xref: /linux/kernel/sys.c (revision f3d9478b2ce468c3115b02ecae7e975990697f15)
1 /*
2  *  linux/kernel/sys.c
3  *
4  *  Copyright (C) 1991, 1992  Linus Torvalds
5  */
6 
7 #include <linux/config.h>
8 #include <linux/module.h>
9 #include <linux/mm.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>
18 #include <linux/fs.h>
19 #include <linux/kernel.h>
20 #include <linux/kexec.h>
21 #include <linux/workqueue.h>
22 #include <linux/capability.h>
23 #include <linux/device.h>
24 #include <linux/key.h>
25 #include <linux/times.h>
26 #include <linux/posix-timers.h>
27 #include <linux/security.h>
28 #include <linux/dcookies.h>
29 #include <linux/suspend.h>
30 #include <linux/tty.h>
31 #include <linux/signal.h>
32 #include <linux/cn_proc.h>
33 
34 #include <linux/compat.h>
35 #include <linux/syscalls.h>
36 #include <linux/kprobes.h>
37 
38 #include <asm/uaccess.h>
39 #include <asm/io.h>
40 #include <asm/unistd.h>
41 
42 #ifndef SET_UNALIGN_CTL
43 # define SET_UNALIGN_CTL(a,b)	(-EINVAL)
44 #endif
45 #ifndef GET_UNALIGN_CTL
46 # define GET_UNALIGN_CTL(a,b)	(-EINVAL)
47 #endif
48 #ifndef SET_FPEMU_CTL
49 # define SET_FPEMU_CTL(a,b)	(-EINVAL)
50 #endif
51 #ifndef GET_FPEMU_CTL
52 # define GET_FPEMU_CTL(a,b)	(-EINVAL)
53 #endif
54 #ifndef SET_FPEXC_CTL
55 # define SET_FPEXC_CTL(a,b)	(-EINVAL)
56 #endif
57 #ifndef GET_FPEXC_CTL
58 # define GET_FPEXC_CTL(a,b)	(-EINVAL)
59 #endif
60 
61 /*
62  * this is where the system-wide overflow UID and GID are defined, for
63  * architectures that now have 32-bit UID/GID but didn't in the past
64  */
65 
66 int overflowuid = DEFAULT_OVERFLOWUID;
67 int overflowgid = DEFAULT_OVERFLOWGID;
68 
69 #ifdef CONFIG_UID16
70 EXPORT_SYMBOL(overflowuid);
71 EXPORT_SYMBOL(overflowgid);
72 #endif
73 
74 /*
75  * the same as above, but for filesystems which can only store a 16-bit
76  * UID and GID. as such, this is needed on all architectures
77  */
78 
79 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
80 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
81 
82 EXPORT_SYMBOL(fs_overflowuid);
83 EXPORT_SYMBOL(fs_overflowgid);
84 
85 /*
86  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
87  */
88 
89 int C_A_D = 1;
90 int cad_pid = 1;
91 
92 /*
93  *	Notifier list for kernel code which wants to be called
94  *	at shutdown. This is used to stop any idling DMA operations
95  *	and the like.
96  */
97 
98 static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
99 
100 /*
101  *	Notifier chain core routines.  The exported routines below
102  *	are layered on top of these, with appropriate locking added.
103  */
104 
105 static int notifier_chain_register(struct notifier_block **nl,
106 		struct notifier_block *n)
107 {
108 	while ((*nl) != NULL) {
109 		if (n->priority > (*nl)->priority)
110 			break;
111 		nl = &((*nl)->next);
112 	}
113 	n->next = *nl;
114 	rcu_assign_pointer(*nl, n);
115 	return 0;
116 }
117 
118 static int notifier_chain_unregister(struct notifier_block **nl,
119 		struct notifier_block *n)
120 {
121 	while ((*nl) != NULL) {
122 		if ((*nl) == n) {
123 			rcu_assign_pointer(*nl, n->next);
124 			return 0;
125 		}
126 		nl = &((*nl)->next);
127 	}
128 	return -ENOENT;
129 }
130 
131 static int __kprobes notifier_call_chain(struct notifier_block **nl,
132 		unsigned long val, void *v)
133 {
134 	int ret = NOTIFY_DONE;
135 	struct notifier_block *nb;
136 
137 	nb = rcu_dereference(*nl);
138 	while (nb) {
139 		ret = nb->notifier_call(nb, val, v);
140 		if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
141 			break;
142 		nb = rcu_dereference(nb->next);
143 	}
144 	return ret;
145 }
146 
147 /*
148  *	Atomic notifier chain routines.  Registration and unregistration
149  *	use a mutex, and call_chain is synchronized by RCU (no locks).
150  */
151 
152 /**
153  *	atomic_notifier_chain_register - Add notifier to an atomic notifier chain
154  *	@nh: Pointer to head of the atomic notifier chain
155  *	@n: New entry in notifier chain
156  *
157  *	Adds a notifier to an atomic notifier chain.
158  *
159  *	Currently always returns zero.
160  */
161 
162 int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
163 		struct notifier_block *n)
164 {
165 	unsigned long flags;
166 	int ret;
167 
168 	spin_lock_irqsave(&nh->lock, flags);
169 	ret = notifier_chain_register(&nh->head, n);
170 	spin_unlock_irqrestore(&nh->lock, flags);
171 	return ret;
172 }
173 
174 EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
175 
176 /**
177  *	atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
178  *	@nh: Pointer to head of the atomic notifier chain
179  *	@n: Entry to remove from notifier chain
180  *
181  *	Removes a notifier from an atomic notifier chain.
182  *
183  *	Returns zero on success or %-ENOENT on failure.
184  */
185 int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
186 		struct notifier_block *n)
187 {
188 	unsigned long flags;
189 	int ret;
190 
191 	spin_lock_irqsave(&nh->lock, flags);
192 	ret = notifier_chain_unregister(&nh->head, n);
193 	spin_unlock_irqrestore(&nh->lock, flags);
194 	synchronize_rcu();
195 	return ret;
196 }
197 
198 EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
199 
200 /**
201  *	atomic_notifier_call_chain - Call functions in an atomic notifier chain
202  *	@nh: Pointer to head of the atomic notifier chain
203  *	@val: Value passed unmodified to notifier function
204  *	@v: Pointer passed unmodified to notifier function
205  *
206  *	Calls each function in a notifier chain in turn.  The functions
207  *	run in an atomic context, so they must not block.
208  *	This routine uses RCU to synchronize with changes to the chain.
209  *
210  *	If the return value of the notifier can be and'ed
211  *	with %NOTIFY_STOP_MASK then atomic_notifier_call_chain
212  *	will return immediately, with the return value of
213  *	the notifier function which halted execution.
214  *	Otherwise the return value is the return value
215  *	of the last notifier function called.
216  */
217 
218 int atomic_notifier_call_chain(struct atomic_notifier_head *nh,
219 		unsigned long val, void *v)
220 {
221 	int ret;
222 
223 	rcu_read_lock();
224 	ret = notifier_call_chain(&nh->head, val, v);
225 	rcu_read_unlock();
226 	return ret;
227 }
228 
229 EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
230 
231 /*
232  *	Blocking notifier chain routines.  All access to the chain is
233  *	synchronized by an rwsem.
234  */
235 
236 /**
237  *	blocking_notifier_chain_register - Add notifier to a blocking notifier chain
238  *	@nh: Pointer to head of the blocking notifier chain
239  *	@n: New entry in notifier chain
240  *
241  *	Adds a notifier to a blocking notifier chain.
242  *	Must be called in process context.
243  *
244  *	Currently always returns zero.
245  */
246 
247 int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
248 		struct notifier_block *n)
249 {
250 	int ret;
251 
252 	/*
253 	 * This code gets used during boot-up, when task switching is
254 	 * not yet working and interrupts must remain disabled.  At
255 	 * such times we must not call down_write().
256 	 */
257 	if (unlikely(system_state == SYSTEM_BOOTING))
258 		return notifier_chain_register(&nh->head, n);
259 
260 	down_write(&nh->rwsem);
261 	ret = notifier_chain_register(&nh->head, n);
262 	up_write(&nh->rwsem);
263 	return ret;
264 }
265 
266 EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
267 
268 /**
269  *	blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
270  *	@nh: Pointer to head of the blocking notifier chain
271  *	@n: Entry to remove from notifier chain
272  *
273  *	Removes a notifier from a blocking notifier chain.
274  *	Must be called from process context.
275  *
276  *	Returns zero on success or %-ENOENT on failure.
277  */
278 int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
279 		struct notifier_block *n)
280 {
281 	int ret;
282 
283 	/*
284 	 * This code gets used during boot-up, when task switching is
285 	 * not yet working and interrupts must remain disabled.  At
286 	 * such times we must not call down_write().
287 	 */
288 	if (unlikely(system_state == SYSTEM_BOOTING))
289 		return notifier_chain_unregister(&nh->head, n);
290 
291 	down_write(&nh->rwsem);
292 	ret = notifier_chain_unregister(&nh->head, n);
293 	up_write(&nh->rwsem);
294 	return ret;
295 }
296 
297 EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
298 
299 /**
300  *	blocking_notifier_call_chain - Call functions in a blocking notifier chain
301  *	@nh: Pointer to head of the blocking notifier chain
302  *	@val: Value passed unmodified to notifier function
303  *	@v: Pointer passed unmodified to notifier function
304  *
305  *	Calls each function in a notifier chain in turn.  The functions
306  *	run in a process context, so they are allowed to block.
307  *
308  *	If the return value of the notifier can be and'ed
309  *	with %NOTIFY_STOP_MASK then blocking_notifier_call_chain
310  *	will return immediately, with the return value of
311  *	the notifier function which halted execution.
312  *	Otherwise the return value is the return value
313  *	of the last notifier function called.
314  */
315 
316 int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
317 		unsigned long val, void *v)
318 {
319 	int ret;
320 
321 	down_read(&nh->rwsem);
322 	ret = notifier_call_chain(&nh->head, val, v);
323 	up_read(&nh->rwsem);
324 	return ret;
325 }
326 
327 EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
328 
329 /*
330  *	Raw notifier chain routines.  There is no protection;
331  *	the caller must provide it.  Use at your own risk!
332  */
333 
334 /**
335  *	raw_notifier_chain_register - Add notifier to a raw notifier chain
336  *	@nh: Pointer to head of the raw notifier chain
337  *	@n: New entry in notifier chain
338  *
339  *	Adds a notifier to a raw notifier chain.
340  *	All locking must be provided by the caller.
341  *
342  *	Currently always returns zero.
343  */
344 
345 int raw_notifier_chain_register(struct raw_notifier_head *nh,
346 		struct notifier_block *n)
347 {
348 	return notifier_chain_register(&nh->head, n);
349 }
350 
351 EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
352 
353 /**
354  *	raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
355  *	@nh: Pointer to head of the raw notifier chain
356  *	@n: Entry to remove from notifier chain
357  *
358  *	Removes a notifier from a raw notifier chain.
359  *	All locking must be provided by the caller.
360  *
361  *	Returns zero on success or %-ENOENT on failure.
362  */
363 int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
364 		struct notifier_block *n)
365 {
366 	return notifier_chain_unregister(&nh->head, n);
367 }
368 
369 EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
370 
371 /**
372  *	raw_notifier_call_chain - Call functions in a raw notifier chain
373  *	@nh: Pointer to head of the raw notifier chain
374  *	@val: Value passed unmodified to notifier function
375  *	@v: Pointer passed unmodified to notifier function
376  *
377  *	Calls each function in a notifier chain in turn.  The functions
378  *	run in an undefined context.
379  *	All locking must be provided by the caller.
380  *
381  *	If the return value of the notifier can be and'ed
382  *	with %NOTIFY_STOP_MASK then raw_notifier_call_chain
383  *	will return immediately, with the return value of
384  *	the notifier function which halted execution.
385  *	Otherwise the return value is the return value
386  *	of the last notifier function called.
387  */
388 
389 int raw_notifier_call_chain(struct raw_notifier_head *nh,
390 		unsigned long val, void *v)
391 {
392 	return notifier_call_chain(&nh->head, val, v);
393 }
394 
395 EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
396 
397 /**
398  *	register_reboot_notifier - Register function to be called at reboot time
399  *	@nb: Info about notifier function to be called
400  *
401  *	Registers a function with the list of functions
402  *	to be called at reboot time.
403  *
404  *	Currently always returns zero, as blocking_notifier_chain_register
405  *	always returns zero.
406  */
407 
408 int register_reboot_notifier(struct notifier_block * nb)
409 {
410 	return blocking_notifier_chain_register(&reboot_notifier_list, nb);
411 }
412 
413 EXPORT_SYMBOL(register_reboot_notifier);
414 
415 /**
416  *	unregister_reboot_notifier - Unregister previously registered reboot notifier
417  *	@nb: Hook to be unregistered
418  *
419  *	Unregisters a previously registered reboot
420  *	notifier function.
421  *
422  *	Returns zero on success, or %-ENOENT on failure.
423  */
424 
425 int unregister_reboot_notifier(struct notifier_block * nb)
426 {
427 	return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
428 }
429 
430 EXPORT_SYMBOL(unregister_reboot_notifier);
431 
432 static int set_one_prio(struct task_struct *p, int niceval, int error)
433 {
434 	int no_nice;
435 
436 	if (p->uid != current->euid &&
437 		p->euid != current->euid && !capable(CAP_SYS_NICE)) {
438 		error = -EPERM;
439 		goto out;
440 	}
441 	if (niceval < task_nice(p) && !can_nice(p, niceval)) {
442 		error = -EACCES;
443 		goto out;
444 	}
445 	no_nice = security_task_setnice(p, niceval);
446 	if (no_nice) {
447 		error = no_nice;
448 		goto out;
449 	}
450 	if (error == -ESRCH)
451 		error = 0;
452 	set_user_nice(p, niceval);
453 out:
454 	return error;
455 }
456 
457 asmlinkage long sys_setpriority(int which, int who, int niceval)
458 {
459 	struct task_struct *g, *p;
460 	struct user_struct *user;
461 	int error = -EINVAL;
462 
463 	if (which > 2 || which < 0)
464 		goto out;
465 
466 	/* normalize: avoid signed division (rounding problems) */
467 	error = -ESRCH;
468 	if (niceval < -20)
469 		niceval = -20;
470 	if (niceval > 19)
471 		niceval = 19;
472 
473 	read_lock(&tasklist_lock);
474 	switch (which) {
475 		case PRIO_PROCESS:
476 			if (!who)
477 				who = current->pid;
478 			p = find_task_by_pid(who);
479 			if (p)
480 				error = set_one_prio(p, niceval, error);
481 			break;
482 		case PRIO_PGRP:
483 			if (!who)
484 				who = process_group(current);
485 			do_each_task_pid(who, PIDTYPE_PGID, p) {
486 				error = set_one_prio(p, niceval, error);
487 			} while_each_task_pid(who, PIDTYPE_PGID, p);
488 			break;
489 		case PRIO_USER:
490 			user = current->user;
491 			if (!who)
492 				who = current->uid;
493 			else
494 				if ((who != current->uid) && !(user = find_user(who)))
495 					goto out_unlock;	/* No processes for this user */
496 
497 			do_each_thread(g, p)
498 				if (p->uid == who)
499 					error = set_one_prio(p, niceval, error);
500 			while_each_thread(g, p);
501 			if (who != current->uid)
502 				free_uid(user);		/* For find_user() */
503 			break;
504 	}
505 out_unlock:
506 	read_unlock(&tasklist_lock);
507 out:
508 	return error;
509 }
510 
511 /*
512  * Ugh. To avoid negative return values, "getpriority()" will
513  * not return the normal nice-value, but a negated value that
514  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
515  * to stay compatible.
516  */
517 asmlinkage long sys_getpriority(int which, int who)
518 {
519 	struct task_struct *g, *p;
520 	struct user_struct *user;
521 	long niceval, retval = -ESRCH;
522 
523 	if (which > 2 || which < 0)
524 		return -EINVAL;
525 
526 	read_lock(&tasklist_lock);
527 	switch (which) {
528 		case PRIO_PROCESS:
529 			if (!who)
530 				who = current->pid;
531 			p = find_task_by_pid(who);
532 			if (p) {
533 				niceval = 20 - task_nice(p);
534 				if (niceval > retval)
535 					retval = niceval;
536 			}
537 			break;
538 		case PRIO_PGRP:
539 			if (!who)
540 				who = process_group(current);
541 			do_each_task_pid(who, PIDTYPE_PGID, p) {
542 				niceval = 20 - task_nice(p);
543 				if (niceval > retval)
544 					retval = niceval;
545 			} while_each_task_pid(who, PIDTYPE_PGID, p);
546 			break;
547 		case PRIO_USER:
548 			user = current->user;
549 			if (!who)
550 				who = current->uid;
551 			else
552 				if ((who != current->uid) && !(user = find_user(who)))
553 					goto out_unlock;	/* No processes for this user */
554 
555 			do_each_thread(g, p)
556 				if (p->uid == who) {
557 					niceval = 20 - task_nice(p);
558 					if (niceval > retval)
559 						retval = niceval;
560 				}
561 			while_each_thread(g, p);
562 			if (who != current->uid)
563 				free_uid(user);		/* for find_user() */
564 			break;
565 	}
566 out_unlock:
567 	read_unlock(&tasklist_lock);
568 
569 	return retval;
570 }
571 
572 /**
573  *	emergency_restart - reboot the system
574  *
575  *	Without shutting down any hardware or taking any locks
576  *	reboot the system.  This is called when we know we are in
577  *	trouble so this is our best effort to reboot.  This is
578  *	safe to call in interrupt context.
579  */
580 void emergency_restart(void)
581 {
582 	machine_emergency_restart();
583 }
584 EXPORT_SYMBOL_GPL(emergency_restart);
585 
586 void kernel_restart_prepare(char *cmd)
587 {
588 	blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
589 	system_state = SYSTEM_RESTART;
590 	device_shutdown();
591 }
592 
593 /**
594  *	kernel_restart - reboot the system
595  *	@cmd: pointer to buffer containing command to execute for restart
596  *		or %NULL
597  *
598  *	Shutdown everything and perform a clean reboot.
599  *	This is not safe to call in interrupt context.
600  */
601 void kernel_restart(char *cmd)
602 {
603 	kernel_restart_prepare(cmd);
604 	if (!cmd) {
605 		printk(KERN_EMERG "Restarting system.\n");
606 	} else {
607 		printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
608 	}
609 	printk(".\n");
610 	machine_restart(cmd);
611 }
612 EXPORT_SYMBOL_GPL(kernel_restart);
613 
614 /**
615  *	kernel_kexec - reboot the system
616  *
617  *	Move into place and start executing a preloaded standalone
618  *	executable.  If nothing was preloaded return an error.
619  */
620 void kernel_kexec(void)
621 {
622 #ifdef CONFIG_KEXEC
623 	struct kimage *image;
624 	image = xchg(&kexec_image, NULL);
625 	if (!image) {
626 		return;
627 	}
628 	kernel_restart_prepare(NULL);
629 	printk(KERN_EMERG "Starting new kernel\n");
630 	machine_shutdown();
631 	machine_kexec(image);
632 #endif
633 }
634 EXPORT_SYMBOL_GPL(kernel_kexec);
635 
636 void kernel_shutdown_prepare(enum system_states state)
637 {
638 	blocking_notifier_call_chain(&reboot_notifier_list,
639 		(state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
640 	system_state = state;
641 	device_shutdown();
642 }
643 /**
644  *	kernel_halt - halt the system
645  *
646  *	Shutdown everything and perform a clean system halt.
647  */
648 void kernel_halt(void)
649 {
650 	kernel_shutdown_prepare(SYSTEM_HALT);
651 	printk(KERN_EMERG "System halted.\n");
652 	machine_halt();
653 }
654 
655 EXPORT_SYMBOL_GPL(kernel_halt);
656 
657 /**
658  *	kernel_power_off - power_off the system
659  *
660  *	Shutdown everything and perform a clean system power_off.
661  */
662 void kernel_power_off(void)
663 {
664 	kernel_shutdown_prepare(SYSTEM_POWER_OFF);
665 	printk(KERN_EMERG "Power down.\n");
666 	machine_power_off();
667 }
668 EXPORT_SYMBOL_GPL(kernel_power_off);
669 /*
670  * Reboot system call: for obvious reasons only root may call it,
671  * and even root needs to set up some magic numbers in the registers
672  * so that some mistake won't make this reboot the whole machine.
673  * You can also set the meaning of the ctrl-alt-del-key here.
674  *
675  * reboot doesn't sync: do that yourself before calling this.
676  */
677 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
678 {
679 	char buffer[256];
680 
681 	/* We only trust the superuser with rebooting the system. */
682 	if (!capable(CAP_SYS_BOOT))
683 		return -EPERM;
684 
685 	/* For safety, we require "magic" arguments. */
686 	if (magic1 != LINUX_REBOOT_MAGIC1 ||
687 	    (magic2 != LINUX_REBOOT_MAGIC2 &&
688 	                magic2 != LINUX_REBOOT_MAGIC2A &&
689 			magic2 != LINUX_REBOOT_MAGIC2B &&
690 	                magic2 != LINUX_REBOOT_MAGIC2C))
691 		return -EINVAL;
692 
693 	/* Instead of trying to make the power_off code look like
694 	 * halt when pm_power_off is not set do it the easy way.
695 	 */
696 	if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
697 		cmd = LINUX_REBOOT_CMD_HALT;
698 
699 	lock_kernel();
700 	switch (cmd) {
701 	case LINUX_REBOOT_CMD_RESTART:
702 		kernel_restart(NULL);
703 		break;
704 
705 	case LINUX_REBOOT_CMD_CAD_ON:
706 		C_A_D = 1;
707 		break;
708 
709 	case LINUX_REBOOT_CMD_CAD_OFF:
710 		C_A_D = 0;
711 		break;
712 
713 	case LINUX_REBOOT_CMD_HALT:
714 		kernel_halt();
715 		unlock_kernel();
716 		do_exit(0);
717 		break;
718 
719 	case LINUX_REBOOT_CMD_POWER_OFF:
720 		kernel_power_off();
721 		unlock_kernel();
722 		do_exit(0);
723 		break;
724 
725 	case LINUX_REBOOT_CMD_RESTART2:
726 		if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
727 			unlock_kernel();
728 			return -EFAULT;
729 		}
730 		buffer[sizeof(buffer) - 1] = '\0';
731 
732 		kernel_restart(buffer);
733 		break;
734 
735 	case LINUX_REBOOT_CMD_KEXEC:
736 		kernel_kexec();
737 		unlock_kernel();
738 		return -EINVAL;
739 
740 #ifdef CONFIG_SOFTWARE_SUSPEND
741 	case LINUX_REBOOT_CMD_SW_SUSPEND:
742 		{
743 			int ret = software_suspend();
744 			unlock_kernel();
745 			return ret;
746 		}
747 #endif
748 
749 	default:
750 		unlock_kernel();
751 		return -EINVAL;
752 	}
753 	unlock_kernel();
754 	return 0;
755 }
756 
757 static void deferred_cad(void *dummy)
758 {
759 	kernel_restart(NULL);
760 }
761 
762 /*
763  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
764  * As it's called within an interrupt, it may NOT sync: the only choice
765  * is whether to reboot at once, or just ignore the ctrl-alt-del.
766  */
767 void ctrl_alt_del(void)
768 {
769 	static DECLARE_WORK(cad_work, deferred_cad, NULL);
770 
771 	if (C_A_D)
772 		schedule_work(&cad_work);
773 	else
774 		kill_proc(cad_pid, SIGINT, 1);
775 }
776 
777 
778 /*
779  * Unprivileged users may change the real gid to the effective gid
780  * or vice versa.  (BSD-style)
781  *
782  * If you set the real gid at all, or set the effective gid to a value not
783  * equal to the real gid, then the saved gid is set to the new effective gid.
784  *
785  * This makes it possible for a setgid program to completely drop its
786  * privileges, which is often a useful assertion to make when you are doing
787  * a security audit over a program.
788  *
789  * The general idea is that a program which uses just setregid() will be
790  * 100% compatible with BSD.  A program which uses just setgid() will be
791  * 100% compatible with POSIX with saved IDs.
792  *
793  * SMP: There are not races, the GIDs are checked only by filesystem
794  *      operations (as far as semantic preservation is concerned).
795  */
796 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
797 {
798 	int old_rgid = current->gid;
799 	int old_egid = current->egid;
800 	int new_rgid = old_rgid;
801 	int new_egid = old_egid;
802 	int retval;
803 
804 	retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
805 	if (retval)
806 		return retval;
807 
808 	if (rgid != (gid_t) -1) {
809 		if ((old_rgid == rgid) ||
810 		    (current->egid==rgid) ||
811 		    capable(CAP_SETGID))
812 			new_rgid = rgid;
813 		else
814 			return -EPERM;
815 	}
816 	if (egid != (gid_t) -1) {
817 		if ((old_rgid == egid) ||
818 		    (current->egid == egid) ||
819 		    (current->sgid == egid) ||
820 		    capable(CAP_SETGID))
821 			new_egid = egid;
822 		else {
823 			return -EPERM;
824 		}
825 	}
826 	if (new_egid != old_egid)
827 	{
828 		current->mm->dumpable = suid_dumpable;
829 		smp_wmb();
830 	}
831 	if (rgid != (gid_t) -1 ||
832 	    (egid != (gid_t) -1 && egid != old_rgid))
833 		current->sgid = new_egid;
834 	current->fsgid = new_egid;
835 	current->egid = new_egid;
836 	current->gid = new_rgid;
837 	key_fsgid_changed(current);
838 	proc_id_connector(current, PROC_EVENT_GID);
839 	return 0;
840 }
841 
842 /*
843  * setgid() is implemented like SysV w/ SAVED_IDS
844  *
845  * SMP: Same implicit races as above.
846  */
847 asmlinkage long sys_setgid(gid_t gid)
848 {
849 	int old_egid = current->egid;
850 	int retval;
851 
852 	retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
853 	if (retval)
854 		return retval;
855 
856 	if (capable(CAP_SETGID))
857 	{
858 		if(old_egid != gid)
859 		{
860 			current->mm->dumpable = suid_dumpable;
861 			smp_wmb();
862 		}
863 		current->gid = current->egid = current->sgid = current->fsgid = gid;
864 	}
865 	else if ((gid == current->gid) || (gid == current->sgid))
866 	{
867 		if(old_egid != gid)
868 		{
869 			current->mm->dumpable = suid_dumpable;
870 			smp_wmb();
871 		}
872 		current->egid = current->fsgid = gid;
873 	}
874 	else
875 		return -EPERM;
876 
877 	key_fsgid_changed(current);
878 	proc_id_connector(current, PROC_EVENT_GID);
879 	return 0;
880 }
881 
882 static int set_user(uid_t new_ruid, int dumpclear)
883 {
884 	struct user_struct *new_user;
885 
886 	new_user = alloc_uid(new_ruid);
887 	if (!new_user)
888 		return -EAGAIN;
889 
890 	if (atomic_read(&new_user->processes) >=
891 				current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
892 			new_user != &root_user) {
893 		free_uid(new_user);
894 		return -EAGAIN;
895 	}
896 
897 	switch_uid(new_user);
898 
899 	if(dumpclear)
900 	{
901 		current->mm->dumpable = suid_dumpable;
902 		smp_wmb();
903 	}
904 	current->uid = new_ruid;
905 	return 0;
906 }
907 
908 /*
909  * Unprivileged users may change the real uid to the effective uid
910  * or vice versa.  (BSD-style)
911  *
912  * If you set the real uid at all, or set the effective uid to a value not
913  * equal to the real uid, then the saved uid is set to the new effective uid.
914  *
915  * This makes it possible for a setuid program to completely drop its
916  * privileges, which is often a useful assertion to make when you are doing
917  * a security audit over a program.
918  *
919  * The general idea is that a program which uses just setreuid() will be
920  * 100% compatible with BSD.  A program which uses just setuid() will be
921  * 100% compatible with POSIX with saved IDs.
922  */
923 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
924 {
925 	int old_ruid, old_euid, old_suid, new_ruid, new_euid;
926 	int retval;
927 
928 	retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
929 	if (retval)
930 		return retval;
931 
932 	new_ruid = old_ruid = current->uid;
933 	new_euid = old_euid = current->euid;
934 	old_suid = current->suid;
935 
936 	if (ruid != (uid_t) -1) {
937 		new_ruid = ruid;
938 		if ((old_ruid != ruid) &&
939 		    (current->euid != ruid) &&
940 		    !capable(CAP_SETUID))
941 			return -EPERM;
942 	}
943 
944 	if (euid != (uid_t) -1) {
945 		new_euid = euid;
946 		if ((old_ruid != euid) &&
947 		    (current->euid != euid) &&
948 		    (current->suid != euid) &&
949 		    !capable(CAP_SETUID))
950 			return -EPERM;
951 	}
952 
953 	if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
954 		return -EAGAIN;
955 
956 	if (new_euid != old_euid)
957 	{
958 		current->mm->dumpable = suid_dumpable;
959 		smp_wmb();
960 	}
961 	current->fsuid = current->euid = new_euid;
962 	if (ruid != (uid_t) -1 ||
963 	    (euid != (uid_t) -1 && euid != old_ruid))
964 		current->suid = current->euid;
965 	current->fsuid = current->euid;
966 
967 	key_fsuid_changed(current);
968 	proc_id_connector(current, PROC_EVENT_UID);
969 
970 	return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
971 }
972 
973 
974 
975 /*
976  * setuid() is implemented like SysV with SAVED_IDS
977  *
978  * Note that SAVED_ID's is deficient in that a setuid root program
979  * like sendmail, for example, cannot set its uid to be a normal
980  * user and then switch back, because if you're root, setuid() sets
981  * the saved uid too.  If you don't like this, blame the bright people
982  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
983  * will allow a root program to temporarily drop privileges and be able to
984  * regain them by swapping the real and effective uid.
985  */
986 asmlinkage long sys_setuid(uid_t uid)
987 {
988 	int old_euid = current->euid;
989 	int old_ruid, old_suid, new_ruid, new_suid;
990 	int retval;
991 
992 	retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
993 	if (retval)
994 		return retval;
995 
996 	old_ruid = new_ruid = current->uid;
997 	old_suid = current->suid;
998 	new_suid = old_suid;
999 
1000 	if (capable(CAP_SETUID)) {
1001 		if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
1002 			return -EAGAIN;
1003 		new_suid = uid;
1004 	} else if ((uid != current->uid) && (uid != new_suid))
1005 		return -EPERM;
1006 
1007 	if (old_euid != uid)
1008 	{
1009 		current->mm->dumpable = suid_dumpable;
1010 		smp_wmb();
1011 	}
1012 	current->fsuid = current->euid = uid;
1013 	current->suid = new_suid;
1014 
1015 	key_fsuid_changed(current);
1016 	proc_id_connector(current, PROC_EVENT_UID);
1017 
1018 	return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
1019 }
1020 
1021 
1022 /*
1023  * This function implements a generic ability to update ruid, euid,
1024  * and suid.  This allows you to implement the 4.4 compatible seteuid().
1025  */
1026 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1027 {
1028 	int old_ruid = current->uid;
1029 	int old_euid = current->euid;
1030 	int old_suid = current->suid;
1031 	int retval;
1032 
1033 	retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
1034 	if (retval)
1035 		return retval;
1036 
1037 	if (!capable(CAP_SETUID)) {
1038 		if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
1039 		    (ruid != current->euid) && (ruid != current->suid))
1040 			return -EPERM;
1041 		if ((euid != (uid_t) -1) && (euid != current->uid) &&
1042 		    (euid != current->euid) && (euid != current->suid))
1043 			return -EPERM;
1044 		if ((suid != (uid_t) -1) && (suid != current->uid) &&
1045 		    (suid != current->euid) && (suid != current->suid))
1046 			return -EPERM;
1047 	}
1048 	if (ruid != (uid_t) -1) {
1049 		if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
1050 			return -EAGAIN;
1051 	}
1052 	if (euid != (uid_t) -1) {
1053 		if (euid != current->euid)
1054 		{
1055 			current->mm->dumpable = suid_dumpable;
1056 			smp_wmb();
1057 		}
1058 		current->euid = euid;
1059 	}
1060 	current->fsuid = current->euid;
1061 	if (suid != (uid_t) -1)
1062 		current->suid = suid;
1063 
1064 	key_fsuid_changed(current);
1065 	proc_id_connector(current, PROC_EVENT_UID);
1066 
1067 	return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
1068 }
1069 
1070 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
1071 {
1072 	int retval;
1073 
1074 	if (!(retval = put_user(current->uid, ruid)) &&
1075 	    !(retval = put_user(current->euid, euid)))
1076 		retval = put_user(current->suid, suid);
1077 
1078 	return retval;
1079 }
1080 
1081 /*
1082  * Same as above, but for rgid, egid, sgid.
1083  */
1084 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1085 {
1086 	int retval;
1087 
1088 	retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
1089 	if (retval)
1090 		return retval;
1091 
1092 	if (!capable(CAP_SETGID)) {
1093 		if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
1094 		    (rgid != current->egid) && (rgid != current->sgid))
1095 			return -EPERM;
1096 		if ((egid != (gid_t) -1) && (egid != current->gid) &&
1097 		    (egid != current->egid) && (egid != current->sgid))
1098 			return -EPERM;
1099 		if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
1100 		    (sgid != current->egid) && (sgid != current->sgid))
1101 			return -EPERM;
1102 	}
1103 	if (egid != (gid_t) -1) {
1104 		if (egid != current->egid)
1105 		{
1106 			current->mm->dumpable = suid_dumpable;
1107 			smp_wmb();
1108 		}
1109 		current->egid = egid;
1110 	}
1111 	current->fsgid = current->egid;
1112 	if (rgid != (gid_t) -1)
1113 		current->gid = rgid;
1114 	if (sgid != (gid_t) -1)
1115 		current->sgid = sgid;
1116 
1117 	key_fsgid_changed(current);
1118 	proc_id_connector(current, PROC_EVENT_GID);
1119 	return 0;
1120 }
1121 
1122 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
1123 {
1124 	int retval;
1125 
1126 	if (!(retval = put_user(current->gid, rgid)) &&
1127 	    !(retval = put_user(current->egid, egid)))
1128 		retval = put_user(current->sgid, sgid);
1129 
1130 	return retval;
1131 }
1132 
1133 
1134 /*
1135  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1136  * is used for "access()" and for the NFS daemon (letting nfsd stay at
1137  * whatever uid it wants to). It normally shadows "euid", except when
1138  * explicitly set by setfsuid() or for access..
1139  */
1140 asmlinkage long sys_setfsuid(uid_t uid)
1141 {
1142 	int old_fsuid;
1143 
1144 	old_fsuid = current->fsuid;
1145 	if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
1146 		return old_fsuid;
1147 
1148 	if (uid == current->uid || uid == current->euid ||
1149 	    uid == current->suid || uid == current->fsuid ||
1150 	    capable(CAP_SETUID))
1151 	{
1152 		if (uid != old_fsuid)
1153 		{
1154 			current->mm->dumpable = suid_dumpable;
1155 			smp_wmb();
1156 		}
1157 		current->fsuid = uid;
1158 	}
1159 
1160 	key_fsuid_changed(current);
1161 	proc_id_connector(current, PROC_EVENT_UID);
1162 
1163 	security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1164 
1165 	return old_fsuid;
1166 }
1167 
1168 /*
1169  * Samma p� svenska..
1170  */
1171 asmlinkage long sys_setfsgid(gid_t gid)
1172 {
1173 	int old_fsgid;
1174 
1175 	old_fsgid = current->fsgid;
1176 	if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1177 		return old_fsgid;
1178 
1179 	if (gid == current->gid || gid == current->egid ||
1180 	    gid == current->sgid || gid == current->fsgid ||
1181 	    capable(CAP_SETGID))
1182 	{
1183 		if (gid != old_fsgid)
1184 		{
1185 			current->mm->dumpable = suid_dumpable;
1186 			smp_wmb();
1187 		}
1188 		current->fsgid = gid;
1189 		key_fsgid_changed(current);
1190 		proc_id_connector(current, PROC_EVENT_GID);
1191 	}
1192 	return old_fsgid;
1193 }
1194 
1195 asmlinkage long sys_times(struct tms __user * tbuf)
1196 {
1197 	/*
1198 	 *	In the SMP world we might just be unlucky and have one of
1199 	 *	the times increment as we use it. Since the value is an
1200 	 *	atomically safe type this is just fine. Conceptually its
1201 	 *	as if the syscall took an instant longer to occur.
1202 	 */
1203 	if (tbuf) {
1204 		struct tms tmp;
1205 		struct task_struct *tsk = current;
1206 		struct task_struct *t;
1207 		cputime_t utime, stime, cutime, cstime;
1208 
1209 		spin_lock_irq(&tsk->sighand->siglock);
1210 		utime = tsk->signal->utime;
1211 		stime = tsk->signal->stime;
1212 		t = tsk;
1213 		do {
1214 			utime = cputime_add(utime, t->utime);
1215 			stime = cputime_add(stime, t->stime);
1216 			t = next_thread(t);
1217 		} while (t != tsk);
1218 
1219 		cutime = tsk->signal->cutime;
1220 		cstime = tsk->signal->cstime;
1221 		spin_unlock_irq(&tsk->sighand->siglock);
1222 
1223 		tmp.tms_utime = cputime_to_clock_t(utime);
1224 		tmp.tms_stime = cputime_to_clock_t(stime);
1225 		tmp.tms_cutime = cputime_to_clock_t(cutime);
1226 		tmp.tms_cstime = cputime_to_clock_t(cstime);
1227 		if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1228 			return -EFAULT;
1229 	}
1230 	return (long) jiffies_64_to_clock_t(get_jiffies_64());
1231 }
1232 
1233 /*
1234  * This needs some heavy checking ...
1235  * I just haven't the stomach for it. I also don't fully
1236  * understand sessions/pgrp etc. Let somebody who does explain it.
1237  *
1238  * OK, I think I have the protection semantics right.... this is really
1239  * only important on a multi-user system anyway, to make sure one user
1240  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1241  *
1242  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1243  * LBT 04.03.94
1244  */
1245 
1246 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1247 {
1248 	struct task_struct *p;
1249 	struct task_struct *group_leader = current->group_leader;
1250 	int err = -EINVAL;
1251 
1252 	if (!pid)
1253 		pid = group_leader->pid;
1254 	if (!pgid)
1255 		pgid = pid;
1256 	if (pgid < 0)
1257 		return -EINVAL;
1258 
1259 	/* From this point forward we keep holding onto the tasklist lock
1260 	 * so that our parent does not change from under us. -DaveM
1261 	 */
1262 	write_lock_irq(&tasklist_lock);
1263 
1264 	err = -ESRCH;
1265 	p = find_task_by_pid(pid);
1266 	if (!p)
1267 		goto out;
1268 
1269 	err = -EINVAL;
1270 	if (!thread_group_leader(p))
1271 		goto out;
1272 
1273 	if (p->real_parent == group_leader) {
1274 		err = -EPERM;
1275 		if (p->signal->session != group_leader->signal->session)
1276 			goto out;
1277 		err = -EACCES;
1278 		if (p->did_exec)
1279 			goto out;
1280 	} else {
1281 		err = -ESRCH;
1282 		if (p != group_leader)
1283 			goto out;
1284 	}
1285 
1286 	err = -EPERM;
1287 	if (p->signal->leader)
1288 		goto out;
1289 
1290 	if (pgid != pid) {
1291 		struct task_struct *p;
1292 
1293 		do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1294 			if (p->signal->session == group_leader->signal->session)
1295 				goto ok_pgid;
1296 		} while_each_task_pid(pgid, PIDTYPE_PGID, p);
1297 		goto out;
1298 	}
1299 
1300 ok_pgid:
1301 	err = security_task_setpgid(p, pgid);
1302 	if (err)
1303 		goto out;
1304 
1305 	if (process_group(p) != pgid) {
1306 		detach_pid(p, PIDTYPE_PGID);
1307 		p->signal->pgrp = pgid;
1308 		attach_pid(p, PIDTYPE_PGID, pgid);
1309 	}
1310 
1311 	err = 0;
1312 out:
1313 	/* All paths lead to here, thus we are safe. -DaveM */
1314 	write_unlock_irq(&tasklist_lock);
1315 	return err;
1316 }
1317 
1318 asmlinkage long sys_getpgid(pid_t pid)
1319 {
1320 	if (!pid) {
1321 		return process_group(current);
1322 	} else {
1323 		int retval;
1324 		struct task_struct *p;
1325 
1326 		read_lock(&tasklist_lock);
1327 		p = find_task_by_pid(pid);
1328 
1329 		retval = -ESRCH;
1330 		if (p) {
1331 			retval = security_task_getpgid(p);
1332 			if (!retval)
1333 				retval = process_group(p);
1334 		}
1335 		read_unlock(&tasklist_lock);
1336 		return retval;
1337 	}
1338 }
1339 
1340 #ifdef __ARCH_WANT_SYS_GETPGRP
1341 
1342 asmlinkage long sys_getpgrp(void)
1343 {
1344 	/* SMP - assuming writes are word atomic this is fine */
1345 	return process_group(current);
1346 }
1347 
1348 #endif
1349 
1350 asmlinkage long sys_getsid(pid_t pid)
1351 {
1352 	if (!pid) {
1353 		return current->signal->session;
1354 	} else {
1355 		int retval;
1356 		struct task_struct *p;
1357 
1358 		read_lock(&tasklist_lock);
1359 		p = find_task_by_pid(pid);
1360 
1361 		retval = -ESRCH;
1362 		if(p) {
1363 			retval = security_task_getsid(p);
1364 			if (!retval)
1365 				retval = p->signal->session;
1366 		}
1367 		read_unlock(&tasklist_lock);
1368 		return retval;
1369 	}
1370 }
1371 
1372 asmlinkage long sys_setsid(void)
1373 {
1374 	struct task_struct *group_leader = current->group_leader;
1375 	pid_t session;
1376 	int err = -EPERM;
1377 
1378 	mutex_lock(&tty_mutex);
1379 	write_lock_irq(&tasklist_lock);
1380 
1381 	/* Fail if I am already a session leader */
1382 	if (group_leader->signal->leader)
1383 		goto out;
1384 
1385 	session = group_leader->pid;
1386 	/* Fail if a process group id already exists that equals the
1387 	 * proposed session id.
1388 	 *
1389 	 * Don't check if session id == 1 because kernel threads use this
1390 	 * session id and so the check will always fail and make it so
1391 	 * init cannot successfully call setsid.
1392 	 */
1393 	if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
1394 		goto out;
1395 
1396 	group_leader->signal->leader = 1;
1397 	__set_special_pids(session, session);
1398 	group_leader->signal->tty = NULL;
1399 	group_leader->signal->tty_old_pgrp = 0;
1400 	err = process_group(group_leader);
1401 out:
1402 	write_unlock_irq(&tasklist_lock);
1403 	mutex_unlock(&tty_mutex);
1404 	return err;
1405 }
1406 
1407 /*
1408  * Supplementary group IDs
1409  */
1410 
1411 /* init to 2 - one for init_task, one to ensure it is never freed */
1412 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1413 
1414 struct group_info *groups_alloc(int gidsetsize)
1415 {
1416 	struct group_info *group_info;
1417 	int nblocks;
1418 	int i;
1419 
1420 	nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1421 	/* Make sure we always allocate at least one indirect block pointer */
1422 	nblocks = nblocks ? : 1;
1423 	group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1424 	if (!group_info)
1425 		return NULL;
1426 	group_info->ngroups = gidsetsize;
1427 	group_info->nblocks = nblocks;
1428 	atomic_set(&group_info->usage, 1);
1429 
1430 	if (gidsetsize <= NGROUPS_SMALL) {
1431 		group_info->blocks[0] = group_info->small_block;
1432 	} else {
1433 		for (i = 0; i < nblocks; i++) {
1434 			gid_t *b;
1435 			b = (void *)__get_free_page(GFP_USER);
1436 			if (!b)
1437 				goto out_undo_partial_alloc;
1438 			group_info->blocks[i] = b;
1439 		}
1440 	}
1441 	return group_info;
1442 
1443 out_undo_partial_alloc:
1444 	while (--i >= 0) {
1445 		free_page((unsigned long)group_info->blocks[i]);
1446 	}
1447 	kfree(group_info);
1448 	return NULL;
1449 }
1450 
1451 EXPORT_SYMBOL(groups_alloc);
1452 
1453 void groups_free(struct group_info *group_info)
1454 {
1455 	if (group_info->blocks[0] != group_info->small_block) {
1456 		int i;
1457 		for (i = 0; i < group_info->nblocks; i++)
1458 			free_page((unsigned long)group_info->blocks[i]);
1459 	}
1460 	kfree(group_info);
1461 }
1462 
1463 EXPORT_SYMBOL(groups_free);
1464 
1465 /* export the group_info to a user-space array */
1466 static int groups_to_user(gid_t __user *grouplist,
1467     struct group_info *group_info)
1468 {
1469 	int i;
1470 	int count = group_info->ngroups;
1471 
1472 	for (i = 0; i < group_info->nblocks; i++) {
1473 		int cp_count = min(NGROUPS_PER_BLOCK, count);
1474 		int off = i * NGROUPS_PER_BLOCK;
1475 		int len = cp_count * sizeof(*grouplist);
1476 
1477 		if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1478 			return -EFAULT;
1479 
1480 		count -= cp_count;
1481 	}
1482 	return 0;
1483 }
1484 
1485 /* fill a group_info from a user-space array - it must be allocated already */
1486 static int groups_from_user(struct group_info *group_info,
1487     gid_t __user *grouplist)
1488  {
1489 	int i;
1490 	int count = group_info->ngroups;
1491 
1492 	for (i = 0; i < group_info->nblocks; i++) {
1493 		int cp_count = min(NGROUPS_PER_BLOCK, count);
1494 		int off = i * NGROUPS_PER_BLOCK;
1495 		int len = cp_count * sizeof(*grouplist);
1496 
1497 		if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1498 			return -EFAULT;
1499 
1500 		count -= cp_count;
1501 	}
1502 	return 0;
1503 }
1504 
1505 /* a simple Shell sort */
1506 static void groups_sort(struct group_info *group_info)
1507 {
1508 	int base, max, stride;
1509 	int gidsetsize = group_info->ngroups;
1510 
1511 	for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1512 		; /* nothing */
1513 	stride /= 3;
1514 
1515 	while (stride) {
1516 		max = gidsetsize - stride;
1517 		for (base = 0; base < max; base++) {
1518 			int left = base;
1519 			int right = left + stride;
1520 			gid_t tmp = GROUP_AT(group_info, right);
1521 
1522 			while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1523 				GROUP_AT(group_info, right) =
1524 				    GROUP_AT(group_info, left);
1525 				right = left;
1526 				left -= stride;
1527 			}
1528 			GROUP_AT(group_info, right) = tmp;
1529 		}
1530 		stride /= 3;
1531 	}
1532 }
1533 
1534 /* a simple bsearch */
1535 int groups_search(struct group_info *group_info, gid_t grp)
1536 {
1537 	unsigned int left, right;
1538 
1539 	if (!group_info)
1540 		return 0;
1541 
1542 	left = 0;
1543 	right = group_info->ngroups;
1544 	while (left < right) {
1545 		unsigned int mid = (left+right)/2;
1546 		int cmp = grp - GROUP_AT(group_info, mid);
1547 		if (cmp > 0)
1548 			left = mid + 1;
1549 		else if (cmp < 0)
1550 			right = mid;
1551 		else
1552 			return 1;
1553 	}
1554 	return 0;
1555 }
1556 
1557 /* validate and set current->group_info */
1558 int set_current_groups(struct group_info *group_info)
1559 {
1560 	int retval;
1561 	struct group_info *old_info;
1562 
1563 	retval = security_task_setgroups(group_info);
1564 	if (retval)
1565 		return retval;
1566 
1567 	groups_sort(group_info);
1568 	get_group_info(group_info);
1569 
1570 	task_lock(current);
1571 	old_info = current->group_info;
1572 	current->group_info = group_info;
1573 	task_unlock(current);
1574 
1575 	put_group_info(old_info);
1576 
1577 	return 0;
1578 }
1579 
1580 EXPORT_SYMBOL(set_current_groups);
1581 
1582 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1583 {
1584 	int i = 0;
1585 
1586 	/*
1587 	 *	SMP: Nobody else can change our grouplist. Thus we are
1588 	 *	safe.
1589 	 */
1590 
1591 	if (gidsetsize < 0)
1592 		return -EINVAL;
1593 
1594 	/* no need to grab task_lock here; it cannot change */
1595 	i = current->group_info->ngroups;
1596 	if (gidsetsize) {
1597 		if (i > gidsetsize) {
1598 			i = -EINVAL;
1599 			goto out;
1600 		}
1601 		if (groups_to_user(grouplist, current->group_info)) {
1602 			i = -EFAULT;
1603 			goto out;
1604 		}
1605 	}
1606 out:
1607 	return i;
1608 }
1609 
1610 /*
1611  *	SMP: Our groups are copy-on-write. We can set them safely
1612  *	without another task interfering.
1613  */
1614 
1615 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1616 {
1617 	struct group_info *group_info;
1618 	int retval;
1619 
1620 	if (!capable(CAP_SETGID))
1621 		return -EPERM;
1622 	if ((unsigned)gidsetsize > NGROUPS_MAX)
1623 		return -EINVAL;
1624 
1625 	group_info = groups_alloc(gidsetsize);
1626 	if (!group_info)
1627 		return -ENOMEM;
1628 	retval = groups_from_user(group_info, grouplist);
1629 	if (retval) {
1630 		put_group_info(group_info);
1631 		return retval;
1632 	}
1633 
1634 	retval = set_current_groups(group_info);
1635 	put_group_info(group_info);
1636 
1637 	return retval;
1638 }
1639 
1640 /*
1641  * Check whether we're fsgid/egid or in the supplemental group..
1642  */
1643 int in_group_p(gid_t grp)
1644 {
1645 	int retval = 1;
1646 	if (grp != current->fsgid) {
1647 		retval = groups_search(current->group_info, grp);
1648 	}
1649 	return retval;
1650 }
1651 
1652 EXPORT_SYMBOL(in_group_p);
1653 
1654 int in_egroup_p(gid_t grp)
1655 {
1656 	int retval = 1;
1657 	if (grp != current->egid) {
1658 		retval = groups_search(current->group_info, grp);
1659 	}
1660 	return retval;
1661 }
1662 
1663 EXPORT_SYMBOL(in_egroup_p);
1664 
1665 DECLARE_RWSEM(uts_sem);
1666 
1667 EXPORT_SYMBOL(uts_sem);
1668 
1669 asmlinkage long sys_newuname(struct new_utsname __user * name)
1670 {
1671 	int errno = 0;
1672 
1673 	down_read(&uts_sem);
1674 	if (copy_to_user(name,&system_utsname,sizeof *name))
1675 		errno = -EFAULT;
1676 	up_read(&uts_sem);
1677 	return errno;
1678 }
1679 
1680 asmlinkage long sys_sethostname(char __user *name, int len)
1681 {
1682 	int errno;
1683 	char tmp[__NEW_UTS_LEN];
1684 
1685 	if (!capable(CAP_SYS_ADMIN))
1686 		return -EPERM;
1687 	if (len < 0 || len > __NEW_UTS_LEN)
1688 		return -EINVAL;
1689 	down_write(&uts_sem);
1690 	errno = -EFAULT;
1691 	if (!copy_from_user(tmp, name, len)) {
1692 		memcpy(system_utsname.nodename, tmp, len);
1693 		system_utsname.nodename[len] = 0;
1694 		errno = 0;
1695 	}
1696 	up_write(&uts_sem);
1697 	return errno;
1698 }
1699 
1700 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1701 
1702 asmlinkage long sys_gethostname(char __user *name, int len)
1703 {
1704 	int i, errno;
1705 
1706 	if (len < 0)
1707 		return -EINVAL;
1708 	down_read(&uts_sem);
1709 	i = 1 + strlen(system_utsname.nodename);
1710 	if (i > len)
1711 		i = len;
1712 	errno = 0;
1713 	if (copy_to_user(name, system_utsname.nodename, i))
1714 		errno = -EFAULT;
1715 	up_read(&uts_sem);
1716 	return errno;
1717 }
1718 
1719 #endif
1720 
1721 /*
1722  * Only setdomainname; getdomainname can be implemented by calling
1723  * uname()
1724  */
1725 asmlinkage long sys_setdomainname(char __user *name, int len)
1726 {
1727 	int errno;
1728 	char tmp[__NEW_UTS_LEN];
1729 
1730 	if (!capable(CAP_SYS_ADMIN))
1731 		return -EPERM;
1732 	if (len < 0 || len > __NEW_UTS_LEN)
1733 		return -EINVAL;
1734 
1735 	down_write(&uts_sem);
1736 	errno = -EFAULT;
1737 	if (!copy_from_user(tmp, name, len)) {
1738 		memcpy(system_utsname.domainname, tmp, len);
1739 		system_utsname.domainname[len] = 0;
1740 		errno = 0;
1741 	}
1742 	up_write(&uts_sem);
1743 	return errno;
1744 }
1745 
1746 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1747 {
1748 	if (resource >= RLIM_NLIMITS)
1749 		return -EINVAL;
1750 	else {
1751 		struct rlimit value;
1752 		task_lock(current->group_leader);
1753 		value = current->signal->rlim[resource];
1754 		task_unlock(current->group_leader);
1755 		return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1756 	}
1757 }
1758 
1759 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1760 
1761 /*
1762  *	Back compatibility for getrlimit. Needed for some apps.
1763  */
1764 
1765 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1766 {
1767 	struct rlimit x;
1768 	if (resource >= RLIM_NLIMITS)
1769 		return -EINVAL;
1770 
1771 	task_lock(current->group_leader);
1772 	x = current->signal->rlim[resource];
1773 	task_unlock(current->group_leader);
1774 	if(x.rlim_cur > 0x7FFFFFFF)
1775 		x.rlim_cur = 0x7FFFFFFF;
1776 	if(x.rlim_max > 0x7FFFFFFF)
1777 		x.rlim_max = 0x7FFFFFFF;
1778 	return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1779 }
1780 
1781 #endif
1782 
1783 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1784 {
1785 	struct rlimit new_rlim, *old_rlim;
1786 	unsigned long it_prof_secs;
1787 	int retval;
1788 
1789 	if (resource >= RLIM_NLIMITS)
1790 		return -EINVAL;
1791 	if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1792 		return -EFAULT;
1793 	if (new_rlim.rlim_cur > new_rlim.rlim_max)
1794 		return -EINVAL;
1795 	old_rlim = current->signal->rlim + resource;
1796 	if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1797 	    !capable(CAP_SYS_RESOURCE))
1798 		return -EPERM;
1799 	if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1800 		return -EPERM;
1801 
1802 	retval = security_task_setrlimit(resource, &new_rlim);
1803 	if (retval)
1804 		return retval;
1805 
1806 	task_lock(current->group_leader);
1807 	*old_rlim = new_rlim;
1808 	task_unlock(current->group_leader);
1809 
1810 	if (resource != RLIMIT_CPU)
1811 		goto out;
1812 
1813 	/*
1814 	 * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1815 	 * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1816 	 * very long-standing error, and fixing it now risks breakage of
1817 	 * applications, so we live with it
1818 	 */
1819 	if (new_rlim.rlim_cur == RLIM_INFINITY)
1820 		goto out;
1821 
1822 	it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
1823 	if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
1824 		unsigned long rlim_cur = new_rlim.rlim_cur;
1825 		cputime_t cputime;
1826 
1827 		if (rlim_cur == 0) {
1828 			/*
1829 			 * The caller is asking for an immediate RLIMIT_CPU
1830 			 * expiry.  But we use the zero value to mean "it was
1831 			 * never set".  So let's cheat and make it one second
1832 			 * instead
1833 			 */
1834 			rlim_cur = 1;
1835 		}
1836 		cputime = secs_to_cputime(rlim_cur);
1837 		read_lock(&tasklist_lock);
1838 		spin_lock_irq(&current->sighand->siglock);
1839 		set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
1840 		spin_unlock_irq(&current->sighand->siglock);
1841 		read_unlock(&tasklist_lock);
1842 	}
1843 out:
1844 	return 0;
1845 }
1846 
1847 /*
1848  * It would make sense to put struct rusage in the task_struct,
1849  * except that would make the task_struct be *really big*.  After
1850  * task_struct gets moved into malloc'ed memory, it would
1851  * make sense to do this.  It will make moving the rest of the information
1852  * a lot simpler!  (Which we're not doing right now because we're not
1853  * measuring them yet).
1854  *
1855  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1856  * races with threads incrementing their own counters.  But since word
1857  * reads are atomic, we either get new values or old values and we don't
1858  * care which for the sums.  We always take the siglock to protect reading
1859  * the c* fields from p->signal from races with exit.c updating those
1860  * fields when reaping, so a sample either gets all the additions of a
1861  * given child after it's reaped, or none so this sample is before reaping.
1862  *
1863  * tasklist_lock locking optimisation:
1864  * If we are current and single threaded, we do not need to take the tasklist
1865  * lock or the siglock.  No one else can take our signal_struct away,
1866  * no one else can reap the children to update signal->c* counters, and
1867  * no one else can race with the signal-> fields.
1868  * If we do not take the tasklist_lock, the signal-> fields could be read
1869  * out of order while another thread was just exiting. So we place a
1870  * read memory barrier when we avoid the lock.  On the writer side,
1871  * write memory barrier is implied in  __exit_signal as __exit_signal releases
1872  * the siglock spinlock after updating the signal-> fields.
1873  *
1874  * We don't really need the siglock when we access the non c* fields
1875  * of the signal_struct (for RUSAGE_SELF) even in multithreaded
1876  * case, since we take the tasklist lock for read and the non c* signal->
1877  * fields are updated only in __exit_signal, which is called with
1878  * tasklist_lock taken for write, hence these two threads cannot execute
1879  * concurrently.
1880  *
1881  */
1882 
1883 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1884 {
1885 	struct task_struct *t;
1886 	unsigned long flags;
1887 	cputime_t utime, stime;
1888 	int need_lock = 0;
1889 
1890 	memset((char *) r, 0, sizeof *r);
1891 	utime = stime = cputime_zero;
1892 
1893 	if (p != current || !thread_group_empty(p))
1894 		need_lock = 1;
1895 
1896 	if (need_lock) {
1897 		read_lock(&tasklist_lock);
1898 		if (unlikely(!p->signal)) {
1899 			read_unlock(&tasklist_lock);
1900 			return;
1901 		}
1902 	} else
1903 		/* See locking comments above */
1904 		smp_rmb();
1905 
1906 	switch (who) {
1907 		case RUSAGE_BOTH:
1908 		case RUSAGE_CHILDREN:
1909 			spin_lock_irqsave(&p->sighand->siglock, flags);
1910 			utime = p->signal->cutime;
1911 			stime = p->signal->cstime;
1912 			r->ru_nvcsw = p->signal->cnvcsw;
1913 			r->ru_nivcsw = p->signal->cnivcsw;
1914 			r->ru_minflt = p->signal->cmin_flt;
1915 			r->ru_majflt = p->signal->cmaj_flt;
1916 			spin_unlock_irqrestore(&p->sighand->siglock, flags);
1917 
1918 			if (who == RUSAGE_CHILDREN)
1919 				break;
1920 
1921 		case RUSAGE_SELF:
1922 			utime = cputime_add(utime, p->signal->utime);
1923 			stime = cputime_add(stime, p->signal->stime);
1924 			r->ru_nvcsw += p->signal->nvcsw;
1925 			r->ru_nivcsw += p->signal->nivcsw;
1926 			r->ru_minflt += p->signal->min_flt;
1927 			r->ru_majflt += p->signal->maj_flt;
1928 			t = p;
1929 			do {
1930 				utime = cputime_add(utime, t->utime);
1931 				stime = cputime_add(stime, t->stime);
1932 				r->ru_nvcsw += t->nvcsw;
1933 				r->ru_nivcsw += t->nivcsw;
1934 				r->ru_minflt += t->min_flt;
1935 				r->ru_majflt += t->maj_flt;
1936 				t = next_thread(t);
1937 			} while (t != p);
1938 			break;
1939 
1940 		default:
1941 			BUG();
1942 	}
1943 
1944 	if (need_lock)
1945 		read_unlock(&tasklist_lock);
1946 	cputime_to_timeval(utime, &r->ru_utime);
1947 	cputime_to_timeval(stime, &r->ru_stime);
1948 }
1949 
1950 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1951 {
1952 	struct rusage r;
1953 	k_getrusage(p, who, &r);
1954 	return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1955 }
1956 
1957 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1958 {
1959 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1960 		return -EINVAL;
1961 	return getrusage(current, who, ru);
1962 }
1963 
1964 asmlinkage long sys_umask(int mask)
1965 {
1966 	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1967 	return mask;
1968 }
1969 
1970 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1971 			  unsigned long arg4, unsigned long arg5)
1972 {
1973 	long error;
1974 
1975 	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1976 	if (error)
1977 		return error;
1978 
1979 	switch (option) {
1980 		case PR_SET_PDEATHSIG:
1981 			if (!valid_signal(arg2)) {
1982 				error = -EINVAL;
1983 				break;
1984 			}
1985 			current->pdeath_signal = arg2;
1986 			break;
1987 		case PR_GET_PDEATHSIG:
1988 			error = put_user(current->pdeath_signal, (int __user *)arg2);
1989 			break;
1990 		case PR_GET_DUMPABLE:
1991 			error = current->mm->dumpable;
1992 			break;
1993 		case PR_SET_DUMPABLE:
1994 			if (arg2 < 0 || arg2 > 2) {
1995 				error = -EINVAL;
1996 				break;
1997 			}
1998 			current->mm->dumpable = arg2;
1999 			break;
2000 
2001 		case PR_SET_UNALIGN:
2002 			error = SET_UNALIGN_CTL(current, arg2);
2003 			break;
2004 		case PR_GET_UNALIGN:
2005 			error = GET_UNALIGN_CTL(current, arg2);
2006 			break;
2007 		case PR_SET_FPEMU:
2008 			error = SET_FPEMU_CTL(current, arg2);
2009 			break;
2010 		case PR_GET_FPEMU:
2011 			error = GET_FPEMU_CTL(current, arg2);
2012 			break;
2013 		case PR_SET_FPEXC:
2014 			error = SET_FPEXC_CTL(current, arg2);
2015 			break;
2016 		case PR_GET_FPEXC:
2017 			error = GET_FPEXC_CTL(current, arg2);
2018 			break;
2019 		case PR_GET_TIMING:
2020 			error = PR_TIMING_STATISTICAL;
2021 			break;
2022 		case PR_SET_TIMING:
2023 			if (arg2 == PR_TIMING_STATISTICAL)
2024 				error = 0;
2025 			else
2026 				error = -EINVAL;
2027 			break;
2028 
2029 		case PR_GET_KEEPCAPS:
2030 			if (current->keep_capabilities)
2031 				error = 1;
2032 			break;
2033 		case PR_SET_KEEPCAPS:
2034 			if (arg2 != 0 && arg2 != 1) {
2035 				error = -EINVAL;
2036 				break;
2037 			}
2038 			current->keep_capabilities = arg2;
2039 			break;
2040 		case PR_SET_NAME: {
2041 			struct task_struct *me = current;
2042 			unsigned char ncomm[sizeof(me->comm)];
2043 
2044 			ncomm[sizeof(me->comm)-1] = 0;
2045 			if (strncpy_from_user(ncomm, (char __user *)arg2,
2046 						sizeof(me->comm)-1) < 0)
2047 				return -EFAULT;
2048 			set_task_comm(me, ncomm);
2049 			return 0;
2050 		}
2051 		case PR_GET_NAME: {
2052 			struct task_struct *me = current;
2053 			unsigned char tcomm[sizeof(me->comm)];
2054 
2055 			get_task_comm(tcomm, me);
2056 			if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
2057 				return -EFAULT;
2058 			return 0;
2059 		}
2060 		default:
2061 			error = -EINVAL;
2062 			break;
2063 	}
2064 	return error;
2065 }
2066