4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/config.h>
8 #include <linux/compat.h>
9 #include <linux/module.h>
11 #include <linux/utsname.h>
12 #include <linux/mman.h>
13 #include <linux/smp_lock.h>
14 #include <linux/notifier.h>
15 #include <linux/kmod.h>
16 #include <linux/reboot.h>
17 #include <linux/prctl.h>
18 #include <linux/init.h>
19 #include <linux/highuid.h>
21 #include <linux/workqueue.h>
22 #include <linux/device.h>
23 #include <linux/times.h>
24 #include <linux/security.h>
25 #include <linux/dcookies.h>
26 #include <linux/suspend.h>
27 #include <linux/vs_base.h>
28 #include <linux/vs_cvirt.h>
30 #include <asm/uaccess.h>
32 #include <asm/unistd.h>
34 #ifndef SET_UNALIGN_CTL
35 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
37 #ifndef GET_UNALIGN_CTL
38 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
41 # define SET_FPEMU_CTL(a,b) (-EINVAL)
44 # define GET_FPEMU_CTL(a,b) (-EINVAL)
47 # define SET_FPEXC_CTL(a,b) (-EINVAL)
50 # define GET_FPEXC_CTL(a,b) (-EINVAL)
54 * this is where the system-wide overflow UID and GID are defined, for
55 * architectures that now have 32-bit UID/GID but didn't in the past
58 int overflowuid = DEFAULT_OVERFLOWUID;
59 int overflowgid = DEFAULT_OVERFLOWGID;
62 EXPORT_SYMBOL(overflowuid);
63 EXPORT_SYMBOL(overflowgid);
67 * the same as above, but for filesystems which can only store a 16-bit
68 * UID and GID. as such, this is needed on all architectures
71 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
72 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
74 EXPORT_SYMBOL(fs_overflowuid);
75 EXPORT_SYMBOL(fs_overflowgid);
78 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
85 * Notifier list for kernel code which wants to be called
86 * at shutdown. This is used to stop any idling DMA operations
90 static struct notifier_block *reboot_notifier_list;
91 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
94 * notifier_chain_register - Add notifier to a notifier chain
95 * @list: Pointer to root list pointer
96 * @n: New entry in notifier chain
98 * Adds a notifier to a notifier chain.
100 * Currently always returns zero.
103 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
105 write_lock(¬ifier_lock);
108 if(n->priority > (*list)->priority)
110 list= &((*list)->next);
114 write_unlock(¬ifier_lock);
118 EXPORT_SYMBOL(notifier_chain_register);
121 * notifier_chain_unregister - Remove notifier from a notifier chain
122 * @nl: Pointer to root list pointer
123 * @n: New entry in notifier chain
125 * Removes a notifier from a notifier chain.
127 * Returns zero on success, or %-ENOENT on failure.
130 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
132 write_lock(¬ifier_lock);
138 write_unlock(¬ifier_lock);
143 write_unlock(¬ifier_lock);
147 EXPORT_SYMBOL(notifier_chain_unregister);
150 * notifier_call_chain - Call functions in a notifier chain
151 * @n: Pointer to root pointer of notifier chain
152 * @val: Value passed unmodified to notifier function
153 * @v: Pointer passed unmodified to notifier function
155 * Calls each function in a notifier chain in turn.
157 * If the return value of the notifier can be and'd
158 * with %NOTIFY_STOP_MASK, then notifier_call_chain
159 * will return immediately, with the return value of
160 * the notifier function which halted execution.
161 * Otherwise, the return value is the return value
162 * of the last notifier function called.
165 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
168 struct notifier_block *nb = *n;
172 ret=nb->notifier_call(nb,val,v);
173 if(ret&NOTIFY_STOP_MASK)
182 EXPORT_SYMBOL(notifier_call_chain);
185 * register_reboot_notifier - Register function to be called at reboot time
186 * @nb: Info about notifier function to be called
188 * Registers a function with the list of functions
189 * to be called at reboot time.
191 * Currently always returns zero, as notifier_chain_register
192 * always returns zero.
195 int register_reboot_notifier(struct notifier_block * nb)
197 return notifier_chain_register(&reboot_notifier_list, nb);
200 EXPORT_SYMBOL(register_reboot_notifier);
203 * unregister_reboot_notifier - Unregister previously registered reboot notifier
204 * @nb: Hook to be unregistered
206 * Unregisters a previously registered reboot
209 * Returns zero on success, or %-ENOENT on failure.
212 int unregister_reboot_notifier(struct notifier_block * nb)
214 return notifier_chain_unregister(&reboot_notifier_list, nb);
217 EXPORT_SYMBOL(unregister_reboot_notifier);
219 asmlinkage long sys_ni_syscall(void)
224 cond_syscall(sys_nfsservctl)
225 cond_syscall(sys_quotactl)
226 cond_syscall(sys_acct)
227 cond_syscall(sys_lookup_dcookie)
228 cond_syscall(sys_swapon)
229 cond_syscall(sys_swapoff)
230 cond_syscall(sys_init_module)
231 cond_syscall(sys_delete_module)
232 cond_syscall(sys_socketpair)
233 cond_syscall(sys_bind)
234 cond_syscall(sys_listen)
235 cond_syscall(sys_accept)
236 cond_syscall(sys_connect)
237 cond_syscall(sys_getsockname)
238 cond_syscall(sys_getpeername)
239 cond_syscall(sys_sendto)
240 cond_syscall(sys_send)
241 cond_syscall(sys_recvfrom)
242 cond_syscall(sys_recv)
243 cond_syscall(sys_socket)
244 cond_syscall(sys_setsockopt)
245 cond_syscall(sys_getsockopt)
246 cond_syscall(sys_shutdown)
247 cond_syscall(sys_sendmsg)
248 cond_syscall(sys_recvmsg)
249 cond_syscall(sys_socketcall)
250 cond_syscall(sys_futex)
251 cond_syscall(compat_sys_futex)
252 cond_syscall(sys_epoll_create)
253 cond_syscall(sys_epoll_ctl)
254 cond_syscall(sys_epoll_wait)
255 cond_syscall(sys_semget)
256 cond_syscall(sys_semop)
257 cond_syscall(sys_semtimedop)
258 cond_syscall(sys_semctl)
259 cond_syscall(sys_msgget)
260 cond_syscall(sys_msgsnd)
261 cond_syscall(sys_msgrcv)
262 cond_syscall(sys_msgctl)
263 cond_syscall(sys_shmget)
264 cond_syscall(sys_shmdt)
265 cond_syscall(sys_shmctl)
266 cond_syscall(sys_mq_open)
267 cond_syscall(sys_mq_unlink)
268 cond_syscall(sys_mq_timedsend)
269 cond_syscall(sys_mq_timedreceive)
270 cond_syscall(sys_mq_notify)
271 cond_syscall(sys_mq_getsetattr)
272 cond_syscall(compat_sys_mq_open)
273 cond_syscall(compat_sys_mq_timedsend)
274 cond_syscall(compat_sys_mq_timedreceive)
275 cond_syscall(compat_sys_mq_notify)
276 cond_syscall(compat_sys_mq_getsetattr)
277 cond_syscall(sys_mbind)
278 cond_syscall(sys_get_mempolicy)
279 cond_syscall(sys_set_mempolicy)
280 cond_syscall(compat_get_mempolicy)
282 /* arch-specific weak syscall entries */
283 cond_syscall(sys_pciconfig_read)
284 cond_syscall(sys_pciconfig_write)
285 cond_syscall(sys_pciconfig_iobase)
287 static int set_one_prio(struct task_struct *p, int niceval, int error)
291 if (p->uid != current->euid &&
292 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
296 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
300 no_nice = security_task_setnice(p, niceval);
307 set_user_nice(p, niceval);
312 asmlinkage long sys_setpriority(int which, int who, int niceval)
314 struct task_struct *g, *p;
315 struct user_struct *user;
320 if (which > 2 || which < 0)
323 /* normalize: avoid signed division (rounding problems) */
330 read_lock(&tasklist_lock);
335 p = find_task_by_pid(who);
337 error = set_one_prio(p, niceval, error);
341 who = process_group(current);
342 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
343 error = set_one_prio(p, niceval, error);
347 user = current->user;
349 user = find_user(vx_current_xid(), who);
356 error = set_one_prio(p, niceval, error);
357 while_each_thread(g, p);
359 free_uid(user); /* For find_user() */
363 read_unlock(&tasklist_lock);
369 * Ugh. To avoid negative return values, "getpriority()" will
370 * not return the normal nice-value, but a negated value that
371 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
372 * to stay compatible.
374 asmlinkage long sys_getpriority(int which, int who)
376 struct task_struct *g, *p;
379 struct user_struct *user;
380 long niceval, retval = -ESRCH;
382 if (which > 2 || which < 0)
385 read_lock(&tasklist_lock);
390 p = find_task_by_pid(who);
392 niceval = 20 - task_nice(p);
393 if (niceval > retval)
399 who = process_group(current);
400 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
401 niceval = 20 - task_nice(p);
402 if (niceval > retval)
408 user = current->user;
410 user = find_user(vx_current_xid(), who);
417 niceval = 20 - task_nice(p);
418 if (niceval > retval)
421 while_each_thread(g, p);
423 free_uid(user); /* for find_user() */
427 read_unlock(&tasklist_lock);
432 long vs_reboot(unsigned int, void *);
435 * Reboot system call: for obvious reasons only root may call it,
436 * and even root needs to set up some magic numbers in the registers
437 * so that some mistake won't make this reboot the whole machine.
438 * You can also set the meaning of the ctrl-alt-del-key here.
440 * reboot doesn't sync: do that yourself before calling this.
442 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
446 /* We only trust the superuser with rebooting the system. */
447 if (!capable(CAP_SYS_BOOT))
450 /* For safety, we require "magic" arguments. */
451 if (magic1 != LINUX_REBOOT_MAGIC1 ||
452 (magic2 != LINUX_REBOOT_MAGIC2 &&
453 magic2 != LINUX_REBOOT_MAGIC2A &&
454 magic2 != LINUX_REBOOT_MAGIC2B &&
455 magic2 != LINUX_REBOOT_MAGIC2C))
458 if (!vx_check(0, VX_ADMIN|VX_WATCH))
459 return vs_reboot(cmd, arg);
463 case LINUX_REBOOT_CMD_RESTART:
464 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
465 system_state = SYSTEM_RESTART;
467 printk(KERN_EMERG "Restarting system.\n");
468 machine_restart(NULL);
471 case LINUX_REBOOT_CMD_CAD_ON:
475 case LINUX_REBOOT_CMD_CAD_OFF:
479 case LINUX_REBOOT_CMD_HALT:
480 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
481 system_state = SYSTEM_HALT;
483 printk(KERN_EMERG "System halted.\n");
489 case LINUX_REBOOT_CMD_POWER_OFF:
490 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
491 system_state = SYSTEM_POWER_OFF;
493 printk(KERN_EMERG "Power down.\n");
499 case LINUX_REBOOT_CMD_RESTART2:
500 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
504 buffer[sizeof(buffer) - 1] = '\0';
506 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
507 system_state = SYSTEM_RESTART;
509 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
510 machine_restart(buffer);
513 #ifdef CONFIG_SOFTWARE_SUSPEND
514 case LINUX_REBOOT_CMD_SW_SUSPEND:
516 int ret = software_suspend();
530 static void deferred_cad(void *dummy)
532 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
533 machine_restart(NULL);
537 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
538 * As it's called within an interrupt, it may NOT sync: the only choice
539 * is whether to reboot at once, or just ignore the ctrl-alt-del.
541 void ctrl_alt_del(void)
543 static DECLARE_WORK(cad_work, deferred_cad, NULL);
546 schedule_work(&cad_work);
548 kill_proc(cad_pid, SIGINT, 1);
553 * Unprivileged users may change the real gid to the effective gid
554 * or vice versa. (BSD-style)
556 * If you set the real gid at all, or set the effective gid to a value not
557 * equal to the real gid, then the saved gid is set to the new effective gid.
559 * This makes it possible for a setgid program to completely drop its
560 * privileges, which is often a useful assertion to make when you are doing
561 * a security audit over a program.
563 * The general idea is that a program which uses just setregid() will be
564 * 100% compatible with BSD. A program which uses just setgid() will be
565 * 100% compatible with POSIX with saved IDs.
567 * SMP: There are not races, the GIDs are checked only by filesystem
568 * operations (as far as semantic preservation is concerned).
570 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
572 int old_rgid = current->gid;
573 int old_egid = current->egid;
574 int new_rgid = old_rgid;
575 int new_egid = old_egid;
578 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
582 if (rgid != (gid_t) -1) {
583 if ((old_rgid == rgid) ||
584 (current->egid==rgid) ||
590 if (egid != (gid_t) -1) {
591 if ((old_rgid == egid) ||
592 (current->egid == egid) ||
593 (current->sgid == egid) ||
600 if (new_egid != old_egid)
602 current->mm->dumpable = 0;
605 if (rgid != (gid_t) -1 ||
606 (egid != (gid_t) -1 && egid != old_rgid))
607 current->sgid = new_egid;
608 current->fsgid = new_egid;
609 current->egid = new_egid;
610 current->gid = new_rgid;
615 * setgid() is implemented like SysV w/ SAVED_IDS
617 * SMP: Same implicit races as above.
619 asmlinkage long sys_setgid(gid_t gid)
621 int old_egid = current->egid;
624 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
628 if (capable(CAP_SETGID))
632 current->mm->dumpable=0;
635 current->gid = current->egid = current->sgid = current->fsgid = gid;
637 else if ((gid == current->gid) || (gid == current->sgid))
641 current->mm->dumpable=0;
644 current->egid = current->fsgid = gid;
651 static int set_user(uid_t new_ruid, int dumpclear)
653 struct user_struct *new_user;
655 new_user = alloc_uid(vx_current_xid(), new_ruid);
659 if (atomic_read(&new_user->processes) >=
660 current->rlim[RLIMIT_NPROC].rlim_cur &&
661 new_user != &root_user) {
666 switch_uid(new_user);
670 current->mm->dumpable = 0;
673 current->uid = new_ruid;
678 * Unprivileged users may change the real uid to the effective uid
679 * or vice versa. (BSD-style)
681 * If you set the real uid at all, or set the effective uid to a value not
682 * equal to the real uid, then the saved uid is set to the new effective uid.
684 * This makes it possible for a setuid program to completely drop its
685 * privileges, which is often a useful assertion to make when you are doing
686 * a security audit over a program.
688 * The general idea is that a program which uses just setreuid() will be
689 * 100% compatible with BSD. A program which uses just setuid() will be
690 * 100% compatible with POSIX with saved IDs.
692 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
694 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
697 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
701 new_ruid = old_ruid = current->uid;
702 new_euid = old_euid = current->euid;
703 old_suid = current->suid;
705 if (ruid != (uid_t) -1) {
707 if ((old_ruid != ruid) &&
708 (current->euid != ruid) &&
709 !capable(CAP_SETUID))
713 if (euid != (uid_t) -1) {
715 if ((old_ruid != euid) &&
716 (current->euid != euid) &&
717 (current->suid != euid) &&
718 !capable(CAP_SETUID))
722 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
725 if (new_euid != old_euid)
727 current->mm->dumpable=0;
730 current->fsuid = current->euid = new_euid;
731 if (ruid != (uid_t) -1 ||
732 (euid != (uid_t) -1 && euid != old_ruid))
733 current->suid = current->euid;
734 current->fsuid = current->euid;
736 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
742 * setuid() is implemented like SysV with SAVED_IDS
744 * Note that SAVED_ID's is deficient in that a setuid root program
745 * like sendmail, for example, cannot set its uid to be a normal
746 * user and then switch back, because if you're root, setuid() sets
747 * the saved uid too. If you don't like this, blame the bright people
748 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
749 * will allow a root program to temporarily drop privileges and be able to
750 * regain them by swapping the real and effective uid.
752 asmlinkage long sys_setuid(uid_t uid)
754 int old_euid = current->euid;
755 int old_ruid, old_suid, new_ruid, new_suid;
758 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
762 old_ruid = new_ruid = current->uid;
763 old_suid = current->suid;
766 if (capable(CAP_SETUID)) {
767 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
770 } else if ((uid != current->uid) && (uid != new_suid))
775 current->mm->dumpable = 0;
778 current->fsuid = current->euid = uid;
779 current->suid = new_suid;
781 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
786 * This function implements a generic ability to update ruid, euid,
787 * and suid. This allows you to implement the 4.4 compatible seteuid().
789 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
791 int old_ruid = current->uid;
792 int old_euid = current->euid;
793 int old_suid = current->suid;
796 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
800 if (!capable(CAP_SETUID)) {
801 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
802 (ruid != current->euid) && (ruid != current->suid))
804 if ((euid != (uid_t) -1) && (euid != current->uid) &&
805 (euid != current->euid) && (euid != current->suid))
807 if ((suid != (uid_t) -1) && (suid != current->uid) &&
808 (suid != current->euid) && (suid != current->suid))
811 if (ruid != (uid_t) -1) {
812 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
815 if (euid != (uid_t) -1) {
816 if (euid != current->euid)
818 current->mm->dumpable = 0;
821 current->euid = euid;
823 current->fsuid = current->euid;
824 if (suid != (uid_t) -1)
825 current->suid = suid;
827 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
830 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
834 if (!(retval = put_user(current->uid, ruid)) &&
835 !(retval = put_user(current->euid, euid)))
836 retval = put_user(current->suid, suid);
842 * Same as above, but for rgid, egid, sgid.
844 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
848 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
852 if (!capable(CAP_SETGID)) {
853 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
854 (rgid != current->egid) && (rgid != current->sgid))
856 if ((egid != (gid_t) -1) && (egid != current->gid) &&
857 (egid != current->egid) && (egid != current->sgid))
859 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
860 (sgid != current->egid) && (sgid != current->sgid))
863 if (egid != (gid_t) -1) {
864 if (egid != current->egid)
866 current->mm->dumpable = 0;
869 current->egid = egid;
871 current->fsgid = current->egid;
872 if (rgid != (gid_t) -1)
874 if (sgid != (gid_t) -1)
875 current->sgid = sgid;
879 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
883 if (!(retval = put_user(current->gid, rgid)) &&
884 !(retval = put_user(current->egid, egid)))
885 retval = put_user(current->sgid, sgid);
892 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
893 * is used for "access()" and for the NFS daemon (letting nfsd stay at
894 * whatever uid it wants to). It normally shadows "euid", except when
895 * explicitly set by setfsuid() or for access..
897 asmlinkage long sys_setfsuid(uid_t uid)
901 old_fsuid = current->fsuid;
902 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
905 if (uid == current->uid || uid == current->euid ||
906 uid == current->suid || uid == current->fsuid ||
909 if (uid != old_fsuid)
911 current->mm->dumpable = 0;
914 current->fsuid = uid;
917 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
923 * Samma på svenska..
925 asmlinkage long sys_setfsgid(gid_t gid)
929 old_fsgid = current->fsgid;
930 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
933 if (gid == current->gid || gid == current->egid ||
934 gid == current->sgid || gid == current->fsgid ||
937 if (gid != old_fsgid)
939 current->mm->dumpable = 0;
942 current->fsgid = gid;
947 asmlinkage long sys_times(struct tms __user * tbuf)
950 * In the SMP world we might just be unlucky and have one of
951 * the times increment as we use it. Since the value is an
952 * atomically safe type this is just fine. Conceptually its
953 * as if the syscall took an instant longer to occur.
957 tmp.tms_utime = jiffies_to_clock_t(current->utime);
958 tmp.tms_stime = jiffies_to_clock_t(current->stime);
959 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
960 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
961 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
964 return (long) jiffies_64_to_clock_t(get_jiffies_64());
968 * This needs some heavy checking ...
969 * I just haven't the stomach for it. I also don't fully
970 * understand sessions/pgrp etc. Let somebody who does explain it.
972 * OK, I think I have the protection semantics right.... this is really
973 * only important on a multi-user system anyway, to make sure one user
974 * can't send a signal to a process owned by another. -TYT, 12/12/91
976 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
980 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
982 struct task_struct *p;
992 /* From this point forward we keep holding onto the tasklist lock
993 * so that our parent does not change from under us. -DaveM
995 write_lock_irq(&tasklist_lock);
998 p = find_task_by_pid(pid);
1003 if (!thread_group_leader(p))
1006 if (p->parent == current || p->real_parent == current) {
1008 if (p->signal->session != current->signal->session)
1020 if (p->signal->leader)
1024 struct task_struct *p;
1026 struct list_head *l;
1028 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1029 if (p->signal->session == current->signal->session)
1035 err = security_task_setpgid(p, pgid);
1039 if (process_group(p) != pgid) {
1040 detach_pid(p, PIDTYPE_PGID);
1041 p->signal->pgrp = pgid;
1042 attach_pid(p, PIDTYPE_PGID, pgid);
1047 /* All paths lead to here, thus we are safe. -DaveM */
1048 write_unlock_irq(&tasklist_lock);
1052 asmlinkage long sys_getpgid(pid_t pid)
1055 return process_group(current);
1058 struct task_struct *p;
1060 read_lock(&tasklist_lock);
1061 p = find_task_by_pid(pid);
1065 retval = security_task_getpgid(p);
1067 retval = process_group(p);
1069 read_unlock(&tasklist_lock);
1074 #ifdef __ARCH_WANT_SYS_GETPGRP
1076 asmlinkage long sys_getpgrp(void)
1078 /* SMP - assuming writes are word atomic this is fine */
1079 return process_group(current);
1084 asmlinkage long sys_getsid(pid_t pid)
1087 return current->signal->session;
1090 struct task_struct *p;
1092 read_lock(&tasklist_lock);
1093 p = find_task_by_pid(pid);
1097 retval = security_task_getsid(p);
1099 retval = p->signal->session;
1101 read_unlock(&tasklist_lock);
1106 asmlinkage long sys_setsid(void)
1111 if (!thread_group_leader(current))
1114 write_lock_irq(&tasklist_lock);
1116 pid = find_pid(PIDTYPE_PGID, current->pid);
1120 current->signal->leader = 1;
1121 __set_special_pids(current->pid, current->pid);
1122 current->signal->tty = NULL;
1123 current->signal->tty_old_pgrp = 0;
1124 err = process_group(current);
1126 write_unlock_irq(&tasklist_lock);
1131 * Supplementary group IDs
1134 /* init to 2 - one for init_task, one to ensure it is never freed */
1135 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1137 struct group_info *groups_alloc(int gidsetsize)
1139 struct group_info *group_info;
1143 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1144 /* Make sure we always allocate at least one indirect block pointer */
1145 nblocks = nblocks ? : 1;
1146 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1149 group_info->ngroups = gidsetsize;
1150 group_info->nblocks = nblocks;
1151 atomic_set(&group_info->usage, 1);
1153 if (gidsetsize <= NGROUPS_SMALL) {
1154 group_info->blocks[0] = group_info->small_block;
1156 for (i = 0; i < nblocks; i++) {
1158 b = (void *)__get_free_page(GFP_USER);
1160 goto out_undo_partial_alloc;
1161 group_info->blocks[i] = b;
1166 out_undo_partial_alloc:
1168 free_page((unsigned long)group_info->blocks[i]);
1174 EXPORT_SYMBOL(groups_alloc);
1176 void groups_free(struct group_info *group_info)
1178 if (group_info->blocks[0] != group_info->small_block) {
1180 for (i = 0; i < group_info->nblocks; i++)
1181 free_page((unsigned long)group_info->blocks[i]);
1186 EXPORT_SYMBOL(groups_free);
1188 /* export the group_info to a user-space array */
1189 static int groups_to_user(gid_t __user *grouplist,
1190 struct group_info *group_info)
1193 int count = group_info->ngroups;
1195 for (i = 0; i < group_info->nblocks; i++) {
1196 int cp_count = min(NGROUPS_PER_BLOCK, count);
1197 int off = i * NGROUPS_PER_BLOCK;
1198 int len = cp_count * sizeof(*grouplist);
1200 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1208 /* fill a group_info from a user-space array - it must be allocated already */
1209 static int groups_from_user(struct group_info *group_info,
1210 gid_t __user *grouplist)
1213 int count = group_info->ngroups;
1215 for (i = 0; i < group_info->nblocks; i++) {
1216 int cp_count = min(NGROUPS_PER_BLOCK, count);
1217 int off = i * NGROUPS_PER_BLOCK;
1218 int len = cp_count * sizeof(*grouplist);
1220 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1228 /* a simple shell-metzner sort */
1229 static void groups_sort(struct group_info *group_info)
1231 int base, max, stride;
1232 int gidsetsize = group_info->ngroups;
1234 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1239 max = gidsetsize - stride;
1240 for (base = 0; base < max; base++) {
1242 int right = left + stride;
1243 gid_t tmp = GROUP_AT(group_info, right);
1245 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1246 GROUP_AT(group_info, right) =
1247 GROUP_AT(group_info, left);
1251 GROUP_AT(group_info, right) = tmp;
1257 /* a simple bsearch */
1258 static int groups_search(struct group_info *group_info, gid_t grp)
1266 right = group_info->ngroups;
1267 while (left < right) {
1268 int mid = (left+right)/2;
1269 int cmp = grp - GROUP_AT(group_info, mid);
1280 /* validate and set current->group_info */
1281 int set_current_groups(struct group_info *group_info)
1284 struct group_info *old_info;
1286 retval = security_task_setgroups(group_info);
1290 groups_sort(group_info);
1291 get_group_info(group_info);
1294 old_info = current->group_info;
1295 current->group_info = group_info;
1296 task_unlock(current);
1298 put_group_info(old_info);
1303 EXPORT_SYMBOL(set_current_groups);
1305 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1310 * SMP: Nobody else can change our grouplist. Thus we are
1317 /* no need to grab task_lock here; it cannot change */
1318 get_group_info(current->group_info);
1319 i = current->group_info->ngroups;
1321 if (i > gidsetsize) {
1325 if (groups_to_user(grouplist, current->group_info)) {
1331 put_group_info(current->group_info);
1336 * SMP: Our groups are copy-on-write. We can set them safely
1337 * without another task interfering.
1340 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1342 struct group_info *group_info;
1345 if (!capable(CAP_SETGID))
1347 if ((unsigned)gidsetsize > NGROUPS_MAX)
1350 group_info = groups_alloc(gidsetsize);
1353 retval = groups_from_user(group_info, grouplist);
1355 put_group_info(group_info);
1359 retval = set_current_groups(group_info);
1360 put_group_info(group_info);
1366 * Check whether we're fsgid/egid or in the supplemental group..
1368 int in_group_p(gid_t grp)
1371 if (grp != current->fsgid) {
1372 get_group_info(current->group_info);
1373 retval = groups_search(current->group_info, grp);
1374 put_group_info(current->group_info);
1379 EXPORT_SYMBOL(in_group_p);
1381 int in_egroup_p(gid_t grp)
1384 if (grp != current->egid) {
1385 get_group_info(current->group_info);
1386 retval = groups_search(current->group_info, grp);
1387 put_group_info(current->group_info);
1392 EXPORT_SYMBOL(in_egroup_p);
1394 DECLARE_RWSEM(uts_sem);
1396 EXPORT_SYMBOL(uts_sem);
1398 asmlinkage long sys_newuname(struct new_utsname __user * name)
1402 down_read(&uts_sem);
1403 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1409 asmlinkage long sys_sethostname(char __user *name, int len)
1412 char tmp[__NEW_UTS_LEN];
1414 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1416 if (len < 0 || len > __NEW_UTS_LEN)
1418 down_write(&uts_sem);
1420 if (!copy_from_user(tmp, name, len)) {
1421 char *ptr = vx_new_uts(nodename);
1423 memcpy(ptr, tmp, len);
1431 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1433 asmlinkage long sys_gethostname(char __user *name, int len)
1440 down_read(&uts_sem);
1441 ptr = vx_new_uts(nodename);
1442 i = 1 + strlen(ptr);
1446 if (copy_to_user(name, ptr, i))
1455 * Only setdomainname; getdomainname can be implemented by calling
1458 asmlinkage long sys_setdomainname(char __user *name, int len)
1461 char tmp[__NEW_UTS_LEN];
1463 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1465 if (len < 0 || len > __NEW_UTS_LEN)
1468 down_write(&uts_sem);
1470 if (!copy_from_user(tmp, name, len)) {
1471 char *ptr = vx_new_uts(domainname);
1473 memcpy(ptr, tmp, len);
1481 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1483 if (resource >= RLIM_NLIMITS)
1486 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1490 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1493 * Back compatibility for getrlimit. Needed for some apps.
1496 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1499 if (resource >= RLIM_NLIMITS)
1502 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1503 if(x.rlim_cur > 0x7FFFFFFF)
1504 x.rlim_cur = 0x7FFFFFFF;
1505 if(x.rlim_max > 0x7FFFFFFF)
1506 x.rlim_max = 0x7FFFFFFF;
1507 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1512 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1514 struct rlimit new_rlim, *old_rlim;
1517 if (resource >= RLIM_NLIMITS)
1519 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1521 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1523 old_rlim = current->rlim + resource;
1524 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1525 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1526 !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1528 if (resource == RLIMIT_NOFILE) {
1529 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1533 retval = security_task_setrlimit(resource, &new_rlim);
1537 *old_rlim = new_rlim;
1542 * It would make sense to put struct rusage in the task_struct,
1543 * except that would make the task_struct be *really big*. After
1544 * task_struct gets moved into malloc'ed memory, it would
1545 * make sense to do this. It will make moving the rest of the information
1546 * a lot simpler! (Which we're not doing right now because we're not
1547 * measuring them yet).
1549 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1550 * below (we know we aren't going to exit/disappear and only we change our
1551 * rusage counters), or we are called from wait4() on a process which is
1552 * either stopped or zombied. In the zombied case the task won't get
1553 * reaped till shortly after the call to getrusage(), in both cases the
1554 * task being examined is in a frozen state so the counters won't change.
1556 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1560 memset((char *) &r, 0, sizeof(r));
1563 jiffies_to_timeval(p->utime, &r.ru_utime);
1564 jiffies_to_timeval(p->stime, &r.ru_stime);
1565 r.ru_nvcsw = p->nvcsw;
1566 r.ru_nivcsw = p->nivcsw;
1567 r.ru_minflt = p->min_flt;
1568 r.ru_majflt = p->maj_flt;
1570 case RUSAGE_CHILDREN:
1571 jiffies_to_timeval(p->cutime, &r.ru_utime);
1572 jiffies_to_timeval(p->cstime, &r.ru_stime);
1573 r.ru_nvcsw = p->cnvcsw;
1574 r.ru_nivcsw = p->cnivcsw;
1575 r.ru_minflt = p->cmin_flt;
1576 r.ru_majflt = p->cmaj_flt;
1579 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1580 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1581 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1582 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1583 r.ru_minflt = p->min_flt + p->cmin_flt;
1584 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1587 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1590 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1592 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1594 return getrusage(current, who, ru);
1597 asmlinkage long sys_umask(int mask)
1599 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1603 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1604 unsigned long arg4, unsigned long arg5)
1609 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1614 case PR_SET_PDEATHSIG:
1616 if (sig < 0 || sig > _NSIG) {
1620 current->pdeath_signal = sig;
1622 case PR_GET_PDEATHSIG:
1623 error = put_user(current->pdeath_signal, (int __user *)arg2);
1625 case PR_GET_DUMPABLE:
1626 if (current->mm->dumpable)
1629 case PR_SET_DUMPABLE:
1630 if (arg2 != 0 && arg2 != 1) {
1634 current->mm->dumpable = arg2;
1637 case PR_SET_UNALIGN:
1638 error = SET_UNALIGN_CTL(current, arg2);
1640 case PR_GET_UNALIGN:
1641 error = GET_UNALIGN_CTL(current, arg2);
1644 error = SET_FPEMU_CTL(current, arg2);
1647 error = GET_FPEMU_CTL(current, arg2);
1650 error = SET_FPEXC_CTL(current, arg2);
1653 error = GET_FPEXC_CTL(current, arg2);
1656 error = PR_TIMING_STATISTICAL;
1659 if (arg2 == PR_TIMING_STATISTICAL)
1665 case PR_GET_KEEPCAPS:
1666 if (current->keep_capabilities)
1669 case PR_SET_KEEPCAPS:
1670 if (arg2 != 0 && arg2 != 1) {
1674 current->keep_capabilities = arg2;