]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/sys.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
9984de1a | 8 | #include <linux/export.h> |
1da177e4 LT |
9 | #include <linux/mm.h> |
10 | #include <linux/utsname.h> | |
11 | #include <linux/mman.h> | |
1da177e4 LT |
12 | #include <linux/reboot.h> |
13 | #include <linux/prctl.h> | |
1da177e4 LT |
14 | #include <linux/highuid.h> |
15 | #include <linux/fs.h> | |
74da1ff7 | 16 | #include <linux/kmod.h> |
cdd6c482 | 17 | #include <linux/perf_event.h> |
3e88c553 | 18 | #include <linux/resource.h> |
dc009d92 | 19 | #include <linux/kernel.h> |
1da177e4 | 20 | #include <linux/workqueue.h> |
c59ede7b | 21 | #include <linux/capability.h> |
1da177e4 LT |
22 | #include <linux/device.h> |
23 | #include <linux/key.h> | |
24 | #include <linux/times.h> | |
25 | #include <linux/posix-timers.h> | |
26 | #include <linux/security.h> | |
27 | #include <linux/dcookies.h> | |
28 | #include <linux/suspend.h> | |
29 | #include <linux/tty.h> | |
7ed20e1a | 30 | #include <linux/signal.h> |
9f46080c | 31 | #include <linux/cn_proc.h> |
3cfc348b | 32 | #include <linux/getcpu.h> |
6eaeeaba | 33 | #include <linux/task_io_accounting_ops.h> |
1d9d02fe | 34 | #include <linux/seccomp.h> |
4047727e | 35 | #include <linux/cpu.h> |
e28cbf22 | 36 | #include <linux/personality.h> |
e3d5a27d | 37 | #include <linux/ptrace.h> |
5ad4e53b | 38 | #include <linux/fs_struct.h> |
b32dfe37 CG |
39 | #include <linux/file.h> |
40 | #include <linux/mount.h> | |
5a0e3ad6 | 41 | #include <linux/gfp.h> |
40dc166c | 42 | #include <linux/syscore_ops.h> |
be27425d AK |
43 | #include <linux/version.h> |
44 | #include <linux/ctype.h> | |
1da177e4 LT |
45 | |
46 | #include <linux/compat.h> | |
47 | #include <linux/syscalls.h> | |
00d7c05a | 48 | #include <linux/kprobes.h> |
acce292c | 49 | #include <linux/user_namespace.h> |
7fe5e042 | 50 | #include <linux/binfmts.h> |
1da177e4 | 51 | |
4a22f166 | 52 | #include <linux/sched.h> |
4eb5aaa3 | 53 | #include <linux/sched/autogroup.h> |
4f17722c | 54 | #include <linux/sched/loadavg.h> |
03441a34 | 55 | #include <linux/sched/stat.h> |
6e84f315 | 56 | #include <linux/sched/mm.h> |
f7ccbae4 | 57 | #include <linux/sched/coredump.h> |
29930025 | 58 | #include <linux/sched/task.h> |
32ef5517 | 59 | #include <linux/sched/cputime.h> |
4a22f166 SR |
60 | #include <linux/rcupdate.h> |
61 | #include <linux/uidgid.h> | |
62 | #include <linux/cred.h> | |
63 | ||
b617cfc8 TG |
64 | #include <linux/nospec.h> |
65 | ||
04c6862c | 66 | #include <linux/kmsg_dump.h> |
be27425d AK |
67 | /* Move somewhere else to avoid recompiling? */ |
68 | #include <generated/utsrelease.h> | |
04c6862c | 69 | |
7c0f6ba6 | 70 | #include <linux/uaccess.h> |
1da177e4 LT |
71 | #include <asm/io.h> |
72 | #include <asm/unistd.h> | |
73 | ||
e530dca5 DB |
74 | #include "uid16.h" |
75 | ||
1da177e4 | 76 | #ifndef SET_UNALIGN_CTL |
ec94fc3d | 77 | # define SET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
78 | #endif |
79 | #ifndef GET_UNALIGN_CTL | |
ec94fc3d | 80 | # define GET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
81 | #endif |
82 | #ifndef SET_FPEMU_CTL | |
ec94fc3d | 83 | # define SET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
84 | #endif |
85 | #ifndef GET_FPEMU_CTL | |
ec94fc3d | 86 | # define GET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
87 | #endif |
88 | #ifndef SET_FPEXC_CTL | |
ec94fc3d | 89 | # define SET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 LT |
90 | #endif |
91 | #ifndef GET_FPEXC_CTL | |
ec94fc3d | 92 | # define GET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 | 93 | #endif |
651d765d | 94 | #ifndef GET_ENDIAN |
ec94fc3d | 95 | # define GET_ENDIAN(a, b) (-EINVAL) |
651d765d AB |
96 | #endif |
97 | #ifndef SET_ENDIAN | |
ec94fc3d | 98 | # define SET_ENDIAN(a, b) (-EINVAL) |
651d765d | 99 | #endif |
8fb402bc EB |
100 | #ifndef GET_TSC_CTL |
101 | # define GET_TSC_CTL(a) (-EINVAL) | |
102 | #endif | |
103 | #ifndef SET_TSC_CTL | |
104 | # define SET_TSC_CTL(a) (-EINVAL) | |
105 | #endif | |
fe3d197f | 106 | #ifndef MPX_ENABLE_MANAGEMENT |
46a6e0cf | 107 | # define MPX_ENABLE_MANAGEMENT() (-EINVAL) |
fe3d197f DH |
108 | #endif |
109 | #ifndef MPX_DISABLE_MANAGEMENT | |
46a6e0cf | 110 | # define MPX_DISABLE_MANAGEMENT() (-EINVAL) |
fe3d197f | 111 | #endif |
9791554b PB |
112 | #ifndef GET_FP_MODE |
113 | # define GET_FP_MODE(a) (-EINVAL) | |
114 | #endif | |
115 | #ifndef SET_FP_MODE | |
116 | # define SET_FP_MODE(a,b) (-EINVAL) | |
117 | #endif | |
2d2123bc DM |
118 | #ifndef SVE_SET_VL |
119 | # define SVE_SET_VL(a) (-EINVAL) | |
120 | #endif | |
121 | #ifndef SVE_GET_VL | |
122 | # define SVE_GET_VL() (-EINVAL) | |
123 | #endif | |
1da177e4 LT |
124 | |
125 | /* | |
126 | * this is where the system-wide overflow UID and GID are defined, for | |
127 | * architectures that now have 32-bit UID/GID but didn't in the past | |
128 | */ | |
129 | ||
130 | int overflowuid = DEFAULT_OVERFLOWUID; | |
131 | int overflowgid = DEFAULT_OVERFLOWGID; | |
132 | ||
1da177e4 LT |
133 | EXPORT_SYMBOL(overflowuid); |
134 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
135 | |
136 | /* | |
137 | * the same as above, but for filesystems which can only store a 16-bit | |
138 | * UID and GID. as such, this is needed on all architectures | |
139 | */ | |
140 | ||
141 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
8b2770a4 | 142 | int fs_overflowgid = DEFAULT_FS_OVERFLOWGID; |
1da177e4 LT |
143 | |
144 | EXPORT_SYMBOL(fs_overflowuid); | |
145 | EXPORT_SYMBOL(fs_overflowgid); | |
146 | ||
fc832ad3 SH |
147 | /* |
148 | * Returns true if current's euid is same as p's uid or euid, | |
149 | * or has CAP_SYS_NICE to p's user_ns. | |
150 | * | |
151 | * Called with rcu_read_lock, creds are safe | |
152 | */ | |
153 | static bool set_one_prio_perm(struct task_struct *p) | |
154 | { | |
155 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
156 | ||
5af66203 EB |
157 | if (uid_eq(pcred->uid, cred->euid) || |
158 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 159 | return true; |
c4a4d603 | 160 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
161 | return true; |
162 | return false; | |
163 | } | |
164 | ||
c69e8d9c DH |
165 | /* |
166 | * set the priority of a task | |
167 | * - the caller must hold the RCU read lock | |
168 | */ | |
1da177e4 LT |
169 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
170 | { | |
171 | int no_nice; | |
172 | ||
fc832ad3 | 173 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
174 | error = -EPERM; |
175 | goto out; | |
176 | } | |
e43379f1 | 177 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
178 | error = -EACCES; |
179 | goto out; | |
180 | } | |
181 | no_nice = security_task_setnice(p, niceval); | |
182 | if (no_nice) { | |
183 | error = no_nice; | |
184 | goto out; | |
185 | } | |
186 | if (error == -ESRCH) | |
187 | error = 0; | |
188 | set_user_nice(p, niceval); | |
189 | out: | |
190 | return error; | |
191 | } | |
192 | ||
754fe8d2 | 193 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
194 | { |
195 | struct task_struct *g, *p; | |
196 | struct user_struct *user; | |
86a264ab | 197 | const struct cred *cred = current_cred(); |
1da177e4 | 198 | int error = -EINVAL; |
41487c65 | 199 | struct pid *pgrp; |
7b44ab97 | 200 | kuid_t uid; |
1da177e4 | 201 | |
3e88c553 | 202 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
203 | goto out; |
204 | ||
205 | /* normalize: avoid signed division (rounding problems) */ | |
206 | error = -ESRCH; | |
c4a4d2f4 DY |
207 | if (niceval < MIN_NICE) |
208 | niceval = MIN_NICE; | |
209 | if (niceval > MAX_NICE) | |
210 | niceval = MAX_NICE; | |
1da177e4 | 211 | |
d4581a23 | 212 | rcu_read_lock(); |
1da177e4 LT |
213 | read_lock(&tasklist_lock); |
214 | switch (which) { | |
ec94fc3d | 215 | case PRIO_PROCESS: |
216 | if (who) | |
217 | p = find_task_by_vpid(who); | |
218 | else | |
219 | p = current; | |
220 | if (p) | |
221 | error = set_one_prio(p, niceval, error); | |
222 | break; | |
223 | case PRIO_PGRP: | |
224 | if (who) | |
225 | pgrp = find_vpid(who); | |
226 | else | |
227 | pgrp = task_pgrp(current); | |
228 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
229 | error = set_one_prio(p, niceval, error); | |
230 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
231 | break; | |
232 | case PRIO_USER: | |
233 | uid = make_kuid(cred->user_ns, who); | |
234 | user = cred->user; | |
235 | if (!who) | |
236 | uid = cred->uid; | |
237 | else if (!uid_eq(uid, cred->uid)) { | |
238 | user = find_user(uid); | |
239 | if (!user) | |
86a264ab | 240 | goto out_unlock; /* No processes for this user */ |
ec94fc3d | 241 | } |
242 | do_each_thread(g, p) { | |
8639b461 | 243 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) |
ec94fc3d | 244 | error = set_one_prio(p, niceval, error); |
245 | } while_each_thread(g, p); | |
246 | if (!uid_eq(uid, cred->uid)) | |
247 | free_uid(user); /* For find_user() */ | |
248 | break; | |
1da177e4 LT |
249 | } |
250 | out_unlock: | |
251 | read_unlock(&tasklist_lock); | |
d4581a23 | 252 | rcu_read_unlock(); |
1da177e4 LT |
253 | out: |
254 | return error; | |
255 | } | |
256 | ||
257 | /* | |
258 | * Ugh. To avoid negative return values, "getpriority()" will | |
259 | * not return the normal nice-value, but a negated value that | |
260 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
261 | * to stay compatible. | |
262 | */ | |
754fe8d2 | 263 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
264 | { |
265 | struct task_struct *g, *p; | |
266 | struct user_struct *user; | |
86a264ab | 267 | const struct cred *cred = current_cred(); |
1da177e4 | 268 | long niceval, retval = -ESRCH; |
41487c65 | 269 | struct pid *pgrp; |
7b44ab97 | 270 | kuid_t uid; |
1da177e4 | 271 | |
3e88c553 | 272 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
273 | return -EINVAL; |
274 | ||
70118837 | 275 | rcu_read_lock(); |
1da177e4 LT |
276 | read_lock(&tasklist_lock); |
277 | switch (which) { | |
ec94fc3d | 278 | case PRIO_PROCESS: |
279 | if (who) | |
280 | p = find_task_by_vpid(who); | |
281 | else | |
282 | p = current; | |
283 | if (p) { | |
284 | niceval = nice_to_rlimit(task_nice(p)); | |
285 | if (niceval > retval) | |
286 | retval = niceval; | |
287 | } | |
288 | break; | |
289 | case PRIO_PGRP: | |
290 | if (who) | |
291 | pgrp = find_vpid(who); | |
292 | else | |
293 | pgrp = task_pgrp(current); | |
294 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
295 | niceval = nice_to_rlimit(task_nice(p)); | |
296 | if (niceval > retval) | |
297 | retval = niceval; | |
298 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
299 | break; | |
300 | case PRIO_USER: | |
301 | uid = make_kuid(cred->user_ns, who); | |
302 | user = cred->user; | |
303 | if (!who) | |
304 | uid = cred->uid; | |
305 | else if (!uid_eq(uid, cred->uid)) { | |
306 | user = find_user(uid); | |
307 | if (!user) | |
308 | goto out_unlock; /* No processes for this user */ | |
309 | } | |
310 | do_each_thread(g, p) { | |
8639b461 | 311 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) { |
7aa2c016 | 312 | niceval = nice_to_rlimit(task_nice(p)); |
1da177e4 LT |
313 | if (niceval > retval) |
314 | retval = niceval; | |
315 | } | |
ec94fc3d | 316 | } while_each_thread(g, p); |
317 | if (!uid_eq(uid, cred->uid)) | |
318 | free_uid(user); /* for find_user() */ | |
319 | break; | |
1da177e4 LT |
320 | } |
321 | out_unlock: | |
322 | read_unlock(&tasklist_lock); | |
70118837 | 323 | rcu_read_unlock(); |
1da177e4 LT |
324 | |
325 | return retval; | |
326 | } | |
327 | ||
1da177e4 LT |
328 | /* |
329 | * Unprivileged users may change the real gid to the effective gid | |
330 | * or vice versa. (BSD-style) | |
331 | * | |
332 | * If you set the real gid at all, or set the effective gid to a value not | |
333 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
334 | * | |
335 | * This makes it possible for a setgid program to completely drop its | |
336 | * privileges, which is often a useful assertion to make when you are doing | |
337 | * a security audit over a program. | |
338 | * | |
339 | * The general idea is that a program which uses just setregid() will be | |
340 | * 100% compatible with BSD. A program which uses just setgid() will be | |
ec94fc3d | 341 | * 100% compatible with POSIX with saved IDs. |
1da177e4 LT |
342 | * |
343 | * SMP: There are not races, the GIDs are checked only by filesystem | |
344 | * operations (as far as semantic preservation is concerned). | |
345 | */ | |
2813893f | 346 | #ifdef CONFIG_MULTIUSER |
e530dca5 | 347 | long __sys_setregid(gid_t rgid, gid_t egid) |
1da177e4 | 348 | { |
a29c33f4 | 349 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
350 | const struct cred *old; |
351 | struct cred *new; | |
1da177e4 | 352 | int retval; |
a29c33f4 EB |
353 | kgid_t krgid, kegid; |
354 | ||
355 | krgid = make_kgid(ns, rgid); | |
356 | kegid = make_kgid(ns, egid); | |
357 | ||
358 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
359 | return -EINVAL; | |
360 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
361 | return -EINVAL; | |
1da177e4 | 362 | |
d84f4f99 DH |
363 | new = prepare_creds(); |
364 | if (!new) | |
365 | return -ENOMEM; | |
366 | old = current_cred(); | |
367 | ||
d84f4f99 | 368 | retval = -EPERM; |
1da177e4 | 369 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
370 | if (gid_eq(old->gid, krgid) || |
371 | gid_eq(old->egid, krgid) || | |
c7b96acf | 372 | ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 | 373 | new->gid = krgid; |
1da177e4 | 374 | else |
d84f4f99 | 375 | goto error; |
1da177e4 LT |
376 | } |
377 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
378 | if (gid_eq(old->gid, kegid) || |
379 | gid_eq(old->egid, kegid) || | |
380 | gid_eq(old->sgid, kegid) || | |
c7b96acf | 381 | ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 | 382 | new->egid = kegid; |
756184b7 | 383 | else |
d84f4f99 | 384 | goto error; |
1da177e4 | 385 | } |
d84f4f99 | 386 | |
1da177e4 | 387 | if (rgid != (gid_t) -1 || |
a29c33f4 | 388 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
389 | new->sgid = new->egid; |
390 | new->fsgid = new->egid; | |
391 | ||
392 | return commit_creds(new); | |
393 | ||
394 | error: | |
395 | abort_creds(new); | |
396 | return retval; | |
1da177e4 LT |
397 | } |
398 | ||
e530dca5 DB |
399 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
400 | { | |
401 | return __sys_setregid(rgid, egid); | |
402 | } | |
403 | ||
1da177e4 | 404 | /* |
ec94fc3d | 405 | * setgid() is implemented like SysV w/ SAVED_IDS |
1da177e4 LT |
406 | * |
407 | * SMP: Same implicit races as above. | |
408 | */ | |
e530dca5 | 409 | long __sys_setgid(gid_t gid) |
1da177e4 | 410 | { |
a29c33f4 | 411 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
412 | const struct cred *old; |
413 | struct cred *new; | |
1da177e4 | 414 | int retval; |
a29c33f4 EB |
415 | kgid_t kgid; |
416 | ||
417 | kgid = make_kgid(ns, gid); | |
418 | if (!gid_valid(kgid)) | |
419 | return -EINVAL; | |
1da177e4 | 420 | |
d84f4f99 DH |
421 | new = prepare_creds(); |
422 | if (!new) | |
423 | return -ENOMEM; | |
424 | old = current_cred(); | |
425 | ||
d84f4f99 | 426 | retval = -EPERM; |
c7b96acf | 427 | if (ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 EB |
428 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
429 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
430 | new->egid = new->fsgid = kgid; | |
1da177e4 | 431 | else |
d84f4f99 | 432 | goto error; |
1da177e4 | 433 | |
d84f4f99 DH |
434 | return commit_creds(new); |
435 | ||
436 | error: | |
437 | abort_creds(new); | |
438 | return retval; | |
1da177e4 | 439 | } |
54e99124 | 440 | |
e530dca5 DB |
441 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
442 | { | |
443 | return __sys_setgid(gid); | |
444 | } | |
445 | ||
d84f4f99 DH |
446 | /* |
447 | * change the user struct in a credentials set to match the new UID | |
448 | */ | |
449 | static int set_user(struct cred *new) | |
1da177e4 LT |
450 | { |
451 | struct user_struct *new_user; | |
452 | ||
078de5f7 | 453 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
454 | if (!new_user) |
455 | return -EAGAIN; | |
456 | ||
72fa5997 VK |
457 | /* |
458 | * We don't fail in case of NPROC limit excess here because too many | |
459 | * poorly written programs don't check set*uid() return code, assuming | |
460 | * it never fails if called by root. We may still enforce NPROC limit | |
461 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
462 | * failure to the execve() stage. | |
463 | */ | |
78d7d407 | 464 | if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) && |
72fa5997 VK |
465 | new_user != INIT_USER) |
466 | current->flags |= PF_NPROC_EXCEEDED; | |
467 | else | |
468 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 469 | |
d84f4f99 DH |
470 | free_uid(new->user); |
471 | new->user = new_user; | |
1da177e4 LT |
472 | return 0; |
473 | } | |
474 | ||
475 | /* | |
476 | * Unprivileged users may change the real uid to the effective uid | |
477 | * or vice versa. (BSD-style) | |
478 | * | |
479 | * If you set the real uid at all, or set the effective uid to a value not | |
480 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
481 | * | |
482 | * This makes it possible for a setuid program to completely drop its | |
483 | * privileges, which is often a useful assertion to make when you are doing | |
484 | * a security audit over a program. | |
485 | * | |
486 | * The general idea is that a program which uses just setreuid() will be | |
487 | * 100% compatible with BSD. A program which uses just setuid() will be | |
ec94fc3d | 488 | * 100% compatible with POSIX with saved IDs. |
1da177e4 | 489 | */ |
e530dca5 | 490 | long __sys_setreuid(uid_t ruid, uid_t euid) |
1da177e4 | 491 | { |
a29c33f4 | 492 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
493 | const struct cred *old; |
494 | struct cred *new; | |
1da177e4 | 495 | int retval; |
a29c33f4 EB |
496 | kuid_t kruid, keuid; |
497 | ||
498 | kruid = make_kuid(ns, ruid); | |
499 | keuid = make_kuid(ns, euid); | |
500 | ||
501 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
502 | return -EINVAL; | |
503 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
504 | return -EINVAL; | |
1da177e4 | 505 | |
d84f4f99 DH |
506 | new = prepare_creds(); |
507 | if (!new) | |
508 | return -ENOMEM; | |
509 | old = current_cred(); | |
510 | ||
d84f4f99 | 511 | retval = -EPERM; |
1da177e4 | 512 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
513 | new->uid = kruid; |
514 | if (!uid_eq(old->uid, kruid) && | |
515 | !uid_eq(old->euid, kruid) && | |
c7b96acf | 516 | !ns_capable(old->user_ns, CAP_SETUID)) |
d84f4f99 | 517 | goto error; |
1da177e4 LT |
518 | } |
519 | ||
520 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
521 | new->euid = keuid; |
522 | if (!uid_eq(old->uid, keuid) && | |
523 | !uid_eq(old->euid, keuid) && | |
524 | !uid_eq(old->suid, keuid) && | |
c7b96acf | 525 | !ns_capable(old->user_ns, CAP_SETUID)) |
d84f4f99 | 526 | goto error; |
1da177e4 LT |
527 | } |
528 | ||
a29c33f4 | 529 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
530 | retval = set_user(new); |
531 | if (retval < 0) | |
532 | goto error; | |
533 | } | |
1da177e4 | 534 | if (ruid != (uid_t) -1 || |
a29c33f4 | 535 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
536 | new->suid = new->euid; |
537 | new->fsuid = new->euid; | |
1da177e4 | 538 | |
d84f4f99 DH |
539 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
540 | if (retval < 0) | |
541 | goto error; | |
1da177e4 | 542 | |
d84f4f99 | 543 | return commit_creds(new); |
1da177e4 | 544 | |
d84f4f99 DH |
545 | error: |
546 | abort_creds(new); | |
547 | return retval; | |
548 | } | |
ec94fc3d | 549 | |
e530dca5 DB |
550 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
551 | { | |
552 | return __sys_setreuid(ruid, euid); | |
553 | } | |
554 | ||
1da177e4 | 555 | /* |
ec94fc3d | 556 | * setuid() is implemented like SysV with SAVED_IDS |
557 | * | |
1da177e4 | 558 | * Note that SAVED_ID's is deficient in that a setuid root program |
ec94fc3d | 559 | * like sendmail, for example, cannot set its uid to be a normal |
1da177e4 LT |
560 | * user and then switch back, because if you're root, setuid() sets |
561 | * the saved uid too. If you don't like this, blame the bright people | |
562 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
563 | * will allow a root program to temporarily drop privileges and be able to | |
ec94fc3d | 564 | * regain them by swapping the real and effective uid. |
1da177e4 | 565 | */ |
e530dca5 | 566 | long __sys_setuid(uid_t uid) |
1da177e4 | 567 | { |
a29c33f4 | 568 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
569 | const struct cred *old; |
570 | struct cred *new; | |
1da177e4 | 571 | int retval; |
a29c33f4 EB |
572 | kuid_t kuid; |
573 | ||
574 | kuid = make_kuid(ns, uid); | |
575 | if (!uid_valid(kuid)) | |
576 | return -EINVAL; | |
1da177e4 | 577 | |
d84f4f99 DH |
578 | new = prepare_creds(); |
579 | if (!new) | |
580 | return -ENOMEM; | |
581 | old = current_cred(); | |
582 | ||
d84f4f99 | 583 | retval = -EPERM; |
c7b96acf | 584 | if (ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
585 | new->suid = new->uid = kuid; |
586 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
587 | retval = set_user(new); |
588 | if (retval < 0) | |
589 | goto error; | |
d84f4f99 | 590 | } |
a29c33f4 | 591 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 592 | goto error; |
1da177e4 | 593 | } |
1da177e4 | 594 | |
a29c33f4 | 595 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
596 | |
597 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
598 | if (retval < 0) | |
599 | goto error; | |
1da177e4 | 600 | |
d84f4f99 | 601 | return commit_creds(new); |
1da177e4 | 602 | |
d84f4f99 DH |
603 | error: |
604 | abort_creds(new); | |
605 | return retval; | |
1da177e4 LT |
606 | } |
607 | ||
e530dca5 DB |
608 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
609 | { | |
610 | return __sys_setuid(uid); | |
611 | } | |
612 | ||
1da177e4 LT |
613 | |
614 | /* | |
615 | * This function implements a generic ability to update ruid, euid, | |
616 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
617 | */ | |
e530dca5 | 618 | long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) |
1da177e4 | 619 | { |
a29c33f4 | 620 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
621 | const struct cred *old; |
622 | struct cred *new; | |
1da177e4 | 623 | int retval; |
a29c33f4 EB |
624 | kuid_t kruid, keuid, ksuid; |
625 | ||
626 | kruid = make_kuid(ns, ruid); | |
627 | keuid = make_kuid(ns, euid); | |
628 | ksuid = make_kuid(ns, suid); | |
629 | ||
630 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
631 | return -EINVAL; | |
632 | ||
633 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
634 | return -EINVAL; | |
635 | ||
636 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
637 | return -EINVAL; | |
1da177e4 | 638 | |
d84f4f99 DH |
639 | new = prepare_creds(); |
640 | if (!new) | |
641 | return -ENOMEM; | |
642 | ||
d84f4f99 | 643 | old = current_cred(); |
1da177e4 | 644 | |
d84f4f99 | 645 | retval = -EPERM; |
c7b96acf | 646 | if (!ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
647 | if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && |
648 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid)) | |
d84f4f99 | 649 | goto error; |
a29c33f4 EB |
650 | if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && |
651 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid)) | |
d84f4f99 | 652 | goto error; |
a29c33f4 EB |
653 | if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && |
654 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid)) | |
d84f4f99 | 655 | goto error; |
1da177e4 | 656 | } |
d84f4f99 | 657 | |
1da177e4 | 658 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
659 | new->uid = kruid; |
660 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
661 | retval = set_user(new); |
662 | if (retval < 0) | |
663 | goto error; | |
664 | } | |
1da177e4 | 665 | } |
d84f4f99 | 666 | if (euid != (uid_t) -1) |
a29c33f4 | 667 | new->euid = keuid; |
1da177e4 | 668 | if (suid != (uid_t) -1) |
a29c33f4 | 669 | new->suid = ksuid; |
d84f4f99 | 670 | new->fsuid = new->euid; |
1da177e4 | 671 | |
d84f4f99 DH |
672 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
673 | if (retval < 0) | |
674 | goto error; | |
1da177e4 | 675 | |
d84f4f99 | 676 | return commit_creds(new); |
1da177e4 | 677 | |
d84f4f99 DH |
678 | error: |
679 | abort_creds(new); | |
680 | return retval; | |
1da177e4 LT |
681 | } |
682 | ||
e530dca5 DB |
683 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
684 | { | |
685 | return __sys_setresuid(ruid, euid, suid); | |
686 | } | |
687 | ||
a29c33f4 | 688 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 689 | { |
86a264ab | 690 | const struct cred *cred = current_cred(); |
1da177e4 | 691 | int retval; |
a29c33f4 EB |
692 | uid_t ruid, euid, suid; |
693 | ||
694 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
695 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
696 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 697 | |
ec94fc3d | 698 | retval = put_user(ruid, ruidp); |
699 | if (!retval) { | |
700 | retval = put_user(euid, euidp); | |
701 | if (!retval) | |
702 | return put_user(suid, suidp); | |
703 | } | |
1da177e4 LT |
704 | return retval; |
705 | } | |
706 | ||
707 | /* | |
708 | * Same as above, but for rgid, egid, sgid. | |
709 | */ | |
e530dca5 | 710 | long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) |
1da177e4 | 711 | { |
a29c33f4 | 712 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
713 | const struct cred *old; |
714 | struct cred *new; | |
1da177e4 | 715 | int retval; |
a29c33f4 EB |
716 | kgid_t krgid, kegid, ksgid; |
717 | ||
718 | krgid = make_kgid(ns, rgid); | |
719 | kegid = make_kgid(ns, egid); | |
720 | ksgid = make_kgid(ns, sgid); | |
721 | ||
722 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
723 | return -EINVAL; | |
724 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
725 | return -EINVAL; | |
726 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
727 | return -EINVAL; | |
1da177e4 | 728 | |
d84f4f99 DH |
729 | new = prepare_creds(); |
730 | if (!new) | |
731 | return -ENOMEM; | |
732 | old = current_cred(); | |
733 | ||
d84f4f99 | 734 | retval = -EPERM; |
c7b96acf | 735 | if (!ns_capable(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
736 | if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && |
737 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid)) | |
d84f4f99 | 738 | goto error; |
a29c33f4 EB |
739 | if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && |
740 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid)) | |
d84f4f99 | 741 | goto error; |
a29c33f4 EB |
742 | if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && |
743 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid)) | |
d84f4f99 | 744 | goto error; |
1da177e4 | 745 | } |
d84f4f99 | 746 | |
1da177e4 | 747 | if (rgid != (gid_t) -1) |
a29c33f4 | 748 | new->gid = krgid; |
d84f4f99 | 749 | if (egid != (gid_t) -1) |
a29c33f4 | 750 | new->egid = kegid; |
1da177e4 | 751 | if (sgid != (gid_t) -1) |
a29c33f4 | 752 | new->sgid = ksgid; |
d84f4f99 | 753 | new->fsgid = new->egid; |
1da177e4 | 754 | |
d84f4f99 DH |
755 | return commit_creds(new); |
756 | ||
757 | error: | |
758 | abort_creds(new); | |
759 | return retval; | |
1da177e4 LT |
760 | } |
761 | ||
e530dca5 DB |
762 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
763 | { | |
764 | return __sys_setresgid(rgid, egid, sgid); | |
765 | } | |
766 | ||
a29c33f4 | 767 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 768 | { |
86a264ab | 769 | const struct cred *cred = current_cred(); |
1da177e4 | 770 | int retval; |
a29c33f4 EB |
771 | gid_t rgid, egid, sgid; |
772 | ||
773 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
774 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
775 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 776 | |
ec94fc3d | 777 | retval = put_user(rgid, rgidp); |
778 | if (!retval) { | |
779 | retval = put_user(egid, egidp); | |
780 | if (!retval) | |
781 | retval = put_user(sgid, sgidp); | |
782 | } | |
1da177e4 LT |
783 | |
784 | return retval; | |
785 | } | |
786 | ||
787 | ||
788 | /* | |
789 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
790 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
791 | * whatever uid it wants to). It normally shadows "euid", except when | |
792 | * explicitly set by setfsuid() or for access.. | |
793 | */ | |
e530dca5 | 794 | long __sys_setfsuid(uid_t uid) |
1da177e4 | 795 | { |
d84f4f99 DH |
796 | const struct cred *old; |
797 | struct cred *new; | |
798 | uid_t old_fsuid; | |
a29c33f4 EB |
799 | kuid_t kuid; |
800 | ||
801 | old = current_cred(); | |
802 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
803 | ||
804 | kuid = make_kuid(old->user_ns, uid); | |
805 | if (!uid_valid(kuid)) | |
806 | return old_fsuid; | |
1da177e4 | 807 | |
d84f4f99 DH |
808 | new = prepare_creds(); |
809 | if (!new) | |
a29c33f4 | 810 | return old_fsuid; |
1da177e4 | 811 | |
a29c33f4 EB |
812 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
813 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
c7b96acf | 814 | ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
815 | if (!uid_eq(kuid, old->fsuid)) { |
816 | new->fsuid = kuid; | |
d84f4f99 DH |
817 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
818 | goto change_okay; | |
1da177e4 | 819 | } |
1da177e4 LT |
820 | } |
821 | ||
d84f4f99 DH |
822 | abort_creds(new); |
823 | return old_fsuid; | |
1da177e4 | 824 | |
d84f4f99 DH |
825 | change_okay: |
826 | commit_creds(new); | |
1da177e4 LT |
827 | return old_fsuid; |
828 | } | |
829 | ||
e530dca5 DB |
830 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
831 | { | |
832 | return __sys_setfsuid(uid); | |
833 | } | |
834 | ||
1da177e4 | 835 | /* |
f42df9e6 | 836 | * Samma på svenska.. |
1da177e4 | 837 | */ |
e530dca5 | 838 | long __sys_setfsgid(gid_t gid) |
1da177e4 | 839 | { |
d84f4f99 DH |
840 | const struct cred *old; |
841 | struct cred *new; | |
842 | gid_t old_fsgid; | |
a29c33f4 EB |
843 | kgid_t kgid; |
844 | ||
845 | old = current_cred(); | |
846 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
847 | ||
848 | kgid = make_kgid(old->user_ns, gid); | |
849 | if (!gid_valid(kgid)) | |
850 | return old_fsgid; | |
d84f4f99 DH |
851 | |
852 | new = prepare_creds(); | |
853 | if (!new) | |
a29c33f4 | 854 | return old_fsgid; |
1da177e4 | 855 | |
a29c33f4 EB |
856 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
857 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
c7b96acf | 858 | ns_capable(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
859 | if (!gid_eq(kgid, old->fsgid)) { |
860 | new->fsgid = kgid; | |
d84f4f99 | 861 | goto change_okay; |
1da177e4 | 862 | } |
1da177e4 | 863 | } |
d84f4f99 | 864 | |
d84f4f99 DH |
865 | abort_creds(new); |
866 | return old_fsgid; | |
867 | ||
868 | change_okay: | |
869 | commit_creds(new); | |
1da177e4 LT |
870 | return old_fsgid; |
871 | } | |
e530dca5 DB |
872 | |
873 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) | |
874 | { | |
875 | return __sys_setfsgid(gid); | |
876 | } | |
2813893f | 877 | #endif /* CONFIG_MULTIUSER */ |
1da177e4 | 878 | |
4a22f166 SR |
879 | /** |
880 | * sys_getpid - return the thread group id of the current process | |
881 | * | |
882 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
883 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
884 | * which case the tgid is the same in all threads of the same group. | |
885 | * | |
886 | * This is SMP safe as current->tgid does not change. | |
887 | */ | |
888 | SYSCALL_DEFINE0(getpid) | |
889 | { | |
890 | return task_tgid_vnr(current); | |
891 | } | |
892 | ||
893 | /* Thread ID - the internal kernel "pid" */ | |
894 | SYSCALL_DEFINE0(gettid) | |
895 | { | |
896 | return task_pid_vnr(current); | |
897 | } | |
898 | ||
899 | /* | |
900 | * Accessing ->real_parent is not SMP-safe, it could | |
901 | * change from under us. However, we can use a stale | |
902 | * value of ->real_parent under rcu_read_lock(), see | |
903 | * release_task()->call_rcu(delayed_put_task_struct). | |
904 | */ | |
905 | SYSCALL_DEFINE0(getppid) | |
906 | { | |
907 | int pid; | |
908 | ||
909 | rcu_read_lock(); | |
910 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
911 | rcu_read_unlock(); | |
912 | ||
913 | return pid; | |
914 | } | |
915 | ||
916 | SYSCALL_DEFINE0(getuid) | |
917 | { | |
918 | /* Only we change this so SMP safe */ | |
919 | return from_kuid_munged(current_user_ns(), current_uid()); | |
920 | } | |
921 | ||
922 | SYSCALL_DEFINE0(geteuid) | |
923 | { | |
924 | /* Only we change this so SMP safe */ | |
925 | return from_kuid_munged(current_user_ns(), current_euid()); | |
926 | } | |
927 | ||
928 | SYSCALL_DEFINE0(getgid) | |
929 | { | |
930 | /* Only we change this so SMP safe */ | |
931 | return from_kgid_munged(current_user_ns(), current_gid()); | |
932 | } | |
933 | ||
934 | SYSCALL_DEFINE0(getegid) | |
935 | { | |
936 | /* Only we change this so SMP safe */ | |
937 | return from_kgid_munged(current_user_ns(), current_egid()); | |
938 | } | |
939 | ||
ca2406ed | 940 | static void do_sys_times(struct tms *tms) |
f06febc9 | 941 | { |
5613fda9 | 942 | u64 tgutime, tgstime, cutime, cstime; |
f06febc9 | 943 | |
e80d0a1a | 944 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
945 | cutime = current->signal->cutime; |
946 | cstime = current->signal->cstime; | |
5613fda9 FW |
947 | tms->tms_utime = nsec_to_clock_t(tgutime); |
948 | tms->tms_stime = nsec_to_clock_t(tgstime); | |
949 | tms->tms_cutime = nsec_to_clock_t(cutime); | |
950 | tms->tms_cstime = nsec_to_clock_t(cstime); | |
f06febc9 FM |
951 | } |
952 | ||
58fd3aa2 | 953 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 954 | { |
1da177e4 LT |
955 | if (tbuf) { |
956 | struct tms tmp; | |
f06febc9 FM |
957 | |
958 | do_sys_times(&tmp); | |
1da177e4 LT |
959 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
960 | return -EFAULT; | |
961 | } | |
e3d5a27d | 962 | force_successful_syscall_return(); |
1da177e4 LT |
963 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
964 | } | |
965 | ||
ca2406ed AV |
966 | #ifdef CONFIG_COMPAT |
967 | static compat_clock_t clock_t_to_compat_clock_t(clock_t x) | |
968 | { | |
969 | return compat_jiffies_to_clock_t(clock_t_to_jiffies(x)); | |
970 | } | |
971 | ||
972 | COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf) | |
973 | { | |
974 | if (tbuf) { | |
975 | struct tms tms; | |
976 | struct compat_tms tmp; | |
977 | ||
978 | do_sys_times(&tms); | |
979 | /* Convert our struct tms to the compat version. */ | |
980 | tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime); | |
981 | tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime); | |
982 | tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime); | |
983 | tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime); | |
984 | if (copy_to_user(tbuf, &tmp, sizeof(tmp))) | |
985 | return -EFAULT; | |
986 | } | |
987 | force_successful_syscall_return(); | |
988 | return compat_jiffies_to_clock_t(jiffies); | |
989 | } | |
990 | #endif | |
991 | ||
1da177e4 LT |
992 | /* |
993 | * This needs some heavy checking ... | |
994 | * I just haven't the stomach for it. I also don't fully | |
995 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
996 | * | |
997 | * OK, I think I have the protection semantics right.... this is really | |
998 | * only important on a multi-user system anyway, to make sure one user | |
999 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
1000 | * | |
98611e4e | 1001 | * !PF_FORKNOEXEC check to conform completely to POSIX. |
1da177e4 | 1002 | */ |
b290ebe2 | 1003 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
1004 | { |
1005 | struct task_struct *p; | |
ee0acf90 | 1006 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
1007 | struct pid *pgrp; |
1008 | int err; | |
1da177e4 LT |
1009 | |
1010 | if (!pid) | |
b488893a | 1011 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
1012 | if (!pgid) |
1013 | pgid = pid; | |
1014 | if (pgid < 0) | |
1015 | return -EINVAL; | |
950eaaca | 1016 | rcu_read_lock(); |
1da177e4 LT |
1017 | |
1018 | /* From this point forward we keep holding onto the tasklist lock | |
1019 | * so that our parent does not change from under us. -DaveM | |
1020 | */ | |
1021 | write_lock_irq(&tasklist_lock); | |
1022 | ||
1023 | err = -ESRCH; | |
4e021306 | 1024 | p = find_task_by_vpid(pid); |
1da177e4 LT |
1025 | if (!p) |
1026 | goto out; | |
1027 | ||
1028 | err = -EINVAL; | |
1029 | if (!thread_group_leader(p)) | |
1030 | goto out; | |
1031 | ||
4e021306 | 1032 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 1033 | err = -EPERM; |
41487c65 | 1034 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
1035 | goto out; |
1036 | err = -EACCES; | |
98611e4e | 1037 | if (!(p->flags & PF_FORKNOEXEC)) |
1da177e4 LT |
1038 | goto out; |
1039 | } else { | |
1040 | err = -ESRCH; | |
ee0acf90 | 1041 | if (p != group_leader) |
1da177e4 LT |
1042 | goto out; |
1043 | } | |
1044 | ||
1045 | err = -EPERM; | |
1046 | if (p->signal->leader) | |
1047 | goto out; | |
1048 | ||
4e021306 | 1049 | pgrp = task_pid(p); |
1da177e4 | 1050 | if (pgid != pid) { |
b488893a | 1051 | struct task_struct *g; |
1da177e4 | 1052 | |
4e021306 ON |
1053 | pgrp = find_vpid(pgid); |
1054 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 1055 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 1056 | goto out; |
1da177e4 LT |
1057 | } |
1058 | ||
1da177e4 LT |
1059 | err = security_task_setpgid(p, pgid); |
1060 | if (err) | |
1061 | goto out; | |
1062 | ||
1b0f7ffd | 1063 | if (task_pgrp(p) != pgrp) |
83beaf3c | 1064 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
1065 | |
1066 | err = 0; | |
1067 | out: | |
1068 | /* All paths lead to here, thus we are safe. -DaveM */ | |
1069 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 1070 | rcu_read_unlock(); |
1da177e4 LT |
1071 | return err; |
1072 | } | |
1073 | ||
192c5807 | 1074 | static int do_getpgid(pid_t pid) |
1da177e4 | 1075 | { |
12a3de0a ON |
1076 | struct task_struct *p; |
1077 | struct pid *grp; | |
1078 | int retval; | |
1079 | ||
1080 | rcu_read_lock(); | |
756184b7 | 1081 | if (!pid) |
12a3de0a | 1082 | grp = task_pgrp(current); |
756184b7 | 1083 | else { |
1da177e4 | 1084 | retval = -ESRCH; |
12a3de0a ON |
1085 | p = find_task_by_vpid(pid); |
1086 | if (!p) | |
1087 | goto out; | |
1088 | grp = task_pgrp(p); | |
1089 | if (!grp) | |
1090 | goto out; | |
1091 | ||
1092 | retval = security_task_getpgid(p); | |
1093 | if (retval) | |
1094 | goto out; | |
1da177e4 | 1095 | } |
12a3de0a ON |
1096 | retval = pid_vnr(grp); |
1097 | out: | |
1098 | rcu_read_unlock(); | |
1099 | return retval; | |
1da177e4 LT |
1100 | } |
1101 | ||
192c5807 DB |
1102 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1103 | { | |
1104 | return do_getpgid(pid); | |
1105 | } | |
1106 | ||
1da177e4 LT |
1107 | #ifdef __ARCH_WANT_SYS_GETPGRP |
1108 | ||
dbf040d9 | 1109 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1110 | { |
192c5807 | 1111 | return do_getpgid(0); |
1da177e4 LT |
1112 | } |
1113 | ||
1114 | #endif | |
1115 | ||
dbf040d9 | 1116 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1117 | { |
1dd768c0 ON |
1118 | struct task_struct *p; |
1119 | struct pid *sid; | |
1120 | int retval; | |
1121 | ||
1122 | rcu_read_lock(); | |
756184b7 | 1123 | if (!pid) |
1dd768c0 | 1124 | sid = task_session(current); |
756184b7 | 1125 | else { |
1da177e4 | 1126 | retval = -ESRCH; |
1dd768c0 ON |
1127 | p = find_task_by_vpid(pid); |
1128 | if (!p) | |
1129 | goto out; | |
1130 | sid = task_session(p); | |
1131 | if (!sid) | |
1132 | goto out; | |
1133 | ||
1134 | retval = security_task_getsid(p); | |
1135 | if (retval) | |
1136 | goto out; | |
1da177e4 | 1137 | } |
1dd768c0 ON |
1138 | retval = pid_vnr(sid); |
1139 | out: | |
1140 | rcu_read_unlock(); | |
1141 | return retval; | |
1da177e4 LT |
1142 | } |
1143 | ||
81dabb46 ON |
1144 | static void set_special_pids(struct pid *pid) |
1145 | { | |
1146 | struct task_struct *curr = current->group_leader; | |
1147 | ||
1148 | if (task_session(curr) != pid) | |
1149 | change_pid(curr, PIDTYPE_SID, pid); | |
1150 | ||
1151 | if (task_pgrp(curr) != pid) | |
1152 | change_pid(curr, PIDTYPE_PGID, pid); | |
1153 | } | |
1154 | ||
e2aaa9f4 | 1155 | int ksys_setsid(void) |
1da177e4 | 1156 | { |
e19f247a | 1157 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1158 | struct pid *sid = task_pid(group_leader); |
1159 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1160 | int err = -EPERM; |
1161 | ||
1da177e4 | 1162 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1163 | /* Fail if I am already a session leader */ |
1164 | if (group_leader->signal->leader) | |
1165 | goto out; | |
1166 | ||
430c6231 ON |
1167 | /* Fail if a process group id already exists that equals the |
1168 | * proposed session id. | |
390e2ff0 | 1169 | */ |
6806aac6 | 1170 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1171 | goto out; |
1172 | ||
e19f247a | 1173 | group_leader->signal->leader = 1; |
81dabb46 | 1174 | set_special_pids(sid); |
24ec839c | 1175 | |
9c9f4ded | 1176 | proc_clear_tty(group_leader); |
24ec839c | 1177 | |
e4cc0a9c | 1178 | err = session; |
1da177e4 LT |
1179 | out: |
1180 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1181 | if (err > 0) { |
0d0df599 | 1182 | proc_sid_connector(group_leader); |
5091faa4 MG |
1183 | sched_autogroup_create_attach(group_leader); |
1184 | } | |
1da177e4 LT |
1185 | return err; |
1186 | } | |
1187 | ||
e2aaa9f4 DB |
1188 | SYSCALL_DEFINE0(setsid) |
1189 | { | |
1190 | return ksys_setsid(); | |
1191 | } | |
1192 | ||
1da177e4 LT |
1193 | DECLARE_RWSEM(uts_sem); |
1194 | ||
e28cbf22 CH |
1195 | #ifdef COMPAT_UTS_MACHINE |
1196 | #define override_architecture(name) \ | |
46da2766 | 1197 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1198 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1199 | sizeof(COMPAT_UTS_MACHINE))) | |
1200 | #else | |
1201 | #define override_architecture(name) 0 | |
1202 | #endif | |
1203 | ||
be27425d AK |
1204 | /* |
1205 | * Work around broken programs that cannot handle "Linux 3.0". | |
1206 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
39afb5ee | 1207 | * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60. |
be27425d | 1208 | */ |
2702b152 | 1209 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1210 | { |
1211 | int ret = 0; | |
be27425d AK |
1212 | |
1213 | if (current->personality & UNAME26) { | |
2702b152 KC |
1214 | const char *rest = UTS_RELEASE; |
1215 | char buf[65] = { 0 }; | |
be27425d AK |
1216 | int ndots = 0; |
1217 | unsigned v; | |
2702b152 | 1218 | size_t copy; |
be27425d AK |
1219 | |
1220 | while (*rest) { | |
1221 | if (*rest == '.' && ++ndots >= 3) | |
1222 | break; | |
1223 | if (!isdigit(*rest) && *rest != '.') | |
1224 | break; | |
1225 | rest++; | |
1226 | } | |
39afb5ee | 1227 | v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60; |
31fd84b9 | 1228 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1229 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1230 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1231 | } |
1232 | return ret; | |
1233 | } | |
1234 | ||
e48fbb69 | 1235 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 LT |
1236 | { |
1237 | int errno = 0; | |
1238 | ||
1239 | down_read(&uts_sem); | |
e9ff3990 | 1240 | if (copy_to_user(name, utsname(), sizeof *name)) |
1da177e4 LT |
1241 | errno = -EFAULT; |
1242 | up_read(&uts_sem); | |
e28cbf22 | 1243 | |
be27425d AK |
1244 | if (!errno && override_release(name->release, sizeof(name->release))) |
1245 | errno = -EFAULT; | |
e28cbf22 CH |
1246 | if (!errno && override_architecture(name)) |
1247 | errno = -EFAULT; | |
1da177e4 LT |
1248 | return errno; |
1249 | } | |
1250 | ||
5cacdb4a CH |
1251 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1252 | /* | |
1253 | * Old cruft | |
1254 | */ | |
1255 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1256 | { | |
1257 | int error = 0; | |
1258 | ||
1259 | if (!name) | |
1260 | return -EFAULT; | |
1261 | ||
1262 | down_read(&uts_sem); | |
1263 | if (copy_to_user(name, utsname(), sizeof(*name))) | |
1264 | error = -EFAULT; | |
1265 | up_read(&uts_sem); | |
1266 | ||
be27425d AK |
1267 | if (!error && override_release(name->release, sizeof(name->release))) |
1268 | error = -EFAULT; | |
5cacdb4a CH |
1269 | if (!error && override_architecture(name)) |
1270 | error = -EFAULT; | |
1271 | return error; | |
1272 | } | |
1273 | ||
1274 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1275 | { | |
1276 | int error; | |
1277 | ||
1278 | if (!name) | |
1279 | return -EFAULT; | |
1280 | if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname))) | |
1281 | return -EFAULT; | |
1282 | ||
1283 | down_read(&uts_sem); | |
1284 | error = __copy_to_user(&name->sysname, &utsname()->sysname, | |
1285 | __OLD_UTS_LEN); | |
1286 | error |= __put_user(0, name->sysname + __OLD_UTS_LEN); | |
1287 | error |= __copy_to_user(&name->nodename, &utsname()->nodename, | |
1288 | __OLD_UTS_LEN); | |
1289 | error |= __put_user(0, name->nodename + __OLD_UTS_LEN); | |
1290 | error |= __copy_to_user(&name->release, &utsname()->release, | |
1291 | __OLD_UTS_LEN); | |
1292 | error |= __put_user(0, name->release + __OLD_UTS_LEN); | |
1293 | error |= __copy_to_user(&name->version, &utsname()->version, | |
1294 | __OLD_UTS_LEN); | |
1295 | error |= __put_user(0, name->version + __OLD_UTS_LEN); | |
1296 | error |= __copy_to_user(&name->machine, &utsname()->machine, | |
1297 | __OLD_UTS_LEN); | |
1298 | error |= __put_user(0, name->machine + __OLD_UTS_LEN); | |
1299 | up_read(&uts_sem); | |
1300 | ||
1301 | if (!error && override_architecture(name)) | |
1302 | error = -EFAULT; | |
be27425d AK |
1303 | if (!error && override_release(name->release, sizeof(name->release))) |
1304 | error = -EFAULT; | |
5cacdb4a CH |
1305 | return error ? -EFAULT : 0; |
1306 | } | |
1307 | #endif | |
1308 | ||
5a8a82b1 | 1309 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1310 | { |
1311 | int errno; | |
1312 | char tmp[__NEW_UTS_LEN]; | |
1313 | ||
bb96a6f5 | 1314 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1315 | return -EPERM; |
fc832ad3 | 1316 | |
1da177e4 LT |
1317 | if (len < 0 || len > __NEW_UTS_LEN) |
1318 | return -EINVAL; | |
1319 | down_write(&uts_sem); | |
1320 | errno = -EFAULT; | |
1321 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1322 | struct new_utsname *u = utsname(); |
1323 | ||
1324 | memcpy(u->nodename, tmp, len); | |
1325 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1326 | errno = 0; |
499eea6b | 1327 | uts_proc_notify(UTS_PROC_HOSTNAME); |
1da177e4 LT |
1328 | } |
1329 | up_write(&uts_sem); | |
1330 | return errno; | |
1331 | } | |
1332 | ||
1333 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1334 | ||
5a8a82b1 | 1335 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 LT |
1336 | { |
1337 | int i, errno; | |
9679e4dd | 1338 | struct new_utsname *u; |
1da177e4 LT |
1339 | |
1340 | if (len < 0) | |
1341 | return -EINVAL; | |
1342 | down_read(&uts_sem); | |
9679e4dd AM |
1343 | u = utsname(); |
1344 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1345 | if (i > len) |
1346 | i = len; | |
1347 | errno = 0; | |
9679e4dd | 1348 | if (copy_to_user(name, u->nodename, i)) |
1da177e4 LT |
1349 | errno = -EFAULT; |
1350 | up_read(&uts_sem); | |
1351 | return errno; | |
1352 | } | |
1353 | ||
1354 | #endif | |
1355 | ||
1356 | /* | |
1357 | * Only setdomainname; getdomainname can be implemented by calling | |
1358 | * uname() | |
1359 | */ | |
5a8a82b1 | 1360 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1361 | { |
1362 | int errno; | |
1363 | char tmp[__NEW_UTS_LEN]; | |
1364 | ||
fc832ad3 | 1365 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1366 | return -EPERM; |
1367 | if (len < 0 || len > __NEW_UTS_LEN) | |
1368 | return -EINVAL; | |
1369 | ||
1370 | down_write(&uts_sem); | |
1371 | errno = -EFAULT; | |
1372 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1373 | struct new_utsname *u = utsname(); |
1374 | ||
1375 | memcpy(u->domainname, tmp, len); | |
1376 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1377 | errno = 0; |
499eea6b | 1378 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
1da177e4 LT |
1379 | } |
1380 | up_write(&uts_sem); | |
1381 | return errno; | |
1382 | } | |
1383 | ||
e48fbb69 | 1384 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1385 | { |
b9518345 JS |
1386 | struct rlimit value; |
1387 | int ret; | |
1388 | ||
1389 | ret = do_prlimit(current, resource, NULL, &value); | |
1390 | if (!ret) | |
1391 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1392 | ||
1393 | return ret; | |
1da177e4 LT |
1394 | } |
1395 | ||
d9e968cb AV |
1396 | #ifdef CONFIG_COMPAT |
1397 | ||
1398 | COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource, | |
1399 | struct compat_rlimit __user *, rlim) | |
1400 | { | |
1401 | struct rlimit r; | |
1402 | struct compat_rlimit r32; | |
1403 | ||
1404 | if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit))) | |
1405 | return -EFAULT; | |
1406 | ||
1407 | if (r32.rlim_cur == COMPAT_RLIM_INFINITY) | |
1408 | r.rlim_cur = RLIM_INFINITY; | |
1409 | else | |
1410 | r.rlim_cur = r32.rlim_cur; | |
1411 | if (r32.rlim_max == COMPAT_RLIM_INFINITY) | |
1412 | r.rlim_max = RLIM_INFINITY; | |
1413 | else | |
1414 | r.rlim_max = r32.rlim_max; | |
1415 | return do_prlimit(current, resource, &r, NULL); | |
1416 | } | |
1417 | ||
1418 | COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource, | |
1419 | struct compat_rlimit __user *, rlim) | |
1420 | { | |
1421 | struct rlimit r; | |
1422 | int ret; | |
1423 | ||
1424 | ret = do_prlimit(current, resource, NULL, &r); | |
1425 | if (!ret) { | |
58c7ffc0 | 1426 | struct compat_rlimit r32; |
d9e968cb AV |
1427 | if (r.rlim_cur > COMPAT_RLIM_INFINITY) |
1428 | r32.rlim_cur = COMPAT_RLIM_INFINITY; | |
1429 | else | |
1430 | r32.rlim_cur = r.rlim_cur; | |
1431 | if (r.rlim_max > COMPAT_RLIM_INFINITY) | |
1432 | r32.rlim_max = COMPAT_RLIM_INFINITY; | |
1433 | else | |
1434 | r32.rlim_max = r.rlim_max; | |
1435 | ||
1436 | if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit))) | |
1437 | return -EFAULT; | |
1438 | } | |
1439 | return ret; | |
1440 | } | |
1441 | ||
1442 | #endif | |
1443 | ||
1da177e4 LT |
1444 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT |
1445 | ||
1446 | /* | |
1447 | * Back compatibility for getrlimit. Needed for some apps. | |
1448 | */ | |
e48fbb69 HC |
1449 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1450 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1451 | { |
1452 | struct rlimit x; | |
1453 | if (resource >= RLIM_NLIMITS) | |
1454 | return -EINVAL; | |
1455 | ||
1456 | task_lock(current->group_leader); | |
1457 | x = current->signal->rlim[resource]; | |
1458 | task_unlock(current->group_leader); | |
756184b7 | 1459 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1460 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1461 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 | 1462 | x.rlim_max = 0x7FFFFFFF; |
ec94fc3d | 1463 | return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0; |
1da177e4 LT |
1464 | } |
1465 | ||
613763a1 AV |
1466 | #ifdef CONFIG_COMPAT |
1467 | COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, | |
1468 | struct compat_rlimit __user *, rlim) | |
1469 | { | |
1470 | struct rlimit r; | |
1471 | ||
1472 | if (resource >= RLIM_NLIMITS) | |
1473 | return -EINVAL; | |
1474 | ||
1475 | task_lock(current->group_leader); | |
1476 | r = current->signal->rlim[resource]; | |
1477 | task_unlock(current->group_leader); | |
1478 | if (r.rlim_cur > 0x7FFFFFFF) | |
1479 | r.rlim_cur = 0x7FFFFFFF; | |
1480 | if (r.rlim_max > 0x7FFFFFFF) | |
1481 | r.rlim_max = 0x7FFFFFFF; | |
1482 | ||
1483 | if (put_user(r.rlim_cur, &rlim->rlim_cur) || | |
1484 | put_user(r.rlim_max, &rlim->rlim_max)) | |
1485 | return -EFAULT; | |
1486 | return 0; | |
1487 | } | |
1488 | #endif | |
1489 | ||
1da177e4 LT |
1490 | #endif |
1491 | ||
c022a0ac JS |
1492 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1493 | { | |
1494 | #if BITS_PER_LONG < 64 | |
1495 | return rlim64 >= ULONG_MAX; | |
1496 | #else | |
1497 | return rlim64 == RLIM64_INFINITY; | |
1498 | #endif | |
1499 | } | |
1500 | ||
1501 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1502 | { | |
1503 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1504 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1505 | else | |
1506 | rlim64->rlim_cur = rlim->rlim_cur; | |
1507 | if (rlim->rlim_max == RLIM_INFINITY) | |
1508 | rlim64->rlim_max = RLIM64_INFINITY; | |
1509 | else | |
1510 | rlim64->rlim_max = rlim->rlim_max; | |
1511 | } | |
1512 | ||
1513 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1514 | { | |
1515 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1516 | rlim->rlim_cur = RLIM_INFINITY; | |
1517 | else | |
1518 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1519 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1520 | rlim->rlim_max = RLIM_INFINITY; | |
1521 | else | |
1522 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1523 | } | |
1524 | ||
1c1e618d | 1525 | /* make sure you are allowed to change @tsk limits before calling this */ |
5b41535a JS |
1526 | int do_prlimit(struct task_struct *tsk, unsigned int resource, |
1527 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1da177e4 | 1528 | { |
5b41535a | 1529 | struct rlimit *rlim; |
86f162f4 | 1530 | int retval = 0; |
1da177e4 LT |
1531 | |
1532 | if (resource >= RLIM_NLIMITS) | |
1533 | return -EINVAL; | |
5b41535a JS |
1534 | if (new_rlim) { |
1535 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1536 | return -EINVAL; | |
1537 | if (resource == RLIMIT_NOFILE && | |
1538 | new_rlim->rlim_max > sysctl_nr_open) | |
1539 | return -EPERM; | |
1540 | } | |
1da177e4 | 1541 | |
1c1e618d JS |
1542 | /* protect tsk->signal and tsk->sighand from disappearing */ |
1543 | read_lock(&tasklist_lock); | |
1544 | if (!tsk->sighand) { | |
1545 | retval = -ESRCH; | |
1546 | goto out; | |
1547 | } | |
1548 | ||
5b41535a | 1549 | rlim = tsk->signal->rlim + resource; |
86f162f4 | 1550 | task_lock(tsk->group_leader); |
5b41535a | 1551 | if (new_rlim) { |
fc832ad3 SH |
1552 | /* Keep the capable check against init_user_ns until |
1553 | cgroups can contain all limits */ | |
5b41535a JS |
1554 | if (new_rlim->rlim_max > rlim->rlim_max && |
1555 | !capable(CAP_SYS_RESOURCE)) | |
1556 | retval = -EPERM; | |
1557 | if (!retval) | |
cad4ea54 | 1558 | retval = security_task_setrlimit(tsk, resource, new_rlim); |
5b41535a JS |
1559 | if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) { |
1560 | /* | |
1561 | * The caller is asking for an immediate RLIMIT_CPU | |
1562 | * expiry. But we use the zero value to mean "it was | |
1563 | * never set". So let's cheat and make it one second | |
1564 | * instead | |
1565 | */ | |
1566 | new_rlim->rlim_cur = 1; | |
1567 | } | |
1568 | } | |
1569 | if (!retval) { | |
1570 | if (old_rlim) | |
1571 | *old_rlim = *rlim; | |
1572 | if (new_rlim) | |
1573 | *rlim = *new_rlim; | |
9926e4c7 | 1574 | } |
7855c35d | 1575 | task_unlock(tsk->group_leader); |
1da177e4 | 1576 | |
d3561f78 AM |
1577 | /* |
1578 | * RLIMIT_CPU handling. Note that the kernel fails to return an error | |
1579 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a | |
1580 | * very long-standing error, and fixing it now risks breakage of | |
1581 | * applications, so we live with it | |
1582 | */ | |
5b41535a | 1583 | if (!retval && new_rlim && resource == RLIMIT_CPU && |
baa73d9e NP |
1584 | new_rlim->rlim_cur != RLIM_INFINITY && |
1585 | IS_ENABLED(CONFIG_POSIX_TIMERS)) | |
5b41535a | 1586 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); |
ec9e16ba | 1587 | out: |
1c1e618d | 1588 | read_unlock(&tasklist_lock); |
2fb9d268 | 1589 | return retval; |
1da177e4 LT |
1590 | } |
1591 | ||
c022a0ac | 1592 | /* rcu lock must be held */ |
791ec491 SS |
1593 | static int check_prlimit_permission(struct task_struct *task, |
1594 | unsigned int flags) | |
c022a0ac JS |
1595 | { |
1596 | const struct cred *cred = current_cred(), *tcred; | |
791ec491 | 1597 | bool id_match; |
c022a0ac | 1598 | |
fc832ad3 SH |
1599 | if (current == task) |
1600 | return 0; | |
c022a0ac | 1601 | |
fc832ad3 | 1602 | tcred = __task_cred(task); |
791ec491 SS |
1603 | id_match = (uid_eq(cred->uid, tcred->euid) && |
1604 | uid_eq(cred->uid, tcred->suid) && | |
1605 | uid_eq(cred->uid, tcred->uid) && | |
1606 | gid_eq(cred->gid, tcred->egid) && | |
1607 | gid_eq(cred->gid, tcred->sgid) && | |
1608 | gid_eq(cred->gid, tcred->gid)); | |
1609 | if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) | |
1610 | return -EPERM; | |
fc832ad3 | 1611 | |
791ec491 | 1612 | return security_task_prlimit(cred, tcred, flags); |
c022a0ac JS |
1613 | } |
1614 | ||
1615 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1616 | const struct rlimit64 __user *, new_rlim, | |
1617 | struct rlimit64 __user *, old_rlim) | |
1618 | { | |
1619 | struct rlimit64 old64, new64; | |
1620 | struct rlimit old, new; | |
1621 | struct task_struct *tsk; | |
791ec491 | 1622 | unsigned int checkflags = 0; |
c022a0ac JS |
1623 | int ret; |
1624 | ||
791ec491 SS |
1625 | if (old_rlim) |
1626 | checkflags |= LSM_PRLIMIT_READ; | |
1627 | ||
c022a0ac JS |
1628 | if (new_rlim) { |
1629 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1630 | return -EFAULT; | |
1631 | rlim64_to_rlim(&new64, &new); | |
791ec491 | 1632 | checkflags |= LSM_PRLIMIT_WRITE; |
c022a0ac JS |
1633 | } |
1634 | ||
1635 | rcu_read_lock(); | |
1636 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1637 | if (!tsk) { | |
1638 | rcu_read_unlock(); | |
1639 | return -ESRCH; | |
1640 | } | |
791ec491 | 1641 | ret = check_prlimit_permission(tsk, checkflags); |
c022a0ac JS |
1642 | if (ret) { |
1643 | rcu_read_unlock(); | |
1644 | return ret; | |
1645 | } | |
1646 | get_task_struct(tsk); | |
1647 | rcu_read_unlock(); | |
1648 | ||
1649 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1650 | old_rlim ? &old : NULL); | |
1651 | ||
1652 | if (!ret && old_rlim) { | |
1653 | rlim_to_rlim64(&old, &old64); | |
1654 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1655 | ret = -EFAULT; | |
1656 | } | |
1657 | ||
1658 | put_task_struct(tsk); | |
1659 | return ret; | |
1660 | } | |
1661 | ||
7855c35d JS |
1662 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1663 | { | |
1664 | struct rlimit new_rlim; | |
1665 | ||
1666 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1667 | return -EFAULT; | |
5b41535a | 1668 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1669 | } |
1670 | ||
1da177e4 LT |
1671 | /* |
1672 | * It would make sense to put struct rusage in the task_struct, | |
1673 | * except that would make the task_struct be *really big*. After | |
1674 | * task_struct gets moved into malloc'ed memory, it would | |
1675 | * make sense to do this. It will make moving the rest of the information | |
1676 | * a lot simpler! (Which we're not doing right now because we're not | |
1677 | * measuring them yet). | |
1678 | * | |
1da177e4 LT |
1679 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1680 | * races with threads incrementing their own counters. But since word | |
1681 | * reads are atomic, we either get new values or old values and we don't | |
1682 | * care which for the sums. We always take the siglock to protect reading | |
1683 | * the c* fields from p->signal from races with exit.c updating those | |
1684 | * fields when reaping, so a sample either gets all the additions of a | |
1685 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1686 | * |
de047c1b RT |
1687 | * Locking: |
1688 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1689 | * for the cases current multithreaded, non-current single threaded | |
1690 | * non-current multithreaded. Thread traversal is now safe with | |
1691 | * the siglock held. | |
1692 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1693 | * single threaded, as no one else can take our signal_struct away, no one | |
1694 | * else can reap the children to update signal->c* counters, and no one else | |
1695 | * can race with the signal-> fields. If we do not take any lock, the | |
1696 | * signal-> fields could be read out of order while another thread was just | |
1697 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1698 | * On the writer side, write memory barrier is implied in __exit_signal | |
1699 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1700 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1701 | * |
1da177e4 LT |
1702 | */ |
1703 | ||
f06febc9 | 1704 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1705 | { |
679c9cd4 SK |
1706 | r->ru_nvcsw += t->nvcsw; |
1707 | r->ru_nivcsw += t->nivcsw; | |
1708 | r->ru_minflt += t->min_flt; | |
1709 | r->ru_majflt += t->maj_flt; | |
1710 | r->ru_inblock += task_io_get_inblock(t); | |
1711 | r->ru_oublock += task_io_get_oublock(t); | |
1712 | } | |
1713 | ||
ce72a16f | 1714 | void getrusage(struct task_struct *p, int who, struct rusage *r) |
1da177e4 LT |
1715 | { |
1716 | struct task_struct *t; | |
1717 | unsigned long flags; | |
5613fda9 | 1718 | u64 tgutime, tgstime, utime, stime; |
1f10206c | 1719 | unsigned long maxrss = 0; |
1da177e4 | 1720 | |
ec94fc3d | 1721 | memset((char *)r, 0, sizeof (*r)); |
64861634 | 1722 | utime = stime = 0; |
1da177e4 | 1723 | |
679c9cd4 | 1724 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1725 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1726 | accumulate_thread_rusage(p, r); |
1f10206c | 1727 | maxrss = p->signal->maxrss; |
679c9cd4 SK |
1728 | goto out; |
1729 | } | |
1730 | ||
d6cf723a | 1731 | if (!lock_task_sighand(p, &flags)) |
de047c1b | 1732 | return; |
0f59cc4a | 1733 | |
1da177e4 | 1734 | switch (who) { |
ec94fc3d | 1735 | case RUSAGE_BOTH: |
1736 | case RUSAGE_CHILDREN: | |
1737 | utime = p->signal->cutime; | |
1738 | stime = p->signal->cstime; | |
1739 | r->ru_nvcsw = p->signal->cnvcsw; | |
1740 | r->ru_nivcsw = p->signal->cnivcsw; | |
1741 | r->ru_minflt = p->signal->cmin_flt; | |
1742 | r->ru_majflt = p->signal->cmaj_flt; | |
1743 | r->ru_inblock = p->signal->cinblock; | |
1744 | r->ru_oublock = p->signal->coublock; | |
1745 | maxrss = p->signal->cmaxrss; | |
1746 | ||
1747 | if (who == RUSAGE_CHILDREN) | |
1da177e4 | 1748 | break; |
0f59cc4a | 1749 | |
ec94fc3d | 1750 | case RUSAGE_SELF: |
1751 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); | |
1752 | utime += tgutime; | |
1753 | stime += tgstime; | |
1754 | r->ru_nvcsw += p->signal->nvcsw; | |
1755 | r->ru_nivcsw += p->signal->nivcsw; | |
1756 | r->ru_minflt += p->signal->min_flt; | |
1757 | r->ru_majflt += p->signal->maj_flt; | |
1758 | r->ru_inblock += p->signal->inblock; | |
1759 | r->ru_oublock += p->signal->oublock; | |
1760 | if (maxrss < p->signal->maxrss) | |
1761 | maxrss = p->signal->maxrss; | |
1762 | t = p; | |
1763 | do { | |
1764 | accumulate_thread_rusage(t, r); | |
1765 | } while_each_thread(p, t); | |
1766 | break; | |
1767 | ||
1768 | default: | |
1769 | BUG(); | |
1da177e4 | 1770 | } |
de047c1b | 1771 | unlock_task_sighand(p, &flags); |
de047c1b | 1772 | |
679c9cd4 | 1773 | out: |
5613fda9 FW |
1774 | r->ru_utime = ns_to_timeval(utime); |
1775 | r->ru_stime = ns_to_timeval(stime); | |
1f10206c JP |
1776 | |
1777 | if (who != RUSAGE_CHILDREN) { | |
1778 | struct mm_struct *mm = get_task_mm(p); | |
ec94fc3d | 1779 | |
1f10206c JP |
1780 | if (mm) { |
1781 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1782 | mmput(mm); | |
1783 | } | |
1784 | } | |
1785 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1da177e4 LT |
1786 | } |
1787 | ||
ce72a16f | 1788 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 LT |
1789 | { |
1790 | struct rusage r; | |
ec94fc3d | 1791 | |
679c9cd4 SK |
1792 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1793 | who != RUSAGE_THREAD) | |
1da177e4 | 1794 | return -EINVAL; |
ce72a16f AV |
1795 | |
1796 | getrusage(current, who, &r); | |
1797 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; | |
1da177e4 LT |
1798 | } |
1799 | ||
8d2d5c4a AV |
1800 | #ifdef CONFIG_COMPAT |
1801 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1802 | { | |
1803 | struct rusage r; | |
1804 | ||
1805 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1806 | who != RUSAGE_THREAD) | |
1807 | return -EINVAL; | |
1808 | ||
ce72a16f | 1809 | getrusage(current, who, &r); |
8d2d5c4a AV |
1810 | return put_compat_rusage(&r, ru); |
1811 | } | |
1812 | #endif | |
1813 | ||
e48fbb69 | 1814 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
1815 | { |
1816 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1817 | return mask; | |
1818 | } | |
3b7391de | 1819 | |
6e399cd1 | 1820 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
b32dfe37 | 1821 | { |
2903ff01 | 1822 | struct fd exe; |
6e399cd1 | 1823 | struct file *old_exe, *exe_file; |
496ad9aa | 1824 | struct inode *inode; |
2903ff01 | 1825 | int err; |
b32dfe37 | 1826 | |
2903ff01 AV |
1827 | exe = fdget(fd); |
1828 | if (!exe.file) | |
b32dfe37 CG |
1829 | return -EBADF; |
1830 | ||
496ad9aa | 1831 | inode = file_inode(exe.file); |
b32dfe37 CG |
1832 | |
1833 | /* | |
1834 | * Because the original mm->exe_file points to executable file, make | |
1835 | * sure that this one is executable as well, to avoid breaking an | |
1836 | * overall picture. | |
1837 | */ | |
1838 | err = -EACCES; | |
90f8572b | 1839 | if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path)) |
b32dfe37 CG |
1840 | goto exit; |
1841 | ||
496ad9aa | 1842 | err = inode_permission(inode, MAY_EXEC); |
b32dfe37 CG |
1843 | if (err) |
1844 | goto exit; | |
1845 | ||
bafb282d | 1846 | /* |
4229fb1d | 1847 | * Forbid mm->exe_file change if old file still mapped. |
bafb282d | 1848 | */ |
6e399cd1 | 1849 | exe_file = get_mm_exe_file(mm); |
bafb282d | 1850 | err = -EBUSY; |
6e399cd1 | 1851 | if (exe_file) { |
4229fb1d KK |
1852 | struct vm_area_struct *vma; |
1853 | ||
6e399cd1 DB |
1854 | down_read(&mm->mmap_sem); |
1855 | for (vma = mm->mmap; vma; vma = vma->vm_next) { | |
1856 | if (!vma->vm_file) | |
1857 | continue; | |
1858 | if (path_equal(&vma->vm_file->f_path, | |
1859 | &exe_file->f_path)) | |
1860 | goto exit_err; | |
1861 | } | |
1862 | ||
1863 | up_read(&mm->mmap_sem); | |
1864 | fput(exe_file); | |
bafb282d KK |
1865 | } |
1866 | ||
4229fb1d | 1867 | err = 0; |
6e399cd1 DB |
1868 | /* set the new file, lockless */ |
1869 | get_file(exe.file); | |
1870 | old_exe = xchg(&mm->exe_file, exe.file); | |
1871 | if (old_exe) | |
1872 | fput(old_exe); | |
b32dfe37 | 1873 | exit: |
2903ff01 | 1874 | fdput(exe); |
b32dfe37 | 1875 | return err; |
6e399cd1 DB |
1876 | exit_err: |
1877 | up_read(&mm->mmap_sem); | |
1878 | fput(exe_file); | |
1879 | goto exit; | |
b32dfe37 CG |
1880 | } |
1881 | ||
f606b77f CG |
1882 | /* |
1883 | * WARNING: we don't require any capability here so be very careful | |
1884 | * in what is allowed for modification from userspace. | |
1885 | */ | |
1886 | static int validate_prctl_map(struct prctl_mm_map *prctl_map) | |
1887 | { | |
1888 | unsigned long mmap_max_addr = TASK_SIZE; | |
1889 | struct mm_struct *mm = current->mm; | |
1890 | int error = -EINVAL, i; | |
1891 | ||
1892 | static const unsigned char offsets[] = { | |
1893 | offsetof(struct prctl_mm_map, start_code), | |
1894 | offsetof(struct prctl_mm_map, end_code), | |
1895 | offsetof(struct prctl_mm_map, start_data), | |
1896 | offsetof(struct prctl_mm_map, end_data), | |
1897 | offsetof(struct prctl_mm_map, start_brk), | |
1898 | offsetof(struct prctl_mm_map, brk), | |
1899 | offsetof(struct prctl_mm_map, start_stack), | |
1900 | offsetof(struct prctl_mm_map, arg_start), | |
1901 | offsetof(struct prctl_mm_map, arg_end), | |
1902 | offsetof(struct prctl_mm_map, env_start), | |
1903 | offsetof(struct prctl_mm_map, env_end), | |
1904 | }; | |
1905 | ||
1906 | /* | |
1907 | * Make sure the members are not somewhere outside | |
1908 | * of allowed address space. | |
1909 | */ | |
1910 | for (i = 0; i < ARRAY_SIZE(offsets); i++) { | |
1911 | u64 val = *(u64 *)((char *)prctl_map + offsets[i]); | |
1912 | ||
1913 | if ((unsigned long)val >= mmap_max_addr || | |
1914 | (unsigned long)val < mmap_min_addr) | |
1915 | goto out; | |
1916 | } | |
1917 | ||
1918 | /* | |
1919 | * Make sure the pairs are ordered. | |
1920 | */ | |
1921 | #define __prctl_check_order(__m1, __op, __m2) \ | |
1922 | ((unsigned long)prctl_map->__m1 __op \ | |
1923 | (unsigned long)prctl_map->__m2) ? 0 : -EINVAL | |
1924 | error = __prctl_check_order(start_code, <, end_code); | |
1925 | error |= __prctl_check_order(start_data, <, end_data); | |
1926 | error |= __prctl_check_order(start_brk, <=, brk); | |
1927 | error |= __prctl_check_order(arg_start, <=, arg_end); | |
1928 | error |= __prctl_check_order(env_start, <=, env_end); | |
1929 | if (error) | |
1930 | goto out; | |
1931 | #undef __prctl_check_order | |
1932 | ||
1933 | error = -EINVAL; | |
1934 | ||
1935 | /* | |
1936 | * @brk should be after @end_data in traditional maps. | |
1937 | */ | |
1938 | if (prctl_map->start_brk <= prctl_map->end_data || | |
1939 | prctl_map->brk <= prctl_map->end_data) | |
1940 | goto out; | |
1941 | ||
1942 | /* | |
1943 | * Neither we should allow to override limits if they set. | |
1944 | */ | |
1945 | if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk, | |
1946 | prctl_map->start_brk, prctl_map->end_data, | |
1947 | prctl_map->start_data)) | |
1948 | goto out; | |
1949 | ||
1950 | /* | |
1951 | * Someone is trying to cheat the auxv vector. | |
1952 | */ | |
1953 | if (prctl_map->auxv_size) { | |
1954 | if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv)) | |
1955 | goto out; | |
1956 | } | |
1957 | ||
1958 | /* | |
1959 | * Finally, make sure the caller has the rights to | |
4d28df61 | 1960 | * change /proc/pid/exe link: only local sys admin should |
f606b77f CG |
1961 | * be allowed to. |
1962 | */ | |
1963 | if (prctl_map->exe_fd != (u32)-1) { | |
4d28df61 | 1964 | if (!ns_capable(current_user_ns(), CAP_SYS_ADMIN)) |
f606b77f CG |
1965 | goto out; |
1966 | } | |
1967 | ||
1968 | error = 0; | |
1969 | out: | |
1970 | return error; | |
1971 | } | |
1972 | ||
4a00e9df | 1973 | #ifdef CONFIG_CHECKPOINT_RESTORE |
f606b77f CG |
1974 | static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size) |
1975 | { | |
1976 | struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, }; | |
1977 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1978 | struct mm_struct *mm = current->mm; | |
1979 | int error; | |
1980 | ||
1981 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1982 | BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256); | |
1983 | ||
1984 | if (opt == PR_SET_MM_MAP_SIZE) | |
1985 | return put_user((unsigned int)sizeof(prctl_map), | |
1986 | (unsigned int __user *)addr); | |
1987 | ||
1988 | if (data_size != sizeof(prctl_map)) | |
1989 | return -EINVAL; | |
1990 | ||
1991 | if (copy_from_user(&prctl_map, addr, sizeof(prctl_map))) | |
1992 | return -EFAULT; | |
1993 | ||
1994 | error = validate_prctl_map(&prctl_map); | |
1995 | if (error) | |
1996 | return error; | |
1997 | ||
1998 | if (prctl_map.auxv_size) { | |
1999 | memset(user_auxv, 0, sizeof(user_auxv)); | |
2000 | if (copy_from_user(user_auxv, | |
2001 | (const void __user *)prctl_map.auxv, | |
2002 | prctl_map.auxv_size)) | |
2003 | return -EFAULT; | |
2004 | ||
2005 | /* Last entry must be AT_NULL as specification requires */ | |
2006 | user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL; | |
2007 | user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL; | |
2008 | } | |
2009 | ||
ddf1d398 | 2010 | if (prctl_map.exe_fd != (u32)-1) { |
6e399cd1 | 2011 | error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd); |
ddf1d398 MG |
2012 | if (error) |
2013 | return error; | |
2014 | } | |
2015 | ||
2016 | down_write(&mm->mmap_sem); | |
f606b77f CG |
2017 | |
2018 | /* | |
2019 | * We don't validate if these members are pointing to | |
2020 | * real present VMAs because application may have correspond | |
2021 | * VMAs already unmapped and kernel uses these members for statistics | |
2022 | * output in procfs mostly, except | |
2023 | * | |
2024 | * - @start_brk/@brk which are used in do_brk but kernel lookups | |
2025 | * for VMAs when updating these memvers so anything wrong written | |
2026 | * here cause kernel to swear at userspace program but won't lead | |
2027 | * to any problem in kernel itself | |
2028 | */ | |
2029 | ||
2030 | mm->start_code = prctl_map.start_code; | |
2031 | mm->end_code = prctl_map.end_code; | |
2032 | mm->start_data = prctl_map.start_data; | |
2033 | mm->end_data = prctl_map.end_data; | |
2034 | mm->start_brk = prctl_map.start_brk; | |
2035 | mm->brk = prctl_map.brk; | |
2036 | mm->start_stack = prctl_map.start_stack; | |
2037 | mm->arg_start = prctl_map.arg_start; | |
2038 | mm->arg_end = prctl_map.arg_end; | |
2039 | mm->env_start = prctl_map.env_start; | |
2040 | mm->env_end = prctl_map.env_end; | |
2041 | ||
2042 | /* | |
2043 | * Note this update of @saved_auxv is lockless thus | |
2044 | * if someone reads this member in procfs while we're | |
2045 | * updating -- it may get partly updated results. It's | |
2046 | * known and acceptable trade off: we leave it as is to | |
2047 | * not introduce additional locks here making the kernel | |
2048 | * more complex. | |
2049 | */ | |
2050 | if (prctl_map.auxv_size) | |
2051 | memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv)); | |
2052 | ||
ddf1d398 MG |
2053 | up_write(&mm->mmap_sem); |
2054 | return 0; | |
f606b77f CG |
2055 | } |
2056 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
2057 | ||
4a00e9df AD |
2058 | static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr, |
2059 | unsigned long len) | |
2060 | { | |
2061 | /* | |
2062 | * This doesn't move the auxiliary vector itself since it's pinned to | |
2063 | * mm_struct, but it permits filling the vector with new values. It's | |
2064 | * up to the caller to provide sane values here, otherwise userspace | |
2065 | * tools which use this vector might be unhappy. | |
2066 | */ | |
2067 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
2068 | ||
2069 | if (len > sizeof(user_auxv)) | |
2070 | return -EINVAL; | |
2071 | ||
2072 | if (copy_from_user(user_auxv, (const void __user *)addr, len)) | |
2073 | return -EFAULT; | |
2074 | ||
2075 | /* Make sure the last entry is always AT_NULL */ | |
2076 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
2077 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
2078 | ||
2079 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
2080 | ||
2081 | task_lock(current); | |
2082 | memcpy(mm->saved_auxv, user_auxv, len); | |
2083 | task_unlock(current); | |
2084 | ||
2085 | return 0; | |
2086 | } | |
2087 | ||
028ee4be CG |
2088 | static int prctl_set_mm(int opt, unsigned long addr, |
2089 | unsigned long arg4, unsigned long arg5) | |
2090 | { | |
028ee4be | 2091 | struct mm_struct *mm = current->mm; |
4a00e9df | 2092 | struct prctl_mm_map prctl_map; |
fe8c7f5c CG |
2093 | struct vm_area_struct *vma; |
2094 | int error; | |
028ee4be | 2095 | |
f606b77f CG |
2096 | if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV && |
2097 | opt != PR_SET_MM_MAP && | |
2098 | opt != PR_SET_MM_MAP_SIZE))) | |
028ee4be CG |
2099 | return -EINVAL; |
2100 | ||
f606b77f CG |
2101 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2102 | if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE) | |
2103 | return prctl_set_mm_map(opt, (const void __user *)addr, arg4); | |
2104 | #endif | |
2105 | ||
79f0713d | 2106 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
2107 | return -EPERM; |
2108 | ||
6e399cd1 DB |
2109 | if (opt == PR_SET_MM_EXE_FILE) |
2110 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
b32dfe37 | 2111 | |
4a00e9df AD |
2112 | if (opt == PR_SET_MM_AUXV) |
2113 | return prctl_set_auxv(mm, addr, arg4); | |
2114 | ||
1ad75b9e | 2115 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
2116 | return -EINVAL; |
2117 | ||
fe8c7f5c CG |
2118 | error = -EINVAL; |
2119 | ||
ddf1d398 | 2120 | down_write(&mm->mmap_sem); |
028ee4be CG |
2121 | vma = find_vma(mm, addr); |
2122 | ||
4a00e9df AD |
2123 | prctl_map.start_code = mm->start_code; |
2124 | prctl_map.end_code = mm->end_code; | |
2125 | prctl_map.start_data = mm->start_data; | |
2126 | prctl_map.end_data = mm->end_data; | |
2127 | prctl_map.start_brk = mm->start_brk; | |
2128 | prctl_map.brk = mm->brk; | |
2129 | prctl_map.start_stack = mm->start_stack; | |
2130 | prctl_map.arg_start = mm->arg_start; | |
2131 | prctl_map.arg_end = mm->arg_end; | |
2132 | prctl_map.env_start = mm->env_start; | |
2133 | prctl_map.env_end = mm->env_end; | |
2134 | prctl_map.auxv = NULL; | |
2135 | prctl_map.auxv_size = 0; | |
2136 | prctl_map.exe_fd = -1; | |
2137 | ||
028ee4be CG |
2138 | switch (opt) { |
2139 | case PR_SET_MM_START_CODE: | |
4a00e9df | 2140 | prctl_map.start_code = addr; |
fe8c7f5c | 2141 | break; |
028ee4be | 2142 | case PR_SET_MM_END_CODE: |
4a00e9df | 2143 | prctl_map.end_code = addr; |
028ee4be | 2144 | break; |
028ee4be | 2145 | case PR_SET_MM_START_DATA: |
4a00e9df | 2146 | prctl_map.start_data = addr; |
028ee4be | 2147 | break; |
fe8c7f5c | 2148 | case PR_SET_MM_END_DATA: |
4a00e9df AD |
2149 | prctl_map.end_data = addr; |
2150 | break; | |
2151 | case PR_SET_MM_START_STACK: | |
2152 | prctl_map.start_stack = addr; | |
028ee4be | 2153 | break; |
028ee4be | 2154 | case PR_SET_MM_START_BRK: |
4a00e9df | 2155 | prctl_map.start_brk = addr; |
028ee4be | 2156 | break; |
028ee4be | 2157 | case PR_SET_MM_BRK: |
4a00e9df | 2158 | prctl_map.brk = addr; |
028ee4be | 2159 | break; |
4a00e9df AD |
2160 | case PR_SET_MM_ARG_START: |
2161 | prctl_map.arg_start = addr; | |
2162 | break; | |
2163 | case PR_SET_MM_ARG_END: | |
2164 | prctl_map.arg_end = addr; | |
2165 | break; | |
2166 | case PR_SET_MM_ENV_START: | |
2167 | prctl_map.env_start = addr; | |
2168 | break; | |
2169 | case PR_SET_MM_ENV_END: | |
2170 | prctl_map.env_end = addr; | |
2171 | break; | |
2172 | default: | |
2173 | goto out; | |
2174 | } | |
2175 | ||
2176 | error = validate_prctl_map(&prctl_map); | |
2177 | if (error) | |
2178 | goto out; | |
028ee4be | 2179 | |
4a00e9df | 2180 | switch (opt) { |
fe8c7f5c CG |
2181 | /* |
2182 | * If command line arguments and environment | |
2183 | * are placed somewhere else on stack, we can | |
2184 | * set them up here, ARG_START/END to setup | |
2185 | * command line argumets and ENV_START/END | |
2186 | * for environment. | |
2187 | */ | |
2188 | case PR_SET_MM_START_STACK: | |
2189 | case PR_SET_MM_ARG_START: | |
2190 | case PR_SET_MM_ARG_END: | |
2191 | case PR_SET_MM_ENV_START: | |
2192 | case PR_SET_MM_ENV_END: | |
2193 | if (!vma) { | |
2194 | error = -EFAULT; | |
2195 | goto out; | |
2196 | } | |
028ee4be CG |
2197 | } |
2198 | ||
4a00e9df AD |
2199 | mm->start_code = prctl_map.start_code; |
2200 | mm->end_code = prctl_map.end_code; | |
2201 | mm->start_data = prctl_map.start_data; | |
2202 | mm->end_data = prctl_map.end_data; | |
2203 | mm->start_brk = prctl_map.start_brk; | |
2204 | mm->brk = prctl_map.brk; | |
2205 | mm->start_stack = prctl_map.start_stack; | |
2206 | mm->arg_start = prctl_map.arg_start; | |
2207 | mm->arg_end = prctl_map.arg_end; | |
2208 | mm->env_start = prctl_map.env_start; | |
2209 | mm->env_end = prctl_map.env_end; | |
2210 | ||
028ee4be | 2211 | error = 0; |
028ee4be | 2212 | out: |
ddf1d398 | 2213 | up_write(&mm->mmap_sem); |
028ee4be CG |
2214 | return error; |
2215 | } | |
300f786b | 2216 | |
52b36941 | 2217 | #ifdef CONFIG_CHECKPOINT_RESTORE |
300f786b CG |
2218 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2219 | { | |
2220 | return put_user(me->clear_child_tid, tid_addr); | |
2221 | } | |
52b36941 | 2222 | #else |
300f786b CG |
2223 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2224 | { | |
2225 | return -EINVAL; | |
2226 | } | |
028ee4be CG |
2227 | #endif |
2228 | ||
749860ce PT |
2229 | static int propagate_has_child_subreaper(struct task_struct *p, void *data) |
2230 | { | |
2231 | /* | |
2232 | * If task has has_child_subreaper - all its decendants | |
2233 | * already have these flag too and new decendants will | |
2234 | * inherit it on fork, skip them. | |
2235 | * | |
2236 | * If we've found child_reaper - skip descendants in | |
2237 | * it's subtree as they will never get out pidns. | |
2238 | */ | |
2239 | if (p->signal->has_child_subreaper || | |
2240 | is_child_reaper(task_pid(p))) | |
2241 | return 0; | |
2242 | ||
2243 | p->signal->has_child_subreaper = 1; | |
2244 | return 1; | |
2245 | } | |
2246 | ||
b617cfc8 TG |
2247 | int __weak arch_prctl_spec_ctrl_get(unsigned long which) |
2248 | { | |
2249 | return -EINVAL; | |
2250 | } | |
2251 | ||
2252 | int __weak arch_prctl_spec_ctrl_set(unsigned long which, unsigned long ctrl) | |
2253 | { | |
2254 | return -EINVAL; | |
2255 | } | |
2256 | ||
c4ea37c2 HC |
2257 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2258 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2259 | { |
b6dff3ec DH |
2260 | struct task_struct *me = current; |
2261 | unsigned char comm[sizeof(me->comm)]; | |
2262 | long error; | |
1da177e4 | 2263 | |
d84f4f99 DH |
2264 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2265 | if (error != -ENOSYS) | |
1da177e4 LT |
2266 | return error; |
2267 | ||
d84f4f99 | 2268 | error = 0; |
1da177e4 | 2269 | switch (option) { |
f3cbd435 AM |
2270 | case PR_SET_PDEATHSIG: |
2271 | if (!valid_signal(arg2)) { | |
2272 | error = -EINVAL; | |
1da177e4 | 2273 | break; |
f3cbd435 AM |
2274 | } |
2275 | me->pdeath_signal = arg2; | |
2276 | break; | |
2277 | case PR_GET_PDEATHSIG: | |
2278 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2279 | break; | |
2280 | case PR_GET_DUMPABLE: | |
2281 | error = get_dumpable(me->mm); | |
2282 | break; | |
2283 | case PR_SET_DUMPABLE: | |
2284 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2285 | error = -EINVAL; | |
1da177e4 | 2286 | break; |
f3cbd435 AM |
2287 | } |
2288 | set_dumpable(me->mm, arg2); | |
2289 | break; | |
1da177e4 | 2290 | |
f3cbd435 AM |
2291 | case PR_SET_UNALIGN: |
2292 | error = SET_UNALIGN_CTL(me, arg2); | |
2293 | break; | |
2294 | case PR_GET_UNALIGN: | |
2295 | error = GET_UNALIGN_CTL(me, arg2); | |
2296 | break; | |
2297 | case PR_SET_FPEMU: | |
2298 | error = SET_FPEMU_CTL(me, arg2); | |
2299 | break; | |
2300 | case PR_GET_FPEMU: | |
2301 | error = GET_FPEMU_CTL(me, arg2); | |
2302 | break; | |
2303 | case PR_SET_FPEXC: | |
2304 | error = SET_FPEXC_CTL(me, arg2); | |
2305 | break; | |
2306 | case PR_GET_FPEXC: | |
2307 | error = GET_FPEXC_CTL(me, arg2); | |
2308 | break; | |
2309 | case PR_GET_TIMING: | |
2310 | error = PR_TIMING_STATISTICAL; | |
2311 | break; | |
2312 | case PR_SET_TIMING: | |
2313 | if (arg2 != PR_TIMING_STATISTICAL) | |
2314 | error = -EINVAL; | |
2315 | break; | |
2316 | case PR_SET_NAME: | |
2317 | comm[sizeof(me->comm) - 1] = 0; | |
2318 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2319 | sizeof(me->comm) - 1) < 0) | |
2320 | return -EFAULT; | |
2321 | set_task_comm(me, comm); | |
2322 | proc_comm_connector(me); | |
2323 | break; | |
2324 | case PR_GET_NAME: | |
2325 | get_task_comm(comm, me); | |
2326 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2327 | return -EFAULT; | |
2328 | break; | |
2329 | case PR_GET_ENDIAN: | |
2330 | error = GET_ENDIAN(me, arg2); | |
2331 | break; | |
2332 | case PR_SET_ENDIAN: | |
2333 | error = SET_ENDIAN(me, arg2); | |
2334 | break; | |
2335 | case PR_GET_SECCOMP: | |
2336 | error = prctl_get_seccomp(); | |
2337 | break; | |
2338 | case PR_SET_SECCOMP: | |
2339 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2340 | break; | |
2341 | case PR_GET_TSC: | |
2342 | error = GET_TSC_CTL(arg2); | |
2343 | break; | |
2344 | case PR_SET_TSC: | |
2345 | error = SET_TSC_CTL(arg2); | |
2346 | break; | |
2347 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2348 | error = perf_event_task_disable(); | |
2349 | break; | |
2350 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2351 | error = perf_event_task_enable(); | |
2352 | break; | |
2353 | case PR_GET_TIMERSLACK: | |
da8b44d5 JS |
2354 | if (current->timer_slack_ns > ULONG_MAX) |
2355 | error = ULONG_MAX; | |
2356 | else | |
2357 | error = current->timer_slack_ns; | |
f3cbd435 AM |
2358 | break; |
2359 | case PR_SET_TIMERSLACK: | |
2360 | if (arg2 <= 0) | |
2361 | current->timer_slack_ns = | |
6976675d | 2362 | current->default_timer_slack_ns; |
f3cbd435 AM |
2363 | else |
2364 | current->timer_slack_ns = arg2; | |
2365 | break; | |
2366 | case PR_MCE_KILL: | |
2367 | if (arg4 | arg5) | |
2368 | return -EINVAL; | |
2369 | switch (arg2) { | |
2370 | case PR_MCE_KILL_CLEAR: | |
2371 | if (arg3 != 0) | |
4db96cf0 | 2372 | return -EINVAL; |
f3cbd435 | 2373 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2374 | break; |
f3cbd435 AM |
2375 | case PR_MCE_KILL_SET: |
2376 | current->flags |= PF_MCE_PROCESS; | |
2377 | if (arg3 == PR_MCE_KILL_EARLY) | |
2378 | current->flags |= PF_MCE_EARLY; | |
2379 | else if (arg3 == PR_MCE_KILL_LATE) | |
2380 | current->flags &= ~PF_MCE_EARLY; | |
2381 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2382 | current->flags &= | |
2383 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2384 | else |
259e5e6c | 2385 | return -EINVAL; |
259e5e6c | 2386 | break; |
1da177e4 | 2387 | default: |
f3cbd435 AM |
2388 | return -EINVAL; |
2389 | } | |
2390 | break; | |
2391 | case PR_MCE_KILL_GET: | |
2392 | if (arg2 | arg3 | arg4 | arg5) | |
2393 | return -EINVAL; | |
2394 | if (current->flags & PF_MCE_PROCESS) | |
2395 | error = (current->flags & PF_MCE_EARLY) ? | |
2396 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2397 | else | |
2398 | error = PR_MCE_KILL_DEFAULT; | |
2399 | break; | |
2400 | case PR_SET_MM: | |
2401 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2402 | break; | |
2403 | case PR_GET_TID_ADDRESS: | |
2404 | error = prctl_get_tid_address(me, (int __user **)arg2); | |
2405 | break; | |
2406 | case PR_SET_CHILD_SUBREAPER: | |
2407 | me->signal->is_child_subreaper = !!arg2; | |
749860ce PT |
2408 | if (!arg2) |
2409 | break; | |
2410 | ||
2411 | walk_process_tree(me, propagate_has_child_subreaper, NULL); | |
f3cbd435 AM |
2412 | break; |
2413 | case PR_GET_CHILD_SUBREAPER: | |
2414 | error = put_user(me->signal->is_child_subreaper, | |
2415 | (int __user *)arg2); | |
2416 | break; | |
2417 | case PR_SET_NO_NEW_PRIVS: | |
2418 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2419 | return -EINVAL; | |
2420 | ||
1d4457f9 | 2421 | task_set_no_new_privs(current); |
f3cbd435 AM |
2422 | break; |
2423 | case PR_GET_NO_NEW_PRIVS: | |
2424 | if (arg2 || arg3 || arg4 || arg5) | |
2425 | return -EINVAL; | |
1d4457f9 | 2426 | return task_no_new_privs(current) ? 1 : 0; |
a0715cc2 AT |
2427 | case PR_GET_THP_DISABLE: |
2428 | if (arg2 || arg3 || arg4 || arg5) | |
2429 | return -EINVAL; | |
18600332 | 2430 | error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 AT |
2431 | break; |
2432 | case PR_SET_THP_DISABLE: | |
2433 | if (arg3 || arg4 || arg5) | |
2434 | return -EINVAL; | |
17b0573d MH |
2435 | if (down_write_killable(&me->mm->mmap_sem)) |
2436 | return -EINTR; | |
a0715cc2 | 2437 | if (arg2) |
18600332 | 2438 | set_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 | 2439 | else |
18600332 | 2440 | clear_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 AT |
2441 | up_write(&me->mm->mmap_sem); |
2442 | break; | |
fe3d197f | 2443 | case PR_MPX_ENABLE_MANAGEMENT: |
e9d1b4f3 DH |
2444 | if (arg2 || arg3 || arg4 || arg5) |
2445 | return -EINVAL; | |
46a6e0cf | 2446 | error = MPX_ENABLE_MANAGEMENT(); |
fe3d197f DH |
2447 | break; |
2448 | case PR_MPX_DISABLE_MANAGEMENT: | |
e9d1b4f3 DH |
2449 | if (arg2 || arg3 || arg4 || arg5) |
2450 | return -EINVAL; | |
46a6e0cf | 2451 | error = MPX_DISABLE_MANAGEMENT(); |
fe3d197f | 2452 | break; |
9791554b PB |
2453 | case PR_SET_FP_MODE: |
2454 | error = SET_FP_MODE(me, arg2); | |
2455 | break; | |
2456 | case PR_GET_FP_MODE: | |
2457 | error = GET_FP_MODE(me); | |
2458 | break; | |
2d2123bc DM |
2459 | case PR_SVE_SET_VL: |
2460 | error = SVE_SET_VL(arg2); | |
2461 | break; | |
2462 | case PR_SVE_GET_VL: | |
2463 | error = SVE_GET_VL(); | |
2464 | break; | |
b617cfc8 TG |
2465 | case PR_GET_SPECULATION_CTRL: |
2466 | if (arg3 || arg4 || arg5) | |
2467 | return -EINVAL; | |
2468 | error = arch_prctl_spec_ctrl_get(arg2); | |
2469 | break; | |
2470 | case PR_SET_SPECULATION_CTRL: | |
2471 | if (arg4 || arg5) | |
2472 | return -EINVAL; | |
2473 | error = arch_prctl_spec_ctrl_set(arg2, arg3); | |
2474 | break; | |
f3cbd435 AM |
2475 | default: |
2476 | error = -EINVAL; | |
2477 | break; | |
1da177e4 LT |
2478 | } |
2479 | return error; | |
2480 | } | |
3cfc348b | 2481 | |
836f92ad HC |
2482 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2483 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2484 | { |
2485 | int err = 0; | |
2486 | int cpu = raw_smp_processor_id(); | |
ec94fc3d | 2487 | |
3cfc348b AK |
2488 | if (cpup) |
2489 | err |= put_user(cpu, cpup); | |
2490 | if (nodep) | |
2491 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2492 | return err ? -EFAULT : 0; |
2493 | } | |
10a0a8d4 | 2494 | |
4a22f166 SR |
2495 | /** |
2496 | * do_sysinfo - fill in sysinfo struct | |
2497 | * @info: pointer to buffer to fill | |
2498 | */ | |
2499 | static int do_sysinfo(struct sysinfo *info) | |
2500 | { | |
2501 | unsigned long mem_total, sav_total; | |
2502 | unsigned int mem_unit, bitcount; | |
2503 | struct timespec tp; | |
2504 | ||
2505 | memset(info, 0, sizeof(struct sysinfo)); | |
2506 | ||
45c64940 | 2507 | get_monotonic_boottime(&tp); |
4a22f166 SR |
2508 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
2509 | ||
2510 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2511 | ||
2512 | info->procs = nr_threads; | |
2513 | ||
2514 | si_meminfo(info); | |
2515 | si_swapinfo(info); | |
2516 | ||
2517 | /* | |
2518 | * If the sum of all the available memory (i.e. ram + swap) | |
2519 | * is less than can be stored in a 32 bit unsigned long then | |
2520 | * we can be binary compatible with 2.2.x kernels. If not, | |
2521 | * well, in that case 2.2.x was broken anyways... | |
2522 | * | |
2523 | * -Erik Andersen <andersee@debian.org> | |
2524 | */ | |
2525 | ||
2526 | mem_total = info->totalram + info->totalswap; | |
2527 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2528 | goto out; | |
2529 | bitcount = 0; | |
2530 | mem_unit = info->mem_unit; | |
2531 | while (mem_unit > 1) { | |
2532 | bitcount++; | |
2533 | mem_unit >>= 1; | |
2534 | sav_total = mem_total; | |
2535 | mem_total <<= 1; | |
2536 | if (mem_total < sav_total) | |
2537 | goto out; | |
2538 | } | |
2539 | ||
2540 | /* | |
2541 | * If mem_total did not overflow, multiply all memory values by | |
2542 | * info->mem_unit and set it to 1. This leaves things compatible | |
2543 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2544 | * kernels... | |
2545 | */ | |
2546 | ||
2547 | info->mem_unit = 1; | |
2548 | info->totalram <<= bitcount; | |
2549 | info->freeram <<= bitcount; | |
2550 | info->sharedram <<= bitcount; | |
2551 | info->bufferram <<= bitcount; | |
2552 | info->totalswap <<= bitcount; | |
2553 | info->freeswap <<= bitcount; | |
2554 | info->totalhigh <<= bitcount; | |
2555 | info->freehigh <<= bitcount; | |
2556 | ||
2557 | out: | |
2558 | return 0; | |
2559 | } | |
2560 | ||
2561 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2562 | { | |
2563 | struct sysinfo val; | |
2564 | ||
2565 | do_sysinfo(&val); | |
2566 | ||
2567 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2568 | return -EFAULT; | |
2569 | ||
2570 | return 0; | |
2571 | } | |
2572 | ||
2573 | #ifdef CONFIG_COMPAT | |
2574 | struct compat_sysinfo { | |
2575 | s32 uptime; | |
2576 | u32 loads[3]; | |
2577 | u32 totalram; | |
2578 | u32 freeram; | |
2579 | u32 sharedram; | |
2580 | u32 bufferram; | |
2581 | u32 totalswap; | |
2582 | u32 freeswap; | |
2583 | u16 procs; | |
2584 | u16 pad; | |
2585 | u32 totalhigh; | |
2586 | u32 freehigh; | |
2587 | u32 mem_unit; | |
2588 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2589 | }; | |
2590 | ||
2591 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2592 | { | |
2593 | struct sysinfo s; | |
2594 | ||
2595 | do_sysinfo(&s); | |
2596 | ||
2597 | /* Check to see if any memory value is too large for 32-bit and scale | |
2598 | * down if needed | |
2599 | */ | |
0baae41e | 2600 | if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) { |
4a22f166 SR |
2601 | int bitcount = 0; |
2602 | ||
2603 | while (s.mem_unit < PAGE_SIZE) { | |
2604 | s.mem_unit <<= 1; | |
2605 | bitcount++; | |
2606 | } | |
2607 | ||
2608 | s.totalram >>= bitcount; | |
2609 | s.freeram >>= bitcount; | |
2610 | s.sharedram >>= bitcount; | |
2611 | s.bufferram >>= bitcount; | |
2612 | s.totalswap >>= bitcount; | |
2613 | s.freeswap >>= bitcount; | |
2614 | s.totalhigh >>= bitcount; | |
2615 | s.freehigh >>= bitcount; | |
2616 | } | |
2617 | ||
2618 | if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) || | |
2619 | __put_user(s.uptime, &info->uptime) || | |
2620 | __put_user(s.loads[0], &info->loads[0]) || | |
2621 | __put_user(s.loads[1], &info->loads[1]) || | |
2622 | __put_user(s.loads[2], &info->loads[2]) || | |
2623 | __put_user(s.totalram, &info->totalram) || | |
2624 | __put_user(s.freeram, &info->freeram) || | |
2625 | __put_user(s.sharedram, &info->sharedram) || | |
2626 | __put_user(s.bufferram, &info->bufferram) || | |
2627 | __put_user(s.totalswap, &info->totalswap) || | |
2628 | __put_user(s.freeswap, &info->freeswap) || | |
2629 | __put_user(s.procs, &info->procs) || | |
2630 | __put_user(s.totalhigh, &info->totalhigh) || | |
2631 | __put_user(s.freehigh, &info->freehigh) || | |
2632 | __put_user(s.mem_unit, &info->mem_unit)) | |
2633 | return -EFAULT; | |
2634 | ||
2635 | return 0; | |
2636 | } | |
2637 | #endif /* CONFIG_COMPAT */ |