blob: c5cb5b99cb8152808f569f6b993010a33e646bce [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);
Shawn Guof96972f2012-10-04 17:12:23 -0700371 disable_nonboot_cpus();
Cal Peake756184b2006-09-30 23:27:24 -0700372 if (!cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600373 printk(KERN_EMERG "Restarting system.\n");
Cal Peake756184b2006-09-30 23:27:24 -0700374 else
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600375 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
Seiji Aguchi04c68622011-01-12 16:59:30 -0800376 kmsg_dump(KMSG_DUMP_RESTART);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600377 machine_restart(cmd);
378}
379EXPORT_SYMBOL_GPL(kernel_restart);
380
Adrian Bunk4ef72292008-02-04 22:30:06 -0800381static void kernel_shutdown_prepare(enum system_states state)
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500382{
Alan Sterne041c682006-03-27 01:16:30 -0800383 blocking_notifier_call_chain(&reboot_notifier_list,
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500384 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
385 system_state = state;
Kay Sieversb50fa7c2011-05-05 13:32:05 +0200386 usermodehelper_disable();
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500387 device_shutdown();
388}
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700389/**
390 * kernel_halt - halt the system
391 *
392 * Shutdown everything and perform a clean system halt.
393 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700394void kernel_halt(void)
395{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500396 kernel_shutdown_prepare(SYSTEM_HALT);
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100397 syscore_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600398 printk(KERN_EMERG "System halted.\n");
Seiji Aguchi04c68622011-01-12 16:59:30 -0800399 kmsg_dump(KMSG_DUMP_HALT);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600400 machine_halt();
401}
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500402
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600403EXPORT_SYMBOL_GPL(kernel_halt);
404
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700405/**
406 * kernel_power_off - power_off the system
407 *
408 * Shutdown everything and perform a clean system power_off.
409 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700410void kernel_power_off(void)
411{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500412 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
Rafael J. Wysockibd804eb2007-07-19 01:47:40 -0700413 if (pm_power_off_prepare)
414 pm_power_off_prepare();
Mark Lord40477272007-10-01 01:20:10 -0700415 disable_nonboot_cpus();
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +0100416 syscore_shutdown();
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600417 printk(KERN_EMERG "Power down.\n");
Seiji Aguchi04c68622011-01-12 16:59:30 -0800418 kmsg_dump(KMSG_DUMP_POWEROFF);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600419 machine_power_off();
420}
421EXPORT_SYMBOL_GPL(kernel_power_off);
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200422
423static DEFINE_MUTEX(reboot_mutex);
424
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425/*
426 * Reboot system call: for obvious reasons only root may call it,
427 * and even root needs to set up some magic numbers in the registers
428 * so that some mistake won't make this reboot the whole machine.
429 * You can also set the meaning of the ctrl-alt-del-key here.
430 *
431 * reboot doesn't sync: do that yourself before calling this.
432 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100433SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
434 void __user *, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435{
436 char buffer[256];
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700437 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
439 /* We only trust the superuser with rebooting the system. */
440 if (!capable(CAP_SYS_BOOT))
441 return -EPERM;
442
443 /* For safety, we require "magic" arguments. */
444 if (magic1 != LINUX_REBOOT_MAGIC1 ||
445 (magic2 != LINUX_REBOOT_MAGIC2 &&
446 magic2 != LINUX_REBOOT_MAGIC2A &&
447 magic2 != LINUX_REBOOT_MAGIC2B &&
448 magic2 != LINUX_REBOOT_MAGIC2C))
449 return -EINVAL;
450
Daniel Lezcanocf3f8922012-03-28 14:42:51 -0700451 /*
452 * If pid namespaces are enabled and the current task is in a child
453 * pid_namespace, the command is handled by reboot_pid_ns() which will
454 * call do_exit().
455 */
456 ret = reboot_pid_ns(task_active_pid_ns(current), cmd);
457 if (ret)
458 return ret;
459
Eric W. Biederman5e382912006-01-08 01:03:46 -0800460 /* Instead of trying to make the power_off code look like
461 * halt when pm_power_off is not set do it the easy way.
462 */
463 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
464 cmd = LINUX_REBOOT_CMD_HALT;
465
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200466 mutex_lock(&reboot_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 switch (cmd) {
468 case LINUX_REBOOT_CMD_RESTART:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600469 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 break;
471
472 case LINUX_REBOOT_CMD_CAD_ON:
473 C_A_D = 1;
474 break;
475
476 case LINUX_REBOOT_CMD_CAD_OFF:
477 C_A_D = 0;
478 break;
479
480 case LINUX_REBOOT_CMD_HALT:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600481 kernel_halt();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 do_exit(0);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700483 panic("cannot halt");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485 case LINUX_REBOOT_CMD_POWER_OFF:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600486 kernel_power_off();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 do_exit(0);
488 break;
489
490 case LINUX_REBOOT_CMD_RESTART2:
491 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200492 ret = -EFAULT;
493 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 }
495 buffer[sizeof(buffer) - 1] = '\0';
496
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600497 kernel_restart(buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 break;
499
Huang Ying3ab83522008-07-25 19:45:07 -0700500#ifdef CONFIG_KEXEC
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700501 case LINUX_REBOOT_CMD_KEXEC:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700502 ret = kernel_kexec();
503 break;
Huang Ying3ab83522008-07-25 19:45:07 -0700504#endif
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600505
Rafael J. Wysockib0cb1a12007-07-29 23:24:36 +0200506#ifdef CONFIG_HIBERNATION
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 case LINUX_REBOOT_CMD_SW_SUSPEND:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700508 ret = hibernate();
509 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510#endif
511
512 default:
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700513 ret = -EINVAL;
514 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 }
Thomas Gleixner6f15fa52009-10-09 20:31:33 +0200516 mutex_unlock(&reboot_mutex);
Andi Kleen3d26dcf2009-04-13 14:40:08 -0700517 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518}
519
David Howells65f27f32006-11-22 14:55:48 +0000520static void deferred_cad(struct work_struct *dummy)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521{
Eric W. Biedermanabcd9e52005-07-26 11:27:34 -0600522 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523}
524
525/*
526 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
527 * As it's called within an interrupt, it may NOT sync: the only choice
528 * is whether to reboot at once, or just ignore the ctrl-alt-del.
529 */
530void ctrl_alt_del(void)
531{
David Howells65f27f32006-11-22 14:55:48 +0000532 static DECLARE_WORK(cad_work, deferred_cad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534 if (C_A_D)
535 schedule_work(&cad_work);
536 else
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700537 kill_cad_pid(SIGINT, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538}
539
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540/*
541 * Unprivileged users may change the real gid to the effective gid
542 * or vice versa. (BSD-style)
543 *
544 * If you set the real gid at all, or set the effective gid to a value not
545 * equal to the real gid, then the saved gid is set to the new effective gid.
546 *
547 * This makes it possible for a setgid program to completely drop its
548 * privileges, which is often a useful assertion to make when you are doing
549 * a security audit over a program.
550 *
551 * The general idea is that a program which uses just setregid() will be
552 * 100% compatible with BSD. A program which uses just setgid() will be
553 * 100% compatible with POSIX with saved IDs.
554 *
555 * SMP: There are not races, the GIDs are checked only by filesystem
556 * operations (as far as semantic preservation is concerned).
557 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100558SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800560 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100561 const struct cred *old;
562 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800564 kgid_t krgid, kegid;
565
566 krgid = make_kgid(ns, rgid);
567 kegid = make_kgid(ns, egid);
568
569 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
570 return -EINVAL;
571 if ((egid != (gid_t) -1) && !gid_valid(kegid))
572 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
David Howellsd84f4f92008-11-14 10:39:23 +1100574 new = prepare_creds();
575 if (!new)
576 return -ENOMEM;
577 old = current_cred();
578
David Howellsd84f4f92008-11-14 10:39:23 +1100579 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 if (rgid != (gid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800581 if (gid_eq(old->gid, krgid) ||
582 gid_eq(old->egid, krgid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700583 nsown_capable(CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800584 new->gid = krgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 else
David Howellsd84f4f92008-11-14 10:39:23 +1100586 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 }
588 if (egid != (gid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800589 if (gid_eq(old->gid, kegid) ||
590 gid_eq(old->egid, kegid) ||
591 gid_eq(old->sgid, kegid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700592 nsown_capable(CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800593 new->egid = kegid;
Cal Peake756184b2006-09-30 23:27:24 -0700594 else
David Howellsd84f4f92008-11-14 10:39:23 +1100595 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 }
David Howellsd84f4f92008-11-14 10:39:23 +1100597
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 if (rgid != (gid_t) -1 ||
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800599 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
David Howellsd84f4f92008-11-14 10:39:23 +1100600 new->sgid = new->egid;
601 new->fsgid = new->egid;
602
603 return commit_creds(new);
604
605error:
606 abort_creds(new);
607 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608}
609
610/*
611 * setgid() is implemented like SysV w/ SAVED_IDS
612 *
613 * SMP: Same implicit races as above.
614 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100615SYSCALL_DEFINE1(setgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800617 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100618 const struct cred *old;
619 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800621 kgid_t kgid;
622
623 kgid = make_kgid(ns, gid);
624 if (!gid_valid(kgid))
625 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626
David Howellsd84f4f92008-11-14 10:39:23 +1100627 new = prepare_creds();
628 if (!new)
629 return -ENOMEM;
630 old = current_cred();
631
David Howellsd84f4f92008-11-14 10:39:23 +1100632 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700633 if (nsown_capable(CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800634 new->gid = new->egid = new->sgid = new->fsgid = kgid;
635 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
636 new->egid = new->fsgid = kgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 else
David Howellsd84f4f92008-11-14 10:39:23 +1100638 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
David Howellsd84f4f92008-11-14 10:39:23 +1100640 return commit_creds(new);
641
642error:
643 abort_creds(new);
644 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645}
Dhaval Giani54e99122009-02-27 15:13:54 +0530646
David Howellsd84f4f92008-11-14 10:39:23 +1100647/*
648 * change the user struct in a credentials set to match the new UID
649 */
650static int set_user(struct cred *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651{
652 struct user_struct *new_user;
653
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800654 new_user = alloc_uid(new->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 if (!new_user)
656 return -EAGAIN;
657
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400658 /*
659 * We don't fail in case of NPROC limit excess here because too many
660 * poorly written programs don't check set*uid() return code, assuming
661 * it never fails if called by root. We may still enforce NPROC limit
662 * for programs doing set*uid()+execve() by harmlessly deferring the
663 * failure to the execve() stage.
664 */
Jiri Slaby78d7d402010-03-05 13:42:54 -0800665 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400666 new_user != INIT_USER)
667 current->flags |= PF_NPROC_EXCEEDED;
668 else
669 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
David Howellsd84f4f92008-11-14 10:39:23 +1100671 free_uid(new->user);
672 new->user = new_user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 return 0;
674}
675
676/*
677 * Unprivileged users may change the real uid to the effective uid
678 * or vice versa. (BSD-style)
679 *
680 * If you set the real uid at all, or set the effective uid to a value not
681 * equal to the real uid, then the saved uid is set to the new effective uid.
682 *
683 * This makes it possible for a setuid program to completely drop its
684 * privileges, which is often a useful assertion to make when you are doing
685 * a security audit over a program.
686 *
687 * The general idea is that a program which uses just setreuid() will be
688 * 100% compatible with BSD. A program which uses just setuid() will be
689 * 100% compatible with POSIX with saved IDs.
690 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100691SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800693 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100694 const struct cred *old;
695 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800697 kuid_t kruid, keuid;
698
699 kruid = make_kuid(ns, ruid);
700 keuid = make_kuid(ns, euid);
701
702 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
703 return -EINVAL;
704 if ((euid != (uid_t) -1) && !uid_valid(keuid))
705 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
David Howellsd84f4f92008-11-14 10:39:23 +1100707 new = prepare_creds();
708 if (!new)
709 return -ENOMEM;
710 old = current_cred();
711
David Howellsd84f4f92008-11-14 10:39:23 +1100712 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 if (ruid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800714 new->uid = kruid;
715 if (!uid_eq(old->uid, kruid) &&
716 !uid_eq(old->euid, kruid) &&
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700717 !nsown_capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100718 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 }
720
721 if (euid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800722 new->euid = keuid;
723 if (!uid_eq(old->uid, keuid) &&
724 !uid_eq(old->euid, keuid) &&
725 !uid_eq(old->suid, keuid) &&
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700726 !nsown_capable(CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100727 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 }
729
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800730 if (!uid_eq(new->uid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530731 retval = set_user(new);
732 if (retval < 0)
733 goto error;
734 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 if (ruid != (uid_t) -1 ||
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800736 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
David Howellsd84f4f92008-11-14 10:39:23 +1100737 new->suid = new->euid;
738 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
David Howellsd84f4f92008-11-14 10:39:23 +1100740 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
741 if (retval < 0)
742 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
David Howellsd84f4f92008-11-14 10:39:23 +1100744 return commit_creds(new);
745
746error:
747 abort_creds(new);
748 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751/*
752 * setuid() is implemented like SysV with SAVED_IDS
753 *
754 * Note that SAVED_ID's is deficient in that a setuid root program
755 * like sendmail, for example, cannot set its uid to be a normal
756 * user and then switch back, because if you're root, setuid() sets
757 * the saved uid too. If you don't like this, blame the bright people
758 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
759 * will allow a root program to temporarily drop privileges and be able to
760 * regain them by swapping the real and effective uid.
761 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100762SYSCALL_DEFINE1(setuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800764 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100765 const struct cred *old;
766 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800768 kuid_t kuid;
769
770 kuid = make_kuid(ns, uid);
771 if (!uid_valid(kuid))
772 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
David Howellsd84f4f92008-11-14 10:39:23 +1100774 new = prepare_creds();
775 if (!new)
776 return -ENOMEM;
777 old = current_cred();
778
David Howellsd84f4f92008-11-14 10:39:23 +1100779 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700780 if (nsown_capable(CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800781 new->suid = new->uid = kuid;
782 if (!uid_eq(kuid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530783 retval = set_user(new);
784 if (retval < 0)
785 goto error;
David Howellsd84f4f92008-11-14 10:39:23 +1100786 }
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800787 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100788 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800791 new->fsuid = new->euid = kuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
David Howellsd84f4f92008-11-14 10:39:23 +1100793 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
794 if (retval < 0)
795 goto error;
796
797 return commit_creds(new);
798
799error:
800 abort_creds(new);
801 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802}
803
804
805/*
806 * This function implements a generic ability to update ruid, euid,
807 * and suid. This allows you to implement the 4.4 compatible seteuid().
808 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100809SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800811 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100812 const struct cred *old;
813 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800815 kuid_t kruid, keuid, ksuid;
816
817 kruid = make_kuid(ns, ruid);
818 keuid = make_kuid(ns, euid);
819 ksuid = make_kuid(ns, suid);
820
821 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
822 return -EINVAL;
823
824 if ((euid != (uid_t) -1) && !uid_valid(keuid))
825 return -EINVAL;
826
827 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
828 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
David Howellsd84f4f92008-11-14 10:39:23 +1100830 new = prepare_creds();
831 if (!new)
832 return -ENOMEM;
833
David Howellsd84f4f92008-11-14 10:39:23 +1100834 old = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
David Howellsd84f4f92008-11-14 10:39:23 +1100836 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700837 if (!nsown_capable(CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800838 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
839 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100840 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800841 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
842 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100843 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800844 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
845 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100846 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 }
David Howellsd84f4f92008-11-14 10:39:23 +1100848
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 if (ruid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800850 new->uid = kruid;
851 if (!uid_eq(kruid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530852 retval = set_user(new);
853 if (retval < 0)
854 goto error;
855 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 }
David Howellsd84f4f92008-11-14 10:39:23 +1100857 if (euid != (uid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800858 new->euid = keuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 if (suid != (uid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800860 new->suid = ksuid;
David Howellsd84f4f92008-11-14 10:39:23 +1100861 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862
David Howellsd84f4f92008-11-14 10:39:23 +1100863 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
864 if (retval < 0)
865 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
David Howellsd84f4f92008-11-14 10:39:23 +1100867 return commit_creds(new);
868
869error:
870 abort_creds(new);
871 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872}
873
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800874SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875{
David Howells86a264a2008-11-14 10:39:18 +1100876 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800878 uid_t ruid, euid, suid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800880 ruid = from_kuid_munged(cred->user_ns, cred->uid);
881 euid = from_kuid_munged(cred->user_ns, cred->euid);
882 suid = from_kuid_munged(cred->user_ns, cred->suid);
883
884 if (!(retval = put_user(ruid, ruidp)) &&
885 !(retval = put_user(euid, euidp)))
886 retval = put_user(suid, suidp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
888 return retval;
889}
890
891/*
892 * Same as above, but for rgid, egid, sgid.
893 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100894SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800896 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100897 const struct cred *old;
898 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800900 kgid_t krgid, kegid, ksgid;
901
902 krgid = make_kgid(ns, rgid);
903 kegid = make_kgid(ns, egid);
904 ksgid = make_kgid(ns, sgid);
905
906 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
907 return -EINVAL;
908 if ((egid != (gid_t) -1) && !gid_valid(kegid))
909 return -EINVAL;
910 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
911 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
David Howellsd84f4f92008-11-14 10:39:23 +1100913 new = prepare_creds();
914 if (!new)
915 return -ENOMEM;
916 old = current_cred();
917
David Howellsd84f4f92008-11-14 10:39:23 +1100918 retval = -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700919 if (!nsown_capable(CAP_SETGID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800920 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
921 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100922 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800923 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
924 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100925 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800926 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
927 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100928 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
David Howellsd84f4f92008-11-14 10:39:23 +1100931 if (rgid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800932 new->gid = krgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100933 if (egid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800934 new->egid = kegid;
David Howellsd84f4f92008-11-14 10:39:23 +1100935 if (sgid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800936 new->sgid = ksgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100937 new->fsgid = new->egid;
938
939 return commit_creds(new);
940
941error:
942 abort_creds(new);
943 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944}
945
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800946SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947{
David Howells86a264a2008-11-14 10:39:18 +1100948 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800950 gid_t rgid, egid, sgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800952 rgid = from_kgid_munged(cred->user_ns, cred->gid);
953 egid = from_kgid_munged(cred->user_ns, cred->egid);
954 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
955
956 if (!(retval = put_user(rgid, rgidp)) &&
957 !(retval = put_user(egid, egidp)))
958 retval = put_user(sgid, sgidp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960 return retval;
961}
962
963
964/*
965 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
966 * is used for "access()" and for the NFS daemon (letting nfsd stay at
967 * whatever uid it wants to). It normally shadows "euid", except when
968 * explicitly set by setfsuid() or for access..
969 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100970SYSCALL_DEFINE1(setfsuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971{
David Howellsd84f4f92008-11-14 10:39:23 +1100972 const struct cred *old;
973 struct cred *new;
974 uid_t old_fsuid;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800975 kuid_t kuid;
976
977 old = current_cred();
978 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
979
980 kuid = make_kuid(old->user_ns, uid);
981 if (!uid_valid(kuid))
982 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983
David Howellsd84f4f92008-11-14 10:39:23 +1100984 new = prepare_creds();
985 if (!new)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800986 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800988 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
989 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700990 nsown_capable(CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800991 if (!uid_eq(kuid, old->fsuid)) {
992 new->fsuid = kuid;
David Howellsd84f4f92008-11-14 10:39:23 +1100993 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
994 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 }
997
David Howellsd84f4f92008-11-14 10:39:23 +1100998 abort_creds(new);
999 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000
David Howellsd84f4f92008-11-14 10:39:23 +11001001change_okay:
1002 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 return old_fsuid;
1004}
1005
1006/*
John Anthony Kazos Jrf42df9e2007-05-09 08:23:08 +02001007 * Samma på svenska..
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 */
Heiko Carstensae1251a2009-01-14 14:14:05 +01001009SYSCALL_DEFINE1(setfsgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010{
David Howellsd84f4f92008-11-14 10:39:23 +11001011 const struct cred *old;
1012 struct cred *new;
1013 gid_t old_fsgid;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001014 kgid_t kgid;
1015
1016 old = current_cred();
1017 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1018
1019 kgid = make_kgid(old->user_ns, gid);
1020 if (!gid_valid(kgid))
1021 return old_fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022
David Howellsd84f4f92008-11-14 10:39:23 +11001023 new = prepare_creds();
1024 if (!new)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001025 return old_fsgid;
David Howellsd84f4f92008-11-14 10:39:23 +11001026
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001027 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1028 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001029 nsown_capable(CAP_SETGID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -08001030 if (!gid_eq(kgid, old->fsgid)) {
1031 new->fsgid = kgid;
David Howellsd84f4f92008-11-14 10:39:23 +11001032 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 }
David Howellsd84f4f92008-11-14 10:39:23 +11001035
David Howellsd84f4f92008-11-14 10:39:23 +11001036 abort_creds(new);
1037 return old_fsgid;
1038
1039change_okay:
1040 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 return old_fsgid;
1042}
1043
Frank Mayharf06febc2008-09-12 09:54:39 -07001044void do_sys_times(struct tms *tms)
1045{
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001046 cputime_t tgutime, tgstime, cutime, cstime;
Frank Mayharf06febc2008-09-12 09:54:39 -07001047
Oleg Nesterov2b5fe6d2008-11-17 15:40:08 +01001048 spin_lock_irq(&current->sighand->siglock);
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001049 thread_group_times(current, &tgutime, &tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001050 cutime = current->signal->cutime;
1051 cstime = current->signal->cstime;
1052 spin_unlock_irq(&current->sighand->siglock);
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001053 tms->tms_utime = cputime_to_clock_t(tgutime);
1054 tms->tms_stime = cputime_to_clock_t(tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001055 tms->tms_cutime = cputime_to_clock_t(cutime);
1056 tms->tms_cstime = cputime_to_clock_t(cstime);
1057}
1058
Heiko Carstens58fd3aa2009-01-14 14:14:03 +01001059SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 if (tbuf) {
1062 struct tms tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063
Frank Mayharf06febc2008-09-12 09:54:39 -07001064 do_sys_times(&tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1066 return -EFAULT;
1067 }
Paul Mackerrase3d5a272009-01-06 14:41:02 -08001068 force_successful_syscall_return();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1070}
1071
1072/*
1073 * This needs some heavy checking ...
1074 * I just haven't the stomach for it. I also don't fully
1075 * understand sessions/pgrp etc. Let somebody who does explain it.
1076 *
1077 * OK, I think I have the protection semantics right.... this is really
1078 * only important on a multi-user system anyway, to make sure one user
1079 * can't send a signal to a process owned by another. -TYT, 12/12/91
1080 *
1081 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1082 * LBT 04.03.94
1083 */
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001084SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085{
1086 struct task_struct *p;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001087 struct task_struct *group_leader = current->group_leader;
Oleg Nesterov4e021302008-02-08 04:19:08 -08001088 struct pid *pgrp;
1089 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090
1091 if (!pid)
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001092 pid = task_pid_vnr(group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 if (!pgid)
1094 pgid = pid;
1095 if (pgid < 0)
1096 return -EINVAL;
Paul E. McKenney950eaac2010-08-31 17:00:18 -07001097 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
1099 /* From this point forward we keep holding onto the tasklist lock
1100 * so that our parent does not change from under us. -DaveM
1101 */
1102 write_lock_irq(&tasklist_lock);
1103
1104 err = -ESRCH;
Oleg Nesterov4e021302008-02-08 04:19:08 -08001105 p = find_task_by_vpid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 if (!p)
1107 goto out;
1108
1109 err = -EINVAL;
1110 if (!thread_group_leader(p))
1111 goto out;
1112
Oleg Nesterov4e021302008-02-08 04:19:08 -08001113 if (same_thread_group(p->real_parent, group_leader)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 err = -EPERM;
Eric W. Biederman41487c62007-02-12 00:53:01 -08001115 if (task_session(p) != task_session(group_leader))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 goto out;
1117 err = -EACCES;
1118 if (p->did_exec)
1119 goto out;
1120 } else {
1121 err = -ESRCH;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001122 if (p != group_leader)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 goto out;
1124 }
1125
1126 err = -EPERM;
1127 if (p->signal->leader)
1128 goto out;
1129
Oleg Nesterov4e021302008-02-08 04:19:08 -08001130 pgrp = task_pid(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 if (pgid != pid) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001132 struct task_struct *g;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
Oleg Nesterov4e021302008-02-08 04:19:08 -08001134 pgrp = find_vpid(pgid);
1135 g = pid_task(pgrp, PIDTYPE_PGID);
Eric W. Biederman41487c62007-02-12 00:53:01 -08001136 if (!g || task_session(g) != task_session(group_leader))
Oleg Nesterovf020bc42006-12-08 02:38:02 -08001137 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 }
1139
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 err = security_task_setpgid(p, pgid);
1141 if (err)
1142 goto out;
1143
Oleg Nesterov1b0f7ffd2009-04-02 16:58:39 -07001144 if (task_pgrp(p) != pgrp)
Oleg Nesterov83beaf32008-04-30 00:54:27 -07001145 change_pid(p, PIDTYPE_PGID, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
1147 err = 0;
1148out:
1149 /* All paths lead to here, thus we are safe. -DaveM */
1150 write_unlock_irq(&tasklist_lock);
Paul E. McKenney950eaac2010-08-31 17:00:18 -07001151 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 return err;
1153}
1154
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001155SYSCALL_DEFINE1(getpgid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001157 struct task_struct *p;
1158 struct pid *grp;
1159 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001161 rcu_read_lock();
1162 if (!pid)
1163 grp = task_pgrp(current);
1164 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 retval = -ESRCH;
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001166 p = find_task_by_vpid(pid);
1167 if (!p)
1168 goto out;
1169 grp = task_pgrp(p);
1170 if (!grp)
1171 goto out;
1172
1173 retval = security_task_getpgid(p);
1174 if (retval)
1175 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 }
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001177 retval = pid_vnr(grp);
1178out:
1179 rcu_read_unlock();
1180 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181}
1182
1183#ifdef __ARCH_WANT_SYS_GETPGRP
1184
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001185SYSCALL_DEFINE0(getpgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001187 return sys_getpgid(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188}
1189
1190#endif
1191
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001192SYSCALL_DEFINE1(getsid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193{
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001194 struct task_struct *p;
1195 struct pid *sid;
1196 int retval;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001197
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001198 rcu_read_lock();
1199 if (!pid)
1200 sid = task_session(current);
1201 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 retval = -ESRCH;
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001203 p = find_task_by_vpid(pid);
1204 if (!p)
1205 goto out;
1206 sid = task_session(p);
1207 if (!sid)
1208 goto out;
1209
1210 retval = security_task_getsid(p);
1211 if (retval)
1212 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 }
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001214 retval = pid_vnr(sid);
1215out:
1216 rcu_read_unlock();
1217 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218}
1219
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001220SYSCALL_DEFINE0(setsid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221{
Oren Laadane19f2472006-01-08 01:03:58 -08001222 struct task_struct *group_leader = current->group_leader;
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001223 struct pid *sid = task_pid(group_leader);
1224 pid_t session = pid_vnr(sid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 int err = -EPERM;
1226
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 write_lock_irq(&tasklist_lock);
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001228 /* Fail if I am already a session leader */
1229 if (group_leader->signal->leader)
1230 goto out;
1231
Oleg Nesterov430c6232008-02-08 04:19:11 -08001232 /* Fail if a process group id already exists that equals the
1233 * proposed session id.
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001234 */
Oleg Nesterov6806aac2008-02-08 04:19:12 -08001235 if (pid_task(sid, PIDTYPE_PGID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 goto out;
1237
Oren Laadane19f2472006-01-08 01:03:58 -08001238 group_leader->signal->leader = 1;
Oleg Nesterov8520d7c2008-02-08 04:19:09 -08001239 __set_special_pids(sid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001240
Alan Cox9c9f4de2008-10-13 10:37:26 +01001241 proc_clear_tty(group_leader);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001242
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001243 err = session;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244out:
1245 write_unlock_irq(&tasklist_lock);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001246 if (err > 0) {
Christian Borntraeger0d0df592009-10-26 16:49:34 -07001247 proc_sid_connector(group_leader);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001248 sched_autogroup_create_attach(group_leader);
1249 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 return err;
1251}
1252
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253DECLARE_RWSEM(uts_sem);
1254
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001255#ifdef COMPAT_UTS_MACHINE
1256#define override_architecture(name) \
Andreas Schwab46da2762010-04-23 13:17:44 -04001257 (personality(current->personality) == PER_LINUX32 && \
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001258 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1259 sizeof(COMPAT_UTS_MACHINE)))
1260#else
1261#define override_architecture(name) 0
1262#endif
1263
Andi Kleenbe274252011-08-19 16:15:10 -07001264/*
1265 * Work around broken programs that cannot handle "Linux 3.0".
1266 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1267 */
1268static int override_release(char __user *release, int len)
1269{
1270 int ret = 0;
Linus Torvaldsa84a79e2011-10-17 08:24:24 -07001271 char buf[65];
Andi Kleenbe274252011-08-19 16:15:10 -07001272
1273 if (current->personality & UNAME26) {
1274 char *rest = UTS_RELEASE;
1275 int ndots = 0;
1276 unsigned v;
1277
1278 while (*rest) {
1279 if (*rest == '.' && ++ndots >= 3)
1280 break;
1281 if (!isdigit(*rest) && *rest != '.')
1282 break;
1283 rest++;
1284 }
1285 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1286 snprintf(buf, len, "2.6.%u%s", v, rest);
1287 ret = copy_to_user(release, buf, len);
1288 }
1289 return ret;
1290}
1291
Heiko Carstense48fbb62009-01-14 14:14:26 +01001292SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293{
1294 int errno = 0;
1295
1296 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001297 if (copy_to_user(name, utsname(), sizeof *name))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 errno = -EFAULT;
1299 up_read(&uts_sem);
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001300
Andi Kleenbe274252011-08-19 16:15:10 -07001301 if (!errno && override_release(name->release, sizeof(name->release)))
1302 errno = -EFAULT;
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001303 if (!errno && override_architecture(name))
1304 errno = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 return errno;
1306}
1307
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001308#ifdef __ARCH_WANT_SYS_OLD_UNAME
1309/*
1310 * Old cruft
1311 */
1312SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1313{
1314 int error = 0;
1315
1316 if (!name)
1317 return -EFAULT;
1318
1319 down_read(&uts_sem);
1320 if (copy_to_user(name, utsname(), sizeof(*name)))
1321 error = -EFAULT;
1322 up_read(&uts_sem);
1323
Andi Kleenbe274252011-08-19 16:15:10 -07001324 if (!error && override_release(name->release, sizeof(name->release)))
1325 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001326 if (!error && override_architecture(name))
1327 error = -EFAULT;
1328 return error;
1329}
1330
1331SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1332{
1333 int error;
1334
1335 if (!name)
1336 return -EFAULT;
1337 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1338 return -EFAULT;
1339
1340 down_read(&uts_sem);
1341 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1342 __OLD_UTS_LEN);
1343 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1344 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1345 __OLD_UTS_LEN);
1346 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1347 error |= __copy_to_user(&name->release, &utsname()->release,
1348 __OLD_UTS_LEN);
1349 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1350 error |= __copy_to_user(&name->version, &utsname()->version,
1351 __OLD_UTS_LEN);
1352 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1353 error |= __copy_to_user(&name->machine, &utsname()->machine,
1354 __OLD_UTS_LEN);
1355 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1356 up_read(&uts_sem);
1357
1358 if (!error && override_architecture(name))
1359 error = -EFAULT;
Andi Kleenbe274252011-08-19 16:15:10 -07001360 if (!error && override_release(name->release, sizeof(name->release)))
1361 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001362 return error ? -EFAULT : 0;
1363}
1364#endif
1365
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001366SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367{
1368 int errno;
1369 char tmp[__NEW_UTS_LEN];
1370
Serge E. Hallynbb96a6f2011-03-23 16:43:18 -07001371 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 return -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001373
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 if (len < 0 || len > __NEW_UTS_LEN)
1375 return -EINVAL;
1376 down_write(&uts_sem);
1377 errno = -EFAULT;
1378 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001379 struct new_utsname *u = utsname();
1380
1381 memcpy(u->nodename, tmp, len);
1382 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 errno = 0;
Sasikantha babu499eea62012-05-31 16:26:07 -07001384 uts_proc_notify(UTS_PROC_HOSTNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 }
1386 up_write(&uts_sem);
1387 return errno;
1388}
1389
1390#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1391
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001392SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393{
1394 int i, errno;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001395 struct new_utsname *u;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396
1397 if (len < 0)
1398 return -EINVAL;
1399 down_read(&uts_sem);
Andrew Morton9679e4d2008-10-15 22:01:51 -07001400 u = utsname();
1401 i = 1 + strlen(u->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 if (i > len)
1403 i = len;
1404 errno = 0;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001405 if (copy_to_user(name, u->nodename, i))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 errno = -EFAULT;
1407 up_read(&uts_sem);
1408 return errno;
1409}
1410
1411#endif
1412
1413/*
1414 * Only setdomainname; getdomainname can be implemented by calling
1415 * uname()
1416 */
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001417SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418{
1419 int errno;
1420 char tmp[__NEW_UTS_LEN];
1421
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001422 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 return -EPERM;
1424 if (len < 0 || len > __NEW_UTS_LEN)
1425 return -EINVAL;
1426
1427 down_write(&uts_sem);
1428 errno = -EFAULT;
1429 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001430 struct new_utsname *u = utsname();
1431
1432 memcpy(u->domainname, tmp, len);
1433 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 errno = 0;
Sasikantha babu499eea62012-05-31 16:26:07 -07001435 uts_proc_notify(UTS_PROC_DOMAINNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 }
1437 up_write(&uts_sem);
1438 return errno;
1439}
1440
Heiko Carstense48fbb62009-01-14 14:14:26 +01001441SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442{
Jiri Slabyb9518342010-05-04 11:28:25 +02001443 struct rlimit value;
1444 int ret;
1445
1446 ret = do_prlimit(current, resource, NULL, &value);
1447 if (!ret)
1448 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1449
1450 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451}
1452
1453#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1454
1455/*
1456 * Back compatibility for getrlimit. Needed for some apps.
1457 */
1458
Heiko Carstense48fbb62009-01-14 14:14:26 +01001459SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1460 struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461{
1462 struct rlimit x;
1463 if (resource >= RLIM_NLIMITS)
1464 return -EINVAL;
1465
1466 task_lock(current->group_leader);
1467 x = current->signal->rlim[resource];
1468 task_unlock(current->group_leader);
Cal Peake756184b2006-09-30 23:27:24 -07001469 if (x.rlim_cur > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 x.rlim_cur = 0x7FFFFFFF;
Cal Peake756184b2006-09-30 23:27:24 -07001471 if (x.rlim_max > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 x.rlim_max = 0x7FFFFFFF;
1473 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1474}
1475
1476#endif
1477
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001478static inline bool rlim64_is_infinity(__u64 rlim64)
1479{
1480#if BITS_PER_LONG < 64
1481 return rlim64 >= ULONG_MAX;
1482#else
1483 return rlim64 == RLIM64_INFINITY;
1484#endif
1485}
1486
1487static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1488{
1489 if (rlim->rlim_cur == RLIM_INFINITY)
1490 rlim64->rlim_cur = RLIM64_INFINITY;
1491 else
1492 rlim64->rlim_cur = rlim->rlim_cur;
1493 if (rlim->rlim_max == RLIM_INFINITY)
1494 rlim64->rlim_max = RLIM64_INFINITY;
1495 else
1496 rlim64->rlim_max = rlim->rlim_max;
1497}
1498
1499static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1500{
1501 if (rlim64_is_infinity(rlim64->rlim_cur))
1502 rlim->rlim_cur = RLIM_INFINITY;
1503 else
1504 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1505 if (rlim64_is_infinity(rlim64->rlim_max))
1506 rlim->rlim_max = RLIM_INFINITY;
1507 else
1508 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1509}
1510
Jiri Slaby1c1e618d2009-08-28 14:08:17 +02001511/* make sure you are allowed to change @tsk limits before calling this */
Jiri Slaby5b415352010-03-24 16:11:29 +01001512int do_prlimit(struct task_struct *tsk, unsigned int resource,
1513 struct rlimit *new_rlim, struct rlimit *old_rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514{
Jiri Slaby5b415352010-03-24 16:11:29 +01001515 struct rlimit *rlim;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001516 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517
1518 if (resource >= RLIM_NLIMITS)
1519 return -EINVAL;
Jiri Slaby5b415352010-03-24 16:11:29 +01001520 if (new_rlim) {
1521 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1522 return -EINVAL;
1523 if (resource == RLIMIT_NOFILE &&
1524 new_rlim->rlim_max > sysctl_nr_open)
1525 return -EPERM;
1526 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Jiri Slaby1c1e618d2009-08-28 14:08:17 +02001528 /* protect tsk->signal and tsk->sighand from disappearing */
1529 read_lock(&tasklist_lock);
1530 if (!tsk->sighand) {
1531 retval = -ESRCH;
1532 goto out;
1533 }
1534
Jiri Slaby5b415352010-03-24 16:11:29 +01001535 rlim = tsk->signal->rlim + resource;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001536 task_lock(tsk->group_leader);
Jiri Slaby5b415352010-03-24 16:11:29 +01001537 if (new_rlim) {
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001538 /* Keep the capable check against init_user_ns until
1539 cgroups can contain all limits */
Jiri Slaby5b415352010-03-24 16:11:29 +01001540 if (new_rlim->rlim_max > rlim->rlim_max &&
1541 !capable(CAP_SYS_RESOURCE))
1542 retval = -EPERM;
1543 if (!retval)
1544 retval = security_task_setrlimit(tsk->group_leader,
1545 resource, new_rlim);
1546 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1547 /*
1548 * The caller is asking for an immediate RLIMIT_CPU
1549 * expiry. But we use the zero value to mean "it was
1550 * never set". So let's cheat and make it one second
1551 * instead
1552 */
1553 new_rlim->rlim_cur = 1;
1554 }
Tom Alsberg9926e4c2007-05-08 00:30:31 -07001555 }
Jiri Slaby5b415352010-03-24 16:11:29 +01001556 if (!retval) {
1557 if (old_rlim)
1558 *old_rlim = *rlim;
1559 if (new_rlim)
1560 *rlim = *new_rlim;
1561 }
Jiri Slaby7855c352009-08-26 23:45:34 +02001562 task_unlock(tsk->group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
Andrew Mortond3561f72006-03-24 03:18:36 -08001564 /*
1565 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1566 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1567 * very long-standing error, and fixing it now risks breakage of
1568 * applications, so we live with it
1569 */
Jiri Slaby5b415352010-03-24 16:11:29 +01001570 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1571 new_rlim->rlim_cur != RLIM_INFINITY)
1572 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001573out:
Jiri Slaby1c1e618d2009-08-28 14:08:17 +02001574 read_unlock(&tasklist_lock);
Oleg Nesterov2fb9d262009-09-03 19:21:45 +02001575 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576}
1577
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001578/* rcu lock must be held */
1579static int check_prlimit_permission(struct task_struct *task)
1580{
1581 const struct cred *cred = current_cred(), *tcred;
1582
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001583 if (current == task)
1584 return 0;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001585
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001586 tcred = __task_cred(task);
Eric W. Biederman5af66202012-03-03 20:21:47 -08001587 if (uid_eq(cred->uid, tcred->euid) &&
1588 uid_eq(cred->uid, tcred->suid) &&
1589 uid_eq(cred->uid, tcred->uid) &&
1590 gid_eq(cred->gid, tcred->egid) &&
1591 gid_eq(cred->gid, tcred->sgid) &&
1592 gid_eq(cred->gid, tcred->gid))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001593 return 0;
Eric W. Biedermanc4a4d602011-11-16 23:15:31 -08001594 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001595 return 0;
1596
1597 return -EPERM;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001598}
1599
1600SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1601 const struct rlimit64 __user *, new_rlim,
1602 struct rlimit64 __user *, old_rlim)
1603{
1604 struct rlimit64 old64, new64;
1605 struct rlimit old, new;
1606 struct task_struct *tsk;
1607 int ret;
1608
1609 if (new_rlim) {
1610 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1611 return -EFAULT;
1612 rlim64_to_rlim(&new64, &new);
1613 }
1614
1615 rcu_read_lock();
1616 tsk = pid ? find_task_by_vpid(pid) : current;
1617 if (!tsk) {
1618 rcu_read_unlock();
1619 return -ESRCH;
1620 }
1621 ret = check_prlimit_permission(tsk);
1622 if (ret) {
1623 rcu_read_unlock();
1624 return ret;
1625 }
1626 get_task_struct(tsk);
1627 rcu_read_unlock();
1628
1629 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1630 old_rlim ? &old : NULL);
1631
1632 if (!ret && old_rlim) {
1633 rlim_to_rlim64(&old, &old64);
1634 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1635 ret = -EFAULT;
1636 }
1637
1638 put_task_struct(tsk);
1639 return ret;
1640}
1641
Jiri Slaby7855c352009-08-26 23:45:34 +02001642SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1643{
1644 struct rlimit new_rlim;
1645
1646 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1647 return -EFAULT;
Jiri Slaby5b415352010-03-24 16:11:29 +01001648 return do_prlimit(current, resource, &new_rlim, NULL);
Jiri Slaby7855c352009-08-26 23:45:34 +02001649}
1650
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651/*
1652 * It would make sense to put struct rusage in the task_struct,
1653 * except that would make the task_struct be *really big*. After
1654 * task_struct gets moved into malloc'ed memory, it would
1655 * make sense to do this. It will make moving the rest of the information
1656 * a lot simpler! (Which we're not doing right now because we're not
1657 * measuring them yet).
1658 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1660 * races with threads incrementing their own counters. But since word
1661 * reads are atomic, we either get new values or old values and we don't
1662 * care which for the sums. We always take the siglock to protect reading
1663 * the c* fields from p->signal from races with exit.c updating those
1664 * fields when reaping, so a sample either gets all the additions of a
1665 * given child after it's reaped, or none so this sample is before reaping.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001666 *
Ravikiran G Thirumalaide047c1b2006-06-22 14:47:26 -07001667 * Locking:
1668 * We need to take the siglock for CHILDEREN, SELF and BOTH
1669 * for the cases current multithreaded, non-current single threaded
1670 * non-current multithreaded. Thread traversal is now safe with
1671 * the siglock held.
1672 * Strictly speaking, we donot need to take the siglock if we are current and
1673 * single threaded, as no one else can take our signal_struct away, no one
1674 * else can reap the children to update signal->c* counters, and no one else
1675 * can race with the signal-> fields. If we do not take any lock, the
1676 * signal-> fields could be read out of order while another thread was just
1677 * exiting. So we should place a read memory barrier when we avoid the lock.
1678 * On the writer side, write memory barrier is implied in __exit_signal
1679 * as __exit_signal releases the siglock spinlock after updating the signal->
1680 * fields. But we don't do this yet to keep things simple.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001681 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 */
1683
Frank Mayharf06febc2008-09-12 09:54:39 -07001684static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001685{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001686 r->ru_nvcsw += t->nvcsw;
1687 r->ru_nivcsw += t->nivcsw;
1688 r->ru_minflt += t->min_flt;
1689 r->ru_majflt += t->maj_flt;
1690 r->ru_inblock += task_io_get_inblock(t);
1691 r->ru_oublock += task_io_get_oublock(t);
1692}
1693
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1695{
1696 struct task_struct *t;
1697 unsigned long flags;
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001698 cputime_t tgutime, tgstime, utime, stime;
Jiri Pirko1f102062009-09-22 16:44:10 -07001699 unsigned long maxrss = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700
1701 memset((char *) r, 0, sizeof *r);
Martin Schwidefsky64861632011-12-15 14:56:09 +01001702 utime = stime = 0;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001703
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001704 if (who == RUSAGE_THREAD) {
Hidetoshi Setod180c5b2009-11-26 14:48:30 +09001705 task_times(current, &utime, &stime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001706 accumulate_thread_rusage(p, r);
Jiri Pirko1f102062009-09-22 16:44:10 -07001707 maxrss = p->signal->maxrss;
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001708 goto out;
1709 }
1710
Oleg Nesterovd6cf7232008-04-30 00:52:38 -07001711 if (!lock_task_sighand(p, &flags))
Ravikiran G Thirumalaide047c1b2006-06-22 14:47:26 -07001712 return;
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001713
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 switch (who) {
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001715 case RUSAGE_BOTH:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 case RUSAGE_CHILDREN:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 utime = p->signal->cutime;
1718 stime = p->signal->cstime;
1719 r->ru_nvcsw = p->signal->cnvcsw;
1720 r->ru_nivcsw = p->signal->cnivcsw;
1721 r->ru_minflt = p->signal->cmin_flt;
1722 r->ru_majflt = p->signal->cmaj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001723 r->ru_inblock = p->signal->cinblock;
1724 r->ru_oublock = p->signal->coublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001725 maxrss = p->signal->cmaxrss;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001726
1727 if (who == RUSAGE_CHILDREN)
1728 break;
1729
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 case RUSAGE_SELF:
Hidetoshi Seto0cf55e12009-12-02 17:28:07 +09001731 thread_group_times(p, &tgutime, &tgstime);
Martin Schwidefsky64861632011-12-15 14:56:09 +01001732 utime += tgutime;
1733 stime += tgstime;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 r->ru_nvcsw += p->signal->nvcsw;
1735 r->ru_nivcsw += p->signal->nivcsw;
1736 r->ru_minflt += p->signal->min_flt;
1737 r->ru_majflt += p->signal->maj_flt;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -07001738 r->ru_inblock += p->signal->inblock;
1739 r->ru_oublock += p->signal->oublock;
Jiri Pirko1f102062009-09-22 16:44:10 -07001740 if (maxrss < p->signal->maxrss)
1741 maxrss = p->signal->maxrss;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 t = p;
1743 do {
Frank Mayharf06febc2008-09-12 09:54:39 -07001744 accumulate_thread_rusage(t, r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 t = next_thread(t);
1746 } while (t != p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 break;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001748
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 default:
1750 BUG();
1751 }
Ravikiran G Thirumalaide047c1b2006-06-22 14:47:26 -07001752 unlock_task_sighand(p, &flags);
Ravikiran G Thirumalaide047c1b2006-06-22 14:47:26 -07001753
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001754out:
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001755 cputime_to_timeval(utime, &r->ru_utime);
1756 cputime_to_timeval(stime, &r->ru_stime);
Jiri Pirko1f102062009-09-22 16:44:10 -07001757
1758 if (who != RUSAGE_CHILDREN) {
1759 struct mm_struct *mm = get_task_mm(p);
1760 if (mm) {
1761 setmax_mm_hiwater_rss(&maxrss, mm);
1762 mmput(mm);
1763 }
1764 }
1765 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766}
1767
1768int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1769{
1770 struct rusage r;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 k_getrusage(p, who, &r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1773}
1774
Heiko Carstense48fbb62009-01-14 14:14:26 +01001775SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001777 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1778 who != RUSAGE_THREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 return -EINVAL;
1780 return getrusage(current, who, ru);
1781}
1782
Heiko Carstense48fbb62009-01-14 14:14:26 +01001783SYSCALL_DEFINE1(umask, int, mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784{
1785 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1786 return mask;
1787}
Serge E. Hallyn3b7391d2008-02-04 22:29:45 -08001788
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001789#ifdef CONFIG_CHECKPOINT_RESTORE
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001790static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1791{
Al Viro2903ff02012-08-28 12:52:22 -04001792 struct fd exe;
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001793 struct dentry *dentry;
Al Viro2903ff02012-08-28 12:52:22 -04001794 int err;
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001795
Al Viro2903ff02012-08-28 12:52:22 -04001796 exe = fdget(fd);
1797 if (!exe.file)
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001798 return -EBADF;
1799
Al Viro2903ff02012-08-28 12:52:22 -04001800 dentry = exe.file->f_path.dentry;
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001801
1802 /*
1803 * Because the original mm->exe_file points to executable file, make
1804 * sure that this one is executable as well, to avoid breaking an
1805 * overall picture.
1806 */
1807 err = -EACCES;
1808 if (!S_ISREG(dentry->d_inode->i_mode) ||
Al Viro2903ff02012-08-28 12:52:22 -04001809 exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001810 goto exit;
1811
1812 err = inode_permission(dentry->d_inode, MAY_EXEC);
1813 if (err)
1814 goto exit;
1815
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001816 down_write(&mm->mmap_sem);
1817
1818 /*
Konstantin Khlebnikov4229fb1d2012-07-11 14:02:11 -07001819 * Forbid mm->exe_file change if old file still mapped.
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001820 */
1821 err = -EBUSY;
Konstantin Khlebnikov4229fb1d2012-07-11 14:02:11 -07001822 if (mm->exe_file) {
1823 struct vm_area_struct *vma;
1824
1825 for (vma = mm->mmap; vma; vma = vma->vm_next)
1826 if (vma->vm_file &&
1827 path_equal(&vma->vm_file->f_path,
1828 &mm->exe_file->f_path))
1829 goto exit_unlock;
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001830 }
1831
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001832 /*
1833 * The symlink can be changed only once, just to disallow arbitrary
1834 * transitions malicious software might bring in. This means one
1835 * could make a snapshot over all processes running and monitor
1836 * /proc/pid/exe changes to notice unusual activity if needed.
1837 */
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001838 err = -EPERM;
1839 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1840 goto exit_unlock;
1841
Konstantin Khlebnikov4229fb1d2012-07-11 14:02:11 -07001842 err = 0;
Al Viro2903ff02012-08-28 12:52:22 -04001843 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001844exit_unlock:
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001845 up_write(&mm->mmap_sem);
1846
1847exit:
Al Viro2903ff02012-08-28 12:52:22 -04001848 fdput(exe);
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001849 return err;
1850}
1851
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001852static int prctl_set_mm(int opt, unsigned long addr,
1853 unsigned long arg4, unsigned long arg5)
1854{
1855 unsigned long rlim = rlimit(RLIMIT_DATA);
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001856 struct mm_struct *mm = current->mm;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001857 struct vm_area_struct *vma;
1858 int error;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001859
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001860 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001861 return -EINVAL;
1862
Cyrill Gorcunov79f07132012-03-15 15:17:10 -07001863 if (!capable(CAP_SYS_RESOURCE))
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001864 return -EPERM;
1865
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001866 if (opt == PR_SET_MM_EXE_FILE)
1867 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1868
Cyrill Gorcunov1ad75b92012-06-07 14:21:11 -07001869 if (addr >= TASK_SIZE || addr < mmap_min_addr)
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001870 return -EINVAL;
1871
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001872 error = -EINVAL;
1873
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001874 down_read(&mm->mmap_sem);
1875 vma = find_vma(mm, addr);
1876
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001877 switch (opt) {
1878 case PR_SET_MM_START_CODE:
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001879 mm->start_code = addr;
1880 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001881 case PR_SET_MM_END_CODE:
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001882 mm->end_code = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001883 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001884 case PR_SET_MM_START_DATA:
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001885 mm->start_data = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001886 break;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001887 case PR_SET_MM_END_DATA:
1888 mm->end_data = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001889 break;
1890
1891 case PR_SET_MM_START_BRK:
1892 if (addr <= mm->end_data)
1893 goto out;
1894
1895 if (rlim < RLIM_INFINITY &&
1896 (mm->brk - addr) +
1897 (mm->end_data - mm->start_data) > rlim)
1898 goto out;
1899
1900 mm->start_brk = addr;
1901 break;
1902
1903 case PR_SET_MM_BRK:
1904 if (addr <= mm->end_data)
1905 goto out;
1906
1907 if (rlim < RLIM_INFINITY &&
1908 (addr - mm->start_brk) +
1909 (mm->end_data - mm->start_data) > rlim)
1910 goto out;
1911
1912 mm->brk = addr;
1913 break;
1914
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001915 /*
1916 * If command line arguments and environment
1917 * are placed somewhere else on stack, we can
1918 * set them up here, ARG_START/END to setup
1919 * command line argumets and ENV_START/END
1920 * for environment.
1921 */
1922 case PR_SET_MM_START_STACK:
1923 case PR_SET_MM_ARG_START:
1924 case PR_SET_MM_ARG_END:
1925 case PR_SET_MM_ENV_START:
1926 case PR_SET_MM_ENV_END:
1927 if (!vma) {
1928 error = -EFAULT;
1929 goto out;
1930 }
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001931 if (opt == PR_SET_MM_START_STACK)
1932 mm->start_stack = addr;
1933 else if (opt == PR_SET_MM_ARG_START)
1934 mm->arg_start = addr;
1935 else if (opt == PR_SET_MM_ARG_END)
1936 mm->arg_end = addr;
1937 else if (opt == PR_SET_MM_ENV_START)
1938 mm->env_start = addr;
1939 else if (opt == PR_SET_MM_ENV_END)
1940 mm->env_end = addr;
1941 break;
1942
1943 /*
1944 * This doesn't move auxiliary vector itself
1945 * since it's pinned to mm_struct, but allow
1946 * to fill vector with new values. It's up
1947 * to a caller to provide sane values here
1948 * otherwise user space tools which use this
1949 * vector might be unhappy.
1950 */
1951 case PR_SET_MM_AUXV: {
1952 unsigned long user_auxv[AT_VECTOR_SIZE];
1953
1954 if (arg4 > sizeof(user_auxv))
1955 goto out;
1956 up_read(&mm->mmap_sem);
1957
1958 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
1959 return -EFAULT;
1960
1961 /* Make sure the last entry is always AT_NULL */
1962 user_auxv[AT_VECTOR_SIZE - 2] = 0;
1963 user_auxv[AT_VECTOR_SIZE - 1] = 0;
1964
1965 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1966
1967 task_lock(current);
1968 memcpy(mm->saved_auxv, user_auxv, arg4);
1969 task_unlock(current);
1970
1971 return 0;
1972 }
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001973 default:
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001974 goto out;
1975 }
1976
1977 error = 0;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001978out:
1979 up_read(&mm->mmap_sem);
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001980 return error;
1981}
Cyrill Gorcunov300f7862012-06-07 14:21:12 -07001982
1983static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
1984{
1985 return put_user(me->clear_child_tid, tid_addr);
1986}
1987
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001988#else /* CONFIG_CHECKPOINT_RESTORE */
1989static int prctl_set_mm(int opt, unsigned long addr,
1990 unsigned long arg4, unsigned long arg5)
1991{
1992 return -EINVAL;
1993}
Cyrill Gorcunov300f7862012-06-07 14:21:12 -07001994static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
1995{
1996 return -EINVAL;
1997}
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001998#endif
1999
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01002000SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2001 unsigned long, arg4, unsigned long, arg5)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002{
David Howellsb6dff3e2008-11-14 10:39:16 +11002003 struct task_struct *me = current;
2004 unsigned char comm[sizeof(me->comm)];
2005 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006
David Howellsd84f4f92008-11-14 10:39:23 +11002007 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2008 if (error != -ENOSYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 return error;
2010
David Howellsd84f4f92008-11-14 10:39:23 +11002011 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012 switch (option) {
2013 case PR_SET_PDEATHSIG:
Jesper Juhl0730ded2005-09-06 15:17:37 -07002014 if (!valid_signal(arg2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 error = -EINVAL;
2016 break;
2017 }
David Howellsb6dff3e2008-11-14 10:39:16 +11002018 me->pdeath_signal = arg2;
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);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 break;
2033
2034 case PR_SET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002035 error = SET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 break;
2037 case PR_GET_UNALIGN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002038 error = GET_UNALIGN_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 break;
2040 case PR_SET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11002041 error = SET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 break;
2043 case PR_GET_FPEMU:
David Howellsb6dff3e2008-11-14 10:39:16 +11002044 error = GET_FPEMU_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 break;
2046 case PR_SET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11002047 error = SET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048 break;
2049 case PR_GET_FPEXC:
David Howellsb6dff3e2008-11-14 10:39:16 +11002050 error = GET_FPEXC_CTL(me, arg2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 break;
2052 case PR_GET_TIMING:
2053 error = PR_TIMING_STATISTICAL;
2054 break;
2055 case PR_SET_TIMING:
Shi Weihua7b266552008-05-23 13:04:59 -07002056 if (arg2 != PR_TIMING_STATISTICAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 error = -EINVAL;
2058 break;
David Howellsb6dff3e2008-11-14 10:39:16 +11002059 case PR_SET_NAME:
2060 comm[sizeof(me->comm)-1] = 0;
2061 if (strncpy_from_user(comm, (char __user *)arg2,
2062 sizeof(me->comm) - 1) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 return -EFAULT;
David Howellsb6dff3e2008-11-14 10:39:16 +11002064 set_task_comm(me, comm);
Vladimir Zapolskiyf786ecb2011-09-21 09:26:44 +00002065 proc_comm_connector(me);
Sasikantha babuf1fd75b2012-07-30 14:39:08 -07002066 break;
David Howellsb6dff3e2008-11-14 10:39:16 +11002067 case PR_GET_NAME:
2068 get_task_comm(comm, me);
2069 if (copy_to_user((char __user *)arg2, comm,
2070 sizeof(comm)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 return -EFAULT;
Sasikantha babuf1fd75b2012-07-30 14:39:08 -07002072 break;
Anton Blanchard651d7652006-06-07 16:10:19 +10002073 case PR_GET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002074 error = GET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10002075 break;
2076 case PR_SET_ENDIAN:
David Howellsb6dff3e2008-11-14 10:39:16 +11002077 error = SET_ENDIAN(me, arg2);
Anton Blanchard651d7652006-06-07 16:10:19 +10002078 break;
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -07002079 case PR_GET_SECCOMP:
2080 error = prctl_get_seccomp();
2081 break;
2082 case PR_SET_SECCOMP:
Will Drewrye2cfabdf2012-04-12 16:47:57 -05002083 error = prctl_set_seccomp(arg2, (char __user *)arg3);
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -07002084 break;
Erik Bosman8fb402b2008-04-11 18:54:17 +02002085 case PR_GET_TSC:
2086 error = GET_TSC_CTL(arg2);
2087 break;
2088 case PR_SET_TSC:
2089 error = SET_TSC_CTL(arg2);
2090 break;
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002091 case PR_TASK_PERF_EVENTS_DISABLE:
2092 error = perf_event_task_disable();
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01002093 break;
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002094 case PR_TASK_PERF_EVENTS_ENABLE:
2095 error = perf_event_task_enable();
Ingo Molnar1d1c7dd2008-12-11 14:59:31 +01002096 break;
Arjan van de Ven69766752008-09-01 15:52:40 -07002097 case PR_GET_TIMERSLACK:
2098 error = current->timer_slack_ns;
2099 break;
2100 case PR_SET_TIMERSLACK:
2101 if (arg2 <= 0)
2102 current->timer_slack_ns =
2103 current->default_timer_slack_ns;
2104 else
2105 current->timer_slack_ns = arg2;
2106 break;
Andi Kleen4db96cf2009-09-16 11:50:14 +02002107 case PR_MCE_KILL:
2108 if (arg4 | arg5)
2109 return -EINVAL;
2110 switch (arg2) {
Andi Kleen1087e9b2009-10-04 02:20:11 +02002111 case PR_MCE_KILL_CLEAR:
Andi Kleen4db96cf2009-09-16 11:50:14 +02002112 if (arg3 != 0)
2113 return -EINVAL;
2114 current->flags &= ~PF_MCE_PROCESS;
2115 break;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002116 case PR_MCE_KILL_SET:
Andi Kleen4db96cf2009-09-16 11:50:14 +02002117 current->flags |= PF_MCE_PROCESS;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002118 if (arg3 == PR_MCE_KILL_EARLY)
Andi Kleen4db96cf2009-09-16 11:50:14 +02002119 current->flags |= PF_MCE_EARLY;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002120 else if (arg3 == PR_MCE_KILL_LATE)
Andi Kleen4db96cf2009-09-16 11:50:14 +02002121 current->flags &= ~PF_MCE_EARLY;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002122 else if (arg3 == PR_MCE_KILL_DEFAULT)
2123 current->flags &=
2124 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2125 else
2126 return -EINVAL;
Andi Kleen4db96cf2009-09-16 11:50:14 +02002127 break;
2128 default:
2129 return -EINVAL;
2130 }
Andi Kleen4db96cf2009-09-16 11:50:14 +02002131 break;
Andi Kleen1087e9b2009-10-04 02:20:11 +02002132 case PR_MCE_KILL_GET:
2133 if (arg2 | arg3 | arg4 | arg5)
2134 return -EINVAL;
2135 if (current->flags & PF_MCE_PROCESS)
2136 error = (current->flags & PF_MCE_EARLY) ?
2137 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2138 else
2139 error = PR_MCE_KILL_DEFAULT;
2140 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002141 case PR_SET_MM:
2142 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2143 break;
Cyrill Gorcunov5702c5e2012-06-20 12:53:04 -07002144 case PR_GET_TID_ADDRESS:
2145 error = prctl_get_tid_address(me, (int __user **)arg2);
2146 break;
Lennart Poetteringebec18a2012-03-23 15:01:54 -07002147 case PR_SET_CHILD_SUBREAPER:
2148 me->signal->is_child_subreaper = !!arg2;
Lennart Poetteringebec18a2012-03-23 15:01:54 -07002149 break;
2150 case PR_GET_CHILD_SUBREAPER:
2151 error = put_user(me->signal->is_child_subreaper,
2152 (int __user *) arg2);
2153 break;
Andy Lutomirski259e5e62012-04-12 16:47:50 -05002154 case PR_SET_NO_NEW_PRIVS:
2155 if (arg2 != 1 || arg3 || arg4 || arg5)
2156 return -EINVAL;
2157
2158 current->no_new_privs = 1;
2159 break;
2160 case PR_GET_NO_NEW_PRIVS:
2161 if (arg2 || arg3 || arg4 || arg5)
2162 return -EINVAL;
2163 return current->no_new_privs ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164 default:
2165 error = -EINVAL;
2166 break;
2167 }
2168 return error;
2169}
Andi Kleen3cfc3482006-09-26 10:52:28 +02002170
Heiko Carstens836f92a2009-01-14 14:14:33 +01002171SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2172 struct getcpu_cache __user *, unused)
Andi Kleen3cfc3482006-09-26 10:52:28 +02002173{
2174 int err = 0;
2175 int cpu = raw_smp_processor_id();
2176 if (cpup)
2177 err |= put_user(cpu, cpup);
2178 if (nodep)
2179 err |= put_user(cpu_to_node(cpu), nodep);
Andi Kleen3cfc3482006-09-26 10:52:28 +02002180 return err ? -EFAULT : 0;
2181}
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002182
2183char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2184
Neil Hormana06a4dc2010-05-26 14:42:58 -07002185static void argv_cleanup(struct subprocess_info *info)
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002186{
Neil Hormana06a4dc2010-05-26 14:42:58 -07002187 argv_free(info->argv);
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002188}
2189
Andrew Mortonb57b44a2012-07-30 14:40:03 -07002190static int __orderly_poweroff(void)
2191{
2192 int argc;
2193 char **argv;
2194 static char *envp[] = {
2195 "HOME=/",
2196 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2197 NULL
2198 };
2199 int ret;
2200
2201 argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
2202 if (argv == NULL) {
2203 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2204 __func__, poweroff_cmd);
2205 return -ENOMEM;
2206 }
2207
hongfeng6c0c0d42012-10-04 17:12:25 -07002208 ret = call_usermodehelper_fns(argv[0], argv, envp, UMH_WAIT_EXEC,
Andrew Mortonb57b44a2012-07-30 14:40:03 -07002209 NULL, argv_cleanup, NULL);
2210 if (ret == -ENOMEM)
2211 argv_free(argv);
2212
2213 return ret;
2214}
2215
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002216/**
2217 * orderly_poweroff - Trigger an orderly system poweroff
2218 * @force: force poweroff if command execution fails
2219 *
2220 * This may be called from any context to trigger a system shutdown.
2221 * If the orderly shutdown fails, it will force an immediate shutdown.
2222 */
2223int orderly_poweroff(bool force)
2224{
Andrew Mortonb57b44a2012-07-30 14:40:03 -07002225 int ret = __orderly_poweroff();
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002226
Andrew Mortonb57b44a2012-07-30 14:40:03 -07002227 if (ret && force) {
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002228 printk(KERN_WARNING "Failed to start orderly shutdown: "
2229 "forcing the issue\n");
2230
Andrew Mortonb57b44a2012-07-30 14:40:03 -07002231 /*
2232 * I guess this should try to kick off some daemon to sync and
2233 * poweroff asap. Or not even bother syncing if we're doing an
2234 * emergency shutdown?
2235 */
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002236 emergency_sync();
2237 kernel_power_off();
2238 }
2239
2240 return ret;
2241}
2242EXPORT_SYMBOL_GPL(orderly_poweroff);