Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/sys.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
9984de1a | 7 | #include <linux/export.h> |
1da177e4 LT |
8 | #include <linux/mm.h> |
9 | #include <linux/utsname.h> | |
10 | #include <linux/mman.h> | |
1da177e4 LT |
11 | #include <linux/reboot.h> |
12 | #include <linux/prctl.h> | |
1da177e4 LT |
13 | #include <linux/highuid.h> |
14 | #include <linux/fs.h> | |
74da1ff7 | 15 | #include <linux/kmod.h> |
cdd6c482 | 16 | #include <linux/perf_event.h> |
3e88c553 | 17 | #include <linux/resource.h> |
dc009d92 | 18 | #include <linux/kernel.h> |
1da177e4 | 19 | #include <linux/workqueue.h> |
c59ede7b | 20 | #include <linux/capability.h> |
1da177e4 LT |
21 | #include <linux/device.h> |
22 | #include <linux/key.h> | |
23 | #include <linux/times.h> | |
24 | #include <linux/posix-timers.h> | |
25 | #include <linux/security.h> | |
26 | #include <linux/dcookies.h> | |
27 | #include <linux/suspend.h> | |
28 | #include <linux/tty.h> | |
7ed20e1a | 29 | #include <linux/signal.h> |
9f46080c | 30 | #include <linux/cn_proc.h> |
3cfc348b | 31 | #include <linux/getcpu.h> |
6eaeeaba | 32 | #include <linux/task_io_accounting_ops.h> |
1d9d02fe | 33 | #include <linux/seccomp.h> |
4047727e | 34 | #include <linux/cpu.h> |
e28cbf22 | 35 | #include <linux/personality.h> |
e3d5a27d | 36 | #include <linux/ptrace.h> |
5ad4e53b | 37 | #include <linux/fs_struct.h> |
b32dfe37 CG |
38 | #include <linux/file.h> |
39 | #include <linux/mount.h> | |
5a0e3ad6 | 40 | #include <linux/gfp.h> |
40dc166c | 41 | #include <linux/syscore_ops.h> |
be27425d AK |
42 | #include <linux/version.h> |
43 | #include <linux/ctype.h> | |
1da177e4 LT |
44 | |
45 | #include <linux/compat.h> | |
46 | #include <linux/syscalls.h> | |
00d7c05a | 47 | #include <linux/kprobes.h> |
acce292c | 48 | #include <linux/user_namespace.h> |
7fe5e042 | 49 | #include <linux/binfmts.h> |
1da177e4 | 50 | |
4a22f166 | 51 | #include <linux/sched.h> |
4eb5aaa3 | 52 | #include <linux/sched/autogroup.h> |
4f17722c | 53 | #include <linux/sched/loadavg.h> |
03441a34 | 54 | #include <linux/sched/stat.h> |
6e84f315 | 55 | #include <linux/sched/mm.h> |
f7ccbae4 | 56 | #include <linux/sched/coredump.h> |
29930025 | 57 | #include <linux/sched/task.h> |
4a22f166 SR |
58 | #include <linux/rcupdate.h> |
59 | #include <linux/uidgid.h> | |
60 | #include <linux/cred.h> | |
61 | ||
04c6862c | 62 | #include <linux/kmsg_dump.h> |
be27425d AK |
63 | /* Move somewhere else to avoid recompiling? */ |
64 | #include <generated/utsrelease.h> | |
04c6862c | 65 | |
7c0f6ba6 | 66 | #include <linux/uaccess.h> |
1da177e4 LT |
67 | #include <asm/io.h> |
68 | #include <asm/unistd.h> | |
69 | ||
70 | #ifndef SET_UNALIGN_CTL | |
ec94fc3d | 71 | # define SET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
72 | #endif |
73 | #ifndef GET_UNALIGN_CTL | |
ec94fc3d | 74 | # define GET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
75 | #endif |
76 | #ifndef SET_FPEMU_CTL | |
ec94fc3d | 77 | # define SET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
78 | #endif |
79 | #ifndef GET_FPEMU_CTL | |
ec94fc3d | 80 | # define GET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
81 | #endif |
82 | #ifndef SET_FPEXC_CTL | |
ec94fc3d | 83 | # define SET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 LT |
84 | #endif |
85 | #ifndef GET_FPEXC_CTL | |
ec94fc3d | 86 | # define GET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 | 87 | #endif |
651d765d | 88 | #ifndef GET_ENDIAN |
ec94fc3d | 89 | # define GET_ENDIAN(a, b) (-EINVAL) |
651d765d AB |
90 | #endif |
91 | #ifndef SET_ENDIAN | |
ec94fc3d | 92 | # define SET_ENDIAN(a, b) (-EINVAL) |
651d765d | 93 | #endif |
8fb402bc EB |
94 | #ifndef GET_TSC_CTL |
95 | # define GET_TSC_CTL(a) (-EINVAL) | |
96 | #endif | |
97 | #ifndef SET_TSC_CTL | |
98 | # define SET_TSC_CTL(a) (-EINVAL) | |
99 | #endif | |
fe3d197f | 100 | #ifndef MPX_ENABLE_MANAGEMENT |
46a6e0cf | 101 | # define MPX_ENABLE_MANAGEMENT() (-EINVAL) |
fe3d197f DH |
102 | #endif |
103 | #ifndef MPX_DISABLE_MANAGEMENT | |
46a6e0cf | 104 | # define MPX_DISABLE_MANAGEMENT() (-EINVAL) |
fe3d197f | 105 | #endif |
9791554b PB |
106 | #ifndef GET_FP_MODE |
107 | # define GET_FP_MODE(a) (-EINVAL) | |
108 | #endif | |
109 | #ifndef SET_FP_MODE | |
110 | # define SET_FP_MODE(a,b) (-EINVAL) | |
111 | #endif | |
1da177e4 LT |
112 | |
113 | /* | |
114 | * this is where the system-wide overflow UID and GID are defined, for | |
115 | * architectures that now have 32-bit UID/GID but didn't in the past | |
116 | */ | |
117 | ||
118 | int overflowuid = DEFAULT_OVERFLOWUID; | |
119 | int overflowgid = DEFAULT_OVERFLOWGID; | |
120 | ||
1da177e4 LT |
121 | EXPORT_SYMBOL(overflowuid); |
122 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
123 | |
124 | /* | |
125 | * the same as above, but for filesystems which can only store a 16-bit | |
126 | * UID and GID. as such, this is needed on all architectures | |
127 | */ | |
128 | ||
129 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
130 | int fs_overflowgid = DEFAULT_FS_OVERFLOWUID; | |
131 | ||
132 | EXPORT_SYMBOL(fs_overflowuid); | |
133 | EXPORT_SYMBOL(fs_overflowgid); | |
134 | ||
fc832ad3 SH |
135 | /* |
136 | * Returns true if current's euid is same as p's uid or euid, | |
137 | * or has CAP_SYS_NICE to p's user_ns. | |
138 | * | |
139 | * Called with rcu_read_lock, creds are safe | |
140 | */ | |
141 | static bool set_one_prio_perm(struct task_struct *p) | |
142 | { | |
143 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
144 | ||
5af66203 EB |
145 | if (uid_eq(pcred->uid, cred->euid) || |
146 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 147 | return true; |
c4a4d603 | 148 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
149 | return true; |
150 | return false; | |
151 | } | |
152 | ||
c69e8d9c DH |
153 | /* |
154 | * set the priority of a task | |
155 | * - the caller must hold the RCU read lock | |
156 | */ | |
1da177e4 LT |
157 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
158 | { | |
159 | int no_nice; | |
160 | ||
fc832ad3 | 161 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
162 | error = -EPERM; |
163 | goto out; | |
164 | } | |
e43379f1 | 165 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
166 | error = -EACCES; |
167 | goto out; | |
168 | } | |
169 | no_nice = security_task_setnice(p, niceval); | |
170 | if (no_nice) { | |
171 | error = no_nice; | |
172 | goto out; | |
173 | } | |
174 | if (error == -ESRCH) | |
175 | error = 0; | |
176 | set_user_nice(p, niceval); | |
177 | out: | |
178 | return error; | |
179 | } | |
180 | ||
754fe8d2 | 181 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
182 | { |
183 | struct task_struct *g, *p; | |
184 | struct user_struct *user; | |
86a264ab | 185 | const struct cred *cred = current_cred(); |
1da177e4 | 186 | int error = -EINVAL; |
41487c65 | 187 | struct pid *pgrp; |
7b44ab97 | 188 | kuid_t uid; |
1da177e4 | 189 | |
3e88c553 | 190 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
191 | goto out; |
192 | ||
193 | /* normalize: avoid signed division (rounding problems) */ | |
194 | error = -ESRCH; | |
c4a4d2f4 DY |
195 | if (niceval < MIN_NICE) |
196 | niceval = MIN_NICE; | |
197 | if (niceval > MAX_NICE) | |
198 | niceval = MAX_NICE; | |
1da177e4 | 199 | |
d4581a23 | 200 | rcu_read_lock(); |
1da177e4 LT |
201 | read_lock(&tasklist_lock); |
202 | switch (which) { | |
ec94fc3d | 203 | case PRIO_PROCESS: |
204 | if (who) | |
205 | p = find_task_by_vpid(who); | |
206 | else | |
207 | p = current; | |
208 | if (p) | |
209 | error = set_one_prio(p, niceval, error); | |
210 | break; | |
211 | case PRIO_PGRP: | |
212 | if (who) | |
213 | pgrp = find_vpid(who); | |
214 | else | |
215 | pgrp = task_pgrp(current); | |
216 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
217 | error = set_one_prio(p, niceval, error); | |
218 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
219 | break; | |
220 | case PRIO_USER: | |
221 | uid = make_kuid(cred->user_ns, who); | |
222 | user = cred->user; | |
223 | if (!who) | |
224 | uid = cred->uid; | |
225 | else if (!uid_eq(uid, cred->uid)) { | |
226 | user = find_user(uid); | |
227 | if (!user) | |
86a264ab | 228 | goto out_unlock; /* No processes for this user */ |
ec94fc3d | 229 | } |
230 | do_each_thread(g, p) { | |
8639b461 | 231 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) |
ec94fc3d | 232 | error = set_one_prio(p, niceval, error); |
233 | } while_each_thread(g, p); | |
234 | if (!uid_eq(uid, cred->uid)) | |
235 | free_uid(user); /* For find_user() */ | |
236 | break; | |
1da177e4 LT |
237 | } |
238 | out_unlock: | |
239 | read_unlock(&tasklist_lock); | |
d4581a23 | 240 | rcu_read_unlock(); |
1da177e4 LT |
241 | out: |
242 | return error; | |
243 | } | |
244 | ||
245 | /* | |
246 | * Ugh. To avoid negative return values, "getpriority()" will | |
247 | * not return the normal nice-value, but a negated value that | |
248 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
249 | * to stay compatible. | |
250 | */ | |
754fe8d2 | 251 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
252 | { |
253 | struct task_struct *g, *p; | |
254 | struct user_struct *user; | |
86a264ab | 255 | const struct cred *cred = current_cred(); |
1da177e4 | 256 | long niceval, retval = -ESRCH; |
41487c65 | 257 | struct pid *pgrp; |
7b44ab97 | 258 | kuid_t uid; |
1da177e4 | 259 | |
3e88c553 | 260 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
261 | return -EINVAL; |
262 | ||
70118837 | 263 | rcu_read_lock(); |
1da177e4 LT |
264 | read_lock(&tasklist_lock); |
265 | switch (which) { | |
ec94fc3d | 266 | case PRIO_PROCESS: |
267 | if (who) | |
268 | p = find_task_by_vpid(who); | |
269 | else | |
270 | p = current; | |
271 | if (p) { | |
272 | niceval = nice_to_rlimit(task_nice(p)); | |
273 | if (niceval > retval) | |
274 | retval = niceval; | |
275 | } | |
276 | break; | |
277 | case PRIO_PGRP: | |
278 | if (who) | |
279 | pgrp = find_vpid(who); | |
280 | else | |
281 | pgrp = task_pgrp(current); | |
282 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
283 | niceval = nice_to_rlimit(task_nice(p)); | |
284 | if (niceval > retval) | |
285 | retval = niceval; | |
286 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
287 | break; | |
288 | case PRIO_USER: | |
289 | uid = make_kuid(cred->user_ns, who); | |
290 | user = cred->user; | |
291 | if (!who) | |
292 | uid = cred->uid; | |
293 | else if (!uid_eq(uid, cred->uid)) { | |
294 | user = find_user(uid); | |
295 | if (!user) | |
296 | goto out_unlock; /* No processes for this user */ | |
297 | } | |
298 | do_each_thread(g, p) { | |
8639b461 | 299 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) { |
7aa2c016 | 300 | niceval = nice_to_rlimit(task_nice(p)); |
1da177e4 LT |
301 | if (niceval > retval) |
302 | retval = niceval; | |
303 | } | |
ec94fc3d | 304 | } while_each_thread(g, p); |
305 | if (!uid_eq(uid, cred->uid)) | |
306 | free_uid(user); /* for find_user() */ | |
307 | break; | |
1da177e4 LT |
308 | } |
309 | out_unlock: | |
310 | read_unlock(&tasklist_lock); | |
70118837 | 311 | rcu_read_unlock(); |
1da177e4 LT |
312 | |
313 | return retval; | |
314 | } | |
315 | ||
1da177e4 LT |
316 | /* |
317 | * Unprivileged users may change the real gid to the effective gid | |
318 | * or vice versa. (BSD-style) | |
319 | * | |
320 | * If you set the real gid at all, or set the effective gid to a value not | |
321 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
322 | * | |
323 | * This makes it possible for a setgid program to completely drop its | |
324 | * privileges, which is often a useful assertion to make when you are doing | |
325 | * a security audit over a program. | |
326 | * | |
327 | * The general idea is that a program which uses just setregid() will be | |
328 | * 100% compatible with BSD. A program which uses just setgid() will be | |
ec94fc3d | 329 | * 100% compatible with POSIX with saved IDs. |
1da177e4 LT |
330 | * |
331 | * SMP: There are not races, the GIDs are checked only by filesystem | |
332 | * operations (as far as semantic preservation is concerned). | |
333 | */ | |
2813893f | 334 | #ifdef CONFIG_MULTIUSER |
ae1251ab | 335 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
1da177e4 | 336 | { |
a29c33f4 | 337 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
338 | const struct cred *old; |
339 | struct cred *new; | |
1da177e4 | 340 | int retval; |
a29c33f4 EB |
341 | kgid_t krgid, kegid; |
342 | ||
343 | krgid = make_kgid(ns, rgid); | |
344 | kegid = make_kgid(ns, egid); | |
345 | ||
346 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
347 | return -EINVAL; | |
348 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
349 | return -EINVAL; | |
1da177e4 | 350 | |
d84f4f99 DH |
351 | new = prepare_creds(); |
352 | if (!new) | |
353 | return -ENOMEM; | |
354 | old = current_cred(); | |
355 | ||
d84f4f99 | 356 | retval = -EPERM; |
1da177e4 | 357 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
358 | if (gid_eq(old->gid, krgid) || |
359 | gid_eq(old->egid, krgid) || | |
c7b96acf | 360 | ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 | 361 | new->gid = krgid; |
1da177e4 | 362 | else |
d84f4f99 | 363 | goto error; |
1da177e4 LT |
364 | } |
365 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
366 | if (gid_eq(old->gid, kegid) || |
367 | gid_eq(old->egid, kegid) || | |
368 | gid_eq(old->sgid, kegid) || | |
c7b96acf | 369 | ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 | 370 | new->egid = kegid; |
756184b7 | 371 | else |
d84f4f99 | 372 | goto error; |
1da177e4 | 373 | } |
d84f4f99 | 374 | |
1da177e4 | 375 | if (rgid != (gid_t) -1 || |
a29c33f4 | 376 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
377 | new->sgid = new->egid; |
378 | new->fsgid = new->egid; | |
379 | ||
380 | return commit_creds(new); | |
381 | ||
382 | error: | |
383 | abort_creds(new); | |
384 | return retval; | |
1da177e4 LT |
385 | } |
386 | ||
387 | /* | |
ec94fc3d | 388 | * setgid() is implemented like SysV w/ SAVED_IDS |
1da177e4 LT |
389 | * |
390 | * SMP: Same implicit races as above. | |
391 | */ | |
ae1251ab | 392 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
1da177e4 | 393 | { |
a29c33f4 | 394 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
395 | const struct cred *old; |
396 | struct cred *new; | |
1da177e4 | 397 | int retval; |
a29c33f4 EB |
398 | kgid_t kgid; |
399 | ||
400 | kgid = make_kgid(ns, gid); | |
401 | if (!gid_valid(kgid)) | |
402 | return -EINVAL; | |
1da177e4 | 403 | |
d84f4f99 DH |
404 | new = prepare_creds(); |
405 | if (!new) | |
406 | return -ENOMEM; | |
407 | old = current_cred(); | |
408 | ||
d84f4f99 | 409 | retval = -EPERM; |
c7b96acf | 410 | if (ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 EB |
411 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
412 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
413 | new->egid = new->fsgid = kgid; | |
1da177e4 | 414 | else |
d84f4f99 | 415 | goto error; |
1da177e4 | 416 | |
d84f4f99 DH |
417 | return commit_creds(new); |
418 | ||
419 | error: | |
420 | abort_creds(new); | |
421 | return retval; | |
1da177e4 | 422 | } |
54e99124 | 423 | |
d84f4f99 DH |
424 | /* |
425 | * change the user struct in a credentials set to match the new UID | |
426 | */ | |
427 | static int set_user(struct cred *new) | |
1da177e4 LT |
428 | { |
429 | struct user_struct *new_user; | |
430 | ||
078de5f7 | 431 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
432 | if (!new_user) |
433 | return -EAGAIN; | |
434 | ||
72fa5997 VK |
435 | /* |
436 | * We don't fail in case of NPROC limit excess here because too many | |
437 | * poorly written programs don't check set*uid() return code, assuming | |
438 | * it never fails if called by root. We may still enforce NPROC limit | |
439 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
440 | * failure to the execve() stage. | |
441 | */ | |
78d7d407 | 442 | if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) && |
72fa5997 VK |
443 | new_user != INIT_USER) |
444 | current->flags |= PF_NPROC_EXCEEDED; | |
445 | else | |
446 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 447 | |
d84f4f99 DH |
448 | free_uid(new->user); |
449 | new->user = new_user; | |
1da177e4 LT |
450 | return 0; |
451 | } | |
452 | ||
453 | /* | |
454 | * Unprivileged users may change the real uid to the effective uid | |
455 | * or vice versa. (BSD-style) | |
456 | * | |
457 | * If you set the real uid at all, or set the effective uid to a value not | |
458 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
459 | * | |
460 | * This makes it possible for a setuid program to completely drop its | |
461 | * privileges, which is often a useful assertion to make when you are doing | |
462 | * a security audit over a program. | |
463 | * | |
464 | * The general idea is that a program which uses just setreuid() will be | |
465 | * 100% compatible with BSD. A program which uses just setuid() will be | |
ec94fc3d | 466 | * 100% compatible with POSIX with saved IDs. |
1da177e4 | 467 | */ |
ae1251ab | 468 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
1da177e4 | 469 | { |
a29c33f4 | 470 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
471 | const struct cred *old; |
472 | struct cred *new; | |
1da177e4 | 473 | int retval; |
a29c33f4 EB |
474 | kuid_t kruid, keuid; |
475 | ||
476 | kruid = make_kuid(ns, ruid); | |
477 | keuid = make_kuid(ns, euid); | |
478 | ||
479 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
480 | return -EINVAL; | |
481 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
482 | return -EINVAL; | |
1da177e4 | 483 | |
d84f4f99 DH |
484 | new = prepare_creds(); |
485 | if (!new) | |
486 | return -ENOMEM; | |
487 | old = current_cred(); | |
488 | ||
d84f4f99 | 489 | retval = -EPERM; |
1da177e4 | 490 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
491 | new->uid = kruid; |
492 | if (!uid_eq(old->uid, kruid) && | |
493 | !uid_eq(old->euid, kruid) && | |
c7b96acf | 494 | !ns_capable(old->user_ns, CAP_SETUID)) |
d84f4f99 | 495 | goto error; |
1da177e4 LT |
496 | } |
497 | ||
498 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
499 | new->euid = keuid; |
500 | if (!uid_eq(old->uid, keuid) && | |
501 | !uid_eq(old->euid, keuid) && | |
502 | !uid_eq(old->suid, keuid) && | |
c7b96acf | 503 | !ns_capable(old->user_ns, CAP_SETUID)) |
d84f4f99 | 504 | goto error; |
1da177e4 LT |
505 | } |
506 | ||
a29c33f4 | 507 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
508 | retval = set_user(new); |
509 | if (retval < 0) | |
510 | goto error; | |
511 | } | |
1da177e4 | 512 | if (ruid != (uid_t) -1 || |
a29c33f4 | 513 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
514 | new->suid = new->euid; |
515 | new->fsuid = new->euid; | |
1da177e4 | 516 | |
d84f4f99 DH |
517 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
518 | if (retval < 0) | |
519 | goto error; | |
1da177e4 | 520 | |
d84f4f99 | 521 | return commit_creds(new); |
1da177e4 | 522 | |
d84f4f99 DH |
523 | error: |
524 | abort_creds(new); | |
525 | return retval; | |
526 | } | |
ec94fc3d | 527 | |
1da177e4 | 528 | /* |
ec94fc3d | 529 | * setuid() is implemented like SysV with SAVED_IDS |
530 | * | |
1da177e4 | 531 | * Note that SAVED_ID's is deficient in that a setuid root program |
ec94fc3d | 532 | * like sendmail, for example, cannot set its uid to be a normal |
1da177e4 LT |
533 | * user and then switch back, because if you're root, setuid() sets |
534 | * the saved uid too. If you don't like this, blame the bright people | |
535 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
536 | * will allow a root program to temporarily drop privileges and be able to | |
ec94fc3d | 537 | * regain them by swapping the real and effective uid. |
1da177e4 | 538 | */ |
ae1251ab | 539 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
1da177e4 | 540 | { |
a29c33f4 | 541 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
542 | const struct cred *old; |
543 | struct cred *new; | |
1da177e4 | 544 | int retval; |
a29c33f4 EB |
545 | kuid_t kuid; |
546 | ||
547 | kuid = make_kuid(ns, uid); | |
548 | if (!uid_valid(kuid)) | |
549 | return -EINVAL; | |
1da177e4 | 550 | |
d84f4f99 DH |
551 | new = prepare_creds(); |
552 | if (!new) | |
553 | return -ENOMEM; | |
554 | old = current_cred(); | |
555 | ||
d84f4f99 | 556 | retval = -EPERM; |
c7b96acf | 557 | if (ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
558 | new->suid = new->uid = kuid; |
559 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
560 | retval = set_user(new); |
561 | if (retval < 0) | |
562 | goto error; | |
d84f4f99 | 563 | } |
a29c33f4 | 564 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 565 | goto error; |
1da177e4 | 566 | } |
1da177e4 | 567 | |
a29c33f4 | 568 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
569 | |
570 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
571 | if (retval < 0) | |
572 | goto error; | |
1da177e4 | 573 | |
d84f4f99 | 574 | return commit_creds(new); |
1da177e4 | 575 | |
d84f4f99 DH |
576 | error: |
577 | abort_creds(new); | |
578 | return retval; | |
1da177e4 LT |
579 | } |
580 | ||
581 | ||
582 | /* | |
583 | * This function implements a generic ability to update ruid, euid, | |
584 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
585 | */ | |
ae1251ab | 586 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
1da177e4 | 587 | { |
a29c33f4 | 588 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
589 | const struct cred *old; |
590 | struct cred *new; | |
1da177e4 | 591 | int retval; |
a29c33f4 EB |
592 | kuid_t kruid, keuid, ksuid; |
593 | ||
594 | kruid = make_kuid(ns, ruid); | |
595 | keuid = make_kuid(ns, euid); | |
596 | ksuid = make_kuid(ns, suid); | |
597 | ||
598 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
599 | return -EINVAL; | |
600 | ||
601 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
602 | return -EINVAL; | |
603 | ||
604 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
605 | return -EINVAL; | |
1da177e4 | 606 | |
d84f4f99 DH |
607 | new = prepare_creds(); |
608 | if (!new) | |
609 | return -ENOMEM; | |
610 | ||
d84f4f99 | 611 | old = current_cred(); |
1da177e4 | 612 | |
d84f4f99 | 613 | retval = -EPERM; |
c7b96acf | 614 | if (!ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
615 | if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && |
616 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid)) | |
d84f4f99 | 617 | goto error; |
a29c33f4 EB |
618 | if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && |
619 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid)) | |
d84f4f99 | 620 | goto error; |
a29c33f4 EB |
621 | if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && |
622 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid)) | |
d84f4f99 | 623 | goto error; |
1da177e4 | 624 | } |
d84f4f99 | 625 | |
1da177e4 | 626 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
627 | new->uid = kruid; |
628 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
629 | retval = set_user(new); |
630 | if (retval < 0) | |
631 | goto error; | |
632 | } | |
1da177e4 | 633 | } |
d84f4f99 | 634 | if (euid != (uid_t) -1) |
a29c33f4 | 635 | new->euid = keuid; |
1da177e4 | 636 | if (suid != (uid_t) -1) |
a29c33f4 | 637 | new->suid = ksuid; |
d84f4f99 | 638 | new->fsuid = new->euid; |
1da177e4 | 639 | |
d84f4f99 DH |
640 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
641 | if (retval < 0) | |
642 | goto error; | |
1da177e4 | 643 | |
d84f4f99 | 644 | return commit_creds(new); |
1da177e4 | 645 | |
d84f4f99 DH |
646 | error: |
647 | abort_creds(new); | |
648 | return retval; | |
1da177e4 LT |
649 | } |
650 | ||
a29c33f4 | 651 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 652 | { |
86a264ab | 653 | const struct cred *cred = current_cred(); |
1da177e4 | 654 | int retval; |
a29c33f4 EB |
655 | uid_t ruid, euid, suid; |
656 | ||
657 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
658 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
659 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 660 | |
ec94fc3d | 661 | retval = put_user(ruid, ruidp); |
662 | if (!retval) { | |
663 | retval = put_user(euid, euidp); | |
664 | if (!retval) | |
665 | return put_user(suid, suidp); | |
666 | } | |
1da177e4 LT |
667 | return retval; |
668 | } | |
669 | ||
670 | /* | |
671 | * Same as above, but for rgid, egid, sgid. | |
672 | */ | |
ae1251ab | 673 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
1da177e4 | 674 | { |
a29c33f4 | 675 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
676 | const struct cred *old; |
677 | struct cred *new; | |
1da177e4 | 678 | int retval; |
a29c33f4 EB |
679 | kgid_t krgid, kegid, ksgid; |
680 | ||
681 | krgid = make_kgid(ns, rgid); | |
682 | kegid = make_kgid(ns, egid); | |
683 | ksgid = make_kgid(ns, sgid); | |
684 | ||
685 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
686 | return -EINVAL; | |
687 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
688 | return -EINVAL; | |
689 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
690 | return -EINVAL; | |
1da177e4 | 691 | |
d84f4f99 DH |
692 | new = prepare_creds(); |
693 | if (!new) | |
694 | return -ENOMEM; | |
695 | old = current_cred(); | |
696 | ||
d84f4f99 | 697 | retval = -EPERM; |
c7b96acf | 698 | if (!ns_capable(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
699 | if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && |
700 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid)) | |
d84f4f99 | 701 | goto error; |
a29c33f4 EB |
702 | if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && |
703 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid)) | |
d84f4f99 | 704 | goto error; |
a29c33f4 EB |
705 | if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && |
706 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid)) | |
d84f4f99 | 707 | goto error; |
1da177e4 | 708 | } |
d84f4f99 | 709 | |
1da177e4 | 710 | if (rgid != (gid_t) -1) |
a29c33f4 | 711 | new->gid = krgid; |
d84f4f99 | 712 | if (egid != (gid_t) -1) |
a29c33f4 | 713 | new->egid = kegid; |
1da177e4 | 714 | if (sgid != (gid_t) -1) |
a29c33f4 | 715 | new->sgid = ksgid; |
d84f4f99 | 716 | new->fsgid = new->egid; |
1da177e4 | 717 | |
d84f4f99 DH |
718 | return commit_creds(new); |
719 | ||
720 | error: | |
721 | abort_creds(new); | |
722 | return retval; | |
1da177e4 LT |
723 | } |
724 | ||
a29c33f4 | 725 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 726 | { |
86a264ab | 727 | const struct cred *cred = current_cred(); |
1da177e4 | 728 | int retval; |
a29c33f4 EB |
729 | gid_t rgid, egid, sgid; |
730 | ||
731 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
732 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
733 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 734 | |
ec94fc3d | 735 | retval = put_user(rgid, rgidp); |
736 | if (!retval) { | |
737 | retval = put_user(egid, egidp); | |
738 | if (!retval) | |
739 | retval = put_user(sgid, sgidp); | |
740 | } | |
1da177e4 LT |
741 | |
742 | return retval; | |
743 | } | |
744 | ||
745 | ||
746 | /* | |
747 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
748 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
749 | * whatever uid it wants to). It normally shadows "euid", except when | |
750 | * explicitly set by setfsuid() or for access.. | |
751 | */ | |
ae1251ab | 752 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
1da177e4 | 753 | { |
d84f4f99 DH |
754 | const struct cred *old; |
755 | struct cred *new; | |
756 | uid_t old_fsuid; | |
a29c33f4 EB |
757 | kuid_t kuid; |
758 | ||
759 | old = current_cred(); | |
760 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
761 | ||
762 | kuid = make_kuid(old->user_ns, uid); | |
763 | if (!uid_valid(kuid)) | |
764 | return old_fsuid; | |
1da177e4 | 765 | |
d84f4f99 DH |
766 | new = prepare_creds(); |
767 | if (!new) | |
a29c33f4 | 768 | return old_fsuid; |
1da177e4 | 769 | |
a29c33f4 EB |
770 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
771 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
c7b96acf | 772 | ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
773 | if (!uid_eq(kuid, old->fsuid)) { |
774 | new->fsuid = kuid; | |
d84f4f99 DH |
775 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
776 | goto change_okay; | |
1da177e4 | 777 | } |
1da177e4 LT |
778 | } |
779 | ||
d84f4f99 DH |
780 | abort_creds(new); |
781 | return old_fsuid; | |
1da177e4 | 782 | |
d84f4f99 DH |
783 | change_okay: |
784 | commit_creds(new); | |
1da177e4 LT |
785 | return old_fsuid; |
786 | } | |
787 | ||
788 | /* | |
f42df9e6 | 789 | * Samma på svenska.. |
1da177e4 | 790 | */ |
ae1251ab | 791 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) |
1da177e4 | 792 | { |
d84f4f99 DH |
793 | const struct cred *old; |
794 | struct cred *new; | |
795 | gid_t old_fsgid; | |
a29c33f4 EB |
796 | kgid_t kgid; |
797 | ||
798 | old = current_cred(); | |
799 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
800 | ||
801 | kgid = make_kgid(old->user_ns, gid); | |
802 | if (!gid_valid(kgid)) | |
803 | return old_fsgid; | |
d84f4f99 DH |
804 | |
805 | new = prepare_creds(); | |
806 | if (!new) | |
a29c33f4 | 807 | return old_fsgid; |
1da177e4 | 808 | |
a29c33f4 EB |
809 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
810 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
c7b96acf | 811 | ns_capable(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
812 | if (!gid_eq(kgid, old->fsgid)) { |
813 | new->fsgid = kgid; | |
d84f4f99 | 814 | goto change_okay; |
1da177e4 | 815 | } |
1da177e4 | 816 | } |
d84f4f99 | 817 | |
d84f4f99 DH |
818 | abort_creds(new); |
819 | return old_fsgid; | |
820 | ||
821 | change_okay: | |
822 | commit_creds(new); | |
1da177e4 LT |
823 | return old_fsgid; |
824 | } | |
2813893f | 825 | #endif /* CONFIG_MULTIUSER */ |
1da177e4 | 826 | |
4a22f166 SR |
827 | /** |
828 | * sys_getpid - return the thread group id of the current process | |
829 | * | |
830 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
831 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
832 | * which case the tgid is the same in all threads of the same group. | |
833 | * | |
834 | * This is SMP safe as current->tgid does not change. | |
835 | */ | |
836 | SYSCALL_DEFINE0(getpid) | |
837 | { | |
838 | return task_tgid_vnr(current); | |
839 | } | |
840 | ||
841 | /* Thread ID - the internal kernel "pid" */ | |
842 | SYSCALL_DEFINE0(gettid) | |
843 | { | |
844 | return task_pid_vnr(current); | |
845 | } | |
846 | ||
847 | /* | |
848 | * Accessing ->real_parent is not SMP-safe, it could | |
849 | * change from under us. However, we can use a stale | |
850 | * value of ->real_parent under rcu_read_lock(), see | |
851 | * release_task()->call_rcu(delayed_put_task_struct). | |
852 | */ | |
853 | SYSCALL_DEFINE0(getppid) | |
854 | { | |
855 | int pid; | |
856 | ||
857 | rcu_read_lock(); | |
858 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
859 | rcu_read_unlock(); | |
860 | ||
861 | return pid; | |
862 | } | |
863 | ||
864 | SYSCALL_DEFINE0(getuid) | |
865 | { | |
866 | /* Only we change this so SMP safe */ | |
867 | return from_kuid_munged(current_user_ns(), current_uid()); | |
868 | } | |
869 | ||
870 | SYSCALL_DEFINE0(geteuid) | |
871 | { | |
872 | /* Only we change this so SMP safe */ | |
873 | return from_kuid_munged(current_user_ns(), current_euid()); | |
874 | } | |
875 | ||
876 | SYSCALL_DEFINE0(getgid) | |
877 | { | |
878 | /* Only we change this so SMP safe */ | |
879 | return from_kgid_munged(current_user_ns(), current_gid()); | |
880 | } | |
881 | ||
882 | SYSCALL_DEFINE0(getegid) | |
883 | { | |
884 | /* Only we change this so SMP safe */ | |
885 | return from_kgid_munged(current_user_ns(), current_egid()); | |
886 | } | |
887 | ||
f06febc9 FM |
888 | void do_sys_times(struct tms *tms) |
889 | { | |
5613fda9 | 890 | u64 tgutime, tgstime, cutime, cstime; |
f06febc9 | 891 | |
e80d0a1a | 892 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
893 | cutime = current->signal->cutime; |
894 | cstime = current->signal->cstime; | |
5613fda9 FW |
895 | tms->tms_utime = nsec_to_clock_t(tgutime); |
896 | tms->tms_stime = nsec_to_clock_t(tgstime); | |
897 | tms->tms_cutime = nsec_to_clock_t(cutime); | |
898 | tms->tms_cstime = nsec_to_clock_t(cstime); | |
f06febc9 FM |
899 | } |
900 | ||
58fd3aa2 | 901 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 902 | { |
1da177e4 LT |
903 | if (tbuf) { |
904 | struct tms tmp; | |
f06febc9 FM |
905 | |
906 | do_sys_times(&tmp); | |
1da177e4 LT |
907 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
908 | return -EFAULT; | |
909 | } | |
e3d5a27d | 910 | force_successful_syscall_return(); |
1da177e4 LT |
911 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
912 | } | |
913 | ||
914 | /* | |
915 | * This needs some heavy checking ... | |
916 | * I just haven't the stomach for it. I also don't fully | |
917 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
918 | * | |
919 | * OK, I think I have the protection semantics right.... this is really | |
920 | * only important on a multi-user system anyway, to make sure one user | |
921 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
922 | * | |
98611e4e | 923 | * !PF_FORKNOEXEC check to conform completely to POSIX. |
1da177e4 | 924 | */ |
b290ebe2 | 925 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
926 | { |
927 | struct task_struct *p; | |
ee0acf90 | 928 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
929 | struct pid *pgrp; |
930 | int err; | |
1da177e4 LT |
931 | |
932 | if (!pid) | |
b488893a | 933 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
934 | if (!pgid) |
935 | pgid = pid; | |
936 | if (pgid < 0) | |
937 | return -EINVAL; | |
950eaaca | 938 | rcu_read_lock(); |
1da177e4 LT |
939 | |
940 | /* From this point forward we keep holding onto the tasklist lock | |
941 | * so that our parent does not change from under us. -DaveM | |
942 | */ | |
943 | write_lock_irq(&tasklist_lock); | |
944 | ||
945 | err = -ESRCH; | |
4e021306 | 946 | p = find_task_by_vpid(pid); |
1da177e4 LT |
947 | if (!p) |
948 | goto out; | |
949 | ||
950 | err = -EINVAL; | |
951 | if (!thread_group_leader(p)) | |
952 | goto out; | |
953 | ||
4e021306 | 954 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 955 | err = -EPERM; |
41487c65 | 956 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
957 | goto out; |
958 | err = -EACCES; | |
98611e4e | 959 | if (!(p->flags & PF_FORKNOEXEC)) |
1da177e4 LT |
960 | goto out; |
961 | } else { | |
962 | err = -ESRCH; | |
ee0acf90 | 963 | if (p != group_leader) |
1da177e4 LT |
964 | goto out; |
965 | } | |
966 | ||
967 | err = -EPERM; | |
968 | if (p->signal->leader) | |
969 | goto out; | |
970 | ||
4e021306 | 971 | pgrp = task_pid(p); |
1da177e4 | 972 | if (pgid != pid) { |
b488893a | 973 | struct task_struct *g; |
1da177e4 | 974 | |
4e021306 ON |
975 | pgrp = find_vpid(pgid); |
976 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 977 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 978 | goto out; |
1da177e4 LT |
979 | } |
980 | ||
1da177e4 LT |
981 | err = security_task_setpgid(p, pgid); |
982 | if (err) | |
983 | goto out; | |
984 | ||
1b0f7ffd | 985 | if (task_pgrp(p) != pgrp) |
83beaf3c | 986 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
987 | |
988 | err = 0; | |
989 | out: | |
990 | /* All paths lead to here, thus we are safe. -DaveM */ | |
991 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 992 | rcu_read_unlock(); |
1da177e4 LT |
993 | return err; |
994 | } | |
995 | ||
dbf040d9 | 996 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1da177e4 | 997 | { |
12a3de0a ON |
998 | struct task_struct *p; |
999 | struct pid *grp; | |
1000 | int retval; | |
1001 | ||
1002 | rcu_read_lock(); | |
756184b7 | 1003 | if (!pid) |
12a3de0a | 1004 | grp = task_pgrp(current); |
756184b7 | 1005 | else { |
1da177e4 | 1006 | retval = -ESRCH; |
12a3de0a ON |
1007 | p = find_task_by_vpid(pid); |
1008 | if (!p) | |
1009 | goto out; | |
1010 | grp = task_pgrp(p); | |
1011 | if (!grp) | |
1012 | goto out; | |
1013 | ||
1014 | retval = security_task_getpgid(p); | |
1015 | if (retval) | |
1016 | goto out; | |
1da177e4 | 1017 | } |
12a3de0a ON |
1018 | retval = pid_vnr(grp); |
1019 | out: | |
1020 | rcu_read_unlock(); | |
1021 | return retval; | |
1da177e4 LT |
1022 | } |
1023 | ||
1024 | #ifdef __ARCH_WANT_SYS_GETPGRP | |
1025 | ||
dbf040d9 | 1026 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1027 | { |
12a3de0a | 1028 | return sys_getpgid(0); |
1da177e4 LT |
1029 | } |
1030 | ||
1031 | #endif | |
1032 | ||
dbf040d9 | 1033 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1034 | { |
1dd768c0 ON |
1035 | struct task_struct *p; |
1036 | struct pid *sid; | |
1037 | int retval; | |
1038 | ||
1039 | rcu_read_lock(); | |
756184b7 | 1040 | if (!pid) |
1dd768c0 | 1041 | sid = task_session(current); |
756184b7 | 1042 | else { |
1da177e4 | 1043 | retval = -ESRCH; |
1dd768c0 ON |
1044 | p = find_task_by_vpid(pid); |
1045 | if (!p) | |
1046 | goto out; | |
1047 | sid = task_session(p); | |
1048 | if (!sid) | |
1049 | goto out; | |
1050 | ||
1051 | retval = security_task_getsid(p); | |
1052 | if (retval) | |
1053 | goto out; | |
1da177e4 | 1054 | } |
1dd768c0 ON |
1055 | retval = pid_vnr(sid); |
1056 | out: | |
1057 | rcu_read_unlock(); | |
1058 | return retval; | |
1da177e4 LT |
1059 | } |
1060 | ||
81dabb46 ON |
1061 | static void set_special_pids(struct pid *pid) |
1062 | { | |
1063 | struct task_struct *curr = current->group_leader; | |
1064 | ||
1065 | if (task_session(curr) != pid) | |
1066 | change_pid(curr, PIDTYPE_SID, pid); | |
1067 | ||
1068 | if (task_pgrp(curr) != pid) | |
1069 | change_pid(curr, PIDTYPE_PGID, pid); | |
1070 | } | |
1071 | ||
b290ebe2 | 1072 | SYSCALL_DEFINE0(setsid) |
1da177e4 | 1073 | { |
e19f247a | 1074 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1075 | struct pid *sid = task_pid(group_leader); |
1076 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1077 | int err = -EPERM; |
1078 | ||
1da177e4 | 1079 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1080 | /* Fail if I am already a session leader */ |
1081 | if (group_leader->signal->leader) | |
1082 | goto out; | |
1083 | ||
430c6231 ON |
1084 | /* Fail if a process group id already exists that equals the |
1085 | * proposed session id. | |
390e2ff0 | 1086 | */ |
6806aac6 | 1087 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1088 | goto out; |
1089 | ||
e19f247a | 1090 | group_leader->signal->leader = 1; |
81dabb46 | 1091 | set_special_pids(sid); |
24ec839c | 1092 | |
9c9f4ded | 1093 | proc_clear_tty(group_leader); |
24ec839c | 1094 | |
e4cc0a9c | 1095 | err = session; |
1da177e4 LT |
1096 | out: |
1097 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1098 | if (err > 0) { |
0d0df599 | 1099 | proc_sid_connector(group_leader); |
5091faa4 MG |
1100 | sched_autogroup_create_attach(group_leader); |
1101 | } | |
1da177e4 LT |
1102 | return err; |
1103 | } | |
1104 | ||
1da177e4 LT |
1105 | DECLARE_RWSEM(uts_sem); |
1106 | ||
e28cbf22 CH |
1107 | #ifdef COMPAT_UTS_MACHINE |
1108 | #define override_architecture(name) \ | |
46da2766 | 1109 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1110 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1111 | sizeof(COMPAT_UTS_MACHINE))) | |
1112 | #else | |
1113 | #define override_architecture(name) 0 | |
1114 | #endif | |
1115 | ||
be27425d AK |
1116 | /* |
1117 | * Work around broken programs that cannot handle "Linux 3.0". | |
1118 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
39afb5ee | 1119 | * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60. |
be27425d | 1120 | */ |
2702b152 | 1121 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1122 | { |
1123 | int ret = 0; | |
be27425d AK |
1124 | |
1125 | if (current->personality & UNAME26) { | |
2702b152 KC |
1126 | const char *rest = UTS_RELEASE; |
1127 | char buf[65] = { 0 }; | |
be27425d AK |
1128 | int ndots = 0; |
1129 | unsigned v; | |
2702b152 | 1130 | size_t copy; |
be27425d AK |
1131 | |
1132 | while (*rest) { | |
1133 | if (*rest == '.' && ++ndots >= 3) | |
1134 | break; | |
1135 | if (!isdigit(*rest) && *rest != '.') | |
1136 | break; | |
1137 | rest++; | |
1138 | } | |
39afb5ee | 1139 | v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60; |
31fd84b9 | 1140 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1141 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1142 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1143 | } |
1144 | return ret; | |
1145 | } | |
1146 | ||
e48fbb69 | 1147 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 LT |
1148 | { |
1149 | int errno = 0; | |
1150 | ||
1151 | down_read(&uts_sem); | |
e9ff3990 | 1152 | if (copy_to_user(name, utsname(), sizeof *name)) |
1da177e4 LT |
1153 | errno = -EFAULT; |
1154 | up_read(&uts_sem); | |
e28cbf22 | 1155 | |
be27425d AK |
1156 | if (!errno && override_release(name->release, sizeof(name->release))) |
1157 | errno = -EFAULT; | |
e28cbf22 CH |
1158 | if (!errno && override_architecture(name)) |
1159 | errno = -EFAULT; | |
1da177e4 LT |
1160 | return errno; |
1161 | } | |
1162 | ||
5cacdb4a CH |
1163 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1164 | /* | |
1165 | * Old cruft | |
1166 | */ | |
1167 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1168 | { | |
1169 | int error = 0; | |
1170 | ||
1171 | if (!name) | |
1172 | return -EFAULT; | |
1173 | ||
1174 | down_read(&uts_sem); | |
1175 | if (copy_to_user(name, utsname(), sizeof(*name))) | |
1176 | error = -EFAULT; | |
1177 | up_read(&uts_sem); | |
1178 | ||
be27425d AK |
1179 | if (!error && override_release(name->release, sizeof(name->release))) |
1180 | error = -EFAULT; | |
5cacdb4a CH |
1181 | if (!error && override_architecture(name)) |
1182 | error = -EFAULT; | |
1183 | return error; | |
1184 | } | |
1185 | ||
1186 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1187 | { | |
1188 | int error; | |
1189 | ||
1190 | if (!name) | |
1191 | return -EFAULT; | |
1192 | if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname))) | |
1193 | return -EFAULT; | |
1194 | ||
1195 | down_read(&uts_sem); | |
1196 | error = __copy_to_user(&name->sysname, &utsname()->sysname, | |
1197 | __OLD_UTS_LEN); | |
1198 | error |= __put_user(0, name->sysname + __OLD_UTS_LEN); | |
1199 | error |= __copy_to_user(&name->nodename, &utsname()->nodename, | |
1200 | __OLD_UTS_LEN); | |
1201 | error |= __put_user(0, name->nodename + __OLD_UTS_LEN); | |
1202 | error |= __copy_to_user(&name->release, &utsname()->release, | |
1203 | __OLD_UTS_LEN); | |
1204 | error |= __put_user(0, name->release + __OLD_UTS_LEN); | |
1205 | error |= __copy_to_user(&name->version, &utsname()->version, | |
1206 | __OLD_UTS_LEN); | |
1207 | error |= __put_user(0, name->version + __OLD_UTS_LEN); | |
1208 | error |= __copy_to_user(&name->machine, &utsname()->machine, | |
1209 | __OLD_UTS_LEN); | |
1210 | error |= __put_user(0, name->machine + __OLD_UTS_LEN); | |
1211 | up_read(&uts_sem); | |
1212 | ||
1213 | if (!error && override_architecture(name)) | |
1214 | error = -EFAULT; | |
be27425d AK |
1215 | if (!error && override_release(name->release, sizeof(name->release))) |
1216 | error = -EFAULT; | |
5cacdb4a CH |
1217 | return error ? -EFAULT : 0; |
1218 | } | |
1219 | #endif | |
1220 | ||
5a8a82b1 | 1221 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1222 | { |
1223 | int errno; | |
1224 | char tmp[__NEW_UTS_LEN]; | |
1225 | ||
bb96a6f5 | 1226 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1227 | return -EPERM; |
fc832ad3 | 1228 | |
1da177e4 LT |
1229 | if (len < 0 || len > __NEW_UTS_LEN) |
1230 | return -EINVAL; | |
1231 | down_write(&uts_sem); | |
1232 | errno = -EFAULT; | |
1233 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1234 | struct new_utsname *u = utsname(); |
1235 | ||
1236 | memcpy(u->nodename, tmp, len); | |
1237 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1238 | errno = 0; |
499eea6b | 1239 | uts_proc_notify(UTS_PROC_HOSTNAME); |
1da177e4 LT |
1240 | } |
1241 | up_write(&uts_sem); | |
1242 | return errno; | |
1243 | } | |
1244 | ||
1245 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1246 | ||
5a8a82b1 | 1247 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 LT |
1248 | { |
1249 | int i, errno; | |
9679e4dd | 1250 | struct new_utsname *u; |
1da177e4 LT |
1251 | |
1252 | if (len < 0) | |
1253 | return -EINVAL; | |
1254 | down_read(&uts_sem); | |
9679e4dd AM |
1255 | u = utsname(); |
1256 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1257 | if (i > len) |
1258 | i = len; | |
1259 | errno = 0; | |
9679e4dd | 1260 | if (copy_to_user(name, u->nodename, i)) |
1da177e4 LT |
1261 | errno = -EFAULT; |
1262 | up_read(&uts_sem); | |
1263 | return errno; | |
1264 | } | |
1265 | ||
1266 | #endif | |
1267 | ||
1268 | /* | |
1269 | * Only setdomainname; getdomainname can be implemented by calling | |
1270 | * uname() | |
1271 | */ | |
5a8a82b1 | 1272 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1273 | { |
1274 | int errno; | |
1275 | char tmp[__NEW_UTS_LEN]; | |
1276 | ||
fc832ad3 | 1277 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1278 | return -EPERM; |
1279 | if (len < 0 || len > __NEW_UTS_LEN) | |
1280 | return -EINVAL; | |
1281 | ||
1282 | down_write(&uts_sem); | |
1283 | errno = -EFAULT; | |
1284 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1285 | struct new_utsname *u = utsname(); |
1286 | ||
1287 | memcpy(u->domainname, tmp, len); | |
1288 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1289 | errno = 0; |
499eea6b | 1290 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
1da177e4 LT |
1291 | } |
1292 | up_write(&uts_sem); | |
1293 | return errno; | |
1294 | } | |
1295 | ||
e48fbb69 | 1296 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1297 | { |
b9518345 JS |
1298 | struct rlimit value; |
1299 | int ret; | |
1300 | ||
1301 | ret = do_prlimit(current, resource, NULL, &value); | |
1302 | if (!ret) | |
1303 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1304 | ||
1305 | return ret; | |
1da177e4 LT |
1306 | } |
1307 | ||
1308 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT | |
1309 | ||
1310 | /* | |
1311 | * Back compatibility for getrlimit. Needed for some apps. | |
1312 | */ | |
e48fbb69 HC |
1313 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1314 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1315 | { |
1316 | struct rlimit x; | |
1317 | if (resource >= RLIM_NLIMITS) | |
1318 | return -EINVAL; | |
1319 | ||
1320 | task_lock(current->group_leader); | |
1321 | x = current->signal->rlim[resource]; | |
1322 | task_unlock(current->group_leader); | |
756184b7 | 1323 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1324 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1325 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 | 1326 | x.rlim_max = 0x7FFFFFFF; |
ec94fc3d | 1327 | return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0; |
1da177e4 LT |
1328 | } |
1329 | ||
1330 | #endif | |
1331 | ||
c022a0ac JS |
1332 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1333 | { | |
1334 | #if BITS_PER_LONG < 64 | |
1335 | return rlim64 >= ULONG_MAX; | |
1336 | #else | |
1337 | return rlim64 == RLIM64_INFINITY; | |
1338 | #endif | |
1339 | } | |
1340 | ||
1341 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1342 | { | |
1343 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1344 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1345 | else | |
1346 | rlim64->rlim_cur = rlim->rlim_cur; | |
1347 | if (rlim->rlim_max == RLIM_INFINITY) | |
1348 | rlim64->rlim_max = RLIM64_INFINITY; | |
1349 | else | |
1350 | rlim64->rlim_max = rlim->rlim_max; | |
1351 | } | |
1352 | ||
1353 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1354 | { | |
1355 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1356 | rlim->rlim_cur = RLIM_INFINITY; | |
1357 | else | |
1358 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1359 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1360 | rlim->rlim_max = RLIM_INFINITY; | |
1361 | else | |
1362 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1363 | } | |
1364 | ||
1c1e618d | 1365 | /* make sure you are allowed to change @tsk limits before calling this */ |
5b41535a JS |
1366 | int do_prlimit(struct task_struct *tsk, unsigned int resource, |
1367 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1da177e4 | 1368 | { |
5b41535a | 1369 | struct rlimit *rlim; |
86f162f4 | 1370 | int retval = 0; |
1da177e4 LT |
1371 | |
1372 | if (resource >= RLIM_NLIMITS) | |
1373 | return -EINVAL; | |
5b41535a JS |
1374 | if (new_rlim) { |
1375 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1376 | return -EINVAL; | |
1377 | if (resource == RLIMIT_NOFILE && | |
1378 | new_rlim->rlim_max > sysctl_nr_open) | |
1379 | return -EPERM; | |
1380 | } | |
1da177e4 | 1381 | |
1c1e618d JS |
1382 | /* protect tsk->signal and tsk->sighand from disappearing */ |
1383 | read_lock(&tasklist_lock); | |
1384 | if (!tsk->sighand) { | |
1385 | retval = -ESRCH; | |
1386 | goto out; | |
1387 | } | |
1388 | ||
5b41535a | 1389 | rlim = tsk->signal->rlim + resource; |
86f162f4 | 1390 | task_lock(tsk->group_leader); |
5b41535a | 1391 | if (new_rlim) { |
fc832ad3 SH |
1392 | /* Keep the capable check against init_user_ns until |
1393 | cgroups can contain all limits */ | |
5b41535a JS |
1394 | if (new_rlim->rlim_max > rlim->rlim_max && |
1395 | !capable(CAP_SYS_RESOURCE)) | |
1396 | retval = -EPERM; | |
1397 | if (!retval) | |
1398 | retval = security_task_setrlimit(tsk->group_leader, | |
1399 | resource, new_rlim); | |
1400 | if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) { | |
1401 | /* | |
1402 | * The caller is asking for an immediate RLIMIT_CPU | |
1403 | * expiry. But we use the zero value to mean "it was | |
1404 | * never set". So let's cheat and make it one second | |
1405 | * instead | |
1406 | */ | |
1407 | new_rlim->rlim_cur = 1; | |
1408 | } | |
1409 | } | |
1410 | if (!retval) { | |
1411 | if (old_rlim) | |
1412 | *old_rlim = *rlim; | |
1413 | if (new_rlim) | |
1414 | *rlim = *new_rlim; | |
9926e4c7 | 1415 | } |
7855c35d | 1416 | task_unlock(tsk->group_leader); |
1da177e4 | 1417 | |
d3561f78 AM |
1418 | /* |
1419 | * RLIMIT_CPU handling. Note that the kernel fails to return an error | |
1420 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a | |
1421 | * very long-standing error, and fixing it now risks breakage of | |
1422 | * applications, so we live with it | |
1423 | */ | |
5b41535a | 1424 | if (!retval && new_rlim && resource == RLIMIT_CPU && |
baa73d9e NP |
1425 | new_rlim->rlim_cur != RLIM_INFINITY && |
1426 | IS_ENABLED(CONFIG_POSIX_TIMERS)) | |
5b41535a | 1427 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); |
ec9e16ba | 1428 | out: |
1c1e618d | 1429 | read_unlock(&tasklist_lock); |
2fb9d268 | 1430 | return retval; |
1da177e4 LT |
1431 | } |
1432 | ||
c022a0ac JS |
1433 | /* rcu lock must be held */ |
1434 | static int check_prlimit_permission(struct task_struct *task) | |
1435 | { | |
1436 | const struct cred *cred = current_cred(), *tcred; | |
1437 | ||
fc832ad3 SH |
1438 | if (current == task) |
1439 | return 0; | |
c022a0ac | 1440 | |
fc832ad3 | 1441 | tcred = __task_cred(task); |
5af66203 EB |
1442 | if (uid_eq(cred->uid, tcred->euid) && |
1443 | uid_eq(cred->uid, tcred->suid) && | |
1444 | uid_eq(cred->uid, tcred->uid) && | |
1445 | gid_eq(cred->gid, tcred->egid) && | |
1446 | gid_eq(cred->gid, tcred->sgid) && | |
1447 | gid_eq(cred->gid, tcred->gid)) | |
fc832ad3 | 1448 | return 0; |
c4a4d603 | 1449 | if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) |
fc832ad3 SH |
1450 | return 0; |
1451 | ||
1452 | return -EPERM; | |
c022a0ac JS |
1453 | } |
1454 | ||
1455 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1456 | const struct rlimit64 __user *, new_rlim, | |
1457 | struct rlimit64 __user *, old_rlim) | |
1458 | { | |
1459 | struct rlimit64 old64, new64; | |
1460 | struct rlimit old, new; | |
1461 | struct task_struct *tsk; | |
1462 | int ret; | |
1463 | ||
1464 | if (new_rlim) { | |
1465 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1466 | return -EFAULT; | |
1467 | rlim64_to_rlim(&new64, &new); | |
1468 | } | |
1469 | ||
1470 | rcu_read_lock(); | |
1471 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1472 | if (!tsk) { | |
1473 | rcu_read_unlock(); | |
1474 | return -ESRCH; | |
1475 | } | |
1476 | ret = check_prlimit_permission(tsk); | |
1477 | if (ret) { | |
1478 | rcu_read_unlock(); | |
1479 | return ret; | |
1480 | } | |
1481 | get_task_struct(tsk); | |
1482 | rcu_read_unlock(); | |
1483 | ||
1484 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1485 | old_rlim ? &old : NULL); | |
1486 | ||
1487 | if (!ret && old_rlim) { | |
1488 | rlim_to_rlim64(&old, &old64); | |
1489 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1490 | ret = -EFAULT; | |
1491 | } | |
1492 | ||
1493 | put_task_struct(tsk); | |
1494 | return ret; | |
1495 | } | |
1496 | ||
7855c35d JS |
1497 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1498 | { | |
1499 | struct rlimit new_rlim; | |
1500 | ||
1501 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1502 | return -EFAULT; | |
5b41535a | 1503 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1504 | } |
1505 | ||
1da177e4 LT |
1506 | /* |
1507 | * It would make sense to put struct rusage in the task_struct, | |
1508 | * except that would make the task_struct be *really big*. After | |
1509 | * task_struct gets moved into malloc'ed memory, it would | |
1510 | * make sense to do this. It will make moving the rest of the information | |
1511 | * a lot simpler! (Which we're not doing right now because we're not | |
1512 | * measuring them yet). | |
1513 | * | |
1da177e4 LT |
1514 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1515 | * races with threads incrementing their own counters. But since word | |
1516 | * reads are atomic, we either get new values or old values and we don't | |
1517 | * care which for the sums. We always take the siglock to protect reading | |
1518 | * the c* fields from p->signal from races with exit.c updating those | |
1519 | * fields when reaping, so a sample either gets all the additions of a | |
1520 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1521 | * |
de047c1b RT |
1522 | * Locking: |
1523 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1524 | * for the cases current multithreaded, non-current single threaded | |
1525 | * non-current multithreaded. Thread traversal is now safe with | |
1526 | * the siglock held. | |
1527 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1528 | * single threaded, as no one else can take our signal_struct away, no one | |
1529 | * else can reap the children to update signal->c* counters, and no one else | |
1530 | * can race with the signal-> fields. If we do not take any lock, the | |
1531 | * signal-> fields could be read out of order while another thread was just | |
1532 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1533 | * On the writer side, write memory barrier is implied in __exit_signal | |
1534 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1535 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1536 | * |
1da177e4 LT |
1537 | */ |
1538 | ||
f06febc9 | 1539 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1540 | { |
679c9cd4 SK |
1541 | r->ru_nvcsw += t->nvcsw; |
1542 | r->ru_nivcsw += t->nivcsw; | |
1543 | r->ru_minflt += t->min_flt; | |
1544 | r->ru_majflt += t->maj_flt; | |
1545 | r->ru_inblock += task_io_get_inblock(t); | |
1546 | r->ru_oublock += task_io_get_oublock(t); | |
1547 | } | |
1548 | ||
1da177e4 LT |
1549 | static void k_getrusage(struct task_struct *p, int who, struct rusage *r) |
1550 | { | |
1551 | struct task_struct *t; | |
1552 | unsigned long flags; | |
5613fda9 | 1553 | u64 tgutime, tgstime, utime, stime; |
1f10206c | 1554 | unsigned long maxrss = 0; |
1da177e4 | 1555 | |
ec94fc3d | 1556 | memset((char *)r, 0, sizeof (*r)); |
64861634 | 1557 | utime = stime = 0; |
1da177e4 | 1558 | |
679c9cd4 | 1559 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1560 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1561 | accumulate_thread_rusage(p, r); |
1f10206c | 1562 | maxrss = p->signal->maxrss; |
679c9cd4 SK |
1563 | goto out; |
1564 | } | |
1565 | ||
d6cf723a | 1566 | if (!lock_task_sighand(p, &flags)) |
de047c1b | 1567 | return; |
0f59cc4a | 1568 | |
1da177e4 | 1569 | switch (who) { |
ec94fc3d | 1570 | case RUSAGE_BOTH: |
1571 | case RUSAGE_CHILDREN: | |
1572 | utime = p->signal->cutime; | |
1573 | stime = p->signal->cstime; | |
1574 | r->ru_nvcsw = p->signal->cnvcsw; | |
1575 | r->ru_nivcsw = p->signal->cnivcsw; | |
1576 | r->ru_minflt = p->signal->cmin_flt; | |
1577 | r->ru_majflt = p->signal->cmaj_flt; | |
1578 | r->ru_inblock = p->signal->cinblock; | |
1579 | r->ru_oublock = p->signal->coublock; | |
1580 | maxrss = p->signal->cmaxrss; | |
1581 | ||
1582 | if (who == RUSAGE_CHILDREN) | |
1da177e4 | 1583 | break; |
0f59cc4a | 1584 | |
ec94fc3d | 1585 | case RUSAGE_SELF: |
1586 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); | |
1587 | utime += tgutime; | |
1588 | stime += tgstime; | |
1589 | r->ru_nvcsw += p->signal->nvcsw; | |
1590 | r->ru_nivcsw += p->signal->nivcsw; | |
1591 | r->ru_minflt += p->signal->min_flt; | |
1592 | r->ru_majflt += p->signal->maj_flt; | |
1593 | r->ru_inblock += p->signal->inblock; | |
1594 | r->ru_oublock += p->signal->oublock; | |
1595 | if (maxrss < p->signal->maxrss) | |
1596 | maxrss = p->signal->maxrss; | |
1597 | t = p; | |
1598 | do { | |
1599 | accumulate_thread_rusage(t, r); | |
1600 | } while_each_thread(p, t); | |
1601 | break; | |
1602 | ||
1603 | default: | |
1604 | BUG(); | |
1da177e4 | 1605 | } |
de047c1b | 1606 | unlock_task_sighand(p, &flags); |
de047c1b | 1607 | |
679c9cd4 | 1608 | out: |
5613fda9 FW |
1609 | r->ru_utime = ns_to_timeval(utime); |
1610 | r->ru_stime = ns_to_timeval(stime); | |
1f10206c JP |
1611 | |
1612 | if (who != RUSAGE_CHILDREN) { | |
1613 | struct mm_struct *mm = get_task_mm(p); | |
ec94fc3d | 1614 | |
1f10206c JP |
1615 | if (mm) { |
1616 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1617 | mmput(mm); | |
1618 | } | |
1619 | } | |
1620 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1da177e4 LT |
1621 | } |
1622 | ||
1623 | int getrusage(struct task_struct *p, int who, struct rusage __user *ru) | |
1624 | { | |
1625 | struct rusage r; | |
ec94fc3d | 1626 | |
1da177e4 | 1627 | k_getrusage(p, who, &r); |
1da177e4 LT |
1628 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; |
1629 | } | |
1630 | ||
e48fbb69 | 1631 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 | 1632 | { |
679c9cd4 SK |
1633 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1634 | who != RUSAGE_THREAD) | |
1da177e4 LT |
1635 | return -EINVAL; |
1636 | return getrusage(current, who, ru); | |
1637 | } | |
1638 | ||
8d2d5c4a AV |
1639 | #ifdef CONFIG_COMPAT |
1640 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1641 | { | |
1642 | struct rusage r; | |
1643 | ||
1644 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1645 | who != RUSAGE_THREAD) | |
1646 | return -EINVAL; | |
1647 | ||
1648 | k_getrusage(current, who, &r); | |
1649 | return put_compat_rusage(&r, ru); | |
1650 | } | |
1651 | #endif | |
1652 | ||
e48fbb69 | 1653 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
1654 | { |
1655 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1656 | return mask; | |
1657 | } | |
3b7391de | 1658 | |
6e399cd1 | 1659 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
b32dfe37 | 1660 | { |
2903ff01 | 1661 | struct fd exe; |
6e399cd1 | 1662 | struct file *old_exe, *exe_file; |
496ad9aa | 1663 | struct inode *inode; |
2903ff01 | 1664 | int err; |
b32dfe37 | 1665 | |
2903ff01 AV |
1666 | exe = fdget(fd); |
1667 | if (!exe.file) | |
b32dfe37 CG |
1668 | return -EBADF; |
1669 | ||
496ad9aa | 1670 | inode = file_inode(exe.file); |
b32dfe37 CG |
1671 | |
1672 | /* | |
1673 | * Because the original mm->exe_file points to executable file, make | |
1674 | * sure that this one is executable as well, to avoid breaking an | |
1675 | * overall picture. | |
1676 | */ | |
1677 | err = -EACCES; | |
90f8572b | 1678 | if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path)) |
b32dfe37 CG |
1679 | goto exit; |
1680 | ||
496ad9aa | 1681 | err = inode_permission(inode, MAY_EXEC); |
b32dfe37 CG |
1682 | if (err) |
1683 | goto exit; | |
1684 | ||
bafb282d | 1685 | /* |
4229fb1d | 1686 | * Forbid mm->exe_file change if old file still mapped. |
bafb282d | 1687 | */ |
6e399cd1 | 1688 | exe_file = get_mm_exe_file(mm); |
bafb282d | 1689 | err = -EBUSY; |
6e399cd1 | 1690 | if (exe_file) { |
4229fb1d KK |
1691 | struct vm_area_struct *vma; |
1692 | ||
6e399cd1 DB |
1693 | down_read(&mm->mmap_sem); |
1694 | for (vma = mm->mmap; vma; vma = vma->vm_next) { | |
1695 | if (!vma->vm_file) | |
1696 | continue; | |
1697 | if (path_equal(&vma->vm_file->f_path, | |
1698 | &exe_file->f_path)) | |
1699 | goto exit_err; | |
1700 | } | |
1701 | ||
1702 | up_read(&mm->mmap_sem); | |
1703 | fput(exe_file); | |
bafb282d KK |
1704 | } |
1705 | ||
4229fb1d | 1706 | err = 0; |
6e399cd1 DB |
1707 | /* set the new file, lockless */ |
1708 | get_file(exe.file); | |
1709 | old_exe = xchg(&mm->exe_file, exe.file); | |
1710 | if (old_exe) | |
1711 | fput(old_exe); | |
b32dfe37 | 1712 | exit: |
2903ff01 | 1713 | fdput(exe); |
b32dfe37 | 1714 | return err; |
6e399cd1 DB |
1715 | exit_err: |
1716 | up_read(&mm->mmap_sem); | |
1717 | fput(exe_file); | |
1718 | goto exit; | |
b32dfe37 CG |
1719 | } |
1720 | ||
f606b77f CG |
1721 | /* |
1722 | * WARNING: we don't require any capability here so be very careful | |
1723 | * in what is allowed for modification from userspace. | |
1724 | */ | |
1725 | static int validate_prctl_map(struct prctl_mm_map *prctl_map) | |
1726 | { | |
1727 | unsigned long mmap_max_addr = TASK_SIZE; | |
1728 | struct mm_struct *mm = current->mm; | |
1729 | int error = -EINVAL, i; | |
1730 | ||
1731 | static const unsigned char offsets[] = { | |
1732 | offsetof(struct prctl_mm_map, start_code), | |
1733 | offsetof(struct prctl_mm_map, end_code), | |
1734 | offsetof(struct prctl_mm_map, start_data), | |
1735 | offsetof(struct prctl_mm_map, end_data), | |
1736 | offsetof(struct prctl_mm_map, start_brk), | |
1737 | offsetof(struct prctl_mm_map, brk), | |
1738 | offsetof(struct prctl_mm_map, start_stack), | |
1739 | offsetof(struct prctl_mm_map, arg_start), | |
1740 | offsetof(struct prctl_mm_map, arg_end), | |
1741 | offsetof(struct prctl_mm_map, env_start), | |
1742 | offsetof(struct prctl_mm_map, env_end), | |
1743 | }; | |
1744 | ||
1745 | /* | |
1746 | * Make sure the members are not somewhere outside | |
1747 | * of allowed address space. | |
1748 | */ | |
1749 | for (i = 0; i < ARRAY_SIZE(offsets); i++) { | |
1750 | u64 val = *(u64 *)((char *)prctl_map + offsets[i]); | |
1751 | ||
1752 | if ((unsigned long)val >= mmap_max_addr || | |
1753 | (unsigned long)val < mmap_min_addr) | |
1754 | goto out; | |
1755 | } | |
1756 | ||
1757 | /* | |
1758 | * Make sure the pairs are ordered. | |
1759 | */ | |
1760 | #define __prctl_check_order(__m1, __op, __m2) \ | |
1761 | ((unsigned long)prctl_map->__m1 __op \ | |
1762 | (unsigned long)prctl_map->__m2) ? 0 : -EINVAL | |
1763 | error = __prctl_check_order(start_code, <, end_code); | |
1764 | error |= __prctl_check_order(start_data, <, end_data); | |
1765 | error |= __prctl_check_order(start_brk, <=, brk); | |
1766 | error |= __prctl_check_order(arg_start, <=, arg_end); | |
1767 | error |= __prctl_check_order(env_start, <=, env_end); | |
1768 | if (error) | |
1769 | goto out; | |
1770 | #undef __prctl_check_order | |
1771 | ||
1772 | error = -EINVAL; | |
1773 | ||
1774 | /* | |
1775 | * @brk should be after @end_data in traditional maps. | |
1776 | */ | |
1777 | if (prctl_map->start_brk <= prctl_map->end_data || | |
1778 | prctl_map->brk <= prctl_map->end_data) | |
1779 | goto out; | |
1780 | ||
1781 | /* | |
1782 | * Neither we should allow to override limits if they set. | |
1783 | */ | |
1784 | if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk, | |
1785 | prctl_map->start_brk, prctl_map->end_data, | |
1786 | prctl_map->start_data)) | |
1787 | goto out; | |
1788 | ||
1789 | /* | |
1790 | * Someone is trying to cheat the auxv vector. | |
1791 | */ | |
1792 | if (prctl_map->auxv_size) { | |
1793 | if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv)) | |
1794 | goto out; | |
1795 | } | |
1796 | ||
1797 | /* | |
1798 | * Finally, make sure the caller has the rights to | |
1799 | * change /proc/pid/exe link: only local root should | |
1800 | * be allowed to. | |
1801 | */ | |
1802 | if (prctl_map->exe_fd != (u32)-1) { | |
1803 | struct user_namespace *ns = current_user_ns(); | |
1804 | const struct cred *cred = current_cred(); | |
1805 | ||
1806 | if (!uid_eq(cred->uid, make_kuid(ns, 0)) || | |
1807 | !gid_eq(cred->gid, make_kgid(ns, 0))) | |
1808 | goto out; | |
1809 | } | |
1810 | ||
1811 | error = 0; | |
1812 | out: | |
1813 | return error; | |
1814 | } | |
1815 | ||
4a00e9df | 1816 | #ifdef CONFIG_CHECKPOINT_RESTORE |
f606b77f CG |
1817 | static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size) |
1818 | { | |
1819 | struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, }; | |
1820 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1821 | struct mm_struct *mm = current->mm; | |
1822 | int error; | |
1823 | ||
1824 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1825 | BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256); | |
1826 | ||
1827 | if (opt == PR_SET_MM_MAP_SIZE) | |
1828 | return put_user((unsigned int)sizeof(prctl_map), | |
1829 | (unsigned int __user *)addr); | |
1830 | ||
1831 | if (data_size != sizeof(prctl_map)) | |
1832 | return -EINVAL; | |
1833 | ||
1834 | if (copy_from_user(&prctl_map, addr, sizeof(prctl_map))) | |
1835 | return -EFAULT; | |
1836 | ||
1837 | error = validate_prctl_map(&prctl_map); | |
1838 | if (error) | |
1839 | return error; | |
1840 | ||
1841 | if (prctl_map.auxv_size) { | |
1842 | memset(user_auxv, 0, sizeof(user_auxv)); | |
1843 | if (copy_from_user(user_auxv, | |
1844 | (const void __user *)prctl_map.auxv, | |
1845 | prctl_map.auxv_size)) | |
1846 | return -EFAULT; | |
1847 | ||
1848 | /* Last entry must be AT_NULL as specification requires */ | |
1849 | user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL; | |
1850 | user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL; | |
1851 | } | |
1852 | ||
ddf1d398 | 1853 | if (prctl_map.exe_fd != (u32)-1) { |
6e399cd1 | 1854 | error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd); |
ddf1d398 MG |
1855 | if (error) |
1856 | return error; | |
1857 | } | |
1858 | ||
1859 | down_write(&mm->mmap_sem); | |
f606b77f CG |
1860 | |
1861 | /* | |
1862 | * We don't validate if these members are pointing to | |
1863 | * real present VMAs because application may have correspond | |
1864 | * VMAs already unmapped and kernel uses these members for statistics | |
1865 | * output in procfs mostly, except | |
1866 | * | |
1867 | * - @start_brk/@brk which are used in do_brk but kernel lookups | |
1868 | * for VMAs when updating these memvers so anything wrong written | |
1869 | * here cause kernel to swear at userspace program but won't lead | |
1870 | * to any problem in kernel itself | |
1871 | */ | |
1872 | ||
1873 | mm->start_code = prctl_map.start_code; | |
1874 | mm->end_code = prctl_map.end_code; | |
1875 | mm->start_data = prctl_map.start_data; | |
1876 | mm->end_data = prctl_map.end_data; | |
1877 | mm->start_brk = prctl_map.start_brk; | |
1878 | mm->brk = prctl_map.brk; | |
1879 | mm->start_stack = prctl_map.start_stack; | |
1880 | mm->arg_start = prctl_map.arg_start; | |
1881 | mm->arg_end = prctl_map.arg_end; | |
1882 | mm->env_start = prctl_map.env_start; | |
1883 | mm->env_end = prctl_map.env_end; | |
1884 | ||
1885 | /* | |
1886 | * Note this update of @saved_auxv is lockless thus | |
1887 | * if someone reads this member in procfs while we're | |
1888 | * updating -- it may get partly updated results. It's | |
1889 | * known and acceptable trade off: we leave it as is to | |
1890 | * not introduce additional locks here making the kernel | |
1891 | * more complex. | |
1892 | */ | |
1893 | if (prctl_map.auxv_size) | |
1894 | memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv)); | |
1895 | ||
ddf1d398 MG |
1896 | up_write(&mm->mmap_sem); |
1897 | return 0; | |
f606b77f CG |
1898 | } |
1899 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
1900 | ||
4a00e9df AD |
1901 | static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr, |
1902 | unsigned long len) | |
1903 | { | |
1904 | /* | |
1905 | * This doesn't move the auxiliary vector itself since it's pinned to | |
1906 | * mm_struct, but it permits filling the vector with new values. It's | |
1907 | * up to the caller to provide sane values here, otherwise userspace | |
1908 | * tools which use this vector might be unhappy. | |
1909 | */ | |
1910 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1911 | ||
1912 | if (len > sizeof(user_auxv)) | |
1913 | return -EINVAL; | |
1914 | ||
1915 | if (copy_from_user(user_auxv, (const void __user *)addr, len)) | |
1916 | return -EFAULT; | |
1917 | ||
1918 | /* Make sure the last entry is always AT_NULL */ | |
1919 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
1920 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
1921 | ||
1922 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1923 | ||
1924 | task_lock(current); | |
1925 | memcpy(mm->saved_auxv, user_auxv, len); | |
1926 | task_unlock(current); | |
1927 | ||
1928 | return 0; | |
1929 | } | |
1930 | ||
028ee4be CG |
1931 | static int prctl_set_mm(int opt, unsigned long addr, |
1932 | unsigned long arg4, unsigned long arg5) | |
1933 | { | |
028ee4be | 1934 | struct mm_struct *mm = current->mm; |
4a00e9df | 1935 | struct prctl_mm_map prctl_map; |
fe8c7f5c CG |
1936 | struct vm_area_struct *vma; |
1937 | int error; | |
028ee4be | 1938 | |
f606b77f CG |
1939 | if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV && |
1940 | opt != PR_SET_MM_MAP && | |
1941 | opt != PR_SET_MM_MAP_SIZE))) | |
028ee4be CG |
1942 | return -EINVAL; |
1943 | ||
f606b77f CG |
1944 | #ifdef CONFIG_CHECKPOINT_RESTORE |
1945 | if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE) | |
1946 | return prctl_set_mm_map(opt, (const void __user *)addr, arg4); | |
1947 | #endif | |
1948 | ||
79f0713d | 1949 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
1950 | return -EPERM; |
1951 | ||
6e399cd1 DB |
1952 | if (opt == PR_SET_MM_EXE_FILE) |
1953 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
b32dfe37 | 1954 | |
4a00e9df AD |
1955 | if (opt == PR_SET_MM_AUXV) |
1956 | return prctl_set_auxv(mm, addr, arg4); | |
1957 | ||
1ad75b9e | 1958 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
1959 | return -EINVAL; |
1960 | ||
fe8c7f5c CG |
1961 | error = -EINVAL; |
1962 | ||
ddf1d398 | 1963 | down_write(&mm->mmap_sem); |
028ee4be CG |
1964 | vma = find_vma(mm, addr); |
1965 | ||
4a00e9df AD |
1966 | prctl_map.start_code = mm->start_code; |
1967 | prctl_map.end_code = mm->end_code; | |
1968 | prctl_map.start_data = mm->start_data; | |
1969 | prctl_map.end_data = mm->end_data; | |
1970 | prctl_map.start_brk = mm->start_brk; | |
1971 | prctl_map.brk = mm->brk; | |
1972 | prctl_map.start_stack = mm->start_stack; | |
1973 | prctl_map.arg_start = mm->arg_start; | |
1974 | prctl_map.arg_end = mm->arg_end; | |
1975 | prctl_map.env_start = mm->env_start; | |
1976 | prctl_map.env_end = mm->env_end; | |
1977 | prctl_map.auxv = NULL; | |
1978 | prctl_map.auxv_size = 0; | |
1979 | prctl_map.exe_fd = -1; | |
1980 | ||
028ee4be CG |
1981 | switch (opt) { |
1982 | case PR_SET_MM_START_CODE: | |
4a00e9df | 1983 | prctl_map.start_code = addr; |
fe8c7f5c | 1984 | break; |
028ee4be | 1985 | case PR_SET_MM_END_CODE: |
4a00e9df | 1986 | prctl_map.end_code = addr; |
028ee4be | 1987 | break; |
028ee4be | 1988 | case PR_SET_MM_START_DATA: |
4a00e9df | 1989 | prctl_map.start_data = addr; |
028ee4be | 1990 | break; |
fe8c7f5c | 1991 | case PR_SET_MM_END_DATA: |
4a00e9df AD |
1992 | prctl_map.end_data = addr; |
1993 | break; | |
1994 | case PR_SET_MM_START_STACK: | |
1995 | prctl_map.start_stack = addr; | |
028ee4be | 1996 | break; |
028ee4be | 1997 | case PR_SET_MM_START_BRK: |
4a00e9df | 1998 | prctl_map.start_brk = addr; |
028ee4be | 1999 | break; |
028ee4be | 2000 | case PR_SET_MM_BRK: |
4a00e9df | 2001 | prctl_map.brk = addr; |
028ee4be | 2002 | break; |
4a00e9df AD |
2003 | case PR_SET_MM_ARG_START: |
2004 | prctl_map.arg_start = addr; | |
2005 | break; | |
2006 | case PR_SET_MM_ARG_END: | |
2007 | prctl_map.arg_end = addr; | |
2008 | break; | |
2009 | case PR_SET_MM_ENV_START: | |
2010 | prctl_map.env_start = addr; | |
2011 | break; | |
2012 | case PR_SET_MM_ENV_END: | |
2013 | prctl_map.env_end = addr; | |
2014 | break; | |
2015 | default: | |
2016 | goto out; | |
2017 | } | |
2018 | ||
2019 | error = validate_prctl_map(&prctl_map); | |
2020 | if (error) | |
2021 | goto out; | |
028ee4be | 2022 | |
4a00e9df | 2023 | switch (opt) { |
fe8c7f5c CG |
2024 | /* |
2025 | * If command line arguments and environment | |
2026 | * are placed somewhere else on stack, we can | |
2027 | * set them up here, ARG_START/END to setup | |
2028 | * command line argumets and ENV_START/END | |
2029 | * for environment. | |
2030 | */ | |
2031 | case PR_SET_MM_START_STACK: | |
2032 | case PR_SET_MM_ARG_START: | |
2033 | case PR_SET_MM_ARG_END: | |
2034 | case PR_SET_MM_ENV_START: | |
2035 | case PR_SET_MM_ENV_END: | |
2036 | if (!vma) { | |
2037 | error = -EFAULT; | |
2038 | goto out; | |
2039 | } | |
028ee4be CG |
2040 | } |
2041 | ||
4a00e9df AD |
2042 | mm->start_code = prctl_map.start_code; |
2043 | mm->end_code = prctl_map.end_code; | |
2044 | mm->start_data = prctl_map.start_data; | |
2045 | mm->end_data = prctl_map.end_data; | |
2046 | mm->start_brk = prctl_map.start_brk; | |
2047 | mm->brk = prctl_map.brk; | |
2048 | mm->start_stack = prctl_map.start_stack; | |
2049 | mm->arg_start = prctl_map.arg_start; | |
2050 | mm->arg_end = prctl_map.arg_end; | |
2051 | mm->env_start = prctl_map.env_start; | |
2052 | mm->env_end = prctl_map.env_end; | |
2053 | ||
028ee4be | 2054 | error = 0; |
028ee4be | 2055 | out: |
ddf1d398 | 2056 | up_write(&mm->mmap_sem); |
028ee4be CG |
2057 | return error; |
2058 | } | |
300f786b | 2059 | |
52b36941 | 2060 | #ifdef CONFIG_CHECKPOINT_RESTORE |
300f786b CG |
2061 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2062 | { | |
2063 | return put_user(me->clear_child_tid, tid_addr); | |
2064 | } | |
52b36941 | 2065 | #else |
300f786b CG |
2066 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2067 | { | |
2068 | return -EINVAL; | |
2069 | } | |
028ee4be CG |
2070 | #endif |
2071 | ||
749860ce PT |
2072 | static int propagate_has_child_subreaper(struct task_struct *p, void *data) |
2073 | { | |
2074 | /* | |
2075 | * If task has has_child_subreaper - all its decendants | |
2076 | * already have these flag too and new decendants will | |
2077 | * inherit it on fork, skip them. | |
2078 | * | |
2079 | * If we've found child_reaper - skip descendants in | |
2080 | * it's subtree as they will never get out pidns. | |
2081 | */ | |
2082 | if (p->signal->has_child_subreaper || | |
2083 | is_child_reaper(task_pid(p))) | |
2084 | return 0; | |
2085 | ||
2086 | p->signal->has_child_subreaper = 1; | |
2087 | return 1; | |
2088 | } | |
2089 | ||
c4ea37c2 HC |
2090 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2091 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2092 | { |
b6dff3ec DH |
2093 | struct task_struct *me = current; |
2094 | unsigned char comm[sizeof(me->comm)]; | |
2095 | long error; | |
1da177e4 | 2096 | |
d84f4f99 DH |
2097 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2098 | if (error != -ENOSYS) | |
1da177e4 LT |
2099 | return error; |
2100 | ||
d84f4f99 | 2101 | error = 0; |
1da177e4 | 2102 | switch (option) { |
f3cbd435 AM |
2103 | case PR_SET_PDEATHSIG: |
2104 | if (!valid_signal(arg2)) { | |
2105 | error = -EINVAL; | |
1da177e4 | 2106 | break; |
f3cbd435 AM |
2107 | } |
2108 | me->pdeath_signal = arg2; | |
2109 | break; | |
2110 | case PR_GET_PDEATHSIG: | |
2111 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2112 | break; | |
2113 | case PR_GET_DUMPABLE: | |
2114 | error = get_dumpable(me->mm); | |
2115 | break; | |
2116 | case PR_SET_DUMPABLE: | |
2117 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2118 | error = -EINVAL; | |
1da177e4 | 2119 | break; |
f3cbd435 AM |
2120 | } |
2121 | set_dumpable(me->mm, arg2); | |
2122 | break; | |
1da177e4 | 2123 | |
f3cbd435 AM |
2124 | case PR_SET_UNALIGN: |
2125 | error = SET_UNALIGN_CTL(me, arg2); | |
2126 | break; | |
2127 | case PR_GET_UNALIGN: | |
2128 | error = GET_UNALIGN_CTL(me, arg2); | |
2129 | break; | |
2130 | case PR_SET_FPEMU: | |
2131 | error = SET_FPEMU_CTL(me, arg2); | |
2132 | break; | |
2133 | case PR_GET_FPEMU: | |
2134 | error = GET_FPEMU_CTL(me, arg2); | |
2135 | break; | |
2136 | case PR_SET_FPEXC: | |
2137 | error = SET_FPEXC_CTL(me, arg2); | |
2138 | break; | |
2139 | case PR_GET_FPEXC: | |
2140 | error = GET_FPEXC_CTL(me, arg2); | |
2141 | break; | |
2142 | case PR_GET_TIMING: | |
2143 | error = PR_TIMING_STATISTICAL; | |
2144 | break; | |
2145 | case PR_SET_TIMING: | |
2146 | if (arg2 != PR_TIMING_STATISTICAL) | |
2147 | error = -EINVAL; | |
2148 | break; | |
2149 | case PR_SET_NAME: | |
2150 | comm[sizeof(me->comm) - 1] = 0; | |
2151 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2152 | sizeof(me->comm) - 1) < 0) | |
2153 | return -EFAULT; | |
2154 | set_task_comm(me, comm); | |
2155 | proc_comm_connector(me); | |
2156 | break; | |
2157 | case PR_GET_NAME: | |
2158 | get_task_comm(comm, me); | |
2159 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2160 | return -EFAULT; | |
2161 | break; | |
2162 | case PR_GET_ENDIAN: | |
2163 | error = GET_ENDIAN(me, arg2); | |
2164 | break; | |
2165 | case PR_SET_ENDIAN: | |
2166 | error = SET_ENDIAN(me, arg2); | |
2167 | break; | |
2168 | case PR_GET_SECCOMP: | |
2169 | error = prctl_get_seccomp(); | |
2170 | break; | |
2171 | case PR_SET_SECCOMP: | |
2172 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2173 | break; | |
2174 | case PR_GET_TSC: | |
2175 | error = GET_TSC_CTL(arg2); | |
2176 | break; | |
2177 | case PR_SET_TSC: | |
2178 | error = SET_TSC_CTL(arg2); | |
2179 | break; | |
2180 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2181 | error = perf_event_task_disable(); | |
2182 | break; | |
2183 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2184 | error = perf_event_task_enable(); | |
2185 | break; | |
2186 | case PR_GET_TIMERSLACK: | |
da8b44d5 JS |
2187 | if (current->timer_slack_ns > ULONG_MAX) |
2188 | error = ULONG_MAX; | |
2189 | else | |
2190 | error = current->timer_slack_ns; | |
f3cbd435 AM |
2191 | break; |
2192 | case PR_SET_TIMERSLACK: | |
2193 | if (arg2 <= 0) | |
2194 | current->timer_slack_ns = | |
6976675d | 2195 | current->default_timer_slack_ns; |
f3cbd435 AM |
2196 | else |
2197 | current->timer_slack_ns = arg2; | |
2198 | break; | |
2199 | case PR_MCE_KILL: | |
2200 | if (arg4 | arg5) | |
2201 | return -EINVAL; | |
2202 | switch (arg2) { | |
2203 | case PR_MCE_KILL_CLEAR: | |
2204 | if (arg3 != 0) | |
4db96cf0 | 2205 | return -EINVAL; |
f3cbd435 | 2206 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2207 | break; |
f3cbd435 AM |
2208 | case PR_MCE_KILL_SET: |
2209 | current->flags |= PF_MCE_PROCESS; | |
2210 | if (arg3 == PR_MCE_KILL_EARLY) | |
2211 | current->flags |= PF_MCE_EARLY; | |
2212 | else if (arg3 == PR_MCE_KILL_LATE) | |
2213 | current->flags &= ~PF_MCE_EARLY; | |
2214 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2215 | current->flags &= | |
2216 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2217 | else |
259e5e6c | 2218 | return -EINVAL; |
259e5e6c | 2219 | break; |
1da177e4 | 2220 | default: |
f3cbd435 AM |
2221 | return -EINVAL; |
2222 | } | |
2223 | break; | |
2224 | case PR_MCE_KILL_GET: | |
2225 | if (arg2 | arg3 | arg4 | arg5) | |
2226 | return -EINVAL; | |
2227 | if (current->flags & PF_MCE_PROCESS) | |
2228 | error = (current->flags & PF_MCE_EARLY) ? | |
2229 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2230 | else | |
2231 | error = PR_MCE_KILL_DEFAULT; | |
2232 | break; | |
2233 | case PR_SET_MM: | |
2234 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2235 | break; | |
2236 | case PR_GET_TID_ADDRESS: | |
2237 | error = prctl_get_tid_address(me, (int __user **)arg2); | |
2238 | break; | |
2239 | case PR_SET_CHILD_SUBREAPER: | |
2240 | me->signal->is_child_subreaper = !!arg2; | |
749860ce PT |
2241 | if (!arg2) |
2242 | break; | |
2243 | ||
2244 | walk_process_tree(me, propagate_has_child_subreaper, NULL); | |
f3cbd435 AM |
2245 | break; |
2246 | case PR_GET_CHILD_SUBREAPER: | |
2247 | error = put_user(me->signal->is_child_subreaper, | |
2248 | (int __user *)arg2); | |
2249 | break; | |
2250 | case PR_SET_NO_NEW_PRIVS: | |
2251 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2252 | return -EINVAL; | |
2253 | ||
1d4457f9 | 2254 | task_set_no_new_privs(current); |
f3cbd435 AM |
2255 | break; |
2256 | case PR_GET_NO_NEW_PRIVS: | |
2257 | if (arg2 || arg3 || arg4 || arg5) | |
2258 | return -EINVAL; | |
1d4457f9 | 2259 | return task_no_new_privs(current) ? 1 : 0; |
a0715cc2 AT |
2260 | case PR_GET_THP_DISABLE: |
2261 | if (arg2 || arg3 || arg4 || arg5) | |
2262 | return -EINVAL; | |
2263 | error = !!(me->mm->def_flags & VM_NOHUGEPAGE); | |
2264 | break; | |
2265 | case PR_SET_THP_DISABLE: | |
2266 | if (arg3 || arg4 || arg5) | |
2267 | return -EINVAL; | |
17b0573d MH |
2268 | if (down_write_killable(&me->mm->mmap_sem)) |
2269 | return -EINTR; | |
a0715cc2 AT |
2270 | if (arg2) |
2271 | me->mm->def_flags |= VM_NOHUGEPAGE; | |
2272 | else | |
2273 | me->mm->def_flags &= ~VM_NOHUGEPAGE; | |
2274 | up_write(&me->mm->mmap_sem); | |
2275 | break; | |
fe3d197f | 2276 | case PR_MPX_ENABLE_MANAGEMENT: |
e9d1b4f3 DH |
2277 | if (arg2 || arg3 || arg4 || arg5) |
2278 | return -EINVAL; | |
46a6e0cf | 2279 | error = MPX_ENABLE_MANAGEMENT(); |
fe3d197f DH |
2280 | break; |
2281 | case PR_MPX_DISABLE_MANAGEMENT: | |
e9d1b4f3 DH |
2282 | if (arg2 || arg3 || arg4 || arg5) |
2283 | return -EINVAL; | |
46a6e0cf | 2284 | error = MPX_DISABLE_MANAGEMENT(); |
fe3d197f | 2285 | break; |
9791554b PB |
2286 | case PR_SET_FP_MODE: |
2287 | error = SET_FP_MODE(me, arg2); | |
2288 | break; | |
2289 | case PR_GET_FP_MODE: | |
2290 | error = GET_FP_MODE(me); | |
2291 | break; | |
f3cbd435 AM |
2292 | default: |
2293 | error = -EINVAL; | |
2294 | break; | |
1da177e4 LT |
2295 | } |
2296 | return error; | |
2297 | } | |
3cfc348b | 2298 | |
836f92ad HC |
2299 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2300 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2301 | { |
2302 | int err = 0; | |
2303 | int cpu = raw_smp_processor_id(); | |
ec94fc3d | 2304 | |
3cfc348b AK |
2305 | if (cpup) |
2306 | err |= put_user(cpu, cpup); | |
2307 | if (nodep) | |
2308 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2309 | return err ? -EFAULT : 0; |
2310 | } | |
10a0a8d4 | 2311 | |
4a22f166 SR |
2312 | /** |
2313 | * do_sysinfo - fill in sysinfo struct | |
2314 | * @info: pointer to buffer to fill | |
2315 | */ | |
2316 | static int do_sysinfo(struct sysinfo *info) | |
2317 | { | |
2318 | unsigned long mem_total, sav_total; | |
2319 | unsigned int mem_unit, bitcount; | |
2320 | struct timespec tp; | |
2321 | ||
2322 | memset(info, 0, sizeof(struct sysinfo)); | |
2323 | ||
45c64940 | 2324 | get_monotonic_boottime(&tp); |
4a22f166 SR |
2325 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
2326 | ||
2327 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2328 | ||
2329 | info->procs = nr_threads; | |
2330 | ||
2331 | si_meminfo(info); | |
2332 | si_swapinfo(info); | |
2333 | ||
2334 | /* | |
2335 | * If the sum of all the available memory (i.e. ram + swap) | |
2336 | * is less than can be stored in a 32 bit unsigned long then | |
2337 | * we can be binary compatible with 2.2.x kernels. If not, | |
2338 | * well, in that case 2.2.x was broken anyways... | |
2339 | * | |
2340 | * -Erik Andersen <andersee@debian.org> | |
2341 | */ | |
2342 | ||
2343 | mem_total = info->totalram + info->totalswap; | |
2344 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2345 | goto out; | |
2346 | bitcount = 0; | |
2347 | mem_unit = info->mem_unit; | |
2348 | while (mem_unit > 1) { | |
2349 | bitcount++; | |
2350 | mem_unit >>= 1; | |
2351 | sav_total = mem_total; | |
2352 | mem_total <<= 1; | |
2353 | if (mem_total < sav_total) | |
2354 | goto out; | |
2355 | } | |
2356 | ||
2357 | /* | |
2358 | * If mem_total did not overflow, multiply all memory values by | |
2359 | * info->mem_unit and set it to 1. This leaves things compatible | |
2360 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2361 | * kernels... | |
2362 | */ | |
2363 | ||
2364 | info->mem_unit = 1; | |
2365 | info->totalram <<= bitcount; | |
2366 | info->freeram <<= bitcount; | |
2367 | info->sharedram <<= bitcount; | |
2368 | info->bufferram <<= bitcount; | |
2369 | info->totalswap <<= bitcount; | |
2370 | info->freeswap <<= bitcount; | |
2371 | info->totalhigh <<= bitcount; | |
2372 | info->freehigh <<= bitcount; | |
2373 | ||
2374 | out: | |
2375 | return 0; | |
2376 | } | |
2377 | ||
2378 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2379 | { | |
2380 | struct sysinfo val; | |
2381 | ||
2382 | do_sysinfo(&val); | |
2383 | ||
2384 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2385 | return -EFAULT; | |
2386 | ||
2387 | return 0; | |
2388 | } | |
2389 | ||
2390 | #ifdef CONFIG_COMPAT | |
2391 | struct compat_sysinfo { | |
2392 | s32 uptime; | |
2393 | u32 loads[3]; | |
2394 | u32 totalram; | |
2395 | u32 freeram; | |
2396 | u32 sharedram; | |
2397 | u32 bufferram; | |
2398 | u32 totalswap; | |
2399 | u32 freeswap; | |
2400 | u16 procs; | |
2401 | u16 pad; | |
2402 | u32 totalhigh; | |
2403 | u32 freehigh; | |
2404 | u32 mem_unit; | |
2405 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2406 | }; | |
2407 | ||
2408 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2409 | { | |
2410 | struct sysinfo s; | |
2411 | ||
2412 | do_sysinfo(&s); | |
2413 | ||
2414 | /* Check to see if any memory value is too large for 32-bit and scale | |
2415 | * down if needed | |
2416 | */ | |
0baae41e | 2417 | if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) { |
4a22f166 SR |
2418 | int bitcount = 0; |
2419 | ||
2420 | while (s.mem_unit < PAGE_SIZE) { | |
2421 | s.mem_unit <<= 1; | |
2422 | bitcount++; | |
2423 | } | |
2424 | ||
2425 | s.totalram >>= bitcount; | |
2426 | s.freeram >>= bitcount; | |
2427 | s.sharedram >>= bitcount; | |
2428 | s.bufferram >>= bitcount; | |
2429 | s.totalswap >>= bitcount; | |
2430 | s.freeswap >>= bitcount; | |
2431 | s.totalhigh >>= bitcount; | |
2432 | s.freehigh >>= bitcount; | |
2433 | } | |
2434 | ||
2435 | if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) || | |
2436 | __put_user(s.uptime, &info->uptime) || | |
2437 | __put_user(s.loads[0], &info->loads[0]) || | |
2438 | __put_user(s.loads[1], &info->loads[1]) || | |
2439 | __put_user(s.loads[2], &info->loads[2]) || | |
2440 | __put_user(s.totalram, &info->totalram) || | |
2441 | __put_user(s.freeram, &info->freeram) || | |
2442 | __put_user(s.sharedram, &info->sharedram) || | |
2443 | __put_user(s.bufferram, &info->bufferram) || | |
2444 | __put_user(s.totalswap, &info->totalswap) || | |
2445 | __put_user(s.freeswap, &info->freeswap) || | |
2446 | __put_user(s.procs, &info->procs) || | |
2447 | __put_user(s.totalhigh, &info->totalhigh) || | |
2448 | __put_user(s.freehigh, &info->freehigh) || | |
2449 | __put_user(s.mem_unit, &info->mem_unit)) | |
2450 | return -EFAULT; | |
2451 | ||
2452 | return 0; | |
2453 | } | |
2454 | #endif /* CONFIG_COMPAT */ |