blob: 7ff6d1b10cecac8e66583f3bc233fd6fcb445b5c [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/workqueue.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080020#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/device.h>
22#include <linux/key.h>
23#include <linux/times.h>
24#include <linux/posix-timers.h>
25#include <linux/security.h>
26#include <linux/dcookies.h>
27#include <linux/suspend.h>
28#include <linux/tty.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070029#include <linux/signal.h>
Matt Helsley9f460802005-11-07 00:59:16 -080030#include <linux/cn_proc.h>
Andi Kleen3cfc3482006-09-26 10:52:28 +020031#include <linux/getcpu.h>
Eric Dumazet6eaeeab2007-05-10 22:22:37 -070032#include <linux/task_io_accounting_ops.h>
Andrea Arcangeli1d9d02f2007-07-15 23:41:32 -070033#include <linux/seccomp.h>
Mark Lord40477272007-10-01 01:20:10 -070034#include <linux/cpu.h>
Christoph Hellwige28cbf22010-03-10 15:21:19 -080035#include <linux/personality.h>
Paul Mackerrase3d5a272009-01-06 14:41:02 -080036#include <linux/ptrace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040037#include <linux/fs_struct.h>
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -070038#include <linux/file.h>
39#include <linux/mount.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090040#include <linux/gfp.h>
Rafael J. Wysocki40dc1662011-03-15 00:43:46 +010041#include <linux/syscore_ops.h>
Andi Kleenbe274252011-08-19 16:15:10 -070042#include <linux/version.h>
43#include <linux/ctype.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45#include <linux/compat.h>
46#include <linux/syscalls.h>
Keshavamurthy Anil S00d7c052005-12-12 00:37:33 -080047#include <linux/kprobes.h>
Cedric Le Goateracce2922007-07-15 23:40:59 -070048#include <linux/user_namespace.h>
Chen Gang7fe5e042013-02-21 16:43:06 -080049#include <linux/binfmts.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
Stephen Rothwell4a22f162013-04-30 15:27:37 -070051#include <linux/sched.h>
Ingo Molnar4eb5aaa2017-02-08 18:51:29 +010052#include <linux/sched/autogroup.h>
Ingo Molnar4f177222017-02-08 08:45:17 +010053#include <linux/sched/loadavg.h>
Ingo Molnar03441a32017-02-08 18:51:35 +010054#include <linux/sched/stat.h>
Ingo Molnar6e84f312017-02-08 18:51:29 +010055#include <linux/sched/mm.h>
Ingo Molnarf7ccbae2017-02-08 18:51:30 +010056#include <linux/sched/coredump.h>
Ingo Molnar29930022017-02-08 18:51:36 +010057#include <linux/sched/task.h>
Ingo Molnar32ef5512017-02-05 11:48:36 +010058#include <linux/sched/cputime.h>
Stephen Rothwell4a22f162013-04-30 15:27:37 -070059#include <linux/rcupdate.h>
60#include <linux/uidgid.h>
61#include <linux/cred.h>
62
Seiji Aguchi04c68622011-01-12 16:59:30 -080063#include <linux/kmsg_dump.h>
Andi Kleenbe274252011-08-19 16:15:10 -070064/* Move somewhere else to avoid recompiling? */
65#include <generated/utsrelease.h>
Seiji Aguchi04c68622011-01-12 16:59:30 -080066
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080067#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <asm/io.h>
69#include <asm/unistd.h>
70
71#ifndef SET_UNALIGN_CTL
vishnu.psec94fc32014-10-09 15:30:23 -070072# define SET_UNALIGN_CTL(a, b) (-EINVAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#endif
74#ifndef GET_UNALIGN_CTL
vishnu.psec94fc32014-10-09 15:30:23 -070075# define GET_UNALIGN_CTL(a, b) (-EINVAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#endif
77#ifndef SET_FPEMU_CTL
vishnu.psec94fc32014-10-09 15:30:23 -070078# define SET_FPEMU_CTL(a, b) (-EINVAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#endif
80#ifndef GET_FPEMU_CTL
vishnu.psec94fc32014-10-09 15:30:23 -070081# define GET_FPEMU_CTL(a, b) (-EINVAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#endif
83#ifndef SET_FPEXC_CTL
vishnu.psec94fc32014-10-09 15:30:23 -070084# define SET_FPEXC_CTL(a, b) (-EINVAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#endif
86#ifndef GET_FPEXC_CTL
vishnu.psec94fc32014-10-09 15:30:23 -070087# define GET_FPEXC_CTL(a, b) (-EINVAL)
Linus Torvalds1da177e2005-04-16 15:20:36 -070088#endif
Anton Blanchard651d7652006-06-07 16:10:19 +100089#ifndef GET_ENDIAN
vishnu.psec94fc32014-10-09 15:30:23 -070090# define GET_ENDIAN(a, b) (-EINVAL)
Anton Blanchard651d7652006-06-07 16:10:19 +100091#endif
92#ifndef SET_ENDIAN
vishnu.psec94fc32014-10-09 15:30:23 -070093# define SET_ENDIAN(a, b) (-EINVAL)
Anton Blanchard651d7652006-06-07 16:10:19 +100094#endif
Erik Bosman8fb402b2008-04-11 18:54:17 +020095#ifndef GET_TSC_CTL
96# define GET_TSC_CTL(a) (-EINVAL)
97#endif
98#ifndef SET_TSC_CTL
99# define SET_TSC_CTL(a) (-EINVAL)
100#endif
Dave Hansenfe3d1972014-11-14 07:18:29 -0800101#ifndef MPX_ENABLE_MANAGEMENT
Dave Hansen46a6e0c2015-06-07 11:37:02 -0700102# define MPX_ENABLE_MANAGEMENT() (-EINVAL)
Dave Hansenfe3d1972014-11-14 07:18:29 -0800103#endif
104#ifndef MPX_DISABLE_MANAGEMENT
Dave Hansen46a6e0c2015-06-07 11:37:02 -0700105# define MPX_DISABLE_MANAGEMENT() (-EINVAL)
Dave Hansenfe3d1972014-11-14 07:18:29 -0800106#endif
Paul Burton97915542015-01-08 12:17:37 +0000107#ifndef GET_FP_MODE
108# define GET_FP_MODE(a) (-EINVAL)
109#endif
110#ifndef SET_FP_MODE
111# define SET_FP_MODE(a,b) (-EINVAL)
112#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
114/*
115 * this is where the system-wide overflow UID and GID are defined, for
116 * architectures that now have 32-bit UID/GID but didn't in the past
117 */
118
119int overflowuid = DEFAULT_OVERFLOWUID;
120int overflowgid = DEFAULT_OVERFLOWGID;
121
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122EXPORT_SYMBOL(overflowuid);
123EXPORT_SYMBOL(overflowgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
125/*
126 * the same as above, but for filesystems which can only store a 16-bit
127 * UID and GID. as such, this is needed on all architectures
128 */
129
130int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
131int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
132
133EXPORT_SYMBOL(fs_overflowuid);
134EXPORT_SYMBOL(fs_overflowgid);
135
136/*
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700137 * Returns true if current's euid is same as p's uid or euid,
138 * or has CAP_SYS_NICE to p's user_ns.
139 *
140 * Called with rcu_read_lock, creds are safe
141 */
142static bool set_one_prio_perm(struct task_struct *p)
143{
144 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
145
Eric W. Biederman5af66202012-03-03 20:21:47 -0800146 if (uid_eq(pcred->uid, cred->euid) ||
147 uid_eq(pcred->euid, cred->euid))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700148 return true;
Eric W. Biedermanc4a4d602011-11-16 23:15:31 -0800149 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700150 return true;
151 return false;
152}
153
154/*
David Howellsc69e8d92008-11-14 10:39:19 +1100155 * set the priority of a task
156 * - the caller must hold the RCU read lock
157 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158static int set_one_prio(struct task_struct *p, int niceval, int error)
159{
160 int no_nice;
161
Serge E. Hallynfc832ad2011-03-23 16:43:22 -0700162 if (!set_one_prio_perm(p)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 error = -EPERM;
164 goto out;
165 }
Matt Mackalle43379f2005-05-01 08:59:00 -0700166 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 error = -EACCES;
168 goto out;
169 }
170 no_nice = security_task_setnice(p, niceval);
171 if (no_nice) {
172 error = no_nice;
173 goto out;
174 }
175 if (error == -ESRCH)
176 error = 0;
177 set_user_nice(p, niceval);
178out:
179 return error;
180}
181
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100182SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183{
184 struct task_struct *g, *p;
185 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100186 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 int error = -EINVAL;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800188 struct pid *pgrp;
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800189 kuid_t uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
Daniel Walker3e88c552007-05-10 22:22:53 -0700191 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 goto out;
193
194 /* normalize: avoid signed division (rounding problems) */
195 error = -ESRCH;
Dongsheng Yangc4a4d2f2014-02-11 15:34:51 +0800196 if (niceval < MIN_NICE)
197 niceval = MIN_NICE;
198 if (niceval > MAX_NICE)
199 niceval = MAX_NICE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200
Thomas Gleixnerd4581a22009-12-10 00:52:51 +0000201 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 read_lock(&tasklist_lock);
203 switch (which) {
vishnu.psec94fc32014-10-09 15:30:23 -0700204 case PRIO_PROCESS:
205 if (who)
206 p = find_task_by_vpid(who);
207 else
208 p = current;
209 if (p)
210 error = set_one_prio(p, niceval, error);
211 break;
212 case PRIO_PGRP:
213 if (who)
214 pgrp = find_vpid(who);
215 else
216 pgrp = task_pgrp(current);
217 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
218 error = set_one_prio(p, niceval, error);
219 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
220 break;
221 case PRIO_USER:
222 uid = make_kuid(cred->user_ns, who);
223 user = cred->user;
224 if (!who)
225 uid = cred->uid;
226 else if (!uid_eq(uid, cred->uid)) {
227 user = find_user(uid);
228 if (!user)
David Howells86a264a2008-11-14 10:39:18 +1100229 goto out_unlock; /* No processes for this user */
vishnu.psec94fc32014-10-09 15:30:23 -0700230 }
231 do_each_thread(g, p) {
Ben Segall8639b462015-11-06 16:32:48 -0800232 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
vishnu.psec94fc32014-10-09 15:30:23 -0700233 error = set_one_prio(p, niceval, error);
234 } while_each_thread(g, p);
235 if (!uid_eq(uid, cred->uid))
236 free_uid(user); /* For find_user() */
237 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 }
239out_unlock:
240 read_unlock(&tasklist_lock);
Thomas Gleixnerd4581a22009-12-10 00:52:51 +0000241 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242out:
243 return error;
244}
245
246/*
247 * Ugh. To avoid negative return values, "getpriority()" will
248 * not return the normal nice-value, but a negated value that
249 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
250 * to stay compatible.
251 */
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100252SYSCALL_DEFINE2(getpriority, int, which, int, who)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253{
254 struct task_struct *g, *p;
255 struct user_struct *user;
David Howells86a264a2008-11-14 10:39:18 +1100256 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 long niceval, retval = -ESRCH;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800258 struct pid *pgrp;
Eric W. Biederman7b44ab92011-11-16 23:20:58 -0800259 kuid_t uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
Daniel Walker3e88c552007-05-10 22:22:53 -0700261 if (which > PRIO_USER || which < PRIO_PROCESS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 return -EINVAL;
263
Tetsuo Handa70118832010-02-22 12:44:16 -0800264 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 read_lock(&tasklist_lock);
266 switch (which) {
vishnu.psec94fc32014-10-09 15:30:23 -0700267 case PRIO_PROCESS:
268 if (who)
269 p = find_task_by_vpid(who);
270 else
271 p = current;
272 if (p) {
273 niceval = nice_to_rlimit(task_nice(p));
274 if (niceval > retval)
275 retval = niceval;
276 }
277 break;
278 case PRIO_PGRP:
279 if (who)
280 pgrp = find_vpid(who);
281 else
282 pgrp = task_pgrp(current);
283 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
284 niceval = nice_to_rlimit(task_nice(p));
285 if (niceval > retval)
286 retval = niceval;
287 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
288 break;
289 case PRIO_USER:
290 uid = make_kuid(cred->user_ns, who);
291 user = cred->user;
292 if (!who)
293 uid = cred->uid;
294 else if (!uid_eq(uid, cred->uid)) {
295 user = find_user(uid);
296 if (!user)
297 goto out_unlock; /* No processes for this user */
298 }
299 do_each_thread(g, p) {
Ben Segall8639b462015-11-06 16:32:48 -0800300 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
Dongsheng Yang7aa2c012014-05-08 18:33:49 +0900301 niceval = nice_to_rlimit(task_nice(p));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 if (niceval > retval)
303 retval = niceval;
304 }
vishnu.psec94fc32014-10-09 15:30:23 -0700305 } while_each_thread(g, p);
306 if (!uid_eq(uid, cred->uid))
307 free_uid(user); /* for find_user() */
308 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 }
310out_unlock:
311 read_unlock(&tasklist_lock);
Tetsuo Handa70118832010-02-22 12:44:16 -0800312 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314 return retval;
315}
316
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317/*
318 * Unprivileged users may change the real gid to the effective gid
319 * or vice versa. (BSD-style)
320 *
321 * If you set the real gid at all, or set the effective gid to a value not
322 * equal to the real gid, then the saved gid is set to the new effective gid.
323 *
324 * This makes it possible for a setgid program to completely drop its
325 * privileges, which is often a useful assertion to make when you are doing
326 * a security audit over a program.
327 *
328 * The general idea is that a program which uses just setregid() will be
329 * 100% compatible with BSD. A program which uses just setgid() will be
vishnu.psec94fc32014-10-09 15:30:23 -0700330 * 100% compatible with POSIX with saved IDs.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 *
332 * SMP: There are not races, the GIDs are checked only by filesystem
333 * operations (as far as semantic preservation is concerned).
334 */
Iulia Manda28138932015-04-15 16:16:41 -0700335#ifdef CONFIG_MULTIUSER
Heiko Carstensae1251a2009-01-14 14:14:05 +0100336SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800338 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100339 const struct cred *old;
340 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800342 kgid_t krgid, kegid;
343
344 krgid = make_kgid(ns, rgid);
345 kegid = make_kgid(ns, egid);
346
347 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
348 return -EINVAL;
349 if ((egid != (gid_t) -1) && !gid_valid(kegid))
350 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
David Howellsd84f4f92008-11-14 10:39:23 +1100352 new = prepare_creds();
353 if (!new)
354 return -ENOMEM;
355 old = current_cred();
356
David Howellsd84f4f92008-11-14 10:39:23 +1100357 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 if (rgid != (gid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800359 if (gid_eq(old->gid, krgid) ||
360 gid_eq(old->egid, krgid) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700361 ns_capable(old->user_ns, CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800362 new->gid = krgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 else
David Howellsd84f4f92008-11-14 10:39:23 +1100364 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 }
366 if (egid != (gid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800367 if (gid_eq(old->gid, kegid) ||
368 gid_eq(old->egid, kegid) ||
369 gid_eq(old->sgid, kegid) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700370 ns_capable(old->user_ns, CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800371 new->egid = kegid;
Cal Peake756184b2006-09-30 23:27:24 -0700372 else
David Howellsd84f4f92008-11-14 10:39:23 +1100373 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 }
David Howellsd84f4f92008-11-14 10:39:23 +1100375
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 if (rgid != (gid_t) -1 ||
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800377 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
David Howellsd84f4f92008-11-14 10:39:23 +1100378 new->sgid = new->egid;
379 new->fsgid = new->egid;
380
381 return commit_creds(new);
382
383error:
384 abort_creds(new);
385 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386}
387
388/*
vishnu.psec94fc32014-10-09 15:30:23 -0700389 * setgid() is implemented like SysV w/ SAVED_IDS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 *
391 * SMP: Same implicit races as above.
392 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100393SYSCALL_DEFINE1(setgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800395 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100396 const struct cred *old;
397 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800399 kgid_t kgid;
400
401 kgid = make_kgid(ns, gid);
402 if (!gid_valid(kgid))
403 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
David Howellsd84f4f92008-11-14 10:39:23 +1100405 new = prepare_creds();
406 if (!new)
407 return -ENOMEM;
408 old = current_cred();
409
David Howellsd84f4f92008-11-14 10:39:23 +1100410 retval = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700411 if (ns_capable(old->user_ns, CAP_SETGID))
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800412 new->gid = new->egid = new->sgid = new->fsgid = kgid;
413 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
414 new->egid = new->fsgid = kgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 else
David Howellsd84f4f92008-11-14 10:39:23 +1100416 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
David Howellsd84f4f92008-11-14 10:39:23 +1100418 return commit_creds(new);
419
420error:
421 abort_creds(new);
422 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423}
Dhaval Giani54e99122009-02-27 15:13:54 +0530424
David Howellsd84f4f92008-11-14 10:39:23 +1100425/*
426 * change the user struct in a credentials set to match the new UID
427 */
428static int set_user(struct cred *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429{
430 struct user_struct *new_user;
431
Eric W. Biederman078de5f2012-02-08 07:00:08 -0800432 new_user = alloc_uid(new->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 if (!new_user)
434 return -EAGAIN;
435
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400436 /*
437 * We don't fail in case of NPROC limit excess here because too many
438 * poorly written programs don't check set*uid() return code, assuming
439 * it never fails if called by root. We may still enforce NPROC limit
440 * for programs doing set*uid()+execve() by harmlessly deferring the
441 * failure to the execve() stage.
442 */
Jiri Slaby78d7d402010-03-05 13:42:54 -0800443 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +0400444 new_user != INIT_USER)
445 current->flags |= PF_NPROC_EXCEEDED;
446 else
447 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
David Howellsd84f4f92008-11-14 10:39:23 +1100449 free_uid(new->user);
450 new->user = new_user;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 return 0;
452}
453
454/*
455 * Unprivileged users may change the real uid to the effective uid
456 * or vice versa. (BSD-style)
457 *
458 * If you set the real uid at all, or set the effective uid to a value not
459 * equal to the real uid, then the saved uid is set to the new effective uid.
460 *
461 * This makes it possible for a setuid program to completely drop its
462 * privileges, which is often a useful assertion to make when you are doing
463 * a security audit over a program.
464 *
465 * The general idea is that a program which uses just setreuid() will be
466 * 100% compatible with BSD. A program which uses just setuid() will be
vishnu.psec94fc32014-10-09 15:30:23 -0700467 * 100% compatible with POSIX with saved IDs.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100469SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800471 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100472 const struct cred *old;
473 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800475 kuid_t kruid, keuid;
476
477 kruid = make_kuid(ns, ruid);
478 keuid = make_kuid(ns, euid);
479
480 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
481 return -EINVAL;
482 if ((euid != (uid_t) -1) && !uid_valid(keuid))
483 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
David Howellsd84f4f92008-11-14 10:39:23 +1100485 new = prepare_creds();
486 if (!new)
487 return -ENOMEM;
488 old = current_cred();
489
David Howellsd84f4f92008-11-14 10:39:23 +1100490 retval = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 if (ruid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800492 new->uid = kruid;
493 if (!uid_eq(old->uid, kruid) &&
494 !uid_eq(old->euid, kruid) &&
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700495 !ns_capable(old->user_ns, CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100496 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 }
498
499 if (euid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800500 new->euid = keuid;
501 if (!uid_eq(old->uid, keuid) &&
502 !uid_eq(old->euid, keuid) &&
503 !uid_eq(old->suid, keuid) &&
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700504 !ns_capable(old->user_ns, CAP_SETUID))
David Howellsd84f4f92008-11-14 10:39:23 +1100505 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 }
507
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800508 if (!uid_eq(new->uid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530509 retval = set_user(new);
510 if (retval < 0)
511 goto error;
512 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 if (ruid != (uid_t) -1 ||
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800514 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
David Howellsd84f4f92008-11-14 10:39:23 +1100515 new->suid = new->euid;
516 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
David Howellsd84f4f92008-11-14 10:39:23 +1100518 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
519 if (retval < 0)
520 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
David Howellsd84f4f92008-11-14 10:39:23 +1100522 return commit_creds(new);
523
524error:
525 abort_creds(new);
526 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527}
vishnu.psec94fc32014-10-09 15:30:23 -0700528
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529/*
vishnu.psec94fc32014-10-09 15:30:23 -0700530 * setuid() is implemented like SysV with SAVED_IDS
531 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 * Note that SAVED_ID's is deficient in that a setuid root program
vishnu.psec94fc32014-10-09 15:30:23 -0700533 * like sendmail, for example, cannot set its uid to be a normal
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 * user and then switch back, because if you're root, setuid() sets
535 * the saved uid too. If you don't like this, blame the bright people
536 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
537 * will allow a root program to temporarily drop privileges and be able to
vishnu.psec94fc32014-10-09 15:30:23 -0700538 * regain them by swapping the real and effective uid.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100540SYSCALL_DEFINE1(setuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800542 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100543 const struct cred *old;
544 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800546 kuid_t kuid;
547
548 kuid = make_kuid(ns, uid);
549 if (!uid_valid(kuid))
550 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
David Howellsd84f4f92008-11-14 10:39:23 +1100552 new = prepare_creds();
553 if (!new)
554 return -ENOMEM;
555 old = current_cred();
556
David Howellsd84f4f92008-11-14 10:39:23 +1100557 retval = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700558 if (ns_capable(old->user_ns, CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800559 new->suid = new->uid = kuid;
560 if (!uid_eq(kuid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530561 retval = set_user(new);
562 if (retval < 0)
563 goto error;
David Howellsd84f4f92008-11-14 10:39:23 +1100564 }
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800565 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
David Howellsd84f4f92008-11-14 10:39:23 +1100566 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800569 new->fsuid = new->euid = kuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
David Howellsd84f4f92008-11-14 10:39:23 +1100571 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
572 if (retval < 0)
573 goto error;
574
575 return commit_creds(new);
576
577error:
578 abort_creds(new);
579 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580}
581
582
583/*
584 * This function implements a generic ability to update ruid, euid,
585 * and suid. This allows you to implement the 4.4 compatible seteuid().
586 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100587SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800589 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100590 const struct cred *old;
591 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800593 kuid_t kruid, keuid, ksuid;
594
595 kruid = make_kuid(ns, ruid);
596 keuid = make_kuid(ns, euid);
597 ksuid = make_kuid(ns, suid);
598
599 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
600 return -EINVAL;
601
602 if ((euid != (uid_t) -1) && !uid_valid(keuid))
603 return -EINVAL;
604
605 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
606 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
David Howellsd84f4f92008-11-14 10:39:23 +1100608 new = prepare_creds();
609 if (!new)
610 return -ENOMEM;
611
David Howellsd84f4f92008-11-14 10:39:23 +1100612 old = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
David Howellsd84f4f92008-11-14 10:39:23 +1100614 retval = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700615 if (!ns_capable(old->user_ns, CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800616 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
617 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100618 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800619 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
620 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100621 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800622 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
623 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
David Howellsd84f4f92008-11-14 10:39:23 +1100624 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 }
David Howellsd84f4f92008-11-14 10:39:23 +1100626
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 if (ruid != (uid_t) -1) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800628 new->uid = kruid;
629 if (!uid_eq(kruid, old->uid)) {
Dhaval Giani54e99122009-02-27 15:13:54 +0530630 retval = set_user(new);
631 if (retval < 0)
632 goto error;
633 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 }
David Howellsd84f4f92008-11-14 10:39:23 +1100635 if (euid != (uid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800636 new->euid = keuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 if (suid != (uid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800638 new->suid = ksuid;
David Howellsd84f4f92008-11-14 10:39:23 +1100639 new->fsuid = new->euid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640
David Howellsd84f4f92008-11-14 10:39:23 +1100641 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
642 if (retval < 0)
643 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
David Howellsd84f4f92008-11-14 10:39:23 +1100645 return commit_creds(new);
646
647error:
648 abort_creds(new);
649 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650}
651
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800652SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653{
David Howells86a264a2008-11-14 10:39:18 +1100654 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800656 uid_t ruid, euid, suid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800658 ruid = from_kuid_munged(cred->user_ns, cred->uid);
659 euid = from_kuid_munged(cred->user_ns, cred->euid);
660 suid = from_kuid_munged(cred->user_ns, cred->suid);
661
vishnu.psec94fc32014-10-09 15:30:23 -0700662 retval = put_user(ruid, ruidp);
663 if (!retval) {
664 retval = put_user(euid, euidp);
665 if (!retval)
666 return put_user(suid, suidp);
667 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 return retval;
669}
670
671/*
672 * Same as above, but for rgid, egid, sgid.
673 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100674SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675{
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800676 struct user_namespace *ns = current_user_ns();
David Howellsd84f4f92008-11-14 10:39:23 +1100677 const struct cred *old;
678 struct cred *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800680 kgid_t krgid, kegid, ksgid;
681
682 krgid = make_kgid(ns, rgid);
683 kegid = make_kgid(ns, egid);
684 ksgid = make_kgid(ns, sgid);
685
686 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
687 return -EINVAL;
688 if ((egid != (gid_t) -1) && !gid_valid(kegid))
689 return -EINVAL;
690 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
691 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
David Howellsd84f4f92008-11-14 10:39:23 +1100693 new = prepare_creds();
694 if (!new)
695 return -ENOMEM;
696 old = current_cred();
697
David Howellsd84f4f92008-11-14 10:39:23 +1100698 retval = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700699 if (!ns_capable(old->user_ns, CAP_SETGID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800700 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
701 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100702 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800703 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
704 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100705 goto error;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800706 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
707 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
David Howellsd84f4f92008-11-14 10:39:23 +1100708 goto error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710
David Howellsd84f4f92008-11-14 10:39:23 +1100711 if (rgid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800712 new->gid = krgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100713 if (egid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800714 new->egid = kegid;
David Howellsd84f4f92008-11-14 10:39:23 +1100715 if (sgid != (gid_t) -1)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800716 new->sgid = ksgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100717 new->fsgid = new->egid;
718
719 return commit_creds(new);
720
721error:
722 abort_creds(new);
723 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724}
725
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800726SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727{
David Howells86a264a2008-11-14 10:39:18 +1100728 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 int retval;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800730 gid_t rgid, egid, sgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800732 rgid = from_kgid_munged(cred->user_ns, cred->gid);
733 egid = from_kgid_munged(cred->user_ns, cred->egid);
734 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
735
vishnu.psec94fc32014-10-09 15:30:23 -0700736 retval = put_user(rgid, rgidp);
737 if (!retval) {
738 retval = put_user(egid, egidp);
739 if (!retval)
740 retval = put_user(sgid, sgidp);
741 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742
743 return retval;
744}
745
746
747/*
748 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
749 * is used for "access()" and for the NFS daemon (letting nfsd stay at
750 * whatever uid it wants to). It normally shadows "euid", except when
751 * explicitly set by setfsuid() or for access..
752 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100753SYSCALL_DEFINE1(setfsuid, uid_t, uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754{
David Howellsd84f4f92008-11-14 10:39:23 +1100755 const struct cred *old;
756 struct cred *new;
757 uid_t old_fsuid;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800758 kuid_t kuid;
759
760 old = current_cred();
761 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
762
763 kuid = make_kuid(old->user_ns, uid);
764 if (!uid_valid(kuid))
765 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
David Howellsd84f4f92008-11-14 10:39:23 +1100767 new = prepare_creds();
768 if (!new)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800769 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800771 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
772 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700773 ns_capable(old->user_ns, CAP_SETUID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800774 if (!uid_eq(kuid, old->fsuid)) {
775 new->fsuid = kuid;
David Howellsd84f4f92008-11-14 10:39:23 +1100776 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
777 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 }
780
David Howellsd84f4f92008-11-14 10:39:23 +1100781 abort_creds(new);
782 return old_fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
David Howellsd84f4f92008-11-14 10:39:23 +1100784change_okay:
785 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 return old_fsuid;
787}
788
789/*
John Anthony Kazos Jrf42df9e2007-05-09 08:23:08 +0200790 * Samma på svenska..
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 */
Heiko Carstensae1251a2009-01-14 14:14:05 +0100792SYSCALL_DEFINE1(setfsgid, gid_t, gid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793{
David Howellsd84f4f92008-11-14 10:39:23 +1100794 const struct cred *old;
795 struct cred *new;
796 gid_t old_fsgid;
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800797 kgid_t kgid;
798
799 old = current_cred();
800 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
801
802 kgid = make_kgid(old->user_ns, gid);
803 if (!gid_valid(kgid))
804 return old_fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
David Howellsd84f4f92008-11-14 10:39:23 +1100806 new = prepare_creds();
807 if (!new)
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800808 return old_fsgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100809
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800810 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
811 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700812 ns_capable(old->user_ns, CAP_SETGID)) {
Eric W. Biedermana29c33f2012-02-07 18:51:01 -0800813 if (!gid_eq(kgid, old->fsgid)) {
814 new->fsgid = kgid;
David Howellsd84f4f92008-11-14 10:39:23 +1100815 goto change_okay;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 }
David Howellsd84f4f92008-11-14 10:39:23 +1100818
David Howellsd84f4f92008-11-14 10:39:23 +1100819 abort_creds(new);
820 return old_fsgid;
821
822change_okay:
823 commit_creds(new);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 return old_fsgid;
825}
Iulia Manda28138932015-04-15 16:16:41 -0700826#endif /* CONFIG_MULTIUSER */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
Stephen Rothwell4a22f162013-04-30 15:27:37 -0700828/**
829 * sys_getpid - return the thread group id of the current process
830 *
831 * Note, despite the name, this returns the tgid not the pid. The tgid and
832 * the pid are identical unless CLONE_THREAD was specified on clone() in
833 * which case the tgid is the same in all threads of the same group.
834 *
835 * This is SMP safe as current->tgid does not change.
836 */
837SYSCALL_DEFINE0(getpid)
838{
839 return task_tgid_vnr(current);
840}
841
842/* Thread ID - the internal kernel "pid" */
843SYSCALL_DEFINE0(gettid)
844{
845 return task_pid_vnr(current);
846}
847
848/*
849 * Accessing ->real_parent is not SMP-safe, it could
850 * change from under us. However, we can use a stale
851 * value of ->real_parent under rcu_read_lock(), see
852 * release_task()->call_rcu(delayed_put_task_struct).
853 */
854SYSCALL_DEFINE0(getppid)
855{
856 int pid;
857
858 rcu_read_lock();
859 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
860 rcu_read_unlock();
861
862 return pid;
863}
864
865SYSCALL_DEFINE0(getuid)
866{
867 /* Only we change this so SMP safe */
868 return from_kuid_munged(current_user_ns(), current_uid());
869}
870
871SYSCALL_DEFINE0(geteuid)
872{
873 /* Only we change this so SMP safe */
874 return from_kuid_munged(current_user_ns(), current_euid());
875}
876
877SYSCALL_DEFINE0(getgid)
878{
879 /* Only we change this so SMP safe */
880 return from_kgid_munged(current_user_ns(), current_gid());
881}
882
883SYSCALL_DEFINE0(getegid)
884{
885 /* Only we change this so SMP safe */
886 return from_kgid_munged(current_user_ns(), current_egid());
887}
888
Frank Mayharf06febc2008-09-12 09:54:39 -0700889void do_sys_times(struct tms *tms)
890{
Frederic Weisbecker5613fda2017-01-31 04:09:23 +0100891 u64 tgutime, tgstime, cutime, cstime;
Frank Mayharf06febc2008-09-12 09:54:39 -0700892
Frederic Weisbeckere80d0a1a2012-11-21 16:26:44 +0100893 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
Frank Mayharf06febc2008-09-12 09:54:39 -0700894 cutime = current->signal->cutime;
895 cstime = current->signal->cstime;
Frederic Weisbecker5613fda2017-01-31 04:09:23 +0100896 tms->tms_utime = nsec_to_clock_t(tgutime);
897 tms->tms_stime = nsec_to_clock_t(tgstime);
898 tms->tms_cutime = nsec_to_clock_t(cutime);
899 tms->tms_cstime = nsec_to_clock_t(cstime);
Frank Mayharf06febc2008-09-12 09:54:39 -0700900}
901
Heiko Carstens58fd3aa2009-01-14 14:14:03 +0100902SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 if (tbuf) {
905 struct tms tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
Frank Mayharf06febc2008-09-12 09:54:39 -0700907 do_sys_times(&tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
909 return -EFAULT;
910 }
Paul Mackerrase3d5a272009-01-06 14:41:02 -0800911 force_successful_syscall_return();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 return (long) jiffies_64_to_clock_t(get_jiffies_64());
913}
914
915/*
916 * This needs some heavy checking ...
917 * I just haven't the stomach for it. I also don't fully
918 * understand sessions/pgrp etc. Let somebody who does explain it.
919 *
920 * OK, I think I have the protection semantics right.... this is really
921 * only important on a multi-user system anyway, to make sure one user
922 * can't send a signal to a process owned by another. -TYT, 12/12/91
923 *
Oleg Nesterov98611e42014-01-23 15:55:52 -0800924 * !PF_FORKNOEXEC check to conform completely to POSIX.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 */
Heiko Carstensb290ebe2009-01-14 14:14:06 +0100926SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927{
928 struct task_struct *p;
Oleg Nesterovee0acf92006-01-08 01:03:53 -0800929 struct task_struct *group_leader = current->group_leader;
Oleg Nesterov4e021302008-02-08 04:19:08 -0800930 struct pid *pgrp;
931 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932
933 if (!pid)
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700934 pid = task_pid_vnr(group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 if (!pgid)
936 pgid = pid;
937 if (pgid < 0)
938 return -EINVAL;
Paul E. McKenney950eaac2010-08-31 17:00:18 -0700939 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940
941 /* From this point forward we keep holding onto the tasklist lock
942 * so that our parent does not change from under us. -DaveM
943 */
944 write_lock_irq(&tasklist_lock);
945
946 err = -ESRCH;
Oleg Nesterov4e021302008-02-08 04:19:08 -0800947 p = find_task_by_vpid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 if (!p)
949 goto out;
950
951 err = -EINVAL;
952 if (!thread_group_leader(p))
953 goto out;
954
Oleg Nesterov4e021302008-02-08 04:19:08 -0800955 if (same_thread_group(p->real_parent, group_leader)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 err = -EPERM;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800957 if (task_session(p) != task_session(group_leader))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 goto out;
959 err = -EACCES;
Oleg Nesterov98611e42014-01-23 15:55:52 -0800960 if (!(p->flags & PF_FORKNOEXEC))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 goto out;
962 } else {
963 err = -ESRCH;
Oleg Nesterovee0acf92006-01-08 01:03:53 -0800964 if (p != group_leader)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 goto out;
966 }
967
968 err = -EPERM;
969 if (p->signal->leader)
970 goto out;
971
Oleg Nesterov4e021302008-02-08 04:19:08 -0800972 pgrp = task_pid(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 if (pgid != pid) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700974 struct task_struct *g;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
Oleg Nesterov4e021302008-02-08 04:19:08 -0800976 pgrp = find_vpid(pgid);
977 g = pid_task(pgrp, PIDTYPE_PGID);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800978 if (!g || task_session(g) != task_session(group_leader))
Oleg Nesterovf020bc42006-12-08 02:38:02 -0800979 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 }
981
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 err = security_task_setpgid(p, pgid);
983 if (err)
984 goto out;
985
Oleg Nesterov1b0f7ff2009-04-02 16:58:39 -0700986 if (task_pgrp(p) != pgrp)
Oleg Nesterov83beaf32008-04-30 00:54:27 -0700987 change_pid(p, PIDTYPE_PGID, pgrp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988
989 err = 0;
990out:
991 /* All paths lead to here, thus we are safe. -DaveM */
992 write_unlock_irq(&tasklist_lock);
Paul E. McKenney950eaac2010-08-31 17:00:18 -0700993 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 return err;
995}
996
Heiko Carstensdbf040d2009-01-14 14:14:04 +0100997SYSCALL_DEFINE1(getpgid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -0700999 struct task_struct *p;
1000 struct pid *grp;
1001 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001003 rcu_read_lock();
1004 if (!pid)
1005 grp = task_pgrp(current);
1006 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 retval = -ESRCH;
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001008 p = find_task_by_vpid(pid);
1009 if (!p)
1010 goto out;
1011 grp = task_pgrp(p);
1012 if (!grp)
1013 goto out;
1014
1015 retval = security_task_getpgid(p);
1016 if (retval)
1017 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 }
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001019 retval = pid_vnr(grp);
1020out:
1021 rcu_read_unlock();
1022 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023}
1024
1025#ifdef __ARCH_WANT_SYS_GETPGRP
1026
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001027SYSCALL_DEFINE0(getpgrp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028{
Oleg Nesterov12a3de0a2008-04-30 00:54:29 -07001029 return sys_getpgid(0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030}
1031
1032#endif
1033
Heiko Carstensdbf040d2009-01-14 14:14:04 +01001034SYSCALL_DEFINE1(getsid, pid_t, pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035{
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001036 struct task_struct *p;
1037 struct pid *sid;
1038 int retval;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001039
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001040 rcu_read_lock();
1041 if (!pid)
1042 sid = task_session(current);
1043 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 retval = -ESRCH;
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001045 p = find_task_by_vpid(pid);
1046 if (!p)
1047 goto out;
1048 sid = task_session(p);
1049 if (!sid)
1050 goto out;
1051
1052 retval = security_task_getsid(p);
1053 if (retval)
1054 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 }
Oleg Nesterov1dd768c02008-04-30 00:54:28 -07001056 retval = pid_vnr(sid);
1057out:
1058 rcu_read_unlock();
1059 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060}
1061
Oleg Nesterov81dabb42013-07-03 15:08:26 -07001062static void set_special_pids(struct pid *pid)
1063{
1064 struct task_struct *curr = current->group_leader;
1065
1066 if (task_session(curr) != pid)
1067 change_pid(curr, PIDTYPE_SID, pid);
1068
1069 if (task_pgrp(curr) != pid)
1070 change_pid(curr, PIDTYPE_PGID, pid);
1071}
1072
Heiko Carstensb290ebe2009-01-14 14:14:06 +01001073SYSCALL_DEFINE0(setsid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
Oren Laadane19f2472006-01-08 01:03:58 -08001075 struct task_struct *group_leader = current->group_leader;
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001076 struct pid *sid = task_pid(group_leader);
1077 pid_t session = pid_vnr(sid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 int err = -EPERM;
1079
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 write_lock_irq(&tasklist_lock);
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001081 /* Fail if I am already a session leader */
1082 if (group_leader->signal->leader)
1083 goto out;
1084
Oleg Nesterov430c6232008-02-08 04:19:11 -08001085 /* Fail if a process group id already exists that equals the
1086 * proposed session id.
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001087 */
Oleg Nesterov6806aac2008-02-08 04:19:12 -08001088 if (pid_task(sid, PIDTYPE_PGID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 goto out;
1090
Oren Laadane19f2472006-01-08 01:03:58 -08001091 group_leader->signal->leader = 1;
Oleg Nesterov81dabb42013-07-03 15:08:26 -07001092 set_special_pids(sid);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001093
Alan Cox9c9f4de2008-10-13 10:37:26 +01001094 proc_clear_tty(group_leader);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001095
Oleg Nesterove4cc0a92008-02-08 04:19:09 -08001096 err = session;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097out:
1098 write_unlock_irq(&tasklist_lock);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001099 if (err > 0) {
Christian Borntraeger0d0df592009-10-26 16:49:34 -07001100 proc_sid_connector(group_leader);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001101 sched_autogroup_create_attach(group_leader);
1102 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 return err;
1104}
1105
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106DECLARE_RWSEM(uts_sem);
1107
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001108#ifdef COMPAT_UTS_MACHINE
1109#define override_architecture(name) \
Andreas Schwab46da2762010-04-23 13:17:44 -04001110 (personality(current->personality) == PER_LINUX32 && \
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001111 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1112 sizeof(COMPAT_UTS_MACHINE)))
1113#else
1114#define override_architecture(name) 0
1115#endif
1116
Andi Kleenbe274252011-08-19 16:15:10 -07001117/*
1118 * Work around broken programs that cannot handle "Linux 3.0".
1119 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
Jon DeVree39afb5e2015-02-27 15:52:07 -08001120 * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
Andi Kleenbe274252011-08-19 16:15:10 -07001121 */
Kees Cook2702b152012-10-19 13:56:51 -07001122static int override_release(char __user *release, size_t len)
Andi Kleenbe274252011-08-19 16:15:10 -07001123{
1124 int ret = 0;
Andi Kleenbe274252011-08-19 16:15:10 -07001125
1126 if (current->personality & UNAME26) {
Kees Cook2702b152012-10-19 13:56:51 -07001127 const char *rest = UTS_RELEASE;
1128 char buf[65] = { 0 };
Andi Kleenbe274252011-08-19 16:15:10 -07001129 int ndots = 0;
1130 unsigned v;
Kees Cook2702b152012-10-19 13:56:51 -07001131 size_t copy;
Andi Kleenbe274252011-08-19 16:15:10 -07001132
1133 while (*rest) {
1134 if (*rest == '.' && ++ndots >= 3)
1135 break;
1136 if (!isdigit(*rest) && *rest != '.')
1137 break;
1138 rest++;
1139 }
Jon DeVree39afb5e2015-02-27 15:52:07 -08001140 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
Kees Cook31fd84b92012-10-19 18:45:53 -07001141 copy = clamp_t(size_t, len, 1, sizeof(buf));
Kees Cook2702b152012-10-19 13:56:51 -07001142 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1143 ret = copy_to_user(release, buf, copy + 1);
Andi Kleenbe274252011-08-19 16:15:10 -07001144 }
1145 return ret;
1146}
1147
Heiko Carstense48fbb62009-01-14 14:14:26 +01001148SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149{
1150 int errno = 0;
1151
1152 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001153 if (copy_to_user(name, utsname(), sizeof *name))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 errno = -EFAULT;
1155 up_read(&uts_sem);
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001156
Andi Kleenbe274252011-08-19 16:15:10 -07001157 if (!errno && override_release(name->release, sizeof(name->release)))
1158 errno = -EFAULT;
Christoph Hellwige28cbf22010-03-10 15:21:19 -08001159 if (!errno && override_architecture(name))
1160 errno = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 return errno;
1162}
1163
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001164#ifdef __ARCH_WANT_SYS_OLD_UNAME
1165/*
1166 * Old cruft
1167 */
1168SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1169{
1170 int error = 0;
1171
1172 if (!name)
1173 return -EFAULT;
1174
1175 down_read(&uts_sem);
1176 if (copy_to_user(name, utsname(), sizeof(*name)))
1177 error = -EFAULT;
1178 up_read(&uts_sem);
1179
Andi Kleenbe274252011-08-19 16:15:10 -07001180 if (!error && override_release(name->release, sizeof(name->release)))
1181 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001182 if (!error && override_architecture(name))
1183 error = -EFAULT;
1184 return error;
1185}
1186
1187SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1188{
1189 int error;
1190
1191 if (!name)
1192 return -EFAULT;
1193 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1194 return -EFAULT;
1195
1196 down_read(&uts_sem);
1197 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1198 __OLD_UTS_LEN);
1199 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1200 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1201 __OLD_UTS_LEN);
1202 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1203 error |= __copy_to_user(&name->release, &utsname()->release,
1204 __OLD_UTS_LEN);
1205 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1206 error |= __copy_to_user(&name->version, &utsname()->version,
1207 __OLD_UTS_LEN);
1208 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1209 error |= __copy_to_user(&name->machine, &utsname()->machine,
1210 __OLD_UTS_LEN);
1211 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1212 up_read(&uts_sem);
1213
1214 if (!error && override_architecture(name))
1215 error = -EFAULT;
Andi Kleenbe274252011-08-19 16:15:10 -07001216 if (!error && override_release(name->release, sizeof(name->release)))
1217 error = -EFAULT;
Christoph Hellwig5cacdb42010-03-10 15:21:21 -08001218 return error ? -EFAULT : 0;
1219}
1220#endif
1221
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001222SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223{
1224 int errno;
1225 char tmp[__NEW_UTS_LEN];
1226
Serge E. Hallynbb96a6f2011-03-23 16:43:18 -07001227 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 return -EPERM;
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 if (len < 0 || len > __NEW_UTS_LEN)
1231 return -EINVAL;
1232 down_write(&uts_sem);
1233 errno = -EFAULT;
1234 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001235 struct new_utsname *u = utsname();
1236
1237 memcpy(u->nodename, tmp, len);
1238 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 errno = 0;
Sasikantha babu499eea62012-05-31 16:26:07 -07001240 uts_proc_notify(UTS_PROC_HOSTNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 }
1242 up_write(&uts_sem);
1243 return errno;
1244}
1245
1246#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1247
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001248SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249{
1250 int i, errno;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001251 struct new_utsname *u;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252
1253 if (len < 0)
1254 return -EINVAL;
1255 down_read(&uts_sem);
Andrew Morton9679e4d2008-10-15 22:01:51 -07001256 u = utsname();
1257 i = 1 + strlen(u->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 if (i > len)
1259 i = len;
1260 errno = 0;
Andrew Morton9679e4d2008-10-15 22:01:51 -07001261 if (copy_to_user(name, u->nodename, i))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 errno = -EFAULT;
1263 up_read(&uts_sem);
1264 return errno;
1265}
1266
1267#endif
1268
1269/*
1270 * Only setdomainname; getdomainname can be implemented by calling
1271 * uname()
1272 */
Heiko Carstens5a8a82b2009-01-14 14:14:25 +01001273SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274{
1275 int errno;
1276 char tmp[__NEW_UTS_LEN];
1277
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001278 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 return -EPERM;
1280 if (len < 0 || len > __NEW_UTS_LEN)
1281 return -EINVAL;
1282
1283 down_write(&uts_sem);
1284 errno = -EFAULT;
1285 if (!copy_from_user(tmp, name, len)) {
Andrew Morton9679e4d2008-10-15 22:01:51 -07001286 struct new_utsname *u = utsname();
1287
1288 memcpy(u->domainname, tmp, len);
1289 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 errno = 0;
Sasikantha babu499eea62012-05-31 16:26:07 -07001291 uts_proc_notify(UTS_PROC_DOMAINNAME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 }
1293 up_write(&uts_sem);
1294 return errno;
1295}
1296
Heiko Carstense48fbb62009-01-14 14:14:26 +01001297SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298{
Jiri Slabyb9518342010-05-04 11:28:25 +02001299 struct rlimit value;
1300 int ret;
1301
1302 ret = do_prlimit(current, resource, NULL, &value);
1303 if (!ret)
1304 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1305
1306 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307}
1308
1309#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1310
1311/*
1312 * Back compatibility for getrlimit. Needed for some apps.
1313 */
Heiko Carstense48fbb62009-01-14 14:14:26 +01001314SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1315 struct rlimit __user *, rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316{
1317 struct rlimit x;
1318 if (resource >= RLIM_NLIMITS)
1319 return -EINVAL;
1320
1321 task_lock(current->group_leader);
1322 x = current->signal->rlim[resource];
1323 task_unlock(current->group_leader);
Cal Peake756184b2006-09-30 23:27:24 -07001324 if (x.rlim_cur > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 x.rlim_cur = 0x7FFFFFFF;
Cal Peake756184b2006-09-30 23:27:24 -07001326 if (x.rlim_max > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 x.rlim_max = 0x7FFFFFFF;
vishnu.psec94fc32014-10-09 15:30:23 -07001328 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329}
1330
1331#endif
1332
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001333static inline bool rlim64_is_infinity(__u64 rlim64)
1334{
1335#if BITS_PER_LONG < 64
1336 return rlim64 >= ULONG_MAX;
1337#else
1338 return rlim64 == RLIM64_INFINITY;
1339#endif
1340}
1341
1342static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1343{
1344 if (rlim->rlim_cur == RLIM_INFINITY)
1345 rlim64->rlim_cur = RLIM64_INFINITY;
1346 else
1347 rlim64->rlim_cur = rlim->rlim_cur;
1348 if (rlim->rlim_max == RLIM_INFINITY)
1349 rlim64->rlim_max = RLIM64_INFINITY;
1350 else
1351 rlim64->rlim_max = rlim->rlim_max;
1352}
1353
1354static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1355{
1356 if (rlim64_is_infinity(rlim64->rlim_cur))
1357 rlim->rlim_cur = RLIM_INFINITY;
1358 else
1359 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1360 if (rlim64_is_infinity(rlim64->rlim_max))
1361 rlim->rlim_max = RLIM_INFINITY;
1362 else
1363 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1364}
1365
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001366/* make sure you are allowed to change @tsk limits before calling this */
Jiri Slaby5b415352010-03-24 16:11:29 +01001367int do_prlimit(struct task_struct *tsk, unsigned int resource,
1368 struct rlimit *new_rlim, struct rlimit *old_rlim)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369{
Jiri Slaby5b415352010-03-24 16:11:29 +01001370 struct rlimit *rlim;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001371 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
1373 if (resource >= RLIM_NLIMITS)
1374 return -EINVAL;
Jiri Slaby5b415352010-03-24 16:11:29 +01001375 if (new_rlim) {
1376 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1377 return -EINVAL;
1378 if (resource == RLIMIT_NOFILE &&
1379 new_rlim->rlim_max > sysctl_nr_open)
1380 return -EPERM;
1381 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001383 /* protect tsk->signal and tsk->sighand from disappearing */
1384 read_lock(&tasklist_lock);
1385 if (!tsk->sighand) {
1386 retval = -ESRCH;
1387 goto out;
1388 }
1389
Jiri Slaby5b415352010-03-24 16:11:29 +01001390 rlim = tsk->signal->rlim + resource;
Jiri Slaby86f162f2009-11-14 17:37:04 +01001391 task_lock(tsk->group_leader);
Jiri Slaby5b415352010-03-24 16:11:29 +01001392 if (new_rlim) {
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001393 /* Keep the capable check against init_user_ns until
1394 cgroups can contain all limits */
Jiri Slaby5b415352010-03-24 16:11:29 +01001395 if (new_rlim->rlim_max > rlim->rlim_max &&
1396 !capable(CAP_SYS_RESOURCE))
1397 retval = -EPERM;
1398 if (!retval)
1399 retval = security_task_setrlimit(tsk->group_leader,
1400 resource, new_rlim);
1401 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1402 /*
1403 * The caller is asking for an immediate RLIMIT_CPU
1404 * expiry. But we use the zero value to mean "it was
1405 * never set". So let's cheat and make it one second
1406 * instead
1407 */
1408 new_rlim->rlim_cur = 1;
1409 }
Tom Alsberg9926e4c2007-05-08 00:30:31 -07001410 }
Jiri Slaby5b415352010-03-24 16:11:29 +01001411 if (!retval) {
1412 if (old_rlim)
1413 *old_rlim = *rlim;
1414 if (new_rlim)
1415 *rlim = *new_rlim;
1416 }
Jiri Slaby7855c352009-08-26 23:45:34 +02001417 task_unlock(tsk->group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
Andrew Mortond3561f72006-03-24 03:18:36 -08001419 /*
1420 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1421 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1422 * very long-standing error, and fixing it now risks breakage of
1423 * applications, so we live with it
1424 */
Jiri Slaby5b415352010-03-24 16:11:29 +01001425 if (!retval && new_rlim && resource == RLIMIT_CPU &&
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001426 new_rlim->rlim_cur != RLIM_INFINITY &&
1427 IS_ENABLED(CONFIG_POSIX_TIMERS))
Jiri Slaby5b415352010-03-24 16:11:29 +01001428 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001429out:
Jiri Slaby1c1e6182009-08-28 14:08:17 +02001430 read_unlock(&tasklist_lock);
Oleg Nesterov2fb9d262009-09-03 19:21:45 +02001431 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432}
1433
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001434/* rcu lock must be held */
1435static int check_prlimit_permission(struct task_struct *task)
1436{
1437 const struct cred *cred = current_cred(), *tcred;
1438
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001439 if (current == task)
1440 return 0;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001441
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001442 tcred = __task_cred(task);
Eric W. Biederman5af66202012-03-03 20:21:47 -08001443 if (uid_eq(cred->uid, tcred->euid) &&
1444 uid_eq(cred->uid, tcred->suid) &&
1445 uid_eq(cred->uid, tcred->uid) &&
1446 gid_eq(cred->gid, tcred->egid) &&
1447 gid_eq(cred->gid, tcred->sgid) &&
1448 gid_eq(cred->gid, tcred->gid))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001449 return 0;
Eric W. Biedermanc4a4d602011-11-16 23:15:31 -08001450 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
Serge E. Hallynfc832ad2011-03-23 16:43:22 -07001451 return 0;
1452
1453 return -EPERM;
Jiri Slabyc022a0a2010-05-04 18:03:50 +02001454}
1455
1456SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1457 const struct rlimit64 __user *, new_rlim,
1458 struct rlimit64 __user *, old_rlim)
1459{
1460 struct rlimit64 old64, new64;
1461 struct rlimit old, new;
1462 struct task_struct *tsk;
1463 int ret;
1464
1465 if (new_rlim) {
1466 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1467 return -EFAULT;
1468 rlim64_to_rlim(&new64, &new);
1469 }
1470
1471 rcu_read_lock();
1472 tsk = pid ? find_task_by_vpid(pid) : current;
1473 if (!tsk) {
1474 rcu_read_unlock();
1475 return -ESRCH;
1476 }
1477 ret = check_prlimit_permission(tsk);
1478 if (ret) {
1479 rcu_read_unlock();
1480 return ret;
1481 }
1482 get_task_struct(tsk);
1483 rcu_read_unlock();
1484
1485 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1486 old_rlim ? &old : NULL);
1487
1488 if (!ret && old_rlim) {
1489 rlim_to_rlim64(&old, &old64);
1490 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1491 ret = -EFAULT;
1492 }
1493
1494 put_task_struct(tsk);
1495 return ret;
1496}
1497
Jiri Slaby7855c352009-08-26 23:45:34 +02001498SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1499{
1500 struct rlimit new_rlim;
1501
1502 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1503 return -EFAULT;
Jiri Slaby5b415352010-03-24 16:11:29 +01001504 return do_prlimit(current, resource, &new_rlim, NULL);
Jiri Slaby7855c352009-08-26 23:45:34 +02001505}
1506
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507/*
1508 * It would make sense to put struct rusage in the task_struct,
1509 * except that would make the task_struct be *really big*. After
1510 * task_struct gets moved into malloc'ed memory, it would
1511 * make sense to do this. It will make moving the rest of the information
1512 * a lot simpler! (Which we're not doing right now because we're not
1513 * measuring them yet).
1514 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1516 * races with threads incrementing their own counters. But since word
1517 * reads are atomic, we either get new values or old values and we don't
1518 * care which for the sums. We always take the siglock to protect reading
1519 * the c* fields from p->signal from races with exit.c updating those
1520 * fields when reaping, so a sample either gets all the additions of a
1521 * given child after it's reaped, or none so this sample is before reaping.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001522 *
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001523 * Locking:
1524 * We need to take the siglock for CHILDEREN, SELF and BOTH
1525 * for the cases current multithreaded, non-current single threaded
1526 * non-current multithreaded. Thread traversal is now safe with
1527 * the siglock held.
1528 * Strictly speaking, we donot need to take the siglock if we are current and
1529 * single threaded, as no one else can take our signal_struct away, no one
1530 * else can reap the children to update signal->c* counters, and no one else
1531 * can race with the signal-> fields. If we do not take any lock, the
1532 * signal-> fields could be read out of order while another thread was just
1533 * exiting. So we should place a read memory barrier when we avoid the lock.
1534 * On the writer side, write memory barrier is implied in __exit_signal
1535 * as __exit_signal releases the siglock spinlock after updating the signal->
1536 * fields. But we don't do this yet to keep things simple.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001537 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 */
1539
Frank Mayharf06febc2008-09-12 09:54:39 -07001540static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001541{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001542 r->ru_nvcsw += t->nvcsw;
1543 r->ru_nivcsw += t->nivcsw;
1544 r->ru_minflt += t->min_flt;
1545 r->ru_majflt += t->maj_flt;
1546 r->ru_inblock += task_io_get_inblock(t);
1547 r->ru_oublock += task_io_get_oublock(t);
1548}
1549
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1551{
1552 struct task_struct *t;
1553 unsigned long flags;
Frederic Weisbecker5613fda2017-01-31 04:09:23 +01001554 u64 tgutime, tgstime, utime, stime;
Jiri Pirko1f102062009-09-22 16:44:10 -07001555 unsigned long maxrss = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556
vishnu.psec94fc32014-10-09 15:30:23 -07001557 memset((char *)r, 0, sizeof (*r));
Martin Schwidefsky64861632011-12-15 14:56:09 +01001558 utime = stime = 0;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001559
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001560 if (who == RUSAGE_THREAD) {
Frederic Weisbeckere80d0a1a2012-11-21 16:26:44 +01001561 task_cputime_adjusted(current, &utime, &stime);
Frank Mayharf06febc2008-09-12 09:54:39 -07001562 accumulate_thread_rusage(p, r);
Jiri Pirko1f102062009-09-22 16:44:10 -07001563 maxrss = p->signal->maxrss;
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001564 goto out;
1565 }
1566
Oleg Nesterovd6cf7232008-04-30 00:52:38 -07001567 if (!lock_task_sighand(p, &flags))
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001568 return;
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001569
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 switch (who) {
vishnu.psec94fc32014-10-09 15:30:23 -07001571 case RUSAGE_BOTH:
1572 case RUSAGE_CHILDREN:
1573 utime = p->signal->cutime;
1574 stime = p->signal->cstime;
1575 r->ru_nvcsw = p->signal->cnvcsw;
1576 r->ru_nivcsw = p->signal->cnivcsw;
1577 r->ru_minflt = p->signal->cmin_flt;
1578 r->ru_majflt = p->signal->cmaj_flt;
1579 r->ru_inblock = p->signal->cinblock;
1580 r->ru_oublock = p->signal->coublock;
1581 maxrss = p->signal->cmaxrss;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001582
vishnu.psec94fc32014-10-09 15:30:23 -07001583 if (who == RUSAGE_CHILDREN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 break;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08001585
vishnu.psec94fc32014-10-09 15:30:23 -07001586 case RUSAGE_SELF:
1587 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1588 utime += tgutime;
1589 stime += tgstime;
1590 r->ru_nvcsw += p->signal->nvcsw;
1591 r->ru_nivcsw += p->signal->nivcsw;
1592 r->ru_minflt += p->signal->min_flt;
1593 r->ru_majflt += p->signal->maj_flt;
1594 r->ru_inblock += p->signal->inblock;
1595 r->ru_oublock += p->signal->oublock;
1596 if (maxrss < p->signal->maxrss)
1597 maxrss = p->signal->maxrss;
1598 t = p;
1599 do {
1600 accumulate_thread_rusage(t, r);
1601 } while_each_thread(p, t);
1602 break;
1603
1604 default:
1605 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 }
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001607 unlock_task_sighand(p, &flags);
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001608
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001609out:
Frederic Weisbecker5613fda2017-01-31 04:09:23 +01001610 r->ru_utime = ns_to_timeval(utime);
1611 r->ru_stime = ns_to_timeval(stime);
Jiri Pirko1f102062009-09-22 16:44:10 -07001612
1613 if (who != RUSAGE_CHILDREN) {
1614 struct mm_struct *mm = get_task_mm(p);
vishnu.psec94fc32014-10-09 15:30:23 -07001615
Jiri Pirko1f102062009-09-22 16:44:10 -07001616 if (mm) {
1617 setmax_mm_hiwater_rss(&maxrss, mm);
1618 mmput(mm);
1619 }
1620 }
1621 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622}
1623
1624int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1625{
1626 struct rusage r;
vishnu.psec94fc32014-10-09 15:30:23 -07001627
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 k_getrusage(p, who, &r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1630}
1631
Heiko Carstense48fbb62009-01-14 14:14:26 +01001632SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633{
Sripathi Kodi679c9cd2008-04-29 00:58:42 -07001634 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1635 who != RUSAGE_THREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 return -EINVAL;
1637 return getrusage(current, who, ru);
1638}
1639
Al Viro8d2d5c42013-03-03 12:49:06 -05001640#ifdef CONFIG_COMPAT
1641COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1642{
1643 struct rusage r;
1644
1645 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1646 who != RUSAGE_THREAD)
1647 return -EINVAL;
1648
1649 k_getrusage(current, who, &r);
1650 return put_compat_rusage(&r, ru);
1651}
1652#endif
1653
Heiko Carstense48fbb62009-01-14 14:14:26 +01001654SYSCALL_DEFINE1(umask, int, mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655{
1656 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1657 return mask;
1658}
Serge E. Hallyn3b7391d2008-02-04 22:29:45 -08001659
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001660static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001661{
Al Viro2903ff02012-08-28 12:52:22 -04001662 struct fd exe;
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001663 struct file *old_exe, *exe_file;
Al Viro496ad9a2013-01-23 17:07:38 -05001664 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -04001665 int err;
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001666
Al Viro2903ff02012-08-28 12:52:22 -04001667 exe = fdget(fd);
1668 if (!exe.file)
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001669 return -EBADF;
1670
Al Viro496ad9a2013-01-23 17:07:38 -05001671 inode = file_inode(exe.file);
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001672
1673 /*
1674 * Because the original mm->exe_file points to executable file, make
1675 * sure that this one is executable as well, to avoid breaking an
1676 * overall picture.
1677 */
1678 err = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -05001679 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001680 goto exit;
1681
Al Viro496ad9a2013-01-23 17:07:38 -05001682 err = inode_permission(inode, MAY_EXEC);
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001683 if (err)
1684 goto exit;
1685
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001686 /*
Konstantin Khlebnikov4229fb1d2012-07-11 14:02:11 -07001687 * Forbid mm->exe_file change if old file still mapped.
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001688 */
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001689 exe_file = get_mm_exe_file(mm);
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001690 err = -EBUSY;
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001691 if (exe_file) {
Konstantin Khlebnikov4229fb1d2012-07-11 14:02:11 -07001692 struct vm_area_struct *vma;
1693
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001694 down_read(&mm->mmap_sem);
1695 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1696 if (!vma->vm_file)
1697 continue;
1698 if (path_equal(&vma->vm_file->f_path,
1699 &exe_file->f_path))
1700 goto exit_err;
1701 }
1702
1703 up_read(&mm->mmap_sem);
1704 fput(exe_file);
Konstantin Khlebnikovbafb2822012-06-07 14:21:11 -07001705 }
1706
Konstantin Khlebnikov4229fb1d2012-07-11 14:02:11 -07001707 err = 0;
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001708 /* set the new file, lockless */
1709 get_file(exe.file);
1710 old_exe = xchg(&mm->exe_file, exe.file);
1711 if (old_exe)
1712 fput(old_exe);
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001713exit:
Al Viro2903ff02012-08-28 12:52:22 -04001714 fdput(exe);
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001715 return err;
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001716exit_err:
1717 up_read(&mm->mmap_sem);
1718 fput(exe_file);
1719 goto exit;
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001720}
1721
Cyrill Gorcunovf606b772014-10-09 15:27:37 -07001722/*
1723 * WARNING: we don't require any capability here so be very careful
1724 * in what is allowed for modification from userspace.
1725 */
1726static int validate_prctl_map(struct prctl_mm_map *prctl_map)
1727{
1728 unsigned long mmap_max_addr = TASK_SIZE;
1729 struct mm_struct *mm = current->mm;
1730 int error = -EINVAL, i;
1731
1732 static const unsigned char offsets[] = {
1733 offsetof(struct prctl_mm_map, start_code),
1734 offsetof(struct prctl_mm_map, end_code),
1735 offsetof(struct prctl_mm_map, start_data),
1736 offsetof(struct prctl_mm_map, end_data),
1737 offsetof(struct prctl_mm_map, start_brk),
1738 offsetof(struct prctl_mm_map, brk),
1739 offsetof(struct prctl_mm_map, start_stack),
1740 offsetof(struct prctl_mm_map, arg_start),
1741 offsetof(struct prctl_mm_map, arg_end),
1742 offsetof(struct prctl_mm_map, env_start),
1743 offsetof(struct prctl_mm_map, env_end),
1744 };
1745
1746 /*
1747 * Make sure the members are not somewhere outside
1748 * of allowed address space.
1749 */
1750 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1751 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1752
1753 if ((unsigned long)val >= mmap_max_addr ||
1754 (unsigned long)val < mmap_min_addr)
1755 goto out;
1756 }
1757
1758 /*
1759 * Make sure the pairs are ordered.
1760 */
1761#define __prctl_check_order(__m1, __op, __m2) \
1762 ((unsigned long)prctl_map->__m1 __op \
1763 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1764 error = __prctl_check_order(start_code, <, end_code);
1765 error |= __prctl_check_order(start_data, <, end_data);
1766 error |= __prctl_check_order(start_brk, <=, brk);
1767 error |= __prctl_check_order(arg_start, <=, arg_end);
1768 error |= __prctl_check_order(env_start, <=, env_end);
1769 if (error)
1770 goto out;
1771#undef __prctl_check_order
1772
1773 error = -EINVAL;
1774
1775 /*
1776 * @brk should be after @end_data in traditional maps.
1777 */
1778 if (prctl_map->start_brk <= prctl_map->end_data ||
1779 prctl_map->brk <= prctl_map->end_data)
1780 goto out;
1781
1782 /*
1783 * Neither we should allow to override limits if they set.
1784 */
1785 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1786 prctl_map->start_brk, prctl_map->end_data,
1787 prctl_map->start_data))
1788 goto out;
1789
1790 /*
1791 * Someone is trying to cheat the auxv vector.
1792 */
1793 if (prctl_map->auxv_size) {
1794 if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv))
1795 goto out;
1796 }
1797
1798 /*
1799 * Finally, make sure the caller has the rights to
1800 * change /proc/pid/exe link: only local root should
1801 * be allowed to.
1802 */
1803 if (prctl_map->exe_fd != (u32)-1) {
1804 struct user_namespace *ns = current_user_ns();
1805 const struct cred *cred = current_cred();
1806
1807 if (!uid_eq(cred->uid, make_kuid(ns, 0)) ||
1808 !gid_eq(cred->gid, make_kgid(ns, 0)))
1809 goto out;
1810 }
1811
1812 error = 0;
1813out:
1814 return error;
1815}
1816
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001817#ifdef CONFIG_CHECKPOINT_RESTORE
Cyrill Gorcunovf606b772014-10-09 15:27:37 -07001818static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1819{
1820 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1821 unsigned long user_auxv[AT_VECTOR_SIZE];
1822 struct mm_struct *mm = current->mm;
1823 int error;
1824
1825 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1826 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1827
1828 if (opt == PR_SET_MM_MAP_SIZE)
1829 return put_user((unsigned int)sizeof(prctl_map),
1830 (unsigned int __user *)addr);
1831
1832 if (data_size != sizeof(prctl_map))
1833 return -EINVAL;
1834
1835 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1836 return -EFAULT;
1837
1838 error = validate_prctl_map(&prctl_map);
1839 if (error)
1840 return error;
1841
1842 if (prctl_map.auxv_size) {
1843 memset(user_auxv, 0, sizeof(user_auxv));
1844 if (copy_from_user(user_auxv,
1845 (const void __user *)prctl_map.auxv,
1846 prctl_map.auxv_size))
1847 return -EFAULT;
1848
1849 /* Last entry must be AT_NULL as specification requires */
1850 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
1851 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
1852 }
1853
Mateusz Guzikddf1d392016-01-20 15:01:02 -08001854 if (prctl_map.exe_fd != (u32)-1) {
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001855 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
Mateusz Guzikddf1d392016-01-20 15:01:02 -08001856 if (error)
1857 return error;
1858 }
1859
1860 down_write(&mm->mmap_sem);
Cyrill Gorcunovf606b772014-10-09 15:27:37 -07001861
1862 /*
1863 * We don't validate if these members are pointing to
1864 * real present VMAs because application may have correspond
1865 * VMAs already unmapped and kernel uses these members for statistics
1866 * output in procfs mostly, except
1867 *
1868 * - @start_brk/@brk which are used in do_brk but kernel lookups
1869 * for VMAs when updating these memvers so anything wrong written
1870 * here cause kernel to swear at userspace program but won't lead
1871 * to any problem in kernel itself
1872 */
1873
1874 mm->start_code = prctl_map.start_code;
1875 mm->end_code = prctl_map.end_code;
1876 mm->start_data = prctl_map.start_data;
1877 mm->end_data = prctl_map.end_data;
1878 mm->start_brk = prctl_map.start_brk;
1879 mm->brk = prctl_map.brk;
1880 mm->start_stack = prctl_map.start_stack;
1881 mm->arg_start = prctl_map.arg_start;
1882 mm->arg_end = prctl_map.arg_end;
1883 mm->env_start = prctl_map.env_start;
1884 mm->env_end = prctl_map.env_end;
1885
1886 /*
1887 * Note this update of @saved_auxv is lockless thus
1888 * if someone reads this member in procfs while we're
1889 * updating -- it may get partly updated results. It's
1890 * known and acceptable trade off: we leave it as is to
1891 * not introduce additional locks here making the kernel
1892 * more complex.
1893 */
1894 if (prctl_map.auxv_size)
1895 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
1896
Mateusz Guzikddf1d392016-01-20 15:01:02 -08001897 up_write(&mm->mmap_sem);
1898 return 0;
Cyrill Gorcunovf606b772014-10-09 15:27:37 -07001899}
1900#endif /* CONFIG_CHECKPOINT_RESTORE */
1901
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001902static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
1903 unsigned long len)
1904{
1905 /*
1906 * This doesn't move the auxiliary vector itself since it's pinned to
1907 * mm_struct, but it permits filling the vector with new values. It's
1908 * up to the caller to provide sane values here, otherwise userspace
1909 * tools which use this vector might be unhappy.
1910 */
1911 unsigned long user_auxv[AT_VECTOR_SIZE];
1912
1913 if (len > sizeof(user_auxv))
1914 return -EINVAL;
1915
1916 if (copy_from_user(user_auxv, (const void __user *)addr, len))
1917 return -EFAULT;
1918
1919 /* Make sure the last entry is always AT_NULL */
1920 user_auxv[AT_VECTOR_SIZE - 2] = 0;
1921 user_auxv[AT_VECTOR_SIZE - 1] = 0;
1922
1923 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1924
1925 task_lock(current);
1926 memcpy(mm->saved_auxv, user_auxv, len);
1927 task_unlock(current);
1928
1929 return 0;
1930}
1931
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001932static int prctl_set_mm(int opt, unsigned long addr,
1933 unsigned long arg4, unsigned long arg5)
1934{
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001935 struct mm_struct *mm = current->mm;
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001936 struct prctl_mm_map prctl_map;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001937 struct vm_area_struct *vma;
1938 int error;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001939
Cyrill Gorcunovf606b772014-10-09 15:27:37 -07001940 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
1941 opt != PR_SET_MM_MAP &&
1942 opt != PR_SET_MM_MAP_SIZE)))
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001943 return -EINVAL;
1944
Cyrill Gorcunovf606b772014-10-09 15:27:37 -07001945#ifdef CONFIG_CHECKPOINT_RESTORE
1946 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
1947 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
1948#endif
1949
Cyrill Gorcunov79f07132012-03-15 15:17:10 -07001950 if (!capable(CAP_SYS_RESOURCE))
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001951 return -EPERM;
1952
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001953 if (opt == PR_SET_MM_EXE_FILE)
1954 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
Cyrill Gorcunovb32dfe32012-05-31 16:26:46 -07001955
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001956 if (opt == PR_SET_MM_AUXV)
1957 return prctl_set_auxv(mm, addr, arg4);
1958
Cyrill Gorcunov1ad75b92012-06-07 14:21:11 -07001959 if (addr >= TASK_SIZE || addr < mmap_min_addr)
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001960 return -EINVAL;
1961
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001962 error = -EINVAL;
1963
Mateusz Guzikddf1d392016-01-20 15:01:02 -08001964 down_write(&mm->mmap_sem);
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001965 vma = find_vma(mm, addr);
1966
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001967 prctl_map.start_code = mm->start_code;
1968 prctl_map.end_code = mm->end_code;
1969 prctl_map.start_data = mm->start_data;
1970 prctl_map.end_data = mm->end_data;
1971 prctl_map.start_brk = mm->start_brk;
1972 prctl_map.brk = mm->brk;
1973 prctl_map.start_stack = mm->start_stack;
1974 prctl_map.arg_start = mm->arg_start;
1975 prctl_map.arg_end = mm->arg_end;
1976 prctl_map.env_start = mm->env_start;
1977 prctl_map.env_end = mm->env_end;
1978 prctl_map.auxv = NULL;
1979 prctl_map.auxv_size = 0;
1980 prctl_map.exe_fd = -1;
1981
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001982 switch (opt) {
1983 case PR_SET_MM_START_CODE:
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001984 prctl_map.start_code = addr;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001985 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001986 case PR_SET_MM_END_CODE:
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001987 prctl_map.end_code = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001988 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001989 case PR_SET_MM_START_DATA:
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001990 prctl_map.start_data = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001991 break;
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07001992 case PR_SET_MM_END_DATA:
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001993 prctl_map.end_data = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001994 break;
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001995 case PR_SET_MM_START_STACK:
1996 prctl_map.start_stack = addr;
1997 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08001998 case PR_SET_MM_START_BRK:
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07001999 prctl_map.start_brk = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002000 break;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002001 case PR_SET_MM_BRK:
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07002002 prctl_map.brk = addr;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002003 break;
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07002004 case PR_SET_MM_ARG_START:
2005 prctl_map.arg_start = addr;
2006 break;
2007 case PR_SET_MM_ARG_END:
2008 prctl_map.arg_end = addr;
2009 break;
2010 case PR_SET_MM_ENV_START:
2011 prctl_map.env_start = addr;
2012 break;
2013 case PR_SET_MM_ENV_END:
2014 prctl_map.env_end = addr;
2015 break;
2016 default:
2017 goto out;
2018 }
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002019
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07002020 error = validate_prctl_map(&prctl_map);
2021 if (error)
2022 goto out;
2023
2024 switch (opt) {
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07002025 /*
2026 * If command line arguments and environment
2027 * are placed somewhere else on stack, we can
2028 * set them up here, ARG_START/END to setup
2029 * command line argumets and ENV_START/END
2030 * for environment.
2031 */
2032 case PR_SET_MM_START_STACK:
2033 case PR_SET_MM_ARG_START:
2034 case PR_SET_MM_ARG_END:
2035 case PR_SET_MM_ENV_START:
2036 case PR_SET_MM_ENV_END:
2037 if (!vma) {
2038 error = -EFAULT;
2039 goto out;
2040 }
Cyrill Gorcunovfe8c7f52012-05-31 16:26:45 -07002041 }
Alexey Dobriyan4a00e9d2015-06-25 15:00:51 -07002042
2043 mm->start_code = prctl_map.start_code;
2044 mm->end_code = prctl_map.end_code;
2045 mm->start_data = prctl_map.start_data;
2046 mm->end_data = prctl_map.end_data;
2047 mm->start_brk = prctl_map.start_brk;
2048 mm->brk = prctl_map.brk;
2049 mm->start_stack = prctl_map.start_stack;
2050 mm->arg_start = prctl_map.arg_start;
2051 mm->arg_end = prctl_map.arg_end;
2052 mm->env_start = prctl_map.env_start;
2053 mm->env_end = prctl_map.env_end;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002054
2055 error = 0;
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002056out:
Mateusz Guzikddf1d392016-01-20 15:01:02 -08002057 up_write(&mm->mmap_sem);
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002058 return error;
2059}
Cyrill Gorcunov300f7862012-06-07 14:21:12 -07002060
Amnon Shiloh52b36942013-04-30 15:28:48 -07002061#ifdef CONFIG_CHECKPOINT_RESTORE
Cyrill Gorcunov300f7862012-06-07 14:21:12 -07002062static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2063{
2064 return put_user(me->clear_child_tid, tid_addr);
2065}
Amnon Shiloh52b36942013-04-30 15:28:48 -07002066#else
Cyrill Gorcunov300f7862012-06-07 14:21:12 -07002067static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2068{
2069 return -EINVAL;
2070}
Cyrill Gorcunov028ee4b2012-01-12 17:20:55 -08002071#endif
2072
Pavel Tikhomirov749860c2017-01-30 18:06:12 +03002073static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2074{
2075 /*
2076 * If task has has_child_subreaper - all its decendants
2077 * already have these flag too and new decendants will
2078 * inherit it on fork, skip them.
2079 *
2080 * If we've found child_reaper - skip descendants in
2081 * it's subtree as they will never get out pidns.
2082 */
2083 if (p->signal->has_child_subreaper ||
2084 is_child_reaper(task_pid(p)))
2085 return 0;
2086
2087 p->signal->has_child_subreaper = 1;
2088 return 1;
2089}
2090
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01002091SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2092 unsigned long, arg4, unsigned long, arg5)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093{
David Howellsb6dff3e2008-11-14 10:39:16 +11002094 struct task_struct *me = current;
2095 unsigned char comm[sizeof(me->comm)];
2096 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
David Howellsd84f4f92008-11-14 10:39:23 +11002098 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2099 if (error != -ENOSYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 return error;
2101
David Howellsd84f4f92008-11-14 10:39:23 +11002102 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 switch (option) {
Andrew Mortonf3cbd432013-02-21 16:43:07 -08002104 case PR_SET_PDEATHSIG:
2105 if (!valid_signal(arg2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106 error = -EINVAL;
2107 break;
Andrew Mortonf3cbd432013-02-21 16:43:07 -08002108 }
2109 me->pdeath_signal = arg2;
2110 break;
2111 case PR_GET_PDEATHSIG:
2112 error = put_user(me->pdeath_signal, (int __user *)arg2);
2113 break;
2114 case PR_GET_DUMPABLE:
2115 error = get_dumpable(me->mm);
2116 break;
2117 case PR_SET_DUMPABLE:
2118 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2119 error = -EINVAL;
2120 break;
2121 }
2122 set_dumpable(me->mm, arg2);
2123 break;
2124
2125 case PR_SET_UNALIGN:
2126 error = SET_UNALIGN_CTL(me, arg2);
2127 break;
2128 case PR_GET_UNALIGN:
2129 error = GET_UNALIGN_CTL(me, arg2);
2130 break;
2131 case PR_SET_FPEMU:
2132 error = SET_FPEMU_CTL(me, arg2);
2133 break;
2134 case PR_GET_FPEMU:
2135 error = GET_FPEMU_CTL(me, arg2);
2136 break;
2137 case PR_SET_FPEXC:
2138 error = SET_FPEXC_CTL(me, arg2);
2139 break;
2140 case PR_GET_FPEXC:
2141 error = GET_FPEXC_CTL(me, arg2);
2142 break;
2143 case PR_GET_TIMING:
2144 error = PR_TIMING_STATISTICAL;
2145 break;
2146 case PR_SET_TIMING:
2147 if (arg2 != PR_TIMING_STATISTICAL)
2148 error = -EINVAL;
2149 break;
2150 case PR_SET_NAME:
2151 comm[sizeof(me->comm) - 1] = 0;
2152 if (strncpy_from_user(comm, (char __user *)arg2,
2153 sizeof(me->comm) - 1) < 0)
2154 return -EFAULT;
2155 set_task_comm(me, comm);
2156 proc_comm_connector(me);
2157 break;
2158 case PR_GET_NAME:
2159 get_task_comm(comm, me);
2160 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2161 return -EFAULT;
2162 break;
2163 case PR_GET_ENDIAN:
2164 error = GET_ENDIAN(me, arg2);
2165 break;
2166 case PR_SET_ENDIAN:
2167 error = SET_ENDIAN(me, arg2);
2168 break;
2169 case PR_GET_SECCOMP:
2170 error = prctl_get_seccomp();
2171 break;
2172 case PR_SET_SECCOMP:
2173 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2174 break;
2175 case PR_GET_TSC:
2176 error = GET_TSC_CTL(arg2);
2177 break;
2178 case PR_SET_TSC:
2179 error = SET_TSC_CTL(arg2);
2180 break;
2181 case PR_TASK_PERF_EVENTS_DISABLE:
2182 error = perf_event_task_disable();
2183 break;
2184 case PR_TASK_PERF_EVENTS_ENABLE:
2185 error = perf_event_task_enable();
2186 break;
2187 case PR_GET_TIMERSLACK:
John Stultzda8b44d2016-03-17 14:20:51 -07002188 if (current->timer_slack_ns > ULONG_MAX)
2189 error = ULONG_MAX;
2190 else
2191 error = current->timer_slack_ns;
Andrew Mortonf3cbd432013-02-21 16:43:07 -08002192 break;
2193 case PR_SET_TIMERSLACK:
2194 if (arg2 <= 0)
2195 current->timer_slack_ns =
2196 current->default_timer_slack_ns;
2197 else
2198 current->timer_slack_ns = arg2;
2199 break;
2200 case PR_MCE_KILL:
2201 if (arg4 | arg5)
2202 return -EINVAL;
2203 switch (arg2) {
2204 case PR_MCE_KILL_CLEAR:
2205 if (arg3 != 0)
2206 return -EINVAL;
2207 current->flags &= ~PF_MCE_PROCESS;
2208 break;
2209 case PR_MCE_KILL_SET:
2210 current->flags |= PF_MCE_PROCESS;
2211 if (arg3 == PR_MCE_KILL_EARLY)
2212 current->flags |= PF_MCE_EARLY;
2213 else if (arg3 == PR_MCE_KILL_LATE)
2214 current->flags &= ~PF_MCE_EARLY;
2215 else if (arg3 == PR_MCE_KILL_DEFAULT)
2216 current->flags &=
2217 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2218 else
2219 return -EINVAL;
2220 break;
2221 default:
2222 return -EINVAL;
2223 }
2224 break;
2225 case PR_MCE_KILL_GET:
2226 if (arg2 | arg3 | arg4 | arg5)
2227 return -EINVAL;
2228 if (current->flags & PF_MCE_PROCESS)
2229 error = (current->flags & PF_MCE_EARLY) ?
2230 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2231 else
2232 error = PR_MCE_KILL_DEFAULT;
2233 break;
2234 case PR_SET_MM:
2235 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2236 break;
2237 case PR_GET_TID_ADDRESS:
2238 error = prctl_get_tid_address(me, (int __user **)arg2);
2239 break;
2240 case PR_SET_CHILD_SUBREAPER:
2241 me->signal->is_child_subreaper = !!arg2;
Pavel Tikhomirov749860c2017-01-30 18:06:12 +03002242 if (!arg2)
2243 break;
2244
2245 walk_process_tree(me, propagate_has_child_subreaper, NULL);
Andrew Mortonf3cbd432013-02-21 16:43:07 -08002246 break;
2247 case PR_GET_CHILD_SUBREAPER:
2248 error = put_user(me->signal->is_child_subreaper,
2249 (int __user *)arg2);
2250 break;
2251 case PR_SET_NO_NEW_PRIVS:
2252 if (arg2 != 1 || arg3 || arg4 || arg5)
2253 return -EINVAL;
2254
Kees Cook1d4457f2014-05-21 15:23:46 -07002255 task_set_no_new_privs(current);
Andrew Mortonf3cbd432013-02-21 16:43:07 -08002256 break;
2257 case PR_GET_NO_NEW_PRIVS:
2258 if (arg2 || arg3 || arg4 || arg5)
2259 return -EINVAL;
Kees Cook1d4457f2014-05-21 15:23:46 -07002260 return task_no_new_privs(current) ? 1 : 0;
Alex Thorltona0715cc2014-04-07 15:37:10 -07002261 case PR_GET_THP_DISABLE:
2262 if (arg2 || arg3 || arg4 || arg5)
2263 return -EINVAL;
2264 error = !!(me->mm->def_flags & VM_NOHUGEPAGE);
2265 break;
2266 case PR_SET_THP_DISABLE:
2267 if (arg3 || arg4 || arg5)
2268 return -EINVAL;
Michal Hocko17b05732016-05-23 16:26:05 -07002269 if (down_write_killable(&me->mm->mmap_sem))
2270 return -EINTR;
Alex Thorltona0715cc2014-04-07 15:37:10 -07002271 if (arg2)
2272 me->mm->def_flags |= VM_NOHUGEPAGE;
2273 else
2274 me->mm->def_flags &= ~VM_NOHUGEPAGE;
2275 up_write(&me->mm->mmap_sem);
2276 break;
Dave Hansenfe3d1972014-11-14 07:18:29 -08002277 case PR_MPX_ENABLE_MANAGEMENT:
Dave Hansene9d1b4f2015-01-08 14:30:22 -08002278 if (arg2 || arg3 || arg4 || arg5)
2279 return -EINVAL;
Dave Hansen46a6e0c2015-06-07 11:37:02 -07002280 error = MPX_ENABLE_MANAGEMENT();
Dave Hansenfe3d1972014-11-14 07:18:29 -08002281 break;
2282 case PR_MPX_DISABLE_MANAGEMENT:
Dave Hansene9d1b4f2015-01-08 14:30:22 -08002283 if (arg2 || arg3 || arg4 || arg5)
2284 return -EINVAL;
Dave Hansen46a6e0c2015-06-07 11:37:02 -07002285 error = MPX_DISABLE_MANAGEMENT();
Dave Hansenfe3d1972014-11-14 07:18:29 -08002286 break;
Paul Burton97915542015-01-08 12:17:37 +00002287 case PR_SET_FP_MODE:
2288 error = SET_FP_MODE(me, arg2);
2289 break;
2290 case PR_GET_FP_MODE:
2291 error = GET_FP_MODE(me);
2292 break;
Andrew Mortonf3cbd432013-02-21 16:43:07 -08002293 default:
2294 error = -EINVAL;
2295 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296 }
2297 return error;
2298}
Andi Kleen3cfc3482006-09-26 10:52:28 +02002299
Heiko Carstens836f92a2009-01-14 14:14:33 +01002300SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2301 struct getcpu_cache __user *, unused)
Andi Kleen3cfc3482006-09-26 10:52:28 +02002302{
2303 int err = 0;
2304 int cpu = raw_smp_processor_id();
vishnu.psec94fc32014-10-09 15:30:23 -07002305
Andi Kleen3cfc3482006-09-26 10:52:28 +02002306 if (cpup)
2307 err |= put_user(cpu, cpup);
2308 if (nodep)
2309 err |= put_user(cpu_to_node(cpu), nodep);
Andi Kleen3cfc3482006-09-26 10:52:28 +02002310 return err ? -EFAULT : 0;
2311}
Jeremy Fitzhardinge10a0a8d2007-07-17 18:37:02 -07002312
Stephen Rothwell4a22f162013-04-30 15:27:37 -07002313/**
2314 * do_sysinfo - fill in sysinfo struct
2315 * @info: pointer to buffer to fill
2316 */
2317static int do_sysinfo(struct sysinfo *info)
2318{
2319 unsigned long mem_total, sav_total;
2320 unsigned int mem_unit, bitcount;
2321 struct timespec tp;
2322
2323 memset(info, 0, sizeof(struct sysinfo));
2324
Oleg Nesterov45c64942013-07-03 15:05:01 -07002325 get_monotonic_boottime(&tp);
Stephen Rothwell4a22f162013-04-30 15:27:37 -07002326 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2327
2328 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2329
2330 info->procs = nr_threads;
2331
2332 si_meminfo(info);
2333 si_swapinfo(info);
2334
2335 /*
2336 * If the sum of all the available memory (i.e. ram + swap)
2337 * is less than can be stored in a 32 bit unsigned long then
2338 * we can be binary compatible with 2.2.x kernels. If not,
2339 * well, in that case 2.2.x was broken anyways...
2340 *
2341 * -Erik Andersen <andersee@debian.org>
2342 */
2343
2344 mem_total = info->totalram + info->totalswap;
2345 if (mem_total < info->totalram || mem_total < info->totalswap)
2346 goto out;
2347 bitcount = 0;
2348 mem_unit = info->mem_unit;
2349 while (mem_unit > 1) {
2350 bitcount++;
2351 mem_unit >>= 1;
2352 sav_total = mem_total;
2353 mem_total <<= 1;
2354 if (mem_total < sav_total)
2355 goto out;
2356 }
2357
2358 /*
2359 * If mem_total did not overflow, multiply all memory values by
2360 * info->mem_unit and set it to 1. This leaves things compatible
2361 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2362 * kernels...
2363 */
2364
2365 info->mem_unit = 1;
2366 info->totalram <<= bitcount;
2367 info->freeram <<= bitcount;
2368 info->sharedram <<= bitcount;
2369 info->bufferram <<= bitcount;
2370 info->totalswap <<= bitcount;
2371 info->freeswap <<= bitcount;
2372 info->totalhigh <<= bitcount;
2373 info->freehigh <<= bitcount;
2374
2375out:
2376 return 0;
2377}
2378
2379SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2380{
2381 struct sysinfo val;
2382
2383 do_sysinfo(&val);
2384
2385 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2386 return -EFAULT;
2387
2388 return 0;
2389}
2390
2391#ifdef CONFIG_COMPAT
2392struct compat_sysinfo {
2393 s32 uptime;
2394 u32 loads[3];
2395 u32 totalram;
2396 u32 freeram;
2397 u32 sharedram;
2398 u32 bufferram;
2399 u32 totalswap;
2400 u32 freeswap;
2401 u16 procs;
2402 u16 pad;
2403 u32 totalhigh;
2404 u32 freehigh;
2405 u32 mem_unit;
2406 char _f[20-2*sizeof(u32)-sizeof(int)];
2407};
2408
2409COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2410{
2411 struct sysinfo s;
2412
2413 do_sysinfo(&s);
2414
2415 /* Check to see if any memory value is too large for 32-bit and scale
2416 * down if needed
2417 */
Scotty Bauer0baae412014-10-09 15:30:26 -07002418 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
Stephen Rothwell4a22f162013-04-30 15:27:37 -07002419 int bitcount = 0;
2420
2421 while (s.mem_unit < PAGE_SIZE) {
2422 s.mem_unit <<= 1;
2423 bitcount++;
2424 }
2425
2426 s.totalram >>= bitcount;
2427 s.freeram >>= bitcount;
2428 s.sharedram >>= bitcount;
2429 s.bufferram >>= bitcount;
2430 s.totalswap >>= bitcount;
2431 s.freeswap >>= bitcount;
2432 s.totalhigh >>= bitcount;
2433 s.freehigh >>= bitcount;
2434 }
2435
2436 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2437 __put_user(s.uptime, &info->uptime) ||
2438 __put_user(s.loads[0], &info->loads[0]) ||
2439 __put_user(s.loads[1], &info->loads[1]) ||
2440 __put_user(s.loads[2], &info->loads[2]) ||
2441 __put_user(s.totalram, &info->totalram) ||
2442 __put_user(s.freeram, &info->freeram) ||
2443 __put_user(s.sharedram, &info->sharedram) ||
2444 __put_user(s.bufferram, &info->bufferram) ||
2445 __put_user(s.totalswap, &info->totalswap) ||
2446 __put_user(s.freeswap, &info->freeswap) ||
2447 __put_user(s.procs, &info->procs) ||
2448 __put_user(s.totalhigh, &info->totalhigh) ||
2449 __put_user(s.freehigh, &info->freehigh) ||
2450 __put_user(s.mem_unit, &info->mem_unit))
2451 return -EFAULT;
2452
2453 return 0;
2454}
2455#endif /* CONFIG_COMPAT */