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> | |
586278d7 CC |
44 | #include <linux/mm.h> |
45 | #include <linux/mempolicy.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> |
7fe5e042 | 51 | #include <linux/binfmts.h> |
1da177e4 | 52 | |
4a22f166 SR |
53 | #include <linux/sched.h> |
54 | #include <linux/rcupdate.h> | |
55 | #include <linux/uidgid.h> | |
56 | #include <linux/cred.h> | |
57 | ||
04c6862c | 58 | #include <linux/kmsg_dump.h> |
be27425d AK |
59 | /* Move somewhere else to avoid recompiling? */ |
60 | #include <generated/utsrelease.h> | |
04c6862c | 61 | |
1da177e4 LT |
62 | #include <asm/uaccess.h> |
63 | #include <asm/io.h> | |
64 | #include <asm/unistd.h> | |
65 | ||
66 | #ifndef SET_UNALIGN_CTL | |
ec94fc3d | 67 | # define SET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
68 | #endif |
69 | #ifndef GET_UNALIGN_CTL | |
ec94fc3d | 70 | # define GET_UNALIGN_CTL(a, b) (-EINVAL) |
1da177e4 LT |
71 | #endif |
72 | #ifndef SET_FPEMU_CTL | |
ec94fc3d | 73 | # define SET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
74 | #endif |
75 | #ifndef GET_FPEMU_CTL | |
ec94fc3d | 76 | # define GET_FPEMU_CTL(a, b) (-EINVAL) |
1da177e4 LT |
77 | #endif |
78 | #ifndef SET_FPEXC_CTL | |
ec94fc3d | 79 | # define SET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 LT |
80 | #endif |
81 | #ifndef GET_FPEXC_CTL | |
ec94fc3d | 82 | # define GET_FPEXC_CTL(a, b) (-EINVAL) |
1da177e4 | 83 | #endif |
651d765d | 84 | #ifndef GET_ENDIAN |
ec94fc3d | 85 | # define GET_ENDIAN(a, b) (-EINVAL) |
651d765d AB |
86 | #endif |
87 | #ifndef SET_ENDIAN | |
ec94fc3d | 88 | # define SET_ENDIAN(a, b) (-EINVAL) |
651d765d | 89 | #endif |
8fb402bc EB |
90 | #ifndef GET_TSC_CTL |
91 | # define GET_TSC_CTL(a) (-EINVAL) | |
92 | #endif | |
93 | #ifndef SET_TSC_CTL | |
94 | # define SET_TSC_CTL(a) (-EINVAL) | |
95 | #endif | |
fe3d197f | 96 | #ifndef MPX_ENABLE_MANAGEMENT |
46a6e0cf | 97 | # define MPX_ENABLE_MANAGEMENT() (-EINVAL) |
fe3d197f DH |
98 | #endif |
99 | #ifndef MPX_DISABLE_MANAGEMENT | |
46a6e0cf | 100 | # define MPX_DISABLE_MANAGEMENT() (-EINVAL) |
fe3d197f | 101 | #endif |
9791554b PB |
102 | #ifndef GET_FP_MODE |
103 | # define GET_FP_MODE(a) (-EINVAL) | |
104 | #endif | |
105 | #ifndef SET_FP_MODE | |
106 | # define SET_FP_MODE(a,b) (-EINVAL) | |
107 | #endif | |
1da177e4 LT |
108 | |
109 | /* | |
110 | * this is where the system-wide overflow UID and GID are defined, for | |
111 | * architectures that now have 32-bit UID/GID but didn't in the past | |
112 | */ | |
113 | ||
114 | int overflowuid = DEFAULT_OVERFLOWUID; | |
115 | int overflowgid = DEFAULT_OVERFLOWGID; | |
116 | ||
1da177e4 LT |
117 | EXPORT_SYMBOL(overflowuid); |
118 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
119 | |
120 | /* | |
121 | * the same as above, but for filesystems which can only store a 16-bit | |
122 | * UID and GID. as such, this is needed on all architectures | |
123 | */ | |
124 | ||
125 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
126 | int fs_overflowgid = DEFAULT_FS_OVERFLOWUID; | |
127 | ||
128 | EXPORT_SYMBOL(fs_overflowuid); | |
129 | EXPORT_SYMBOL(fs_overflowgid); | |
130 | ||
fc832ad3 SH |
131 | /* |
132 | * Returns true if current's euid is same as p's uid or euid, | |
133 | * or has CAP_SYS_NICE to p's user_ns. | |
134 | * | |
135 | * Called with rcu_read_lock, creds are safe | |
136 | */ | |
137 | static bool set_one_prio_perm(struct task_struct *p) | |
138 | { | |
139 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
140 | ||
5af66203 EB |
141 | if (uid_eq(pcred->uid, cred->euid) || |
142 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 143 | return true; |
c4a4d603 | 144 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
145 | return true; |
146 | return false; | |
147 | } | |
148 | ||
c69e8d9c DH |
149 | /* |
150 | * set the priority of a task | |
151 | * - the caller must hold the RCU read lock | |
152 | */ | |
1da177e4 LT |
153 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
154 | { | |
155 | int no_nice; | |
156 | ||
fc832ad3 | 157 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
158 | error = -EPERM; |
159 | goto out; | |
160 | } | |
e43379f1 | 161 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
162 | error = -EACCES; |
163 | goto out; | |
164 | } | |
165 | no_nice = security_task_setnice(p, niceval); | |
166 | if (no_nice) { | |
167 | error = no_nice; | |
168 | goto out; | |
169 | } | |
170 | if (error == -ESRCH) | |
171 | error = 0; | |
172 | set_user_nice(p, niceval); | |
173 | out: | |
174 | return error; | |
175 | } | |
176 | ||
754fe8d2 | 177 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
178 | { |
179 | struct task_struct *g, *p; | |
180 | struct user_struct *user; | |
86a264ab | 181 | const struct cred *cred = current_cred(); |
1da177e4 | 182 | int error = -EINVAL; |
41487c65 | 183 | struct pid *pgrp; |
7b44ab97 | 184 | kuid_t uid; |
1da177e4 | 185 | |
3e88c553 | 186 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
187 | goto out; |
188 | ||
189 | /* normalize: avoid signed division (rounding problems) */ | |
190 | error = -ESRCH; | |
c4a4d2f4 DY |
191 | if (niceval < MIN_NICE) |
192 | niceval = MIN_NICE; | |
193 | if (niceval > MAX_NICE) | |
194 | niceval = MAX_NICE; | |
1da177e4 | 195 | |
d4581a23 | 196 | rcu_read_lock(); |
1da177e4 LT |
197 | read_lock(&tasklist_lock); |
198 | switch (which) { | |
ec94fc3d | 199 | case PRIO_PROCESS: |
200 | if (who) | |
201 | p = find_task_by_vpid(who); | |
202 | else | |
203 | p = current; | |
204 | if (p) | |
205 | error = set_one_prio(p, niceval, error); | |
206 | break; | |
207 | case PRIO_PGRP: | |
208 | if (who) | |
209 | pgrp = find_vpid(who); | |
210 | else | |
211 | pgrp = task_pgrp(current); | |
212 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
213 | error = set_one_prio(p, niceval, error); | |
214 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
215 | break; | |
216 | case PRIO_USER: | |
217 | uid = make_kuid(cred->user_ns, who); | |
218 | user = cred->user; | |
219 | if (!who) | |
220 | uid = cred->uid; | |
221 | else if (!uid_eq(uid, cred->uid)) { | |
222 | user = find_user(uid); | |
223 | if (!user) | |
86a264ab | 224 | goto out_unlock; /* No processes for this user */ |
ec94fc3d | 225 | } |
226 | do_each_thread(g, p) { | |
8639b461 | 227 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) |
ec94fc3d | 228 | error = set_one_prio(p, niceval, error); |
229 | } while_each_thread(g, p); | |
230 | if (!uid_eq(uid, cred->uid)) | |
231 | free_uid(user); /* For find_user() */ | |
232 | break; | |
1da177e4 LT |
233 | } |
234 | out_unlock: | |
235 | read_unlock(&tasklist_lock); | |
d4581a23 | 236 | rcu_read_unlock(); |
1da177e4 LT |
237 | out: |
238 | return error; | |
239 | } | |
240 | ||
241 | /* | |
242 | * Ugh. To avoid negative return values, "getpriority()" will | |
243 | * not return the normal nice-value, but a negated value that | |
244 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
245 | * to stay compatible. | |
246 | */ | |
754fe8d2 | 247 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
248 | { |
249 | struct task_struct *g, *p; | |
250 | struct user_struct *user; | |
86a264ab | 251 | const struct cred *cred = current_cred(); |
1da177e4 | 252 | long niceval, retval = -ESRCH; |
41487c65 | 253 | struct pid *pgrp; |
7b44ab97 | 254 | kuid_t uid; |
1da177e4 | 255 | |
3e88c553 | 256 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
257 | return -EINVAL; |
258 | ||
70118837 | 259 | rcu_read_lock(); |
1da177e4 LT |
260 | read_lock(&tasklist_lock); |
261 | switch (which) { | |
ec94fc3d | 262 | case PRIO_PROCESS: |
263 | if (who) | |
264 | p = find_task_by_vpid(who); | |
265 | else | |
266 | p = current; | |
267 | if (p) { | |
268 | niceval = nice_to_rlimit(task_nice(p)); | |
269 | if (niceval > retval) | |
270 | retval = niceval; | |
271 | } | |
272 | break; | |
273 | case PRIO_PGRP: | |
274 | if (who) | |
275 | pgrp = find_vpid(who); | |
276 | else | |
277 | pgrp = task_pgrp(current); | |
278 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
279 | niceval = nice_to_rlimit(task_nice(p)); | |
280 | if (niceval > retval) | |
281 | retval = niceval; | |
282 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
283 | break; | |
284 | case PRIO_USER: | |
285 | uid = make_kuid(cred->user_ns, who); | |
286 | user = cred->user; | |
287 | if (!who) | |
288 | uid = cred->uid; | |
289 | else if (!uid_eq(uid, cred->uid)) { | |
290 | user = find_user(uid); | |
291 | if (!user) | |
292 | goto out_unlock; /* No processes for this user */ | |
293 | } | |
294 | do_each_thread(g, p) { | |
8639b461 | 295 | if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) { |
7aa2c016 | 296 | niceval = nice_to_rlimit(task_nice(p)); |
1da177e4 LT |
297 | if (niceval > retval) |
298 | retval = niceval; | |
299 | } | |
ec94fc3d | 300 | } while_each_thread(g, p); |
301 | if (!uid_eq(uid, cred->uid)) | |
302 | free_uid(user); /* for find_user() */ | |
303 | break; | |
1da177e4 LT |
304 | } |
305 | out_unlock: | |
306 | read_unlock(&tasklist_lock); | |
70118837 | 307 | rcu_read_unlock(); |
1da177e4 LT |
308 | |
309 | return retval; | |
310 | } | |
311 | ||
1da177e4 LT |
312 | /* |
313 | * Unprivileged users may change the real gid to the effective gid | |
314 | * or vice versa. (BSD-style) | |
315 | * | |
316 | * If you set the real gid at all, or set the effective gid to a value not | |
317 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
318 | * | |
319 | * This makes it possible for a setgid program to completely drop its | |
320 | * privileges, which is often a useful assertion to make when you are doing | |
321 | * a security audit over a program. | |
322 | * | |
323 | * The general idea is that a program which uses just setregid() will be | |
324 | * 100% compatible with BSD. A program which uses just setgid() will be | |
ec94fc3d | 325 | * 100% compatible with POSIX with saved IDs. |
1da177e4 LT |
326 | * |
327 | * SMP: There are not races, the GIDs are checked only by filesystem | |
328 | * operations (as far as semantic preservation is concerned). | |
329 | */ | |
2813893f | 330 | #ifdef CONFIG_MULTIUSER |
ae1251ab | 331 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
1da177e4 | 332 | { |
a29c33f4 | 333 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
334 | const struct cred *old; |
335 | struct cred *new; | |
1da177e4 | 336 | int retval; |
a29c33f4 EB |
337 | kgid_t krgid, kegid; |
338 | ||
339 | krgid = make_kgid(ns, rgid); | |
340 | kegid = make_kgid(ns, egid); | |
341 | ||
342 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
343 | return -EINVAL; | |
344 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
345 | return -EINVAL; | |
1da177e4 | 346 | |
d84f4f99 DH |
347 | new = prepare_creds(); |
348 | if (!new) | |
349 | return -ENOMEM; | |
350 | old = current_cred(); | |
351 | ||
d84f4f99 | 352 | retval = -EPERM; |
1da177e4 | 353 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
354 | if (gid_eq(old->gid, krgid) || |
355 | gid_eq(old->egid, krgid) || | |
c7b96acf | 356 | ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 | 357 | new->gid = krgid; |
1da177e4 | 358 | else |
d84f4f99 | 359 | goto error; |
1da177e4 LT |
360 | } |
361 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
362 | if (gid_eq(old->gid, kegid) || |
363 | gid_eq(old->egid, kegid) || | |
364 | gid_eq(old->sgid, kegid) || | |
c7b96acf | 365 | ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 | 366 | new->egid = kegid; |
756184b7 | 367 | else |
d84f4f99 | 368 | goto error; |
1da177e4 | 369 | } |
d84f4f99 | 370 | |
1da177e4 | 371 | if (rgid != (gid_t) -1 || |
a29c33f4 | 372 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
373 | new->sgid = new->egid; |
374 | new->fsgid = new->egid; | |
375 | ||
376 | return commit_creds(new); | |
377 | ||
378 | error: | |
379 | abort_creds(new); | |
380 | return retval; | |
1da177e4 LT |
381 | } |
382 | ||
383 | /* | |
ec94fc3d | 384 | * setgid() is implemented like SysV w/ SAVED_IDS |
1da177e4 LT |
385 | * |
386 | * SMP: Same implicit races as above. | |
387 | */ | |
ae1251ab | 388 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
1da177e4 | 389 | { |
a29c33f4 | 390 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
391 | const struct cred *old; |
392 | struct cred *new; | |
1da177e4 | 393 | int retval; |
a29c33f4 EB |
394 | kgid_t kgid; |
395 | ||
396 | kgid = make_kgid(ns, gid); | |
397 | if (!gid_valid(kgid)) | |
398 | return -EINVAL; | |
1da177e4 | 399 | |
d84f4f99 DH |
400 | new = prepare_creds(); |
401 | if (!new) | |
402 | return -ENOMEM; | |
403 | old = current_cred(); | |
404 | ||
d84f4f99 | 405 | retval = -EPERM; |
c7b96acf | 406 | if (ns_capable(old->user_ns, CAP_SETGID)) |
a29c33f4 EB |
407 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
408 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
409 | new->egid = new->fsgid = kgid; | |
1da177e4 | 410 | else |
d84f4f99 | 411 | goto error; |
1da177e4 | 412 | |
d84f4f99 DH |
413 | return commit_creds(new); |
414 | ||
415 | error: | |
416 | abort_creds(new); | |
417 | return retval; | |
1da177e4 | 418 | } |
54e99124 | 419 | |
d84f4f99 DH |
420 | /* |
421 | * change the user struct in a credentials set to match the new UID | |
422 | */ | |
423 | static int set_user(struct cred *new) | |
1da177e4 LT |
424 | { |
425 | struct user_struct *new_user; | |
426 | ||
078de5f7 | 427 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
428 | if (!new_user) |
429 | return -EAGAIN; | |
430 | ||
72fa5997 VK |
431 | /* |
432 | * We don't fail in case of NPROC limit excess here because too many | |
433 | * poorly written programs don't check set*uid() return code, assuming | |
434 | * it never fails if called by root. We may still enforce NPROC limit | |
435 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
436 | * failure to the execve() stage. | |
437 | */ | |
78d7d407 | 438 | if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) && |
72fa5997 VK |
439 | new_user != INIT_USER) |
440 | current->flags |= PF_NPROC_EXCEEDED; | |
441 | else | |
442 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 443 | |
d84f4f99 DH |
444 | free_uid(new->user); |
445 | new->user = new_user; | |
1da177e4 LT |
446 | return 0; |
447 | } | |
448 | ||
449 | /* | |
450 | * Unprivileged users may change the real uid to the effective uid | |
451 | * or vice versa. (BSD-style) | |
452 | * | |
453 | * If you set the real uid at all, or set the effective uid to a value not | |
454 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
455 | * | |
456 | * This makes it possible for a setuid program to completely drop its | |
457 | * privileges, which is often a useful assertion to make when you are doing | |
458 | * a security audit over a program. | |
459 | * | |
460 | * The general idea is that a program which uses just setreuid() will be | |
461 | * 100% compatible with BSD. A program which uses just setuid() will be | |
ec94fc3d | 462 | * 100% compatible with POSIX with saved IDs. |
1da177e4 | 463 | */ |
ae1251ab | 464 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
1da177e4 | 465 | { |
a29c33f4 | 466 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
467 | const struct cred *old; |
468 | struct cred *new; | |
1da177e4 | 469 | int retval; |
a29c33f4 EB |
470 | kuid_t kruid, keuid; |
471 | ||
472 | kruid = make_kuid(ns, ruid); | |
473 | keuid = make_kuid(ns, euid); | |
474 | ||
475 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
476 | return -EINVAL; | |
477 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
478 | return -EINVAL; | |
1da177e4 | 479 | |
d84f4f99 DH |
480 | new = prepare_creds(); |
481 | if (!new) | |
482 | return -ENOMEM; | |
483 | old = current_cred(); | |
484 | ||
d84f4f99 | 485 | retval = -EPERM; |
1da177e4 | 486 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
487 | new->uid = kruid; |
488 | if (!uid_eq(old->uid, kruid) && | |
489 | !uid_eq(old->euid, kruid) && | |
c7b96acf | 490 | !ns_capable(old->user_ns, CAP_SETUID)) |
d84f4f99 | 491 | goto error; |
1da177e4 LT |
492 | } |
493 | ||
494 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
495 | new->euid = keuid; |
496 | if (!uid_eq(old->uid, keuid) && | |
497 | !uid_eq(old->euid, keuid) && | |
498 | !uid_eq(old->suid, keuid) && | |
c7b96acf | 499 | !ns_capable(old->user_ns, CAP_SETUID)) |
d84f4f99 | 500 | goto error; |
1da177e4 LT |
501 | } |
502 | ||
a29c33f4 | 503 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
504 | retval = set_user(new); |
505 | if (retval < 0) | |
506 | goto error; | |
507 | } | |
1da177e4 | 508 | if (ruid != (uid_t) -1 || |
a29c33f4 | 509 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
510 | new->suid = new->euid; |
511 | new->fsuid = new->euid; | |
1da177e4 | 512 | |
d84f4f99 DH |
513 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
514 | if (retval < 0) | |
515 | goto error; | |
1da177e4 | 516 | |
d84f4f99 | 517 | return commit_creds(new); |
1da177e4 | 518 | |
d84f4f99 DH |
519 | error: |
520 | abort_creds(new); | |
521 | return retval; | |
522 | } | |
ec94fc3d | 523 | |
1da177e4 | 524 | /* |
ec94fc3d | 525 | * setuid() is implemented like SysV with SAVED_IDS |
526 | * | |
1da177e4 | 527 | * Note that SAVED_ID's is deficient in that a setuid root program |
ec94fc3d | 528 | * like sendmail, for example, cannot set its uid to be a normal |
1da177e4 LT |
529 | * user and then switch back, because if you're root, setuid() sets |
530 | * the saved uid too. If you don't like this, blame the bright people | |
531 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
532 | * will allow a root program to temporarily drop privileges and be able to | |
ec94fc3d | 533 | * regain them by swapping the real and effective uid. |
1da177e4 | 534 | */ |
ae1251ab | 535 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
1da177e4 | 536 | { |
a29c33f4 | 537 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
538 | const struct cred *old; |
539 | struct cred *new; | |
1da177e4 | 540 | int retval; |
a29c33f4 EB |
541 | kuid_t kuid; |
542 | ||
543 | kuid = make_kuid(ns, uid); | |
544 | if (!uid_valid(kuid)) | |
545 | return -EINVAL; | |
1da177e4 | 546 | |
d84f4f99 DH |
547 | new = prepare_creds(); |
548 | if (!new) | |
549 | return -ENOMEM; | |
550 | old = current_cred(); | |
551 | ||
d84f4f99 | 552 | retval = -EPERM; |
c7b96acf | 553 | if (ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
554 | new->suid = new->uid = kuid; |
555 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
556 | retval = set_user(new); |
557 | if (retval < 0) | |
558 | goto error; | |
d84f4f99 | 559 | } |
a29c33f4 | 560 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 561 | goto error; |
1da177e4 | 562 | } |
1da177e4 | 563 | |
a29c33f4 | 564 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
565 | |
566 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
567 | if (retval < 0) | |
568 | goto error; | |
1da177e4 | 569 | |
d84f4f99 | 570 | return commit_creds(new); |
1da177e4 | 571 | |
d84f4f99 DH |
572 | error: |
573 | abort_creds(new); | |
574 | return retval; | |
1da177e4 LT |
575 | } |
576 | ||
577 | ||
578 | /* | |
579 | * This function implements a generic ability to update ruid, euid, | |
580 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
581 | */ | |
ae1251ab | 582 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
1da177e4 | 583 | { |
a29c33f4 | 584 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
585 | const struct cred *old; |
586 | struct cred *new; | |
1da177e4 | 587 | int retval; |
a29c33f4 EB |
588 | kuid_t kruid, keuid, ksuid; |
589 | ||
590 | kruid = make_kuid(ns, ruid); | |
591 | keuid = make_kuid(ns, euid); | |
592 | ksuid = make_kuid(ns, suid); | |
593 | ||
594 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
595 | return -EINVAL; | |
596 | ||
597 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
598 | return -EINVAL; | |
599 | ||
600 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
601 | return -EINVAL; | |
1da177e4 | 602 | |
d84f4f99 DH |
603 | new = prepare_creds(); |
604 | if (!new) | |
605 | return -ENOMEM; | |
606 | ||
d84f4f99 | 607 | old = current_cred(); |
1da177e4 | 608 | |
d84f4f99 | 609 | retval = -EPERM; |
c7b96acf | 610 | if (!ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
611 | if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && |
612 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid)) | |
d84f4f99 | 613 | goto error; |
a29c33f4 EB |
614 | if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && |
615 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid)) | |
d84f4f99 | 616 | goto error; |
a29c33f4 EB |
617 | if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && |
618 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid)) | |
d84f4f99 | 619 | goto error; |
1da177e4 | 620 | } |
d84f4f99 | 621 | |
1da177e4 | 622 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
623 | new->uid = kruid; |
624 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
625 | retval = set_user(new); |
626 | if (retval < 0) | |
627 | goto error; | |
628 | } | |
1da177e4 | 629 | } |
d84f4f99 | 630 | if (euid != (uid_t) -1) |
a29c33f4 | 631 | new->euid = keuid; |
1da177e4 | 632 | if (suid != (uid_t) -1) |
a29c33f4 | 633 | new->suid = ksuid; |
d84f4f99 | 634 | new->fsuid = new->euid; |
1da177e4 | 635 | |
d84f4f99 DH |
636 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
637 | if (retval < 0) | |
638 | goto error; | |
1da177e4 | 639 | |
d84f4f99 | 640 | return commit_creds(new); |
1da177e4 | 641 | |
d84f4f99 DH |
642 | error: |
643 | abort_creds(new); | |
644 | return retval; | |
1da177e4 LT |
645 | } |
646 | ||
a29c33f4 | 647 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 648 | { |
86a264ab | 649 | const struct cred *cred = current_cred(); |
1da177e4 | 650 | int retval; |
a29c33f4 EB |
651 | uid_t ruid, euid, suid; |
652 | ||
653 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
654 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
655 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 656 | |
ec94fc3d | 657 | retval = put_user(ruid, ruidp); |
658 | if (!retval) { | |
659 | retval = put_user(euid, euidp); | |
660 | if (!retval) | |
661 | return put_user(suid, suidp); | |
662 | } | |
1da177e4 LT |
663 | return retval; |
664 | } | |
665 | ||
666 | /* | |
667 | * Same as above, but for rgid, egid, sgid. | |
668 | */ | |
ae1251ab | 669 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
1da177e4 | 670 | { |
a29c33f4 | 671 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
672 | const struct cred *old; |
673 | struct cred *new; | |
1da177e4 | 674 | int retval; |
a29c33f4 EB |
675 | kgid_t krgid, kegid, ksgid; |
676 | ||
677 | krgid = make_kgid(ns, rgid); | |
678 | kegid = make_kgid(ns, egid); | |
679 | ksgid = make_kgid(ns, sgid); | |
680 | ||
681 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
682 | return -EINVAL; | |
683 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
684 | return -EINVAL; | |
685 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
686 | return -EINVAL; | |
1da177e4 | 687 | |
d84f4f99 DH |
688 | new = prepare_creds(); |
689 | if (!new) | |
690 | return -ENOMEM; | |
691 | old = current_cred(); | |
692 | ||
d84f4f99 | 693 | retval = -EPERM; |
c7b96acf | 694 | if (!ns_capable(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
695 | if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && |
696 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid)) | |
d84f4f99 | 697 | goto error; |
a29c33f4 EB |
698 | if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && |
699 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid)) | |
d84f4f99 | 700 | goto error; |
a29c33f4 EB |
701 | if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && |
702 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid)) | |
d84f4f99 | 703 | goto error; |
1da177e4 | 704 | } |
d84f4f99 | 705 | |
1da177e4 | 706 | if (rgid != (gid_t) -1) |
a29c33f4 | 707 | new->gid = krgid; |
d84f4f99 | 708 | if (egid != (gid_t) -1) |
a29c33f4 | 709 | new->egid = kegid; |
1da177e4 | 710 | if (sgid != (gid_t) -1) |
a29c33f4 | 711 | new->sgid = ksgid; |
d84f4f99 | 712 | new->fsgid = new->egid; |
1da177e4 | 713 | |
d84f4f99 DH |
714 | return commit_creds(new); |
715 | ||
716 | error: | |
717 | abort_creds(new); | |
718 | return retval; | |
1da177e4 LT |
719 | } |
720 | ||
a29c33f4 | 721 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 722 | { |
86a264ab | 723 | const struct cred *cred = current_cred(); |
1da177e4 | 724 | int retval; |
a29c33f4 EB |
725 | gid_t rgid, egid, sgid; |
726 | ||
727 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
728 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
729 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 730 | |
ec94fc3d | 731 | retval = put_user(rgid, rgidp); |
732 | if (!retval) { | |
733 | retval = put_user(egid, egidp); | |
734 | if (!retval) | |
735 | retval = put_user(sgid, sgidp); | |
736 | } | |
1da177e4 LT |
737 | |
738 | return retval; | |
739 | } | |
740 | ||
741 | ||
742 | /* | |
743 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
744 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
745 | * whatever uid it wants to). It normally shadows "euid", except when | |
746 | * explicitly set by setfsuid() or for access.. | |
747 | */ | |
ae1251ab | 748 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
1da177e4 | 749 | { |
d84f4f99 DH |
750 | const struct cred *old; |
751 | struct cred *new; | |
752 | uid_t old_fsuid; | |
a29c33f4 EB |
753 | kuid_t kuid; |
754 | ||
755 | old = current_cred(); | |
756 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
757 | ||
758 | kuid = make_kuid(old->user_ns, uid); | |
759 | if (!uid_valid(kuid)) | |
760 | return old_fsuid; | |
1da177e4 | 761 | |
d84f4f99 DH |
762 | new = prepare_creds(); |
763 | if (!new) | |
a29c33f4 | 764 | return old_fsuid; |
1da177e4 | 765 | |
a29c33f4 EB |
766 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
767 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
c7b96acf | 768 | ns_capable(old->user_ns, CAP_SETUID)) { |
a29c33f4 EB |
769 | if (!uid_eq(kuid, old->fsuid)) { |
770 | new->fsuid = kuid; | |
d84f4f99 DH |
771 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
772 | goto change_okay; | |
1da177e4 | 773 | } |
1da177e4 LT |
774 | } |
775 | ||
d84f4f99 DH |
776 | abort_creds(new); |
777 | return old_fsuid; | |
1da177e4 | 778 | |
d84f4f99 DH |
779 | change_okay: |
780 | commit_creds(new); | |
1da177e4 LT |
781 | return old_fsuid; |
782 | } | |
783 | ||
784 | /* | |
f42df9e6 | 785 | * Samma på svenska.. |
1da177e4 | 786 | */ |
ae1251ab | 787 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) |
1da177e4 | 788 | { |
d84f4f99 DH |
789 | const struct cred *old; |
790 | struct cred *new; | |
791 | gid_t old_fsgid; | |
a29c33f4 EB |
792 | kgid_t kgid; |
793 | ||
794 | old = current_cred(); | |
795 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
796 | ||
797 | kgid = make_kgid(old->user_ns, gid); | |
798 | if (!gid_valid(kgid)) | |
799 | return old_fsgid; | |
d84f4f99 DH |
800 | |
801 | new = prepare_creds(); | |
802 | if (!new) | |
a29c33f4 | 803 | return old_fsgid; |
1da177e4 | 804 | |
a29c33f4 EB |
805 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
806 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
c7b96acf | 807 | ns_capable(old->user_ns, CAP_SETGID)) { |
a29c33f4 EB |
808 | if (!gid_eq(kgid, old->fsgid)) { |
809 | new->fsgid = kgid; | |
d84f4f99 | 810 | goto change_okay; |
1da177e4 | 811 | } |
1da177e4 | 812 | } |
d84f4f99 | 813 | |
d84f4f99 DH |
814 | abort_creds(new); |
815 | return old_fsgid; | |
816 | ||
817 | change_okay: | |
818 | commit_creds(new); | |
1da177e4 LT |
819 | return old_fsgid; |
820 | } | |
2813893f | 821 | #endif /* CONFIG_MULTIUSER */ |
1da177e4 | 822 | |
4a22f166 SR |
823 | /** |
824 | * sys_getpid - return the thread group id of the current process | |
825 | * | |
826 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
827 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
828 | * which case the tgid is the same in all threads of the same group. | |
829 | * | |
830 | * This is SMP safe as current->tgid does not change. | |
831 | */ | |
832 | SYSCALL_DEFINE0(getpid) | |
833 | { | |
834 | return task_tgid_vnr(current); | |
835 | } | |
836 | ||
837 | /* Thread ID - the internal kernel "pid" */ | |
838 | SYSCALL_DEFINE0(gettid) | |
839 | { | |
840 | return task_pid_vnr(current); | |
841 | } | |
842 | ||
843 | /* | |
844 | * Accessing ->real_parent is not SMP-safe, it could | |
845 | * change from under us. However, we can use a stale | |
846 | * value of ->real_parent under rcu_read_lock(), see | |
847 | * release_task()->call_rcu(delayed_put_task_struct). | |
848 | */ | |
849 | SYSCALL_DEFINE0(getppid) | |
850 | { | |
851 | int pid; | |
852 | ||
853 | rcu_read_lock(); | |
854 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
855 | rcu_read_unlock(); | |
856 | ||
857 | return pid; | |
858 | } | |
859 | ||
860 | SYSCALL_DEFINE0(getuid) | |
861 | { | |
862 | /* Only we change this so SMP safe */ | |
863 | return from_kuid_munged(current_user_ns(), current_uid()); | |
864 | } | |
865 | ||
866 | SYSCALL_DEFINE0(geteuid) | |
867 | { | |
868 | /* Only we change this so SMP safe */ | |
869 | return from_kuid_munged(current_user_ns(), current_euid()); | |
870 | } | |
871 | ||
872 | SYSCALL_DEFINE0(getgid) | |
873 | { | |
874 | /* Only we change this so SMP safe */ | |
875 | return from_kgid_munged(current_user_ns(), current_gid()); | |
876 | } | |
877 | ||
878 | SYSCALL_DEFINE0(getegid) | |
879 | { | |
880 | /* Only we change this so SMP safe */ | |
881 | return from_kgid_munged(current_user_ns(), current_egid()); | |
882 | } | |
883 | ||
f06febc9 FM |
884 | void do_sys_times(struct tms *tms) |
885 | { | |
0cf55e1e | 886 | cputime_t tgutime, tgstime, cutime, cstime; |
f06febc9 | 887 | |
e80d0a1a | 888 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
889 | cutime = current->signal->cutime; |
890 | cstime = current->signal->cstime; | |
0cf55e1e HS |
891 | tms->tms_utime = cputime_to_clock_t(tgutime); |
892 | tms->tms_stime = cputime_to_clock_t(tgstime); | |
f06febc9 FM |
893 | tms->tms_cutime = cputime_to_clock_t(cutime); |
894 | tms->tms_cstime = cputime_to_clock_t(cstime); | |
895 | } | |
896 | ||
58fd3aa2 | 897 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 898 | { |
1da177e4 LT |
899 | if (tbuf) { |
900 | struct tms tmp; | |
f06febc9 FM |
901 | |
902 | do_sys_times(&tmp); | |
1da177e4 LT |
903 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
904 | return -EFAULT; | |
905 | } | |
e3d5a27d | 906 | force_successful_syscall_return(); |
1da177e4 LT |
907 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
908 | } | |
909 | ||
910 | /* | |
911 | * This needs some heavy checking ... | |
912 | * I just haven't the stomach for it. I also don't fully | |
913 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
914 | * | |
915 | * OK, I think I have the protection semantics right.... this is really | |
916 | * only important on a multi-user system anyway, to make sure one user | |
917 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
918 | * | |
98611e4e | 919 | * !PF_FORKNOEXEC check to conform completely to POSIX. |
1da177e4 | 920 | */ |
b290ebe2 | 921 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
922 | { |
923 | struct task_struct *p; | |
ee0acf90 | 924 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
925 | struct pid *pgrp; |
926 | int err; | |
1da177e4 LT |
927 | |
928 | if (!pid) | |
b488893a | 929 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
930 | if (!pgid) |
931 | pgid = pid; | |
932 | if (pgid < 0) | |
933 | return -EINVAL; | |
950eaaca | 934 | rcu_read_lock(); |
1da177e4 LT |
935 | |
936 | /* From this point forward we keep holding onto the tasklist lock | |
937 | * so that our parent does not change from under us. -DaveM | |
938 | */ | |
939 | write_lock_irq(&tasklist_lock); | |
940 | ||
941 | err = -ESRCH; | |
4e021306 | 942 | p = find_task_by_vpid(pid); |
1da177e4 LT |
943 | if (!p) |
944 | goto out; | |
945 | ||
946 | err = -EINVAL; | |
947 | if (!thread_group_leader(p)) | |
948 | goto out; | |
949 | ||
4e021306 | 950 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 951 | err = -EPERM; |
41487c65 | 952 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
953 | goto out; |
954 | err = -EACCES; | |
98611e4e | 955 | if (!(p->flags & PF_FORKNOEXEC)) |
1da177e4 LT |
956 | goto out; |
957 | } else { | |
958 | err = -ESRCH; | |
ee0acf90 | 959 | if (p != group_leader) |
1da177e4 LT |
960 | goto out; |
961 | } | |
962 | ||
963 | err = -EPERM; | |
964 | if (p->signal->leader) | |
965 | goto out; | |
966 | ||
4e021306 | 967 | pgrp = task_pid(p); |
1da177e4 | 968 | if (pgid != pid) { |
b488893a | 969 | struct task_struct *g; |
1da177e4 | 970 | |
4e021306 ON |
971 | pgrp = find_vpid(pgid); |
972 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 973 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 974 | goto out; |
1da177e4 LT |
975 | } |
976 | ||
1da177e4 LT |
977 | err = security_task_setpgid(p, pgid); |
978 | if (err) | |
979 | goto out; | |
980 | ||
1b0f7ffd | 981 | if (task_pgrp(p) != pgrp) |
83beaf3c | 982 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
983 | |
984 | err = 0; | |
985 | out: | |
986 | /* All paths lead to here, thus we are safe. -DaveM */ | |
987 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 988 | rcu_read_unlock(); |
1da177e4 LT |
989 | return err; |
990 | } | |
991 | ||
dbf040d9 | 992 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1da177e4 | 993 | { |
12a3de0a ON |
994 | struct task_struct *p; |
995 | struct pid *grp; | |
996 | int retval; | |
997 | ||
998 | rcu_read_lock(); | |
756184b7 | 999 | if (!pid) |
12a3de0a | 1000 | grp = task_pgrp(current); |
756184b7 | 1001 | else { |
1da177e4 | 1002 | retval = -ESRCH; |
12a3de0a ON |
1003 | p = find_task_by_vpid(pid); |
1004 | if (!p) | |
1005 | goto out; | |
1006 | grp = task_pgrp(p); | |
1007 | if (!grp) | |
1008 | goto out; | |
1009 | ||
1010 | retval = security_task_getpgid(p); | |
1011 | if (retval) | |
1012 | goto out; | |
1da177e4 | 1013 | } |
12a3de0a ON |
1014 | retval = pid_vnr(grp); |
1015 | out: | |
1016 | rcu_read_unlock(); | |
1017 | return retval; | |
1da177e4 LT |
1018 | } |
1019 | ||
1020 | #ifdef __ARCH_WANT_SYS_GETPGRP | |
1021 | ||
dbf040d9 | 1022 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1023 | { |
12a3de0a | 1024 | return sys_getpgid(0); |
1da177e4 LT |
1025 | } |
1026 | ||
1027 | #endif | |
1028 | ||
dbf040d9 | 1029 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1030 | { |
1dd768c0 ON |
1031 | struct task_struct *p; |
1032 | struct pid *sid; | |
1033 | int retval; | |
1034 | ||
1035 | rcu_read_lock(); | |
756184b7 | 1036 | if (!pid) |
1dd768c0 | 1037 | sid = task_session(current); |
756184b7 | 1038 | else { |
1da177e4 | 1039 | retval = -ESRCH; |
1dd768c0 ON |
1040 | p = find_task_by_vpid(pid); |
1041 | if (!p) | |
1042 | goto out; | |
1043 | sid = task_session(p); | |
1044 | if (!sid) | |
1045 | goto out; | |
1046 | ||
1047 | retval = security_task_getsid(p); | |
1048 | if (retval) | |
1049 | goto out; | |
1da177e4 | 1050 | } |
1dd768c0 ON |
1051 | retval = pid_vnr(sid); |
1052 | out: | |
1053 | rcu_read_unlock(); | |
1054 | return retval; | |
1da177e4 LT |
1055 | } |
1056 | ||
81dabb46 ON |
1057 | static void set_special_pids(struct pid *pid) |
1058 | { | |
1059 | struct task_struct *curr = current->group_leader; | |
1060 | ||
1061 | if (task_session(curr) != pid) | |
1062 | change_pid(curr, PIDTYPE_SID, pid); | |
1063 | ||
1064 | if (task_pgrp(curr) != pid) | |
1065 | change_pid(curr, PIDTYPE_PGID, pid); | |
1066 | } | |
1067 | ||
b290ebe2 | 1068 | SYSCALL_DEFINE0(setsid) |
1da177e4 | 1069 | { |
e19f247a | 1070 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1071 | struct pid *sid = task_pid(group_leader); |
1072 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1073 | int err = -EPERM; |
1074 | ||
1da177e4 | 1075 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1076 | /* Fail if I am already a session leader */ |
1077 | if (group_leader->signal->leader) | |
1078 | goto out; | |
1079 | ||
430c6231 ON |
1080 | /* Fail if a process group id already exists that equals the |
1081 | * proposed session id. | |
390e2ff0 | 1082 | */ |
6806aac6 | 1083 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1084 | goto out; |
1085 | ||
e19f247a | 1086 | group_leader->signal->leader = 1; |
81dabb46 | 1087 | set_special_pids(sid); |
24ec839c | 1088 | |
9c9f4ded | 1089 | proc_clear_tty(group_leader); |
24ec839c | 1090 | |
e4cc0a9c | 1091 | err = session; |
1da177e4 LT |
1092 | out: |
1093 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1094 | if (err > 0) { |
0d0df599 | 1095 | proc_sid_connector(group_leader); |
5091faa4 MG |
1096 | sched_autogroup_create_attach(group_leader); |
1097 | } | |
1da177e4 LT |
1098 | return err; |
1099 | } | |
1100 | ||
1da177e4 LT |
1101 | DECLARE_RWSEM(uts_sem); |
1102 | ||
e28cbf22 CH |
1103 | #ifdef COMPAT_UTS_MACHINE |
1104 | #define override_architecture(name) \ | |
46da2766 | 1105 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1106 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1107 | sizeof(COMPAT_UTS_MACHINE))) | |
1108 | #else | |
1109 | #define override_architecture(name) 0 | |
1110 | #endif | |
1111 | ||
be27425d AK |
1112 | /* |
1113 | * Work around broken programs that cannot handle "Linux 3.0". | |
1114 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
39afb5ee | 1115 | * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60. |
be27425d | 1116 | */ |
2702b152 | 1117 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1118 | { |
1119 | int ret = 0; | |
be27425d AK |
1120 | |
1121 | if (current->personality & UNAME26) { | |
2702b152 KC |
1122 | const char *rest = UTS_RELEASE; |
1123 | char buf[65] = { 0 }; | |
be27425d AK |
1124 | int ndots = 0; |
1125 | unsigned v; | |
2702b152 | 1126 | size_t copy; |
be27425d AK |
1127 | |
1128 | while (*rest) { | |
1129 | if (*rest == '.' && ++ndots >= 3) | |
1130 | break; | |
1131 | if (!isdigit(*rest) && *rest != '.') | |
1132 | break; | |
1133 | rest++; | |
1134 | } | |
39afb5ee | 1135 | v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60; |
31fd84b9 | 1136 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1137 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1138 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1139 | } |
1140 | return ret; | |
1141 | } | |
1142 | ||
e48fbb69 | 1143 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 LT |
1144 | { |
1145 | int errno = 0; | |
1146 | ||
1147 | down_read(&uts_sem); | |
e9ff3990 | 1148 | if (copy_to_user(name, utsname(), sizeof *name)) |
1da177e4 LT |
1149 | errno = -EFAULT; |
1150 | up_read(&uts_sem); | |
e28cbf22 | 1151 | |
be27425d AK |
1152 | if (!errno && override_release(name->release, sizeof(name->release))) |
1153 | errno = -EFAULT; | |
e28cbf22 CH |
1154 | if (!errno && override_architecture(name)) |
1155 | errno = -EFAULT; | |
1da177e4 LT |
1156 | return errno; |
1157 | } | |
1158 | ||
5cacdb4a CH |
1159 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1160 | /* | |
1161 | * Old cruft | |
1162 | */ | |
1163 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1164 | { | |
1165 | int error = 0; | |
1166 | ||
1167 | if (!name) | |
1168 | return -EFAULT; | |
1169 | ||
1170 | down_read(&uts_sem); | |
1171 | if (copy_to_user(name, utsname(), sizeof(*name))) | |
1172 | error = -EFAULT; | |
1173 | up_read(&uts_sem); | |
1174 | ||
be27425d AK |
1175 | if (!error && override_release(name->release, sizeof(name->release))) |
1176 | error = -EFAULT; | |
5cacdb4a CH |
1177 | if (!error && override_architecture(name)) |
1178 | error = -EFAULT; | |
1179 | return error; | |
1180 | } | |
1181 | ||
1182 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1183 | { | |
1184 | int error; | |
1185 | ||
1186 | if (!name) | |
1187 | return -EFAULT; | |
1188 | if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname))) | |
1189 | return -EFAULT; | |
1190 | ||
1191 | down_read(&uts_sem); | |
1192 | error = __copy_to_user(&name->sysname, &utsname()->sysname, | |
1193 | __OLD_UTS_LEN); | |
1194 | error |= __put_user(0, name->sysname + __OLD_UTS_LEN); | |
1195 | error |= __copy_to_user(&name->nodename, &utsname()->nodename, | |
1196 | __OLD_UTS_LEN); | |
1197 | error |= __put_user(0, name->nodename + __OLD_UTS_LEN); | |
1198 | error |= __copy_to_user(&name->release, &utsname()->release, | |
1199 | __OLD_UTS_LEN); | |
1200 | error |= __put_user(0, name->release + __OLD_UTS_LEN); | |
1201 | error |= __copy_to_user(&name->version, &utsname()->version, | |
1202 | __OLD_UTS_LEN); | |
1203 | error |= __put_user(0, name->version + __OLD_UTS_LEN); | |
1204 | error |= __copy_to_user(&name->machine, &utsname()->machine, | |
1205 | __OLD_UTS_LEN); | |
1206 | error |= __put_user(0, name->machine + __OLD_UTS_LEN); | |
1207 | up_read(&uts_sem); | |
1208 | ||
1209 | if (!error && override_architecture(name)) | |
1210 | error = -EFAULT; | |
be27425d AK |
1211 | if (!error && override_release(name->release, sizeof(name->release))) |
1212 | error = -EFAULT; | |
5cacdb4a CH |
1213 | return error ? -EFAULT : 0; |
1214 | } | |
1215 | #endif | |
1216 | ||
5a8a82b1 | 1217 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1218 | { |
1219 | int errno; | |
1220 | char tmp[__NEW_UTS_LEN]; | |
1221 | ||
bb96a6f5 | 1222 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1223 | return -EPERM; |
fc832ad3 | 1224 | |
1da177e4 LT |
1225 | if (len < 0 || len > __NEW_UTS_LEN) |
1226 | return -EINVAL; | |
1227 | down_write(&uts_sem); | |
1228 | errno = -EFAULT; | |
1229 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1230 | struct new_utsname *u = utsname(); |
1231 | ||
1232 | memcpy(u->nodename, tmp, len); | |
1233 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1234 | errno = 0; |
499eea6b | 1235 | uts_proc_notify(UTS_PROC_HOSTNAME); |
1da177e4 LT |
1236 | } |
1237 | up_write(&uts_sem); | |
1238 | return errno; | |
1239 | } | |
1240 | ||
1241 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1242 | ||
5a8a82b1 | 1243 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 LT |
1244 | { |
1245 | int i, errno; | |
9679e4dd | 1246 | struct new_utsname *u; |
1da177e4 LT |
1247 | |
1248 | if (len < 0) | |
1249 | return -EINVAL; | |
1250 | down_read(&uts_sem); | |
9679e4dd AM |
1251 | u = utsname(); |
1252 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1253 | if (i > len) |
1254 | i = len; | |
1255 | errno = 0; | |
9679e4dd | 1256 | if (copy_to_user(name, u->nodename, i)) |
1da177e4 LT |
1257 | errno = -EFAULT; |
1258 | up_read(&uts_sem); | |
1259 | return errno; | |
1260 | } | |
1261 | ||
1262 | #endif | |
1263 | ||
1264 | /* | |
1265 | * Only setdomainname; getdomainname can be implemented by calling | |
1266 | * uname() | |
1267 | */ | |
5a8a82b1 | 1268 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1269 | { |
1270 | int errno; | |
1271 | char tmp[__NEW_UTS_LEN]; | |
1272 | ||
fc832ad3 | 1273 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1274 | return -EPERM; |
1275 | if (len < 0 || len > __NEW_UTS_LEN) | |
1276 | return -EINVAL; | |
1277 | ||
1278 | down_write(&uts_sem); | |
1279 | errno = -EFAULT; | |
1280 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1281 | struct new_utsname *u = utsname(); |
1282 | ||
1283 | memcpy(u->domainname, tmp, len); | |
1284 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1285 | errno = 0; |
499eea6b | 1286 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
1da177e4 LT |
1287 | } |
1288 | up_write(&uts_sem); | |
1289 | return errno; | |
1290 | } | |
1291 | ||
e48fbb69 | 1292 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1293 | { |
b9518345 JS |
1294 | struct rlimit value; |
1295 | int ret; | |
1296 | ||
1297 | ret = do_prlimit(current, resource, NULL, &value); | |
1298 | if (!ret) | |
1299 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1300 | ||
1301 | return ret; | |
1da177e4 LT |
1302 | } |
1303 | ||
1304 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT | |
1305 | ||
1306 | /* | |
1307 | * Back compatibility for getrlimit. Needed for some apps. | |
1308 | */ | |
e48fbb69 HC |
1309 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1310 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1311 | { |
1312 | struct rlimit x; | |
1313 | if (resource >= RLIM_NLIMITS) | |
1314 | return -EINVAL; | |
1315 | ||
1316 | task_lock(current->group_leader); | |
1317 | x = current->signal->rlim[resource]; | |
1318 | task_unlock(current->group_leader); | |
756184b7 | 1319 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1320 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1321 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 | 1322 | x.rlim_max = 0x7FFFFFFF; |
ec94fc3d | 1323 | return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0; |
1da177e4 LT |
1324 | } |
1325 | ||
1326 | #endif | |
1327 | ||
c022a0ac JS |
1328 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1329 | { | |
1330 | #if BITS_PER_LONG < 64 | |
1331 | return rlim64 >= ULONG_MAX; | |
1332 | #else | |
1333 | return rlim64 == RLIM64_INFINITY; | |
1334 | #endif | |
1335 | } | |
1336 | ||
1337 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1338 | { | |
1339 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1340 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1341 | else | |
1342 | rlim64->rlim_cur = rlim->rlim_cur; | |
1343 | if (rlim->rlim_max == RLIM_INFINITY) | |
1344 | rlim64->rlim_max = RLIM64_INFINITY; | |
1345 | else | |
1346 | rlim64->rlim_max = rlim->rlim_max; | |
1347 | } | |
1348 | ||
1349 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1350 | { | |
1351 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1352 | rlim->rlim_cur = RLIM_INFINITY; | |
1353 | else | |
1354 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1355 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1356 | rlim->rlim_max = RLIM_INFINITY; | |
1357 | else | |
1358 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1359 | } | |
1360 | ||
1c1e618d | 1361 | /* make sure you are allowed to change @tsk limits before calling this */ |
5b41535a JS |
1362 | int do_prlimit(struct task_struct *tsk, unsigned int resource, |
1363 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1da177e4 | 1364 | { |
5b41535a | 1365 | struct rlimit *rlim; |
86f162f4 | 1366 | int retval = 0; |
1da177e4 LT |
1367 | |
1368 | if (resource >= RLIM_NLIMITS) | |
1369 | return -EINVAL; | |
5b41535a JS |
1370 | if (new_rlim) { |
1371 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1372 | return -EINVAL; | |
1373 | if (resource == RLIMIT_NOFILE && | |
1374 | new_rlim->rlim_max > sysctl_nr_open) | |
1375 | return -EPERM; | |
1376 | } | |
1da177e4 | 1377 | |
1c1e618d JS |
1378 | /* protect tsk->signal and tsk->sighand from disappearing */ |
1379 | read_lock(&tasklist_lock); | |
1380 | if (!tsk->sighand) { | |
1381 | retval = -ESRCH; | |
1382 | goto out; | |
1383 | } | |
1384 | ||
5b41535a | 1385 | rlim = tsk->signal->rlim + resource; |
86f162f4 | 1386 | task_lock(tsk->group_leader); |
5b41535a | 1387 | if (new_rlim) { |
fc832ad3 SH |
1388 | /* Keep the capable check against init_user_ns until |
1389 | cgroups can contain all limits */ | |
5b41535a JS |
1390 | if (new_rlim->rlim_max > rlim->rlim_max && |
1391 | !capable(CAP_SYS_RESOURCE)) | |
1392 | retval = -EPERM; | |
1393 | if (!retval) | |
1394 | retval = security_task_setrlimit(tsk->group_leader, | |
1395 | resource, new_rlim); | |
1396 | if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) { | |
1397 | /* | |
1398 | * The caller is asking for an immediate RLIMIT_CPU | |
1399 | * expiry. But we use the zero value to mean "it was | |
1400 | * never set". So let's cheat and make it one second | |
1401 | * instead | |
1402 | */ | |
1403 | new_rlim->rlim_cur = 1; | |
1404 | } | |
1405 | } | |
1406 | if (!retval) { | |
1407 | if (old_rlim) | |
1408 | *old_rlim = *rlim; | |
1409 | if (new_rlim) | |
1410 | *rlim = *new_rlim; | |
9926e4c7 | 1411 | } |
7855c35d | 1412 | task_unlock(tsk->group_leader); |
1da177e4 | 1413 | |
d3561f78 AM |
1414 | /* |
1415 | * RLIMIT_CPU handling. Note that the kernel fails to return an error | |
1416 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a | |
1417 | * very long-standing error, and fixing it now risks breakage of | |
1418 | * applications, so we live with it | |
1419 | */ | |
5b41535a JS |
1420 | if (!retval && new_rlim && resource == RLIMIT_CPU && |
1421 | new_rlim->rlim_cur != RLIM_INFINITY) | |
1422 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); | |
ec9e16ba | 1423 | out: |
1c1e618d | 1424 | read_unlock(&tasklist_lock); |
2fb9d268 | 1425 | return retval; |
1da177e4 LT |
1426 | } |
1427 | ||
c022a0ac JS |
1428 | /* rcu lock must be held */ |
1429 | static int check_prlimit_permission(struct task_struct *task) | |
1430 | { | |
1431 | const struct cred *cred = current_cred(), *tcred; | |
1432 | ||
fc832ad3 SH |
1433 | if (current == task) |
1434 | return 0; | |
c022a0ac | 1435 | |
fc832ad3 | 1436 | tcred = __task_cred(task); |
5af66203 EB |
1437 | if (uid_eq(cred->uid, tcred->euid) && |
1438 | uid_eq(cred->uid, tcred->suid) && | |
1439 | uid_eq(cred->uid, tcred->uid) && | |
1440 | gid_eq(cred->gid, tcred->egid) && | |
1441 | gid_eq(cred->gid, tcred->sgid) && | |
1442 | gid_eq(cred->gid, tcred->gid)) | |
fc832ad3 | 1443 | return 0; |
c4a4d603 | 1444 | if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) |
fc832ad3 SH |
1445 | return 0; |
1446 | ||
1447 | return -EPERM; | |
c022a0ac JS |
1448 | } |
1449 | ||
1450 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1451 | const struct rlimit64 __user *, new_rlim, | |
1452 | struct rlimit64 __user *, old_rlim) | |
1453 | { | |
1454 | struct rlimit64 old64, new64; | |
1455 | struct rlimit old, new; | |
1456 | struct task_struct *tsk; | |
1457 | int ret; | |
1458 | ||
1459 | if (new_rlim) { | |
1460 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1461 | return -EFAULT; | |
1462 | rlim64_to_rlim(&new64, &new); | |
1463 | } | |
1464 | ||
1465 | rcu_read_lock(); | |
1466 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1467 | if (!tsk) { | |
1468 | rcu_read_unlock(); | |
1469 | return -ESRCH; | |
1470 | } | |
1471 | ret = check_prlimit_permission(tsk); | |
1472 | if (ret) { | |
1473 | rcu_read_unlock(); | |
1474 | return ret; | |
1475 | } | |
1476 | get_task_struct(tsk); | |
1477 | rcu_read_unlock(); | |
1478 | ||
1479 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1480 | old_rlim ? &old : NULL); | |
1481 | ||
1482 | if (!ret && old_rlim) { | |
1483 | rlim_to_rlim64(&old, &old64); | |
1484 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1485 | ret = -EFAULT; | |
1486 | } | |
1487 | ||
1488 | put_task_struct(tsk); | |
1489 | return ret; | |
1490 | } | |
1491 | ||
7855c35d JS |
1492 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1493 | { | |
1494 | struct rlimit new_rlim; | |
1495 | ||
1496 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1497 | return -EFAULT; | |
5b41535a | 1498 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1499 | } |
1500 | ||
1da177e4 LT |
1501 | /* |
1502 | * It would make sense to put struct rusage in the task_struct, | |
1503 | * except that would make the task_struct be *really big*. After | |
1504 | * task_struct gets moved into malloc'ed memory, it would | |
1505 | * make sense to do this. It will make moving the rest of the information | |
1506 | * a lot simpler! (Which we're not doing right now because we're not | |
1507 | * measuring them yet). | |
1508 | * | |
1da177e4 LT |
1509 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1510 | * races with threads incrementing their own counters. But since word | |
1511 | * reads are atomic, we either get new values or old values and we don't | |
1512 | * care which for the sums. We always take the siglock to protect reading | |
1513 | * the c* fields from p->signal from races with exit.c updating those | |
1514 | * fields when reaping, so a sample either gets all the additions of a | |
1515 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1516 | * |
de047c1b RT |
1517 | * Locking: |
1518 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1519 | * for the cases current multithreaded, non-current single threaded | |
1520 | * non-current multithreaded. Thread traversal is now safe with | |
1521 | * the siglock held. | |
1522 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1523 | * single threaded, as no one else can take our signal_struct away, no one | |
1524 | * else can reap the children to update signal->c* counters, and no one else | |
1525 | * can race with the signal-> fields. If we do not take any lock, the | |
1526 | * signal-> fields could be read out of order while another thread was just | |
1527 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1528 | * On the writer side, write memory barrier is implied in __exit_signal | |
1529 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1530 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1531 | * |
1da177e4 LT |
1532 | */ |
1533 | ||
f06febc9 | 1534 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1535 | { |
679c9cd4 SK |
1536 | r->ru_nvcsw += t->nvcsw; |
1537 | r->ru_nivcsw += t->nivcsw; | |
1538 | r->ru_minflt += t->min_flt; | |
1539 | r->ru_majflt += t->maj_flt; | |
1540 | r->ru_inblock += task_io_get_inblock(t); | |
1541 | r->ru_oublock += task_io_get_oublock(t); | |
1542 | } | |
1543 | ||
1da177e4 LT |
1544 | static void k_getrusage(struct task_struct *p, int who, struct rusage *r) |
1545 | { | |
1546 | struct task_struct *t; | |
1547 | unsigned long flags; | |
0cf55e1e | 1548 | cputime_t tgutime, tgstime, utime, stime; |
1f10206c | 1549 | unsigned long maxrss = 0; |
1da177e4 | 1550 | |
ec94fc3d | 1551 | memset((char *)r, 0, sizeof (*r)); |
64861634 | 1552 | utime = stime = 0; |
1da177e4 | 1553 | |
679c9cd4 | 1554 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1555 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1556 | accumulate_thread_rusage(p, r); |
1f10206c | 1557 | maxrss = p->signal->maxrss; |
679c9cd4 SK |
1558 | goto out; |
1559 | } | |
1560 | ||
d6cf723a | 1561 | if (!lock_task_sighand(p, &flags)) |
de047c1b | 1562 | return; |
0f59cc4a | 1563 | |
1da177e4 | 1564 | switch (who) { |
ec94fc3d | 1565 | case RUSAGE_BOTH: |
1566 | case RUSAGE_CHILDREN: | |
1567 | utime = p->signal->cutime; | |
1568 | stime = p->signal->cstime; | |
1569 | r->ru_nvcsw = p->signal->cnvcsw; | |
1570 | r->ru_nivcsw = p->signal->cnivcsw; | |
1571 | r->ru_minflt = p->signal->cmin_flt; | |
1572 | r->ru_majflt = p->signal->cmaj_flt; | |
1573 | r->ru_inblock = p->signal->cinblock; | |
1574 | r->ru_oublock = p->signal->coublock; | |
1575 | maxrss = p->signal->cmaxrss; | |
1576 | ||
1577 | if (who == RUSAGE_CHILDREN) | |
1da177e4 | 1578 | break; |
0f59cc4a | 1579 | |
ec94fc3d | 1580 | case RUSAGE_SELF: |
1581 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); | |
1582 | utime += tgutime; | |
1583 | stime += tgstime; | |
1584 | r->ru_nvcsw += p->signal->nvcsw; | |
1585 | r->ru_nivcsw += p->signal->nivcsw; | |
1586 | r->ru_minflt += p->signal->min_flt; | |
1587 | r->ru_majflt += p->signal->maj_flt; | |
1588 | r->ru_inblock += p->signal->inblock; | |
1589 | r->ru_oublock += p->signal->oublock; | |
1590 | if (maxrss < p->signal->maxrss) | |
1591 | maxrss = p->signal->maxrss; | |
1592 | t = p; | |
1593 | do { | |
1594 | accumulate_thread_rusage(t, r); | |
1595 | } while_each_thread(p, t); | |
1596 | break; | |
1597 | ||
1598 | default: | |
1599 | BUG(); | |
1da177e4 | 1600 | } |
de047c1b | 1601 | unlock_task_sighand(p, &flags); |
de047c1b | 1602 | |
679c9cd4 | 1603 | out: |
0f59cc4a ON |
1604 | cputime_to_timeval(utime, &r->ru_utime); |
1605 | cputime_to_timeval(stime, &r->ru_stime); | |
1f10206c JP |
1606 | |
1607 | if (who != RUSAGE_CHILDREN) { | |
1608 | struct mm_struct *mm = get_task_mm(p); | |
ec94fc3d | 1609 | |
1f10206c JP |
1610 | if (mm) { |
1611 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1612 | mmput(mm); | |
1613 | } | |
1614 | } | |
1615 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1da177e4 LT |
1616 | } |
1617 | ||
1618 | int getrusage(struct task_struct *p, int who, struct rusage __user *ru) | |
1619 | { | |
1620 | struct rusage r; | |
ec94fc3d | 1621 | |
1da177e4 | 1622 | k_getrusage(p, who, &r); |
1da177e4 LT |
1623 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; |
1624 | } | |
1625 | ||
e48fbb69 | 1626 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 | 1627 | { |
679c9cd4 SK |
1628 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1629 | who != RUSAGE_THREAD) | |
1da177e4 LT |
1630 | return -EINVAL; |
1631 | return getrusage(current, who, ru); | |
1632 | } | |
1633 | ||
8d2d5c4a AV |
1634 | #ifdef CONFIG_COMPAT |
1635 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1636 | { | |
1637 | struct rusage r; | |
1638 | ||
1639 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1640 | who != RUSAGE_THREAD) | |
1641 | return -EINVAL; | |
1642 | ||
1643 | k_getrusage(current, who, &r); | |
1644 | return put_compat_rusage(&r, ru); | |
1645 | } | |
1646 | #endif | |
1647 | ||
e48fbb69 | 1648 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
1649 | { |
1650 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1651 | return mask; | |
1652 | } | |
3b7391de | 1653 | |
6e399cd1 | 1654 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
b32dfe37 | 1655 | { |
2903ff01 | 1656 | struct fd exe; |
6e399cd1 | 1657 | struct file *old_exe, *exe_file; |
496ad9aa | 1658 | struct inode *inode; |
2903ff01 | 1659 | int err; |
b32dfe37 | 1660 | |
2903ff01 AV |
1661 | exe = fdget(fd); |
1662 | if (!exe.file) | |
b32dfe37 CG |
1663 | return -EBADF; |
1664 | ||
496ad9aa | 1665 | inode = file_inode(exe.file); |
b32dfe37 CG |
1666 | |
1667 | /* | |
1668 | * Because the original mm->exe_file points to executable file, make | |
1669 | * sure that this one is executable as well, to avoid breaking an | |
1670 | * overall picture. | |
1671 | */ | |
1672 | err = -EACCES; | |
90f8572b | 1673 | if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path)) |
b32dfe37 CG |
1674 | goto exit; |
1675 | ||
496ad9aa | 1676 | err = inode_permission(inode, MAY_EXEC); |
b32dfe37 CG |
1677 | if (err) |
1678 | goto exit; | |
1679 | ||
bafb282d | 1680 | /* |
4229fb1d | 1681 | * Forbid mm->exe_file change if old file still mapped. |
bafb282d | 1682 | */ |
6e399cd1 | 1683 | exe_file = get_mm_exe_file(mm); |
bafb282d | 1684 | err = -EBUSY; |
6e399cd1 | 1685 | if (exe_file) { |
4229fb1d KK |
1686 | struct vm_area_struct *vma; |
1687 | ||
6e399cd1 DB |
1688 | down_read(&mm->mmap_sem); |
1689 | for (vma = mm->mmap; vma; vma = vma->vm_next) { | |
1690 | if (!vma->vm_file) | |
1691 | continue; | |
1692 | if (path_equal(&vma->vm_file->f_path, | |
1693 | &exe_file->f_path)) | |
1694 | goto exit_err; | |
1695 | } | |
1696 | ||
1697 | up_read(&mm->mmap_sem); | |
1698 | fput(exe_file); | |
bafb282d KK |
1699 | } |
1700 | ||
b32dfe37 CG |
1701 | /* |
1702 | * The symlink can be changed only once, just to disallow arbitrary | |
1703 | * transitions malicious software might bring in. This means one | |
1704 | * could make a snapshot over all processes running and monitor | |
1705 | * /proc/pid/exe changes to notice unusual activity if needed. | |
1706 | */ | |
bafb282d KK |
1707 | err = -EPERM; |
1708 | if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags)) | |
71fe97e1 | 1709 | goto exit; |
bafb282d | 1710 | |
4229fb1d | 1711 | err = 0; |
6e399cd1 DB |
1712 | /* set the new file, lockless */ |
1713 | get_file(exe.file); | |
1714 | old_exe = xchg(&mm->exe_file, exe.file); | |
1715 | if (old_exe) | |
1716 | fput(old_exe); | |
b32dfe37 | 1717 | exit: |
2903ff01 | 1718 | fdput(exe); |
b32dfe37 | 1719 | return err; |
6e399cd1 DB |
1720 | exit_err: |
1721 | up_read(&mm->mmap_sem); | |
1722 | fput(exe_file); | |
1723 | goto exit; | |
b32dfe37 CG |
1724 | } |
1725 | ||
f606b77f CG |
1726 | /* |
1727 | * WARNING: we don't require any capability here so be very careful | |
1728 | * in what is allowed for modification from userspace. | |
1729 | */ | |
1730 | static int validate_prctl_map(struct prctl_mm_map *prctl_map) | |
1731 | { | |
1732 | unsigned long mmap_max_addr = TASK_SIZE; | |
1733 | struct mm_struct *mm = current->mm; | |
1734 | int error = -EINVAL, i; | |
1735 | ||
1736 | static const unsigned char offsets[] = { | |
1737 | offsetof(struct prctl_mm_map, start_code), | |
1738 | offsetof(struct prctl_mm_map, end_code), | |
1739 | offsetof(struct prctl_mm_map, start_data), | |
1740 | offsetof(struct prctl_mm_map, end_data), | |
1741 | offsetof(struct prctl_mm_map, start_brk), | |
1742 | offsetof(struct prctl_mm_map, brk), | |
1743 | offsetof(struct prctl_mm_map, start_stack), | |
1744 | offsetof(struct prctl_mm_map, arg_start), | |
1745 | offsetof(struct prctl_mm_map, arg_end), | |
1746 | offsetof(struct prctl_mm_map, env_start), | |
1747 | offsetof(struct prctl_mm_map, env_end), | |
1748 | }; | |
1749 | ||
1750 | /* | |
1751 | * Make sure the members are not somewhere outside | |
1752 | * of allowed address space. | |
1753 | */ | |
1754 | for (i = 0; i < ARRAY_SIZE(offsets); i++) { | |
1755 | u64 val = *(u64 *)((char *)prctl_map + offsets[i]); | |
1756 | ||
1757 | if ((unsigned long)val >= mmap_max_addr || | |
1758 | (unsigned long)val < mmap_min_addr) | |
1759 | goto out; | |
1760 | } | |
1761 | ||
1762 | /* | |
1763 | * Make sure the pairs are ordered. | |
1764 | */ | |
1765 | #define __prctl_check_order(__m1, __op, __m2) \ | |
1766 | ((unsigned long)prctl_map->__m1 __op \ | |
1767 | (unsigned long)prctl_map->__m2) ? 0 : -EINVAL | |
1768 | error = __prctl_check_order(start_code, <, end_code); | |
1769 | error |= __prctl_check_order(start_data, <, end_data); | |
1770 | error |= __prctl_check_order(start_brk, <=, brk); | |
1771 | error |= __prctl_check_order(arg_start, <=, arg_end); | |
1772 | error |= __prctl_check_order(env_start, <=, env_end); | |
1773 | if (error) | |
1774 | goto out; | |
1775 | #undef __prctl_check_order | |
1776 | ||
1777 | error = -EINVAL; | |
1778 | ||
1779 | /* | |
1780 | * @brk should be after @end_data in traditional maps. | |
1781 | */ | |
1782 | if (prctl_map->start_brk <= prctl_map->end_data || | |
1783 | prctl_map->brk <= prctl_map->end_data) | |
1784 | goto out; | |
1785 | ||
1786 | /* | |
1787 | * Neither we should allow to override limits if they set. | |
1788 | */ | |
1789 | if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk, | |
1790 | prctl_map->start_brk, prctl_map->end_data, | |
1791 | prctl_map->start_data)) | |
1792 | goto out; | |
1793 | ||
1794 | /* | |
1795 | * Someone is trying to cheat the auxv vector. | |
1796 | */ | |
1797 | if (prctl_map->auxv_size) { | |
1798 | if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv)) | |
1799 | goto out; | |
1800 | } | |
1801 | ||
1802 | /* | |
1803 | * Finally, make sure the caller has the rights to | |
1804 | * change /proc/pid/exe link: only local root should | |
1805 | * be allowed to. | |
1806 | */ | |
1807 | if (prctl_map->exe_fd != (u32)-1) { | |
1808 | struct user_namespace *ns = current_user_ns(); | |
1809 | const struct cred *cred = current_cred(); | |
1810 | ||
1811 | if (!uid_eq(cred->uid, make_kuid(ns, 0)) || | |
1812 | !gid_eq(cred->gid, make_kgid(ns, 0))) | |
1813 | goto out; | |
1814 | } | |
1815 | ||
1816 | error = 0; | |
1817 | out: | |
1818 | return error; | |
1819 | } | |
1820 | ||
4a00e9df | 1821 | #ifdef CONFIG_CHECKPOINT_RESTORE |
f606b77f CG |
1822 | static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size) |
1823 | { | |
1824 | struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, }; | |
1825 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1826 | struct mm_struct *mm = current->mm; | |
1827 | int error; | |
1828 | ||
1829 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1830 | BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256); | |
1831 | ||
1832 | if (opt == PR_SET_MM_MAP_SIZE) | |
1833 | return put_user((unsigned int)sizeof(prctl_map), | |
1834 | (unsigned int __user *)addr); | |
1835 | ||
1836 | if (data_size != sizeof(prctl_map)) | |
1837 | return -EINVAL; | |
1838 | ||
1839 | if (copy_from_user(&prctl_map, addr, sizeof(prctl_map))) | |
1840 | return -EFAULT; | |
1841 | ||
1842 | error = validate_prctl_map(&prctl_map); | |
1843 | if (error) | |
1844 | return error; | |
1845 | ||
1846 | if (prctl_map.auxv_size) { | |
1847 | memset(user_auxv, 0, sizeof(user_auxv)); | |
1848 | if (copy_from_user(user_auxv, | |
1849 | (const void __user *)prctl_map.auxv, | |
1850 | prctl_map.auxv_size)) | |
1851 | return -EFAULT; | |
1852 | ||
1853 | /* Last entry must be AT_NULL as specification requires */ | |
1854 | user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL; | |
1855 | user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL; | |
1856 | } | |
1857 | ||
e5e99792 | 1858 | if (prctl_map.exe_fd != (u32)-1) { |
6e399cd1 | 1859 | error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd); |
e5e99792 MG |
1860 | if (error) |
1861 | return error; | |
1862 | } | |
1863 | ||
1864 | down_write(&mm->mmap_sem); | |
f606b77f CG |
1865 | |
1866 | /* | |
1867 | * We don't validate if these members are pointing to | |
1868 | * real present VMAs because application may have correspond | |
1869 | * VMAs already unmapped and kernel uses these members for statistics | |
1870 | * output in procfs mostly, except | |
1871 | * | |
1872 | * - @start_brk/@brk which are used in do_brk but kernel lookups | |
1873 | * for VMAs when updating these memvers so anything wrong written | |
1874 | * here cause kernel to swear at userspace program but won't lead | |
1875 | * to any problem in kernel itself | |
1876 | */ | |
1877 | ||
1878 | mm->start_code = prctl_map.start_code; | |
1879 | mm->end_code = prctl_map.end_code; | |
1880 | mm->start_data = prctl_map.start_data; | |
1881 | mm->end_data = prctl_map.end_data; | |
1882 | mm->start_brk = prctl_map.start_brk; | |
1883 | mm->brk = prctl_map.brk; | |
1884 | mm->start_stack = prctl_map.start_stack; | |
1885 | mm->arg_start = prctl_map.arg_start; | |
1886 | mm->arg_end = prctl_map.arg_end; | |
1887 | mm->env_start = prctl_map.env_start; | |
1888 | mm->env_end = prctl_map.env_end; | |
1889 | ||
1890 | /* | |
1891 | * Note this update of @saved_auxv is lockless thus | |
1892 | * if someone reads this member in procfs while we're | |
1893 | * updating -- it may get partly updated results. It's | |
1894 | * known and acceptable trade off: we leave it as is to | |
1895 | * not introduce additional locks here making the kernel | |
1896 | * more complex. | |
1897 | */ | |
1898 | if (prctl_map.auxv_size) | |
1899 | memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv)); | |
1900 | ||
e5e99792 MG |
1901 | up_write(&mm->mmap_sem); |
1902 | return 0; | |
f606b77f CG |
1903 | } |
1904 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
1905 | ||
4a00e9df AD |
1906 | static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr, |
1907 | unsigned long len) | |
1908 | { | |
1909 | /* | |
1910 | * This doesn't move the auxiliary vector itself since it's pinned to | |
1911 | * mm_struct, but it permits filling the vector with new values. It's | |
1912 | * up to the caller to provide sane values here, otherwise userspace | |
1913 | * tools which use this vector might be unhappy. | |
1914 | */ | |
1915 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
1916 | ||
1917 | if (len > sizeof(user_auxv)) | |
1918 | return -EINVAL; | |
1919 | ||
1920 | if (copy_from_user(user_auxv, (const void __user *)addr, len)) | |
1921 | return -EFAULT; | |
1922 | ||
1923 | /* Make sure the last entry is always AT_NULL */ | |
1924 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
1925 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
1926 | ||
1927 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
1928 | ||
1929 | task_lock(current); | |
1930 | memcpy(mm->saved_auxv, user_auxv, len); | |
1931 | task_unlock(current); | |
1932 | ||
1933 | return 0; | |
1934 | } | |
1935 | ||
028ee4be CG |
1936 | static int prctl_set_mm(int opt, unsigned long addr, |
1937 | unsigned long arg4, unsigned long arg5) | |
1938 | { | |
028ee4be | 1939 | struct mm_struct *mm = current->mm; |
4a00e9df | 1940 | struct prctl_mm_map prctl_map; |
fe8c7f5c CG |
1941 | struct vm_area_struct *vma; |
1942 | int error; | |
028ee4be | 1943 | |
f606b77f CG |
1944 | if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV && |
1945 | opt != PR_SET_MM_MAP && | |
1946 | opt != PR_SET_MM_MAP_SIZE))) | |
028ee4be CG |
1947 | return -EINVAL; |
1948 | ||
f606b77f CG |
1949 | #ifdef CONFIG_CHECKPOINT_RESTORE |
1950 | if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE) | |
1951 | return prctl_set_mm_map(opt, (const void __user *)addr, arg4); | |
1952 | #endif | |
1953 | ||
79f0713d | 1954 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
1955 | return -EPERM; |
1956 | ||
6e399cd1 DB |
1957 | if (opt == PR_SET_MM_EXE_FILE) |
1958 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
b32dfe37 | 1959 | |
4a00e9df AD |
1960 | if (opt == PR_SET_MM_AUXV) |
1961 | return prctl_set_auxv(mm, addr, arg4); | |
1962 | ||
1ad75b9e | 1963 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
1964 | return -EINVAL; |
1965 | ||
fe8c7f5c CG |
1966 | error = -EINVAL; |
1967 | ||
e5e99792 | 1968 | down_write(&mm->mmap_sem); |
028ee4be CG |
1969 | vma = find_vma(mm, addr); |
1970 | ||
4a00e9df AD |
1971 | prctl_map.start_code = mm->start_code; |
1972 | prctl_map.end_code = mm->end_code; | |
1973 | prctl_map.start_data = mm->start_data; | |
1974 | prctl_map.end_data = mm->end_data; | |
1975 | prctl_map.start_brk = mm->start_brk; | |
1976 | prctl_map.brk = mm->brk; | |
1977 | prctl_map.start_stack = mm->start_stack; | |
1978 | prctl_map.arg_start = mm->arg_start; | |
1979 | prctl_map.arg_end = mm->arg_end; | |
1980 | prctl_map.env_start = mm->env_start; | |
1981 | prctl_map.env_end = mm->env_end; | |
1982 | prctl_map.auxv = NULL; | |
1983 | prctl_map.auxv_size = 0; | |
1984 | prctl_map.exe_fd = -1; | |
1985 | ||
028ee4be CG |
1986 | switch (opt) { |
1987 | case PR_SET_MM_START_CODE: | |
4a00e9df | 1988 | prctl_map.start_code = addr; |
fe8c7f5c | 1989 | break; |
028ee4be | 1990 | case PR_SET_MM_END_CODE: |
4a00e9df | 1991 | prctl_map.end_code = addr; |
028ee4be | 1992 | break; |
028ee4be | 1993 | case PR_SET_MM_START_DATA: |
4a00e9df | 1994 | prctl_map.start_data = addr; |
028ee4be | 1995 | break; |
fe8c7f5c | 1996 | case PR_SET_MM_END_DATA: |
4a00e9df AD |
1997 | prctl_map.end_data = addr; |
1998 | break; | |
1999 | case PR_SET_MM_START_STACK: | |
2000 | prctl_map.start_stack = addr; | |
028ee4be | 2001 | break; |
028ee4be | 2002 | case PR_SET_MM_START_BRK: |
4a00e9df | 2003 | prctl_map.start_brk = addr; |
028ee4be | 2004 | break; |
028ee4be | 2005 | case PR_SET_MM_BRK: |
4a00e9df | 2006 | prctl_map.brk = addr; |
028ee4be | 2007 | break; |
4a00e9df AD |
2008 | case PR_SET_MM_ARG_START: |
2009 | prctl_map.arg_start = addr; | |
2010 | break; | |
2011 | case PR_SET_MM_ARG_END: | |
2012 | prctl_map.arg_end = addr; | |
2013 | break; | |
2014 | case PR_SET_MM_ENV_START: | |
2015 | prctl_map.env_start = addr; | |
2016 | break; | |
2017 | case PR_SET_MM_ENV_END: | |
2018 | prctl_map.env_end = addr; | |
2019 | break; | |
2020 | default: | |
2021 | goto out; | |
2022 | } | |
2023 | ||
2024 | error = validate_prctl_map(&prctl_map); | |
2025 | if (error) | |
2026 | goto out; | |
028ee4be | 2027 | |
4a00e9df | 2028 | switch (opt) { |
fe8c7f5c CG |
2029 | /* |
2030 | * If command line arguments and environment | |
2031 | * are placed somewhere else on stack, we can | |
2032 | * set them up here, ARG_START/END to setup | |
2033 | * command line argumets and ENV_START/END | |
2034 | * for environment. | |
2035 | */ | |
2036 | case PR_SET_MM_START_STACK: | |
2037 | case PR_SET_MM_ARG_START: | |
2038 | case PR_SET_MM_ARG_END: | |
2039 | case PR_SET_MM_ENV_START: | |
2040 | case PR_SET_MM_ENV_END: | |
2041 | if (!vma) { | |
2042 | error = -EFAULT; | |
2043 | goto out; | |
2044 | } | |
028ee4be CG |
2045 | } |
2046 | ||
4a00e9df AD |
2047 | mm->start_code = prctl_map.start_code; |
2048 | mm->end_code = prctl_map.end_code; | |
2049 | mm->start_data = prctl_map.start_data; | |
2050 | mm->end_data = prctl_map.end_data; | |
2051 | mm->start_brk = prctl_map.start_brk; | |
2052 | mm->brk = prctl_map.brk; | |
2053 | mm->start_stack = prctl_map.start_stack; | |
2054 | mm->arg_start = prctl_map.arg_start; | |
2055 | mm->arg_end = prctl_map.arg_end; | |
2056 | mm->env_start = prctl_map.env_start; | |
2057 | mm->env_end = prctl_map.env_end; | |
2058 | ||
028ee4be | 2059 | error = 0; |
028ee4be | 2060 | out: |
e5e99792 | 2061 | up_write(&mm->mmap_sem); |
028ee4be CG |
2062 | return error; |
2063 | } | |
300f786b | 2064 | |
52b36941 | 2065 | #ifdef CONFIG_CHECKPOINT_RESTORE |
300f786b CG |
2066 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2067 | { | |
2068 | return put_user(me->clear_child_tid, tid_addr); | |
2069 | } | |
52b36941 | 2070 | #else |
300f786b CG |
2071 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2072 | { | |
2073 | return -EINVAL; | |
2074 | } | |
028ee4be CG |
2075 | #endif |
2076 | ||
586278d7 CC |
2077 | #ifdef CONFIG_MMU |
2078 | static int prctl_update_vma_anon_name(struct vm_area_struct *vma, | |
2079 | struct vm_area_struct **prev, | |
2080 | unsigned long start, unsigned long end, | |
2081 | const char __user *name_addr) | |
2082 | { | |
2083 | struct mm_struct *mm = vma->vm_mm; | |
2084 | int error = 0; | |
2085 | pgoff_t pgoff; | |
2086 | ||
2087 | if (name_addr == vma_get_anon_name(vma)) { | |
2088 | *prev = vma; | |
2089 | goto out; | |
2090 | } | |
2091 | ||
2092 | pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT); | |
2093 | *prev = vma_merge(mm, *prev, start, end, vma->vm_flags, vma->anon_vma, | |
2094 | vma->vm_file, pgoff, vma_policy(vma), | |
29a4f01d | 2095 | vma->vm_userfaultfd_ctx, name_addr); |
586278d7 CC |
2096 | if (*prev) { |
2097 | vma = *prev; | |
2098 | goto success; | |
2099 | } | |
2100 | ||
2101 | *prev = vma; | |
2102 | ||
2103 | if (start != vma->vm_start) { | |
2104 | error = split_vma(mm, vma, start, 1); | |
2105 | if (error) | |
2106 | goto out; | |
2107 | } | |
2108 | ||
2109 | if (end != vma->vm_end) { | |
2110 | error = split_vma(mm, vma, end, 0); | |
2111 | if (error) | |
2112 | goto out; | |
2113 | } | |
2114 | ||
2115 | success: | |
2116 | if (!vma->vm_file) | |
2117 | vma->anon_name = name_addr; | |
2118 | ||
2119 | out: | |
2120 | if (error == -ENOMEM) | |
2121 | error = -EAGAIN; | |
2122 | return error; | |
2123 | } | |
2124 | ||
2125 | static int prctl_set_vma_anon_name(unsigned long start, unsigned long end, | |
2126 | unsigned long arg) | |
2127 | { | |
2128 | unsigned long tmp; | |
2129 | struct vm_area_struct *vma, *prev; | |
2130 | int unmapped_error = 0; | |
2131 | int error = -EINVAL; | |
2132 | ||
2133 | /* | |
2134 | * If the interval [start,end) covers some unmapped address | |
2135 | * ranges, just ignore them, but return -ENOMEM at the end. | |
2136 | * - this matches the handling in madvise. | |
2137 | */ | |
2138 | vma = find_vma_prev(current->mm, start, &prev); | |
2139 | if (vma && start > vma->vm_start) | |
2140 | prev = vma; | |
2141 | ||
2142 | for (;;) { | |
2143 | /* Still start < end. */ | |
2144 | error = -ENOMEM; | |
2145 | if (!vma) | |
2146 | return error; | |
2147 | ||
2148 | /* Here start < (end|vma->vm_end). */ | |
2149 | if (start < vma->vm_start) { | |
2150 | unmapped_error = -ENOMEM; | |
2151 | start = vma->vm_start; | |
2152 | if (start >= end) | |
2153 | return error; | |
2154 | } | |
2155 | ||
2156 | /* Here vma->vm_start <= start < (end|vma->vm_end) */ | |
2157 | tmp = vma->vm_end; | |
2158 | if (end < tmp) | |
2159 | tmp = end; | |
2160 | ||
2161 | /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */ | |
2162 | error = prctl_update_vma_anon_name(vma, &prev, start, tmp, | |
2163 | (const char __user *)arg); | |
2164 | if (error) | |
2165 | return error; | |
2166 | start = tmp; | |
2167 | if (prev && start < prev->vm_end) | |
2168 | start = prev->vm_end; | |
2169 | error = unmapped_error; | |
2170 | if (start >= end) | |
2171 | return error; | |
2172 | if (prev) | |
2173 | vma = prev->vm_next; | |
2174 | else /* madvise_remove dropped mmap_sem */ | |
2175 | vma = find_vma(current->mm, start); | |
2176 | } | |
2177 | } | |
2178 | ||
2179 | static int prctl_set_vma(unsigned long opt, unsigned long start, | |
2180 | unsigned long len_in, unsigned long arg) | |
2181 | { | |
2182 | struct mm_struct *mm = current->mm; | |
2183 | int error; | |
2184 | unsigned long len; | |
2185 | unsigned long end; | |
2186 | ||
2187 | if (start & ~PAGE_MASK) | |
2188 | return -EINVAL; | |
2189 | len = (len_in + ~PAGE_MASK) & PAGE_MASK; | |
2190 | ||
2191 | /* Check to see whether len was rounded up from small -ve to zero */ | |
2192 | if (len_in && !len) | |
2193 | return -EINVAL; | |
2194 | ||
2195 | end = start + len; | |
2196 | if (end < start) | |
2197 | return -EINVAL; | |
2198 | ||
2199 | if (end == start) | |
2200 | return 0; | |
2201 | ||
2202 | down_write(&mm->mmap_sem); | |
2203 | ||
2204 | switch (opt) { | |
2205 | case PR_SET_VMA_ANON_NAME: | |
2206 | error = prctl_set_vma_anon_name(start, end, arg); | |
2207 | break; | |
2208 | default: | |
2209 | error = -EINVAL; | |
2210 | } | |
2211 | ||
2212 | up_write(&mm->mmap_sem); | |
2213 | ||
2214 | return error; | |
2215 | } | |
2216 | #else /* CONFIG_MMU */ | |
2217 | static int prctl_set_vma(unsigned long opt, unsigned long start, | |
2218 | unsigned long len_in, unsigned long arg) | |
2219 | { | |
2220 | return -EINVAL; | |
2221 | } | |
2222 | #endif | |
2223 | ||
c4ea37c2 HC |
2224 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2225 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2226 | { |
b6dff3ec | 2227 | struct task_struct *me = current; |
f2902f90 | 2228 | struct task_struct *tsk; |
b6dff3ec DH |
2229 | unsigned char comm[sizeof(me->comm)]; |
2230 | long error; | |
1da177e4 | 2231 | |
d84f4f99 DH |
2232 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2233 | if (error != -ENOSYS) | |
1da177e4 LT |
2234 | return error; |
2235 | ||
d84f4f99 | 2236 | error = 0; |
1da177e4 | 2237 | switch (option) { |
f3cbd435 AM |
2238 | case PR_SET_PDEATHSIG: |
2239 | if (!valid_signal(arg2)) { | |
2240 | error = -EINVAL; | |
1da177e4 | 2241 | break; |
f3cbd435 AM |
2242 | } |
2243 | me->pdeath_signal = arg2; | |
2244 | break; | |
2245 | case PR_GET_PDEATHSIG: | |
2246 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2247 | break; | |
2248 | case PR_GET_DUMPABLE: | |
2249 | error = get_dumpable(me->mm); | |
2250 | break; | |
2251 | case PR_SET_DUMPABLE: | |
2252 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2253 | error = -EINVAL; | |
1da177e4 | 2254 | break; |
f3cbd435 AM |
2255 | } |
2256 | set_dumpable(me->mm, arg2); | |
2257 | break; | |
1da177e4 | 2258 | |
f3cbd435 AM |
2259 | case PR_SET_UNALIGN: |
2260 | error = SET_UNALIGN_CTL(me, arg2); | |
2261 | break; | |
2262 | case PR_GET_UNALIGN: | |
2263 | error = GET_UNALIGN_CTL(me, arg2); | |
2264 | break; | |
2265 | case PR_SET_FPEMU: | |
2266 | error = SET_FPEMU_CTL(me, arg2); | |
2267 | break; | |
2268 | case PR_GET_FPEMU: | |
2269 | error = GET_FPEMU_CTL(me, arg2); | |
2270 | break; | |
2271 | case PR_SET_FPEXC: | |
2272 | error = SET_FPEXC_CTL(me, arg2); | |
2273 | break; | |
2274 | case PR_GET_FPEXC: | |
2275 | error = GET_FPEXC_CTL(me, arg2); | |
2276 | break; | |
2277 | case PR_GET_TIMING: | |
2278 | error = PR_TIMING_STATISTICAL; | |
2279 | break; | |
2280 | case PR_SET_TIMING: | |
2281 | if (arg2 != PR_TIMING_STATISTICAL) | |
2282 | error = -EINVAL; | |
2283 | break; | |
2284 | case PR_SET_NAME: | |
2285 | comm[sizeof(me->comm) - 1] = 0; | |
2286 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2287 | sizeof(me->comm) - 1) < 0) | |
2288 | return -EFAULT; | |
2289 | set_task_comm(me, comm); | |
2290 | proc_comm_connector(me); | |
2291 | break; | |
2292 | case PR_GET_NAME: | |
2293 | get_task_comm(comm, me); | |
2294 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2295 | return -EFAULT; | |
2296 | break; | |
2297 | case PR_GET_ENDIAN: | |
2298 | error = GET_ENDIAN(me, arg2); | |
2299 | break; | |
2300 | case PR_SET_ENDIAN: | |
2301 | error = SET_ENDIAN(me, arg2); | |
2302 | break; | |
2303 | case PR_GET_SECCOMP: | |
2304 | error = prctl_get_seccomp(); | |
2305 | break; | |
2306 | case PR_SET_SECCOMP: | |
2307 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2308 | break; | |
2309 | case PR_GET_TSC: | |
2310 | error = GET_TSC_CTL(arg2); | |
2311 | break; | |
2312 | case PR_SET_TSC: | |
2313 | error = SET_TSC_CTL(arg2); | |
2314 | break; | |
2315 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2316 | error = perf_event_task_disable(); | |
2317 | break; | |
2318 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2319 | error = perf_event_task_enable(); | |
2320 | break; | |
2321 | case PR_GET_TIMERSLACK: | |
286d25ba JS |
2322 | if (current->timer_slack_ns > ULONG_MAX) |
2323 | error = ULONG_MAX; | |
2324 | else | |
2325 | error = current->timer_slack_ns; | |
f3cbd435 AM |
2326 | break; |
2327 | case PR_SET_TIMERSLACK: | |
2328 | if (arg2 <= 0) | |
2329 | current->timer_slack_ns = | |
6976675d | 2330 | current->default_timer_slack_ns; |
f3cbd435 AM |
2331 | else |
2332 | current->timer_slack_ns = arg2; | |
2333 | break; | |
2334 | case PR_MCE_KILL: | |
2335 | if (arg4 | arg5) | |
2336 | return -EINVAL; | |
2337 | switch (arg2) { | |
2338 | case PR_MCE_KILL_CLEAR: | |
2339 | if (arg3 != 0) | |
4db96cf0 | 2340 | return -EINVAL; |
f3cbd435 | 2341 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2342 | break; |
f3cbd435 AM |
2343 | case PR_MCE_KILL_SET: |
2344 | current->flags |= PF_MCE_PROCESS; | |
2345 | if (arg3 == PR_MCE_KILL_EARLY) | |
2346 | current->flags |= PF_MCE_EARLY; | |
2347 | else if (arg3 == PR_MCE_KILL_LATE) | |
2348 | current->flags &= ~PF_MCE_EARLY; | |
2349 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2350 | current->flags &= | |
2351 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2352 | else |
259e5e6c | 2353 | return -EINVAL; |
259e5e6c | 2354 | break; |
1da177e4 | 2355 | default: |
f3cbd435 AM |
2356 | return -EINVAL; |
2357 | } | |
2358 | break; | |
2359 | case PR_MCE_KILL_GET: | |
2360 | if (arg2 | arg3 | arg4 | arg5) | |
2361 | return -EINVAL; | |
2362 | if (current->flags & PF_MCE_PROCESS) | |
2363 | error = (current->flags & PF_MCE_EARLY) ? | |
2364 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2365 | else | |
2366 | error = PR_MCE_KILL_DEFAULT; | |
2367 | break; | |
2368 | case PR_SET_MM: | |
2369 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2370 | break; | |
2371 | case PR_GET_TID_ADDRESS: | |
2372 | error = prctl_get_tid_address(me, (int __user **)arg2); | |
2373 | break; | |
18f42f60 | 2374 | case PR_SET_TIMERSLACK_PID: |
d4d049c5 | 2375 | if (task_pid_vnr(current) != (pid_t)arg3 && |
18f42f60 MK |
2376 | !capable(CAP_SYS_NICE)) |
2377 | return -EPERM; | |
2378 | rcu_read_lock(); | |
d4d049c5 | 2379 | tsk = find_task_by_vpid((pid_t)arg3); |
18f42f60 MK |
2380 | if (tsk == NULL) { |
2381 | rcu_read_unlock(); | |
2382 | return -EINVAL; | |
2383 | } | |
2384 | get_task_struct(tsk); | |
2385 | rcu_read_unlock(); | |
2386 | if (arg2 <= 0) | |
2387 | tsk->timer_slack_ns = | |
2388 | tsk->default_timer_slack_ns; | |
2389 | else | |
2390 | tsk->timer_slack_ns = arg2; | |
2391 | put_task_struct(tsk); | |
2392 | error = 0; | |
2393 | break; | |
f3cbd435 AM |
2394 | case PR_SET_CHILD_SUBREAPER: |
2395 | me->signal->is_child_subreaper = !!arg2; | |
2396 | break; | |
2397 | case PR_GET_CHILD_SUBREAPER: | |
2398 | error = put_user(me->signal->is_child_subreaper, | |
2399 | (int __user *)arg2); | |
2400 | break; | |
2401 | case PR_SET_NO_NEW_PRIVS: | |
2402 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2403 | return -EINVAL; | |
2404 | ||
1d4457f9 | 2405 | task_set_no_new_privs(current); |
f3cbd435 AM |
2406 | break; |
2407 | case PR_GET_NO_NEW_PRIVS: | |
2408 | if (arg2 || arg3 || arg4 || arg5) | |
2409 | return -EINVAL; | |
1d4457f9 | 2410 | return task_no_new_privs(current) ? 1 : 0; |
a0715cc2 AT |
2411 | case PR_GET_THP_DISABLE: |
2412 | if (arg2 || arg3 || arg4 || arg5) | |
2413 | return -EINVAL; | |
2414 | error = !!(me->mm->def_flags & VM_NOHUGEPAGE); | |
2415 | break; | |
2416 | case PR_SET_THP_DISABLE: | |
2417 | if (arg3 || arg4 || arg5) | |
2418 | return -EINVAL; | |
2419 | down_write(&me->mm->mmap_sem); | |
2420 | if (arg2) | |
2421 | me->mm->def_flags |= VM_NOHUGEPAGE; | |
2422 | else | |
2423 | me->mm->def_flags &= ~VM_NOHUGEPAGE; | |
2424 | up_write(&me->mm->mmap_sem); | |
2425 | break; | |
fe3d197f | 2426 | case PR_MPX_ENABLE_MANAGEMENT: |
e9d1b4f3 DH |
2427 | if (arg2 || arg3 || arg4 || arg5) |
2428 | return -EINVAL; | |
46a6e0cf | 2429 | error = MPX_ENABLE_MANAGEMENT(); |
fe3d197f DH |
2430 | break; |
2431 | case PR_MPX_DISABLE_MANAGEMENT: | |
e9d1b4f3 DH |
2432 | if (arg2 || arg3 || arg4 || arg5) |
2433 | return -EINVAL; | |
46a6e0cf | 2434 | error = MPX_DISABLE_MANAGEMENT(); |
fe3d197f | 2435 | break; |
9791554b PB |
2436 | case PR_SET_FP_MODE: |
2437 | error = SET_FP_MODE(me, arg2); | |
2438 | break; | |
2439 | case PR_GET_FP_MODE: | |
2440 | error = GET_FP_MODE(me); | |
2441 | break; | |
586278d7 CC |
2442 | case PR_SET_VMA: |
2443 | error = prctl_set_vma(arg2, arg3, arg4, arg5); | |
2444 | break; | |
f3cbd435 AM |
2445 | default: |
2446 | error = -EINVAL; | |
2447 | break; | |
1da177e4 LT |
2448 | } |
2449 | return error; | |
2450 | } | |
3cfc348b | 2451 | |
836f92ad HC |
2452 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2453 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2454 | { |
2455 | int err = 0; | |
2456 | int cpu = raw_smp_processor_id(); | |
ec94fc3d | 2457 | |
3cfc348b AK |
2458 | if (cpup) |
2459 | err |= put_user(cpu, cpup); | |
2460 | if (nodep) | |
2461 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2462 | return err ? -EFAULT : 0; |
2463 | } | |
10a0a8d4 | 2464 | |
4a22f166 SR |
2465 | /** |
2466 | * do_sysinfo - fill in sysinfo struct | |
2467 | * @info: pointer to buffer to fill | |
2468 | */ | |
2469 | static int do_sysinfo(struct sysinfo *info) | |
2470 | { | |
2471 | unsigned long mem_total, sav_total; | |
2472 | unsigned int mem_unit, bitcount; | |
2473 | struct timespec tp; | |
2474 | ||
2475 | memset(info, 0, sizeof(struct sysinfo)); | |
2476 | ||
45c64940 | 2477 | get_monotonic_boottime(&tp); |
4a22f166 SR |
2478 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); |
2479 | ||
2480 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2481 | ||
2482 | info->procs = nr_threads; | |
2483 | ||
2484 | si_meminfo(info); | |
2485 | si_swapinfo(info); | |
2486 | ||
2487 | /* | |
2488 | * If the sum of all the available memory (i.e. ram + swap) | |
2489 | * is less than can be stored in a 32 bit unsigned long then | |
2490 | * we can be binary compatible with 2.2.x kernels. If not, | |
2491 | * well, in that case 2.2.x was broken anyways... | |
2492 | * | |
2493 | * -Erik Andersen <andersee@debian.org> | |
2494 | */ | |
2495 | ||
2496 | mem_total = info->totalram + info->totalswap; | |
2497 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2498 | goto out; | |
2499 | bitcount = 0; | |
2500 | mem_unit = info->mem_unit; | |
2501 | while (mem_unit > 1) { | |
2502 | bitcount++; | |
2503 | mem_unit >>= 1; | |
2504 | sav_total = mem_total; | |
2505 | mem_total <<= 1; | |
2506 | if (mem_total < sav_total) | |
2507 | goto out; | |
2508 | } | |
2509 | ||
2510 | /* | |
2511 | * If mem_total did not overflow, multiply all memory values by | |
2512 | * info->mem_unit and set it to 1. This leaves things compatible | |
2513 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2514 | * kernels... | |
2515 | */ | |
2516 | ||
2517 | info->mem_unit = 1; | |
2518 | info->totalram <<= bitcount; | |
2519 | info->freeram <<= bitcount; | |
2520 | info->sharedram <<= bitcount; | |
2521 | info->bufferram <<= bitcount; | |
2522 | info->totalswap <<= bitcount; | |
2523 | info->freeswap <<= bitcount; | |
2524 | info->totalhigh <<= bitcount; | |
2525 | info->freehigh <<= bitcount; | |
2526 | ||
2527 | out: | |
2528 | return 0; | |
2529 | } | |
2530 | ||
2531 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2532 | { | |
2533 | struct sysinfo val; | |
2534 | ||
2535 | do_sysinfo(&val); | |
2536 | ||
2537 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2538 | return -EFAULT; | |
2539 | ||
2540 | return 0; | |
2541 | } | |
2542 | ||
2543 | #ifdef CONFIG_COMPAT | |
2544 | struct compat_sysinfo { | |
2545 | s32 uptime; | |
2546 | u32 loads[3]; | |
2547 | u32 totalram; | |
2548 | u32 freeram; | |
2549 | u32 sharedram; | |
2550 | u32 bufferram; | |
2551 | u32 totalswap; | |
2552 | u32 freeswap; | |
2553 | u16 procs; | |
2554 | u16 pad; | |
2555 | u32 totalhigh; | |
2556 | u32 freehigh; | |
2557 | u32 mem_unit; | |
2558 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2559 | }; | |
2560 | ||
2561 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2562 | { | |
2563 | struct sysinfo s; | |
2564 | ||
2565 | do_sysinfo(&s); | |
2566 | ||
2567 | /* Check to see if any memory value is too large for 32-bit and scale | |
2568 | * down if needed | |
2569 | */ | |
0baae41e | 2570 | if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) { |
4a22f166 SR |
2571 | int bitcount = 0; |
2572 | ||
2573 | while (s.mem_unit < PAGE_SIZE) { | |
2574 | s.mem_unit <<= 1; | |
2575 | bitcount++; | |
2576 | } | |
2577 | ||
2578 | s.totalram >>= bitcount; | |
2579 | s.freeram >>= bitcount; | |
2580 | s.sharedram >>= bitcount; | |
2581 | s.bufferram >>= bitcount; | |
2582 | s.totalswap >>= bitcount; | |
2583 | s.freeswap >>= bitcount; | |
2584 | s.totalhigh >>= bitcount; | |
2585 | s.freehigh >>= bitcount; | |
2586 | } | |
2587 | ||
2588 | if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) || | |
2589 | __put_user(s.uptime, &info->uptime) || | |
2590 | __put_user(s.loads[0], &info->loads[0]) || | |
2591 | __put_user(s.loads[1], &info->loads[1]) || | |
2592 | __put_user(s.loads[2], &info->loads[2]) || | |
2593 | __put_user(s.totalram, &info->totalram) || | |
2594 | __put_user(s.freeram, &info->freeram) || | |
2595 | __put_user(s.sharedram, &info->sharedram) || | |
2596 | __put_user(s.bufferram, &info->bufferram) || | |
2597 | __put_user(s.totalswap, &info->totalswap) || | |
2598 | __put_user(s.freeswap, &info->freeswap) || | |
2599 | __put_user(s.procs, &info->procs) || | |
2600 | __put_user(s.totalhigh, &info->totalhigh) || | |
2601 | __put_user(s.freehigh, &info->freehigh) || | |
2602 | __put_user(s.mem_unit, &info->mem_unit)) | |
2603 | return -EFAULT; | |
2604 | ||
2605 | return 0; | |
2606 | } | |
2607 | #endif /* CONFIG_COMPAT */ |