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_mbind
)
278 cond_syscall(compat_get_mempolicy
)
279 cond_syscall(compat_set_mempolicy
)
281 /* arch-specific weak syscall entries */
282 cond_syscall(sys_pciconfig_read
)
283 cond_syscall(sys_pciconfig_write
)
284 cond_syscall(sys_pciconfig_iobase
)
286 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
290 if (p
->uid
!= current
->euid
&&
291 p
->uid
!= current
->uid
&& !capable(CAP_SYS_NICE
)) {
295 if (niceval
< task_nice(p
) && !capable(CAP_SYS_NICE
)) {
299 no_nice
= security_task_setnice(p
, niceval
);
306 set_user_nice(p
, niceval
);
311 asmlinkage
long sys_setpriority(int which
, int who
, int niceval
)
313 struct task_struct
*g
, *p
;
314 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 do_each_task_pid(who
, PIDTYPE_PGID
, p
) {
340 error
= set_one_prio(p
, niceval
, error
);
341 } while_each_task_pid(who
, PIDTYPE_PGID
, p
);
345 user
= current
->user
;
347 user
= find_user(who
);
354 error
= set_one_prio(p
, niceval
, error
);
355 while_each_thread(g
, p
);
357 free_uid(user
); /* For find_user() */
361 read_unlock(&tasklist_lock
);
367 * Ugh. To avoid negative return values, "getpriority()" will
368 * not return the normal nice-value, but a negated value that
369 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
370 * to stay compatible.
372 asmlinkage
long sys_getpriority(int which
, int who
)
374 struct task_struct
*g
, *p
;
375 struct user_struct
*user
;
376 long niceval
, retval
= -ESRCH
;
378 if (which
> 2 || which
< 0)
381 read_lock(&tasklist_lock
);
386 p
= find_task_by_pid(who
);
388 niceval
= 20 - task_nice(p
);
389 if (niceval
> retval
)
395 who
= process_group(current
);
396 do_each_task_pid(who
, PIDTYPE_PGID
, p
) {
397 niceval
= 20 - task_nice(p
);
398 if (niceval
> retval
)
400 } while_each_task_pid(who
, PIDTYPE_PGID
, p
);
404 user
= current
->user
;
406 user
= find_user(who
);
413 niceval
= 20 - task_nice(p
);
414 if (niceval
> retval
)
417 while_each_thread(g
, p
);
419 free_uid(user
); /* for find_user() */
423 read_unlock(&tasklist_lock
);
430 * Reboot system call: for obvious reasons only root may call it,
431 * and even root needs to set up some magic numbers in the registers
432 * so that some mistake won't make this reboot the whole machine.
433 * You can also set the meaning of the ctrl-alt-del-key here.
435 * reboot doesn't sync: do that yourself before calling this.
437 asmlinkage
long sys_reboot(int magic1
, int magic2
, unsigned int cmd
, void __user
* arg
)
441 /* We only trust the superuser with rebooting the system. */
442 if (!capable(CAP_SYS_BOOT
))
445 /* For safety, we require "magic" arguments. */
446 if (magic1
!= LINUX_REBOOT_MAGIC1
||
447 (magic2
!= LINUX_REBOOT_MAGIC2
&&
448 magic2
!= LINUX_REBOOT_MAGIC2A
&&
449 magic2
!= LINUX_REBOOT_MAGIC2B
&&
450 magic2
!= LINUX_REBOOT_MAGIC2C
))
455 case LINUX_REBOOT_CMD_RESTART
:
456 notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, NULL
);
457 system_state
= SYSTEM_RESTART
;
459 printk(KERN_EMERG
"Restarting system.\n");
460 machine_restart(NULL
);
463 case LINUX_REBOOT_CMD_CAD_ON
:
467 case LINUX_REBOOT_CMD_CAD_OFF
:
471 case LINUX_REBOOT_CMD_HALT
:
472 notifier_call_chain(&reboot_notifier_list
, SYS_HALT
, NULL
);
473 system_state
= SYSTEM_HALT
;
475 printk(KERN_EMERG
"System halted.\n");
481 case LINUX_REBOOT_CMD_POWER_OFF
:
482 notifier_call_chain(&reboot_notifier_list
, SYS_POWER_OFF
, NULL
);
483 system_state
= SYSTEM_POWER_OFF
;
485 printk(KERN_EMERG
"Power down.\n");
491 case LINUX_REBOOT_CMD_RESTART2
:
492 if (strncpy_from_user(&buffer
[0], arg
, sizeof(buffer
) - 1) < 0) {
496 buffer
[sizeof(buffer
) - 1] = '\0';
498 notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, buffer
);
499 system_state
= SYSTEM_RESTART
;
501 printk(KERN_EMERG
"Restarting system with command '%s'.\n", buffer
);
502 machine_restart(buffer
);
505 #ifdef CONFIG_SOFTWARE_SUSPEND
506 case LINUX_REBOOT_CMD_SW_SUSPEND
:
508 int ret
= software_suspend();
522 static void deferred_cad(void *dummy
)
524 notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, NULL
);
525 machine_restart(NULL
);
529 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
530 * As it's called within an interrupt, it may NOT sync: the only choice
531 * is whether to reboot at once, or just ignore the ctrl-alt-del.
533 void ctrl_alt_del(void)
535 static DECLARE_WORK(cad_work
, deferred_cad
, NULL
);
538 schedule_work(&cad_work
);
540 kill_proc(cad_pid
, SIGINT
, 1);
545 * Unprivileged users may change the real gid to the effective gid
546 * or vice versa. (BSD-style)
548 * If you set the real gid at all, or set the effective gid to a value not
549 * equal to the real gid, then the saved gid is set to the new effective gid.
551 * This makes it possible for a setgid program to completely drop its
552 * privileges, which is often a useful assertion to make when you are doing
553 * a security audit over a program.
555 * The general idea is that a program which uses just setregid() will be
556 * 100% compatible with BSD. A program which uses just setgid() will be
557 * 100% compatible with POSIX with saved IDs.
559 * SMP: There are not races, the GIDs are checked only by filesystem
560 * operations (as far as semantic preservation is concerned).
562 asmlinkage
long sys_setregid(gid_t rgid
, gid_t egid
)
564 int old_rgid
= current
->gid
;
565 int old_egid
= current
->egid
;
566 int new_rgid
= old_rgid
;
567 int new_egid
= old_egid
;
570 retval
= security_task_setgid(rgid
, egid
, (gid_t
)-1, LSM_SETID_RE
);
574 if (rgid
!= (gid_t
) -1) {
575 if ((old_rgid
== rgid
) ||
576 (current
->egid
==rgid
) ||
582 if (egid
!= (gid_t
) -1) {
583 if ((old_rgid
== egid
) ||
584 (current
->egid
== egid
) ||
585 (current
->sgid
== egid
) ||
592 if (new_egid
!= old_egid
)
594 current
->mm
->dumpable
= 0;
597 if (rgid
!= (gid_t
) -1 ||
598 (egid
!= (gid_t
) -1 && egid
!= old_rgid
))
599 current
->sgid
= new_egid
;
600 current
->fsgid
= new_egid
;
601 current
->egid
= new_egid
;
602 current
->gid
= new_rgid
;
607 * setgid() is implemented like SysV w/ SAVED_IDS
609 * SMP: Same implicit races as above.
611 asmlinkage
long sys_setgid(gid_t gid
)
613 int old_egid
= current
->egid
;
616 retval
= security_task_setgid(gid
, (gid_t
)-1, (gid_t
)-1, LSM_SETID_ID
);
620 if (capable(CAP_SETGID
))
624 current
->mm
->dumpable
=0;
627 current
->gid
= current
->egid
= current
->sgid
= current
->fsgid
= gid
;
629 else if ((gid
== current
->gid
) || (gid
== current
->sgid
))
633 current
->mm
->dumpable
=0;
636 current
->egid
= current
->fsgid
= gid
;
643 static int set_user(uid_t new_ruid
, int dumpclear
)
645 struct user_struct
*new_user
;
647 new_user
= alloc_uid(new_ruid
);
651 if (atomic_read(&new_user
->processes
) >=
652 current
->rlim
[RLIMIT_NPROC
].rlim_cur
&&
653 new_user
!= &root_user
) {
658 switch_uid(new_user
);
662 current
->mm
->dumpable
= 0;
665 current
->uid
= new_ruid
;
670 * Unprivileged users may change the real uid to the effective uid
671 * or vice versa. (BSD-style)
673 * If you set the real uid at all, or set the effective uid to a value not
674 * equal to the real uid, then the saved uid is set to the new effective uid.
676 * This makes it possible for a setuid program to completely drop its
677 * privileges, which is often a useful assertion to make when you are doing
678 * a security audit over a program.
680 * The general idea is that a program which uses just setreuid() will be
681 * 100% compatible with BSD. A program which uses just setuid() will be
682 * 100% compatible with POSIX with saved IDs.
684 asmlinkage
long sys_setreuid(uid_t ruid
, uid_t euid
)
686 int old_ruid
, old_euid
, old_suid
, new_ruid
, new_euid
;
689 retval
= security_task_setuid(ruid
, euid
, (uid_t
)-1, LSM_SETID_RE
);
693 new_ruid
= old_ruid
= current
->uid
;
694 new_euid
= old_euid
= current
->euid
;
695 old_suid
= current
->suid
;
697 if (ruid
!= (uid_t
) -1) {
699 if ((old_ruid
!= ruid
) &&
700 (current
->euid
!= ruid
) &&
701 !capable(CAP_SETUID
))
705 if (euid
!= (uid_t
) -1) {
707 if ((old_ruid
!= euid
) &&
708 (current
->euid
!= euid
) &&
709 (current
->suid
!= euid
) &&
710 !capable(CAP_SETUID
))
714 if (new_ruid
!= old_ruid
&& set_user(new_ruid
, new_euid
!= old_euid
) < 0)
717 if (new_euid
!= old_euid
)
719 current
->mm
->dumpable
=0;
722 current
->fsuid
= current
->euid
= new_euid
;
723 if (ruid
!= (uid_t
) -1 ||
724 (euid
!= (uid_t
) -1 && euid
!= old_ruid
))
725 current
->suid
= current
->euid
;
726 current
->fsuid
= current
->euid
;
728 return security_task_post_setuid(old_ruid
, old_euid
, old_suid
, LSM_SETID_RE
);
734 * setuid() is implemented like SysV with SAVED_IDS
736 * Note that SAVED_ID's is deficient in that a setuid root program
737 * like sendmail, for example, cannot set its uid to be a normal
738 * user and then switch back, because if you're root, setuid() sets
739 * the saved uid too. If you don't like this, blame the bright people
740 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
741 * will allow a root program to temporarily drop privileges and be able to
742 * regain them by swapping the real and effective uid.
744 asmlinkage
long sys_setuid(uid_t uid
)
746 int old_euid
= current
->euid
;
747 int old_ruid
, old_suid
, new_ruid
, new_suid
;
750 retval
= security_task_setuid(uid
, (uid_t
)-1, (uid_t
)-1, LSM_SETID_ID
);
754 old_ruid
= new_ruid
= current
->uid
;
755 old_suid
= current
->suid
;
758 if (capable(CAP_SETUID
)) {
759 if (uid
!= old_ruid
&& set_user(uid
, old_euid
!= uid
) < 0)
762 } else if ((uid
!= current
->uid
) && (uid
!= new_suid
))
767 current
->mm
->dumpable
= 0;
770 current
->fsuid
= current
->euid
= uid
;
771 current
->suid
= new_suid
;
773 return security_task_post_setuid(old_ruid
, old_euid
, old_suid
, LSM_SETID_ID
);
778 * This function implements a generic ability to update ruid, euid,
779 * and suid. This allows you to implement the 4.4 compatible seteuid().
781 asmlinkage
long sys_setresuid(uid_t ruid
, uid_t euid
, uid_t suid
)
783 int old_ruid
= current
->uid
;
784 int old_euid
= current
->euid
;
785 int old_suid
= current
->suid
;
788 retval
= security_task_setuid(ruid
, euid
, suid
, LSM_SETID_RES
);
792 if (!capable(CAP_SETUID
)) {
793 if ((ruid
!= (uid_t
) -1) && (ruid
!= current
->uid
) &&
794 (ruid
!= current
->euid
) && (ruid
!= current
->suid
))
796 if ((euid
!= (uid_t
) -1) && (euid
!= current
->uid
) &&
797 (euid
!= current
->euid
) && (euid
!= current
->suid
))
799 if ((suid
!= (uid_t
) -1) && (suid
!= current
->uid
) &&
800 (suid
!= current
->euid
) && (suid
!= current
->suid
))
803 if (ruid
!= (uid_t
) -1) {
804 if (ruid
!= current
->uid
&& set_user(ruid
, euid
!= current
->euid
) < 0)
807 if (euid
!= (uid_t
) -1) {
808 if (euid
!= current
->euid
)
810 current
->mm
->dumpable
= 0;
813 current
->euid
= euid
;
815 current
->fsuid
= current
->euid
;
816 if (suid
!= (uid_t
) -1)
817 current
->suid
= suid
;
819 return security_task_post_setuid(old_ruid
, old_euid
, old_suid
, LSM_SETID_RES
);
822 asmlinkage
long sys_getresuid(uid_t __user
*ruid
, uid_t __user
*euid
, uid_t __user
*suid
)
826 if (!(retval
= put_user(current
->uid
, ruid
)) &&
827 !(retval
= put_user(current
->euid
, euid
)))
828 retval
= put_user(current
->suid
, suid
);
834 * Same as above, but for rgid, egid, sgid.
836 asmlinkage
long sys_setresgid(gid_t rgid
, gid_t egid
, gid_t sgid
)
840 retval
= security_task_setgid(rgid
, egid
, sgid
, LSM_SETID_RES
);
844 if (!capable(CAP_SETGID
)) {
845 if ((rgid
!= (gid_t
) -1) && (rgid
!= current
->gid
) &&
846 (rgid
!= current
->egid
) && (rgid
!= current
->sgid
))
848 if ((egid
!= (gid_t
) -1) && (egid
!= current
->gid
) &&
849 (egid
!= current
->egid
) && (egid
!= current
->sgid
))
851 if ((sgid
!= (gid_t
) -1) && (sgid
!= current
->gid
) &&
852 (sgid
!= current
->egid
) && (sgid
!= current
->sgid
))
855 if (egid
!= (gid_t
) -1) {
856 if (egid
!= current
->egid
)
858 current
->mm
->dumpable
= 0;
861 current
->egid
= egid
;
863 current
->fsgid
= current
->egid
;
864 if (rgid
!= (gid_t
) -1)
866 if (sgid
!= (gid_t
) -1)
867 current
->sgid
= sgid
;
871 asmlinkage
long sys_getresgid(gid_t __user
*rgid
, gid_t __user
*egid
, gid_t __user
*sgid
)
875 if (!(retval
= put_user(current
->gid
, rgid
)) &&
876 !(retval
= put_user(current
->egid
, egid
)))
877 retval
= put_user(current
->sgid
, sgid
);
884 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
885 * is used for "access()" and for the NFS daemon (letting nfsd stay at
886 * whatever uid it wants to). It normally shadows "euid", except when
887 * explicitly set by setfsuid() or for access..
889 asmlinkage
long sys_setfsuid(uid_t uid
)
893 old_fsuid
= current
->fsuid
;
894 if (security_task_setuid(uid
, (uid_t
)-1, (uid_t
)-1, LSM_SETID_FS
))
897 if (uid
== current
->uid
|| uid
== current
->euid
||
898 uid
== current
->suid
|| uid
== current
->fsuid
||
901 if (uid
!= old_fsuid
)
903 current
->mm
->dumpable
= 0;
906 current
->fsuid
= uid
;
909 security_task_post_setuid(old_fsuid
, (uid_t
)-1, (uid_t
)-1, LSM_SETID_FS
);
915 * Samma på svenska..
917 asmlinkage
long sys_setfsgid(gid_t gid
)
921 old_fsgid
= current
->fsgid
;
922 if (security_task_setgid(gid
, (gid_t
)-1, (gid_t
)-1, LSM_SETID_FS
))
925 if (gid
== current
->gid
|| gid
== current
->egid
||
926 gid
== current
->sgid
|| gid
== current
->fsgid
||
929 if (gid
!= old_fsgid
)
931 current
->mm
->dumpable
= 0;
934 current
->fsgid
= gid
;
939 asmlinkage
long sys_times(struct tms __user
* tbuf
)
942 * In the SMP world we might just be unlucky and have one of
943 * the times increment as we use it. Since the value is an
944 * atomically safe type this is just fine. Conceptually its
945 * as if the syscall took an instant longer to occur.
949 struct task_struct
*tsk
= current
;
950 struct task_struct
*t
;
951 unsigned long utime
, stime
, cutime
, cstime
;
953 read_lock(&tasklist_lock
);
954 utime
= tsk
->signal
->utime
;
955 stime
= tsk
->signal
->stime
;
964 * While we have tasklist_lock read-locked, no dying thread
965 * can be updating current->signal->[us]time. Instead,
966 * we got their counts included in the live thread loop.
967 * However, another thread can come in right now and
968 * do a wait call that updates current->signal->c[us]time.
969 * To make sure we always see that pair updated atomically,
970 * we take the siglock around fetching them.
972 spin_lock_irq(&tsk
->sighand
->siglock
);
973 cutime
= tsk
->signal
->cutime
;
974 cstime
= tsk
->signal
->cstime
;
975 spin_unlock_irq(&tsk
->sighand
->siglock
);
976 read_unlock(&tasklist_lock
);
978 tmp
.tms_utime
= jiffies_to_clock_t(utime
);
979 tmp
.tms_stime
= jiffies_to_clock_t(stime
);
980 tmp
.tms_cutime
= jiffies_to_clock_t(cutime
);
981 tmp
.tms_cstime
= jiffies_to_clock_t(cstime
);
982 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
985 return (long) jiffies_64_to_clock_t(get_jiffies_64());
989 * This needs some heavy checking ...
990 * I just haven't the stomach for it. I also don't fully
991 * understand sessions/pgrp etc. Let somebody who does explain it.
993 * OK, I think I have the protection semantics right.... this is really
994 * only important on a multi-user system anyway, to make sure one user
995 * can't send a signal to a process owned by another. -TYT, 12/12/91
997 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1001 asmlinkage
long sys_setpgid(pid_t pid
, pid_t pgid
)
1003 struct task_struct
*p
;
1013 /* From this point forward we keep holding onto the tasklist lock
1014 * so that our parent does not change from under us. -DaveM
1016 write_lock_irq(&tasklist_lock
);
1019 p
= find_task_by_pid(pid
);
1024 if (!thread_group_leader(p
))
1027 if (p
->parent
== current
|| p
->real_parent
== current
) {
1029 if (p
->signal
->session
!= current
->signal
->session
)
1041 if (p
->signal
->leader
)
1045 struct task_struct
*p
;
1047 do_each_task_pid(pgid
, PIDTYPE_PGID
, p
) {
1048 if (p
->signal
->session
== current
->signal
->session
)
1050 } while_each_task_pid(pgid
, PIDTYPE_PGID
, p
);
1055 err
= security_task_setpgid(p
, pgid
);
1059 if (process_group(p
) != pgid
) {
1060 detach_pid(p
, PIDTYPE_PGID
);
1061 p
->signal
->pgrp
= pgid
;
1062 attach_pid(p
, PIDTYPE_PGID
, pgid
);
1067 /* All paths lead to here, thus we are safe. -DaveM */
1068 write_unlock_irq(&tasklist_lock
);
1072 asmlinkage
long sys_getpgid(pid_t pid
)
1075 return process_group(current
);
1078 struct task_struct
*p
;
1080 read_lock(&tasklist_lock
);
1081 p
= find_task_by_pid(pid
);
1085 retval
= security_task_getpgid(p
);
1087 retval
= process_group(p
);
1089 read_unlock(&tasklist_lock
);
1094 #ifdef __ARCH_WANT_SYS_GETPGRP
1096 asmlinkage
long sys_getpgrp(void)
1098 /* SMP - assuming writes are word atomic this is fine */
1099 return process_group(current
);
1104 asmlinkage
long sys_getsid(pid_t pid
)
1107 return current
->signal
->session
;
1110 struct task_struct
*p
;
1112 read_lock(&tasklist_lock
);
1113 p
= find_task_by_pid(pid
);
1117 retval
= security_task_getsid(p
);
1119 retval
= p
->signal
->session
;
1121 read_unlock(&tasklist_lock
);
1126 asmlinkage
long sys_setsid(void)
1131 if (!thread_group_leader(current
))
1134 write_lock_irq(&tasklist_lock
);
1136 pid
= find_pid(PIDTYPE_PGID
, current
->pid
);
1140 current
->signal
->leader
= 1;
1141 __set_special_pids(current
->pid
, current
->pid
);
1142 current
->signal
->tty
= NULL
;
1143 current
->signal
->tty_old_pgrp
= 0;
1144 err
= process_group(current
);
1146 write_unlock_irq(&tasklist_lock
);
1151 * Supplementary group IDs
1154 /* init to 2 - one for init_task, one to ensure it is never freed */
1155 struct group_info init_groups
= { .usage
= ATOMIC_INIT(2) };
1157 struct group_info
*groups_alloc(int gidsetsize
)
1159 struct group_info
*group_info
;
1163 nblocks
= (gidsetsize
+ NGROUPS_PER_BLOCK
- 1) / NGROUPS_PER_BLOCK
;
1164 /* Make sure we always allocate at least one indirect block pointer */
1165 nblocks
= nblocks
? : 1;
1166 group_info
= kmalloc(sizeof(*group_info
) + nblocks
*sizeof(gid_t
*), GFP_USER
);
1169 group_info
->ngroups
= gidsetsize
;
1170 group_info
->nblocks
= nblocks
;
1171 atomic_set(&group_info
->usage
, 1);
1173 if (gidsetsize
<= NGROUPS_SMALL
) {
1174 group_info
->blocks
[0] = group_info
->small_block
;
1176 for (i
= 0; i
< nblocks
; i
++) {
1178 b
= (void *)__get_free_page(GFP_USER
);
1180 goto out_undo_partial_alloc
;
1181 group_info
->blocks
[i
] = b
;
1186 out_undo_partial_alloc
:
1188 free_page((unsigned long)group_info
->blocks
[i
]);
1194 EXPORT_SYMBOL(groups_alloc
);
1196 void groups_free(struct group_info
*group_info
)
1198 if (group_info
->blocks
[0] != group_info
->small_block
) {
1200 for (i
= 0; i
< group_info
->nblocks
; i
++)
1201 free_page((unsigned long)group_info
->blocks
[i
]);
1206 EXPORT_SYMBOL(groups_free
);
1208 /* export the group_info to a user-space array */
1209 static int groups_to_user(gid_t __user
*grouplist
,
1210 struct group_info
*group_info
)
1213 int count
= group_info
->ngroups
;
1215 for (i
= 0; i
< group_info
->nblocks
; i
++) {
1216 int cp_count
= min(NGROUPS_PER_BLOCK
, count
);
1217 int off
= i
* NGROUPS_PER_BLOCK
;
1218 int len
= cp_count
* sizeof(*grouplist
);
1220 if (copy_to_user(grouplist
+off
, group_info
->blocks
[i
], len
))
1228 /* fill a group_info from a user-space array - it must be allocated already */
1229 static int groups_from_user(struct group_info
*group_info
,
1230 gid_t __user
*grouplist
)
1233 int count
= group_info
->ngroups
;
1235 for (i
= 0; i
< group_info
->nblocks
; i
++) {
1236 int cp_count
= min(NGROUPS_PER_BLOCK
, count
);
1237 int off
= i
* NGROUPS_PER_BLOCK
;
1238 int len
= cp_count
* sizeof(*grouplist
);
1240 if (copy_from_user(group_info
->blocks
[i
], grouplist
+off
, len
))
1248 /* a simple shell-metzner sort */
1249 static void groups_sort(struct group_info
*group_info
)
1251 int base
, max
, stride
;
1252 int gidsetsize
= group_info
->ngroups
;
1254 for (stride
= 1; stride
< gidsetsize
; stride
= 3 * stride
+ 1)
1259 max
= gidsetsize
- stride
;
1260 for (base
= 0; base
< max
; base
++) {
1262 int right
= left
+ stride
;
1263 gid_t tmp
= GROUP_AT(group_info
, right
);
1265 while (left
>= 0 && GROUP_AT(group_info
, left
) > tmp
) {
1266 GROUP_AT(group_info
, right
) =
1267 GROUP_AT(group_info
, left
);
1271 GROUP_AT(group_info
, right
) = tmp
;
1277 /* a simple bsearch */
1278 static int groups_search(struct group_info
*group_info
, gid_t grp
)
1286 right
= group_info
->ngroups
;
1287 while (left
< right
) {
1288 int mid
= (left
+right
)/2;
1289 int cmp
= grp
- GROUP_AT(group_info
, mid
);
1300 /* validate and set current->group_info */
1301 int set_current_groups(struct group_info
*group_info
)
1304 struct group_info
*old_info
;
1306 retval
= security_task_setgroups(group_info
);
1310 groups_sort(group_info
);
1311 get_group_info(group_info
);
1314 old_info
= current
->group_info
;
1315 current
->group_info
= group_info
;
1316 task_unlock(current
);
1318 put_group_info(old_info
);
1323 EXPORT_SYMBOL(set_current_groups
);
1325 asmlinkage
long sys_getgroups(int gidsetsize
, gid_t __user
*grouplist
)
1330 * SMP: Nobody else can change our grouplist. Thus we are
1337 /* no need to grab task_lock here; it cannot change */
1338 get_group_info(current
->group_info
);
1339 i
= current
->group_info
->ngroups
;
1341 if (i
> gidsetsize
) {
1345 if (groups_to_user(grouplist
, current
->group_info
)) {
1351 put_group_info(current
->group_info
);
1356 * SMP: Our groups are copy-on-write. We can set them safely
1357 * without another task interfering.
1360 asmlinkage
long sys_setgroups(int gidsetsize
, gid_t __user
*grouplist
)
1362 struct group_info
*group_info
;
1365 if (!capable(CAP_SETGID
))
1367 if ((unsigned)gidsetsize
> NGROUPS_MAX
)
1370 group_info
= groups_alloc(gidsetsize
);
1373 retval
= groups_from_user(group_info
, grouplist
);
1375 put_group_info(group_info
);
1379 retval
= set_current_groups(group_info
);
1380 put_group_info(group_info
);
1386 * Check whether we're fsgid/egid or in the supplemental group..
1388 int in_group_p(gid_t grp
)
1391 if (grp
!= current
->fsgid
) {
1392 get_group_info(current
->group_info
);
1393 retval
= groups_search(current
->group_info
, grp
);
1394 put_group_info(current
->group_info
);
1399 EXPORT_SYMBOL(in_group_p
);
1401 int in_egroup_p(gid_t grp
)
1404 if (grp
!= current
->egid
) {
1405 get_group_info(current
->group_info
);
1406 retval
= groups_search(current
->group_info
, grp
);
1407 put_group_info(current
->group_info
);
1412 EXPORT_SYMBOL(in_egroup_p
);
1414 DECLARE_RWSEM(uts_sem
);
1416 EXPORT_SYMBOL(uts_sem
);
1418 asmlinkage
long sys_newuname(struct new_utsname __user
* name
)
1422 down_read(&uts_sem
);
1423 if (copy_to_user(name
,&system_utsname
,sizeof *name
))
1429 asmlinkage
long sys_sethostname(char __user
*name
, int len
)
1432 char tmp
[__NEW_UTS_LEN
];
1434 if (!capable(CAP_SYS_ADMIN
))
1436 if (len
< 0 || len
> __NEW_UTS_LEN
)
1438 down_write(&uts_sem
);
1440 if (!copy_from_user(tmp
, name
, len
)) {
1441 memcpy(system_utsname
.nodename
, tmp
, len
);
1442 system_utsname
.nodename
[len
] = 0;
1449 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1451 asmlinkage
long sys_gethostname(char __user
*name
, int len
)
1457 down_read(&uts_sem
);
1458 i
= 1 + strlen(system_utsname
.nodename
);
1462 if (copy_to_user(name
, system_utsname
.nodename
, i
))
1471 * Only setdomainname; getdomainname can be implemented by calling
1474 asmlinkage
long sys_setdomainname(char __user
*name
, int len
)
1477 char tmp
[__NEW_UTS_LEN
];
1479 if (!capable(CAP_SYS_ADMIN
))
1481 if (len
< 0 || len
> __NEW_UTS_LEN
)
1484 down_write(&uts_sem
);
1486 if (!copy_from_user(tmp
, name
, len
)) {
1487 memcpy(system_utsname
.domainname
, tmp
, len
);
1488 system_utsname
.domainname
[len
] = 0;
1495 asmlinkage
long sys_getrlimit(unsigned int resource
, struct rlimit __user
*rlim
)
1497 if (resource
>= RLIM_NLIMITS
)
1500 return copy_to_user(rlim
, current
->rlim
+ resource
, sizeof(*rlim
))
1504 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1507 * Back compatibility for getrlimit. Needed for some apps.
1510 asmlinkage
long sys_old_getrlimit(unsigned int resource
, struct rlimit __user
*rlim
)
1513 if (resource
>= RLIM_NLIMITS
)
1516 memcpy(&x
, current
->rlim
+ resource
, sizeof(*rlim
));
1517 if(x
.rlim_cur
> 0x7FFFFFFF)
1518 x
.rlim_cur
= 0x7FFFFFFF;
1519 if(x
.rlim_max
> 0x7FFFFFFF)
1520 x
.rlim_max
= 0x7FFFFFFF;
1521 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1526 asmlinkage
long sys_setrlimit(unsigned int resource
, struct rlimit __user
*rlim
)
1528 struct rlimit new_rlim
, *old_rlim
;
1531 if (resource
>= RLIM_NLIMITS
)
1533 if(copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1535 if (new_rlim
.rlim_cur
> new_rlim
.rlim_max
)
1537 old_rlim
= current
->rlim
+ resource
;
1538 if (((new_rlim
.rlim_cur
> old_rlim
->rlim_max
) ||
1539 (new_rlim
.rlim_max
> old_rlim
->rlim_max
)) &&
1540 !capable(CAP_SYS_RESOURCE
))
1542 if (resource
== RLIMIT_NOFILE
) {
1543 if (new_rlim
.rlim_cur
> NR_OPEN
|| new_rlim
.rlim_max
> NR_OPEN
)
1547 retval
= security_task_setrlimit(resource
, &new_rlim
);
1551 *old_rlim
= new_rlim
;
1556 * It would make sense to put struct rusage in the task_struct,
1557 * except that would make the task_struct be *really big*. After
1558 * task_struct gets moved into malloc'ed memory, it would
1559 * make sense to do this. It will make moving the rest of the information
1560 * a lot simpler! (Which we're not doing right now because we're not
1561 * measuring them yet).
1563 * This expects to be called with tasklist_lock read-locked or better,
1564 * and the siglock not locked. It may momentarily take the siglock.
1566 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1567 * races with threads incrementing their own counters. But since word
1568 * reads are atomic, we either get new values or old values and we don't
1569 * care which for the sums. We always take the siglock to protect reading
1570 * the c* fields from p->signal from races with exit.c updating those
1571 * fields when reaping, so a sample either gets all the additions of a
1572 * given child after it's reaped, or none so this sample is before reaping.
1575 void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1577 struct task_struct
*t
;
1578 unsigned long flags
;
1579 unsigned long utime
, stime
;
1581 memset((char *) r
, 0, sizeof *r
);
1583 if (unlikely(!p
->signal
))
1587 case RUSAGE_CHILDREN
:
1588 spin_lock_irqsave(&p
->sighand
->siglock
, flags
);
1589 utime
= p
->signal
->cutime
;
1590 stime
= p
->signal
->cstime
;
1591 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1592 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1593 r
->ru_minflt
= p
->signal
->cmin_flt
;
1594 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1595 spin_unlock_irqrestore(&p
->sighand
->siglock
, flags
);
1596 jiffies_to_timeval(utime
, &r
->ru_utime
);
1597 jiffies_to_timeval(stime
, &r
->ru_stime
);
1600 spin_lock_irqsave(&p
->sighand
->siglock
, flags
);
1604 spin_lock_irqsave(&p
->sighand
->siglock
, flags
);
1605 utime
= p
->signal
->cutime
;
1606 stime
= p
->signal
->cstime
;
1607 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1608 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1609 r
->ru_minflt
= p
->signal
->cmin_flt
;
1610 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1612 utime
+= p
->signal
->utime
;
1613 stime
+= p
->signal
->stime
;
1614 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1615 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1616 r
->ru_minflt
+= p
->signal
->min_flt
;
1617 r
->ru_majflt
+= p
->signal
->maj_flt
;
1622 r
->ru_nvcsw
+= t
->nvcsw
;
1623 r
->ru_nivcsw
+= t
->nivcsw
;
1624 r
->ru_minflt
+= t
->min_flt
;
1625 r
->ru_majflt
+= t
->maj_flt
;
1628 spin_unlock_irqrestore(&p
->sighand
->siglock
, flags
);
1629 jiffies_to_timeval(utime
, &r
->ru_utime
);
1630 jiffies_to_timeval(stime
, &r
->ru_stime
);
1637 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1640 read_lock(&tasklist_lock
);
1641 k_getrusage(p
, who
, &r
);
1642 read_unlock(&tasklist_lock
);
1643 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1646 asmlinkage
long sys_getrusage(int who
, struct rusage __user
*ru
)
1648 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
)
1650 return getrusage(current
, who
, ru
);
1653 asmlinkage
long sys_umask(int mask
)
1655 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1659 asmlinkage
long sys_prctl(int option
, unsigned long arg2
, unsigned long arg3
,
1660 unsigned long arg4
, unsigned long arg5
)
1665 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
1670 case PR_SET_PDEATHSIG
:
1672 if (sig
< 0 || sig
> _NSIG
) {
1676 current
->pdeath_signal
= sig
;
1678 case PR_GET_PDEATHSIG
:
1679 error
= put_user(current
->pdeath_signal
, (int __user
*)arg2
);
1681 case PR_GET_DUMPABLE
:
1682 if (current
->mm
->dumpable
)
1685 case PR_SET_DUMPABLE
:
1686 if (arg2
!= 0 && arg2
!= 1) {
1690 current
->mm
->dumpable
= arg2
;
1693 case PR_SET_UNALIGN
:
1694 error
= SET_UNALIGN_CTL(current
, arg2
);
1696 case PR_GET_UNALIGN
:
1697 error
= GET_UNALIGN_CTL(current
, arg2
);
1700 error
= SET_FPEMU_CTL(current
, arg2
);
1703 error
= GET_FPEMU_CTL(current
, arg2
);
1706 error
= SET_FPEXC_CTL(current
, arg2
);
1709 error
= GET_FPEXC_CTL(current
, arg2
);
1712 error
= PR_TIMING_STATISTICAL
;
1715 if (arg2
== PR_TIMING_STATISTICAL
)
1721 case PR_GET_KEEPCAPS
:
1722 if (current
->keep_capabilities
)
1725 case PR_SET_KEEPCAPS
:
1726 if (arg2
!= 0 && arg2
!= 1) {
1730 current
->keep_capabilities
= arg2
;
1733 struct task_struct
*me
= current
;
1734 unsigned char ncomm
[sizeof(me
->comm
)];
1736 ncomm
[sizeof(me
->comm
)-1] = 0;
1737 if (strncpy_from_user(ncomm
, (char __user
*)arg2
,
1738 sizeof(me
->comm
)-1) < 0)
1740 set_task_comm(me
, ncomm
);