]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * linux/kernel/sys.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | #include <linux/module.h> | |
8 | #include <linux/mm.h> | |
9 | #include <linux/utsname.h> | |
10 | #include <linux/mman.h> | |
11 | #include <linux/notifier.h> | |
12 | #include <linux/reboot.h> | |
13 | #include <linux/prctl.h> | |
14 | #include <linux/highuid.h> | |
15 | #include <linux/fs.h> | |
16 | #include <linux/perf_event.h> | |
17 | #include <linux/resource.h> | |
18 | #include <linux/kernel.h> | |
19 | #include <linux/kexec.h> | |
20 | #include <linux/workqueue.h> | |
21 | #include <linux/capability.h> | |
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> | |
30 | #include <linux/signal.h> | |
31 | #include <linux/cn_proc.h> | |
32 | #include <linux/getcpu.h> | |
33 | #include <linux/task_io_accounting_ops.h> | |
34 | #include <linux/seccomp.h> | |
35 | #include <linux/cpu.h> | |
36 | #include <linux/personality.h> | |
37 | #include <linux/ptrace.h> | |
38 | #include <linux/fs_struct.h> | |
39 | #include <linux/gfp.h> | |
40 | ||
41 | #include <linux/compat.h> | |
42 | #include <linux/syscalls.h> | |
43 | #include <linux/kprobes.h> | |
44 | #include <linux/user_namespace.h> | |
45 | ||
46 | #include <asm/uaccess.h> | |
47 | #include <asm/io.h> | |
48 | #include <asm/unistd.h> | |
49 | ||
50 | #ifndef SET_UNALIGN_CTL | |
51 | # define SET_UNALIGN_CTL(a,b) (-EINVAL) | |
52 | #endif | |
53 | #ifndef GET_UNALIGN_CTL | |
54 | # define GET_UNALIGN_CTL(a,b) (-EINVAL) | |
55 | #endif | |
56 | #ifndef SET_FPEMU_CTL | |
57 | # define SET_FPEMU_CTL(a,b) (-EINVAL) | |
58 | #endif | |
59 | #ifndef GET_FPEMU_CTL | |
60 | # define GET_FPEMU_CTL(a,b) (-EINVAL) | |
61 | #endif | |
62 | #ifndef SET_FPEXC_CTL | |
63 | # define SET_FPEXC_CTL(a,b) (-EINVAL) | |
64 | #endif | |
65 | #ifndef GET_FPEXC_CTL | |
66 | # define GET_FPEXC_CTL(a,b) (-EINVAL) | |
67 | #endif | |
68 | #ifndef GET_ENDIAN | |
69 | # define GET_ENDIAN(a,b) (-EINVAL) | |
70 | #endif | |
71 | #ifndef SET_ENDIAN | |
72 | # define SET_ENDIAN(a,b) (-EINVAL) | |
73 | #endif | |
74 | #ifndef GET_TSC_CTL | |
75 | # define GET_TSC_CTL(a) (-EINVAL) | |
76 | #endif | |
77 | #ifndef SET_TSC_CTL | |
78 | # define SET_TSC_CTL(a) (-EINVAL) | |
79 | #endif | |
80 | ||
81 | /* | |
82 | * this is where the system-wide overflow UID and GID are defined, for | |
83 | * architectures that now have 32-bit UID/GID but didn't in the past | |
84 | */ | |
85 | ||
86 | int overflowuid = DEFAULT_OVERFLOWUID; | |
87 | int overflowgid = DEFAULT_OVERFLOWGID; | |
88 | ||
89 | #ifdef CONFIG_UID16 | |
90 | EXPORT_SYMBOL(overflowuid); | |
91 | EXPORT_SYMBOL(overflowgid); | |
92 | #endif | |
93 | ||
94 | /* | |
95 | * the same as above, but for filesystems which can only store a 16-bit | |
96 | * UID and GID. as such, this is needed on all architectures | |
97 | */ | |
98 | ||
99 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; | |
100 | int fs_overflowgid = DEFAULT_FS_OVERFLOWUID; | |
101 | ||
102 | EXPORT_SYMBOL(fs_overflowuid); | |
103 | EXPORT_SYMBOL(fs_overflowgid); | |
104 | ||
105 | /* | |
106 | * this indicates whether you can reboot with ctrl-alt-del: the default is yes | |
107 | */ | |
108 | ||
109 | int C_A_D = 1; | |
110 | struct pid *cad_pid; | |
111 | EXPORT_SYMBOL(cad_pid); | |
112 | ||
113 | /* | |
114 | * If set, this is used for preparing the system to power off. | |
115 | */ | |
116 | ||
117 | void (*pm_power_off_prepare)(void); | |
118 | ||
119 | /* | |
120 | * set the priority of a task | |
121 | * - the caller must hold the RCU read lock | |
122 | */ | |
123 | static int set_one_prio(struct task_struct *p, int niceval, int error) | |
124 | { | |
125 | const struct cred *cred = current_cred(), *pcred = __task_cred(p); | |
126 | int no_nice; | |
127 | ||
128 | if (pcred->uid != cred->euid && | |
129 | pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) { | |
130 | error = -EPERM; | |
131 | goto out; | |
132 | } | |
133 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { | |
134 | error = -EACCES; | |
135 | goto out; | |
136 | } | |
137 | no_nice = security_task_setnice(p, niceval); | |
138 | if (no_nice) { | |
139 | error = no_nice; | |
140 | goto out; | |
141 | } | |
142 | if (error == -ESRCH) | |
143 | error = 0; | |
144 | set_user_nice(p, niceval); | |
145 | out: | |
146 | return error; | |
147 | } | |
148 | ||
149 | SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval) | |
150 | { | |
151 | struct task_struct *g, *p; | |
152 | struct user_struct *user; | |
153 | const struct cred *cred = current_cred(); | |
154 | int error = -EINVAL; | |
155 | struct pid *pgrp; | |
156 | ||
157 | if (which > PRIO_USER || which < PRIO_PROCESS) | |
158 | goto out; | |
159 | ||
160 | /* normalize: avoid signed division (rounding problems) */ | |
161 | error = -ESRCH; | |
162 | if (niceval < -20) | |
163 | niceval = -20; | |
164 | if (niceval > 19) | |
165 | niceval = 19; | |
166 | ||
167 | rcu_read_lock(); | |
168 | read_lock(&tasklist_lock); | |
169 | switch (which) { | |
170 | case PRIO_PROCESS: | |
171 | if (who) | |
172 | p = find_task_by_vpid(who); | |
173 | else | |
174 | p = current; | |
175 | if (p) | |
176 | error = set_one_prio(p, niceval, error); | |
177 | break; | |
178 | case PRIO_PGRP: | |
179 | if (who) | |
180 | pgrp = find_vpid(who); | |
181 | else | |
182 | pgrp = task_pgrp(current); | |
183 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
184 | error = set_one_prio(p, niceval, error); | |
185 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
186 | break; | |
187 | case PRIO_USER: | |
188 | user = (struct user_struct *) cred->user; | |
189 | if (!who) | |
190 | who = cred->uid; | |
191 | else if ((who != cred->uid) && | |
192 | !(user = find_user(who))) | |
193 | goto out_unlock; /* No processes for this user */ | |
194 | ||
195 | do_each_thread(g, p) { | |
196 | if (__task_cred(p)->uid == who) | |
197 | error = set_one_prio(p, niceval, error); | |
198 | } while_each_thread(g, p); | |
199 | if (who != cred->uid) | |
200 | free_uid(user); /* For find_user() */ | |
201 | break; | |
202 | } | |
203 | out_unlock: | |
204 | read_unlock(&tasklist_lock); | |
205 | rcu_read_unlock(); | |
206 | out: | |
207 | return error; | |
208 | } | |
209 | ||
210 | /* | |
211 | * Ugh. To avoid negative return values, "getpriority()" will | |
212 | * not return the normal nice-value, but a negated value that | |
213 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) | |
214 | * to stay compatible. | |
215 | */ | |
216 | SYSCALL_DEFINE2(getpriority, int, which, int, who) | |
217 | { | |
218 | struct task_struct *g, *p; | |
219 | struct user_struct *user; | |
220 | const struct cred *cred = current_cred(); | |
221 | long niceval, retval = -ESRCH; | |
222 | struct pid *pgrp; | |
223 | ||
224 | if (which > PRIO_USER || which < PRIO_PROCESS) | |
225 | return -EINVAL; | |
226 | ||
227 | rcu_read_lock(); | |
228 | read_lock(&tasklist_lock); | |
229 | switch (which) { | |
230 | case PRIO_PROCESS: | |
231 | if (who) | |
232 | p = find_task_by_vpid(who); | |
233 | else | |
234 | p = current; | |
235 | if (p) { | |
236 | niceval = 20 - task_nice(p); | |
237 | if (niceval > retval) | |
238 | retval = niceval; | |
239 | } | |
240 | break; | |
241 | case PRIO_PGRP: | |
242 | if (who) | |
243 | pgrp = find_vpid(who); | |
244 | else | |
245 | pgrp = task_pgrp(current); | |
246 | do_each_pid_thread(pgrp, PIDTYPE_PGID, p) { | |
247 | niceval = 20 - task_nice(p); | |
248 | if (niceval > retval) | |
249 | retval = niceval; | |
250 | } while_each_pid_thread(pgrp, PIDTYPE_PGID, p); | |
251 | break; | |
252 | case PRIO_USER: | |
253 | user = (struct user_struct *) cred->user; | |
254 | if (!who) | |
255 | who = cred->uid; | |
256 | else if ((who != cred->uid) && | |
257 | !(user = find_user(who))) | |
258 | goto out_unlock; /* No processes for this user */ | |
259 | ||
260 | do_each_thread(g, p) { | |
261 | if (__task_cred(p)->uid == who) { | |
262 | niceval = 20 - task_nice(p); | |
263 | if (niceval > retval) | |
264 | retval = niceval; | |
265 | } | |
266 | } while_each_thread(g, p); | |
267 | if (who != cred->uid) | |
268 | free_uid(user); /* for find_user() */ | |
269 | break; | |
270 | } | |
271 | out_unlock: | |
272 | read_unlock(&tasklist_lock); | |
273 | rcu_read_unlock(); | |
274 | ||
275 | return retval; | |
276 | } | |
277 | ||
278 | /** | |
279 | * emergency_restart - reboot the system | |
280 | * | |
281 | * Without shutting down any hardware or taking any locks | |
282 | * reboot the system. This is called when we know we are in | |
283 | * trouble so this is our best effort to reboot. This is | |
284 | * safe to call in interrupt context. | |
285 | */ | |
286 | void emergency_restart(void) | |
287 | { | |
288 | machine_emergency_restart(); | |
289 | } | |
290 | EXPORT_SYMBOL_GPL(emergency_restart); | |
291 | ||
292 | void kernel_restart_prepare(char *cmd) | |
293 | { | |
294 | blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd); | |
295 | system_state = SYSTEM_RESTART; | |
296 | device_shutdown(); | |
297 | sysdev_shutdown(); | |
298 | } | |
299 | ||
300 | /** | |
301 | * kernel_restart - reboot the system | |
302 | * @cmd: pointer to buffer containing command to execute for restart | |
303 | * or %NULL | |
304 | * | |
305 | * Shutdown everything and perform a clean reboot. | |
306 | * This is not safe to call in interrupt context. | |
307 | */ | |
308 | void kernel_restart(char *cmd) | |
309 | { | |
310 | kernel_restart_prepare(cmd); | |
311 | if (!cmd) | |
312 | printk(KERN_EMERG "Restarting system.\n"); | |
313 | else | |
314 | printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd); | |
315 | machine_restart(cmd); | |
316 | } | |
317 | EXPORT_SYMBOL_GPL(kernel_restart); | |
318 | ||
319 | static void kernel_shutdown_prepare(enum system_states state) | |
320 | { | |
321 | blocking_notifier_call_chain(&reboot_notifier_list, | |
322 | (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL); | |
323 | system_state = state; | |
324 | device_shutdown(); | |
325 | } | |
326 | /** | |
327 | * kernel_halt - halt the system | |
328 | * | |
329 | * Shutdown everything and perform a clean system halt. | |
330 | */ | |
331 | void kernel_halt(void) | |
332 | { | |
333 | kernel_shutdown_prepare(SYSTEM_HALT); | |
334 | sysdev_shutdown(); | |
335 | printk(KERN_EMERG "System halted.\n"); | |
336 | machine_halt(); | |
337 | } | |
338 | ||
339 | EXPORT_SYMBOL_GPL(kernel_halt); | |
340 | ||
341 | /** | |
342 | * kernel_power_off - power_off the system | |
343 | * | |
344 | * Shutdown everything and perform a clean system power_off. | |
345 | */ | |
346 | void kernel_power_off(void) | |
347 | { | |
348 | kernel_shutdown_prepare(SYSTEM_POWER_OFF); | |
349 | if (pm_power_off_prepare) | |
350 | pm_power_off_prepare(); | |
351 | disable_nonboot_cpus(); | |
352 | sysdev_shutdown(); | |
353 | printk(KERN_EMERG "Power down.\n"); | |
354 | machine_power_off(); | |
355 | } | |
356 | EXPORT_SYMBOL_GPL(kernel_power_off); | |
357 | ||
358 | static DEFINE_MUTEX(reboot_mutex); | |
359 | ||
360 | /* | |
361 | * Reboot system call: for obvious reasons only root may call it, | |
362 | * and even root needs to set up some magic numbers in the registers | |
363 | * so that some mistake won't make this reboot the whole machine. | |
364 | * You can also set the meaning of the ctrl-alt-del-key here. | |
365 | * | |
366 | * reboot doesn't sync: do that yourself before calling this. | |
367 | */ | |
368 | SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd, | |
369 | void __user *, arg) | |
370 | { | |
371 | char buffer[256]; | |
372 | int ret = 0; | |
373 | ||
374 | /* We only trust the superuser with rebooting the system. */ | |
375 | if (!capable(CAP_SYS_BOOT)) | |
376 | return -EPERM; | |
377 | ||
378 | /* For safety, we require "magic" arguments. */ | |
379 | if (magic1 != LINUX_REBOOT_MAGIC1 || | |
380 | (magic2 != LINUX_REBOOT_MAGIC2 && | |
381 | magic2 != LINUX_REBOOT_MAGIC2A && | |
382 | magic2 != LINUX_REBOOT_MAGIC2B && | |
383 | magic2 != LINUX_REBOOT_MAGIC2C)) | |
384 | return -EINVAL; | |
385 | ||
386 | /* Instead of trying to make the power_off code look like | |
387 | * halt when pm_power_off is not set do it the easy way. | |
388 | */ | |
389 | if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off) | |
390 | cmd = LINUX_REBOOT_CMD_HALT; | |
391 | ||
392 | mutex_lock(&reboot_mutex); | |
393 | switch (cmd) { | |
394 | case LINUX_REBOOT_CMD_RESTART: | |
395 | kernel_restart(NULL); | |
396 | break; | |
397 | ||
398 | case LINUX_REBOOT_CMD_CAD_ON: | |
399 | C_A_D = 1; | |
400 | break; | |
401 | ||
402 | case LINUX_REBOOT_CMD_CAD_OFF: | |
403 | C_A_D = 0; | |
404 | break; | |
405 | ||
406 | case LINUX_REBOOT_CMD_HALT: | |
407 | kernel_halt(); | |
408 | do_exit(0); | |
409 | panic("cannot halt"); | |
410 | ||
411 | case LINUX_REBOOT_CMD_POWER_OFF: | |
412 | kernel_power_off(); | |
413 | do_exit(0); | |
414 | break; | |
415 | ||
416 | case LINUX_REBOOT_CMD_RESTART2: | |
417 | if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) { | |
418 | ret = -EFAULT; | |
419 | break; | |
420 | } | |
421 | buffer[sizeof(buffer) - 1] = '\0'; | |
422 | ||
423 | kernel_restart(buffer); | |
424 | break; | |
425 | ||
426 | #ifdef CONFIG_KEXEC | |
427 | case LINUX_REBOOT_CMD_KEXEC: | |
428 | ret = kernel_kexec(); | |
429 | break; | |
430 | #endif | |
431 | ||
432 | #ifdef CONFIG_HIBERNATION | |
433 | case LINUX_REBOOT_CMD_SW_SUSPEND: | |
434 | ret = hibernate(); | |
435 | break; | |
436 | #endif | |
437 | ||
438 | default: | |
439 | ret = -EINVAL; | |
440 | break; | |
441 | } | |
442 | mutex_unlock(&reboot_mutex); | |
443 | return ret; | |
444 | } | |
445 | ||
446 | static void deferred_cad(struct work_struct *dummy) | |
447 | { | |
448 | kernel_restart(NULL); | |
449 | } | |
450 | ||
451 | /* | |
452 | * This function gets called by ctrl-alt-del - ie the keyboard interrupt. | |
453 | * As it's called within an interrupt, it may NOT sync: the only choice | |
454 | * is whether to reboot at once, or just ignore the ctrl-alt-del. | |
455 | */ | |
456 | void ctrl_alt_del(void) | |
457 | { | |
458 | static DECLARE_WORK(cad_work, deferred_cad); | |
459 | ||
460 | if (C_A_D) | |
461 | schedule_work(&cad_work); | |
462 | else | |
463 | kill_cad_pid(SIGINT, 1); | |
464 | } | |
465 | ||
466 | /* | |
467 | * Unprivileged users may change the real gid to the effective gid | |
468 | * or vice versa. (BSD-style) | |
469 | * | |
470 | * If you set the real gid at all, or set the effective gid to a value not | |
471 | * equal to the real gid, then the saved gid is set to the new effective gid. | |
472 | * | |
473 | * This makes it possible for a setgid program to completely drop its | |
474 | * privileges, which is often a useful assertion to make when you are doing | |
475 | * a security audit over a program. | |
476 | * | |
477 | * The general idea is that a program which uses just setregid() will be | |
478 | * 100% compatible with BSD. A program which uses just setgid() will be | |
479 | * 100% compatible with POSIX with saved IDs. | |
480 | * | |
481 | * SMP: There are not races, the GIDs are checked only by filesystem | |
482 | * operations (as far as semantic preservation is concerned). | |
483 | */ | |
484 | SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid) | |
485 | { | |
486 | const struct cred *old; | |
487 | struct cred *new; | |
488 | int retval; | |
489 | ||
490 | new = prepare_creds(); | |
491 | if (!new) | |
492 | return -ENOMEM; | |
493 | old = current_cred(); | |
494 | ||
495 | retval = -EPERM; | |
496 | if (rgid != (gid_t) -1) { | |
497 | if (old->gid == rgid || | |
498 | old->egid == rgid || | |
499 | capable(CAP_SETGID)) | |
500 | new->gid = rgid; | |
501 | else | |
502 | goto error; | |
503 | } | |
504 | if (egid != (gid_t) -1) { | |
505 | if (old->gid == egid || | |
506 | old->egid == egid || | |
507 | old->sgid == egid || | |
508 | capable(CAP_SETGID)) | |
509 | new->egid = egid; | |
510 | else | |
511 | goto error; | |
512 | } | |
513 | ||
514 | if (rgid != (gid_t) -1 || | |
515 | (egid != (gid_t) -1 && egid != old->gid)) | |
516 | new->sgid = new->egid; | |
517 | new->fsgid = new->egid; | |
518 | ||
519 | return commit_creds(new); | |
520 | ||
521 | error: | |
522 | abort_creds(new); | |
523 | return retval; | |
524 | } | |
525 | ||
526 | /* | |
527 | * setgid() is implemented like SysV w/ SAVED_IDS | |
528 | * | |
529 | * SMP: Same implicit races as above. | |
530 | */ | |
531 | SYSCALL_DEFINE1(setgid, gid_t, gid) | |
532 | { | |
533 | const struct cred *old; | |
534 | struct cred *new; | |
535 | int retval; | |
536 | ||
537 | new = prepare_creds(); | |
538 | if (!new) | |
539 | return -ENOMEM; | |
540 | old = current_cred(); | |
541 | ||
542 | retval = -EPERM; | |
543 | if (capable(CAP_SETGID)) | |
544 | new->gid = new->egid = new->sgid = new->fsgid = gid; | |
545 | else if (gid == old->gid || gid == old->sgid) | |
546 | new->egid = new->fsgid = gid; | |
547 | else | |
548 | goto error; | |
549 | ||
550 | return commit_creds(new); | |
551 | ||
552 | error: | |
553 | abort_creds(new); | |
554 | return retval; | |
555 | } | |
556 | ||
557 | /* | |
558 | * change the user struct in a credentials set to match the new UID | |
559 | */ | |
560 | static int set_user(struct cred *new) | |
561 | { | |
562 | struct user_struct *new_user; | |
563 | ||
564 | new_user = alloc_uid(current_user_ns(), new->uid); | |
565 | if (!new_user) | |
566 | return -EAGAIN; | |
567 | ||
568 | if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) && | |
569 | new_user != INIT_USER) { | |
570 | free_uid(new_user); | |
571 | return -EAGAIN; | |
572 | } | |
573 | ||
574 | free_uid(new->user); | |
575 | new->user = new_user; | |
576 | return 0; | |
577 | } | |
578 | ||
579 | /* | |
580 | * Unprivileged users may change the real uid to the effective uid | |
581 | * or vice versa. (BSD-style) | |
582 | * | |
583 | * If you set the real uid at all, or set the effective uid to a value not | |
584 | * equal to the real uid, then the saved uid is set to the new effective uid. | |
585 | * | |
586 | * This makes it possible for a setuid program to completely drop its | |
587 | * privileges, which is often a useful assertion to make when you are doing | |
588 | * a security audit over a program. | |
589 | * | |
590 | * The general idea is that a program which uses just setreuid() will be | |
591 | * 100% compatible with BSD. A program which uses just setuid() will be | |
592 | * 100% compatible with POSIX with saved IDs. | |
593 | */ | |
594 | SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid) | |
595 | { | |
596 | const struct cred *old; | |
597 | struct cred *new; | |
598 | int retval; | |
599 | ||
600 | new = prepare_creds(); | |
601 | if (!new) | |
602 | return -ENOMEM; | |
603 | old = current_cred(); | |
604 | ||
605 | retval = -EPERM; | |
606 | if (ruid != (uid_t) -1) { | |
607 | new->uid = ruid; | |
608 | if (old->uid != ruid && | |
609 | old->euid != ruid && | |
610 | !capable(CAP_SETUID)) | |
611 | goto error; | |
612 | } | |
613 | ||
614 | if (euid != (uid_t) -1) { | |
615 | new->euid = euid; | |
616 | if (old->uid != euid && | |
617 | old->euid != euid && | |
618 | old->suid != euid && | |
619 | !capable(CAP_SETUID)) | |
620 | goto error; | |
621 | } | |
622 | ||
623 | if (new->uid != old->uid) { | |
624 | retval = set_user(new); | |
625 | if (retval < 0) | |
626 | goto error; | |
627 | } | |
628 | if (ruid != (uid_t) -1 || | |
629 | (euid != (uid_t) -1 && euid != old->uid)) | |
630 | new->suid = new->euid; | |
631 | new->fsuid = new->euid; | |
632 | ||
633 | retval = security_task_fix_setuid(new, old, LSM_SETID_RE); | |
634 | if (retval < 0) | |
635 | goto error; | |
636 | ||
637 | return commit_creds(new); | |
638 | ||
639 | error: | |
640 | abort_creds(new); | |
641 | return retval; | |
642 | } | |
643 | ||
644 | /* | |
645 | * setuid() is implemented like SysV with SAVED_IDS | |
646 | * | |
647 | * Note that SAVED_ID's is deficient in that a setuid root program | |
648 | * like sendmail, for example, cannot set its uid to be a normal | |
649 | * user and then switch back, because if you're root, setuid() sets | |
650 | * the saved uid too. If you don't like this, blame the bright people | |
651 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() | |
652 | * will allow a root program to temporarily drop privileges and be able to | |
653 | * regain them by swapping the real and effective uid. | |
654 | */ | |
655 | SYSCALL_DEFINE1(setuid, uid_t, uid) | |
656 | { | |
657 | const struct cred *old; | |
658 | struct cred *new; | |
659 | int retval; | |
660 | ||
661 | new = prepare_creds(); | |
662 | if (!new) | |
663 | return -ENOMEM; | |
664 | old = current_cred(); | |
665 | ||
666 | retval = -EPERM; | |
667 | if (capable(CAP_SETUID)) { | |
668 | new->suid = new->uid = uid; | |
669 | if (uid != old->uid) { | |
670 | retval = set_user(new); | |
671 | if (retval < 0) | |
672 | goto error; | |
673 | } | |
674 | } else if (uid != old->uid && uid != new->suid) { | |
675 | goto error; | |
676 | } | |
677 | ||
678 | new->fsuid = new->euid = uid; | |
679 | ||
680 | retval = security_task_fix_setuid(new, old, LSM_SETID_ID); | |
681 | if (retval < 0) | |
682 | goto error; | |
683 | ||
684 | return commit_creds(new); | |
685 | ||
686 | error: | |
687 | abort_creds(new); | |
688 | return retval; | |
689 | } | |
690 | ||
691 | ||
692 | /* | |
693 | * This function implements a generic ability to update ruid, euid, | |
694 | * and suid. This allows you to implement the 4.4 compatible seteuid(). | |
695 | */ | |
696 | SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid) | |
697 | { | |
698 | const struct cred *old; | |
699 | struct cred *new; | |
700 | int retval; | |
701 | ||
702 | new = prepare_creds(); | |
703 | if (!new) | |
704 | return -ENOMEM; | |
705 | ||
706 | old = current_cred(); | |
707 | ||
708 | retval = -EPERM; | |
709 | if (!capable(CAP_SETUID)) { | |
710 | if (ruid != (uid_t) -1 && ruid != old->uid && | |
711 | ruid != old->euid && ruid != old->suid) | |
712 | goto error; | |
713 | if (euid != (uid_t) -1 && euid != old->uid && | |
714 | euid != old->euid && euid != old->suid) | |
715 | goto error; | |
716 | if (suid != (uid_t) -1 && suid != old->uid && | |
717 | suid != old->euid && suid != old->suid) | |
718 | goto error; | |
719 | } | |
720 | ||
721 | if (ruid != (uid_t) -1) { | |
722 | new->uid = ruid; | |
723 | if (ruid != old->uid) { | |
724 | retval = set_user(new); | |
725 | if (retval < 0) | |
726 | goto error; | |
727 | } | |
728 | } | |
729 | if (euid != (uid_t) -1) | |
730 | new->euid = euid; | |
731 | if (suid != (uid_t) -1) | |
732 | new->suid = suid; | |
733 | new->fsuid = new->euid; | |
734 | ||
735 | retval = security_task_fix_setuid(new, old, LSM_SETID_RES); | |
736 | if (retval < 0) | |
737 | goto error; | |
738 | ||
739 | return commit_creds(new); | |
740 | ||
741 | error: | |
742 | abort_creds(new); | |
743 | return retval; | |
744 | } | |
745 | ||
746 | SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid) | |
747 | { | |
748 | const struct cred *cred = current_cred(); | |
749 | int retval; | |
750 | ||
751 | if (!(retval = put_user(cred->uid, ruid)) && | |
752 | !(retval = put_user(cred->euid, euid))) | |
753 | retval = put_user(cred->suid, suid); | |
754 | ||
755 | return retval; | |
756 | } | |
757 | ||
758 | /* | |
759 | * Same as above, but for rgid, egid, sgid. | |
760 | */ | |
761 | SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid) | |
762 | { | |
763 | const struct cred *old; | |
764 | struct cred *new; | |
765 | int retval; | |
766 | ||
767 | new = prepare_creds(); | |
768 | if (!new) | |
769 | return -ENOMEM; | |
770 | old = current_cred(); | |
771 | ||
772 | retval = -EPERM; | |
773 | if (!capable(CAP_SETGID)) { | |
774 | if (rgid != (gid_t) -1 && rgid != old->gid && | |
775 | rgid != old->egid && rgid != old->sgid) | |
776 | goto error; | |
777 | if (egid != (gid_t) -1 && egid != old->gid && | |
778 | egid != old->egid && egid != old->sgid) | |
779 | goto error; | |
780 | if (sgid != (gid_t) -1 && sgid != old->gid && | |
781 | sgid != old->egid && sgid != old->sgid) | |
782 | goto error; | |
783 | } | |
784 | ||
785 | if (rgid != (gid_t) -1) | |
786 | new->gid = rgid; | |
787 | if (egid != (gid_t) -1) | |
788 | new->egid = egid; | |
789 | if (sgid != (gid_t) -1) | |
790 | new->sgid = sgid; | |
791 | new->fsgid = new->egid; | |
792 | ||
793 | return commit_creds(new); | |
794 | ||
795 | error: | |
796 | abort_creds(new); | |
797 | return retval; | |
798 | } | |
799 | ||
800 | SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid) | |
801 | { | |
802 | const struct cred *cred = current_cred(); | |
803 | int retval; | |
804 | ||
805 | if (!(retval = put_user(cred->gid, rgid)) && | |
806 | !(retval = put_user(cred->egid, egid))) | |
807 | retval = put_user(cred->sgid, sgid); | |
808 | ||
809 | return retval; | |
810 | } | |
811 | ||
812 | ||
813 | /* | |
814 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This | |
815 | * is used for "access()" and for the NFS daemon (letting nfsd stay at | |
816 | * whatever uid it wants to). It normally shadows "euid", except when | |
817 | * explicitly set by setfsuid() or for access.. | |
818 | */ | |
819 | SYSCALL_DEFINE1(setfsuid, uid_t, uid) | |
820 | { | |
821 | const struct cred *old; | |
822 | struct cred *new; | |
823 | uid_t old_fsuid; | |
824 | ||
825 | new = prepare_creds(); | |
826 | if (!new) | |
827 | return current_fsuid(); | |
828 | old = current_cred(); | |
829 | old_fsuid = old->fsuid; | |
830 | ||
831 | if (uid == old->uid || uid == old->euid || | |
832 | uid == old->suid || uid == old->fsuid || | |
833 | capable(CAP_SETUID)) { | |
834 | if (uid != old_fsuid) { | |
835 | new->fsuid = uid; | |
836 | if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0) | |
837 | goto change_okay; | |
838 | } | |
839 | } | |
840 | ||
841 | abort_creds(new); | |
842 | return old_fsuid; | |
843 | ||
844 | change_okay: | |
845 | commit_creds(new); | |
846 | return old_fsuid; | |
847 | } | |
848 | ||
849 | /* | |
850 | * Samma på svenska.. | |
851 | */ | |
852 | SYSCALL_DEFINE1(setfsgid, gid_t, gid) | |
853 | { | |
854 | const struct cred *old; | |
855 | struct cred *new; | |
856 | gid_t old_fsgid; | |
857 | ||
858 | new = prepare_creds(); | |
859 | if (!new) | |
860 | return current_fsgid(); | |
861 | old = current_cred(); | |
862 | old_fsgid = old->fsgid; | |
863 | ||
864 | if (gid == old->gid || gid == old->egid || | |
865 | gid == old->sgid || gid == old->fsgid || | |
866 | capable(CAP_SETGID)) { | |
867 | if (gid != old_fsgid) { | |
868 | new->fsgid = gid; | |
869 | goto change_okay; | |
870 | } | |
871 | } | |
872 | ||
873 | abort_creds(new); | |
874 | return old_fsgid; | |
875 | ||
876 | change_okay: | |
877 | commit_creds(new); | |
878 | return old_fsgid; | |
879 | } | |
880 | ||
881 | void do_sys_times(struct tms *tms) | |
882 | { | |
883 | cputime_t tgutime, tgstime, cutime, cstime; | |
884 | ||
885 | spin_lock_irq(¤t->sighand->siglock); | |
886 | thread_group_times(current, &tgutime, &tgstime); | |
887 | cutime = current->signal->cutime; | |
888 | cstime = current->signal->cstime; | |
889 | spin_unlock_irq(¤t->sighand->siglock); | |
890 | tms->tms_utime = cputime_to_clock_t(tgutime); | |
891 | tms->tms_stime = cputime_to_clock_t(tgstime); | |
892 | tms->tms_cutime = cputime_to_clock_t(cutime); | |
893 | tms->tms_cstime = cputime_to_clock_t(cstime); | |
894 | } | |
895 | ||
896 | SYSCALL_DEFINE1(times, struct tms __user *, tbuf) | |
897 | { | |
898 | if (tbuf) { | |
899 | struct tms tmp; | |
900 | ||
901 | do_sys_times(&tmp); | |
902 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) | |
903 | return -EFAULT; | |
904 | } | |
905 | force_successful_syscall_return(); | |
906 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); | |
907 | } | |
908 | ||
909 | /* | |
910 | * This needs some heavy checking ... | |
911 | * I just haven't the stomach for it. I also don't fully | |
912 | * understand sessions/pgrp etc. Let somebody who does explain it. | |
913 | * | |
914 | * OK, I think I have the protection semantics right.... this is really | |
915 | * only important on a multi-user system anyway, to make sure one user | |
916 | * can't send a signal to a process owned by another. -TYT, 12/12/91 | |
917 | * | |
918 | * Auch. Had to add the 'did_exec' flag to conform completely to POSIX. | |
919 | * LBT 04.03.94 | |
920 | */ | |
921 | SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid) | |
922 | { | |
923 | struct task_struct *p; | |
924 | struct task_struct *group_leader = current->group_leader; | |
925 | struct pid *pgrp; | |
926 | int err; | |
927 | ||
928 | if (!pid) | |
929 | pid = task_pid_vnr(group_leader); | |
930 | if (!pgid) | |
931 | pgid = pid; | |
932 | if (pgid < 0) | |
933 | return -EINVAL; | |
934 | ||
935 | /* From this point forward we keep holding onto the tasklist lock | |
936 | * so that our parent does not change from under us. -DaveM | |
937 | */ | |
938 | write_lock_irq(&tasklist_lock); | |
939 | ||
940 | err = -ESRCH; | |
941 | p = find_task_by_vpid(pid); | |
942 | if (!p) | |
943 | goto out; | |
944 | ||
945 | err = -EINVAL; | |
946 | if (!thread_group_leader(p)) | |
947 | goto out; | |
948 | ||
949 | if (same_thread_group(p->real_parent, group_leader)) { | |
950 | err = -EPERM; | |
951 | if (task_session(p) != task_session(group_leader)) | |
952 | goto out; | |
953 | err = -EACCES; | |
954 | if (p->did_exec) | |
955 | goto out; | |
956 | } else { | |
957 | err = -ESRCH; | |
958 | if (p != group_leader) | |
959 | goto out; | |
960 | } | |
961 | ||
962 | err = -EPERM; | |
963 | if (p->signal->leader) | |
964 | goto out; | |
965 | ||
966 | pgrp = task_pid(p); | |
967 | if (pgid != pid) { | |
968 | struct task_struct *g; | |
969 | ||
970 | pgrp = find_vpid(pgid); | |
971 | g = pid_task(pgrp, PIDTYPE_PGID); | |
972 | if (!g || task_session(g) != task_session(group_leader)) | |
973 | goto out; | |
974 | } | |
975 | ||
976 | err = security_task_setpgid(p, pgid); | |
977 | if (err) | |
978 | goto out; | |
979 | ||
980 | if (task_pgrp(p) != pgrp) | |
981 | change_pid(p, PIDTYPE_PGID, pgrp); | |
982 | ||
983 | err = 0; | |
984 | out: | |
985 | /* All paths lead to here, thus we are safe. -DaveM */ | |
986 | write_unlock_irq(&tasklist_lock); | |
987 | return err; | |
988 | } | |
989 | ||
990 | SYSCALL_DEFINE1(getpgid, pid_t, pid) | |
991 | { | |
992 | struct task_struct *p; | |
993 | struct pid *grp; | |
994 | int retval; | |
995 | ||
996 | rcu_read_lock(); | |
997 | if (!pid) | |
998 | grp = task_pgrp(current); | |
999 | else { | |
1000 | retval = -ESRCH; | |
1001 | p = find_task_by_vpid(pid); | |
1002 | if (!p) | |
1003 | goto out; | |
1004 | grp = task_pgrp(p); | |
1005 | if (!grp) | |
1006 | goto out; | |
1007 | ||
1008 | retval = security_task_getpgid(p); | |
1009 | if (retval) | |
1010 | goto out; | |
1011 | } | |
1012 | retval = pid_vnr(grp); | |
1013 | out: | |
1014 | rcu_read_unlock(); | |
1015 | return retval; | |
1016 | } | |
1017 | ||
1018 | #ifdef __ARCH_WANT_SYS_GETPGRP | |
1019 | ||
1020 | SYSCALL_DEFINE0(getpgrp) | |
1021 | { | |
1022 | return sys_getpgid(0); | |
1023 | } | |
1024 | ||
1025 | #endif | |
1026 | ||
1027 | SYSCALL_DEFINE1(getsid, pid_t, pid) | |
1028 | { | |
1029 | struct task_struct *p; | |
1030 | struct pid *sid; | |
1031 | int retval; | |
1032 | ||
1033 | rcu_read_lock(); | |
1034 | if (!pid) | |
1035 | sid = task_session(current); | |
1036 | else { | |
1037 | retval = -ESRCH; | |
1038 | p = find_task_by_vpid(pid); | |
1039 | if (!p) | |
1040 | goto out; | |
1041 | sid = task_session(p); | |
1042 | if (!sid) | |
1043 | goto out; | |
1044 | ||
1045 | retval = security_task_getsid(p); | |
1046 | if (retval) | |
1047 | goto out; | |
1048 | } | |
1049 | retval = pid_vnr(sid); | |
1050 | out: | |
1051 | rcu_read_unlock(); | |
1052 | return retval; | |
1053 | } | |
1054 | ||
1055 | SYSCALL_DEFINE0(setsid) | |
1056 | { | |
1057 | struct task_struct *group_leader = current->group_leader; | |
1058 | struct pid *sid = task_pid(group_leader); | |
1059 | pid_t session = pid_vnr(sid); | |
1060 | int err = -EPERM; | |
1061 | ||
1062 | write_lock_irq(&tasklist_lock); | |
1063 | /* Fail if I am already a session leader */ | |
1064 | if (group_leader->signal->leader) | |
1065 | goto out; | |
1066 | ||
1067 | /* Fail if a process group id already exists that equals the | |
1068 | * proposed session id. | |
1069 | */ | |
1070 | if (pid_task(sid, PIDTYPE_PGID)) | |
1071 | goto out; | |
1072 | ||
1073 | group_leader->signal->leader = 1; | |
1074 | __set_special_pids(sid); | |
1075 | ||
1076 | proc_clear_tty(group_leader); | |
1077 | ||
1078 | err = session; | |
1079 | out: | |
1080 | write_unlock_irq(&tasklist_lock); | |
1081 | if (err > 0) | |
1082 | proc_sid_connector(group_leader); | |
1083 | return err; | |
1084 | } | |
1085 | ||
1086 | DECLARE_RWSEM(uts_sem); | |
1087 | ||
1088 | #ifdef COMPAT_UTS_MACHINE | |
1089 | #define override_architecture(name) \ | |
1090 | (personality(current->personality) == PER_LINUX32 && \ | |
1091 | copy_to_user(name->machine, COMPAT_UTS_MACHINE, \ | |
1092 | sizeof(COMPAT_UTS_MACHINE))) | |
1093 | #else | |
1094 | #define override_architecture(name) 0 | |
1095 | #endif | |
1096 | ||
1097 | SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name) | |
1098 | { | |
1099 | int errno = 0; | |
1100 | ||
1101 | down_read(&uts_sem); | |
1102 | if (copy_to_user(name, utsname(), sizeof *name)) | |
1103 | errno = -EFAULT; | |
1104 | up_read(&uts_sem); | |
1105 | ||
1106 | if (!errno && override_architecture(name)) | |
1107 | errno = -EFAULT; | |
1108 | return errno; | |
1109 | } | |
1110 | ||
1111 | #ifdef __ARCH_WANT_SYS_OLD_UNAME | |
1112 | /* | |
1113 | * Old cruft | |
1114 | */ | |
1115 | SYSCALL_DEFINE1(uname, struct old_utsname __user *, name) | |
1116 | { | |
1117 | int error = 0; | |
1118 | ||
1119 | if (!name) | |
1120 | return -EFAULT; | |
1121 | ||
1122 | down_read(&uts_sem); | |
1123 | if (copy_to_user(name, utsname(), sizeof(*name))) | |
1124 | error = -EFAULT; | |
1125 | up_read(&uts_sem); | |
1126 | ||
1127 | if (!error && override_architecture(name)) | |
1128 | error = -EFAULT; | |
1129 | return error; | |
1130 | } | |
1131 | ||
1132 | SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name) | |
1133 | { | |
1134 | int error; | |
1135 | ||
1136 | if (!name) | |
1137 | return -EFAULT; | |
1138 | if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname))) | |
1139 | return -EFAULT; | |
1140 | ||
1141 | down_read(&uts_sem); | |
1142 | error = __copy_to_user(&name->sysname, &utsname()->sysname, | |
1143 | __OLD_UTS_LEN); | |
1144 | error |= __put_user(0, name->sysname + __OLD_UTS_LEN); | |
1145 | error |= __copy_to_user(&name->nodename, &utsname()->nodename, | |
1146 | __OLD_UTS_LEN); | |
1147 | error |= __put_user(0, name->nodename + __OLD_UTS_LEN); | |
1148 | error |= __copy_to_user(&name->release, &utsname()->release, | |
1149 | __OLD_UTS_LEN); | |
1150 | error |= __put_user(0, name->release + __OLD_UTS_LEN); | |
1151 | error |= __copy_to_user(&name->version, &utsname()->version, | |
1152 | __OLD_UTS_LEN); | |
1153 | error |= __put_user(0, name->version + __OLD_UTS_LEN); | |
1154 | error |= __copy_to_user(&name->machine, &utsname()->machine, | |
1155 | __OLD_UTS_LEN); | |
1156 | error |= __put_user(0, name->machine + __OLD_UTS_LEN); | |
1157 | up_read(&uts_sem); | |
1158 | ||
1159 | if (!error && override_architecture(name)) | |
1160 | error = -EFAULT; | |
1161 | return error ? -EFAULT : 0; | |
1162 | } | |
1163 | #endif | |
1164 | ||
1165 | SYSCALL_DEFINE2(sethostname, char __user *, name, int, len) | |
1166 | { | |
1167 | int errno; | |
1168 | char tmp[__NEW_UTS_LEN]; | |
1169 | ||
1170 | if (!capable(CAP_SYS_ADMIN)) | |
1171 | return -EPERM; | |
1172 | if (len < 0 || len > __NEW_UTS_LEN) | |
1173 | return -EINVAL; | |
1174 | down_write(&uts_sem); | |
1175 | errno = -EFAULT; | |
1176 | if (!copy_from_user(tmp, name, len)) { | |
1177 | struct new_utsname *u = utsname(); | |
1178 | ||
1179 | memcpy(u->nodename, tmp, len); | |
1180 | memset(u->nodename + len, 0, sizeof(u->nodename) - len); | |
1181 | errno = 0; | |
1182 | } | |
1183 | up_write(&uts_sem); | |
1184 | return errno; | |
1185 | } | |
1186 | ||
1187 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME | |
1188 | ||
1189 | SYSCALL_DEFINE2(gethostname, char __user *, name, int, len) | |
1190 | { | |
1191 | int i, errno; | |
1192 | struct new_utsname *u; | |
1193 | ||
1194 | if (len < 0) | |
1195 | return -EINVAL; | |
1196 | down_read(&uts_sem); | |
1197 | u = utsname(); | |
1198 | i = 1 + strlen(u->nodename); | |
1199 | if (i > len) | |
1200 | i = len; | |
1201 | errno = 0; | |
1202 | if (copy_to_user(name, u->nodename, i)) | |
1203 | errno = -EFAULT; | |
1204 | up_read(&uts_sem); | |
1205 | return errno; | |
1206 | } | |
1207 | ||
1208 | #endif | |
1209 | ||
1210 | /* | |
1211 | * Only setdomainname; getdomainname can be implemented by calling | |
1212 | * uname() | |
1213 | */ | |
1214 | SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len) | |
1215 | { | |
1216 | int errno; | |
1217 | char tmp[__NEW_UTS_LEN]; | |
1218 | ||
1219 | if (!capable(CAP_SYS_ADMIN)) | |
1220 | return -EPERM; | |
1221 | if (len < 0 || len > __NEW_UTS_LEN) | |
1222 | return -EINVAL; | |
1223 | ||
1224 | down_write(&uts_sem); | |
1225 | errno = -EFAULT; | |
1226 | if (!copy_from_user(tmp, name, len)) { | |
1227 | struct new_utsname *u = utsname(); | |
1228 | ||
1229 | memcpy(u->domainname, tmp, len); | |
1230 | memset(u->domainname + len, 0, sizeof(u->domainname) - len); | |
1231 | errno = 0; | |
1232 | } | |
1233 | up_write(&uts_sem); | |
1234 | return errno; | |
1235 | } | |
1236 | ||
1237 | SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim) | |
1238 | { | |
1239 | struct rlimit value; | |
1240 | int ret; | |
1241 | ||
1242 | ret = do_prlimit(current, resource, NULL, &value); | |
1243 | if (!ret) | |
1244 | ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; | |
1245 | ||
1246 | return ret; | |
1247 | } | |
1248 | ||
1249 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT | |
1250 | ||
1251 | /* | |
1252 | * Back compatibility for getrlimit. Needed for some apps. | |
1253 | */ | |
1254 | ||
1255 | SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource, | |
1256 | struct rlimit __user *, rlim) | |
1257 | { | |
1258 | struct rlimit x; | |
1259 | if (resource >= RLIM_NLIMITS) | |
1260 | return -EINVAL; | |
1261 | ||
1262 | task_lock(current->group_leader); | |
1263 | x = current->signal->rlim[resource]; | |
1264 | task_unlock(current->group_leader); | |
1265 | if (x.rlim_cur > 0x7FFFFFFF) | |
1266 | x.rlim_cur = 0x7FFFFFFF; | |
1267 | if (x.rlim_max > 0x7FFFFFFF) | |
1268 | x.rlim_max = 0x7FFFFFFF; | |
1269 | return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0; | |
1270 | } | |
1271 | ||
1272 | #endif | |
1273 | ||
1274 | static inline bool rlim64_is_infinity(__u64 rlim64) | |
1275 | { | |
1276 | #if BITS_PER_LONG < 64 | |
1277 | return rlim64 >= ULONG_MAX; | |
1278 | #else | |
1279 | return rlim64 == RLIM64_INFINITY; | |
1280 | #endif | |
1281 | } | |
1282 | ||
1283 | static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64) | |
1284 | { | |
1285 | if (rlim->rlim_cur == RLIM_INFINITY) | |
1286 | rlim64->rlim_cur = RLIM64_INFINITY; | |
1287 | else | |
1288 | rlim64->rlim_cur = rlim->rlim_cur; | |
1289 | if (rlim->rlim_max == RLIM_INFINITY) | |
1290 | rlim64->rlim_max = RLIM64_INFINITY; | |
1291 | else | |
1292 | rlim64->rlim_max = rlim->rlim_max; | |
1293 | } | |
1294 | ||
1295 | static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim) | |
1296 | { | |
1297 | if (rlim64_is_infinity(rlim64->rlim_cur)) | |
1298 | rlim->rlim_cur = RLIM_INFINITY; | |
1299 | else | |
1300 | rlim->rlim_cur = (unsigned long)rlim64->rlim_cur; | |
1301 | if (rlim64_is_infinity(rlim64->rlim_max)) | |
1302 | rlim->rlim_max = RLIM_INFINITY; | |
1303 | else | |
1304 | rlim->rlim_max = (unsigned long)rlim64->rlim_max; | |
1305 | } | |
1306 | ||
1307 | /* make sure you are allowed to change @tsk limits before calling this */ | |
1308 | int do_prlimit(struct task_struct *tsk, unsigned int resource, | |
1309 | struct rlimit *new_rlim, struct rlimit *old_rlim) | |
1310 | { | |
1311 | struct rlimit *rlim; | |
1312 | int retval = 0; | |
1313 | ||
1314 | if (resource >= RLIM_NLIMITS) | |
1315 | return -EINVAL; | |
1316 | if (new_rlim) { | |
1317 | if (new_rlim->rlim_cur > new_rlim->rlim_max) | |
1318 | return -EINVAL; | |
1319 | if (resource == RLIMIT_NOFILE && | |
1320 | new_rlim->rlim_max > sysctl_nr_open) | |
1321 | return -EPERM; | |
1322 | } | |
1323 | ||
1324 | /* protect tsk->signal and tsk->sighand from disappearing */ | |
1325 | read_lock(&tasklist_lock); | |
1326 | if (!tsk->sighand) { | |
1327 | retval = -ESRCH; | |
1328 | goto out; | |
1329 | } | |
1330 | ||
1331 | rlim = tsk->signal->rlim + resource; | |
1332 | task_lock(tsk->group_leader); | |
1333 | if (new_rlim) { | |
1334 | if (new_rlim->rlim_max > rlim->rlim_max && | |
1335 | !capable(CAP_SYS_RESOURCE)) | |
1336 | retval = -EPERM; | |
1337 | if (!retval) | |
1338 | retval = security_task_setrlimit(tsk->group_leader, | |
1339 | resource, new_rlim); | |
1340 | if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) { | |
1341 | /* | |
1342 | * The caller is asking for an immediate RLIMIT_CPU | |
1343 | * expiry. But we use the zero value to mean "it was | |
1344 | * never set". So let's cheat and make it one second | |
1345 | * instead | |
1346 | */ | |
1347 | new_rlim->rlim_cur = 1; | |
1348 | } | |
1349 | } | |
1350 | if (!retval) { | |
1351 | if (old_rlim) | |
1352 | *old_rlim = *rlim; | |
1353 | if (new_rlim) | |
1354 | *rlim = *new_rlim; | |
1355 | } | |
1356 | task_unlock(tsk->group_leader); | |
1357 | ||
1358 | /* | |
1359 | * RLIMIT_CPU handling. Note that the kernel fails to return an error | |
1360 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a | |
1361 | * very long-standing error, and fixing it now risks breakage of | |
1362 | * applications, so we live with it | |
1363 | */ | |
1364 | if (!retval && new_rlim && resource == RLIMIT_CPU && | |
1365 | new_rlim->rlim_cur != RLIM_INFINITY) | |
1366 | update_rlimit_cpu(tsk, new_rlim->rlim_cur); | |
1367 | out: | |
1368 | read_unlock(&tasklist_lock); | |
1369 | return retval; | |
1370 | } | |
1371 | ||
1372 | /* rcu lock must be held */ | |
1373 | static int check_prlimit_permission(struct task_struct *task) | |
1374 | { | |
1375 | const struct cred *cred = current_cred(), *tcred; | |
1376 | ||
1377 | tcred = __task_cred(task); | |
1378 | if ((cred->uid != tcred->euid || | |
1379 | cred->uid != tcred->suid || | |
1380 | cred->uid != tcred->uid || | |
1381 | cred->gid != tcred->egid || | |
1382 | cred->gid != tcred->sgid || | |
1383 | cred->gid != tcred->gid) && | |
1384 | !capable(CAP_SYS_RESOURCE)) { | |
1385 | return -EPERM; | |
1386 | } | |
1387 | ||
1388 | return 0; | |
1389 | } | |
1390 | ||
1391 | SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource, | |
1392 | const struct rlimit64 __user *, new_rlim, | |
1393 | struct rlimit64 __user *, old_rlim) | |
1394 | { | |
1395 | struct rlimit64 old64, new64; | |
1396 | struct rlimit old, new; | |
1397 | struct task_struct *tsk; | |
1398 | int ret; | |
1399 | ||
1400 | if (new_rlim) { | |
1401 | if (copy_from_user(&new64, new_rlim, sizeof(new64))) | |
1402 | return -EFAULT; | |
1403 | rlim64_to_rlim(&new64, &new); | |
1404 | } | |
1405 | ||
1406 | rcu_read_lock(); | |
1407 | tsk = pid ? find_task_by_vpid(pid) : current; | |
1408 | if (!tsk) { | |
1409 | rcu_read_unlock(); | |
1410 | return -ESRCH; | |
1411 | } | |
1412 | ret = check_prlimit_permission(tsk); | |
1413 | if (ret) { | |
1414 | rcu_read_unlock(); | |
1415 | return ret; | |
1416 | } | |
1417 | get_task_struct(tsk); | |
1418 | rcu_read_unlock(); | |
1419 | ||
1420 | ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL, | |
1421 | old_rlim ? &old : NULL); | |
1422 | ||
1423 | if (!ret && old_rlim) { | |
1424 | rlim_to_rlim64(&old, &old64); | |
1425 | if (copy_to_user(old_rlim, &old64, sizeof(old64))) | |
1426 | ret = -EFAULT; | |
1427 | } | |
1428 | ||
1429 | put_task_struct(tsk); | |
1430 | return ret; | |
1431 | } | |
1432 | ||
1433 | SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim) | |
1434 | { | |
1435 | struct rlimit new_rlim; | |
1436 | ||
1437 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) | |
1438 | return -EFAULT; | |
1439 | return do_prlimit(current, resource, &new_rlim, NULL); | |
1440 | } | |
1441 | ||
1442 | /* | |
1443 | * It would make sense to put struct rusage in the task_struct, | |
1444 | * except that would make the task_struct be *really big*. After | |
1445 | * task_struct gets moved into malloc'ed memory, it would | |
1446 | * make sense to do this. It will make moving the rest of the information | |
1447 | * a lot simpler! (Which we're not doing right now because we're not | |
1448 | * measuring them yet). | |
1449 | * | |
1450 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have | |
1451 | * races with threads incrementing their own counters. But since word | |
1452 | * reads are atomic, we either get new values or old values and we don't | |
1453 | * care which for the sums. We always take the siglock to protect reading | |
1454 | * the c* fields from p->signal from races with exit.c updating those | |
1455 | * fields when reaping, so a sample either gets all the additions of a | |
1456 | * given child after it's reaped, or none so this sample is before reaping. | |
1457 | * | |
1458 | * Locking: | |
1459 | * We need to take the siglock for CHILDEREN, SELF and BOTH | |
1460 | * for the cases current multithreaded, non-current single threaded | |
1461 | * non-current multithreaded. Thread traversal is now safe with | |
1462 | * the siglock held. | |
1463 | * Strictly speaking, we donot need to take the siglock if we are current and | |
1464 | * single threaded, as no one else can take our signal_struct away, no one | |
1465 | * else can reap the children to update signal->c* counters, and no one else | |
1466 | * can race with the signal-> fields. If we do not take any lock, the | |
1467 | * signal-> fields could be read out of order while another thread was just | |
1468 | * exiting. So we should place a read memory barrier when we avoid the lock. | |
1469 | * On the writer side, write memory barrier is implied in __exit_signal | |
1470 | * as __exit_signal releases the siglock spinlock after updating the signal-> | |
1471 | * fields. But we don't do this yet to keep things simple. | |
1472 | * | |
1473 | */ | |
1474 | ||
1475 | static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r) | |
1476 | { | |
1477 | r->ru_nvcsw += t->nvcsw; | |
1478 | r->ru_nivcsw += t->nivcsw; | |
1479 | r->ru_minflt += t->min_flt; | |
1480 | r->ru_majflt += t->maj_flt; | |
1481 | r->ru_inblock += task_io_get_inblock(t); | |
1482 | r->ru_oublock += task_io_get_oublock(t); | |
1483 | } | |
1484 | ||
1485 | static void k_getrusage(struct task_struct *p, int who, struct rusage *r) | |
1486 | { | |
1487 | struct task_struct *t; | |
1488 | unsigned long flags; | |
1489 | cputime_t tgutime, tgstime, utime, stime; | |
1490 | unsigned long maxrss = 0; | |
1491 | ||
1492 | memset((char *) r, 0, sizeof *r); | |
1493 | utime = stime = cputime_zero; | |
1494 | ||
1495 | if (who == RUSAGE_THREAD) { | |
1496 | task_times(current, &utime, &stime); | |
1497 | accumulate_thread_rusage(p, r); | |
1498 | maxrss = p->signal->maxrss; | |
1499 | goto out; | |
1500 | } | |
1501 | ||
1502 | if (!lock_task_sighand(p, &flags)) | |
1503 | return; | |
1504 | ||
1505 | switch (who) { | |
1506 | case RUSAGE_BOTH: | |
1507 | case RUSAGE_CHILDREN: | |
1508 | utime = p->signal->cutime; | |
1509 | stime = p->signal->cstime; | |
1510 | r->ru_nvcsw = p->signal->cnvcsw; | |
1511 | r->ru_nivcsw = p->signal->cnivcsw; | |
1512 | r->ru_minflt = p->signal->cmin_flt; | |
1513 | r->ru_majflt = p->signal->cmaj_flt; | |
1514 | r->ru_inblock = p->signal->cinblock; | |
1515 | r->ru_oublock = p->signal->coublock; | |
1516 | maxrss = p->signal->cmaxrss; | |
1517 | ||
1518 | if (who == RUSAGE_CHILDREN) | |
1519 | break; | |
1520 | ||
1521 | case RUSAGE_SELF: | |
1522 | thread_group_times(p, &tgutime, &tgstime); | |
1523 | utime = cputime_add(utime, tgutime); | |
1524 | stime = cputime_add(stime, tgstime); | |
1525 | r->ru_nvcsw += p->signal->nvcsw; | |
1526 | r->ru_nivcsw += p->signal->nivcsw; | |
1527 | r->ru_minflt += p->signal->min_flt; | |
1528 | r->ru_majflt += p->signal->maj_flt; | |
1529 | r->ru_inblock += p->signal->inblock; | |
1530 | r->ru_oublock += p->signal->oublock; | |
1531 | if (maxrss < p->signal->maxrss) | |
1532 | maxrss = p->signal->maxrss; | |
1533 | t = p; | |
1534 | do { | |
1535 | accumulate_thread_rusage(t, r); | |
1536 | t = next_thread(t); | |
1537 | } while (t != p); | |
1538 | break; | |
1539 | ||
1540 | default: | |
1541 | BUG(); | |
1542 | } | |
1543 | unlock_task_sighand(p, &flags); | |
1544 | ||
1545 | out: | |
1546 | cputime_to_timeval(utime, &r->ru_utime); | |
1547 | cputime_to_timeval(stime, &r->ru_stime); | |
1548 | ||
1549 | if (who != RUSAGE_CHILDREN) { | |
1550 | struct mm_struct *mm = get_task_mm(p); | |
1551 | if (mm) { | |
1552 | setmax_mm_hiwater_rss(&maxrss, mm); | |
1553 | mmput(mm); | |
1554 | } | |
1555 | } | |
1556 | r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */ | |
1557 | } | |
1558 | ||
1559 | int getrusage(struct task_struct *p, int who, struct rusage __user *ru) | |
1560 | { | |
1561 | struct rusage r; | |
1562 | k_getrusage(p, who, &r); | |
1563 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; | |
1564 | } | |
1565 | ||
1566 | SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru) | |
1567 | { | |
1568 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN && | |
1569 | who != RUSAGE_THREAD) | |
1570 | return -EINVAL; | |
1571 | return getrusage(current, who, ru); | |
1572 | } | |
1573 | ||
1574 | SYSCALL_DEFINE1(umask, int, mask) | |
1575 | { | |
1576 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); | |
1577 | return mask; | |
1578 | } | |
1579 | ||
1580 | SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, | |
1581 | unsigned long, arg4, unsigned long, arg5) | |
1582 | { | |
1583 | struct task_struct *me = current; | |
1584 | unsigned char comm[sizeof(me->comm)]; | |
1585 | long error; | |
1586 | ||
1587 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); | |
1588 | if (error != -ENOSYS) | |
1589 | return error; | |
1590 | ||
1591 | error = 0; | |
1592 | switch (option) { | |
1593 | case PR_SET_PDEATHSIG: | |
1594 | if (!valid_signal(arg2)) { | |
1595 | error = -EINVAL; | |
1596 | break; | |
1597 | } | |
1598 | me->pdeath_signal = arg2; | |
1599 | error = 0; | |
1600 | break; | |
1601 | case PR_GET_PDEATHSIG: | |
1602 | error = put_user(me->pdeath_signal, (int __user *)arg2); | |
1603 | break; | |
1604 | case PR_GET_DUMPABLE: | |
1605 | error = get_dumpable(me->mm); | |
1606 | break; | |
1607 | case PR_SET_DUMPABLE: | |
1608 | if (arg2 < 0 || arg2 > 1) { | |
1609 | error = -EINVAL; | |
1610 | break; | |
1611 | } | |
1612 | set_dumpable(me->mm, arg2); | |
1613 | error = 0; | |
1614 | break; | |
1615 | ||
1616 | case PR_SET_UNALIGN: | |
1617 | error = SET_UNALIGN_CTL(me, arg2); | |
1618 | break; | |
1619 | case PR_GET_UNALIGN: | |
1620 | error = GET_UNALIGN_CTL(me, arg2); | |
1621 | break; | |
1622 | case PR_SET_FPEMU: | |
1623 | error = SET_FPEMU_CTL(me, arg2); | |
1624 | break; | |
1625 | case PR_GET_FPEMU: | |
1626 | error = GET_FPEMU_CTL(me, arg2); | |
1627 | break; | |
1628 | case PR_SET_FPEXC: | |
1629 | error = SET_FPEXC_CTL(me, arg2); | |
1630 | break; | |
1631 | case PR_GET_FPEXC: | |
1632 | error = GET_FPEXC_CTL(me, arg2); | |
1633 | break; | |
1634 | case PR_GET_TIMING: | |
1635 | error = PR_TIMING_STATISTICAL; | |
1636 | break; | |
1637 | case PR_SET_TIMING: | |
1638 | if (arg2 != PR_TIMING_STATISTICAL) | |
1639 | error = -EINVAL; | |
1640 | else | |
1641 | error = 0; | |
1642 | break; | |
1643 | ||
1644 | case PR_SET_NAME: | |
1645 | comm[sizeof(me->comm)-1] = 0; | |
1646 | if (strncpy_from_user(comm, (char __user *)arg2, | |
1647 | sizeof(me->comm) - 1) < 0) | |
1648 | return -EFAULT; | |
1649 | set_task_comm(me, comm); | |
1650 | return 0; | |
1651 | case PR_GET_NAME: | |
1652 | get_task_comm(comm, me); | |
1653 | if (copy_to_user((char __user *)arg2, comm, | |
1654 | sizeof(comm))) | |
1655 | return -EFAULT; | |
1656 | return 0; | |
1657 | case PR_GET_ENDIAN: | |
1658 | error = GET_ENDIAN(me, arg2); | |
1659 | break; | |
1660 | case PR_SET_ENDIAN: | |
1661 | error = SET_ENDIAN(me, arg2); | |
1662 | break; | |
1663 | ||
1664 | case PR_GET_SECCOMP: | |
1665 | error = prctl_get_seccomp(); | |
1666 | break; | |
1667 | case PR_SET_SECCOMP: | |
1668 | error = prctl_set_seccomp(arg2); | |
1669 | break; | |
1670 | case PR_GET_TSC: | |
1671 | error = GET_TSC_CTL(arg2); | |
1672 | break; | |
1673 | case PR_SET_TSC: | |
1674 | error = SET_TSC_CTL(arg2); | |
1675 | break; | |
1676 | case PR_TASK_PERF_EVENTS_DISABLE: | |
1677 | error = perf_event_task_disable(); | |
1678 | break; | |
1679 | case PR_TASK_PERF_EVENTS_ENABLE: | |
1680 | error = perf_event_task_enable(); | |
1681 | break; | |
1682 | case PR_GET_TIMERSLACK: | |
1683 | error = current->timer_slack_ns; | |
1684 | break; | |
1685 | case PR_SET_TIMERSLACK: | |
1686 | if (arg2 <= 0) | |
1687 | current->timer_slack_ns = | |
1688 | current->default_timer_slack_ns; | |
1689 | else | |
1690 | current->timer_slack_ns = arg2; | |
1691 | error = 0; | |
1692 | break; | |
1693 | case PR_MCE_KILL: | |
1694 | if (arg4 | arg5) | |
1695 | return -EINVAL; | |
1696 | switch (arg2) { | |
1697 | case PR_MCE_KILL_CLEAR: | |
1698 | if (arg3 != 0) | |
1699 | return -EINVAL; | |
1700 | current->flags &= ~PF_MCE_PROCESS; | |
1701 | break; | |
1702 | case PR_MCE_KILL_SET: | |
1703 | current->flags |= PF_MCE_PROCESS; | |
1704 | if (arg3 == PR_MCE_KILL_EARLY) | |
1705 | current->flags |= PF_MCE_EARLY; | |
1706 | else if (arg3 == PR_MCE_KILL_LATE) | |
1707 | current->flags &= ~PF_MCE_EARLY; | |
1708 | else if (arg3 == PR_MCE_KILL_DEFAULT) | |
1709 | current->flags &= | |
1710 | ~(PF_MCE_EARLY|PF_MCE_PROCESS); | |
1711 | else | |
1712 | return -EINVAL; | |
1713 | break; | |
1714 | default: | |
1715 | return -EINVAL; | |
1716 | } | |
1717 | error = 0; | |
1718 | break; | |
1719 | case PR_MCE_KILL_GET: | |
1720 | if (arg2 | arg3 | arg4 | arg5) | |
1721 | return -EINVAL; | |
1722 | if (current->flags & PF_MCE_PROCESS) | |
1723 | error = (current->flags & PF_MCE_EARLY) ? | |
1724 | PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE; | |
1725 | else | |
1726 | error = PR_MCE_KILL_DEFAULT; | |
1727 | break; | |
1728 | default: | |
1729 | error = -EINVAL; | |
1730 | break; | |
1731 | } | |
1732 | return error; | |
1733 | } | |
1734 | ||
1735 | SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep, | |
1736 | struct getcpu_cache __user *, unused) | |
1737 | { | |
1738 | int err = 0; | |
1739 | int cpu = raw_smp_processor_id(); | |
1740 | if (cpup) | |
1741 | err |= put_user(cpu, cpup); | |
1742 | if (nodep) | |
1743 | err |= put_user(cpu_to_node(cpu), nodep); | |
1744 | return err ? -EFAULT : 0; | |
1745 | } | |
1746 | ||
1747 | char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff"; | |
1748 | ||
1749 | static void argv_cleanup(struct subprocess_info *info) | |
1750 | { | |
1751 | argv_free(info->argv); | |
1752 | } | |
1753 | ||
1754 | /** | |
1755 | * orderly_poweroff - Trigger an orderly system poweroff | |
1756 | * @force: force poweroff if command execution fails | |
1757 | * | |
1758 | * This may be called from any context to trigger a system shutdown. | |
1759 | * If the orderly shutdown fails, it will force an immediate shutdown. | |
1760 | */ | |
1761 | int orderly_poweroff(bool force) | |
1762 | { | |
1763 | int argc; | |
1764 | char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc); | |
1765 | static char *envp[] = { | |
1766 | "HOME=/", | |
1767 | "PATH=/sbin:/bin:/usr/sbin:/usr/bin", | |
1768 | NULL | |
1769 | }; | |
1770 | int ret = -ENOMEM; | |
1771 | struct subprocess_info *info; | |
1772 | ||
1773 | if (argv == NULL) { | |
1774 | printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n", | |
1775 | __func__, poweroff_cmd); | |
1776 | goto out; | |
1777 | } | |
1778 | ||
1779 | info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC); | |
1780 | if (info == NULL) { | |
1781 | argv_free(argv); | |
1782 | goto out; | |
1783 | } | |
1784 | ||
1785 | call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL); | |
1786 | ||
1787 | ret = call_usermodehelper_exec(info, UMH_NO_WAIT); | |
1788 | ||
1789 | out: | |
1790 | if (ret && force) { | |
1791 | printk(KERN_WARNING "Failed to start orderly shutdown: " | |
1792 | "forcing the issue\n"); | |
1793 | ||
1794 | /* I guess this should try to kick off some daemon to | |
1795 | sync and poweroff asap. Or not even bother syncing | |
1796 | if we're doing an emergency shutdown? */ | |
1797 | emergency_sync(); | |
1798 | kernel_power_off(); | |
1799 | } | |
1800 | ||
1801 | return ret; | |
1802 | } | |
1803 | EXPORT_SYMBOL_GPL(orderly_poweroff); |