]>
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 | 9 | #include <linux/mm.h> |
5c26f6ac | 10 | #include <linux/mm_inline.h> |
1da177e4 LT |
11 | #include <linux/utsname.h> |
12 | #include <linux/mman.h> | |
1da177e4 LT |
13 | #include <linux/reboot.h> |
14 | #include <linux/prctl.h> | |
1da177e4 LT |
15 | #include <linux/highuid.h> |
16 | #include <linux/fs.h> | |
74da1ff7 | 17 | #include <linux/kmod.h> |
d7597f59 | 18 | #include <linux/ksm.h> |
cdd6c482 | 19 | #include <linux/perf_event.h> |
3e88c553 | 20 | #include <linux/resource.h> |
dc009d92 | 21 | #include <linux/kernel.h> |
1da177e4 | 22 | #include <linux/workqueue.h> |
c59ede7b | 23 | #include <linux/capability.h> |
1da177e4 LT |
24 | #include <linux/device.h> |
25 | #include <linux/key.h> | |
26 | #include <linux/times.h> | |
27 | #include <linux/posix-timers.h> | |
28 | #include <linux/security.h> | |
37608ba3 | 29 | #include <linux/random.h> |
1da177e4 LT |
30 | #include <linux/suspend.h> |
31 | #include <linux/tty.h> | |
7ed20e1a | 32 | #include <linux/signal.h> |
9f46080c | 33 | #include <linux/cn_proc.h> |
3cfc348b | 34 | #include <linux/getcpu.h> |
6eaeeaba | 35 | #include <linux/task_io_accounting_ops.h> |
1d9d02fe | 36 | #include <linux/seccomp.h> |
4047727e | 37 | #include <linux/cpu.h> |
e28cbf22 | 38 | #include <linux/personality.h> |
e3d5a27d | 39 | #include <linux/ptrace.h> |
5ad4e53b | 40 | #include <linux/fs_struct.h> |
b32dfe37 CG |
41 | #include <linux/file.h> |
42 | #include <linux/mount.h> | |
5a0e3ad6 | 43 | #include <linux/gfp.h> |
40dc166c | 44 | #include <linux/syscore_ops.h> |
be27425d AK |
45 | #include <linux/version.h> |
46 | #include <linux/ctype.h> | |
1446e1df | 47 | #include <linux/syscall_user_dispatch.h> |
1da177e4 LT |
48 | |
49 | #include <linux/compat.h> | |
50 | #include <linux/syscalls.h> | |
00d7c05a | 51 | #include <linux/kprobes.h> |
acce292c | 52 | #include <linux/user_namespace.h> |
ecc421e0 | 53 | #include <linux/time_namespace.h> |
7fe5e042 | 54 | #include <linux/binfmts.h> |
1da177e4 | 55 | |
4a22f166 | 56 | #include <linux/sched.h> |
4eb5aaa3 | 57 | #include <linux/sched/autogroup.h> |
4f17722c | 58 | #include <linux/sched/loadavg.h> |
03441a34 | 59 | #include <linux/sched/stat.h> |
6e84f315 | 60 | #include <linux/sched/mm.h> |
f7ccbae4 | 61 | #include <linux/sched/coredump.h> |
29930025 | 62 | #include <linux/sched/task.h> |
32ef5517 | 63 | #include <linux/sched/cputime.h> |
4a22f166 SR |
64 | #include <linux/rcupdate.h> |
65 | #include <linux/uidgid.h> | |
66 | #include <linux/cred.h> | |
67 | ||
b617cfc8 TG |
68 | #include <linux/nospec.h> |
69 | ||
04c6862c | 70 | #include <linux/kmsg_dump.h> |
be27425d AK |
71 | /* Move somewhere else to avoid recompiling? */ |
72 | #include <generated/utsrelease.h> | |
04c6862c | 73 | |
7c0f6ba6 | 74 | #include <linux/uaccess.h> |
1da177e4 LT |
75 | #include <asm/io.h> |
76 | #include <asm/unistd.h> | |
77 | ||
e530dca5 DB |
78 | #include "uid16.h" |
79 | ||
1da177e4 | 80 | #ifndef SET_UNALIGN_CTL |
ec94fc3d | 81 | # define SET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
82 | #endif |
83 | #ifndef GET_UNALIGN_CTL | |
ec94fc3d | 84 | # define GET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
85 | #endif |
86 | #ifndef SET_FPEMU_CTL | |
ec94fc3d | 87 | # define SET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
88 | #endif |
89 | #ifndef GET_FPEMU_CTL | |
ec94fc3d | 90 | # define GET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
91 | #endif |
92 | #ifndef SET_FPEXC_CTL | |
ec94fc3d | 93 | # define SET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 LT |
94 | #endif |
95 | #ifndef GET_FPEXC_CTL | |
ec94fc3d | 96 | # define GET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 | 97 | #endif |
651d765d | 98 | #ifndef GET_ENDIAN |
ec94fc3d | 99 | # define GET_ENDIAN(a, b) (-EINVAL) |
651d765d AB |
100 | #endif |
101 | #ifndef SET_ENDIAN | |
ec94fc3d | 102 | # define SET_ENDIAN(a, b) (-EINVAL) |
651d765d | 103 | #endif |
8fb402bc EB |
104 | #ifndef GET_TSC_CTL |
105 | # define GET_TSC_CTL(a) (-EINVAL) | |
106 | #endif | |
107 | #ifndef SET_TSC_CTL | |
108 | # define SET_TSC_CTL(a) (-EINVAL) | |
109 | #endif | |
9791554b PB |
110 | #ifndef GET_FP_MODE |
111 | # define GET_FP_MODE(a) (-EINVAL) | |
112 | #endif | |
113 | #ifndef SET_FP_MODE | |
114 | # define SET_FP_MODE(a,b) (-EINVAL) | |
115 | #endif | |
2d2123bc DM |
116 | #ifndef SVE_SET_VL |
117 | # define SVE_SET_VL(a) (-EINVAL) | |
118 | #endif | |
119 | #ifndef SVE_GET_VL | |
120 | # define SVE_GET_VL() (-EINVAL) | |
121 | #endif | |
9e4ab6c8 MB |
122 | #ifndef SME_SET_VL |
123 | # define SME_SET_VL(a) (-EINVAL) | |
124 | #endif | |
125 | #ifndef SME_GET_VL | |
126 | # define SME_GET_VL() (-EINVAL) | |
127 | #endif | |
ba830885 KM |
128 | #ifndef PAC_RESET_KEYS |
129 | # define PAC_RESET_KEYS(a, b) (-EINVAL) | |
130 | #endif | |
20169862 PC |
131 | #ifndef PAC_SET_ENABLED_KEYS |
132 | # define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL) | |
133 | #endif | |
134 | #ifndef PAC_GET_ENABLED_KEYS | |
135 | # define PAC_GET_ENABLED_KEYS(a) (-EINVAL) | |
136 | #endif | |
63f0c603 CM |
137 | #ifndef SET_TAGGED_ADDR_CTRL |
138 | # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL) | |
139 | #endif | |
140 | #ifndef GET_TAGGED_ADDR_CTRL | |
141 | # define GET_TAGGED_ADDR_CTRL() (-EINVAL) | |
142 | #endif | |
1fd96a3e AC |
143 | #ifndef RISCV_V_SET_CONTROL |
144 | # define RISCV_V_SET_CONTROL(a) (-EINVAL) | |
145 | #endif | |
146 | #ifndef RISCV_V_GET_CONTROL | |
147 | # define RISCV_V_GET_CONTROL() (-EINVAL) | |
148 | #endif | |
6b9391b5 CJ |
149 | #ifndef RISCV_SET_ICACHE_FLUSH_CTX |
150 | # define RISCV_SET_ICACHE_FLUSH_CTX(a, b) (-EINVAL) | |
151 | #endif | |
628d701f BG |
152 | #ifndef PPC_GET_DEXCR_ASPECT |
153 | # define PPC_GET_DEXCR_ASPECT(a, b) (-EINVAL) | |
154 | #endif | |
155 | #ifndef PPC_SET_DEXCR_ASPECT | |
156 | # define PPC_SET_DEXCR_ASPECT(a, b, c) (-EINVAL) | |
157 | #endif | |
1da177e4 LT |
158 | |
159 | /* | |
160 | * this is where the system-wide overflow UID and GID are defined, for | |
161 | * architectures that now have 32-bit UID/GID but didn't in the past | |
162 | */ | |
163 | ||
164 | int overflowuid = DEFAULT_OVERFLOWUID; | |
165 | int overflowgid = DEFAULT_OVERFLOWGID; | |
166 | ||
1da177e4 LT |
167 | EXPORT_SYMBOL(overflowuid); |
168 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
169 | |
170 | /* | |
171 | * the same as above, but for filesystems which can only store a 16-bit | |
172 | * UID and GID. as such, this is needed on all architectures | |
173 | */ | |
174 | ||
175 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
8b2770a4 | 176 | int fs_overflowgid = DEFAULT_FS_OVERFLOWGID; |
1da177e4 LT |
177 | |
178 | EXPORT_SYMBOL(fs_overflowuid); | |
179 | EXPORT_SYMBOL(fs_overflowgid); | |
180 | ||
fc832ad3 SH |
181 | /* |
182 | * Returns true if current's euid is same as p's uid or euid, | |
183 | * or has CAP_SYS_NICE to p's user_ns. | |
184 | * | |
185 | * Called with rcu_read_lock, creds are safe | |
186 | */ | |
187 | static bool set_one_prio_perm(struct task_struct *p) | |
188 | { | |
189 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
190 | ||
5af66203 EB |
191 | if (uid_eq(pcred->uid, cred->euid) || |
192 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 193 | return true; |
c4a4d603 | 194 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
195 | return true; |
196 | return false; | |
197 | } | |
198 | ||
c69e8d9c DH |
199 | /* |
200 | * set the priority of a task | |
201 | * - the caller must hold the RCU read lock | |
202 | */ | |
1da177e4 LT |
203 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
204 | { | |
205 | int no_nice; | |
206 | ||
fc832ad3 | 207 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
208 | error = -EPERM; |
209 | goto out; | |
210 | } | |
e43379f1 | 211 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
212 | error = -EACCES; |
213 | goto out; | |
214 | } | |
215 | no_nice = security_task_setnice(p, niceval); | |
216 | if (no_nice) { | |
217 | error = no_nice; | |
218 | goto out; | |
219 | } | |
220 | if (error == -ESRCH) | |
221 | error = 0; | |
222 | set_user_nice(p, niceval); | |
223 | out: | |
224 | return error; | |
225 | } | |
226 | ||
754fe8d2 | 227 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
228 | { |
229 | struct task_struct *g, *p; | |
230 | struct user_struct *user; | |
86a264ab | 231 | const struct cred *cred = current_cred(); |
1da177e4 | 232 | int error = -EINVAL; |
41487c65 | 233 | struct pid *pgrp; |
7b44ab97 | 234 | kuid_t uid; |
1da177e4 | 235 | |
3e88c553 | 236 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
237 | goto out; |
238 | ||
239 | /* normalize: avoid signed division (rounding problems) */ | |
240 | error = -ESRCH; | |
c4a4d2f4 DY |
241 | if (niceval < MIN_NICE) |
242 | niceval = MIN_NICE; | |
243 | if (niceval > MAX_NICE) | |
244 | niceval = MAX_NICE; | |
1da177e4 | 245 | |
d4581a23 | 246 | rcu_read_lock(); |
1da177e4 | 247 | switch (which) { |
ec94fc3d | 248 | case PRIO_PROCESS: |
249 | if (who) | |
250 | p = find_task_by_vpid(who); | |
251 | else | |
252 | p = current; | |
253 | if (p) | |
254 | error = set_one_prio(p, niceval, error); | |
255 | break; | |
256 | case PRIO_PGRP: | |
257 | if (who) | |
258 | pgrp = find_vpid(who); | |
259 | else | |
260 | pgrp = task_pgrp(current); | |
7f8ca0ed | 261 | read_lock(&tasklist_lock); |
ec94fc3d | 262 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { |
263 | error = set_one_prio(p, niceval, error); | |
264 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
7f8ca0ed | 265 | read_unlock(&tasklist_lock); |
ec94fc3d | 266 | break; |
267 | case PRIO_USER: | |
268 | uid = make_kuid(cred->user_ns, who); | |
269 | user = cred->user; | |
270 | if (!who) | |
271 | uid = cred->uid; | |
272 | else if (!uid_eq(uid, cred->uid)) { | |
273 | user = find_user(uid); | |
274 | if (!user) | |
86a264ab | 275 | goto out_unlock; /* No processes for this user */ |
ec94fc3d | 276 | } |
7f8ca0ed | 277 | for_each_process_thread(g, p) { |
8639b461 | 278 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) |
ec94fc3d | 279 | error = set_one_prio(p, niceval, error); |
7f8ca0ed | 280 | } |
ec94fc3d | 281 | if (!uid_eq(uid, cred->uid)) |
282 | free_uid(user); /* For find_user() */ | |
283 | break; | |
1da177e4 LT |
284 | } |
285 | out_unlock: | |
d4581a23 | 286 | rcu_read_unlock(); |
1da177e4 LT |
287 | out: |
288 | return error; | |
289 | } | |
290 | ||
291 | /* | |
292 | * Ugh. To avoid negative return values, "getpriority()" will | |
293 | * not return the normal nice-value, but a negated value that | |
294 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
295 | * to stay compatible. | |
296 | */ | |
754fe8d2 | 297 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
298 | { |
299 | struct task_struct *g, *p; | |
300 | struct user_struct *user; | |
86a264ab | 301 | const struct cred *cred = current_cred(); |
1da177e4 | 302 | long niceval, retval = -ESRCH; |
41487c65 | 303 | struct pid *pgrp; |
7b44ab97 | 304 | kuid_t uid; |
1da177e4 | 305 | |
3e88c553 | 306 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
307 | return -EINVAL; |
308 | ||
70118837 | 309 | rcu_read_lock(); |
1da177e4 | 310 | switch (which) { |
ec94fc3d | 311 | case PRIO_PROCESS: |
312 | if (who) | |
313 | p = find_task_by_vpid(who); | |
314 | else | |
315 | p = current; | |
316 | if (p) { | |
317 | niceval = nice_to_rlimit(task_nice(p)); | |
318 | if (niceval > retval) | |
319 | retval = niceval; | |
320 | } | |
321 | break; | |
322 | case PRIO_PGRP: | |
323 | if (who) | |
324 | pgrp = find_vpid(who); | |
325 | else | |
326 | pgrp = task_pgrp(current); | |
7f8ca0ed | 327 | read_lock(&tasklist_lock); |
ec94fc3d | 328 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { |
329 | niceval = nice_to_rlimit(task_nice(p)); | |
330 | if (niceval > retval) | |
331 | retval = niceval; | |
332 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
7f8ca0ed | 333 | read_unlock(&tasklist_lock); |
ec94fc3d | 334 | break; |
335 | case PRIO_USER: | |
336 | uid = make_kuid(cred->user_ns, who); | |
337 | user = cred->user; | |
338 | if (!who) | |
339 | uid = cred->uid; | |
340 | else if (!uid_eq(uid, cred->uid)) { | |
341 | user = find_user(uid); | |
342 | if (!user) | |
343 | goto out_unlock; /* No processes for this user */ | |
344 | } | |
7f8ca0ed | 345 | for_each_process_thread(g, p) { |
8639b461 | 346 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) { |
7aa2c016 | 347 | niceval = nice_to_rlimit(task_nice(p)); |
1da177e4 LT |
348 | if (niceval > retval) |
349 | retval = niceval; | |
350 | } | |
7f8ca0ed | 351 | } |
ec94fc3d | 352 | if (!uid_eq(uid, cred->uid)) |
353 | free_uid(user); /* for find_user() */ | |
354 | break; | |
1da177e4 LT |
355 | } |
356 | out_unlock: | |
70118837 | 357 | rcu_read_unlock(); |
1da177e4 LT |
358 | |
359 | return retval; | |
360 | } | |
361 | ||
1da177e4 LT |
362 | /* |
363 | * Unprivileged users may change the real gid to the effective gid | |
364 | * or vice versa. (BSD-style) | |
365 | * | |
366 | * If you set the real gid at all, or set the effective gid to a value not | |
367 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
368 | * | |
369 | * This makes it possible for a setgid program to completely drop its | |
370 | * privileges, which is often a useful assertion to make when you are doing | |
371 | * a security audit over a program. | |
372 | * | |
373 | * The general idea is that a program which uses just setregid() will be | |
374 | * 100% compatible with BSD. A program which uses just setgid() will be | |
ec94fc3d | 375 | * 100% compatible with POSIX with saved IDs. |
1da177e4 LT |
376 | * |
377 | * SMP: There are not races, the GIDs are checked only by filesystem | |
378 | * operations (as far as semantic preservation is concerned). | |
379 | */ | |
2813893f | 380 | #ifdef CONFIG_MULTIUSER |
e530dca5 | 381 | long __sys_setregid(gid_t rgid, gid_t egid) |
1da177e4 | 382 | { |
a29c33f4 | 383 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
384 | const struct cred *old; |
385 | struct cred *new; | |
1da177e4 | 386 | int retval; |
a29c33f4 EB |
387 | kgid_t krgid, kegid; |
388 | ||
389 | krgid = make_kgid(ns, rgid); | |
390 | kegid = make_kgid(ns, egid); | |
391 | ||
392 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
393 | return -EINVAL; | |
394 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
395 | return -EINVAL; | |
1da177e4 | 396 | |
d84f4f99 DH |
397 | new = prepare_creds(); |
398 | if (!new) | |
399 | return -ENOMEM; | |
400 | old = current_cred(); | |
401 | ||
d84f4f99 | 402 | retval = -EPERM; |
1da177e4 | 403 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
404 | if (gid_eq(old->gid, krgid) || |
405 | gid_eq(old->egid, krgid) || | |
111767c1 | 406 | ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 | 407 | new->gid = krgid; |
1da177e4 | 408 | else |
d84f4f99 | 409 | goto error; |
1da177e4 LT |
410 | } |
411 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
412 | if (gid_eq(old->gid, kegid) || |
413 | gid_eq(old->egid, kegid) || | |
414 | gid_eq(old->sgid, kegid) || | |
111767c1 | 415 | ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 | 416 | new->egid = kegid; |
756184b7 | 417 | else |
d84f4f99 | 418 | goto error; |
1da177e4 | 419 | } |
d84f4f99 | 420 | |
1da177e4 | 421 | if (rgid != (gid_t) -1 || |
a29c33f4 | 422 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
423 | new->sgid = new->egid; |
424 | new->fsgid = new->egid; | |
425 | ||
39030e13 TC |
426 | retval = security_task_fix_setgid(new, old, LSM_SETID_RE); |
427 | if (retval < 0) | |
428 | goto error; | |
429 | ||
d84f4f99 DH |
430 | return commit_creds(new); |
431 | ||
432 | error: | |
433 | abort_creds(new); | |
434 | return retval; | |
1da177e4 LT |
435 | } |
436 | ||
e530dca5 DB |
437 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
438 | { | |
439 | return __sys_setregid(rgid, egid); | |
440 | } | |
441 | ||
1da177e4 | 442 | /* |
ec94fc3d | 443 | * setgid() is implemented like SysV w/ SAVED_IDS |
1da177e4 LT |
444 | * |
445 | * SMP: Same implicit races as above. | |
446 | */ | |
e530dca5 | 447 | long __sys_setgid(gid_t gid) |
1da177e4 | 448 | { |
a29c33f4 | 449 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
450 | const struct cred *old; |
451 | struct cred *new; | |
1da177e4 | 452 | int retval; |
a29c33f4 EB |
453 | kgid_t kgid; |
454 | ||
455 | kgid = make_kgid(ns, gid); | |
456 | if (!gid_valid(kgid)) | |
457 | return -EINVAL; | |
1da177e4 | 458 | |
d84f4f99 DH |
459 | new = prepare_creds(); |
460 | if (!new) | |
461 | return -ENOMEM; | |
462 | old = current_cred(); | |
463 | ||
d84f4f99 | 464 | retval = -EPERM; |
111767c1 | 465 | if (ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 EB |
466 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
467 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
468 | new->egid = new->fsgid = kgid; | |
1da177e4 | 469 | else |
d84f4f99 | 470 | goto error; |
1da177e4 | 471 | |
39030e13 TC |
472 | retval = security_task_fix_setgid(new, old, LSM_SETID_ID); |
473 | if (retval < 0) | |
474 | goto error; | |
475 | ||
d84f4f99 DH |
476 | return commit_creds(new); |
477 | ||
478 | error: | |
479 | abort_creds(new); | |
480 | return retval; | |
1da177e4 | 481 | } |
54e99124 | 482 | |
e530dca5 DB |
483 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
484 | { | |
485 | return __sys_setgid(gid); | |
486 | } | |
487 | ||
d84f4f99 DH |
488 | /* |
489 | * change the user struct in a credentials set to match the new UID | |
490 | */ | |
491 | static int set_user(struct cred *new) | |
1da177e4 LT |
492 | { |
493 | struct user_struct *new_user; | |
494 | ||
078de5f7 | 495 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
496 | if (!new_user) |
497 | return -EAGAIN; | |
498 | ||
c923a8e7 EB |
499 | free_uid(new->user); |
500 | new->user = new_user; | |
501 | return 0; | |
502 | } | |
503 | ||
504 | static void flag_nproc_exceeded(struct cred *new) | |
505 | { | |
506 | if (new->ucounts == current_ucounts()) | |
507 | return; | |
508 | ||
72fa5997 VK |
509 | /* |
510 | * We don't fail in case of NPROC limit excess here because too many | |
511 | * poorly written programs don't check set*uid() return code, assuming | |
512 | * it never fails if called by root. We may still enforce NPROC limit | |
513 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
514 | * failure to the execve() stage. | |
515 | */ | |
de399236 | 516 | if (is_rlimit_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) && |
c923a8e7 | 517 | new->user != INIT_USER) |
72fa5997 VK |
518 | current->flags |= PF_NPROC_EXCEEDED; |
519 | else | |
520 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 LT |
521 | } |
522 | ||
523 | /* | |
524 | * Unprivileged users may change the real uid to the effective uid | |
525 | * or vice versa. (BSD-style) | |
526 | * | |
527 | * If you set the real uid at all, or set the effective uid to a value not | |
528 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
529 | * | |
530 | * This makes it possible for a setuid program to completely drop its | |
531 | * privileges, which is often a useful assertion to make when you are doing | |
532 | * a security audit over a program. | |
533 | * | |
534 | * The general idea is that a program which uses just setreuid() will be | |
535 | * 100% compatible with BSD. A program which uses just setuid() will be | |
ec94fc3d | 536 | * 100% compatible with POSIX with saved IDs. |
1da177e4 | 537 | */ |
e530dca5 | 538 | long __sys_setreuid(uid_t ruid, uid_t euid) |
1da177e4 | 539 | { |
a29c33f4 | 540 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
541 | const struct cred *old; |
542 | struct cred *new; | |
1da177e4 | 543 | int retval; |
a29c33f4 EB |
544 | kuid_t kruid, keuid; |
545 | ||
546 | kruid = make_kuid(ns, ruid); | |
547 | keuid = make_kuid(ns, euid); | |
548 | ||
549 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
550 | return -EINVAL; | |
551 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
552 | return -EINVAL; | |
1da177e4 | 553 | |
d84f4f99 DH |
554 | new = prepare_creds(); |
555 | if (!new) | |
556 | return -ENOMEM; | |
557 | old = current_cred(); | |
558 | ||
d84f4f99 | 559 | retval = -EPERM; |
1da177e4 | 560 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
561 | new->uid = kruid; |
562 | if (!uid_eq(old->uid, kruid) && | |
563 | !uid_eq(old->euid, kruid) && | |
40852275 | 564 | !ns_capable_setid(old->user_ns, CAP_SETUID)) |
d84f4f99 | 565 | goto error; |
1da177e4 LT |
566 | } |
567 | ||
568 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
569 | new->euid = keuid; |
570 | if (!uid_eq(old->uid, keuid) && | |
571 | !uid_eq(old->euid, keuid) && | |
572 | !uid_eq(old->suid, keuid) && | |
40852275 | 573 | !ns_capable_setid(old->user_ns, CAP_SETUID)) |
d84f4f99 | 574 | goto error; |
1da177e4 LT |
575 | } |
576 | ||
a29c33f4 | 577 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
578 | retval = set_user(new); |
579 | if (retval < 0) | |
580 | goto error; | |
581 | } | |
1da177e4 | 582 | if (ruid != (uid_t) -1 || |
a29c33f4 | 583 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
584 | new->suid = new->euid; |
585 | new->fsuid = new->euid; | |
1da177e4 | 586 | |
d84f4f99 DH |
587 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
588 | if (retval < 0) | |
589 | goto error; | |
1da177e4 | 590 | |
905ae01c AG |
591 | retval = set_cred_ucounts(new); |
592 | if (retval < 0) | |
593 | goto error; | |
594 | ||
c923a8e7 | 595 | flag_nproc_exceeded(new); |
d84f4f99 | 596 | return commit_creds(new); |
1da177e4 | 597 | |
d84f4f99 DH |
598 | error: |
599 | abort_creds(new); | |
600 | return retval; | |
601 | } | |
ec94fc3d | 602 | |
e530dca5 DB |
603 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
604 | { | |
605 | return __sys_setreuid(ruid, euid); | |
606 | } | |
607 | ||
1da177e4 | 608 | /* |
ec94fc3d | 609 | * setuid() is implemented like SysV with SAVED_IDS |
610 | * | |
1da177e4 | 611 | * Note that SAVED_ID's is deficient in that a setuid root program |
ec94fc3d | 612 | * like sendmail, for example, cannot set its uid to be a normal |
1da177e4 LT |
613 | * user and then switch back, because if you're root, setuid() sets |
614 | * the saved uid too. If you don't like this, blame the bright people | |
615 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
616 | * will allow a root program to temporarily drop privileges and be able to | |
ec94fc3d | 617 | * regain them by swapping the real and effective uid. |
1da177e4 | 618 | */ |
e530dca5 | 619 | long __sys_setuid(uid_t uid) |
1da177e4 | 620 | { |
a29c33f4 | 621 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
622 | const struct cred *old; |
623 | struct cred *new; | |
1da177e4 | 624 | int retval; |
a29c33f4 EB |
625 | kuid_t kuid; |
626 | ||
627 | kuid = make_kuid(ns, uid); | |
628 | if (!uid_valid(kuid)) | |
629 | return -EINVAL; | |
1da177e4 | 630 | |
d84f4f99 DH |
631 | new = prepare_creds(); |
632 | if (!new) | |
633 | return -ENOMEM; | |
634 | old = current_cred(); | |
635 | ||
d84f4f99 | 636 | retval = -EPERM; |
40852275 | 637 | if (ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
638 | new->suid = new->uid = kuid; |
639 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
640 | retval = set_user(new); |
641 | if (retval < 0) | |
642 | goto error; | |
d84f4f99 | 643 | } |
a29c33f4 | 644 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 645 | goto error; |
1da177e4 | 646 | } |
1da177e4 | 647 | |
a29c33f4 | 648 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
649 | |
650 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
651 | if (retval < 0) | |
652 | goto error; | |
1da177e4 | 653 | |
905ae01c AG |
654 | retval = set_cred_ucounts(new); |
655 | if (retval < 0) | |
656 | goto error; | |
657 | ||
c923a8e7 | 658 | flag_nproc_exceeded(new); |
d84f4f99 | 659 | return commit_creds(new); |
1da177e4 | 660 | |
d84f4f99 DH |
661 | error: |
662 | abort_creds(new); | |
663 | return retval; | |
1da177e4 LT |
664 | } |
665 | ||
e530dca5 DB |
666 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
667 | { | |
668 | return __sys_setuid(uid); | |
669 | } | |
670 | ||
1da177e4 LT |
671 | |
672 | /* | |
673 | * This function implements a generic ability to update ruid, euid, | |
674 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
675 | */ | |
e530dca5 | 676 | long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) |
1da177e4 | 677 | { |
a29c33f4 | 678 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
679 | const struct cred *old; |
680 | struct cred *new; | |
1da177e4 | 681 | int retval; |
a29c33f4 | 682 | kuid_t kruid, keuid, ksuid; |
659c0ce1 | 683 | bool ruid_new, euid_new, suid_new; |
a29c33f4 EB |
684 | |
685 | kruid = make_kuid(ns, ruid); | |
686 | keuid = make_kuid(ns, euid); | |
687 | ksuid = make_kuid(ns, suid); | |
688 | ||
689 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
690 | return -EINVAL; | |
691 | ||
692 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
693 | return -EINVAL; | |
694 | ||
695 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
696 | return -EINVAL; | |
1da177e4 | 697 | |
659c0ce1 OM |
698 | old = current_cred(); |
699 | ||
700 | /* check for no-op */ | |
701 | if ((ruid == (uid_t) -1 || uid_eq(kruid, old->uid)) && | |
702 | (euid == (uid_t) -1 || (uid_eq(keuid, old->euid) && | |
703 | uid_eq(keuid, old->fsuid))) && | |
704 | (suid == (uid_t) -1 || uid_eq(ksuid, old->suid))) | |
705 | return 0; | |
706 | ||
707 | ruid_new = ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && | |
708 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid); | |
709 | euid_new = euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && | |
710 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid); | |
711 | suid_new = suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && | |
712 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid); | |
713 | if ((ruid_new || euid_new || suid_new) && | |
714 | !ns_capable_setid(old->user_ns, CAP_SETUID)) | |
715 | return -EPERM; | |
716 | ||
d84f4f99 DH |
717 | new = prepare_creds(); |
718 | if (!new) | |
719 | return -ENOMEM; | |
720 | ||
1da177e4 | 721 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
722 | new->uid = kruid; |
723 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
724 | retval = set_user(new); |
725 | if (retval < 0) | |
726 | goto error; | |
727 | } | |
1da177e4 | 728 | } |
d84f4f99 | 729 | if (euid != (uid_t) -1) |
a29c33f4 | 730 | new->euid = keuid; |
1da177e4 | 731 | if (suid != (uid_t) -1) |
a29c33f4 | 732 | new->suid = ksuid; |
d84f4f99 | 733 | new->fsuid = new->euid; |
1da177e4 | 734 | |
d84f4f99 DH |
735 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
736 | if (retval < 0) | |
737 | goto error; | |
1da177e4 | 738 | |
905ae01c AG |
739 | retval = set_cred_ucounts(new); |
740 | if (retval < 0) | |
741 | goto error; | |
742 | ||
c923a8e7 | 743 | flag_nproc_exceeded(new); |
d84f4f99 | 744 | return commit_creds(new); |
1da177e4 | 745 | |
d84f4f99 DH |
746 | error: |
747 | abort_creds(new); | |
748 | return retval; | |
1da177e4 LT |
749 | } |
750 | ||
e530dca5 DB |
751 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
752 | { | |
753 | return __sys_setresuid(ruid, euid, suid); | |
754 | } | |
755 | ||
a29c33f4 | 756 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 757 | { |
86a264ab | 758 | const struct cred *cred = current_cred(); |
1da177e4 | 759 | int retval; |
a29c33f4 EB |
760 | uid_t ruid, euid, suid; |
761 | ||
762 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
763 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
764 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 765 | |
ec94fc3d | 766 | retval = put_user(ruid, ruidp); |
767 | if (!retval) { | |
768 | retval = put_user(euid, euidp); | |
769 | if (!retval) | |
770 | return put_user(suid, suidp); | |
771 | } | |
1da177e4 LT |
772 | return retval; |
773 | } | |
774 | ||
775 | /* | |
776 | * Same as above, but for rgid, egid, sgid. | |
777 | */ | |
e530dca5 | 778 | long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) |
1da177e4 | 779 | { |
a29c33f4 | 780 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
781 | const struct cred *old; |
782 | struct cred *new; | |
1da177e4 | 783 | int retval; |
a29c33f4 | 784 | kgid_t krgid, kegid, ksgid; |
659c0ce1 | 785 | bool rgid_new, egid_new, sgid_new; |
a29c33f4 EB |
786 | |
787 | krgid = make_kgid(ns, rgid); | |
788 | kegid = make_kgid(ns, egid); | |
789 | ksgid = make_kgid(ns, sgid); | |
790 | ||
791 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
792 | return -EINVAL; | |
793 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
794 | return -EINVAL; | |
795 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
796 | return -EINVAL; | |
1da177e4 | 797 | |
659c0ce1 OM |
798 | old = current_cred(); |
799 | ||
800 | /* check for no-op */ | |
801 | if ((rgid == (gid_t) -1 || gid_eq(krgid, old->gid)) && | |
802 | (egid == (gid_t) -1 || (gid_eq(kegid, old->egid) && | |
803 | gid_eq(kegid, old->fsgid))) && | |
804 | (sgid == (gid_t) -1 || gid_eq(ksgid, old->sgid))) | |
805 | return 0; | |
806 | ||
807 | rgid_new = rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && | |
808 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid); | |
809 | egid_new = egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && | |
810 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid); | |
811 | sgid_new = sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && | |
812 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid); | |
813 | if ((rgid_new || egid_new || sgid_new) && | |
814 | !ns_capable_setid(old->user_ns, CAP_SETGID)) | |
815 | return -EPERM; | |
816 | ||
d84f4f99 DH |
817 | new = prepare_creds(); |
818 | if (!new) | |
819 | return -ENOMEM; | |
d84f4f99 | 820 | |
1da177e4 | 821 | if (rgid != (gid_t) -1) |
a29c33f4 | 822 | new->gid = krgid; |
d84f4f99 | 823 | if (egid != (gid_t) -1) |
a29c33f4 | 824 | new->egid = kegid; |
1da177e4 | 825 | if (sgid != (gid_t) -1) |
a29c33f4 | 826 | new->sgid = ksgid; |
d84f4f99 | 827 | new->fsgid = new->egid; |
1da177e4 | 828 | |
39030e13 TC |
829 | retval = security_task_fix_setgid(new, old, LSM_SETID_RES); |
830 | if (retval < 0) | |
831 | goto error; | |
832 | ||
d84f4f99 DH |
833 | return commit_creds(new); |
834 | ||
835 | error: | |
836 | abort_creds(new); | |
837 | return retval; | |
1da177e4 LT |
838 | } |
839 | ||
e530dca5 DB |
840 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
841 | { | |
842 | return __sys_setresgid(rgid, egid, sgid); | |
843 | } | |
844 | ||
a29c33f4 | 845 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 846 | { |
86a264ab | 847 | const struct cred *cred = current_cred(); |
1da177e4 | 848 | int retval; |
a29c33f4 EB |
849 | gid_t rgid, egid, sgid; |
850 | ||
851 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
852 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
853 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 854 | |
ec94fc3d | 855 | retval = put_user(rgid, rgidp); |
856 | if (!retval) { | |
857 | retval = put_user(egid, egidp); | |
858 | if (!retval) | |
859 | retval = put_user(sgid, sgidp); | |
860 | } | |
1da177e4 LT |
861 | |
862 | return retval; | |
863 | } | |
864 | ||
865 | ||
866 | /* | |
867 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
868 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
869 | * whatever uid it wants to). It normally shadows "euid", except when | |
870 | * explicitly set by setfsuid() or for access.. | |
871 | */ | |
e530dca5 | 872 | long __sys_setfsuid(uid_t uid) |
1da177e4 | 873 | { |
d84f4f99 DH |
874 | const struct cred *old; |
875 | struct cred *new; | |
876 | uid_t old_fsuid; | |
a29c33f4 EB |
877 | kuid_t kuid; |
878 | ||
879 | old = current_cred(); | |
880 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
881 | ||
882 | kuid = make_kuid(old->user_ns, uid); | |
883 | if (!uid_valid(kuid)) | |
884 | return old_fsuid; | |
1da177e4 | 885 | |
d84f4f99 DH |
886 | new = prepare_creds(); |
887 | if (!new) | |
a29c33f4 | 888 | return old_fsuid; |
1da177e4 | 889 | |
a29c33f4 EB |
890 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
891 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
40852275 | 892 | ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
893 | if (!uid_eq(kuid, old->fsuid)) { |
894 | new->fsuid = kuid; | |
d84f4f99 DH |
895 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
896 | goto change_okay; | |
1da177e4 | 897 | } |
1da177e4 LT |
898 | } |
899 | ||
d84f4f99 DH |
900 | abort_creds(new); |
901 | return old_fsuid; | |
1da177e4 | 902 | |
d84f4f99 DH |
903 | change_okay: |
904 | commit_creds(new); | |
1da177e4 LT |
905 | return old_fsuid; |
906 | } | |
907 | ||
e530dca5 DB |
908 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
909 | { | |
910 | return __sys_setfsuid(uid); | |
911 | } | |
912 | ||
1da177e4 | 913 | /* |
f42df9e6 | 914 | * Samma på svenska.. |
1da177e4 | 915 | */ |
e530dca5 | 916 | long __sys_setfsgid(gid_t gid) |
1da177e4 | 917 | { |
d84f4f99 DH |
918 | const struct cred *old; |
919 | struct cred *new; | |
920 | gid_t old_fsgid; | |
a29c33f4 EB |
921 | kgid_t kgid; |
922 | ||
923 | old = current_cred(); | |
924 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
925 | ||
926 | kgid = make_kgid(old->user_ns, gid); | |
927 | if (!gid_valid(kgid)) | |
928 | return old_fsgid; | |
d84f4f99 DH |
929 | |
930 | new = prepare_creds(); | |
931 | if (!new) | |
a29c33f4 | 932 | return old_fsgid; |
1da177e4 | 933 | |
a29c33f4 EB |
934 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
935 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
111767c1 | 936 | ns_capable_setid(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
937 | if (!gid_eq(kgid, old->fsgid)) { |
938 | new->fsgid = kgid; | |
39030e13 TC |
939 | if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0) |
940 | goto change_okay; | |
1da177e4 | 941 | } |
1da177e4 | 942 | } |
d84f4f99 | 943 | |
d84f4f99 DH |
944 | abort_creds(new); |
945 | return old_fsgid; | |
946 | ||
947 | change_okay: | |
948 | commit_creds(new); | |
1da177e4 LT |
949 | return old_fsgid; |
950 | } | |
e530dca5 DB |
951 | |
952 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) | |
953 | { | |
954 | return __sys_setfsgid(gid); | |
955 | } | |
2813893f | 956 | #endif /* CONFIG_MULTIUSER */ |
1da177e4 | 957 | |
4a22f166 SR |
958 | /** |
959 | * sys_getpid - return the thread group id of the current process | |
960 | * | |
961 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
962 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
963 | * which case the tgid is the same in all threads of the same group. | |
964 | * | |
965 | * This is SMP safe as current->tgid does not change. | |
966 | */ | |
967 | SYSCALL_DEFINE0(getpid) | |
968 | { | |
969 | return task_tgid_vnr(current); | |
970 | } | |
971 | ||
972 | /* Thread ID - the internal kernel "pid" */ | |
973 | SYSCALL_DEFINE0(gettid) | |
974 | { | |
975 | return task_pid_vnr(current); | |
976 | } | |
977 | ||
978 | /* | |
979 | * Accessing ->real_parent is not SMP-safe, it could | |
980 | * change from under us. However, we can use a stale | |
981 | * value of ->real_parent under rcu_read_lock(), see | |
982 | * release_task()->call_rcu(delayed_put_task_struct). | |
983 | */ | |
984 | SYSCALL_DEFINE0(getppid) | |
985 | { | |
986 | int pid; | |
987 | ||
988 | rcu_read_lock(); | |
989 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
990 | rcu_read_unlock(); | |
991 | ||
992 | return pid; | |
993 | } | |
994 | ||
995 | SYSCALL_DEFINE0(getuid) | |
996 | { | |
997 | /* Only we change this so SMP safe */ | |
998 | return from_kuid_munged(current_user_ns(), current_uid()); | |
999 | } | |
1000 | ||
1001 | SYSCALL_DEFINE0(geteuid) | |
1002 | { | |
1003 | /* Only we change this so SMP safe */ | |
1004 | return from_kuid_munged(current_user_ns(), current_euid()); | |
1005 | } | |
1006 | ||
1007 | SYSCALL_DEFINE0(getgid) | |
1008 | { | |
1009 | /* Only we change this so SMP safe */ | |
1010 | return from_kgid_munged(current_user_ns(), current_gid()); | |
1011 | } | |
1012 | ||
1013 | SYSCALL_DEFINE0(getegid) | |
1014 | { | |
1015 | /* Only we change this so SMP safe */ | |
1016 | return from_kgid_munged(current_user_ns(), current_egid()); | |
1017 | } | |
1018 | ||
ca2406ed | 1019 | static void do_sys_times(struct tms *tms) |
f06febc9 | 1020 | { |
5613fda9 | 1021 | u64 tgutime, tgstime, cutime, cstime; |
f06febc9 | 1022 | |
e80d0a1a | 1023 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
1024 | cutime = current->signal->cutime; |
1025 | cstime = current->signal->cstime; | |
5613fda9 FW |
1026 | tms->tms_utime = nsec_to_clock_t(tgutime); |
1027 | tms->tms_stime = nsec_to_clock_t(tgstime); | |
1028 | tms->tms_cutime = nsec_to_clock_t(cutime); | |
1029 | tms->tms_cstime = nsec_to_clock_t(cstime); | |
f06febc9 FM |
1030 | } |
1031 | ||
58fd3aa2 | 1032 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 1033 | { |
1da177e4 LT |
1034 | if (tbuf) { |
1035 | struct tms tmp; | |
f06febc9 FM |
1036 | |
1037 | do_sys_times(&tmp); | |
1da177e4 LT |
1038 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
1039 | return -EFAULT; | |
1040 | } | |
e3d5a27d | 1041 | force_successful_syscall_return(); |
1da177e4 LT |
1042 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
1043 | } | |
1044 | ||
ca2406ed AV |
1045 | #ifdef CONFIG_COMPAT |
1046 | static compat_clock_t clock_t_to_compat_clock_t(clock_t x) | |
1047 | { | |
1048 | return compat_jiffies_to_clock_t(clock_t_to_jiffies(x)); | |
1049 | } | |
1050 | ||
1051 | COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf) | |
1052 | { | |
1053 | if (tbuf) { | |
1054 | struct tms tms; | |
1055 | struct compat_tms tmp; | |
1056 | ||
1057 | do_sys_times(&tms); | |
1058 | /* Convert our struct tms to the compat version. */ | |
1059 | tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime); | |
1060 | tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime); | |
1061 | tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime); | |
1062 | tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime); | |
1063 | if (copy_to_user(tbuf, &tmp, sizeof(tmp))) | |
1064 | return -EFAULT; | |
1065 | } | |
1066 | force_successful_syscall_return(); | |
1067 | return compat_jiffies_to_clock_t(jiffies); | |
1068 | } | |
1069 | #endif | |
1070 | ||
1da177e4 LT |
1071 | /* |
1072 | * This needs some heavy checking ... | |
1073 | * I just haven't the stomach for it. I also don't fully | |
1074 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
1075 | * | |
1076 | * OK, I think I have the protection semantics right.... this is really | |
1077 | * only important on a multi-user system anyway, to make sure one user | |
1078 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
1079 | * | |
98611e4e | 1080 | * !PF_FORKNOEXEC check to conform completely to POSIX. |
1da177e4 | 1081 | */ |
b290ebe2 | 1082 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
1083 | { |
1084 | struct task_struct *p; | |
ee0acf90 | 1085 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
1086 | struct pid *pgrp; |
1087 | int err; | |
1da177e4 LT |
1088 | |
1089 | if (!pid) | |
b488893a | 1090 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
1091 | if (!pgid) |
1092 | pgid = pid; | |
1093 | if (pgid < 0) | |
1094 | return -EINVAL; | |
950eaaca | 1095 | rcu_read_lock(); |
1da177e4 LT |
1096 | |
1097 | /* From this point forward we keep holding onto the tasklist lock | |
1098 | * so that our parent does not change from under us. -DaveM | |
1099 | */ | |
1100 | write_lock_irq(&tasklist_lock); | |
1101 | ||
1102 | err = -ESRCH; | |
4e021306 | 1103 | p = find_task_by_vpid(pid); |
1da177e4 LT |
1104 | if (!p) |
1105 | goto out; | |
1106 | ||
1107 | err = -EINVAL; | |
1108 | if (!thread_group_leader(p)) | |
1109 | goto out; | |
1110 | ||
4e021306 | 1111 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 1112 | err = -EPERM; |
41487c65 | 1113 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
1114 | goto out; |
1115 | err = -EACCES; | |
98611e4e | 1116 | if (!(p->flags & PF_FORKNOEXEC)) |
1da177e4 LT |
1117 | goto out; |
1118 | } else { | |
1119 | err = -ESRCH; | |
ee0acf90 | 1120 | if (p != group_leader) |
1da177e4 LT |
1121 | goto out; |
1122 | } | |
1123 | ||
1124 | err = -EPERM; | |
1125 | if (p->signal->leader) | |
1126 | goto out; | |
1127 | ||
4e021306 | 1128 | pgrp = task_pid(p); |
1da177e4 | 1129 | if (pgid != pid) { |
b488893a | 1130 | struct task_struct *g; |
1da177e4 | 1131 | |
4e021306 ON |
1132 | pgrp = find_vpid(pgid); |
1133 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 1134 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 1135 | goto out; |
1da177e4 LT |
1136 | } |
1137 | ||
1da177e4 LT |
1138 | err = security_task_setpgid(p, pgid); |
1139 | if (err) | |
1140 | goto out; | |
1141 | ||
1b0f7ffd | 1142 | if (task_pgrp(p) != pgrp) |
83beaf3c | 1143 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
1144 | |
1145 | err = 0; | |
1146 | out: | |
1147 | /* All paths lead to here, thus we are safe. -DaveM */ | |
1148 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 1149 | rcu_read_unlock(); |
1da177e4 LT |
1150 | return err; |
1151 | } | |
1152 | ||
192c5807 | 1153 | static int do_getpgid(pid_t pid) |
1da177e4 | 1154 | { |
12a3de0a ON |
1155 | struct task_struct *p; |
1156 | struct pid *grp; | |
1157 | int retval; | |
1158 | ||
1159 | rcu_read_lock(); | |
756184b7 | 1160 | if (!pid) |
12a3de0a | 1161 | grp = task_pgrp(current); |
756184b7 | 1162 | else { |
1da177e4 | 1163 | retval = -ESRCH; |
12a3de0a ON |
1164 | p = find_task_by_vpid(pid); |
1165 | if (!p) | |
1166 | goto out; | |
1167 | grp = task_pgrp(p); | |
1168 | if (!grp) | |
1169 | goto out; | |
1170 | ||
1171 | retval = security_task_getpgid(p); | |
1172 | if (retval) | |
1173 | goto out; | |
1da177e4 | 1174 | } |
12a3de0a ON |
1175 | retval = pid_vnr(grp); |
1176 | out: | |
1177 | rcu_read_unlock(); | |
1178 | return retval; | |
1da177e4 LT |
1179 | } |
1180 | ||
192c5807 DB |
1181 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1182 | { | |
1183 | return do_getpgid(pid); | |
1184 | } | |
1185 | ||
1da177e4 LT |
1186 | #ifdef __ARCH_WANT_SYS_GETPGRP |
1187 | ||
dbf040d9 | 1188 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1189 | { |
192c5807 | 1190 | return do_getpgid(0); |
1da177e4 LT |
1191 | } |
1192 | ||
1193 | #endif | |
1194 | ||
dbf040d9 | 1195 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1196 | { |
1dd768c0 ON |
1197 | struct task_struct *p; |
1198 | struct pid *sid; | |
1199 | int retval; | |
1200 | ||
1201 | rcu_read_lock(); | |
756184b7 | 1202 | if (!pid) |
1dd768c0 | 1203 | sid = task_session(current); |
756184b7 | 1204 | else { |
1da177e4 | 1205 | retval = -ESRCH; |
1dd768c0 ON |
1206 | p = find_task_by_vpid(pid); |
1207 | if (!p) | |
1208 | goto out; | |
1209 | sid = task_session(p); | |
1210 | if (!sid) | |
1211 | goto out; | |
1212 | ||
1213 | retval = security_task_getsid(p); | |
1214 | if (retval) | |
1215 | goto out; | |
1da177e4 | 1216 | } |
1dd768c0 ON |
1217 | retval = pid_vnr(sid); |
1218 | out: | |
1219 | rcu_read_unlock(); | |
1220 | return retval; | |
1da177e4 LT |
1221 | } |
1222 | ||
81dabb46 ON |
1223 | static void set_special_pids(struct pid *pid) |
1224 | { | |
1225 | struct task_struct *curr = current->group_leader; | |
1226 | ||
1227 | if (task_session(curr) != pid) | |
1228 | change_pid(curr, PIDTYPE_SID, pid); | |
1229 | ||
1230 | if (task_pgrp(curr) != pid) | |
1231 | change_pid(curr, PIDTYPE_PGID, pid); | |
1232 | } | |
1233 | ||
e2aaa9f4 | 1234 | int ksys_setsid(void) |
1da177e4 | 1235 | { |
e19f247a | 1236 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1237 | struct pid *sid = task_pid(group_leader); |
1238 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1239 | int err = -EPERM; |
1240 | ||
1da177e4 | 1241 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1242 | /* Fail if I am already a session leader */ |
1243 | if (group_leader->signal->leader) | |
1244 | goto out; | |
1245 | ||
430c6231 ON |
1246 | /* Fail if a process group id already exists that equals the |
1247 | * proposed session id. | |
390e2ff0 | 1248 | */ |
6806aac6 | 1249 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1250 | goto out; |
1251 | ||
e19f247a | 1252 | group_leader->signal->leader = 1; |
81dabb46 | 1253 | set_special_pids(sid); |
24ec839c | 1254 | |
9c9f4ded | 1255 | proc_clear_tty(group_leader); |
24ec839c | 1256 | |
e4cc0a9c | 1257 | err = session; |
1da177e4 LT |
1258 | out: |
1259 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1260 | if (err > 0) { |
0d0df599 | 1261 | proc_sid_connector(group_leader); |
5091faa4 MG |
1262 | sched_autogroup_create_attach(group_leader); |
1263 | } | |
1da177e4 LT |
1264 | return err; |
1265 | } | |
1266 | ||
e2aaa9f4 DB |
1267 | SYSCALL_DEFINE0(setsid) |
1268 | { | |
1269 | return ksys_setsid(); | |
1270 | } | |
1271 | ||
1da177e4 LT |
1272 | DECLARE_RWSEM(uts_sem); |
1273 | ||
e28cbf22 CH |
1274 | #ifdef COMPAT_UTS_MACHINE |
1275 | #define override_architecture(name) \ | |
46da2766 | 1276 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1277 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1278 | sizeof(COMPAT_UTS_MACHINE))) | |
1279 | #else | |
1280 | #define override_architecture(name) 0 | |
1281 | #endif | |
1282 | ||
be27425d AK |
1283 | /* |
1284 | * Work around broken programs that cannot handle "Linux 3.0". | |
1285 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
b7285b42 JN |
1286 | * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be |
1287 | * 2.6.60. | |
be27425d | 1288 | */ |
2702b152 | 1289 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1290 | { |
1291 | int ret = 0; | |
be27425d AK |
1292 | |
1293 | if (current->personality & UNAME26) { | |
2702b152 KC |
1294 | const char *rest = UTS_RELEASE; |
1295 | char buf[65] = { 0 }; | |
be27425d AK |
1296 | int ndots = 0; |
1297 | unsigned v; | |
2702b152 | 1298 | size_t copy; |
be27425d AK |
1299 | |
1300 | while (*rest) { | |
1301 | if (*rest == '.' && ++ndots >= 3) | |
1302 | break; | |
1303 | if (!isdigit(*rest) && *rest != '.') | |
1304 | break; | |
1305 | rest++; | |
1306 | } | |
88a68672 | 1307 | v = LINUX_VERSION_PATCHLEVEL + 60; |
31fd84b9 | 1308 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1309 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1310 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1311 | } |
1312 | return ret; | |
1313 | } | |
1314 | ||
e48fbb69 | 1315 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 | 1316 | { |
42a0cc34 | 1317 | struct new_utsname tmp; |
1da177e4 LT |
1318 | |
1319 | down_read(&uts_sem); | |
42a0cc34 | 1320 | memcpy(&tmp, utsname(), sizeof(tmp)); |
1da177e4 | 1321 | up_read(&uts_sem); |
42a0cc34 JH |
1322 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1323 | return -EFAULT; | |
e28cbf22 | 1324 | |
42a0cc34 JH |
1325 | if (override_release(name->release, sizeof(name->release))) |
1326 | return -EFAULT; | |
1327 | if (override_architecture(name)) | |
1328 | return -EFAULT; | |
1329 | return 0; | |
1da177e4 LT |
1330 | } |
1331 | ||
5cacdb4a CH |
1332 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1333 | /* | |
1334 | * Old cruft | |
1335 | */ | |
1336 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1337 | { | |
42a0cc34 | 1338 | struct old_utsname tmp; |
5cacdb4a CH |
1339 | |
1340 | if (!name) | |
1341 | return -EFAULT; | |
1342 | ||
1343 | down_read(&uts_sem); | |
42a0cc34 | 1344 | memcpy(&tmp, utsname(), sizeof(tmp)); |
5cacdb4a | 1345 | up_read(&uts_sem); |
42a0cc34 JH |
1346 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1347 | return -EFAULT; | |
5cacdb4a | 1348 | |
42a0cc34 JH |
1349 | if (override_release(name->release, sizeof(name->release))) |
1350 | return -EFAULT; | |
1351 | if (override_architecture(name)) | |
1352 | return -EFAULT; | |
1353 | return 0; | |
5cacdb4a CH |
1354 | } |
1355 | ||
1356 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1357 | { | |
5e1aada0 | 1358 | struct oldold_utsname tmp; |
5cacdb4a CH |
1359 | |
1360 | if (!name) | |
1361 | return -EFAULT; | |
5cacdb4a | 1362 | |
5e1aada0 JP |
1363 | memset(&tmp, 0, sizeof(tmp)); |
1364 | ||
5cacdb4a | 1365 | down_read(&uts_sem); |
42a0cc34 JH |
1366 | memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN); |
1367 | memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN); | |
1368 | memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN); | |
1369 | memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN); | |
1370 | memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN); | |
5cacdb4a | 1371 | up_read(&uts_sem); |
42a0cc34 JH |
1372 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1373 | return -EFAULT; | |
5cacdb4a | 1374 | |
42a0cc34 JH |
1375 | if (override_architecture(name)) |
1376 | return -EFAULT; | |
1377 | if (override_release(name->release, sizeof(name->release))) | |
1378 | return -EFAULT; | |
1379 | return 0; | |
5cacdb4a CH |
1380 | } |
1381 | #endif | |
1382 | ||
5a8a82b1 | 1383 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1384 | { |
1385 | int errno; | |
1386 | char tmp[__NEW_UTS_LEN]; | |
1387 | ||
bb96a6f5 | 1388 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1389 | return -EPERM; |
fc832ad3 | 1390 | |
1da177e4 LT |
1391 | if (len < 0 || len > __NEW_UTS_LEN) |
1392 | return -EINVAL; | |
1da177e4 LT |
1393 | errno = -EFAULT; |
1394 | if (!copy_from_user(tmp, name, len)) { | |
42a0cc34 | 1395 | struct new_utsname *u; |
9679e4dd | 1396 | |
37608ba3 | 1397 | add_device_randomness(tmp, len); |
42a0cc34 JH |
1398 | down_write(&uts_sem); |
1399 | u = utsname(); | |
9679e4dd AM |
1400 | memcpy(u->nodename, tmp, len); |
1401 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1402 | errno = 0; |
499eea6b | 1403 | uts_proc_notify(UTS_PROC_HOSTNAME); |
42a0cc34 | 1404 | up_write(&uts_sem); |
1da177e4 | 1405 | } |
1da177e4 LT |
1406 | return errno; |
1407 | } | |
1408 | ||
1409 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1410 | ||
5a8a82b1 | 1411 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 | 1412 | { |
42a0cc34 | 1413 | int i; |
9679e4dd | 1414 | struct new_utsname *u; |
42a0cc34 | 1415 | char tmp[__NEW_UTS_LEN + 1]; |
1da177e4 LT |
1416 | |
1417 | if (len < 0) | |
1418 | return -EINVAL; | |
1419 | down_read(&uts_sem); | |
9679e4dd AM |
1420 | u = utsname(); |
1421 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1422 | if (i > len) |
1423 | i = len; | |
42a0cc34 | 1424 | memcpy(tmp, u->nodename, i); |
1da177e4 | 1425 | up_read(&uts_sem); |
42a0cc34 JH |
1426 | if (copy_to_user(name, tmp, i)) |
1427 | return -EFAULT; | |
1428 | return 0; | |
1da177e4 LT |
1429 | } |
1430 | ||
1431 | #endif | |
1432 | ||
1433 | /* | |
1434 | * Only setdomainname; getdomainname can be implemented by calling | |
1435 | * uname() | |
1436 | */ | |
5a8a82b1 | 1437 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1438 | { |
1439 | int errno; | |
1440 | char tmp[__NEW_UTS_LEN]; | |
1441 | ||
fc832ad3 | 1442 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1443 | return -EPERM; |
1444 | if (len < 0 || len > __NEW_UTS_LEN) | |
1445 | return -EINVAL; | |
1446 | ||
1da177e4 LT |
1447 | errno = -EFAULT; |
1448 | if (!copy_from_user(tmp, name, len)) { | |
42a0cc34 | 1449 | struct new_utsname *u; |
9679e4dd | 1450 | |
37608ba3 | 1451 | add_device_randomness(tmp, len); |
42a0cc34 JH |
1452 | down_write(&uts_sem); |
1453 | u = utsname(); | |
9679e4dd AM |
1454 | memcpy(u->domainname, tmp, len); |
1455 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1456 | errno = 0; |
499eea6b | 1457 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
42a0cc34 | 1458 | up_write(&uts_sem); |
1da177e4 | 1459 | } |
1da177e4 LT |
1460 | return errno; |
1461 | } | |
1462 | ||
c57bef02 BR |
1463 | /* make sure you are allowed to change @tsk limits before calling this */ |
1464 | static int do_prlimit(struct task_struct *tsk, unsigned int resource, | |
1465 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1466 | { | |
1467 | struct rlimit *rlim; | |
1468 | int retval = 0; | |
1469 | ||
1470 | if (resource >= RLIM_NLIMITS) | |
1471 | return -EINVAL; | |
73979060 GKH |
1472 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
1473 | ||
c57bef02 BR |
1474 | if (new_rlim) { |
1475 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1476 | return -EINVAL; | |
1477 | if (resource == RLIMIT_NOFILE && | |
1478 | new_rlim->rlim_max > sysctl_nr_open) | |
1479 | return -EPERM; | |
1480 | } | |
1481 | ||
18c91bb2 | 1482 | /* Holding a refcount on tsk protects tsk->signal from disappearing. */ |
c57bef02 BR |
1483 | rlim = tsk->signal->rlim + resource; |
1484 | task_lock(tsk->group_leader); | |
1485 | if (new_rlim) { | |
1486 | /* | |
1487 | * Keep the capable check against init_user_ns until cgroups can | |
1488 | * contain all limits. | |
1489 | */ | |
1490 | if (new_rlim->rlim_max > rlim->rlim_max && | |
1491 | !capable(CAP_SYS_RESOURCE)) | |
1492 | retval = -EPERM; | |
1493 | if (!retval) | |
1494 | retval = security_task_setrlimit(tsk, resource, new_rlim); | |
1495 | } | |
1496 | if (!retval) { | |
1497 | if (old_rlim) | |
1498 | *old_rlim = *rlim; | |
1499 | if (new_rlim) | |
1500 | *rlim = *new_rlim; | |
1501 | } | |
1502 | task_unlock(tsk->group_leader); | |
1503 | ||
1504 | /* | |
1505 | * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not | |
1506 | * infinite. In case of RLIM_INFINITY the posix CPU timer code | |
1507 | * ignores the rlimit. | |
1508 | */ | |
1509 | if (!retval && new_rlim && resource == RLIMIT_CPU && | |
1510 | new_rlim->rlim_cur != RLIM_INFINITY && | |
18c91bb2 BR |
1511 | IS_ENABLED(CONFIG_POSIX_TIMERS)) { |
1512 | /* | |
1513 | * update_rlimit_cpu can fail if the task is exiting, but there | |
1514 | * may be other tasks in the thread group that are not exiting, | |
1515 | * and they need their cpu timers adjusted. | |
1516 | * | |
1517 | * The group_leader is the last task to be released, so if we | |
1518 | * cannot update_rlimit_cpu on it, then the entire process is | |
1519 | * exiting and we do not need to update at all. | |
1520 | */ | |
1521 | update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur); | |
1522 | } | |
1523 | ||
c57bef02 BR |
1524 | return retval; |
1525 | } | |
1526 | ||
e48fbb69 | 1527 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1528 | { |
b9518345 JS |
1529 | struct rlimit value; |
1530 | int ret; | |
1531 | ||
1532 | ret = do_prlimit(current, resource, NULL, &value); | |
1533 | if (!ret) | |
1534 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1535 | ||
1536 | return ret; | |
1da177e4 LT |
1537 | } |
1538 | ||
d9e968cb AV |
1539 | #ifdef CONFIG_COMPAT |
1540 | ||
1541 | COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource, | |
1542 | struct compat_rlimit __user *, rlim) | |
1543 | { | |
1544 | struct rlimit r; | |
1545 | struct compat_rlimit r32; | |
1546 | ||
1547 | if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit))) | |
1548 | return -EFAULT; | |
1549 | ||
1550 | if (r32.rlim_cur == COMPAT_RLIM_INFINITY) | |
1551 | r.rlim_cur = RLIM_INFINITY; | |
1552 | else | |
1553 | r.rlim_cur = r32.rlim_cur; | |
1554 | if (r32.rlim_max == COMPAT_RLIM_INFINITY) | |
1555 | r.rlim_max = RLIM_INFINITY; | |
1556 | else | |
1557 | r.rlim_max = r32.rlim_max; | |
1558 | return do_prlimit(current, resource, &r, NULL); | |
1559 | } | |
1560 | ||
1561 | COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource, | |
1562 | struct compat_rlimit __user *, rlim) | |
1563 | { | |
1564 | struct rlimit r; | |
1565 | int ret; | |
1566 | ||
1567 | ret = do_prlimit(current, resource, NULL, &r); | |
1568 | if (!ret) { | |
58c7ffc0 | 1569 | struct compat_rlimit r32; |
d9e968cb AV |
1570 | if (r.rlim_cur > COMPAT_RLIM_INFINITY) |
1571 | r32.rlim_cur = COMPAT_RLIM_INFINITY; | |
1572 | else | |
1573 | r32.rlim_cur = r.rlim_cur; | |
1574 | if (r.rlim_max > COMPAT_RLIM_INFINITY) | |
1575 | r32.rlim_max = COMPAT_RLIM_INFINITY; | |
1576 | else | |
1577 | r32.rlim_max = r.rlim_max; | |
1578 | ||
1579 | if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit))) | |
1580 | return -EFAULT; | |
1581 | } | |
1582 | return ret; | |
1583 | } | |
1584 | ||
1585 | #endif | |
1586 | ||
1da177e4 LT |
1587 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT |
1588 | ||
1589 | /* | |
1590 | * Back compatibility for getrlimit. Needed for some apps. | |
1591 | */ | |
e48fbb69 HC |
1592 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1593 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1594 | { |
1595 | struct rlimit x; | |
1596 | if (resource >= RLIM_NLIMITS) | |
1597 | return -EINVAL; | |
1598 | ||
23d6aef7 | 1599 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
1da177e4 LT |
1600 | task_lock(current->group_leader); |
1601 | x = current->signal->rlim[resource]; | |
1602 | task_unlock(current->group_leader); | |
756184b7 | 1603 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1604 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1605 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 | 1606 | x.rlim_max = 0x7FFFFFFF; |
ec94fc3d | 1607 | return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0; |
1da177e4 LT |
1608 | } |
1609 | ||
613763a1 AV |
1610 | #ifdef CONFIG_COMPAT |
1611 | COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, | |
1612 | struct compat_rlimit __user *, rlim) | |
1613 | { | |
1614 | struct rlimit r; | |
1615 | ||
1616 | if (resource >= RLIM_NLIMITS) | |
1617 | return -EINVAL; | |
1618 | ||
23d6aef7 | 1619 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
613763a1 AV |
1620 | task_lock(current->group_leader); |
1621 | r = current->signal->rlim[resource]; | |
1622 | task_unlock(current->group_leader); | |
1623 | if (r.rlim_cur > 0x7FFFFFFF) | |
1624 | r.rlim_cur = 0x7FFFFFFF; | |
1625 | if (r.rlim_max > 0x7FFFFFFF) | |
1626 | r.rlim_max = 0x7FFFFFFF; | |
1627 | ||
1628 | if (put_user(r.rlim_cur, &rlim->rlim_cur) || | |
1629 | put_user(r.rlim_max, &rlim->rlim_max)) | |
1630 | return -EFAULT; | |
1631 | return 0; | |
1632 | } | |
1633 | #endif | |
1634 | ||
1da177e4 LT |
1635 | #endif |
1636 | ||
c022a0ac JS |
1637 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1638 | { | |
1639 | #if BITS_PER_LONG < 64 | |
1640 | return rlim64 >= ULONG_MAX; | |
1641 | #else | |
1642 | return rlim64 == RLIM64_INFINITY; | |
1643 | #endif | |
1644 | } | |
1645 | ||
1646 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1647 | { | |
1648 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1649 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1650 | else | |
1651 | rlim64->rlim_cur = rlim->rlim_cur; | |
1652 | if (rlim->rlim_max == RLIM_INFINITY) | |
1653 | rlim64->rlim_max = RLIM64_INFINITY; | |
1654 | else | |
1655 | rlim64->rlim_max = rlim->rlim_max; | |
1656 | } | |
1657 | ||
1658 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1659 | { | |
1660 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1661 | rlim->rlim_cur = RLIM_INFINITY; | |
1662 | else | |
1663 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1664 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1665 | rlim->rlim_max = RLIM_INFINITY; | |
1666 | else | |
1667 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1668 | } | |
1669 | ||
c022a0ac | 1670 | /* rcu lock must be held */ |
791ec491 SS |
1671 | static int check_prlimit_permission(struct task_struct *task, |
1672 | unsigned int flags) | |
c022a0ac JS |
1673 | { |
1674 | const struct cred *cred = current_cred(), *tcred; | |
791ec491 | 1675 | bool id_match; |
c022a0ac | 1676 | |
fc832ad3 SH |
1677 | if (current == task) |
1678 | return 0; | |
c022a0ac | 1679 | |
fc832ad3 | 1680 | tcred = __task_cred(task); |
791ec491 SS |
1681 | id_match = (uid_eq(cred->uid, tcred->euid) && |
1682 | uid_eq(cred->uid, tcred->suid) && | |
1683 | uid_eq(cred->uid, tcred->uid) && | |
1684 | gid_eq(cred->gid, tcred->egid) && | |
1685 | gid_eq(cred->gid, tcred->sgid) && | |
1686 | gid_eq(cred->gid, tcred->gid)); | |
1687 | if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) | |
1688 | return -EPERM; | |
fc832ad3 | 1689 | |
791ec491 | 1690 | return security_task_prlimit(cred, tcred, flags); |
c022a0ac JS |
1691 | } |
1692 | ||
1693 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1694 | const struct rlimit64 __user *, new_rlim, | |
1695 | struct rlimit64 __user *, old_rlim) | |
1696 | { | |
1697 | struct rlimit64 old64, new64; | |
1698 | struct rlimit old, new; | |
1699 | struct task_struct *tsk; | |
791ec491 | 1700 | unsigned int checkflags = 0; |
c022a0ac JS |
1701 | int ret; |
1702 | ||
791ec491 SS |
1703 | if (old_rlim) |
1704 | checkflags |= LSM_PRLIMIT_READ; | |
1705 | ||
c022a0ac JS |
1706 | if (new_rlim) { |
1707 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1708 | return -EFAULT; | |
1709 | rlim64_to_rlim(&new64, &new); | |
791ec491 | 1710 | checkflags |= LSM_PRLIMIT_WRITE; |
c022a0ac JS |
1711 | } |
1712 | ||
1713 | rcu_read_lock(); | |
1714 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1715 | if (!tsk) { | |
1716 | rcu_read_unlock(); | |
1717 | return -ESRCH; | |
1718 | } | |
791ec491 | 1719 | ret = check_prlimit_permission(tsk, checkflags); |
c022a0ac JS |
1720 | if (ret) { |
1721 | rcu_read_unlock(); | |
1722 | return ret; | |
1723 | } | |
1724 | get_task_struct(tsk); | |
1725 | rcu_read_unlock(); | |
1726 | ||
1727 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1728 | old_rlim ? &old : NULL); | |
1729 | ||
1730 | if (!ret && old_rlim) { | |
1731 | rlim_to_rlim64(&old, &old64); | |
1732 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1733 | ret = -EFAULT; | |
1734 | } | |
1735 | ||
1736 | put_task_struct(tsk); | |
1737 | return ret; | |
1738 | } | |
1739 | ||
7855c35d JS |
1740 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1741 | { | |
1742 | struct rlimit new_rlim; | |
1743 | ||
1744 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1745 | return -EFAULT; | |
5b41535a | 1746 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1747 | } |
1748 | ||
1da177e4 LT |
1749 | /* |
1750 | * It would make sense to put struct rusage in the task_struct, | |
1751 | * except that would make the task_struct be *really big*. After | |
1752 | * task_struct gets moved into malloc'ed memory, it would | |
1753 | * make sense to do this. It will make moving the rest of the information | |
1754 | * a lot simpler! (Which we're not doing right now because we're not | |
1755 | * measuring them yet). | |
1756 | * | |
1da177e4 LT |
1757 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1758 | * races with threads incrementing their own counters. But since word | |
1759 | * reads are atomic, we either get new values or old values and we don't | |
1760 | * care which for the sums. We always take the siglock to protect reading | |
1761 | * the c* fields from p->signal from races with exit.c updating those | |
1762 | * fields when reaping, so a sample either gets all the additions of a | |
1763 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1764 | * |
de047c1b RT |
1765 | * Locking: |
1766 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1767 | * for the cases current multithreaded, non-current single threaded | |
1768 | * non-current multithreaded. Thread traversal is now safe with | |
1769 | * the siglock held. | |
1770 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1771 | * single threaded, as no one else can take our signal_struct away, no one | |
1772 | * else can reap the children to update signal->c* counters, and no one else | |
1773 | * can race with the signal-> fields. If we do not take any lock, the | |
1774 | * signal-> fields could be read out of order while another thread was just | |
1775 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1776 | * On the writer side, write memory barrier is implied in __exit_signal | |
1777 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1778 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1779 | * |
1da177e4 LT |
1780 | */ |
1781 | ||
f06febc9 | 1782 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1783 | { |
679c9cd4 SK |
1784 | r->ru_nvcsw += t->nvcsw; |
1785 | r->ru_nivcsw += t->nivcsw; | |
1786 | r->ru_minflt += t->min_flt; | |
1787 | r->ru_majflt += t->maj_flt; | |
1788 | r->ru_inblock += task_io_get_inblock(t); | |
1789 | r->ru_oublock += task_io_get_oublock(t); | |
1790 | } | |
1791 | ||
ce72a16f | 1792 | void getrusage(struct task_struct *p, int who, struct rusage *r) |
1da177e4 LT |
1793 | { |
1794 | struct task_struct *t; | |
1795 | unsigned long flags; | |
5613fda9 | 1796 | u64 tgutime, tgstime, utime, stime; |
daa694e4 ON |
1797 | unsigned long maxrss; |
1798 | struct mm_struct *mm; | |
c7ac8231 | 1799 | struct signal_struct *sig = p->signal; |
f7ec1cd5 | 1800 | unsigned int seq = 0; |
1da177e4 | 1801 | |
f7ec1cd5 | 1802 | retry: |
daa694e4 | 1803 | memset(r, 0, sizeof(*r)); |
64861634 | 1804 | utime = stime = 0; |
daa694e4 | 1805 | maxrss = 0; |
1da177e4 | 1806 | |
679c9cd4 | 1807 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1808 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1809 | accumulate_thread_rusage(p, r); |
c7ac8231 | 1810 | maxrss = sig->maxrss; |
daa694e4 | 1811 | goto out_thread; |
679c9cd4 SK |
1812 | } |
1813 | ||
f7ec1cd5 | 1814 | flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq); |
0f59cc4a | 1815 | |
1da177e4 | 1816 | switch (who) { |
ec94fc3d | 1817 | case RUSAGE_BOTH: |
1818 | case RUSAGE_CHILDREN: | |
c7ac8231 ON |
1819 | utime = sig->cutime; |
1820 | stime = sig->cstime; | |
1821 | r->ru_nvcsw = sig->cnvcsw; | |
1822 | r->ru_nivcsw = sig->cnivcsw; | |
1823 | r->ru_minflt = sig->cmin_flt; | |
1824 | r->ru_majflt = sig->cmaj_flt; | |
1825 | r->ru_inblock = sig->cinblock; | |
1826 | r->ru_oublock = sig->coublock; | |
1827 | maxrss = sig->cmaxrss; | |
ec94fc3d | 1828 | |
1829 | if (who == RUSAGE_CHILDREN) | |
1da177e4 | 1830 | break; |
df561f66 | 1831 | fallthrough; |
0f59cc4a | 1832 | |
ec94fc3d | 1833 | case RUSAGE_SELF: |
c7ac8231 ON |
1834 | r->ru_nvcsw += sig->nvcsw; |
1835 | r->ru_nivcsw += sig->nivcsw; | |
1836 | r->ru_minflt += sig->min_flt; | |
1837 | r->ru_majflt += sig->maj_flt; | |
1838 | r->ru_inblock += sig->inblock; | |
1839 | r->ru_oublock += sig->oublock; | |
1840 | if (maxrss < sig->maxrss) | |
1841 | maxrss = sig->maxrss; | |
f7ec1cd5 ON |
1842 | |
1843 | rcu_read_lock(); | |
13b7bc60 | 1844 | __for_each_thread(sig, t) |
ec94fc3d | 1845 | accumulate_thread_rusage(t, r); |
f7ec1cd5 ON |
1846 | rcu_read_unlock(); |
1847 | ||
ec94fc3d | 1848 | break; |
1849 | ||
1850 | default: | |
1851 | BUG(); | |
1da177e4 | 1852 | } |
f7ec1cd5 ON |
1853 | |
1854 | if (need_seqretry(&sig->stats_lock, seq)) { | |
1855 | seq = 1; | |
1856 | goto retry; | |
1857 | } | |
1858 | done_seqretry_irqrestore(&sig->stats_lock, seq, flags); | |
de047c1b | 1859 | |
daa694e4 ON |
1860 | if (who == RUSAGE_CHILDREN) |
1861 | goto out_children; | |
1f10206c | 1862 | |
daa694e4 ON |
1863 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
1864 | utime += tgutime; | |
1865 | stime += tgstime; | |
ec94fc3d | 1866 | |
daa694e4 ON |
1867 | out_thread: |
1868 | mm = get_task_mm(p); | |
1869 | if (mm) { | |
1870 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1871 | mmput(mm); | |
1f10206c | 1872 | } |
daa694e4 ON |
1873 | |
1874 | out_children: | |
1f10206c | 1875 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ |
daa694e4 ON |
1876 | r->ru_utime = ns_to_kernel_old_timeval(utime); |
1877 | r->ru_stime = ns_to_kernel_old_timeval(stime); | |
1da177e4 LT |
1878 | } |
1879 | ||
ce72a16f | 1880 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 LT |
1881 | { |
1882 | struct rusage r; | |
ec94fc3d | 1883 | |
679c9cd4 SK |
1884 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1885 | who != RUSAGE_THREAD) | |
1da177e4 | 1886 | return -EINVAL; |
ce72a16f AV |
1887 | |
1888 | getrusage(current, who, &r); | |
1889 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; | |
1da177e4 LT |
1890 | } |
1891 | ||
8d2d5c4a AV |
1892 | #ifdef CONFIG_COMPAT |
1893 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1894 | { | |
1895 | struct rusage r; | |
1896 | ||
1897 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1898 | who != RUSAGE_THREAD) | |
1899 | return -EINVAL; | |
1900 | ||
ce72a16f | 1901 | getrusage(current, who, &r); |
8d2d5c4a AV |
1902 | return put_compat_rusage(&r, ru); |
1903 | } | |
1904 | #endif | |
1905 | ||
e48fbb69 | 1906 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
1907 | { |
1908 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1909 | return mask; | |
1910 | } | |
3b7391de | 1911 | |
6e399cd1 | 1912 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
b32dfe37 | 1913 | { |
2903ff01 | 1914 | struct fd exe; |
496ad9aa | 1915 | struct inode *inode; |
2903ff01 | 1916 | int err; |
b32dfe37 | 1917 | |
2903ff01 AV |
1918 | exe = fdget(fd); |
1919 | if (!exe.file) | |
b32dfe37 CG |
1920 | return -EBADF; |
1921 | ||
496ad9aa | 1922 | inode = file_inode(exe.file); |
b32dfe37 CG |
1923 | |
1924 | /* | |
1925 | * Because the original mm->exe_file points to executable file, make | |
1926 | * sure that this one is executable as well, to avoid breaking an | |
1927 | * overall picture. | |
1928 | */ | |
1929 | err = -EACCES; | |
90f8572b | 1930 | if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path)) |
b32dfe37 CG |
1931 | goto exit; |
1932 | ||
02f92b38 | 1933 | err = file_permission(exe.file, MAY_EXEC); |
b32dfe37 CG |
1934 | if (err) |
1935 | goto exit; | |
1936 | ||
35d7bdc8 | 1937 | err = replace_mm_exe_file(mm, exe.file); |
b32dfe37 | 1938 | exit: |
2903ff01 | 1939 | fdput(exe); |
b32dfe37 CG |
1940 | return err; |
1941 | } | |
1942 | ||
f606b77f | 1943 | /* |
11bbd8b4 MK |
1944 | * Check arithmetic relations of passed addresses. |
1945 | * | |
f606b77f CG |
1946 | * WARNING: we don't require any capability here so be very careful |
1947 | * in what is allowed for modification from userspace. | |
1948 | */ | |
11bbd8b4 | 1949 | static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map) |
f606b77f CG |
1950 | { |
1951 | unsigned long mmap_max_addr = TASK_SIZE; | |
f606b77f CG |
1952 | int error = -EINVAL, i; |
1953 | ||
1954 | static const unsigned char offsets[] = { | |
1955 | offsetof(struct prctl_mm_map, start_code), | |
1956 | offsetof(struct prctl_mm_map, end_code), | |
1957 | offsetof(struct prctl_mm_map, start_data), | |
1958 | offsetof(struct prctl_mm_map, end_data), | |
1959 | offsetof(struct prctl_mm_map, start_brk), | |
1960 | offsetof(struct prctl_mm_map, brk), | |
1961 | offsetof(struct prctl_mm_map, start_stack), | |
1962 | offsetof(struct prctl_mm_map, arg_start), | |
1963 | offsetof(struct prctl_mm_map, arg_end), | |
1964 | offsetof(struct prctl_mm_map, env_start), | |
1965 | offsetof(struct prctl_mm_map, env_end), | |
1966 | }; | |
1967 | ||
1968 | /* | |
1969 | * Make sure the members are not somewhere outside | |
1970 | * of allowed address space. | |
1971 | */ | |
1972 | for (i = 0; i < ARRAY_SIZE(offsets); i++) { | |
1973 | u64 val = *(u64 *)((char *)prctl_map + offsets[i]); | |
1974 | ||
1975 | if ((unsigned long)val >= mmap_max_addr || | |
1976 | (unsigned long)val < mmap_min_addr) | |
1977 | goto out; | |
1978 | } | |
1979 | ||
1980 | /* | |
1981 | * Make sure the pairs are ordered. | |
1982 | */ | |
1983 | #define __prctl_check_order(__m1, __op, __m2) \ | |
1984 | ((unsigned long)prctl_map->__m1 __op \ | |
1985 | (unsigned long)prctl_map->__m2) ? 0 : -EINVAL | |
1986 | error = __prctl_check_order(start_code, <, end_code); | |
a9e73998 | 1987 | error |= __prctl_check_order(start_data,<=, end_data); |
f606b77f CG |
1988 | error |= __prctl_check_order(start_brk, <=, brk); |
1989 | error |= __prctl_check_order(arg_start, <=, arg_end); | |
1990 | error |= __prctl_check_order(env_start, <=, env_end); | |
1991 | if (error) | |
1992 | goto out; | |
1993 | #undef __prctl_check_order | |
1994 | ||
1995 | error = -EINVAL; | |
1996 | ||
f606b77f CG |
1997 | /* |
1998 | * Neither we should allow to override limits if they set. | |
1999 | */ | |
2000 | if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk, | |
2001 | prctl_map->start_brk, prctl_map->end_data, | |
2002 | prctl_map->start_data)) | |
2003 | goto out; | |
2004 | ||
f606b77f CG |
2005 | error = 0; |
2006 | out: | |
2007 | return error; | |
2008 | } | |
2009 | ||
4a00e9df | 2010 | #ifdef CONFIG_CHECKPOINT_RESTORE |
f606b77f CG |
2011 | static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size) |
2012 | { | |
2013 | struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, }; | |
2014 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
2015 | struct mm_struct *mm = current->mm; | |
2016 | int error; | |
2017 | ||
2018 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
2019 | BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256); | |
2020 | ||
2021 | if (opt == PR_SET_MM_MAP_SIZE) | |
2022 | return put_user((unsigned int)sizeof(prctl_map), | |
2023 | (unsigned int __user *)addr); | |
2024 | ||
2025 | if (data_size != sizeof(prctl_map)) | |
2026 | return -EINVAL; | |
2027 | ||
2028 | if (copy_from_user(&prctl_map, addr, sizeof(prctl_map))) | |
2029 | return -EFAULT; | |
2030 | ||
11bbd8b4 | 2031 | error = validate_prctl_map_addr(&prctl_map); |
f606b77f CG |
2032 | if (error) |
2033 | return error; | |
2034 | ||
2035 | if (prctl_map.auxv_size) { | |
11bbd8b4 MK |
2036 | /* |
2037 | * Someone is trying to cheat the auxv vector. | |
2038 | */ | |
2039 | if (!prctl_map.auxv || | |
2040 | prctl_map.auxv_size > sizeof(mm->saved_auxv)) | |
2041 | return -EINVAL; | |
2042 | ||
f606b77f CG |
2043 | memset(user_auxv, 0, sizeof(user_auxv)); |
2044 | if (copy_from_user(user_auxv, | |
2045 | (const void __user *)prctl_map.auxv, | |
2046 | prctl_map.auxv_size)) | |
2047 | return -EFAULT; | |
2048 | ||
2049 | /* Last entry must be AT_NULL as specification requires */ | |
2050 | user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL; | |
2051 | user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL; | |
2052 | } | |
2053 | ||
ddf1d398 | 2054 | if (prctl_map.exe_fd != (u32)-1) { |
11bbd8b4 | 2055 | /* |
ebd6de68 NV |
2056 | * Check if the current user is checkpoint/restore capable. |
2057 | * At the time of this writing, it checks for CAP_SYS_ADMIN | |
2058 | * or CAP_CHECKPOINT_RESTORE. | |
2059 | * Note that a user with access to ptrace can masquerade an | |
2060 | * arbitrary program as any executable, even setuid ones. | |
2061 | * This may have implications in the tomoyo subsystem. | |
11bbd8b4 | 2062 | */ |
ebd6de68 | 2063 | if (!checkpoint_restore_ns_capable(current_user_ns())) |
227175b2 | 2064 | return -EPERM; |
11bbd8b4 | 2065 | |
6e399cd1 | 2066 | error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd); |
ddf1d398 MG |
2067 | if (error) |
2068 | return error; | |
2069 | } | |
2070 | ||
88aa7cc6 | 2071 | /* |
5afe69c2 | 2072 | * arg_lock protects concurrent updates but we still need mmap_lock for |
88aa7cc6 YS |
2073 | * read to exclude races with sys_brk. |
2074 | */ | |
d8ed45c5 | 2075 | mmap_read_lock(mm); |
f606b77f CG |
2076 | |
2077 | /* | |
2078 | * We don't validate if these members are pointing to | |
2079 | * real present VMAs because application may have correspond | |
2080 | * VMAs already unmapped and kernel uses these members for statistics | |
2081 | * output in procfs mostly, except | |
2082 | * | |
15ec0fcf | 2083 | * - @start_brk/@brk which are used in do_brk_flags but kernel lookups |
5afe69c2 | 2084 | * for VMAs when updating these members so anything wrong written |
f606b77f CG |
2085 | * here cause kernel to swear at userspace program but won't lead |
2086 | * to any problem in kernel itself | |
2087 | */ | |
2088 | ||
88aa7cc6 | 2089 | spin_lock(&mm->arg_lock); |
f606b77f CG |
2090 | mm->start_code = prctl_map.start_code; |
2091 | mm->end_code = prctl_map.end_code; | |
2092 | mm->start_data = prctl_map.start_data; | |
2093 | mm->end_data = prctl_map.end_data; | |
2094 | mm->start_brk = prctl_map.start_brk; | |
2095 | mm->brk = prctl_map.brk; | |
2096 | mm->start_stack = prctl_map.start_stack; | |
2097 | mm->arg_start = prctl_map.arg_start; | |
2098 | mm->arg_end = prctl_map.arg_end; | |
2099 | mm->env_start = prctl_map.env_start; | |
2100 | mm->env_end = prctl_map.env_end; | |
88aa7cc6 | 2101 | spin_unlock(&mm->arg_lock); |
f606b77f CG |
2102 | |
2103 | /* | |
2104 | * Note this update of @saved_auxv is lockless thus | |
2105 | * if someone reads this member in procfs while we're | |
2106 | * updating -- it may get partly updated results. It's | |
2107 | * known and acceptable trade off: we leave it as is to | |
2108 | * not introduce additional locks here making the kernel | |
2109 | * more complex. | |
2110 | */ | |
2111 | if (prctl_map.auxv_size) | |
2112 | memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv)); | |
2113 | ||
d8ed45c5 | 2114 | mmap_read_unlock(mm); |
ddf1d398 | 2115 | return 0; |
f606b77f CG |
2116 | } |
2117 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
2118 | ||
4a00e9df AD |
2119 | static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr, |
2120 | unsigned long len) | |
2121 | { | |
2122 | /* | |
2123 | * This doesn't move the auxiliary vector itself since it's pinned to | |
2124 | * mm_struct, but it permits filling the vector with new values. It's | |
2125 | * up to the caller to provide sane values here, otherwise userspace | |
2126 | * tools which use this vector might be unhappy. | |
2127 | */ | |
c995f12a | 2128 | unsigned long user_auxv[AT_VECTOR_SIZE] = {}; |
4a00e9df AD |
2129 | |
2130 | if (len > sizeof(user_auxv)) | |
2131 | return -EINVAL; | |
2132 | ||
2133 | if (copy_from_user(user_auxv, (const void __user *)addr, len)) | |
2134 | return -EFAULT; | |
2135 | ||
2136 | /* Make sure the last entry is always AT_NULL */ | |
2137 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
2138 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
2139 | ||
2140 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
2141 | ||
2142 | task_lock(current); | |
2143 | memcpy(mm->saved_auxv, user_auxv, len); | |
2144 | task_unlock(current); | |
2145 | ||
2146 | return 0; | |
2147 | } | |
2148 | ||
028ee4be CG |
2149 | static int prctl_set_mm(int opt, unsigned long addr, |
2150 | unsigned long arg4, unsigned long arg5) | |
2151 | { | |
028ee4be | 2152 | struct mm_struct *mm = current->mm; |
11bbd8b4 MK |
2153 | struct prctl_mm_map prctl_map = { |
2154 | .auxv = NULL, | |
2155 | .auxv_size = 0, | |
2156 | .exe_fd = -1, | |
2157 | }; | |
fe8c7f5c CG |
2158 | struct vm_area_struct *vma; |
2159 | int error; | |
028ee4be | 2160 | |
f606b77f CG |
2161 | if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV && |
2162 | opt != PR_SET_MM_MAP && | |
2163 | opt != PR_SET_MM_MAP_SIZE))) | |
028ee4be CG |
2164 | return -EINVAL; |
2165 | ||
f606b77f CG |
2166 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2167 | if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE) | |
2168 | return prctl_set_mm_map(opt, (const void __user *)addr, arg4); | |
2169 | #endif | |
2170 | ||
79f0713d | 2171 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
2172 | return -EPERM; |
2173 | ||
6e399cd1 DB |
2174 | if (opt == PR_SET_MM_EXE_FILE) |
2175 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
b32dfe37 | 2176 | |
4a00e9df AD |
2177 | if (opt == PR_SET_MM_AUXV) |
2178 | return prctl_set_auxv(mm, addr, arg4); | |
2179 | ||
1ad75b9e | 2180 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
2181 | return -EINVAL; |
2182 | ||
fe8c7f5c CG |
2183 | error = -EINVAL; |
2184 | ||
bc81426f | 2185 | /* |
5afe69c2 | 2186 | * arg_lock protects concurrent updates of arg boundaries, we need |
c1e8d7c6 | 2187 | * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr |
bc81426f MK |
2188 | * validation. |
2189 | */ | |
d8ed45c5 | 2190 | mmap_read_lock(mm); |
028ee4be CG |
2191 | vma = find_vma(mm, addr); |
2192 | ||
bc81426f | 2193 | spin_lock(&mm->arg_lock); |
4a00e9df AD |
2194 | prctl_map.start_code = mm->start_code; |
2195 | prctl_map.end_code = mm->end_code; | |
2196 | prctl_map.start_data = mm->start_data; | |
2197 | prctl_map.end_data = mm->end_data; | |
2198 | prctl_map.start_brk = mm->start_brk; | |
2199 | prctl_map.brk = mm->brk; | |
2200 | prctl_map.start_stack = mm->start_stack; | |
2201 | prctl_map.arg_start = mm->arg_start; | |
2202 | prctl_map.arg_end = mm->arg_end; | |
2203 | prctl_map.env_start = mm->env_start; | |
2204 | prctl_map.env_end = mm->env_end; | |
4a00e9df | 2205 | |
028ee4be CG |
2206 | switch (opt) { |
2207 | case PR_SET_MM_START_CODE: | |
4a00e9df | 2208 | prctl_map.start_code = addr; |
fe8c7f5c | 2209 | break; |
028ee4be | 2210 | case PR_SET_MM_END_CODE: |
4a00e9df | 2211 | prctl_map.end_code = addr; |
028ee4be | 2212 | break; |
028ee4be | 2213 | case PR_SET_MM_START_DATA: |
4a00e9df | 2214 | prctl_map.start_data = addr; |
028ee4be | 2215 | break; |
fe8c7f5c | 2216 | case PR_SET_MM_END_DATA: |
4a00e9df AD |
2217 | prctl_map.end_data = addr; |
2218 | break; | |
2219 | case PR_SET_MM_START_STACK: | |
2220 | prctl_map.start_stack = addr; | |
028ee4be | 2221 | break; |
028ee4be | 2222 | case PR_SET_MM_START_BRK: |
4a00e9df | 2223 | prctl_map.start_brk = addr; |
028ee4be | 2224 | break; |
028ee4be | 2225 | case PR_SET_MM_BRK: |
4a00e9df | 2226 | prctl_map.brk = addr; |
028ee4be | 2227 | break; |
4a00e9df AD |
2228 | case PR_SET_MM_ARG_START: |
2229 | prctl_map.arg_start = addr; | |
2230 | break; | |
2231 | case PR_SET_MM_ARG_END: | |
2232 | prctl_map.arg_end = addr; | |
2233 | break; | |
2234 | case PR_SET_MM_ENV_START: | |
2235 | prctl_map.env_start = addr; | |
2236 | break; | |
2237 | case PR_SET_MM_ENV_END: | |
2238 | prctl_map.env_end = addr; | |
2239 | break; | |
2240 | default: | |
2241 | goto out; | |
2242 | } | |
2243 | ||
11bbd8b4 | 2244 | error = validate_prctl_map_addr(&prctl_map); |
4a00e9df AD |
2245 | if (error) |
2246 | goto out; | |
028ee4be | 2247 | |
4a00e9df | 2248 | switch (opt) { |
fe8c7f5c CG |
2249 | /* |
2250 | * If command line arguments and environment | |
2251 | * are placed somewhere else on stack, we can | |
2252 | * set them up here, ARG_START/END to setup | |
5afe69c2 | 2253 | * command line arguments and ENV_START/END |
fe8c7f5c CG |
2254 | * for environment. |
2255 | */ | |
2256 | case PR_SET_MM_START_STACK: | |
2257 | case PR_SET_MM_ARG_START: | |
2258 | case PR_SET_MM_ARG_END: | |
2259 | case PR_SET_MM_ENV_START: | |
2260 | case PR_SET_MM_ENV_END: | |
2261 | if (!vma) { | |
2262 | error = -EFAULT; | |
2263 | goto out; | |
2264 | } | |
028ee4be CG |
2265 | } |
2266 | ||
4a00e9df AD |
2267 | mm->start_code = prctl_map.start_code; |
2268 | mm->end_code = prctl_map.end_code; | |
2269 | mm->start_data = prctl_map.start_data; | |
2270 | mm->end_data = prctl_map.end_data; | |
2271 | mm->start_brk = prctl_map.start_brk; | |
2272 | mm->brk = prctl_map.brk; | |
2273 | mm->start_stack = prctl_map.start_stack; | |
2274 | mm->arg_start = prctl_map.arg_start; | |
2275 | mm->arg_end = prctl_map.arg_end; | |
2276 | mm->env_start = prctl_map.env_start; | |
2277 | mm->env_end = prctl_map.env_end; | |
2278 | ||
028ee4be | 2279 | error = 0; |
028ee4be | 2280 | out: |
bc81426f | 2281 | spin_unlock(&mm->arg_lock); |
d8ed45c5 | 2282 | mmap_read_unlock(mm); |
028ee4be CG |
2283 | return error; |
2284 | } | |
300f786b | 2285 | |
52b36941 | 2286 | #ifdef CONFIG_CHECKPOINT_RESTORE |
986b9eac | 2287 | static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr) |
300f786b CG |
2288 | { |
2289 | return put_user(me->clear_child_tid, tid_addr); | |
2290 | } | |
52b36941 | 2291 | #else |
986b9eac | 2292 | static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr) |
300f786b CG |
2293 | { |
2294 | return -EINVAL; | |
2295 | } | |
028ee4be CG |
2296 | #endif |
2297 | ||
749860ce PT |
2298 | static int propagate_has_child_subreaper(struct task_struct *p, void *data) |
2299 | { | |
2300 | /* | |
5afe69c2 XC |
2301 | * If task has has_child_subreaper - all its descendants |
2302 | * already have these flag too and new descendants will | |
749860ce PT |
2303 | * inherit it on fork, skip them. |
2304 | * | |
2305 | * If we've found child_reaper - skip descendants in | |
2306 | * it's subtree as they will never get out pidns. | |
2307 | */ | |
2308 | if (p->signal->has_child_subreaper || | |
2309 | is_child_reaper(task_pid(p))) | |
2310 | return 0; | |
2311 | ||
2312 | p->signal->has_child_subreaper = 1; | |
2313 | return 1; | |
2314 | } | |
2315 | ||
7bbf1373 | 2316 | int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which) |
b617cfc8 TG |
2317 | { |
2318 | return -EINVAL; | |
2319 | } | |
2320 | ||
7bbf1373 KC |
2321 | int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which, |
2322 | unsigned long ctrl) | |
b617cfc8 TG |
2323 | { |
2324 | return -EINVAL; | |
2325 | } | |
2326 | ||
a37b0715 | 2327 | #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE) |
8d19f1c8 | 2328 | |
9a10064f CC |
2329 | #ifdef CONFIG_ANON_VMA_NAME |
2330 | ||
2331 | #define ANON_VMA_NAME_MAX_LEN 80 | |
2332 | #define ANON_VMA_NAME_INVALID_CHARS "\\`$[]" | |
2333 | ||
2334 | static inline bool is_valid_name_char(char ch) | |
2335 | { | |
2336 | /* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */ | |
2337 | return ch > 0x1f && ch < 0x7f && | |
2338 | !strchr(ANON_VMA_NAME_INVALID_CHARS, ch); | |
2339 | } | |
2340 | ||
2341 | static int prctl_set_vma(unsigned long opt, unsigned long addr, | |
2342 | unsigned long size, unsigned long arg) | |
2343 | { | |
2344 | struct mm_struct *mm = current->mm; | |
2345 | const char __user *uname; | |
5c26f6ac | 2346 | struct anon_vma_name *anon_name = NULL; |
9a10064f CC |
2347 | int error; |
2348 | ||
2349 | switch (opt) { | |
2350 | case PR_SET_VMA_ANON_NAME: | |
2351 | uname = (const char __user *)arg; | |
2352 | if (uname) { | |
5c26f6ac | 2353 | char *name, *pch; |
9a10064f | 2354 | |
5c26f6ac | 2355 | name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN); |
9a10064f CC |
2356 | if (IS_ERR(name)) |
2357 | return PTR_ERR(name); | |
2358 | ||
2359 | for (pch = name; *pch != '\0'; pch++) { | |
2360 | if (!is_valid_name_char(*pch)) { | |
2361 | kfree(name); | |
2362 | return -EINVAL; | |
2363 | } | |
2364 | } | |
5c26f6ac SB |
2365 | /* anon_vma has its own copy */ |
2366 | anon_name = anon_vma_name_alloc(name); | |
2367 | kfree(name); | |
2368 | if (!anon_name) | |
2369 | return -ENOMEM; | |
2370 | ||
9a10064f CC |
2371 | } |
2372 | ||
2373 | mmap_write_lock(mm); | |
5c26f6ac | 2374 | error = madvise_set_anon_name(mm, addr, size, anon_name); |
9a10064f | 2375 | mmap_write_unlock(mm); |
5c26f6ac | 2376 | anon_vma_name_put(anon_name); |
9a10064f CC |
2377 | break; |
2378 | default: | |
2379 | error = -EINVAL; | |
2380 | } | |
2381 | ||
2382 | return error; | |
2383 | } | |
2384 | ||
2385 | #else /* CONFIG_ANON_VMA_NAME */ | |
2386 | static int prctl_set_vma(unsigned long opt, unsigned long start, | |
2387 | unsigned long size, unsigned long arg) | |
2388 | { | |
2389 | return -EINVAL; | |
2390 | } | |
2391 | #endif /* CONFIG_ANON_VMA_NAME */ | |
2392 | ||
24e41bf8 FR |
2393 | static inline unsigned long get_current_mdwe(void) |
2394 | { | |
2395 | unsigned long ret = 0; | |
2396 | ||
2397 | if (test_bit(MMF_HAS_MDWE, ¤t->mm->flags)) | |
2398 | ret |= PR_MDWE_REFUSE_EXEC_GAIN; | |
2399 | if (test_bit(MMF_HAS_MDWE_NO_INHERIT, ¤t->mm->flags)) | |
2400 | ret |= PR_MDWE_NO_INHERIT; | |
2401 | ||
2402 | return ret; | |
2403 | } | |
2404 | ||
b507808e JG |
2405 | static inline int prctl_set_mdwe(unsigned long bits, unsigned long arg3, |
2406 | unsigned long arg4, unsigned long arg5) | |
2407 | { | |
24e41bf8 FR |
2408 | unsigned long current_bits; |
2409 | ||
b507808e JG |
2410 | if (arg3 || arg4 || arg5) |
2411 | return -EINVAL; | |
2412 | ||
24e41bf8 FR |
2413 | if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN | PR_MDWE_NO_INHERIT)) |
2414 | return -EINVAL; | |
2415 | ||
2416 | /* NO_INHERIT only makes sense with REFUSE_EXEC_GAIN */ | |
2417 | if (bits & PR_MDWE_NO_INHERIT && !(bits & PR_MDWE_REFUSE_EXEC_GAIN)) | |
b507808e JG |
2418 | return -EINVAL; |
2419 | ||
d5aad4c2 ZW |
2420 | /* |
2421 | * EOPNOTSUPP might be more appropriate here in principle, but | |
2422 | * existing userspace depends on EINVAL specifically. | |
2423 | */ | |
2424 | if (!arch_memory_deny_write_exec_supported()) | |
79383813 HD |
2425 | return -EINVAL; |
2426 | ||
24e41bf8 FR |
2427 | current_bits = get_current_mdwe(); |
2428 | if (current_bits && current_bits != bits) | |
2429 | return -EPERM; /* Cannot unset the flags */ | |
2430 | ||
2431 | if (bits & PR_MDWE_NO_INHERIT) | |
2432 | set_bit(MMF_HAS_MDWE_NO_INHERIT, ¤t->mm->flags); | |
b507808e JG |
2433 | if (bits & PR_MDWE_REFUSE_EXEC_GAIN) |
2434 | set_bit(MMF_HAS_MDWE, ¤t->mm->flags); | |
b507808e JG |
2435 | |
2436 | return 0; | |
2437 | } | |
2438 | ||
2439 | static inline int prctl_get_mdwe(unsigned long arg2, unsigned long arg3, | |
2440 | unsigned long arg4, unsigned long arg5) | |
2441 | { | |
2442 | if (arg2 || arg3 || arg4 || arg5) | |
2443 | return -EINVAL; | |
24e41bf8 | 2444 | return get_current_mdwe(); |
b507808e JG |
2445 | } |
2446 | ||
ddc65971 JT |
2447 | static int prctl_get_auxv(void __user *addr, unsigned long len) |
2448 | { | |
2449 | struct mm_struct *mm = current->mm; | |
2450 | unsigned long size = min_t(unsigned long, sizeof(mm->saved_auxv), len); | |
2451 | ||
2452 | if (size && copy_to_user(addr, mm->saved_auxv, size)) | |
2453 | return -EFAULT; | |
2454 | return sizeof(mm->saved_auxv); | |
2455 | } | |
2456 | ||
c4ea37c2 HC |
2457 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2458 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2459 | { |
b6dff3ec DH |
2460 | struct task_struct *me = current; |
2461 | unsigned char comm[sizeof(me->comm)]; | |
2462 | long error; | |
1da177e4 | 2463 | |
d84f4f99 DH |
2464 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2465 | if (error != -ENOSYS) | |
1da177e4 LT |
2466 | return error; |
2467 | ||
d84f4f99 | 2468 | error = 0; |
1da177e4 | 2469 | switch (option) { |
f3cbd435 AM |
2470 | case PR_SET_PDEATHSIG: |
2471 | if (!valid_signal(arg2)) { | |
2472 | error = -EINVAL; | |
1da177e4 | 2473 | break; |
f3cbd435 AM |
2474 | } |
2475 | me->pdeath_signal = arg2; | |
2476 | break; | |
2477 | case PR_GET_PDEATHSIG: | |
2478 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2479 | break; | |
2480 | case PR_GET_DUMPABLE: | |
2481 | error = get_dumpable(me->mm); | |
2482 | break; | |
2483 | case PR_SET_DUMPABLE: | |
2484 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2485 | error = -EINVAL; | |
1da177e4 | 2486 | break; |
f3cbd435 AM |
2487 | } |
2488 | set_dumpable(me->mm, arg2); | |
2489 | break; | |
1da177e4 | 2490 | |
f3cbd435 AM |
2491 | case PR_SET_UNALIGN: |
2492 | error = SET_UNALIGN_CTL(me, arg2); | |
2493 | break; | |
2494 | case PR_GET_UNALIGN: | |
2495 | error = GET_UNALIGN_CTL(me, arg2); | |
2496 | break; | |
2497 | case PR_SET_FPEMU: | |
2498 | error = SET_FPEMU_CTL(me, arg2); | |
2499 | break; | |
2500 | case PR_GET_FPEMU: | |
2501 | error = GET_FPEMU_CTL(me, arg2); | |
2502 | break; | |
2503 | case PR_SET_FPEXC: | |
2504 | error = SET_FPEXC_CTL(me, arg2); | |
2505 | break; | |
2506 | case PR_GET_FPEXC: | |
2507 | error = GET_FPEXC_CTL(me, arg2); | |
2508 | break; | |
2509 | case PR_GET_TIMING: | |
2510 | error = PR_TIMING_STATISTICAL; | |
2511 | break; | |
2512 | case PR_SET_TIMING: | |
2513 | if (arg2 != PR_TIMING_STATISTICAL) | |
2514 | error = -EINVAL; | |
2515 | break; | |
2516 | case PR_SET_NAME: | |
2517 | comm[sizeof(me->comm) - 1] = 0; | |
2518 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2519 | sizeof(me->comm) - 1) < 0) | |
2520 | return -EFAULT; | |
2521 | set_task_comm(me, comm); | |
2522 | proc_comm_connector(me); | |
2523 | break; | |
2524 | case PR_GET_NAME: | |
2525 | get_task_comm(comm, me); | |
2526 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2527 | return -EFAULT; | |
2528 | break; | |
2529 | case PR_GET_ENDIAN: | |
2530 | error = GET_ENDIAN(me, arg2); | |
2531 | break; | |
2532 | case PR_SET_ENDIAN: | |
2533 | error = SET_ENDIAN(me, arg2); | |
2534 | break; | |
2535 | case PR_GET_SECCOMP: | |
2536 | error = prctl_get_seccomp(); | |
2537 | break; | |
2538 | case PR_SET_SECCOMP: | |
2539 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2540 | break; | |
2541 | case PR_GET_TSC: | |
2542 | error = GET_TSC_CTL(arg2); | |
2543 | break; | |
2544 | case PR_SET_TSC: | |
2545 | error = SET_TSC_CTL(arg2); | |
2546 | break; | |
2547 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2548 | error = perf_event_task_disable(); | |
2549 | break; | |
2550 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2551 | error = perf_event_task_enable(); | |
2552 | break; | |
2553 | case PR_GET_TIMERSLACK: | |
da8b44d5 JS |
2554 | if (current->timer_slack_ns > ULONG_MAX) |
2555 | error = ULONG_MAX; | |
2556 | else | |
2557 | error = current->timer_slack_ns; | |
f3cbd435 AM |
2558 | break; |
2559 | case PR_SET_TIMERSLACK: | |
2560 | if (arg2 <= 0) | |
2561 | current->timer_slack_ns = | |
6976675d | 2562 | current->default_timer_slack_ns; |
f3cbd435 AM |
2563 | else |
2564 | current->timer_slack_ns = arg2; | |
2565 | break; | |
2566 | case PR_MCE_KILL: | |
2567 | if (arg4 | arg5) | |
2568 | return -EINVAL; | |
2569 | switch (arg2) { | |
2570 | case PR_MCE_KILL_CLEAR: | |
2571 | if (arg3 != 0) | |
4db96cf0 | 2572 | return -EINVAL; |
f3cbd435 | 2573 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2574 | break; |
f3cbd435 AM |
2575 | case PR_MCE_KILL_SET: |
2576 | current->flags |= PF_MCE_PROCESS; | |
2577 | if (arg3 == PR_MCE_KILL_EARLY) | |
2578 | current->flags |= PF_MCE_EARLY; | |
2579 | else if (arg3 == PR_MCE_KILL_LATE) | |
2580 | current->flags &= ~PF_MCE_EARLY; | |
2581 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2582 | current->flags &= | |
2583 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2584 | else |
259e5e6c | 2585 | return -EINVAL; |
259e5e6c | 2586 | break; |
1da177e4 | 2587 | default: |
f3cbd435 AM |
2588 | return -EINVAL; |
2589 | } | |
2590 | break; | |
2591 | case PR_MCE_KILL_GET: | |
2592 | if (arg2 | arg3 | arg4 | arg5) | |
2593 | return -EINVAL; | |
2594 | if (current->flags & PF_MCE_PROCESS) | |
2595 | error = (current->flags & PF_MCE_EARLY) ? | |
2596 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2597 | else | |
2598 | error = PR_MCE_KILL_DEFAULT; | |
2599 | break; | |
2600 | case PR_SET_MM: | |
2601 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2602 | break; | |
2603 | case PR_GET_TID_ADDRESS: | |
986b9eac | 2604 | error = prctl_get_tid_address(me, (int __user * __user *)arg2); |
f3cbd435 AM |
2605 | break; |
2606 | case PR_SET_CHILD_SUBREAPER: | |
2607 | me->signal->is_child_subreaper = !!arg2; | |
749860ce PT |
2608 | if (!arg2) |
2609 | break; | |
2610 | ||
2611 | walk_process_tree(me, propagate_has_child_subreaper, NULL); | |
f3cbd435 AM |
2612 | break; |
2613 | case PR_GET_CHILD_SUBREAPER: | |
2614 | error = put_user(me->signal->is_child_subreaper, | |
2615 | (int __user *)arg2); | |
2616 | break; | |
2617 | case PR_SET_NO_NEW_PRIVS: | |
2618 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2619 | return -EINVAL; | |
2620 | ||
1d4457f9 | 2621 | task_set_no_new_privs(current); |
f3cbd435 AM |
2622 | break; |
2623 | case PR_GET_NO_NEW_PRIVS: | |
2624 | if (arg2 || arg3 || arg4 || arg5) | |
2625 | return -EINVAL; | |
1d4457f9 | 2626 | return task_no_new_privs(current) ? 1 : 0; |
a0715cc2 AT |
2627 | case PR_GET_THP_DISABLE: |
2628 | if (arg2 || arg3 || arg4 || arg5) | |
2629 | return -EINVAL; | |
18600332 | 2630 | error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 AT |
2631 | break; |
2632 | case PR_SET_THP_DISABLE: | |
2633 | if (arg3 || arg4 || arg5) | |
2634 | return -EINVAL; | |
d8ed45c5 | 2635 | if (mmap_write_lock_killable(me->mm)) |
17b0573d | 2636 | return -EINTR; |
a0715cc2 | 2637 | if (arg2) |
18600332 | 2638 | set_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 | 2639 | else |
18600332 | 2640 | clear_bit(MMF_DISABLE_THP, &me->mm->flags); |
d8ed45c5 | 2641 | mmap_write_unlock(me->mm); |
a0715cc2 | 2642 | break; |
fe3d197f | 2643 | case PR_MPX_ENABLE_MANAGEMENT: |
fe3d197f | 2644 | case PR_MPX_DISABLE_MANAGEMENT: |
f240652b DH |
2645 | /* No longer implemented: */ |
2646 | return -EINVAL; | |
9791554b PB |
2647 | case PR_SET_FP_MODE: |
2648 | error = SET_FP_MODE(me, arg2); | |
2649 | break; | |
2650 | case PR_GET_FP_MODE: | |
2651 | error = GET_FP_MODE(me); | |
2652 | break; | |
2d2123bc DM |
2653 | case PR_SVE_SET_VL: |
2654 | error = SVE_SET_VL(arg2); | |
2655 | break; | |
2656 | case PR_SVE_GET_VL: | |
2657 | error = SVE_GET_VL(); | |
2658 | break; | |
9e4ab6c8 MB |
2659 | case PR_SME_SET_VL: |
2660 | error = SME_SET_VL(arg2); | |
2661 | break; | |
2662 | case PR_SME_GET_VL: | |
2663 | error = SME_GET_VL(); | |
2664 | break; | |
b617cfc8 TG |
2665 | case PR_GET_SPECULATION_CTRL: |
2666 | if (arg3 || arg4 || arg5) | |
2667 | return -EINVAL; | |
7bbf1373 | 2668 | error = arch_prctl_spec_ctrl_get(me, arg2); |
b617cfc8 TG |
2669 | break; |
2670 | case PR_SET_SPECULATION_CTRL: | |
2671 | if (arg4 || arg5) | |
2672 | return -EINVAL; | |
7bbf1373 | 2673 | error = arch_prctl_spec_ctrl_set(me, arg2, arg3); |
b617cfc8 | 2674 | break; |
ba830885 KM |
2675 | case PR_PAC_RESET_KEYS: |
2676 | if (arg3 || arg4 || arg5) | |
2677 | return -EINVAL; | |
2678 | error = PAC_RESET_KEYS(me, arg2); | |
2679 | break; | |
20169862 PC |
2680 | case PR_PAC_SET_ENABLED_KEYS: |
2681 | if (arg4 || arg5) | |
2682 | return -EINVAL; | |
2683 | error = PAC_SET_ENABLED_KEYS(me, arg2, arg3); | |
2684 | break; | |
2685 | case PR_PAC_GET_ENABLED_KEYS: | |
2686 | if (arg2 || arg3 || arg4 || arg5) | |
2687 | return -EINVAL; | |
2688 | error = PAC_GET_ENABLED_KEYS(me); | |
2689 | break; | |
63f0c603 | 2690 | case PR_SET_TAGGED_ADDR_CTRL: |
3e91ec89 CM |
2691 | if (arg3 || arg4 || arg5) |
2692 | return -EINVAL; | |
63f0c603 CM |
2693 | error = SET_TAGGED_ADDR_CTRL(arg2); |
2694 | break; | |
2695 | case PR_GET_TAGGED_ADDR_CTRL: | |
3e91ec89 CM |
2696 | if (arg2 || arg3 || arg4 || arg5) |
2697 | return -EINVAL; | |
63f0c603 CM |
2698 | error = GET_TAGGED_ADDR_CTRL(); |
2699 | break; | |
8d19f1c8 MC |
2700 | case PR_SET_IO_FLUSHER: |
2701 | if (!capable(CAP_SYS_RESOURCE)) | |
2702 | return -EPERM; | |
2703 | ||
2704 | if (arg3 || arg4 || arg5) | |
2705 | return -EINVAL; | |
2706 | ||
2707 | if (arg2 == 1) | |
2708 | current->flags |= PR_IO_FLUSHER; | |
2709 | else if (!arg2) | |
2710 | current->flags &= ~PR_IO_FLUSHER; | |
2711 | else | |
2712 | return -EINVAL; | |
2713 | break; | |
2714 | case PR_GET_IO_FLUSHER: | |
2715 | if (!capable(CAP_SYS_RESOURCE)) | |
2716 | return -EPERM; | |
2717 | ||
2718 | if (arg2 || arg3 || arg4 || arg5) | |
2719 | return -EINVAL; | |
2720 | ||
2721 | error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER; | |
2722 | break; | |
1446e1df GKB |
2723 | case PR_SET_SYSCALL_USER_DISPATCH: |
2724 | error = set_syscall_user_dispatch(arg2, arg3, arg4, | |
2725 | (char __user *) arg5); | |
2726 | break; | |
7ac592aa CH |
2727 | #ifdef CONFIG_SCHED_CORE |
2728 | case PR_SCHED_CORE: | |
2729 | error = sched_core_share_pid(arg2, arg3, arg4, arg5); | |
2730 | break; | |
2731 | #endif | |
b507808e JG |
2732 | case PR_SET_MDWE: |
2733 | error = prctl_set_mdwe(arg2, arg3, arg4, arg5); | |
2734 | break; | |
2735 | case PR_GET_MDWE: | |
2736 | error = prctl_get_mdwe(arg2, arg3, arg4, arg5); | |
2737 | break; | |
628d701f BG |
2738 | case PR_PPC_GET_DEXCR: |
2739 | if (arg3 || arg4 || arg5) | |
2740 | return -EINVAL; | |
2741 | error = PPC_GET_DEXCR_ASPECT(me, arg2); | |
2742 | break; | |
2743 | case PR_PPC_SET_DEXCR: | |
2744 | if (arg4 || arg5) | |
2745 | return -EINVAL; | |
2746 | error = PPC_SET_DEXCR_ASPECT(me, arg2, arg3); | |
2747 | break; | |
9a10064f CC |
2748 | case PR_SET_VMA: |
2749 | error = prctl_set_vma(arg2, arg3, arg4, arg5); | |
2750 | break; | |
636e3483 MO |
2751 | case PR_GET_AUXV: |
2752 | if (arg4 || arg5) | |
2753 | return -EINVAL; | |
2754 | error = prctl_get_auxv((void __user *)arg2, arg3); | |
2755 | break; | |
d7597f59 SR |
2756 | #ifdef CONFIG_KSM |
2757 | case PR_SET_MEMORY_MERGE: | |
2758 | if (arg3 || arg4 || arg5) | |
2759 | return -EINVAL; | |
2760 | if (mmap_write_lock_killable(me->mm)) | |
2761 | return -EINTR; | |
2762 | ||
24139c07 | 2763 | if (arg2) |
d7597f59 | 2764 | error = ksm_enable_merge_any(me->mm); |
24139c07 DH |
2765 | else |
2766 | error = ksm_disable_merge_any(me->mm); | |
d7597f59 SR |
2767 | mmap_write_unlock(me->mm); |
2768 | break; | |
2769 | case PR_GET_MEMORY_MERGE: | |
2770 | if (arg2 || arg3 || arg4 || arg5) | |
2771 | return -EINVAL; | |
2772 | ||
2773 | error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags); | |
2774 | break; | |
2775 | #endif | |
1fd96a3e AC |
2776 | case PR_RISCV_V_SET_CONTROL: |
2777 | error = RISCV_V_SET_CONTROL(arg2); | |
2778 | break; | |
2779 | case PR_RISCV_V_GET_CONTROL: | |
2780 | error = RISCV_V_GET_CONTROL(); | |
2781 | break; | |
6b9391b5 CJ |
2782 | case PR_RISCV_SET_ICACHE_FLUSH_CTX: |
2783 | error = RISCV_SET_ICACHE_FLUSH_CTX(arg2, arg3); | |
2784 | break; | |
f3cbd435 AM |
2785 | default: |
2786 | error = -EINVAL; | |
2787 | break; | |
1da177e4 LT |
2788 | } |
2789 | return error; | |
2790 | } | |
3cfc348b | 2791 | |
836f92ad HC |
2792 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2793 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2794 | { |
2795 | int err = 0; | |
2796 | int cpu = raw_smp_processor_id(); | |
ec94fc3d | 2797 | |
3cfc348b AK |
2798 | if (cpup) |
2799 | err |= put_user(cpu, cpup); | |
2800 | if (nodep) | |
2801 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2802 | return err ? -EFAULT : 0; |
2803 | } | |
10a0a8d4 | 2804 | |
4a22f166 SR |
2805 | /** |
2806 | * do_sysinfo - fill in sysinfo struct | |
2807 | * @info: pointer to buffer to fill | |
2808 | */ | |
2809 | static int do_sysinfo(struct sysinfo *info) | |
2810 | { | |
2811 | unsigned long mem_total, sav_total; | |
2812 | unsigned int mem_unit, bitcount; | |
dc1b7b6c | 2813 | struct timespec64 tp; |
4a22f166 SR |
2814 | |
2815 | memset(info, 0, sizeof(struct sysinfo)); | |
2816 | ||
dc1b7b6c | 2817 | ktime_get_boottime_ts64(&tp); |
ecc421e0 | 2818 | timens_add_boottime(&tp); |
4a22f166 SR |
2819 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
2820 | ||
2821 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2822 | ||
2823 | info->procs = nr_threads; | |
2824 | ||
2825 | si_meminfo(info); | |
2826 | si_swapinfo(info); | |
2827 | ||
2828 | /* | |
2829 | * If the sum of all the available memory (i.e. ram + swap) | |
2830 | * is less than can be stored in a 32 bit unsigned long then | |
2831 | * we can be binary compatible with 2.2.x kernels. If not, | |
2832 | * well, in that case 2.2.x was broken anyways... | |
2833 | * | |
2834 | * -Erik Andersen <[email protected]> | |
2835 | */ | |
2836 | ||
2837 | mem_total = info->totalram + info->totalswap; | |
2838 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2839 | goto out; | |
2840 | bitcount = 0; | |
2841 | mem_unit = info->mem_unit; | |
2842 | while (mem_unit > 1) { | |
2843 | bitcount++; | |
2844 | mem_unit >>= 1; | |
2845 | sav_total = mem_total; | |
2846 | mem_total <<= 1; | |
2847 | if (mem_total < sav_total) | |
2848 | goto out; | |
2849 | } | |
2850 | ||
2851 | /* | |
2852 | * If mem_total did not overflow, multiply all memory values by | |
2853 | * info->mem_unit and set it to 1. This leaves things compatible | |
2854 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2855 | * kernels... | |
2856 | */ | |
2857 | ||
2858 | info->mem_unit = 1; | |
2859 | info->totalram <<= bitcount; | |
2860 | info->freeram <<= bitcount; | |
2861 | info->sharedram <<= bitcount; | |
2862 | info->bufferram <<= bitcount; | |
2863 | info->totalswap <<= bitcount; | |
2864 | info->freeswap <<= bitcount; | |
2865 | info->totalhigh <<= bitcount; | |
2866 | info->freehigh <<= bitcount; | |
2867 | ||
2868 | out: | |
2869 | return 0; | |
2870 | } | |
2871 | ||
2872 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2873 | { | |
2874 | struct sysinfo val; | |
2875 | ||
2876 | do_sysinfo(&val); | |
2877 | ||
2878 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2879 | return -EFAULT; | |
2880 | ||
2881 | return 0; | |
2882 | } | |
2883 | ||
2884 | #ifdef CONFIG_COMPAT | |
2885 | struct compat_sysinfo { | |
2886 | s32 uptime; | |
2887 | u32 loads[3]; | |
2888 | u32 totalram; | |
2889 | u32 freeram; | |
2890 | u32 sharedram; | |
2891 | u32 bufferram; | |
2892 | u32 totalswap; | |
2893 | u32 freeswap; | |
2894 | u16 procs; | |
2895 | u16 pad; | |
2896 | u32 totalhigh; | |
2897 | u32 freehigh; | |
2898 | u32 mem_unit; | |
2899 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2900 | }; | |
2901 | ||
2902 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2903 | { | |
2904 | struct sysinfo s; | |
ce5155c4 | 2905 | struct compat_sysinfo s_32; |
4a22f166 SR |
2906 | |
2907 | do_sysinfo(&s); | |
2908 | ||
2909 | /* Check to see if any memory value is too large for 32-bit and scale | |
2910 | * down if needed | |
2911 | */ | |
0baae41e | 2912 | if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) { |
4a22f166 SR |
2913 | int bitcount = 0; |
2914 | ||
2915 | while (s.mem_unit < PAGE_SIZE) { | |
2916 | s.mem_unit <<= 1; | |
2917 | bitcount++; | |
2918 | } | |
2919 | ||
2920 | s.totalram >>= bitcount; | |
2921 | s.freeram >>= bitcount; | |
2922 | s.sharedram >>= bitcount; | |
2923 | s.bufferram >>= bitcount; | |
2924 | s.totalswap >>= bitcount; | |
2925 | s.freeswap >>= bitcount; | |
2926 | s.totalhigh >>= bitcount; | |
2927 | s.freehigh >>= bitcount; | |
2928 | } | |
2929 | ||
ce5155c4 AV |
2930 | memset(&s_32, 0, sizeof(s_32)); |
2931 | s_32.uptime = s.uptime; | |
2932 | s_32.loads[0] = s.loads[0]; | |
2933 | s_32.loads[1] = s.loads[1]; | |
2934 | s_32.loads[2] = s.loads[2]; | |
2935 | s_32.totalram = s.totalram; | |
2936 | s_32.freeram = s.freeram; | |
2937 | s_32.sharedram = s.sharedram; | |
2938 | s_32.bufferram = s.bufferram; | |
2939 | s_32.totalswap = s.totalswap; | |
2940 | s_32.freeswap = s.freeswap; | |
2941 | s_32.procs = s.procs; | |
2942 | s_32.totalhigh = s.totalhigh; | |
2943 | s_32.freehigh = s.freehigh; | |
2944 | s_32.mem_unit = s.mem_unit; | |
2945 | if (copy_to_user(info, &s_32, sizeof(s_32))) | |
4a22f166 | 2946 | return -EFAULT; |
4a22f166 SR |
2947 | return 0; |
2948 | } | |
2949 | #endif /* CONFIG_COMPAT */ |