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)
275 /* arch-specific weak syscall entries */
276 cond_syscall(sys_pciconfig_read)
277 cond_syscall(sys_pciconfig_write)
278 cond_syscall(sys_pciconfig_iobase)
280 static int set_one_prio(struct task_struct *p, int niceval, int error)
284 if (p->uid != current->euid &&
285 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
289 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
293 no_nice = security_task_setnice(p, niceval);
300 set_user_nice(p, niceval);
305 asmlinkage long sys_setpriority(int which, int who, int niceval)
307 struct task_struct *g, *p;
308 struct user_struct *user;
313 if (which > 2 || which < 0)
316 /* normalize: avoid signed division (rounding problems) */
323 read_lock(&tasklist_lock);
328 p = find_task_by_pid(who);
330 error = set_one_prio(p, niceval, error);
334 who = process_group(current);
335 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
336 error = set_one_prio(p, niceval, error);
340 user = current->user;
342 user = find_user(who);
349 error = set_one_prio(p, niceval, error);
350 while_each_thread(g, p);
354 read_unlock(&tasklist_lock);
360 * Ugh. To avoid negative return values, "getpriority()" will
361 * not return the normal nice-value, but a negated value that
362 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
363 * to stay compatible.
365 asmlinkage long sys_getpriority(int which, int who)
367 struct task_struct *g, *p;
370 struct user_struct *user;
371 long niceval, retval = -ESRCH;
373 if (which > 2 || which < 0)
376 read_lock(&tasklist_lock);
381 p = find_task_by_pid(who);
383 niceval = 20 - task_nice(p);
384 if (niceval > retval)
390 who = process_group(current);
391 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
392 niceval = 20 - task_nice(p);
393 if (niceval > retval)
399 user = current->user;
401 user = find_user(who);
408 niceval = 20 - task_nice(p);
409 if (niceval > retval)
412 while_each_thread(g, p);
416 read_unlock(&tasklist_lock);
423 * Reboot system call: for obvious reasons only root may call it,
424 * and even root needs to set up some magic numbers in the registers
425 * so that some mistake won't make this reboot the whole machine.
426 * You can also set the meaning of the ctrl-alt-del-key here.
428 * reboot doesn't sync: do that yourself before calling this.
430 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
434 /* We only trust the superuser with rebooting the system. */
435 if (!capable(CAP_SYS_BOOT))
438 /* For safety, we require "magic" arguments. */
439 if (magic1 != LINUX_REBOOT_MAGIC1 ||
440 (magic2 != LINUX_REBOOT_MAGIC2 &&
441 magic2 != LINUX_REBOOT_MAGIC2A &&
442 magic2 != LINUX_REBOOT_MAGIC2B &&
443 magic2 != LINUX_REBOOT_MAGIC2C))
448 case LINUX_REBOOT_CMD_RESTART:
449 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
450 system_state = SYSTEM_SHUTDOWN;
452 printk(KERN_EMERG "Restarting system.\n");
453 machine_restart(NULL);
456 case LINUX_REBOOT_CMD_CAD_ON:
460 case LINUX_REBOOT_CMD_CAD_OFF:
464 case LINUX_REBOOT_CMD_HALT:
465 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
466 system_state = SYSTEM_SHUTDOWN;
468 printk(KERN_EMERG "System halted.\n");
474 case LINUX_REBOOT_CMD_POWER_OFF:
475 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
476 system_state = SYSTEM_SHUTDOWN;
478 printk(KERN_EMERG "Power down.\n");
484 case LINUX_REBOOT_CMD_RESTART2:
485 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
489 buffer[sizeof(buffer) - 1] = '\0';
491 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
492 system_state = SYSTEM_SHUTDOWN;
494 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
495 machine_restart(buffer);
498 #ifdef CONFIG_SOFTWARE_SUSPEND
499 case LINUX_REBOOT_CMD_SW_SUSPEND:
501 int ret = software_suspend();
515 static void deferred_cad(void *dummy)
517 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
518 machine_restart(NULL);
522 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
523 * As it's called within an interrupt, it may NOT sync: the only choice
524 * is whether to reboot at once, or just ignore the ctrl-alt-del.
526 void ctrl_alt_del(void)
528 static DECLARE_WORK(cad_work, deferred_cad, NULL);
531 schedule_work(&cad_work);
533 kill_proc(cad_pid, SIGINT, 1);
538 * Unprivileged users may change the real gid to the effective gid
539 * or vice versa. (BSD-style)
541 * If you set the real gid at all, or set the effective gid to a value not
542 * equal to the real gid, then the saved gid is set to the new effective gid.
544 * This makes it possible for a setgid program to completely drop its
545 * privileges, which is often a useful assertion to make when you are doing
546 * a security audit over a program.
548 * The general idea is that a program which uses just setregid() will be
549 * 100% compatible with BSD. A program which uses just setgid() will be
550 * 100% compatible with POSIX with saved IDs.
552 * SMP: There are not races, the GIDs are checked only by filesystem
553 * operations (as far as semantic preservation is concerned).
555 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
557 int old_rgid = current->gid;
558 int old_egid = current->egid;
559 int new_rgid = old_rgid;
560 int new_egid = old_egid;
563 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
567 if (rgid != (gid_t) -1) {
568 if ((old_rgid == rgid) ||
569 (current->egid==rgid) ||
575 if (egid != (gid_t) -1) {
576 if ((old_rgid == egid) ||
577 (current->egid == egid) ||
578 (current->sgid == egid) ||
585 if (new_egid != old_egid)
587 current->mm->dumpable = 0;
590 if (rgid != (gid_t) -1 ||
591 (egid != (gid_t) -1 && egid != old_rgid))
592 current->sgid = new_egid;
593 current->fsgid = new_egid;
594 current->egid = new_egid;
595 current->gid = new_rgid;
600 * setgid() is implemented like SysV w/ SAVED_IDS
602 * SMP: Same implicit races as above.
604 asmlinkage long sys_setgid(gid_t gid)
606 int old_egid = current->egid;
609 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
613 if (capable(CAP_SETGID))
617 current->mm->dumpable=0;
620 current->gid = current->egid = current->sgid = current->fsgid = gid;
622 else if ((gid == current->gid) || (gid == current->sgid))
626 current->mm->dumpable=0;
629 current->egid = current->fsgid = gid;
636 static int set_user(uid_t new_ruid, int dumpclear)
638 struct user_struct *new_user;
640 new_user = alloc_uid(new_ruid);
644 if (atomic_read(&new_user->processes) >=
645 current->rlim[RLIMIT_NPROC].rlim_cur &&
646 new_user != &root_user) {
651 switch_uid(new_user);
655 current->mm->dumpable = 0;
658 current->uid = new_ruid;
663 * Unprivileged users may change the real uid to the effective uid
664 * or vice versa. (BSD-style)
666 * If you set the real uid at all, or set the effective uid to a value not
667 * equal to the real uid, then the saved uid is set to the new effective uid.
669 * This makes it possible for a setuid program to completely drop its
670 * privileges, which is often a useful assertion to make when you are doing
671 * a security audit over a program.
673 * The general idea is that a program which uses just setreuid() will be
674 * 100% compatible with BSD. A program which uses just setuid() will be
675 * 100% compatible with POSIX with saved IDs.
677 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
679 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
682 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
686 new_ruid = old_ruid = current->uid;
687 new_euid = old_euid = current->euid;
688 old_suid = current->suid;
690 if (ruid != (uid_t) -1) {
692 if ((old_ruid != ruid) &&
693 (current->euid != ruid) &&
694 !capable(CAP_SETUID))
698 if (euid != (uid_t) -1) {
700 if ((old_ruid != euid) &&
701 (current->euid != euid) &&
702 (current->suid != euid) &&
703 !capable(CAP_SETUID))
707 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
710 if (new_euid != old_euid)
712 current->mm->dumpable=0;
715 current->fsuid = current->euid = new_euid;
716 if (ruid != (uid_t) -1 ||
717 (euid != (uid_t) -1 && euid != old_ruid))
718 current->suid = current->euid;
719 current->fsuid = current->euid;
721 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
727 * setuid() is implemented like SysV with SAVED_IDS
729 * Note that SAVED_ID's is deficient in that a setuid root program
730 * like sendmail, for example, cannot set its uid to be a normal
731 * user and then switch back, because if you're root, setuid() sets
732 * the saved uid too. If you don't like this, blame the bright people
733 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
734 * will allow a root program to temporarily drop privileges and be able to
735 * regain them by swapping the real and effective uid.
737 asmlinkage long sys_setuid(uid_t uid)
739 int old_euid = current->euid;
740 int old_ruid, old_suid, new_ruid, new_suid;
743 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
747 old_ruid = new_ruid = current->uid;
748 old_suid = current->suid;
751 if (capable(CAP_SETUID)) {
752 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
755 } else if ((uid != current->uid) && (uid != new_suid))
760 current->mm->dumpable = 0;
763 current->fsuid = current->euid = uid;
764 current->suid = new_suid;
766 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
771 * This function implements a generic ability to update ruid, euid,
772 * and suid. This allows you to implement the 4.4 compatible seteuid().
774 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
776 int old_ruid = current->uid;
777 int old_euid = current->euid;
778 int old_suid = current->suid;
781 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
785 if (!capable(CAP_SETUID)) {
786 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
787 (ruid != current->euid) && (ruid != current->suid))
789 if ((euid != (uid_t) -1) && (euid != current->uid) &&
790 (euid != current->euid) && (euid != current->suid))
792 if ((suid != (uid_t) -1) && (suid != current->uid) &&
793 (suid != current->euid) && (suid != current->suid))
796 if (ruid != (uid_t) -1) {
797 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
800 if (euid != (uid_t) -1) {
801 if (euid != current->euid)
803 current->mm->dumpable = 0;
806 current->euid = euid;
808 current->fsuid = current->euid;
809 if (suid != (uid_t) -1)
810 current->suid = suid;
812 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
815 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
819 if (!(retval = put_user(current->uid, ruid)) &&
820 !(retval = put_user(current->euid, euid)))
821 retval = put_user(current->suid, suid);
827 * Same as above, but for rgid, egid, sgid.
829 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
833 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
837 if (!capable(CAP_SETGID)) {
838 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
839 (rgid != current->egid) && (rgid != current->sgid))
841 if ((egid != (gid_t) -1) && (egid != current->gid) &&
842 (egid != current->egid) && (egid != current->sgid))
844 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
845 (sgid != current->egid) && (sgid != current->sgid))
848 if (egid != (gid_t) -1) {
849 if (egid != current->egid)
851 current->mm->dumpable = 0;
854 current->egid = egid;
856 current->fsgid = current->egid;
857 if (rgid != (gid_t) -1)
859 if (sgid != (gid_t) -1)
860 current->sgid = sgid;
864 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
868 if (!(retval = put_user(current->gid, rgid)) &&
869 !(retval = put_user(current->egid, egid)))
870 retval = put_user(current->sgid, sgid);
877 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
878 * is used for "access()" and for the NFS daemon (letting nfsd stay at
879 * whatever uid it wants to). It normally shadows "euid", except when
880 * explicitly set by setfsuid() or for access..
882 asmlinkage long sys_setfsuid(uid_t uid)
886 old_fsuid = current->fsuid;
887 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
890 if (uid == current->uid || uid == current->euid ||
891 uid == current->suid || uid == current->fsuid ||
894 if (uid != old_fsuid)
896 current->mm->dumpable = 0;
899 current->fsuid = uid;
902 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
908 * Samma på svenska..
910 asmlinkage long sys_setfsgid(gid_t gid)
914 old_fsgid = current->fsgid;
915 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
918 if (gid == current->gid || gid == current->egid ||
919 gid == current->sgid || gid == current->fsgid ||
922 if (gid != old_fsgid)
924 current->mm->dumpable = 0;
927 current->fsgid = gid;
932 asmlinkage long sys_times(struct tms __user * tbuf)
935 * In the SMP world we might just be unlucky and have one of
936 * the times increment as we use it. Since the value is an
937 * atomically safe type this is just fine. Conceptually its
938 * as if the syscall took an instant longer to occur.
942 tmp.tms_utime = jiffies_to_clock_t(current->utime);
943 tmp.tms_stime = jiffies_to_clock_t(current->stime);
944 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
945 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
946 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
949 return (long) jiffies_64_to_clock_t(get_jiffies_64());
953 * This needs some heavy checking ...
954 * I just haven't the stomach for it. I also don't fully
955 * understand sessions/pgrp etc. Let somebody who does explain it.
957 * OK, I think I have the protection semantics right.... this is really
958 * only important on a multi-user system anyway, to make sure one user
959 * can't send a signal to a process owned by another. -TYT, 12/12/91
961 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
965 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
967 struct task_struct *p;
977 /* From this point forward we keep holding onto the tasklist lock
978 * so that our parent does not change from under us. -DaveM
980 write_lock_irq(&tasklist_lock);
983 p = find_task_by_pid(pid);
988 if (!thread_group_leader(p))
991 if (p->parent == current || p->real_parent == current) {
993 if (p->signal->session != current->signal->session)
1005 if (p->signal->leader)
1009 struct task_struct *p;
1011 struct list_head *l;
1013 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1014 if (p->signal->session == current->signal->session)
1020 err = security_task_setpgid(p, pgid);
1024 if (process_group(p) != pgid) {
1025 detach_pid(p, PIDTYPE_PGID);
1026 p->signal->pgrp = pgid;
1027 attach_pid(p, PIDTYPE_PGID, pgid);
1032 /* All paths lead to here, thus we are safe. -DaveM */
1033 write_unlock_irq(&tasklist_lock);
1037 asmlinkage long sys_getpgid(pid_t pid)
1040 return process_group(current);
1043 struct task_struct *p;
1045 read_lock(&tasklist_lock);
1046 p = find_task_by_pid(pid);
1050 retval = security_task_getpgid(p);
1052 retval = process_group(p);
1054 read_unlock(&tasklist_lock);
1059 asmlinkage long sys_getpgrp(void)
1061 /* SMP - assuming writes are word atomic this is fine */
1062 return process_group(current);
1065 asmlinkage long sys_getsid(pid_t pid)
1068 return current->signal->session;
1071 struct task_struct *p;
1073 read_lock(&tasklist_lock);
1074 p = find_task_by_pid(pid);
1078 retval = security_task_getsid(p);
1080 retval = p->signal->session;
1082 read_unlock(&tasklist_lock);
1087 asmlinkage long sys_setsid(void)
1092 if (!thread_group_leader(current))
1095 write_lock_irq(&tasklist_lock);
1097 pid = find_pid(PIDTYPE_PGID, current->pid);
1101 current->signal->leader = 1;
1102 __set_special_pids(current->pid, current->pid);
1103 current->signal->tty = NULL;
1104 current->signal->tty_old_pgrp = 0;
1105 err = process_group(current);
1107 write_unlock_irq(&tasklist_lock);
1112 * Supplementary group IDs
1115 /* init to 2 - one for init_task, one to ensure it is never freed */
1116 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1118 struct group_info *groups_alloc(int gidsetsize)
1120 struct group_info *group_info;
1124 nblocks = (gidsetsize/NGROUPS_PER_BLOCK) +
1125 (gidsetsize%NGROUPS_PER_BLOCK?1:0);
1126 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *),
1130 group_info->ngroups = gidsetsize;
1131 group_info->nblocks = nblocks;
1132 atomic_set(&group_info->usage, 1);
1134 if (gidsetsize <= NGROUPS_SMALL) {
1135 group_info->blocks[0] = group_info->small_block;
1137 for (i = 0; i < nblocks; i++) {
1139 b = (void *)__get_free_page(GFP_USER);
1141 goto out_undo_partial_alloc;
1142 group_info->blocks[i] = b;
1147 out_undo_partial_alloc:
1149 free_page((unsigned long)group_info->blocks[i]);
1155 EXPORT_SYMBOL(groups_alloc);
1157 void groups_free(struct group_info *group_info)
1159 if (group_info->blocks[0] != group_info->small_block) {
1161 for (i = 0; i < group_info->nblocks; i++)
1162 free_page((unsigned long)group_info->blocks[i]);
1167 EXPORT_SYMBOL(groups_free);
1169 /* export the group_info to a user-space array */
1170 static int groups_to_user(gid_t __user *grouplist,
1171 struct group_info *group_info)
1174 int count = group_info->ngroups;
1176 for (i = 0; i < group_info->nblocks; i++) {
1177 int cp_count = min(NGROUPS_PER_BLOCK, count);
1178 int off = i * NGROUPS_PER_BLOCK;
1179 int len = cp_count * sizeof(*grouplist);
1181 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1189 /* fill a group_info from a user-space array - it must be allocated already */
1190 static int groups_from_user(struct group_info *group_info,
1191 gid_t __user *grouplist)
1194 int count = group_info->ngroups;
1196 for (i = 0; i < group_info->nblocks; i++) {
1197 int cp_count = min(NGROUPS_PER_BLOCK, count);
1198 int off = i * NGROUPS_PER_BLOCK;
1199 int len = cp_count * sizeof(*grouplist);
1201 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1209 /* a simple shell-metzner sort */
1210 static void groups_sort(struct group_info *group_info)
1212 int base, max, stride;
1213 int gidsetsize = group_info->ngroups;
1215 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1220 max = gidsetsize - stride;
1221 for (base = 0; base < max; base++) {
1223 int right = left + stride;
1224 gid_t tmp = GROUP_AT(group_info, right);
1226 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1227 GROUP_AT(group_info, right) =
1228 GROUP_AT(group_info, left);
1232 GROUP_AT(group_info, right) = tmp;
1238 /* a simple bsearch */
1239 static int groups_search(struct group_info *group_info, gid_t grp)
1247 right = group_info->ngroups;
1248 while (left < right) {
1249 int mid = (left+right)/2;
1250 int cmp = grp - GROUP_AT(group_info, mid);
1261 /* validate and set current->group_info */
1262 int set_current_groups(struct group_info *group_info)
1265 struct group_info *old_info;
1267 retval = security_task_setgroups(group_info);
1271 groups_sort(group_info);
1272 get_group_info(group_info);
1273 old_info = current->group_info;
1274 current->group_info = group_info;
1275 put_group_info(old_info);
1280 EXPORT_SYMBOL(set_current_groups);
1282 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1287 * SMP: Nobody else can change our grouplist. Thus we are
1294 get_group_info(current->group_info);
1295 i = current->group_info->ngroups;
1297 if (i > gidsetsize) {
1301 if (groups_to_user(grouplist, current->group_info)) {
1307 put_group_info(current->group_info);
1312 * SMP: Our groups are copy-on-write. We can set them safely
1313 * without another task interfering.
1316 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1318 struct group_info *group_info;
1321 if (!capable(CAP_SETGID))
1323 if ((unsigned)gidsetsize > NGROUPS_MAX)
1326 group_info = groups_alloc(gidsetsize);
1329 retval = groups_from_user(group_info, grouplist);
1331 put_group_info(group_info);
1335 retval = set_current_groups(group_info);
1336 put_group_info(group_info);
1342 * Check whether we're fsgid/egid or in the supplemental group..
1344 int in_group_p(gid_t grp)
1347 if (grp != current->fsgid) {
1348 get_group_info(current->group_info);
1349 retval = groups_search(current->group_info, grp);
1350 put_group_info(current->group_info);
1355 EXPORT_SYMBOL(in_group_p);
1357 int in_egroup_p(gid_t grp)
1360 if (grp != current->egid) {
1361 get_group_info(current->group_info);
1362 retval = groups_search(current->group_info, grp);
1363 put_group_info(current->group_info);
1368 EXPORT_SYMBOL(in_egroup_p);
1370 DECLARE_RWSEM(uts_sem);
1372 EXPORT_SYMBOL(uts_sem);
1374 asmlinkage long sys_newuname(struct new_utsname __user * name)
1378 down_read(&uts_sem);
1379 if (copy_to_user(name,&system_utsname,sizeof *name))
1385 asmlinkage long sys_sethostname(char __user *name, int len)
1388 char tmp[__NEW_UTS_LEN];
1390 if (!capable(CAP_SYS_ADMIN))
1392 if (len < 0 || len > __NEW_UTS_LEN)
1394 down_write(&uts_sem);
1396 if (!copy_from_user(tmp, name, len)) {
1397 memcpy(system_utsname.nodename, tmp, len);
1398 system_utsname.nodename[len] = 0;
1405 asmlinkage long sys_gethostname(char __user *name, int len)
1411 down_read(&uts_sem);
1412 i = 1 + strlen(system_utsname.nodename);
1416 if (copy_to_user(name, system_utsname.nodename, i))
1423 * Only setdomainname; getdomainname can be implemented by calling
1426 asmlinkage long sys_setdomainname(char __user *name, int len)
1429 char tmp[__NEW_UTS_LEN];
1431 if (!capable(CAP_SYS_ADMIN))
1433 if (len < 0 || len > __NEW_UTS_LEN)
1436 down_write(&uts_sem);
1438 if (!copy_from_user(tmp, name, len)) {
1439 memcpy(system_utsname.domainname, tmp, len);
1440 system_utsname.domainname[len] = 0;
1447 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1449 if (resource >= RLIM_NLIMITS)
1452 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1456 #if defined(COMPAT_RLIM_OLD_INFINITY) || !(defined(CONFIG_IA64) || defined(CONFIG_V850))
1459 * Back compatibility for getrlimit. Needed for some apps.
1462 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1465 if (resource >= RLIM_NLIMITS)
1468 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1469 if(x.rlim_cur > 0x7FFFFFFF)
1470 x.rlim_cur = 0x7FFFFFFF;
1471 if(x.rlim_max > 0x7FFFFFFF)
1472 x.rlim_max = 0x7FFFFFFF;
1473 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1478 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1480 struct rlimit new_rlim, *old_rlim;
1483 if (resource >= RLIM_NLIMITS)
1485 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1487 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1489 old_rlim = current->rlim + resource;
1490 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1491 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1492 !capable(CAP_SYS_RESOURCE))
1494 if (resource == RLIMIT_NOFILE) {
1495 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1499 retval = security_task_setrlimit(resource, &new_rlim);
1503 *old_rlim = new_rlim;
1508 * It would make sense to put struct rusage in the task_struct,
1509 * except that would make the task_struct be *really big*. After
1510 * task_struct gets moved into malloc'ed memory, it would
1511 * make sense to do this. It will make moving the rest of the information
1512 * a lot simpler! (Which we're not doing right now because we're not
1513 * measuring them yet).
1515 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1516 * below (we know we aren't going to exit/disappear and only we change our
1517 * rusage counters), or we are called from wait4() on a process which is
1518 * either stopped or zombied. In the zombied case the task won't get
1519 * reaped till shortly after the call to getrusage(), in both cases the
1520 * task being examined is in a frozen state so the counters won't change.
1522 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1526 memset((char *) &r, 0, sizeof(r));
1529 jiffies_to_timeval(p->utime, &r.ru_utime);
1530 jiffies_to_timeval(p->stime, &r.ru_stime);
1531 r.ru_nvcsw = p->nvcsw;
1532 r.ru_nivcsw = p->nivcsw;
1533 r.ru_minflt = p->min_flt;
1534 r.ru_majflt = p->maj_flt;
1536 case RUSAGE_CHILDREN:
1537 jiffies_to_timeval(p->cutime, &r.ru_utime);
1538 jiffies_to_timeval(p->cstime, &r.ru_stime);
1539 r.ru_nvcsw = p->cnvcsw;
1540 r.ru_nivcsw = p->cnivcsw;
1541 r.ru_minflt = p->cmin_flt;
1542 r.ru_majflt = p->cmaj_flt;
1545 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1546 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1547 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1548 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1549 r.ru_minflt = p->min_flt + p->cmin_flt;
1550 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1553 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1556 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1558 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1560 return getrusage(current, who, ru);
1563 asmlinkage long sys_umask(int mask)
1565 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1569 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1570 unsigned long arg4, unsigned long arg5)
1575 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1580 case PR_SET_PDEATHSIG:
1582 if (sig < 0 || sig > _NSIG) {
1586 current->pdeath_signal = sig;
1588 case PR_GET_PDEATHSIG:
1589 error = put_user(current->pdeath_signal, (int __user *)arg2);
1591 case PR_GET_DUMPABLE:
1592 if (current->mm->dumpable)
1595 case PR_SET_DUMPABLE:
1596 if (arg2 != 0 && arg2 != 1) {
1600 current->mm->dumpable = arg2;
1603 case PR_SET_UNALIGN:
1604 error = SET_UNALIGN_CTL(current, arg2);
1606 case PR_GET_UNALIGN:
1607 error = GET_UNALIGN_CTL(current, arg2);
1610 error = SET_FPEMU_CTL(current, arg2);
1613 error = GET_FPEMU_CTL(current, arg2);
1616 error = SET_FPEXC_CTL(current, arg2);
1619 error = GET_FPEXC_CTL(current, arg2);
1622 error = PR_TIMING_STATISTICAL;
1625 if (arg2 == PR_TIMING_STATISTICAL)
1631 case PR_GET_KEEPCAPS:
1632 if (current->keep_capabilities)
1635 case PR_SET_KEEPCAPS:
1636 if (arg2 != 0 && arg2 != 1) {
1640 current->keep_capabilities = arg2;