]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/sys.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
9984de1a | 8 | #include <linux/export.h> |
1da177e4 LT |
9 | #include <linux/mm.h> |
10 | #include <linux/utsname.h> | |
11 | #include <linux/mman.h> | |
1da177e4 LT |
12 | #include <linux/reboot.h> |
13 | #include <linux/prctl.h> | |
1da177e4 LT |
14 | #include <linux/highuid.h> |
15 | #include <linux/fs.h> | |
74da1ff7 | 16 | #include <linux/kmod.h> |
cdd6c482 | 17 | #include <linux/perf_event.h> |
3e88c553 | 18 | #include <linux/resource.h> |
dc009d92 | 19 | #include <linux/kernel.h> |
1da177e4 | 20 | #include <linux/workqueue.h> |
c59ede7b | 21 | #include <linux/capability.h> |
1da177e4 LT |
22 | #include <linux/device.h> |
23 | #include <linux/key.h> | |
24 | #include <linux/times.h> | |
25 | #include <linux/posix-timers.h> | |
26 | #include <linux/security.h> | |
1da177e4 LT |
27 | #include <linux/suspend.h> |
28 | #include <linux/tty.h> | |
7ed20e1a | 29 | #include <linux/signal.h> |
9f46080c | 30 | #include <linux/cn_proc.h> |
3cfc348b | 31 | #include <linux/getcpu.h> |
6eaeeaba | 32 | #include <linux/task_io_accounting_ops.h> |
1d9d02fe | 33 | #include <linux/seccomp.h> |
4047727e | 34 | #include <linux/cpu.h> |
e28cbf22 | 35 | #include <linux/personality.h> |
e3d5a27d | 36 | #include <linux/ptrace.h> |
5ad4e53b | 37 | #include <linux/fs_struct.h> |
b32dfe37 CG |
38 | #include <linux/file.h> |
39 | #include <linux/mount.h> | |
5a0e3ad6 | 40 | #include <linux/gfp.h> |
40dc166c | 41 | #include <linux/syscore_ops.h> |
be27425d AK |
42 | #include <linux/version.h> |
43 | #include <linux/ctype.h> | |
1446e1df | 44 | #include <linux/syscall_user_dispatch.h> |
1da177e4 LT |
45 | |
46 | #include <linux/compat.h> | |
47 | #include <linux/syscalls.h> | |
00d7c05a | 48 | #include <linux/kprobes.h> |
acce292c | 49 | #include <linux/user_namespace.h> |
ecc421e0 | 50 | #include <linux/time_namespace.h> |
7fe5e042 | 51 | #include <linux/binfmts.h> |
1da177e4 | 52 | |
4a22f166 | 53 | #include <linux/sched.h> |
4eb5aaa3 | 54 | #include <linux/sched/autogroup.h> |
4f17722c | 55 | #include <linux/sched/loadavg.h> |
03441a34 | 56 | #include <linux/sched/stat.h> |
6e84f315 | 57 | #include <linux/sched/mm.h> |
f7ccbae4 | 58 | #include <linux/sched/coredump.h> |
29930025 | 59 | #include <linux/sched/task.h> |
32ef5517 | 60 | #include <linux/sched/cputime.h> |
4a22f166 SR |
61 | #include <linux/rcupdate.h> |
62 | #include <linux/uidgid.h> | |
63 | #include <linux/cred.h> | |
64 | ||
b617cfc8 TG |
65 | #include <linux/nospec.h> |
66 | ||
04c6862c | 67 | #include <linux/kmsg_dump.h> |
be27425d AK |
68 | /* Move somewhere else to avoid recompiling? */ |
69 | #include <generated/utsrelease.h> | |
04c6862c | 70 | |
7c0f6ba6 | 71 | #include <linux/uaccess.h> |
1da177e4 LT |
72 | #include <asm/io.h> |
73 | #include <asm/unistd.h> | |
74 | ||
e530dca5 DB |
75 | #include "uid16.h" |
76 | ||
1da177e4 | 77 | #ifndef SET_UNALIGN_CTL |
ec94fc3d | 78 | # define SET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
79 | #endif |
80 | #ifndef GET_UNALIGN_CTL | |
ec94fc3d | 81 | # define GET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
82 | #endif |
83 | #ifndef SET_FPEMU_CTL | |
ec94fc3d | 84 | # define SET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
85 | #endif |
86 | #ifndef GET_FPEMU_CTL | |
ec94fc3d | 87 | # define GET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
88 | #endif |
89 | #ifndef SET_FPEXC_CTL | |
ec94fc3d | 90 | # define SET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 LT |
91 | #endif |
92 | #ifndef GET_FPEXC_CTL | |
ec94fc3d | 93 | # define GET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 | 94 | #endif |
651d765d | 95 | #ifndef GET_ENDIAN |
ec94fc3d | 96 | # define GET_ENDIAN(a, b) (-EINVAL) |
651d765d AB |
97 | #endif |
98 | #ifndef SET_ENDIAN | |
ec94fc3d | 99 | # define SET_ENDIAN(a, b) (-EINVAL) |
651d765d | 100 | #endif |
8fb402bc EB |
101 | #ifndef GET_TSC_CTL |
102 | # define GET_TSC_CTL(a) (-EINVAL) | |
103 | #endif | |
104 | #ifndef SET_TSC_CTL | |
105 | # define SET_TSC_CTL(a) (-EINVAL) | |
106 | #endif | |
9791554b PB |
107 | #ifndef GET_FP_MODE |
108 | # define GET_FP_MODE(a) (-EINVAL) | |
109 | #endif | |
110 | #ifndef SET_FP_MODE | |
111 | # define SET_FP_MODE(a,b) (-EINVAL) | |
112 | #endif | |
2d2123bc DM |
113 | #ifndef SVE_SET_VL |
114 | # define SVE_SET_VL(a) (-EINVAL) | |
115 | #endif | |
116 | #ifndef SVE_GET_VL | |
117 | # define SVE_GET_VL() (-EINVAL) | |
118 | #endif | |
ba830885 KM |
119 | #ifndef PAC_RESET_KEYS |
120 | # define PAC_RESET_KEYS(a, b) (-EINVAL) | |
121 | #endif | |
20169862 PC |
122 | #ifndef PAC_SET_ENABLED_KEYS |
123 | # define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL) | |
124 | #endif | |
125 | #ifndef PAC_GET_ENABLED_KEYS | |
126 | # define PAC_GET_ENABLED_KEYS(a) (-EINVAL) | |
127 | #endif | |
63f0c603 CM |
128 | #ifndef SET_TAGGED_ADDR_CTRL |
129 | # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL) | |
130 | #endif | |
131 | #ifndef GET_TAGGED_ADDR_CTRL | |
132 | # define GET_TAGGED_ADDR_CTRL() (-EINVAL) | |
133 | #endif | |
1da177e4 LT |
134 | |
135 | /* | |
136 | * this is where the system-wide overflow UID and GID are defined, for | |
137 | * architectures that now have 32-bit UID/GID but didn't in the past | |
138 | */ | |
139 | ||
140 | int overflowuid = DEFAULT_OVERFLOWUID; | |
141 | int overflowgid = DEFAULT_OVERFLOWGID; | |
142 | ||
1da177e4 LT |
143 | EXPORT_SYMBOL(overflowuid); |
144 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
145 | |
146 | /* | |
147 | * the same as above, but for filesystems which can only store a 16-bit | |
148 | * UID and GID. as such, this is needed on all architectures | |
149 | */ | |
150 | ||
151 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
8b2770a4 | 152 | int fs_overflowgid = DEFAULT_FS_OVERFLOWGID; |
1da177e4 LT |
153 | |
154 | EXPORT_SYMBOL(fs_overflowuid); | |
155 | EXPORT_SYMBOL(fs_overflowgid); | |
156 | ||
fc832ad3 SH |
157 | /* |
158 | * Returns true if current's euid is same as p's uid or euid, | |
159 | * or has CAP_SYS_NICE to p's user_ns. | |
160 | * | |
161 | * Called with rcu_read_lock, creds are safe | |
162 | */ | |
163 | static bool set_one_prio_perm(struct task_struct *p) | |
164 | { | |
165 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
166 | ||
5af66203 EB |
167 | if (uid_eq(pcred->uid, cred->euid) || |
168 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 169 | return true; |
c4a4d603 | 170 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
171 | return true; |
172 | return false; | |
173 | } | |
174 | ||
c69e8d9c DH |
175 | /* |
176 | * set the priority of a task | |
177 | * - the caller must hold the RCU read lock | |
178 | */ | |
1da177e4 LT |
179 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
180 | { | |
181 | int no_nice; | |
182 | ||
fc832ad3 | 183 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
184 | error = -EPERM; |
185 | goto out; | |
186 | } | |
e43379f1 | 187 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
188 | error = -EACCES; |
189 | goto out; | |
190 | } | |
191 | no_nice = security_task_setnice(p, niceval); | |
192 | if (no_nice) { | |
193 | error = no_nice; | |
194 | goto out; | |
195 | } | |
196 | if (error == -ESRCH) | |
197 | error = 0; | |
198 | set_user_nice(p, niceval); | |
199 | out: | |
200 | return error; | |
201 | } | |
202 | ||
754fe8d2 | 203 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
204 | { |
205 | struct task_struct *g, *p; | |
206 | struct user_struct *user; | |
86a264ab | 207 | const struct cred *cred = current_cred(); |
1da177e4 | 208 | int error = -EINVAL; |
41487c65 | 209 | struct pid *pgrp; |
7b44ab97 | 210 | kuid_t uid; |
1da177e4 | 211 | |
3e88c553 | 212 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
213 | goto out; |
214 | ||
215 | /* normalize: avoid signed division (rounding problems) */ | |
216 | error = -ESRCH; | |
c4a4d2f4 DY |
217 | if (niceval < MIN_NICE) |
218 | niceval = MIN_NICE; | |
219 | if (niceval > MAX_NICE) | |
220 | niceval = MAX_NICE; | |
1da177e4 | 221 | |
d4581a23 | 222 | rcu_read_lock(); |
1da177e4 LT |
223 | read_lock(&tasklist_lock); |
224 | switch (which) { | |
ec94fc3d | 225 | case PRIO_PROCESS: |
226 | if (who) | |
227 | p = find_task_by_vpid(who); | |
228 | else | |
229 | p = current; | |
230 | if (p) | |
231 | error = set_one_prio(p, niceval, error); | |
232 | break; | |
233 | case PRIO_PGRP: | |
234 | if (who) | |
235 | pgrp = find_vpid(who); | |
236 | else | |
237 | pgrp = task_pgrp(current); | |
238 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
239 | error = set_one_prio(p, niceval, error); | |
240 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
241 | break; | |
242 | case PRIO_USER: | |
243 | uid = make_kuid(cred->user_ns, who); | |
244 | user = cred->user; | |
245 | if (!who) | |
246 | uid = cred->uid; | |
247 | else if (!uid_eq(uid, cred->uid)) { | |
248 | user = find_user(uid); | |
249 | if (!user) | |
86a264ab | 250 | goto out_unlock; /* No processes for this user */ |
ec94fc3d | 251 | } |
252 | do_each_thread(g, p) { | |
8639b461 | 253 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) |
ec94fc3d | 254 | error = set_one_prio(p, niceval, error); |
255 | } while_each_thread(g, p); | |
256 | if (!uid_eq(uid, cred->uid)) | |
257 | free_uid(user); /* For find_user() */ | |
258 | break; | |
1da177e4 LT |
259 | } |
260 | out_unlock: | |
261 | read_unlock(&tasklist_lock); | |
d4581a23 | 262 | rcu_read_unlock(); |
1da177e4 LT |
263 | out: |
264 | return error; | |
265 | } | |
266 | ||
267 | /* | |
268 | * Ugh. To avoid negative return values, "getpriority()" will | |
269 | * not return the normal nice-value, but a negated value that | |
270 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
271 | * to stay compatible. | |
272 | */ | |
754fe8d2 | 273 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
274 | { |
275 | struct task_struct *g, *p; | |
276 | struct user_struct *user; | |
86a264ab | 277 | const struct cred *cred = current_cred(); |
1da177e4 | 278 | long niceval, retval = -ESRCH; |
41487c65 | 279 | struct pid *pgrp; |
7b44ab97 | 280 | kuid_t uid; |
1da177e4 | 281 | |
3e88c553 | 282 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
283 | return -EINVAL; |
284 | ||
70118837 | 285 | rcu_read_lock(); |
1da177e4 LT |
286 | read_lock(&tasklist_lock); |
287 | switch (which) { | |
ec94fc3d | 288 | case PRIO_PROCESS: |
289 | if (who) | |
290 | p = find_task_by_vpid(who); | |
291 | else | |
292 | p = current; | |
293 | if (p) { | |
294 | niceval = nice_to_rlimit(task_nice(p)); | |
295 | if (niceval > retval) | |
296 | retval = niceval; | |
297 | } | |
298 | break; | |
299 | case PRIO_PGRP: | |
300 | if (who) | |
301 | pgrp = find_vpid(who); | |
302 | else | |
303 | pgrp = task_pgrp(current); | |
304 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
305 | niceval = nice_to_rlimit(task_nice(p)); | |
306 | if (niceval > retval) | |
307 | retval = niceval; | |
308 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
309 | break; | |
310 | case PRIO_USER: | |
311 | uid = make_kuid(cred->user_ns, who); | |
312 | user = cred->user; | |
313 | if (!who) | |
314 | uid = cred->uid; | |
315 | else if (!uid_eq(uid, cred->uid)) { | |
316 | user = find_user(uid); | |
317 | if (!user) | |
318 | goto out_unlock; /* No processes for this user */ | |
319 | } | |
320 | do_each_thread(g, p) { | |
8639b461 | 321 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) { |
7aa2c016 | 322 | niceval = nice_to_rlimit(task_nice(p)); |
1da177e4 LT |
323 | if (niceval > retval) |
324 | retval = niceval; | |
325 | } | |
ec94fc3d | 326 | } while_each_thread(g, p); |
327 | if (!uid_eq(uid, cred->uid)) | |
328 | free_uid(user); /* for find_user() */ | |
329 | break; | |
1da177e4 LT |
330 | } |
331 | out_unlock: | |
332 | read_unlock(&tasklist_lock); | |
70118837 | 333 | rcu_read_unlock(); |
1da177e4 LT |
334 | |
335 | return retval; | |
336 | } | |
337 | ||
1da177e4 LT |
338 | /* |
339 | * Unprivileged users may change the real gid to the effective gid | |
340 | * or vice versa. (BSD-style) | |
341 | * | |
342 | * If you set the real gid at all, or set the effective gid to a value not | |
343 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
344 | * | |
345 | * This makes it possible for a setgid program to completely drop its | |
346 | * privileges, which is often a useful assertion to make when you are doing | |
347 | * a security audit over a program. | |
348 | * | |
349 | * The general idea is that a program which uses just setregid() will be | |
350 | * 100% compatible with BSD. A program which uses just setgid() will be | |
ec94fc3d | 351 | * 100% compatible with POSIX with saved IDs. |
1da177e4 LT |
352 | * |
353 | * SMP: There are not races, the GIDs are checked only by filesystem | |
354 | * operations (as far as semantic preservation is concerned). | |
355 | */ | |
2813893f | 356 | #ifdef CONFIG_MULTIUSER |
e530dca5 | 357 | long __sys_setregid(gid_t rgid, gid_t egid) |
1da177e4 | 358 | { |
a29c33f4 | 359 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
360 | const struct cred *old; |
361 | struct cred *new; | |
1da177e4 | 362 | int retval; |
a29c33f4 EB |
363 | kgid_t krgid, kegid; |
364 | ||
365 | krgid = make_kgid(ns, rgid); | |
366 | kegid = make_kgid(ns, egid); | |
367 | ||
368 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
369 | return -EINVAL; | |
370 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
371 | return -EINVAL; | |
1da177e4 | 372 | |
d84f4f99 DH |
373 | new = prepare_creds(); |
374 | if (!new) | |
375 | return -ENOMEM; | |
376 | old = current_cred(); | |
377 | ||
d84f4f99 | 378 | retval = -EPERM; |
1da177e4 | 379 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
380 | if (gid_eq(old->gid, krgid) || |
381 | gid_eq(old->egid, krgid) || | |
111767c1 | 382 | ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 | 383 | new->gid = krgid; |
1da177e4 | 384 | else |
d84f4f99 | 385 | goto error; |
1da177e4 LT |
386 | } |
387 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
388 | if (gid_eq(old->gid, kegid) || |
389 | gid_eq(old->egid, kegid) || | |
390 | gid_eq(old->sgid, kegid) || | |
111767c1 | 391 | ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 | 392 | new->egid = kegid; |
756184b7 | 393 | else |
d84f4f99 | 394 | goto error; |
1da177e4 | 395 | } |
d84f4f99 | 396 | |
1da177e4 | 397 | if (rgid != (gid_t) -1 || |
a29c33f4 | 398 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
399 | new->sgid = new->egid; |
400 | new->fsgid = new->egid; | |
401 | ||
39030e13 TC |
402 | retval = security_task_fix_setgid(new, old, LSM_SETID_RE); |
403 | if (retval < 0) | |
404 | goto error; | |
405 | ||
d84f4f99 DH |
406 | return commit_creds(new); |
407 | ||
408 | error: | |
409 | abort_creds(new); | |
410 | return retval; | |
1da177e4 LT |
411 | } |
412 | ||
e530dca5 DB |
413 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
414 | { | |
415 | return __sys_setregid(rgid, egid); | |
416 | } | |
417 | ||
1da177e4 | 418 | /* |
ec94fc3d | 419 | * setgid() is implemented like SysV w/ SAVED_IDS |
1da177e4 LT |
420 | * |
421 | * SMP: Same implicit races as above. | |
422 | */ | |
e530dca5 | 423 | long __sys_setgid(gid_t gid) |
1da177e4 | 424 | { |
a29c33f4 | 425 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
426 | const struct cred *old; |
427 | struct cred *new; | |
1da177e4 | 428 | int retval; |
a29c33f4 EB |
429 | kgid_t kgid; |
430 | ||
431 | kgid = make_kgid(ns, gid); | |
432 | if (!gid_valid(kgid)) | |
433 | return -EINVAL; | |
1da177e4 | 434 | |
d84f4f99 DH |
435 | new = prepare_creds(); |
436 | if (!new) | |
437 | return -ENOMEM; | |
438 | old = current_cred(); | |
439 | ||
d84f4f99 | 440 | retval = -EPERM; |
111767c1 | 441 | if (ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 EB |
442 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
443 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
444 | new->egid = new->fsgid = kgid; | |
1da177e4 | 445 | else |
d84f4f99 | 446 | goto error; |
1da177e4 | 447 | |
39030e13 TC |
448 | retval = security_task_fix_setgid(new, old, LSM_SETID_ID); |
449 | if (retval < 0) | |
450 | goto error; | |
451 | ||
d84f4f99 DH |
452 | return commit_creds(new); |
453 | ||
454 | error: | |
455 | abort_creds(new); | |
456 | return retval; | |
1da177e4 | 457 | } |
54e99124 | 458 | |
e530dca5 DB |
459 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
460 | { | |
461 | return __sys_setgid(gid); | |
462 | } | |
463 | ||
d84f4f99 DH |
464 | /* |
465 | * change the user struct in a credentials set to match the new UID | |
466 | */ | |
467 | static int set_user(struct cred *new) | |
1da177e4 LT |
468 | { |
469 | struct user_struct *new_user; | |
470 | ||
078de5f7 | 471 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
472 | if (!new_user) |
473 | return -EAGAIN; | |
474 | ||
a38965fc EB |
475 | free_uid(new->user); |
476 | new->user = new_user; | |
477 | return 0; | |
478 | } | |
479 | ||
480 | static void flag_nproc_exceeded(struct cred *new) | |
481 | { | |
482 | if (new->ucounts == current_ucounts()) | |
483 | return; | |
484 | ||
72fa5997 VK |
485 | /* |
486 | * We don't fail in case of NPROC limit excess here because too many | |
487 | * poorly written programs don't check set*uid() return code, assuming | |
488 | * it never fails if called by root. We may still enforce NPROC limit | |
489 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
490 | * failure to the execve() stage. | |
491 | */ | |
21d1c5e3 | 492 | if (is_ucounts_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) && |
a38965fc | 493 | new->user != INIT_USER) |
72fa5997 VK |
494 | current->flags |= PF_NPROC_EXCEEDED; |
495 | else | |
496 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 LT |
497 | } |
498 | ||
499 | /* | |
500 | * Unprivileged users may change the real uid to the effective uid | |
501 | * or vice versa. (BSD-style) | |
502 | * | |
503 | * If you set the real uid at all, or set the effective uid to a value not | |
504 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
505 | * | |
506 | * This makes it possible for a setuid program to completely drop its | |
507 | * privileges, which is often a useful assertion to make when you are doing | |
508 | * a security audit over a program. | |
509 | * | |
510 | * The general idea is that a program which uses just setreuid() will be | |
511 | * 100% compatible with BSD. A program which uses just setuid() will be | |
ec94fc3d | 512 | * 100% compatible with POSIX with saved IDs. |
1da177e4 | 513 | */ |
e530dca5 | 514 | long __sys_setreuid(uid_t ruid, uid_t euid) |
1da177e4 | 515 | { |
a29c33f4 | 516 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
517 | const struct cred *old; |
518 | struct cred *new; | |
1da177e4 | 519 | int retval; |
a29c33f4 EB |
520 | kuid_t kruid, keuid; |
521 | ||
522 | kruid = make_kuid(ns, ruid); | |
523 | keuid = make_kuid(ns, euid); | |
524 | ||
525 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
526 | return -EINVAL; | |
527 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
528 | return -EINVAL; | |
1da177e4 | 529 | |
d84f4f99 DH |
530 | new = prepare_creds(); |
531 | if (!new) | |
532 | return -ENOMEM; | |
533 | old = current_cred(); | |
534 | ||
d84f4f99 | 535 | retval = -EPERM; |
1da177e4 | 536 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
537 | new->uid = kruid; |
538 | if (!uid_eq(old->uid, kruid) && | |
539 | !uid_eq(old->euid, kruid) && | |
40852275 | 540 | !ns_capable_setid(old->user_ns, CAP_SETUID)) |
d84f4f99 | 541 | goto error; |
1da177e4 LT |
542 | } |
543 | ||
544 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
545 | new->euid = keuid; |
546 | if (!uid_eq(old->uid, keuid) && | |
547 | !uid_eq(old->euid, keuid) && | |
548 | !uid_eq(old->suid, keuid) && | |
40852275 | 549 | !ns_capable_setid(old->user_ns, CAP_SETUID)) |
d84f4f99 | 550 | goto error; |
1da177e4 LT |
551 | } |
552 | ||
a29c33f4 | 553 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
554 | retval = set_user(new); |
555 | if (retval < 0) | |
556 | goto error; | |
557 | } | |
1da177e4 | 558 | if (ruid != (uid_t) -1 || |
a29c33f4 | 559 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
560 | new->suid = new->euid; |
561 | new->fsuid = new->euid; | |
1da177e4 | 562 | |
d84f4f99 DH |
563 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
564 | if (retval < 0) | |
565 | goto error; | |
1da177e4 | 566 | |
905ae01c AG |
567 | retval = set_cred_ucounts(new); |
568 | if (retval < 0) | |
569 | goto error; | |
570 | ||
a38965fc | 571 | flag_nproc_exceeded(new); |
d84f4f99 | 572 | return commit_creds(new); |
1da177e4 | 573 | |
d84f4f99 DH |
574 | error: |
575 | abort_creds(new); | |
576 | return retval; | |
577 | } | |
ec94fc3d | 578 | |
e530dca5 DB |
579 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
580 | { | |
581 | return __sys_setreuid(ruid, euid); | |
582 | } | |
583 | ||
1da177e4 | 584 | /* |
ec94fc3d | 585 | * setuid() is implemented like SysV with SAVED_IDS |
586 | * | |
1da177e4 | 587 | * Note that SAVED_ID's is deficient in that a setuid root program |
ec94fc3d | 588 | * like sendmail, for example, cannot set its uid to be a normal |
1da177e4 LT |
589 | * user and then switch back, because if you're root, setuid() sets |
590 | * the saved uid too. If you don't like this, blame the bright people | |
591 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
592 | * will allow a root program to temporarily drop privileges and be able to | |
ec94fc3d | 593 | * regain them by swapping the real and effective uid. |
1da177e4 | 594 | */ |
e530dca5 | 595 | long __sys_setuid(uid_t uid) |
1da177e4 | 596 | { |
a29c33f4 | 597 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
598 | const struct cred *old; |
599 | struct cred *new; | |
1da177e4 | 600 | int retval; |
a29c33f4 EB |
601 | kuid_t kuid; |
602 | ||
603 | kuid = make_kuid(ns, uid); | |
604 | if (!uid_valid(kuid)) | |
605 | return -EINVAL; | |
1da177e4 | 606 | |
d84f4f99 DH |
607 | new = prepare_creds(); |
608 | if (!new) | |
609 | return -ENOMEM; | |
610 | old = current_cred(); | |
611 | ||
d84f4f99 | 612 | retval = -EPERM; |
40852275 | 613 | if (ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
614 | new->suid = new->uid = kuid; |
615 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
616 | retval = set_user(new); |
617 | if (retval < 0) | |
618 | goto error; | |
d84f4f99 | 619 | } |
a29c33f4 | 620 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 621 | goto error; |
1da177e4 | 622 | } |
1da177e4 | 623 | |
a29c33f4 | 624 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
625 | |
626 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
627 | if (retval < 0) | |
628 | goto error; | |
1da177e4 | 629 | |
905ae01c AG |
630 | retval = set_cred_ucounts(new); |
631 | if (retval < 0) | |
632 | goto error; | |
633 | ||
a38965fc | 634 | flag_nproc_exceeded(new); |
d84f4f99 | 635 | return commit_creds(new); |
1da177e4 | 636 | |
d84f4f99 DH |
637 | error: |
638 | abort_creds(new); | |
639 | return retval; | |
1da177e4 LT |
640 | } |
641 | ||
e530dca5 DB |
642 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
643 | { | |
644 | return __sys_setuid(uid); | |
645 | } | |
646 | ||
1da177e4 LT |
647 | |
648 | /* | |
649 | * This function implements a generic ability to update ruid, euid, | |
650 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
651 | */ | |
e530dca5 | 652 | long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) |
1da177e4 | 653 | { |
a29c33f4 | 654 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
655 | const struct cred *old; |
656 | struct cred *new; | |
1da177e4 | 657 | int retval; |
a29c33f4 EB |
658 | kuid_t kruid, keuid, ksuid; |
659 | ||
660 | kruid = make_kuid(ns, ruid); | |
661 | keuid = make_kuid(ns, euid); | |
662 | ksuid = make_kuid(ns, suid); | |
663 | ||
664 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
665 | return -EINVAL; | |
666 | ||
667 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
668 | return -EINVAL; | |
669 | ||
670 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
671 | return -EINVAL; | |
1da177e4 | 672 | |
d84f4f99 DH |
673 | new = prepare_creds(); |
674 | if (!new) | |
675 | return -ENOMEM; | |
676 | ||
d84f4f99 | 677 | old = current_cred(); |
1da177e4 | 678 | |
d84f4f99 | 679 | retval = -EPERM; |
40852275 | 680 | if (!ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
681 | if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && |
682 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid)) | |
d84f4f99 | 683 | goto error; |
a29c33f4 EB |
684 | if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && |
685 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid)) | |
d84f4f99 | 686 | goto error; |
a29c33f4 EB |
687 | if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && |
688 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid)) | |
d84f4f99 | 689 | goto error; |
1da177e4 | 690 | } |
d84f4f99 | 691 | |
1da177e4 | 692 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
693 | new->uid = kruid; |
694 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
695 | retval = set_user(new); |
696 | if (retval < 0) | |
697 | goto error; | |
698 | } | |
1da177e4 | 699 | } |
d84f4f99 | 700 | if (euid != (uid_t) -1) |
a29c33f4 | 701 | new->euid = keuid; |
1da177e4 | 702 | if (suid != (uid_t) -1) |
a29c33f4 | 703 | new->suid = ksuid; |
d84f4f99 | 704 | new->fsuid = new->euid; |
1da177e4 | 705 | |
d84f4f99 DH |
706 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
707 | if (retval < 0) | |
708 | goto error; | |
1da177e4 | 709 | |
905ae01c AG |
710 | retval = set_cred_ucounts(new); |
711 | if (retval < 0) | |
712 | goto error; | |
713 | ||
a38965fc | 714 | flag_nproc_exceeded(new); |
d84f4f99 | 715 | return commit_creds(new); |
1da177e4 | 716 | |
d84f4f99 DH |
717 | error: |
718 | abort_creds(new); | |
719 | return retval; | |
1da177e4 LT |
720 | } |
721 | ||
e530dca5 DB |
722 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
723 | { | |
724 | return __sys_setresuid(ruid, euid, suid); | |
725 | } | |
726 | ||
a29c33f4 | 727 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 728 | { |
86a264ab | 729 | const struct cred *cred = current_cred(); |
1da177e4 | 730 | int retval; |
a29c33f4 EB |
731 | uid_t ruid, euid, suid; |
732 | ||
733 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
734 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
735 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 736 | |
ec94fc3d | 737 | retval = put_user(ruid, ruidp); |
738 | if (!retval) { | |
739 | retval = put_user(euid, euidp); | |
740 | if (!retval) | |
741 | return put_user(suid, suidp); | |
742 | } | |
1da177e4 LT |
743 | return retval; |
744 | } | |
745 | ||
746 | /* | |
747 | * Same as above, but for rgid, egid, sgid. | |
748 | */ | |
e530dca5 | 749 | long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) |
1da177e4 | 750 | { |
a29c33f4 | 751 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
752 | const struct cred *old; |
753 | struct cred *new; | |
1da177e4 | 754 | int retval; |
a29c33f4 EB |
755 | kgid_t krgid, kegid, ksgid; |
756 | ||
757 | krgid = make_kgid(ns, rgid); | |
758 | kegid = make_kgid(ns, egid); | |
759 | ksgid = make_kgid(ns, sgid); | |
760 | ||
761 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
762 | return -EINVAL; | |
763 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
764 | return -EINVAL; | |
765 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
766 | return -EINVAL; | |
1da177e4 | 767 | |
d84f4f99 DH |
768 | new = prepare_creds(); |
769 | if (!new) | |
770 | return -ENOMEM; | |
771 | old = current_cred(); | |
772 | ||
d84f4f99 | 773 | retval = -EPERM; |
111767c1 | 774 | if (!ns_capable_setid(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
775 | if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && |
776 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid)) | |
d84f4f99 | 777 | goto error; |
a29c33f4 EB |
778 | if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && |
779 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid)) | |
d84f4f99 | 780 | goto error; |
a29c33f4 EB |
781 | if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && |
782 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid)) | |
d84f4f99 | 783 | goto error; |
1da177e4 | 784 | } |
d84f4f99 | 785 | |
1da177e4 | 786 | if (rgid != (gid_t) -1) |
a29c33f4 | 787 | new->gid = krgid; |
d84f4f99 | 788 | if (egid != (gid_t) -1) |
a29c33f4 | 789 | new->egid = kegid; |
1da177e4 | 790 | if (sgid != (gid_t) -1) |
a29c33f4 | 791 | new->sgid = ksgid; |
d84f4f99 | 792 | new->fsgid = new->egid; |
1da177e4 | 793 | |
39030e13 TC |
794 | retval = security_task_fix_setgid(new, old, LSM_SETID_RES); |
795 | if (retval < 0) | |
796 | goto error; | |
797 | ||
d84f4f99 DH |
798 | return commit_creds(new); |
799 | ||
800 | error: | |
801 | abort_creds(new); | |
802 | return retval; | |
1da177e4 LT |
803 | } |
804 | ||
e530dca5 DB |
805 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
806 | { | |
807 | return __sys_setresgid(rgid, egid, sgid); | |
808 | } | |
809 | ||
a29c33f4 | 810 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 811 | { |
86a264ab | 812 | const struct cred *cred = current_cred(); |
1da177e4 | 813 | int retval; |
a29c33f4 EB |
814 | gid_t rgid, egid, sgid; |
815 | ||
816 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
817 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
818 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 819 | |
ec94fc3d | 820 | retval = put_user(rgid, rgidp); |
821 | if (!retval) { | |
822 | retval = put_user(egid, egidp); | |
823 | if (!retval) | |
824 | retval = put_user(sgid, sgidp); | |
825 | } | |
1da177e4 LT |
826 | |
827 | return retval; | |
828 | } | |
829 | ||
830 | ||
831 | /* | |
832 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
833 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
834 | * whatever uid it wants to). It normally shadows "euid", except when | |
835 | * explicitly set by setfsuid() or for access.. | |
836 | */ | |
e530dca5 | 837 | long __sys_setfsuid(uid_t uid) |
1da177e4 | 838 | { |
d84f4f99 DH |
839 | const struct cred *old; |
840 | struct cred *new; | |
841 | uid_t old_fsuid; | |
a29c33f4 EB |
842 | kuid_t kuid; |
843 | ||
844 | old = current_cred(); | |
845 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
846 | ||
847 | kuid = make_kuid(old->user_ns, uid); | |
848 | if (!uid_valid(kuid)) | |
849 | return old_fsuid; | |
1da177e4 | 850 | |
d84f4f99 DH |
851 | new = prepare_creds(); |
852 | if (!new) | |
a29c33f4 | 853 | return old_fsuid; |
1da177e4 | 854 | |
a29c33f4 EB |
855 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
856 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
40852275 | 857 | ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
858 | if (!uid_eq(kuid, old->fsuid)) { |
859 | new->fsuid = kuid; | |
d84f4f99 DH |
860 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
861 | goto change_okay; | |
1da177e4 | 862 | } |
1da177e4 LT |
863 | } |
864 | ||
d84f4f99 DH |
865 | abort_creds(new); |
866 | return old_fsuid; | |
1da177e4 | 867 | |
d84f4f99 DH |
868 | change_okay: |
869 | commit_creds(new); | |
1da177e4 LT |
870 | return old_fsuid; |
871 | } | |
872 | ||
e530dca5 DB |
873 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
874 | { | |
875 | return __sys_setfsuid(uid); | |
876 | } | |
877 | ||
1da177e4 | 878 | /* |
f42df9e6 | 879 | * Samma på svenska.. |
1da177e4 | 880 | */ |
e530dca5 | 881 | long __sys_setfsgid(gid_t gid) |
1da177e4 | 882 | { |
d84f4f99 DH |
883 | const struct cred *old; |
884 | struct cred *new; | |
885 | gid_t old_fsgid; | |
a29c33f4 EB |
886 | kgid_t kgid; |
887 | ||
888 | old = current_cred(); | |
889 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
890 | ||
891 | kgid = make_kgid(old->user_ns, gid); | |
892 | if (!gid_valid(kgid)) | |
893 | return old_fsgid; | |
d84f4f99 DH |
894 | |
895 | new = prepare_creds(); | |
896 | if (!new) | |
a29c33f4 | 897 | return old_fsgid; |
1da177e4 | 898 | |
a29c33f4 EB |
899 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
900 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
111767c1 | 901 | ns_capable_setid(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
902 | if (!gid_eq(kgid, old->fsgid)) { |
903 | new->fsgid = kgid; | |
39030e13 TC |
904 | if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0) |
905 | goto change_okay; | |
1da177e4 | 906 | } |
1da177e4 | 907 | } |
d84f4f99 | 908 | |
d84f4f99 DH |
909 | abort_creds(new); |
910 | return old_fsgid; | |
911 | ||
912 | change_okay: | |
913 | commit_creds(new); | |
1da177e4 LT |
914 | return old_fsgid; |
915 | } | |
e530dca5 DB |
916 | |
917 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) | |
918 | { | |
919 | return __sys_setfsgid(gid); | |
920 | } | |
2813893f | 921 | #endif /* CONFIG_MULTIUSER */ |
1da177e4 | 922 | |
4a22f166 SR |
923 | /** |
924 | * sys_getpid - return the thread group id of the current process | |
925 | * | |
926 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
927 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
928 | * which case the tgid is the same in all threads of the same group. | |
929 | * | |
930 | * This is SMP safe as current->tgid does not change. | |
931 | */ | |
932 | SYSCALL_DEFINE0(getpid) | |
933 | { | |
934 | return task_tgid_vnr(current); | |
935 | } | |
936 | ||
937 | /* Thread ID - the internal kernel "pid" */ | |
938 | SYSCALL_DEFINE0(gettid) | |
939 | { | |
940 | return task_pid_vnr(current); | |
941 | } | |
942 | ||
943 | /* | |
944 | * Accessing ->real_parent is not SMP-safe, it could | |
945 | * change from under us. However, we can use a stale | |
946 | * value of ->real_parent under rcu_read_lock(), see | |
947 | * release_task()->call_rcu(delayed_put_task_struct). | |
948 | */ | |
949 | SYSCALL_DEFINE0(getppid) | |
950 | { | |
951 | int pid; | |
952 | ||
953 | rcu_read_lock(); | |
954 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
955 | rcu_read_unlock(); | |
956 | ||
957 | return pid; | |
958 | } | |
959 | ||
960 | SYSCALL_DEFINE0(getuid) | |
961 | { | |
962 | /* Only we change this so SMP safe */ | |
963 | return from_kuid_munged(current_user_ns(), current_uid()); | |
964 | } | |
965 | ||
966 | SYSCALL_DEFINE0(geteuid) | |
967 | { | |
968 | /* Only we change this so SMP safe */ | |
969 | return from_kuid_munged(current_user_ns(), current_euid()); | |
970 | } | |
971 | ||
972 | SYSCALL_DEFINE0(getgid) | |
973 | { | |
974 | /* Only we change this so SMP safe */ | |
975 | return from_kgid_munged(current_user_ns(), current_gid()); | |
976 | } | |
977 | ||
978 | SYSCALL_DEFINE0(getegid) | |
979 | { | |
980 | /* Only we change this so SMP safe */ | |
981 | return from_kgid_munged(current_user_ns(), current_egid()); | |
982 | } | |
983 | ||
ca2406ed | 984 | static void do_sys_times(struct tms *tms) |
f06febc9 | 985 | { |
5613fda9 | 986 | u64 tgutime, tgstime, cutime, cstime; |
f06febc9 | 987 | |
e80d0a1a | 988 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
989 | cutime = current->signal->cutime; |
990 | cstime = current->signal->cstime; | |
5613fda9 FW |
991 | tms->tms_utime = nsec_to_clock_t(tgutime); |
992 | tms->tms_stime = nsec_to_clock_t(tgstime); | |
993 | tms->tms_cutime = nsec_to_clock_t(cutime); | |
994 | tms->tms_cstime = nsec_to_clock_t(cstime); | |
f06febc9 FM |
995 | } |
996 | ||
58fd3aa2 | 997 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 998 | { |
1da177e4 LT |
999 | if (tbuf) { |
1000 | struct tms tmp; | |
f06febc9 FM |
1001 | |
1002 | do_sys_times(&tmp); | |
1da177e4 LT |
1003 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
1004 | return -EFAULT; | |
1005 | } | |
e3d5a27d | 1006 | force_successful_syscall_return(); |
1da177e4 LT |
1007 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
1008 | } | |
1009 | ||
ca2406ed AV |
1010 | #ifdef CONFIG_COMPAT |
1011 | static compat_clock_t clock_t_to_compat_clock_t(clock_t x) | |
1012 | { | |
1013 | return compat_jiffies_to_clock_t(clock_t_to_jiffies(x)); | |
1014 | } | |
1015 | ||
1016 | COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf) | |
1017 | { | |
1018 | if (tbuf) { | |
1019 | struct tms tms; | |
1020 | struct compat_tms tmp; | |
1021 | ||
1022 | do_sys_times(&tms); | |
1023 | /* Convert our struct tms to the compat version. */ | |
1024 | tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime); | |
1025 | tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime); | |
1026 | tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime); | |
1027 | tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime); | |
1028 | if (copy_to_user(tbuf, &tmp, sizeof(tmp))) | |
1029 | return -EFAULT; | |
1030 | } | |
1031 | force_successful_syscall_return(); | |
1032 | return compat_jiffies_to_clock_t(jiffies); | |
1033 | } | |
1034 | #endif | |
1035 | ||
1da177e4 LT |
1036 | /* |
1037 | * This needs some heavy checking ... | |
1038 | * I just haven't the stomach for it. I also don't fully | |
1039 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
1040 | * | |
1041 | * OK, I think I have the protection semantics right.... this is really | |
1042 | * only important on a multi-user system anyway, to make sure one user | |
1043 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
1044 | * | |
98611e4e | 1045 | * !PF_FORKNOEXEC check to conform completely to POSIX. |
1da177e4 | 1046 | */ |
b290ebe2 | 1047 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
1048 | { |
1049 | struct task_struct *p; | |
ee0acf90 | 1050 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
1051 | struct pid *pgrp; |
1052 | int err; | |
1da177e4 LT |
1053 | |
1054 | if (!pid) | |
b488893a | 1055 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
1056 | if (!pgid) |
1057 | pgid = pid; | |
1058 | if (pgid < 0) | |
1059 | return -EINVAL; | |
950eaaca | 1060 | rcu_read_lock(); |
1da177e4 LT |
1061 | |
1062 | /* From this point forward we keep holding onto the tasklist lock | |
1063 | * so that our parent does not change from under us. -DaveM | |
1064 | */ | |
1065 | write_lock_irq(&tasklist_lock); | |
1066 | ||
1067 | err = -ESRCH; | |
4e021306 | 1068 | p = find_task_by_vpid(pid); |
1da177e4 LT |
1069 | if (!p) |
1070 | goto out; | |
1071 | ||
1072 | err = -EINVAL; | |
1073 | if (!thread_group_leader(p)) | |
1074 | goto out; | |
1075 | ||
4e021306 | 1076 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 1077 | err = -EPERM; |
41487c65 | 1078 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
1079 | goto out; |
1080 | err = -EACCES; | |
98611e4e | 1081 | if (!(p->flags & PF_FORKNOEXEC)) |
1da177e4 LT |
1082 | goto out; |
1083 | } else { | |
1084 | err = -ESRCH; | |
ee0acf90 | 1085 | if (p != group_leader) |
1da177e4 LT |
1086 | goto out; |
1087 | } | |
1088 | ||
1089 | err = -EPERM; | |
1090 | if (p->signal->leader) | |
1091 | goto out; | |
1092 | ||
4e021306 | 1093 | pgrp = task_pid(p); |
1da177e4 | 1094 | if (pgid != pid) { |
b488893a | 1095 | struct task_struct *g; |
1da177e4 | 1096 | |
4e021306 ON |
1097 | pgrp = find_vpid(pgid); |
1098 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 1099 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 1100 | goto out; |
1da177e4 LT |
1101 | } |
1102 | ||
1da177e4 LT |
1103 | err = security_task_setpgid(p, pgid); |
1104 | if (err) | |
1105 | goto out; | |
1106 | ||
1b0f7ffd | 1107 | if (task_pgrp(p) != pgrp) |
83beaf3c | 1108 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
1109 | |
1110 | err = 0; | |
1111 | out: | |
1112 | /* All paths lead to here, thus we are safe. -DaveM */ | |
1113 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 1114 | rcu_read_unlock(); |
1da177e4 LT |
1115 | return err; |
1116 | } | |
1117 | ||
192c5807 | 1118 | static int do_getpgid(pid_t pid) |
1da177e4 | 1119 | { |
12a3de0a ON |
1120 | struct task_struct *p; |
1121 | struct pid *grp; | |
1122 | int retval; | |
1123 | ||
1124 | rcu_read_lock(); | |
756184b7 | 1125 | if (!pid) |
12a3de0a | 1126 | grp = task_pgrp(current); |
756184b7 | 1127 | else { |
1da177e4 | 1128 | retval = -ESRCH; |
12a3de0a ON |
1129 | p = find_task_by_vpid(pid); |
1130 | if (!p) | |
1131 | goto out; | |
1132 | grp = task_pgrp(p); | |
1133 | if (!grp) | |
1134 | goto out; | |
1135 | ||
1136 | retval = security_task_getpgid(p); | |
1137 | if (retval) | |
1138 | goto out; | |
1da177e4 | 1139 | } |
12a3de0a ON |
1140 | retval = pid_vnr(grp); |
1141 | out: | |
1142 | rcu_read_unlock(); | |
1143 | return retval; | |
1da177e4 LT |
1144 | } |
1145 | ||
192c5807 DB |
1146 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1147 | { | |
1148 | return do_getpgid(pid); | |
1149 | } | |
1150 | ||
1da177e4 LT |
1151 | #ifdef __ARCH_WANT_SYS_GETPGRP |
1152 | ||
dbf040d9 | 1153 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1154 | { |
192c5807 | 1155 | return do_getpgid(0); |
1da177e4 LT |
1156 | } |
1157 | ||
1158 | #endif | |
1159 | ||
dbf040d9 | 1160 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1161 | { |
1dd768c0 ON |
1162 | struct task_struct *p; |
1163 | struct pid *sid; | |
1164 | int retval; | |
1165 | ||
1166 | rcu_read_lock(); | |
756184b7 | 1167 | if (!pid) |
1dd768c0 | 1168 | sid = task_session(current); |
756184b7 | 1169 | else { |
1da177e4 | 1170 | retval = -ESRCH; |
1dd768c0 ON |
1171 | p = find_task_by_vpid(pid); |
1172 | if (!p) | |
1173 | goto out; | |
1174 | sid = task_session(p); | |
1175 | if (!sid) | |
1176 | goto out; | |
1177 | ||
1178 | retval = security_task_getsid(p); | |
1179 | if (retval) | |
1180 | goto out; | |
1da177e4 | 1181 | } |
1dd768c0 ON |
1182 | retval = pid_vnr(sid); |
1183 | out: | |
1184 | rcu_read_unlock(); | |
1185 | return retval; | |
1da177e4 LT |
1186 | } |
1187 | ||
81dabb46 ON |
1188 | static void set_special_pids(struct pid *pid) |
1189 | { | |
1190 | struct task_struct *curr = current->group_leader; | |
1191 | ||
1192 | if (task_session(curr) != pid) | |
1193 | change_pid(curr, PIDTYPE_SID, pid); | |
1194 | ||
1195 | if (task_pgrp(curr) != pid) | |
1196 | change_pid(curr, PIDTYPE_PGID, pid); | |
1197 | } | |
1198 | ||
e2aaa9f4 | 1199 | int ksys_setsid(void) |
1da177e4 | 1200 | { |
e19f247a | 1201 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1202 | struct pid *sid = task_pid(group_leader); |
1203 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1204 | int err = -EPERM; |
1205 | ||
1da177e4 | 1206 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1207 | /* Fail if I am already a session leader */ |
1208 | if (group_leader->signal->leader) | |
1209 | goto out; | |
1210 | ||
430c6231 ON |
1211 | /* Fail if a process group id already exists that equals the |
1212 | * proposed session id. | |
390e2ff0 | 1213 | */ |
6806aac6 | 1214 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1215 | goto out; |
1216 | ||
e19f247a | 1217 | group_leader->signal->leader = 1; |
81dabb46 | 1218 | set_special_pids(sid); |
24ec839c | 1219 | |
9c9f4ded | 1220 | proc_clear_tty(group_leader); |
24ec839c | 1221 | |
e4cc0a9c | 1222 | err = session; |
1da177e4 LT |
1223 | out: |
1224 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1225 | if (err > 0) { |
0d0df599 | 1226 | proc_sid_connector(group_leader); |
5091faa4 MG |
1227 | sched_autogroup_create_attach(group_leader); |
1228 | } | |
1da177e4 LT |
1229 | return err; |
1230 | } | |
1231 | ||
e2aaa9f4 DB |
1232 | SYSCALL_DEFINE0(setsid) |
1233 | { | |
1234 | return ksys_setsid(); | |
1235 | } | |
1236 | ||
1da177e4 LT |
1237 | DECLARE_RWSEM(uts_sem); |
1238 | ||
c1da50fa AW |
1239 | #ifdef COMPAT_UTS_MACHINE |
1240 | static char compat_uts_machine[__OLD_UTS_LEN+1] = COMPAT_UTS_MACHINE; | |
1241 | ||
1242 | static int __init parse_compat_uts_machine(char *arg) | |
1243 | { | |
1244 | strncpy(compat_uts_machine, arg, __OLD_UTS_LEN); | |
1245 | compat_uts_machine[__OLD_UTS_LEN] = 0; | |
1246 | return 0; | |
1247 | } | |
1248 | early_param("compat_uts_machine", parse_compat_uts_machine); | |
1249 | ||
1250 | #undef COMPAT_UTS_MACHINE | |
1251 | #define COMPAT_UTS_MACHINE compat_uts_machine | |
1252 | #endif | |
1253 | ||
e28cbf22 CH |
1254 | #ifdef COMPAT_UTS_MACHINE |
1255 | #define override_architecture(name) \ | |
46da2766 | 1256 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1257 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1258 | sizeof(COMPAT_UTS_MACHINE))) | |
1259 | #else | |
1260 | #define override_architecture(name) 0 | |
1261 | #endif | |
1262 | ||
be27425d AK |
1263 | /* |
1264 | * Work around broken programs that cannot handle "Linux 3.0". | |
1265 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
b7285b42 JN |
1266 | * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be |
1267 | * 2.6.60. | |
be27425d | 1268 | */ |
2702b152 | 1269 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1270 | { |
1271 | int ret = 0; | |
be27425d AK |
1272 | |
1273 | if (current->personality & UNAME26) { | |
2702b152 KC |
1274 | const char *rest = UTS_RELEASE; |
1275 | char buf[65] = { 0 }; | |
be27425d AK |
1276 | int ndots = 0; |
1277 | unsigned v; | |
2702b152 | 1278 | size_t copy; |
be27425d AK |
1279 | |
1280 | while (*rest) { | |
1281 | if (*rest == '.' && ++ndots >= 3) | |
1282 | break; | |
1283 | if (!isdigit(*rest) && *rest != '.') | |
1284 | break; | |
1285 | rest++; | |
1286 | } | |
88a68672 | 1287 | v = LINUX_VERSION_PATCHLEVEL + 60; |
31fd84b9 | 1288 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1289 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1290 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1291 | } |
1292 | return ret; | |
1293 | } | |
1294 | ||
e48fbb69 | 1295 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 | 1296 | { |
42a0cc34 | 1297 | struct new_utsname tmp; |
1da177e4 LT |
1298 | |
1299 | down_read(&uts_sem); | |
42a0cc34 | 1300 | memcpy(&tmp, utsname(), sizeof(tmp)); |
1da177e4 | 1301 | up_read(&uts_sem); |
42a0cc34 JH |
1302 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1303 | return -EFAULT; | |
e28cbf22 | 1304 | |
42a0cc34 JH |
1305 | if (override_release(name->release, sizeof(name->release))) |
1306 | return -EFAULT; | |
1307 | if (override_architecture(name)) | |
1308 | return -EFAULT; | |
1309 | return 0; | |
1da177e4 LT |
1310 | } |
1311 | ||
5cacdb4a CH |
1312 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1313 | /* | |
1314 | * Old cruft | |
1315 | */ | |
1316 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1317 | { | |
42a0cc34 | 1318 | struct old_utsname tmp; |
5cacdb4a CH |
1319 | |
1320 | if (!name) | |
1321 | return -EFAULT; | |
1322 | ||
1323 | down_read(&uts_sem); | |
42a0cc34 | 1324 | memcpy(&tmp, utsname(), sizeof(tmp)); |
5cacdb4a | 1325 | up_read(&uts_sem); |
42a0cc34 JH |
1326 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1327 | return -EFAULT; | |
5cacdb4a | 1328 | |
42a0cc34 JH |
1329 | if (override_release(name->release, sizeof(name->release))) |
1330 | return -EFAULT; | |
1331 | if (override_architecture(name)) | |
1332 | return -EFAULT; | |
1333 | return 0; | |
5cacdb4a CH |
1334 | } |
1335 | ||
1336 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1337 | { | |
5e1aada0 | 1338 | struct oldold_utsname tmp; |
5cacdb4a CH |
1339 | |
1340 | if (!name) | |
1341 | return -EFAULT; | |
5cacdb4a | 1342 | |
5e1aada0 JP |
1343 | memset(&tmp, 0, sizeof(tmp)); |
1344 | ||
5cacdb4a | 1345 | down_read(&uts_sem); |
42a0cc34 JH |
1346 | memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN); |
1347 | memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN); | |
1348 | memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN); | |
1349 | memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN); | |
1350 | memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN); | |
5cacdb4a | 1351 | up_read(&uts_sem); |
42a0cc34 JH |
1352 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1353 | return -EFAULT; | |
5cacdb4a | 1354 | |
42a0cc34 JH |
1355 | if (override_architecture(name)) |
1356 | return -EFAULT; | |
1357 | if (override_release(name->release, sizeof(name->release))) | |
1358 | return -EFAULT; | |
1359 | return 0; | |
5cacdb4a CH |
1360 | } |
1361 | #endif | |
1362 | ||
5a8a82b1 | 1363 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1364 | { |
1365 | int errno; | |
1366 | char tmp[__NEW_UTS_LEN]; | |
1367 | ||
bb96a6f5 | 1368 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1369 | return -EPERM; |
fc832ad3 | 1370 | |
1da177e4 LT |
1371 | if (len < 0 || len > __NEW_UTS_LEN) |
1372 | return -EINVAL; | |
1da177e4 LT |
1373 | errno = -EFAULT; |
1374 | if (!copy_from_user(tmp, name, len)) { | |
42a0cc34 | 1375 | struct new_utsname *u; |
9679e4dd | 1376 | |
42a0cc34 JH |
1377 | down_write(&uts_sem); |
1378 | u = utsname(); | |
9679e4dd AM |
1379 | memcpy(u->nodename, tmp, len); |
1380 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1381 | errno = 0; |
499eea6b | 1382 | uts_proc_notify(UTS_PROC_HOSTNAME); |
42a0cc34 | 1383 | up_write(&uts_sem); |
1da177e4 | 1384 | } |
1da177e4 LT |
1385 | return errno; |
1386 | } | |
1387 | ||
1388 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1389 | ||
5a8a82b1 | 1390 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 | 1391 | { |
42a0cc34 | 1392 | int i; |
9679e4dd | 1393 | struct new_utsname *u; |
42a0cc34 | 1394 | char tmp[__NEW_UTS_LEN + 1]; |
1da177e4 LT |
1395 | |
1396 | if (len < 0) | |
1397 | return -EINVAL; | |
1398 | down_read(&uts_sem); | |
9679e4dd AM |
1399 | u = utsname(); |
1400 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1401 | if (i > len) |
1402 | i = len; | |
42a0cc34 | 1403 | memcpy(tmp, u->nodename, i); |
1da177e4 | 1404 | up_read(&uts_sem); |
42a0cc34 JH |
1405 | if (copy_to_user(name, tmp, i)) |
1406 | return -EFAULT; | |
1407 | return 0; | |
1da177e4 LT |
1408 | } |
1409 | ||
1410 | #endif | |
1411 | ||
1412 | /* | |
1413 | * Only setdomainname; getdomainname can be implemented by calling | |
1414 | * uname() | |
1415 | */ | |
5a8a82b1 | 1416 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1417 | { |
1418 | int errno; | |
1419 | char tmp[__NEW_UTS_LEN]; | |
1420 | ||
fc832ad3 | 1421 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1422 | return -EPERM; |
1423 | if (len < 0 || len > __NEW_UTS_LEN) | |
1424 | return -EINVAL; | |
1425 | ||
1da177e4 LT |
1426 | errno = -EFAULT; |
1427 | if (!copy_from_user(tmp, name, len)) { | |
42a0cc34 | 1428 | struct new_utsname *u; |
9679e4dd | 1429 | |
42a0cc34 JH |
1430 | down_write(&uts_sem); |
1431 | u = utsname(); | |
9679e4dd AM |
1432 | memcpy(u->domainname, tmp, len); |
1433 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1434 | errno = 0; |
499eea6b | 1435 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
42a0cc34 | 1436 | up_write(&uts_sem); |
1da177e4 | 1437 | } |
1da177e4 LT |
1438 | return errno; |
1439 | } | |
1440 | ||
e48fbb69 | 1441 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1442 | { |
b9518345 JS |
1443 | struct rlimit value; |
1444 | int ret; | |
1445 | ||
1446 | ret = do_prlimit(current, resource, NULL, &value); | |
1447 | if (!ret) | |
1448 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1449 | ||
1450 | return ret; | |
1da177e4 LT |
1451 | } |
1452 | ||
d9e968cb AV |
1453 | #ifdef CONFIG_COMPAT |
1454 | ||
1455 | COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource, | |
1456 | struct compat_rlimit __user *, rlim) | |
1457 | { | |
1458 | struct rlimit r; | |
1459 | struct compat_rlimit r32; | |
1460 | ||
1461 | if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit))) | |
1462 | return -EFAULT; | |
1463 | ||
1464 | if (r32.rlim_cur == COMPAT_RLIM_INFINITY) | |
1465 | r.rlim_cur = RLIM_INFINITY; | |
1466 | else | |
1467 | r.rlim_cur = r32.rlim_cur; | |
1468 | if (r32.rlim_max == COMPAT_RLIM_INFINITY) | |
1469 | r.rlim_max = RLIM_INFINITY; | |
1470 | else | |
1471 | r.rlim_max = r32.rlim_max; | |
1472 | return do_prlimit(current, resource, &r, NULL); | |
1473 | } | |
1474 | ||
1475 | COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource, | |
1476 | struct compat_rlimit __user *, rlim) | |
1477 | { | |
1478 | struct rlimit r; | |
1479 | int ret; | |
1480 | ||
1481 | ret = do_prlimit(current, resource, NULL, &r); | |
1482 | if (!ret) { | |
58c7ffc0 | 1483 | struct compat_rlimit r32; |
d9e968cb AV |
1484 | if (r.rlim_cur > COMPAT_RLIM_INFINITY) |
1485 | r32.rlim_cur = COMPAT_RLIM_INFINITY; | |
1486 | else | |
1487 | r32.rlim_cur = r.rlim_cur; | |
1488 | if (r.rlim_max > COMPAT_RLIM_INFINITY) | |
1489 | r32.rlim_max = COMPAT_RLIM_INFINITY; | |
1490 | else | |
1491 | r32.rlim_max = r.rlim_max; | |
1492 | ||
1493 | if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit))) | |
1494 | return -EFAULT; | |
1495 | } | |
1496 | return ret; | |
1497 | } | |
1498 | ||
1499 | #endif | |
1500 | ||
1da177e4 LT |
1501 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT |
1502 | ||
1503 | /* | |
1504 | * Back compatibility for getrlimit. Needed for some apps. | |
1505 | */ | |
e48fbb69 HC |
1506 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1507 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1508 | { |
1509 | struct rlimit x; | |
1510 | if (resource >= RLIM_NLIMITS) | |
1511 | return -EINVAL; | |
1512 | ||
23d6aef7 | 1513 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
1da177e4 LT |
1514 | task_lock(current->group_leader); |
1515 | x = current->signal->rlim[resource]; | |
1516 | task_unlock(current->group_leader); | |
756184b7 | 1517 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1518 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1519 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 | 1520 | x.rlim_max = 0x7FFFFFFF; |
ec94fc3d | 1521 | return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0; |
1da177e4 LT |
1522 | } |
1523 | ||
613763a1 AV |
1524 | #ifdef CONFIG_COMPAT |
1525 | COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, | |
1526 | struct compat_rlimit __user *, rlim) | |
1527 | { | |
1528 | struct rlimit r; | |
1529 | ||
1530 | if (resource >= RLIM_NLIMITS) | |
1531 | return -EINVAL; | |
1532 | ||
23d6aef7 | 1533 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
613763a1 AV |
1534 | task_lock(current->group_leader); |
1535 | r = current->signal->rlim[resource]; | |
1536 | task_unlock(current->group_leader); | |
1537 | if (r.rlim_cur > 0x7FFFFFFF) | |
1538 | r.rlim_cur = 0x7FFFFFFF; | |
1539 | if (r.rlim_max > 0x7FFFFFFF) | |
1540 | r.rlim_max = 0x7FFFFFFF; | |
1541 | ||
1542 | if (put_user(r.rlim_cur, &rlim->rlim_cur) || | |
1543 | put_user(r.rlim_max, &rlim->rlim_max)) | |
1544 | return -EFAULT; | |
1545 | return 0; | |
1546 | } | |
1547 | #endif | |
1548 | ||
1da177e4 LT |
1549 | #endif |
1550 | ||
c022a0ac JS |
1551 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1552 | { | |
1553 | #if BITS_PER_LONG < 64 | |
1554 | return rlim64 >= ULONG_MAX; | |
1555 | #else | |
1556 | return rlim64 == RLIM64_INFINITY; | |
1557 | #endif | |
1558 | } | |
1559 | ||
1560 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1561 | { | |
1562 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1563 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1564 | else | |
1565 | rlim64->rlim_cur = rlim->rlim_cur; | |
1566 | if (rlim->rlim_max == RLIM_INFINITY) | |
1567 | rlim64->rlim_max = RLIM64_INFINITY; | |
1568 | else | |
1569 | rlim64->rlim_max = rlim->rlim_max; | |
1570 | } | |
1571 | ||
1572 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1573 | { | |
1574 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1575 | rlim->rlim_cur = RLIM_INFINITY; | |
1576 | else | |
1577 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1578 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1579 | rlim->rlim_max = RLIM_INFINITY; | |
1580 | else | |
1581 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1582 | } | |
1583 | ||
1c1e618d | 1584 | /* make sure you are allowed to change @tsk limits before calling this */ |
5b41535a JS |
1585 | int do_prlimit(struct task_struct *tsk, unsigned int resource, |
1586 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1da177e4 | 1587 | { |
5b41535a | 1588 | struct rlimit *rlim; |
86f162f4 | 1589 | int retval = 0; |
1da177e4 LT |
1590 | |
1591 | if (resource >= RLIM_NLIMITS) | |
1592 | return -EINVAL; | |
5b41535a JS |
1593 | if (new_rlim) { |
1594 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1595 | return -EINVAL; | |
1596 | if (resource == RLIMIT_NOFILE && | |
1597 | new_rlim->rlim_max > sysctl_nr_open) | |
1598 | return -EPERM; | |
1599 | } | |
1da177e4 | 1600 | |
1c1e618d JS |
1601 | /* protect tsk->signal and tsk->sighand from disappearing */ |
1602 | read_lock(&tasklist_lock); | |
1603 | if (!tsk->sighand) { | |
1604 | retval = -ESRCH; | |
1605 | goto out; | |
1606 | } | |
1607 | ||
5b41535a | 1608 | rlim = tsk->signal->rlim + resource; |
86f162f4 | 1609 | task_lock(tsk->group_leader); |
5b41535a | 1610 | if (new_rlim) { |
fc832ad3 SH |
1611 | /* Keep the capable check against init_user_ns until |
1612 | cgroups can contain all limits */ | |
5b41535a JS |
1613 | if (new_rlim->rlim_max > rlim->rlim_max && |
1614 | !capable(CAP_SYS_RESOURCE)) | |
1615 | retval = -EPERM; | |
1616 | if (!retval) | |
cad4ea54 | 1617 | retval = security_task_setrlimit(tsk, resource, new_rlim); |
5b41535a JS |
1618 | } |
1619 | if (!retval) { | |
1620 | if (old_rlim) | |
1621 | *old_rlim = *rlim; | |
1622 | if (new_rlim) | |
1623 | *rlim = *new_rlim; | |
9926e4c7 | 1624 | } |
7855c35d | 1625 | task_unlock(tsk->group_leader); |
1da177e4 | 1626 | |
d3561f78 | 1627 | /* |
24db4dd9 | 1628 | * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not |
5afe69c2 | 1629 | * infinite. In case of RLIM_INFINITY the posix CPU timer code |
24db4dd9 | 1630 | * ignores the rlimit. |
d3561f78 | 1631 | */ |
5b41535a | 1632 | if (!retval && new_rlim && resource == RLIMIT_CPU && |
baa73d9e NP |
1633 | new_rlim->rlim_cur != RLIM_INFINITY && |
1634 | IS_ENABLED(CONFIG_POSIX_TIMERS)) | |
5b41535a | 1635 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); |
ec9e16ba | 1636 | out: |
1c1e618d | 1637 | read_unlock(&tasklist_lock); |
2fb9d268 | 1638 | return retval; |
1da177e4 LT |
1639 | } |
1640 | ||
c022a0ac | 1641 | /* rcu lock must be held */ |
791ec491 SS |
1642 | static int check_prlimit_permission(struct task_struct *task, |
1643 | unsigned int flags) | |
c022a0ac JS |
1644 | { |
1645 | const struct cred *cred = current_cred(), *tcred; | |
791ec491 | 1646 | bool id_match; |
c022a0ac | 1647 | |
fc832ad3 SH |
1648 | if (current == task) |
1649 | return 0; | |
c022a0ac | 1650 | |
fc832ad3 | 1651 | tcred = __task_cred(task); |
791ec491 SS |
1652 | id_match = (uid_eq(cred->uid, tcred->euid) && |
1653 | uid_eq(cred->uid, tcred->suid) && | |
1654 | uid_eq(cred->uid, tcred->uid) && | |
1655 | gid_eq(cred->gid, tcred->egid) && | |
1656 | gid_eq(cred->gid, tcred->sgid) && | |
1657 | gid_eq(cred->gid, tcred->gid)); | |
1658 | if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) | |
1659 | return -EPERM; | |
fc832ad3 | 1660 | |
791ec491 | 1661 | return security_task_prlimit(cred, tcred, flags); |
c022a0ac JS |
1662 | } |
1663 | ||
1664 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1665 | const struct rlimit64 __user *, new_rlim, | |
1666 | struct rlimit64 __user *, old_rlim) | |
1667 | { | |
1668 | struct rlimit64 old64, new64; | |
1669 | struct rlimit old, new; | |
1670 | struct task_struct *tsk; | |
791ec491 | 1671 | unsigned int checkflags = 0; |
c022a0ac JS |
1672 | int ret; |
1673 | ||
791ec491 SS |
1674 | if (old_rlim) |
1675 | checkflags |= LSM_PRLIMIT_READ; | |
1676 | ||
c022a0ac JS |
1677 | if (new_rlim) { |
1678 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1679 | return -EFAULT; | |
1680 | rlim64_to_rlim(&new64, &new); | |
791ec491 | 1681 | checkflags |= LSM_PRLIMIT_WRITE; |
c022a0ac JS |
1682 | } |
1683 | ||
1684 | rcu_read_lock(); | |
1685 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1686 | if (!tsk) { | |
1687 | rcu_read_unlock(); | |
1688 | return -ESRCH; | |
1689 | } | |
791ec491 | 1690 | ret = check_prlimit_permission(tsk, checkflags); |
c022a0ac JS |
1691 | if (ret) { |
1692 | rcu_read_unlock(); | |
1693 | return ret; | |
1694 | } | |
1695 | get_task_struct(tsk); | |
1696 | rcu_read_unlock(); | |
1697 | ||
1698 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1699 | old_rlim ? &old : NULL); | |
1700 | ||
1701 | if (!ret && old_rlim) { | |
1702 | rlim_to_rlim64(&old, &old64); | |
1703 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1704 | ret = -EFAULT; | |
1705 | } | |
1706 | ||
1707 | put_task_struct(tsk); | |
1708 | return ret; | |
1709 | } | |
1710 | ||
7855c35d JS |
1711 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1712 | { | |
1713 | struct rlimit new_rlim; | |
1714 | ||
1715 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1716 | return -EFAULT; | |
5b41535a | 1717 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1718 | } |
1719 | ||
1da177e4 LT |
1720 | /* |
1721 | * It would make sense to put struct rusage in the task_struct, | |
1722 | * except that would make the task_struct be *really big*. After | |
1723 | * task_struct gets moved into malloc'ed memory, it would | |
1724 | * make sense to do this. It will make moving the rest of the information | |
1725 | * a lot simpler! (Which we're not doing right now because we're not | |
1726 | * measuring them yet). | |
1727 | * | |
1da177e4 LT |
1728 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1729 | * races with threads incrementing their own counters. But since word | |
1730 | * reads are atomic, we either get new values or old values and we don't | |
1731 | * care which for the sums. We always take the siglock to protect reading | |
1732 | * the c* fields from p->signal from races with exit.c updating those | |
1733 | * fields when reaping, so a sample either gets all the additions of a | |
1734 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1735 | * |
de047c1b RT |
1736 | * Locking: |
1737 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1738 | * for the cases current multithreaded, non-current single threaded | |
1739 | * non-current multithreaded. Thread traversal is now safe with | |
1740 | * the siglock held. | |
1741 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1742 | * single threaded, as no one else can take our signal_struct away, no one | |
1743 | * else can reap the children to update signal->c* counters, and no one else | |
1744 | * can race with the signal-> fields. If we do not take any lock, the | |
1745 | * signal-> fields could be read out of order while another thread was just | |
1746 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1747 | * On the writer side, write memory barrier is implied in __exit_signal | |
1748 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1749 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1750 | * |
1da177e4 LT |
1751 | */ |
1752 | ||
f06febc9 | 1753 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1754 | { |
679c9cd4 SK |
1755 | r->ru_nvcsw += t->nvcsw; |
1756 | r->ru_nivcsw += t->nivcsw; | |
1757 | r->ru_minflt += t->min_flt; | |
1758 | r->ru_majflt += t->maj_flt; | |
1759 | r->ru_inblock += task_io_get_inblock(t); | |
1760 | r->ru_oublock += task_io_get_oublock(t); | |
1761 | } | |
1762 | ||
ce72a16f | 1763 | void getrusage(struct task_struct *p, int who, struct rusage *r) |
1da177e4 LT |
1764 | { |
1765 | struct task_struct *t; | |
1766 | unsigned long flags; | |
5613fda9 | 1767 | u64 tgutime, tgstime, utime, stime; |
1f10206c | 1768 | unsigned long maxrss = 0; |
1da177e4 | 1769 | |
ec94fc3d | 1770 | memset((char *)r, 0, sizeof (*r)); |
64861634 | 1771 | utime = stime = 0; |
1da177e4 | 1772 | |
679c9cd4 | 1773 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1774 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1775 | accumulate_thread_rusage(p, r); |
1f10206c | 1776 | maxrss = p->signal->maxrss; |
679c9cd4 SK |
1777 | goto out; |
1778 | } | |
1779 | ||
d6cf723a | 1780 | if (!lock_task_sighand(p, &flags)) |
de047c1b | 1781 | return; |
0f59cc4a | 1782 | |
1da177e4 | 1783 | switch (who) { |
ec94fc3d | 1784 | case RUSAGE_BOTH: |
1785 | case RUSAGE_CHILDREN: | |
1786 | utime = p->signal->cutime; | |
1787 | stime = p->signal->cstime; | |
1788 | r->ru_nvcsw = p->signal->cnvcsw; | |
1789 | r->ru_nivcsw = p->signal->cnivcsw; | |
1790 | r->ru_minflt = p->signal->cmin_flt; | |
1791 | r->ru_majflt = p->signal->cmaj_flt; | |
1792 | r->ru_inblock = p->signal->cinblock; | |
1793 | r->ru_oublock = p->signal->coublock; | |
1794 | maxrss = p->signal->cmaxrss; | |
1795 | ||
1796 | if (who == RUSAGE_CHILDREN) | |
1da177e4 | 1797 | break; |
df561f66 | 1798 | fallthrough; |
0f59cc4a | 1799 | |
ec94fc3d | 1800 | case RUSAGE_SELF: |
1801 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); | |
1802 | utime += tgutime; | |
1803 | stime += tgstime; | |
1804 | r->ru_nvcsw += p->signal->nvcsw; | |
1805 | r->ru_nivcsw += p->signal->nivcsw; | |
1806 | r->ru_minflt += p->signal->min_flt; | |
1807 | r->ru_majflt += p->signal->maj_flt; | |
1808 | r->ru_inblock += p->signal->inblock; | |
1809 | r->ru_oublock += p->signal->oublock; | |
1810 | if (maxrss < p->signal->maxrss) | |
1811 | maxrss = p->signal->maxrss; | |
1812 | t = p; | |
1813 | do { | |
1814 | accumulate_thread_rusage(t, r); | |
1815 | } while_each_thread(p, t); | |
1816 | break; | |
1817 | ||
1818 | default: | |
1819 | BUG(); | |
1da177e4 | 1820 | } |
de047c1b | 1821 | unlock_task_sighand(p, &flags); |
de047c1b | 1822 | |
679c9cd4 | 1823 | out: |
bdd565f8 AB |
1824 | r->ru_utime = ns_to_kernel_old_timeval(utime); |
1825 | r->ru_stime = ns_to_kernel_old_timeval(stime); | |
1f10206c JP |
1826 | |
1827 | if (who != RUSAGE_CHILDREN) { | |
1828 | struct mm_struct *mm = get_task_mm(p); | |
ec94fc3d | 1829 | |
1f10206c JP |
1830 | if (mm) { |
1831 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1832 | mmput(mm); | |
1833 | } | |
1834 | } | |
1835 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1da177e4 LT |
1836 | } |
1837 | ||
ce72a16f | 1838 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 LT |
1839 | { |
1840 | struct rusage r; | |
ec94fc3d | 1841 | |
679c9cd4 SK |
1842 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1843 | who != RUSAGE_THREAD) | |
1da177e4 | 1844 | return -EINVAL; |
ce72a16f AV |
1845 | |
1846 | getrusage(current, who, &r); | |
1847 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; | |
1da177e4 LT |
1848 | } |
1849 | ||
8d2d5c4a AV |
1850 | #ifdef CONFIG_COMPAT |
1851 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1852 | { | |
1853 | struct rusage r; | |
1854 | ||
1855 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1856 | who != RUSAGE_THREAD) | |
1857 | return -EINVAL; | |
1858 | ||
ce72a16f | 1859 | getrusage(current, who, &r); |
8d2d5c4a AV |
1860 | return put_compat_rusage(&r, ru); |
1861 | } | |
1862 | #endif | |
1863 | ||
e48fbb69 | 1864 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
1865 | { |
1866 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1867 | return mask; | |
1868 | } | |
3b7391de | 1869 | |
6e399cd1 | 1870 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
b32dfe37 | 1871 | { |
2903ff01 | 1872 | struct fd exe; |
496ad9aa | 1873 | struct inode *inode; |
2903ff01 | 1874 | int err; |
b32dfe37 | 1875 | |
2903ff01 AV |
1876 | exe = fdget(fd); |
1877 | if (!exe.file) | |
b32dfe37 CG |
1878 | return -EBADF; |
1879 | ||
496ad9aa | 1880 | inode = file_inode(exe.file); |
b32dfe37 CG |
1881 | |
1882 | /* | |
1883 | * Because the original mm->exe_file points to executable file, make | |
1884 | * sure that this one is executable as well, to avoid breaking an | |
1885 | * overall picture. | |
1886 | */ | |
1887 | err = -EACCES; | |
90f8572b | 1888 | if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path)) |
b32dfe37 CG |
1889 | goto exit; |
1890 | ||
02f92b38 | 1891 | err = file_permission(exe.file, MAY_EXEC); |
b32dfe37 CG |
1892 | if (err) |
1893 | goto exit; | |
1894 | ||
35d7bdc8 | 1895 | err = replace_mm_exe_file(mm, exe.file); |
b32dfe37 | 1896 | exit: |
2903ff01 | 1897 | fdput(exe); |
b32dfe37 CG |
1898 | return err; |
1899 | } | |
1900 | ||
f606b77f | 1901 | /* |
11bbd8b4 MK |
1902 | * Check arithmetic relations of passed addresses. |
1903 | * | |
f606b77f CG |
1904 | * WARNING: we don't require any capability here so be very careful |
1905 | * in what is allowed for modification from userspace. | |
1906 | */ | |
11bbd8b4 | 1907 | static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map) |
f606b77f CG |
1908 | { |
1909 | unsigned long mmap_max_addr = TASK_SIZE; | |
f606b77f CG |
1910 | int error = -EINVAL, i; |
1911 | ||
1912 | static const unsigned char offsets[] = { | |
1913 | offsetof(struct prctl_mm_map, start_code), | |
1914 | offsetof(struct prctl_mm_map, end_code), | |
1915 | offsetof(struct prctl_mm_map, start_data), | |
1916 | offsetof(struct prctl_mm_map, end_data), | |
1917 | offsetof(struct prctl_mm_map, start_brk), | |
1918 | offsetof(struct prctl_mm_map, brk), | |
1919 | offsetof(struct prctl_mm_map, start_stack), | |
1920 | offsetof(struct prctl_mm_map, arg_start), | |
1921 | offsetof(struct prctl_mm_map, arg_end), | |
1922 | offsetof(struct prctl_mm_map, env_start), | |
1923 | offsetof(struct prctl_mm_map, env_end), | |
1924 | }; | |
1925 | ||
1926 | /* | |
1927 | * Make sure the members are not somewhere outside | |
1928 | * of allowed address space. | |
1929 | */ | |
1930 | for (i = 0; i < ARRAY_SIZE(offsets); i++) { | |
1931 | u64 val = *(u64 *)((char *)prctl_map + offsets[i]); | |
1932 | ||
1933 | if ((unsigned long)val >= mmap_max_addr || | |
1934 | (unsigned long)val < mmap_min_addr) | |
1935 | goto out; | |
1936 | } | |
1937 | ||
1938 | /* | |
1939 | * Make sure the pairs are ordered. | |
1940 | */ | |
1941 | #define __prctl_check_order(__m1, __op, __m2) \ | |
1942 | ((unsigned long)prctl_map->__m1 __op \ | |
1943 | (unsigned long)prctl_map->__m2) ? 0 : -EINVAL | |
1944 | error = __prctl_check_order(start_code, <, end_code); | |
a9e73998 | 1945 | error |= __prctl_check_order(start_data,<=, end_data); |
f606b77f CG |
1946 | error |= __prctl_check_order(start_brk, <=, brk); |
1947 | error |= __prctl_check_order(arg_start, <=, arg_end); | |
1948 | error |= __prctl_check_order(env_start, <=, env_end); | |
1949 | if (error) | |
1950 | goto out; | |
1951 | #undef __prctl_check_order | |
1952 | ||
1953 | error = -EINVAL; | |
1954 | ||
f606b77f CG |
1955 | /* |
1956 | * Neither we should allow to override limits if they set. | |
1957 | */ | |
1958 | if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk, | |
1959 | prctl_map->start_brk, prctl_map->end_data, | |
1960 | prctl_map->start_data)) | |
1961 | goto out; | |
1962 | ||
f606b77f CG |
1963 | error = 0; |
1964 | out: | |
1965 | return error; | |
1966 | } | |
1967 | ||
4a00e9df | 1968 | #ifdef CONFIG_CHECKPOINT_RESTORE |
f606b77f CG |
1969 | static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size) |
1970 | { | |
1971 | struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, }; | |
1972 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1973 | struct mm_struct *mm = current->mm; | |
1974 | int error; | |
1975 | ||
1976 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1977 | BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256); | |
1978 | ||
1979 | if (opt == PR_SET_MM_MAP_SIZE) | |
1980 | return put_user((unsigned int)sizeof(prctl_map), | |
1981 | (unsigned int __user *)addr); | |
1982 | ||
1983 | if (data_size != sizeof(prctl_map)) | |
1984 | return -EINVAL; | |
1985 | ||
1986 | if (copy_from_user(&prctl_map, addr, sizeof(prctl_map))) | |
1987 | return -EFAULT; | |
1988 | ||
11bbd8b4 | 1989 | error = validate_prctl_map_addr(&prctl_map); |
f606b77f CG |
1990 | if (error) |
1991 | return error; | |
1992 | ||
1993 | if (prctl_map.auxv_size) { | |
11bbd8b4 MK |
1994 | /* |
1995 | * Someone is trying to cheat the auxv vector. | |
1996 | */ | |
1997 | if (!prctl_map.auxv || | |
1998 | prctl_map.auxv_size > sizeof(mm->saved_auxv)) | |
1999 | return -EINVAL; | |
2000 | ||
f606b77f CG |
2001 | memset(user_auxv, 0, sizeof(user_auxv)); |
2002 | if (copy_from_user(user_auxv, | |
2003 | (const void __user *)prctl_map.auxv, | |
2004 | prctl_map.auxv_size)) | |
2005 | return -EFAULT; | |
2006 | ||
2007 | /* Last entry must be AT_NULL as specification requires */ | |
2008 | user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL; | |
2009 | user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL; | |
2010 | } | |
2011 | ||
ddf1d398 | 2012 | if (prctl_map.exe_fd != (u32)-1) { |
11bbd8b4 | 2013 | /* |
ebd6de68 NV |
2014 | * Check if the current user is checkpoint/restore capable. |
2015 | * At the time of this writing, it checks for CAP_SYS_ADMIN | |
2016 | * or CAP_CHECKPOINT_RESTORE. | |
2017 | * Note that a user with access to ptrace can masquerade an | |
2018 | * arbitrary program as any executable, even setuid ones. | |
2019 | * This may have implications in the tomoyo subsystem. | |
11bbd8b4 | 2020 | */ |
ebd6de68 | 2021 | if (!checkpoint_restore_ns_capable(current_user_ns())) |
227175b2 | 2022 | return -EPERM; |
11bbd8b4 | 2023 | |
6e399cd1 | 2024 | error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd); |
ddf1d398 MG |
2025 | if (error) |
2026 | return error; | |
2027 | } | |
2028 | ||
88aa7cc6 | 2029 | /* |
5afe69c2 | 2030 | * arg_lock protects concurrent updates but we still need mmap_lock for |
88aa7cc6 YS |
2031 | * read to exclude races with sys_brk. |
2032 | */ | |
d8ed45c5 | 2033 | mmap_read_lock(mm); |
f606b77f CG |
2034 | |
2035 | /* | |
2036 | * We don't validate if these members are pointing to | |
2037 | * real present VMAs because application may have correspond | |
2038 | * VMAs already unmapped and kernel uses these members for statistics | |
2039 | * output in procfs mostly, except | |
2040 | * | |
15ec0fcf | 2041 | * - @start_brk/@brk which are used in do_brk_flags but kernel lookups |
5afe69c2 | 2042 | * for VMAs when updating these members so anything wrong written |
f606b77f CG |
2043 | * here cause kernel to swear at userspace program but won't lead |
2044 | * to any problem in kernel itself | |
2045 | */ | |
2046 | ||
88aa7cc6 | 2047 | spin_lock(&mm->arg_lock); |
f606b77f CG |
2048 | mm->start_code = prctl_map.start_code; |
2049 | mm->end_code = prctl_map.end_code; | |
2050 | mm->start_data = prctl_map.start_data; | |
2051 | mm->end_data = prctl_map.end_data; | |
2052 | mm->start_brk = prctl_map.start_brk; | |
2053 | mm->brk = prctl_map.brk; | |
2054 | mm->start_stack = prctl_map.start_stack; | |
2055 | mm->arg_start = prctl_map.arg_start; | |
2056 | mm->arg_end = prctl_map.arg_end; | |
2057 | mm->env_start = prctl_map.env_start; | |
2058 | mm->env_end = prctl_map.env_end; | |
88aa7cc6 | 2059 | spin_unlock(&mm->arg_lock); |
f606b77f CG |
2060 | |
2061 | /* | |
2062 | * Note this update of @saved_auxv is lockless thus | |
2063 | * if someone reads this member in procfs while we're | |
2064 | * updating -- it may get partly updated results. It's | |
2065 | * known and acceptable trade off: we leave it as is to | |
2066 | * not introduce additional locks here making the kernel | |
2067 | * more complex. | |
2068 | */ | |
2069 | if (prctl_map.auxv_size) | |
2070 | memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv)); | |
2071 | ||
d8ed45c5 | 2072 | mmap_read_unlock(mm); |
ddf1d398 | 2073 | return 0; |
f606b77f CG |
2074 | } |
2075 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
2076 | ||
4a00e9df AD |
2077 | static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr, |
2078 | unsigned long len) | |
2079 | { | |
2080 | /* | |
2081 | * This doesn't move the auxiliary vector itself since it's pinned to | |
2082 | * mm_struct, but it permits filling the vector with new values. It's | |
2083 | * up to the caller to provide sane values here, otherwise userspace | |
2084 | * tools which use this vector might be unhappy. | |
2085 | */ | |
c995f12a | 2086 | unsigned long user_auxv[AT_VECTOR_SIZE] = {}; |
4a00e9df AD |
2087 | |
2088 | if (len > sizeof(user_auxv)) | |
2089 | return -EINVAL; | |
2090 | ||
2091 | if (copy_from_user(user_auxv, (const void __user *)addr, len)) | |
2092 | return -EFAULT; | |
2093 | ||
2094 | /* Make sure the last entry is always AT_NULL */ | |
2095 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
2096 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
2097 | ||
2098 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
2099 | ||
2100 | task_lock(current); | |
2101 | memcpy(mm->saved_auxv, user_auxv, len); | |
2102 | task_unlock(current); | |
2103 | ||
2104 | return 0; | |
2105 | } | |
2106 | ||
028ee4be CG |
2107 | static int prctl_set_mm(int opt, unsigned long addr, |
2108 | unsigned long arg4, unsigned long arg5) | |
2109 | { | |
028ee4be | 2110 | struct mm_struct *mm = current->mm; |
11bbd8b4 MK |
2111 | struct prctl_mm_map prctl_map = { |
2112 | .auxv = NULL, | |
2113 | .auxv_size = 0, | |
2114 | .exe_fd = -1, | |
2115 | }; | |
fe8c7f5c CG |
2116 | struct vm_area_struct *vma; |
2117 | int error; | |
028ee4be | 2118 | |
f606b77f CG |
2119 | if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV && |
2120 | opt != PR_SET_MM_MAP && | |
2121 | opt != PR_SET_MM_MAP_SIZE))) | |
028ee4be CG |
2122 | return -EINVAL; |
2123 | ||
f606b77f CG |
2124 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2125 | if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE) | |
2126 | return prctl_set_mm_map(opt, (const void __user *)addr, arg4); | |
2127 | #endif | |
2128 | ||
79f0713d | 2129 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
2130 | return -EPERM; |
2131 | ||
6e399cd1 DB |
2132 | if (opt == PR_SET_MM_EXE_FILE) |
2133 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
b32dfe37 | 2134 | |
4a00e9df AD |
2135 | if (opt == PR_SET_MM_AUXV) |
2136 | return prctl_set_auxv(mm, addr, arg4); | |
2137 | ||
1ad75b9e | 2138 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
2139 | return -EINVAL; |
2140 | ||
fe8c7f5c CG |
2141 | error = -EINVAL; |
2142 | ||
bc81426f | 2143 | /* |
5afe69c2 | 2144 | * arg_lock protects concurrent updates of arg boundaries, we need |
c1e8d7c6 | 2145 | * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr |
bc81426f MK |
2146 | * validation. |
2147 | */ | |
d8ed45c5 | 2148 | mmap_read_lock(mm); |
028ee4be CG |
2149 | vma = find_vma(mm, addr); |
2150 | ||
bc81426f | 2151 | spin_lock(&mm->arg_lock); |
4a00e9df AD |
2152 | prctl_map.start_code = mm->start_code; |
2153 | prctl_map.end_code = mm->end_code; | |
2154 | prctl_map.start_data = mm->start_data; | |
2155 | prctl_map.end_data = mm->end_data; | |
2156 | prctl_map.start_brk = mm->start_brk; | |
2157 | prctl_map.brk = mm->brk; | |
2158 | prctl_map.start_stack = mm->start_stack; | |
2159 | prctl_map.arg_start = mm->arg_start; | |
2160 | prctl_map.arg_end = mm->arg_end; | |
2161 | prctl_map.env_start = mm->env_start; | |
2162 | prctl_map.env_end = mm->env_end; | |
4a00e9df | 2163 | |
028ee4be CG |
2164 | switch (opt) { |
2165 | case PR_SET_MM_START_CODE: | |
4a00e9df | 2166 | prctl_map.start_code = addr; |
fe8c7f5c | 2167 | break; |
028ee4be | 2168 | case PR_SET_MM_END_CODE: |
4a00e9df | 2169 | prctl_map.end_code = addr; |
028ee4be | 2170 | break; |
028ee4be | 2171 | case PR_SET_MM_START_DATA: |
4a00e9df | 2172 | prctl_map.start_data = addr; |
028ee4be | 2173 | break; |
fe8c7f5c | 2174 | case PR_SET_MM_END_DATA: |
4a00e9df AD |
2175 | prctl_map.end_data = addr; |
2176 | break; | |
2177 | case PR_SET_MM_START_STACK: | |
2178 | prctl_map.start_stack = addr; | |
028ee4be | 2179 | break; |
028ee4be | 2180 | case PR_SET_MM_START_BRK: |
4a00e9df | 2181 | prctl_map.start_brk = addr; |
028ee4be | 2182 | break; |
028ee4be | 2183 | case PR_SET_MM_BRK: |
4a00e9df | 2184 | prctl_map.brk = addr; |
028ee4be | 2185 | break; |
4a00e9df AD |
2186 | case PR_SET_MM_ARG_START: |
2187 | prctl_map.arg_start = addr; | |
2188 | break; | |
2189 | case PR_SET_MM_ARG_END: | |
2190 | prctl_map.arg_end = addr; | |
2191 | break; | |
2192 | case PR_SET_MM_ENV_START: | |
2193 | prctl_map.env_start = addr; | |
2194 | break; | |
2195 | case PR_SET_MM_ENV_END: | |
2196 | prctl_map.env_end = addr; | |
2197 | break; | |
2198 | default: | |
2199 | goto out; | |
2200 | } | |
2201 | ||
11bbd8b4 | 2202 | error = validate_prctl_map_addr(&prctl_map); |
4a00e9df AD |
2203 | if (error) |
2204 | goto out; | |
028ee4be | 2205 | |
4a00e9df | 2206 | switch (opt) { |
fe8c7f5c CG |
2207 | /* |
2208 | * If command line arguments and environment | |
2209 | * are placed somewhere else on stack, we can | |
2210 | * set them up here, ARG_START/END to setup | |
5afe69c2 | 2211 | * command line arguments and ENV_START/END |
fe8c7f5c CG |
2212 | * for environment. |
2213 | */ | |
2214 | case PR_SET_MM_START_STACK: | |
2215 | case PR_SET_MM_ARG_START: | |
2216 | case PR_SET_MM_ARG_END: | |
2217 | case PR_SET_MM_ENV_START: | |
2218 | case PR_SET_MM_ENV_END: | |
2219 | if (!vma) { | |
2220 | error = -EFAULT; | |
2221 | goto out; | |
2222 | } | |
028ee4be CG |
2223 | } |
2224 | ||
4a00e9df AD |
2225 | mm->start_code = prctl_map.start_code; |
2226 | mm->end_code = prctl_map.end_code; | |
2227 | mm->start_data = prctl_map.start_data; | |
2228 | mm->end_data = prctl_map.end_data; | |
2229 | mm->start_brk = prctl_map.start_brk; | |
2230 | mm->brk = prctl_map.brk; | |
2231 | mm->start_stack = prctl_map.start_stack; | |
2232 | mm->arg_start = prctl_map.arg_start; | |
2233 | mm->arg_end = prctl_map.arg_end; | |
2234 | mm->env_start = prctl_map.env_start; | |
2235 | mm->env_end = prctl_map.env_end; | |
2236 | ||
028ee4be | 2237 | error = 0; |
028ee4be | 2238 | out: |
bc81426f | 2239 | spin_unlock(&mm->arg_lock); |
d8ed45c5 | 2240 | mmap_read_unlock(mm); |
028ee4be CG |
2241 | return error; |
2242 | } | |
300f786b | 2243 | |
52b36941 | 2244 | #ifdef CONFIG_CHECKPOINT_RESTORE |
986b9eac | 2245 | static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr) |
300f786b CG |
2246 | { |
2247 | return put_user(me->clear_child_tid, tid_addr); | |
2248 | } | |
52b36941 | 2249 | #else |
986b9eac | 2250 | static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr) |
300f786b CG |
2251 | { |
2252 | return -EINVAL; | |
2253 | } | |
028ee4be CG |
2254 | #endif |
2255 | ||
749860ce PT |
2256 | static int propagate_has_child_subreaper(struct task_struct *p, void *data) |
2257 | { | |
2258 | /* | |
5afe69c2 XC |
2259 | * If task has has_child_subreaper - all its descendants |
2260 | * already have these flag too and new descendants will | |
749860ce PT |
2261 | * inherit it on fork, skip them. |
2262 | * | |
2263 | * If we've found child_reaper - skip descendants in | |
2264 | * it's subtree as they will never get out pidns. | |
2265 | */ | |
2266 | if (p->signal->has_child_subreaper || | |
2267 | is_child_reaper(task_pid(p))) | |
2268 | return 0; | |
2269 | ||
2270 | p->signal->has_child_subreaper = 1; | |
2271 | return 1; | |
2272 | } | |
2273 | ||
7bbf1373 | 2274 | int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which) |
b617cfc8 TG |
2275 | { |
2276 | return -EINVAL; | |
2277 | } | |
2278 | ||
7bbf1373 KC |
2279 | int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which, |
2280 | unsigned long ctrl) | |
b617cfc8 TG |
2281 | { |
2282 | return -EINVAL; | |
2283 | } | |
2284 | ||
a37b0715 | 2285 | #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE) |
8d19f1c8 | 2286 | |
c4ea37c2 HC |
2287 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2288 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2289 | { |
b6dff3ec DH |
2290 | struct task_struct *me = current; |
2291 | unsigned char comm[sizeof(me->comm)]; | |
2292 | long error; | |
1da177e4 | 2293 | |
d84f4f99 DH |
2294 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2295 | if (error != -ENOSYS) | |
1da177e4 LT |
2296 | return error; |
2297 | ||
d84f4f99 | 2298 | error = 0; |
1da177e4 | 2299 | switch (option) { |
f3cbd435 AM |
2300 | case PR_SET_PDEATHSIG: |
2301 | if (!valid_signal(arg2)) { | |
2302 | error = -EINVAL; | |
1da177e4 | 2303 | break; |
f3cbd435 AM |
2304 | } |
2305 | me->pdeath_signal = arg2; | |
2306 | break; | |
2307 | case PR_GET_PDEATHSIG: | |
2308 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2309 | break; | |
2310 | case PR_GET_DUMPABLE: | |
2311 | error = get_dumpable(me->mm); | |
2312 | break; | |
2313 | case PR_SET_DUMPABLE: | |
2314 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2315 | error = -EINVAL; | |
1da177e4 | 2316 | break; |
f3cbd435 AM |
2317 | } |
2318 | set_dumpable(me->mm, arg2); | |
2319 | break; | |
1da177e4 | 2320 | |
f3cbd435 AM |
2321 | case PR_SET_UNALIGN: |
2322 | error = SET_UNALIGN_CTL(me, arg2); | |
2323 | break; | |
2324 | case PR_GET_UNALIGN: | |
2325 | error = GET_UNALIGN_CTL(me, arg2); | |
2326 | break; | |
2327 | case PR_SET_FPEMU: | |
2328 | error = SET_FPEMU_CTL(me, arg2); | |
2329 | break; | |
2330 | case PR_GET_FPEMU: | |
2331 | error = GET_FPEMU_CTL(me, arg2); | |
2332 | break; | |
2333 | case PR_SET_FPEXC: | |
2334 | error = SET_FPEXC_CTL(me, arg2); | |
2335 | break; | |
2336 | case PR_GET_FPEXC: | |
2337 | error = GET_FPEXC_CTL(me, arg2); | |
2338 | break; | |
2339 | case PR_GET_TIMING: | |
2340 | error = PR_TIMING_STATISTICAL; | |
2341 | break; | |
2342 | case PR_SET_TIMING: | |
2343 | if (arg2 != PR_TIMING_STATISTICAL) | |
2344 | error = -EINVAL; | |
2345 | break; | |
2346 | case PR_SET_NAME: | |
2347 | comm[sizeof(me->comm) - 1] = 0; | |
2348 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2349 | sizeof(me->comm) - 1) < 0) | |
2350 | return -EFAULT; | |
2351 | set_task_comm(me, comm); | |
2352 | proc_comm_connector(me); | |
2353 | break; | |
2354 | case PR_GET_NAME: | |
2355 | get_task_comm(comm, me); | |
2356 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2357 | return -EFAULT; | |
2358 | break; | |
2359 | case PR_GET_ENDIAN: | |
2360 | error = GET_ENDIAN(me, arg2); | |
2361 | break; | |
2362 | case PR_SET_ENDIAN: | |
2363 | error = SET_ENDIAN(me, arg2); | |
2364 | break; | |
2365 | case PR_GET_SECCOMP: | |
2366 | error = prctl_get_seccomp(); | |
2367 | break; | |
2368 | case PR_SET_SECCOMP: | |
2369 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2370 | break; | |
2371 | case PR_GET_TSC: | |
2372 | error = GET_TSC_CTL(arg2); | |
2373 | break; | |
2374 | case PR_SET_TSC: | |
2375 | error = SET_TSC_CTL(arg2); | |
2376 | break; | |
2377 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2378 | error = perf_event_task_disable(); | |
2379 | break; | |
2380 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2381 | error = perf_event_task_enable(); | |
2382 | break; | |
2383 | case PR_GET_TIMERSLACK: | |
da8b44d5 JS |
2384 | if (current->timer_slack_ns > ULONG_MAX) |
2385 | error = ULONG_MAX; | |
2386 | else | |
2387 | error = current->timer_slack_ns; | |
f3cbd435 AM |
2388 | break; |
2389 | case PR_SET_TIMERSLACK: | |
2390 | if (arg2 <= 0) | |
2391 | current->timer_slack_ns = | |
6976675d | 2392 | current->default_timer_slack_ns; |
f3cbd435 AM |
2393 | else |
2394 | current->timer_slack_ns = arg2; | |
2395 | break; | |
2396 | case PR_MCE_KILL: | |
2397 | if (arg4 | arg5) | |
2398 | return -EINVAL; | |
2399 | switch (arg2) { | |
2400 | case PR_MCE_KILL_CLEAR: | |
2401 | if (arg3 != 0) | |
4db96cf0 | 2402 | return -EINVAL; |
f3cbd435 | 2403 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2404 | break; |
f3cbd435 AM |
2405 | case PR_MCE_KILL_SET: |
2406 | current->flags |= PF_MCE_PROCESS; | |
2407 | if (arg3 == PR_MCE_KILL_EARLY) | |
2408 | current->flags |= PF_MCE_EARLY; | |
2409 | else if (arg3 == PR_MCE_KILL_LATE) | |
2410 | current->flags &= ~PF_MCE_EARLY; | |
2411 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2412 | current->flags &= | |
2413 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2414 | else |
259e5e6c | 2415 | return -EINVAL; |
259e5e6c | 2416 | break; |
1da177e4 | 2417 | default: |
f3cbd435 AM |
2418 | return -EINVAL; |
2419 | } | |
2420 | break; | |
2421 | case PR_MCE_KILL_GET: | |
2422 | if (arg2 | arg3 | arg4 | arg5) | |
2423 | return -EINVAL; | |
2424 | if (current->flags & PF_MCE_PROCESS) | |
2425 | error = (current->flags & PF_MCE_EARLY) ? | |
2426 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2427 | else | |
2428 | error = PR_MCE_KILL_DEFAULT; | |
2429 | break; | |
2430 | case PR_SET_MM: | |
2431 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2432 | break; | |
2433 | case PR_GET_TID_ADDRESS: | |
986b9eac | 2434 | error = prctl_get_tid_address(me, (int __user * __user *)arg2); |
f3cbd435 AM |
2435 | break; |
2436 | case PR_SET_CHILD_SUBREAPER: | |
2437 | me->signal->is_child_subreaper = !!arg2; | |
749860ce PT |
2438 | if (!arg2) |
2439 | break; | |
2440 | ||
2441 | walk_process_tree(me, propagate_has_child_subreaper, NULL); | |
f3cbd435 AM |
2442 | break; |
2443 | case PR_GET_CHILD_SUBREAPER: | |
2444 | error = put_user(me->signal->is_child_subreaper, | |
2445 | (int __user *)arg2); | |
2446 | break; | |
2447 | case PR_SET_NO_NEW_PRIVS: | |
2448 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2449 | return -EINVAL; | |
2450 | ||
1d4457f9 | 2451 | task_set_no_new_privs(current); |
f3cbd435 AM |
2452 | break; |
2453 | case PR_GET_NO_NEW_PRIVS: | |
2454 | if (arg2 || arg3 || arg4 || arg5) | |
2455 | return -EINVAL; | |
1d4457f9 | 2456 | return task_no_new_privs(current) ? 1 : 0; |
a0715cc2 AT |
2457 | case PR_GET_THP_DISABLE: |
2458 | if (arg2 || arg3 || arg4 || arg5) | |
2459 | return -EINVAL; | |
18600332 | 2460 | error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 AT |
2461 | break; |
2462 | case PR_SET_THP_DISABLE: | |
2463 | if (arg3 || arg4 || arg5) | |
2464 | return -EINVAL; | |
d8ed45c5 | 2465 | if (mmap_write_lock_killable(me->mm)) |
17b0573d | 2466 | return -EINTR; |
a0715cc2 | 2467 | if (arg2) |
18600332 | 2468 | set_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 | 2469 | else |
18600332 | 2470 | clear_bit(MMF_DISABLE_THP, &me->mm->flags); |
d8ed45c5 | 2471 | mmap_write_unlock(me->mm); |
a0715cc2 | 2472 | break; |
fe3d197f | 2473 | case PR_MPX_ENABLE_MANAGEMENT: |
fe3d197f | 2474 | case PR_MPX_DISABLE_MANAGEMENT: |
f240652b DH |
2475 | /* No longer implemented: */ |
2476 | return -EINVAL; | |
9791554b PB |
2477 | case PR_SET_FP_MODE: |
2478 | error = SET_FP_MODE(me, arg2); | |
2479 | break; | |
2480 | case PR_GET_FP_MODE: | |
2481 | error = GET_FP_MODE(me); | |
2482 | break; | |
2d2123bc DM |
2483 | case PR_SVE_SET_VL: |
2484 | error = SVE_SET_VL(arg2); | |
2485 | break; | |
2486 | case PR_SVE_GET_VL: | |
2487 | error = SVE_GET_VL(); | |
2488 | break; | |
b617cfc8 TG |
2489 | case PR_GET_SPECULATION_CTRL: |
2490 | if (arg3 || arg4 || arg5) | |
2491 | return -EINVAL; | |
7bbf1373 | 2492 | error = arch_prctl_spec_ctrl_get(me, arg2); |
b617cfc8 TG |
2493 | break; |
2494 | case PR_SET_SPECULATION_CTRL: | |
2495 | if (arg4 || arg5) | |
2496 | return -EINVAL; | |
7bbf1373 | 2497 | error = arch_prctl_spec_ctrl_set(me, arg2, arg3); |
b617cfc8 | 2498 | break; |
ba830885 KM |
2499 | case PR_PAC_RESET_KEYS: |
2500 | if (arg3 || arg4 || arg5) | |
2501 | return -EINVAL; | |
2502 | error = PAC_RESET_KEYS(me, arg2); | |
2503 | break; | |
20169862 PC |
2504 | case PR_PAC_SET_ENABLED_KEYS: |
2505 | if (arg4 || arg5) | |
2506 | return -EINVAL; | |
2507 | error = PAC_SET_ENABLED_KEYS(me, arg2, arg3); | |
2508 | break; | |
2509 | case PR_PAC_GET_ENABLED_KEYS: | |
2510 | if (arg2 || arg3 || arg4 || arg5) | |
2511 | return -EINVAL; | |
2512 | error = PAC_GET_ENABLED_KEYS(me); | |
2513 | break; | |
63f0c603 | 2514 | case PR_SET_TAGGED_ADDR_CTRL: |
3e91ec89 CM |
2515 | if (arg3 || arg4 || arg5) |
2516 | return -EINVAL; | |
63f0c603 CM |
2517 | error = SET_TAGGED_ADDR_CTRL(arg2); |
2518 | break; | |
2519 | case PR_GET_TAGGED_ADDR_CTRL: | |
3e91ec89 CM |
2520 | if (arg2 || arg3 || arg4 || arg5) |
2521 | return -EINVAL; | |
63f0c603 CM |
2522 | error = GET_TAGGED_ADDR_CTRL(); |
2523 | break; | |
8d19f1c8 MC |
2524 | case PR_SET_IO_FLUSHER: |
2525 | if (!capable(CAP_SYS_RESOURCE)) | |
2526 | return -EPERM; | |
2527 | ||
2528 | if (arg3 || arg4 || arg5) | |
2529 | return -EINVAL; | |
2530 | ||
2531 | if (arg2 == 1) | |
2532 | current->flags |= PR_IO_FLUSHER; | |
2533 | else if (!arg2) | |
2534 | current->flags &= ~PR_IO_FLUSHER; | |
2535 | else | |
2536 | return -EINVAL; | |
2537 | break; | |
2538 | case PR_GET_IO_FLUSHER: | |
2539 | if (!capable(CAP_SYS_RESOURCE)) | |
2540 | return -EPERM; | |
2541 | ||
2542 | if (arg2 || arg3 || arg4 || arg5) | |
2543 | return -EINVAL; | |
2544 | ||
2545 | error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER; | |
2546 | break; | |
1446e1df GKB |
2547 | case PR_SET_SYSCALL_USER_DISPATCH: |
2548 | error = set_syscall_user_dispatch(arg2, arg3, arg4, | |
2549 | (char __user *) arg5); | |
2550 | break; | |
7ac592aa CH |
2551 | #ifdef CONFIG_SCHED_CORE |
2552 | case PR_SCHED_CORE: | |
2553 | error = sched_core_share_pid(arg2, arg3, arg4, arg5); | |
2554 | break; | |
2555 | #endif | |
f3cbd435 AM |
2556 | default: |
2557 | error = -EINVAL; | |
2558 | break; | |
1da177e4 LT |
2559 | } |
2560 | return error; | |
2561 | } | |
3cfc348b | 2562 | |
836f92ad HC |
2563 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2564 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2565 | { |
2566 | int err = 0; | |
2567 | int cpu = raw_smp_processor_id(); | |
ec94fc3d | 2568 | |
3cfc348b AK |
2569 | if (cpup) |
2570 | err |= put_user(cpu, cpup); | |
2571 | if (nodep) | |
2572 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2573 | return err ? -EFAULT : 0; |
2574 | } | |
10a0a8d4 | 2575 | |
4a22f166 SR |
2576 | /** |
2577 | * do_sysinfo - fill in sysinfo struct | |
2578 | * @info: pointer to buffer to fill | |
2579 | */ | |
2580 | static int do_sysinfo(struct sysinfo *info) | |
2581 | { | |
2582 | unsigned long mem_total, sav_total; | |
2583 | unsigned int mem_unit, bitcount; | |
dc1b7b6c | 2584 | struct timespec64 tp; |
4a22f166 SR |
2585 | |
2586 | memset(info, 0, sizeof(struct sysinfo)); | |
2587 | ||
dc1b7b6c | 2588 | ktime_get_boottime_ts64(&tp); |
ecc421e0 | 2589 | timens_add_boottime(&tp); |
4a22f166 SR |
2590 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
2591 | ||
2592 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2593 | ||
2594 | info->procs = nr_threads; | |
2595 | ||
2596 | si_meminfo(info); | |
2597 | si_swapinfo(info); | |
2598 | ||
2599 | /* | |
2600 | * If the sum of all the available memory (i.e. ram + swap) | |
2601 | * is less than can be stored in a 32 bit unsigned long then | |
2602 | * we can be binary compatible with 2.2.x kernels. If not, | |
2603 | * well, in that case 2.2.x was broken anyways... | |
2604 | * | |
2605 | * -Erik Andersen <andersee@debian.org> | |
2606 | */ | |
2607 | ||
2608 | mem_total = info->totalram + info->totalswap; | |
2609 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2610 | goto out; | |
2611 | bitcount = 0; | |
2612 | mem_unit = info->mem_unit; | |
2613 | while (mem_unit > 1) { | |
2614 | bitcount++; | |
2615 | mem_unit >>= 1; | |
2616 | sav_total = mem_total; | |
2617 | mem_total <<= 1; | |
2618 | if (mem_total < sav_total) | |
2619 | goto out; | |
2620 | } | |
2621 | ||
2622 | /* | |
2623 | * If mem_total did not overflow, multiply all memory values by | |
2624 | * info->mem_unit and set it to 1. This leaves things compatible | |
2625 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2626 | * kernels... | |
2627 | */ | |
2628 | ||
2629 | info->mem_unit = 1; | |
2630 | info->totalram <<= bitcount; | |
2631 | info->freeram <<= bitcount; | |
2632 | info->sharedram <<= bitcount; | |
2633 | info->bufferram <<= bitcount; | |
2634 | info->totalswap <<= bitcount; | |
2635 | info->freeswap <<= bitcount; | |
2636 | info->totalhigh <<= bitcount; | |
2637 | info->freehigh <<= bitcount; | |
2638 | ||
2639 | out: | |
2640 | return 0; | |
2641 | } | |
2642 | ||
2643 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2644 | { | |
2645 | struct sysinfo val; | |
2646 | ||
2647 | do_sysinfo(&val); | |
2648 | ||
2649 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2650 | return -EFAULT; | |
2651 | ||
2652 | return 0; | |
2653 | } | |
2654 | ||
2655 | #ifdef CONFIG_COMPAT | |
2656 | struct compat_sysinfo { | |
2657 | s32 uptime; | |
2658 | u32 loads[3]; | |
2659 | u32 totalram; | |
2660 | u32 freeram; | |
2661 | u32 sharedram; | |
2662 | u32 bufferram; | |
2663 | u32 totalswap; | |
2664 | u32 freeswap; | |
2665 | u16 procs; | |
2666 | u16 pad; | |
2667 | u32 totalhigh; | |
2668 | u32 freehigh; | |
2669 | u32 mem_unit; | |
2670 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2671 | }; | |
2672 | ||
2673 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2674 | { | |
2675 | struct sysinfo s; | |
ce5155c4 | 2676 | struct compat_sysinfo s_32; |
4a22f166 SR |
2677 | |
2678 | do_sysinfo(&s); | |
2679 | ||
2680 | /* Check to see if any memory value is too large for 32-bit and scale | |
2681 | * down if needed | |
2682 | */ | |
0baae41e | 2683 | if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) { |
4a22f166 SR |
2684 | int bitcount = 0; |
2685 | ||
2686 | while (s.mem_unit < PAGE_SIZE) { | |
2687 | s.mem_unit <<= 1; | |
2688 | bitcount++; | |
2689 | } | |
2690 | ||
2691 | s.totalram >>= bitcount; | |
2692 | s.freeram >>= bitcount; | |
2693 | s.sharedram >>= bitcount; | |
2694 | s.bufferram >>= bitcount; | |
2695 | s.totalswap >>= bitcount; | |
2696 | s.freeswap >>= bitcount; | |
2697 | s.totalhigh >>= bitcount; | |
2698 | s.freehigh >>= bitcount; | |
2699 | } | |
2700 | ||
ce5155c4 AV |
2701 | memset(&s_32, 0, sizeof(s_32)); |
2702 | s_32.uptime = s.uptime; | |
2703 | s_32.loads[0] = s.loads[0]; | |
2704 | s_32.loads[1] = s.loads[1]; | |
2705 | s_32.loads[2] = s.loads[2]; | |
2706 | s_32.totalram = s.totalram; | |
2707 | s_32.freeram = s.freeram; | |
2708 | s_32.sharedram = s.sharedram; | |
2709 | s_32.bufferram = s.bufferram; | |
2710 | s_32.totalswap = s.totalswap; | |
2711 | s_32.freeswap = s.freeswap; | |
2712 | s_32.procs = s.procs; | |
2713 | s_32.totalhigh = s.totalhigh; | |
2714 | s_32.freehigh = s.freehigh; | |
2715 | s_32.mem_unit = s.mem_unit; | |
2716 | if (copy_to_user(info, &s_32, sizeof(s_32))) | |
4a22f166 | 2717 | return -EFAULT; |
4a22f166 SR |
2718 | return 0; |
2719 | } | |
2720 | #endif /* CONFIG_COMPAT */ |