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 EB |
18 | #include <linux/kernel.h> |
19 | #include <linux/kexec.h> | |
1da177e4 | 20 | #include <linux/workqueue.h> |
c59ede7b | 21 | #include <linux/capability.h> |
1da177e4 LT |
22 | #include <linux/device.h> |
23 | #include <linux/key.h> | |
24 | #include <linux/times.h> | |
25 | #include <linux/posix-timers.h> | |
26 | #include <linux/security.h> | |
27 | #include <linux/dcookies.h> | |
28 | #include <linux/suspend.h> | |
29 | #include <linux/tty.h> | |
7ed20e1a | 30 | #include <linux/signal.h> |
9f46080c | 31 | #include <linux/cn_proc.h> |
3cfc348b | 32 | #include <linux/getcpu.h> |
6eaeeaba | 33 | #include <linux/task_io_accounting_ops.h> |
1d9d02fe | 34 | #include <linux/seccomp.h> |
4047727e | 35 | #include <linux/cpu.h> |
e28cbf22 | 36 | #include <linux/personality.h> |
e3d5a27d | 37 | #include <linux/ptrace.h> |
5ad4e53b | 38 | #include <linux/fs_struct.h> |
b32dfe37 CG |
39 | #include <linux/file.h> |
40 | #include <linux/mount.h> | |
5a0e3ad6 | 41 | #include <linux/gfp.h> |
40dc166c | 42 | #include <linux/syscore_ops.h> |
be27425d AK |
43 | #include <linux/version.h> |
44 | #include <linux/ctype.h> | |
3c2a0909 S |
45 | #include <linux/mm.h> |
46 | #include <linux/mempolicy.h> | |
47 | #include <linux/sched.h> | |
1da177e4 LT |
48 | |
49 | #include <linux/compat.h> | |
50 | #include <linux/syscalls.h> | |
00d7c05a | 51 | #include <linux/kprobes.h> |
acce292c | 52 | #include <linux/user_namespace.h> |
7fe5e042 | 53 | #include <linux/binfmts.h> |
1da177e4 | 54 | |
4a22f166 SR |
55 | #include <linux/sched.h> |
56 | #include <linux/rcupdate.h> | |
57 | #include <linux/uidgid.h> | |
58 | #include <linux/cred.h> | |
59 | ||
04c6862c | 60 | #include <linux/kmsg_dump.h> |
be27425d AK |
61 | /* Move somewhere else to avoid recompiling? */ |
62 | #include <generated/utsrelease.h> | |
04c6862c | 63 | |
1da177e4 LT |
64 | #include <asm/uaccess.h> |
65 | #include <asm/io.h> | |
66 | #include <asm/unistd.h> | |
67 | ||
68 | #ifndef SET_UNALIGN_CTL | |
69 | # define SET_UNALIGN_CTL(a,b) (-EINVAL) | |
70 | #endif | |
71 | #ifndef GET_UNALIGN_CTL | |
72 | # define GET_UNALIGN_CTL(a,b) (-EINVAL) | |
73 | #endif | |
74 | #ifndef SET_FPEMU_CTL | |
75 | # define SET_FPEMU_CTL(a,b) (-EINVAL) | |
76 | #endif | |
77 | #ifndef GET_FPEMU_CTL | |
78 | # define GET_FPEMU_CTL(a,b) (-EINVAL) | |
79 | #endif | |
80 | #ifndef SET_FPEXC_CTL | |
81 | # define SET_FPEXC_CTL(a,b) (-EINVAL) | |
82 | #endif | |
83 | #ifndef GET_FPEXC_CTL | |
84 | # define GET_FPEXC_CTL(a,b) (-EINVAL) | |
85 | #endif | |
651d765d AB |
86 | #ifndef GET_ENDIAN |
87 | # define GET_ENDIAN(a,b) (-EINVAL) | |
88 | #endif | |
89 | #ifndef SET_ENDIAN | |
90 | # define SET_ENDIAN(a,b) (-EINVAL) | |
91 | #endif | |
8fb402bc EB |
92 | #ifndef GET_TSC_CTL |
93 | # define GET_TSC_CTL(a) (-EINVAL) | |
94 | #endif | |
95 | #ifndef SET_TSC_CTL | |
96 | # define SET_TSC_CTL(a) (-EINVAL) | |
97 | #endif | |
1da177e4 LT |
98 | |
99 | /* | |
100 | * this is where the system-wide overflow UID and GID are defined, for | |
101 | * architectures that now have 32-bit UID/GID but didn't in the past | |
102 | */ | |
103 | ||
104 | int overflowuid = DEFAULT_OVERFLOWUID; | |
105 | int overflowgid = DEFAULT_OVERFLOWGID; | |
106 | ||
1da177e4 LT |
107 | EXPORT_SYMBOL(overflowuid); |
108 | EXPORT_SYMBOL(overflowgid); | |
1da177e4 LT |
109 | |
110 | /* | |
111 | * the same as above, but for filesystems which can only store a 16-bit | |
112 | * UID and GID. as such, this is needed on all architectures | |
113 | */ | |
114 | ||
115 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
116 | int fs_overflowgid = DEFAULT_FS_OVERFLOWUID; | |
117 | ||
118 | EXPORT_SYMBOL(fs_overflowuid); | |
119 | EXPORT_SYMBOL(fs_overflowgid); | |
120 | ||
121 | /* | |
122 | * this indicates whether you can reboot with ctrl-alt-del: the default is yes | |
123 | */ | |
124 | ||
125 | int C_A_D = 1; | |
9ec52099 CLG |
126 | struct pid *cad_pid; |
127 | EXPORT_SYMBOL(cad_pid); | |
1da177e4 | 128 | |
3c2a0909 S |
129 | int ignore_fs_panic = 0; // To prevent kernel panic by EIO during shutdown |
130 | ||
131 | #if defined CONFIG_SEC_RESTRICT_SETUID | |
132 | int sec_check_execpath(struct mm_struct *mm, char *denypath); | |
133 | #if defined CONFIG_SEC_RESTRICT_ROOTING_LOG | |
134 | #define PRINT_LOG(...) printk(KERN_ERR __VA_ARGS__) | |
135 | #else | |
136 | #define PRINT_LOG(...) | |
137 | #endif // End of CONFIG_SEC_RESTRICT_ROOTING_LOG | |
138 | ||
139 | static int sec_restrict_uid(void) | |
140 | { | |
141 | int ret = 0; | |
142 | struct task_struct *parent_tsk; | |
143 | const struct cred *parent_cred; | |
144 | ||
145 | read_lock(&tasklist_lock); | |
146 | parent_tsk = current->parent; | |
147 | if (!parent_tsk) { | |
148 | read_unlock(&tasklist_lock); | |
149 | return 0; | |
150 | } | |
151 | ||
152 | get_task_struct(parent_tsk); | |
153 | /* holding on to the task struct is enough so just release | |
154 | * the tasklist lock here */ | |
155 | read_unlock(&tasklist_lock); | |
156 | ||
157 | parent_cred = get_task_cred(parent_tsk); | |
158 | if (!parent_cred) | |
159 | goto out; | |
160 | if (parent_cred->euid == 0 || parent_tsk->pid == 1) { | |
161 | ret = 0; | |
162 | } else if (sec_check_execpath(current->mm, "/system/bin/pppd")) { | |
163 | PRINT_LOG("VPN allowed to use root permission"); | |
164 | ret = 0; | |
165 | } else { | |
166 | PRINT_LOG("Restricted changing UID. PID = %d(%s) PPID = %d(%s)\n", | |
167 | current->pid, current->comm, | |
168 | parent_tsk->pid, parent_tsk->comm); | |
169 | ret = 1; | |
170 | } | |
171 | put_cred(parent_cred); | |
172 | out: | |
173 | put_task_struct(parent_tsk); | |
174 | ||
175 | return ret; | |
176 | } | |
177 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
178 | ||
bd804eba RW |
179 | /* |
180 | * If set, this is used for preparing the system to power off. | |
181 | */ | |
182 | ||
183 | void (*pm_power_off_prepare)(void); | |
bd804eba | 184 | |
fc832ad3 SH |
185 | /* |
186 | * Returns true if current's euid is same as p's uid or euid, | |
187 | * or has CAP_SYS_NICE to p's user_ns. | |
188 | * | |
189 | * Called with rcu_read_lock, creds are safe | |
190 | */ | |
191 | static bool set_one_prio_perm(struct task_struct *p) | |
192 | { | |
193 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
194 | ||
5af66203 EB |
195 | if (uid_eq(pcred->uid, cred->euid) || |
196 | uid_eq(pcred->euid, cred->euid)) | |
fc832ad3 | 197 | return true; |
c4a4d603 | 198 | if (ns_capable(pcred->user_ns, CAP_SYS_NICE)) |
fc832ad3 SH |
199 | return true; |
200 | return false; | |
201 | } | |
202 | ||
c69e8d9c DH |
203 | /* |
204 | * set the priority of a task | |
205 | * - the caller must hold the RCU read lock | |
206 | */ | |
1da177e4 LT |
207 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
208 | { | |
209 | int no_nice; | |
210 | ||
fc832ad3 | 211 | if (!set_one_prio_perm(p)) { |
1da177e4 LT |
212 | error = -EPERM; |
213 | goto out; | |
214 | } | |
e43379f1 | 215 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
1da177e4 LT |
216 | error = -EACCES; |
217 | goto out; | |
218 | } | |
219 | no_nice = security_task_setnice(p, niceval); | |
220 | if (no_nice) { | |
221 | error = no_nice; | |
222 | goto out; | |
223 | } | |
224 | if (error == -ESRCH) | |
225 | error = 0; | |
226 | set_user_nice(p, niceval); | |
227 | out: | |
228 | return error; | |
229 | } | |
230 | ||
754fe8d2 | 231 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) |
1da177e4 LT |
232 | { |
233 | struct task_struct *g, *p; | |
234 | struct user_struct *user; | |
86a264ab | 235 | const struct cred *cred = current_cred(); |
1da177e4 | 236 | int error = -EINVAL; |
41487c65 | 237 | struct pid *pgrp; |
7b44ab97 | 238 | kuid_t uid; |
1da177e4 | 239 | |
3e88c553 | 240 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
241 | goto out; |
242 | ||
243 | /* normalize: avoid signed division (rounding problems) */ | |
244 | error = -ESRCH; | |
245 | if (niceval < -20) | |
246 | niceval = -20; | |
247 | if (niceval > 19) | |
248 | niceval = 19; | |
249 | ||
d4581a23 | 250 | rcu_read_lock(); |
1da177e4 LT |
251 | read_lock(&tasklist_lock); |
252 | switch (which) { | |
253 | case PRIO_PROCESS: | |
41487c65 | 254 | if (who) |
228ebcbe | 255 | p = find_task_by_vpid(who); |
41487c65 EB |
256 | else |
257 | p = current; | |
1da177e4 LT |
258 | if (p) |
259 | error = set_one_prio(p, niceval, error); | |
260 | break; | |
261 | case PRIO_PGRP: | |
41487c65 | 262 | if (who) |
b488893a | 263 | pgrp = find_vpid(who); |
41487c65 EB |
264 | else |
265 | pgrp = task_pgrp(current); | |
2d70b68d | 266 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { |
1da177e4 | 267 | error = set_one_prio(p, niceval, error); |
2d70b68d | 268 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
269 | break; |
270 | case PRIO_USER: | |
7b44ab97 | 271 | uid = make_kuid(cred->user_ns, who); |
74ba508f | 272 | user = cred->user; |
1da177e4 | 273 | if (!who) |
078de5f7 EB |
274 | uid = cred->uid; |
275 | else if (!uid_eq(uid, cred->uid) && | |
7b44ab97 | 276 | !(user = find_user(uid))) |
86a264ab | 277 | goto out_unlock; /* No processes for this user */ |
1da177e4 | 278 | |
dfc6a736 | 279 | do_each_thread(g, p) { |
078de5f7 | 280 | if (uid_eq(task_uid(p), uid)) |
1da177e4 | 281 | error = set_one_prio(p, niceval, error); |
dfc6a736 | 282 | } while_each_thread(g, p); |
078de5f7 | 283 | if (!uid_eq(uid, cred->uid)) |
1da177e4 LT |
284 | free_uid(user); /* For find_user() */ |
285 | break; | |
286 | } | |
287 | out_unlock: | |
288 | read_unlock(&tasklist_lock); | |
d4581a23 | 289 | rcu_read_unlock(); |
1da177e4 LT |
290 | out: |
291 | return error; | |
292 | } | |
293 | ||
294 | /* | |
295 | * Ugh. To avoid negative return values, "getpriority()" will | |
296 | * not return the normal nice-value, but a negated value that | |
297 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
298 | * to stay compatible. | |
299 | */ | |
754fe8d2 | 300 | SYSCALL_DEFINE2(getpriority, int, which, int, who) |
1da177e4 LT |
301 | { |
302 | struct task_struct *g, *p; | |
303 | struct user_struct *user; | |
86a264ab | 304 | const struct cred *cred = current_cred(); |
1da177e4 | 305 | long niceval, retval = -ESRCH; |
41487c65 | 306 | struct pid *pgrp; |
7b44ab97 | 307 | kuid_t uid; |
1da177e4 | 308 | |
3e88c553 | 309 | if (which > PRIO_USER || which < PRIO_PROCESS) |
1da177e4 LT |
310 | return -EINVAL; |
311 | ||
70118837 | 312 | rcu_read_lock(); |
1da177e4 LT |
313 | read_lock(&tasklist_lock); |
314 | switch (which) { | |
315 | case PRIO_PROCESS: | |
41487c65 | 316 | if (who) |
228ebcbe | 317 | p = find_task_by_vpid(who); |
41487c65 EB |
318 | else |
319 | p = current; | |
1da177e4 LT |
320 | if (p) { |
321 | niceval = 20 - task_nice(p); | |
322 | if (niceval > retval) | |
323 | retval = niceval; | |
324 | } | |
325 | break; | |
326 | case PRIO_PGRP: | |
41487c65 | 327 | if (who) |
b488893a | 328 | pgrp = find_vpid(who); |
41487c65 EB |
329 | else |
330 | pgrp = task_pgrp(current); | |
2d70b68d | 331 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { |
1da177e4 LT |
332 | niceval = 20 - task_nice(p); |
333 | if (niceval > retval) | |
334 | retval = niceval; | |
2d70b68d | 335 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); |
1da177e4 LT |
336 | break; |
337 | case PRIO_USER: | |
7b44ab97 | 338 | uid = make_kuid(cred->user_ns, who); |
74ba508f | 339 | user = cred->user; |
1da177e4 | 340 | if (!who) |
078de5f7 EB |
341 | uid = cred->uid; |
342 | else if (!uid_eq(uid, cred->uid) && | |
7b44ab97 | 343 | !(user = find_user(uid))) |
86a264ab | 344 | goto out_unlock; /* No processes for this user */ |
1da177e4 | 345 | |
dfc6a736 | 346 | do_each_thread(g, p) { |
078de5f7 | 347 | if (uid_eq(task_uid(p), uid)) { |
1da177e4 LT |
348 | niceval = 20 - task_nice(p); |
349 | if (niceval > retval) | |
350 | retval = niceval; | |
351 | } | |
dfc6a736 | 352 | } while_each_thread(g, p); |
078de5f7 | 353 | if (!uid_eq(uid, cred->uid)) |
1da177e4 LT |
354 | free_uid(user); /* for find_user() */ |
355 | break; | |
356 | } | |
357 | out_unlock: | |
358 | read_unlock(&tasklist_lock); | |
70118837 | 359 | rcu_read_unlock(); |
1da177e4 LT |
360 | |
361 | return retval; | |
362 | } | |
363 | ||
e4c94330 EB |
364 | /** |
365 | * emergency_restart - reboot the system | |
366 | * | |
367 | * Without shutting down any hardware or taking any locks | |
368 | * reboot the system. This is called when we know we are in | |
369 | * trouble so this is our best effort to reboot. This is | |
370 | * safe to call in interrupt context. | |
371 | */ | |
7c903473 EB |
372 | void emergency_restart(void) |
373 | { | |
04c6862c | 374 | kmsg_dump(KMSG_DUMP_EMERG); |
7c903473 EB |
375 | machine_emergency_restart(); |
376 | } | |
377 | EXPORT_SYMBOL_GPL(emergency_restart); | |
378 | ||
ca195b7f | 379 | void kernel_restart_prepare(char *cmd) |
4a00ea1e | 380 | { |
e041c683 | 381 | blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd); |
4a00ea1e | 382 | system_state = SYSTEM_RESTART; |
3c2a0909 S |
383 | |
384 | /* user process freeze before device shutdown */ | |
385 | freeze_processes(); | |
b50fa7c8 | 386 | usermodehelper_disable(); |
3c2a0909 | 387 | ignore_fs_panic = 1; |
4a00ea1e | 388 | device_shutdown(); |
e4c94330 | 389 | } |
1e5d5331 | 390 | |
c5f41752 AW |
391 | /** |
392 | * register_reboot_notifier - Register function to be called at reboot time | |
393 | * @nb: Info about notifier function to be called | |
394 | * | |
395 | * Registers a function with the list of functions | |
396 | * to be called at reboot time. | |
397 | * | |
398 | * Currently always returns zero, as blocking_notifier_chain_register() | |
399 | * always returns zero. | |
400 | */ | |
401 | int register_reboot_notifier(struct notifier_block *nb) | |
402 | { | |
403 | return blocking_notifier_chain_register(&reboot_notifier_list, nb); | |
404 | } | |
405 | EXPORT_SYMBOL(register_reboot_notifier); | |
406 | ||
407 | /** | |
408 | * unregister_reboot_notifier - Unregister previously registered reboot notifier | |
409 | * @nb: Hook to be unregistered | |
410 | * | |
411 | * Unregisters a previously registered reboot | |
412 | * notifier function. | |
413 | * | |
414 | * Returns zero on success, or %-ENOENT on failure. | |
415 | */ | |
416 | int unregister_reboot_notifier(struct notifier_block *nb) | |
417 | { | |
418 | return blocking_notifier_chain_unregister(&reboot_notifier_list, nb); | |
419 | } | |
420 | EXPORT_SYMBOL(unregister_reboot_notifier); | |
421 | ||
cf7df378 RH |
422 | /* Add backwards compatibility for stable trees. */ |
423 | #ifndef PF_NO_SETAFFINITY | |
424 | #define PF_NO_SETAFFINITY PF_THREAD_BOUND | |
425 | #endif | |
426 | ||
427 | static void migrate_to_reboot_cpu(void) | |
428 | { | |
429 | /* The boot cpu is always logical cpu 0 */ | |
430 | int cpu = 0; | |
431 | ||
432 | cpu_hotplug_disable(); | |
433 | ||
434 | /* Make certain the cpu I'm about to reboot on is online */ | |
435 | if (!cpu_online(cpu)) | |
436 | cpu = cpumask_first(cpu_online_mask); | |
437 | ||
438 | /* Prevent races with other tasks migrating this task */ | |
439 | current->flags |= PF_NO_SETAFFINITY; | |
440 | ||
441 | /* Make certain I only run on the appropriate processor */ | |
442 | set_cpus_allowed_ptr(current, cpumask_of(cpu)); | |
443 | } | |
444 | ||
1e5d5331 RD |
445 | /** |
446 | * kernel_restart - reboot the system | |
447 | * @cmd: pointer to buffer containing command to execute for restart | |
b8887e6e | 448 | * or %NULL |
1e5d5331 RD |
449 | * |
450 | * Shutdown everything and perform a clean reboot. | |
451 | * This is not safe to call in interrupt context. | |
452 | */ | |
e4c94330 EB |
453 | void kernel_restart(char *cmd) |
454 | { | |
455 | kernel_restart_prepare(cmd); | |
cf7df378 | 456 | migrate_to_reboot_cpu(); |
6f389a8f | 457 | syscore_shutdown(); |
756184b7 | 458 | if (!cmd) |
4a00ea1e | 459 | printk(KERN_EMERG "Restarting system.\n"); |
756184b7 | 460 | else |
4a00ea1e | 461 | printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd); |
04c6862c | 462 | kmsg_dump(KMSG_DUMP_RESTART); |
4a00ea1e EB |
463 | machine_restart(cmd); |
464 | } | |
465 | EXPORT_SYMBOL_GPL(kernel_restart); | |
466 | ||
4ef7229f | 467 | static void kernel_shutdown_prepare(enum system_states state) |
729b4d4c | 468 | { |
e041c683 | 469 | blocking_notifier_call_chain(&reboot_notifier_list, |
729b4d4c AS |
470 | (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL); |
471 | system_state = state; | |
3c2a0909 S |
472 | |
473 | /* user process freeze before device shutdown */ | |
474 | freeze_processes(); | |
b50fa7c8 | 475 | usermodehelper_disable(); |
3c2a0909 | 476 | ignore_fs_panic = 1; |
729b4d4c AS |
477 | device_shutdown(); |
478 | } | |
e4c94330 EB |
479 | /** |
480 | * kernel_halt - halt the system | |
481 | * | |
482 | * Shutdown everything and perform a clean system halt. | |
483 | */ | |
e4c94330 EB |
484 | void kernel_halt(void) |
485 | { | |
729b4d4c | 486 | kernel_shutdown_prepare(SYSTEM_HALT); |
cf7df378 | 487 | migrate_to_reboot_cpu(); |
40dc166c | 488 | syscore_shutdown(); |
4a00ea1e | 489 | printk(KERN_EMERG "System halted.\n"); |
04c6862c | 490 | kmsg_dump(KMSG_DUMP_HALT); |
4a00ea1e EB |
491 | machine_halt(); |
492 | } | |
729b4d4c | 493 | |
4a00ea1e EB |
494 | EXPORT_SYMBOL_GPL(kernel_halt); |
495 | ||
e4c94330 EB |
496 | /** |
497 | * kernel_power_off - power_off the system | |
498 | * | |
499 | * Shutdown everything and perform a clean system power_off. | |
500 | */ | |
e4c94330 EB |
501 | void kernel_power_off(void) |
502 | { | |
729b4d4c | 503 | kernel_shutdown_prepare(SYSTEM_POWER_OFF); |
bd804eba RW |
504 | if (pm_power_off_prepare) |
505 | pm_power_off_prepare(); | |
cf7df378 | 506 | migrate_to_reboot_cpu(); |
40dc166c | 507 | syscore_shutdown(); |
4a00ea1e | 508 | printk(KERN_EMERG "Power down.\n"); |
04c6862c | 509 | kmsg_dump(KMSG_DUMP_POWEROFF); |
4a00ea1e EB |
510 | machine_power_off(); |
511 | } | |
512 | EXPORT_SYMBOL_GPL(kernel_power_off); | |
6f15fa50 TG |
513 | |
514 | static DEFINE_MUTEX(reboot_mutex); | |
515 | ||
1da177e4 LT |
516 | /* |
517 | * Reboot system call: for obvious reasons only root may call it, | |
518 | * and even root needs to set up some magic numbers in the registers | |
519 | * so that some mistake won't make this reboot the whole machine. | |
520 | * You can also set the meaning of the ctrl-alt-del-key here. | |
521 | * | |
522 | * reboot doesn't sync: do that yourself before calling this. | |
523 | */ | |
754fe8d2 HC |
524 | SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd, |
525 | void __user *, arg) | |
1da177e4 | 526 | { |
923c7538 | 527 | struct pid_namespace *pid_ns = task_active_pid_ns(current); |
1da177e4 | 528 | char buffer[256]; |
3d26dcf7 | 529 | int ret = 0; |
1da177e4 LT |
530 | |
531 | /* We only trust the superuser with rebooting the system. */ | |
923c7538 | 532 | if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT)) |
1da177e4 LT |
533 | return -EPERM; |
534 | ||
535 | /* For safety, we require "magic" arguments. */ | |
536 | if (magic1 != LINUX_REBOOT_MAGIC1 || | |
537 | (magic2 != LINUX_REBOOT_MAGIC2 && | |
538 | magic2 != LINUX_REBOOT_MAGIC2A && | |
539 | magic2 != LINUX_REBOOT_MAGIC2B && | |
540 | magic2 != LINUX_REBOOT_MAGIC2C)) | |
541 | return -EINVAL; | |
542 | ||
cf3f8921 DL |
543 | /* |
544 | * If pid namespaces are enabled and the current task is in a child | |
545 | * pid_namespace, the command is handled by reboot_pid_ns() which will | |
546 | * call do_exit(). | |
547 | */ | |
923c7538 | 548 | ret = reboot_pid_ns(pid_ns, cmd); |
cf3f8921 DL |
549 | if (ret) |
550 | return ret; | |
551 | ||
5e38291d EB |
552 | /* Instead of trying to make the power_off code look like |
553 | * halt when pm_power_off is not set do it the easy way. | |
554 | */ | |
555 | if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off) | |
556 | cmd = LINUX_REBOOT_CMD_HALT; | |
557 | ||
6f15fa50 | 558 | mutex_lock(&reboot_mutex); |
1da177e4 LT |
559 | switch (cmd) { |
560 | case LINUX_REBOOT_CMD_RESTART: | |
4a00ea1e | 561 | kernel_restart(NULL); |
1da177e4 LT |
562 | break; |
563 | ||
564 | case LINUX_REBOOT_CMD_CAD_ON: | |
565 | C_A_D = 1; | |
566 | break; | |
567 | ||
568 | case LINUX_REBOOT_CMD_CAD_OFF: | |
569 | C_A_D = 0; | |
570 | break; | |
571 | ||
572 | case LINUX_REBOOT_CMD_HALT: | |
3c2a0909 | 573 | /* kernel_halt(); |
1da177e4 | 574 | do_exit(0); |
3c2a0909 S |
575 | panic("cannot halt"); */ |
576 | kernel_restart(NULL); | |
577 | break; | |
1da177e4 LT |
578 | |
579 | case LINUX_REBOOT_CMD_POWER_OFF: | |
4a00ea1e | 580 | kernel_power_off(); |
1da177e4 LT |
581 | do_exit(0); |
582 | break; | |
583 | ||
584 | case LINUX_REBOOT_CMD_RESTART2: | |
585 | if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) { | |
6f15fa50 TG |
586 | ret = -EFAULT; |
587 | break; | |
1da177e4 LT |
588 | } |
589 | buffer[sizeof(buffer) - 1] = '\0'; | |
590 | ||
4a00ea1e | 591 | kernel_restart(buffer); |
1da177e4 LT |
592 | break; |
593 | ||
3ab83521 | 594 | #ifdef CONFIG_KEXEC |
dc009d92 | 595 | case LINUX_REBOOT_CMD_KEXEC: |
3d26dcf7 AK |
596 | ret = kernel_kexec(); |
597 | break; | |
3ab83521 | 598 | #endif |
4a00ea1e | 599 | |
b0cb1a19 | 600 | #ifdef CONFIG_HIBERNATION |
1da177e4 | 601 | case LINUX_REBOOT_CMD_SW_SUSPEND: |
3d26dcf7 AK |
602 | ret = hibernate(); |
603 | break; | |
1da177e4 LT |
604 | #endif |
605 | ||
606 | default: | |
3d26dcf7 AK |
607 | ret = -EINVAL; |
608 | break; | |
1da177e4 | 609 | } |
6f15fa50 | 610 | mutex_unlock(&reboot_mutex); |
3d26dcf7 | 611 | return ret; |
1da177e4 LT |
612 | } |
613 | ||
3c2a0909 S |
614 | extern void do_emergency_remount(struct work_struct *work); |
615 | ||
65f27f38 | 616 | static void deferred_cad(struct work_struct *dummy) |
1da177e4 | 617 | { |
3c2a0909 | 618 | do_emergency_remount(NULL); |
abcd9e51 | 619 | kernel_restart(NULL); |
1da177e4 LT |
620 | } |
621 | ||
622 | /* | |
623 | * This function gets called by ctrl-alt-del - ie the keyboard interrupt. | |
624 | * As it's called within an interrupt, it may NOT sync: the only choice | |
625 | * is whether to reboot at once, or just ignore the ctrl-alt-del. | |
626 | */ | |
627 | void ctrl_alt_del(void) | |
628 | { | |
65f27f38 | 629 | static DECLARE_WORK(cad_work, deferred_cad); |
1da177e4 LT |
630 | |
631 | if (C_A_D) | |
632 | schedule_work(&cad_work); | |
633 | else | |
9ec52099 | 634 | kill_cad_pid(SIGINT, 1); |
1da177e4 LT |
635 | } |
636 | ||
1da177e4 LT |
637 | /* |
638 | * Unprivileged users may change the real gid to the effective gid | |
639 | * or vice versa. (BSD-style) | |
640 | * | |
641 | * If you set the real gid at all, or set the effective gid to a value not | |
642 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
643 | * | |
644 | * This makes it possible for a setgid program to completely drop its | |
645 | * privileges, which is often a useful assertion to make when you are doing | |
646 | * a security audit over a program. | |
647 | * | |
648 | * The general idea is that a program which uses just setregid() will be | |
649 | * 100% compatible with BSD. A program which uses just setgid() will be | |
650 | * 100% compatible with POSIX with saved IDs. | |
651 | * | |
652 | * SMP: There are not races, the GIDs are checked only by filesystem | |
653 | * operations (as far as semantic preservation is concerned). | |
654 | */ | |
ae1251ab | 655 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) |
1da177e4 | 656 | { |
a29c33f4 | 657 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
658 | const struct cred *old; |
659 | struct cred *new; | |
1da177e4 | 660 | int retval; |
a29c33f4 EB |
661 | kgid_t krgid, kegid; |
662 | ||
3c2a0909 S |
663 | #if defined CONFIG_SEC_RESTRICT_SETUID |
664 | if(rgid == 0 || egid == 0) | |
665 | { | |
666 | if(sec_restrict_uid()) | |
667 | return -EACCES; | |
668 | } | |
669 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
670 | ||
a29c33f4 EB |
671 | krgid = make_kgid(ns, rgid); |
672 | kegid = make_kgid(ns, egid); | |
673 | ||
674 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
675 | return -EINVAL; | |
676 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
677 | return -EINVAL; | |
1da177e4 | 678 | |
d84f4f99 DH |
679 | new = prepare_creds(); |
680 | if (!new) | |
681 | return -ENOMEM; | |
682 | old = current_cred(); | |
683 | ||
d84f4f99 | 684 | retval = -EPERM; |
1da177e4 | 685 | if (rgid != (gid_t) -1) { |
a29c33f4 EB |
686 | if (gid_eq(old->gid, krgid) || |
687 | gid_eq(old->egid, krgid) || | |
fc832ad3 | 688 | nsown_capable(CAP_SETGID)) |
a29c33f4 | 689 | new->gid = krgid; |
1da177e4 | 690 | else |
d84f4f99 | 691 | goto error; |
1da177e4 LT |
692 | } |
693 | if (egid != (gid_t) -1) { | |
a29c33f4 EB |
694 | if (gid_eq(old->gid, kegid) || |
695 | gid_eq(old->egid, kegid) || | |
696 | gid_eq(old->sgid, kegid) || | |
fc832ad3 | 697 | nsown_capable(CAP_SETGID)) |
a29c33f4 | 698 | new->egid = kegid; |
756184b7 | 699 | else |
d84f4f99 | 700 | goto error; |
1da177e4 | 701 | } |
d84f4f99 | 702 | |
1da177e4 | 703 | if (rgid != (gid_t) -1 || |
a29c33f4 | 704 | (egid != (gid_t) -1 && !gid_eq(kegid, old->gid))) |
d84f4f99 DH |
705 | new->sgid = new->egid; |
706 | new->fsgid = new->egid; | |
707 | ||
708 | return commit_creds(new); | |
709 | ||
710 | error: | |
711 | abort_creds(new); | |
712 | return retval; | |
1da177e4 LT |
713 | } |
714 | ||
715 | /* | |
716 | * setgid() is implemented like SysV w/ SAVED_IDS | |
717 | * | |
718 | * SMP: Same implicit races as above. | |
719 | */ | |
ae1251ab | 720 | SYSCALL_DEFINE1(setgid, gid_t, gid) |
1da177e4 | 721 | { |
a29c33f4 | 722 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
723 | const struct cred *old; |
724 | struct cred *new; | |
1da177e4 | 725 | int retval; |
a29c33f4 EB |
726 | kgid_t kgid; |
727 | ||
3c2a0909 S |
728 | #if defined CONFIG_SEC_RESTRICT_SETUID |
729 | if(gid == 0) | |
730 | { | |
731 | if(sec_restrict_uid()) | |
732 | return -EACCES; | |
733 | } | |
734 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
735 | ||
a29c33f4 EB |
736 | kgid = make_kgid(ns, gid); |
737 | if (!gid_valid(kgid)) | |
738 | return -EINVAL; | |
1da177e4 | 739 | |
d84f4f99 DH |
740 | new = prepare_creds(); |
741 | if (!new) | |
742 | return -ENOMEM; | |
743 | old = current_cred(); | |
744 | ||
d84f4f99 | 745 | retval = -EPERM; |
fc832ad3 | 746 | if (nsown_capable(CAP_SETGID)) |
a29c33f4 EB |
747 | new->gid = new->egid = new->sgid = new->fsgid = kgid; |
748 | else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid)) | |
749 | new->egid = new->fsgid = kgid; | |
1da177e4 | 750 | else |
d84f4f99 | 751 | goto error; |
1da177e4 | 752 | |
d84f4f99 DH |
753 | return commit_creds(new); |
754 | ||
755 | error: | |
756 | abort_creds(new); | |
757 | return retval; | |
1da177e4 | 758 | } |
54e99124 | 759 | |
d84f4f99 DH |
760 | /* |
761 | * change the user struct in a credentials set to match the new UID | |
762 | */ | |
763 | static int set_user(struct cred *new) | |
1da177e4 LT |
764 | { |
765 | struct user_struct *new_user; | |
766 | ||
078de5f7 | 767 | new_user = alloc_uid(new->uid); |
1da177e4 LT |
768 | if (!new_user) |
769 | return -EAGAIN; | |
770 | ||
72fa5997 VK |
771 | /* |
772 | * We don't fail in case of NPROC limit excess here because too many | |
773 | * poorly written programs don't check set*uid() return code, assuming | |
774 | * it never fails if called by root. We may still enforce NPROC limit | |
775 | * for programs doing set*uid()+execve() by harmlessly deferring the | |
776 | * failure to the execve() stage. | |
777 | */ | |
78d7d407 | 778 | if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) && |
72fa5997 VK |
779 | new_user != INIT_USER) |
780 | current->flags |= PF_NPROC_EXCEEDED; | |
781 | else | |
782 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 783 | |
d84f4f99 DH |
784 | free_uid(new->user); |
785 | new->user = new_user; | |
1da177e4 LT |
786 | return 0; |
787 | } | |
788 | ||
789 | /* | |
790 | * Unprivileged users may change the real uid to the effective uid | |
791 | * or vice versa. (BSD-style) | |
792 | * | |
793 | * If you set the real uid at all, or set the effective uid to a value not | |
794 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
795 | * | |
796 | * This makes it possible for a setuid program to completely drop its | |
797 | * privileges, which is often a useful assertion to make when you are doing | |
798 | * a security audit over a program. | |
799 | * | |
800 | * The general idea is that a program which uses just setreuid() will be | |
801 | * 100% compatible with BSD. A program which uses just setuid() will be | |
802 | * 100% compatible with POSIX with saved IDs. | |
803 | */ | |
ae1251ab | 804 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) |
1da177e4 | 805 | { |
a29c33f4 | 806 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
807 | const struct cred *old; |
808 | struct cred *new; | |
1da177e4 | 809 | int retval; |
a29c33f4 EB |
810 | kuid_t kruid, keuid; |
811 | ||
3c2a0909 S |
812 | #if defined CONFIG_SEC_RESTRICT_SETUID |
813 | if(ruid == 0 || euid == 0) | |
814 | { | |
815 | if(sec_restrict_uid()) | |
816 | return -EACCES; | |
817 | } | |
818 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
819 | ||
a29c33f4 EB |
820 | kruid = make_kuid(ns, ruid); |
821 | keuid = make_kuid(ns, euid); | |
822 | ||
823 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
824 | return -EINVAL; | |
825 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
826 | return -EINVAL; | |
1da177e4 | 827 | |
d84f4f99 DH |
828 | new = prepare_creds(); |
829 | if (!new) | |
830 | return -ENOMEM; | |
831 | old = current_cred(); | |
832 | ||
d84f4f99 | 833 | retval = -EPERM; |
1da177e4 | 834 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
835 | new->uid = kruid; |
836 | if (!uid_eq(old->uid, kruid) && | |
837 | !uid_eq(old->euid, kruid) && | |
fc832ad3 | 838 | !nsown_capable(CAP_SETUID)) |
d84f4f99 | 839 | goto error; |
1da177e4 LT |
840 | } |
841 | ||
842 | if (euid != (uid_t) -1) { | |
a29c33f4 EB |
843 | new->euid = keuid; |
844 | if (!uid_eq(old->uid, keuid) && | |
845 | !uid_eq(old->euid, keuid) && | |
846 | !uid_eq(old->suid, keuid) && | |
fc832ad3 | 847 | !nsown_capable(CAP_SETUID)) |
d84f4f99 | 848 | goto error; |
1da177e4 LT |
849 | } |
850 | ||
a29c33f4 | 851 | if (!uid_eq(new->uid, old->uid)) { |
54e99124 DG |
852 | retval = set_user(new); |
853 | if (retval < 0) | |
854 | goto error; | |
855 | } | |
1da177e4 | 856 | if (ruid != (uid_t) -1 || |
a29c33f4 | 857 | (euid != (uid_t) -1 && !uid_eq(keuid, old->uid))) |
d84f4f99 DH |
858 | new->suid = new->euid; |
859 | new->fsuid = new->euid; | |
1da177e4 | 860 | |
d84f4f99 DH |
861 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); |
862 | if (retval < 0) | |
863 | goto error; | |
1da177e4 | 864 | |
d84f4f99 | 865 | return commit_creds(new); |
1da177e4 | 866 | |
d84f4f99 DH |
867 | error: |
868 | abort_creds(new); | |
869 | return retval; | |
870 | } | |
1da177e4 LT |
871 | |
872 | /* | |
873 | * setuid() is implemented like SysV with SAVED_IDS | |
874 | * | |
875 | * Note that SAVED_ID's is deficient in that a setuid root program | |
876 | * like sendmail, for example, cannot set its uid to be a normal | |
877 | * user and then switch back, because if you're root, setuid() sets | |
878 | * the saved uid too. If you don't like this, blame the bright people | |
879 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
880 | * will allow a root program to temporarily drop privileges and be able to | |
881 | * regain them by swapping the real and effective uid. | |
882 | */ | |
ae1251ab | 883 | SYSCALL_DEFINE1(setuid, uid_t, uid) |
1da177e4 | 884 | { |
a29c33f4 | 885 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
886 | const struct cred *old; |
887 | struct cred *new; | |
1da177e4 | 888 | int retval; |
a29c33f4 EB |
889 | kuid_t kuid; |
890 | ||
3c2a0909 S |
891 | #if defined CONFIG_SEC_RESTRICT_SETUID |
892 | if(uid == 0) | |
893 | { | |
894 | if(sec_restrict_uid()) | |
895 | return -EACCES; | |
896 | } | |
897 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
898 | ||
a29c33f4 EB |
899 | kuid = make_kuid(ns, uid); |
900 | if (!uid_valid(kuid)) | |
901 | return -EINVAL; | |
1da177e4 | 902 | |
d84f4f99 DH |
903 | new = prepare_creds(); |
904 | if (!new) | |
905 | return -ENOMEM; | |
906 | old = current_cred(); | |
907 | ||
d84f4f99 | 908 | retval = -EPERM; |
fc832ad3 | 909 | if (nsown_capable(CAP_SETUID)) { |
a29c33f4 EB |
910 | new->suid = new->uid = kuid; |
911 | if (!uid_eq(kuid, old->uid)) { | |
54e99124 DG |
912 | retval = set_user(new); |
913 | if (retval < 0) | |
914 | goto error; | |
d84f4f99 | 915 | } |
a29c33f4 | 916 | } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) { |
d84f4f99 | 917 | goto error; |
1da177e4 | 918 | } |
1da177e4 | 919 | |
a29c33f4 | 920 | new->fsuid = new->euid = kuid; |
d84f4f99 DH |
921 | |
922 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
923 | if (retval < 0) | |
924 | goto error; | |
1da177e4 | 925 | |
d84f4f99 | 926 | return commit_creds(new); |
1da177e4 | 927 | |
d84f4f99 DH |
928 | error: |
929 | abort_creds(new); | |
930 | return retval; | |
1da177e4 LT |
931 | } |
932 | ||
933 | ||
934 | /* | |
935 | * This function implements a generic ability to update ruid, euid, | |
936 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
937 | */ | |
ae1251ab | 938 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) |
1da177e4 | 939 | { |
a29c33f4 | 940 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
941 | const struct cred *old; |
942 | struct cred *new; | |
1da177e4 | 943 | int retval; |
a29c33f4 EB |
944 | kuid_t kruid, keuid, ksuid; |
945 | ||
3c2a0909 S |
946 | #if defined CONFIG_SEC_RESTRICT_SETUID |
947 | if(ruid == 0 || euid == 0 || suid == 0) | |
948 | { | |
949 | if(sec_restrict_uid()) | |
950 | return -EACCES; | |
951 | } | |
952 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
953 | ||
a29c33f4 EB |
954 | kruid = make_kuid(ns, ruid); |
955 | keuid = make_kuid(ns, euid); | |
956 | ksuid = make_kuid(ns, suid); | |
957 | ||
958 | if ((ruid != (uid_t) -1) && !uid_valid(kruid)) | |
959 | return -EINVAL; | |
960 | ||
961 | if ((euid != (uid_t) -1) && !uid_valid(keuid)) | |
962 | return -EINVAL; | |
963 | ||
964 | if ((suid != (uid_t) -1) && !uid_valid(ksuid)) | |
965 | return -EINVAL; | |
1da177e4 | 966 | |
d84f4f99 DH |
967 | new = prepare_creds(); |
968 | if (!new) | |
969 | return -ENOMEM; | |
970 | ||
d84f4f99 | 971 | old = current_cred(); |
1da177e4 | 972 | |
d84f4f99 | 973 | retval = -EPERM; |
fc832ad3 | 974 | if (!nsown_capable(CAP_SETUID)) { |
a29c33f4 EB |
975 | if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) && |
976 | !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid)) | |
d84f4f99 | 977 | goto error; |
a29c33f4 EB |
978 | if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) && |
979 | !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid)) | |
d84f4f99 | 980 | goto error; |
a29c33f4 EB |
981 | if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) && |
982 | !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid)) | |
d84f4f99 | 983 | goto error; |
1da177e4 | 984 | } |
d84f4f99 | 985 | |
1da177e4 | 986 | if (ruid != (uid_t) -1) { |
a29c33f4 EB |
987 | new->uid = kruid; |
988 | if (!uid_eq(kruid, old->uid)) { | |
54e99124 DG |
989 | retval = set_user(new); |
990 | if (retval < 0) | |
991 | goto error; | |
992 | } | |
1da177e4 | 993 | } |
d84f4f99 | 994 | if (euid != (uid_t) -1) |
a29c33f4 | 995 | new->euid = keuid; |
1da177e4 | 996 | if (suid != (uid_t) -1) |
a29c33f4 | 997 | new->suid = ksuid; |
d84f4f99 | 998 | new->fsuid = new->euid; |
1da177e4 | 999 | |
d84f4f99 DH |
1000 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); |
1001 | if (retval < 0) | |
1002 | goto error; | |
1da177e4 | 1003 | |
d84f4f99 | 1004 | return commit_creds(new); |
1da177e4 | 1005 | |
d84f4f99 DH |
1006 | error: |
1007 | abort_creds(new); | |
1008 | return retval; | |
1da177e4 LT |
1009 | } |
1010 | ||
a29c33f4 | 1011 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp) |
1da177e4 | 1012 | { |
86a264ab | 1013 | const struct cred *cred = current_cred(); |
1da177e4 | 1014 | int retval; |
a29c33f4 EB |
1015 | uid_t ruid, euid, suid; |
1016 | ||
1017 | ruid = from_kuid_munged(cred->user_ns, cred->uid); | |
1018 | euid = from_kuid_munged(cred->user_ns, cred->euid); | |
1019 | suid = from_kuid_munged(cred->user_ns, cred->suid); | |
1da177e4 | 1020 | |
a29c33f4 EB |
1021 | if (!(retval = put_user(ruid, ruidp)) && |
1022 | !(retval = put_user(euid, euidp))) | |
1023 | retval = put_user(suid, suidp); | |
1da177e4 LT |
1024 | |
1025 | return retval; | |
1026 | } | |
1027 | ||
1028 | /* | |
1029 | * Same as above, but for rgid, egid, sgid. | |
1030 | */ | |
ae1251ab | 1031 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) |
1da177e4 | 1032 | { |
a29c33f4 | 1033 | struct user_namespace *ns = current_user_ns(); |
d84f4f99 DH |
1034 | const struct cred *old; |
1035 | struct cred *new; | |
1da177e4 | 1036 | int retval; |
a29c33f4 EB |
1037 | kgid_t krgid, kegid, ksgid; |
1038 | ||
3c2a0909 S |
1039 | #if defined CONFIG_SEC_RESTRICT_SETUID |
1040 | if(rgid == 0 || egid == 0 || sgid == 0) | |
1041 | { | |
1042 | if(sec_restrict_uid()) | |
1043 | return -EACCES; | |
1044 | } | |
1045 | #endif // End of CONFIG_SEC_RESTRICT_SETUID | |
1046 | ||
a29c33f4 EB |
1047 | krgid = make_kgid(ns, rgid); |
1048 | kegid = make_kgid(ns, egid); | |
1049 | ksgid = make_kgid(ns, sgid); | |
1050 | ||
1051 | if ((rgid != (gid_t) -1) && !gid_valid(krgid)) | |
1052 | return -EINVAL; | |
1053 | if ((egid != (gid_t) -1) && !gid_valid(kegid)) | |
1054 | return -EINVAL; | |
1055 | if ((sgid != (gid_t) -1) && !gid_valid(ksgid)) | |
1056 | return -EINVAL; | |
1da177e4 | 1057 | |
d84f4f99 DH |
1058 | new = prepare_creds(); |
1059 | if (!new) | |
1060 | return -ENOMEM; | |
1061 | old = current_cred(); | |
1062 | ||
d84f4f99 | 1063 | retval = -EPERM; |
fc832ad3 | 1064 | if (!nsown_capable(CAP_SETGID)) { |
a29c33f4 EB |
1065 | if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) && |
1066 | !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid)) | |
d84f4f99 | 1067 | goto error; |
a29c33f4 EB |
1068 | if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) && |
1069 | !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid)) | |
d84f4f99 | 1070 | goto error; |
a29c33f4 EB |
1071 | if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) && |
1072 | !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid)) | |
d84f4f99 | 1073 | goto error; |
1da177e4 | 1074 | } |
d84f4f99 | 1075 | |
1da177e4 | 1076 | if (rgid != (gid_t) -1) |
a29c33f4 | 1077 | new->gid = krgid; |
d84f4f99 | 1078 | if (egid != (gid_t) -1) |
a29c33f4 | 1079 | new->egid = kegid; |
1da177e4 | 1080 | if (sgid != (gid_t) -1) |
a29c33f4 | 1081 | new->sgid = ksgid; |
d84f4f99 | 1082 | new->fsgid = new->egid; |
1da177e4 | 1083 | |
d84f4f99 DH |
1084 | return commit_creds(new); |
1085 | ||
1086 | error: | |
1087 | abort_creds(new); | |
1088 | return retval; | |
1da177e4 LT |
1089 | } |
1090 | ||
a29c33f4 | 1091 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp) |
1da177e4 | 1092 | { |
86a264ab | 1093 | const struct cred *cred = current_cred(); |
1da177e4 | 1094 | int retval; |
a29c33f4 EB |
1095 | gid_t rgid, egid, sgid; |
1096 | ||
1097 | rgid = from_kgid_munged(cred->user_ns, cred->gid); | |
1098 | egid = from_kgid_munged(cred->user_ns, cred->egid); | |
1099 | sgid = from_kgid_munged(cred->user_ns, cred->sgid); | |
1da177e4 | 1100 | |
a29c33f4 EB |
1101 | if (!(retval = put_user(rgid, rgidp)) && |
1102 | !(retval = put_user(egid, egidp))) | |
1103 | retval = put_user(sgid, sgidp); | |
1da177e4 LT |
1104 | |
1105 | return retval; | |
1106 | } | |
1107 | ||
1108 | ||
1109 | /* | |
1110 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
1111 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
1112 | * whatever uid it wants to). It normally shadows "euid", except when | |
1113 | * explicitly set by setfsuid() or for access.. | |
1114 | */ | |
ae1251ab | 1115 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) |
1da177e4 | 1116 | { |
d84f4f99 DH |
1117 | const struct cred *old; |
1118 | struct cred *new; | |
1119 | uid_t old_fsuid; | |
a29c33f4 EB |
1120 | kuid_t kuid; |
1121 | ||
1122 | old = current_cred(); | |
1123 | old_fsuid = from_kuid_munged(old->user_ns, old->fsuid); | |
1124 | ||
1125 | kuid = make_kuid(old->user_ns, uid); | |
1126 | if (!uid_valid(kuid)) | |
1127 | return old_fsuid; | |
1da177e4 | 1128 | |
d84f4f99 DH |
1129 | new = prepare_creds(); |
1130 | if (!new) | |
a29c33f4 | 1131 | return old_fsuid; |
1da177e4 | 1132 | |
a29c33f4 EB |
1133 | if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) || |
1134 | uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) || | |
fc832ad3 | 1135 | nsown_capable(CAP_SETUID)) { |
a29c33f4 EB |
1136 | if (!uid_eq(kuid, old->fsuid)) { |
1137 | new->fsuid = kuid; | |
d84f4f99 DH |
1138 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) |
1139 | goto change_okay; | |
1da177e4 | 1140 | } |
1da177e4 LT |
1141 | } |
1142 | ||
d84f4f99 DH |
1143 | abort_creds(new); |
1144 | return old_fsuid; | |
1da177e4 | 1145 | |
d84f4f99 DH |
1146 | change_okay: |
1147 | commit_creds(new); | |
1da177e4 LT |
1148 | return old_fsuid; |
1149 | } | |
1150 | ||
1151 | /* | |
f42df9e6 | 1152 | * Samma på svenska.. |
1da177e4 | 1153 | */ |
ae1251ab | 1154 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) |
1da177e4 | 1155 | { |
d84f4f99 DH |
1156 | const struct cred *old; |
1157 | struct cred *new; | |
1158 | gid_t old_fsgid; | |
a29c33f4 EB |
1159 | kgid_t kgid; |
1160 | ||
1161 | old = current_cred(); | |
1162 | old_fsgid = from_kgid_munged(old->user_ns, old->fsgid); | |
1163 | ||
1164 | kgid = make_kgid(old->user_ns, gid); | |
1165 | if (!gid_valid(kgid)) | |
1166 | return old_fsgid; | |
d84f4f99 DH |
1167 | |
1168 | new = prepare_creds(); | |
1169 | if (!new) | |
a29c33f4 | 1170 | return old_fsgid; |
1da177e4 | 1171 | |
a29c33f4 EB |
1172 | if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) || |
1173 | gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) || | |
fc832ad3 | 1174 | nsown_capable(CAP_SETGID)) { |
a29c33f4 EB |
1175 | if (!gid_eq(kgid, old->fsgid)) { |
1176 | new->fsgid = kgid; | |
d84f4f99 | 1177 | goto change_okay; |
1da177e4 | 1178 | } |
1da177e4 | 1179 | } |
d84f4f99 | 1180 | |
d84f4f99 DH |
1181 | abort_creds(new); |
1182 | return old_fsgid; | |
1183 | ||
1184 | change_okay: | |
1185 | commit_creds(new); | |
1da177e4 LT |
1186 | return old_fsgid; |
1187 | } | |
1188 | ||
4a22f166 SR |
1189 | /** |
1190 | * sys_getpid - return the thread group id of the current process | |
1191 | * | |
1192 | * Note, despite the name, this returns the tgid not the pid. The tgid and | |
1193 | * the pid are identical unless CLONE_THREAD was specified on clone() in | |
1194 | * which case the tgid is the same in all threads of the same group. | |
1195 | * | |
1196 | * This is SMP safe as current->tgid does not change. | |
1197 | */ | |
1198 | SYSCALL_DEFINE0(getpid) | |
1199 | { | |
1200 | return task_tgid_vnr(current); | |
1201 | } | |
1202 | ||
1203 | /* Thread ID - the internal kernel "pid" */ | |
1204 | SYSCALL_DEFINE0(gettid) | |
1205 | { | |
1206 | return task_pid_vnr(current); | |
1207 | } | |
1208 | ||
1209 | /* | |
1210 | * Accessing ->real_parent is not SMP-safe, it could | |
1211 | * change from under us. However, we can use a stale | |
1212 | * value of ->real_parent under rcu_read_lock(), see | |
1213 | * release_task()->call_rcu(delayed_put_task_struct). | |
1214 | */ | |
1215 | SYSCALL_DEFINE0(getppid) | |
1216 | { | |
1217 | int pid; | |
1218 | ||
1219 | rcu_read_lock(); | |
1220 | pid = task_tgid_vnr(rcu_dereference(current->real_parent)); | |
1221 | rcu_read_unlock(); | |
1222 | ||
1223 | return pid; | |
1224 | } | |
1225 | ||
1226 | SYSCALL_DEFINE0(getuid) | |
1227 | { | |
1228 | /* Only we change this so SMP safe */ | |
1229 | return from_kuid_munged(current_user_ns(), current_uid()); | |
1230 | } | |
1231 | ||
1232 | SYSCALL_DEFINE0(geteuid) | |
1233 | { | |
1234 | /* Only we change this so SMP safe */ | |
1235 | return from_kuid_munged(current_user_ns(), current_euid()); | |
1236 | } | |
1237 | ||
1238 | SYSCALL_DEFINE0(getgid) | |
1239 | { | |
1240 | /* Only we change this so SMP safe */ | |
1241 | return from_kgid_munged(current_user_ns(), current_gid()); | |
1242 | } | |
1243 | ||
1244 | SYSCALL_DEFINE0(getegid) | |
1245 | { | |
1246 | /* Only we change this so SMP safe */ | |
1247 | return from_kgid_munged(current_user_ns(), current_egid()); | |
1248 | } | |
1249 | ||
f06febc9 FM |
1250 | void do_sys_times(struct tms *tms) |
1251 | { | |
0cf55e1e | 1252 | cputime_t tgutime, tgstime, cutime, cstime; |
f06febc9 | 1253 | |
2b5fe6de | 1254 | spin_lock_irq(¤t->sighand->siglock); |
e80d0a1a | 1255 | thread_group_cputime_adjusted(current, &tgutime, &tgstime); |
f06febc9 FM |
1256 | cutime = current->signal->cutime; |
1257 | cstime = current->signal->cstime; | |
1258 | spin_unlock_irq(¤t->sighand->siglock); | |
0cf55e1e HS |
1259 | tms->tms_utime = cputime_to_clock_t(tgutime); |
1260 | tms->tms_stime = cputime_to_clock_t(tgstime); | |
f06febc9 FM |
1261 | tms->tms_cutime = cputime_to_clock_t(cutime); |
1262 | tms->tms_cstime = cputime_to_clock_t(cstime); | |
1263 | } | |
1264 | ||
58fd3aa2 | 1265 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) |
1da177e4 | 1266 | { |
1da177e4 LT |
1267 | if (tbuf) { |
1268 | struct tms tmp; | |
f06febc9 FM |
1269 | |
1270 | do_sys_times(&tmp); | |
1da177e4 LT |
1271 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
1272 | return -EFAULT; | |
1273 | } | |
e3d5a27d | 1274 | force_successful_syscall_return(); |
1da177e4 LT |
1275 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
1276 | } | |
1277 | ||
1278 | /* | |
1279 | * This needs some heavy checking ... | |
1280 | * I just haven't the stomach for it. I also don't fully | |
1281 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
1282 | * | |
1283 | * OK, I think I have the protection semantics right.... this is really | |
1284 | * only important on a multi-user system anyway, to make sure one user | |
1285 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
1286 | * | |
1287 | * Auch. Had to add the 'did_exec' flag to conform completely to POSIX. | |
1288 | * LBT 04.03.94 | |
1289 | */ | |
b290ebe2 | 1290 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) |
1da177e4 LT |
1291 | { |
1292 | struct task_struct *p; | |
ee0acf90 | 1293 | struct task_struct *group_leader = current->group_leader; |
4e021306 ON |
1294 | struct pid *pgrp; |
1295 | int err; | |
1da177e4 LT |
1296 | |
1297 | if (!pid) | |
b488893a | 1298 | pid = task_pid_vnr(group_leader); |
1da177e4 LT |
1299 | if (!pgid) |
1300 | pgid = pid; | |
1301 | if (pgid < 0) | |
1302 | return -EINVAL; | |
950eaaca | 1303 | rcu_read_lock(); |
1da177e4 LT |
1304 | |
1305 | /* From this point forward we keep holding onto the tasklist lock | |
1306 | * so that our parent does not change from under us. -DaveM | |
1307 | */ | |
1308 | write_lock_irq(&tasklist_lock); | |
1309 | ||
1310 | err = -ESRCH; | |
4e021306 | 1311 | p = find_task_by_vpid(pid); |
1da177e4 LT |
1312 | if (!p) |
1313 | goto out; | |
1314 | ||
1315 | err = -EINVAL; | |
1316 | if (!thread_group_leader(p)) | |
1317 | goto out; | |
1318 | ||
4e021306 | 1319 | if (same_thread_group(p->real_parent, group_leader)) { |
1da177e4 | 1320 | err = -EPERM; |
41487c65 | 1321 | if (task_session(p) != task_session(group_leader)) |
1da177e4 LT |
1322 | goto out; |
1323 | err = -EACCES; | |
1324 | if (p->did_exec) | |
1325 | goto out; | |
1326 | } else { | |
1327 | err = -ESRCH; | |
ee0acf90 | 1328 | if (p != group_leader) |
1da177e4 LT |
1329 | goto out; |
1330 | } | |
1331 | ||
1332 | err = -EPERM; | |
1333 | if (p->signal->leader) | |
1334 | goto out; | |
1335 | ||
4e021306 | 1336 | pgrp = task_pid(p); |
1da177e4 | 1337 | if (pgid != pid) { |
b488893a | 1338 | struct task_struct *g; |
1da177e4 | 1339 | |
4e021306 ON |
1340 | pgrp = find_vpid(pgid); |
1341 | g = pid_task(pgrp, PIDTYPE_PGID); | |
41487c65 | 1342 | if (!g || task_session(g) != task_session(group_leader)) |
f020bc46 | 1343 | goto out; |
1da177e4 LT |
1344 | } |
1345 | ||
1da177e4 LT |
1346 | err = security_task_setpgid(p, pgid); |
1347 | if (err) | |
1348 | goto out; | |
1349 | ||
1b0f7ffd | 1350 | if (task_pgrp(p) != pgrp) |
83beaf3c | 1351 | change_pid(p, PIDTYPE_PGID, pgrp); |
1da177e4 LT |
1352 | |
1353 | err = 0; | |
1354 | out: | |
1355 | /* All paths lead to here, thus we are safe. -DaveM */ | |
1356 | write_unlock_irq(&tasklist_lock); | |
950eaaca | 1357 | rcu_read_unlock(); |
1da177e4 LT |
1358 | return err; |
1359 | } | |
1360 | ||
dbf040d9 | 1361 | SYSCALL_DEFINE1(getpgid, pid_t, pid) |
1da177e4 | 1362 | { |
12a3de0a ON |
1363 | struct task_struct *p; |
1364 | struct pid *grp; | |
1365 | int retval; | |
1366 | ||
1367 | rcu_read_lock(); | |
756184b7 | 1368 | if (!pid) |
12a3de0a | 1369 | grp = task_pgrp(current); |
756184b7 | 1370 | else { |
1da177e4 | 1371 | retval = -ESRCH; |
12a3de0a ON |
1372 | p = find_task_by_vpid(pid); |
1373 | if (!p) | |
1374 | goto out; | |
1375 | grp = task_pgrp(p); | |
1376 | if (!grp) | |
1377 | goto out; | |
1378 | ||
1379 | retval = security_task_getpgid(p); | |
1380 | if (retval) | |
1381 | goto out; | |
1da177e4 | 1382 | } |
12a3de0a ON |
1383 | retval = pid_vnr(grp); |
1384 | out: | |
1385 | rcu_read_unlock(); | |
1386 | return retval; | |
1da177e4 LT |
1387 | } |
1388 | ||
1389 | #ifdef __ARCH_WANT_SYS_GETPGRP | |
1390 | ||
dbf040d9 | 1391 | SYSCALL_DEFINE0(getpgrp) |
1da177e4 | 1392 | { |
12a3de0a | 1393 | return sys_getpgid(0); |
1da177e4 LT |
1394 | } |
1395 | ||
1396 | #endif | |
1397 | ||
dbf040d9 | 1398 | SYSCALL_DEFINE1(getsid, pid_t, pid) |
1da177e4 | 1399 | { |
1dd768c0 ON |
1400 | struct task_struct *p; |
1401 | struct pid *sid; | |
1402 | int retval; | |
1403 | ||
1404 | rcu_read_lock(); | |
756184b7 | 1405 | if (!pid) |
1dd768c0 | 1406 | sid = task_session(current); |
756184b7 | 1407 | else { |
1da177e4 | 1408 | retval = -ESRCH; |
1dd768c0 ON |
1409 | p = find_task_by_vpid(pid); |
1410 | if (!p) | |
1411 | goto out; | |
1412 | sid = task_session(p); | |
1413 | if (!sid) | |
1414 | goto out; | |
1415 | ||
1416 | retval = security_task_getsid(p); | |
1417 | if (retval) | |
1418 | goto out; | |
1da177e4 | 1419 | } |
1dd768c0 ON |
1420 | retval = pid_vnr(sid); |
1421 | out: | |
1422 | rcu_read_unlock(); | |
1423 | return retval; | |
1da177e4 LT |
1424 | } |
1425 | ||
b290ebe2 | 1426 | SYSCALL_DEFINE0(setsid) |
1da177e4 | 1427 | { |
e19f247a | 1428 | struct task_struct *group_leader = current->group_leader; |
e4cc0a9c ON |
1429 | struct pid *sid = task_pid(group_leader); |
1430 | pid_t session = pid_vnr(sid); | |
1da177e4 LT |
1431 | int err = -EPERM; |
1432 | ||
1da177e4 | 1433 | write_lock_irq(&tasklist_lock); |
390e2ff0 EB |
1434 | /* Fail if I am already a session leader */ |
1435 | if (group_leader->signal->leader) | |
1436 | goto out; | |
1437 | ||
430c6231 ON |
1438 | /* Fail if a process group id already exists that equals the |
1439 | * proposed session id. | |
390e2ff0 | 1440 | */ |
6806aac6 | 1441 | if (pid_task(sid, PIDTYPE_PGID)) |
1da177e4 LT |
1442 | goto out; |
1443 | ||
e19f247a | 1444 | group_leader->signal->leader = 1; |
8520d7c7 | 1445 | __set_special_pids(sid); |
24ec839c | 1446 | |
9c9f4ded | 1447 | proc_clear_tty(group_leader); |
24ec839c | 1448 | |
e4cc0a9c | 1449 | err = session; |
1da177e4 LT |
1450 | out: |
1451 | write_unlock_irq(&tasklist_lock); | |
5091faa4 | 1452 | if (err > 0) { |
0d0df599 | 1453 | proc_sid_connector(group_leader); |
5091faa4 MG |
1454 | sched_autogroup_create_attach(group_leader); |
1455 | } | |
1da177e4 LT |
1456 | return err; |
1457 | } | |
1458 | ||
1da177e4 LT |
1459 | DECLARE_RWSEM(uts_sem); |
1460 | ||
e28cbf22 CH |
1461 | #ifdef COMPAT_UTS_MACHINE |
1462 | #define override_architecture(name) \ | |
46da2766 | 1463 | (personality(current->personality) == PER_LINUX32 && \ |
e28cbf22 CH |
1464 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ |
1465 | sizeof(COMPAT_UTS_MACHINE))) | |
1466 | #else | |
1467 | #define override_architecture(name) 0 | |
1468 | #endif | |
1469 | ||
be27425d AK |
1470 | /* |
1471 | * Work around broken programs that cannot handle "Linux 3.0". | |
1472 | * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 | |
1473 | */ | |
2702b152 | 1474 | static int override_release(char __user *release, size_t len) |
be27425d AK |
1475 | { |
1476 | int ret = 0; | |
be27425d AK |
1477 | |
1478 | if (current->personality & UNAME26) { | |
2702b152 KC |
1479 | const char *rest = UTS_RELEASE; |
1480 | char buf[65] = { 0 }; | |
be27425d AK |
1481 | int ndots = 0; |
1482 | unsigned v; | |
2702b152 | 1483 | size_t copy; |
be27425d AK |
1484 | |
1485 | while (*rest) { | |
1486 | if (*rest == '.' && ++ndots >= 3) | |
1487 | break; | |
1488 | if (!isdigit(*rest) && *rest != '.') | |
1489 | break; | |
1490 | rest++; | |
1491 | } | |
1492 | v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40; | |
31fd84b9 | 1493 | copy = clamp_t(size_t, len, 1, sizeof(buf)); |
2702b152 KC |
1494 | copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); |
1495 | ret = copy_to_user(release, buf, copy + 1); | |
be27425d AK |
1496 | } |
1497 | return ret; | |
1498 | } | |
1499 | ||
e48fbb69 | 1500 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) |
1da177e4 LT |
1501 | { |
1502 | int errno = 0; | |
1503 | ||
1504 | down_read(&uts_sem); | |
e9ff3990 | 1505 | if (copy_to_user(name, utsname(), sizeof *name)) |
1da177e4 LT |
1506 | errno = -EFAULT; |
1507 | up_read(&uts_sem); | |
e28cbf22 | 1508 | |
be27425d AK |
1509 | if (!errno && override_release(name->release, sizeof(name->release))) |
1510 | errno = -EFAULT; | |
e28cbf22 CH |
1511 | if (!errno && override_architecture(name)) |
1512 | errno = -EFAULT; | |
1da177e4 LT |
1513 | return errno; |
1514 | } | |
1515 | ||
5cacdb4a CH |
1516 | #ifdef __ARCH_WANT_SYS_OLD_UNAME |
1517 | /* | |
1518 | * Old cruft | |
1519 | */ | |
1520 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1521 | { | |
1522 | int error = 0; | |
1523 | ||
1524 | if (!name) | |
1525 | return -EFAULT; | |
1526 | ||
1527 | down_read(&uts_sem); | |
1528 | if (copy_to_user(name, utsname(), sizeof(*name))) | |
1529 | error = -EFAULT; | |
1530 | up_read(&uts_sem); | |
1531 | ||
be27425d AK |
1532 | if (!error && override_release(name->release, sizeof(name->release))) |
1533 | error = -EFAULT; | |
5cacdb4a CH |
1534 | if (!error && override_architecture(name)) |
1535 | error = -EFAULT; | |
1536 | return error; | |
1537 | } | |
1538 | ||
1539 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1540 | { | |
1541 | int error; | |
1542 | ||
1543 | if (!name) | |
1544 | return -EFAULT; | |
1545 | if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname))) | |
1546 | return -EFAULT; | |
1547 | ||
1548 | down_read(&uts_sem); | |
1549 | error = __copy_to_user(&name->sysname, &utsname()->sysname, | |
1550 | __OLD_UTS_LEN); | |
1551 | error |= __put_user(0, name->sysname + __OLD_UTS_LEN); | |
1552 | error |= __copy_to_user(&name->nodename, &utsname()->nodename, | |
1553 | __OLD_UTS_LEN); | |
1554 | error |= __put_user(0, name->nodename + __OLD_UTS_LEN); | |
1555 | error |= __copy_to_user(&name->release, &utsname()->release, | |
1556 | __OLD_UTS_LEN); | |
1557 | error |= __put_user(0, name->release + __OLD_UTS_LEN); | |
1558 | error |= __copy_to_user(&name->version, &utsname()->version, | |
1559 | __OLD_UTS_LEN); | |
1560 | error |= __put_user(0, name->version + __OLD_UTS_LEN); | |
1561 | error |= __copy_to_user(&name->machine, &utsname()->machine, | |
1562 | __OLD_UTS_LEN); | |
1563 | error |= __put_user(0, name->machine + __OLD_UTS_LEN); | |
1564 | up_read(&uts_sem); | |
1565 | ||
1566 | if (!error && override_architecture(name)) | |
1567 | error = -EFAULT; | |
be27425d AK |
1568 | if (!error && override_release(name->release, sizeof(name->release))) |
1569 | error = -EFAULT; | |
5cacdb4a CH |
1570 | return error ? -EFAULT : 0; |
1571 | } | |
1572 | #endif | |
1573 | ||
5a8a82b1 | 1574 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) |
1da177e4 LT |
1575 | { |
1576 | int errno; | |
1577 | char tmp[__NEW_UTS_LEN]; | |
1578 | ||
bb96a6f5 | 1579 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 | 1580 | return -EPERM; |
fc832ad3 | 1581 | |
1da177e4 LT |
1582 | if (len < 0 || len > __NEW_UTS_LEN) |
1583 | return -EINVAL; | |
1584 | down_write(&uts_sem); | |
1585 | errno = -EFAULT; | |
1586 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1587 | struct new_utsname *u = utsname(); |
1588 | ||
1589 | memcpy(u->nodename, tmp, len); | |
1590 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1da177e4 | 1591 | errno = 0; |
499eea6b | 1592 | uts_proc_notify(UTS_PROC_HOSTNAME); |
1da177e4 LT |
1593 | } |
1594 | up_write(&uts_sem); | |
1595 | return errno; | |
1596 | } | |
1597 | ||
1598 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1599 | ||
5a8a82b1 | 1600 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) |
1da177e4 LT |
1601 | { |
1602 | int i, errno; | |
9679e4dd | 1603 | struct new_utsname *u; |
1da177e4 LT |
1604 | |
1605 | if (len < 0) | |
1606 | return -EINVAL; | |
1607 | down_read(&uts_sem); | |
9679e4dd AM |
1608 | u = utsname(); |
1609 | i = 1 + strlen(u->nodename); | |
1da177e4 LT |
1610 | if (i > len) |
1611 | i = len; | |
1612 | errno = 0; | |
9679e4dd | 1613 | if (copy_to_user(name, u->nodename, i)) |
1da177e4 LT |
1614 | errno = -EFAULT; |
1615 | up_read(&uts_sem); | |
1616 | return errno; | |
1617 | } | |
1618 | ||
1619 | #endif | |
1620 | ||
1621 | /* | |
1622 | * Only setdomainname; getdomainname can be implemented by calling | |
1623 | * uname() | |
1624 | */ | |
5a8a82b1 | 1625 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) |
1da177e4 LT |
1626 | { |
1627 | int errno; | |
1628 | char tmp[__NEW_UTS_LEN]; | |
1629 | ||
fc832ad3 | 1630 | if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN)) |
1da177e4 LT |
1631 | return -EPERM; |
1632 | if (len < 0 || len > __NEW_UTS_LEN) | |
1633 | return -EINVAL; | |
1634 | ||
1635 | down_write(&uts_sem); | |
1636 | errno = -EFAULT; | |
1637 | if (!copy_from_user(tmp, name, len)) { | |
9679e4dd AM |
1638 | struct new_utsname *u = utsname(); |
1639 | ||
1640 | memcpy(u->domainname, tmp, len); | |
1641 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1da177e4 | 1642 | errno = 0; |
499eea6b | 1643 | uts_proc_notify(UTS_PROC_DOMAINNAME); |
1da177e4 LT |
1644 | } |
1645 | up_write(&uts_sem); | |
1646 | return errno; | |
1647 | } | |
1648 | ||
e48fbb69 | 1649 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1da177e4 | 1650 | { |
b9518345 JS |
1651 | struct rlimit value; |
1652 | int ret; | |
1653 | ||
1654 | ret = do_prlimit(current, resource, NULL, &value); | |
1655 | if (!ret) | |
1656 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1657 | ||
1658 | return ret; | |
1da177e4 LT |
1659 | } |
1660 | ||
1661 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT | |
1662 | ||
1663 | /* | |
1664 | * Back compatibility for getrlimit. Needed for some apps. | |
1665 | */ | |
1666 | ||
e48fbb69 HC |
1667 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, |
1668 | struct rlimit __user *, rlim) | |
1da177e4 LT |
1669 | { |
1670 | struct rlimit x; | |
1671 | if (resource >= RLIM_NLIMITS) | |
1672 | return -EINVAL; | |
1673 | ||
1674 | task_lock(current->group_leader); | |
1675 | x = current->signal->rlim[resource]; | |
1676 | task_unlock(current->group_leader); | |
756184b7 | 1677 | if (x.rlim_cur > 0x7FFFFFFF) |
1da177e4 | 1678 | x.rlim_cur = 0x7FFFFFFF; |
756184b7 | 1679 | if (x.rlim_max > 0x7FFFFFFF) |
1da177e4 LT |
1680 | x.rlim_max = 0x7FFFFFFF; |
1681 | return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0; | |
1682 | } | |
1683 | ||
1684 | #endif | |
1685 | ||
c022a0ac JS |
1686 | static inline bool rlim64_is_infinity(__u64 rlim64) |
1687 | { | |
1688 | #if BITS_PER_LONG < 64 | |
1689 | return rlim64 >= ULONG_MAX; | |
1690 | #else | |
1691 | return rlim64 == RLIM64_INFINITY; | |
1692 | #endif | |
1693 | } | |
1694 | ||
1695 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1696 | { | |
1697 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1698 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1699 | else | |
1700 | rlim64->rlim_cur = rlim->rlim_cur; | |
1701 | if (rlim->rlim_max == RLIM_INFINITY) | |
1702 | rlim64->rlim_max = RLIM64_INFINITY; | |
1703 | else | |
1704 | rlim64->rlim_max = rlim->rlim_max; | |
1705 | } | |
1706 | ||
1707 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1708 | { | |
1709 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1710 | rlim->rlim_cur = RLIM_INFINITY; | |
1711 | else | |
1712 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1713 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1714 | rlim->rlim_max = RLIM_INFINITY; | |
1715 | else | |
1716 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1717 | } | |
1718 | ||
1c1e618d | 1719 | /* make sure you are allowed to change @tsk limits before calling this */ |
5b41535a JS |
1720 | int do_prlimit(struct task_struct *tsk, unsigned int resource, |
1721 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1da177e4 | 1722 | { |
5b41535a | 1723 | struct rlimit *rlim; |
86f162f4 | 1724 | int retval = 0; |
1da177e4 LT |
1725 | |
1726 | if (resource >= RLIM_NLIMITS) | |
1727 | return -EINVAL; | |
5b41535a JS |
1728 | if (new_rlim) { |
1729 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1730 | return -EINVAL; | |
1731 | if (resource == RLIMIT_NOFILE && | |
1732 | new_rlim->rlim_max > sysctl_nr_open) | |
1733 | return -EPERM; | |
1734 | } | |
1da177e4 | 1735 | |
1c1e618d JS |
1736 | /* protect tsk->signal and tsk->sighand from disappearing */ |
1737 | read_lock(&tasklist_lock); | |
1738 | if (!tsk->sighand) { | |
1739 | retval = -ESRCH; | |
1740 | goto out; | |
1741 | } | |
1742 | ||
5b41535a | 1743 | rlim = tsk->signal->rlim + resource; |
86f162f4 | 1744 | task_lock(tsk->group_leader); |
5b41535a | 1745 | if (new_rlim) { |
fc832ad3 SH |
1746 | /* Keep the capable check against init_user_ns until |
1747 | cgroups can contain all limits */ | |
5b41535a JS |
1748 | if (new_rlim->rlim_max > rlim->rlim_max && |
1749 | !capable(CAP_SYS_RESOURCE)) | |
1750 | retval = -EPERM; | |
1751 | if (!retval) | |
1752 | retval = security_task_setrlimit(tsk->group_leader, | |
1753 | resource, new_rlim); | |
1754 | if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) { | |
1755 | /* | |
1756 | * The caller is asking for an immediate RLIMIT_CPU | |
1757 | * expiry. But we use the zero value to mean "it was | |
1758 | * never set". So let's cheat and make it one second | |
1759 | * instead | |
1760 | */ | |
1761 | new_rlim->rlim_cur = 1; | |
1762 | } | |
1763 | } | |
1764 | if (!retval) { | |
1765 | if (old_rlim) | |
1766 | *old_rlim = *rlim; | |
1767 | if (new_rlim) | |
1768 | *rlim = *new_rlim; | |
9926e4c7 | 1769 | } |
7855c35d | 1770 | task_unlock(tsk->group_leader); |
1da177e4 | 1771 | |
d3561f78 AM |
1772 | /* |
1773 | * RLIMIT_CPU handling. Note that the kernel fails to return an error | |
1774 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a | |
1775 | * very long-standing error, and fixing it now risks breakage of | |
1776 | * applications, so we live with it | |
1777 | */ | |
5b41535a JS |
1778 | if (!retval && new_rlim && resource == RLIMIT_CPU && |
1779 | new_rlim->rlim_cur != RLIM_INFINITY) | |
1780 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); | |
ec9e16ba | 1781 | out: |
1c1e618d | 1782 | read_unlock(&tasklist_lock); |
2fb9d268 | 1783 | return retval; |
1da177e4 LT |
1784 | } |
1785 | ||
c022a0ac JS |
1786 | /* rcu lock must be held */ |
1787 | static int check_prlimit_permission(struct task_struct *task) | |
1788 | { | |
1789 | const struct cred *cred = current_cred(), *tcred; | |
1790 | ||
fc832ad3 SH |
1791 | if (current == task) |
1792 | return 0; | |
c022a0ac | 1793 | |
fc832ad3 | 1794 | tcred = __task_cred(task); |
5af66203 EB |
1795 | if (uid_eq(cred->uid, tcred->euid) && |
1796 | uid_eq(cred->uid, tcred->suid) && | |
1797 | uid_eq(cred->uid, tcred->uid) && | |
1798 | gid_eq(cred->gid, tcred->egid) && | |
1799 | gid_eq(cred->gid, tcred->sgid) && | |
1800 | gid_eq(cred->gid, tcred->gid)) | |
fc832ad3 | 1801 | return 0; |
c4a4d603 | 1802 | if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) |
fc832ad3 SH |
1803 | return 0; |
1804 | ||
1805 | return -EPERM; | |
c022a0ac JS |
1806 | } |
1807 | ||
1808 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1809 | const struct rlimit64 __user *, new_rlim, | |
1810 | struct rlimit64 __user *, old_rlim) | |
1811 | { | |
1812 | struct rlimit64 old64, new64; | |
1813 | struct rlimit old, new; | |
1814 | struct task_struct *tsk; | |
1815 | int ret; | |
1816 | ||
1817 | if (new_rlim) { | |
1818 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1819 | return -EFAULT; | |
1820 | rlim64_to_rlim(&new64, &new); | |
1821 | } | |
1822 | ||
1823 | rcu_read_lock(); | |
1824 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1825 | if (!tsk) { | |
1826 | rcu_read_unlock(); | |
1827 | return -ESRCH; | |
1828 | } | |
1829 | ret = check_prlimit_permission(tsk); | |
1830 | if (ret) { | |
1831 | rcu_read_unlock(); | |
1832 | return ret; | |
1833 | } | |
1834 | get_task_struct(tsk); | |
1835 | rcu_read_unlock(); | |
1836 | ||
1837 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1838 | old_rlim ? &old : NULL); | |
1839 | ||
1840 | if (!ret && old_rlim) { | |
1841 | rlim_to_rlim64(&old, &old64); | |
1842 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1843 | ret = -EFAULT; | |
1844 | } | |
1845 | ||
1846 | put_task_struct(tsk); | |
1847 | return ret; | |
1848 | } | |
1849 | ||
7855c35d JS |
1850 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) |
1851 | { | |
1852 | struct rlimit new_rlim; | |
1853 | ||
1854 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1855 | return -EFAULT; | |
5b41535a | 1856 | return do_prlimit(current, resource, &new_rlim, NULL); |
7855c35d JS |
1857 | } |
1858 | ||
1da177e4 LT |
1859 | /* |
1860 | * It would make sense to put struct rusage in the task_struct, | |
1861 | * except that would make the task_struct be *really big*. After | |
1862 | * task_struct gets moved into malloc'ed memory, it would | |
1863 | * make sense to do this. It will make moving the rest of the information | |
1864 | * a lot simpler! (Which we're not doing right now because we're not | |
1865 | * measuring them yet). | |
1866 | * | |
1da177e4 LT |
1867 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
1868 | * races with threads incrementing their own counters. But since word | |
1869 | * reads are atomic, we either get new values or old values and we don't | |
1870 | * care which for the sums. We always take the siglock to protect reading | |
1871 | * the c* fields from p->signal from races with exit.c updating those | |
1872 | * fields when reaping, so a sample either gets all the additions of a | |
1873 | * given child after it's reaped, or none so this sample is before reaping. | |
2dd0ebcd | 1874 | * |
de047c1b RT |
1875 | * Locking: |
1876 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1877 | * for the cases current multithreaded, non-current single threaded | |
1878 | * non-current multithreaded. Thread traversal is now safe with | |
1879 | * the siglock held. | |
1880 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1881 | * single threaded, as no one else can take our signal_struct away, no one | |
1882 | * else can reap the children to update signal->c* counters, and no one else | |
1883 | * can race with the signal-> fields. If we do not take any lock, the | |
1884 | * signal-> fields could be read out of order while another thread was just | |
1885 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1886 | * On the writer side, write memory barrier is implied in __exit_signal | |
1887 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1888 | * fields. But we don't do this yet to keep things simple. | |
2dd0ebcd | 1889 | * |
1da177e4 LT |
1890 | */ |
1891 | ||
f06febc9 | 1892 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) |
679c9cd4 | 1893 | { |
679c9cd4 SK |
1894 | r->ru_nvcsw += t->nvcsw; |
1895 | r->ru_nivcsw += t->nivcsw; | |
1896 | r->ru_minflt += t->min_flt; | |
1897 | r->ru_majflt += t->maj_flt; | |
1898 | r->ru_inblock += task_io_get_inblock(t); | |
1899 | r->ru_oublock += task_io_get_oublock(t); | |
1900 | } | |
1901 | ||
1da177e4 LT |
1902 | static void k_getrusage(struct task_struct *p, int who, struct rusage *r) |
1903 | { | |
1904 | struct task_struct *t; | |
1905 | unsigned long flags; | |
0cf55e1e | 1906 | cputime_t tgutime, tgstime, utime, stime; |
1f10206c | 1907 | unsigned long maxrss = 0; |
1da177e4 LT |
1908 | |
1909 | memset((char *) r, 0, sizeof *r); | |
64861634 | 1910 | utime = stime = 0; |
1da177e4 | 1911 | |
679c9cd4 | 1912 | if (who == RUSAGE_THREAD) { |
e80d0a1a | 1913 | task_cputime_adjusted(current, &utime, &stime); |
f06febc9 | 1914 | accumulate_thread_rusage(p, r); |
1f10206c | 1915 | maxrss = p->signal->maxrss; |
679c9cd4 SK |
1916 | goto out; |
1917 | } | |
1918 | ||
d6cf723a | 1919 | if (!lock_task_sighand(p, &flags)) |
de047c1b | 1920 | return; |
0f59cc4a | 1921 | |
1da177e4 | 1922 | switch (who) { |
0f59cc4a | 1923 | case RUSAGE_BOTH: |
1da177e4 | 1924 | case RUSAGE_CHILDREN: |
1da177e4 LT |
1925 | utime = p->signal->cutime; |
1926 | stime = p->signal->cstime; | |
1927 | r->ru_nvcsw = p->signal->cnvcsw; | |
1928 | r->ru_nivcsw = p->signal->cnivcsw; | |
1929 | r->ru_minflt = p->signal->cmin_flt; | |
1930 | r->ru_majflt = p->signal->cmaj_flt; | |
6eaeeaba ED |
1931 | r->ru_inblock = p->signal->cinblock; |
1932 | r->ru_oublock = p->signal->coublock; | |
1f10206c | 1933 | maxrss = p->signal->cmaxrss; |
0f59cc4a ON |
1934 | |
1935 | if (who == RUSAGE_CHILDREN) | |
1936 | break; | |
1937 | ||
1da177e4 | 1938 | case RUSAGE_SELF: |
e80d0a1a | 1939 | thread_group_cputime_adjusted(p, &tgutime, &tgstime); |
64861634 MS |
1940 | utime += tgutime; |
1941 | stime += tgstime; | |
1da177e4 LT |
1942 | r->ru_nvcsw += p->signal->nvcsw; |
1943 | r->ru_nivcsw += p->signal->nivcsw; | |
1944 | r->ru_minflt += p->signal->min_flt; | |
1945 | r->ru_majflt += p->signal->maj_flt; | |
6eaeeaba ED |
1946 | r->ru_inblock += p->signal->inblock; |
1947 | r->ru_oublock += p->signal->oublock; | |
1f10206c JP |
1948 | if (maxrss < p->signal->maxrss) |
1949 | maxrss = p->signal->maxrss; | |
1da177e4 LT |
1950 | t = p; |
1951 | do { | |
f06febc9 | 1952 | accumulate_thread_rusage(t, r); |
1da177e4 LT |
1953 | t = next_thread(t); |
1954 | } while (t != p); | |
1da177e4 | 1955 | break; |
0f59cc4a | 1956 | |
1da177e4 LT |
1957 | default: |
1958 | BUG(); | |
1959 | } | |
de047c1b | 1960 | unlock_task_sighand(p, &flags); |
de047c1b | 1961 | |
679c9cd4 | 1962 | out: |
0f59cc4a ON |
1963 | cputime_to_timeval(utime, &r->ru_utime); |
1964 | cputime_to_timeval(stime, &r->ru_stime); | |
1f10206c JP |
1965 | |
1966 | if (who != RUSAGE_CHILDREN) { | |
1967 | struct mm_struct *mm = get_task_mm(p); | |
1968 | if (mm) { | |
1969 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1970 | mmput(mm); | |
1971 | } | |
1972 | } | |
1973 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1da177e4 LT |
1974 | } |
1975 | ||
1976 | int getrusage(struct task_struct *p, int who, struct rusage __user *ru) | |
1977 | { | |
1978 | struct rusage r; | |
1da177e4 | 1979 | k_getrusage(p, who, &r); |
1da177e4 LT |
1980 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; |
1981 | } | |
1982 | ||
e48fbb69 | 1983 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) |
1da177e4 | 1984 | { |
679c9cd4 SK |
1985 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && |
1986 | who != RUSAGE_THREAD) | |
1da177e4 LT |
1987 | return -EINVAL; |
1988 | return getrusage(current, who, ru); | |
1989 | } | |
1990 | ||
8d2d5c4a AV |
1991 | #ifdef CONFIG_COMPAT |
1992 | COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru) | |
1993 | { | |
1994 | struct rusage r; | |
1995 | ||
1996 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1997 | who != RUSAGE_THREAD) | |
1998 | return -EINVAL; | |
1999 | ||
2000 | k_getrusage(current, who, &r); | |
2001 | return put_compat_rusage(&r, ru); | |
2002 | } | |
2003 | #endif | |
2004 | ||
e48fbb69 | 2005 | SYSCALL_DEFINE1(umask, int, mask) |
1da177e4 LT |
2006 | { |
2007 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
2008 | return mask; | |
2009 | } | |
3b7391de | 2010 | |
b32dfe37 CG |
2011 | static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd) |
2012 | { | |
2903ff01 | 2013 | struct fd exe; |
496ad9aa | 2014 | struct inode *inode; |
2903ff01 | 2015 | int err; |
b32dfe37 | 2016 | |
2903ff01 AV |
2017 | exe = fdget(fd); |
2018 | if (!exe.file) | |
b32dfe37 CG |
2019 | return -EBADF; |
2020 | ||
496ad9aa | 2021 | inode = file_inode(exe.file); |
b32dfe37 CG |
2022 | |
2023 | /* | |
2024 | * Because the original mm->exe_file points to executable file, make | |
2025 | * sure that this one is executable as well, to avoid breaking an | |
2026 | * overall picture. | |
2027 | */ | |
2028 | err = -EACCES; | |
496ad9aa | 2029 | if (!S_ISREG(inode->i_mode) || |
2903ff01 | 2030 | exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC) |
b32dfe37 CG |
2031 | goto exit; |
2032 | ||
496ad9aa | 2033 | err = inode_permission(inode, MAY_EXEC); |
b32dfe37 CG |
2034 | if (err) |
2035 | goto exit; | |
2036 | ||
bafb282d KK |
2037 | down_write(&mm->mmap_sem); |
2038 | ||
2039 | /* | |
4229fb1d | 2040 | * Forbid mm->exe_file change if old file still mapped. |
bafb282d KK |
2041 | */ |
2042 | err = -EBUSY; | |
4229fb1d KK |
2043 | if (mm->exe_file) { |
2044 | struct vm_area_struct *vma; | |
2045 | ||
2046 | for (vma = mm->mmap; vma; vma = vma->vm_next) | |
2047 | if (vma->vm_file && | |
2048 | path_equal(&vma->vm_file->f_path, | |
2049 | &mm->exe_file->f_path)) | |
2050 | goto exit_unlock; | |
bafb282d KK |
2051 | } |
2052 | ||
b32dfe37 CG |
2053 | /* |
2054 | * The symlink can be changed only once, just to disallow arbitrary | |
2055 | * transitions malicious software might bring in. This means one | |
2056 | * could make a snapshot over all processes running and monitor | |
2057 | * /proc/pid/exe changes to notice unusual activity if needed. | |
2058 | */ | |
bafb282d KK |
2059 | err = -EPERM; |
2060 | if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags)) | |
2061 | goto exit_unlock; | |
2062 | ||
4229fb1d | 2063 | err = 0; |
2903ff01 | 2064 | set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */ |
bafb282d | 2065 | exit_unlock: |
b32dfe37 CG |
2066 | up_write(&mm->mmap_sem); |
2067 | ||
2068 | exit: | |
2903ff01 | 2069 | fdput(exe); |
b32dfe37 CG |
2070 | return err; |
2071 | } | |
2072 | ||
028ee4be CG |
2073 | static int prctl_set_mm(int opt, unsigned long addr, |
2074 | unsigned long arg4, unsigned long arg5) | |
2075 | { | |
2076 | unsigned long rlim = rlimit(RLIMIT_DATA); | |
028ee4be | 2077 | struct mm_struct *mm = current->mm; |
fe8c7f5c CG |
2078 | struct vm_area_struct *vma; |
2079 | int error; | |
028ee4be | 2080 | |
fe8c7f5c | 2081 | if (arg5 || (arg4 && opt != PR_SET_MM_AUXV)) |
028ee4be CG |
2082 | return -EINVAL; |
2083 | ||
79f0713d | 2084 | if (!capable(CAP_SYS_RESOURCE)) |
028ee4be CG |
2085 | return -EPERM; |
2086 | ||
b32dfe37 CG |
2087 | if (opt == PR_SET_MM_EXE_FILE) |
2088 | return prctl_set_mm_exe_file(mm, (unsigned int)addr); | |
2089 | ||
1ad75b9e | 2090 | if (addr >= TASK_SIZE || addr < mmap_min_addr) |
028ee4be CG |
2091 | return -EINVAL; |
2092 | ||
fe8c7f5c CG |
2093 | error = -EINVAL; |
2094 | ||
028ee4be CG |
2095 | down_read(&mm->mmap_sem); |
2096 | vma = find_vma(mm, addr); | |
2097 | ||
028ee4be CG |
2098 | switch (opt) { |
2099 | case PR_SET_MM_START_CODE: | |
fe8c7f5c CG |
2100 | mm->start_code = addr; |
2101 | break; | |
028ee4be | 2102 | case PR_SET_MM_END_CODE: |
fe8c7f5c | 2103 | mm->end_code = addr; |
028ee4be | 2104 | break; |
028ee4be | 2105 | case PR_SET_MM_START_DATA: |
fe8c7f5c | 2106 | mm->start_data = addr; |
028ee4be | 2107 | break; |
fe8c7f5c CG |
2108 | case PR_SET_MM_END_DATA: |
2109 | mm->end_data = addr; | |
028ee4be CG |
2110 | break; |
2111 | ||
2112 | case PR_SET_MM_START_BRK: | |
2113 | if (addr <= mm->end_data) | |
2114 | goto out; | |
2115 | ||
2116 | if (rlim < RLIM_INFINITY && | |
2117 | (mm->brk - addr) + | |
2118 | (mm->end_data - mm->start_data) > rlim) | |
2119 | goto out; | |
2120 | ||
2121 | mm->start_brk = addr; | |
2122 | break; | |
2123 | ||
2124 | case PR_SET_MM_BRK: | |
2125 | if (addr <= mm->end_data) | |
2126 | goto out; | |
2127 | ||
2128 | if (rlim < RLIM_INFINITY && | |
2129 | (addr - mm->start_brk) + | |
2130 | (mm->end_data - mm->start_data) > rlim) | |
2131 | goto out; | |
2132 | ||
2133 | mm->brk = addr; | |
2134 | break; | |
2135 | ||
fe8c7f5c CG |
2136 | /* |
2137 | * If command line arguments and environment | |
2138 | * are placed somewhere else on stack, we can | |
2139 | * set them up here, ARG_START/END to setup | |
2140 | * command line argumets and ENV_START/END | |
2141 | * for environment. | |
2142 | */ | |
2143 | case PR_SET_MM_START_STACK: | |
2144 | case PR_SET_MM_ARG_START: | |
2145 | case PR_SET_MM_ARG_END: | |
2146 | case PR_SET_MM_ENV_START: | |
2147 | case PR_SET_MM_ENV_END: | |
2148 | if (!vma) { | |
2149 | error = -EFAULT; | |
2150 | goto out; | |
2151 | } | |
fe8c7f5c CG |
2152 | if (opt == PR_SET_MM_START_STACK) |
2153 | mm->start_stack = addr; | |
2154 | else if (opt == PR_SET_MM_ARG_START) | |
2155 | mm->arg_start = addr; | |
2156 | else if (opt == PR_SET_MM_ARG_END) | |
2157 | mm->arg_end = addr; | |
2158 | else if (opt == PR_SET_MM_ENV_START) | |
2159 | mm->env_start = addr; | |
2160 | else if (opt == PR_SET_MM_ENV_END) | |
2161 | mm->env_end = addr; | |
2162 | break; | |
2163 | ||
2164 | /* | |
2165 | * This doesn't move auxiliary vector itself | |
2166 | * since it's pinned to mm_struct, but allow | |
2167 | * to fill vector with new values. It's up | |
2168 | * to a caller to provide sane values here | |
2169 | * otherwise user space tools which use this | |
2170 | * vector might be unhappy. | |
2171 | */ | |
2172 | case PR_SET_MM_AUXV: { | |
2173 | unsigned long user_auxv[AT_VECTOR_SIZE]; | |
2174 | ||
2175 | if (arg4 > sizeof(user_auxv)) | |
2176 | goto out; | |
2177 | up_read(&mm->mmap_sem); | |
2178 | ||
2179 | if (copy_from_user(user_auxv, (const void __user *)addr, arg4)) | |
2180 | return -EFAULT; | |
2181 | ||
2182 | /* Make sure the last entry is always AT_NULL */ | |
2183 | user_auxv[AT_VECTOR_SIZE - 2] = 0; | |
2184 | user_auxv[AT_VECTOR_SIZE - 1] = 0; | |
2185 | ||
2186 | BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv)); | |
2187 | ||
2188 | task_lock(current); | |
2189 | memcpy(mm->saved_auxv, user_auxv, arg4); | |
2190 | task_unlock(current); | |
2191 | ||
2192 | return 0; | |
2193 | } | |
028ee4be | 2194 | default: |
028ee4be CG |
2195 | goto out; |
2196 | } | |
2197 | ||
2198 | error = 0; | |
028ee4be CG |
2199 | out: |
2200 | up_read(&mm->mmap_sem); | |
028ee4be CG |
2201 | return error; |
2202 | } | |
300f786b | 2203 | |
52b36941 | 2204 | #ifdef CONFIG_CHECKPOINT_RESTORE |
300f786b CG |
2205 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2206 | { | |
2207 | return put_user(me->clear_child_tid, tid_addr); | |
2208 | } | |
52b36941 | 2209 | #else |
300f786b CG |
2210 | static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr) |
2211 | { | |
2212 | return -EINVAL; | |
2213 | } | |
028ee4be CG |
2214 | #endif |
2215 | ||
3c2a0909 S |
2216 | #ifdef CONFIG_MMU |
2217 | static int prctl_update_vma_anon_name(struct vm_area_struct *vma, | |
2218 | struct vm_area_struct **prev, | |
2219 | unsigned long start, unsigned long end, | |
2220 | const char __user *name_addr) | |
2221 | { | |
2222 | struct mm_struct * mm = vma->vm_mm; | |
2223 | int error = 0; | |
2224 | pgoff_t pgoff; | |
2225 | ||
2226 | if (name_addr == vma_get_anon_name(vma)) { | |
2227 | *prev = vma; | |
2228 | goto out; | |
2229 | } | |
2230 | ||
2231 | pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT); | |
2232 | *prev = vma_merge(mm, *prev, start, end, vma->vm_flags, vma->anon_vma, | |
2233 | vma->vm_file, pgoff, vma_policy(vma), | |
2234 | name_addr); | |
2235 | if (*prev) { | |
2236 | vma = *prev; | |
2237 | goto success; | |
2238 | } | |
2239 | ||
2240 | *prev = vma; | |
2241 | ||
2242 | if (start != vma->vm_start) { | |
2243 | error = split_vma(mm, vma, start, 1); | |
2244 | if (error) | |
2245 | goto out; | |
2246 | } | |
2247 | ||
2248 | if (end != vma->vm_end) { | |
2249 | error = split_vma(mm, vma, end, 0); | |
2250 | if (error) | |
2251 | goto out; | |
2252 | } | |
2253 | ||
2254 | success: | |
2255 | if (!vma->vm_file) | |
2256 | vma->shared.anon_name = name_addr; | |
2257 | ||
2258 | out: | |
2259 | if (error == -ENOMEM) | |
2260 | error = -EAGAIN; | |
2261 | return error; | |
2262 | } | |
2263 | ||
2264 | static int prctl_set_vma_anon_name(unsigned long start, unsigned long end, | |
2265 | unsigned long arg) | |
2266 | { | |
2267 | unsigned long tmp; | |
2268 | struct vm_area_struct * vma, *prev; | |
2269 | int unmapped_error = 0; | |
2270 | int error = -EINVAL; | |
2271 | ||
2272 | /* | |
2273 | * If the interval [start,end) covers some unmapped address | |
2274 | * ranges, just ignore them, but return -ENOMEM at the end. | |
2275 | * - this matches the handling in madvise. | |
2276 | */ | |
2277 | vma = find_vma_prev(current->mm, start, &prev); | |
2278 | if (vma && start > vma->vm_start) | |
2279 | prev = vma; | |
2280 | ||
2281 | for (;;) { | |
2282 | /* Still start < end. */ | |
2283 | error = -ENOMEM; | |
2284 | if (!vma) | |
2285 | return error; | |
2286 | ||
2287 | /* Here start < (end|vma->vm_end). */ | |
2288 | if (start < vma->vm_start) { | |
2289 | unmapped_error = -ENOMEM; | |
2290 | start = vma->vm_start; | |
2291 | if (start >= end) | |
2292 | return error; | |
2293 | } | |
2294 | ||
2295 | /* Here vma->vm_start <= start < (end|vma->vm_end) */ | |
2296 | tmp = vma->vm_end; | |
2297 | if (end < tmp) | |
2298 | tmp = end; | |
2299 | ||
2300 | /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */ | |
2301 | error = prctl_update_vma_anon_name(vma, &prev, start, tmp, | |
2302 | (const char __user *)arg); | |
2303 | if (error) | |
2304 | return error; | |
2305 | start = tmp; | |
2306 | if (prev && start < prev->vm_end) | |
2307 | start = prev->vm_end; | |
2308 | error = unmapped_error; | |
2309 | if (start >= end) | |
2310 | return error; | |
2311 | if (prev) | |
2312 | vma = prev->vm_next; | |
2313 | else /* madvise_remove dropped mmap_sem */ | |
2314 | vma = find_vma(current->mm, start); | |
2315 | } | |
2316 | } | |
2317 | ||
2318 | static int prctl_set_vma(unsigned long opt, unsigned long start, | |
2319 | unsigned long len_in, unsigned long arg) | |
2320 | { | |
2321 | struct mm_struct *mm = current->mm; | |
2322 | int error; | |
2323 | unsigned long len; | |
2324 | unsigned long end; | |
2325 | ||
2326 | if (start & ~PAGE_MASK) | |
2327 | return -EINVAL; | |
2328 | len = (len_in + ~PAGE_MASK) & PAGE_MASK; | |
2329 | ||
2330 | /* Check to see whether len was rounded up from small -ve to zero */ | |
2331 | if (len_in && !len) | |
2332 | return -EINVAL; | |
2333 | ||
2334 | end = start + len; | |
2335 | if (end < start) | |
2336 | return -EINVAL; | |
2337 | ||
2338 | if (end == start) | |
2339 | return 0; | |
2340 | ||
2341 | down_write(&mm->mmap_sem); | |
2342 | ||
2343 | switch (opt) { | |
2344 | case PR_SET_VMA_ANON_NAME: | |
2345 | error = prctl_set_vma_anon_name(start, end, arg); | |
2346 | break; | |
2347 | default: | |
2348 | error = -EINVAL; | |
2349 | } | |
2350 | ||
2351 | up_write(&mm->mmap_sem); | |
2352 | ||
2353 | return error; | |
2354 | } | |
2355 | #else /* CONFIG_MMU */ | |
2356 | static int prctl_set_vma(unsigned long opt, unsigned long start, | |
2357 | unsigned long len_in, unsigned long arg) | |
2358 | { | |
2359 | return -EINVAL; | |
2360 | } | |
2361 | #endif | |
2362 | ||
c4ea37c2 HC |
2363 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, |
2364 | unsigned long, arg4, unsigned long, arg5) | |
1da177e4 | 2365 | { |
b6dff3ec | 2366 | struct task_struct *me = current; |
3c2a0909 | 2367 | struct task_struct *tsk; |
b6dff3ec DH |
2368 | unsigned char comm[sizeof(me->comm)]; |
2369 | long error; | |
1da177e4 | 2370 | |
d84f4f99 DH |
2371 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
2372 | if (error != -ENOSYS) | |
1da177e4 LT |
2373 | return error; |
2374 | ||
d84f4f99 | 2375 | error = 0; |
1da177e4 | 2376 | switch (option) { |
f3cbd435 AM |
2377 | case PR_SET_PDEATHSIG: |
2378 | if (!valid_signal(arg2)) { | |
2379 | error = -EINVAL; | |
1da177e4 | 2380 | break; |
f3cbd435 AM |
2381 | } |
2382 | me->pdeath_signal = arg2; | |
2383 | break; | |
2384 | case PR_GET_PDEATHSIG: | |
2385 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
2386 | break; | |
2387 | case PR_GET_DUMPABLE: | |
2388 | error = get_dumpable(me->mm); | |
2389 | break; | |
2390 | case PR_SET_DUMPABLE: | |
2391 | if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) { | |
2392 | error = -EINVAL; | |
1da177e4 | 2393 | break; |
f3cbd435 AM |
2394 | } |
2395 | set_dumpable(me->mm, arg2); | |
2396 | break; | |
1da177e4 | 2397 | |
f3cbd435 AM |
2398 | case PR_SET_UNALIGN: |
2399 | error = SET_UNALIGN_CTL(me, arg2); | |
2400 | break; | |
2401 | case PR_GET_UNALIGN: | |
2402 | error = GET_UNALIGN_CTL(me, arg2); | |
2403 | break; | |
2404 | case PR_SET_FPEMU: | |
2405 | error = SET_FPEMU_CTL(me, arg2); | |
2406 | break; | |
2407 | case PR_GET_FPEMU: | |
2408 | error = GET_FPEMU_CTL(me, arg2); | |
2409 | break; | |
2410 | case PR_SET_FPEXC: | |
2411 | error = SET_FPEXC_CTL(me, arg2); | |
2412 | break; | |
2413 | case PR_GET_FPEXC: | |
2414 | error = GET_FPEXC_CTL(me, arg2); | |
2415 | break; | |
2416 | case PR_GET_TIMING: | |
2417 | error = PR_TIMING_STATISTICAL; | |
2418 | break; | |
2419 | case PR_SET_TIMING: | |
2420 | if (arg2 != PR_TIMING_STATISTICAL) | |
2421 | error = -EINVAL; | |
2422 | break; | |
2423 | case PR_SET_NAME: | |
2424 | comm[sizeof(me->comm) - 1] = 0; | |
2425 | if (strncpy_from_user(comm, (char __user *)arg2, | |
2426 | sizeof(me->comm) - 1) < 0) | |
2427 | return -EFAULT; | |
2428 | set_task_comm(me, comm); | |
2429 | proc_comm_connector(me); | |
2430 | break; | |
2431 | case PR_GET_NAME: | |
2432 | get_task_comm(comm, me); | |
2433 | if (copy_to_user((char __user *)arg2, comm, sizeof(comm))) | |
2434 | return -EFAULT; | |
2435 | break; | |
2436 | case PR_GET_ENDIAN: | |
2437 | error = GET_ENDIAN(me, arg2); | |
2438 | break; | |
2439 | case PR_SET_ENDIAN: | |
2440 | error = SET_ENDIAN(me, arg2); | |
2441 | break; | |
2442 | case PR_GET_SECCOMP: | |
2443 | error = prctl_get_seccomp(); | |
2444 | break; | |
2445 | case PR_SET_SECCOMP: | |
2446 | error = prctl_set_seccomp(arg2, (char __user *)arg3); | |
2447 | break; | |
2448 | case PR_GET_TSC: | |
2449 | error = GET_TSC_CTL(arg2); | |
2450 | break; | |
2451 | case PR_SET_TSC: | |
2452 | error = SET_TSC_CTL(arg2); | |
2453 | break; | |
2454 | case PR_TASK_PERF_EVENTS_DISABLE: | |
2455 | error = perf_event_task_disable(); | |
2456 | break; | |
2457 | case PR_TASK_PERF_EVENTS_ENABLE: | |
2458 | error = perf_event_task_enable(); | |
2459 | break; | |
2460 | case PR_GET_TIMERSLACK: | |
2461 | error = current->timer_slack_ns; | |
2462 | break; | |
2463 | case PR_SET_TIMERSLACK: | |
2464 | if (arg2 <= 0) | |
2465 | current->timer_slack_ns = | |
6976675d | 2466 | current->default_timer_slack_ns; |
f3cbd435 AM |
2467 | else |
2468 | current->timer_slack_ns = arg2; | |
2469 | break; | |
2470 | case PR_MCE_KILL: | |
2471 | if (arg4 | arg5) | |
2472 | return -EINVAL; | |
2473 | switch (arg2) { | |
2474 | case PR_MCE_KILL_CLEAR: | |
2475 | if (arg3 != 0) | |
4db96cf0 | 2476 | return -EINVAL; |
f3cbd435 | 2477 | current->flags &= ~PF_MCE_PROCESS; |
4db96cf0 | 2478 | break; |
f3cbd435 AM |
2479 | case PR_MCE_KILL_SET: |
2480 | current->flags |= PF_MCE_PROCESS; | |
2481 | if (arg3 == PR_MCE_KILL_EARLY) | |
2482 | current->flags |= PF_MCE_EARLY; | |
2483 | else if (arg3 == PR_MCE_KILL_LATE) | |
2484 | current->flags &= ~PF_MCE_EARLY; | |
2485 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
2486 | current->flags &= | |
2487 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1087e9b4 | 2488 | else |
259e5e6c | 2489 | return -EINVAL; |
259e5e6c | 2490 | break; |
3c2a0909 S |
2491 | case PR_SET_TIMERSLACK_PID: |
2492 | if (current->pid != (pid_t)arg3 && | |
2493 | !capable(CAP_SYS_NICE)) | |
2494 | return -EPERM; | |
2495 | rcu_read_lock(); | |
2496 | tsk = find_task_by_pid_ns((pid_t)arg3, &init_pid_ns); | |
2497 | if (tsk == NULL) { | |
2498 | rcu_read_unlock(); | |
2499 | return -EINVAL; | |
2500 | } | |
2501 | get_task_struct(tsk); | |
2502 | rcu_read_unlock(); | |
2503 | if (arg2 <= 0) | |
2504 | tsk->timer_slack_ns = | |
2505 | tsk->default_timer_slack_ns; | |
2506 | else | |
2507 | tsk->timer_slack_ns = arg2; | |
2508 | put_task_struct(tsk); | |
2509 | error = 0; | |
2510 | break; | |
1da177e4 | 2511 | default: |
f3cbd435 AM |
2512 | return -EINVAL; |
2513 | } | |
2514 | break; | |
2515 | case PR_MCE_KILL_GET: | |
2516 | if (arg2 | arg3 | arg4 | arg5) | |
2517 | return -EINVAL; | |
2518 | if (current->flags & PF_MCE_PROCESS) | |
2519 | error = (current->flags & PF_MCE_EARLY) ? | |
2520 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
2521 | else | |
2522 | error = PR_MCE_KILL_DEFAULT; | |
2523 | break; | |
2524 | case PR_SET_MM: | |
2525 | error = prctl_set_mm(arg2, arg3, arg4, arg5); | |
2526 | break; | |
2527 | case PR_GET_TID_ADDRESS: | |
2528 | error = prctl_get_tid_address(me, (int __user **)arg2); | |
2529 | break; | |
2530 | case PR_SET_CHILD_SUBREAPER: | |
2531 | me->signal->is_child_subreaper = !!arg2; | |
2532 | break; | |
2533 | case PR_GET_CHILD_SUBREAPER: | |
2534 | error = put_user(me->signal->is_child_subreaper, | |
2535 | (int __user *)arg2); | |
2536 | break; | |
2537 | case PR_SET_NO_NEW_PRIVS: | |
2538 | if (arg2 != 1 || arg3 || arg4 || arg5) | |
2539 | return -EINVAL; | |
2540 | ||
3c2a0909 | 2541 | task_set_no_new_privs(current); |
f3cbd435 AM |
2542 | break; |
2543 | case PR_GET_NO_NEW_PRIVS: | |
2544 | if (arg2 || arg3 || arg4 || arg5) | |
2545 | return -EINVAL; | |
3c2a0909 S |
2546 | return task_no_new_privs(current) ? 1 : 0; |
2547 | case PR_SET_VMA: | |
2548 | error = prctl_set_vma(arg2, arg3, arg4, arg5); | |
2549 | break; | |
f3cbd435 AM |
2550 | default: |
2551 | error = -EINVAL; | |
2552 | break; | |
1da177e4 LT |
2553 | } |
2554 | return error; | |
2555 | } | |
3cfc348b | 2556 | |
836f92ad HC |
2557 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, |
2558 | struct getcpu_cache __user *, unused) | |
3cfc348b AK |
2559 | { |
2560 | int err = 0; | |
2561 | int cpu = raw_smp_processor_id(); | |
2562 | if (cpup) | |
2563 | err |= put_user(cpu, cpup); | |
2564 | if (nodep) | |
2565 | err |= put_user(cpu_to_node(cpu), nodep); | |
3cfc348b AK |
2566 | return err ? -EFAULT : 0; |
2567 | } | |
10a0a8d4 JF |
2568 | |
2569 | char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff"; | |
2570 | ||
2ca067ef | 2571 | static int __orderly_poweroff(bool force) |
10a0a8d4 | 2572 | { |
b57b44ae | 2573 | char **argv; |
10a0a8d4 JF |
2574 | static char *envp[] = { |
2575 | "HOME=/", | |
2576 | "PATH=/sbin:/bin:/usr/sbin:/usr/bin", | |
2577 | NULL | |
2578 | }; | |
b57b44ae | 2579 | int ret; |
10a0a8d4 | 2580 | |
2ca067ef ON |
2581 | argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL); |
2582 | if (argv) { | |
2583 | ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC); | |
2584 | argv_free(argv); | |
2585 | } else { | |
10a0a8d4 | 2586 | printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n", |
2ca067ef ON |
2587 | __func__, poweroff_cmd); |
2588 | ret = -ENOMEM; | |
10a0a8d4 JF |
2589 | } |
2590 | ||
2ca067ef ON |
2591 | if (ret && force) { |
2592 | printk(KERN_WARNING "Failed to start orderly shutdown: " | |
2593 | "forcing the issue\n"); | |
2594 | /* | |
2595 | * I guess this should try to kick off some daemon to sync and | |
2596 | * poweroff asap. Or not even bother syncing if we're doing an | |
2597 | * emergency shutdown? | |
2598 | */ | |
2599 | emergency_sync(); | |
2600 | kernel_power_off(); | |
2601 | } | |
10a0a8d4 | 2602 | |
b57b44ae AM |
2603 | return ret; |
2604 | } | |
2605 | ||
2ca067ef ON |
2606 | static bool poweroff_force; |
2607 | ||
2608 | static void poweroff_work_func(struct work_struct *work) | |
2609 | { | |
2610 | __orderly_poweroff(poweroff_force); | |
2611 | } | |
2612 | ||
2613 | static DECLARE_WORK(poweroff_work, poweroff_work_func); | |
2614 | ||
b57b44ae AM |
2615 | /** |
2616 | * orderly_poweroff - Trigger an orderly system poweroff | |
2617 | * @force: force poweroff if command execution fails | |
2618 | * | |
2619 | * This may be called from any context to trigger a system shutdown. | |
2620 | * If the orderly shutdown fails, it will force an immediate shutdown. | |
2621 | */ | |
2622 | int orderly_poweroff(bool force) | |
2623 | { | |
2ca067ef ON |
2624 | if (force) /* do not override the pending "true" */ |
2625 | poweroff_force = true; | |
2626 | schedule_work(&poweroff_work); | |
2627 | return 0; | |
10a0a8d4 JF |
2628 | } |
2629 | EXPORT_SYMBOL_GPL(orderly_poweroff); | |
4a22f166 SR |
2630 | |
2631 | /** | |
2632 | * do_sysinfo - fill in sysinfo struct | |
2633 | * @info: pointer to buffer to fill | |
2634 | */ | |
2635 | static int do_sysinfo(struct sysinfo *info) | |
2636 | { | |
2637 | unsigned long mem_total, sav_total; | |
2638 | unsigned int mem_unit, bitcount; | |
2639 | struct timespec tp; | |
2640 | ||
2641 | memset(info, 0, sizeof(struct sysinfo)); | |
2642 | ||
2643 | ktime_get_ts(&tp); | |
2644 | monotonic_to_bootbased(&tp); | |
2645 | info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); | |
2646 | ||
2647 | get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); | |
2648 | ||
2649 | info->procs = nr_threads; | |
2650 | ||
2651 | si_meminfo(info); | |
2652 | si_swapinfo(info); | |
2653 | ||
2654 | /* | |
2655 | * If the sum of all the available memory (i.e. ram + swap) | |
2656 | * is less than can be stored in a 32 bit unsigned long then | |
2657 | * we can be binary compatible with 2.2.x kernels. If not, | |
2658 | * well, in that case 2.2.x was broken anyways... | |
2659 | * | |
2660 | * -Erik Andersen <andersee@debian.org> | |
2661 | */ | |
2662 | ||
2663 | mem_total = info->totalram + info->totalswap; | |
2664 | if (mem_total < info->totalram || mem_total < info->totalswap) | |
2665 | goto out; | |
2666 | bitcount = 0; | |
2667 | mem_unit = info->mem_unit; | |
2668 | while (mem_unit > 1) { | |
2669 | bitcount++; | |
2670 | mem_unit >>= 1; | |
2671 | sav_total = mem_total; | |
2672 | mem_total <<= 1; | |
2673 | if (mem_total < sav_total) | |
2674 | goto out; | |
2675 | } | |
2676 | ||
2677 | /* | |
2678 | * If mem_total did not overflow, multiply all memory values by | |
2679 | * info->mem_unit and set it to 1. This leaves things compatible | |
2680 | * with 2.2.x, and also retains compatibility with earlier 2.4.x | |
2681 | * kernels... | |
2682 | */ | |
2683 | ||
2684 | info->mem_unit = 1; | |
2685 | info->totalram <<= bitcount; | |
2686 | info->freeram <<= bitcount; | |
2687 | info->sharedram <<= bitcount; | |
2688 | info->bufferram <<= bitcount; | |
2689 | info->totalswap <<= bitcount; | |
2690 | info->freeswap <<= bitcount; | |
2691 | info->totalhigh <<= bitcount; | |
2692 | info->freehigh <<= bitcount; | |
2693 | ||
2694 | out: | |
2695 | return 0; | |
2696 | } | |
2697 | ||
2698 | SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) | |
2699 | { | |
2700 | struct sysinfo val; | |
2701 | ||
2702 | do_sysinfo(&val); | |
2703 | ||
2704 | if (copy_to_user(info, &val, sizeof(struct sysinfo))) | |
2705 | return -EFAULT; | |
2706 | ||
2707 | return 0; | |
2708 | } | |
2709 | ||
2710 | #ifdef CONFIG_COMPAT | |
2711 | struct compat_sysinfo { | |
2712 | s32 uptime; | |
2713 | u32 loads[3]; | |
2714 | u32 totalram; | |
2715 | u32 freeram; | |
2716 | u32 sharedram; | |
2717 | u32 bufferram; | |
2718 | u32 totalswap; | |
2719 | u32 freeswap; | |
2720 | u16 procs; | |
2721 | u16 pad; | |
2722 | u32 totalhigh; | |
2723 | u32 freehigh; | |
2724 | u32 mem_unit; | |
2725 | char _f[20-2*sizeof(u32)-sizeof(int)]; | |
2726 | }; | |
2727 | ||
2728 | COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info) | |
2729 | { | |
2730 | struct sysinfo s; | |
2731 | ||
2732 | do_sysinfo(&s); | |
2733 | ||
2734 | /* Check to see if any memory value is too large for 32-bit and scale | |
2735 | * down if needed | |
2736 | */ | |
2737 | if ((s.totalram >> 32) || (s.totalswap >> 32)) { | |
2738 | int bitcount = 0; | |
2739 | ||
2740 | while (s.mem_unit < PAGE_SIZE) { | |
2741 | s.mem_unit <<= 1; | |
2742 | bitcount++; | |
2743 | } | |
2744 | ||
2745 | s.totalram >>= bitcount; | |
2746 | s.freeram >>= bitcount; | |
2747 | s.sharedram >>= bitcount; | |
2748 | s.bufferram >>= bitcount; | |
2749 | s.totalswap >>= bitcount; | |
2750 | s.freeswap >>= bitcount; | |
2751 | s.totalhigh >>= bitcount; | |
2752 | s.freehigh >>= bitcount; | |
2753 | } | |
2754 | ||
2755 | if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) || | |
2756 | __put_user(s.uptime, &info->uptime) || | |
2757 | __put_user(s.loads[0], &info->loads[0]) || | |
2758 | __put_user(s.loads[1], &info->loads[1]) || | |
2759 | __put_user(s.loads[2], &info->loads[2]) || | |
2760 | __put_user(s.totalram, &info->totalram) || | |
2761 | __put_user(s.freeram, &info->freeram) || | |
2762 | __put_user(s.sharedram, &info->sharedram) || | |
2763 | __put_user(s.bufferram, &info->bufferram) || | |
2764 | __put_user(s.totalswap, &info->totalswap) || | |
2765 | __put_user(s.freeswap, &info->freeswap) || | |
2766 | __put_user(s.procs, &info->procs) || | |
2767 | __put_user(s.totalhigh, &info->totalhigh) || | |
2768 | __put_user(s.freehigh, &info->freehigh) || | |
2769 | __put_user(s.mem_unit, &info->mem_unit)) | |
2770 | return -EFAULT; | |
2771 | ||
2772 | return 0; | |
2773 | } | |
2774 | #endif /* CONFIG_COMPAT */ |