]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/sys.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
9984de1a | 8 | #include <linux/export.h> |
1da177e4 LT |
9 | #include <linux/mm.h> |
10 | #include <linux/utsname.h> | |
11 | #include <linux/mman.h> | |
1da177e4 LT |
12 | #include <linux/reboot.h> |
13 | #include <linux/prctl.h> | |
1da177e4 LT |
14 | #include <linux/highuid.h> |
15 | #include <linux/fs.h> | |
74da1ff7 | 16 | #include <linux/kmod.h> |
cdd6c482 | 17 | #include <linux/perf_event.h> |
3e88c553 | 18 | #include <linux/resource.h> |
dc009d92 | 19 | #include <linux/kernel.h> |
1da177e4 | 20 | #include <linux/workqueue.h> |
c59ede7b | 21 | #include <linux/capability.h> |
1da177e4 LT |
22 | #include <linux/device.h> |
23 | #include <linux/key.h> | |
24 | #include <linux/times.h> | |
25 | #include <linux/posix-timers.h> | |
26 | #include <linux/security.h> | |
27 | #include <linux/dcookies.h> | |
28 | #include <linux/suspend.h> | |
29 | #include <linux/tty.h> | |
7ed20e1a | 30 | #include <linux/signal.h> |
9f46080c | 31 | #include <linux/cn_proc.h> |
3cfc348b | 32 | #include <linux/getcpu.h> |
6eaeeaba | 33 | #include <linux/task_io_accounting_ops.h> |
1d9d02fe | 34 | #include <linux/seccomp.h> |
4047727e | 35 | #include <linux/cpu.h> |
e28cbf22 | 36 | #include <linux/personality.h> |
e3d5a27d | 37 | #include <linux/ptrace.h> |
5ad4e53b | 38 | #include <linux/fs_struct.h> |
b32dfe37 CG |
39 | #include <linux/file.h> |
40 | #include <linux/mount.h> | |
5a0e3ad6 | 41 | #include <linux/gfp.h> |
40dc166c | 42 | #include <linux/syscore_ops.h> |
be27425d AK |
43 | #include <linux/version.h> |
44 | #include <linux/ctype.h> | |
1446e1df | 45 | #include <linux/syscall_user_dispatch.h> |
1da177e4 LT |
46 | |
47 | #include <linux/compat.h> | |
48 | #include <linux/syscalls.h> | |
00d7c05a | 49 | #include <linux/kprobes.h> |
acce292c | 50 | #include <linux/user_namespace.h> |
ecc421e0 | 51 | #include <linux/time_namespace.h> |
7fe5e042 | 52 | #include <linux/binfmts.h> |
1da177e4 | 53 | |
4a22f166 | 54 | #include <linux/sched.h> |
4eb5aaa3 | 55 | #include <linux/sched/autogroup.h> |
4f17722c | 56 | #include <linux/sched/loadavg.h> |
03441a34 | 57 | #include <linux/sched/stat.h> |
6e84f315 | 58 | #include <linux/sched/mm.h> |
f7ccbae4 | 59 | #include <linux/sched/coredump.h> |
29930025 | 60 | #include <linux/sched/task.h> |
32ef5517 | 61 | #include <linux/sched/cputime.h> |
4a22f166 SR |
62 | #include <linux/rcupdate.h> |
63 | #include <linux/uidgid.h> | |
64 | #include <linux/cred.h> | |
65 | ||
b617cfc8 TG |
66 | #include <linux/nospec.h> |
67 | ||
04c6862c | 68 | #include <linux/kmsg_dump.h> |
be27425d AK |
69 | /* Move somewhere else to avoid recompiling? */ |
70 | #include <generated/utsrelease.h> | |
04c6862c | 71 | |
7c0f6ba6 | 72 | #include <linux/uaccess.h> |
1da177e4 LT |
73 | #include <asm/io.h> |
74 | #include <asm/unistd.h> | |
75 | ||
e530dca5 DB |
76 | #include "uid16.h" |
77 | ||
1da177e4 | 78 | #ifndef SET_UNALIGN_CTL |
ec94fc3d | 79 | # define SET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
80 | #endif |
81 | #ifndef GET_UNALIGN_CTL | |
ec94fc3d | 82 | # define GET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
83 | #endif |
84 | #ifndef SET_FPEMU_CTL | |
ec94fc3d | 85 | # define SET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
86 | #endif |
87 | #ifndef GET_FPEMU_CTL | |
ec94fc3d | 88 | # define GET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
89 | #endif |
90 | #ifndef SET_FPEXC_CTL | |
ec94fc3d | 91 | # define SET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 LT |
92 | #endif |
93 | #ifndef GET_FPEXC_CTL | |
ec94fc3d | 94 | # define GET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 | 95 | #endif |
651d765d | 96 | #ifndef GET_ENDIAN |
ec94fc3d | 97 | # define GET_ENDIAN(a, b) (-EINVAL) |
651d765d AB |
98 | #endif |
99 | #ifndef SET_ENDIAN | |
ec94fc3d | 100 | # define SET_ENDIAN(a, b) (-EINVAL) |
651d765d | 101 | #endif |
8fb402bc EB |
102 | #ifndef GET_TSC_CTL |
103 | # define GET_TSC_CTL(a) (-EINVAL) | |
104 | #endif | |
105 | #ifndef SET_TSC_CTL | |
106 | # define SET_TSC_CTL(a) (-EINVAL) | |
107 | #endif | |
9791554b PB |
108 | #ifndef GET_FP_MODE |
109 | # define GET_FP_MODE(a) (-EINVAL) | |
110 | #endif | |
111 | #ifndef SET_FP_MODE | |
112 | # define SET_FP_MODE(a,b) (-EINVAL) | |
113 | #endif | |
2d2123bc DM |
114 | #ifndef SVE_SET_VL |
115 | # define SVE_SET_VL(a) (-EINVAL) | |
116 | #endif | |
117 | #ifndef SVE_GET_VL | |
118 | # define SVE_GET_VL() (-EINVAL) | |
119 | #endif | |
ba830885 KM |
120 | #ifndef PAC_RESET_KEYS |
121 | # define PAC_RESET_KEYS(a, b) (-EINVAL) | |
122 | #endif | |
63f0c603 CM |
123 | #ifndef SET_TAGGED_ADDR_CTRL |
124 | # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL) | |
125 | #endif | |
126 | #ifndef GET_TAGGED_ADDR_CTRL | |
127 | # define GET_TAGGED_ADDR_CTRL() (-EINVAL) | |
128 | #endif | |
1da177e4 LT |
129 | |
130 | /* | |
131 | * this is where the system-wide overflow UID and GID are defined, for | |
132 | * architectures that now have 32-bit UID/GID but didn't in the past | |
133 | */ | |
134 | ||
135 | int overflowuid = DEFAULT_OVERFLOWUID; | |
136 | int overflowgid = DEFAULT_OVERFLOWGID; | |
137 | ||
1da177e4 LT |
138 | EXPORT_SYMBOL(overflowuid); |
139 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
140 | |
141 | /* | |
142 | * the same as above, but for filesystems which can only store a 16-bit | |
143 | * UID and GID. as such, this is needed on all architectures | |
144 | */ | |
145 | ||
146 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
8b2770a4 | 147 | int fs_overflowgid = DEFAULT_FS_OVERFLOWGID; |
1da177e4 LT |
148 | |
149 | EXPORT_SYMBOL(fs_overflowuid); | |
150 | EXPORT_SYMBOL(fs_overflowgid); | |
151 | ||
fc832ad3 SH |
152 | /* |
153 | * Returns true if current's euid is same as p's uid or euid, | |
154 | * or has CAP_SYS_NICE to p's user_ns. | |
155 | * | |
156 | * Called with rcu_read_lock, creds are safe | |
157 | */ | |
158 | static bool set_one_prio_perm(struct task_struct *p) | |
159 | { | |
160 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
161 | ||
5af66203 EB |
162 | if (uid_eq(pcred->uid, cred->euid) || |
163 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 164 | return true; |
c4a4d603 | 165 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
166 | return true; |
167 | return false; | |
168 | } | |
169 | ||
c69e8d9c DH |
170 | /* |
171 | * set the priority of a task | |
172 | * - the caller must hold the RCU read lock | |
173 | */ | |
1da177e4 LT |
174 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
175 | { | |
176 | int no_nice; | |
177 | ||
fc832ad3 | 178 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
179 | error = -EPERM; |
180 | goto out; | |
181 | } | |
e43379f1 | 182 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
183 | error = -EACCES; |
184 | goto out; | |
185 | } | |
186 | no_nice = security_task_setnice(p, niceval); | |
187 | if (no_nice) { | |
188 | error = no_nice; | |
189 | goto out; | |
190 | } | |
191 | if (error == -ESRCH) | |
192 | error = 0; | |
193 | set_user_nice(p, niceval); | |
194 | out: | |
195 | return error; | |
196 | } | |
197 | ||
754fe8d2 | 198 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
199 | { |
200 | struct task_struct *g, *p; | |
201 | struct user_struct *user; | |
86a264ab | 202 | const struct cred *cred = current_cred(); |
1da177e4 | 203 | int error = -EINVAL; |
41487c65 | 204 | struct pid *pgrp; |
7b44ab97 | 205 | kuid_t uid; |
1da177e4 | 206 | |
3e88c553 | 207 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
208 | goto out; |
209 | ||
210 | /* normalize: avoid signed division (rounding problems) */ | |
211 | error = -ESRCH; | |
c4a4d2f4 DY |
212 | if (niceval < MIN_NICE) |
213 | niceval = MIN_NICE; | |
214 | if (niceval > MAX_NICE) | |
215 | niceval = MAX_NICE; | |
1da177e4 | 216 | |
d4581a23 | 217 | rcu_read_lock(); |
1da177e4 LT |
218 | read_lock(&tasklist_lock); |
219 | switch (which) { | |
ec94fc3d | 220 | case PRIO_PROCESS: |
221 | if (who) | |
222 | p = find_task_by_vpid(who); | |
223 | else | |
224 | p = current; | |
225 | if (p) | |
226 | error = set_one_prio(p, niceval, error); | |
227 | break; | |
228 | case PRIO_PGRP: | |
229 | if (who) | |
230 | pgrp = find_vpid(who); | |
231 | else | |
232 | pgrp = task_pgrp(current); | |
233 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
234 | error = set_one_prio(p, niceval, error); | |
235 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
236 | break; | |
237 | case PRIO_USER: | |
238 | uid = make_kuid(cred->user_ns, who); | |
239 | user = cred->user; | |
240 | if (!who) | |
241 | uid = cred->uid; | |
242 | else if (!uid_eq(uid, cred->uid)) { | |
243 | user = find_user(uid); | |
244 | if (!user) | |
86a264ab | 245 | goto out_unlock; /* No processes for this user */ |
ec94fc3d | 246 | } |
247 | do_each_thread(g, p) { | |
8639b461 | 248 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) |
ec94fc3d | 249 | error = set_one_prio(p, niceval, error); |
250 | } while_each_thread(g, p); | |
251 | if (!uid_eq(uid, cred->uid)) | |
252 | free_uid(user); /* For find_user() */ | |
253 | break; | |
1da177e4 LT |
254 | } |
255 | out_unlock: | |
256 | read_unlock(&tasklist_lock); | |
d4581a23 | 257 | rcu_read_unlock(); |
1da177e4 LT |
258 | out: |
259 | return error; | |
260 | } | |
261 | ||
262 | /* | |
263 | * Ugh. To avoid negative return values, "getpriority()" will | |
264 | * not return the normal nice-value, but a negated value that | |
265 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
266 | * to stay compatible. | |
267 | */ | |
754fe8d2 | 268 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
269 | { |
270 | struct task_struct *g, *p; | |
271 | struct user_struct *user; | |
86a264ab | 272 | const struct cred *cred = current_cred(); |
1da177e4 | 273 | long niceval, retval = -ESRCH; |
41487c65 | 274 | struct pid *pgrp; |
7b44ab97 | 275 | kuid_t uid; |
1da177e4 | 276 | |
3e88c553 | 277 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
278 | return -EINVAL; |
279 | ||
70118837 | 280 | rcu_read_lock(); |
1da177e4 LT |
281 | read_lock(&tasklist_lock); |
282 | switch (which) { | |
ec94fc3d | 283 | case PRIO_PROCESS: |
284 | if (who) | |
285 | p = find_task_by_vpid(who); | |
286 | else | |
287 | p = current; | |
288 | if (p) { | |
289 | niceval = nice_to_rlimit(task_nice(p)); | |
290 | if (niceval > retval) | |
291 | retval = niceval; | |
292 | } | |
293 | break; | |
294 | case PRIO_PGRP: | |
295 | if (who) | |
296 | pgrp = find_vpid(who); | |
297 | else | |
298 | pgrp = task_pgrp(current); | |
299 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
300 | niceval = nice_to_rlimit(task_nice(p)); | |
301 | if (niceval > retval) | |
302 | retval = niceval; | |
303 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
304 | break; | |
305 | case PRIO_USER: | |
306 | uid = make_kuid(cred->user_ns, who); | |
307 | user = cred->user; | |
308 | if (!who) | |
309 | uid = cred->uid; | |
310 | else if (!uid_eq(uid, cred->uid)) { | |
311 | user = find_user(uid); | |
312 | if (!user) | |
313 | goto out_unlock; /* No processes for this user */ | |
314 | } | |
315 | do_each_thread(g, p) { | |
8639b461 | 316 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) { |
7aa2c016 | 317 | niceval = nice_to_rlimit(task_nice(p)); |
1da177e4 LT |
318 | if (niceval > retval) |
319 | retval = niceval; | |
320 | } | |
ec94fc3d | 321 | } while_each_thread(g, p); |
322 | if (!uid_eq(uid, cred->uid)) | |
323 | free_uid(user); /* for find_user() */ | |
324 | break; | |
1da177e4 LT |
325 | } |
326 | out_unlock: | |
327 | read_unlock(&tasklist_lock); | |
70118837 | 328 | rcu_read_unlock(); |
1da177e4 LT |
329 | |
330 | return retval; | |
331 | } | |
332 | ||
1da177e4 LT |
333 | /* |
334 | * Unprivileged users may change the real gid to the effective gid | |
335 | * or vice versa. (BSD-style) | |
336 | * | |
337 | * If you set the real gid at all, or set the effective gid to a value not | |
338 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
339 | * | |
340 | * This makes it possible for a setgid program to completely drop its | |
341 | * privileges, which is often a useful assertion to make when you are doing | |
342 | * a security audit over a program. | |
343 | * | |
344 | * The general idea is that a program which uses just setregid() will be | |
345 | * 100% compatible with BSD. A program which uses just setgid() will be | |
ec94fc3d | 346 | * 100% compatible with POSIX with saved IDs. |
1da177e4 LT |
347 | * |
348 | * SMP: There are not races, the GIDs are checked only by filesystem | |
349 | * operations (as far as semantic preservation is concerned). | |
350 | */ | |
2813893f | 351 | #ifdef CONFIG_MULTIUSER |
e530dca5 | 352 | long __sys_setregid(gid_t rgid, gid_t egid) |
1da177e4 | 353 | { |
a29c33f4 | 354 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
355 | const struct cred *old; |
356 | struct cred *new; | |
1da177e4 | 357 | int retval; |
a29c33f4 EB |
358 | kgid_t krgid, kegid; |
359 | ||
360 | krgid = make_kgid(ns, rgid); | |
361 | kegid = make_kgid(ns, egid); | |
362 | ||
363 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
364 | return -EINVAL; | |
365 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
366 | return -EINVAL; | |
1da177e4 | 367 | |
d84f4f99 DH |
368 | new = prepare_creds(); |
369 | if (!new) | |
370 | return -ENOMEM; | |
371 | old = current_cred(); | |
372 | ||
d84f4f99 | 373 | retval = -EPERM; |
1da177e4 | 374 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
375 | if (gid_eq(old->gid, krgid) || |
376 | gid_eq(old->egid, krgid) || | |
111767c1 | 377 | ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 | 378 | new->gid = krgid; |
1da177e4 | 379 | else |
d84f4f99 | 380 | goto error; |
1da177e4 LT |
381 | } |
382 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
383 | if (gid_eq(old->gid, kegid) || |
384 | gid_eq(old->egid, kegid) || | |
385 | gid_eq(old->sgid, kegid) || | |
111767c1 | 386 | ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 | 387 | new->egid = kegid; |
756184b7 | 388 | else |
d84f4f99 | 389 | goto error; |
1da177e4 | 390 | } |
d84f4f99 | 391 | |
1da177e4 | 392 | if (rgid != (gid_t) -1 || |
a29c33f4 | 393 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
394 | new->sgid = new->egid; |
395 | new->fsgid = new->egid; | |
396 | ||
39030e13 TC |
397 | retval = security_task_fix_setgid(new, old, LSM_SETID_RE); |
398 | if (retval < 0) | |
399 | goto error; | |
400 | ||
d84f4f99 DH |
401 | return commit_creds(new); |
402 | ||
403 | error: | |
404 | abort_creds(new); | |
405 | return retval; | |
1da177e4 LT |
406 | } |
407 | ||
e530dca5 DB |
408 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
409 | { | |
410 | return __sys_setregid(rgid, egid); | |
411 | } | |
412 | ||
1da177e4 | 413 | /* |
ec94fc3d | 414 | * setgid() is implemented like SysV w/ SAVED_IDS |
1da177e4 LT |
415 | * |
416 | * SMP: Same implicit races as above. | |
417 | */ | |
e530dca5 | 418 | long __sys_setgid(gid_t gid) |
1da177e4 | 419 | { |
a29c33f4 | 420 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
421 | const struct cred *old; |
422 | struct cred *new; | |
1da177e4 | 423 | int retval; |
a29c33f4 EB |
424 | kgid_t kgid; |
425 | ||
426 | kgid = make_kgid(ns, gid); | |
427 | if (!gid_valid(kgid)) | |
428 | return -EINVAL; | |
1da177e4 | 429 | |
d84f4f99 DH |
430 | new = prepare_creds(); |
431 | if (!new) | |
432 | return -ENOMEM; | |
433 | old = current_cred(); | |
434 | ||
d84f4f99 | 435 | retval = -EPERM; |
111767c1 | 436 | if (ns_capable_setid(old->user_ns, CAP_SETGID)) |
a29c33f4 EB |
437 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
438 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
439 | new->egid = new->fsgid = kgid; | |
1da177e4 | 440 | else |
d84f4f99 | 441 | goto error; |
1da177e4 | 442 | |
39030e13 TC |
443 | retval = security_task_fix_setgid(new, old, LSM_SETID_ID); |
444 | if (retval < 0) | |
445 | goto error; | |
446 | ||
d84f4f99 DH |
447 | return commit_creds(new); |
448 | ||
449 | error: | |
450 | abort_creds(new); | |
451 | return retval; | |
1da177e4 | 452 | } |
54e99124 | 453 | |
e530dca5 DB |
454 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
455 | { | |
456 | return __sys_setgid(gid); | |
457 | } | |
458 | ||
d84f4f99 DH |
459 | /* |
460 | * change the user struct in a credentials set to match the new UID | |
461 | */ | |
462 | static int set_user(struct cred *new) | |
1da177e4 LT |
463 | { |
464 | struct user_struct *new_user; | |
465 | ||
078de5f7 | 466 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
467 | if (!new_user) |
468 | return -EAGAIN; | |
469 | ||
72fa5997 VK |
470 | /* |
471 | * We don't fail in case of NPROC limit excess here because too many | |
472 | * poorly written programs don't check set*uid() return code, assuming | |
473 | * it never fails if called by root. We may still enforce NPROC limit | |
474 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
475 | * failure to the execve() stage. | |
476 | */ | |
78d7d407 | 477 | if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) && |
72fa5997 VK |
478 | new_user != INIT_USER) |
479 | current->flags |= PF_NPROC_EXCEEDED; | |
480 | else | |
481 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 482 | |
d84f4f99 DH |
483 | free_uid(new->user); |
484 | new->user = new_user; | |
1da177e4 LT |
485 | return 0; |
486 | } | |
487 | ||
488 | /* | |
489 | * Unprivileged users may change the real uid to the effective uid | |
490 | * or vice versa. (BSD-style) | |
491 | * | |
492 | * If you set the real uid at all, or set the effective uid to a value not | |
493 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
494 | * | |
495 | * This makes it possible for a setuid program to completely drop its | |
496 | * privileges, which is often a useful assertion to make when you are doing | |
497 | * a security audit over a program. | |
498 | * | |
499 | * The general idea is that a program which uses just setreuid() will be | |
500 | * 100% compatible with BSD. A program which uses just setuid() will be | |
ec94fc3d | 501 | * 100% compatible with POSIX with saved IDs. |
1da177e4 | 502 | */ |
e530dca5 | 503 | long __sys_setreuid(uid_t ruid, uid_t euid) |
1da177e4 | 504 | { |
a29c33f4 | 505 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
506 | const struct cred *old; |
507 | struct cred *new; | |
1da177e4 | 508 | int retval; |
a29c33f4 EB |
509 | kuid_t kruid, keuid; |
510 | ||
511 | kruid = make_kuid(ns, ruid); | |
512 | keuid = make_kuid(ns, euid); | |
513 | ||
514 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
515 | return -EINVAL; | |
516 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
517 | return -EINVAL; | |
1da177e4 | 518 | |
d84f4f99 DH |
519 | new = prepare_creds(); |
520 | if (!new) | |
521 | return -ENOMEM; | |
522 | old = current_cred(); | |
523 | ||
d84f4f99 | 524 | retval = -EPERM; |
1da177e4 | 525 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
526 | new->uid = kruid; |
527 | if (!uid_eq(old->uid, kruid) && | |
528 | !uid_eq(old->euid, kruid) && | |
40852275 | 529 | !ns_capable_setid(old->user_ns, CAP_SETUID)) |
d84f4f99 | 530 | goto error; |
1da177e4 LT |
531 | } |
532 | ||
533 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
534 | new->euid = keuid; |
535 | if (!uid_eq(old->uid, keuid) && | |
536 | !uid_eq(old->euid, keuid) && | |
537 | !uid_eq(old->suid, keuid) && | |
40852275 | 538 | !ns_capable_setid(old->user_ns, CAP_SETUID)) |
d84f4f99 | 539 | goto error; |
1da177e4 LT |
540 | } |
541 | ||
a29c33f4 | 542 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
543 | retval = set_user(new); |
544 | if (retval < 0) | |
545 | goto error; | |
546 | } | |
1da177e4 | 547 | if (ruid != (uid_t) -1 || |
a29c33f4 | 548 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
549 | new->suid = new->euid; |
550 | new->fsuid = new->euid; | |
1da177e4 | 551 | |
d84f4f99 DH |
552 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
553 | if (retval < 0) | |
554 | goto error; | |
1da177e4 | 555 | |
d84f4f99 | 556 | return commit_creds(new); |
1da177e4 | 557 | |
d84f4f99 DH |
558 | error: |
559 | abort_creds(new); | |
560 | return retval; | |
561 | } | |
ec94fc3d | 562 | |
e530dca5 DB |
563 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
564 | { | |
565 | return __sys_setreuid(ruid, euid); | |
566 | } | |
567 | ||
1da177e4 | 568 | /* |
ec94fc3d | 569 | * setuid() is implemented like SysV with SAVED_IDS |
570 | * | |
1da177e4 | 571 | * Note that SAVED_ID's is deficient in that a setuid root program |
ec94fc3d | 572 | * like sendmail, for example, cannot set its uid to be a normal |
1da177e4 LT |
573 | * user and then switch back, because if you're root, setuid() sets |
574 | * the saved uid too. If you don't like this, blame the bright people | |
575 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
576 | * will allow a root program to temporarily drop privileges and be able to | |
ec94fc3d | 577 | * regain them by swapping the real and effective uid. |
1da177e4 | 578 | */ |
e530dca5 | 579 | long __sys_setuid(uid_t uid) |
1da177e4 | 580 | { |
a29c33f4 | 581 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
582 | const struct cred *old; |
583 | struct cred *new; | |
1da177e4 | 584 | int retval; |
a29c33f4 EB |
585 | kuid_t kuid; |
586 | ||
587 | kuid = make_kuid(ns, uid); | |
588 | if (!uid_valid(kuid)) | |
589 | return -EINVAL; | |
1da177e4 | 590 | |
d84f4f99 DH |
591 | new = prepare_creds(); |
592 | if (!new) | |
593 | return -ENOMEM; | |
594 | old = current_cred(); | |
595 | ||
d84f4f99 | 596 | retval = -EPERM; |
40852275 | 597 | if (ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
598 | new->suid = new->uid = kuid; |
599 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
600 | retval = set_user(new); |
601 | if (retval < 0) | |
602 | goto error; | |
d84f4f99 | 603 | } |
a29c33f4 | 604 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 605 | goto error; |
1da177e4 | 606 | } |
1da177e4 | 607 | |
a29c33f4 | 608 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
609 | |
610 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
611 | if (retval < 0) | |
612 | goto error; | |
1da177e4 | 613 | |
d84f4f99 | 614 | return commit_creds(new); |
1da177e4 | 615 | |
d84f4f99 DH |
616 | error: |
617 | abort_creds(new); | |
618 | return retval; | |
1da177e4 LT |
619 | } |
620 | ||
e530dca5 DB |
621 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
622 | { | |
623 | return __sys_setuid(uid); | |
624 | } | |
625 | ||
1da177e4 LT |
626 | |
627 | /* | |
628 | * This function implements a generic ability to update ruid, euid, | |
629 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
630 | */ | |
e530dca5 | 631 | long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) |
1da177e4 | 632 | { |
a29c33f4 | 633 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
634 | const struct cred *old; |
635 | struct cred *new; | |
1da177e4 | 636 | int retval; |
a29c33f4 EB |
637 | kuid_t kruid, keuid, ksuid; |
638 | ||
639 | kruid = make_kuid(ns, ruid); | |
640 | keuid = make_kuid(ns, euid); | |
641 | ksuid = make_kuid(ns, suid); | |
642 | ||
643 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
644 | return -EINVAL; | |
645 | ||
646 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
647 | return -EINVAL; | |
648 | ||
649 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
650 | return -EINVAL; | |
1da177e4 | 651 | |
d84f4f99 DH |
652 | new = prepare_creds(); |
653 | if (!new) | |
654 | return -ENOMEM; | |
655 | ||
d84f4f99 | 656 | old = current_cred(); |
1da177e4 | 657 | |
d84f4f99 | 658 | retval = -EPERM; |
40852275 | 659 | if (!ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
660 | if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && |
661 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid)) | |
d84f4f99 | 662 | goto error; |
a29c33f4 EB |
663 | if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && |
664 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid)) | |
d84f4f99 | 665 | goto error; |
a29c33f4 EB |
666 | if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && |
667 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid)) | |
d84f4f99 | 668 | goto error; |
1da177e4 | 669 | } |
d84f4f99 | 670 | |
1da177e4 | 671 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
672 | new->uid = kruid; |
673 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
674 | retval = set_user(new); |
675 | if (retval < 0) | |
676 | goto error; | |
677 | } | |
1da177e4 | 678 | } |
d84f4f99 | 679 | if (euid != (uid_t) -1) |
a29c33f4 | 680 | new->euid = keuid; |
1da177e4 | 681 | if (suid != (uid_t) -1) |
a29c33f4 | 682 | new->suid = ksuid; |
d84f4f99 | 683 | new->fsuid = new->euid; |
1da177e4 | 684 | |
d84f4f99 DH |
685 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
686 | if (retval < 0) | |
687 | goto error; | |
1da177e4 | 688 | |
d84f4f99 | 689 | return commit_creds(new); |
1da177e4 | 690 | |
d84f4f99 DH |
691 | error: |
692 | abort_creds(new); | |
693 | return retval; | |
1da177e4 LT |
694 | } |
695 | ||
e530dca5 DB |
696 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
697 | { | |
698 | return __sys_setresuid(ruid, euid, suid); | |
699 | } | |
700 | ||
a29c33f4 | 701 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 702 | { |
86a264ab | 703 | const struct cred *cred = current_cred(); |
1da177e4 | 704 | int retval; |
a29c33f4 EB |
705 | uid_t ruid, euid, suid; |
706 | ||
707 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
708 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
709 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 710 | |
ec94fc3d | 711 | retval = put_user(ruid, ruidp); |
712 | if (!retval) { | |
713 | retval = put_user(euid, euidp); | |
714 | if (!retval) | |
715 | return put_user(suid, suidp); | |
716 | } | |
1da177e4 LT |
717 | return retval; |
718 | } | |
719 | ||
720 | /* | |
721 | * Same as above, but for rgid, egid, sgid. | |
722 | */ | |
e530dca5 | 723 | long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) |
1da177e4 | 724 | { |
a29c33f4 | 725 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
726 | const struct cred *old; |
727 | struct cred *new; | |
1da177e4 | 728 | int retval; |
a29c33f4 EB |
729 | kgid_t krgid, kegid, ksgid; |
730 | ||
731 | krgid = make_kgid(ns, rgid); | |
732 | kegid = make_kgid(ns, egid); | |
733 | ksgid = make_kgid(ns, sgid); | |
734 | ||
735 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
736 | return -EINVAL; | |
737 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
738 | return -EINVAL; | |
739 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
740 | return -EINVAL; | |
1da177e4 | 741 | |
d84f4f99 DH |
742 | new = prepare_creds(); |
743 | if (!new) | |
744 | return -ENOMEM; | |
745 | old = current_cred(); | |
746 | ||
d84f4f99 | 747 | retval = -EPERM; |
111767c1 | 748 | if (!ns_capable_setid(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
749 | if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && |
750 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid)) | |
d84f4f99 | 751 | goto error; |
a29c33f4 EB |
752 | if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && |
753 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid)) | |
d84f4f99 | 754 | goto error; |
a29c33f4 EB |
755 | if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && |
756 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid)) | |
d84f4f99 | 757 | goto error; |
1da177e4 | 758 | } |
d84f4f99 | 759 | |
1da177e4 | 760 | if (rgid != (gid_t) -1) |
a29c33f4 | 761 | new->gid = krgid; |
d84f4f99 | 762 | if (egid != (gid_t) -1) |
a29c33f4 | 763 | new->egid = kegid; |
1da177e4 | 764 | if (sgid != (gid_t) -1) |
a29c33f4 | 765 | new->sgid = ksgid; |
d84f4f99 | 766 | new->fsgid = new->egid; |
1da177e4 | 767 | |
39030e13 TC |
768 | retval = security_task_fix_setgid(new, old, LSM_SETID_RES); |
769 | if (retval < 0) | |
770 | goto error; | |
771 | ||
d84f4f99 DH |
772 | return commit_creds(new); |
773 | ||
774 | error: | |
775 | abort_creds(new); | |
776 | return retval; | |
1da177e4 LT |
777 | } |
778 | ||
e530dca5 DB |
779 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
780 | { | |
781 | return __sys_setresgid(rgid, egid, sgid); | |
782 | } | |
783 | ||
a29c33f4 | 784 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 785 | { |
86a264ab | 786 | const struct cred *cred = current_cred(); |
1da177e4 | 787 | int retval; |
a29c33f4 EB |
788 | gid_t rgid, egid, sgid; |
789 | ||
790 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
791 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
792 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 793 | |
ec94fc3d | 794 | retval = put_user(rgid, rgidp); |
795 | if (!retval) { | |
796 | retval = put_user(egid, egidp); | |
797 | if (!retval) | |
798 | retval = put_user(sgid, sgidp); | |
799 | } | |
1da177e4 LT |
800 | |
801 | return retval; | |
802 | } | |
803 | ||
804 | ||
805 | /* | |
806 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
807 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
808 | * whatever uid it wants to). It normally shadows "euid", except when | |
809 | * explicitly set by setfsuid() or for access.. | |
810 | */ | |
e530dca5 | 811 | long __sys_setfsuid(uid_t uid) |
1da177e4 | 812 | { |
d84f4f99 DH |
813 | const struct cred *old; |
814 | struct cred *new; | |
815 | uid_t old_fsuid; | |
a29c33f4 EB |
816 | kuid_t kuid; |
817 | ||
818 | old = current_cred(); | |
819 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
820 | ||
821 | kuid = make_kuid(old->user_ns, uid); | |
822 | if (!uid_valid(kuid)) | |
823 | return old_fsuid; | |
1da177e4 | 824 | |
d84f4f99 DH |
825 | new = prepare_creds(); |
826 | if (!new) | |
a29c33f4 | 827 | return old_fsuid; |
1da177e4 | 828 | |
a29c33f4 EB |
829 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
830 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
40852275 | 831 | ns_capable_setid(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
832 | if (!uid_eq(kuid, old->fsuid)) { |
833 | new->fsuid = kuid; | |
d84f4f99 DH |
834 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
835 | goto change_okay; | |
1da177e4 | 836 | } |
1da177e4 LT |
837 | } |
838 | ||
d84f4f99 DH |
839 | abort_creds(new); |
840 | return old_fsuid; | |
1da177e4 | 841 | |
d84f4f99 DH |
842 | change_okay: |
843 | commit_creds(new); | |
1da177e4 LT |
844 | return old_fsuid; |
845 | } | |
846 | ||
e530dca5 DB |
847 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
848 | { | |
849 | return __sys_setfsuid(uid); | |
850 | } | |
851 | ||
1da177e4 | 852 | /* |
f42df9e6 | 853 | * Samma på svenska.. |
1da177e4 | 854 | */ |
e530dca5 | 855 | long __sys_setfsgid(gid_t gid) |
1da177e4 | 856 | { |
d84f4f99 DH |
857 | const struct cred *old; |
858 | struct cred *new; | |
859 | gid_t old_fsgid; | |
a29c33f4 EB |
860 | kgid_t kgid; |
861 | ||
862 | old = current_cred(); | |
863 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
864 | ||
865 | kgid = make_kgid(old->user_ns, gid); | |
866 | if (!gid_valid(kgid)) | |
867 | return old_fsgid; | |
d84f4f99 DH |
868 | |
869 | new = prepare_creds(); | |
870 | if (!new) | |
a29c33f4 | 871 | return old_fsgid; |
1da177e4 | 872 | |
a29c33f4 EB |
873 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
874 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
111767c1 | 875 | ns_capable_setid(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
876 | if (!gid_eq(kgid, old->fsgid)) { |
877 | new->fsgid = kgid; | |
39030e13 TC |
878 | if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0) |
879 | goto change_okay; | |
1da177e4 | 880 | } |
1da177e4 | 881 | } |
d84f4f99 | 882 | |
d84f4f99 DH |
883 | abort_creds(new); |
884 | return old_fsgid; | |
885 | ||
886 | change_okay: | |
887 | commit_creds(new); | |
1da177e4 LT |
888 | return old_fsgid; |
889 | } | |
e530dca5 DB |
890 | |
891 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) | |
892 | { | |
893 | return __sys_setfsgid(gid); | |
894 | } | |
2813893f | 895 | #endif /* CONFIG_MULTIUSER */ |
1da177e4 | 896 | |
4a22f166 SR |
897 | /** |
898 | * sys_getpid - return the thread group id of the current process | |
899 | * | |
900 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
901 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
902 | * which case the tgid is the same in all threads of the same group. | |
903 | * | |
904 | * This is SMP safe as current->tgid does not change. | |
905 | */ | |
906 | SYSCALL_DEFINE0(getpid) | |
907 | { | |
908 | return task_tgid_vnr(current); | |
909 | } | |
910 | ||
911 | /* Thread ID - the internal kernel "pid" */ | |
912 | SYSCALL_DEFINE0(gettid) | |
913 | { | |
914 | return task_pid_vnr(current); | |
915 | } | |
916 | ||
917 | /* | |
918 | * Accessing ->real_parent is not SMP-safe, it could | |
919 | * change from under us. However, we can use a stale | |
920 | * value of ->real_parent under rcu_read_lock(), see | |
921 | * release_task()->call_rcu(delayed_put_task_struct). | |
922 | */ | |
923 | SYSCALL_DEFINE0(getppid) | |
924 | { | |
925 | int pid; | |
926 | ||
927 | rcu_read_lock(); | |
928 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
929 | rcu_read_unlock(); | |
930 | ||
931 | return pid; | |
932 | } | |
933 | ||
934 | SYSCALL_DEFINE0(getuid) | |
935 | { | |
936 | /* Only we change this so SMP safe */ | |
937 | return from_kuid_munged(current_user_ns(), current_uid()); | |
938 | } | |
939 | ||
940 | SYSCALL_DEFINE0(geteuid) | |
941 | { | |
942 | /* Only we change this so SMP safe */ | |
943 | return from_kuid_munged(current_user_ns(), current_euid()); | |
944 | } | |
945 | ||
946 | SYSCALL_DEFINE0(getgid) | |
947 | { | |
948 | /* Only we change this so SMP safe */ | |
949 | return from_kgid_munged(current_user_ns(), current_gid()); | |
950 | } | |
951 | ||
952 | SYSCALL_DEFINE0(getegid) | |
953 | { | |
954 | /* Only we change this so SMP safe */ | |
955 | return from_kgid_munged(current_user_ns(), current_egid()); | |
956 | } | |
957 | ||
ca2406ed | 958 | static void do_sys_times(struct tms *tms) |
f06febc9 | 959 | { |
5613fda9 | 960 | u64 tgutime, tgstime, cutime, cstime; |
f06febc9 | 961 | |
e80d0a1a | 962 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
963 | cutime = current->signal->cutime; |
964 | cstime = current->signal->cstime; | |
5613fda9 FW |
965 | tms->tms_utime = nsec_to_clock_t(tgutime); |
966 | tms->tms_stime = nsec_to_clock_t(tgstime); | |
967 | tms->tms_cutime = nsec_to_clock_t(cutime); | |
968 | tms->tms_cstime = nsec_to_clock_t(cstime); | |
f06febc9 FM |
969 | } |
970 | ||
58fd3aa2 | 971 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 972 | { |
1da177e4 LT |
973 | if (tbuf) { |
974 | struct tms tmp; | |
f06febc9 FM |
975 | |
976 | do_sys_times(&tmp); | |
1da177e4 LT |
977 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
978 | return -EFAULT; | |
979 | } | |
e3d5a27d | 980 | force_successful_syscall_return(); |
1da177e4 LT |
981 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
982 | } | |
983 | ||
ca2406ed AV |
984 | #ifdef CONFIG_COMPAT |
985 | static compat_clock_t clock_t_to_compat_clock_t(clock_t x) | |
986 | { | |
987 | return compat_jiffies_to_clock_t(clock_t_to_jiffies(x)); | |
988 | } | |
989 | ||
990 | COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf) | |
991 | { | |
992 | if (tbuf) { | |
993 | struct tms tms; | |
994 | struct compat_tms tmp; | |
995 | ||
996 | do_sys_times(&tms); | |
997 | /* Convert our struct tms to the compat version. */ | |
998 | tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime); | |
999 | tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime); | |
1000 | tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime); | |
1001 | tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime); | |
1002 | if (copy_to_user(tbuf, &tmp, sizeof(tmp))) | |
1003 | return -EFAULT; | |
1004 | } | |
1005 | force_successful_syscall_return(); | |
1006 | return compat_jiffies_to_clock_t(jiffies); | |
1007 | } | |
1008 | #endif | |
1009 | ||
1da177e4 LT |
1010 | /* |
1011 | * This needs some heavy checking ... | |
1012 | * I just haven't the stomach for it. I also don't fully | |
1013 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
1014 | * | |
1015 | * OK, I think I have the protection semantics right.... this is really | |
1016 | * only important on a multi-user system anyway, to make sure one user | |
1017 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
1018 | * | |
98611e4e | 1019 | * !PF_FORKNOEXEC check to conform completely to POSIX. |
1da177e4 | 1020 | */ |
b290ebe2 | 1021 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
1022 | { |
1023 | struct task_struct *p; | |
ee0acf90 | 1024 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
1025 | struct pid *pgrp; |
1026 | int err; | |
1da177e4 LT |
1027 | |
1028 | if (!pid) | |
b488893a | 1029 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
1030 | if (!pgid) |
1031 | pgid = pid; | |
1032 | if (pgid < 0) | |
1033 | return -EINVAL; | |
950eaaca | 1034 | rcu_read_lock(); |
1da177e4 LT |
1035 | |
1036 | /* From this point forward we keep holding onto the tasklist lock | |
1037 | * so that our parent does not change from under us. -DaveM | |
1038 | */ | |
1039 | write_lock_irq(&tasklist_lock); | |
1040 | ||
1041 | err = -ESRCH; | |
4e021306 | 1042 | p = find_task_by_vpid(pid); |
1da177e4 LT |
1043 | if (!p) |
1044 | goto out; | |
1045 | ||
1046 | err = -EINVAL; | |
1047 | if (!thread_group_leader(p)) | |
1048 | goto out; | |
1049 | ||
4e021306 | 1050 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 1051 | err = -EPERM; |
41487c65 | 1052 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
1053 | goto out; |
1054 | err = -EACCES; | |
98611e4e | 1055 | if (!(p->flags & PF_FORKNOEXEC)) |
1da177e4 LT |
1056 | goto out; |
1057 | } else { | |
1058 | err = -ESRCH; | |
ee0acf90 | 1059 | if (p != group_leader) |
1da177e4 LT |
1060 | goto out; |
1061 | } | |
1062 | ||
1063 | err = -EPERM; | |
1064 | if (p->signal->leader) | |
1065 | goto out; | |
1066 | ||
4e021306 | 1067 | pgrp = task_pid(p); |
1da177e4 | 1068 | if (pgid != pid) { |
b488893a | 1069 | struct task_struct *g; |
1da177e4 | 1070 | |
4e021306 ON |
1071 | pgrp = find_vpid(pgid); |
1072 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 1073 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 1074 | goto out; |
1da177e4 LT |
1075 | } |
1076 | ||
1da177e4 LT |
1077 | err = security_task_setpgid(p, pgid); |
1078 | if (err) | |
1079 | goto out; | |
1080 | ||
1b0f7ffd | 1081 | if (task_pgrp(p) != pgrp) |
83beaf3c | 1082 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
1083 | |
1084 | err = 0; | |
1085 | out: | |
1086 | /* All paths lead to here, thus we are safe. -DaveM */ | |
1087 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 1088 | rcu_read_unlock(); |
1da177e4 LT |
1089 | return err; |
1090 | } | |
1091 | ||
192c5807 | 1092 | static int do_getpgid(pid_t pid) |
1da177e4 | 1093 | { |
12a3de0a ON |
1094 | struct task_struct *p; |
1095 | struct pid *grp; | |
1096 | int retval; | |
1097 | ||
1098 | rcu_read_lock(); | |
756184b7 | 1099 | if (!pid) |
12a3de0a | 1100 | grp = task_pgrp(current); |
756184b7 | 1101 | else { |
1da177e4 | 1102 | retval = -ESRCH; |
12a3de0a ON |
1103 | p = find_task_by_vpid(pid); |
1104 | if (!p) | |
1105 | goto out; | |
1106 | grp = task_pgrp(p); | |
1107 | if (!grp) | |
1108 | goto out; | |
1109 | ||
1110 | retval = security_task_getpgid(p); | |
1111 | if (retval) | |
1112 | goto out; | |
1da177e4 | 1113 | } |
12a3de0a ON |
1114 | retval = pid_vnr(grp); |
1115 | out: | |
1116 | rcu_read_unlock(); | |
1117 | return retval; | |
1da177e4 LT |
1118 | } |
1119 | ||
192c5807 DB |
1120 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1121 | { | |
1122 | return do_getpgid(pid); | |
1123 | } | |
1124 | ||
1da177e4 LT |
1125 | #ifdef __ARCH_WANT_SYS_GETPGRP |
1126 | ||
dbf040d9 | 1127 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1128 | { |
192c5807 | 1129 | return do_getpgid(0); |
1da177e4 LT |
1130 | } |
1131 | ||
1132 | #endif | |
1133 | ||
dbf040d9 | 1134 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1135 | { |
1dd768c0 ON |
1136 | struct task_struct *p; |
1137 | struct pid *sid; | |
1138 | int retval; | |
1139 | ||
1140 | rcu_read_lock(); | |
756184b7 | 1141 | if (!pid) |
1dd768c0 | 1142 | sid = task_session(current); |
756184b7 | 1143 | else { |
1da177e4 | 1144 | retval = -ESRCH; |
1dd768c0 ON |
1145 | p = find_task_by_vpid(pid); |
1146 | if (!p) | |
1147 | goto out; | |
1148 | sid = task_session(p); | |
1149 | if (!sid) | |
1150 | goto out; | |
1151 | ||
1152 | retval = security_task_getsid(p); | |
1153 | if (retval) | |
1154 | goto out; | |
1da177e4 | 1155 | } |
1dd768c0 ON |
1156 | retval = pid_vnr(sid); |
1157 | out: | |
1158 | rcu_read_unlock(); | |
1159 | return retval; | |
1da177e4 LT |
1160 | } |
1161 | ||
81dabb46 ON |
1162 | static void set_special_pids(struct pid *pid) |
1163 | { | |
1164 | struct task_struct *curr = current->group_leader; | |
1165 | ||
1166 | if (task_session(curr) != pid) | |
1167 | change_pid(curr, PIDTYPE_SID, pid); | |
1168 | ||
1169 | if (task_pgrp(curr) != pid) | |
1170 | change_pid(curr, PIDTYPE_PGID, pid); | |
1171 | } | |
1172 | ||
e2aaa9f4 | 1173 | int ksys_setsid(void) |
1da177e4 | 1174 | { |
e19f247a | 1175 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1176 | struct pid *sid = task_pid(group_leader); |
1177 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1178 | int err = -EPERM; |
1179 | ||
1da177e4 | 1180 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1181 | /* Fail if I am already a session leader */ |
1182 | if (group_leader->signal->leader) | |
1183 | goto out; | |
1184 | ||
430c6231 ON |
1185 | /* Fail if a process group id already exists that equals the |
1186 | * proposed session id. | |
390e2ff0 | 1187 | */ |
6806aac6 | 1188 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1189 | goto out; |
1190 | ||
e19f247a | 1191 | group_leader->signal->leader = 1; |
81dabb46 | 1192 | set_special_pids(sid); |
24ec839c | 1193 | |
9c9f4ded | 1194 | proc_clear_tty(group_leader); |
24ec839c | 1195 | |
e4cc0a9c | 1196 | err = session; |
1da177e4 LT |
1197 | out: |
1198 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1199 | if (err > 0) { |
0d0df599 | 1200 | proc_sid_connector(group_leader); |
5091faa4 MG |
1201 | sched_autogroup_create_attach(group_leader); |
1202 | } | |
1da177e4 LT |
1203 | return err; |
1204 | } | |
1205 | ||
e2aaa9f4 DB |
1206 | SYSCALL_DEFINE0(setsid) |
1207 | { | |
1208 | return ksys_setsid(); | |
1209 | } | |
1210 | ||
1da177e4 LT |
1211 | DECLARE_RWSEM(uts_sem); |
1212 | ||
09f9bbbf AW |
1213 | #ifdef COMPAT_UTS_MACHINE |
1214 | static char compat_uts_machine[__OLD_UTS_LEN+1] = COMPAT_UTS_MACHINE; | |
1215 | ||
1216 | static int __init parse_compat_uts_machine(char *arg) | |
1217 | { | |
1218 | strncpy(compat_uts_machine, arg, __OLD_UTS_LEN); | |
1219 | compat_uts_machine[__OLD_UTS_LEN] = 0; | |
1220 | return 0; | |
1221 | } | |
1222 | early_param("compat_uts_machine", parse_compat_uts_machine); | |
1223 | ||
1224 | #undef COMPAT_UTS_MACHINE | |
1225 | #define COMPAT_UTS_MACHINE compat_uts_machine | |
1226 | #endif | |
1227 | ||
e28cbf22 CH |
1228 | #ifdef COMPAT_UTS_MACHINE |
1229 | #define override_architecture(name) \ | |
46da2766 | 1230 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1231 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1232 | sizeof(COMPAT_UTS_MACHINE))) | |
1233 | #else | |
1234 | #define override_architecture(name) 0 | |
1235 | #endif | |
1236 | ||
be27425d AK |
1237 | /* |
1238 | * Work around broken programs that cannot handle "Linux 3.0". | |
1239 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
b7285b42 JN |
1240 | * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be |
1241 | * 2.6.60. | |
be27425d | 1242 | */ |
2702b152 | 1243 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1244 | { |
1245 | int ret = 0; | |
be27425d AK |
1246 | |
1247 | if (current->personality & UNAME26) { | |
2702b152 KC |
1248 | const char *rest = UTS_RELEASE; |
1249 | char buf[65] = { 0 }; | |
be27425d AK |
1250 | int ndots = 0; |
1251 | unsigned v; | |
2702b152 | 1252 | size_t copy; |
be27425d AK |
1253 | |
1254 | while (*rest) { | |
1255 | if (*rest == '.' && ++ndots >= 3) | |
1256 | break; | |
1257 | if (!isdigit(*rest) && *rest != '.') | |
1258 | break; | |
1259 | rest++; | |
1260 | } | |
39afb5ee | 1261 | v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60; |
31fd84b9 | 1262 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1263 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1264 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1265 | } |
1266 | return ret; | |
1267 | } | |
1268 | ||
e48fbb69 | 1269 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 | 1270 | { |
42a0cc34 | 1271 | struct new_utsname tmp; |
1da177e4 LT |
1272 | |
1273 | down_read(&uts_sem); | |
42a0cc34 | 1274 | memcpy(&tmp, utsname(), sizeof(tmp)); |
1da177e4 | 1275 | up_read(&uts_sem); |
42a0cc34 JH |
1276 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1277 | return -EFAULT; | |
e28cbf22 | 1278 | |
42a0cc34 JH |
1279 | if (override_release(name->release, sizeof(name->release))) |
1280 | return -EFAULT; | |
1281 | if (override_architecture(name)) | |
1282 | return -EFAULT; | |
1283 | return 0; | |
1da177e4 LT |
1284 | } |
1285 | ||
5cacdb4a CH |
1286 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1287 | /* | |
1288 | * Old cruft | |
1289 | */ | |
1290 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1291 | { | |
42a0cc34 | 1292 | struct old_utsname tmp; |
5cacdb4a CH |
1293 | |
1294 | if (!name) | |
1295 | return -EFAULT; | |
1296 | ||
1297 | down_read(&uts_sem); | |
42a0cc34 | 1298 | memcpy(&tmp, utsname(), sizeof(tmp)); |
5cacdb4a | 1299 | up_read(&uts_sem); |
42a0cc34 JH |
1300 | if (copy_to_user(name, &tmp, sizeof(tmp))) |
1301 | return -EFAULT; | |
5cacdb4a | 1302 | |
42a0cc34 JH |
1303 | if (override_release(name->release, sizeof(name->release))) |
1304 | return -EFAULT; | |
1305 | if (override_architecture(name)) | |
1306 | return -EFAULT; | |
1307 | return 0; | |
5cacdb4a CH |
1308 | } |
1309 | ||
1310 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1311 | { | |
5e1aada0 | 1312 | struct oldold_utsname tmp; |
5cacdb4a CH |
1313 | |
1314 | if (!name) | |
1315 | return -EFAULT; | |
5cacdb4a | 1316 | |
5e1aada0 JP |
1317 | memset(&tmp, 0, sizeof(tmp)); |
1318 | ||
5cacdb4a | 1319 | down_read(&uts_sem); |
42a0cc34 JH |
1320 | memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN); |
1321 | memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN); | |
1322 | memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN); | |
1323 | memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN); | |
1324 | memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN); | |
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_architecture(name)) |
1330 | return -EFAULT; | |
1331 | if (override_release(name->release, sizeof(name->release))) | |
1332 | return -EFAULT; | |
1333 | return 0; | |
5cacdb4a CH |
1334 | } |
1335 | #endif | |
1336 | ||
5a8a82b1 | 1337 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1338 | { |
1339 | int errno; | |
1340 | char tmp[__NEW_UTS_LEN]; | |
1341 | ||
bb96a6f5 | 1342 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1343 | return -EPERM; |
fc832ad3 | 1344 | |
1da177e4 LT |
1345 | if (len < 0 || len > __NEW_UTS_LEN) |
1346 | return -EINVAL; | |
1da177e4 LT |
1347 | errno = -EFAULT; |
1348 | if (!copy_from_user(tmp, name, len)) { | |
42a0cc34 | 1349 | struct new_utsname *u; |
9679e4dd | 1350 | |
42a0cc34 JH |
1351 | down_write(&uts_sem); |
1352 | u = utsname(); | |
9679e4dd AM |
1353 | memcpy(u->nodename, tmp, len); |
1354 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1355 | errno = 0; |
499eea6b | 1356 | uts_proc_notify(UTS_PROC_HOSTNAME); |
42a0cc34 | 1357 | up_write(&uts_sem); |
1da177e4 | 1358 | } |
1da177e4 LT |
1359 | return errno; |
1360 | } | |
1361 | ||
1362 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1363 | ||
5a8a82b1 | 1364 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 | 1365 | { |
42a0cc34 | 1366 | int i; |
9679e4dd | 1367 | struct new_utsname *u; |
42a0cc34 | 1368 | char tmp[__NEW_UTS_LEN + 1]; |
1da177e4 LT |
1369 | |
1370 | if (len < 0) | |
1371 | return -EINVAL; | |
1372 | down_read(&uts_sem); | |
9679e4dd AM |
1373 | u = utsname(); |
1374 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1375 | if (i > len) |
1376 | i = len; | |
42a0cc34 | 1377 | memcpy(tmp, u->nodename, i); |
1da177e4 | 1378 | up_read(&uts_sem); |
42a0cc34 JH |
1379 | if (copy_to_user(name, tmp, i)) |
1380 | return -EFAULT; | |
1381 | return 0; | |
1da177e4 LT |
1382 | } |
1383 | ||
1384 | #endif | |
1385 | ||
1386 | /* | |
1387 | * Only setdomainname; getdomainname can be implemented by calling | |
1388 | * uname() | |
1389 | */ | |
5a8a82b1 | 1390 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1391 | { |
1392 | int errno; | |
1393 | char tmp[__NEW_UTS_LEN]; | |
1394 | ||
fc832ad3 | 1395 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1396 | return -EPERM; |
1397 | if (len < 0 || len > __NEW_UTS_LEN) | |
1398 | return -EINVAL; | |
1399 | ||
1da177e4 LT |
1400 | errno = -EFAULT; |
1401 | if (!copy_from_user(tmp, name, len)) { | |
42a0cc34 | 1402 | struct new_utsname *u; |
9679e4dd | 1403 | |
42a0cc34 JH |
1404 | down_write(&uts_sem); |
1405 | u = utsname(); | |
9679e4dd AM |
1406 | memcpy(u->domainname, tmp, len); |
1407 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1408 | errno = 0; |
499eea6b | 1409 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
42a0cc34 | 1410 | up_write(&uts_sem); |
1da177e4 | 1411 | } |
1da177e4 LT |
1412 | return errno; |
1413 | } | |
1414 | ||
e48fbb69 | 1415 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1416 | { |
b9518345 JS |
1417 | struct rlimit value; |
1418 | int ret; | |
1419 | ||
1420 | ret = do_prlimit(current, resource, NULL, &value); | |
1421 | if (!ret) | |
1422 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1423 | ||
1424 | return ret; | |
1da177e4 LT |
1425 | } |
1426 | ||
d9e968cb AV |
1427 | #ifdef CONFIG_COMPAT |
1428 | ||
1429 | COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource, | |
1430 | struct compat_rlimit __user *, rlim) | |
1431 | { | |
1432 | struct rlimit r; | |
1433 | struct compat_rlimit r32; | |
1434 | ||
1435 | if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit))) | |
1436 | return -EFAULT; | |
1437 | ||
1438 | if (r32.rlim_cur == COMPAT_RLIM_INFINITY) | |
1439 | r.rlim_cur = RLIM_INFINITY; | |
1440 | else | |
1441 | r.rlim_cur = r32.rlim_cur; | |
1442 | if (r32.rlim_max == COMPAT_RLIM_INFINITY) | |
1443 | r.rlim_max = RLIM_INFINITY; | |
1444 | else | |
1445 | r.rlim_max = r32.rlim_max; | |
1446 | return do_prlimit(current, resource, &r, NULL); | |
1447 | } | |
1448 | ||
1449 | COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource, | |
1450 | struct compat_rlimit __user *, rlim) | |
1451 | { | |
1452 | struct rlimit r; | |
1453 | int ret; | |
1454 | ||
1455 | ret = do_prlimit(current, resource, NULL, &r); | |
1456 | if (!ret) { | |
58c7ffc0 | 1457 | struct compat_rlimit r32; |
d9e968cb AV |
1458 | if (r.rlim_cur > COMPAT_RLIM_INFINITY) |
1459 | r32.rlim_cur = COMPAT_RLIM_INFINITY; | |
1460 | else | |
1461 | r32.rlim_cur = r.rlim_cur; | |
1462 | if (r.rlim_max > COMPAT_RLIM_INFINITY) | |
1463 | r32.rlim_max = COMPAT_RLIM_INFINITY; | |
1464 | else | |
1465 | r32.rlim_max = r.rlim_max; | |
1466 | ||
1467 | if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit))) | |
1468 | return -EFAULT; | |
1469 | } | |
1470 | return ret; | |
1471 | } | |
1472 | ||
1473 | #endif | |
1474 | ||
1da177e4 LT |
1475 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT |
1476 | ||
1477 | /* | |
1478 | * Back compatibility for getrlimit. Needed for some apps. | |
1479 | */ | |
e48fbb69 HC |
1480 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1481 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1482 | { |
1483 | struct rlimit x; | |
1484 | if (resource >= RLIM_NLIMITS) | |
1485 | return -EINVAL; | |
1486 | ||
23d6aef7 | 1487 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
1da177e4 LT |
1488 | task_lock(current->group_leader); |
1489 | x = current->signal->rlim[resource]; | |
1490 | task_unlock(current->group_leader); | |
756184b7 | 1491 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1492 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1493 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 | 1494 | x.rlim_max = 0x7FFFFFFF; |
ec94fc3d | 1495 | return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0; |
1da177e4 LT |
1496 | } |
1497 | ||
613763a1 AV |
1498 | #ifdef CONFIG_COMPAT |
1499 | COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, | |
1500 | struct compat_rlimit __user *, rlim) | |
1501 | { | |
1502 | struct rlimit r; | |
1503 | ||
1504 | if (resource >= RLIM_NLIMITS) | |
1505 | return -EINVAL; | |
1506 | ||
23d6aef7 | 1507 | resource = array_index_nospec(resource, RLIM_NLIMITS); |
613763a1 AV |
1508 | task_lock(current->group_leader); |
1509 | r = current->signal->rlim[resource]; | |
1510 | task_unlock(current->group_leader); | |
1511 | if (r.rlim_cur > 0x7FFFFFFF) | |
1512 | r.rlim_cur = 0x7FFFFFFF; | |
1513 | if (r.rlim_max > 0x7FFFFFFF) | |
1514 | r.rlim_max = 0x7FFFFFFF; | |
1515 | ||
1516 | if (put_user(r.rlim_cur, &rlim->rlim_cur) || | |
1517 | put_user(r.rlim_max, &rlim->rlim_max)) | |
1518 | return -EFAULT; | |
1519 | return 0; | |
1520 | } | |
1521 | #endif | |
1522 | ||
1da177e4 LT |
1523 | #endif |
1524 | ||
c022a0ac JS |
1525 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1526 | { | |
1527 | #if BITS_PER_LONG < 64 | |
1528 | return rlim64 >= ULONG_MAX; | |
1529 | #else | |
1530 | return rlim64 == RLIM64_INFINITY; | |
1531 | #endif | |
1532 | } | |
1533 | ||
1534 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1535 | { | |
1536 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1537 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1538 | else | |
1539 | rlim64->rlim_cur = rlim->rlim_cur; | |
1540 | if (rlim->rlim_max == RLIM_INFINITY) | |
1541 | rlim64->rlim_max = RLIM64_INFINITY; | |
1542 | else | |
1543 | rlim64->rlim_max = rlim->rlim_max; | |
1544 | } | |
1545 | ||
1546 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1547 | { | |
1548 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1549 | rlim->rlim_cur = RLIM_INFINITY; | |
1550 | else | |
1551 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1552 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1553 | rlim->rlim_max = RLIM_INFINITY; | |
1554 | else | |
1555 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1556 | } | |
1557 | ||
1c1e618d | 1558 | /* make sure you are allowed to change @tsk limits before calling this */ |
5b41535a JS |
1559 | int do_prlimit(struct task_struct *tsk, unsigned int resource, |
1560 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1da177e4 | 1561 | { |
5b41535a | 1562 | struct rlimit *rlim; |
86f162f4 | 1563 | int retval = 0; |
1da177e4 LT |
1564 | |
1565 | if (resource >= RLIM_NLIMITS) | |
1566 | return -EINVAL; | |
5b41535a JS |
1567 | if (new_rlim) { |
1568 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1569 | return -EINVAL; | |
1570 | if (resource == RLIMIT_NOFILE && | |
1571 | new_rlim->rlim_max > sysctl_nr_open) | |
1572 | return -EPERM; | |
1573 | } | |
1da177e4 | 1574 | |
1c1e618d JS |
1575 | /* protect tsk->signal and tsk->sighand from disappearing */ |
1576 | read_lock(&tasklist_lock); | |
1577 | if (!tsk->sighand) { | |
1578 | retval = -ESRCH; | |
1579 | goto out; | |
1580 | } | |
1581 | ||
5b41535a | 1582 | rlim = tsk->signal->rlim + resource; |
86f162f4 | 1583 | task_lock(tsk->group_leader); |
5b41535a | 1584 | if (new_rlim) { |
fc832ad3 SH |
1585 | /* Keep the capable check against init_user_ns until |
1586 | cgroups can contain all limits */ | |
5b41535a JS |
1587 | if (new_rlim->rlim_max > rlim->rlim_max && |
1588 | !capable(CAP_SYS_RESOURCE)) | |
1589 | retval = -EPERM; | |
1590 | if (!retval) | |
cad4ea54 | 1591 | retval = security_task_setrlimit(tsk, resource, new_rlim); |
5b41535a JS |
1592 | } |
1593 | if (!retval) { | |
1594 | if (old_rlim) | |
1595 | *old_rlim = *rlim; | |
1596 | if (new_rlim) | |
1597 | *rlim = *new_rlim; | |
9926e4c7 | 1598 | } |
7855c35d | 1599 | task_unlock(tsk->group_leader); |
1da177e4 | 1600 | |
d3561f78 | 1601 | /* |
24db4dd9 TG |
1602 | * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not |
1603 | * infite. In case of RLIM_INFINITY the posix CPU timer code | |
1604 | * ignores the rlimit. | |
d3561f78 | 1605 | */ |
5b41535a | 1606 | if (!retval && new_rlim && resource == RLIMIT_CPU && |
baa73d9e NP |
1607 | new_rlim->rlim_cur != RLIM_INFINITY && |
1608 | IS_ENABLED(CONFIG_POSIX_TIMERS)) | |
5b41535a | 1609 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); |
ec9e16ba | 1610 | out: |
1c1e618d | 1611 | read_unlock(&tasklist_lock); |
2fb9d268 | 1612 | return retval; |
1da177e4 LT |
1613 | } |
1614 | ||
c022a0ac | 1615 | /* rcu lock must be held */ |
791ec491 SS |
1616 | static int check_prlimit_permission(struct task_struct *task, |
1617 | unsigned int flags) | |
c022a0ac JS |
1618 | { |
1619 | const struct cred *cred = current_cred(), *tcred; | |
791ec491 | 1620 | bool id_match; |
c022a0ac | 1621 | |
fc832ad3 SH |
1622 | if (current == task) |
1623 | return 0; | |
c022a0ac | 1624 | |
fc832ad3 | 1625 | tcred = __task_cred(task); |
791ec491 SS |
1626 | id_match = (uid_eq(cred->uid, tcred->euid) && |
1627 | uid_eq(cred->uid, tcred->suid) && | |
1628 | uid_eq(cred->uid, tcred->uid) && | |
1629 | gid_eq(cred->gid, tcred->egid) && | |
1630 | gid_eq(cred->gid, tcred->sgid) && | |
1631 | gid_eq(cred->gid, tcred->gid)); | |
1632 | if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) | |
1633 | return -EPERM; | |
fc832ad3 | 1634 | |
791ec491 | 1635 | return security_task_prlimit(cred, tcred, flags); |
c022a0ac JS |
1636 | } |
1637 | ||
1638 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1639 | const struct rlimit64 __user *, new_rlim, | |
1640 | struct rlimit64 __user *, old_rlim) | |
1641 | { | |
1642 | struct rlimit64 old64, new64; | |
1643 | struct rlimit old, new; | |
1644 | struct task_struct *tsk; | |
791ec491 | 1645 | unsigned int checkflags = 0; |
c022a0ac JS |
1646 | int ret; |
1647 | ||
791ec491 SS |
1648 | if (old_rlim) |
1649 | checkflags |= LSM_PRLIMIT_READ; | |
1650 | ||
c022a0ac JS |
1651 | if (new_rlim) { |
1652 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1653 | return -EFAULT; | |
1654 | rlim64_to_rlim(&new64, &new); | |
791ec491 | 1655 | checkflags |= LSM_PRLIMIT_WRITE; |
c022a0ac JS |
1656 | } |
1657 | ||
1658 | rcu_read_lock(); | |
1659 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1660 | if (!tsk) { | |
1661 | rcu_read_unlock(); | |
1662 | return -ESRCH; | |
1663 | } | |
791ec491 | 1664 | ret = check_prlimit_permission(tsk, checkflags); |
c022a0ac JS |
1665 | if (ret) { |
1666 | rcu_read_unlock(); | |
1667 | return ret; | |
1668 | } | |
1669 | get_task_struct(tsk); | |
1670 | rcu_read_unlock(); | |
1671 | ||
1672 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1673 | old_rlim ? &old : NULL); | |
1674 | ||
1675 | if (!ret && old_rlim) { | |
1676 | rlim_to_rlim64(&old, &old64); | |
1677 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1678 | ret = -EFAULT; | |
1679 | } | |
1680 | ||
1681 | put_task_struct(tsk); | |
1682 | return ret; | |
1683 | } | |
1684 | ||
7855c35d JS |
1685 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1686 | { | |
1687 | struct rlimit new_rlim; | |
1688 | ||
1689 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1690 | return -EFAULT; | |
5b41535a | 1691 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1692 | } |
1693 | ||
1da177e4 LT |
1694 | /* |
1695 | * It would make sense to put struct rusage in the task_struct, | |
1696 | * except that would make the task_struct be *really big*. After | |
1697 | * task_struct gets moved into malloc'ed memory, it would | |
1698 | * make sense to do this. It will make moving the rest of the information | |
1699 | * a lot simpler! (Which we're not doing right now because we're not | |
1700 | * measuring them yet). | |
1701 | * | |
1da177e4 LT |
1702 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1703 | * races with threads incrementing their own counters. But since word | |
1704 | * reads are atomic, we either get new values or old values and we don't | |
1705 | * care which for the sums. We always take the siglock to protect reading | |
1706 | * the c* fields from p->signal from races with exit.c updating those | |
1707 | * fields when reaping, so a sample either gets all the additions of a | |
1708 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1709 | * |
de047c1b RT |
1710 | * Locking: |
1711 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1712 | * for the cases current multithreaded, non-current single threaded | |
1713 | * non-current multithreaded. Thread traversal is now safe with | |
1714 | * the siglock held. | |
1715 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1716 | * single threaded, as no one else can take our signal_struct away, no one | |
1717 | * else can reap the children to update signal->c* counters, and no one else | |
1718 | * can race with the signal-> fields. If we do not take any lock, the | |
1719 | * signal-> fields could be read out of order while another thread was just | |
1720 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1721 | * On the writer side, write memory barrier is implied in __exit_signal | |
1722 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1723 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1724 | * |
1da177e4 LT |
1725 | */ |
1726 | ||
f06febc9 | 1727 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1728 | { |
679c9cd4 SK |
1729 | r->ru_nvcsw += t->nvcsw; |
1730 | r->ru_nivcsw += t->nivcsw; | |
1731 | r->ru_minflt += t->min_flt; | |
1732 | r->ru_majflt += t->maj_flt; | |
1733 | r->ru_inblock += task_io_get_inblock(t); | |
1734 | r->ru_oublock += task_io_get_oublock(t); | |
1735 | } | |
1736 | ||
ce72a16f | 1737 | void getrusage(struct task_struct *p, int who, struct rusage *r) |
1da177e4 LT |
1738 | { |
1739 | struct task_struct *t; | |
1740 | unsigned long flags; | |
5613fda9 | 1741 | u64 tgutime, tgstime, utime, stime; |
1f10206c | 1742 | unsigned long maxrss = 0; |
1da177e4 | 1743 | |
ec94fc3d | 1744 | memset((char *)r, 0, sizeof (*r)); |
64861634 | 1745 | utime = stime = 0; |
1da177e4 | 1746 | |
679c9cd4 | 1747 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1748 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1749 | accumulate_thread_rusage(p, r); |
1f10206c | 1750 | maxrss = p->signal->maxrss; |
679c9cd4 SK |
1751 | goto out; |
1752 | } | |
1753 | ||
d6cf723a | 1754 | if (!lock_task_sighand(p, &flags)) |
de047c1b | 1755 | return; |
0f59cc4a | 1756 | |
1da177e4 | 1757 | switch (who) { |
ec94fc3d | 1758 | case RUSAGE_BOTH: |
1759 | case RUSAGE_CHILDREN: | |
1760 | utime = p->signal->cutime; | |
1761 | stime = p->signal->cstime; | |
1762 | r->ru_nvcsw = p->signal->cnvcsw; | |
1763 | r->ru_nivcsw = p->signal->cnivcsw; | |
1764 | r->ru_minflt = p->signal->cmin_flt; | |
1765 | r->ru_majflt = p->signal->cmaj_flt; | |
1766 | r->ru_inblock = p->signal->cinblock; | |
1767 | r->ru_oublock = p->signal->coublock; | |
1768 | maxrss = p->signal->cmaxrss; | |
1769 | ||
1770 | if (who == RUSAGE_CHILDREN) | |
1da177e4 | 1771 | break; |
df561f66 | 1772 | fallthrough; |
0f59cc4a | 1773 | |
ec94fc3d | 1774 | case RUSAGE_SELF: |
1775 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); | |
1776 | utime += tgutime; | |
1777 | stime += tgstime; | |
1778 | r->ru_nvcsw += p->signal->nvcsw; | |
1779 | r->ru_nivcsw += p->signal->nivcsw; | |
1780 | r->ru_minflt += p->signal->min_flt; | |
1781 | r->ru_majflt += p->signal->maj_flt; | |
1782 | r->ru_inblock += p->signal->inblock; | |
1783 | r->ru_oublock += p->signal->oublock; | |
1784 | if (maxrss < p->signal->maxrss) | |
1785 | maxrss = p->signal->maxrss; | |
1786 | t = p; | |
1787 | do { | |
1788 | accumulate_thread_rusage(t, r); | |
1789 | } while_each_thread(p, t); | |
1790 | break; | |
1791 | ||
1792 | default: | |
1793 | BUG(); | |
1da177e4 | 1794 | } |
de047c1b | 1795 | unlock_task_sighand(p, &flags); |
de047c1b | 1796 | |
679c9cd4 | 1797 | out: |
bdd565f8 AB |
1798 | r->ru_utime = ns_to_kernel_old_timeval(utime); |
1799 | r->ru_stime = ns_to_kernel_old_timeval(stime); | |
1f10206c JP |
1800 | |
1801 | if (who != RUSAGE_CHILDREN) { | |
1802 | struct mm_struct *mm = get_task_mm(p); | |
ec94fc3d | 1803 | |
1f10206c JP |
1804 | if (mm) { |
1805 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1806 | mmput(mm); | |
1807 | } | |
1808 | } | |
1809 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1da177e4 LT |
1810 | } |
1811 | ||
ce72a16f | 1812 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 LT |
1813 | { |
1814 | struct rusage r; | |
ec94fc3d | 1815 | |
679c9cd4 SK |
1816 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1817 | who != RUSAGE_THREAD) | |
1da177e4 | 1818 | return -EINVAL; |
ce72a16f AV |
1819 | |
1820 | getrusage(current, who, &r); | |
1821 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; | |
1da177e4 LT |
1822 | } |
1823 | ||
8d2d5c4a AV |
1824 | #ifdef CONFIG_COMPAT |
1825 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1826 | { | |
1827 | struct rusage r; | |
1828 | ||
1829 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1830 | who != RUSAGE_THREAD) | |
1831 | return -EINVAL; | |
1832 | ||
ce72a16f | 1833 | getrusage(current, who, &r); |
8d2d5c4a AV |
1834 | return put_compat_rusage(&r, ru); |
1835 | } | |
1836 | #endif | |
1837 | ||
e48fbb69 | 1838 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
1839 | { |
1840 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1841 | return mask; | |
1842 | } | |
3b7391de | 1843 | |
6e399cd1 | 1844 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
b32dfe37 | 1845 | { |
2903ff01 | 1846 | struct fd exe; |
6e399cd1 | 1847 | struct file *old_exe, *exe_file; |
496ad9aa | 1848 | struct inode *inode; |
2903ff01 | 1849 | int err; |
b32dfe37 | 1850 | |
2903ff01 AV |
1851 | exe = fdget(fd); |
1852 | if (!exe.file) | |
b32dfe37 CG |
1853 | return -EBADF; |
1854 | ||
496ad9aa | 1855 | inode = file_inode(exe.file); |
b32dfe37 CG |
1856 | |
1857 | /* | |
1858 | * Because the original mm->exe_file points to executable file, make | |
1859 | * sure that this one is executable as well, to avoid breaking an | |
1860 | * overall picture. | |
1861 | */ | |
1862 | err = -EACCES; | |
90f8572b | 1863 | if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path)) |
b32dfe37 CG |
1864 | goto exit; |
1865 | ||
496ad9aa | 1866 | err = inode_permission(inode, MAY_EXEC); |
b32dfe37 CG |
1867 | if (err) |
1868 | goto exit; | |
1869 | ||
bafb282d | 1870 | /* |
4229fb1d | 1871 | * Forbid mm->exe_file change if old file still mapped. |
bafb282d | 1872 | */ |
6e399cd1 | 1873 | exe_file = get_mm_exe_file(mm); |
bafb282d | 1874 | err = -EBUSY; |
6e399cd1 | 1875 | if (exe_file) { |
4229fb1d KK |
1876 | struct vm_area_struct *vma; |
1877 | ||
d8ed45c5 | 1878 | mmap_read_lock(mm); |
6e399cd1 DB |
1879 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
1880 | if (!vma->vm_file) | |
1881 | continue; | |
1882 | if (path_equal(&vma->vm_file->f_path, | |
1883 | &exe_file->f_path)) | |
1884 | goto exit_err; | |
1885 | } | |
1886 | ||
d8ed45c5 | 1887 | mmap_read_unlock(mm); |
6e399cd1 | 1888 | fput(exe_file); |
bafb282d KK |
1889 | } |
1890 | ||
4229fb1d | 1891 | err = 0; |
6e399cd1 DB |
1892 | /* set the new file, lockless */ |
1893 | get_file(exe.file); | |
1894 | old_exe = xchg(&mm->exe_file, exe.file); | |
1895 | if (old_exe) | |
1896 | fput(old_exe); | |
b32dfe37 | 1897 | exit: |
2903ff01 | 1898 | fdput(exe); |
b32dfe37 | 1899 | return err; |
6e399cd1 | 1900 | exit_err: |
d8ed45c5 | 1901 | mmap_read_unlock(mm); |
6e399cd1 DB |
1902 | fput(exe_file); |
1903 | goto exit; | |
b32dfe37 CG |
1904 | } |
1905 | ||
f606b77f | 1906 | /* |
11bbd8b4 MK |
1907 | * Check arithmetic relations of passed addresses. |
1908 | * | |
f606b77f CG |
1909 | * WARNING: we don't require any capability here so be very careful |
1910 | * in what is allowed for modification from userspace. | |
1911 | */ | |
11bbd8b4 | 1912 | static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map) |
f606b77f CG |
1913 | { |
1914 | unsigned long mmap_max_addr = TASK_SIZE; | |
f606b77f CG |
1915 | int error = -EINVAL, i; |
1916 | ||
1917 | static const unsigned char offsets[] = { | |
1918 | offsetof(struct prctl_mm_map, start_code), | |
1919 | offsetof(struct prctl_mm_map, end_code), | |
1920 | offsetof(struct prctl_mm_map, start_data), | |
1921 | offsetof(struct prctl_mm_map, end_data), | |
1922 | offsetof(struct prctl_mm_map, start_brk), | |
1923 | offsetof(struct prctl_mm_map, brk), | |
1924 | offsetof(struct prctl_mm_map, start_stack), | |
1925 | offsetof(struct prctl_mm_map, arg_start), | |
1926 | offsetof(struct prctl_mm_map, arg_end), | |
1927 | offsetof(struct prctl_mm_map, env_start), | |
1928 | offsetof(struct prctl_mm_map, env_end), | |
1929 | }; | |
1930 | ||
1931 | /* | |
1932 | * Make sure the members are not somewhere outside | |
1933 | * of allowed address space. | |
1934 | */ | |
1935 | for (i = 0; i < ARRAY_SIZE(offsets); i++) { | |
1936 | u64 val = *(u64 *)((char *)prctl_map + offsets[i]); | |
1937 | ||
1938 | if ((unsigned long)val >= mmap_max_addr || | |
1939 | (unsigned long)val < mmap_min_addr) | |
1940 | goto out; | |
1941 | } | |
1942 | ||
1943 | /* | |
1944 | * Make sure the pairs are ordered. | |
1945 | */ | |
1946 | #define __prctl_check_order(__m1, __op, __m2) \ | |
1947 | ((unsigned long)prctl_map->__m1 __op \ | |
1948 | (unsigned long)prctl_map->__m2) ? 0 : -EINVAL | |
1949 | error = __prctl_check_order(start_code, <, end_code); | |
a9e73998 | 1950 | error |= __prctl_check_order(start_data,<=, end_data); |
f606b77f CG |
1951 | error |= __prctl_check_order(start_brk, <=, brk); |
1952 | error |= __prctl_check_order(arg_start, <=, arg_end); | |
1953 | error |= __prctl_check_order(env_start, <=, env_end); | |
1954 | if (error) | |
1955 | goto out; | |
1956 | #undef __prctl_check_order | |
1957 | ||
1958 | error = -EINVAL; | |
1959 | ||
1960 | /* | |
1961 | * @brk should be after @end_data in traditional maps. | |
1962 | */ | |
1963 | if (prctl_map->start_brk <= prctl_map->end_data || | |
1964 | prctl_map->brk <= prctl_map->end_data) | |
1965 | goto out; | |
1966 | ||
1967 | /* | |
1968 | * Neither we should allow to override limits if they set. | |
1969 | */ | |
1970 | if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk, | |
1971 | prctl_map->start_brk, prctl_map->end_data, | |
1972 | prctl_map->start_data)) | |
1973 | goto out; | |
1974 | ||
f606b77f CG |
1975 | error = 0; |
1976 | out: | |
1977 | return error; | |
1978 | } | |
1979 | ||
4a00e9df | 1980 | #ifdef CONFIG_CHECKPOINT_RESTORE |
f606b77f CG |
1981 | static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size) |
1982 | { | |
1983 | struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, }; | |
1984 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1985 | struct mm_struct *mm = current->mm; | |
1986 | int error; | |
1987 | ||
1988 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1989 | BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256); | |
1990 | ||
1991 | if (opt == PR_SET_MM_MAP_SIZE) | |
1992 | return put_user((unsigned int)sizeof(prctl_map), | |
1993 | (unsigned int __user *)addr); | |
1994 | ||
1995 | if (data_size != sizeof(prctl_map)) | |
1996 | return -EINVAL; | |
1997 | ||
1998 | if (copy_from_user(&prctl_map, addr, sizeof(prctl_map))) | |
1999 | return -EFAULT; | |
2000 | ||
11bbd8b4 | 2001 | error = validate_prctl_map_addr(&prctl_map); |
f606b77f CG |
2002 | if (error) |
2003 | return error; | |
2004 | ||
2005 | if (prctl_map.auxv_size) { | |
11bbd8b4 MK |
2006 | /* |
2007 | * Someone is trying to cheat the auxv vector. | |
2008 | */ | |
2009 | if (!prctl_map.auxv || | |
2010 | prctl_map.auxv_size > sizeof(mm->saved_auxv)) | |
2011 | return -EINVAL; | |
2012 | ||
f606b77f CG |
2013 | memset(user_auxv, 0, sizeof(user_auxv)); |
2014 | if (copy_from_user(user_auxv, | |
2015 | (const void __user *)prctl_map.auxv, | |
2016 | prctl_map.auxv_size)) | |
2017 | return -EFAULT; | |
2018 | ||
2019 | /* Last entry must be AT_NULL as specification requires */ | |
2020 | user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL; | |
2021 | user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL; | |
2022 | } | |
2023 | ||
ddf1d398 | 2024 | if (prctl_map.exe_fd != (u32)-1) { |
11bbd8b4 | 2025 | /* |
ebd6de68 NV |
2026 | * Check if the current user is checkpoint/restore capable. |
2027 | * At the time of this writing, it checks for CAP_SYS_ADMIN | |
2028 | * or CAP_CHECKPOINT_RESTORE. | |
2029 | * Note that a user with access to ptrace can masquerade an | |
2030 | * arbitrary program as any executable, even setuid ones. | |
2031 | * This may have implications in the tomoyo subsystem. | |
11bbd8b4 | 2032 | */ |
ebd6de68 | 2033 | if (!checkpoint_restore_ns_capable(current_user_ns())) |
227175b2 | 2034 | return -EPERM; |
11bbd8b4 | 2035 | |
6e399cd1 | 2036 | error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd); |
ddf1d398 MG |
2037 | if (error) |
2038 | return error; | |
2039 | } | |
2040 | ||
88aa7cc6 | 2041 | /* |
c1e8d7c6 | 2042 | * arg_lock protects concurent updates but we still need mmap_lock for |
88aa7cc6 YS |
2043 | * read to exclude races with sys_brk. |
2044 | */ | |
d8ed45c5 | 2045 | mmap_read_lock(mm); |
f606b77f CG |
2046 | |
2047 | /* | |
2048 | * We don't validate if these members are pointing to | |
2049 | * real present VMAs because application may have correspond | |
2050 | * VMAs already unmapped and kernel uses these members for statistics | |
2051 | * output in procfs mostly, except | |
2052 | * | |
15ec0fcf | 2053 | * - @start_brk/@brk which are used in do_brk_flags but kernel lookups |
f606b77f CG |
2054 | * for VMAs when updating these memvers so anything wrong written |
2055 | * here cause kernel to swear at userspace program but won't lead | |
2056 | * to any problem in kernel itself | |
2057 | */ | |
2058 | ||
88aa7cc6 | 2059 | spin_lock(&mm->arg_lock); |
f606b77f CG |
2060 | mm->start_code = prctl_map.start_code; |
2061 | mm->end_code = prctl_map.end_code; | |
2062 | mm->start_data = prctl_map.start_data; | |
2063 | mm->end_data = prctl_map.end_data; | |
2064 | mm->start_brk = prctl_map.start_brk; | |
2065 | mm->brk = prctl_map.brk; | |
2066 | mm->start_stack = prctl_map.start_stack; | |
2067 | mm->arg_start = prctl_map.arg_start; | |
2068 | mm->arg_end = prctl_map.arg_end; | |
2069 | mm->env_start = prctl_map.env_start; | |
2070 | mm->env_end = prctl_map.env_end; | |
88aa7cc6 | 2071 | spin_unlock(&mm->arg_lock); |
f606b77f CG |
2072 | |
2073 | /* | |
2074 | * Note this update of @saved_auxv is lockless thus | |
2075 | * if someone reads this member in procfs while we're | |
2076 | * updating -- it may get partly updated results. It's | |
2077 | * known and acceptable trade off: we leave it as is to | |
2078 | * not introduce additional locks here making the kernel | |
2079 | * more complex. | |
2080 | */ | |
2081 | if (prctl_map.auxv_size) | |
2082 | memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv)); | |
2083 | ||
d8ed45c5 | 2084 | mmap_read_unlock(mm); |
ddf1d398 | 2085 | return 0; |
f606b77f CG |
2086 | } |
2087 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
2088 | ||
4a00e9df AD |
2089 | static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr, |
2090 | unsigned long len) | |
2091 | { | |
2092 | /* | |
2093 | * This doesn't move the auxiliary vector itself since it's pinned to | |
2094 | * mm_struct, but it permits filling the vector with new values. It's | |
2095 | * up to the caller to provide sane values here, otherwise userspace | |
2096 | * tools which use this vector might be unhappy. | |
2097 | */ | |
2098 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
2099 | ||
2100 | if (len > sizeof(user_auxv)) | |
2101 | return -EINVAL; | |
2102 | ||
2103 | if (copy_from_user(user_auxv, (const void __user *)addr, len)) | |
2104 | return -EFAULT; | |
2105 | ||
2106 | /* Make sure the last entry is always AT_NULL */ | |
2107 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
2108 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
2109 | ||
2110 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
2111 | ||
2112 | task_lock(current); | |
2113 | memcpy(mm->saved_auxv, user_auxv, len); | |
2114 | task_unlock(current); | |
2115 | ||
2116 | return 0; | |
2117 | } | |
2118 | ||
028ee4be CG |
2119 | static int prctl_set_mm(int opt, unsigned long addr, |
2120 | unsigned long arg4, unsigned long arg5) | |
2121 | { | |
028ee4be | 2122 | struct mm_struct *mm = current->mm; |
11bbd8b4 MK |
2123 | struct prctl_mm_map prctl_map = { |
2124 | .auxv = NULL, | |
2125 | .auxv_size = 0, | |
2126 | .exe_fd = -1, | |
2127 | }; | |
fe8c7f5c CG |
2128 | struct vm_area_struct *vma; |
2129 | int error; | |
028ee4be | 2130 | |
f606b77f CG |
2131 | if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV && |
2132 | opt != PR_SET_MM_MAP && | |
2133 | opt != PR_SET_MM_MAP_SIZE))) | |
028ee4be CG |
2134 | return -EINVAL; |
2135 | ||
f606b77f CG |
2136 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2137 | if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE) | |
2138 | return prctl_set_mm_map(opt, (const void __user *)addr, arg4); | |
2139 | #endif | |
2140 | ||
79f0713d | 2141 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
2142 | return -EPERM; |
2143 | ||
6e399cd1 DB |
2144 | if (opt == PR_SET_MM_EXE_FILE) |
2145 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
b32dfe37 | 2146 | |
4a00e9df AD |
2147 | if (opt == PR_SET_MM_AUXV) |
2148 | return prctl_set_auxv(mm, addr, arg4); | |
2149 | ||
1ad75b9e | 2150 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
2151 | return -EINVAL; |
2152 | ||
fe8c7f5c CG |
2153 | error = -EINVAL; |
2154 | ||
bc81426f MK |
2155 | /* |
2156 | * arg_lock protects concurent updates of arg boundaries, we need | |
c1e8d7c6 | 2157 | * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr |
bc81426f MK |
2158 | * validation. |
2159 | */ | |
d8ed45c5 | 2160 | mmap_read_lock(mm); |
028ee4be CG |
2161 | vma = find_vma(mm, addr); |
2162 | ||
bc81426f | 2163 | spin_lock(&mm->arg_lock); |
4a00e9df AD |
2164 | prctl_map.start_code = mm->start_code; |
2165 | prctl_map.end_code = mm->end_code; | |
2166 | prctl_map.start_data = mm->start_data; | |
2167 | prctl_map.end_data = mm->end_data; | |
2168 | prctl_map.start_brk = mm->start_brk; | |
2169 | prctl_map.brk = mm->brk; | |
2170 | prctl_map.start_stack = mm->start_stack; | |
2171 | prctl_map.arg_start = mm->arg_start; | |
2172 | prctl_map.arg_end = mm->arg_end; | |
2173 | prctl_map.env_start = mm->env_start; | |
2174 | prctl_map.env_end = mm->env_end; | |
4a00e9df | 2175 | |
028ee4be CG |
2176 | switch (opt) { |
2177 | case PR_SET_MM_START_CODE: | |
4a00e9df | 2178 | prctl_map.start_code = addr; |
fe8c7f5c | 2179 | break; |
028ee4be | 2180 | case PR_SET_MM_END_CODE: |
4a00e9df | 2181 | prctl_map.end_code = addr; |
028ee4be | 2182 | break; |
028ee4be | 2183 | case PR_SET_MM_START_DATA: |
4a00e9df | 2184 | prctl_map.start_data = addr; |
028ee4be | 2185 | break; |
fe8c7f5c | 2186 | case PR_SET_MM_END_DATA: |
4a00e9df AD |
2187 | prctl_map.end_data = addr; |
2188 | break; | |
2189 | case PR_SET_MM_START_STACK: | |
2190 | prctl_map.start_stack = addr; | |
028ee4be | 2191 | break; |
028ee4be | 2192 | case PR_SET_MM_START_BRK: |
4a00e9df | 2193 | prctl_map.start_brk = addr; |
028ee4be | 2194 | break; |
028ee4be | 2195 | case PR_SET_MM_BRK: |
4a00e9df | 2196 | prctl_map.brk = addr; |
028ee4be | 2197 | break; |
4a00e9df AD |
2198 | case PR_SET_MM_ARG_START: |
2199 | prctl_map.arg_start = addr; | |
2200 | break; | |
2201 | case PR_SET_MM_ARG_END: | |
2202 | prctl_map.arg_end = addr; | |
2203 | break; | |
2204 | case PR_SET_MM_ENV_START: | |
2205 | prctl_map.env_start = addr; | |
2206 | break; | |
2207 | case PR_SET_MM_ENV_END: | |
2208 | prctl_map.env_end = addr; | |
2209 | break; | |
2210 | default: | |
2211 | goto out; | |
2212 | } | |
2213 | ||
11bbd8b4 | 2214 | error = validate_prctl_map_addr(&prctl_map); |
4a00e9df AD |
2215 | if (error) |
2216 | goto out; | |
028ee4be | 2217 | |
4a00e9df | 2218 | switch (opt) { |
fe8c7f5c CG |
2219 | /* |
2220 | * If command line arguments and environment | |
2221 | * are placed somewhere else on stack, we can | |
2222 | * set them up here, ARG_START/END to setup | |
2223 | * command line argumets and ENV_START/END | |
2224 | * for environment. | |
2225 | */ | |
2226 | case PR_SET_MM_START_STACK: | |
2227 | case PR_SET_MM_ARG_START: | |
2228 | case PR_SET_MM_ARG_END: | |
2229 | case PR_SET_MM_ENV_START: | |
2230 | case PR_SET_MM_ENV_END: | |
2231 | if (!vma) { | |
2232 | error = -EFAULT; | |
2233 | goto out; | |
2234 | } | |
028ee4be CG |
2235 | } |
2236 | ||
4a00e9df AD |
2237 | mm->start_code = prctl_map.start_code; |
2238 | mm->end_code = prctl_map.end_code; | |
2239 | mm->start_data = prctl_map.start_data; | |
2240 | mm->end_data = prctl_map.end_data; | |
2241 | mm->start_brk = prctl_map.start_brk; | |
2242 | mm->brk = prctl_map.brk; | |
2243 | mm->start_stack = prctl_map.start_stack; | |
2244 | mm->arg_start = prctl_map.arg_start; | |
2245 | mm->arg_end = prctl_map.arg_end; | |
2246 | mm->env_start = prctl_map.env_start; | |
2247 | mm->env_end = prctl_map.env_end; | |
2248 | ||
028ee4be | 2249 | error = 0; |
028ee4be | 2250 | out: |
bc81426f | 2251 | spin_unlock(&mm->arg_lock); |
d8ed45c5 | 2252 | mmap_read_unlock(mm); |
028ee4be CG |
2253 | return error; |
2254 | } | |
300f786b | 2255 | |
52b36941 | 2256 | #ifdef CONFIG_CHECKPOINT_RESTORE |
986b9eac | 2257 | static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr) |
300f786b CG |
2258 | { |
2259 | return put_user(me->clear_child_tid, tid_addr); | |
2260 | } | |
52b36941 | 2261 | #else |
986b9eac | 2262 | static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr) |
300f786b CG |
2263 | { |
2264 | return -EINVAL; | |
2265 | } | |
028ee4be CG |
2266 | #endif |
2267 | ||
749860ce PT |
2268 | static int propagate_has_child_subreaper(struct task_struct *p, void *data) |
2269 | { | |
2270 | /* | |
2271 | * If task has has_child_subreaper - all its decendants | |
2272 | * already have these flag too and new decendants will | |
2273 | * inherit it on fork, skip them. | |
2274 | * | |
2275 | * If we've found child_reaper - skip descendants in | |
2276 | * it's subtree as they will never get out pidns. | |
2277 | */ | |
2278 | if (p->signal->has_child_subreaper || | |
2279 | is_child_reaper(task_pid(p))) | |
2280 | return 0; | |
2281 | ||
2282 | p->signal->has_child_subreaper = 1; | |
2283 | return 1; | |
2284 | } | |
2285 | ||
7bbf1373 | 2286 | int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which) |
b617cfc8 TG |
2287 | { |
2288 | return -EINVAL; | |
2289 | } | |
2290 | ||
7bbf1373 KC |
2291 | int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which, |
2292 | unsigned long ctrl) | |
b617cfc8 TG |
2293 | { |
2294 | return -EINVAL; | |
2295 | } | |
2296 | ||
a37b0715 | 2297 | #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE) |
8d19f1c8 | 2298 | |
c4ea37c2 HC |
2299 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2300 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2301 | { |
b6dff3ec DH |
2302 | struct task_struct *me = current; |
2303 | unsigned char comm[sizeof(me->comm)]; | |
2304 | long error; | |
1da177e4 | 2305 | |
d84f4f99 DH |
2306 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2307 | if (error != -ENOSYS) | |
1da177e4 LT |
2308 | return error; |
2309 | ||
d84f4f99 | 2310 | error = 0; |
1da177e4 | 2311 | switch (option) { |
f3cbd435 AM |
2312 | case PR_SET_PDEATHSIG: |
2313 | if (!valid_signal(arg2)) { | |
2314 | error = -EINVAL; | |
1da177e4 | 2315 | break; |
f3cbd435 AM |
2316 | } |
2317 | me->pdeath_signal = arg2; | |
2318 | break; | |
2319 | case PR_GET_PDEATHSIG: | |
2320 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2321 | break; | |
2322 | case PR_GET_DUMPABLE: | |
2323 | error = get_dumpable(me->mm); | |
2324 | break; | |
2325 | case PR_SET_DUMPABLE: | |
2326 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2327 | error = -EINVAL; | |
1da177e4 | 2328 | break; |
f3cbd435 AM |
2329 | } |
2330 | set_dumpable(me->mm, arg2); | |
2331 | break; | |
1da177e4 | 2332 | |
f3cbd435 AM |
2333 | case PR_SET_UNALIGN: |
2334 | error = SET_UNALIGN_CTL(me, arg2); | |
2335 | break; | |
2336 | case PR_GET_UNALIGN: | |
2337 | error = GET_UNALIGN_CTL(me, arg2); | |
2338 | break; | |
2339 | case PR_SET_FPEMU: | |
2340 | error = SET_FPEMU_CTL(me, arg2); | |
2341 | break; | |
2342 | case PR_GET_FPEMU: | |
2343 | error = GET_FPEMU_CTL(me, arg2); | |
2344 | break; | |
2345 | case PR_SET_FPEXC: | |
2346 | error = SET_FPEXC_CTL(me, arg2); | |
2347 | break; | |
2348 | case PR_GET_FPEXC: | |
2349 | error = GET_FPEXC_CTL(me, arg2); | |
2350 | break; | |
2351 | case PR_GET_TIMING: | |
2352 | error = PR_TIMING_STATISTICAL; | |
2353 | break; | |
2354 | case PR_SET_TIMING: | |
2355 | if (arg2 != PR_TIMING_STATISTICAL) | |
2356 | error = -EINVAL; | |
2357 | break; | |
2358 | case PR_SET_NAME: | |
2359 | comm[sizeof(me->comm) - 1] = 0; | |
2360 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2361 | sizeof(me->comm) - 1) < 0) | |
2362 | return -EFAULT; | |
2363 | set_task_comm(me, comm); | |
2364 | proc_comm_connector(me); | |
2365 | break; | |
2366 | case PR_GET_NAME: | |
2367 | get_task_comm(comm, me); | |
2368 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2369 | return -EFAULT; | |
2370 | break; | |
2371 | case PR_GET_ENDIAN: | |
2372 | error = GET_ENDIAN(me, arg2); | |
2373 | break; | |
2374 | case PR_SET_ENDIAN: | |
2375 | error = SET_ENDIAN(me, arg2); | |
2376 | break; | |
2377 | case PR_GET_SECCOMP: | |
2378 | error = prctl_get_seccomp(); | |
2379 | break; | |
2380 | case PR_SET_SECCOMP: | |
2381 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2382 | break; | |
2383 | case PR_GET_TSC: | |
2384 | error = GET_TSC_CTL(arg2); | |
2385 | break; | |
2386 | case PR_SET_TSC: | |
2387 | error = SET_TSC_CTL(arg2); | |
2388 | break; | |
2389 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2390 | error = perf_event_task_disable(); | |
2391 | break; | |
2392 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2393 | error = perf_event_task_enable(); | |
2394 | break; | |
2395 | case PR_GET_TIMERSLACK: | |
da8b44d5 JS |
2396 | if (current->timer_slack_ns > ULONG_MAX) |
2397 | error = ULONG_MAX; | |
2398 | else | |
2399 | error = current->timer_slack_ns; | |
f3cbd435 AM |
2400 | break; |
2401 | case PR_SET_TIMERSLACK: | |
2402 | if (arg2 <= 0) | |
2403 | current->timer_slack_ns = | |
6976675d | 2404 | current->default_timer_slack_ns; |
f3cbd435 AM |
2405 | else |
2406 | current->timer_slack_ns = arg2; | |
2407 | break; | |
2408 | case PR_MCE_KILL: | |
2409 | if (arg4 | arg5) | |
2410 | return -EINVAL; | |
2411 | switch (arg2) { | |
2412 | case PR_MCE_KILL_CLEAR: | |
2413 | if (arg3 != 0) | |
4db96cf0 | 2414 | return -EINVAL; |
f3cbd435 | 2415 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2416 | break; |
f3cbd435 AM |
2417 | case PR_MCE_KILL_SET: |
2418 | current->flags |= PF_MCE_PROCESS; | |
2419 | if (arg3 == PR_MCE_KILL_EARLY) | |
2420 | current->flags |= PF_MCE_EARLY; | |
2421 | else if (arg3 == PR_MCE_KILL_LATE) | |
2422 | current->flags &= ~PF_MCE_EARLY; | |
2423 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2424 | current->flags &= | |
2425 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2426 | else |
259e5e6c | 2427 | return -EINVAL; |
259e5e6c | 2428 | break; |
1da177e4 | 2429 | default: |
f3cbd435 AM |
2430 | return -EINVAL; |
2431 | } | |
2432 | break; | |
2433 | case PR_MCE_KILL_GET: | |
2434 | if (arg2 | arg3 | arg4 | arg5) | |
2435 | return -EINVAL; | |
2436 | if (current->flags & PF_MCE_PROCESS) | |
2437 | error = (current->flags & PF_MCE_EARLY) ? | |
2438 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2439 | else | |
2440 | error = PR_MCE_KILL_DEFAULT; | |
2441 | break; | |
2442 | case PR_SET_MM: | |
2443 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2444 | break; | |
2445 | case PR_GET_TID_ADDRESS: | |
986b9eac | 2446 | error = prctl_get_tid_address(me, (int __user * __user *)arg2); |
f3cbd435 AM |
2447 | break; |
2448 | case PR_SET_CHILD_SUBREAPER: | |
2449 | me->signal->is_child_subreaper = !!arg2; | |
749860ce PT |
2450 | if (!arg2) |
2451 | break; | |
2452 | ||
2453 | walk_process_tree(me, propagate_has_child_subreaper, NULL); | |
f3cbd435 AM |
2454 | break; |
2455 | case PR_GET_CHILD_SUBREAPER: | |
2456 | error = put_user(me->signal->is_child_subreaper, | |
2457 | (int __user *)arg2); | |
2458 | break; | |
2459 | case PR_SET_NO_NEW_PRIVS: | |
2460 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2461 | return -EINVAL; | |
2462 | ||
1d4457f9 | 2463 | task_set_no_new_privs(current); |
f3cbd435 AM |
2464 | break; |
2465 | case PR_GET_NO_NEW_PRIVS: | |
2466 | if (arg2 || arg3 || arg4 || arg5) | |
2467 | return -EINVAL; | |
1d4457f9 | 2468 | return task_no_new_privs(current) ? 1 : 0; |
a0715cc2 AT |
2469 | case PR_GET_THP_DISABLE: |
2470 | if (arg2 || arg3 || arg4 || arg5) | |
2471 | return -EINVAL; | |
18600332 | 2472 | error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 AT |
2473 | break; |
2474 | case PR_SET_THP_DISABLE: | |
2475 | if (arg3 || arg4 || arg5) | |
2476 | return -EINVAL; | |
d8ed45c5 | 2477 | if (mmap_write_lock_killable(me->mm)) |
17b0573d | 2478 | return -EINTR; |
a0715cc2 | 2479 | if (arg2) |
18600332 | 2480 | set_bit(MMF_DISABLE_THP, &me->mm->flags); |
a0715cc2 | 2481 | else |
18600332 | 2482 | clear_bit(MMF_DISABLE_THP, &me->mm->flags); |
d8ed45c5 | 2483 | mmap_write_unlock(me->mm); |
a0715cc2 | 2484 | break; |
fe3d197f | 2485 | case PR_MPX_ENABLE_MANAGEMENT: |
fe3d197f | 2486 | case PR_MPX_DISABLE_MANAGEMENT: |
f240652b DH |
2487 | /* No longer implemented: */ |
2488 | return -EINVAL; | |
9791554b PB |
2489 | case PR_SET_FP_MODE: |
2490 | error = SET_FP_MODE(me, arg2); | |
2491 | break; | |
2492 | case PR_GET_FP_MODE: | |
2493 | error = GET_FP_MODE(me); | |
2494 | break; | |
2d2123bc DM |
2495 | case PR_SVE_SET_VL: |
2496 | error = SVE_SET_VL(arg2); | |
2497 | break; | |
2498 | case PR_SVE_GET_VL: | |
2499 | error = SVE_GET_VL(); | |
2500 | break; | |
b617cfc8 TG |
2501 | case PR_GET_SPECULATION_CTRL: |
2502 | if (arg3 || arg4 || arg5) | |
2503 | return -EINVAL; | |
7bbf1373 | 2504 | error = arch_prctl_spec_ctrl_get(me, arg2); |
b617cfc8 TG |
2505 | break; |
2506 | case PR_SET_SPECULATION_CTRL: | |
2507 | if (arg4 || arg5) | |
2508 | return -EINVAL; | |
7bbf1373 | 2509 | error = arch_prctl_spec_ctrl_set(me, arg2, arg3); |
b617cfc8 | 2510 | break; |
ba830885 KM |
2511 | case PR_PAC_RESET_KEYS: |
2512 | if (arg3 || arg4 || arg5) | |
2513 | return -EINVAL; | |
2514 | error = PAC_RESET_KEYS(me, arg2); | |
2515 | break; | |
63f0c603 | 2516 | case PR_SET_TAGGED_ADDR_CTRL: |
3e91ec89 CM |
2517 | if (arg3 || arg4 || arg5) |
2518 | return -EINVAL; | |
63f0c603 CM |
2519 | error = SET_TAGGED_ADDR_CTRL(arg2); |
2520 | break; | |
2521 | case PR_GET_TAGGED_ADDR_CTRL: | |
3e91ec89 CM |
2522 | if (arg2 || arg3 || arg4 || arg5) |
2523 | return -EINVAL; | |
63f0c603 CM |
2524 | error = GET_TAGGED_ADDR_CTRL(); |
2525 | break; | |
8d19f1c8 MC |
2526 | case PR_SET_IO_FLUSHER: |
2527 | if (!capable(CAP_SYS_RESOURCE)) | |
2528 | return -EPERM; | |
2529 | ||
2530 | if (arg3 || arg4 || arg5) | |
2531 | return -EINVAL; | |
2532 | ||
2533 | if (arg2 == 1) | |
2534 | current->flags |= PR_IO_FLUSHER; | |
2535 | else if (!arg2) | |
2536 | current->flags &= ~PR_IO_FLUSHER; | |
2537 | else | |
2538 | return -EINVAL; | |
2539 | break; | |
2540 | case PR_GET_IO_FLUSHER: | |
2541 | if (!capable(CAP_SYS_RESOURCE)) | |
2542 | return -EPERM; | |
2543 | ||
2544 | if (arg2 || arg3 || arg4 || arg5) | |
2545 | return -EINVAL; | |
2546 | ||
2547 | error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER; | |
2548 | break; | |
1446e1df GKB |
2549 | case PR_SET_SYSCALL_USER_DISPATCH: |
2550 | error = set_syscall_user_dispatch(arg2, arg3, arg4, | |
2551 | (char __user *) arg5); | |
2552 | break; | |
f3cbd435 AM |
2553 | default: |
2554 | error = -EINVAL; | |
2555 | break; | |
1da177e4 LT |
2556 | } |
2557 | return error; | |
2558 | } | |
3cfc348b | 2559 | |
836f92ad HC |
2560 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2561 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2562 | { |
2563 | int err = 0; | |
2564 | int cpu = raw_smp_processor_id(); | |
ec94fc3d | 2565 | |
3cfc348b AK |
2566 | if (cpup) |
2567 | err |= put_user(cpu, cpup); | |
2568 | if (nodep) | |
2569 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2570 | return err ? -EFAULT : 0; |
2571 | } | |
10a0a8d4 | 2572 | |
4a22f166 SR |
2573 | /** |
2574 | * do_sysinfo - fill in sysinfo struct | |
2575 | * @info: pointer to buffer to fill | |
2576 | */ | |
2577 | static int do_sysinfo(struct sysinfo *info) | |
2578 | { | |
2579 | unsigned long mem_total, sav_total; | |
2580 | unsigned int mem_unit, bitcount; | |
dc1b7b6c | 2581 | struct timespec64 tp; |
4a22f166 SR |
2582 | |
2583 | memset(info, 0, sizeof(struct sysinfo)); | |
2584 | ||
dc1b7b6c | 2585 | ktime_get_boottime_ts64(&tp); |
ecc421e0 | 2586 | timens_add_boottime(&tp); |
4a22f166 SR |
2587 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
2588 | ||
2589 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2590 | ||
2591 | info->procs = nr_threads; | |
2592 | ||
2593 | si_meminfo(info); | |
2594 | si_swapinfo(info); | |
2595 | ||
2596 | /* | |
2597 | * If the sum of all the available memory (i.e. ram + swap) | |
2598 | * is less than can be stored in a 32 bit unsigned long then | |
2599 | * we can be binary compatible with 2.2.x kernels. If not, | |
2600 | * well, in that case 2.2.x was broken anyways... | |
2601 | * | |
2602 | * -Erik Andersen <andersee@debian.org> | |
2603 | */ | |
2604 | ||
2605 | mem_total = info->totalram + info->totalswap; | |
2606 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2607 | goto out; | |
2608 | bitcount = 0; | |
2609 | mem_unit = info->mem_unit; | |
2610 | while (mem_unit > 1) { | |
2611 | bitcount++; | |
2612 | mem_unit >>= 1; | |
2613 | sav_total = mem_total; | |
2614 | mem_total <<= 1; | |
2615 | if (mem_total < sav_total) | |
2616 | goto out; | |
2617 | } | |
2618 | ||
2619 | /* | |
2620 | * If mem_total did not overflow, multiply all memory values by | |
2621 | * info->mem_unit and set it to 1. This leaves things compatible | |
2622 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2623 | * kernels... | |
2624 | */ | |
2625 | ||
2626 | info->mem_unit = 1; | |
2627 | info->totalram <<= bitcount; | |
2628 | info->freeram <<= bitcount; | |
2629 | info->sharedram <<= bitcount; | |
2630 | info->bufferram <<= bitcount; | |
2631 | info->totalswap <<= bitcount; | |
2632 | info->freeswap <<= bitcount; | |
2633 | info->totalhigh <<= bitcount; | |
2634 | info->freehigh <<= bitcount; | |
2635 | ||
2636 | out: | |
2637 | return 0; | |
2638 | } | |
2639 | ||
2640 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2641 | { | |
2642 | struct sysinfo val; | |
2643 | ||
2644 | do_sysinfo(&val); | |
2645 | ||
2646 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2647 | return -EFAULT; | |
2648 | ||
2649 | return 0; | |
2650 | } | |
2651 | ||
2652 | #ifdef CONFIG_COMPAT | |
2653 | struct compat_sysinfo { | |
2654 | s32 uptime; | |
2655 | u32 loads[3]; | |
2656 | u32 totalram; | |
2657 | u32 freeram; | |
2658 | u32 sharedram; | |
2659 | u32 bufferram; | |
2660 | u32 totalswap; | |
2661 | u32 freeswap; | |
2662 | u16 procs; | |
2663 | u16 pad; | |
2664 | u32 totalhigh; | |
2665 | u32 freehigh; | |
2666 | u32 mem_unit; | |
2667 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2668 | }; | |
2669 | ||
2670 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2671 | { | |
2672 | struct sysinfo s; | |
ce5155c4 | 2673 | struct compat_sysinfo s_32; |
4a22f166 SR |
2674 | |
2675 | do_sysinfo(&s); | |
2676 | ||
2677 | /* Check to see if any memory value is too large for 32-bit and scale | |
2678 | * down if needed | |
2679 | */ | |
0baae41e | 2680 | if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) { |
4a22f166 SR |
2681 | int bitcount = 0; |
2682 | ||
2683 | while (s.mem_unit < PAGE_SIZE) { | |
2684 | s.mem_unit <<= 1; | |
2685 | bitcount++; | |
2686 | } | |
2687 | ||
2688 | s.totalram >>= bitcount; | |
2689 | s.freeram >>= bitcount; | |
2690 | s.sharedram >>= bitcount; | |
2691 | s.bufferram >>= bitcount; | |
2692 | s.totalswap >>= bitcount; | |
2693 | s.freeswap >>= bitcount; | |
2694 | s.totalhigh >>= bitcount; | |
2695 | s.freehigh >>= bitcount; | |
2696 | } | |
2697 | ||
ce5155c4 AV |
2698 | memset(&s_32, 0, sizeof(s_32)); |
2699 | s_32.uptime = s.uptime; | |
2700 | s_32.loads[0] = s.loads[0]; | |
2701 | s_32.loads[1] = s.loads[1]; | |
2702 | s_32.loads[2] = s.loads[2]; | |
2703 | s_32.totalram = s.totalram; | |
2704 | s_32.freeram = s.freeram; | |
2705 | s_32.sharedram = s.sharedram; | |
2706 | s_32.bufferram = s.bufferram; | |
2707 | s_32.totalswap = s.totalswap; | |
2708 | s_32.freeswap = s.freeswap; | |
2709 | s_32.procs = s.procs; | |
2710 | s_32.totalhigh = s.totalhigh; | |
2711 | s_32.freehigh = s.freehigh; | |
2712 | s_32.mem_unit = s.mem_unit; | |
2713 | if (copy_to_user(info, &s_32, sizeof(s_32))) | |
4a22f166 | 2714 | return -EFAULT; |
4a22f166 SR |
2715 | return 0; |
2716 | } | |
2717 | #endif /* CONFIG_COMPAT */ |