blob: 54f20fdee93c300a57556e004ffbf7f0b5e83cff [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>
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -070039#include <linux/file.h>
40#include <linux/mount.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090041#include <linux/gfp.h>
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +010042#include <linux/syscore_ops.h>
Andi Kleenbe274252011-08-19 16:15:10 -070043#include <linux/version.h>
44#include <linux/ctype.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
46#include <linux/compat.h>
47#include <linux/syscalls.h>
Keshavamurthy Anil S00d7c052005-12-12 00:37:33 -080048#include <linux/kprobes.h>
Cedric Le Goateracce2922007-07-15 23:40:59 -070049#include <linux/user_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
Seiji Aguchi04c68622011-01-12 16:59:30 -080051#include <linux/kmsg_dump.h>
Andi Kleenbe274252011-08-19 16:15:10 -070052/* Move somewhere else to avoid recompiling? */
53#include <generated/utsrelease.h>
Seiji Aguchi04c68622011-01-12 16:59:30 -080054
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#include <asm/uaccess.h>
56#include <asm/io.h>
57#include <asm/unistd.h>
58
59#ifndef SET_UNALIGN_CTL
60# define SET_UNALIGN_CTL(a,b) (-EINVAL)
61#endif
62#ifndef GET_UNALIGN_CTL
63# define GET_UNALIGN_CTL(a,b) (-EINVAL)
64#endif
65#ifndef SET_FPEMU_CTL
66# define SET_FPEMU_CTL(a,b) (-EINVAL)
67#endif
68#ifndef GET_FPEMU_CTL
69# define GET_FPEMU_CTL(a,b) (-EINVAL)
70#endif
71#ifndef SET_FPEXC_CTL
72# define SET_FPEXC_CTL(a,b) (-EINVAL)
73#endif
74#ifndef GET_FPEXC_CTL
75# define GET_FPEXC_CTL(a,b) (-EINVAL)
76#endif
Anton Blanchard651d7652006-06-07 16:10:19 +100077#ifndef GET_ENDIAN
78# define GET_ENDIAN(a,b) (-EINVAL)
79#endif
80#ifndef SET_ENDIAN
81# define SET_ENDIAN(a,b) (-EINVAL)
82#endif
Erik Bosman8fb402b2008-04-11 18:54:17 +020083#ifndef GET_TSC_CTL
84# define GET_TSC_CTL(a) (-EINVAL)
85#endif
86#ifndef SET_TSC_CTL
87# define SET_TSC_CTL(a) (-EINVAL)
88#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
90/*
91 * this is where the system-wide overflow UID and GID are defined, for
92 * architectures that now have 32-bit UID/GID but didn't in the past
93 */
94
95int overflowuid = DEFAULT_OVERFLOWUID;
96int overflowgid = DEFAULT_OVERFLOWGID;
97
Linus Torvalds1da177e2005-04-16 15:20:36 -070098EXPORT_SYMBOL(overflowuid);
99EXPORT_SYMBOL(overflowgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
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
Eric W. Biederman5af66202012-03-03 20:21:47 -0800136 if (uid_eq(pcred->uid, cred->euid) ||
137 uid_eq(pcred->euid, cred->euid))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700138 return true;
Eric W. Biedermanc4a4d602011-11-16 23:15:31 -0800139 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700140 return true;
141 return false;
142}
143
144/*
David Howellsc69e8d92008-11-14 10:39:19 +1100145 * set the priority of a task
146 * - the caller must hold the RCU read lock
147 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148static int set_one_prio(struct task_struct *p, int niceval, int error)
149{
150 int no_nice;
151
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700152 if (!set_one_prio_perm(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 error = -EPERM;
154 goto out;
155 }
Matt Mackalle43379f2005-05-01 08:59:00 -0700156 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 error = -EACCES;
158 goto out;
159 }
160 no_nice = security_task_setnice(p, niceval);
161 if (no_nice) {
162 error = no_nice;
163 goto out;
164 }
165 if (error == -ESRCH)
166 error = 0;
167 set_user_nice(p, niceval);
168out:
169 return error;
170}
171
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100172SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173{
174 struct task_struct *g, *p;
175 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100176 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 int error = -EINVAL;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800178 struct pid *pgrp;
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800179 kuid_t uid;
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:
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800212 uid = make_kuid(cred->user_ns, who);
Eric W. Biederman74ba5082012-03-03 18:58:11 -0800213 user = cred->user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 if (!who)
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800215 uid = cred->uid;
216 else if (!uid_eq(uid, cred->uid) &&
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800217 !(user = find_user(uid)))
David Howells86a264a2008-11-14 10:39:18 +1100218 goto out_unlock; /* No processes for this user */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800220 do_each_thread(g, p) {
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800221 if (uid_eq(task_uid(p), uid))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 error = set_one_prio(p, niceval, error);
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800223 } while_each_thread(g, p);
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800224 if (!uid_eq(uid, cred->uid))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 free_uid(user); /* For find_user() */
226 break;
227 }
228out_unlock:
229 read_unlock(&tasklist_lock);
Thomas Gleixnerd4581a22009-12-10 00:52:51 +0000230 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231out:
232 return error;
233}
234
235/*
236 * Ugh. To avoid negative return values, "getpriority()" will
237 * not return the normal nice-value, but a negated value that
238 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
239 * to stay compatible.
240 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100241SYSCALL_DEFINE2(getpriority, int, which, int, who)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242{
243 struct task_struct *g, *p;
244 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100245 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 long niceval, retval = -ESRCH;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800247 struct pid *pgrp;
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800248 kuid_t uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Daniel Walker3e88c552007-05-10 22:22:53 -0700250 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 return -EINVAL;
252
Tetsuo Handa70118832010-02-22 12:44:16 -0800253 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 read_lock(&tasklist_lock);
255 switch (which) {
256 case PRIO_PROCESS:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800257 if (who)
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700258 p = find_task_by_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800259 else
260 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 if (p) {
262 niceval = 20 - task_nice(p);
263 if (niceval > retval)
264 retval = niceval;
265 }
266 break;
267 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800268 if (who)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700269 pgrp = find_vpid(who);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800270 else
271 pgrp = task_pgrp(current);
Ken Chen2d70b682008-08-20 14:09:17 -0700272 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 niceval = 20 - task_nice(p);
274 if (niceval > retval)
275 retval = niceval;
Ken Chen2d70b682008-08-20 14:09:17 -0700276 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 break;
278 case PRIO_USER:
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800279 uid = make_kuid(cred->user_ns, who);
Eric W. Biederman74ba5082012-03-03 18:58:11 -0800280 user = cred->user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 if (!who)
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800282 uid = cred->uid;
283 else if (!uid_eq(uid, cred->uid) &&
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800284 !(user = find_user(uid)))
David Howells86a264a2008-11-14 10:39:18 +1100285 goto out_unlock; /* No processes for this user */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800287 do_each_thread(g, p) {
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800288 if (uid_eq(task_uid(p), uid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 niceval = 20 - task_nice(p);
290 if (niceval > retval)
291 retval = niceval;
292 }
H Hartley Sweetendfc6a732009-12-14 18:00:22 -0800293 } while_each_thread(g, p);
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800294 if (!uid_eq(uid, cred->uid))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 free_uid(user); /* for find_user() */
296 break;
297 }
298out_unlock:
299 read_unlock(&tasklist_lock);
Tetsuo Handa70118832010-02-22 12:44:16 -0800300 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302 return retval;
303}
304
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700305/**
306 * emergency_restart - reboot the system
307 *
308 * Without shutting down any hardware or taking any locks
309 * reboot the system. This is called when we know we are in
310 * trouble so this is our best effort to reboot. This is
311 * safe to call in interrupt context.
312 */
Eric W. Biederman7c903472005-07-26 11:29:55 -0600313void emergency_restart(void)
314{
Seiji Aguchi04c68622011-01-12 16:59:30 -0800315 kmsg_dump(KMSG_DUMP_EMERG);
Eric W. Biederman7c903472005-07-26 11:29:55 -0600316 machine_emergency_restart();
317}
318EXPORT_SYMBOL_GPL(emergency_restart);
319
Huang Yingca195b72008-08-15 00:40:24 -0700320void kernel_restart_prepare(char *cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600321{
Alan Sterne041c682006-03-27 01:16:30 -0800322 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600323 system_state = SYSTEM_RESTART;
Kay Sieversb50fa7c2011-05-05 13:32:05 +0200324 usermodehelper_disable();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600325 device_shutdown();
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100326 syscore_shutdown();
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700327}
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800328
329/**
Amerigo Wangc5f41752011-07-25 17:13:10 -0700330 * register_reboot_notifier - Register function to be called at reboot time
331 * @nb: Info about notifier function to be called
332 *
333 * Registers a function with the list of functions
334 * to be called at reboot time.
335 *
336 * Currently always returns zero, as blocking_notifier_chain_register()
337 * always returns zero.
338 */
339int register_reboot_notifier(struct notifier_block *nb)
340{
341 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
342}
343EXPORT_SYMBOL(register_reboot_notifier);
344
345/**
346 * unregister_reboot_notifier - Unregister previously registered reboot notifier
347 * @nb: Hook to be unregistered
348 *
349 * Unregisters a previously registered reboot
350 * notifier function.
351 *
352 * Returns zero on success, or %-ENOENT on failure.
353 */
354int unregister_reboot_notifier(struct notifier_block *nb)
355{
356 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
357}
358EXPORT_SYMBOL(unregister_reboot_notifier);
359
360/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800361 * kernel_restart - reboot the system
362 * @cmd: pointer to buffer containing command to execute for restart
Randy Dunlapb8887e62005-11-07 01:01:07 -0800363 * or %NULL
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800364 *
365 * Shutdown everything and perform a clean reboot.
366 * This is not safe to call in interrupt context.
367 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700368void kernel_restart(char *cmd)
369{
370 kernel_restart_prepare(cmd);
Cal Peake756184b2006-09-30 23:27:24 -0700371 if (!cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600372 printk(KERN_EMERG "Restarting system.\n");
Cal Peake756184b2006-09-30 23:27:24 -0700373 else
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600374 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
Seiji Aguchi04c68622011-01-12 16:59:30 -0800375 kmsg_dump(KMSG_DUMP_RESTART);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600376 machine_restart(cmd);
377}
378EXPORT_SYMBOL_GPL(kernel_restart);
379
Adrian Bunk4ef72292008-02-04 22:30:06 -0800380static void kernel_shutdown_prepare(enum system_states state)
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500381{
Alan Sterne041c682006-03-27 01:16:30 -0800382 blocking_notifier_call_chain(&reboot_notifier_list,
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500383 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
384 system_state = state;
Kay Sieversb50fa7c2011-05-05 13:32:05 +0200385 usermodehelper_disable();
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500386 device_shutdown();
387}
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700388/**
389 * kernel_halt - halt the system
390 *
391 * Shutdown everything and perform a clean system halt.
392 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700393void kernel_halt(void)
394{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500395 kernel_shutdown_prepare(SYSTEM_HALT);
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100396 syscore_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600397 printk(KERN_EMERG "System halted.\n");
Seiji Aguchi04c68622011-01-12 16:59:30 -0800398 kmsg_dump(KMSG_DUMP_HALT);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600399 machine_halt();
400}
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500401
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600402EXPORT_SYMBOL_GPL(kernel_halt);
403
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700404/**
405 * kernel_power_off - power_off the system
406 *
407 * Shutdown everything and perform a clean system power_off.
408 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700409void kernel_power_off(void)
410{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500411 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700412 if (pm_power_off_prepare)
413 pm_power_off_prepare();
Mark Lord40477272007-10-01 01:20:10 -0700414 disable_nonboot_cpus();
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100415 syscore_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600416 printk(KERN_EMERG "Power down.\n");
Seiji Aguchi04c68622011-01-12 16:59:30 -0800417 kmsg_dump(KMSG_DUMP_POWEROFF);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600418 machine_power_off();
419}
420EXPORT_SYMBOL_GPL(kernel_power_off);
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200421
422static DEFINE_MUTEX(reboot_mutex);
423
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424/*
425 * Reboot system call: for obvious reasons only root may call it,
426 * and even root needs to set up some magic numbers in the registers
427 * so that some mistake won't make this reboot the whole machine.
428 * You can also set the meaning of the ctrl-alt-del-key here.
429 *
430 * reboot doesn't sync: do that yourself before calling this.
431 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100432SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
433 void __user *, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434{
435 char buffer[256];
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700436 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438 /* We only trust the superuser with rebooting the system. */
439 if (!capable(CAP_SYS_BOOT))
440 return -EPERM;
441
442 /* For safety, we require "magic" arguments. */
443 if (magic1 != LINUX_REBOOT_MAGIC1 ||
444 (magic2 != LINUX_REBOOT_MAGIC2 &&
445 magic2 != LINUX_REBOOT_MAGIC2A &&
446 magic2 != LINUX_REBOOT_MAGIC2B &&
447 magic2 != LINUX_REBOOT_MAGIC2C))
448 return -EINVAL;
449
Daniel Lezcanocf3f8922012-03-28 14:42:51 -0700450 /*
451 * If pid namespaces are enabled and the current task is in a child
452 * pid_namespace, the command is handled by reboot_pid_ns() which will
453 * call do_exit().
454 */
455 ret = reboot_pid_ns(task_active_pid_ns(current), cmd);
456 if (ret)
457 return ret;
458
Eric W. Biederman5e382912006-01-08 01:03:46 -0800459 /* Instead of trying to make the power_off code look like
460 * halt when pm_power_off is not set do it the easy way.
461 */
462 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
463 cmd = LINUX_REBOOT_CMD_HALT;
464
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200465 mutex_lock(&reboot_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 switch (cmd) {
467 case LINUX_REBOOT_CMD_RESTART:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600468 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 break;
470
471 case LINUX_REBOOT_CMD_CAD_ON:
472 C_A_D = 1;
473 break;
474
475 case LINUX_REBOOT_CMD_CAD_OFF:
476 C_A_D = 0;
477 break;
478
479 case LINUX_REBOOT_CMD_HALT:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600480 kernel_halt();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 do_exit(0);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700482 panic("cannot halt");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
484 case LINUX_REBOOT_CMD_POWER_OFF:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600485 kernel_power_off();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 do_exit(0);
487 break;
488
489 case LINUX_REBOOT_CMD_RESTART2:
490 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200491 ret = -EFAULT;
492 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 }
494 buffer[sizeof(buffer) - 1] = '\0';
495
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600496 kernel_restart(buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 break;
498
Huang Ying3ab83522008-07-25 19:45:07 -0700499#ifdef CONFIG_KEXEC
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700500 case LINUX_REBOOT_CMD_KEXEC:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700501 ret = kernel_kexec();
502 break;
Huang Ying3ab83522008-07-25 19:45:07 -0700503#endif
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600504
Rafael J. Wysockib0cb1a12007-07-29 23:24:36 +0200505#ifdef CONFIG_HIBERNATION
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 case LINUX_REBOOT_CMD_SW_SUSPEND:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700507 ret = hibernate();
508 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509#endif
510
511 default:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700512 ret = -EINVAL;
513 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 }
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200515 mutex_unlock(&reboot_mutex);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700516 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517}
518
David Howells65f27f32006-11-22 14:55:48 +0000519static void deferred_cad(struct work_struct *dummy)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520{
Eric W. Biedermanabcd9e52005-07-26 11:27:34 -0600521 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522}
523
524/*
525 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
526 * As it's called within an interrupt, it may NOT sync: the only choice
527 * is whether to reboot at once, or just ignore the ctrl-alt-del.
528 */
529void ctrl_alt_del(void)
530{
David Howells65f27f32006-11-22 14:55:48 +0000531 static DECLARE_WORK(cad_work, deferred_cad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
533 if (C_A_D)
534 schedule_work(&cad_work);
535 else
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700536 kill_cad_pid(SIGINT, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537}
538
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539/*
540 * Unprivileged users may change the real gid to the effective gid
541 * or vice versa. (BSD-style)
542 *
543 * If you set the real gid at all, or set the effective gid to a value not
544 * equal to the real gid, then the saved gid is set to the new effective gid.
545 *
546 * This makes it possible for a setgid program to completely drop its
547 * privileges, which is often a useful assertion to make when you are doing
548 * a security audit over a program.
549 *
550 * The general idea is that a program which uses just setregid() will be
551 * 100% compatible with BSD. A program which uses just setgid() will be
552 * 100% compatible with POSIX with saved IDs.
553 *
554 * SMP: There are not races, the GIDs are checked only by filesystem
555 * operations (as far as semantic preservation is concerned).
556 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100557SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800559 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100560 const struct cred *old;
561 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800563 kgid_t krgid, kegid;
564
565 krgid = make_kgid(ns, rgid);
566 kegid = make_kgid(ns, egid);
567
568 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
569 return -EINVAL;
570 if ((egid != (gid_t) -1) && !gid_valid(kegid))
571 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
David Howellsd84f4f92008-11-14 10:39:23 +1100573 new = prepare_creds();
574 if (!new)
575 return -ENOMEM;
576 old = current_cred();
577
David Howellsd84f4f92008-11-14 10:39:23 +1100578 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 if (rgid != (gid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800580 if (gid_eq(old->gid, krgid) ||
581 gid_eq(old->egid, krgid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700582 nsown_capable(CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800583 new->gid = krgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 else
David Howellsd84f4f92008-11-14 10:39:23 +1100585 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 }
587 if (egid != (gid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800588 if (gid_eq(old->gid, kegid) ||
589 gid_eq(old->egid, kegid) ||
590 gid_eq(old->sgid, kegid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700591 nsown_capable(CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800592 new->egid = kegid;
Cal Peake756184b2006-09-30 23:27:24 -0700593 else
David Howellsd84f4f92008-11-14 10:39:23 +1100594 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 }
David Howellsd84f4f92008-11-14 10:39:23 +1100596
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 if (rgid != (gid_t) -1 ||
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800598 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
David Howellsd84f4f92008-11-14 10:39:23 +1100599 new->sgid = new->egid;
600 new->fsgid = new->egid;
601
602 return commit_creds(new);
603
604error:
605 abort_creds(new);
606 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607}
608
609/*
610 * setgid() is implemented like SysV w/ SAVED_IDS
611 *
612 * SMP: Same implicit races as above.
613 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100614SYSCALL_DEFINE1(setgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800616 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100617 const struct cred *old;
618 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800620 kgid_t kgid;
621
622 kgid = make_kgid(ns, gid);
623 if (!gid_valid(kgid))
624 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
David Howellsd84f4f92008-11-14 10:39:23 +1100626 new = prepare_creds();
627 if (!new)
628 return -ENOMEM;
629 old = current_cred();
630
David Howellsd84f4f92008-11-14 10:39:23 +1100631 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700632 if (nsown_capable(CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800633 new->gid = new->egid = new->sgid = new->fsgid = kgid;
634 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
635 new->egid = new->fsgid = kgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 else
David Howellsd84f4f92008-11-14 10:39:23 +1100637 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
David Howellsd84f4f92008-11-14 10:39:23 +1100639 return commit_creds(new);
640
641error:
642 abort_creds(new);
643 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644}
Dhaval Giani54e99122009-02-27 15:13:54 +0530645
David Howellsd84f4f92008-11-14 10:39:23 +1100646/*
647 * change the user struct in a credentials set to match the new UID
648 */
649static int set_user(struct cred *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650{
651 struct user_struct *new_user;
652
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800653 new_user = alloc_uid(new->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 if (!new_user)
655 return -EAGAIN;
656
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400657 /*
658 * We don't fail in case of NPROC limit excess here because too many
659 * poorly written programs don't check set*uid() return code, assuming
660 * it never fails if called by root. We may still enforce NPROC limit
661 * for programs doing set*uid()+execve() by harmlessly deferring the
662 * failure to the execve() stage.
663 */
Jiri Slaby78d7d402010-03-05 13:42:54 -0800664 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400665 new_user != INIT_USER)
666 current->flags |= PF_NPROC_EXCEEDED;
667 else
668 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
David Howellsd84f4f92008-11-14 10:39:23 +1100670 free_uid(new->user);
671 new->user = new_user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 return 0;
673}
674
675/*
676 * Unprivileged users may change the real uid to the effective uid
677 * or vice versa. (BSD-style)
678 *
679 * If you set the real uid at all, or set the effective uid to a value not
680 * equal to the real uid, then the saved uid is set to the new effective uid.
681 *
682 * This makes it possible for a setuid program to completely drop its
683 * privileges, which is often a useful assertion to make when you are doing
684 * a security audit over a program.
685 *
686 * The general idea is that a program which uses just setreuid() will be
687 * 100% compatible with BSD. A program which uses just setuid() will be
688 * 100% compatible with POSIX with saved IDs.
689 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100690SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800692 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100693 const struct cred *old;
694 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800696 kuid_t kruid, keuid;
697
698 kruid = make_kuid(ns, ruid);
699 keuid = make_kuid(ns, euid);
700
701 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
702 return -EINVAL;
703 if ((euid != (uid_t) -1) && !uid_valid(keuid))
704 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705
David Howellsd84f4f92008-11-14 10:39:23 +1100706 new = prepare_creds();
707 if (!new)
708 return -ENOMEM;
709 old = current_cred();
710
David Howellsd84f4f92008-11-14 10:39:23 +1100711 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 if (ruid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800713 new->uid = kruid;
714 if (!uid_eq(old->uid, kruid) &&
715 !uid_eq(old->euid, kruid) &&
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700716 !nsown_capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100717 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 }
719
720 if (euid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800721 new->euid = keuid;
722 if (!uid_eq(old->uid, keuid) &&
723 !uid_eq(old->euid, keuid) &&
724 !uid_eq(old->suid, keuid) &&
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700725 !nsown_capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100726 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 }
728
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800729 if (!uid_eq(new->uid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530730 retval = set_user(new);
731 if (retval < 0)
732 goto error;
733 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 if (ruid != (uid_t) -1 ||
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800735 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
David Howellsd84f4f92008-11-14 10:39:23 +1100736 new->suid = new->euid;
737 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
David Howellsd84f4f92008-11-14 10:39:23 +1100739 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
740 if (retval < 0)
741 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742
David Howellsd84f4f92008-11-14 10:39:23 +1100743 return commit_creds(new);
744
745error:
746 abort_creds(new);
747 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750/*
751 * setuid() is implemented like SysV with SAVED_IDS
752 *
753 * Note that SAVED_ID's is deficient in that a setuid root program
754 * like sendmail, for example, cannot set its uid to be a normal
755 * user and then switch back, because if you're root, setuid() sets
756 * the saved uid too. If you don't like this, blame the bright people
757 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
758 * will allow a root program to temporarily drop privileges and be able to
759 * regain them by swapping the real and effective uid.
760 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100761SYSCALL_DEFINE1(setuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800763 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100764 const struct cred *old;
765 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800767 kuid_t kuid;
768
769 kuid = make_kuid(ns, uid);
770 if (!uid_valid(kuid))
771 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
David Howellsd84f4f92008-11-14 10:39:23 +1100773 new = prepare_creds();
774 if (!new)
775 return -ENOMEM;
776 old = current_cred();
777
David Howellsd84f4f92008-11-14 10:39:23 +1100778 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700779 if (nsown_capable(CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800780 new->suid = new->uid = kuid;
781 if (!uid_eq(kuid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530782 retval = set_user(new);
783 if (retval < 0)
784 goto error;
David Howellsd84f4f92008-11-14 10:39:23 +1100785 }
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800786 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100787 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800790 new->fsuid = new->euid = kuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
David Howellsd84f4f92008-11-14 10:39:23 +1100792 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
793 if (retval < 0)
794 goto error;
795
796 return commit_creds(new);
797
798error:
799 abort_creds(new);
800 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801}
802
803
804/*
805 * This function implements a generic ability to update ruid, euid,
806 * and suid. This allows you to implement the 4.4 compatible seteuid().
807 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100808SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800810 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100811 const struct cred *old;
812 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800814 kuid_t kruid, keuid, ksuid;
815
816 kruid = make_kuid(ns, ruid);
817 keuid = make_kuid(ns, euid);
818 ksuid = make_kuid(ns, suid);
819
820 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
821 return -EINVAL;
822
823 if ((euid != (uid_t) -1) && !uid_valid(keuid))
824 return -EINVAL;
825
826 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
827 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
David Howellsd84f4f92008-11-14 10:39:23 +1100829 new = prepare_creds();
830 if (!new)
831 return -ENOMEM;
832
David Howellsd84f4f92008-11-14 10:39:23 +1100833 old = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
David Howellsd84f4f92008-11-14 10:39:23 +1100835 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700836 if (!nsown_capable(CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800837 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
838 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100839 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800840 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
841 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100842 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800843 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
844 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100845 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 }
David Howellsd84f4f92008-11-14 10:39:23 +1100847
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 if (ruid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800849 new->uid = kruid;
850 if (!uid_eq(kruid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530851 retval = set_user(new);
852 if (retval < 0)
853 goto error;
854 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 }
David Howellsd84f4f92008-11-14 10:39:23 +1100856 if (euid != (uid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800857 new->euid = keuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 if (suid != (uid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800859 new->suid = ksuid;
David Howellsd84f4f92008-11-14 10:39:23 +1100860 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861
David Howellsd84f4f92008-11-14 10:39:23 +1100862 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
863 if (retval < 0)
864 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
David Howellsd84f4f92008-11-14 10:39:23 +1100866 return commit_creds(new);
867
868error:
869 abort_creds(new);
870 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871}
872
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800873SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874{
David Howells86a264a2008-11-14 10:39:18 +1100875 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800877 uid_t ruid, euid, suid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800879 ruid = from_kuid_munged(cred->user_ns, cred->uid);
880 euid = from_kuid_munged(cred->user_ns, cred->euid);
881 suid = from_kuid_munged(cred->user_ns, cred->suid);
882
883 if (!(retval = put_user(ruid, ruidp)) &&
884 !(retval = put_user(euid, euidp)))
885 retval = put_user(suid, suidp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886
887 return retval;
888}
889
890/*
891 * Same as above, but for rgid, egid, sgid.
892 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100893SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800895 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100896 const struct cred *old;
897 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800899 kgid_t krgid, kegid, ksgid;
900
901 krgid = make_kgid(ns, rgid);
902 kegid = make_kgid(ns, egid);
903 ksgid = make_kgid(ns, sgid);
904
905 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
906 return -EINVAL;
907 if ((egid != (gid_t) -1) && !gid_valid(kegid))
908 return -EINVAL;
909 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
910 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911
David Howellsd84f4f92008-11-14 10:39:23 +1100912 new = prepare_creds();
913 if (!new)
914 return -ENOMEM;
915 old = current_cred();
916
David Howellsd84f4f92008-11-14 10:39:23 +1100917 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700918 if (!nsown_capable(CAP_SETGID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800919 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
920 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100921 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800922 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
923 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100924 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800925 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
926 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100927 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929
David Howellsd84f4f92008-11-14 10:39:23 +1100930 if (rgid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800931 new->gid = krgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100932 if (egid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800933 new->egid = kegid;
David Howellsd84f4f92008-11-14 10:39:23 +1100934 if (sgid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800935 new->sgid = ksgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100936 new->fsgid = new->egid;
937
938 return commit_creds(new);
939
940error:
941 abort_creds(new);
942 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943}
944
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800945SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946{
David Howells86a264a2008-11-14 10:39:18 +1100947 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800949 gid_t rgid, egid, sgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800951 rgid = from_kgid_munged(cred->user_ns, cred->gid);
952 egid = from_kgid_munged(cred->user_ns, cred->egid);
953 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
954
955 if (!(retval = put_user(rgid, rgidp)) &&
956 !(retval = put_user(egid, egidp)))
957 retval = put_user(sgid, sgidp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958
959 return retval;
960}
961
962
963/*
964 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
965 * is used for "access()" and for the NFS daemon (letting nfsd stay at
966 * whatever uid it wants to). It normally shadows "euid", except when
967 * explicitly set by setfsuid() or for access..
968 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100969SYSCALL_DEFINE1(setfsuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970{
David Howellsd84f4f92008-11-14 10:39:23 +1100971 const struct cred *old;
972 struct cred *new;
973 uid_t old_fsuid;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800974 kuid_t kuid;
975
976 old = current_cred();
977 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
978
979 kuid = make_kuid(old->user_ns, uid);
980 if (!uid_valid(kuid))
981 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
David Howellsd84f4f92008-11-14 10:39:23 +1100983 new = prepare_creds();
984 if (!new)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800985 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800987 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
988 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700989 nsown_capable(CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800990 if (!uid_eq(kuid, old->fsuid)) {
991 new->fsuid = kuid;
David Howellsd84f4f92008-11-14 10:39:23 +1100992 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
993 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 }
996
David Howellsd84f4f92008-11-14 10:39:23 +1100997 abort_creds(new);
998 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
David Howellsd84f4f92008-11-14 10:39:23 +11001000change_okay:
1001 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 return old_fsuid;
1003}
1004
1005/*
John Anthony Kazos Jrf42df9e2007-05-09 08:23:08 +02001006 * Samma på svenska..
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 */
Heiko Carstensae1251a2009-01-14 14:14:05 +01001008SYSCALL_DEFINE1(setfsgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009{
David Howellsd84f4f92008-11-14 10:39:23 +11001010 const struct cred *old;
1011 struct cred *new;
1012 gid_t old_fsgid;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001013 kgid_t kgid;
1014
1015 old = current_cred();
1016 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1017
1018 kgid = make_kgid(old->user_ns, gid);
1019 if (!gid_valid(kgid))
1020 return old_fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
David Howellsd84f4f92008-11-14 10:39:23 +11001022 new = prepare_creds();
1023 if (!new)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001024 return old_fsgid;
David Howellsd84f4f92008-11-14 10:39:23 +11001025
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001026 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1027 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001028 nsown_capable(CAP_SETGID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001029 if (!gid_eq(kgid, old->fsgid)) {
1030 new->fsgid = kgid;
David Howellsd84f4f92008-11-14 10:39:23 +11001031 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 }
David Howellsd84f4f92008-11-14 10:39:23 +11001034
David Howellsd84f4f92008-11-14 10:39:23 +11001035 abort_creds(new);
1036 return old_fsgid;
1037
1038change_okay:
1039 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 return old_fsgid;
1041}
1042
Frank Mayharf06febc2008-09-12 09:54:39 -07001043void do_sys_times(struct tms *tms)
1044{
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001045 cputime_t tgutime, tgstime, cutime, cstime;
Frank Mayharf06febc2008-09-12 09:54:39 -07001046
Oleg Nesterov2b5fe6d2008-11-17 15:40:08 +01001047 spin_lock_irq(&current->sighand->siglock);
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001048 thread_group_times(current, &tgutime, &tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001049 cutime = current->signal->cutime;
1050 cstime = current->signal->cstime;
1051 spin_unlock_irq(&current->sighand->siglock);
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001052 tms->tms_utime = cputime_to_clock_t(tgutime);
1053 tms->tms_stime = cputime_to_clock_t(tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001054 tms->tms_cutime = cputime_to_clock_t(cutime);
1055 tms->tms_cstime = cputime_to_clock_t(cstime);
1056}
1057
Heiko Carstens58fd3aa2009-01-14 14:14:03 +01001058SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 if (tbuf) {
1061 struct tms tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
Frank Mayharf06febc2008-09-12 09:54:39 -07001063 do_sys_times(&tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1065 return -EFAULT;
1066 }
Paul Mackerrase3d5a272009-01-06 14:41:02 -08001067 force_successful_syscall_return();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1069}
1070
1071/*
1072 * This needs some heavy checking ...
1073 * I just haven't the stomach for it. I also don't fully
1074 * understand sessions/pgrp etc. Let somebody who does explain it.
1075 *
1076 * OK, I think I have the protection semantics right.... this is really
1077 * only important on a multi-user system anyway, to make sure one user
1078 * can't send a signal to a process owned by another. -TYT, 12/12/91
1079 *
1080 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1081 * LBT 04.03.94
1082 */
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001083SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
1085 struct task_struct *p;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001086 struct task_struct *group_leader = current->group_leader;
Oleg Nesterov4e021302008-02-08 04:19:08 -08001087 struct pid *pgrp;
1088 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
1090 if (!pid)
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001091 pid = task_pid_vnr(group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 if (!pgid)
1093 pgid = pid;
1094 if (pgid < 0)
1095 return -EINVAL;
Paul E. McKenney950eaac2010-08-31 17:00:18 -07001096 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
1098 /* From this point forward we keep holding onto the tasklist lock
1099 * so that our parent does not change from under us. -DaveM
1100 */
1101 write_lock_irq(&tasklist_lock);
1102
1103 err = -ESRCH;
Oleg Nesterov4e021302008-02-08 04:19:08 -08001104 p = find_task_by_vpid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 if (!p)
1106 goto out;
1107
1108 err = -EINVAL;
1109 if (!thread_group_leader(p))
1110 goto out;
1111
Oleg Nesterov4e021302008-02-08 04:19:08 -08001112 if (same_thread_group(p->real_parent, group_leader)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 err = -EPERM;
Eric W. Biederman41487c62007-02-12 00:53:01 -08001114 if (task_session(p) != task_session(group_leader))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 goto out;
1116 err = -EACCES;
1117 if (p->did_exec)
1118 goto out;
1119 } else {
1120 err = -ESRCH;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001121 if (p != group_leader)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 goto out;
1123 }
1124
1125 err = -EPERM;
1126 if (p->signal->leader)
1127 goto out;
1128
Oleg Nesterov4e021302008-02-08 04:19:08 -08001129 pgrp = task_pid(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 if (pgid != pid) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001131 struct task_struct *g;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132
Oleg Nesterov4e021302008-02-08 04:19:08 -08001133 pgrp = find_vpid(pgid);
1134 g = pid_task(pgrp, PIDTYPE_PGID);
Eric W. Biederman41487c62007-02-12 00:53:01 -08001135 if (!g || task_session(g) != task_session(group_leader))
Oleg Nesterovf020bc42006-12-08 02:38:02 -08001136 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 }
1138
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 err = security_task_setpgid(p, pgid);
1140 if (err)
1141 goto out;
1142
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -07001143 if (task_pgrp(p) != pgrp)
Oleg Nesterov83beaf32008-04-30 00:54:27 -07001144 change_pid(p, PIDTYPE_PGID, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
1146 err = 0;
1147out:
1148 /* All paths lead to here, thus we are safe. -DaveM */
1149 write_unlock_irq(&tasklist_lock);
Paul E. McKenney950eaac2010-08-31 17:00:18 -07001150 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 return err;
1152}
1153
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001154SYSCALL_DEFINE1(getpgid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001156 struct task_struct *p;
1157 struct pid *grp;
1158 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001160 rcu_read_lock();
1161 if (!pid)
1162 grp = task_pgrp(current);
1163 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 retval = -ESRCH;
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001165 p = find_task_by_vpid(pid);
1166 if (!p)
1167 goto out;
1168 grp = task_pgrp(p);
1169 if (!grp)
1170 goto out;
1171
1172 retval = security_task_getpgid(p);
1173 if (retval)
1174 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 }
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001176 retval = pid_vnr(grp);
1177out:
1178 rcu_read_unlock();
1179 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180}
1181
1182#ifdef __ARCH_WANT_SYS_GETPGRP
1183
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001184SYSCALL_DEFINE0(getpgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001186 return sys_getpgid(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187}
1188
1189#endif
1190
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001191SYSCALL_DEFINE1(getsid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192{
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001193 struct task_struct *p;
1194 struct pid *sid;
1195 int retval;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001196
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001197 rcu_read_lock();
1198 if (!pid)
1199 sid = task_session(current);
1200 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 retval = -ESRCH;
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001202 p = find_task_by_vpid(pid);
1203 if (!p)
1204 goto out;
1205 sid = task_session(p);
1206 if (!sid)
1207 goto out;
1208
1209 retval = security_task_getsid(p);
1210 if (retval)
1211 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 }
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001213 retval = pid_vnr(sid);
1214out:
1215 rcu_read_unlock();
1216 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217}
1218
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001219SYSCALL_DEFINE0(setsid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220{
Oren Laadane19f2472006-01-08 01:03:58 -08001221 struct task_struct *group_leader = current->group_leader;
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001222 struct pid *sid = task_pid(group_leader);
1223 pid_t session = pid_vnr(sid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 int err = -EPERM;
1225
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 write_lock_irq(&tasklist_lock);
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001227 /* Fail if I am already a session leader */
1228 if (group_leader->signal->leader)
1229 goto out;
1230
Oleg Nesterov430c6232008-02-08 04:19:11 -08001231 /* Fail if a process group id already exists that equals the
1232 * proposed session id.
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001233 */
Oleg Nesterov6806aac2008-02-08 04:19:12 -08001234 if (pid_task(sid, PIDTYPE_PGID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 goto out;
1236
Oren Laadane19f2472006-01-08 01:03:58 -08001237 group_leader->signal->leader = 1;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -08001238 __set_special_pids(sid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001239
Alan Cox9c9f4de2008-10-13 10:37:26 +01001240 proc_clear_tty(group_leader);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001241
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001242 err = session;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243out:
1244 write_unlock_irq(&tasklist_lock);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001245 if (err > 0) {
Christian Borntraeger0d0df592009-10-26 16:49:34 -07001246 proc_sid_connector(group_leader);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001247 sched_autogroup_create_attach(group_leader);
1248 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 return err;
1250}
1251
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252DECLARE_RWSEM(uts_sem);
1253
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001254#ifdef COMPAT_UTS_MACHINE
1255#define override_architecture(name) \
Andreas Schwab46da2762010-04-23 13:17:44 -04001256 (personality(current->personality) == PER_LINUX32 && \
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001257 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1258 sizeof(COMPAT_UTS_MACHINE)))
1259#else
1260#define override_architecture(name) 0
1261#endif
1262
Andi Kleenbe274252011-08-19 16:15:10 -07001263/*
1264 * Work around broken programs that cannot handle "Linux 3.0".
1265 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1266 */
1267static int override_release(char __user *release, int len)
1268{
1269 int ret = 0;
Linus Torvaldsa84a79e2011-10-17 08:24:24 -07001270 char buf[65];
Andi Kleenbe274252011-08-19 16:15:10 -07001271
1272 if (current->personality & UNAME26) {
1273 char *rest = UTS_RELEASE;
1274 int ndots = 0;
1275 unsigned v;
1276
1277 while (*rest) {
1278 if (*rest == '.' && ++ndots >= 3)
1279 break;
1280 if (!isdigit(*rest) && *rest != '.')
1281 break;
1282 rest++;
1283 }
1284 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1285 snprintf(buf, len, "2.6.%u%s", v, rest);
1286 ret = copy_to_user(release, buf, len);
1287 }
1288 return ret;
1289}
1290
Heiko Carstense48fbb62009-01-14 14:14:26 +01001291SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292{
1293 int errno = 0;
1294
1295 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001296 if (copy_to_user(name, utsname(), sizeof *name))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 errno = -EFAULT;
1298 up_read(&uts_sem);
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001299
Andi Kleenbe274252011-08-19 16:15:10 -07001300 if (!errno && override_release(name->release, sizeof(name->release)))
1301 errno = -EFAULT;
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001302 if (!errno && override_architecture(name))
1303 errno = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 return errno;
1305}
1306
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001307#ifdef __ARCH_WANT_SYS_OLD_UNAME
1308/*
1309 * Old cruft
1310 */
1311SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1312{
1313 int error = 0;
1314
1315 if (!name)
1316 return -EFAULT;
1317
1318 down_read(&uts_sem);
1319 if (copy_to_user(name, utsname(), sizeof(*name)))
1320 error = -EFAULT;
1321 up_read(&uts_sem);
1322
Andi Kleenbe274252011-08-19 16:15:10 -07001323 if (!error && override_release(name->release, sizeof(name->release)))
1324 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001325 if (!error && override_architecture(name))
1326 error = -EFAULT;
1327 return error;
1328}
1329
1330SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1331{
1332 int error;
1333
1334 if (!name)
1335 return -EFAULT;
1336 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1337 return -EFAULT;
1338
1339 down_read(&uts_sem);
1340 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1341 __OLD_UTS_LEN);
1342 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1343 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1344 __OLD_UTS_LEN);
1345 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1346 error |= __copy_to_user(&name->release, &utsname()->release,
1347 __OLD_UTS_LEN);
1348 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1349 error |= __copy_to_user(&name->version, &utsname()->version,
1350 __OLD_UTS_LEN);
1351 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1352 error |= __copy_to_user(&name->machine, &utsname()->machine,
1353 __OLD_UTS_LEN);
1354 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1355 up_read(&uts_sem);
1356
1357 if (!error && override_architecture(name))
1358 error = -EFAULT;
Andi Kleenbe274252011-08-19 16:15:10 -07001359 if (!error && override_release(name->release, sizeof(name->release)))
1360 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001361 return error ? -EFAULT : 0;
1362}
1363#endif
1364
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001365SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366{
1367 int errno;
1368 char tmp[__NEW_UTS_LEN];
1369
Serge E. Hallynbb96a6f2011-03-23 16:43:18 -07001370 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 return -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001372
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 if (len < 0 || len > __NEW_UTS_LEN)
1374 return -EINVAL;
1375 down_write(&uts_sem);
1376 errno = -EFAULT;
1377 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001378 struct new_utsname *u = utsname();
1379
1380 memcpy(u->nodename, tmp, len);
1381 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 errno = 0;
Sasikantha babu499eea62012-05-31 16:26:07 -07001383 uts_proc_notify(UTS_PROC_HOSTNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 }
1385 up_write(&uts_sem);
1386 return errno;
1387}
1388
1389#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1390
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001391SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392{
1393 int i, errno;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001394 struct new_utsname *u;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395
1396 if (len < 0)
1397 return -EINVAL;
1398 down_read(&uts_sem);
Andrew Morton9679e4d2008-10-15 22:01:51 -07001399 u = utsname();
1400 i = 1 + strlen(u->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 if (i > len)
1402 i = len;
1403 errno = 0;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001404 if (copy_to_user(name, u->nodename, i))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 errno = -EFAULT;
1406 up_read(&uts_sem);
1407 return errno;
1408}
1409
1410#endif
1411
1412/*
1413 * Only setdomainname; getdomainname can be implemented by calling
1414 * uname()
1415 */
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001416SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
1418 int errno;
1419 char tmp[__NEW_UTS_LEN];
1420
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001421 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 return -EPERM;
1423 if (len < 0 || len > __NEW_UTS_LEN)
1424 return -EINVAL;
1425
1426 down_write(&uts_sem);
1427 errno = -EFAULT;
1428 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001429 struct new_utsname *u = utsname();
1430
1431 memcpy(u->domainname, tmp, len);
1432 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 errno = 0;
Sasikantha babu499eea62012-05-31 16:26:07 -07001434 uts_proc_notify(UTS_PROC_DOMAINNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 }
1436 up_write(&uts_sem);
1437 return errno;
1438}
1439
Heiko Carstense48fbb62009-01-14 14:14:26 +01001440SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441{
Jiri Slabyb9518342010-05-04 11:28:25 +02001442 struct rlimit value;
1443 int ret;
1444
1445 ret = do_prlimit(current, resource, NULL, &value);
1446 if (!ret)
1447 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1448
1449 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450}
1451
1452#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1453
1454/*
1455 * Back compatibility for getrlimit. Needed for some apps.
1456 */
1457
Heiko Carstense48fbb62009-01-14 14:14:26 +01001458SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1459 struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460{
1461 struct rlimit x;
1462 if (resource >= RLIM_NLIMITS)
1463 return -EINVAL;
1464
1465 task_lock(current->group_leader);
1466 x = current->signal->rlim[resource];
1467 task_unlock(current->group_leader);
Cal Peake756184b2006-09-30 23:27:24 -07001468 if (x.rlim_cur > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 x.rlim_cur = 0x7FFFFFFF;
Cal Peake756184b2006-09-30 23:27:24 -07001470 if (x.rlim_max > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 x.rlim_max = 0x7FFFFFFF;
1472 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1473}
1474
1475#endif
1476
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001477static inline bool rlim64_is_infinity(__u64 rlim64)
1478{
1479#if BITS_PER_LONG < 64
1480 return rlim64 >= ULONG_MAX;
1481#else
1482 return rlim64 == RLIM64_INFINITY;
1483#endif
1484}
1485
1486static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1487{
1488 if (rlim->rlim_cur == RLIM_INFINITY)
1489 rlim64->rlim_cur = RLIM64_INFINITY;
1490 else
1491 rlim64->rlim_cur = rlim->rlim_cur;
1492 if (rlim->rlim_max == RLIM_INFINITY)
1493 rlim64->rlim_max = RLIM64_INFINITY;
1494 else
1495 rlim64->rlim_max = rlim->rlim_max;
1496}
1497
1498static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1499{
1500 if (rlim64_is_infinity(rlim64->rlim_cur))
1501 rlim->rlim_cur = RLIM_INFINITY;
1502 else
1503 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1504 if (rlim64_is_infinity(rlim64->rlim_max))
1505 rlim->rlim_max = RLIM_INFINITY;
1506 else
1507 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1508}
1509
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001510/* make sure you are allowed to change @tsk limits before calling this */
Jiri Slaby5b415352010-03-24 16:11:29 +01001511int do_prlimit(struct task_struct *tsk, unsigned int resource,
1512 struct rlimit *new_rlim, struct rlimit *old_rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513{
Jiri Slaby5b415352010-03-24 16:11:29 +01001514 struct rlimit *rlim;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001515 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516
1517 if (resource >= RLIM_NLIMITS)
1518 return -EINVAL;
Jiri Slaby5b415352010-03-24 16:11:29 +01001519 if (new_rlim) {
1520 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1521 return -EINVAL;
1522 if (resource == RLIMIT_NOFILE &&
1523 new_rlim->rlim_max > sysctl_nr_open)
1524 return -EPERM;
1525 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001527 /* protect tsk->signal and tsk->sighand from disappearing */
1528 read_lock(&tasklist_lock);
1529 if (!tsk->sighand) {
1530 retval = -ESRCH;
1531 goto out;
1532 }
1533
Jiri Slaby5b415352010-03-24 16:11:29 +01001534 rlim = tsk->signal->rlim + resource;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001535 task_lock(tsk->group_leader);
Jiri Slaby5b415352010-03-24 16:11:29 +01001536 if (new_rlim) {
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001537 /* Keep the capable check against init_user_ns until
1538 cgroups can contain all limits */
Jiri Slaby5b415352010-03-24 16:11:29 +01001539 if (new_rlim->rlim_max > rlim->rlim_max &&
1540 !capable(CAP_SYS_RESOURCE))
1541 retval = -EPERM;
1542 if (!retval)
1543 retval = security_task_setrlimit(tsk->group_leader,
1544 resource, new_rlim);
1545 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1546 /*
1547 * The caller is asking for an immediate RLIMIT_CPU
1548 * expiry. But we use the zero value to mean "it was
1549 * never set". So let's cheat and make it one second
1550 * instead
1551 */
1552 new_rlim->rlim_cur = 1;
1553 }
Tom Alsberg9926e4c2007-05-08 00:30:31 -07001554 }
Jiri Slaby5b415352010-03-24 16:11:29 +01001555 if (!retval) {
1556 if (old_rlim)
1557 *old_rlim = *rlim;
1558 if (new_rlim)
1559 *rlim = *new_rlim;
1560 }
Jiri Slaby7855c352009-08-26 23:45:34 +02001561 task_unlock(tsk->group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Andrew Mortond3561f72006-03-24 03:18:36 -08001563 /*
1564 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1565 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1566 * very long-standing error, and fixing it now risks breakage of
1567 * applications, so we live with it
1568 */
Jiri Slaby5b415352010-03-24 16:11:29 +01001569 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1570 new_rlim->rlim_cur != RLIM_INFINITY)
1571 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001572out:
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001573 read_unlock(&tasklist_lock);
Oleg Nesterov2fb9d262009-09-03 19:21:45 +02001574 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575}
1576
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001577/* rcu lock must be held */
1578static int check_prlimit_permission(struct task_struct *task)
1579{
1580 const struct cred *cred = current_cred(), *tcred;
1581
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001582 if (current == task)
1583 return 0;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001584
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001585 tcred = __task_cred(task);
Eric W. Biederman5af66202012-03-03 20:21:47 -08001586 if (uid_eq(cred->uid, tcred->euid) &&
1587 uid_eq(cred->uid, tcred->suid) &&
1588 uid_eq(cred->uid, tcred->uid) &&
1589 gid_eq(cred->gid, tcred->egid) &&
1590 gid_eq(cred->gid, tcred->sgid) &&
1591 gid_eq(cred->gid, tcred->gid))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001592 return 0;
Eric W. Biedermanc4a4d602011-11-16 23:15:31 -08001593 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001594 return 0;
1595
1596 return -EPERM;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001597}
1598
1599SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1600 const struct rlimit64 __user *, new_rlim,
1601 struct rlimit64 __user *, old_rlim)
1602{
1603 struct rlimit64 old64, new64;
1604 struct rlimit old, new;
1605 struct task_struct *tsk;
1606 int ret;
1607
1608 if (new_rlim) {
1609 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1610 return -EFAULT;
1611 rlim64_to_rlim(&new64, &new);
1612 }
1613
1614 rcu_read_lock();
1615 tsk = pid ? find_task_by_vpid(pid) : current;
1616 if (!tsk) {
1617 rcu_read_unlock();
1618 return -ESRCH;
1619 }
1620 ret = check_prlimit_permission(tsk);
1621 if (ret) {
1622 rcu_read_unlock();
1623 return ret;
1624 }
1625 get_task_struct(tsk);
1626 rcu_read_unlock();
1627
1628 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1629 old_rlim ? &old : NULL);
1630
1631 if (!ret && old_rlim) {
1632 rlim_to_rlim64(&old, &old64);
1633 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1634 ret = -EFAULT;
1635 }
1636
1637 put_task_struct(tsk);
1638 return ret;
1639}
1640
Jiri Slaby7855c352009-08-26 23:45:34 +02001641SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1642{
1643 struct rlimit new_rlim;
1644
1645 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1646 return -EFAULT;
Jiri Slaby5b415352010-03-24 16:11:29 +01001647 return do_prlimit(current, resource, &new_rlim, NULL);
Jiri Slaby7855c352009-08-26 23:45:34 +02001648}
1649
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650/*
1651 * It would make sense to put struct rusage in the task_struct,
1652 * except that would make the task_struct be *really big*. After
1653 * task_struct gets moved into malloc'ed memory, it would
1654 * make sense to do this. It will make moving the rest of the information
1655 * a lot simpler! (Which we're not doing right now because we're not
1656 * measuring them yet).
1657 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1659 * races with threads incrementing their own counters. But since word
1660 * reads are atomic, we either get new values or old values and we don't
1661 * care which for the sums. We always take the siglock to protect reading
1662 * the c* fields from p->signal from races with exit.c updating those
1663 * fields when reaping, so a sample either gets all the additions of a
1664 * given child after it's reaped, or none so this sample is before reaping.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001665 *
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001666 * Locking:
1667 * We need to take the siglock for CHILDEREN, SELF and BOTH
1668 * for the cases current multithreaded, non-current single threaded
1669 * non-current multithreaded. Thread traversal is now safe with
1670 * the siglock held.
1671 * Strictly speaking, we donot need to take the siglock if we are current and
1672 * single threaded, as no one else can take our signal_struct away, no one
1673 * else can reap the children to update signal->c* counters, and no one else
1674 * can race with the signal-> fields. If we do not take any lock, the
1675 * signal-> fields could be read out of order while another thread was just
1676 * exiting. So we should place a read memory barrier when we avoid the lock.
1677 * On the writer side, write memory barrier is implied in __exit_signal
1678 * as __exit_signal releases the siglock spinlock after updating the signal->
1679 * fields. But we don't do this yet to keep things simple.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001680 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 */
1682
Frank Mayharf06febc2008-09-12 09:54:39 -07001683static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001684{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001685 r->ru_nvcsw += t->nvcsw;
1686 r->ru_nivcsw += t->nivcsw;
1687 r->ru_minflt += t->min_flt;
1688 r->ru_majflt += t->maj_flt;
1689 r->ru_inblock += task_io_get_inblock(t);
1690 r->ru_oublock += task_io_get_oublock(t);
1691}
1692
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1694{
1695 struct task_struct *t;
1696 unsigned long flags;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001697 cputime_t tgutime, tgstime, utime, stime;
Jiri Pirko1f102062009-09-22 16:44:10 -07001698 unsigned long maxrss = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
1700 memset((char *) r, 0, sizeof *r);
Martin Schwidefsky64861632011-12-15 14:56:09 +01001701 utime = stime = 0;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001702
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001703 if (who == RUSAGE_THREAD) {
Hidetoshi Setod180c5b2009-11-26 14:48:30 +09001704 task_times(current, &utime, &stime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001705 accumulate_thread_rusage(p, r);
Jiri Pirko1f102062009-09-22 16:44:10 -07001706 maxrss = p->signal->maxrss;
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001707 goto out;
1708 }
1709
Oleg Nesterovd6cf7232008-04-30 00:52:38 -07001710 if (!lock_task_sighand(p, &flags))
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001711 return;
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001712
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 switch (who) {
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001714 case RUSAGE_BOTH:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 case RUSAGE_CHILDREN:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 utime = p->signal->cutime;
1717 stime = p->signal->cstime;
1718 r->ru_nvcsw = p->signal->cnvcsw;
1719 r->ru_nivcsw = p->signal->cnivcsw;
1720 r->ru_minflt = p->signal->cmin_flt;
1721 r->ru_majflt = p->signal->cmaj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001722 r->ru_inblock = p->signal->cinblock;
1723 r->ru_oublock = p->signal->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001724 maxrss = p->signal->cmaxrss;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001725
1726 if (who == RUSAGE_CHILDREN)
1727 break;
1728
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 case RUSAGE_SELF:
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001730 thread_group_times(p, &tgutime, &tgstime);
Martin Schwidefsky64861632011-12-15 14:56:09 +01001731 utime += tgutime;
1732 stime += tgstime;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 r->ru_nvcsw += p->signal->nvcsw;
1734 r->ru_nivcsw += p->signal->nivcsw;
1735 r->ru_minflt += p->signal->min_flt;
1736 r->ru_majflt += p->signal->maj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001737 r->ru_inblock += p->signal->inblock;
1738 r->ru_oublock += p->signal->oublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001739 if (maxrss < p->signal->maxrss)
1740 maxrss = p->signal->maxrss;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 t = p;
1742 do {
Frank Mayharf06febc2008-09-12 09:54:39 -07001743 accumulate_thread_rusage(t, r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 t = next_thread(t);
1745 } while (t != p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 break;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001747
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 default:
1749 BUG();
1750 }
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001751 unlock_task_sighand(p, &flags);
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001752
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001753out:
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001754 cputime_to_timeval(utime, &r->ru_utime);
1755 cputime_to_timeval(stime, &r->ru_stime);
Jiri Pirko1f102062009-09-22 16:44:10 -07001756
1757 if (who != RUSAGE_CHILDREN) {
1758 struct mm_struct *mm = get_task_mm(p);
1759 if (mm) {
1760 setmax_mm_hiwater_rss(&maxrss, mm);
1761 mmput(mm);
1762 }
1763 }
1764 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765}
1766
1767int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1768{
1769 struct rusage r;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 k_getrusage(p, who, &r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1772}
1773
Heiko Carstense48fbb62009-01-14 14:14:26 +01001774SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001776 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1777 who != RUSAGE_THREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 return -EINVAL;
1779 return getrusage(current, who, ru);
1780}
1781
Heiko Carstense48fbb62009-01-14 14:14:26 +01001782SYSCALL_DEFINE1(umask, int, mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783{
1784 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1785 return mask;
1786}
Serge E. Hallyn3b7391d2008-02-04 22:29:45 -08001787
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001788#ifdef CONFIG_CHECKPOINT_RESTORE
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001789static bool vma_flags_mismatch(struct vm_area_struct *vma,
1790 unsigned long required,
1791 unsigned long banned)
1792{
1793 return (vma->vm_flags & required) != required ||
1794 (vma->vm_flags & banned);
1795}
1796
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001797static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1798{
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001799 struct vm_area_struct *vma;
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001800 struct file *exe_file;
1801 struct dentry *dentry;
1802 int err;
1803
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001804 exe_file = fget(fd);
1805 if (!exe_file)
1806 return -EBADF;
1807
1808 dentry = exe_file->f_path.dentry;
1809
1810 /*
1811 * Because the original mm->exe_file points to executable file, make
1812 * sure that this one is executable as well, to avoid breaking an
1813 * overall picture.
1814 */
1815 err = -EACCES;
1816 if (!S_ISREG(dentry->d_inode->i_mode) ||
1817 exe_file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1818 goto exit;
1819
1820 err = inode_permission(dentry->d_inode, MAY_EXEC);
1821 if (err)
1822 goto exit;
1823
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001824 down_write(&mm->mmap_sem);
1825
1826 /*
1827 * Forbid mm->exe_file change if there are mapped other files.
1828 */
1829 err = -EBUSY;
1830 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1831 if (vma->vm_file && !path_equal(&vma->vm_file->f_path,
1832 &exe_file->f_path))
1833 goto exit_unlock;
1834 }
1835
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001836 /*
1837 * The symlink can be changed only once, just to disallow arbitrary
1838 * transitions malicious software might bring in. This means one
1839 * could make a snapshot over all processes running and monitor
1840 * /proc/pid/exe changes to notice unusual activity if needed.
1841 */
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001842 err = -EPERM;
1843 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1844 goto exit_unlock;
1845
1846 set_mm_exe_file(mm, exe_file);
1847exit_unlock:
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001848 up_write(&mm->mmap_sem);
1849
1850exit:
1851 fput(exe_file);
1852 return err;
1853}
1854
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001855static int prctl_set_mm(int opt, unsigned long addr,
1856 unsigned long arg4, unsigned long arg5)
1857{
1858 unsigned long rlim = rlimit(RLIMIT_DATA);
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001859 struct mm_struct *mm = current->mm;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001860 struct vm_area_struct *vma;
1861 int error;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001862
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001863 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001864 return -EINVAL;
1865
Cyrill Gorcunov79f07132012-03-15 15:17:10 -07001866 if (!capable(CAP_SYS_RESOURCE))
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001867 return -EPERM;
1868
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001869 if (opt == PR_SET_MM_EXE_FILE)
1870 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1871
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001872 if (addr >= TASK_SIZE)
1873 return -EINVAL;
1874
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001875 error = -EINVAL;
1876
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001877 down_read(&mm->mmap_sem);
1878 vma = find_vma(mm, addr);
1879
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001880 switch (opt) {
1881 case PR_SET_MM_START_CODE:
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001882 mm->start_code = addr;
1883 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001884 case PR_SET_MM_END_CODE:
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001885 mm->end_code = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001886 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001887 case PR_SET_MM_START_DATA:
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001888 mm->start_data = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001889 break;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001890 case PR_SET_MM_END_DATA:
1891 mm->end_data = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001892 break;
1893
1894 case PR_SET_MM_START_BRK:
1895 if (addr <= mm->end_data)
1896 goto out;
1897
1898 if (rlim < RLIM_INFINITY &&
1899 (mm->brk - addr) +
1900 (mm->end_data - mm->start_data) > rlim)
1901 goto out;
1902
1903 mm->start_brk = addr;
1904 break;
1905
1906 case PR_SET_MM_BRK:
1907 if (addr <= mm->end_data)
1908 goto out;
1909
1910 if (rlim < RLIM_INFINITY &&
1911 (addr - mm->start_brk) +
1912 (mm->end_data - mm->start_data) > rlim)
1913 goto out;
1914
1915 mm->brk = addr;
1916 break;
1917
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001918 /*
1919 * If command line arguments and environment
1920 * are placed somewhere else on stack, we can
1921 * set them up here, ARG_START/END to setup
1922 * command line argumets and ENV_START/END
1923 * for environment.
1924 */
1925 case PR_SET_MM_START_STACK:
1926 case PR_SET_MM_ARG_START:
1927 case PR_SET_MM_ARG_END:
1928 case PR_SET_MM_ENV_START:
1929 case PR_SET_MM_ENV_END:
1930 if (!vma) {
1931 error = -EFAULT;
1932 goto out;
1933 }
1934#ifdef CONFIG_STACK_GROWSUP
1935 if (vma_flags_mismatch(vma, VM_READ | VM_WRITE | VM_GROWSUP, 0))
1936#else
1937 if (vma_flags_mismatch(vma, VM_READ | VM_WRITE | VM_GROWSDOWN, 0))
1938#endif
1939 goto out;
1940 if (opt == PR_SET_MM_START_STACK)
1941 mm->start_stack = addr;
1942 else if (opt == PR_SET_MM_ARG_START)
1943 mm->arg_start = addr;
1944 else if (opt == PR_SET_MM_ARG_END)
1945 mm->arg_end = addr;
1946 else if (opt == PR_SET_MM_ENV_START)
1947 mm->env_start = addr;
1948 else if (opt == PR_SET_MM_ENV_END)
1949 mm->env_end = addr;
1950 break;
1951
1952 /*
1953 * This doesn't move auxiliary vector itself
1954 * since it's pinned to mm_struct, but allow
1955 * to fill vector with new values. It's up
1956 * to a caller to provide sane values here
1957 * otherwise user space tools which use this
1958 * vector might be unhappy.
1959 */
1960 case PR_SET_MM_AUXV: {
1961 unsigned long user_auxv[AT_VECTOR_SIZE];
1962
1963 if (arg4 > sizeof(user_auxv))
1964 goto out;
1965 up_read(&mm->mmap_sem);
1966
1967 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
1968 return -EFAULT;
1969
1970 /* Make sure the last entry is always AT_NULL */
1971 user_auxv[AT_VECTOR_SIZE - 2] = 0;
1972 user_auxv[AT_VECTOR_SIZE - 1] = 0;
1973
1974 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1975
1976 task_lock(current);
1977 memcpy(mm->saved_auxv, user_auxv, arg4);
1978 task_unlock(current);
1979
1980 return 0;
1981 }
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001982 default:
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001983 goto out;
1984 }
1985
1986 error = 0;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001987out:
1988 up_read(&mm->mmap_sem);
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001989 return error;
1990}
1991#else /* CONFIG_CHECKPOINT_RESTORE */
1992static int prctl_set_mm(int opt, unsigned long addr,
1993 unsigned long arg4, unsigned long arg5)
1994{
1995 return -EINVAL;
1996}
1997#endif
1998
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001999SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2000 unsigned long, arg4, unsigned long, arg5)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001{
David Howellsb6dff3e2008-11-14 10:39:16 +11002002 struct task_struct *me = current;
2003 unsigned char comm[sizeof(me->comm)];
2004 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005
David Howellsd84f4f92008-11-14 10:39:23 +11002006 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2007 if (error != -ENOSYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 return error;
2009
David Howellsd84f4f92008-11-14 10:39:23 +11002010 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 switch (option) {
2012 case PR_SET_PDEATHSIG:
Jesper Juhl0730ded2005-09-06 15:17:37 -07002013 if (!valid_signal(arg2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 error = -EINVAL;
2015 break;
2016 }
David Howellsb6dff3e2008-11-14 10:39:16 +11002017 me->pdeath_signal = arg2;
2018 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 break;
2020 case PR_GET_PDEATHSIG:
David Howellsb6dff3e2008-11-14 10:39:16 +11002021 error = put_user(me->pdeath_signal, (int __user *)arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 break;
2023 case PR_GET_DUMPABLE:
David Howellsb6dff3e2008-11-14 10:39:16 +11002024 error = get_dumpable(me->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 break;
2026 case PR_SET_DUMPABLE:
Marcel Holtmannabf75a52006-07-12 13:12:00 +02002027 if (arg2 < 0 || arg2 > 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 error = -EINVAL;
2029 break;
2030 }
David Howellsb6dff3e2008-11-14 10:39:16 +11002031 set_dumpable(me->mm, arg2);
2032 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 break;
2034
2035 case PR_SET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002036 error = SET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 break;
2038 case PR_GET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002039 error = GET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 break;
2041 case PR_SET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11002042 error = SET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 break;
2044 case PR_GET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11002045 error = GET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 break;
2047 case PR_SET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11002048 error = SET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 break;
2050 case PR_GET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11002051 error = GET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 break;
2053 case PR_GET_TIMING:
2054 error = PR_TIMING_STATISTICAL;
2055 break;
2056 case PR_SET_TIMING:
Shi Weihua7b266552008-05-23 13:04:59 -07002057 if (arg2 != PR_TIMING_STATISTICAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 error = -EINVAL;
David Howellsb6dff3e2008-11-14 10:39:16 +11002059 else
2060 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 break;
2062
David Howellsb6dff3e2008-11-14 10:39:16 +11002063 case PR_SET_NAME:
2064 comm[sizeof(me->comm)-1] = 0;
2065 if (strncpy_from_user(comm, (char __user *)arg2,
2066 sizeof(me->comm) - 1) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067 return -EFAULT;
David Howellsb6dff3e2008-11-14 10:39:16 +11002068 set_task_comm(me, comm);
Vladimir Zapolskiyf786ecb2011-09-21 09:26:44 +00002069 proc_comm_connector(me);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070 return 0;
David Howellsb6dff3e2008-11-14 10:39:16 +11002071 case PR_GET_NAME:
2072 get_task_comm(comm, me);
2073 if (copy_to_user((char __user *)arg2, comm,
2074 sizeof(comm)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 return -EFAULT;
2076 return 0;
Anton Blanchard651d7652006-06-07 16:10:19 +10002077 case PR_GET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002078 error = GET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10002079 break;
2080 case PR_SET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002081 error = SET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10002082 break;
2083
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -07002084 case PR_GET_SECCOMP:
2085 error = prctl_get_seccomp();
2086 break;
2087 case PR_SET_SECCOMP:
Will Drewrye2cfabdf2012-04-12 16:47:57 -05002088 error = prctl_set_seccomp(arg2, (char __user *)arg3);
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -07002089 break;
Erik Bosman8fb402b2008-04-11 18:54:17 +02002090 case PR_GET_TSC:
2091 error = GET_TSC_CTL(arg2);
2092 break;
2093 case PR_SET_TSC:
2094 error = SET_TSC_CTL(arg2);
2095 break;
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002096 case PR_TASK_PERF_EVENTS_DISABLE:
2097 error = perf_event_task_disable();
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01002098 break;
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002099 case PR_TASK_PERF_EVENTS_ENABLE:
2100 error = perf_event_task_enable();
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01002101 break;
Arjan van de Ven69766752008-09-01 15:52:40 -07002102 case PR_GET_TIMERSLACK:
2103 error = current->timer_slack_ns;
2104 break;
2105 case PR_SET_TIMERSLACK:
2106 if (arg2 <= 0)
2107 current->timer_slack_ns =
2108 current->default_timer_slack_ns;
2109 else
2110 current->timer_slack_ns = arg2;
David Howellsb6dff3e2008-11-14 10:39:16 +11002111 error = 0;
Arjan van de Ven69766752008-09-01 15:52:40 -07002112 break;
Andi Kleen4db96cf2009-09-16 11:50:14 +02002113 case PR_MCE_KILL:
2114 if (arg4 | arg5)
2115 return -EINVAL;
2116 switch (arg2) {
Andi Kleen1087e9b2009-10-04 02:20:11 +02002117 case PR_MCE_KILL_CLEAR:
Andi Kleen4db96cf2009-09-16 11:50:14 +02002118 if (arg3 != 0)
2119 return -EINVAL;
2120 current->flags &= ~PF_MCE_PROCESS;
2121 break;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002122 case PR_MCE_KILL_SET:
Andi Kleen4db96cf2009-09-16 11:50:14 +02002123 current->flags |= PF_MCE_PROCESS;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002124 if (arg3 == PR_MCE_KILL_EARLY)
Andi Kleen4db96cf2009-09-16 11:50:14 +02002125 current->flags |= PF_MCE_EARLY;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002126 else if (arg3 == PR_MCE_KILL_LATE)
Andi Kleen4db96cf2009-09-16 11:50:14 +02002127 current->flags &= ~PF_MCE_EARLY;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002128 else if (arg3 == PR_MCE_KILL_DEFAULT)
2129 current->flags &=
2130 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2131 else
2132 return -EINVAL;
Andi Kleen4db96cf2009-09-16 11:50:14 +02002133 break;
2134 default:
2135 return -EINVAL;
2136 }
2137 error = 0;
2138 break;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002139 case PR_MCE_KILL_GET:
2140 if (arg2 | arg3 | arg4 | arg5)
2141 return -EINVAL;
2142 if (current->flags & PF_MCE_PROCESS)
2143 error = (current->flags & PF_MCE_EARLY) ?
2144 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2145 else
2146 error = PR_MCE_KILL_DEFAULT;
2147 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002148 case PR_SET_MM:
2149 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2150 break;
Lennart Poetteringebec18a2012-03-23 15:01:54 -07002151 case PR_SET_CHILD_SUBREAPER:
2152 me->signal->is_child_subreaper = !!arg2;
2153 error = 0;
2154 break;
2155 case PR_GET_CHILD_SUBREAPER:
2156 error = put_user(me->signal->is_child_subreaper,
2157 (int __user *) arg2);
2158 break;
Andy Lutomirski259e5e62012-04-12 16:47:50 -05002159 case PR_SET_NO_NEW_PRIVS:
2160 if (arg2 != 1 || arg3 || arg4 || arg5)
2161 return -EINVAL;
2162
2163 current->no_new_privs = 1;
2164 break;
2165 case PR_GET_NO_NEW_PRIVS:
2166 if (arg2 || arg3 || arg4 || arg5)
2167 return -EINVAL;
2168 return current->no_new_privs ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169 default:
2170 error = -EINVAL;
2171 break;
2172 }
2173 return error;
2174}
Andi Kleen3cfc3482006-09-26 10:52:28 +02002175
Heiko Carstens836f92a2009-01-14 14:14:33 +01002176SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2177 struct getcpu_cache __user *, unused)
Andi Kleen3cfc3482006-09-26 10:52:28 +02002178{
2179 int err = 0;
2180 int cpu = raw_smp_processor_id();
2181 if (cpup)
2182 err |= put_user(cpu, cpup);
2183 if (nodep)
2184 err |= put_user(cpu_to_node(cpu), nodep);
Andi Kleen3cfc3482006-09-26 10:52:28 +02002185 return err ? -EFAULT : 0;
2186}
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002187
2188char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2189
Neil Hormana06a4dc2010-05-26 14:42:58 -07002190static void argv_cleanup(struct subprocess_info *info)
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002191{
Neil Hormana06a4dc2010-05-26 14:42:58 -07002192 argv_free(info->argv);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002193}
2194
2195/**
2196 * orderly_poweroff - Trigger an orderly system poweroff
2197 * @force: force poweroff if command execution fails
2198 *
2199 * This may be called from any context to trigger a system shutdown.
2200 * If the orderly shutdown fails, it will force an immediate shutdown.
2201 */
2202int orderly_poweroff(bool force)
2203{
2204 int argc;
2205 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
2206 static char *envp[] = {
2207 "HOME=/",
2208 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2209 NULL
2210 };
2211 int ret = -ENOMEM;
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002212
2213 if (argv == NULL) {
2214 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2215 __func__, poweroff_cmd);
2216 goto out;
2217 }
2218
Boaz Harrosh81ab6e72012-05-31 16:26:15 -07002219 ret = call_usermodehelper_fns(argv[0], argv, envp, UMH_NO_WAIT,
2220 NULL, argv_cleanup, NULL);
2221out:
2222 if (likely(!ret))
2223 return 0;
2224
2225 if (ret == -ENOMEM)
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002226 argv_free(argv);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002227
Boaz Harrosh81ab6e72012-05-31 16:26:15 -07002228 if (force) {
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002229 printk(KERN_WARNING "Failed to start orderly shutdown: "
2230 "forcing the issue\n");
2231
2232 /* I guess this should try to kick off some daemon to
2233 sync and poweroff asap. Or not even bother syncing
2234 if we're doing an emergency shutdown? */
2235 emergency_sync();
2236 kernel_power_off();
2237 }
2238
2239 return ret;
2240}
2241EXPORT_SYMBOL_GPL(orderly_poweroff);