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/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/times.h>
23 #include <linux/security.h>
24 #include <linux/dcookies.h>
25 #include <linux/suspend.h>
27 #include <asm/uaccess.h>
29 #include <asm/unistd.h>
31 #ifndef SET_UNALIGN_CTL
32 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
34 #ifndef GET_UNALIGN_CTL
35 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
38 # define SET_FPEMU_CTL(a,b) (-EINVAL)
41 # define GET_FPEMU_CTL(a,b) (-EINVAL)
44 # define SET_FPEXC_CTL(a,b) (-EINVAL)
47 # define GET_FPEXC_CTL(a,b) (-EINVAL)
51 * this is where the system-wide overflow UID and GID are defined, for
52 * architectures that now have 32-bit UID/GID but didn't in the past
55 int overflowuid = DEFAULT_OVERFLOWUID;
56 int overflowgid = DEFAULT_OVERFLOWGID;
59 EXPORT_SYMBOL(overflowuid);
60 EXPORT_SYMBOL(overflowgid);
64 * the same as above, but for filesystems which can only store a 16-bit
65 * UID and GID. as such, this is needed on all architectures
68 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
69 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
71 EXPORT_SYMBOL(fs_overflowuid);
72 EXPORT_SYMBOL(fs_overflowgid);
75 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
82 * Notifier list for kernel code which wants to be called
83 * at shutdown. This is used to stop any idling DMA operations
87 static struct notifier_block *reboot_notifier_list;
88 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
91 * notifier_chain_register - Add notifier to a notifier chain
92 * @list: Pointer to root list pointer
93 * @n: New entry in notifier chain
95 * Adds a notifier to a notifier chain.
97 * Currently always returns zero.
100 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
102 write_lock(¬ifier_lock);
105 if(n->priority > (*list)->priority)
107 list= &((*list)->next);
111 write_unlock(¬ifier_lock);
115 EXPORT_SYMBOL(notifier_chain_register);
118 * notifier_chain_unregister - Remove notifier from a notifier chain
119 * @nl: Pointer to root list pointer
120 * @n: New entry in notifier chain
122 * Removes a notifier from a notifier chain.
124 * Returns zero on success, or %-ENOENT on failure.
127 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
129 write_lock(¬ifier_lock);
135 write_unlock(¬ifier_lock);
140 write_unlock(¬ifier_lock);
144 EXPORT_SYMBOL(notifier_chain_unregister);
147 * notifier_call_chain - Call functions in a notifier chain
148 * @n: Pointer to root pointer of notifier chain
149 * @val: Value passed unmodified to notifier function
150 * @v: Pointer passed unmodified to notifier function
152 * Calls each function in a notifier chain in turn.
154 * If the return value of the notifier can be and'd
155 * with %NOTIFY_STOP_MASK, then notifier_call_chain
156 * will return immediately, with the return value of
157 * the notifier function which halted execution.
158 * Otherwise, the return value is the return value
159 * of the last notifier function called.
162 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
165 struct notifier_block *nb = *n;
169 ret=nb->notifier_call(nb,val,v);
170 if(ret&NOTIFY_STOP_MASK)
179 EXPORT_SYMBOL(notifier_call_chain);
182 * register_reboot_notifier - Register function to be called at reboot time
183 * @nb: Info about notifier function to be called
185 * Registers a function with the list of functions
186 * to be called at reboot time.
188 * Currently always returns zero, as notifier_chain_register
189 * always returns zero.
192 int register_reboot_notifier(struct notifier_block * nb)
194 return notifier_chain_register(&reboot_notifier_list, nb);
197 EXPORT_SYMBOL(register_reboot_notifier);
200 * unregister_reboot_notifier - Unregister previously registered reboot notifier
201 * @nb: Hook to be unregistered
203 * Unregisters a previously registered reboot
206 * Returns zero on success, or %-ENOENT on failure.
209 int unregister_reboot_notifier(struct notifier_block * nb)
211 return notifier_chain_unregister(&reboot_notifier_list, nb);
214 EXPORT_SYMBOL(unregister_reboot_notifier);
216 asmlinkage long sys_ni_syscall(void)
221 cond_syscall(sys_nfsservctl)
222 cond_syscall(sys_quotactl)
223 cond_syscall(sys_acct)
224 cond_syscall(sys_lookup_dcookie)
225 cond_syscall(sys_swapon)
226 cond_syscall(sys_swapoff)
227 cond_syscall(sys_init_module)
228 cond_syscall(sys_delete_module)
229 cond_syscall(sys_socketpair)
230 cond_syscall(sys_bind)
231 cond_syscall(sys_listen)
232 cond_syscall(sys_accept)
233 cond_syscall(sys_connect)
234 cond_syscall(sys_getsockname)
235 cond_syscall(sys_getpeername)
236 cond_syscall(sys_sendto)
237 cond_syscall(sys_send)
238 cond_syscall(sys_recvfrom)
239 cond_syscall(sys_recv)
240 cond_syscall(sys_socket)
241 cond_syscall(sys_setsockopt)
242 cond_syscall(sys_getsockopt)
243 cond_syscall(sys_shutdown)
244 cond_syscall(sys_sendmsg)
245 cond_syscall(sys_recvmsg)
246 cond_syscall(sys_socketcall)
247 cond_syscall(sys_futex)
248 cond_syscall(compat_sys_futex)
249 cond_syscall(sys_epoll_create)
250 cond_syscall(sys_epoll_ctl)
251 cond_syscall(sys_epoll_wait)
252 cond_syscall(sys_semget)
253 cond_syscall(sys_semop)
254 cond_syscall(sys_semtimedop)
255 cond_syscall(sys_semctl)
256 cond_syscall(sys_msgget)
257 cond_syscall(sys_msgsnd)
258 cond_syscall(sys_msgrcv)
259 cond_syscall(sys_msgctl)
260 cond_syscall(sys_shmget)
261 cond_syscall(sys_shmdt)
262 cond_syscall(sys_shmctl)
263 cond_syscall(sys_mq_open)
264 cond_syscall(sys_mq_unlink)
265 cond_syscall(sys_mq_timedsend)
266 cond_syscall(sys_mq_timedreceive)
267 cond_syscall(sys_mq_notify)
268 cond_syscall(sys_mq_getsetattr)
269 cond_syscall(compat_sys_mq_open)
270 cond_syscall(compat_sys_mq_timedsend)
271 cond_syscall(compat_sys_mq_timedreceive)
272 cond_syscall(compat_sys_mq_notify)
273 cond_syscall(compat_sys_mq_getsetattr)
274 cond_syscall(sys_mbind)
275 cond_syscall(sys_get_mempolicy)
276 cond_syscall(sys_set_mempolicy)
277 cond_syscall(compat_get_mempolicy)
279 /* arch-specific weak syscall entries */
280 cond_syscall(sys_pciconfig_read)
281 cond_syscall(sys_pciconfig_write)
282 cond_syscall(sys_pciconfig_iobase)
284 static int set_one_prio(struct task_struct *p, int niceval, int error)
288 if (p->uid != current->euid &&
289 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
293 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
297 no_nice = security_task_setnice(p, niceval);
304 set_user_nice(p, niceval);
309 asmlinkage long sys_setpriority(int which, int who, int niceval)
311 struct task_struct *g, *p;
312 struct user_struct *user;
317 if (which > 2 || which < 0)
320 /* normalize: avoid signed division (rounding problems) */
327 read_lock(&tasklist_lock);
332 p = find_task_by_pid(who);
334 error = set_one_prio(p, niceval, error);
338 who = process_group(current);
339 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
340 error = set_one_prio(p, niceval, error);
344 user = current->user;
346 user = find_user(who);
353 error = set_one_prio(p, niceval, error);
354 while_each_thread(g, p);
356 free_uid(user); /* For find_user() */
360 read_unlock(&tasklist_lock);
366 * Ugh. To avoid negative return values, "getpriority()" will
367 * not return the normal nice-value, but a negated value that
368 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
369 * to stay compatible.
371 asmlinkage long sys_getpriority(int which, int who)
373 struct task_struct *g, *p;
376 struct user_struct *user;
377 long niceval, retval = -ESRCH;
379 if (which > 2 || which < 0)
382 read_lock(&tasklist_lock);
387 p = find_task_by_pid(who);
389 niceval = 20 - task_nice(p);
390 if (niceval > retval)
396 who = process_group(current);
397 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
398 niceval = 20 - task_nice(p);
399 if (niceval > retval)
405 user = current->user;
407 user = find_user(who);
414 niceval = 20 - task_nice(p);
415 if (niceval > retval)
418 while_each_thread(g, p);
420 free_uid(user); /* for find_user() */
424 read_unlock(&tasklist_lock);
431 * Reboot system call: for obvious reasons only root may call it,
432 * and even root needs to set up some magic numbers in the registers
433 * so that some mistake won't make this reboot the whole machine.
434 * You can also set the meaning of the ctrl-alt-del-key here.
436 * reboot doesn't sync: do that yourself before calling this.
438 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
442 /* We only trust the superuser with rebooting the system. */
443 if (!capable(CAP_SYS_BOOT))
446 /* For safety, we require "magic" arguments. */
447 if (magic1 != LINUX_REBOOT_MAGIC1 ||
448 (magic2 != LINUX_REBOOT_MAGIC2 &&
449 magic2 != LINUX_REBOOT_MAGIC2A &&
450 magic2 != LINUX_REBOOT_MAGIC2B &&
451 magic2 != LINUX_REBOOT_MAGIC2C))
456 case LINUX_REBOOT_CMD_RESTART:
457 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
458 system_state = SYSTEM_RESTART;
460 printk(KERN_EMERG "Restarting system.\n");
461 machine_restart(NULL);
464 case LINUX_REBOOT_CMD_CAD_ON:
468 case LINUX_REBOOT_CMD_CAD_OFF:
472 case LINUX_REBOOT_CMD_HALT:
473 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
474 system_state = SYSTEM_HALT;
476 printk(KERN_EMERG "System halted.\n");
482 case LINUX_REBOOT_CMD_POWER_OFF:
483 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
484 system_state = SYSTEM_POWER_OFF;
486 printk(KERN_EMERG "Power down.\n");
492 case LINUX_REBOOT_CMD_RESTART2:
493 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
497 buffer[sizeof(buffer) - 1] = '\0';
499 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
500 system_state = SYSTEM_RESTART;
502 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
503 machine_restart(buffer);
506 #ifdef CONFIG_SOFTWARE_SUSPEND
507 case LINUX_REBOOT_CMD_SW_SUSPEND:
509 int ret = software_suspend();
523 static void deferred_cad(void *dummy)
525 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
526 machine_restart(NULL);
530 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
531 * As it's called within an interrupt, it may NOT sync: the only choice
532 * is whether to reboot at once, or just ignore the ctrl-alt-del.
534 void ctrl_alt_del(void)
536 static DECLARE_WORK(cad_work, deferred_cad, NULL);
539 schedule_work(&cad_work);
541 kill_proc(cad_pid, SIGINT, 1);
546 * Unprivileged users may change the real gid to the effective gid
547 * or vice versa. (BSD-style)
549 * If you set the real gid at all, or set the effective gid to a value not
550 * equal to the real gid, then the saved gid is set to the new effective gid.
552 * This makes it possible for a setgid program to completely drop its
553 * privileges, which is often a useful assertion to make when you are doing
554 * a security audit over a program.
556 * The general idea is that a program which uses just setregid() will be
557 * 100% compatible with BSD. A program which uses just setgid() will be
558 * 100% compatible with POSIX with saved IDs.
560 * SMP: There are not races, the GIDs are checked only by filesystem
561 * operations (as far as semantic preservation is concerned).
563 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
565 int old_rgid = current->gid;
566 int old_egid = current->egid;
567 int new_rgid = old_rgid;
568 int new_egid = old_egid;
571 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
575 if (rgid != (gid_t) -1) {
576 if ((old_rgid == rgid) ||
577 (current->egid==rgid) ||
583 if (egid != (gid_t) -1) {
584 if ((old_rgid == egid) ||
585 (current->egid == egid) ||
586 (current->sgid == egid) ||
593 if (new_egid != old_egid)
595 current->mm->dumpable = 0;
598 if (rgid != (gid_t) -1 ||
599 (egid != (gid_t) -1 && egid != old_rgid))
600 current->sgid = new_egid;
601 current->fsgid = new_egid;
602 current->egid = new_egid;
603 current->gid = new_rgid;
608 * setgid() is implemented like SysV w/ SAVED_IDS
610 * SMP: Same implicit races as above.
612 asmlinkage long sys_setgid(gid_t gid)
614 int old_egid = current->egid;
617 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
621 if (capable(CAP_SETGID))
625 current->mm->dumpable=0;
628 current->gid = current->egid = current->sgid = current->fsgid = gid;
630 else if ((gid == current->gid) || (gid == current->sgid))
634 current->mm->dumpable=0;
637 current->egid = current->fsgid = gid;
644 static int set_user(uid_t new_ruid, int dumpclear)
646 struct user_struct *new_user;
648 new_user = alloc_uid(new_ruid);
652 if (atomic_read(&new_user->processes) >=
653 current->rlim[RLIMIT_NPROC].rlim_cur &&
654 new_user != &root_user) {
659 switch_uid(new_user);
663 current->mm->dumpable = 0;
666 current->uid = new_ruid;
671 * Unprivileged users may change the real uid to the effective uid
672 * or vice versa. (BSD-style)
674 * If you set the real uid at all, or set the effective uid to a value not
675 * equal to the real uid, then the saved uid is set to the new effective uid.
677 * This makes it possible for a setuid program to completely drop its
678 * privileges, which is often a useful assertion to make when you are doing
679 * a security audit over a program.
681 * The general idea is that a program which uses just setreuid() will be
682 * 100% compatible with BSD. A program which uses just setuid() will be
683 * 100% compatible with POSIX with saved IDs.
685 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
687 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
690 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
694 new_ruid = old_ruid = current->uid;
695 new_euid = old_euid = current->euid;
696 old_suid = current->suid;
698 if (ruid != (uid_t) -1) {
700 if ((old_ruid != ruid) &&
701 (current->euid != ruid) &&
702 !capable(CAP_SETUID))
706 if (euid != (uid_t) -1) {
708 if ((old_ruid != euid) &&
709 (current->euid != euid) &&
710 (current->suid != euid) &&
711 !capable(CAP_SETUID))
715 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
718 if (new_euid != old_euid)
720 current->mm->dumpable=0;
723 current->fsuid = current->euid = new_euid;
724 if (ruid != (uid_t) -1 ||
725 (euid != (uid_t) -1 && euid != old_ruid))
726 current->suid = current->euid;
727 current->fsuid = current->euid;
729 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
735 * setuid() is implemented like SysV with SAVED_IDS
737 * Note that SAVED_ID's is deficient in that a setuid root program
738 * like sendmail, for example, cannot set its uid to be a normal
739 * user and then switch back, because if you're root, setuid() sets
740 * the saved uid too. If you don't like this, blame the bright people
741 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
742 * will allow a root program to temporarily drop privileges and be able to
743 * regain them by swapping the real and effective uid.
745 asmlinkage long sys_setuid(uid_t uid)
747 int old_euid = current->euid;
748 int old_ruid, old_suid, new_ruid, new_suid;
751 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
755 old_ruid = new_ruid = current->uid;
756 old_suid = current->suid;
759 if (capable(CAP_SETUID)) {
760 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
763 } else if ((uid != current->uid) && (uid != new_suid))
768 current->mm->dumpable = 0;
771 current->fsuid = current->euid = uid;
772 current->suid = new_suid;
774 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
779 * This function implements a generic ability to update ruid, euid,
780 * and suid. This allows you to implement the 4.4 compatible seteuid().
782 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
784 int old_ruid = current->uid;
785 int old_euid = current->euid;
786 int old_suid = current->suid;
789 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
793 if (!capable(CAP_SETUID)) {
794 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
795 (ruid != current->euid) && (ruid != current->suid))
797 if ((euid != (uid_t) -1) && (euid != current->uid) &&
798 (euid != current->euid) && (euid != current->suid))
800 if ((suid != (uid_t) -1) && (suid != current->uid) &&
801 (suid != current->euid) && (suid != current->suid))
804 if (ruid != (uid_t) -1) {
805 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
808 if (euid != (uid_t) -1) {
809 if (euid != current->euid)
811 current->mm->dumpable = 0;
814 current->euid = euid;
816 current->fsuid = current->euid;
817 if (suid != (uid_t) -1)
818 current->suid = suid;
820 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
823 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
827 if (!(retval = put_user(current->uid, ruid)) &&
828 !(retval = put_user(current->euid, euid)))
829 retval = put_user(current->suid, suid);
835 * Same as above, but for rgid, egid, sgid.
837 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
841 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
845 if (!capable(CAP_SETGID)) {
846 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
847 (rgid != current->egid) && (rgid != current->sgid))
849 if ((egid != (gid_t) -1) && (egid != current->gid) &&
850 (egid != current->egid) && (egid != current->sgid))
852 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
853 (sgid != current->egid) && (sgid != current->sgid))
856 if (egid != (gid_t) -1) {
857 if (egid != current->egid)
859 current->mm->dumpable = 0;
862 current->egid = egid;
864 current->fsgid = current->egid;
865 if (rgid != (gid_t) -1)
867 if (sgid != (gid_t) -1)
868 current->sgid = sgid;
872 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
876 if (!(retval = put_user(current->gid, rgid)) &&
877 !(retval = put_user(current->egid, egid)))
878 retval = put_user(current->sgid, sgid);
885 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
886 * is used for "access()" and for the NFS daemon (letting nfsd stay at
887 * whatever uid it wants to). It normally shadows "euid", except when
888 * explicitly set by setfsuid() or for access..
890 asmlinkage long sys_setfsuid(uid_t uid)
894 old_fsuid = current->fsuid;
895 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
898 if (uid == current->uid || uid == current->euid ||
899 uid == current->suid || uid == current->fsuid ||
902 if (uid != old_fsuid)
904 current->mm->dumpable = 0;
907 current->fsuid = uid;
910 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
916 * Samma på svenska..
918 asmlinkage long sys_setfsgid(gid_t gid)
922 old_fsgid = current->fsgid;
923 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
926 if (gid == current->gid || gid == current->egid ||
927 gid == current->sgid || gid == current->fsgid ||
930 if (gid != old_fsgid)
932 current->mm->dumpable = 0;
935 current->fsgid = gid;
940 asmlinkage long sys_times(struct tms __user * tbuf)
943 * In the SMP world we might just be unlucky and have one of
944 * the times increment as we use it. Since the value is an
945 * atomically safe type this is just fine. Conceptually its
946 * as if the syscall took an instant longer to occur.
950 tmp.tms_utime = jiffies_to_clock_t(current->utime);
951 tmp.tms_stime = jiffies_to_clock_t(current->stime);
952 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
953 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
954 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
957 return (long) jiffies_64_to_clock_t(get_jiffies_64());
961 * This needs some heavy checking ...
962 * I just haven't the stomach for it. I also don't fully
963 * understand sessions/pgrp etc. Let somebody who does explain it.
965 * OK, I think I have the protection semantics right.... this is really
966 * only important on a multi-user system anyway, to make sure one user
967 * can't send a signal to a process owned by another. -TYT, 12/12/91
969 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
973 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
975 struct task_struct *p;
985 /* From this point forward we keep holding onto the tasklist lock
986 * so that our parent does not change from under us. -DaveM
988 write_lock_irq(&tasklist_lock);
991 p = find_task_by_pid(pid);
996 if (!thread_group_leader(p))
999 if (p->parent == current || p->real_parent == current) {
1001 if (p->signal->session != current->signal->session)
1013 if (p->signal->leader)
1017 struct task_struct *p;
1019 struct list_head *l;
1021 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1022 if (p->signal->session == current->signal->session)
1028 err = security_task_setpgid(p, pgid);
1032 if (process_group(p) != pgid) {
1033 detach_pid(p, PIDTYPE_PGID);
1034 p->signal->pgrp = pgid;
1035 attach_pid(p, PIDTYPE_PGID, pgid);
1040 /* All paths lead to here, thus we are safe. -DaveM */
1041 write_unlock_irq(&tasklist_lock);
1045 asmlinkage long sys_getpgid(pid_t pid)
1048 return process_group(current);
1051 struct task_struct *p;
1053 read_lock(&tasklist_lock);
1054 p = find_task_by_pid(pid);
1058 retval = security_task_getpgid(p);
1060 retval = process_group(p);
1062 read_unlock(&tasklist_lock);
1067 #ifdef __ARCH_WANT_SYS_GETPGRP
1069 asmlinkage long sys_getpgrp(void)
1071 /* SMP - assuming writes are word atomic this is fine */
1072 return process_group(current);
1077 asmlinkage long sys_getsid(pid_t pid)
1080 return current->signal->session;
1083 struct task_struct *p;
1085 read_lock(&tasklist_lock);
1086 p = find_task_by_pid(pid);
1090 retval = security_task_getsid(p);
1092 retval = p->signal->session;
1094 read_unlock(&tasklist_lock);
1099 asmlinkage long sys_setsid(void)
1104 if (!thread_group_leader(current))
1107 write_lock_irq(&tasklist_lock);
1109 pid = find_pid(PIDTYPE_PGID, current->pid);
1113 current->signal->leader = 1;
1114 __set_special_pids(current->pid, current->pid);
1115 current->signal->tty = NULL;
1116 current->signal->tty_old_pgrp = 0;
1117 err = process_group(current);
1119 write_unlock_irq(&tasklist_lock);
1124 * Supplementary group IDs
1127 /* init to 2 - one for init_task, one to ensure it is never freed */
1128 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1130 struct group_info *groups_alloc(int gidsetsize)
1132 struct group_info *group_info;
1136 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1137 /* Make sure we always allocate at least one indirect block pointer */
1138 nblocks = nblocks ? : 1;
1139 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1142 group_info->ngroups = gidsetsize;
1143 group_info->nblocks = nblocks;
1144 atomic_set(&group_info->usage, 1);
1146 if (gidsetsize <= NGROUPS_SMALL) {
1147 group_info->blocks[0] = group_info->small_block;
1149 for (i = 0; i < nblocks; i++) {
1151 b = (void *)__get_free_page(GFP_USER);
1153 goto out_undo_partial_alloc;
1154 group_info->blocks[i] = b;
1159 out_undo_partial_alloc:
1161 free_page((unsigned long)group_info->blocks[i]);
1167 EXPORT_SYMBOL(groups_alloc);
1169 void groups_free(struct group_info *group_info)
1171 if (group_info->blocks[0] != group_info->small_block) {
1173 for (i = 0; i < group_info->nblocks; i++)
1174 free_page((unsigned long)group_info->blocks[i]);
1179 EXPORT_SYMBOL(groups_free);
1181 /* export the group_info to a user-space array */
1182 static int groups_to_user(gid_t __user *grouplist,
1183 struct group_info *group_info)
1186 int count = group_info->ngroups;
1188 for (i = 0; i < group_info->nblocks; i++) {
1189 int cp_count = min(NGROUPS_PER_BLOCK, count);
1190 int off = i * NGROUPS_PER_BLOCK;
1191 int len = cp_count * sizeof(*grouplist);
1193 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1201 /* fill a group_info from a user-space array - it must be allocated already */
1202 static int groups_from_user(struct group_info *group_info,
1203 gid_t __user *grouplist)
1206 int count = group_info->ngroups;
1208 for (i = 0; i < group_info->nblocks; i++) {
1209 int cp_count = min(NGROUPS_PER_BLOCK, count);
1210 int off = i * NGROUPS_PER_BLOCK;
1211 int len = cp_count * sizeof(*grouplist);
1213 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1221 /* a simple shell-metzner sort */
1222 static void groups_sort(struct group_info *group_info)
1224 int base, max, stride;
1225 int gidsetsize = group_info->ngroups;
1227 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1232 max = gidsetsize - stride;
1233 for (base = 0; base < max; base++) {
1235 int right = left + stride;
1236 gid_t tmp = GROUP_AT(group_info, right);
1238 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1239 GROUP_AT(group_info, right) =
1240 GROUP_AT(group_info, left);
1244 GROUP_AT(group_info, right) = tmp;
1250 /* a simple bsearch */
1251 static int groups_search(struct group_info *group_info, gid_t grp)
1259 right = group_info->ngroups;
1260 while (left < right) {
1261 int mid = (left+right)/2;
1262 int cmp = grp - GROUP_AT(group_info, mid);
1273 /* validate and set current->group_info */
1274 int set_current_groups(struct group_info *group_info)
1277 struct group_info *old_info;
1279 retval = security_task_setgroups(group_info);
1283 groups_sort(group_info);
1284 get_group_info(group_info);
1287 old_info = current->group_info;
1288 current->group_info = group_info;
1289 task_unlock(current);
1291 put_group_info(old_info);
1296 EXPORT_SYMBOL(set_current_groups);
1298 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1303 * SMP: Nobody else can change our grouplist. Thus we are
1310 /* no need to grab task_lock here; it cannot change */
1311 get_group_info(current->group_info);
1312 i = current->group_info->ngroups;
1314 if (i > gidsetsize) {
1318 if (groups_to_user(grouplist, current->group_info)) {
1324 put_group_info(current->group_info);
1329 * SMP: Our groups are copy-on-write. We can set them safely
1330 * without another task interfering.
1333 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1335 struct group_info *group_info;
1338 if (!capable(CAP_SETGID))
1340 if ((unsigned)gidsetsize > NGROUPS_MAX)
1343 group_info = groups_alloc(gidsetsize);
1346 retval = groups_from_user(group_info, grouplist);
1348 put_group_info(group_info);
1352 retval = set_current_groups(group_info);
1353 put_group_info(group_info);
1359 * Check whether we're fsgid/egid or in the supplemental group..
1361 int in_group_p(gid_t grp)
1364 if (grp != current->fsgid) {
1365 get_group_info(current->group_info);
1366 retval = groups_search(current->group_info, grp);
1367 put_group_info(current->group_info);
1372 EXPORT_SYMBOL(in_group_p);
1374 int in_egroup_p(gid_t grp)
1377 if (grp != current->egid) {
1378 get_group_info(current->group_info);
1379 retval = groups_search(current->group_info, grp);
1380 put_group_info(current->group_info);
1385 EXPORT_SYMBOL(in_egroup_p);
1387 DECLARE_RWSEM(uts_sem);
1389 EXPORT_SYMBOL(uts_sem);
1391 asmlinkage long sys_newuname(struct new_utsname __user * name)
1395 down_read(&uts_sem);
1396 if (copy_to_user(name,&system_utsname,sizeof *name))
1402 asmlinkage long sys_sethostname(char __user *name, int len)
1405 char tmp[__NEW_UTS_LEN];
1407 if (!capable(CAP_SYS_ADMIN))
1409 if (len < 0 || len > __NEW_UTS_LEN)
1411 down_write(&uts_sem);
1413 if (!copy_from_user(tmp, name, len)) {
1414 memcpy(system_utsname.nodename, tmp, len);
1415 system_utsname.nodename[len] = 0;
1422 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1424 asmlinkage long sys_gethostname(char __user *name, int len)
1430 down_read(&uts_sem);
1431 i = 1 + strlen(system_utsname.nodename);
1435 if (copy_to_user(name, system_utsname.nodename, i))
1444 * Only setdomainname; getdomainname can be implemented by calling
1447 asmlinkage long sys_setdomainname(char __user *name, int len)
1450 char tmp[__NEW_UTS_LEN];
1452 if (!capable(CAP_SYS_ADMIN))
1454 if (len < 0 || len > __NEW_UTS_LEN)
1457 down_write(&uts_sem);
1459 if (!copy_from_user(tmp, name, len)) {
1460 memcpy(system_utsname.domainname, tmp, len);
1461 system_utsname.domainname[len] = 0;
1468 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1470 if (resource >= RLIM_NLIMITS)
1473 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1477 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1480 * Back compatibility for getrlimit. Needed for some apps.
1483 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1486 if (resource >= RLIM_NLIMITS)
1489 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1490 if(x.rlim_cur > 0x7FFFFFFF)
1491 x.rlim_cur = 0x7FFFFFFF;
1492 if(x.rlim_max > 0x7FFFFFFF)
1493 x.rlim_max = 0x7FFFFFFF;
1494 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1499 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1501 struct rlimit new_rlim, *old_rlim;
1504 if (resource >= RLIM_NLIMITS)
1506 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1508 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1510 old_rlim = current->rlim + resource;
1511 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1512 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1513 !capable(CAP_SYS_RESOURCE))
1515 if (resource == RLIMIT_NOFILE) {
1516 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1520 retval = security_task_setrlimit(resource, &new_rlim);
1524 *old_rlim = new_rlim;
1529 * It would make sense to put struct rusage in the task_struct,
1530 * except that would make the task_struct be *really big*. After
1531 * task_struct gets moved into malloc'ed memory, it would
1532 * make sense to do this. It will make moving the rest of the information
1533 * a lot simpler! (Which we're not doing right now because we're not
1534 * measuring them yet).
1536 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1537 * below (we know we aren't going to exit/disappear and only we change our
1538 * rusage counters), or we are called from wait4() on a process which is
1539 * either stopped or zombied. In the zombied case the task won't get
1540 * reaped till shortly after the call to getrusage(), in both cases the
1541 * task being examined is in a frozen state so the counters won't change.
1543 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1547 memset((char *) &r, 0, sizeof(r));
1550 jiffies_to_timeval(p->utime, &r.ru_utime);
1551 jiffies_to_timeval(p->stime, &r.ru_stime);
1552 r.ru_nvcsw = p->nvcsw;
1553 r.ru_nivcsw = p->nivcsw;
1554 r.ru_minflt = p->min_flt;
1555 r.ru_majflt = p->maj_flt;
1557 case RUSAGE_CHILDREN:
1558 jiffies_to_timeval(p->cutime, &r.ru_utime);
1559 jiffies_to_timeval(p->cstime, &r.ru_stime);
1560 r.ru_nvcsw = p->cnvcsw;
1561 r.ru_nivcsw = p->cnivcsw;
1562 r.ru_minflt = p->cmin_flt;
1563 r.ru_majflt = p->cmaj_flt;
1566 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1567 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1568 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1569 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1570 r.ru_minflt = p->min_flt + p->cmin_flt;
1571 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1574 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1577 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1579 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1581 return getrusage(current, who, ru);
1584 asmlinkage long sys_umask(int mask)
1586 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1590 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1591 unsigned long arg4, unsigned long arg5)
1596 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1601 case PR_SET_PDEATHSIG:
1603 if (sig < 0 || sig > _NSIG) {
1607 current->pdeath_signal = sig;
1609 case PR_GET_PDEATHSIG:
1610 error = put_user(current->pdeath_signal, (int __user *)arg2);
1612 case PR_GET_DUMPABLE:
1613 if (current->mm->dumpable)
1616 case PR_SET_DUMPABLE:
1617 if (arg2 != 0 && arg2 != 1) {
1621 current->mm->dumpable = arg2;
1624 case PR_SET_UNALIGN:
1625 error = SET_UNALIGN_CTL(current, arg2);
1627 case PR_GET_UNALIGN:
1628 error = GET_UNALIGN_CTL(current, arg2);
1631 error = SET_FPEMU_CTL(current, arg2);
1634 error = GET_FPEMU_CTL(current, arg2);
1637 error = SET_FPEXC_CTL(current, arg2);
1640 error = GET_FPEXC_CTL(current, arg2);
1643 error = PR_TIMING_STATISTICAL;
1646 if (arg2 == PR_TIMING_STATISTICAL)
1652 case PR_GET_KEEPCAPS:
1653 if (current->keep_capabilities)
1656 case PR_SET_KEEPCAPS:
1657 if (arg2 != 0 && arg2 != 1) {
1661 current->keep_capabilities = arg2;