ia64/linux-2.6.18-xen.hg

view kernel/sys.c @ 562:66faefe721eb

pvSCSI backend driver

Signed-off-by: Tomonari Horikoshi <t.horikoshi@jp.fujitsu.com>
Signed-off-by: Jun Kamada <kama@jp.fujitsu.com>
author Keir Fraser <keir.fraser@citrix.com>
date Mon Jun 02 09:58:27 2008 +0100 (2008-06-02)
parents 831230e53067
children
line source
1 /*
2 * linux/kernel/sys.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
7 #include <linux/module.h>
8 #include <linux/mm.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/smp_lock.h>
12 #include <linux/notifier.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.h>
16 #include <linux/fs.h>
17 #include <linux/kernel.h>
18 #include <linux/kexec.h>
19 #include <linux/workqueue.h>
20 #include <linux/capability.h>
21 #include <linux/device.h>
22 #include <linux/key.h>
23 #include <linux/times.h>
24 #include <linux/posix-timers.h>
25 #include <linux/security.h>
26 #include <linux/dcookies.h>
27 #include <linux/suspend.h>
28 #include <linux/tty.h>
29 #include <linux/signal.h>
30 #include <linux/cn_proc.h>
32 #include <linux/compat.h>
33 #include <linux/syscalls.h>
34 #include <linux/kprobes.h>
36 #include <asm/uaccess.h>
37 #include <asm/io.h>
38 #include <asm/unistd.h>
40 #ifndef SET_UNALIGN_CTL
41 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
42 #endif
43 #ifndef GET_UNALIGN_CTL
44 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
45 #endif
46 #ifndef SET_FPEMU_CTL
47 # define SET_FPEMU_CTL(a,b) (-EINVAL)
48 #endif
49 #ifndef GET_FPEMU_CTL
50 # define GET_FPEMU_CTL(a,b) (-EINVAL)
51 #endif
52 #ifndef SET_FPEXC_CTL
53 # define SET_FPEXC_CTL(a,b) (-EINVAL)
54 #endif
55 #ifndef GET_FPEXC_CTL
56 # define GET_FPEXC_CTL(a,b) (-EINVAL)
57 #endif
58 #ifndef GET_ENDIAN
59 # define GET_ENDIAN(a,b) (-EINVAL)
60 #endif
61 #ifndef SET_ENDIAN
62 # define SET_ENDIAN(a,b) (-EINVAL)
63 #endif
65 /*
66 * this is where the system-wide overflow UID and GID are defined, for
67 * architectures that now have 32-bit UID/GID but didn't in the past
68 */
70 int overflowuid = DEFAULT_OVERFLOWUID;
71 int overflowgid = DEFAULT_OVERFLOWGID;
73 #ifdef CONFIG_UID16
74 EXPORT_SYMBOL(overflowuid);
75 EXPORT_SYMBOL(overflowgid);
76 #endif
78 /*
79 * the same as above, but for filesystems which can only store a 16-bit
80 * UID and GID. as such, this is needed on all architectures
81 */
83 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
84 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
86 EXPORT_SYMBOL(fs_overflowuid);
87 EXPORT_SYMBOL(fs_overflowgid);
89 /*
90 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
91 */
93 int C_A_D = 1;
94 int cad_pid = 1;
96 /*
97 * Notifier list for kernel code which wants to be called
98 * at shutdown. This is used to stop any idling DMA operations
99 * and the like.
100 */
102 static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
104 /*
105 * Notifier chain core routines. The exported routines below
106 * are layered on top of these, with appropriate locking added.
107 */
109 static int notifier_chain_register(struct notifier_block **nl,
110 struct notifier_block *n)
111 {
112 while ((*nl) != NULL) {
113 if (n->priority > (*nl)->priority)
114 break;
115 nl = &((*nl)->next);
116 }
117 n->next = *nl;
118 rcu_assign_pointer(*nl, n);
119 return 0;
120 }
122 static int notifier_chain_unregister(struct notifier_block **nl,
123 struct notifier_block *n)
124 {
125 while ((*nl) != NULL) {
126 if ((*nl) == n) {
127 rcu_assign_pointer(*nl, n->next);
128 return 0;
129 }
130 nl = &((*nl)->next);
131 }
132 return -ENOENT;
133 }
135 static int __kprobes notifier_call_chain(struct notifier_block **nl,
136 unsigned long val, void *v)
137 {
138 int ret = NOTIFY_DONE;
139 struct notifier_block *nb, *next_nb;
141 nb = rcu_dereference(*nl);
142 while (nb) {
143 next_nb = rcu_dereference(nb->next);
144 ret = nb->notifier_call(nb, val, v);
145 if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
146 break;
147 nb = next_nb;
148 }
149 return ret;
150 }
152 /*
153 * Atomic notifier chain routines. Registration and unregistration
154 * use a mutex, and call_chain is synchronized by RCU (no locks).
155 */
157 /**
158 * atomic_notifier_chain_register - Add notifier to an atomic notifier chain
159 * @nh: Pointer to head of the atomic notifier chain
160 * @n: New entry in notifier chain
161 *
162 * Adds a notifier to an atomic notifier chain.
163 *
164 * Currently always returns zero.
165 */
167 int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
168 struct notifier_block *n)
169 {
170 unsigned long flags;
171 int ret;
173 spin_lock_irqsave(&nh->lock, flags);
174 ret = notifier_chain_register(&nh->head, n);
175 spin_unlock_irqrestore(&nh->lock, flags);
176 return ret;
177 }
179 EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
181 /**
182 * atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
183 * @nh: Pointer to head of the atomic notifier chain
184 * @n: Entry to remove from notifier chain
185 *
186 * Removes a notifier from an atomic notifier chain.
187 *
188 * Returns zero on success or %-ENOENT on failure.
189 */
190 int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
191 struct notifier_block *n)
192 {
193 unsigned long flags;
194 int ret;
196 spin_lock_irqsave(&nh->lock, flags);
197 ret = notifier_chain_unregister(&nh->head, n);
198 spin_unlock_irqrestore(&nh->lock, flags);
199 synchronize_rcu();
200 return ret;
201 }
203 EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
205 /**
206 * atomic_notifier_call_chain - Call functions in an atomic notifier chain
207 * @nh: Pointer to head of the atomic notifier chain
208 * @val: Value passed unmodified to notifier function
209 * @v: Pointer passed unmodified to notifier function
210 *
211 * Calls each function in a notifier chain in turn. The functions
212 * run in an atomic context, so they must not block.
213 * This routine uses RCU to synchronize with changes to the chain.
214 *
215 * If the return value of the notifier can be and'ed
216 * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain
217 * will return immediately, with the return value of
218 * the notifier function which halted execution.
219 * Otherwise the return value is the return value
220 * of the last notifier function called.
221 */
223 int atomic_notifier_call_chain(struct atomic_notifier_head *nh,
224 unsigned long val, void *v)
225 {
226 int ret;
228 rcu_read_lock();
229 ret = notifier_call_chain(&nh->head, val, v);
230 rcu_read_unlock();
231 return ret;
232 }
234 EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
236 /*
237 * Blocking notifier chain routines. All access to the chain is
238 * synchronized by an rwsem.
239 */
241 /**
242 * blocking_notifier_chain_register - Add notifier to a blocking notifier chain
243 * @nh: Pointer to head of the blocking notifier chain
244 * @n: New entry in notifier chain
245 *
246 * Adds a notifier to a blocking notifier chain.
247 * Must be called in process context.
248 *
249 * Currently always returns zero.
250 */
252 int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
253 struct notifier_block *n)
254 {
255 int ret;
257 /*
258 * This code gets used during boot-up, when task switching is
259 * not yet working and interrupts must remain disabled. At
260 * such times we must not call down_write().
261 */
262 if (unlikely(system_state == SYSTEM_BOOTING))
263 return notifier_chain_register(&nh->head, n);
265 down_write(&nh->rwsem);
266 ret = notifier_chain_register(&nh->head, n);
267 up_write(&nh->rwsem);
268 return ret;
269 }
271 EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
273 /**
274 * blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
275 * @nh: Pointer to head of the blocking notifier chain
276 * @n: Entry to remove from notifier chain
277 *
278 * Removes a notifier from a blocking notifier chain.
279 * Must be called from process context.
280 *
281 * Returns zero on success or %-ENOENT on failure.
282 */
283 int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
284 struct notifier_block *n)
285 {
286 int ret;
288 /*
289 * This code gets used during boot-up, when task switching is
290 * not yet working and interrupts must remain disabled. At
291 * such times we must not call down_write().
292 */
293 if (unlikely(system_state == SYSTEM_BOOTING))
294 return notifier_chain_unregister(&nh->head, n);
296 down_write(&nh->rwsem);
297 ret = notifier_chain_unregister(&nh->head, n);
298 up_write(&nh->rwsem);
299 return ret;
300 }
302 EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
304 /**
305 * blocking_notifier_call_chain - Call functions in a blocking notifier chain
306 * @nh: Pointer to head of the blocking notifier chain
307 * @val: Value passed unmodified to notifier function
308 * @v: Pointer passed unmodified to notifier function
309 *
310 * Calls each function in a notifier chain in turn. The functions
311 * run in a process context, so they are allowed to block.
312 *
313 * If the return value of the notifier can be and'ed
314 * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain
315 * will return immediately, with the return value of
316 * the notifier function which halted execution.
317 * Otherwise the return value is the return value
318 * of the last notifier function called.
319 */
321 int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
322 unsigned long val, void *v)
323 {
324 int ret;
326 down_read(&nh->rwsem);
327 ret = notifier_call_chain(&nh->head, val, v);
328 up_read(&nh->rwsem);
329 return ret;
330 }
332 EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
334 /*
335 * Raw notifier chain routines. There is no protection;
336 * the caller must provide it. Use at your own risk!
337 */
339 /**
340 * raw_notifier_chain_register - Add notifier to a raw notifier chain
341 * @nh: Pointer to head of the raw notifier chain
342 * @n: New entry in notifier chain
343 *
344 * Adds a notifier to a raw notifier chain.
345 * All locking must be provided by the caller.
346 *
347 * Currently always returns zero.
348 */
350 int raw_notifier_chain_register(struct raw_notifier_head *nh,
351 struct notifier_block *n)
352 {
353 return notifier_chain_register(&nh->head, n);
354 }
356 EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
358 /**
359 * raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
360 * @nh: Pointer to head of the raw notifier chain
361 * @n: Entry to remove from notifier chain
362 *
363 * Removes a notifier from a raw notifier chain.
364 * All locking must be provided by the caller.
365 *
366 * Returns zero on success or %-ENOENT on failure.
367 */
368 int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
369 struct notifier_block *n)
370 {
371 return notifier_chain_unregister(&nh->head, n);
372 }
374 EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
376 /**
377 * raw_notifier_call_chain - Call functions in a raw notifier chain
378 * @nh: Pointer to head of the raw notifier chain
379 * @val: Value passed unmodified to notifier function
380 * @v: Pointer passed unmodified to notifier function
381 *
382 * Calls each function in a notifier chain in turn. The functions
383 * run in an undefined context.
384 * All locking must be provided by the caller.
385 *
386 * If the return value of the notifier can be and'ed
387 * with %NOTIFY_STOP_MASK then raw_notifier_call_chain
388 * will return immediately, with the return value of
389 * the notifier function which halted execution.
390 * Otherwise the return value is the return value
391 * of the last notifier function called.
392 */
394 int raw_notifier_call_chain(struct raw_notifier_head *nh,
395 unsigned long val, void *v)
396 {
397 return notifier_call_chain(&nh->head, val, v);
398 }
400 EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
402 /**
403 * register_reboot_notifier - Register function to be called at reboot time
404 * @nb: Info about notifier function to be called
405 *
406 * Registers a function with the list of functions
407 * to be called at reboot time.
408 *
409 * Currently always returns zero, as blocking_notifier_chain_register
410 * always returns zero.
411 */
413 int register_reboot_notifier(struct notifier_block * nb)
414 {
415 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
416 }
418 EXPORT_SYMBOL(register_reboot_notifier);
420 /**
421 * unregister_reboot_notifier - Unregister previously registered reboot notifier
422 * @nb: Hook to be unregistered
423 *
424 * Unregisters a previously registered reboot
425 * notifier function.
426 *
427 * Returns zero on success, or %-ENOENT on failure.
428 */
430 int unregister_reboot_notifier(struct notifier_block * nb)
431 {
432 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
433 }
435 EXPORT_SYMBOL(unregister_reboot_notifier);
437 static int set_one_prio(struct task_struct *p, int niceval, int error)
438 {
439 int no_nice;
441 if (p->uid != current->euid &&
442 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
443 error = -EPERM;
444 goto out;
445 }
446 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
447 error = -EACCES;
448 goto out;
449 }
450 no_nice = security_task_setnice(p, niceval);
451 if (no_nice) {
452 error = no_nice;
453 goto out;
454 }
455 if (error == -ESRCH)
456 error = 0;
457 set_user_nice(p, niceval);
458 out:
459 return error;
460 }
462 asmlinkage long sys_setpriority(int which, int who, int niceval)
463 {
464 struct task_struct *g, *p;
465 struct user_struct *user;
466 int error = -EINVAL;
468 if (which > 2 || which < 0)
469 goto out;
471 /* normalize: avoid signed division (rounding problems) */
472 error = -ESRCH;
473 if (niceval < -20)
474 niceval = -20;
475 if (niceval > 19)
476 niceval = 19;
478 read_lock(&tasklist_lock);
479 switch (which) {
480 case PRIO_PROCESS:
481 if (!who)
482 who = current->pid;
483 p = find_task_by_pid(who);
484 if (p)
485 error = set_one_prio(p, niceval, error);
486 break;
487 case PRIO_PGRP:
488 if (!who)
489 who = process_group(current);
490 do_each_task_pid(who, PIDTYPE_PGID, p) {
491 error = set_one_prio(p, niceval, error);
492 } while_each_task_pid(who, PIDTYPE_PGID, p);
493 break;
494 case PRIO_USER:
495 user = current->user;
496 if (!who)
497 who = current->uid;
498 else
499 if ((who != current->uid) && !(user = find_user(who)))
500 goto out_unlock; /* No processes for this user */
502 do_each_thread(g, p)
503 if (p->uid == who)
504 error = set_one_prio(p, niceval, error);
505 while_each_thread(g, p);
506 if (who != current->uid)
507 free_uid(user); /* For find_user() */
508 break;
509 }
510 out_unlock:
511 read_unlock(&tasklist_lock);
512 out:
513 return error;
514 }
516 /*
517 * Ugh. To avoid negative return values, "getpriority()" will
518 * not return the normal nice-value, but a negated value that
519 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
520 * to stay compatible.
521 */
522 asmlinkage long sys_getpriority(int which, int who)
523 {
524 struct task_struct *g, *p;
525 struct user_struct *user;
526 long niceval, retval = -ESRCH;
528 if (which > 2 || which < 0)
529 return -EINVAL;
531 read_lock(&tasklist_lock);
532 switch (which) {
533 case PRIO_PROCESS:
534 if (!who)
535 who = current->pid;
536 p = find_task_by_pid(who);
537 if (p) {
538 niceval = 20 - task_nice(p);
539 if (niceval > retval)
540 retval = niceval;
541 }
542 break;
543 case PRIO_PGRP:
544 if (!who)
545 who = process_group(current);
546 do_each_task_pid(who, PIDTYPE_PGID, p) {
547 niceval = 20 - task_nice(p);
548 if (niceval > retval)
549 retval = niceval;
550 } while_each_task_pid(who, PIDTYPE_PGID, p);
551 break;
552 case PRIO_USER:
553 user = current->user;
554 if (!who)
555 who = current->uid;
556 else
557 if ((who != current->uid) && !(user = find_user(who)))
558 goto out_unlock; /* No processes for this user */
560 do_each_thread(g, p)
561 if (p->uid == who) {
562 niceval = 20 - task_nice(p);
563 if (niceval > retval)
564 retval = niceval;
565 }
566 while_each_thread(g, p);
567 if (who != current->uid)
568 free_uid(user); /* for find_user() */
569 break;
570 }
571 out_unlock:
572 read_unlock(&tasklist_lock);
574 return retval;
575 }
577 /**
578 * emergency_restart - reboot the system
579 *
580 * Without shutting down any hardware or taking any locks
581 * reboot the system. This is called when we know we are in
582 * trouble so this is our best effort to reboot. This is
583 * safe to call in interrupt context.
584 */
585 void emergency_restart(void)
586 {
587 machine_emergency_restart();
588 }
589 EXPORT_SYMBOL_GPL(emergency_restart);
591 static void kernel_restart_prepare(char *cmd)
592 {
593 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
594 system_state = SYSTEM_RESTART;
595 device_shutdown();
596 }
598 /**
599 * kernel_restart - reboot the system
600 * @cmd: pointer to buffer containing command to execute for restart
601 * or %NULL
602 *
603 * Shutdown everything and perform a clean reboot.
604 * This is not safe to call in interrupt context.
605 */
606 void kernel_restart(char *cmd)
607 {
608 kernel_restart_prepare(cmd);
609 if (!cmd) {
610 printk(KERN_EMERG "Restarting system.\n");
611 } else {
612 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
613 }
614 printk(".\n");
615 machine_restart(cmd);
616 }
617 EXPORT_SYMBOL_GPL(kernel_restart);
619 /**
620 * kernel_kexec - reboot the system
621 *
622 * Move into place and start executing a preloaded standalone
623 * executable. If nothing was preloaded return an error.
624 */
625 static void kernel_kexec(void)
626 {
627 #ifdef CONFIG_KEXEC
628 struct kimage *image;
629 image = xchg(&kexec_image, NULL);
630 if (!image) {
631 return;
632 }
633 kernel_restart_prepare(NULL);
634 printk(KERN_EMERG "Starting new kernel\n");
635 machine_shutdown();
636 machine_kexec(image);
637 #endif
638 }
640 void kernel_shutdown_prepare(enum system_states state)
641 {
642 blocking_notifier_call_chain(&reboot_notifier_list,
643 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
644 system_state = state;
645 device_shutdown();
646 }
647 /**
648 * kernel_halt - halt the system
649 *
650 * Shutdown everything and perform a clean system halt.
651 */
652 void kernel_halt(void)
653 {
654 kernel_shutdown_prepare(SYSTEM_HALT);
655 printk(KERN_EMERG "System halted.\n");
656 machine_halt();
657 }
659 EXPORT_SYMBOL_GPL(kernel_halt);
661 /**
662 * kernel_power_off - power_off the system
663 *
664 * Shutdown everything and perform a clean system power_off.
665 */
666 void kernel_power_off(void)
667 {
668 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
669 printk(KERN_EMERG "Power down.\n");
670 machine_power_off();
671 }
672 EXPORT_SYMBOL_GPL(kernel_power_off);
673 /*
674 * Reboot system call: for obvious reasons only root may call it,
675 * and even root needs to set up some magic numbers in the registers
676 * so that some mistake won't make this reboot the whole machine.
677 * You can also set the meaning of the ctrl-alt-del-key here.
678 *
679 * reboot doesn't sync: do that yourself before calling this.
680 */
681 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
682 {
683 char buffer[256];
685 /* We only trust the superuser with rebooting the system. */
686 if (!capable(CAP_SYS_BOOT))
687 return -EPERM;
689 /* For safety, we require "magic" arguments. */
690 if (magic1 != LINUX_REBOOT_MAGIC1 ||
691 (magic2 != LINUX_REBOOT_MAGIC2 &&
692 magic2 != LINUX_REBOOT_MAGIC2A &&
693 magic2 != LINUX_REBOOT_MAGIC2B &&
694 magic2 != LINUX_REBOOT_MAGIC2C))
695 return -EINVAL;
697 /* Instead of trying to make the power_off code look like
698 * halt when pm_power_off is not set do it the easy way.
699 */
700 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
701 cmd = LINUX_REBOOT_CMD_HALT;
703 lock_kernel();
704 switch (cmd) {
705 case LINUX_REBOOT_CMD_RESTART:
706 kernel_restart(NULL);
707 break;
709 case LINUX_REBOOT_CMD_CAD_ON:
710 C_A_D = 1;
711 break;
713 case LINUX_REBOOT_CMD_CAD_OFF:
714 C_A_D = 0;
715 break;
717 case LINUX_REBOOT_CMD_HALT:
718 kernel_halt();
719 unlock_kernel();
720 do_exit(0);
721 break;
723 case LINUX_REBOOT_CMD_POWER_OFF:
724 kernel_power_off();
725 unlock_kernel();
726 do_exit(0);
727 break;
729 case LINUX_REBOOT_CMD_RESTART2:
730 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
731 unlock_kernel();
732 return -EFAULT;
733 }
734 buffer[sizeof(buffer) - 1] = '\0';
736 kernel_restart(buffer);
737 break;
739 case LINUX_REBOOT_CMD_KEXEC:
740 kernel_kexec();
741 unlock_kernel();
742 return -EINVAL;
744 #ifdef CONFIG_SOFTWARE_SUSPEND
745 case LINUX_REBOOT_CMD_SW_SUSPEND:
746 {
747 int ret = software_suspend();
748 unlock_kernel();
749 return ret;
750 }
751 #endif
753 default:
754 unlock_kernel();
755 return -EINVAL;
756 }
757 unlock_kernel();
758 return 0;
759 }
761 static void deferred_cad(void *dummy)
762 {
763 kernel_restart(NULL);
764 }
766 /*
767 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
768 * As it's called within an interrupt, it may NOT sync: the only choice
769 * is whether to reboot at once, or just ignore the ctrl-alt-del.
770 */
771 void ctrl_alt_del(void)
772 {
773 static DECLARE_WORK(cad_work, deferred_cad, NULL);
775 if (C_A_D)
776 schedule_work(&cad_work);
777 else
778 kill_proc(cad_pid, SIGINT, 1);
779 }
782 /*
783 * Unprivileged users may change the real gid to the effective gid
784 * or vice versa. (BSD-style)
785 *
786 * If you set the real gid at all, or set the effective gid to a value not
787 * equal to the real gid, then the saved gid is set to the new effective gid.
788 *
789 * This makes it possible for a setgid program to completely drop its
790 * privileges, which is often a useful assertion to make when you are doing
791 * a security audit over a program.
792 *
793 * The general idea is that a program which uses just setregid() will be
794 * 100% compatible with BSD. A program which uses just setgid() will be
795 * 100% compatible with POSIX with saved IDs.
796 *
797 * SMP: There are not races, the GIDs are checked only by filesystem
798 * operations (as far as semantic preservation is concerned).
799 */
800 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
801 {
802 int old_rgid = current->gid;
803 int old_egid = current->egid;
804 int new_rgid = old_rgid;
805 int new_egid = old_egid;
806 int retval;
808 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
809 if (retval)
810 return retval;
812 if (rgid != (gid_t) -1) {
813 if ((old_rgid == rgid) ||
814 (current->egid==rgid) ||
815 capable(CAP_SETGID))
816 new_rgid = rgid;
817 else
818 return -EPERM;
819 }
820 if (egid != (gid_t) -1) {
821 if ((old_rgid == egid) ||
822 (current->egid == egid) ||
823 (current->sgid == egid) ||
824 capable(CAP_SETGID))
825 new_egid = egid;
826 else {
827 return -EPERM;
828 }
829 }
830 if (new_egid != old_egid)
831 {
832 current->mm->dumpable = suid_dumpable;
833 smp_wmb();
834 }
835 if (rgid != (gid_t) -1 ||
836 (egid != (gid_t) -1 && egid != old_rgid))
837 current->sgid = new_egid;
838 current->fsgid = new_egid;
839 current->egid = new_egid;
840 current->gid = new_rgid;
841 key_fsgid_changed(current);
842 proc_id_connector(current, PROC_EVENT_GID);
843 return 0;
844 }
846 /*
847 * setgid() is implemented like SysV w/ SAVED_IDS
848 *
849 * SMP: Same implicit races as above.
850 */
851 asmlinkage long sys_setgid(gid_t gid)
852 {
853 int old_egid = current->egid;
854 int retval;
856 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
857 if (retval)
858 return retval;
860 if (capable(CAP_SETGID))
861 {
862 if(old_egid != gid)
863 {
864 current->mm->dumpable = suid_dumpable;
865 smp_wmb();
866 }
867 current->gid = current->egid = current->sgid = current->fsgid = gid;
868 }
869 else if ((gid == current->gid) || (gid == current->sgid))
870 {
871 if(old_egid != gid)
872 {
873 current->mm->dumpable = suid_dumpable;
874 smp_wmb();
875 }
876 current->egid = current->fsgid = gid;
877 }
878 else
879 return -EPERM;
881 key_fsgid_changed(current);
882 proc_id_connector(current, PROC_EVENT_GID);
883 return 0;
884 }
886 static int set_user(uid_t new_ruid, int dumpclear)
887 {
888 struct user_struct *new_user;
890 new_user = alloc_uid(new_ruid);
891 if (!new_user)
892 return -EAGAIN;
894 if (atomic_read(&new_user->processes) >=
895 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
896 new_user != &root_user) {
897 free_uid(new_user);
898 return -EAGAIN;
899 }
901 switch_uid(new_user);
903 if(dumpclear)
904 {
905 current->mm->dumpable = suid_dumpable;
906 smp_wmb();
907 }
908 current->uid = new_ruid;
909 return 0;
910 }
912 /*
913 * Unprivileged users may change the real uid to the effective uid
914 * or vice versa. (BSD-style)
915 *
916 * If you set the real uid at all, or set the effective uid to a value not
917 * equal to the real uid, then the saved uid is set to the new effective uid.
918 *
919 * This makes it possible for a setuid program to completely drop its
920 * privileges, which is often a useful assertion to make when you are doing
921 * a security audit over a program.
922 *
923 * The general idea is that a program which uses just setreuid() will be
924 * 100% compatible with BSD. A program which uses just setuid() will be
925 * 100% compatible with POSIX with saved IDs.
926 */
927 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
928 {
929 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
930 int retval;
932 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
933 if (retval)
934 return retval;
936 new_ruid = old_ruid = current->uid;
937 new_euid = old_euid = current->euid;
938 old_suid = current->suid;
940 if (ruid != (uid_t) -1) {
941 new_ruid = ruid;
942 if ((old_ruid != ruid) &&
943 (current->euid != ruid) &&
944 !capable(CAP_SETUID))
945 return -EPERM;
946 }
948 if (euid != (uid_t) -1) {
949 new_euid = euid;
950 if ((old_ruid != euid) &&
951 (current->euid != euid) &&
952 (current->suid != euid) &&
953 !capable(CAP_SETUID))
954 return -EPERM;
955 }
957 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
958 return -EAGAIN;
960 if (new_euid != old_euid)
961 {
962 current->mm->dumpable = suid_dumpable;
963 smp_wmb();
964 }
965 current->fsuid = current->euid = new_euid;
966 if (ruid != (uid_t) -1 ||
967 (euid != (uid_t) -1 && euid != old_ruid))
968 current->suid = current->euid;
969 current->fsuid = current->euid;
971 key_fsuid_changed(current);
972 proc_id_connector(current, PROC_EVENT_UID);
974 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
975 }
979 /*
980 * setuid() is implemented like SysV with SAVED_IDS
981 *
982 * Note that SAVED_ID's is deficient in that a setuid root program
983 * like sendmail, for example, cannot set its uid to be a normal
984 * user and then switch back, because if you're root, setuid() sets
985 * the saved uid too. If you don't like this, blame the bright people
986 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
987 * will allow a root program to temporarily drop privileges and be able to
988 * regain them by swapping the real and effective uid.
989 */
990 asmlinkage long sys_setuid(uid_t uid)
991 {
992 int old_euid = current->euid;
993 int old_ruid, old_suid, new_ruid, new_suid;
994 int retval;
996 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
997 if (retval)
998 return retval;
1000 old_ruid = new_ruid = current->uid;
1001 old_suid = current->suid;
1002 new_suid = old_suid;
1004 if (capable(CAP_SETUID)) {
1005 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
1006 return -EAGAIN;
1007 new_suid = uid;
1008 } else if ((uid != current->uid) && (uid != new_suid))
1009 return -EPERM;
1011 if (old_euid != uid)
1013 current->mm->dumpable = suid_dumpable;
1014 smp_wmb();
1016 current->fsuid = current->euid = uid;
1017 current->suid = new_suid;
1019 key_fsuid_changed(current);
1020 proc_id_connector(current, PROC_EVENT_UID);
1022 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
1026 /*
1027 * This function implements a generic ability to update ruid, euid,
1028 * and suid. This allows you to implement the 4.4 compatible seteuid().
1029 */
1030 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1032 int old_ruid = current->uid;
1033 int old_euid = current->euid;
1034 int old_suid = current->suid;
1035 int retval;
1037 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
1038 if (retval)
1039 return retval;
1041 if (!capable(CAP_SETUID)) {
1042 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
1043 (ruid != current->euid) && (ruid != current->suid))
1044 return -EPERM;
1045 if ((euid != (uid_t) -1) && (euid != current->uid) &&
1046 (euid != current->euid) && (euid != current->suid))
1047 return -EPERM;
1048 if ((suid != (uid_t) -1) && (suid != current->uid) &&
1049 (suid != current->euid) && (suid != current->suid))
1050 return -EPERM;
1052 if (ruid != (uid_t) -1) {
1053 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
1054 return -EAGAIN;
1056 if (euid != (uid_t) -1) {
1057 if (euid != current->euid)
1059 current->mm->dumpable = suid_dumpable;
1060 smp_wmb();
1062 current->euid = euid;
1064 current->fsuid = current->euid;
1065 if (suid != (uid_t) -1)
1066 current->suid = suid;
1068 key_fsuid_changed(current);
1069 proc_id_connector(current, PROC_EVENT_UID);
1071 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
1074 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
1076 int retval;
1078 if (!(retval = put_user(current->uid, ruid)) &&
1079 !(retval = put_user(current->euid, euid)))
1080 retval = put_user(current->suid, suid);
1082 return retval;
1085 /*
1086 * Same as above, but for rgid, egid, sgid.
1087 */
1088 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1090 int retval;
1092 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
1093 if (retval)
1094 return retval;
1096 if (!capable(CAP_SETGID)) {
1097 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
1098 (rgid != current->egid) && (rgid != current->sgid))
1099 return -EPERM;
1100 if ((egid != (gid_t) -1) && (egid != current->gid) &&
1101 (egid != current->egid) && (egid != current->sgid))
1102 return -EPERM;
1103 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
1104 (sgid != current->egid) && (sgid != current->sgid))
1105 return -EPERM;
1107 if (egid != (gid_t) -1) {
1108 if (egid != current->egid)
1110 current->mm->dumpable = suid_dumpable;
1111 smp_wmb();
1113 current->egid = egid;
1115 current->fsgid = current->egid;
1116 if (rgid != (gid_t) -1)
1117 current->gid = rgid;
1118 if (sgid != (gid_t) -1)
1119 current->sgid = sgid;
1121 key_fsgid_changed(current);
1122 proc_id_connector(current, PROC_EVENT_GID);
1123 return 0;
1126 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
1128 int retval;
1130 if (!(retval = put_user(current->gid, rgid)) &&
1131 !(retval = put_user(current->egid, egid)))
1132 retval = put_user(current->sgid, sgid);
1134 return retval;
1138 /*
1139 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1140 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1141 * whatever uid it wants to). It normally shadows "euid", except when
1142 * explicitly set by setfsuid() or for access..
1143 */
1144 asmlinkage long sys_setfsuid(uid_t uid)
1146 int old_fsuid;
1148 old_fsuid = current->fsuid;
1149 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
1150 return old_fsuid;
1152 if (uid == current->uid || uid == current->euid ||
1153 uid == current->suid || uid == current->fsuid ||
1154 capable(CAP_SETUID))
1156 if (uid != old_fsuid)
1158 current->mm->dumpable = suid_dumpable;
1159 smp_wmb();
1161 current->fsuid = uid;
1164 key_fsuid_changed(current);
1165 proc_id_connector(current, PROC_EVENT_UID);
1167 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1169 return old_fsuid;
1172 /*
1173 * Samma på svenska..
1174 */
1175 asmlinkage long sys_setfsgid(gid_t gid)
1177 int old_fsgid;
1179 old_fsgid = current->fsgid;
1180 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1181 return old_fsgid;
1183 if (gid == current->gid || gid == current->egid ||
1184 gid == current->sgid || gid == current->fsgid ||
1185 capable(CAP_SETGID))
1187 if (gid != old_fsgid)
1189 current->mm->dumpable = suid_dumpable;
1190 smp_wmb();
1192 current->fsgid = gid;
1193 key_fsgid_changed(current);
1194 proc_id_connector(current, PROC_EVENT_GID);
1196 return old_fsgid;
1199 asmlinkage long sys_times(struct tms __user * tbuf)
1201 /*
1202 * In the SMP world we might just be unlucky and have one of
1203 * the times increment as we use it. Since the value is an
1204 * atomically safe type this is just fine. Conceptually its
1205 * as if the syscall took an instant longer to occur.
1206 */
1207 if (tbuf) {
1208 struct tms tmp;
1209 struct task_struct *tsk = current;
1210 struct task_struct *t;
1211 cputime_t utime, stime, cutime, cstime;
1213 spin_lock_irq(&tsk->sighand->siglock);
1214 utime = tsk->signal->utime;
1215 stime = tsk->signal->stime;
1216 t = tsk;
1217 do {
1218 utime = cputime_add(utime, t->utime);
1219 stime = cputime_add(stime, t->stime);
1220 t = next_thread(t);
1221 } while (t != tsk);
1223 cutime = tsk->signal->cutime;
1224 cstime = tsk->signal->cstime;
1225 spin_unlock_irq(&tsk->sighand->siglock);
1227 tmp.tms_utime = cputime_to_clock_t(utime);
1228 tmp.tms_stime = cputime_to_clock_t(stime);
1229 tmp.tms_cutime = cputime_to_clock_t(cutime);
1230 tmp.tms_cstime = cputime_to_clock_t(cstime);
1231 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1232 return -EFAULT;
1234 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1237 /*
1238 * This needs some heavy checking ...
1239 * I just haven't the stomach for it. I also don't fully
1240 * understand sessions/pgrp etc. Let somebody who does explain it.
1242 * OK, I think I have the protection semantics right.... this is really
1243 * only important on a multi-user system anyway, to make sure one user
1244 * can't send a signal to a process owned by another. -TYT, 12/12/91
1246 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1247 * LBT 04.03.94
1248 */
1250 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1252 struct task_struct *p;
1253 struct task_struct *group_leader = current->group_leader;
1254 int err = -EINVAL;
1256 if (!pid)
1257 pid = group_leader->pid;
1258 if (!pgid)
1259 pgid = pid;
1260 if (pgid < 0)
1261 return -EINVAL;
1263 /* From this point forward we keep holding onto the tasklist lock
1264 * so that our parent does not change from under us. -DaveM
1265 */
1266 write_lock_irq(&tasklist_lock);
1268 err = -ESRCH;
1269 p = find_task_by_pid(pid);
1270 if (!p)
1271 goto out;
1273 err = -EINVAL;
1274 if (!thread_group_leader(p))
1275 goto out;
1277 if (p->real_parent == group_leader) {
1278 err = -EPERM;
1279 if (p->signal->session != group_leader->signal->session)
1280 goto out;
1281 err = -EACCES;
1282 if (p->did_exec)
1283 goto out;
1284 } else {
1285 err = -ESRCH;
1286 if (p != group_leader)
1287 goto out;
1290 err = -EPERM;
1291 if (p->signal->leader)
1292 goto out;
1294 if (pgid != pid) {
1295 struct task_struct *p;
1297 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1298 if (p->signal->session == group_leader->signal->session)
1299 goto ok_pgid;
1300 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1301 goto out;
1304 ok_pgid:
1305 err = security_task_setpgid(p, pgid);
1306 if (err)
1307 goto out;
1309 if (process_group(p) != pgid) {
1310 detach_pid(p, PIDTYPE_PGID);
1311 p->signal->pgrp = pgid;
1312 attach_pid(p, PIDTYPE_PGID, pgid);
1315 err = 0;
1316 out:
1317 /* All paths lead to here, thus we are safe. -DaveM */
1318 write_unlock_irq(&tasklist_lock);
1319 return err;
1322 asmlinkage long sys_getpgid(pid_t pid)
1324 if (!pid) {
1325 return process_group(current);
1326 } else {
1327 int retval;
1328 struct task_struct *p;
1330 read_lock(&tasklist_lock);
1331 p = find_task_by_pid(pid);
1333 retval = -ESRCH;
1334 if (p) {
1335 retval = security_task_getpgid(p);
1336 if (!retval)
1337 retval = process_group(p);
1339 read_unlock(&tasklist_lock);
1340 return retval;
1344 #ifdef __ARCH_WANT_SYS_GETPGRP
1346 asmlinkage long sys_getpgrp(void)
1348 /* SMP - assuming writes are word atomic this is fine */
1349 return process_group(current);
1352 #endif
1354 asmlinkage long sys_getsid(pid_t pid)
1356 if (!pid) {
1357 return current->signal->session;
1358 } else {
1359 int retval;
1360 struct task_struct *p;
1362 read_lock(&tasklist_lock);
1363 p = find_task_by_pid(pid);
1365 retval = -ESRCH;
1366 if(p) {
1367 retval = security_task_getsid(p);
1368 if (!retval)
1369 retval = p->signal->session;
1371 read_unlock(&tasklist_lock);
1372 return retval;
1376 asmlinkage long sys_setsid(void)
1378 struct task_struct *group_leader = current->group_leader;
1379 pid_t session;
1380 int err = -EPERM;
1382 mutex_lock(&tty_mutex);
1383 write_lock_irq(&tasklist_lock);
1385 /* Fail if I am already a session leader */
1386 if (group_leader->signal->leader)
1387 goto out;
1389 session = group_leader->pid;
1390 /* Fail if a process group id already exists that equals the
1391 * proposed session id.
1393 * Don't check if session id == 1 because kernel threads use this
1394 * session id and so the check will always fail and make it so
1395 * init cannot successfully call setsid.
1396 */
1397 if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
1398 goto out;
1400 group_leader->signal->leader = 1;
1401 __set_special_pids(session, session);
1402 group_leader->signal->tty = NULL;
1403 group_leader->signal->tty_old_pgrp = 0;
1404 err = process_group(group_leader);
1405 out:
1406 write_unlock_irq(&tasklist_lock);
1407 mutex_unlock(&tty_mutex);
1408 return err;
1411 /*
1412 * Supplementary group IDs
1413 */
1415 /* init to 2 - one for init_task, one to ensure it is never freed */
1416 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1418 struct group_info *groups_alloc(int gidsetsize)
1420 struct group_info *group_info;
1421 int nblocks;
1422 int i;
1424 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1425 /* Make sure we always allocate at least one indirect block pointer */
1426 nblocks = nblocks ? : 1;
1427 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1428 if (!group_info)
1429 return NULL;
1430 group_info->ngroups = gidsetsize;
1431 group_info->nblocks = nblocks;
1432 atomic_set(&group_info->usage, 1);
1434 if (gidsetsize <= NGROUPS_SMALL) {
1435 group_info->blocks[0] = group_info->small_block;
1436 } else {
1437 for (i = 0; i < nblocks; i++) {
1438 gid_t *b;
1439 b = (void *)__get_free_page(GFP_USER);
1440 if (!b)
1441 goto out_undo_partial_alloc;
1442 group_info->blocks[i] = b;
1445 return group_info;
1447 out_undo_partial_alloc:
1448 while (--i >= 0) {
1449 free_page((unsigned long)group_info->blocks[i]);
1451 kfree(group_info);
1452 return NULL;
1455 EXPORT_SYMBOL(groups_alloc);
1457 void groups_free(struct group_info *group_info)
1459 if (group_info->blocks[0] != group_info->small_block) {
1460 int i;
1461 for (i = 0; i < group_info->nblocks; i++)
1462 free_page((unsigned long)group_info->blocks[i]);
1464 kfree(group_info);
1467 EXPORT_SYMBOL(groups_free);
1469 /* export the group_info to a user-space array */
1470 static int groups_to_user(gid_t __user *grouplist,
1471 struct group_info *group_info)
1473 int i;
1474 int count = group_info->ngroups;
1476 for (i = 0; i < group_info->nblocks; i++) {
1477 int cp_count = min(NGROUPS_PER_BLOCK, count);
1478 int off = i * NGROUPS_PER_BLOCK;
1479 int len = cp_count * sizeof(*grouplist);
1481 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1482 return -EFAULT;
1484 count -= cp_count;
1486 return 0;
1489 /* fill a group_info from a user-space array - it must be allocated already */
1490 static int groups_from_user(struct group_info *group_info,
1491 gid_t __user *grouplist)
1493 int i;
1494 int count = group_info->ngroups;
1496 for (i = 0; i < group_info->nblocks; i++) {
1497 int cp_count = min(NGROUPS_PER_BLOCK, count);
1498 int off = i * NGROUPS_PER_BLOCK;
1499 int len = cp_count * sizeof(*grouplist);
1501 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1502 return -EFAULT;
1504 count -= cp_count;
1506 return 0;
1509 /* a simple Shell sort */
1510 static void groups_sort(struct group_info *group_info)
1512 int base, max, stride;
1513 int gidsetsize = group_info->ngroups;
1515 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1516 ; /* nothing */
1517 stride /= 3;
1519 while (stride) {
1520 max = gidsetsize - stride;
1521 for (base = 0; base < max; base++) {
1522 int left = base;
1523 int right = left + stride;
1524 gid_t tmp = GROUP_AT(group_info, right);
1526 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1527 GROUP_AT(group_info, right) =
1528 GROUP_AT(group_info, left);
1529 right = left;
1530 left -= stride;
1532 GROUP_AT(group_info, right) = tmp;
1534 stride /= 3;
1538 /* a simple bsearch */
1539 int groups_search(struct group_info *group_info, gid_t grp)
1541 unsigned int left, right;
1543 if (!group_info)
1544 return 0;
1546 left = 0;
1547 right = group_info->ngroups;
1548 while (left < right) {
1549 unsigned int mid = (left+right)/2;
1550 int cmp = grp - GROUP_AT(group_info, mid);
1551 if (cmp > 0)
1552 left = mid + 1;
1553 else if (cmp < 0)
1554 right = mid;
1555 else
1556 return 1;
1558 return 0;
1561 /* validate and set current->group_info */
1562 int set_current_groups(struct group_info *group_info)
1564 int retval;
1565 struct group_info *old_info;
1567 retval = security_task_setgroups(group_info);
1568 if (retval)
1569 return retval;
1571 groups_sort(group_info);
1572 get_group_info(group_info);
1574 task_lock(current);
1575 old_info = current->group_info;
1576 current->group_info = group_info;
1577 task_unlock(current);
1579 put_group_info(old_info);
1581 return 0;
1584 EXPORT_SYMBOL(set_current_groups);
1586 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1588 int i = 0;
1590 /*
1591 * SMP: Nobody else can change our grouplist. Thus we are
1592 * safe.
1593 */
1595 if (gidsetsize < 0)
1596 return -EINVAL;
1598 /* no need to grab task_lock here; it cannot change */
1599 i = current->group_info->ngroups;
1600 if (gidsetsize) {
1601 if (i > gidsetsize) {
1602 i = -EINVAL;
1603 goto out;
1605 if (groups_to_user(grouplist, current->group_info)) {
1606 i = -EFAULT;
1607 goto out;
1610 out:
1611 return i;
1614 /*
1615 * SMP: Our groups are copy-on-write. We can set them safely
1616 * without another task interfering.
1617 */
1619 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1621 struct group_info *group_info;
1622 int retval;
1624 if (!capable(CAP_SETGID))
1625 return -EPERM;
1626 if ((unsigned)gidsetsize > NGROUPS_MAX)
1627 return -EINVAL;
1629 group_info = groups_alloc(gidsetsize);
1630 if (!group_info)
1631 return -ENOMEM;
1632 retval = groups_from_user(group_info, grouplist);
1633 if (retval) {
1634 put_group_info(group_info);
1635 return retval;
1638 retval = set_current_groups(group_info);
1639 put_group_info(group_info);
1641 return retval;
1644 /*
1645 * Check whether we're fsgid/egid or in the supplemental group..
1646 */
1647 int in_group_p(gid_t grp)
1649 int retval = 1;
1650 if (grp != current->fsgid) {
1651 retval = groups_search(current->group_info, grp);
1653 return retval;
1656 EXPORT_SYMBOL(in_group_p);
1658 int in_egroup_p(gid_t grp)
1660 int retval = 1;
1661 if (grp != current->egid) {
1662 retval = groups_search(current->group_info, grp);
1664 return retval;
1667 EXPORT_SYMBOL(in_egroup_p);
1669 DECLARE_RWSEM(uts_sem);
1671 EXPORT_SYMBOL(uts_sem);
1673 asmlinkage long sys_newuname(struct new_utsname __user * name)
1675 int errno = 0;
1677 down_read(&uts_sem);
1678 if (copy_to_user(name,&system_utsname,sizeof *name))
1679 errno = -EFAULT;
1680 up_read(&uts_sem);
1681 return errno;
1684 asmlinkage long sys_sethostname(char __user *name, int len)
1686 int errno;
1687 char tmp[__NEW_UTS_LEN];
1689 if (!capable(CAP_SYS_ADMIN))
1690 return -EPERM;
1691 if (len < 0 || len > __NEW_UTS_LEN)
1692 return -EINVAL;
1693 down_write(&uts_sem);
1694 errno = -EFAULT;
1695 if (!copy_from_user(tmp, name, len)) {
1696 memcpy(system_utsname.nodename, tmp, len);
1697 system_utsname.nodename[len] = 0;
1698 errno = 0;
1700 up_write(&uts_sem);
1701 return errno;
1704 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1706 asmlinkage long sys_gethostname(char __user *name, int len)
1708 int i, errno;
1710 if (len < 0)
1711 return -EINVAL;
1712 down_read(&uts_sem);
1713 i = 1 + strlen(system_utsname.nodename);
1714 if (i > len)
1715 i = len;
1716 errno = 0;
1717 if (copy_to_user(name, system_utsname.nodename, i))
1718 errno = -EFAULT;
1719 up_read(&uts_sem);
1720 return errno;
1723 #endif
1725 /*
1726 * Only setdomainname; getdomainname can be implemented by calling
1727 * uname()
1728 */
1729 asmlinkage long sys_setdomainname(char __user *name, int len)
1731 int errno;
1732 char tmp[__NEW_UTS_LEN];
1734 if (!capable(CAP_SYS_ADMIN))
1735 return -EPERM;
1736 if (len < 0 || len > __NEW_UTS_LEN)
1737 return -EINVAL;
1739 down_write(&uts_sem);
1740 errno = -EFAULT;
1741 if (!copy_from_user(tmp, name, len)) {
1742 memcpy(system_utsname.domainname, tmp, len);
1743 system_utsname.domainname[len] = 0;
1744 errno = 0;
1746 up_write(&uts_sem);
1747 return errno;
1750 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1752 if (resource >= RLIM_NLIMITS)
1753 return -EINVAL;
1754 else {
1755 struct rlimit value;
1756 task_lock(current->group_leader);
1757 value = current->signal->rlim[resource];
1758 task_unlock(current->group_leader);
1759 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1763 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1765 /*
1766 * Back compatibility for getrlimit. Needed for some apps.
1767 */
1769 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1771 struct rlimit x;
1772 if (resource >= RLIM_NLIMITS)
1773 return -EINVAL;
1775 task_lock(current->group_leader);
1776 x = current->signal->rlim[resource];
1777 task_unlock(current->group_leader);
1778 if(x.rlim_cur > 0x7FFFFFFF)
1779 x.rlim_cur = 0x7FFFFFFF;
1780 if(x.rlim_max > 0x7FFFFFFF)
1781 x.rlim_max = 0x7FFFFFFF;
1782 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1785 #endif
1787 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1789 struct rlimit new_rlim, *old_rlim;
1790 unsigned long it_prof_secs;
1791 int retval;
1793 if (resource >= RLIM_NLIMITS)
1794 return -EINVAL;
1795 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1796 return -EFAULT;
1797 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1798 return -EINVAL;
1799 old_rlim = current->signal->rlim + resource;
1800 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1801 !capable(CAP_SYS_RESOURCE))
1802 return -EPERM;
1803 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1804 return -EPERM;
1806 retval = security_task_setrlimit(resource, &new_rlim);
1807 if (retval)
1808 return retval;
1810 task_lock(current->group_leader);
1811 *old_rlim = new_rlim;
1812 task_unlock(current->group_leader);
1814 if (resource != RLIMIT_CPU)
1815 goto out;
1817 /*
1818 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1819 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1820 * very long-standing error, and fixing it now risks breakage of
1821 * applications, so we live with it
1822 */
1823 if (new_rlim.rlim_cur == RLIM_INFINITY)
1824 goto out;
1826 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
1827 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
1828 unsigned long rlim_cur = new_rlim.rlim_cur;
1829 cputime_t cputime;
1831 if (rlim_cur == 0) {
1832 /*
1833 * The caller is asking for an immediate RLIMIT_CPU
1834 * expiry. But we use the zero value to mean "it was
1835 * never set". So let's cheat and make it one second
1836 * instead
1837 */
1838 rlim_cur = 1;
1840 cputime = secs_to_cputime(rlim_cur);
1841 read_lock(&tasklist_lock);
1842 spin_lock_irq(&current->sighand->siglock);
1843 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
1844 spin_unlock_irq(&current->sighand->siglock);
1845 read_unlock(&tasklist_lock);
1847 out:
1848 return 0;
1851 /*
1852 * It would make sense to put struct rusage in the task_struct,
1853 * except that would make the task_struct be *really big*. After
1854 * task_struct gets moved into malloc'ed memory, it would
1855 * make sense to do this. It will make moving the rest of the information
1856 * a lot simpler! (Which we're not doing right now because we're not
1857 * measuring them yet).
1859 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1860 * races with threads incrementing their own counters. But since word
1861 * reads are atomic, we either get new values or old values and we don't
1862 * care which for the sums. We always take the siglock to protect reading
1863 * the c* fields from p->signal from races with exit.c updating those
1864 * fields when reaping, so a sample either gets all the additions of a
1865 * given child after it's reaped, or none so this sample is before reaping.
1867 * Locking:
1868 * We need to take the siglock for CHILDEREN, SELF and BOTH
1869 * for the cases current multithreaded, non-current single threaded
1870 * non-current multithreaded. Thread traversal is now safe with
1871 * the siglock held.
1872 * Strictly speaking, we donot need to take the siglock if we are current and
1873 * single threaded, as no one else can take our signal_struct away, no one
1874 * else can reap the children to update signal->c* counters, and no one else
1875 * can race with the signal-> fields. If we do not take any lock, the
1876 * signal-> fields could be read out of order while another thread was just
1877 * exiting. So we should place a read memory barrier when we avoid the lock.
1878 * On the writer side, write memory barrier is implied in __exit_signal
1879 * as __exit_signal releases the siglock spinlock after updating the signal->
1880 * fields. But we don't do this yet to keep things simple.
1882 */
1884 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1886 struct task_struct *t;
1887 unsigned long flags;
1888 cputime_t utime, stime;
1890 memset((char *) r, 0, sizeof *r);
1891 utime = stime = cputime_zero;
1893 rcu_read_lock();
1894 if (!lock_task_sighand(p, &flags)) {
1895 rcu_read_unlock();
1896 return;
1899 switch (who) {
1900 case RUSAGE_BOTH:
1901 case RUSAGE_CHILDREN:
1902 utime = p->signal->cutime;
1903 stime = p->signal->cstime;
1904 r->ru_nvcsw = p->signal->cnvcsw;
1905 r->ru_nivcsw = p->signal->cnivcsw;
1906 r->ru_minflt = p->signal->cmin_flt;
1907 r->ru_majflt = p->signal->cmaj_flt;
1909 if (who == RUSAGE_CHILDREN)
1910 break;
1912 case RUSAGE_SELF:
1913 utime = cputime_add(utime, p->signal->utime);
1914 stime = cputime_add(stime, p->signal->stime);
1915 r->ru_nvcsw += p->signal->nvcsw;
1916 r->ru_nivcsw += p->signal->nivcsw;
1917 r->ru_minflt += p->signal->min_flt;
1918 r->ru_majflt += p->signal->maj_flt;
1919 t = p;
1920 do {
1921 utime = cputime_add(utime, t->utime);
1922 stime = cputime_add(stime, t->stime);
1923 r->ru_nvcsw += t->nvcsw;
1924 r->ru_nivcsw += t->nivcsw;
1925 r->ru_minflt += t->min_flt;
1926 r->ru_majflt += t->maj_flt;
1927 t = next_thread(t);
1928 } while (t != p);
1929 break;
1931 default:
1932 BUG();
1935 unlock_task_sighand(p, &flags);
1936 rcu_read_unlock();
1938 cputime_to_timeval(utime, &r->ru_utime);
1939 cputime_to_timeval(stime, &r->ru_stime);
1942 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1944 struct rusage r;
1945 k_getrusage(p, who, &r);
1946 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1949 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1951 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1952 return -EINVAL;
1953 return getrusage(current, who, ru);
1956 asmlinkage long sys_umask(int mask)
1958 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1959 return mask;
1962 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1963 unsigned long arg4, unsigned long arg5)
1965 long error;
1967 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1968 if (error)
1969 return error;
1971 switch (option) {
1972 case PR_SET_PDEATHSIG:
1973 if (!valid_signal(arg2)) {
1974 error = -EINVAL;
1975 break;
1977 current->pdeath_signal = arg2;
1978 break;
1979 case PR_GET_PDEATHSIG:
1980 error = put_user(current->pdeath_signal, (int __user *)arg2);
1981 break;
1982 case PR_GET_DUMPABLE:
1983 error = current->mm->dumpable;
1984 break;
1985 case PR_SET_DUMPABLE:
1986 if (arg2 < 0 || arg2 > 1) {
1987 error = -EINVAL;
1988 break;
1990 current->mm->dumpable = arg2;
1991 break;
1993 case PR_SET_UNALIGN:
1994 error = SET_UNALIGN_CTL(current, arg2);
1995 break;
1996 case PR_GET_UNALIGN:
1997 error = GET_UNALIGN_CTL(current, arg2);
1998 break;
1999 case PR_SET_FPEMU:
2000 error = SET_FPEMU_CTL(current, arg2);
2001 break;
2002 case PR_GET_FPEMU:
2003 error = GET_FPEMU_CTL(current, arg2);
2004 break;
2005 case PR_SET_FPEXC:
2006 error = SET_FPEXC_CTL(current, arg2);
2007 break;
2008 case PR_GET_FPEXC:
2009 error = GET_FPEXC_CTL(current, arg2);
2010 break;
2011 case PR_GET_TIMING:
2012 error = PR_TIMING_STATISTICAL;
2013 break;
2014 case PR_SET_TIMING:
2015 if (arg2 == PR_TIMING_STATISTICAL)
2016 error = 0;
2017 else
2018 error = -EINVAL;
2019 break;
2021 case PR_GET_KEEPCAPS:
2022 if (current->keep_capabilities)
2023 error = 1;
2024 break;
2025 case PR_SET_KEEPCAPS:
2026 if (arg2 != 0 && arg2 != 1) {
2027 error = -EINVAL;
2028 break;
2030 current->keep_capabilities = arg2;
2031 break;
2032 case PR_SET_NAME: {
2033 struct task_struct *me = current;
2034 unsigned char ncomm[sizeof(me->comm)];
2036 ncomm[sizeof(me->comm)-1] = 0;
2037 if (strncpy_from_user(ncomm, (char __user *)arg2,
2038 sizeof(me->comm)-1) < 0)
2039 return -EFAULT;
2040 set_task_comm(me, ncomm);
2041 return 0;
2043 case PR_GET_NAME: {
2044 struct task_struct *me = current;
2045 unsigned char tcomm[sizeof(me->comm)];
2047 get_task_comm(tcomm, me);
2048 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
2049 return -EFAULT;
2050 return 0;
2052 case PR_GET_ENDIAN:
2053 error = GET_ENDIAN(current, arg2);
2054 break;
2055 case PR_SET_ENDIAN:
2056 error = SET_ENDIAN(current, arg2);
2057 break;
2059 default:
2060 error = -EINVAL;
2061 break;
2063 return error;