blob: 438d99a38c87f343dd318c3966799ec73bb6072a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/sys.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/highuid.h>
16#include <linux/fs.h>
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +010017#include <linux/perf_counter.h>
Daniel Walker3e88c552007-05-10 22:22:53 -070018#include <linux/resource.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070019#include <linux/kernel.h>
20#include <linux/kexec.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/workqueue.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080022#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#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>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070031#include <linux/signal.h>
Matt Helsley9f460802005-11-07 00:59:16 -080032#include <linux/cn_proc.h>
Andi Kleen3cfc3482006-09-26 10:52:28 +020033#include <linux/getcpu.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070034#include <linux/task_io_accounting_ops.h>
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -070035#include <linux/seccomp.h>
Mark Lord40477272007-10-01 01:20:10 -070036#include <linux/cpu.h>
Paul Mackerrase3d5a272009-01-06 14:41:02 -080037#include <linux/ptrace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040038#include <linux/fs_struct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
40#include <linux/compat.h>
41#include <linux/syscalls.h>
Keshavamurthy Anil S00d7c052005-12-12 00:37:33 -080042#include <linux/kprobes.h>
Cedric Le Goateracce2922007-07-15 23:40:59 -070043#include <linux/user_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45#include <asm/uaccess.h>
46#include <asm/io.h>
47#include <asm/unistd.h>
48
49#ifndef SET_UNALIGN_CTL
50# define SET_UNALIGN_CTL(a,b) (-EINVAL)
51#endif
52#ifndef GET_UNALIGN_CTL
53# define GET_UNALIGN_CTL(a,b) (-EINVAL)
54#endif
55#ifndef SET_FPEMU_CTL
56# define SET_FPEMU_CTL(a,b) (-EINVAL)
57#endif
58#ifndef GET_FPEMU_CTL
59# define GET_FPEMU_CTL(a,b) (-EINVAL)
60#endif
61#ifndef SET_FPEXC_CTL
62# define SET_FPEXC_CTL(a,b) (-EINVAL)
63#endif
64#ifndef GET_FPEXC_CTL
65# define GET_FPEXC_CTL(a,b) (-EINVAL)
66#endif
Anton Blanchard651d7652006-06-07 16:10:19 +100067#ifndef GET_ENDIAN
68# define GET_ENDIAN(a,b) (-EINVAL)
69#endif
70#ifndef SET_ENDIAN
71# define SET_ENDIAN(a,b) (-EINVAL)
72#endif
Erik Bosman8fb402b2008-04-11 18:54:17 +020073#ifndef GET_TSC_CTL
74# define GET_TSC_CTL(a) (-EINVAL)
75#endif
76#ifndef SET_TSC_CTL
77# define SET_TSC_CTL(a) (-EINVAL)
78#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070079
80/*
81 * this is where the system-wide overflow UID and GID are defined, for
82 * architectures that now have 32-bit UID/GID but didn't in the past
83 */
84
85int overflowuid = DEFAULT_OVERFLOWUID;
86int overflowgid = DEFAULT_OVERFLOWGID;
87
88#ifdef CONFIG_UID16
89EXPORT_SYMBOL(overflowuid);
90EXPORT_SYMBOL(overflowgid);
91#endif
92
93/*
94 * the same as above, but for filesystems which can only store a 16-bit
95 * UID and GID. as such, this is needed on all architectures
96 */
97
98int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
99int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
100
101EXPORT_SYMBOL(fs_overflowuid);
102EXPORT_SYMBOL(fs_overflowgid);
103
104/*
105 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
106 */
107
108int C_A_D = 1;
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700109struct pid *cad_pid;
110EXPORT_SYMBOL(cad_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
112/*
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700113 * If set, this is used for preparing the system to power off.
114 */
115
116void (*pm_power_off_prepare)(void);
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700117
David Howellsc69e8d92008-11-14 10:39:19 +1100118/*
119 * set the priority of a task
120 * - the caller must hold the RCU read lock
121 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122static int set_one_prio(struct task_struct *p, int niceval, int error)
123{
David Howellsc69e8d92008-11-14 10:39:19 +1100124 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 int no_nice;
126
David Howellsc69e8d92008-11-14 10:39:19 +1100127 if (pcred->uid != cred->euid &&
128 pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 error = -EPERM;
130 goto out;
131 }
Matt Mackalle43379f2005-05-01 08:59:00 -0700132 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 error = -EACCES;
134 goto out;
135 }
136 no_nice = security_task_setnice(p, niceval);
137 if (no_nice) {
138 error = no_nice;
139 goto out;
140 }
141 if (error == -ESRCH)
142 error = 0;
143 set_user_nice(p, niceval);
144out:
145 return error;
146}
147
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100148SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149{
150 struct task_struct *g, *p;
151 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100152 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 int error = -EINVAL;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800154 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
Daniel Walker3e88c552007-05-10 22:22:53 -0700156 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 goto out;
158
159 /* normalize: avoid signed division (rounding problems) */
160 error = -ESRCH;
161 if (niceval < -20)
162 niceval = -20;
163 if (niceval > 19)
164 niceval = 19;
165
166 read_lock(&tasklist_lock);
167 switch (which) {
168 case PRIO_PROCESS:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800169 if (who)
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700170 p = find_task_by_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800171 else
172 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 if (p)
174 error = set_one_prio(p, niceval, error);
175 break;
176 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800177 if (who)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700178 pgrp = find_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800179 else
180 pgrp = task_pgrp(current);
Ken Chen2d70b682008-08-20 14:09:17 -0700181 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 error = set_one_prio(p, niceval, error);
Ken Chen2d70b682008-08-20 14:09:17 -0700183 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 break;
185 case PRIO_USER:
David Howellsd84f4f92008-11-14 10:39:23 +1100186 user = (struct user_struct *) cred->user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 if (!who)
David Howells86a264a2008-11-14 10:39:18 +1100188 who = cred->uid;
189 else if ((who != cred->uid) &&
190 !(user = find_user(who)))
191 goto out_unlock; /* No processes for this user */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
193 do_each_thread(g, p)
David Howells86a264a2008-11-14 10:39:18 +1100194 if (__task_cred(p)->uid == who)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 error = set_one_prio(p, niceval, error);
196 while_each_thread(g, p);
David Howells86a264a2008-11-14 10:39:18 +1100197 if (who != cred->uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 free_uid(user); /* For find_user() */
199 break;
200 }
201out_unlock:
202 read_unlock(&tasklist_lock);
203out:
204 return error;
205}
206
207/*
208 * Ugh. To avoid negative return values, "getpriority()" will
209 * not return the normal nice-value, but a negated value that
210 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
211 * to stay compatible.
212 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100213SYSCALL_DEFINE2(getpriority, int, which, int, who)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 struct task_struct *g, *p;
216 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100217 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 long niceval, retval = -ESRCH;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800219 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Daniel Walker3e88c552007-05-10 22:22:53 -0700221 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 return -EINVAL;
223
224 read_lock(&tasklist_lock);
225 switch (which) {
226 case PRIO_PROCESS:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800227 if (who)
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700228 p = find_task_by_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800229 else
230 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 if (p) {
232 niceval = 20 - task_nice(p);
233 if (niceval > retval)
234 retval = niceval;
235 }
236 break;
237 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800238 if (who)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700239 pgrp = find_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800240 else
241 pgrp = task_pgrp(current);
Ken Chen2d70b682008-08-20 14:09:17 -0700242 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 niceval = 20 - task_nice(p);
244 if (niceval > retval)
245 retval = niceval;
Ken Chen2d70b682008-08-20 14:09:17 -0700246 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 break;
248 case PRIO_USER:
David Howells86a264a2008-11-14 10:39:18 +1100249 user = (struct user_struct *) cred->user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 if (!who)
David Howells86a264a2008-11-14 10:39:18 +1100251 who = cred->uid;
252 else if ((who != cred->uid) &&
253 !(user = find_user(who)))
254 goto out_unlock; /* No processes for this user */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256 do_each_thread(g, p)
David Howells86a264a2008-11-14 10:39:18 +1100257 if (__task_cred(p)->uid == who) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 niceval = 20 - task_nice(p);
259 if (niceval > retval)
260 retval = niceval;
261 }
262 while_each_thread(g, p);
David Howells86a264a2008-11-14 10:39:18 +1100263 if (who != cred->uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 free_uid(user); /* for find_user() */
265 break;
266 }
267out_unlock:
268 read_unlock(&tasklist_lock);
269
270 return retval;
271}
272
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700273/**
274 * emergency_restart - reboot the system
275 *
276 * Without shutting down any hardware or taking any locks
277 * reboot the system. This is called when we know we are in
278 * trouble so this is our best effort to reboot. This is
279 * safe to call in interrupt context.
280 */
Eric W. Biederman7c903472005-07-26 11:29:55 -0600281void emergency_restart(void)
282{
283 machine_emergency_restart();
284}
285EXPORT_SYMBOL_GPL(emergency_restart);
286
Huang Yingca195b72008-08-15 00:40:24 -0700287void kernel_restart_prepare(char *cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600288{
Alan Sterne041c682006-03-27 01:16:30 -0800289 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600290 system_state = SYSTEM_RESTART;
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600291 device_shutdown();
Rafael J. Wysocki58b3b712007-07-26 16:29:55 +0200292 sysdev_shutdown();
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700293}
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800294
295/**
296 * kernel_restart - reboot the system
297 * @cmd: pointer to buffer containing command to execute for restart
Randy Dunlapb8887e62005-11-07 01:01:07 -0800298 * or %NULL
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800299 *
300 * Shutdown everything and perform a clean reboot.
301 * This is not safe to call in interrupt context.
302 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700303void kernel_restart(char *cmd)
304{
305 kernel_restart_prepare(cmd);
Cal Peake756184b2006-09-30 23:27:24 -0700306 if (!cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600307 printk(KERN_EMERG "Restarting system.\n");
Cal Peake756184b2006-09-30 23:27:24 -0700308 else
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600309 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600310 machine_restart(cmd);
311}
312EXPORT_SYMBOL_GPL(kernel_restart);
313
Adrian Bunk4ef72292008-02-04 22:30:06 -0800314static void kernel_shutdown_prepare(enum system_states state)
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500315{
Alan Sterne041c682006-03-27 01:16:30 -0800316 blocking_notifier_call_chain(&reboot_notifier_list,
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500317 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
318 system_state = state;
319 device_shutdown();
320}
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700321/**
322 * kernel_halt - halt the system
323 *
324 * Shutdown everything and perform a clean system halt.
325 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700326void kernel_halt(void)
327{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500328 kernel_shutdown_prepare(SYSTEM_HALT);
Rafael J. Wysocki58b3b712007-07-26 16:29:55 +0200329 sysdev_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600330 printk(KERN_EMERG "System halted.\n");
331 machine_halt();
332}
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500333
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600334EXPORT_SYMBOL_GPL(kernel_halt);
335
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700336/**
337 * kernel_power_off - power_off the system
338 *
339 * Shutdown everything and perform a clean system power_off.
340 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700341void kernel_power_off(void)
342{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500343 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700344 if (pm_power_off_prepare)
345 pm_power_off_prepare();
Mark Lord40477272007-10-01 01:20:10 -0700346 disable_nonboot_cpus();
Rafael J. Wysocki58b3b712007-07-26 16:29:55 +0200347 sysdev_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600348 printk(KERN_EMERG "Power down.\n");
349 machine_power_off();
350}
351EXPORT_SYMBOL_GPL(kernel_power_off);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352/*
353 * Reboot system call: for obvious reasons only root may call it,
354 * and even root needs to set up some magic numbers in the registers
355 * so that some mistake won't make this reboot the whole machine.
356 * You can also set the meaning of the ctrl-alt-del-key here.
357 *
358 * reboot doesn't sync: do that yourself before calling this.
359 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100360SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
361 void __user *, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362{
363 char buffer[256];
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700364 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
366 /* We only trust the superuser with rebooting the system. */
367 if (!capable(CAP_SYS_BOOT))
368 return -EPERM;
369
370 /* For safety, we require "magic" arguments. */
371 if (magic1 != LINUX_REBOOT_MAGIC1 ||
372 (magic2 != LINUX_REBOOT_MAGIC2 &&
373 magic2 != LINUX_REBOOT_MAGIC2A &&
374 magic2 != LINUX_REBOOT_MAGIC2B &&
375 magic2 != LINUX_REBOOT_MAGIC2C))
376 return -EINVAL;
377
Eric W. Biederman5e382912006-01-08 01:03:46 -0800378 /* Instead of trying to make the power_off code look like
379 * halt when pm_power_off is not set do it the easy way.
380 */
381 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
382 cmd = LINUX_REBOOT_CMD_HALT;
383
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 lock_kernel();
385 switch (cmd) {
386 case LINUX_REBOOT_CMD_RESTART:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600387 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 break;
389
390 case LINUX_REBOOT_CMD_CAD_ON:
391 C_A_D = 1;
392 break;
393
394 case LINUX_REBOOT_CMD_CAD_OFF:
395 C_A_D = 0;
396 break;
397
398 case LINUX_REBOOT_CMD_HALT:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600399 kernel_halt();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 unlock_kernel();
401 do_exit(0);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700402 panic("cannot halt");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403
404 case LINUX_REBOOT_CMD_POWER_OFF:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600405 kernel_power_off();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 unlock_kernel();
407 do_exit(0);
408 break;
409
410 case LINUX_REBOOT_CMD_RESTART2:
411 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
412 unlock_kernel();
413 return -EFAULT;
414 }
415 buffer[sizeof(buffer) - 1] = '\0';
416
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600417 kernel_restart(buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 break;
419
Huang Ying3ab83522008-07-25 19:45:07 -0700420#ifdef CONFIG_KEXEC
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700421 case LINUX_REBOOT_CMD_KEXEC:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700422 ret = kernel_kexec();
423 break;
Huang Ying3ab83522008-07-25 19:45:07 -0700424#endif
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600425
Rafael J. Wysockib0cb1a12007-07-29 23:24:36 +0200426#ifdef CONFIG_HIBERNATION
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 case LINUX_REBOOT_CMD_SW_SUSPEND:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700428 ret = hibernate();
429 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430#endif
431
432 default:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700433 ret = -EINVAL;
434 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 }
436 unlock_kernel();
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700437 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438}
439
David Howells65f27f32006-11-22 14:55:48 +0000440static void deferred_cad(struct work_struct *dummy)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441{
Eric W. Biedermanabcd9e52005-07-26 11:27:34 -0600442 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443}
444
445/*
446 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
447 * As it's called within an interrupt, it may NOT sync: the only choice
448 * is whether to reboot at once, or just ignore the ctrl-alt-del.
449 */
450void ctrl_alt_del(void)
451{
David Howells65f27f32006-11-22 14:55:48 +0000452 static DECLARE_WORK(cad_work, deferred_cad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453
454 if (C_A_D)
455 schedule_work(&cad_work);
456 else
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700457 kill_cad_pid(SIGINT, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458}
459
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460/*
461 * Unprivileged users may change the real gid to the effective gid
462 * or vice versa. (BSD-style)
463 *
464 * If you set the real gid at all, or set the effective gid to a value not
465 * equal to the real gid, then the saved gid is set to the new effective gid.
466 *
467 * This makes it possible for a setgid program to completely drop its
468 * privileges, which is often a useful assertion to make when you are doing
469 * a security audit over a program.
470 *
471 * The general idea is that a program which uses just setregid() will be
472 * 100% compatible with BSD. A program which uses just setgid() will be
473 * 100% compatible with POSIX with saved IDs.
474 *
475 * SMP: There are not races, the GIDs are checked only by filesystem
476 * operations (as far as semantic preservation is concerned).
477 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100478SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479{
David Howellsd84f4f92008-11-14 10:39:23 +1100480 const struct cred *old;
481 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 int retval;
483
David Howellsd84f4f92008-11-14 10:39:23 +1100484 new = prepare_creds();
485 if (!new)
486 return -ENOMEM;
487 old = current_cred();
488
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
490 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +1100491 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
David Howellsd84f4f92008-11-14 10:39:23 +1100493 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 if (rgid != (gid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100495 if (old->gid == rgid ||
496 old->egid == rgid ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 capable(CAP_SETGID))
David Howellsd84f4f92008-11-14 10:39:23 +1100498 new->gid = rgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 else
David Howellsd84f4f92008-11-14 10:39:23 +1100500 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 }
502 if (egid != (gid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100503 if (old->gid == egid ||
504 old->egid == egid ||
505 old->sgid == egid ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 capable(CAP_SETGID))
David Howellsd84f4f92008-11-14 10:39:23 +1100507 new->egid = egid;
Cal Peake756184b2006-09-30 23:27:24 -0700508 else
David Howellsd84f4f92008-11-14 10:39:23 +1100509 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 }
David Howellsd84f4f92008-11-14 10:39:23 +1100511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 if (rgid != (gid_t) -1 ||
David Howellsd84f4f92008-11-14 10:39:23 +1100513 (egid != (gid_t) -1 && egid != old->gid))
514 new->sgid = new->egid;
515 new->fsgid = new->egid;
516
517 return commit_creds(new);
518
519error:
520 abort_creds(new);
521 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522}
523
524/*
525 * setgid() is implemented like SysV w/ SAVED_IDS
526 *
527 * SMP: Same implicit races as above.
528 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100529SYSCALL_DEFINE1(setgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530{
David Howellsd84f4f92008-11-14 10:39:23 +1100531 const struct cred *old;
532 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 int retval;
534
David Howellsd84f4f92008-11-14 10:39:23 +1100535 new = prepare_creds();
536 if (!new)
537 return -ENOMEM;
538 old = current_cred();
539
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
541 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +1100542 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543
David Howellsd84f4f92008-11-14 10:39:23 +1100544 retval = -EPERM;
545 if (capable(CAP_SETGID))
546 new->gid = new->egid = new->sgid = new->fsgid = gid;
547 else if (gid == old->gid || gid == old->sgid)
548 new->egid = new->fsgid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 else
David Howellsd84f4f92008-11-14 10:39:23 +1100550 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
David Howellsd84f4f92008-11-14 10:39:23 +1100552 return commit_creds(new);
553
554error:
555 abort_creds(new);
556 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557}
Dhaval Giani54e99122009-02-27 15:13:54 +0530558
David Howellsd84f4f92008-11-14 10:39:23 +1100559/*
560 * change the user struct in a credentials set to match the new UID
561 */
562static int set_user(struct cred *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563{
564 struct user_struct *new_user;
565
Serge Hallyn18b6e042008-10-15 16:38:45 -0500566 new_user = alloc_uid(current_user_ns(), new->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 if (!new_user)
568 return -EAGAIN;
569
Dhaval Giani54e99122009-02-27 15:13:54 +0530570 if (!task_can_switch_user(new_user, current)) {
571 free_uid(new_user);
572 return -EINVAL;
573 }
574
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 if (atomic_read(&new_user->processes) >=
576 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
Serge Hallyn18b6e042008-10-15 16:38:45 -0500577 new_user != INIT_USER) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 free_uid(new_user);
579 return -EAGAIN;
580 }
581
David Howellsd84f4f92008-11-14 10:39:23 +1100582 free_uid(new->user);
583 new->user = new_user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 return 0;
585}
586
587/*
588 * Unprivileged users may change the real uid to the effective uid
589 * or vice versa. (BSD-style)
590 *
591 * If you set the real uid at all, or set the effective uid to a value not
592 * equal to the real uid, then the saved uid is set to the new effective uid.
593 *
594 * This makes it possible for a setuid program to completely drop its
595 * privileges, which is often a useful assertion to make when you are doing
596 * a security audit over a program.
597 *
598 * The general idea is that a program which uses just setreuid() will be
599 * 100% compatible with BSD. A program which uses just setuid() will be
600 * 100% compatible with POSIX with saved IDs.
601 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100602SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
David Howellsd84f4f92008-11-14 10:39:23 +1100604 const struct cred *old;
605 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 int retval;
607
David Howellsd84f4f92008-11-14 10:39:23 +1100608 new = prepare_creds();
609 if (!new)
610 return -ENOMEM;
611 old = current_cred();
612
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
614 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +1100615 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
David Howellsd84f4f92008-11-14 10:39:23 +1100617 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 if (ruid != (uid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100619 new->uid = ruid;
620 if (old->uid != ruid &&
621 old->euid != ruid &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 !capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100623 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 }
625
626 if (euid != (uid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100627 new->euid = euid;
628 if (old->uid != euid &&
629 old->euid != euid &&
630 old->suid != euid &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 !capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100632 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 }
634
Dhaval Giani54e99122009-02-27 15:13:54 +0530635 if (new->uid != old->uid) {
636 retval = set_user(new);
637 if (retval < 0)
638 goto error;
639 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 if (ruid != (uid_t) -1 ||
David Howellsd84f4f92008-11-14 10:39:23 +1100641 (euid != (uid_t) -1 && euid != old->uid))
642 new->suid = new->euid;
643 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
David Howellsd84f4f92008-11-14 10:39:23 +1100645 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
646 if (retval < 0)
647 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
David Howellsd84f4f92008-11-14 10:39:23 +1100649 return commit_creds(new);
650
651error:
652 abort_creds(new);
653 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656/*
657 * setuid() is implemented like SysV with SAVED_IDS
658 *
659 * Note that SAVED_ID's is deficient in that a setuid root program
660 * like sendmail, for example, cannot set its uid to be a normal
661 * user and then switch back, because if you're root, setuid() sets
662 * the saved uid too. If you don't like this, blame the bright people
663 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
664 * will allow a root program to temporarily drop privileges and be able to
665 * regain them by swapping the real and effective uid.
666 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100667SYSCALL_DEFINE1(setuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668{
David Howellsd84f4f92008-11-14 10:39:23 +1100669 const struct cred *old;
670 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 int retval;
672
David Howellsd84f4f92008-11-14 10:39:23 +1100673 new = prepare_creds();
674 if (!new)
675 return -ENOMEM;
676 old = current_cred();
677
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
679 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +1100680 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681
David Howellsd84f4f92008-11-14 10:39:23 +1100682 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 if (capable(CAP_SETUID)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100684 new->suid = new->uid = uid;
Dhaval Giani54e99122009-02-27 15:13:54 +0530685 if (uid != old->uid) {
686 retval = set_user(new);
687 if (retval < 0)
688 goto error;
David Howellsd84f4f92008-11-14 10:39:23 +1100689 }
690 } else if (uid != old->uid && uid != new->suid) {
691 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
David Howellsd84f4f92008-11-14 10:39:23 +1100694 new->fsuid = new->euid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
David Howellsd84f4f92008-11-14 10:39:23 +1100696 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
697 if (retval < 0)
698 goto error;
699
700 return commit_creds(new);
701
702error:
703 abort_creds(new);
704 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705}
706
707
708/*
709 * This function implements a generic ability to update ruid, euid,
710 * and suid. This allows you to implement the 4.4 compatible seteuid().
711 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100712SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713{
David Howellsd84f4f92008-11-14 10:39:23 +1100714 const struct cred *old;
715 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 int retval;
717
David Howellsd84f4f92008-11-14 10:39:23 +1100718 new = prepare_creds();
719 if (!new)
720 return -ENOMEM;
721
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
723 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +1100724 goto error;
725 old = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726
David Howellsd84f4f92008-11-14 10:39:23 +1100727 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 if (!capable(CAP_SETUID)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100729 if (ruid != (uid_t) -1 && ruid != old->uid &&
730 ruid != old->euid && ruid != old->suid)
731 goto error;
732 if (euid != (uid_t) -1 && euid != old->uid &&
733 euid != old->euid && euid != old->suid)
734 goto error;
735 if (suid != (uid_t) -1 && suid != old->uid &&
736 suid != old->euid && suid != old->suid)
737 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 }
David Howellsd84f4f92008-11-14 10:39:23 +1100739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 if (ruid != (uid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100741 new->uid = ruid;
Dhaval Giani54e99122009-02-27 15:13:54 +0530742 if (ruid != old->uid) {
743 retval = set_user(new);
744 if (retval < 0)
745 goto error;
746 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 }
David Howellsd84f4f92008-11-14 10:39:23 +1100748 if (euid != (uid_t) -1)
749 new->euid = euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 if (suid != (uid_t) -1)
David Howellsd84f4f92008-11-14 10:39:23 +1100751 new->suid = suid;
752 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
David Howellsd84f4f92008-11-14 10:39:23 +1100754 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
755 if (retval < 0)
756 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
David Howellsd84f4f92008-11-14 10:39:23 +1100758 return commit_creds(new);
759
760error:
761 abort_creds(new);
762 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763}
764
Heiko Carstensdbf040d2009-01-14 14:14:04 +0100765SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766{
David Howells86a264a2008-11-14 10:39:18 +1100767 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 int retval;
769
David Howells86a264a2008-11-14 10:39:18 +1100770 if (!(retval = put_user(cred->uid, ruid)) &&
771 !(retval = put_user(cred->euid, euid)))
David Howellsb6dff3e2008-11-14 10:39:16 +1100772 retval = put_user(cred->suid, suid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 return retval;
775}
776
777/*
778 * Same as above, but for rgid, egid, sgid.
779 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100780SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781{
David Howellsd84f4f92008-11-14 10:39:23 +1100782 const struct cred *old;
783 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 int retval;
785
David Howellsd84f4f92008-11-14 10:39:23 +1100786 new = prepare_creds();
787 if (!new)
788 return -ENOMEM;
789 old = current_cred();
790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
792 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +1100793 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
David Howellsd84f4f92008-11-14 10:39:23 +1100795 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 if (!capable(CAP_SETGID)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100797 if (rgid != (gid_t) -1 && rgid != old->gid &&
798 rgid != old->egid && rgid != old->sgid)
799 goto error;
800 if (egid != (gid_t) -1 && egid != old->gid &&
801 egid != old->egid && egid != old->sgid)
802 goto error;
803 if (sgid != (gid_t) -1 && sgid != old->gid &&
804 sgid != old->egid && sgid != old->sgid)
805 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
David Howellsd84f4f92008-11-14 10:39:23 +1100808 if (rgid != (gid_t) -1)
809 new->gid = rgid;
810 if (egid != (gid_t) -1)
811 new->egid = egid;
812 if (sgid != (gid_t) -1)
813 new->sgid = sgid;
814 new->fsgid = new->egid;
815
816 return commit_creds(new);
817
818error:
819 abort_creds(new);
820 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821}
822
Heiko Carstensdbf040d2009-01-14 14:14:04 +0100823SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
David Howells86a264a2008-11-14 10:39:18 +1100825 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 int retval;
827
David Howells86a264a2008-11-14 10:39:18 +1100828 if (!(retval = put_user(cred->gid, rgid)) &&
829 !(retval = put_user(cred->egid, egid)))
David Howellsb6dff3e2008-11-14 10:39:16 +1100830 retval = put_user(cred->sgid, sgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
832 return retval;
833}
834
835
836/*
837 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
838 * is used for "access()" and for the NFS daemon (letting nfsd stay at
839 * whatever uid it wants to). It normally shadows "euid", except when
840 * explicitly set by setfsuid() or for access..
841 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100842SYSCALL_DEFINE1(setfsuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843{
David Howellsd84f4f92008-11-14 10:39:23 +1100844 const struct cred *old;
845 struct cred *new;
846 uid_t old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
David Howellsd84f4f92008-11-14 10:39:23 +1100848 new = prepare_creds();
849 if (!new)
850 return current_fsuid();
851 old = current_cred();
852 old_fsuid = old->fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
David Howellsd84f4f92008-11-14 10:39:23 +1100854 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
855 goto error;
856
857 if (uid == old->uid || uid == old->euid ||
858 uid == old->suid || uid == old->fsuid ||
Cal Peake756184b2006-09-30 23:27:24 -0700859 capable(CAP_SETUID)) {
860 if (uid != old_fsuid) {
David Howellsd84f4f92008-11-14 10:39:23 +1100861 new->fsuid = uid;
862 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
863 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 }
866
David Howellsd84f4f92008-11-14 10:39:23 +1100867error:
868 abort_creds(new);
869 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
David Howellsd84f4f92008-11-14 10:39:23 +1100871change_okay:
872 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 return old_fsuid;
874}
875
876/*
John Anthony Kazos Jrf42df9e2007-05-09 08:23:08 +0200877 * Samma på svenska..
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100879SYSCALL_DEFINE1(setfsgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880{
David Howellsd84f4f92008-11-14 10:39:23 +1100881 const struct cred *old;
882 struct cred *new;
883 gid_t old_fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884
David Howellsd84f4f92008-11-14 10:39:23 +1100885 new = prepare_creds();
886 if (!new)
887 return current_fsgid();
888 old = current_cred();
889 old_fsgid = old->fsgid;
890
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
David Howellsd84f4f92008-11-14 10:39:23 +1100892 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
David Howellsd84f4f92008-11-14 10:39:23 +1100894 if (gid == old->gid || gid == old->egid ||
895 gid == old->sgid || gid == old->fsgid ||
Cal Peake756184b2006-09-30 23:27:24 -0700896 capable(CAP_SETGID)) {
897 if (gid != old_fsgid) {
David Howellsd84f4f92008-11-14 10:39:23 +1100898 new->fsgid = gid;
899 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 }
David Howellsd84f4f92008-11-14 10:39:23 +1100902
903error:
904 abort_creds(new);
905 return old_fsgid;
906
907change_okay:
908 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 return old_fsgid;
910}
911
Frank Mayharf06febc2008-09-12 09:54:39 -0700912void do_sys_times(struct tms *tms)
913{
914 struct task_cputime cputime;
915 cputime_t cutime, cstime;
916
Frank Mayharf06febc2008-09-12 09:54:39 -0700917 thread_group_cputime(current, &cputime);
Oleg Nesterov2b5fe6d2008-11-17 15:40:08 +0100918 spin_lock_irq(&current->sighand->siglock);
Frank Mayharf06febc2008-09-12 09:54:39 -0700919 cutime = current->signal->cutime;
920 cstime = current->signal->cstime;
921 spin_unlock_irq(&current->sighand->siglock);
922 tms->tms_utime = cputime_to_clock_t(cputime.utime);
923 tms->tms_stime = cputime_to_clock_t(cputime.stime);
924 tms->tms_cutime = cputime_to_clock_t(cutime);
925 tms->tms_cstime = cputime_to_clock_t(cstime);
926}
927
Heiko Carstens58fd3aa2009-01-14 14:14:03 +0100928SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 if (tbuf) {
931 struct tms tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932
Frank Mayharf06febc2008-09-12 09:54:39 -0700933 do_sys_times(&tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
935 return -EFAULT;
936 }
Paul Mackerrase3d5a272009-01-06 14:41:02 -0800937 force_successful_syscall_return();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 return (long) jiffies_64_to_clock_t(get_jiffies_64());
939}
940
941/*
942 * This needs some heavy checking ...
943 * I just haven't the stomach for it. I also don't fully
944 * understand sessions/pgrp etc. Let somebody who does explain it.
945 *
946 * OK, I think I have the protection semantics right.... this is really
947 * only important on a multi-user system anyway, to make sure one user
948 * can't send a signal to a process owned by another. -TYT, 12/12/91
949 *
950 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
951 * LBT 04.03.94
952 */
Heiko Carstensb290ebe2009-01-14 14:14:06 +0100953SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954{
955 struct task_struct *p;
Oleg Nesterovee0acf92006-01-08 01:03:53 -0800956 struct task_struct *group_leader = current->group_leader;
Oleg Nesterov4e021302008-02-08 04:19:08 -0800957 struct pid *pgrp;
958 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960 if (!pid)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700961 pid = task_pid_vnr(group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 if (!pgid)
963 pgid = pid;
964 if (pgid < 0)
965 return -EINVAL;
966
967 /* From this point forward we keep holding onto the tasklist lock
968 * so that our parent does not change from under us. -DaveM
969 */
970 write_lock_irq(&tasklist_lock);
971
972 err = -ESRCH;
Oleg Nesterov4e021302008-02-08 04:19:08 -0800973 p = find_task_by_vpid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 if (!p)
975 goto out;
976
977 err = -EINVAL;
978 if (!thread_group_leader(p))
979 goto out;
980
Oleg Nesterov4e021302008-02-08 04:19:08 -0800981 if (same_thread_group(p->real_parent, group_leader)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 err = -EPERM;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800983 if (task_session(p) != task_session(group_leader))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 goto out;
985 err = -EACCES;
986 if (p->did_exec)
987 goto out;
988 } else {
989 err = -ESRCH;
Oleg Nesterovee0acf92006-01-08 01:03:53 -0800990 if (p != group_leader)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 goto out;
992 }
993
994 err = -EPERM;
995 if (p->signal->leader)
996 goto out;
997
Oleg Nesterov4e021302008-02-08 04:19:08 -0800998 pgrp = task_pid(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 if (pgid != pid) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001000 struct task_struct *g;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
Oleg Nesterov4e021302008-02-08 04:19:08 -08001002 pgrp = find_vpid(pgid);
1003 g = pid_task(pgrp, PIDTYPE_PGID);
Eric W. Biederman41487c62007-02-12 00:53:01 -08001004 if (!g || task_session(g) != task_session(group_leader))
Oleg Nesterovf020bc42006-12-08 02:38:02 -08001005 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 }
1007
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 err = security_task_setpgid(p, pgid);
1009 if (err)
1010 goto out;
1011
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -07001012 if (task_pgrp(p) != pgrp)
Oleg Nesterov83beaf32008-04-30 00:54:27 -07001013 change_pid(p, PIDTYPE_PGID, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
1015 err = 0;
1016out:
1017 /* All paths lead to here, thus we are safe. -DaveM */
1018 write_unlock_irq(&tasklist_lock);
1019 return err;
1020}
1021
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001022SYSCALL_DEFINE1(getpgid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001024 struct task_struct *p;
1025 struct pid *grp;
1026 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001028 rcu_read_lock();
1029 if (!pid)
1030 grp = task_pgrp(current);
1031 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 retval = -ESRCH;
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001033 p = find_task_by_vpid(pid);
1034 if (!p)
1035 goto out;
1036 grp = task_pgrp(p);
1037 if (!grp)
1038 goto out;
1039
1040 retval = security_task_getpgid(p);
1041 if (retval)
1042 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 }
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001044 retval = pid_vnr(grp);
1045out:
1046 rcu_read_unlock();
1047 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048}
1049
1050#ifdef __ARCH_WANT_SYS_GETPGRP
1051
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001052SYSCALL_DEFINE0(getpgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001054 return sys_getpgid(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055}
1056
1057#endif
1058
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001059SYSCALL_DEFINE1(getsid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060{
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001061 struct task_struct *p;
1062 struct pid *sid;
1063 int retval;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001064
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001065 rcu_read_lock();
1066 if (!pid)
1067 sid = task_session(current);
1068 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 retval = -ESRCH;
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001070 p = find_task_by_vpid(pid);
1071 if (!p)
1072 goto out;
1073 sid = task_session(p);
1074 if (!sid)
1075 goto out;
1076
1077 retval = security_task_getsid(p);
1078 if (retval)
1079 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 }
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001081 retval = pid_vnr(sid);
1082out:
1083 rcu_read_unlock();
1084 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085}
1086
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001087SYSCALL_DEFINE0(setsid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088{
Oren Laadane19f2472006-01-08 01:03:58 -08001089 struct task_struct *group_leader = current->group_leader;
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001090 struct pid *sid = task_pid(group_leader);
1091 pid_t session = pid_vnr(sid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 int err = -EPERM;
1093
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 write_lock_irq(&tasklist_lock);
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001095 /* Fail if I am already a session leader */
1096 if (group_leader->signal->leader)
1097 goto out;
1098
Oleg Nesterov430c6232008-02-08 04:19:11 -08001099 /* Fail if a process group id already exists that equals the
1100 * proposed session id.
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001101 */
Oleg Nesterov6806aac2008-02-08 04:19:12 -08001102 if (pid_task(sid, PIDTYPE_PGID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 goto out;
1104
Oren Laadane19f2472006-01-08 01:03:58 -08001105 group_leader->signal->leader = 1;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -08001106 __set_special_pids(sid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001107
Alan Cox9c9f4de2008-10-13 10:37:26 +01001108 proc_clear_tty(group_leader);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001109
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001110 err = session;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111out:
1112 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 return err;
1114}
1115
1116/*
1117 * Supplementary group IDs
1118 */
1119
1120/* init to 2 - one for init_task, one to ensure it is never freed */
1121struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1122
1123struct group_info *groups_alloc(int gidsetsize)
1124{
1125 struct group_info *group_info;
1126 int nblocks;
1127 int i;
1128
1129 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1130 /* Make sure we always allocate at least one indirect block pointer */
1131 nblocks = nblocks ? : 1;
1132 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1133 if (!group_info)
1134 return NULL;
1135 group_info->ngroups = gidsetsize;
1136 group_info->nblocks = nblocks;
1137 atomic_set(&group_info->usage, 1);
1138
Cal Peake756184b2006-09-30 23:27:24 -07001139 if (gidsetsize <= NGROUPS_SMALL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 group_info->blocks[0] = group_info->small_block;
Cal Peake756184b2006-09-30 23:27:24 -07001141 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 for (i = 0; i < nblocks; i++) {
1143 gid_t *b;
1144 b = (void *)__get_free_page(GFP_USER);
1145 if (!b)
1146 goto out_undo_partial_alloc;
1147 group_info->blocks[i] = b;
1148 }
1149 }
1150 return group_info;
1151
1152out_undo_partial_alloc:
1153 while (--i >= 0) {
1154 free_page((unsigned long)group_info->blocks[i]);
1155 }
1156 kfree(group_info);
1157 return NULL;
1158}
1159
1160EXPORT_SYMBOL(groups_alloc);
1161
1162void groups_free(struct group_info *group_info)
1163{
1164 if (group_info->blocks[0] != group_info->small_block) {
1165 int i;
1166 for (i = 0; i < group_info->nblocks; i++)
1167 free_page((unsigned long)group_info->blocks[i]);
1168 }
1169 kfree(group_info);
1170}
1171
1172EXPORT_SYMBOL(groups_free);
1173
1174/* export the group_info to a user-space array */
1175static int groups_to_user(gid_t __user *grouplist,
David Howellsd84f4f92008-11-14 10:39:23 +11001176 const struct group_info *group_info)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177{
1178 int i;
Eric Dumazet1bf47342008-02-06 01:37:56 -08001179 unsigned int count = group_info->ngroups;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
1181 for (i = 0; i < group_info->nblocks; i++) {
Eric Dumazet1bf47342008-02-06 01:37:56 -08001182 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1183 unsigned int len = cp_count * sizeof(*grouplist);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184
Eric Dumazet1bf47342008-02-06 01:37:56 -08001185 if (copy_to_user(grouplist, group_info->blocks[i], len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 return -EFAULT;
1187
Eric Dumazet1bf47342008-02-06 01:37:56 -08001188 grouplist += NGROUPS_PER_BLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 count -= cp_count;
1190 }
1191 return 0;
1192}
1193
1194/* fill a group_info from a user-space array - it must be allocated already */
1195static int groups_from_user(struct group_info *group_info,
1196 gid_t __user *grouplist)
Cal Peake756184b2006-09-30 23:27:24 -07001197{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 int i;
Eric Dumazet1bf47342008-02-06 01:37:56 -08001199 unsigned int count = group_info->ngroups;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
1201 for (i = 0; i < group_info->nblocks; i++) {
Eric Dumazet1bf47342008-02-06 01:37:56 -08001202 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1203 unsigned int len = cp_count * sizeof(*grouplist);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204
Eric Dumazet1bf47342008-02-06 01:37:56 -08001205 if (copy_from_user(group_info->blocks[i], grouplist, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 return -EFAULT;
1207
Eric Dumazet1bf47342008-02-06 01:37:56 -08001208 grouplist += NGROUPS_PER_BLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 count -= cp_count;
1210 }
1211 return 0;
1212}
1213
Domen Puncerebe8b542005-05-05 16:16:19 -07001214/* a simple Shell sort */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215static void groups_sort(struct group_info *group_info)
1216{
1217 int base, max, stride;
1218 int gidsetsize = group_info->ngroups;
1219
1220 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1221 ; /* nothing */
1222 stride /= 3;
1223
1224 while (stride) {
1225 max = gidsetsize - stride;
1226 for (base = 0; base < max; base++) {
1227 int left = base;
1228 int right = left + stride;
1229 gid_t tmp = GROUP_AT(group_info, right);
1230
1231 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1232 GROUP_AT(group_info, right) =
1233 GROUP_AT(group_info, left);
1234 right = left;
1235 left -= stride;
1236 }
1237 GROUP_AT(group_info, right) = tmp;
1238 }
1239 stride /= 3;
1240 }
1241}
1242
1243/* a simple bsearch */
David Howells86a264a2008-11-14 10:39:18 +11001244int groups_search(const struct group_info *group_info, gid_t grp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245{
Eric Dumazetd74beb92006-03-25 03:08:19 -08001246 unsigned int left, right;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
1248 if (!group_info)
1249 return 0;
1250
1251 left = 0;
1252 right = group_info->ngroups;
1253 while (left < right) {
Eric Dumazetd74beb92006-03-25 03:08:19 -08001254 unsigned int mid = (left+right)/2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 int cmp = grp - GROUP_AT(group_info, mid);
1256 if (cmp > 0)
1257 left = mid + 1;
1258 else if (cmp < 0)
1259 right = mid;
1260 else
1261 return 1;
1262 }
1263 return 0;
1264}
1265
David Howellsb6dff3e2008-11-14 10:39:16 +11001266/**
David Howellsd84f4f92008-11-14 10:39:23 +11001267 * set_groups - Change a group subscription in a set of credentials
1268 * @new: The newly prepared set of credentials to alter
1269 * @group_info: The group list to install
David Howellsb6dff3e2008-11-14 10:39:16 +11001270 *
David Howellsd84f4f92008-11-14 10:39:23 +11001271 * Validate a group subscription and, if valid, insert it into a set
1272 * of credentials.
David Howellsb6dff3e2008-11-14 10:39:16 +11001273 */
David Howellsd84f4f92008-11-14 10:39:23 +11001274int set_groups(struct cred *new, struct group_info *group_info)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275{
1276 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
1278 retval = security_task_setgroups(group_info);
1279 if (retval)
1280 return retval;
1281
David Howellsd84f4f92008-11-14 10:39:23 +11001282 put_group_info(new->group_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 groups_sort(group_info);
1284 get_group_info(group_info);
David Howellsd84f4f92008-11-14 10:39:23 +11001285 new->group_info = group_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 return 0;
1287}
1288
David Howellsb6dff3e2008-11-14 10:39:16 +11001289EXPORT_SYMBOL(set_groups);
1290
1291/**
1292 * set_current_groups - Change current's group subscription
1293 * @group_info: The group list to impose
1294 *
1295 * Validate a group subscription and, if valid, impose it upon current's task
1296 * security record.
1297 */
1298int set_current_groups(struct group_info *group_info)
1299{
David Howellsd84f4f92008-11-14 10:39:23 +11001300 struct cred *new;
1301 int ret;
1302
1303 new = prepare_creds();
1304 if (!new)
1305 return -ENOMEM;
1306
1307 ret = set_groups(new, group_info);
1308 if (ret < 0) {
1309 abort_creds(new);
1310 return ret;
1311 }
1312
1313 return commit_creds(new);
David Howellsb6dff3e2008-11-14 10:39:16 +11001314}
1315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316EXPORT_SYMBOL(set_current_groups);
1317
Heiko Carstensae1251a2009-01-14 14:14:05 +01001318SYSCALL_DEFINE2(getgroups, int, gidsetsize, gid_t __user *, grouplist)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319{
David Howells86a264a2008-11-14 10:39:18 +11001320 const struct cred *cred = current_cred();
1321 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322
1323 if (gidsetsize < 0)
1324 return -EINVAL;
1325
1326 /* no need to grab task_lock here; it cannot change */
David Howellsb6dff3e2008-11-14 10:39:16 +11001327 i = cred->group_info->ngroups;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 if (gidsetsize) {
1329 if (i > gidsetsize) {
1330 i = -EINVAL;
1331 goto out;
1332 }
David Howellsb6dff3e2008-11-14 10:39:16 +11001333 if (groups_to_user(grouplist, cred->group_info)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 i = -EFAULT;
1335 goto out;
1336 }
1337 }
1338out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 return i;
1340}
1341
1342/*
1343 * SMP: Our groups are copy-on-write. We can set them safely
1344 * without another task interfering.
1345 */
1346
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001347SYSCALL_DEFINE2(setgroups, int, gidsetsize, gid_t __user *, grouplist)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348{
1349 struct group_info *group_info;
1350 int retval;
1351
1352 if (!capable(CAP_SETGID))
1353 return -EPERM;
1354 if ((unsigned)gidsetsize > NGROUPS_MAX)
1355 return -EINVAL;
1356
1357 group_info = groups_alloc(gidsetsize);
1358 if (!group_info)
1359 return -ENOMEM;
1360 retval = groups_from_user(group_info, grouplist);
1361 if (retval) {
1362 put_group_info(group_info);
1363 return retval;
1364 }
1365
1366 retval = set_current_groups(group_info);
1367 put_group_info(group_info);
1368
1369 return retval;
1370}
1371
1372/*
1373 * Check whether we're fsgid/egid or in the supplemental group..
1374 */
1375int in_group_p(gid_t grp)
1376{
David Howells86a264a2008-11-14 10:39:18 +11001377 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 int retval = 1;
David Howells86a264a2008-11-14 10:39:18 +11001379
David Howellsb6dff3e2008-11-14 10:39:16 +11001380 if (grp != cred->fsgid)
1381 retval = groups_search(cred->group_info, grp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 return retval;
1383}
1384
1385EXPORT_SYMBOL(in_group_p);
1386
1387int in_egroup_p(gid_t grp)
1388{
David Howells86a264a2008-11-14 10:39:18 +11001389 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 int retval = 1;
David Howells86a264a2008-11-14 10:39:18 +11001391
David Howellsb6dff3e2008-11-14 10:39:16 +11001392 if (grp != cred->egid)
1393 retval = groups_search(cred->group_info, grp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 return retval;
1395}
1396
1397EXPORT_SYMBOL(in_egroup_p);
1398
1399DECLARE_RWSEM(uts_sem);
1400
Heiko Carstense48fbb62009-01-14 14:14:26 +01001401SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402{
1403 int errno = 0;
1404
1405 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001406 if (copy_to_user(name, utsname(), sizeof *name))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 errno = -EFAULT;
1408 up_read(&uts_sem);
1409 return errno;
1410}
1411
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001412SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413{
1414 int errno;
1415 char tmp[__NEW_UTS_LEN];
1416
1417 if (!capable(CAP_SYS_ADMIN))
1418 return -EPERM;
1419 if (len < 0 || len > __NEW_UTS_LEN)
1420 return -EINVAL;
1421 down_write(&uts_sem);
1422 errno = -EFAULT;
1423 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001424 struct new_utsname *u = utsname();
1425
1426 memcpy(u->nodename, tmp, len);
1427 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 errno = 0;
1429 }
1430 up_write(&uts_sem);
1431 return errno;
1432}
1433
1434#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1435
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001436SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437{
1438 int i, errno;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001439 struct new_utsname *u;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440
1441 if (len < 0)
1442 return -EINVAL;
1443 down_read(&uts_sem);
Andrew Morton9679e4d2008-10-15 22:01:51 -07001444 u = utsname();
1445 i = 1 + strlen(u->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 if (i > len)
1447 i = len;
1448 errno = 0;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001449 if (copy_to_user(name, u->nodename, i))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 errno = -EFAULT;
1451 up_read(&uts_sem);
1452 return errno;
1453}
1454
1455#endif
1456
1457/*
1458 * Only setdomainname; getdomainname can be implemented by calling
1459 * uname()
1460 */
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001461SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462{
1463 int errno;
1464 char tmp[__NEW_UTS_LEN];
1465
1466 if (!capable(CAP_SYS_ADMIN))
1467 return -EPERM;
1468 if (len < 0 || len > __NEW_UTS_LEN)
1469 return -EINVAL;
1470
1471 down_write(&uts_sem);
1472 errno = -EFAULT;
1473 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001474 struct new_utsname *u = utsname();
1475
1476 memcpy(u->domainname, tmp, len);
1477 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 errno = 0;
1479 }
1480 up_write(&uts_sem);
1481 return errno;
1482}
1483
Heiko Carstense48fbb62009-01-14 14:14:26 +01001484SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485{
1486 if (resource >= RLIM_NLIMITS)
1487 return -EINVAL;
1488 else {
1489 struct rlimit value;
1490 task_lock(current->group_leader);
1491 value = current->signal->rlim[resource];
1492 task_unlock(current->group_leader);
1493 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1494 }
1495}
1496
1497#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1498
1499/*
1500 * Back compatibility for getrlimit. Needed for some apps.
1501 */
1502
Heiko Carstense48fbb62009-01-14 14:14:26 +01001503SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1504 struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505{
1506 struct rlimit x;
1507 if (resource >= RLIM_NLIMITS)
1508 return -EINVAL;
1509
1510 task_lock(current->group_leader);
1511 x = current->signal->rlim[resource];
1512 task_unlock(current->group_leader);
Cal Peake756184b2006-09-30 23:27:24 -07001513 if (x.rlim_cur > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 x.rlim_cur = 0x7FFFFFFF;
Cal Peake756184b2006-09-30 23:27:24 -07001515 if (x.rlim_max > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 x.rlim_max = 0x7FFFFFFF;
1517 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1518}
1519
1520#endif
1521
Heiko Carstense48fbb62009-01-14 14:14:26 +01001522SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523{
1524 struct rlimit new_rlim, *old_rlim;
1525 int retval;
1526
1527 if (resource >= RLIM_NLIMITS)
1528 return -EINVAL;
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001529 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 return -EFAULT;
Andrew Morton60fd7602009-02-04 15:12:06 -08001531 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1532 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 old_rlim = current->signal->rlim + resource;
1534 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1535 !capable(CAP_SYS_RESOURCE))
1536 return -EPERM;
Andrew Morton60fd7602009-02-04 15:12:06 -08001537 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1538 return -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539
1540 retval = security_task_setrlimit(resource, &new_rlim);
1541 if (retval)
1542 return retval;
1543
Tom Alsberg9926e4c2007-05-08 00:30:31 -07001544 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1545 /*
1546 * The caller is asking for an immediate RLIMIT_CPU
1547 * expiry. But we use the zero value to mean "it was
1548 * never set". So let's cheat and make it one second
1549 * instead
1550 */
1551 new_rlim.rlim_cur = 1;
1552 }
1553
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 task_lock(current->group_leader);
1555 *old_rlim = new_rlim;
1556 task_unlock(current->group_leader);
1557
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001558 if (resource != RLIMIT_CPU)
1559 goto out;
Andrew Mortond3561f72006-03-24 03:18:36 -08001560
1561 /*
1562 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1563 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1564 * very long-standing error, and fixing it now risks breakage of
1565 * applications, so we live with it
1566 */
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001567 if (new_rlim.rlim_cur == RLIM_INFINITY)
1568 goto out;
1569
Frank Mayharf06febc2008-09-12 09:54:39 -07001570 update_rlimit_cpu(new_rlim.rlim_cur);
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001571out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 return 0;
1573}
1574
1575/*
1576 * It would make sense to put struct rusage in the task_struct,
1577 * except that would make the task_struct be *really big*. After
1578 * task_struct gets moved into malloc'ed memory, it would
1579 * make sense to do this. It will make moving the rest of the information
1580 * a lot simpler! (Which we're not doing right now because we're not
1581 * measuring them yet).
1582 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1584 * races with threads incrementing their own counters. But since word
1585 * reads are atomic, we either get new values or old values and we don't
1586 * care which for the sums. We always take the siglock to protect reading
1587 * the c* fields from p->signal from races with exit.c updating those
1588 * fields when reaping, so a sample either gets all the additions of a
1589 * given child after it's reaped, or none so this sample is before reaping.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001590 *
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001591 * Locking:
1592 * We need to take the siglock for CHILDEREN, SELF and BOTH
1593 * for the cases current multithreaded, non-current single threaded
1594 * non-current multithreaded. Thread traversal is now safe with
1595 * the siglock held.
1596 * Strictly speaking, we donot need to take the siglock if we are current and
1597 * single threaded, as no one else can take our signal_struct away, no one
1598 * else can reap the children to update signal->c* counters, and no one else
1599 * can race with the signal-> fields. If we do not take any lock, the
1600 * signal-> fields could be read out of order while another thread was just
1601 * exiting. So we should place a read memory barrier when we avoid the lock.
1602 * On the writer side, write memory barrier is implied in __exit_signal
1603 * as __exit_signal releases the siglock spinlock after updating the signal->
1604 * fields. But we don't do this yet to keep things simple.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001605 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 */
1607
Frank Mayharf06febc2008-09-12 09:54:39 -07001608static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001609{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001610 r->ru_nvcsw += t->nvcsw;
1611 r->ru_nivcsw += t->nivcsw;
1612 r->ru_minflt += t->min_flt;
1613 r->ru_majflt += t->maj_flt;
1614 r->ru_inblock += task_io_get_inblock(t);
1615 r->ru_oublock += task_io_get_oublock(t);
1616}
1617
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1619{
1620 struct task_struct *t;
1621 unsigned long flags;
1622 cputime_t utime, stime;
Frank Mayharf06febc2008-09-12 09:54:39 -07001623 struct task_cputime cputime;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624
1625 memset((char *) r, 0, sizeof *r);
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001626 utime = stime = cputime_zero;
1627
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001628 if (who == RUSAGE_THREAD) {
KOSAKI Motohiro8916ede2009-01-04 05:40:37 +09001629 utime = task_utime(current);
1630 stime = task_stime(current);
Frank Mayharf06febc2008-09-12 09:54:39 -07001631 accumulate_thread_rusage(p, r);
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001632 goto out;
1633 }
1634
Oleg Nesterovd6cf7232008-04-30 00:52:38 -07001635 if (!lock_task_sighand(p, &flags))
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001636 return;
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001637
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 switch (who) {
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001639 case RUSAGE_BOTH:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 case RUSAGE_CHILDREN:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 utime = p->signal->cutime;
1642 stime = p->signal->cstime;
1643 r->ru_nvcsw = p->signal->cnvcsw;
1644 r->ru_nivcsw = p->signal->cnivcsw;
1645 r->ru_minflt = p->signal->cmin_flt;
1646 r->ru_majflt = p->signal->cmaj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001647 r->ru_inblock = p->signal->cinblock;
1648 r->ru_oublock = p->signal->coublock;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001649
1650 if (who == RUSAGE_CHILDREN)
1651 break;
1652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 case RUSAGE_SELF:
Frank Mayharf06febc2008-09-12 09:54:39 -07001654 thread_group_cputime(p, &cputime);
1655 utime = cputime_add(utime, cputime.utime);
1656 stime = cputime_add(stime, cputime.stime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657 r->ru_nvcsw += p->signal->nvcsw;
1658 r->ru_nivcsw += p->signal->nivcsw;
1659 r->ru_minflt += p->signal->min_flt;
1660 r->ru_majflt += p->signal->maj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001661 r->ru_inblock += p->signal->inblock;
1662 r->ru_oublock += p->signal->oublock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 t = p;
1664 do {
Frank Mayharf06febc2008-09-12 09:54:39 -07001665 accumulate_thread_rusage(t, r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 t = next_thread(t);
1667 } while (t != p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 break;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001669
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 default:
1671 BUG();
1672 }
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001673 unlock_task_sighand(p, &flags);
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001674
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001675out:
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001676 cputime_to_timeval(utime, &r->ru_utime);
1677 cputime_to_timeval(stime, &r->ru_stime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678}
1679
1680int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1681{
1682 struct rusage r;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 k_getrusage(p, who, &r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1685}
1686
Heiko Carstense48fbb62009-01-14 14:14:26 +01001687SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001689 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1690 who != RUSAGE_THREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 return -EINVAL;
1692 return getrusage(current, who, ru);
1693}
1694
Heiko Carstense48fbb62009-01-14 14:14:26 +01001695SYSCALL_DEFINE1(umask, int, mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696{
1697 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1698 return mask;
1699}
Serge E. Hallyn3b7391d2008-02-04 22:29:45 -08001700
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001701SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1702 unsigned long, arg4, unsigned long, arg5)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703{
David Howellsb6dff3e2008-11-14 10:39:16 +11001704 struct task_struct *me = current;
1705 unsigned char comm[sizeof(me->comm)];
1706 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707
David Howellsd84f4f92008-11-14 10:39:23 +11001708 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1709 if (error != -ENOSYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 return error;
1711
David Howellsd84f4f92008-11-14 10:39:23 +11001712 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 switch (option) {
1714 case PR_SET_PDEATHSIG:
Jesper Juhl0730ded2005-09-06 15:17:37 -07001715 if (!valid_signal(arg2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 error = -EINVAL;
1717 break;
1718 }
David Howellsb6dff3e2008-11-14 10:39:16 +11001719 me->pdeath_signal = arg2;
1720 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 break;
1722 case PR_GET_PDEATHSIG:
David Howellsb6dff3e2008-11-14 10:39:16 +11001723 error = put_user(me->pdeath_signal, (int __user *)arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 break;
1725 case PR_GET_DUMPABLE:
David Howellsb6dff3e2008-11-14 10:39:16 +11001726 error = get_dumpable(me->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 break;
1728 case PR_SET_DUMPABLE:
Marcel Holtmannabf75a52006-07-12 13:12:00 +02001729 if (arg2 < 0 || arg2 > 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 error = -EINVAL;
1731 break;
1732 }
David Howellsb6dff3e2008-11-14 10:39:16 +11001733 set_dumpable(me->mm, arg2);
1734 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 break;
1736
1737 case PR_SET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001738 error = SET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 break;
1740 case PR_GET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001741 error = GET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 break;
1743 case PR_SET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11001744 error = SET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 break;
1746 case PR_GET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11001747 error = GET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 break;
1749 case PR_SET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11001750 error = SET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 break;
1752 case PR_GET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11001753 error = GET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 break;
1755 case PR_GET_TIMING:
1756 error = PR_TIMING_STATISTICAL;
1757 break;
1758 case PR_SET_TIMING:
Shi Weihua7b266552008-05-23 13:04:59 -07001759 if (arg2 != PR_TIMING_STATISTICAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 error = -EINVAL;
David Howellsb6dff3e2008-11-14 10:39:16 +11001761 else
1762 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 break;
1764
David Howellsb6dff3e2008-11-14 10:39:16 +11001765 case PR_SET_NAME:
1766 comm[sizeof(me->comm)-1] = 0;
1767 if (strncpy_from_user(comm, (char __user *)arg2,
1768 sizeof(me->comm) - 1) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 return -EFAULT;
David Howellsb6dff3e2008-11-14 10:39:16 +11001770 set_task_comm(me, comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 return 0;
David Howellsb6dff3e2008-11-14 10:39:16 +11001772 case PR_GET_NAME:
1773 get_task_comm(comm, me);
1774 if (copy_to_user((char __user *)arg2, comm,
1775 sizeof(comm)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 return -EFAULT;
1777 return 0;
Anton Blanchard651d7652006-06-07 16:10:19 +10001778 case PR_GET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001779 error = GET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10001780 break;
1781 case PR_SET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001782 error = SET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10001783 break;
1784
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -07001785 case PR_GET_SECCOMP:
1786 error = prctl_get_seccomp();
1787 break;
1788 case PR_SET_SECCOMP:
1789 error = prctl_set_seccomp(arg2);
1790 break;
Erik Bosman8fb402b2008-04-11 18:54:17 +02001791 case PR_GET_TSC:
1792 error = GET_TSC_CTL(arg2);
1793 break;
1794 case PR_SET_TSC:
1795 error = SET_TSC_CTL(arg2);
1796 break;
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01001797 case PR_TASK_PERF_COUNTERS_DISABLE:
1798 error = perf_counter_task_disable();
1799 break;
1800 case PR_TASK_PERF_COUNTERS_ENABLE:
1801 error = perf_counter_task_enable();
1802 break;
Arjan van de Ven69766752008-09-01 15:52:40 -07001803 case PR_GET_TIMERSLACK:
1804 error = current->timer_slack_ns;
1805 break;
1806 case PR_SET_TIMERSLACK:
1807 if (arg2 <= 0)
1808 current->timer_slack_ns =
1809 current->default_timer_slack_ns;
1810 else
1811 current->timer_slack_ns = arg2;
David Howellsb6dff3e2008-11-14 10:39:16 +11001812 error = 0;
Arjan van de Ven69766752008-09-01 15:52:40 -07001813 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 default:
1815 error = -EINVAL;
1816 break;
1817 }
1818 return error;
1819}
Andi Kleen3cfc3482006-09-26 10:52:28 +02001820
Heiko Carstens836f92a2009-01-14 14:14:33 +01001821SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1822 struct getcpu_cache __user *, unused)
Andi Kleen3cfc3482006-09-26 10:52:28 +02001823{
1824 int err = 0;
1825 int cpu = raw_smp_processor_id();
1826 if (cpup)
1827 err |= put_user(cpu, cpup);
1828 if (nodep)
1829 err |= put_user(cpu_to_node(cpu), nodep);
Andi Kleen3cfc3482006-09-26 10:52:28 +02001830 return err ? -EFAULT : 0;
1831}
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001832
1833char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1834
1835static void argv_cleanup(char **argv, char **envp)
1836{
1837 argv_free(argv);
1838}
1839
1840/**
1841 * orderly_poweroff - Trigger an orderly system poweroff
1842 * @force: force poweroff if command execution fails
1843 *
1844 * This may be called from any context to trigger a system shutdown.
1845 * If the orderly shutdown fails, it will force an immediate shutdown.
1846 */
1847int orderly_poweroff(bool force)
1848{
1849 int argc;
1850 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1851 static char *envp[] = {
1852 "HOME=/",
1853 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1854 NULL
1855 };
1856 int ret = -ENOMEM;
1857 struct subprocess_info *info;
1858
1859 if (argv == NULL) {
1860 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1861 __func__, poweroff_cmd);
1862 goto out;
1863 }
1864
KOSAKI Motohiroac331d12008-07-25 01:45:38 -07001865 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001866 if (info == NULL) {
1867 argv_free(argv);
1868 goto out;
1869 }
1870
1871 call_usermodehelper_setcleanup(info, argv_cleanup);
1872
Jeremy Fitzhardinge86313c42007-07-17 18:37:03 -07001873 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001874
1875 out:
1876 if (ret && force) {
1877 printk(KERN_WARNING "Failed to start orderly shutdown: "
1878 "forcing the issue\n");
1879
1880 /* I guess this should try to kick off some daemon to
1881 sync and poweroff asap. Or not even bother syncing
1882 if we're doing an emergency shutdown? */
1883 emergency_sync();
1884 kernel_power_off();
1885 }
1886
1887 return ret;
1888}
1889EXPORT_SYMBOL_GPL(orderly_poweroff);