL(current, arg2);
1711 error = GET_FPEXC_CTL(current, arg2);
1714 error = PR_TIMING_STATISTICAL;
1717 if (arg2 == PR_TIMING_STATISTICAL)
1723 case PR_GET_KEEPCAPS:
1724 if (current->keep_capabilities)
1727 case PR_SET_KEEPCAPS:
1728 if (arg2 != 0 && arg2 != 1) {
1732 current->keep_capabilities = arg2;
1735 struct task_struct *me = current;
1736 unsigned char ncomm[sizeof(me->comm)];
1738 ncomm[sizeof(me->comm)-1] = 0;
1739 if (strncpy_from_user(ncomm, (char __user *)arg2,
1740 sizeof(me->comm)-1) < 0)
1742 set_task_comm(me, ncomm);