4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/config.h>
8 #include <linux/module.h>
10 #include <linux/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/smp_lock.h>
13 #include <linux/notifier.h>
14 #include <linux/reboot.h>
15 #include <linux/prctl.h>
16 #include <linux/init.h>
17 #include <linux/highuid.h>
19 #include <linux/workqueue.h>
20 #include <linux/device.h>
21 #include <linux/key.h>
22 #include <linux/times.h>
23 #include <linux/posix-timers.h>
24 #include <linux/security.h>
25 #include <linux/dcookies.h>
26 #include <linux/suspend.h>
27 #include <linux/tty.h>
28 #include <linux/signal.h>
30 #include <linux/compat.h>
31 #include <linux/syscalls.h>
33 #include <asm/uaccess.h>
35 #include <asm/unistd.h>
37 #ifndef SET_UNALIGN_CTL
38 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
40 #ifndef GET_UNALIGN_CTL
41 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
44 # define SET_FPEMU_CTL(a,b) (-EINVAL)
47 # define GET_FPEMU_CTL(a,b) (-EINVAL)
50 # define SET_FPEXC_CTL(a,b) (-EINVAL)
53 # define GET_FPEXC_CTL(a,b) (-EINVAL)
57 * this is where the system-wide overflow UID and GID are defined, for
58 * architectures that now have 32-bit UID/GID but didn't in the past
61 int overflowuid
= DEFAULT_OVERFLOWUID
;
62 int overflowgid
= DEFAULT_OVERFLOWGID
;
65 EXPORT_SYMBOL(overflowuid
);
66 EXPORT_SYMBOL(overflowgid
);
70 * the same as above, but for filesystems which can only store a 16-bit
71 * UID and GID. as such, this is needed on all architectures
74 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
75 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
77 EXPORT_SYMBOL(fs_overflowuid
);
78 EXPORT_SYMBOL(fs_overflowgid
);
81 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
88 * Notifier list for kernel code which wants to be called
89 * at shutdown. This is used to stop any idling DMA operations
93 static struct notifier_block
*reboot_notifier_list
;
94 static DEFINE_RWLOCK(notifier_lock
);
97 * notifier_chain_register - Add notifier to a notifier chain
98 * @list: Pointer to root list pointer
99 * @n: New entry in notifier chain
101 * Adds a notifier to a notifier chain.
103 * Currently always returns zero.
106 int notifier_chain_register(struct notifier_block
**list
, struct notifier_block
*n
)
108 write_lock(¬ifier_lock
);
111 if(n
->priority
> (*list
)->priority
)
113 list
= &((*list
)->next
);
117 write_unlock(¬ifier_lock
);
121 EXPORT_SYMBOL(notifier_chain_register
);
124 * notifier_chain_unregister - Remove notifier from a notifier chain
125 * @nl: Pointer to root list pointer
126 * @n: New entry in notifier chain
128 * Removes a notifier from a notifier chain.
130 * Returns zero on success, or %-ENOENT on failure.
133 int notifier_chain_unregister(struct notifier_block
**nl
, struct notifier_block
*n
)
135 write_lock(¬ifier_lock
);
141 write_unlock(¬ifier_lock
);
146 write_unlock(¬ifier_lock
);
150 EXPORT_SYMBOL(notifier_chain_unregister
);
153 * notifier_call_chain - Call functions in a notifier chain
154 * @n: Pointer to root pointer of notifier chain
155 * @val: Value passed unmodified to notifier function
156 * @v: Pointer passed unmodified to notifier function
158 * Calls each function in a notifier chain in turn.
160 * If the return value of the notifier can be and'd
161 * with %NOTIFY_STOP_MASK, then notifier_call_chain
162 * will return immediately, with the return value of
163 * the notifier function which halted execution.
164 * Otherwise, the return value is the return value
165 * of the last notifier function called.
168 int notifier_call_chain(struct notifier_block
**n
, unsigned long val
, void *v
)
171 struct notifier_block
*nb
= *n
;
175 ret
=nb
->notifier_call(nb
,val
,v
);
176 if(ret
&NOTIFY_STOP_MASK
)
185 EXPORT_SYMBOL(notifier_call_chain
);
188 * register_reboot_notifier - Register function to be called at reboot time
189 * @nb: Info about notifier function to be called
191 * Registers a function with the list of functions
192 * to be called at reboot time.
194 * Currently always returns zero, as notifier_chain_register
195 * always returns zero.
198 int register_reboot_notifier(struct notifier_block
* nb
)
200 return notifier_chain_register(&reboot_notifier_list
, nb
);
203 EXPORT_SYMBOL(register_reboot_notifier
);
206 * unregister_reboot_notifier - Unregister previously registered reboot notifier
207 * @nb: Hook to be unregistered
209 * Unregisters a previously registered reboot
212 * Returns zero on success, or %-ENOENT on failure.
215 int unregister_reboot_notifier(struct notifier_block
* nb
)
217 return notifier_chain_unregister(&reboot_notifier_list
, nb
);
220 EXPORT_SYMBOL(unregister_reboot_notifier
);
222 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
226 if (p
->uid
!= current
->euid
&&
227 p
->euid
!= current
->euid
&& !capable(CAP_SYS_NICE
)) {
231 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
235 no_nice
= security_task_setnice(p
, niceval
);
242 set_user_nice(p
, niceval
);
247 asmlinkage
long sys_setpriority(int which
, int who
, int niceval
)
249 struct task_struct
*g
, *p
;
250 struct user_struct
*user
;
253 if (which
> 2 || which
< 0)
256 /* normalize: avoid signed division (rounding problems) */
263 read_lock(&tasklist_lock
);
268 p
= find_task_by_pid(who
);
270 error
= set_one_prio(p
, niceval
, error
);
274 who
= process_group(current
);
275 do_each_task_pid(who
, PIDTYPE_PGID
, p
) {
276 error
= set_one_prio(p
, niceval
, error
);
277 } while_each_task_pid(who
, PIDTYPE_PGID
, p
);
280 user
= current
->user
;
284 if ((who
!= current
->uid
) && !(user
= find_user(who
)))
285 goto out_unlock
; /* No processes for this user */
289 error
= set_one_prio(p
, niceval
, error
);
290 while_each_thread(g
, p
);
291 if (who
!= current
->uid
)
292 free_uid(user
); /* For find_user() */
296 read_unlock(&tasklist_lock
);
302 * Ugh. To avoid negative return values, "getpriority()" will
303 * not return the normal nice-value, but a negated value that
304 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
305 * to stay compatible.
307 asmlinkage
long sys_getpriority(int which
, int who
)
309 struct task_struct
*g
, *p
;
310 struct user_struct
*user
;
311 long niceval
, retval
= -ESRCH
;
313 if (which
> 2 || which
< 0)
316 read_lock(&tasklist_lock
);
321 p
= find_task_by_pid(who
);
323 niceval
= 20 - task_nice(p
);
324 if (niceval
> retval
)
330 who
= process_group(current
);
331 do_each_task_pid(who
, PIDTYPE_PGID
, p
) {
332 niceval
= 20 - task_nice(p
);
333 if (niceval
> retval
)
335 } while_each_task_pid(who
, PIDTYPE_PGID
, p
);
338 user
= current
->user
;
342 if ((who
!= current
->uid
) && !(user
= find_user(who
)))
343 goto out_unlock
; /* No processes for this user */
347 niceval
= 20 - task_nice(p
);
348 if (niceval
> retval
)
351 while_each_thread(g
, p
);
352 if (who
!= current
->uid
)
353 free_uid(user
); /* for find_user() */
357 read_unlock(&tasklist_lock
);
364 * Reboot system call: for obvious reasons only root may call it,
365 * and even root needs to set up some magic numbers in the registers
366 * so that some mistake won't make this reboot the whole machine.
367 * You can also set the meaning of the ctrl-alt-del-key here.
369 * reboot doesn't sync: do that yourself before calling this.
371 asmlinkage
long sys_reboot(int magic1
, int magic2
, unsigned int cmd
, void __user
* arg
)
375 /* We only trust the superuser with rebooting the system. */
376 if (!capable(CAP_SYS_BOOT
))
379 /* For safety, we require "magic" arguments. */
380 if (magic1
!= LINUX_REBOOT_MAGIC1
||
381 (magic2
!= LINUX_REBOOT_MAGIC2
&&
382 magic2
!= LINUX_REBOOT_MAGIC2A
&&
383 magic2
!= LINUX_REBOOT_MAGIC2B
&&
384 magic2
!= LINUX_REBOOT_MAGIC2C
))
389 case LINUX_REBOOT_CMD_RESTART
:
390 notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, NULL
);
391 system_state
= SYSTEM_RESTART
;
393 printk(KERN_EMERG
"Restarting system.\n");
394 machine_restart(NULL
);
397 case LINUX_REBOOT_CMD_CAD_ON
:
401 case LINUX_REBOOT_CMD_CAD_OFF
:
405 case LINUX_REBOOT_CMD_HALT
:
406 notifier_call_chain(&reboot_notifier_list
, SYS_HALT
, NULL
);
407 system_state
= SYSTEM_HALT
;
409 printk(KERN_EMERG
"System halted.\n");
415 case LINUX_REBOOT_CMD_POWER_OFF
:
416 notifier_call_chain(&reboot_notifier_list
, SYS_POWER_OFF
, NULL
);
417 system_state
= SYSTEM_POWER_OFF
;
419 printk(KERN_EMERG
"Power down.\n");
425 case LINUX_REBOOT_CMD_RESTART2
:
426 if (strncpy_from_user(&buffer
[0], arg
, sizeof(buffer
) - 1) < 0) {
430 buffer
[sizeof(buffer
) - 1] = '\0';
432 notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, buffer
);
433 system_state
= SYSTEM_RESTART
;
435 printk(KERN_EMERG
"Restarting system with command '%s'.\n", buffer
);
436 machine_restart(buffer
);
439 #ifdef CONFIG_SOFTWARE_SUSPEND
440 case LINUX_REBOOT_CMD_SW_SUSPEND
:
442 int ret
= software_suspend();
456 static void deferred_cad(void *dummy
)
458 notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, NULL
);
459 machine_restart(NULL
);
463 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
464 * As it's called within an interrupt, it may NOT sync: the only choice
465 * is whether to reboot at once, or just ignore the ctrl-alt-del.
467 void ctrl_alt_del(void)
469 static DECLARE_WORK(cad_work
, deferred_cad
, NULL
);
472 schedule_work(&cad_work
);
474 kill_proc(cad_pid
, SIGINT
, 1);
479 * Unprivileged users may change the real gid to the effective gid
480 * or vice versa. (BSD-style)
482 * If you set the real gid at all, or set the effective gid to a value not
483 * equal to the real gid, then the saved gid is set to the new effective gid.
485 * This makes it possible for a setgid program to completely drop its
486 * privileges, which is often a useful assertion to make when you are doing
487 * a security audit over a program.
489 * The general idea is that a program which uses just setregid() will be
490 * 100% compatible with BSD. A program which uses just setgid() will be
491 * 100% compatible with POSIX with saved IDs.
493 * SMP: There are not races, the GIDs are checked only by filesystem
494 * operations (as far as semantic preservation is concerned).
496 asmlinkage
long sys_setregid(gid_t rgid
, gid_t egid
)
498 int old_rgid
= current
->gid
;
499 int old_egid
= current
->egid
;
500 int new_rgid
= old_rgid
;
501 int new_egid
= old_egid
;
504 retval
= security_task_setgid(rgid
, egid
, (gid_t
)-1, LSM_SETID_RE
);
508 if (rgid
!= (gid_t
) -1) {
509 if ((old_rgid
== rgid
) ||
510 (current
->egid
==rgid
) ||
516 if (egid
!= (gid_t
) -1) {
517 if ((old_rgid
== egid
) ||
518 (current
->egid
== egid
) ||
519 (current
->sgid
== egid
) ||
526 if (new_egid
!= old_egid
)
528 current
->mm
->dumpable
= 0;
531 if (rgid
!= (gid_t
) -1 ||
532 (egid
!= (gid_t
) -1 && egid
!= old_rgid
))
533 current
->sgid
= new_egid
;
534 current
->fsgid
= new_egid
;
535 current
->egid
= new_egid
;
536 current
->gid
= new_rgid
;
537 key_fsgid_changed(current
);
542 * setgid() is implemented like SysV w/ SAVED_IDS
544 * SMP: Same implicit races as above.
546 asmlinkage
long sys_setgid(gid_t gid
)
548 int old_egid
= current
->egid
;
551 retval
= security_task_setgid(gid
, (gid_t
)-1, (gid_t
)-1, LSM_SETID_ID
);
555 if (capable(CAP_SETGID
))
559 current
->mm
->dumpable
=0;
562 current
->gid
= current
->egid
= current
->sgid
= current
->fsgid
= gid
;
564 else if ((gid
== current
->gid
) || (gid
== current
->sgid
))
568 current
->mm
->dumpable
=0;
571 current
->egid
= current
->fsgid
= gid
;
576 key_fsgid_changed(current
);
580 static int set_user(uid_t new_ruid
, int dumpclear
)
582 struct user_struct
*new_user
;
584 new_user
= alloc_uid(new_ruid
);
588 if (atomic_read(&new_user
->processes
) >=
589 current
->signal
->rlim
[RLIMIT_NPROC
].rlim_cur
&&
590 new_user
!= &root_user
) {
595 switch_uid(new_user
);
599 current
->mm
->dumpable
= 0;
602 current
->uid
= new_ruid
;
607 * Unprivileged users may change the real uid to the effective uid
608 * or vice versa. (BSD-style)
610 * If you set the real uid at all, or set the effective uid to a value not
611 * equal to the real uid, then the saved uid is set to the new effective uid.
613 * This makes it possible for a setuid program to completely drop its
614 * privileges, which is often a useful assertion to make when you are doing
615 * a security audit over a program.
617 * The general idea is that a program which uses just setreuid() will be
618 * 100% compatible with BSD. A program which uses just setuid() will be
619 * 100% compatible with POSIX with saved IDs.
621 asmlinkage
long sys_setreuid(uid_t ruid
, uid_t euid
)
623 int old_ruid
, old_euid
, old_suid
, new_ruid
, new_euid
;
626 retval
= security_task_setuid(ruid
, euid
, (uid_t
)-1, LSM_SETID_RE
);
630 new_ruid
= old_ruid
= current
->uid
;
631 new_euid
= old_euid
= current
->euid
;
632 old_suid
= current
->suid
;
634 if (ruid
!= (uid_t
) -1) {
636 if ((old_ruid
!= ruid
) &&
637 (current
->euid
!= ruid
) &&
638 !capable(CAP_SETUID
))
642 if (euid
!= (uid_t
) -1) {
644 if ((old_ruid
!= euid
) &&
645 (current
->euid
!= euid
) &&
646 (current
->suid
!= euid
) &&
647 !capable(CAP_SETUID
))
651 if (new_ruid
!= old_ruid
&& set_user(new_ruid
, new_euid
!= old_euid
) < 0)
654 if (new_euid
!= old_euid
)
656 current
->mm
->dumpable
=0;
659 current
->fsuid
= current
->euid
= new_euid
;
660 if (ruid
!= (uid_t
) -1 ||
661 (euid
!= (uid_t
) -1 && euid
!= old_ruid
))
662 current
->suid
= current
->euid
;
663 current
->fsuid
= current
->euid
;
665 key_fsuid_changed(current
);
667 return security_task_post_setuid(old_ruid
, old_euid
, old_suid
, LSM_SETID_RE
);
673 * setuid() is implemented like SysV with SAVED_IDS
675 * Note that SAVED_ID's is deficient in that a setuid root program
676 * like sendmail, for example, cannot set its uid to be a normal
677 * user and then switch back, because if you're root, setuid() sets
678 * the saved uid too. If you don't like this, blame the bright people
679 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
680 * will allow a root program to temporarily drop privileges and be able to
681 * regain them by swapping the real and effective uid.
683 asmlinkage
long sys_setuid(uid_t uid
)
685 int old_euid
= current
->euid
;
686 int old_ruid
, old_suid
, new_ruid
, new_suid
;
689 retval
= security_task_setuid(uid
, (uid_t
)-1, (uid_t
)-1, LSM_SETID_ID
);
693 old_ruid
= new_ruid
= current
->uid
;
694 old_suid
= current
->suid
;
697 if (capable(CAP_SETUID
)) {
698 if (uid
!= old_ruid
&& set_user(uid
, old_euid
!= uid
) < 0)
701 } else if ((uid
!= current
->uid
) && (uid
!= new_suid
))
706 current
->mm
->dumpable
= 0;
709 current
->fsuid
= current
->euid
= uid
;
710 current
->suid
= new_suid
;
712 key_fsuid_changed(current
);
714 return security_task_post_setuid(old_ruid
, old_euid
, old_suid
, LSM_SETID_ID
);
719 * This function implements a generic ability to update ruid, euid,
720 * and suid. This allows you to implement the 4.4 compatible seteuid().
722 asmlinkage
long sys_setresuid(uid_t ruid
, uid_t euid
, uid_t suid
)
724 int old_ruid
= current
->uid
;
725 int old_euid
= current
->euid
;
726 int old_suid
= current
->suid
;
729 retval
= security_task_setuid(ruid
, euid
, suid
, LSM_SETID_RES
);
733 if (!capable(CAP_SETUID
)) {
734 if ((ruid
!= (uid_t
) -1) && (ruid
!= current
->uid
) &&
735 (ruid
!= current
->euid
) && (ruid
!= current
->suid
))
737 if ((euid
!= (uid_t
) -1) && (euid
!= current
->uid
) &&
738 (euid
!= current
->euid
) && (euid
!= current
->suid
))
740 if ((suid
!= (uid_t
) -1) && (suid
!= current
->uid
) &&
741 (suid
!= current
->euid
) && (suid
!= current
->suid
))
744 if (ruid
!= (uid_t
) -1) {
745 if (ruid
!= current
->uid
&& set_user(ruid
, euid
!= current
->euid
) < 0)
748 if (euid
!= (uid_t
) -1) {
749 if (euid
!= current
->euid
)
751 current
->mm
->dumpable
= 0;
754 current
->euid
= euid
;
756 current
->fsuid
= current
->euid
;
757 if (suid
!= (uid_t
) -1)
758 current
->suid
= suid
;
760 key_fsuid_changed(current
);
762 return security_task_post_setuid(old_ruid
, old_euid
, old_suid
, LSM_SETID_RES
);
765 asmlinkage
long sys_getresuid(uid_t __user
*ruid
, uid_t __user
*euid
, uid_t __user
*suid
)
769 if (!(retval
= put_user(current
->uid
, ruid
)) &&
770 !(retval
= put_user(current
->euid
, euid
)))
771 retval
= put_user(current
->suid
, suid
);
777 * Same as above, but for rgid, egid, sgid.
779 asmlinkage
long sys_setresgid(gid_t rgid
, gid_t egid
, gid_t sgid
)
783 retval
= security_task_setgid(rgid
, egid
, sgid
, LSM_SETID_RES
);
787 if (!capable(CAP_SETGID
)) {
788 if ((rgid
!= (gid_t
) -1) && (rgid
!= current
->gid
) &&
789 (rgid
!= current
->egid
) && (rgid
!= current
->sgid
))
791 if ((egid
!= (gid_t
) -1) && (egid
!= current
->gid
) &&
792 (egid
!= current
->egid
) && (egid
!= current
->sgid
))
794 if ((sgid
!= (gid_t
) -1) && (sgid
!= current
->gid
) &&
795 (sgid
!= current
->egid
) && (sgid
!= current
->sgid
))
798 if (egid
!= (gid_t
) -1) {
799 if (egid
!= current
->egid
)
801 current
->mm
->dumpable
= 0;
804 current
->egid
= egid
;
806 current
->fsgid
= current
->egid
;
807 if (rgid
!= (gid_t
) -1)
809 if (sgid
!= (gid_t
) -1)
810 current
->sgid
= sgid
;
812 key_fsgid_changed(current
);
816 asmlinkage
long sys_getresgid(gid_t __user
*rgid
, gid_t __user
*egid
, gid_t __user
*sgid
)
820 if (!(retval
= put_user(current
->gid
, rgid
)) &&
821 !(retval
= put_user(current
->egid
, egid
)))
822 retval
= put_user(current
->sgid
, sgid
);
829 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
830 * is used for "access()" and for the NFS daemon (letting nfsd stay at
831 * whatever uid it wants to). It normally shadows "euid", except when
832 * explicitly set by setfsuid() or for access..
834 asmlinkage
long sys_setfsuid(uid_t uid
)
838 old_fsuid
= current
->fsuid
;
839 if (security_task_setuid(uid
, (uid_t
)-1, (uid_t
)-1, LSM_SETID_FS
))
842 if (uid
== current
->uid
|| uid
== current
->euid
||
843 uid
== current
->suid
|| uid
== current
->fsuid
||
846 if (uid
!= old_fsuid
)
848 current
->mm
->dumpable
= 0;
851 current
->fsuid
= uid
;
854 key_fsuid_changed(current
);
856 security_task_post_setuid(old_fsuid
, (uid_t
)-1, (uid_t
)-1, LSM_SETID_FS
);
862 * Samma på svenska..
864 asmlinkage
long sys_setfsgid(gid_t gid
)
868 old_fsgid
= current
->fsgid
;
869 if (security_task_setgid(gid
, (gid_t
)-1, (gid_t
)-1, LSM_SETID_FS
))
872 if (gid
== current
->gid
|| gid
== current
->egid
||
873 gid
== current
->sgid
|| gid
== current
->fsgid
||
876 if (gid
!= old_fsgid
)
878 current
->mm
->dumpable
= 0;
881 current
->fsgid
= gid
;
882 key_fsgid_changed(current
);
887 asmlinkage
long sys_times(struct tms __user
* tbuf
)
890 * In the SMP world we might just be unlucky and have one of
891 * the times increment as we use it. Since the value is an
892 * atomically safe type this is just fine. Conceptually its
893 * as if the syscall took an instant longer to occur.
897 struct task_struct
*tsk
= current
;
898 struct task_struct
*t
;
899 cputime_t utime
, stime
, cutime
, cstime
;
901 read_lock(&tasklist_lock
);
902 utime
= tsk
->signal
->utime
;
903 stime
= tsk
->signal
->stime
;
906 utime
= cputime_add(utime
, t
->utime
);
907 stime
= cputime_add(stime
, t
->stime
);
912 * While we have tasklist_lock read-locked, no dying thread
913 * can be updating current->signal->[us]time. Instead,
914 * we got their counts included in the live thread loop.
915 * However, another thread can come in right now and
916 * do a wait call that updates current->signal->c[us]time.
917 * To make sure we always see that pair updated atomically,
918 * we take the siglock around fetching them.
920 spin_lock_irq(&tsk
->sighand
->siglock
);
921 cutime
= tsk
->signal
->cutime
;
922 cstime
= tsk
->signal
->cstime
;
923 spin_unlock_irq(&tsk
->sighand
->siglock
);
924 read_unlock(&tasklist_lock
);
926 tmp
.tms_utime
= cputime_to_clock_t(utime
);
927 tmp
.tms_stime
= cputime_to_clock_t(stime
);
928 tmp
.tms_cutime
= cputime_to_clock_t(cutime
);
929 tmp
.tms_cstime
= cputime_to_clock_t(cstime
);
930 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
933 return (long) jiffies_64_to_clock_t(get_jiffies_64());
937 * This needs some heavy checking ...
938 * I just haven't the stomach for it. I also don't fully
939 * understand sessions/pgrp etc. Let somebody who does explain it.
941 * OK, I think I have the protection semantics right.... this is really
942 * only important on a multi-user system anyway, to make sure one user
943 * can't send a signal to a process owned by another. -TYT, 12/12/91
945 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
949 asmlinkage
long sys_setpgid(pid_t pid
, pid_t pgid
)
951 struct task_struct
*p
;
961 /* From this point forward we keep holding onto the tasklist lock
962 * so that our parent does not change from under us. -DaveM
964 write_lock_irq(&tasklist_lock
);
967 p
= find_task_by_pid(pid
);
972 if (!thread_group_leader(p
))
975 if (p
->parent
== current
|| p
->real_parent
== current
) {
977 if (p
->signal
->session
!= current
->signal
->session
)
989 if (p
->signal
->leader
)
993 struct task_struct
*p
;
995 do_each_task_pid(pgid
, PIDTYPE_PGID
, p
) {
996 if (p
->signal
->session
== current
->signal
->session
)
998 } while_each_task_pid(pgid
, PIDTYPE_PGID
, p
);
1003 err
= security_task_setpgid(p
, pgid
);
1007 if (process_group(p
) != pgid
) {
1008 detach_pid(p
, PIDTYPE_PGID
);
1009 p
->signal
->pgrp
= pgid
;
1010 attach_pid(p
, PIDTYPE_PGID
, pgid
);
1015 /* All paths lead to here, thus we are safe. -DaveM */
1016 write_unlock_irq(&tasklist_lock
);
1020 asmlinkage
long sys_getpgid(pid_t pid
)
1023 return process_group(current
);
1026 struct task_struct
*p
;
1028 read_lock(&tasklist_lock
);
1029 p
= find_task_by_pid(pid
);
1033 retval
= security_task_getpgid(p
);
1035 retval
= process_group(p
);
1037 read_unlock(&tasklist_lock
);
1042 #ifdef __ARCH_WANT_SYS_GETPGRP
1044 asmlinkage
long sys_getpgrp(void)
1046 /* SMP - assuming writes are word atomic this is fine */
1047 return process_group(current
);
1052 asmlinkage
long sys_getsid(pid_t pid
)
1055 return current
->signal
->session
;
1058 struct task_struct
*p
;
1060 read_lock(&tasklist_lock
);
1061 p
= find_task_by_pid(pid
);
1065 retval
= security_task_getsid(p
);
1067 retval
= p
->signal
->session
;
1069 read_unlock(&tasklist_lock
);
1074 asmlinkage
long sys_setsid(void)
1079 if (!thread_group_leader(current
))
1083 write_lock_irq(&tasklist_lock
);
1085 pid
= find_pid(PIDTYPE_PGID
, current
->pid
);
1089 current
->signal
->leader
= 1;
1090 __set_special_pids(current
->pid
, current
->pid
);
1091 current
->signal
->tty
= NULL
;
1092 current
->signal
->tty_old_pgrp
= 0;
1093 err
= process_group(current
);
1095 write_unlock_irq(&tasklist_lock
);
1101 * Supplementary group IDs
1104 /* init to 2 - one for init_task, one to ensure it is never freed */
1105 struct group_info init_groups
= { .usage
= ATOMIC_INIT(2) };
1107 struct group_info
*groups_alloc(int gidsetsize
)
1109 struct group_info
*group_info
;
1113 nblocks
= (gidsetsize
+ NGROUPS_PER_BLOCK
- 1) / NGROUPS_PER_BLOCK
;
1114 /* Make sure we always allocate at least one indirect block pointer */
1115 nblocks
= nblocks
? : 1;
1116 group_info
= kmalloc(sizeof(*group_info
) + nblocks
*sizeof(gid_t
*), GFP_USER
);
1119 group_info
->ngroups
= gidsetsize
;
1120 group_info
->nblocks
= nblocks
;
1121 atomic_set(&group_info
->usage
, 1);
1123 if (gidsetsize
<= NGROUPS_SMALL
) {
1124 group_info
->blocks
[0] = group_info
->small_block
;
1126 for (i
= 0; i
< nblocks
; i
++) {
1128 b
= (void *)__get_free_page(GFP_USER
);
1130 goto out_undo_partial_alloc
;
1131 group_info
->blocks
[i
] = b
;
1136 out_undo_partial_alloc
:
1138 free_page((unsigned long)group_info
->blocks
[i
]);
1144 EXPORT_SYMBOL(groups_alloc
);
1146 void groups_free(struct group_info
*group_info
)
1148 if (group_info
->blocks
[0] != group_info
->small_block
) {
1150 for (i
= 0; i
< group_info
->nblocks
; i
++)
1151 free_page((unsigned long)group_info
->blocks
[i
]);
1156 EXPORT_SYMBOL(groups_free
);
1158 /* export the group_info to a user-space array */
1159 static int groups_to_user(gid_t __user
*grouplist
,
1160 struct group_info
*group_info
)
1163 int count
= group_info
->ngroups
;
1165 for (i
= 0; i
< group_info
->nblocks
; i
++) {
1166 int cp_count
= min(NGROUPS_PER_BLOCK
, count
);
1167 int off
= i
* NGROUPS_PER_BLOCK
;
1168 int len
= cp_count
* sizeof(*grouplist
);
1170 if (copy_to_user(grouplist
+off
, group_info
->blocks
[i
], len
))
1178 /* fill a group_info from a user-space array - it must be allocated already */
1179 static int groups_from_user(struct group_info
*group_info
,
1180 gid_t __user
*grouplist
)
1183 int count
= group_info
->ngroups
;
1185 for (i
= 0; i
< group_info
->nblocks
; i
++) {
1186 int cp_count
= min(NGROUPS_PER_BLOCK
, count
);
1187 int off
= i
* NGROUPS_PER_BLOCK
;
1188 int len
= cp_count
* sizeof(*grouplist
);
1190 if (copy_from_user(group_info
->blocks
[i
], grouplist
+off
, len
))
1198 /* a simple Shell sort */
1199 static void groups_sort(struct group_info
*group_info
)
1201 int base
, max
, stride
;
1202 int gidsetsize
= group_info
->ngroups
;
1204 for (stride
= 1; stride
< gidsetsize
; stride
= 3 * stride
+ 1)
1209 max
= gidsetsize
- stride
;
1210 for (base
= 0; base
< max
; base
++) {
1212 int right
= left
+ stride
;
1213 gid_t tmp
= GROUP_AT(group_info
, right
);
1215 while (left
>= 0 && GROUP_AT(group_info
, left
) > tmp
) {
1216 GROUP_AT(group_info
, right
) =
1217 GROUP_AT(group_info
, left
);
1221 GROUP_AT(group_info
, right
) = tmp
;
1227 /* a simple bsearch */
1228 static int groups_search(struct group_info
*group_info
, gid_t grp
)
1236 right
= group_info
->ngroups
;
1237 while (left
< right
) {
1238 int mid
= (left
+right
)/2;
1239 int cmp
= grp
- GROUP_AT(group_info
, mid
);
1250 /* validate and set current->group_info */
1251 int set_current_groups(struct group_info
*group_info
)
1254 struct group_info
*old_info
;
1256 retval
= security_task_setgroups(group_info
);
1260 groups_sort(group_info
);
1261 get_group_info(group_info
);
1264 old_info
= current
->group_info
;
1265 current
->group_info
= group_info
;
1266 task_unlock(current
);
1268 put_group_info(old_info
);
1273 EXPORT_SYMBOL(set_current_groups
);
1275 asmlinkage
long sys_getgroups(int gidsetsize
, gid_t __user
*grouplist
)
1280 * SMP: Nobody else can change our grouplist. Thus we are
1287 /* no need to grab task_lock here; it cannot change */
1288 get_group_info(current
->group_info
);
1289 i
= current
->group_info
->ngroups
;
1291 if (i
> gidsetsize
) {
1295 if (groups_to_user(grouplist
, current
->group_info
)) {
1301 put_group_info(current
->group_info
);
1306 * SMP: Our groups are copy-on-write. We can set them safely
1307 * without another task interfering.
1310 asmlinkage
long sys_setgroups(int gidsetsize
, gid_t __user
*grouplist
)
1312 struct group_info
*group_info
;
1315 if (!capable(CAP_SETGID
))
1317 if ((unsigned)gidsetsize
> NGROUPS_MAX
)
1320 group_info
= groups_alloc(gidsetsize
);
1323 retval
= groups_from_user(group_info
, grouplist
);
1325 put_group_info(group_info
);
1329 retval
= set_current_groups(group_info
);
1330 put_group_info(group_info
);
1336 * Check whether we're fsgid/egid or in the supplemental group..
1338 int in_group_p(gid_t grp
)
1341 if (grp
!= current
->fsgid
) {
1342 get_group_info(current
->group_info
);
1343 retval
= groups_search(current
->group_info
, grp
);
1344 put_group_info(current
->group_info
);
1349 EXPORT_SYMBOL(in_group_p
);
1351 int in_egroup_p(gid_t grp
)
1354 if (grp
!= current
->egid
) {
1355 get_group_info(current
->group_info
);
1356 retval
= groups_search(current
->group_info
, grp
);
1357 put_group_info(current
->group_info
);
1362 EXPORT_SYMBOL(in_egroup_p
);
1364 DECLARE_RWSEM(uts_sem
);
1366 EXPORT_SYMBOL(uts_sem
);
1368 asmlinkage
long sys_newuname(struct new_utsname __user
* name
)
1372 down_read(&uts_sem
);
1373 if (copy_to_user(name
,&system_utsname
,sizeof *name
))
1379 asmlinkage
long sys_sethostname(char __user
*name
, int len
)
1382 char tmp
[__NEW_UTS_LEN
];
1384 if (!capable(CAP_SYS_ADMIN
))
1386 if (len
< 0 || len
> __NEW_UTS_LEN
)
1388 down_write(&uts_sem
);
1390 if (!copy_from_user(tmp
, name
, len
)) {
1391 memcpy(system_utsname
.nodename
, tmp
, len
);
1392 system_utsname
.nodename
[len
] = 0;
1399 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1401 asmlinkage
long sys_gethostname(char __user
*name
, int len
)
1407 down_read(&uts_sem
);
1408 i
= 1 + strlen(system_utsname
.nodename
);
1412 if (copy_to_user(name
, system_utsname
.nodename
, i
))
1421 * Only setdomainname; getdomainname can be implemented by calling
1424 asmlinkage
long sys_setdomainname(char __user
*name
, int len
)
1427 char tmp
[__NEW_UTS_LEN
];
1429 if (!capable(CAP_SYS_ADMIN
))
1431 if (len
< 0 || len
> __NEW_UTS_LEN
)
1434 down_write(&uts_sem
);
1436 if (!copy_from_user(tmp
, name
, len
)) {
1437 memcpy(system_utsname
.domainname
, tmp
, len
);
1438 system_utsname
.domainname
[len
] = 0;
1445 asmlinkage
long sys_getrlimit(unsigned int resource
, struct rlimit __user
*rlim
)
1447 if (resource
>= RLIM_NLIMITS
)
1450 struct rlimit value
;
1451 task_lock(current
->group_leader
);
1452 value
= current
->signal
->rlim
[resource
];
1453 task_unlock(current
->group_leader
);
1454 return copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1458 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1461 * Back compatibility for getrlimit. Needed for some apps.
1464 asmlinkage
long sys_old_getrlimit(unsigned int resource
, struct rlimit __user
*rlim
)
1467 if (resource
>= RLIM_NLIMITS
)
1470 task_lock(current
->group_leader
);
1471 x
= current
->signal
->rlim
[resource
];
1472 task_unlock(current
->group_leader
);
1473 if(x
.rlim_cur
> 0x7FFFFFFF)
1474 x
.rlim_cur
= 0x7FFFFFFF;
1475 if(x
.rlim_max
> 0x7FFFFFFF)
1476 x
.rlim_max
= 0x7FFFFFFF;
1477 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1482 asmlinkage
long sys_setrlimit(unsigned int resource
, struct rlimit __user
*rlim
)
1484 struct rlimit new_rlim
, *old_rlim
;
1487 if (resource
>= RLIM_NLIMITS
)
1489 if(copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1491 if (new_rlim
.rlim_cur
> new_rlim
.rlim_max
)
1493 old_rlim
= current
->signal
->rlim
+ resource
;
1494 if ((new_rlim
.rlim_max
> old_rlim
->rlim_max
) &&
1495 !capable(CAP_SYS_RESOURCE
))
1497 if (resource
== RLIMIT_NOFILE
&& new_rlim
.rlim_max
> NR_OPEN
)
1500 retval
= security_task_setrlimit(resource
, &new_rlim
);
1504 task_lock(current
->group_leader
);
1505 *old_rlim
= new_rlim
;
1506 task_unlock(current
->group_leader
);
1508 if (resource
== RLIMIT_CPU
&& new_rlim
.rlim_cur
!= RLIM_INFINITY
&&
1509 (cputime_eq(current
->signal
->it_prof_expires
, cputime_zero
) ||
1510 new_rlim
.rlim_cur
<= cputime_to_secs(
1511 current
->signal
->it_prof_expires
))) {
1512 cputime_t cputime
= secs_to_cputime(new_rlim
.rlim_cur
);
1513 read_lock(&tasklist_lock
);
1514 spin_lock_irq(¤t
->sighand
->siglock
);
1515 set_process_cpu_timer(current
, CPUCLOCK_PROF
,
1517 spin_unlock_irq(¤t
->sighand
->siglock
);
1518 read_unlock(&tasklist_lock
);
1525 * It would make sense to put struct rusage in the task_struct,
1526 * except that would make the task_struct be *really big*. After
1527 * task_struct gets moved into malloc'ed memory, it would
1528 * make sense to do this. It will make moving the rest of the information
1529 * a lot simpler! (Which we're not doing right now because we're not
1530 * measuring them yet).
1532 * This expects to be called with tasklist_lock read-locked or better,
1533 * and the siglock not locked. It may momentarily take the siglock.
1535 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1536 * races with threads incrementing their own counters. But since word
1537 * reads are atomic, we either get new values or old values and we don't
1538 * care which for the sums. We always take the siglock to protect reading
1539 * the c* fields from p->signal from races with exit.c updating those
1540 * fields when reaping, so a sample either gets all the additions of a
1541 * given child after it's reaped, or none so this sample is before reaping.
1544 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1546 struct task_struct
*t
;
1547 unsigned long flags
;
1548 cputime_t utime
, stime
;
1550 memset((char *) r
, 0, sizeof *r
);
1552 if (unlikely(!p
->signal
))
1556 case RUSAGE_CHILDREN
:
1557 spin_lock_irqsave(&p
->sighand
->siglock
, flags
);
1558 utime
= p
->signal
->cutime
;
1559 stime
= p
->signal
->cstime
;
1560 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1561 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1562 r
->ru_minflt
= p
->signal
->cmin_flt
;
1563 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1564 spin_unlock_irqrestore(&p
->sighand
->siglock
, flags
);
1565 cputime_to_timeval(utime
, &r
->ru_utime
);
1566 cputime_to_timeval(stime
, &r
->ru_stime
);
1569 spin_lock_irqsave(&p
->sighand
->siglock
, flags
);
1570 utime
= stime
= cputime_zero
;
1573 spin_lock_irqsave(&p
->sighand
->siglock
, flags
);
1574 utime
= p
->signal
->cutime
;
1575 stime
= p
->signal
->cstime
;
1576 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1577 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1578 r
->ru_minflt
= p
->signal
->cmin_flt
;
1579 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1581 utime
= cputime_add(utime
, p
->signal
->utime
);
1582 stime
= cputime_add(stime
, p
->signal
->stime
);
1583 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1584 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1585 r
->ru_minflt
+= p
->signal
->min_flt
;
1586 r
->ru_majflt
+= p
->signal
->maj_flt
;
1589 utime
= cputime_add(utime
, t
->utime
);
1590 stime
= cputime_add(stime
, t
->stime
);
1591 r
->ru_nvcsw
+= t
->nvcsw
;
1592 r
->ru_nivcsw
+= t
->nivcsw
;
1593 r
->ru_minflt
+= t
->min_flt
;
1594 r
->ru_majflt
+= t
->maj_flt
;
1597 spin_unlock_irqrestore(&p
->sighand
->siglock
, flags
);
1598 cputime_to_timeval(utime
, &r
->ru_utime
);
1599 cputime_to_timeval(stime
, &r
->ru_stime
);
1606 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1609 read_lock(&tasklist_lock
);
1610 k_getrusage(p
, who
, &r
);
1611 read_unlock(&tasklist_lock
);
1612 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1615 asmlinkage
long sys_getrusage(int who
, struct rusage __user
*ru
)
1617 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
)
1619 return getrusage(current
, who
, ru
);
1622 asmlinkage
long sys_umask(int mask
)
1624 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1628 asmlinkage
long sys_prctl(int option
, unsigned long arg2
, unsigned long arg3
,
1629 unsigned long arg4
, unsigned long arg5
)
1634 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
1639 case PR_SET_PDEATHSIG
:
1641 if (!valid_signal(sig
)) {
1645 current
->pdeath_signal
= sig
;
1647 case PR_GET_PDEATHSIG
:
1648 error
= put_user(current
->pdeath_signal
, (int __user
*)arg2
);
1650 case PR_GET_DUMPABLE
:
1651 if (current
->mm
->dumpable
)
1654 case PR_SET_DUMPABLE
:
1655 if (arg2
!= 0 && arg2
!= 1) {
1659 current
->mm
->dumpable
= arg2
;
1662 case PR_SET_UNALIGN
:
1663 error
= SET_UNALIGN_CTL(current
, arg2
);
1665 case PR_GET_UNALIGN
:
1666 error
= GET_UNALIGN_CTL(current
, arg2
);
1669 error
= SET_FPEMU_CTL(current
, arg2
);
1672 error
= GET_FPEMU_CTL(current
, arg2
);
1675 error
= SET_FPEXC_CTL(current
, arg2
);
1678 error
= GET_FPEXC_CTL(current
, arg2
);
1681 error
= PR_TIMING_STATISTICAL
;
1684 if (arg2
== PR_TIMING_STATISTICAL
)
1690 case PR_GET_KEEPCAPS
:
1691 if (current
->keep_capabilities
)
1694 case PR_SET_KEEPCAPS
:
1695 if (arg2
!= 0 && arg2
!= 1) {
1699 current
->keep_capabilities
= arg2
;
1702 struct task_struct
*me
= current
;
1703 unsigned char ncomm
[sizeof(me
->comm
)];
1705 ncomm
[sizeof(me
->comm
)-1] = 0;
1706 if (strncpy_from_user(ncomm
, (char __user
*)arg2
,
1707 sizeof(me
->comm
)-1) < 0)
1709 set_task_comm(me
, ncomm
);
1713 struct task_struct
*me
= current
;
1714 unsigned char tcomm
[sizeof(me
->comm
)];
1716 get_task_comm(tcomm
, me
);
1717 if (copy_to_user((char __user
*)arg2
, tcomm
, sizeof(tcomm
)))