blob: 481611fbd079a90094876bd2ef9743803fa45576 [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
Paul Gortmaker9984de12011-05-23 14:51:41 -04007#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/mm.h>
9#include <linux/utsname.h>
10#include <linux/mman.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/reboot.h>
12#include <linux/prctl.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/highuid.h>
14#include <linux/fs.h>
Paul Gortmaker74da1ff2011-05-26 12:48:41 -040015#include <linux/kmod.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020016#include <linux/perf_event.h>
Daniel Walker3e88c552007-05-10 22:22:53 -070017#include <linux/resource.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070018#include <linux/kernel.h>
19#include <linux/kexec.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/workqueue.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080021#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/device.h>
23#include <linux/key.h>
24#include <linux/times.h>
25#include <linux/posix-timers.h>
26#include <linux/security.h>
27#include <linux/dcookies.h>
28#include <linux/suspend.h>
29#include <linux/tty.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070030#include <linux/signal.h>
Matt Helsley9f460802005-11-07 00:59:16 -080031#include <linux/cn_proc.h>
Andi Kleen3cfc3482006-09-26 10:52:28 +020032#include <linux/getcpu.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070033#include <linux/task_io_accounting_ops.h>
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -070034#include <linux/seccomp.h>
Mark Lord40477272007-10-01 01:20:10 -070035#include <linux/cpu.h>
Christoph Hellwige28cbf22010-03-10 15:21:19 -080036#include <linux/personality.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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090039#include <linux/gfp.h>
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +010040#include <linux/syscore_ops.h>
Andi Kleenbe274252011-08-19 16:15:10 -070041#include <linux/version.h>
42#include <linux/ctype.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44#include <linux/compat.h>
45#include <linux/syscalls.h>
Keshavamurthy Anil S00d7c052005-12-12 00:37:33 -080046#include <linux/kprobes.h>
Cedric Le Goateracce2922007-07-15 23:40:59 -070047#include <linux/user_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
Seiji Aguchi04c68622011-01-12 16:59:30 -080049#include <linux/kmsg_dump.h>
Andi Kleenbe274252011-08-19 16:15:10 -070050/* Move somewhere else to avoid recompiling? */
51#include <generated/utsrelease.h>
Seiji Aguchi04c68622011-01-12 16:59:30 -080052
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <asm/uaccess.h>
54#include <asm/io.h>
55#include <asm/unistd.h>
56
57#ifndef SET_UNALIGN_CTL
58# define SET_UNALIGN_CTL(a,b) (-EINVAL)
59#endif
60#ifndef GET_UNALIGN_CTL
61# define GET_UNALIGN_CTL(a,b) (-EINVAL)
62#endif
63#ifndef SET_FPEMU_CTL
64# define SET_FPEMU_CTL(a,b) (-EINVAL)
65#endif
66#ifndef GET_FPEMU_CTL
67# define GET_FPEMU_CTL(a,b) (-EINVAL)
68#endif
69#ifndef SET_FPEXC_CTL
70# define SET_FPEXC_CTL(a,b) (-EINVAL)
71#endif
72#ifndef GET_FPEXC_CTL
73# define GET_FPEXC_CTL(a,b) (-EINVAL)
74#endif
Anton Blanchard651d7652006-06-07 16:10:19 +100075#ifndef GET_ENDIAN
76# define GET_ENDIAN(a,b) (-EINVAL)
77#endif
78#ifndef SET_ENDIAN
79# define SET_ENDIAN(a,b) (-EINVAL)
80#endif
Erik Bosman8fb402b2008-04-11 18:54:17 +020081#ifndef GET_TSC_CTL
82# define GET_TSC_CTL(a) (-EINVAL)
83#endif
84#ifndef SET_TSC_CTL
85# define SET_TSC_CTL(a) (-EINVAL)
86#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
88/*
89 * this is where the system-wide overflow UID and GID are defined, for
90 * architectures that now have 32-bit UID/GID but didn't in the past
91 */
92
93int overflowuid = DEFAULT_OVERFLOWUID;
94int overflowgid = DEFAULT_OVERFLOWGID;
95
96#ifdef CONFIG_UID16
97EXPORT_SYMBOL(overflowuid);
98EXPORT_SYMBOL(overflowgid);
99#endif
100
101/*
102 * the same as above, but for filesystems which can only store a 16-bit
103 * UID and GID. as such, this is needed on all architectures
104 */
105
106int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
107int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
108
109EXPORT_SYMBOL(fs_overflowuid);
110EXPORT_SYMBOL(fs_overflowgid);
111
112/*
113 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
114 */
115
116int C_A_D = 1;
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700117struct pid *cad_pid;
118EXPORT_SYMBOL(cad_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
120/*
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700121 * If set, this is used for preparing the system to power off.
122 */
123
124void (*pm_power_off_prepare)(void);
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700125
David Howellsc69e8d92008-11-14 10:39:19 +1100126/*
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700127 * Returns true if current's euid is same as p's uid or euid,
128 * or has CAP_SYS_NICE to p's user_ns.
129 *
130 * Called with rcu_read_lock, creds are safe
131 */
132static bool set_one_prio_perm(struct task_struct *p)
133{
134 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
135
136 if (pcred->user->user_ns == cred->user->user_ns &&
137 (pcred->uid == cred->euid ||
138 pcred->euid == cred->euid))
139 return true;
140 if (ns_capable(pcred->user->user_ns, CAP_SYS_NICE))
141 return true;
142 return false;
143}
144
145/*
David Howellsc69e8d92008-11-14 10:39:19 +1100146 * set the priority of a task
147 * - the caller must hold the RCU read lock
148 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149static int set_one_prio(struct task_struct *p, int niceval, int error)
150{
151 int no_nice;
152
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700153 if (!set_one_prio_perm(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 error = -EPERM;
155 goto out;
156 }
Matt Mackalle43379f2005-05-01 08:59:00 -0700157 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 error = -EACCES;
159 goto out;
160 }
161 no_nice = security_task_setnice(p, niceval);
162 if (no_nice) {
163 error = no_nice;
164 goto out;
165 }
166 if (error == -ESRCH)
167 error = 0;
168 set_user_nice(p, niceval);
169out:
170 return error;
171}
172
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100173SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174{
175 struct task_struct *g, *p;
176 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100177 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 int error = -EINVAL;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800179 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
Daniel Walker3e88c552007-05-10 22:22:53 -0700181 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 goto out;
183
184 /* normalize: avoid signed division (rounding problems) */
185 error = -ESRCH;
186 if (niceval < -20)
187 niceval = -20;
188 if (niceval > 19)
189 niceval = 19;
190
Thomas Gleixnerd4581a22009-12-10 00:52:51 +0000191 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 read_lock(&tasklist_lock);
193 switch (which) {
194 case PRIO_PROCESS:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800195 if (who)
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700196 p = find_task_by_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800197 else
198 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 if (p)
200 error = set_one_prio(p, niceval, error);
201 break;
202 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800203 if (who)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700204 pgrp = find_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800205 else
206 pgrp = task_pgrp(current);
Ken Chen2d70b682008-08-20 14:09:17 -0700207 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 error = set_one_prio(p, niceval, error);
Ken Chen2d70b682008-08-20 14:09:17 -0700209 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 break;
211 case PRIO_USER:
David Howellsd84f4f92008-11-14 10:39:23 +1100212 user = (struct user_struct *) cred->user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 if (!who)
David Howells86a264a2008-11-14 10:39:18 +1100214 who = cred->uid;
215 else if ((who != cred->uid) &&
216 !(user = find_user(who)))
217 goto out_unlock; /* No processes for this user */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800219 do_each_thread(g, p) {
David Howells86a264a2008-11-14 10:39:18 +1100220 if (__task_cred(p)->uid == who)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 error = set_one_prio(p, niceval, error);
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800222 } while_each_thread(g, p);
David Howells86a264a2008-11-14 10:39:18 +1100223 if (who != cred->uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 free_uid(user); /* For find_user() */
225 break;
226 }
227out_unlock:
228 read_unlock(&tasklist_lock);
Thomas Gleixnerd4581a22009-12-10 00:52:51 +0000229 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230out:
231 return error;
232}
233
234/*
235 * Ugh. To avoid negative return values, "getpriority()" will
236 * not return the normal nice-value, but a negated value that
237 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
238 * to stay compatible.
239 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100240SYSCALL_DEFINE2(getpriority, int, which, int, who)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241{
242 struct task_struct *g, *p;
243 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100244 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 long niceval, retval = -ESRCH;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800246 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
Daniel Walker3e88c552007-05-10 22:22:53 -0700248 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 return -EINVAL;
250
Tetsuo Handa70118832010-02-22 12:44:16 -0800251 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 read_lock(&tasklist_lock);
253 switch (which) {
254 case PRIO_PROCESS:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800255 if (who)
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700256 p = find_task_by_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800257 else
258 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 if (p) {
260 niceval = 20 - task_nice(p);
261 if (niceval > retval)
262 retval = niceval;
263 }
264 break;
265 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800266 if (who)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700267 pgrp = find_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800268 else
269 pgrp = task_pgrp(current);
Ken Chen2d70b682008-08-20 14:09:17 -0700270 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 niceval = 20 - task_nice(p);
272 if (niceval > retval)
273 retval = niceval;
Ken Chen2d70b682008-08-20 14:09:17 -0700274 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 break;
276 case PRIO_USER:
David Howells86a264a2008-11-14 10:39:18 +1100277 user = (struct user_struct *) cred->user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 if (!who)
David Howells86a264a2008-11-14 10:39:18 +1100279 who = cred->uid;
280 else if ((who != cred->uid) &&
281 !(user = find_user(who)))
282 goto out_unlock; /* No processes for this user */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800284 do_each_thread(g, p) {
David Howells86a264a2008-11-14 10:39:18 +1100285 if (__task_cred(p)->uid == who) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 niceval = 20 - task_nice(p);
287 if (niceval > retval)
288 retval = niceval;
289 }
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800290 } while_each_thread(g, p);
David Howells86a264a2008-11-14 10:39:18 +1100291 if (who != cred->uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 free_uid(user); /* for find_user() */
293 break;
294 }
295out_unlock:
296 read_unlock(&tasklist_lock);
Tetsuo Handa70118832010-02-22 12:44:16 -0800297 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
299 return retval;
300}
301
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700302/**
303 * emergency_restart - reboot the system
304 *
305 * Without shutting down any hardware or taking any locks
306 * reboot the system. This is called when we know we are in
307 * trouble so this is our best effort to reboot. This is
308 * safe to call in interrupt context.
309 */
Eric W. Biederman7c903472005-07-26 11:29:55 -0600310void emergency_restart(void)
311{
Seiji Aguchi04c68622011-01-12 16:59:30 -0800312 kmsg_dump(KMSG_DUMP_EMERG);
Eric W. Biederman7c903472005-07-26 11:29:55 -0600313 machine_emergency_restart();
314}
315EXPORT_SYMBOL_GPL(emergency_restart);
316
Huang Yingca195b72008-08-15 00:40:24 -0700317void kernel_restart_prepare(char *cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600318{
Alan Sterne041c682006-03-27 01:16:30 -0800319 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600320 system_state = SYSTEM_RESTART;
Kay Sieversb50fa7c2011-05-05 13:32:05 +0200321 usermodehelper_disable();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600322 device_shutdown();
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100323 syscore_shutdown();
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700324}
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800325
326/**
Amerigo Wangc5f41752011-07-25 17:13:10 -0700327 * register_reboot_notifier - Register function to be called at reboot time
328 * @nb: Info about notifier function to be called
329 *
330 * Registers a function with the list of functions
331 * to be called at reboot time.
332 *
333 * Currently always returns zero, as blocking_notifier_chain_register()
334 * always returns zero.
335 */
336int register_reboot_notifier(struct notifier_block *nb)
337{
338 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
339}
340EXPORT_SYMBOL(register_reboot_notifier);
341
342/**
343 * unregister_reboot_notifier - Unregister previously registered reboot notifier
344 * @nb: Hook to be unregistered
345 *
346 * Unregisters a previously registered reboot
347 * notifier function.
348 *
349 * Returns zero on success, or %-ENOENT on failure.
350 */
351int unregister_reboot_notifier(struct notifier_block *nb)
352{
353 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
354}
355EXPORT_SYMBOL(unregister_reboot_notifier);
356
357/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800358 * kernel_restart - reboot the system
359 * @cmd: pointer to buffer containing command to execute for restart
Randy Dunlapb8887e62005-11-07 01:01:07 -0800360 * or %NULL
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800361 *
362 * Shutdown everything and perform a clean reboot.
363 * This is not safe to call in interrupt context.
364 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700365void kernel_restart(char *cmd)
366{
367 kernel_restart_prepare(cmd);
Cal Peake756184b2006-09-30 23:27:24 -0700368 if (!cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600369 printk(KERN_EMERG "Restarting system.\n");
Cal Peake756184b2006-09-30 23:27:24 -0700370 else
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600371 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
Seiji Aguchi04c68622011-01-12 16:59:30 -0800372 kmsg_dump(KMSG_DUMP_RESTART);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600373 machine_restart(cmd);
374}
375EXPORT_SYMBOL_GPL(kernel_restart);
376
Adrian Bunk4ef72292008-02-04 22:30:06 -0800377static void kernel_shutdown_prepare(enum system_states state)
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500378{
Alan Sterne041c682006-03-27 01:16:30 -0800379 blocking_notifier_call_chain(&reboot_notifier_list,
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500380 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
381 system_state = state;
Kay Sieversb50fa7c2011-05-05 13:32:05 +0200382 usermodehelper_disable();
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500383 device_shutdown();
384}
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700385/**
386 * kernel_halt - halt the system
387 *
388 * Shutdown everything and perform a clean system halt.
389 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700390void kernel_halt(void)
391{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500392 kernel_shutdown_prepare(SYSTEM_HALT);
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100393 syscore_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600394 printk(KERN_EMERG "System halted.\n");
Seiji Aguchi04c68622011-01-12 16:59:30 -0800395 kmsg_dump(KMSG_DUMP_HALT);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600396 machine_halt();
397}
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500398
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600399EXPORT_SYMBOL_GPL(kernel_halt);
400
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700401/**
402 * kernel_power_off - power_off the system
403 *
404 * Shutdown everything and perform a clean system power_off.
405 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700406void kernel_power_off(void)
407{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500408 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700409 if (pm_power_off_prepare)
410 pm_power_off_prepare();
Mark Lord40477272007-10-01 01:20:10 -0700411 disable_nonboot_cpus();
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100412 syscore_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600413 printk(KERN_EMERG "Power down.\n");
Seiji Aguchi04c68622011-01-12 16:59:30 -0800414 kmsg_dump(KMSG_DUMP_POWEROFF);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600415 machine_power_off();
416}
417EXPORT_SYMBOL_GPL(kernel_power_off);
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200418
419static DEFINE_MUTEX(reboot_mutex);
420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421/*
422 * Reboot system call: for obvious reasons only root may call it,
423 * and even root needs to set up some magic numbers in the registers
424 * so that some mistake won't make this reboot the whole machine.
425 * You can also set the meaning of the ctrl-alt-del-key here.
426 *
427 * reboot doesn't sync: do that yourself before calling this.
428 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100429SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
430 void __user *, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431{
432 char buffer[256];
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700433 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
435 /* We only trust the superuser with rebooting the system. */
436 if (!capable(CAP_SYS_BOOT))
437 return -EPERM;
438
439 /* For safety, we require "magic" arguments. */
440 if (magic1 != LINUX_REBOOT_MAGIC1 ||
441 (magic2 != LINUX_REBOOT_MAGIC2 &&
442 magic2 != LINUX_REBOOT_MAGIC2A &&
443 magic2 != LINUX_REBOOT_MAGIC2B &&
444 magic2 != LINUX_REBOOT_MAGIC2C))
445 return -EINVAL;
446
Eric W. Biederman5e382912006-01-08 01:03:46 -0800447 /* Instead of trying to make the power_off code look like
448 * halt when pm_power_off is not set do it the easy way.
449 */
450 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
451 cmd = LINUX_REBOOT_CMD_HALT;
452
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200453 mutex_lock(&reboot_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 switch (cmd) {
455 case LINUX_REBOOT_CMD_RESTART:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600456 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 break;
458
459 case LINUX_REBOOT_CMD_CAD_ON:
460 C_A_D = 1;
461 break;
462
463 case LINUX_REBOOT_CMD_CAD_OFF:
464 C_A_D = 0;
465 break;
466
467 case LINUX_REBOOT_CMD_HALT:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600468 kernel_halt();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 do_exit(0);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700470 panic("cannot halt");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472 case LINUX_REBOOT_CMD_POWER_OFF:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600473 kernel_power_off();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 do_exit(0);
475 break;
476
477 case LINUX_REBOOT_CMD_RESTART2:
478 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200479 ret = -EFAULT;
480 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 }
482 buffer[sizeof(buffer) - 1] = '\0';
483
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600484 kernel_restart(buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 break;
486
Huang Ying3ab83522008-07-25 19:45:07 -0700487#ifdef CONFIG_KEXEC
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700488 case LINUX_REBOOT_CMD_KEXEC:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700489 ret = kernel_kexec();
490 break;
Huang Ying3ab83522008-07-25 19:45:07 -0700491#endif
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600492
Rafael J. Wysockib0cb1a12007-07-29 23:24:36 +0200493#ifdef CONFIG_HIBERNATION
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 case LINUX_REBOOT_CMD_SW_SUSPEND:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700495 ret = hibernate();
496 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497#endif
498
499 default:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700500 ret = -EINVAL;
501 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 }
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200503 mutex_unlock(&reboot_mutex);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700504 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505}
506
David Howells65f27f32006-11-22 14:55:48 +0000507static void deferred_cad(struct work_struct *dummy)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508{
Eric W. Biedermanabcd9e52005-07-26 11:27:34 -0600509 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510}
511
512/*
513 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
514 * As it's called within an interrupt, it may NOT sync: the only choice
515 * is whether to reboot at once, or just ignore the ctrl-alt-del.
516 */
517void ctrl_alt_del(void)
518{
David Howells65f27f32006-11-22 14:55:48 +0000519 static DECLARE_WORK(cad_work, deferred_cad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
521 if (C_A_D)
522 schedule_work(&cad_work);
523 else
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700524 kill_cad_pid(SIGINT, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525}
526
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527/*
528 * Unprivileged users may change the real gid to the effective gid
529 * or vice versa. (BSD-style)
530 *
531 * If you set the real gid at all, or set the effective gid to a value not
532 * equal to the real gid, then the saved gid is set to the new effective gid.
533 *
534 * This makes it possible for a setgid program to completely drop its
535 * privileges, which is often a useful assertion to make when you are doing
536 * a security audit over a program.
537 *
538 * The general idea is that a program which uses just setregid() will be
539 * 100% compatible with BSD. A program which uses just setgid() will be
540 * 100% compatible with POSIX with saved IDs.
541 *
542 * SMP: There are not races, the GIDs are checked only by filesystem
543 * operations (as far as semantic preservation is concerned).
544 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100545SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546{
David Howellsd84f4f92008-11-14 10:39:23 +1100547 const struct cred *old;
548 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 int retval;
550
David Howellsd84f4f92008-11-14 10:39:23 +1100551 new = prepare_creds();
552 if (!new)
553 return -ENOMEM;
554 old = current_cred();
555
David Howellsd84f4f92008-11-14 10:39:23 +1100556 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 if (rgid != (gid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100558 if (old->gid == rgid ||
559 old->egid == rgid ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700560 nsown_capable(CAP_SETGID))
David Howellsd84f4f92008-11-14 10:39:23 +1100561 new->gid = rgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 else
David Howellsd84f4f92008-11-14 10:39:23 +1100563 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 }
565 if (egid != (gid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100566 if (old->gid == egid ||
567 old->egid == egid ||
568 old->sgid == egid ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700569 nsown_capable(CAP_SETGID))
David Howellsd84f4f92008-11-14 10:39:23 +1100570 new->egid = egid;
Cal Peake756184b2006-09-30 23:27:24 -0700571 else
David Howellsd84f4f92008-11-14 10:39:23 +1100572 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 }
David Howellsd84f4f92008-11-14 10:39:23 +1100574
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 if (rgid != (gid_t) -1 ||
David Howellsd84f4f92008-11-14 10:39:23 +1100576 (egid != (gid_t) -1 && egid != old->gid))
577 new->sgid = new->egid;
578 new->fsgid = new->egid;
579
580 return commit_creds(new);
581
582error:
583 abort_creds(new);
584 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585}
586
587/*
588 * setgid() is implemented like SysV w/ SAVED_IDS
589 *
590 * SMP: Same implicit races as above.
591 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100592SYSCALL_DEFINE1(setgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593{
David Howellsd84f4f92008-11-14 10:39:23 +1100594 const struct cred *old;
595 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 int retval;
597
David Howellsd84f4f92008-11-14 10:39:23 +1100598 new = prepare_creds();
599 if (!new)
600 return -ENOMEM;
601 old = current_cred();
602
David Howellsd84f4f92008-11-14 10:39:23 +1100603 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700604 if (nsown_capable(CAP_SETGID))
David Howellsd84f4f92008-11-14 10:39:23 +1100605 new->gid = new->egid = new->sgid = new->fsgid = gid;
606 else if (gid == old->gid || gid == old->sgid)
607 new->egid = new->fsgid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 else
David Howellsd84f4f92008-11-14 10:39:23 +1100609 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610
David Howellsd84f4f92008-11-14 10:39:23 +1100611 return commit_creds(new);
612
613error:
614 abort_creds(new);
615 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616}
Dhaval Giani54e99122009-02-27 15:13:54 +0530617
David Howellsd84f4f92008-11-14 10:39:23 +1100618/*
619 * change the user struct in a credentials set to match the new UID
620 */
621static int set_user(struct cred *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622{
623 struct user_struct *new_user;
624
Serge Hallyn18b6e042008-10-15 16:38:45 -0500625 new_user = alloc_uid(current_user_ns(), new->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 if (!new_user)
627 return -EAGAIN;
628
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400629 /*
630 * We don't fail in case of NPROC limit excess here because too many
631 * poorly written programs don't check set*uid() return code, assuming
632 * it never fails if called by root. We may still enforce NPROC limit
633 * for programs doing set*uid()+execve() by harmlessly deferring the
634 * failure to the execve() stage.
635 */
Jiri Slaby78d7d402010-03-05 13:42:54 -0800636 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400637 new_user != INIT_USER)
638 current->flags |= PF_NPROC_EXCEEDED;
639 else
640 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
David Howellsd84f4f92008-11-14 10:39:23 +1100642 free_uid(new->user);
643 new->user = new_user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 return 0;
645}
646
647/*
648 * Unprivileged users may change the real uid to the effective uid
649 * or vice versa. (BSD-style)
650 *
651 * If you set the real uid at all, or set the effective uid to a value not
652 * equal to the real uid, then the saved uid is set to the new effective uid.
653 *
654 * This makes it possible for a setuid program to completely drop its
655 * privileges, which is often a useful assertion to make when you are doing
656 * a security audit over a program.
657 *
658 * The general idea is that a program which uses just setreuid() will be
659 * 100% compatible with BSD. A program which uses just setuid() will be
660 * 100% compatible with POSIX with saved IDs.
661 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100662SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663{
David Howellsd84f4f92008-11-14 10:39:23 +1100664 const struct cred *old;
665 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 int retval;
667
David Howellsd84f4f92008-11-14 10:39:23 +1100668 new = prepare_creds();
669 if (!new)
670 return -ENOMEM;
671 old = current_cred();
672
David Howellsd84f4f92008-11-14 10:39:23 +1100673 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 if (ruid != (uid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100675 new->uid = ruid;
676 if (old->uid != ruid &&
677 old->euid != ruid &&
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700678 !nsown_capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100679 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 }
681
682 if (euid != (uid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100683 new->euid = euid;
684 if (old->uid != euid &&
685 old->euid != euid &&
686 old->suid != euid &&
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700687 !nsown_capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100688 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 }
690
Dhaval Giani54e99122009-02-27 15:13:54 +0530691 if (new->uid != old->uid) {
692 retval = set_user(new);
693 if (retval < 0)
694 goto error;
695 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 if (ruid != (uid_t) -1 ||
David Howellsd84f4f92008-11-14 10:39:23 +1100697 (euid != (uid_t) -1 && euid != old->uid))
698 new->suid = new->euid;
699 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
David Howellsd84f4f92008-11-14 10:39:23 +1100701 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
702 if (retval < 0)
703 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704
David Howellsd84f4f92008-11-14 10:39:23 +1100705 return commit_creds(new);
706
707error:
708 abort_creds(new);
709 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712/*
713 * setuid() is implemented like SysV with SAVED_IDS
714 *
715 * Note that SAVED_ID's is deficient in that a setuid root program
716 * like sendmail, for example, cannot set its uid to be a normal
717 * user and then switch back, because if you're root, setuid() sets
718 * the saved uid too. If you don't like this, blame the bright people
719 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
720 * will allow a root program to temporarily drop privileges and be able to
721 * regain them by swapping the real and effective uid.
722 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100723SYSCALL_DEFINE1(setuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724{
David Howellsd84f4f92008-11-14 10:39:23 +1100725 const struct cred *old;
726 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 int retval;
728
David Howellsd84f4f92008-11-14 10:39:23 +1100729 new = prepare_creds();
730 if (!new)
731 return -ENOMEM;
732 old = current_cred();
733
David Howellsd84f4f92008-11-14 10:39:23 +1100734 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700735 if (nsown_capable(CAP_SETUID)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100736 new->suid = new->uid = uid;
Dhaval Giani54e99122009-02-27 15:13:54 +0530737 if (uid != old->uid) {
738 retval = set_user(new);
739 if (retval < 0)
740 goto error;
David Howellsd84f4f92008-11-14 10:39:23 +1100741 }
742 } else if (uid != old->uid && uid != new->suid) {
743 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
David Howellsd84f4f92008-11-14 10:39:23 +1100746 new->fsuid = new->euid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
David Howellsd84f4f92008-11-14 10:39:23 +1100748 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
749 if (retval < 0)
750 goto error;
751
752 return commit_creds(new);
753
754error:
755 abort_creds(new);
756 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757}
758
759
760/*
761 * This function implements a generic ability to update ruid, euid,
762 * and suid. This allows you to implement the 4.4 compatible seteuid().
763 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100764SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765{
David Howellsd84f4f92008-11-14 10:39:23 +1100766 const struct cred *old;
767 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 int retval;
769
David Howellsd84f4f92008-11-14 10:39:23 +1100770 new = prepare_creds();
771 if (!new)
772 return -ENOMEM;
773
David Howellsd84f4f92008-11-14 10:39:23 +1100774 old = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
David Howellsd84f4f92008-11-14 10:39:23 +1100776 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700777 if (!nsown_capable(CAP_SETUID)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100778 if (ruid != (uid_t) -1 && ruid != old->uid &&
779 ruid != old->euid && ruid != old->suid)
780 goto error;
781 if (euid != (uid_t) -1 && euid != old->uid &&
782 euid != old->euid && euid != old->suid)
783 goto error;
784 if (suid != (uid_t) -1 && suid != old->uid &&
785 suid != old->euid && suid != old->suid)
786 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 }
David Howellsd84f4f92008-11-14 10:39:23 +1100788
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 if (ruid != (uid_t) -1) {
David Howellsd84f4f92008-11-14 10:39:23 +1100790 new->uid = ruid;
Dhaval Giani54e99122009-02-27 15:13:54 +0530791 if (ruid != old->uid) {
792 retval = set_user(new);
793 if (retval < 0)
794 goto error;
795 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 }
David Howellsd84f4f92008-11-14 10:39:23 +1100797 if (euid != (uid_t) -1)
798 new->euid = euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 if (suid != (uid_t) -1)
David Howellsd84f4f92008-11-14 10:39:23 +1100800 new->suid = suid;
801 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
David Howellsd84f4f92008-11-14 10:39:23 +1100803 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
804 if (retval < 0)
805 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806
David Howellsd84f4f92008-11-14 10:39:23 +1100807 return commit_creds(new);
808
809error:
810 abort_creds(new);
811 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812}
813
Heiko Carstensdbf040d2009-01-14 14:14:04 +0100814SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815{
David Howells86a264a2008-11-14 10:39:18 +1100816 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 int retval;
818
David Howells86a264a2008-11-14 10:39:18 +1100819 if (!(retval = put_user(cred->uid, ruid)) &&
820 !(retval = put_user(cred->euid, euid)))
David Howellsb6dff3e2008-11-14 10:39:16 +1100821 retval = put_user(cred->suid, suid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
823 return retval;
824}
825
826/*
827 * Same as above, but for rgid, egid, sgid.
828 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100829SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830{
David Howellsd84f4f92008-11-14 10:39:23 +1100831 const struct cred *old;
832 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 int retval;
834
David Howellsd84f4f92008-11-14 10:39:23 +1100835 new = prepare_creds();
836 if (!new)
837 return -ENOMEM;
838 old = current_cred();
839
David Howellsd84f4f92008-11-14 10:39:23 +1100840 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700841 if (!nsown_capable(CAP_SETGID)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100842 if (rgid != (gid_t) -1 && rgid != old->gid &&
843 rgid != old->egid && rgid != old->sgid)
844 goto error;
845 if (egid != (gid_t) -1 && egid != old->gid &&
846 egid != old->egid && egid != old->sgid)
847 goto error;
848 if (sgid != (gid_t) -1 && sgid != old->gid &&
849 sgid != old->egid && sgid != old->sgid)
850 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
David Howellsd84f4f92008-11-14 10:39:23 +1100853 if (rgid != (gid_t) -1)
854 new->gid = rgid;
855 if (egid != (gid_t) -1)
856 new->egid = egid;
857 if (sgid != (gid_t) -1)
858 new->sgid = sgid;
859 new->fsgid = new->egid;
860
861 return commit_creds(new);
862
863error:
864 abort_creds(new);
865 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866}
867
Heiko Carstensdbf040d2009-01-14 14:14:04 +0100868SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869{
David Howells86a264a2008-11-14 10:39:18 +1100870 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 int retval;
872
David Howells86a264a2008-11-14 10:39:18 +1100873 if (!(retval = put_user(cred->gid, rgid)) &&
874 !(retval = put_user(cred->egid, egid)))
David Howellsb6dff3e2008-11-14 10:39:16 +1100875 retval = put_user(cred->sgid, sgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876
877 return retval;
878}
879
880
881/*
882 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
883 * is used for "access()" and for the NFS daemon (letting nfsd stay at
884 * whatever uid it wants to). It normally shadows "euid", except when
885 * explicitly set by setfsuid() or for access..
886 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100887SYSCALL_DEFINE1(setfsuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888{
David Howellsd84f4f92008-11-14 10:39:23 +1100889 const struct cred *old;
890 struct cred *new;
891 uid_t old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
David Howellsd84f4f92008-11-14 10:39:23 +1100893 new = prepare_creds();
894 if (!new)
895 return current_fsuid();
896 old = current_cred();
897 old_fsuid = old->fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
David Howellsd84f4f92008-11-14 10:39:23 +1100899 if (uid == old->uid || uid == old->euid ||
900 uid == old->suid || uid == old->fsuid ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700901 nsown_capable(CAP_SETUID)) {
Cal Peake756184b2006-09-30 23:27:24 -0700902 if (uid != old_fsuid) {
David Howellsd84f4f92008-11-14 10:39:23 +1100903 new->fsuid = uid;
904 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
905 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 }
908
David Howellsd84f4f92008-11-14 10:39:23 +1100909 abort_creds(new);
910 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911
David Howellsd84f4f92008-11-14 10:39:23 +1100912change_okay:
913 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 return old_fsuid;
915}
916
917/*
John Anthony Kazos Jrf42df9e2007-05-09 08:23:08 +0200918 * Samma på svenska..
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100920SYSCALL_DEFINE1(setfsgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921{
David Howellsd84f4f92008-11-14 10:39:23 +1100922 const struct cred *old;
923 struct cred *new;
924 gid_t old_fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925
David Howellsd84f4f92008-11-14 10:39:23 +1100926 new = prepare_creds();
927 if (!new)
928 return current_fsgid();
929 old = current_cred();
930 old_fsgid = old->fsgid;
931
David Howellsd84f4f92008-11-14 10:39:23 +1100932 if (gid == old->gid || gid == old->egid ||
933 gid == old->sgid || gid == old->fsgid ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700934 nsown_capable(CAP_SETGID)) {
Cal Peake756184b2006-09-30 23:27:24 -0700935 if (gid != old_fsgid) {
David Howellsd84f4f92008-11-14 10:39:23 +1100936 new->fsgid = gid;
937 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 }
David Howellsd84f4f92008-11-14 10:39:23 +1100940
David Howellsd84f4f92008-11-14 10:39:23 +1100941 abort_creds(new);
942 return old_fsgid;
943
944change_okay:
945 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 return old_fsgid;
947}
948
Frank Mayharf06febc2008-09-12 09:54:39 -0700949void do_sys_times(struct tms *tms)
950{
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +0900951 cputime_t tgutime, tgstime, cutime, cstime;
Frank Mayharf06febc2008-09-12 09:54:39 -0700952
Oleg Nesterov2b5fe6d2008-11-17 15:40:08 +0100953 spin_lock_irq(&current->sighand->siglock);
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +0900954 thread_group_times(current, &tgutime, &tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -0700955 cutime = current->signal->cutime;
956 cstime = current->signal->cstime;
957 spin_unlock_irq(&current->sighand->siglock);
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +0900958 tms->tms_utime = cputime_to_clock_t(tgutime);
959 tms->tms_stime = cputime_to_clock_t(tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -0700960 tms->tms_cutime = cputime_to_clock_t(cutime);
961 tms->tms_cstime = cputime_to_clock_t(cstime);
962}
963
Heiko Carstens58fd3aa2009-01-14 14:14:03 +0100964SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 if (tbuf) {
967 struct tms tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
Frank Mayharf06febc2008-09-12 09:54:39 -0700969 do_sys_times(&tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
971 return -EFAULT;
972 }
Paul Mackerrase3d5a272009-01-06 14:41:02 -0800973 force_successful_syscall_return();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 return (long) jiffies_64_to_clock_t(get_jiffies_64());
975}
976
977/*
978 * This needs some heavy checking ...
979 * I just haven't the stomach for it. I also don't fully
980 * understand sessions/pgrp etc. Let somebody who does explain it.
981 *
982 * OK, I think I have the protection semantics right.... this is really
983 * only important on a multi-user system anyway, to make sure one user
984 * can't send a signal to a process owned by another. -TYT, 12/12/91
985 *
986 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
987 * LBT 04.03.94
988 */
Heiko Carstensb290ebe2009-01-14 14:14:06 +0100989SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990{
991 struct task_struct *p;
Oleg Nesterovee0acf92006-01-08 01:03:53 -0800992 struct task_struct *group_leader = current->group_leader;
Oleg Nesterov4e021302008-02-08 04:19:08 -0800993 struct pid *pgrp;
994 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 if (!pid)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700997 pid = task_pid_vnr(group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 if (!pgid)
999 pgid = pid;
1000 if (pgid < 0)
1001 return -EINVAL;
Paul E. McKenney950eaac2010-08-31 17:00:18 -07001002 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
1004 /* From this point forward we keep holding onto the tasklist lock
1005 * so that our parent does not change from under us. -DaveM
1006 */
1007 write_lock_irq(&tasklist_lock);
1008
1009 err = -ESRCH;
Oleg Nesterov4e021302008-02-08 04:19:08 -08001010 p = find_task_by_vpid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 if (!p)
1012 goto out;
1013
1014 err = -EINVAL;
1015 if (!thread_group_leader(p))
1016 goto out;
1017
Oleg Nesterov4e021302008-02-08 04:19:08 -08001018 if (same_thread_group(p->real_parent, group_leader)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 err = -EPERM;
Eric W. Biederman41487c62007-02-12 00:53:01 -08001020 if (task_session(p) != task_session(group_leader))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 goto out;
1022 err = -EACCES;
1023 if (p->did_exec)
1024 goto out;
1025 } else {
1026 err = -ESRCH;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001027 if (p != group_leader)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 goto out;
1029 }
1030
1031 err = -EPERM;
1032 if (p->signal->leader)
1033 goto out;
1034
Oleg Nesterov4e021302008-02-08 04:19:08 -08001035 pgrp = task_pid(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 if (pgid != pid) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001037 struct task_struct *g;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038
Oleg Nesterov4e021302008-02-08 04:19:08 -08001039 pgrp = find_vpid(pgid);
1040 g = pid_task(pgrp, PIDTYPE_PGID);
Eric W. Biederman41487c62007-02-12 00:53:01 -08001041 if (!g || task_session(g) != task_session(group_leader))
Oleg Nesterovf020bc42006-12-08 02:38:02 -08001042 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 }
1044
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 err = security_task_setpgid(p, pgid);
1046 if (err)
1047 goto out;
1048
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -07001049 if (task_pgrp(p) != pgrp)
Oleg Nesterov83beaf32008-04-30 00:54:27 -07001050 change_pid(p, PIDTYPE_PGID, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
1052 err = 0;
1053out:
1054 /* All paths lead to here, thus we are safe. -DaveM */
1055 write_unlock_irq(&tasklist_lock);
Paul E. McKenney950eaac2010-08-31 17:00:18 -07001056 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 return err;
1058}
1059
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001060SYSCALL_DEFINE1(getpgid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001062 struct task_struct *p;
1063 struct pid *grp;
1064 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001066 rcu_read_lock();
1067 if (!pid)
1068 grp = task_pgrp(current);
1069 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 retval = -ESRCH;
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001071 p = find_task_by_vpid(pid);
1072 if (!p)
1073 goto out;
1074 grp = task_pgrp(p);
1075 if (!grp)
1076 goto out;
1077
1078 retval = security_task_getpgid(p);
1079 if (retval)
1080 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 }
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001082 retval = pid_vnr(grp);
1083out:
1084 rcu_read_unlock();
1085 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086}
1087
1088#ifdef __ARCH_WANT_SYS_GETPGRP
1089
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001090SYSCALL_DEFINE0(getpgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001092 return sys_getpgid(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093}
1094
1095#endif
1096
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001097SYSCALL_DEFINE1(getsid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001099 struct task_struct *p;
1100 struct pid *sid;
1101 int retval;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001102
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001103 rcu_read_lock();
1104 if (!pid)
1105 sid = task_session(current);
1106 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 retval = -ESRCH;
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001108 p = find_task_by_vpid(pid);
1109 if (!p)
1110 goto out;
1111 sid = task_session(p);
1112 if (!sid)
1113 goto out;
1114
1115 retval = security_task_getsid(p);
1116 if (retval)
1117 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 }
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001119 retval = pid_vnr(sid);
1120out:
1121 rcu_read_unlock();
1122 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123}
1124
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001125SYSCALL_DEFINE0(setsid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126{
Oren Laadane19f2472006-01-08 01:03:58 -08001127 struct task_struct *group_leader = current->group_leader;
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001128 struct pid *sid = task_pid(group_leader);
1129 pid_t session = pid_vnr(sid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 int err = -EPERM;
1131
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 write_lock_irq(&tasklist_lock);
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001133 /* Fail if I am already a session leader */
1134 if (group_leader->signal->leader)
1135 goto out;
1136
Oleg Nesterov430c6232008-02-08 04:19:11 -08001137 /* Fail if a process group id already exists that equals the
1138 * proposed session id.
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001139 */
Oleg Nesterov6806aac2008-02-08 04:19:12 -08001140 if (pid_task(sid, PIDTYPE_PGID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 goto out;
1142
Oren Laadane19f2472006-01-08 01:03:58 -08001143 group_leader->signal->leader = 1;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -08001144 __set_special_pids(sid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001145
Alan Cox9c9f4de2008-10-13 10:37:26 +01001146 proc_clear_tty(group_leader);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001147
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001148 err = session;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149out:
1150 write_unlock_irq(&tasklist_lock);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001151 if (err > 0) {
Christian Borntraeger0d0df592009-10-26 16:49:34 -07001152 proc_sid_connector(group_leader);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001153 sched_autogroup_create_attach(group_leader);
1154 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 return err;
1156}
1157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158DECLARE_RWSEM(uts_sem);
1159
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001160#ifdef COMPAT_UTS_MACHINE
1161#define override_architecture(name) \
Andreas Schwab46da2762010-04-23 13:17:44 -04001162 (personality(current->personality) == PER_LINUX32 && \
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001163 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1164 sizeof(COMPAT_UTS_MACHINE)))
1165#else
1166#define override_architecture(name) 0
1167#endif
1168
Andi Kleenbe274252011-08-19 16:15:10 -07001169/*
1170 * Work around broken programs that cannot handle "Linux 3.0".
1171 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1172 */
1173static int override_release(char __user *release, int len)
1174{
1175 int ret = 0;
Linus Torvaldsa84a79e2011-10-17 08:24:24 -07001176 char buf[65];
Andi Kleenbe274252011-08-19 16:15:10 -07001177
1178 if (current->personality & UNAME26) {
1179 char *rest = UTS_RELEASE;
1180 int ndots = 0;
1181 unsigned v;
1182
1183 while (*rest) {
1184 if (*rest == '.' && ++ndots >= 3)
1185 break;
1186 if (!isdigit(*rest) && *rest != '.')
1187 break;
1188 rest++;
1189 }
1190 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1191 snprintf(buf, len, "2.6.%u%s", v, rest);
1192 ret = copy_to_user(release, buf, len);
1193 }
1194 return ret;
1195}
1196
Heiko Carstense48fbb62009-01-14 14:14:26 +01001197SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198{
1199 int errno = 0;
1200
1201 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001202 if (copy_to_user(name, utsname(), sizeof *name))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 errno = -EFAULT;
1204 up_read(&uts_sem);
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001205
Andi Kleenbe274252011-08-19 16:15:10 -07001206 if (!errno && override_release(name->release, sizeof(name->release)))
1207 errno = -EFAULT;
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001208 if (!errno && override_architecture(name))
1209 errno = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 return errno;
1211}
1212
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001213#ifdef __ARCH_WANT_SYS_OLD_UNAME
1214/*
1215 * Old cruft
1216 */
1217SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1218{
1219 int error = 0;
1220
1221 if (!name)
1222 return -EFAULT;
1223
1224 down_read(&uts_sem);
1225 if (copy_to_user(name, utsname(), sizeof(*name)))
1226 error = -EFAULT;
1227 up_read(&uts_sem);
1228
Andi Kleenbe274252011-08-19 16:15:10 -07001229 if (!error && override_release(name->release, sizeof(name->release)))
1230 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001231 if (!error && override_architecture(name))
1232 error = -EFAULT;
1233 return error;
1234}
1235
1236SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1237{
1238 int error;
1239
1240 if (!name)
1241 return -EFAULT;
1242 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1243 return -EFAULT;
1244
1245 down_read(&uts_sem);
1246 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1247 __OLD_UTS_LEN);
1248 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1249 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1250 __OLD_UTS_LEN);
1251 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1252 error |= __copy_to_user(&name->release, &utsname()->release,
1253 __OLD_UTS_LEN);
1254 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1255 error |= __copy_to_user(&name->version, &utsname()->version,
1256 __OLD_UTS_LEN);
1257 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1258 error |= __copy_to_user(&name->machine, &utsname()->machine,
1259 __OLD_UTS_LEN);
1260 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1261 up_read(&uts_sem);
1262
1263 if (!error && override_architecture(name))
1264 error = -EFAULT;
Andi Kleenbe274252011-08-19 16:15:10 -07001265 if (!error && override_release(name->release, sizeof(name->release)))
1266 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001267 return error ? -EFAULT : 0;
1268}
1269#endif
1270
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001271SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272{
1273 int errno;
1274 char tmp[__NEW_UTS_LEN];
1275
Serge E. Hallynbb96a6f2011-03-23 16:43:18 -07001276 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 return -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001278
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 if (len < 0 || len > __NEW_UTS_LEN)
1280 return -EINVAL;
1281 down_write(&uts_sem);
1282 errno = -EFAULT;
1283 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001284 struct new_utsname *u = utsname();
1285
1286 memcpy(u->nodename, tmp, len);
1287 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 errno = 0;
1289 }
Lucas De Marchif1ecf062011-11-02 13:39:22 -07001290 uts_proc_notify(UTS_PROC_HOSTNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 up_write(&uts_sem);
1292 return errno;
1293}
1294
1295#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1296
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001297SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298{
1299 int i, errno;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001300 struct new_utsname *u;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301
1302 if (len < 0)
1303 return -EINVAL;
1304 down_read(&uts_sem);
Andrew Morton9679e4d2008-10-15 22:01:51 -07001305 u = utsname();
1306 i = 1 + strlen(u->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 if (i > len)
1308 i = len;
1309 errno = 0;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001310 if (copy_to_user(name, u->nodename, i))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 errno = -EFAULT;
1312 up_read(&uts_sem);
1313 return errno;
1314}
1315
1316#endif
1317
1318/*
1319 * Only setdomainname; getdomainname can be implemented by calling
1320 * uname()
1321 */
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001322SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323{
1324 int errno;
1325 char tmp[__NEW_UTS_LEN];
1326
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001327 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 return -EPERM;
1329 if (len < 0 || len > __NEW_UTS_LEN)
1330 return -EINVAL;
1331
1332 down_write(&uts_sem);
1333 errno = -EFAULT;
1334 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001335 struct new_utsname *u = utsname();
1336
1337 memcpy(u->domainname, tmp, len);
1338 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 errno = 0;
1340 }
Lucas De Marchif1ecf062011-11-02 13:39:22 -07001341 uts_proc_notify(UTS_PROC_DOMAINNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 up_write(&uts_sem);
1343 return errno;
1344}
1345
Heiko Carstense48fbb62009-01-14 14:14:26 +01001346SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347{
Jiri Slabyb9518342010-05-04 11:28:25 +02001348 struct rlimit value;
1349 int ret;
1350
1351 ret = do_prlimit(current, resource, NULL, &value);
1352 if (!ret)
1353 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1354
1355 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356}
1357
1358#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1359
1360/*
1361 * Back compatibility for getrlimit. Needed for some apps.
1362 */
1363
Heiko Carstense48fbb62009-01-14 14:14:26 +01001364SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1365 struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366{
1367 struct rlimit x;
1368 if (resource >= RLIM_NLIMITS)
1369 return -EINVAL;
1370
1371 task_lock(current->group_leader);
1372 x = current->signal->rlim[resource];
1373 task_unlock(current->group_leader);
Cal Peake756184b2006-09-30 23:27:24 -07001374 if (x.rlim_cur > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 x.rlim_cur = 0x7FFFFFFF;
Cal Peake756184b2006-09-30 23:27:24 -07001376 if (x.rlim_max > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 x.rlim_max = 0x7FFFFFFF;
1378 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1379}
1380
1381#endif
1382
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001383static inline bool rlim64_is_infinity(__u64 rlim64)
1384{
1385#if BITS_PER_LONG < 64
1386 return rlim64 >= ULONG_MAX;
1387#else
1388 return rlim64 == RLIM64_INFINITY;
1389#endif
1390}
1391
1392static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1393{
1394 if (rlim->rlim_cur == RLIM_INFINITY)
1395 rlim64->rlim_cur = RLIM64_INFINITY;
1396 else
1397 rlim64->rlim_cur = rlim->rlim_cur;
1398 if (rlim->rlim_max == RLIM_INFINITY)
1399 rlim64->rlim_max = RLIM64_INFINITY;
1400 else
1401 rlim64->rlim_max = rlim->rlim_max;
1402}
1403
1404static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1405{
1406 if (rlim64_is_infinity(rlim64->rlim_cur))
1407 rlim->rlim_cur = RLIM_INFINITY;
1408 else
1409 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1410 if (rlim64_is_infinity(rlim64->rlim_max))
1411 rlim->rlim_max = RLIM_INFINITY;
1412 else
1413 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1414}
1415
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001416/* make sure you are allowed to change @tsk limits before calling this */
Jiri Slaby5b415352010-03-24 16:11:29 +01001417int do_prlimit(struct task_struct *tsk, unsigned int resource,
1418 struct rlimit *new_rlim, struct rlimit *old_rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419{
Jiri Slaby5b415352010-03-24 16:11:29 +01001420 struct rlimit *rlim;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001421 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422
1423 if (resource >= RLIM_NLIMITS)
1424 return -EINVAL;
Jiri Slaby5b415352010-03-24 16:11:29 +01001425 if (new_rlim) {
1426 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1427 return -EINVAL;
1428 if (resource == RLIMIT_NOFILE &&
1429 new_rlim->rlim_max > sysctl_nr_open)
1430 return -EPERM;
1431 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001433 /* protect tsk->signal and tsk->sighand from disappearing */
1434 read_lock(&tasklist_lock);
1435 if (!tsk->sighand) {
1436 retval = -ESRCH;
1437 goto out;
1438 }
1439
Jiri Slaby5b415352010-03-24 16:11:29 +01001440 rlim = tsk->signal->rlim + resource;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001441 task_lock(tsk->group_leader);
Jiri Slaby5b415352010-03-24 16:11:29 +01001442 if (new_rlim) {
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001443 /* Keep the capable check against init_user_ns until
1444 cgroups can contain all limits */
Jiri Slaby5b415352010-03-24 16:11:29 +01001445 if (new_rlim->rlim_max > rlim->rlim_max &&
1446 !capable(CAP_SYS_RESOURCE))
1447 retval = -EPERM;
1448 if (!retval)
1449 retval = security_task_setrlimit(tsk->group_leader,
1450 resource, new_rlim);
1451 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1452 /*
1453 * The caller is asking for an immediate RLIMIT_CPU
1454 * expiry. But we use the zero value to mean "it was
1455 * never set". So let's cheat and make it one second
1456 * instead
1457 */
1458 new_rlim->rlim_cur = 1;
1459 }
Tom Alsberg9926e4c2007-05-08 00:30:31 -07001460 }
Jiri Slaby5b415352010-03-24 16:11:29 +01001461 if (!retval) {
1462 if (old_rlim)
1463 *old_rlim = *rlim;
1464 if (new_rlim)
1465 *rlim = *new_rlim;
1466 }
Jiri Slaby7855c352009-08-26 23:45:34 +02001467 task_unlock(tsk->group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
Andrew Mortond3561f72006-03-24 03:18:36 -08001469 /*
1470 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1471 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1472 * very long-standing error, and fixing it now risks breakage of
1473 * applications, so we live with it
1474 */
Jiri Slaby5b415352010-03-24 16:11:29 +01001475 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1476 new_rlim->rlim_cur != RLIM_INFINITY)
1477 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001478out:
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001479 read_unlock(&tasklist_lock);
Oleg Nesterov2fb9d262009-09-03 19:21:45 +02001480 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481}
1482
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001483/* rcu lock must be held */
1484static int check_prlimit_permission(struct task_struct *task)
1485{
1486 const struct cred *cred = current_cred(), *tcred;
1487
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001488 if (current == task)
1489 return 0;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001490
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001491 tcred = __task_cred(task);
1492 if (cred->user->user_ns == tcred->user->user_ns &&
1493 (cred->uid == tcred->euid &&
1494 cred->uid == tcred->suid &&
1495 cred->uid == tcred->uid &&
1496 cred->gid == tcred->egid &&
1497 cred->gid == tcred->sgid &&
1498 cred->gid == tcred->gid))
1499 return 0;
1500 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1501 return 0;
1502
1503 return -EPERM;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001504}
1505
1506SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1507 const struct rlimit64 __user *, new_rlim,
1508 struct rlimit64 __user *, old_rlim)
1509{
1510 struct rlimit64 old64, new64;
1511 struct rlimit old, new;
1512 struct task_struct *tsk;
1513 int ret;
1514
1515 if (new_rlim) {
1516 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1517 return -EFAULT;
1518 rlim64_to_rlim(&new64, &new);
1519 }
1520
1521 rcu_read_lock();
1522 tsk = pid ? find_task_by_vpid(pid) : current;
1523 if (!tsk) {
1524 rcu_read_unlock();
1525 return -ESRCH;
1526 }
1527 ret = check_prlimit_permission(tsk);
1528 if (ret) {
1529 rcu_read_unlock();
1530 return ret;
1531 }
1532 get_task_struct(tsk);
1533 rcu_read_unlock();
1534
1535 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1536 old_rlim ? &old : NULL);
1537
1538 if (!ret && old_rlim) {
1539 rlim_to_rlim64(&old, &old64);
1540 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1541 ret = -EFAULT;
1542 }
1543
1544 put_task_struct(tsk);
1545 return ret;
1546}
1547
Jiri Slaby7855c352009-08-26 23:45:34 +02001548SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1549{
1550 struct rlimit new_rlim;
1551
1552 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1553 return -EFAULT;
Jiri Slaby5b415352010-03-24 16:11:29 +01001554 return do_prlimit(current, resource, &new_rlim, NULL);
Jiri Slaby7855c352009-08-26 23:45:34 +02001555}
1556
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557/*
1558 * It would make sense to put struct rusage in the task_struct,
1559 * except that would make the task_struct be *really big*. After
1560 * task_struct gets moved into malloc'ed memory, it would
1561 * make sense to do this. It will make moving the rest of the information
1562 * a lot simpler! (Which we're not doing right now because we're not
1563 * measuring them yet).
1564 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1566 * races with threads incrementing their own counters. But since word
1567 * reads are atomic, we either get new values or old values and we don't
1568 * care which for the sums. We always take the siglock to protect reading
1569 * the c* fields from p->signal from races with exit.c updating those
1570 * fields when reaping, so a sample either gets all the additions of a
1571 * given child after it's reaped, or none so this sample is before reaping.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001572 *
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001573 * Locking:
1574 * We need to take the siglock for CHILDEREN, SELF and BOTH
1575 * for the cases current multithreaded, non-current single threaded
1576 * non-current multithreaded. Thread traversal is now safe with
1577 * the siglock held.
1578 * Strictly speaking, we donot need to take the siglock if we are current and
1579 * single threaded, as no one else can take our signal_struct away, no one
1580 * else can reap the children to update signal->c* counters, and no one else
1581 * can race with the signal-> fields. If we do not take any lock, the
1582 * signal-> fields could be read out of order while another thread was just
1583 * exiting. So we should place a read memory barrier when we avoid the lock.
1584 * On the writer side, write memory barrier is implied in __exit_signal
1585 * as __exit_signal releases the siglock spinlock after updating the signal->
1586 * fields. But we don't do this yet to keep things simple.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001587 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 */
1589
Frank Mayharf06febc2008-09-12 09:54:39 -07001590static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001591{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001592 r->ru_nvcsw += t->nvcsw;
1593 r->ru_nivcsw += t->nivcsw;
1594 r->ru_minflt += t->min_flt;
1595 r->ru_majflt += t->maj_flt;
1596 r->ru_inblock += task_io_get_inblock(t);
1597 r->ru_oublock += task_io_get_oublock(t);
1598}
1599
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1601{
1602 struct task_struct *t;
1603 unsigned long flags;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001604 cputime_t tgutime, tgstime, utime, stime;
Jiri Pirko1f102062009-09-22 16:44:10 -07001605 unsigned long maxrss = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606
1607 memset((char *) r, 0, sizeof *r);
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001608 utime = stime = cputime_zero;
1609
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001610 if (who == RUSAGE_THREAD) {
Hidetoshi Setod180c5b2009-11-26 14:48:30 +09001611 task_times(current, &utime, &stime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001612 accumulate_thread_rusage(p, r);
Jiri Pirko1f102062009-09-22 16:44:10 -07001613 maxrss = p->signal->maxrss;
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001614 goto out;
1615 }
1616
Oleg Nesterovd6cf7232008-04-30 00:52:38 -07001617 if (!lock_task_sighand(p, &flags))
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001618 return;
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001619
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 switch (who) {
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001621 case RUSAGE_BOTH:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 case RUSAGE_CHILDREN:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 utime = p->signal->cutime;
1624 stime = p->signal->cstime;
1625 r->ru_nvcsw = p->signal->cnvcsw;
1626 r->ru_nivcsw = p->signal->cnivcsw;
1627 r->ru_minflt = p->signal->cmin_flt;
1628 r->ru_majflt = p->signal->cmaj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001629 r->ru_inblock = p->signal->cinblock;
1630 r->ru_oublock = p->signal->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001631 maxrss = p->signal->cmaxrss;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001632
1633 if (who == RUSAGE_CHILDREN)
1634 break;
1635
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 case RUSAGE_SELF:
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001637 thread_group_times(p, &tgutime, &tgstime);
1638 utime = cputime_add(utime, tgutime);
1639 stime = cputime_add(stime, tgstime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 r->ru_nvcsw += p->signal->nvcsw;
1641 r->ru_nivcsw += p->signal->nivcsw;
1642 r->ru_minflt += p->signal->min_flt;
1643 r->ru_majflt += p->signal->maj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001644 r->ru_inblock += p->signal->inblock;
1645 r->ru_oublock += p->signal->oublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001646 if (maxrss < p->signal->maxrss)
1647 maxrss = p->signal->maxrss;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 t = p;
1649 do {
Frank Mayharf06febc2008-09-12 09:54:39 -07001650 accumulate_thread_rusage(t, r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 t = next_thread(t);
1652 } while (t != p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 break;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001654
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 default:
1656 BUG();
1657 }
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001658 unlock_task_sighand(p, &flags);
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001659
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001660out:
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001661 cputime_to_timeval(utime, &r->ru_utime);
1662 cputime_to_timeval(stime, &r->ru_stime);
Jiri Pirko1f102062009-09-22 16:44:10 -07001663
1664 if (who != RUSAGE_CHILDREN) {
1665 struct mm_struct *mm = get_task_mm(p);
1666 if (mm) {
1667 setmax_mm_hiwater_rss(&maxrss, mm);
1668 mmput(mm);
1669 }
1670 }
1671 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672}
1673
1674int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1675{
1676 struct rusage r;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 k_getrusage(p, who, &r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1679}
1680
Heiko Carstense48fbb62009-01-14 14:14:26 +01001681SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001683 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1684 who != RUSAGE_THREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 return -EINVAL;
1686 return getrusage(current, who, ru);
1687}
1688
Heiko Carstense48fbb62009-01-14 14:14:26 +01001689SYSCALL_DEFINE1(umask, int, mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690{
1691 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1692 return mask;
1693}
Serge E. Hallyn3b7391d2008-02-04 22:29:45 -08001694
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001695SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1696 unsigned long, arg4, unsigned long, arg5)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697{
David Howellsb6dff3e2008-11-14 10:39:16 +11001698 struct task_struct *me = current;
1699 unsigned char comm[sizeof(me->comm)];
1700 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
David Howellsd84f4f92008-11-14 10:39:23 +11001702 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1703 if (error != -ENOSYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 return error;
1705
David Howellsd84f4f92008-11-14 10:39:23 +11001706 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 switch (option) {
1708 case PR_SET_PDEATHSIG:
Jesper Juhl0730ded2005-09-06 15:17:37 -07001709 if (!valid_signal(arg2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 error = -EINVAL;
1711 break;
1712 }
David Howellsb6dff3e2008-11-14 10:39:16 +11001713 me->pdeath_signal = arg2;
1714 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 break;
1716 case PR_GET_PDEATHSIG:
David Howellsb6dff3e2008-11-14 10:39:16 +11001717 error = put_user(me->pdeath_signal, (int __user *)arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 break;
1719 case PR_GET_DUMPABLE:
David Howellsb6dff3e2008-11-14 10:39:16 +11001720 error = get_dumpable(me->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 break;
1722 case PR_SET_DUMPABLE:
Marcel Holtmannabf75a52006-07-12 13:12:00 +02001723 if (arg2 < 0 || arg2 > 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 error = -EINVAL;
1725 break;
1726 }
David Howellsb6dff3e2008-11-14 10:39:16 +11001727 set_dumpable(me->mm, arg2);
1728 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 break;
1730
1731 case PR_SET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001732 error = SET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 break;
1734 case PR_GET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001735 error = GET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 break;
1737 case PR_SET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11001738 error = SET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 break;
1740 case PR_GET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11001741 error = GET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 break;
1743 case PR_SET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11001744 error = SET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 break;
1746 case PR_GET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11001747 error = GET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 break;
1749 case PR_GET_TIMING:
1750 error = PR_TIMING_STATISTICAL;
1751 break;
1752 case PR_SET_TIMING:
Shi Weihua7b266552008-05-23 13:04:59 -07001753 if (arg2 != PR_TIMING_STATISTICAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 error = -EINVAL;
David Howellsb6dff3e2008-11-14 10:39:16 +11001755 else
1756 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 break;
1758
David Howellsb6dff3e2008-11-14 10:39:16 +11001759 case PR_SET_NAME:
1760 comm[sizeof(me->comm)-1] = 0;
1761 if (strncpy_from_user(comm, (char __user *)arg2,
1762 sizeof(me->comm) - 1) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 return -EFAULT;
David Howellsb6dff3e2008-11-14 10:39:16 +11001764 set_task_comm(me, comm);
Vladimir Zapolskiyf786ecb2011-09-21 09:26:44 +00001765 proc_comm_connector(me);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 return 0;
David Howellsb6dff3e2008-11-14 10:39:16 +11001767 case PR_GET_NAME:
1768 get_task_comm(comm, me);
1769 if (copy_to_user((char __user *)arg2, comm,
1770 sizeof(comm)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 return -EFAULT;
1772 return 0;
Anton Blanchard651d7652006-06-07 16:10:19 +10001773 case PR_GET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001774 error = GET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10001775 break;
1776 case PR_SET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11001777 error = SET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10001778 break;
1779
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -07001780 case PR_GET_SECCOMP:
1781 error = prctl_get_seccomp();
1782 break;
1783 case PR_SET_SECCOMP:
1784 error = prctl_set_seccomp(arg2);
1785 break;
Erik Bosman8fb402b2008-04-11 18:54:17 +02001786 case PR_GET_TSC:
1787 error = GET_TSC_CTL(arg2);
1788 break;
1789 case PR_SET_TSC:
1790 error = SET_TSC_CTL(arg2);
1791 break;
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001792 case PR_TASK_PERF_EVENTS_DISABLE:
1793 error = perf_event_task_disable();
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01001794 break;
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001795 case PR_TASK_PERF_EVENTS_ENABLE:
1796 error = perf_event_task_enable();
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01001797 break;
Arjan van de Ven69766752008-09-01 15:52:40 -07001798 case PR_GET_TIMERSLACK:
1799 error = current->timer_slack_ns;
1800 break;
1801 case PR_SET_TIMERSLACK:
1802 if (arg2 <= 0)
1803 current->timer_slack_ns =
1804 current->default_timer_slack_ns;
1805 else
1806 current->timer_slack_ns = arg2;
David Howellsb6dff3e2008-11-14 10:39:16 +11001807 error = 0;
Arjan van de Ven69766752008-09-01 15:52:40 -07001808 break;
Andi Kleen4db96cf2009-09-16 11:50:14 +02001809 case PR_MCE_KILL:
1810 if (arg4 | arg5)
1811 return -EINVAL;
1812 switch (arg2) {
Andi Kleen1087e9b2009-10-04 02:20:11 +02001813 case PR_MCE_KILL_CLEAR:
Andi Kleen4db96cf2009-09-16 11:50:14 +02001814 if (arg3 != 0)
1815 return -EINVAL;
1816 current->flags &= ~PF_MCE_PROCESS;
1817 break;
Andi Kleen1087e9b2009-10-04 02:20:11 +02001818 case PR_MCE_KILL_SET:
Andi Kleen4db96cf2009-09-16 11:50:14 +02001819 current->flags |= PF_MCE_PROCESS;
Andi Kleen1087e9b2009-10-04 02:20:11 +02001820 if (arg3 == PR_MCE_KILL_EARLY)
Andi Kleen4db96cf2009-09-16 11:50:14 +02001821 current->flags |= PF_MCE_EARLY;
Andi Kleen1087e9b2009-10-04 02:20:11 +02001822 else if (arg3 == PR_MCE_KILL_LATE)
Andi Kleen4db96cf2009-09-16 11:50:14 +02001823 current->flags &= ~PF_MCE_EARLY;
Andi Kleen1087e9b2009-10-04 02:20:11 +02001824 else if (arg3 == PR_MCE_KILL_DEFAULT)
1825 current->flags &=
1826 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1827 else
1828 return -EINVAL;
Andi Kleen4db96cf2009-09-16 11:50:14 +02001829 break;
1830 default:
1831 return -EINVAL;
1832 }
1833 error = 0;
1834 break;
Andi Kleen1087e9b2009-10-04 02:20:11 +02001835 case PR_MCE_KILL_GET:
1836 if (arg2 | arg3 | arg4 | arg5)
1837 return -EINVAL;
1838 if (current->flags & PF_MCE_PROCESS)
1839 error = (current->flags & PF_MCE_EARLY) ?
1840 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1841 else
1842 error = PR_MCE_KILL_DEFAULT;
1843 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 default:
1845 error = -EINVAL;
1846 break;
1847 }
1848 return error;
1849}
Andi Kleen3cfc3482006-09-26 10:52:28 +02001850
Heiko Carstens836f92a2009-01-14 14:14:33 +01001851SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1852 struct getcpu_cache __user *, unused)
Andi Kleen3cfc3482006-09-26 10:52:28 +02001853{
1854 int err = 0;
1855 int cpu = raw_smp_processor_id();
1856 if (cpup)
1857 err |= put_user(cpu, cpup);
1858 if (nodep)
1859 err |= put_user(cpu_to_node(cpu), nodep);
Andi Kleen3cfc3482006-09-26 10:52:28 +02001860 return err ? -EFAULT : 0;
1861}
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001862
1863char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1864
Neil Hormana06a4dc2010-05-26 14:42:58 -07001865static void argv_cleanup(struct subprocess_info *info)
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001866{
Neil Hormana06a4dc2010-05-26 14:42:58 -07001867 argv_free(info->argv);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001868}
1869
1870/**
1871 * orderly_poweroff - Trigger an orderly system poweroff
1872 * @force: force poweroff if command execution fails
1873 *
1874 * This may be called from any context to trigger a system shutdown.
1875 * If the orderly shutdown fails, it will force an immediate shutdown.
1876 */
1877int orderly_poweroff(bool force)
1878{
1879 int argc;
1880 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1881 static char *envp[] = {
1882 "HOME=/",
1883 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1884 NULL
1885 };
1886 int ret = -ENOMEM;
1887 struct subprocess_info *info;
1888
1889 if (argv == NULL) {
1890 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1891 __func__, poweroff_cmd);
1892 goto out;
1893 }
1894
KOSAKI Motohiroac331d12008-07-25 01:45:38 -07001895 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001896 if (info == NULL) {
1897 argv_free(argv);
1898 goto out;
1899 }
1900
Neil Hormana06a4dc2010-05-26 14:42:58 -07001901 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001902
Jeremy Fitzhardinge86313c42007-07-17 18:37:03 -07001903 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07001904
1905 out:
1906 if (ret && force) {
1907 printk(KERN_WARNING "Failed to start orderly shutdown: "
1908 "forcing the issue\n");
1909
1910 /* I guess this should try to kick off some daemon to
1911 sync and poweroff asap. Or not even bother syncing
1912 if we're doing an emergency shutdown? */
1913 emergency_sync();
1914 kernel_power_off();
1915 }
1916
1917 return ret;
1918}
1919EXPORT_SYMBOL_GPL(orderly_poweroff);