• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/kernel/sys.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  */
7 
8 #include <linux/export.h>
9 #include <linux/mm.h>
10 #include <linux/mm_inline.h>
11 #include <linux/utsname.h>
12 #include <linux/mman.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.h>
16 #include <linux/fs.h>
17 #include <linux/kmod.h>
18 #include <linux/perf_event.h>
19 #include <linux/resource.h>
20 #include <linux/kernel.h>
21 #include <linux/workqueue.h>
22 #include <linux/capability.h>
23 #include <linux/device.h>
24 #include <linux/key.h>
25 #include <linux/times.h>
26 #include <linux/posix-timers.h>
27 #include <linux/security.h>
28 #include <linux/random.h>
29 #include <linux/suspend.h>
30 #include <linux/tty.h>
31 #include <linux/signal.h>
32 #include <linux/cn_proc.h>
33 #include <linux/getcpu.h>
34 #include <linux/task_io_accounting_ops.h>
35 #include <linux/seccomp.h>
36 #include <linux/cpu.h>
37 #include <linux/personality.h>
38 #include <linux/ptrace.h>
39 #include <linux/fs_struct.h>
40 #include <linux/file.h>
41 #include <linux/mount.h>
42 #include <linux/gfp.h>
43 #include <linux/syscore_ops.h>
44 #include <linux/version.h>
45 #include <linux/ctype.h>
46 #include <linux/syscall_user_dispatch.h>
47 
48 #include <linux/compat.h>
49 #include <linux/syscalls.h>
50 #include <linux/kprobes.h>
51 #include <linux/user_namespace.h>
52 #include <linux/time_namespace.h>
53 #include <linux/binfmts.h>
54 
55 #include <linux/sched.h>
56 #include <linux/sched/autogroup.h>
57 #include <linux/sched/loadavg.h>
58 #include <linux/sched/stat.h>
59 #include <linux/sched/mm.h>
60 #include <linux/sched/coredump.h>
61 #include <linux/sched/task.h>
62 #include <linux/sched/cputime.h>
63 #include <linux/rcupdate.h>
64 #include <linux/uidgid.h>
65 #include <linux/cred.h>
66 
67 #include <linux/nospec.h>
68 
69 #include <linux/kmsg_dump.h>
70 /* Move somewhere else to avoid recompiling? */
71 #include <generated/utsrelease.h>
72 
73 #include <linux/uaccess.h>
74 #include <asm/io.h>
75 #include <asm/unistd.h>
76 
77 #include "uid16.h"
78 
79 #include <trace/hooks/sys.h>
80 
81 #ifndef SET_UNALIGN_CTL
82 # define SET_UNALIGN_CTL(a, b)	(-EINVAL)
83 #endif
84 #ifndef GET_UNALIGN_CTL
85 # define GET_UNALIGN_CTL(a, b)	(-EINVAL)
86 #endif
87 #ifndef SET_FPEMU_CTL
88 # define SET_FPEMU_CTL(a, b)	(-EINVAL)
89 #endif
90 #ifndef GET_FPEMU_CTL
91 # define GET_FPEMU_CTL(a, b)	(-EINVAL)
92 #endif
93 #ifndef SET_FPEXC_CTL
94 # define SET_FPEXC_CTL(a, b)	(-EINVAL)
95 #endif
96 #ifndef GET_FPEXC_CTL
97 # define GET_FPEXC_CTL(a, b)	(-EINVAL)
98 #endif
99 #ifndef GET_ENDIAN
100 # define GET_ENDIAN(a, b)	(-EINVAL)
101 #endif
102 #ifndef SET_ENDIAN
103 # define SET_ENDIAN(a, b)	(-EINVAL)
104 #endif
105 #ifndef GET_TSC_CTL
106 # define GET_TSC_CTL(a)		(-EINVAL)
107 #endif
108 #ifndef SET_TSC_CTL
109 # define SET_TSC_CTL(a)		(-EINVAL)
110 #endif
111 #ifndef GET_FP_MODE
112 # define GET_FP_MODE(a)		(-EINVAL)
113 #endif
114 #ifndef SET_FP_MODE
115 # define SET_FP_MODE(a,b)	(-EINVAL)
116 #endif
117 #ifndef SVE_SET_VL
118 # define SVE_SET_VL(a)		(-EINVAL)
119 #endif
120 #ifndef SVE_GET_VL
121 # define SVE_GET_VL()		(-EINVAL)
122 #endif
123 #ifndef SME_SET_VL
124 # define SME_SET_VL(a)		(-EINVAL)
125 #endif
126 #ifndef SME_GET_VL
127 # define SME_GET_VL()		(-EINVAL)
128 #endif
129 #ifndef PAC_RESET_KEYS
130 # define PAC_RESET_KEYS(a, b)	(-EINVAL)
131 #endif
132 #ifndef PAC_SET_ENABLED_KEYS
133 # define PAC_SET_ENABLED_KEYS(a, b, c)	(-EINVAL)
134 #endif
135 #ifndef PAC_GET_ENABLED_KEYS
136 # define PAC_GET_ENABLED_KEYS(a)	(-EINVAL)
137 #endif
138 #ifndef SET_TAGGED_ADDR_CTRL
139 # define SET_TAGGED_ADDR_CTRL(a)	(-EINVAL)
140 #endif
141 #ifndef GET_TAGGED_ADDR_CTRL
142 # define GET_TAGGED_ADDR_CTRL()		(-EINVAL)
143 #endif
144 
145 /*
146  * this is where the system-wide overflow UID and GID are defined, for
147  * architectures that now have 32-bit UID/GID but didn't in the past
148  */
149 
150 int overflowuid = DEFAULT_OVERFLOWUID;
151 int overflowgid = DEFAULT_OVERFLOWGID;
152 
153 EXPORT_SYMBOL(overflowuid);
154 EXPORT_SYMBOL(overflowgid);
155 
156 /*
157  * the same as above, but for filesystems which can only store a 16-bit
158  * UID and GID. as such, this is needed on all architectures
159  */
160 
161 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
162 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
163 
164 EXPORT_SYMBOL(fs_overflowuid);
165 EXPORT_SYMBOL(fs_overflowgid);
166 
167 /*
168  * Returns true if current's euid is same as p's uid or euid,
169  * or has CAP_SYS_NICE to p's user_ns.
170  *
171  * Called with rcu_read_lock, creds are safe
172  */
set_one_prio_perm(struct task_struct * p)173 static bool set_one_prio_perm(struct task_struct *p)
174 {
175 	const struct cred *cred = current_cred(), *pcred = __task_cred(p);
176 
177 	if (uid_eq(pcred->uid,  cred->euid) ||
178 	    uid_eq(pcred->euid, cred->euid))
179 		return true;
180 	if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
181 		return true;
182 	return false;
183 }
184 
185 /*
186  * set the priority of a task
187  * - the caller must hold the RCU read lock
188  */
set_one_prio(struct task_struct * p,int niceval,int error)189 static int set_one_prio(struct task_struct *p, int niceval, int error)
190 {
191 	int no_nice;
192 
193 	if (!set_one_prio_perm(p)) {
194 		error = -EPERM;
195 		goto out;
196 	}
197 	if (niceval < task_nice(p) && !can_nice(p, niceval)) {
198 		error = -EACCES;
199 		goto out;
200 	}
201 	no_nice = security_task_setnice(p, niceval);
202 	if (no_nice) {
203 		error = no_nice;
204 		goto out;
205 	}
206 	if (error == -ESRCH)
207 		error = 0;
208 	set_user_nice(p, niceval);
209 out:
210 	return error;
211 }
212 
SYSCALL_DEFINE3(setpriority,int,which,int,who,int,niceval)213 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
214 {
215 	struct task_struct *g, *p;
216 	struct user_struct *user;
217 	const struct cred *cred = current_cred();
218 	int error = -EINVAL;
219 	struct pid *pgrp;
220 	kuid_t uid;
221 
222 	if (which > PRIO_USER || which < PRIO_PROCESS)
223 		goto out;
224 
225 	/* normalize: avoid signed division (rounding problems) */
226 	error = -ESRCH;
227 	if (niceval < MIN_NICE)
228 		niceval = MIN_NICE;
229 	if (niceval > MAX_NICE)
230 		niceval = MAX_NICE;
231 
232 	rcu_read_lock();
233 	switch (which) {
234 	case PRIO_PROCESS:
235 		if (who)
236 			p = find_task_by_vpid(who);
237 		else
238 			p = current;
239 		if (p)
240 			error = set_one_prio(p, niceval, error);
241 		break;
242 	case PRIO_PGRP:
243 		if (who)
244 			pgrp = find_vpid(who);
245 		else
246 			pgrp = task_pgrp(current);
247 		read_lock(&tasklist_lock);
248 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
249 			error = set_one_prio(p, niceval, error);
250 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
251 		read_unlock(&tasklist_lock);
252 		break;
253 	case PRIO_USER:
254 		uid = make_kuid(cred->user_ns, who);
255 		user = cred->user;
256 		if (!who)
257 			uid = cred->uid;
258 		else if (!uid_eq(uid, cred->uid)) {
259 			user = find_user(uid);
260 			if (!user)
261 				goto out_unlock;	/* No processes for this user */
262 		}
263 		for_each_process_thread(g, p) {
264 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
265 				error = set_one_prio(p, niceval, error);
266 		}
267 		if (!uid_eq(uid, cred->uid))
268 			free_uid(user);		/* For find_user() */
269 		break;
270 	}
271 out_unlock:
272 	rcu_read_unlock();
273 out:
274 	return error;
275 }
276 
277 /*
278  * Ugh. To avoid negative return values, "getpriority()" will
279  * not return the normal nice-value, but a negated value that
280  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
281  * to stay compatible.
282  */
SYSCALL_DEFINE2(getpriority,int,which,int,who)283 SYSCALL_DEFINE2(getpriority, int, which, int, who)
284 {
285 	struct task_struct *g, *p;
286 	struct user_struct *user;
287 	const struct cred *cred = current_cred();
288 	long niceval, retval = -ESRCH;
289 	struct pid *pgrp;
290 	kuid_t uid;
291 
292 	if (which > PRIO_USER || which < PRIO_PROCESS)
293 		return -EINVAL;
294 
295 	rcu_read_lock();
296 	switch (which) {
297 	case PRIO_PROCESS:
298 		if (who)
299 			p = find_task_by_vpid(who);
300 		else
301 			p = current;
302 		if (p) {
303 			niceval = nice_to_rlimit(task_nice(p));
304 			if (niceval > retval)
305 				retval = niceval;
306 		}
307 		break;
308 	case PRIO_PGRP:
309 		if (who)
310 			pgrp = find_vpid(who);
311 		else
312 			pgrp = task_pgrp(current);
313 		read_lock(&tasklist_lock);
314 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
315 			niceval = nice_to_rlimit(task_nice(p));
316 			if (niceval > retval)
317 				retval = niceval;
318 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
319 		read_unlock(&tasklist_lock);
320 		break;
321 	case PRIO_USER:
322 		uid = make_kuid(cred->user_ns, who);
323 		user = cred->user;
324 		if (!who)
325 			uid = cred->uid;
326 		else if (!uid_eq(uid, cred->uid)) {
327 			user = find_user(uid);
328 			if (!user)
329 				goto out_unlock;	/* No processes for this user */
330 		}
331 		for_each_process_thread(g, p) {
332 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
333 				niceval = nice_to_rlimit(task_nice(p));
334 				if (niceval > retval)
335 					retval = niceval;
336 			}
337 		}
338 		if (!uid_eq(uid, cred->uid))
339 			free_uid(user);		/* for find_user() */
340 		break;
341 	}
342 out_unlock:
343 	rcu_read_unlock();
344 
345 	return retval;
346 }
347 
348 /*
349  * Unprivileged users may change the real gid to the effective gid
350  * or vice versa.  (BSD-style)
351  *
352  * If you set the real gid at all, or set the effective gid to a value not
353  * equal to the real gid, then the saved gid is set to the new effective gid.
354  *
355  * This makes it possible for a setgid program to completely drop its
356  * privileges, which is often a useful assertion to make when you are doing
357  * a security audit over a program.
358  *
359  * The general idea is that a program which uses just setregid() will be
360  * 100% compatible with BSD.  A program which uses just setgid() will be
361  * 100% compatible with POSIX with saved IDs.
362  *
363  * SMP: There are not races, the GIDs are checked only by filesystem
364  *      operations (as far as semantic preservation is concerned).
365  */
366 #ifdef CONFIG_MULTIUSER
__sys_setregid(gid_t rgid,gid_t egid)367 long __sys_setregid(gid_t rgid, gid_t egid)
368 {
369 	struct user_namespace *ns = current_user_ns();
370 	const struct cred *old;
371 	struct cred *new;
372 	int retval;
373 	kgid_t krgid, kegid;
374 
375 	krgid = make_kgid(ns, rgid);
376 	kegid = make_kgid(ns, egid);
377 
378 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
379 		return -EINVAL;
380 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
381 		return -EINVAL;
382 
383 	new = prepare_creds();
384 	if (!new)
385 		return -ENOMEM;
386 	old = current_cred();
387 
388 	retval = -EPERM;
389 	if (rgid != (gid_t) -1) {
390 		if (gid_eq(old->gid, krgid) ||
391 		    gid_eq(old->egid, krgid) ||
392 		    ns_capable_setid(old->user_ns, CAP_SETGID))
393 			new->gid = krgid;
394 		else
395 			goto error;
396 	}
397 	if (egid != (gid_t) -1) {
398 		if (gid_eq(old->gid, kegid) ||
399 		    gid_eq(old->egid, kegid) ||
400 		    gid_eq(old->sgid, kegid) ||
401 		    ns_capable_setid(old->user_ns, CAP_SETGID))
402 			new->egid = kegid;
403 		else
404 			goto error;
405 	}
406 
407 	if (rgid != (gid_t) -1 ||
408 	    (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
409 		new->sgid = new->egid;
410 	new->fsgid = new->egid;
411 
412 	retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
413 	if (retval < 0)
414 		goto error;
415 
416 	return commit_creds(new);
417 
418 error:
419 	abort_creds(new);
420 	return retval;
421 }
422 
SYSCALL_DEFINE2(setregid,gid_t,rgid,gid_t,egid)423 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
424 {
425 	return __sys_setregid(rgid, egid);
426 }
427 
428 /*
429  * setgid() is implemented like SysV w/ SAVED_IDS
430  *
431  * SMP: Same implicit races as above.
432  */
__sys_setgid(gid_t gid)433 long __sys_setgid(gid_t gid)
434 {
435 	struct user_namespace *ns = current_user_ns();
436 	const struct cred *old;
437 	struct cred *new;
438 	int retval;
439 	kgid_t kgid;
440 
441 	kgid = make_kgid(ns, gid);
442 	if (!gid_valid(kgid))
443 		return -EINVAL;
444 
445 	new = prepare_creds();
446 	if (!new)
447 		return -ENOMEM;
448 	old = current_cred();
449 
450 	retval = -EPERM;
451 	if (ns_capable_setid(old->user_ns, CAP_SETGID))
452 		new->gid = new->egid = new->sgid = new->fsgid = kgid;
453 	else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
454 		new->egid = new->fsgid = kgid;
455 	else
456 		goto error;
457 
458 	retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
459 	if (retval < 0)
460 		goto error;
461 
462 	return commit_creds(new);
463 
464 error:
465 	abort_creds(new);
466 	return retval;
467 }
468 
SYSCALL_DEFINE1(setgid,gid_t,gid)469 SYSCALL_DEFINE1(setgid, gid_t, gid)
470 {
471 	return __sys_setgid(gid);
472 }
473 
474 /*
475  * change the user struct in a credentials set to match the new UID
476  */
set_user(struct cred * new)477 static int set_user(struct cred *new)
478 {
479 	struct user_struct *new_user;
480 
481 	new_user = alloc_uid(new->uid);
482 	if (!new_user)
483 		return -EAGAIN;
484 
485 	free_uid(new->user);
486 	new->user = new_user;
487 	return 0;
488 }
489 
flag_nproc_exceeded(struct cred * new)490 static void flag_nproc_exceeded(struct cred *new)
491 {
492 	if (new->ucounts == current_ucounts())
493 		return;
494 
495 	/*
496 	 * We don't fail in case of NPROC limit excess here because too many
497 	 * poorly written programs don't check set*uid() return code, assuming
498 	 * it never fails if called by root.  We may still enforce NPROC limit
499 	 * for programs doing set*uid()+execve() by harmlessly deferring the
500 	 * failure to the execve() stage.
501 	 */
502 	if (is_rlimit_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
503 			new->user != INIT_USER)
504 		current->flags |= PF_NPROC_EXCEEDED;
505 	else
506 		current->flags &= ~PF_NPROC_EXCEEDED;
507 }
508 
509 /*
510  * Unprivileged users may change the real uid to the effective uid
511  * or vice versa.  (BSD-style)
512  *
513  * If you set the real uid at all, or set the effective uid to a value not
514  * equal to the real uid, then the saved uid is set to the new effective uid.
515  *
516  * This makes it possible for a setuid program to completely drop its
517  * privileges, which is often a useful assertion to make when you are doing
518  * a security audit over a program.
519  *
520  * The general idea is that a program which uses just setreuid() will be
521  * 100% compatible with BSD.  A program which uses just setuid() will be
522  * 100% compatible with POSIX with saved IDs.
523  */
__sys_setreuid(uid_t ruid,uid_t euid)524 long __sys_setreuid(uid_t ruid, uid_t euid)
525 {
526 	struct user_namespace *ns = current_user_ns();
527 	const struct cred *old;
528 	struct cred *new;
529 	int retval;
530 	kuid_t kruid, keuid;
531 
532 	kruid = make_kuid(ns, ruid);
533 	keuid = make_kuid(ns, euid);
534 
535 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
536 		return -EINVAL;
537 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
538 		return -EINVAL;
539 
540 	new = prepare_creds();
541 	if (!new)
542 		return -ENOMEM;
543 	old = current_cred();
544 
545 	retval = -EPERM;
546 	if (ruid != (uid_t) -1) {
547 		new->uid = kruid;
548 		if (!uid_eq(old->uid, kruid) &&
549 		    !uid_eq(old->euid, kruid) &&
550 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
551 			goto error;
552 	}
553 
554 	if (euid != (uid_t) -1) {
555 		new->euid = keuid;
556 		if (!uid_eq(old->uid, keuid) &&
557 		    !uid_eq(old->euid, keuid) &&
558 		    !uid_eq(old->suid, keuid) &&
559 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
560 			goto error;
561 	}
562 
563 	if (!uid_eq(new->uid, old->uid)) {
564 		retval = set_user(new);
565 		if (retval < 0)
566 			goto error;
567 	}
568 	if (ruid != (uid_t) -1 ||
569 	    (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
570 		new->suid = new->euid;
571 	new->fsuid = new->euid;
572 
573 	retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
574 	if (retval < 0)
575 		goto error;
576 
577 	retval = set_cred_ucounts(new);
578 	if (retval < 0)
579 		goto error;
580 
581 	flag_nproc_exceeded(new);
582 	return commit_creds(new);
583 
584 error:
585 	abort_creds(new);
586 	return retval;
587 }
588 
SYSCALL_DEFINE2(setreuid,uid_t,ruid,uid_t,euid)589 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
590 {
591 	return __sys_setreuid(ruid, euid);
592 }
593 
594 /*
595  * setuid() is implemented like SysV with SAVED_IDS
596  *
597  * Note that SAVED_ID's is deficient in that a setuid root program
598  * like sendmail, for example, cannot set its uid to be a normal
599  * user and then switch back, because if you're root, setuid() sets
600  * the saved uid too.  If you don't like this, blame the bright people
601  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
602  * will allow a root program to temporarily drop privileges and be able to
603  * regain them by swapping the real and effective uid.
604  */
__sys_setuid(uid_t uid)605 long __sys_setuid(uid_t uid)
606 {
607 	struct user_namespace *ns = current_user_ns();
608 	const struct cred *old;
609 	struct cred *new;
610 	int retval;
611 	kuid_t kuid;
612 
613 	kuid = make_kuid(ns, uid);
614 	if (!uid_valid(kuid))
615 		return -EINVAL;
616 
617 	new = prepare_creds();
618 	if (!new)
619 		return -ENOMEM;
620 	old = current_cred();
621 
622 	retval = -EPERM;
623 	if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
624 		new->suid = new->uid = kuid;
625 		if (!uid_eq(kuid, old->uid)) {
626 			retval = set_user(new);
627 			if (retval < 0)
628 				goto error;
629 		}
630 	} else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
631 		goto error;
632 	}
633 
634 	new->fsuid = new->euid = kuid;
635 
636 	retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
637 	if (retval < 0)
638 		goto error;
639 
640 	retval = set_cred_ucounts(new);
641 	if (retval < 0)
642 		goto error;
643 
644 	flag_nproc_exceeded(new);
645 	return commit_creds(new);
646 
647 error:
648 	abort_creds(new);
649 	return retval;
650 }
651 
SYSCALL_DEFINE1(setuid,uid_t,uid)652 SYSCALL_DEFINE1(setuid, uid_t, uid)
653 {
654 	return __sys_setuid(uid);
655 }
656 
657 
658 /*
659  * This function implements a generic ability to update ruid, euid,
660  * and suid.  This allows you to implement the 4.4 compatible seteuid().
661  */
__sys_setresuid(uid_t ruid,uid_t euid,uid_t suid)662 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
663 {
664 	struct user_namespace *ns = current_user_ns();
665 	const struct cred *old;
666 	struct cred *new;
667 	int retval;
668 	kuid_t kruid, keuid, ksuid;
669 	bool ruid_new, euid_new, suid_new;
670 
671 	kruid = make_kuid(ns, ruid);
672 	keuid = make_kuid(ns, euid);
673 	ksuid = make_kuid(ns, suid);
674 
675 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
676 		return -EINVAL;
677 
678 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
679 		return -EINVAL;
680 
681 	if ((suid != (uid_t) -1) && !uid_valid(ksuid))
682 		return -EINVAL;
683 
684 	old = current_cred();
685 
686 	/* check for no-op */
687 	if ((ruid == (uid_t) -1 || uid_eq(kruid, old->uid)) &&
688 	    (euid == (uid_t) -1 || (uid_eq(keuid, old->euid) &&
689 				    uid_eq(keuid, old->fsuid))) &&
690 	    (suid == (uid_t) -1 || uid_eq(ksuid, old->suid)))
691 		return 0;
692 
693 	ruid_new = ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
694 		   !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid);
695 	euid_new = euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
696 		   !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid);
697 	suid_new = suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
698 		   !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid);
699 	if ((ruid_new || euid_new || suid_new) &&
700 	    !ns_capable_setid(old->user_ns, CAP_SETUID))
701 		return -EPERM;
702 
703 	new = prepare_creds();
704 	if (!new)
705 		return -ENOMEM;
706 
707 	if (ruid != (uid_t) -1) {
708 		new->uid = kruid;
709 		if (!uid_eq(kruid, old->uid)) {
710 			retval = set_user(new);
711 			if (retval < 0)
712 				goto error;
713 		}
714 	}
715 	if (euid != (uid_t) -1)
716 		new->euid = keuid;
717 	if (suid != (uid_t) -1)
718 		new->suid = ksuid;
719 	new->fsuid = new->euid;
720 
721 	retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
722 	if (retval < 0)
723 		goto error;
724 
725 	retval = set_cred_ucounts(new);
726 	if (retval < 0)
727 		goto error;
728 
729 	flag_nproc_exceeded(new);
730 	return commit_creds(new);
731 
732 error:
733 	abort_creds(new);
734 	return retval;
735 }
736 
SYSCALL_DEFINE3(setresuid,uid_t,ruid,uid_t,euid,uid_t,suid)737 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
738 {
739 	return __sys_setresuid(ruid, euid, suid);
740 }
741 
SYSCALL_DEFINE3(getresuid,uid_t __user *,ruidp,uid_t __user *,euidp,uid_t __user *,suidp)742 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
743 {
744 	const struct cred *cred = current_cred();
745 	int retval;
746 	uid_t ruid, euid, suid;
747 
748 	ruid = from_kuid_munged(cred->user_ns, cred->uid);
749 	euid = from_kuid_munged(cred->user_ns, cred->euid);
750 	suid = from_kuid_munged(cred->user_ns, cred->suid);
751 
752 	retval = put_user(ruid, ruidp);
753 	if (!retval) {
754 		retval = put_user(euid, euidp);
755 		if (!retval)
756 			return put_user(suid, suidp);
757 	}
758 	return retval;
759 }
760 
761 /*
762  * Same as above, but for rgid, egid, sgid.
763  */
__sys_setresgid(gid_t rgid,gid_t egid,gid_t sgid)764 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
765 {
766 	struct user_namespace *ns = current_user_ns();
767 	const struct cred *old;
768 	struct cred *new;
769 	int retval;
770 	kgid_t krgid, kegid, ksgid;
771 	bool rgid_new, egid_new, sgid_new;
772 
773 	krgid = make_kgid(ns, rgid);
774 	kegid = make_kgid(ns, egid);
775 	ksgid = make_kgid(ns, sgid);
776 
777 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
778 		return -EINVAL;
779 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
780 		return -EINVAL;
781 	if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
782 		return -EINVAL;
783 
784 	old = current_cred();
785 
786 	/* check for no-op */
787 	if ((rgid == (gid_t) -1 || gid_eq(krgid, old->gid)) &&
788 	    (egid == (gid_t) -1 || (gid_eq(kegid, old->egid) &&
789 				    gid_eq(kegid, old->fsgid))) &&
790 	    (sgid == (gid_t) -1 || gid_eq(ksgid, old->sgid)))
791 		return 0;
792 
793 	rgid_new = rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
794 		   !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid);
795 	egid_new = egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
796 		   !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid);
797 	sgid_new = sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
798 		   !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid);
799 	if ((rgid_new || egid_new || sgid_new) &&
800 	    !ns_capable_setid(old->user_ns, CAP_SETGID))
801 		return -EPERM;
802 
803 	new = prepare_creds();
804 	if (!new)
805 		return -ENOMEM;
806 
807 	if (rgid != (gid_t) -1)
808 		new->gid = krgid;
809 	if (egid != (gid_t) -1)
810 		new->egid = kegid;
811 	if (sgid != (gid_t) -1)
812 		new->sgid = ksgid;
813 	new->fsgid = new->egid;
814 
815 	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
816 	if (retval < 0)
817 		goto error;
818 
819 	return commit_creds(new);
820 
821 error:
822 	abort_creds(new);
823 	return retval;
824 }
825 
SYSCALL_DEFINE3(setresgid,gid_t,rgid,gid_t,egid,gid_t,sgid)826 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
827 {
828 	return __sys_setresgid(rgid, egid, sgid);
829 }
830 
SYSCALL_DEFINE3(getresgid,gid_t __user *,rgidp,gid_t __user *,egidp,gid_t __user *,sgidp)831 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
832 {
833 	const struct cred *cred = current_cred();
834 	int retval;
835 	gid_t rgid, egid, sgid;
836 
837 	rgid = from_kgid_munged(cred->user_ns, cred->gid);
838 	egid = from_kgid_munged(cred->user_ns, cred->egid);
839 	sgid = from_kgid_munged(cred->user_ns, cred->sgid);
840 
841 	retval = put_user(rgid, rgidp);
842 	if (!retval) {
843 		retval = put_user(egid, egidp);
844 		if (!retval)
845 			retval = put_user(sgid, sgidp);
846 	}
847 
848 	return retval;
849 }
850 
851 
852 /*
853  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
854  * is used for "access()" and for the NFS daemon (letting nfsd stay at
855  * whatever uid it wants to). It normally shadows "euid", except when
856  * explicitly set by setfsuid() or for access..
857  */
__sys_setfsuid(uid_t uid)858 long __sys_setfsuid(uid_t uid)
859 {
860 	const struct cred *old;
861 	struct cred *new;
862 	uid_t old_fsuid;
863 	kuid_t kuid;
864 
865 	old = current_cred();
866 	old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
867 
868 	kuid = make_kuid(old->user_ns, uid);
869 	if (!uid_valid(kuid))
870 		return old_fsuid;
871 
872 	new = prepare_creds();
873 	if (!new)
874 		return old_fsuid;
875 
876 	if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
877 	    uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
878 	    ns_capable_setid(old->user_ns, CAP_SETUID)) {
879 		if (!uid_eq(kuid, old->fsuid)) {
880 			new->fsuid = kuid;
881 			if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
882 				goto change_okay;
883 		}
884 	}
885 
886 	abort_creds(new);
887 	return old_fsuid;
888 
889 change_okay:
890 	commit_creds(new);
891 	return old_fsuid;
892 }
893 
SYSCALL_DEFINE1(setfsuid,uid_t,uid)894 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
895 {
896 	return __sys_setfsuid(uid);
897 }
898 
899 /*
900  * Samma på svenska..
901  */
__sys_setfsgid(gid_t gid)902 long __sys_setfsgid(gid_t gid)
903 {
904 	const struct cred *old;
905 	struct cred *new;
906 	gid_t old_fsgid;
907 	kgid_t kgid;
908 
909 	old = current_cred();
910 	old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
911 
912 	kgid = make_kgid(old->user_ns, gid);
913 	if (!gid_valid(kgid))
914 		return old_fsgid;
915 
916 	new = prepare_creds();
917 	if (!new)
918 		return old_fsgid;
919 
920 	if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
921 	    gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
922 	    ns_capable_setid(old->user_ns, CAP_SETGID)) {
923 		if (!gid_eq(kgid, old->fsgid)) {
924 			new->fsgid = kgid;
925 			if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
926 				goto change_okay;
927 		}
928 	}
929 
930 	abort_creds(new);
931 	return old_fsgid;
932 
933 change_okay:
934 	commit_creds(new);
935 	return old_fsgid;
936 }
937 
SYSCALL_DEFINE1(setfsgid,gid_t,gid)938 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
939 {
940 	return __sys_setfsgid(gid);
941 }
942 #endif /* CONFIG_MULTIUSER */
943 
944 /**
945  * sys_getpid - return the thread group id of the current process
946  *
947  * Note, despite the name, this returns the tgid not the pid.  The tgid and
948  * the pid are identical unless CLONE_THREAD was specified on clone() in
949  * which case the tgid is the same in all threads of the same group.
950  *
951  * This is SMP safe as current->tgid does not change.
952  */
SYSCALL_DEFINE0(getpid)953 SYSCALL_DEFINE0(getpid)
954 {
955 	return task_tgid_vnr(current);
956 }
957 
958 /* Thread ID - the internal kernel "pid" */
SYSCALL_DEFINE0(gettid)959 SYSCALL_DEFINE0(gettid)
960 {
961 	return task_pid_vnr(current);
962 }
963 
964 /*
965  * Accessing ->real_parent is not SMP-safe, it could
966  * change from under us. However, we can use a stale
967  * value of ->real_parent under rcu_read_lock(), see
968  * release_task()->call_rcu(delayed_put_task_struct).
969  */
SYSCALL_DEFINE0(getppid)970 SYSCALL_DEFINE0(getppid)
971 {
972 	int pid;
973 
974 	rcu_read_lock();
975 	pid = task_tgid_vnr(rcu_dereference(current->real_parent));
976 	rcu_read_unlock();
977 
978 	return pid;
979 }
980 
SYSCALL_DEFINE0(getuid)981 SYSCALL_DEFINE0(getuid)
982 {
983 	/* Only we change this so SMP safe */
984 	return from_kuid_munged(current_user_ns(), current_uid());
985 }
986 
SYSCALL_DEFINE0(geteuid)987 SYSCALL_DEFINE0(geteuid)
988 {
989 	/* Only we change this so SMP safe */
990 	return from_kuid_munged(current_user_ns(), current_euid());
991 }
992 
SYSCALL_DEFINE0(getgid)993 SYSCALL_DEFINE0(getgid)
994 {
995 	/* Only we change this so SMP safe */
996 	return from_kgid_munged(current_user_ns(), current_gid());
997 }
998 
SYSCALL_DEFINE0(getegid)999 SYSCALL_DEFINE0(getegid)
1000 {
1001 	/* Only we change this so SMP safe */
1002 	return from_kgid_munged(current_user_ns(), current_egid());
1003 }
1004 
do_sys_times(struct tms * tms)1005 static void do_sys_times(struct tms *tms)
1006 {
1007 	u64 tgutime, tgstime, cutime, cstime;
1008 
1009 	thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1010 	cutime = current->signal->cutime;
1011 	cstime = current->signal->cstime;
1012 	tms->tms_utime = nsec_to_clock_t(tgutime);
1013 	tms->tms_stime = nsec_to_clock_t(tgstime);
1014 	tms->tms_cutime = nsec_to_clock_t(cutime);
1015 	tms->tms_cstime = nsec_to_clock_t(cstime);
1016 }
1017 
SYSCALL_DEFINE1(times,struct tms __user *,tbuf)1018 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1019 {
1020 	if (tbuf) {
1021 		struct tms tmp;
1022 
1023 		do_sys_times(&tmp);
1024 		if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1025 			return -EFAULT;
1026 	}
1027 	force_successful_syscall_return();
1028 	return (long) jiffies_64_to_clock_t(get_jiffies_64());
1029 }
1030 
1031 #ifdef CONFIG_COMPAT
clock_t_to_compat_clock_t(clock_t x)1032 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1033 {
1034 	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1035 }
1036 
COMPAT_SYSCALL_DEFINE1(times,struct compat_tms __user *,tbuf)1037 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1038 {
1039 	if (tbuf) {
1040 		struct tms tms;
1041 		struct compat_tms tmp;
1042 
1043 		do_sys_times(&tms);
1044 		/* Convert our struct tms to the compat version. */
1045 		tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1046 		tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1047 		tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1048 		tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1049 		if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1050 			return -EFAULT;
1051 	}
1052 	force_successful_syscall_return();
1053 	return compat_jiffies_to_clock_t(jiffies);
1054 }
1055 #endif
1056 
1057 /*
1058  * This needs some heavy checking ...
1059  * I just haven't the stomach for it. I also don't fully
1060  * understand sessions/pgrp etc. Let somebody who does explain it.
1061  *
1062  * OK, I think I have the protection semantics right.... this is really
1063  * only important on a multi-user system anyway, to make sure one user
1064  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1065  *
1066  * !PF_FORKNOEXEC check to conform completely to POSIX.
1067  */
SYSCALL_DEFINE2(setpgid,pid_t,pid,pid_t,pgid)1068 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1069 {
1070 	struct task_struct *p;
1071 	struct task_struct *group_leader = current->group_leader;
1072 	struct pid *pgrp;
1073 	int err;
1074 
1075 	if (!pid)
1076 		pid = task_pid_vnr(group_leader);
1077 	if (!pgid)
1078 		pgid = pid;
1079 	if (pgid < 0)
1080 		return -EINVAL;
1081 	rcu_read_lock();
1082 
1083 	/* From this point forward we keep holding onto the tasklist lock
1084 	 * so that our parent does not change from under us. -DaveM
1085 	 */
1086 	write_lock_irq(&tasklist_lock);
1087 
1088 	err = -ESRCH;
1089 	p = find_task_by_vpid(pid);
1090 	if (!p)
1091 		goto out;
1092 
1093 	err = -EINVAL;
1094 	if (!thread_group_leader(p))
1095 		goto out;
1096 
1097 	if (same_thread_group(p->real_parent, group_leader)) {
1098 		err = -EPERM;
1099 		if (task_session(p) != task_session(group_leader))
1100 			goto out;
1101 		err = -EACCES;
1102 		if (!(p->flags & PF_FORKNOEXEC))
1103 			goto out;
1104 	} else {
1105 		err = -ESRCH;
1106 		if (p != group_leader)
1107 			goto out;
1108 	}
1109 
1110 	err = -EPERM;
1111 	if (p->signal->leader)
1112 		goto out;
1113 
1114 	pgrp = task_pid(p);
1115 	if (pgid != pid) {
1116 		struct task_struct *g;
1117 
1118 		pgrp = find_vpid(pgid);
1119 		g = pid_task(pgrp, PIDTYPE_PGID);
1120 		if (!g || task_session(g) != task_session(group_leader))
1121 			goto out;
1122 	}
1123 
1124 	err = security_task_setpgid(p, pgid);
1125 	if (err)
1126 		goto out;
1127 
1128 	if (task_pgrp(p) != pgrp)
1129 		change_pid(p, PIDTYPE_PGID, pgrp);
1130 
1131 	err = 0;
1132 out:
1133 	/* All paths lead to here, thus we are safe. -DaveM */
1134 	write_unlock_irq(&tasklist_lock);
1135 	rcu_read_unlock();
1136 	return err;
1137 }
1138 
do_getpgid(pid_t pid)1139 static int do_getpgid(pid_t pid)
1140 {
1141 	struct task_struct *p;
1142 	struct pid *grp;
1143 	int retval;
1144 
1145 	rcu_read_lock();
1146 	if (!pid)
1147 		grp = task_pgrp(current);
1148 	else {
1149 		retval = -ESRCH;
1150 		p = find_task_by_vpid(pid);
1151 		if (!p)
1152 			goto out;
1153 		grp = task_pgrp(p);
1154 		if (!grp)
1155 			goto out;
1156 
1157 		retval = security_task_getpgid(p);
1158 		if (retval)
1159 			goto out;
1160 	}
1161 	retval = pid_vnr(grp);
1162 out:
1163 	rcu_read_unlock();
1164 	return retval;
1165 }
1166 
SYSCALL_DEFINE1(getpgid,pid_t,pid)1167 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1168 {
1169 	return do_getpgid(pid);
1170 }
1171 
1172 #ifdef __ARCH_WANT_SYS_GETPGRP
1173 
SYSCALL_DEFINE0(getpgrp)1174 SYSCALL_DEFINE0(getpgrp)
1175 {
1176 	return do_getpgid(0);
1177 }
1178 
1179 #endif
1180 
SYSCALL_DEFINE1(getsid,pid_t,pid)1181 SYSCALL_DEFINE1(getsid, pid_t, pid)
1182 {
1183 	struct task_struct *p;
1184 	struct pid *sid;
1185 	int retval;
1186 
1187 	rcu_read_lock();
1188 	if (!pid)
1189 		sid = task_session(current);
1190 	else {
1191 		retval = -ESRCH;
1192 		p = find_task_by_vpid(pid);
1193 		if (!p)
1194 			goto out;
1195 		sid = task_session(p);
1196 		if (!sid)
1197 			goto out;
1198 
1199 		retval = security_task_getsid(p);
1200 		if (retval)
1201 			goto out;
1202 	}
1203 	retval = pid_vnr(sid);
1204 out:
1205 	rcu_read_unlock();
1206 	return retval;
1207 }
1208 
set_special_pids(struct pid * pid)1209 static void set_special_pids(struct pid *pid)
1210 {
1211 	struct task_struct *curr = current->group_leader;
1212 
1213 	if (task_session(curr) != pid)
1214 		change_pid(curr, PIDTYPE_SID, pid);
1215 
1216 	if (task_pgrp(curr) != pid)
1217 		change_pid(curr, PIDTYPE_PGID, pid);
1218 }
1219 
ksys_setsid(void)1220 int ksys_setsid(void)
1221 {
1222 	struct task_struct *group_leader = current->group_leader;
1223 	struct pid *sid = task_pid(group_leader);
1224 	pid_t session = pid_vnr(sid);
1225 	int err = -EPERM;
1226 
1227 	write_lock_irq(&tasklist_lock);
1228 	/* Fail if I am already a session leader */
1229 	if (group_leader->signal->leader)
1230 		goto out;
1231 
1232 	/* Fail if a process group id already exists that equals the
1233 	 * proposed session id.
1234 	 */
1235 	if (pid_task(sid, PIDTYPE_PGID))
1236 		goto out;
1237 
1238 	group_leader->signal->leader = 1;
1239 	set_special_pids(sid);
1240 
1241 	proc_clear_tty(group_leader);
1242 
1243 	err = session;
1244 out:
1245 	write_unlock_irq(&tasklist_lock);
1246 	if (err > 0) {
1247 		proc_sid_connector(group_leader);
1248 		sched_autogroup_create_attach(group_leader);
1249 	}
1250 	return err;
1251 }
1252 
SYSCALL_DEFINE0(setsid)1253 SYSCALL_DEFINE0(setsid)
1254 {
1255 	return ksys_setsid();
1256 }
1257 
1258 DECLARE_RWSEM(uts_sem);
1259 
1260 #ifdef COMPAT_UTS_MACHINE
1261 #define override_architecture(name) \
1262 	(personality(current->personality) == PER_LINUX32 && \
1263 	 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1264 		      sizeof(COMPAT_UTS_MACHINE)))
1265 #else
1266 #define override_architecture(name)	0
1267 #endif
1268 
1269 /*
1270  * Work around broken programs that cannot handle "Linux 3.0".
1271  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1272  * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1273  * 2.6.60.
1274  */
override_release(char __user * release,size_t len)1275 static int override_release(char __user *release, size_t len)
1276 {
1277 	int ret = 0;
1278 
1279 	if (current->personality & UNAME26) {
1280 		const char *rest = UTS_RELEASE;
1281 		char buf[65] = { 0 };
1282 		int ndots = 0;
1283 		unsigned v;
1284 		size_t copy;
1285 
1286 		while (*rest) {
1287 			if (*rest == '.' && ++ndots >= 3)
1288 				break;
1289 			if (!isdigit(*rest) && *rest != '.')
1290 				break;
1291 			rest++;
1292 		}
1293 		v = LINUX_VERSION_PATCHLEVEL + 60;
1294 		copy = clamp_t(size_t, len, 1, sizeof(buf));
1295 		copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1296 		ret = copy_to_user(release, buf, copy + 1);
1297 	}
1298 	return ret;
1299 }
1300 
SYSCALL_DEFINE1(newuname,struct new_utsname __user *,name)1301 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1302 {
1303 	struct new_utsname tmp;
1304 
1305 	down_read(&uts_sem);
1306 	memcpy(&tmp, utsname(), sizeof(tmp));
1307 	up_read(&uts_sem);
1308 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1309 		return -EFAULT;
1310 
1311 	if (override_release(name->release, sizeof(name->release)))
1312 		return -EFAULT;
1313 	if (override_architecture(name))
1314 		return -EFAULT;
1315 	return 0;
1316 }
1317 
1318 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1319 /*
1320  * Old cruft
1321  */
SYSCALL_DEFINE1(uname,struct old_utsname __user *,name)1322 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1323 {
1324 	struct old_utsname tmp;
1325 
1326 	if (!name)
1327 		return -EFAULT;
1328 
1329 	down_read(&uts_sem);
1330 	memcpy(&tmp, utsname(), sizeof(tmp));
1331 	up_read(&uts_sem);
1332 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1333 		return -EFAULT;
1334 
1335 	if (override_release(name->release, sizeof(name->release)))
1336 		return -EFAULT;
1337 	if (override_architecture(name))
1338 		return -EFAULT;
1339 	return 0;
1340 }
1341 
SYSCALL_DEFINE1(olduname,struct oldold_utsname __user *,name)1342 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1343 {
1344 	struct oldold_utsname tmp;
1345 
1346 	if (!name)
1347 		return -EFAULT;
1348 
1349 	memset(&tmp, 0, sizeof(tmp));
1350 
1351 	down_read(&uts_sem);
1352 	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1353 	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1354 	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1355 	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1356 	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1357 	up_read(&uts_sem);
1358 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1359 		return -EFAULT;
1360 
1361 	if (override_architecture(name))
1362 		return -EFAULT;
1363 	if (override_release(name->release, sizeof(name->release)))
1364 		return -EFAULT;
1365 	return 0;
1366 }
1367 #endif
1368 
SYSCALL_DEFINE2(sethostname,char __user *,name,int,len)1369 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1370 {
1371 	int errno;
1372 	char tmp[__NEW_UTS_LEN];
1373 
1374 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1375 		return -EPERM;
1376 
1377 	if (len < 0 || len > __NEW_UTS_LEN)
1378 		return -EINVAL;
1379 	errno = -EFAULT;
1380 	if (!copy_from_user(tmp, name, len)) {
1381 		struct new_utsname *u;
1382 
1383 		add_device_randomness(tmp, len);
1384 		down_write(&uts_sem);
1385 		u = utsname();
1386 		memcpy(u->nodename, tmp, len);
1387 		memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1388 		errno = 0;
1389 		uts_proc_notify(UTS_PROC_HOSTNAME);
1390 		up_write(&uts_sem);
1391 	}
1392 	return errno;
1393 }
1394 
1395 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1396 
SYSCALL_DEFINE2(gethostname,char __user *,name,int,len)1397 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1398 {
1399 	int i;
1400 	struct new_utsname *u;
1401 	char tmp[__NEW_UTS_LEN + 1];
1402 
1403 	if (len < 0)
1404 		return -EINVAL;
1405 	down_read(&uts_sem);
1406 	u = utsname();
1407 	i = 1 + strlen(u->nodename);
1408 	if (i > len)
1409 		i = len;
1410 	memcpy(tmp, u->nodename, i);
1411 	up_read(&uts_sem);
1412 	if (copy_to_user(name, tmp, i))
1413 		return -EFAULT;
1414 	return 0;
1415 }
1416 
1417 #endif
1418 
1419 /*
1420  * Only setdomainname; getdomainname can be implemented by calling
1421  * uname()
1422  */
SYSCALL_DEFINE2(setdomainname,char __user *,name,int,len)1423 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1424 {
1425 	int errno;
1426 	char tmp[__NEW_UTS_LEN];
1427 
1428 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1429 		return -EPERM;
1430 	if (len < 0 || len > __NEW_UTS_LEN)
1431 		return -EINVAL;
1432 
1433 	errno = -EFAULT;
1434 	if (!copy_from_user(tmp, name, len)) {
1435 		struct new_utsname *u;
1436 
1437 		add_device_randomness(tmp, len);
1438 		down_write(&uts_sem);
1439 		u = utsname();
1440 		memcpy(u->domainname, tmp, len);
1441 		memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1442 		errno = 0;
1443 		uts_proc_notify(UTS_PROC_DOMAINNAME);
1444 		up_write(&uts_sem);
1445 	}
1446 	return errno;
1447 }
1448 
1449 /* make sure you are allowed to change @tsk limits before calling this */
do_prlimit(struct task_struct * tsk,unsigned int resource,struct rlimit * new_rlim,struct rlimit * old_rlim)1450 static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1451 		      struct rlimit *new_rlim, struct rlimit *old_rlim)
1452 {
1453 	struct rlimit *rlim;
1454 	int retval = 0;
1455 
1456 	if (resource >= RLIM_NLIMITS)
1457 		return -EINVAL;
1458 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1459 
1460 	if (new_rlim) {
1461 		if (new_rlim->rlim_cur > new_rlim->rlim_max)
1462 			return -EINVAL;
1463 		if (resource == RLIMIT_NOFILE &&
1464 				new_rlim->rlim_max > sysctl_nr_open)
1465 			return -EPERM;
1466 	}
1467 
1468 	/* Holding a refcount on tsk protects tsk->signal from disappearing. */
1469 	rlim = tsk->signal->rlim + resource;
1470 	task_lock(tsk->group_leader);
1471 	if (new_rlim) {
1472 		/*
1473 		 * Keep the capable check against init_user_ns until cgroups can
1474 		 * contain all limits.
1475 		 */
1476 		if (new_rlim->rlim_max > rlim->rlim_max &&
1477 				!capable(CAP_SYS_RESOURCE))
1478 			retval = -EPERM;
1479 		if (!retval)
1480 			retval = security_task_setrlimit(tsk, resource, new_rlim);
1481 	}
1482 	if (!retval) {
1483 		if (old_rlim)
1484 			*old_rlim = *rlim;
1485 		if (new_rlim)
1486 			*rlim = *new_rlim;
1487 	}
1488 	task_unlock(tsk->group_leader);
1489 
1490 	/*
1491 	 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1492 	 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1493 	 * ignores the rlimit.
1494 	 */
1495 	if (!retval && new_rlim && resource == RLIMIT_CPU &&
1496 	    new_rlim->rlim_cur != RLIM_INFINITY &&
1497 	    IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1498 		/*
1499 		 * update_rlimit_cpu can fail if the task is exiting, but there
1500 		 * may be other tasks in the thread group that are not exiting,
1501 		 * and they need their cpu timers adjusted.
1502 		 *
1503 		 * The group_leader is the last task to be released, so if we
1504 		 * cannot update_rlimit_cpu on it, then the entire process is
1505 		 * exiting and we do not need to update at all.
1506 		 */
1507 		update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1508 	}
1509 
1510 	return retval;
1511 }
1512 
SYSCALL_DEFINE2(getrlimit,unsigned int,resource,struct rlimit __user *,rlim)1513 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1514 {
1515 	struct rlimit value;
1516 	int ret;
1517 
1518 	ret = do_prlimit(current, resource, NULL, &value);
1519 	if (!ret)
1520 		ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1521 
1522 	return ret;
1523 }
1524 
1525 #ifdef CONFIG_COMPAT
1526 
COMPAT_SYSCALL_DEFINE2(setrlimit,unsigned int,resource,struct compat_rlimit __user *,rlim)1527 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1528 		       struct compat_rlimit __user *, rlim)
1529 {
1530 	struct rlimit r;
1531 	struct compat_rlimit r32;
1532 
1533 	if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1534 		return -EFAULT;
1535 
1536 	if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1537 		r.rlim_cur = RLIM_INFINITY;
1538 	else
1539 		r.rlim_cur = r32.rlim_cur;
1540 	if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1541 		r.rlim_max = RLIM_INFINITY;
1542 	else
1543 		r.rlim_max = r32.rlim_max;
1544 	return do_prlimit(current, resource, &r, NULL);
1545 }
1546 
COMPAT_SYSCALL_DEFINE2(getrlimit,unsigned int,resource,struct compat_rlimit __user *,rlim)1547 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1548 		       struct compat_rlimit __user *, rlim)
1549 {
1550 	struct rlimit r;
1551 	int ret;
1552 
1553 	ret = do_prlimit(current, resource, NULL, &r);
1554 	if (!ret) {
1555 		struct compat_rlimit r32;
1556 		if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1557 			r32.rlim_cur = COMPAT_RLIM_INFINITY;
1558 		else
1559 			r32.rlim_cur = r.rlim_cur;
1560 		if (r.rlim_max > COMPAT_RLIM_INFINITY)
1561 			r32.rlim_max = COMPAT_RLIM_INFINITY;
1562 		else
1563 			r32.rlim_max = r.rlim_max;
1564 
1565 		if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1566 			return -EFAULT;
1567 	}
1568 	return ret;
1569 }
1570 
1571 #endif
1572 
1573 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1574 
1575 /*
1576  *	Back compatibility for getrlimit. Needed for some apps.
1577  */
SYSCALL_DEFINE2(old_getrlimit,unsigned int,resource,struct rlimit __user *,rlim)1578 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1579 		struct rlimit __user *, rlim)
1580 {
1581 	struct rlimit x;
1582 	if (resource >= RLIM_NLIMITS)
1583 		return -EINVAL;
1584 
1585 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1586 	task_lock(current->group_leader);
1587 	x = current->signal->rlim[resource];
1588 	task_unlock(current->group_leader);
1589 	if (x.rlim_cur > 0x7FFFFFFF)
1590 		x.rlim_cur = 0x7FFFFFFF;
1591 	if (x.rlim_max > 0x7FFFFFFF)
1592 		x.rlim_max = 0x7FFFFFFF;
1593 	return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1594 }
1595 
1596 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(old_getrlimit,unsigned int,resource,struct compat_rlimit __user *,rlim)1597 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1598 		       struct compat_rlimit __user *, rlim)
1599 {
1600 	struct rlimit r;
1601 
1602 	if (resource >= RLIM_NLIMITS)
1603 		return -EINVAL;
1604 
1605 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1606 	task_lock(current->group_leader);
1607 	r = current->signal->rlim[resource];
1608 	task_unlock(current->group_leader);
1609 	if (r.rlim_cur > 0x7FFFFFFF)
1610 		r.rlim_cur = 0x7FFFFFFF;
1611 	if (r.rlim_max > 0x7FFFFFFF)
1612 		r.rlim_max = 0x7FFFFFFF;
1613 
1614 	if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1615 	    put_user(r.rlim_max, &rlim->rlim_max))
1616 		return -EFAULT;
1617 	return 0;
1618 }
1619 #endif
1620 
1621 #endif
1622 
rlim64_is_infinity(__u64 rlim64)1623 static inline bool rlim64_is_infinity(__u64 rlim64)
1624 {
1625 #if BITS_PER_LONG < 64
1626 	return rlim64 >= ULONG_MAX;
1627 #else
1628 	return rlim64 == RLIM64_INFINITY;
1629 #endif
1630 }
1631 
rlim_to_rlim64(const struct rlimit * rlim,struct rlimit64 * rlim64)1632 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1633 {
1634 	if (rlim->rlim_cur == RLIM_INFINITY)
1635 		rlim64->rlim_cur = RLIM64_INFINITY;
1636 	else
1637 		rlim64->rlim_cur = rlim->rlim_cur;
1638 	if (rlim->rlim_max == RLIM_INFINITY)
1639 		rlim64->rlim_max = RLIM64_INFINITY;
1640 	else
1641 		rlim64->rlim_max = rlim->rlim_max;
1642 }
1643 
rlim64_to_rlim(const struct rlimit64 * rlim64,struct rlimit * rlim)1644 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1645 {
1646 	if (rlim64_is_infinity(rlim64->rlim_cur))
1647 		rlim->rlim_cur = RLIM_INFINITY;
1648 	else
1649 		rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1650 	if (rlim64_is_infinity(rlim64->rlim_max))
1651 		rlim->rlim_max = RLIM_INFINITY;
1652 	else
1653 		rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1654 }
1655 
1656 /* rcu lock must be held */
check_prlimit_permission(struct task_struct * task,unsigned int flags)1657 static int check_prlimit_permission(struct task_struct *task,
1658 				    unsigned int flags)
1659 {
1660 	const struct cred *cred = current_cred(), *tcred;
1661 	bool id_match;
1662 
1663 	if (current == task)
1664 		return 0;
1665 
1666 	tcred = __task_cred(task);
1667 	id_match = (uid_eq(cred->uid, tcred->euid) &&
1668 		    uid_eq(cred->uid, tcred->suid) &&
1669 		    uid_eq(cred->uid, tcred->uid)  &&
1670 		    gid_eq(cred->gid, tcred->egid) &&
1671 		    gid_eq(cred->gid, tcred->sgid) &&
1672 		    gid_eq(cred->gid, tcred->gid));
1673 	if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1674 		return -EPERM;
1675 
1676 	return security_task_prlimit(cred, tcred, flags);
1677 }
1678 
SYSCALL_DEFINE4(prlimit64,pid_t,pid,unsigned int,resource,const struct rlimit64 __user *,new_rlim,struct rlimit64 __user *,old_rlim)1679 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1680 		const struct rlimit64 __user *, new_rlim,
1681 		struct rlimit64 __user *, old_rlim)
1682 {
1683 	struct rlimit64 old64, new64;
1684 	struct rlimit old, new;
1685 	struct task_struct *tsk;
1686 	unsigned int checkflags = 0;
1687 	int ret;
1688 
1689 	if (old_rlim)
1690 		checkflags |= LSM_PRLIMIT_READ;
1691 
1692 	if (new_rlim) {
1693 		if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1694 			return -EFAULT;
1695 		rlim64_to_rlim(&new64, &new);
1696 		checkflags |= LSM_PRLIMIT_WRITE;
1697 	}
1698 
1699 	rcu_read_lock();
1700 	tsk = pid ? find_task_by_vpid(pid) : current;
1701 	if (!tsk) {
1702 		rcu_read_unlock();
1703 		return -ESRCH;
1704 	}
1705 	ret = check_prlimit_permission(tsk, checkflags);
1706 	if (ret) {
1707 		rcu_read_unlock();
1708 		return ret;
1709 	}
1710 	get_task_struct(tsk);
1711 	rcu_read_unlock();
1712 
1713 	ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1714 			old_rlim ? &old : NULL);
1715 
1716 	if (!ret && old_rlim) {
1717 		rlim_to_rlim64(&old, &old64);
1718 		if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1719 			ret = -EFAULT;
1720 	}
1721 
1722 	put_task_struct(tsk);
1723 	return ret;
1724 }
1725 
SYSCALL_DEFINE2(setrlimit,unsigned int,resource,struct rlimit __user *,rlim)1726 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1727 {
1728 	struct rlimit new_rlim;
1729 
1730 	if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1731 		return -EFAULT;
1732 	return do_prlimit(current, resource, &new_rlim, NULL);
1733 }
1734 
1735 /*
1736  * It would make sense to put struct rusage in the task_struct,
1737  * except that would make the task_struct be *really big*.  After
1738  * task_struct gets moved into malloc'ed memory, it would
1739  * make sense to do this.  It will make moving the rest of the information
1740  * a lot simpler!  (Which we're not doing right now because we're not
1741  * measuring them yet).
1742  *
1743  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1744  * races with threads incrementing their own counters.  But since word
1745  * reads are atomic, we either get new values or old values and we don't
1746  * care which for the sums.  We always take the siglock to protect reading
1747  * the c* fields from p->signal from races with exit.c updating those
1748  * fields when reaping, so a sample either gets all the additions of a
1749  * given child after it's reaped, or none so this sample is before reaping.
1750  *
1751  * Locking:
1752  * We need to take the siglock for CHILDEREN, SELF and BOTH
1753  * for  the cases current multithreaded, non-current single threaded
1754  * non-current multithreaded.  Thread traversal is now safe with
1755  * the siglock held.
1756  * Strictly speaking, we donot need to take the siglock if we are current and
1757  * single threaded,  as no one else can take our signal_struct away, no one
1758  * else can  reap the  children to update signal->c* counters, and no one else
1759  * can race with the signal-> fields. If we do not take any lock, the
1760  * signal-> fields could be read out of order while another thread was just
1761  * exiting. So we should  place a read memory barrier when we avoid the lock.
1762  * On the writer side,  write memory barrier is implied in  __exit_signal
1763  * as __exit_signal releases  the siglock spinlock after updating the signal->
1764  * fields. But we don't do this yet to keep things simple.
1765  *
1766  */
1767 
accumulate_thread_rusage(struct task_struct * t,struct rusage * r)1768 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1769 {
1770 	r->ru_nvcsw += t->nvcsw;
1771 	r->ru_nivcsw += t->nivcsw;
1772 	r->ru_minflt += t->min_flt;
1773 	r->ru_majflt += t->maj_flt;
1774 	r->ru_inblock += task_io_get_inblock(t);
1775 	r->ru_oublock += task_io_get_oublock(t);
1776 }
1777 
getrusage(struct task_struct * p,int who,struct rusage * r)1778 void getrusage(struct task_struct *p, int who, struct rusage *r)
1779 {
1780 	struct task_struct *t;
1781 	unsigned long flags;
1782 	u64 tgutime, tgstime, utime, stime;
1783 	unsigned long maxrss = 0;
1784 
1785 	memset((char *)r, 0, sizeof (*r));
1786 	utime = stime = 0;
1787 
1788 	if (who == RUSAGE_THREAD) {
1789 		task_cputime_adjusted(current, &utime, &stime);
1790 		accumulate_thread_rusage(p, r);
1791 		maxrss = p->signal->maxrss;
1792 		goto out;
1793 	}
1794 
1795 	if (!lock_task_sighand(p, &flags))
1796 		return;
1797 
1798 	switch (who) {
1799 	case RUSAGE_BOTH:
1800 	case RUSAGE_CHILDREN:
1801 		utime = p->signal->cutime;
1802 		stime = p->signal->cstime;
1803 		r->ru_nvcsw = p->signal->cnvcsw;
1804 		r->ru_nivcsw = p->signal->cnivcsw;
1805 		r->ru_minflt = p->signal->cmin_flt;
1806 		r->ru_majflt = p->signal->cmaj_flt;
1807 		r->ru_inblock = p->signal->cinblock;
1808 		r->ru_oublock = p->signal->coublock;
1809 		maxrss = p->signal->cmaxrss;
1810 
1811 		if (who == RUSAGE_CHILDREN)
1812 			break;
1813 		fallthrough;
1814 
1815 	case RUSAGE_SELF:
1816 		thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1817 		utime += tgutime;
1818 		stime += tgstime;
1819 		r->ru_nvcsw += p->signal->nvcsw;
1820 		r->ru_nivcsw += p->signal->nivcsw;
1821 		r->ru_minflt += p->signal->min_flt;
1822 		r->ru_majflt += p->signal->maj_flt;
1823 		r->ru_inblock += p->signal->inblock;
1824 		r->ru_oublock += p->signal->oublock;
1825 		if (maxrss < p->signal->maxrss)
1826 			maxrss = p->signal->maxrss;
1827 		t = p;
1828 		do {
1829 			accumulate_thread_rusage(t, r);
1830 		} while_each_thread(p, t);
1831 		break;
1832 
1833 	default:
1834 		BUG();
1835 	}
1836 	unlock_task_sighand(p, &flags);
1837 
1838 out:
1839 	r->ru_utime = ns_to_kernel_old_timeval(utime);
1840 	r->ru_stime = ns_to_kernel_old_timeval(stime);
1841 
1842 	if (who != RUSAGE_CHILDREN) {
1843 		struct mm_struct *mm = get_task_mm(p);
1844 
1845 		if (mm) {
1846 			setmax_mm_hiwater_rss(&maxrss, mm);
1847 			mmput(mm);
1848 		}
1849 	}
1850 	r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1851 }
1852 
SYSCALL_DEFINE2(getrusage,int,who,struct rusage __user *,ru)1853 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1854 {
1855 	struct rusage r;
1856 
1857 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1858 	    who != RUSAGE_THREAD)
1859 		return -EINVAL;
1860 
1861 	getrusage(current, who, &r);
1862 	return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1863 }
1864 
1865 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(getrusage,int,who,struct compat_rusage __user *,ru)1866 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1867 {
1868 	struct rusage r;
1869 
1870 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1871 	    who != RUSAGE_THREAD)
1872 		return -EINVAL;
1873 
1874 	getrusage(current, who, &r);
1875 	return put_compat_rusage(&r, ru);
1876 }
1877 #endif
1878 
SYSCALL_DEFINE1(umask,int,mask)1879 SYSCALL_DEFINE1(umask, int, mask)
1880 {
1881 	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1882 	return mask;
1883 }
1884 
prctl_set_mm_exe_file(struct mm_struct * mm,unsigned int fd)1885 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1886 {
1887 	struct fd exe;
1888 	struct inode *inode;
1889 	int err;
1890 
1891 	exe = fdget(fd);
1892 	if (!exe.file)
1893 		return -EBADF;
1894 
1895 	inode = file_inode(exe.file);
1896 
1897 	/*
1898 	 * Because the original mm->exe_file points to executable file, make
1899 	 * sure that this one is executable as well, to avoid breaking an
1900 	 * overall picture.
1901 	 */
1902 	err = -EACCES;
1903 	if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1904 		goto exit;
1905 
1906 	err = file_permission(exe.file, MAY_EXEC);
1907 	if (err)
1908 		goto exit;
1909 
1910 	err = replace_mm_exe_file(mm, exe.file);
1911 exit:
1912 	fdput(exe);
1913 	return err;
1914 }
1915 
1916 /*
1917  * Check arithmetic relations of passed addresses.
1918  *
1919  * WARNING: we don't require any capability here so be very careful
1920  * in what is allowed for modification from userspace.
1921  */
validate_prctl_map_addr(struct prctl_mm_map * prctl_map)1922 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1923 {
1924 	unsigned long mmap_max_addr = TASK_SIZE;
1925 	int error = -EINVAL, i;
1926 
1927 	static const unsigned char offsets[] = {
1928 		offsetof(struct prctl_mm_map, start_code),
1929 		offsetof(struct prctl_mm_map, end_code),
1930 		offsetof(struct prctl_mm_map, start_data),
1931 		offsetof(struct prctl_mm_map, end_data),
1932 		offsetof(struct prctl_mm_map, start_brk),
1933 		offsetof(struct prctl_mm_map, brk),
1934 		offsetof(struct prctl_mm_map, start_stack),
1935 		offsetof(struct prctl_mm_map, arg_start),
1936 		offsetof(struct prctl_mm_map, arg_end),
1937 		offsetof(struct prctl_mm_map, env_start),
1938 		offsetof(struct prctl_mm_map, env_end),
1939 	};
1940 
1941 	/*
1942 	 * Make sure the members are not somewhere outside
1943 	 * of allowed address space.
1944 	 */
1945 	for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1946 		u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1947 
1948 		if ((unsigned long)val >= mmap_max_addr ||
1949 		    (unsigned long)val < mmap_min_addr)
1950 			goto out;
1951 	}
1952 
1953 	/*
1954 	 * Make sure the pairs are ordered.
1955 	 */
1956 #define __prctl_check_order(__m1, __op, __m2)				\
1957 	((unsigned long)prctl_map->__m1 __op				\
1958 	 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1959 	error  = __prctl_check_order(start_code, <, end_code);
1960 	error |= __prctl_check_order(start_data,<=, end_data);
1961 	error |= __prctl_check_order(start_brk, <=, brk);
1962 	error |= __prctl_check_order(arg_start, <=, arg_end);
1963 	error |= __prctl_check_order(env_start, <=, env_end);
1964 	if (error)
1965 		goto out;
1966 #undef __prctl_check_order
1967 
1968 	error = -EINVAL;
1969 
1970 	/*
1971 	 * Neither we should allow to override limits if they set.
1972 	 */
1973 	if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1974 			      prctl_map->start_brk, prctl_map->end_data,
1975 			      prctl_map->start_data))
1976 			goto out;
1977 
1978 	error = 0;
1979 out:
1980 	return error;
1981 }
1982 
1983 #ifdef CONFIG_CHECKPOINT_RESTORE
prctl_set_mm_map(int opt,const void __user * addr,unsigned long data_size)1984 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1985 {
1986 	struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1987 	unsigned long user_auxv[AT_VECTOR_SIZE];
1988 	struct mm_struct *mm = current->mm;
1989 	int error;
1990 
1991 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1992 	BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1993 
1994 	if (opt == PR_SET_MM_MAP_SIZE)
1995 		return put_user((unsigned int)sizeof(prctl_map),
1996 				(unsigned int __user *)addr);
1997 
1998 	if (data_size != sizeof(prctl_map))
1999 		return -EINVAL;
2000 
2001 	if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
2002 		return -EFAULT;
2003 
2004 	error = validate_prctl_map_addr(&prctl_map);
2005 	if (error)
2006 		return error;
2007 
2008 	if (prctl_map.auxv_size) {
2009 		/*
2010 		 * Someone is trying to cheat the auxv vector.
2011 		 */
2012 		if (!prctl_map.auxv ||
2013 				prctl_map.auxv_size > sizeof(mm->saved_auxv))
2014 			return -EINVAL;
2015 
2016 		memset(user_auxv, 0, sizeof(user_auxv));
2017 		if (copy_from_user(user_auxv,
2018 				   (const void __user *)prctl_map.auxv,
2019 				   prctl_map.auxv_size))
2020 			return -EFAULT;
2021 
2022 		/* Last entry must be AT_NULL as specification requires */
2023 		user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2024 		user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2025 	}
2026 
2027 	if (prctl_map.exe_fd != (u32)-1) {
2028 		/*
2029 		 * Check if the current user is checkpoint/restore capable.
2030 		 * At the time of this writing, it checks for CAP_SYS_ADMIN
2031 		 * or CAP_CHECKPOINT_RESTORE.
2032 		 * Note that a user with access to ptrace can masquerade an
2033 		 * arbitrary program as any executable, even setuid ones.
2034 		 * This may have implications in the tomoyo subsystem.
2035 		 */
2036 		if (!checkpoint_restore_ns_capable(current_user_ns()))
2037 			return -EPERM;
2038 
2039 		error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2040 		if (error)
2041 			return error;
2042 	}
2043 
2044 	/*
2045 	 * arg_lock protects concurrent updates but we still need mmap_lock for
2046 	 * read to exclude races with sys_brk.
2047 	 */
2048 	mmap_read_lock(mm);
2049 
2050 	/*
2051 	 * We don't validate if these members are pointing to
2052 	 * real present VMAs because application may have correspond
2053 	 * VMAs already unmapped and kernel uses these members for statistics
2054 	 * output in procfs mostly, except
2055 	 *
2056 	 *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2057 	 *    for VMAs when updating these members so anything wrong written
2058 	 *    here cause kernel to swear at userspace program but won't lead
2059 	 *    to any problem in kernel itself
2060 	 */
2061 
2062 	spin_lock(&mm->arg_lock);
2063 	mm->start_code	= prctl_map.start_code;
2064 	mm->end_code	= prctl_map.end_code;
2065 	mm->start_data	= prctl_map.start_data;
2066 	mm->end_data	= prctl_map.end_data;
2067 	mm->start_brk	= prctl_map.start_brk;
2068 	mm->brk		= prctl_map.brk;
2069 	mm->start_stack	= prctl_map.start_stack;
2070 	mm->arg_start	= prctl_map.arg_start;
2071 	mm->arg_end	= prctl_map.arg_end;
2072 	mm->env_start	= prctl_map.env_start;
2073 	mm->env_end	= prctl_map.env_end;
2074 	spin_unlock(&mm->arg_lock);
2075 
2076 	/*
2077 	 * Note this update of @saved_auxv is lockless thus
2078 	 * if someone reads this member in procfs while we're
2079 	 * updating -- it may get partly updated results. It's
2080 	 * known and acceptable trade off: we leave it as is to
2081 	 * not introduce additional locks here making the kernel
2082 	 * more complex.
2083 	 */
2084 	if (prctl_map.auxv_size)
2085 		memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2086 
2087 	mmap_read_unlock(mm);
2088 	return 0;
2089 }
2090 #endif /* CONFIG_CHECKPOINT_RESTORE */
2091 
prctl_set_auxv(struct mm_struct * mm,unsigned long addr,unsigned long len)2092 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2093 			  unsigned long len)
2094 {
2095 	/*
2096 	 * This doesn't move the auxiliary vector itself since it's pinned to
2097 	 * mm_struct, but it permits filling the vector with new values.  It's
2098 	 * up to the caller to provide sane values here, otherwise userspace
2099 	 * tools which use this vector might be unhappy.
2100 	 */
2101 	unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2102 
2103 	if (len > sizeof(user_auxv))
2104 		return -EINVAL;
2105 
2106 	if (copy_from_user(user_auxv, (const void __user *)addr, len))
2107 		return -EFAULT;
2108 
2109 	/* Make sure the last entry is always AT_NULL */
2110 	user_auxv[AT_VECTOR_SIZE - 2] = 0;
2111 	user_auxv[AT_VECTOR_SIZE - 1] = 0;
2112 
2113 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2114 
2115 	task_lock(current);
2116 	memcpy(mm->saved_auxv, user_auxv, len);
2117 	task_unlock(current);
2118 
2119 	return 0;
2120 }
2121 
prctl_set_mm(int opt,unsigned long addr,unsigned long arg4,unsigned long arg5)2122 static int prctl_set_mm(int opt, unsigned long addr,
2123 			unsigned long arg4, unsigned long arg5)
2124 {
2125 	struct mm_struct *mm = current->mm;
2126 	struct prctl_mm_map prctl_map = {
2127 		.auxv = NULL,
2128 		.auxv_size = 0,
2129 		.exe_fd = -1,
2130 	};
2131 	struct vm_area_struct *vma;
2132 	int error;
2133 
2134 	if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2135 			      opt != PR_SET_MM_MAP &&
2136 			      opt != PR_SET_MM_MAP_SIZE)))
2137 		return -EINVAL;
2138 
2139 #ifdef CONFIG_CHECKPOINT_RESTORE
2140 	if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2141 		return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2142 #endif
2143 
2144 	if (!capable(CAP_SYS_RESOURCE))
2145 		return -EPERM;
2146 
2147 	if (opt == PR_SET_MM_EXE_FILE)
2148 		return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2149 
2150 	if (opt == PR_SET_MM_AUXV)
2151 		return prctl_set_auxv(mm, addr, arg4);
2152 
2153 	if (addr >= TASK_SIZE || addr < mmap_min_addr)
2154 		return -EINVAL;
2155 
2156 	error = -EINVAL;
2157 
2158 	/*
2159 	 * arg_lock protects concurrent updates of arg boundaries, we need
2160 	 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2161 	 * validation.
2162 	 */
2163 	mmap_read_lock(mm);
2164 	vma = find_vma(mm, addr);
2165 
2166 	spin_lock(&mm->arg_lock);
2167 	prctl_map.start_code	= mm->start_code;
2168 	prctl_map.end_code	= mm->end_code;
2169 	prctl_map.start_data	= mm->start_data;
2170 	prctl_map.end_data	= mm->end_data;
2171 	prctl_map.start_brk	= mm->start_brk;
2172 	prctl_map.brk		= mm->brk;
2173 	prctl_map.start_stack	= mm->start_stack;
2174 	prctl_map.arg_start	= mm->arg_start;
2175 	prctl_map.arg_end	= mm->arg_end;
2176 	prctl_map.env_start	= mm->env_start;
2177 	prctl_map.env_end	= mm->env_end;
2178 
2179 	switch (opt) {
2180 	case PR_SET_MM_START_CODE:
2181 		prctl_map.start_code = addr;
2182 		break;
2183 	case PR_SET_MM_END_CODE:
2184 		prctl_map.end_code = addr;
2185 		break;
2186 	case PR_SET_MM_START_DATA:
2187 		prctl_map.start_data = addr;
2188 		break;
2189 	case PR_SET_MM_END_DATA:
2190 		prctl_map.end_data = addr;
2191 		break;
2192 	case PR_SET_MM_START_STACK:
2193 		prctl_map.start_stack = addr;
2194 		break;
2195 	case PR_SET_MM_START_BRK:
2196 		prctl_map.start_brk = addr;
2197 		break;
2198 	case PR_SET_MM_BRK:
2199 		prctl_map.brk = addr;
2200 		break;
2201 	case PR_SET_MM_ARG_START:
2202 		prctl_map.arg_start = addr;
2203 		break;
2204 	case PR_SET_MM_ARG_END:
2205 		prctl_map.arg_end = addr;
2206 		break;
2207 	case PR_SET_MM_ENV_START:
2208 		prctl_map.env_start = addr;
2209 		break;
2210 	case PR_SET_MM_ENV_END:
2211 		prctl_map.env_end = addr;
2212 		break;
2213 	default:
2214 		goto out;
2215 	}
2216 
2217 	error = validate_prctl_map_addr(&prctl_map);
2218 	if (error)
2219 		goto out;
2220 
2221 	switch (opt) {
2222 	/*
2223 	 * If command line arguments and environment
2224 	 * are placed somewhere else on stack, we can
2225 	 * set them up here, ARG_START/END to setup
2226 	 * command line arguments and ENV_START/END
2227 	 * for environment.
2228 	 */
2229 	case PR_SET_MM_START_STACK:
2230 	case PR_SET_MM_ARG_START:
2231 	case PR_SET_MM_ARG_END:
2232 	case PR_SET_MM_ENV_START:
2233 	case PR_SET_MM_ENV_END:
2234 		if (!vma) {
2235 			error = -EFAULT;
2236 			goto out;
2237 		}
2238 	}
2239 
2240 	mm->start_code	= prctl_map.start_code;
2241 	mm->end_code	= prctl_map.end_code;
2242 	mm->start_data	= prctl_map.start_data;
2243 	mm->end_data	= prctl_map.end_data;
2244 	mm->start_brk	= prctl_map.start_brk;
2245 	mm->brk		= prctl_map.brk;
2246 	mm->start_stack	= prctl_map.start_stack;
2247 	mm->arg_start	= prctl_map.arg_start;
2248 	mm->arg_end	= prctl_map.arg_end;
2249 	mm->env_start	= prctl_map.env_start;
2250 	mm->env_end	= prctl_map.env_end;
2251 
2252 	error = 0;
2253 out:
2254 	spin_unlock(&mm->arg_lock);
2255 	mmap_read_unlock(mm);
2256 	return error;
2257 }
2258 
2259 #ifdef CONFIG_CHECKPOINT_RESTORE
prctl_get_tid_address(struct task_struct * me,int __user * __user * tid_addr)2260 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2261 {
2262 	return put_user(me->clear_child_tid, tid_addr);
2263 }
2264 #else
prctl_get_tid_address(struct task_struct * me,int __user * __user * tid_addr)2265 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2266 {
2267 	return -EINVAL;
2268 }
2269 #endif
2270 
propagate_has_child_subreaper(struct task_struct * p,void * data)2271 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2272 {
2273 	/*
2274 	 * If task has has_child_subreaper - all its descendants
2275 	 * already have these flag too and new descendants will
2276 	 * inherit it on fork, skip them.
2277 	 *
2278 	 * If we've found child_reaper - skip descendants in
2279 	 * it's subtree as they will never get out pidns.
2280 	 */
2281 	if (p->signal->has_child_subreaper ||
2282 	    is_child_reaper(task_pid(p)))
2283 		return 0;
2284 
2285 	p->signal->has_child_subreaper = 1;
2286 	return 1;
2287 }
2288 
arch_prctl_spec_ctrl_get(struct task_struct * t,unsigned long which)2289 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2290 {
2291 	return -EINVAL;
2292 }
2293 
arch_prctl_spec_ctrl_set(struct task_struct * t,unsigned long which,unsigned long ctrl)2294 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2295 				    unsigned long ctrl)
2296 {
2297 	return -EINVAL;
2298 }
2299 
2300 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2301 
2302 #ifdef CONFIG_ANON_VMA_NAME
2303 
2304 #define ANON_VMA_NAME_MAX_LEN		80
2305 #define ANON_VMA_NAME_INVALID_CHARS	"\\`$[]"
2306 
is_valid_name_char(char ch)2307 static inline bool is_valid_name_char(char ch)
2308 {
2309 	/* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2310 	return ch > 0x1f && ch < 0x7f &&
2311 		!strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2312 }
2313 
prctl_set_vma(unsigned long opt,unsigned long addr,unsigned long size,unsigned long arg)2314 static int prctl_set_vma(unsigned long opt, unsigned long addr,
2315 			 unsigned long size, unsigned long arg)
2316 {
2317 	struct mm_struct *mm = current->mm;
2318 	const char __user *uname;
2319 	struct anon_vma_name *anon_name = NULL;
2320 	int error;
2321 
2322 	switch (opt) {
2323 	case PR_SET_VMA_ANON_NAME:
2324 		uname = (const char __user *)arg;
2325 		if (uname) {
2326 			char *name, *pch;
2327 
2328 			name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2329 			if (IS_ERR(name))
2330 				return PTR_ERR(name);
2331 
2332 			for (pch = name; *pch != '\0'; pch++) {
2333 				if (!is_valid_name_char(*pch)) {
2334 					kfree(name);
2335 					return -EINVAL;
2336 				}
2337 			}
2338 			/* anon_vma has its own copy */
2339 			anon_name = anon_vma_name_alloc(name);
2340 			kfree(name);
2341 			if (!anon_name)
2342 				return -ENOMEM;
2343 
2344 		}
2345 
2346 		mmap_write_lock(mm);
2347 		error = madvise_set_anon_name(mm, addr, size, anon_name);
2348 		mmap_write_unlock(mm);
2349 		anon_vma_name_put(anon_name);
2350 		break;
2351 	default:
2352 		error = -EINVAL;
2353 	}
2354 
2355 	return error;
2356 }
2357 
2358 #else /* CONFIG_ANON_VMA_NAME */
prctl_set_vma(unsigned long opt,unsigned long start,unsigned long size,unsigned long arg)2359 static int prctl_set_vma(unsigned long opt, unsigned long start,
2360 			 unsigned long size, unsigned long arg)
2361 {
2362 	return -EINVAL;
2363 }
2364 #endif /* CONFIG_ANON_VMA_NAME */
2365 
SYSCALL_DEFINE5(prctl,int,option,unsigned long,arg2,unsigned long,arg3,unsigned long,arg4,unsigned long,arg5)2366 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2367 		unsigned long, arg4, unsigned long, arg5)
2368 {
2369 	struct task_struct *me = current;
2370 	unsigned char comm[sizeof(me->comm)];
2371 	long error;
2372 
2373 	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2374 	if (error != -ENOSYS)
2375 		return error;
2376 
2377 	error = 0;
2378 	switch (option) {
2379 	case PR_SET_PDEATHSIG:
2380 		if (!valid_signal(arg2)) {
2381 			error = -EINVAL;
2382 			break;
2383 		}
2384 		me->pdeath_signal = arg2;
2385 		break;
2386 	case PR_GET_PDEATHSIG:
2387 		error = put_user(me->pdeath_signal, (int __user *)arg2);
2388 		break;
2389 	case PR_GET_DUMPABLE:
2390 		error = get_dumpable(me->mm);
2391 		break;
2392 	case PR_SET_DUMPABLE:
2393 		if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2394 			error = -EINVAL;
2395 			break;
2396 		}
2397 		set_dumpable(me->mm, arg2);
2398 		break;
2399 
2400 	case PR_SET_UNALIGN:
2401 		error = SET_UNALIGN_CTL(me, arg2);
2402 		break;
2403 	case PR_GET_UNALIGN:
2404 		error = GET_UNALIGN_CTL(me, arg2);
2405 		break;
2406 	case PR_SET_FPEMU:
2407 		error = SET_FPEMU_CTL(me, arg2);
2408 		break;
2409 	case PR_GET_FPEMU:
2410 		error = GET_FPEMU_CTL(me, arg2);
2411 		break;
2412 	case PR_SET_FPEXC:
2413 		error = SET_FPEXC_CTL(me, arg2);
2414 		break;
2415 	case PR_GET_FPEXC:
2416 		error = GET_FPEXC_CTL(me, arg2);
2417 		break;
2418 	case PR_GET_TIMING:
2419 		error = PR_TIMING_STATISTICAL;
2420 		break;
2421 	case PR_SET_TIMING:
2422 		if (arg2 != PR_TIMING_STATISTICAL)
2423 			error = -EINVAL;
2424 		break;
2425 	case PR_SET_NAME:
2426 		comm[sizeof(me->comm) - 1] = 0;
2427 		if (strncpy_from_user(comm, (char __user *)arg2,
2428 				      sizeof(me->comm) - 1) < 0)
2429 			return -EFAULT;
2430 		set_task_comm(me, comm);
2431 		proc_comm_connector(me);
2432 		break;
2433 	case PR_GET_NAME:
2434 		get_task_comm(comm, me);
2435 		if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2436 			return -EFAULT;
2437 		break;
2438 	case PR_GET_ENDIAN:
2439 		error = GET_ENDIAN(me, arg2);
2440 		break;
2441 	case PR_SET_ENDIAN:
2442 		error = SET_ENDIAN(me, arg2);
2443 		break;
2444 	case PR_GET_SECCOMP:
2445 		error = prctl_get_seccomp();
2446 		break;
2447 	case PR_SET_SECCOMP:
2448 		error = prctl_set_seccomp(arg2, (char __user *)arg3);
2449 		break;
2450 	case PR_GET_TSC:
2451 		error = GET_TSC_CTL(arg2);
2452 		break;
2453 	case PR_SET_TSC:
2454 		error = SET_TSC_CTL(arg2);
2455 		break;
2456 	case PR_TASK_PERF_EVENTS_DISABLE:
2457 		error = perf_event_task_disable();
2458 		break;
2459 	case PR_TASK_PERF_EVENTS_ENABLE:
2460 		error = perf_event_task_enable();
2461 		break;
2462 	case PR_GET_TIMERSLACK:
2463 		if (current->timer_slack_ns > ULONG_MAX)
2464 			error = ULONG_MAX;
2465 		else
2466 			error = current->timer_slack_ns;
2467 		break;
2468 	case PR_SET_TIMERSLACK:
2469 		if (arg2 <= 0)
2470 			current->timer_slack_ns =
2471 					current->default_timer_slack_ns;
2472 		else
2473 			current->timer_slack_ns = arg2;
2474 		break;
2475 	case PR_MCE_KILL:
2476 		if (arg4 | arg5)
2477 			return -EINVAL;
2478 		switch (arg2) {
2479 		case PR_MCE_KILL_CLEAR:
2480 			if (arg3 != 0)
2481 				return -EINVAL;
2482 			current->flags &= ~PF_MCE_PROCESS;
2483 			break;
2484 		case PR_MCE_KILL_SET:
2485 			current->flags |= PF_MCE_PROCESS;
2486 			if (arg3 == PR_MCE_KILL_EARLY)
2487 				current->flags |= PF_MCE_EARLY;
2488 			else if (arg3 == PR_MCE_KILL_LATE)
2489 				current->flags &= ~PF_MCE_EARLY;
2490 			else if (arg3 == PR_MCE_KILL_DEFAULT)
2491 				current->flags &=
2492 						~(PF_MCE_EARLY|PF_MCE_PROCESS);
2493 			else
2494 				return -EINVAL;
2495 			break;
2496 		default:
2497 			return -EINVAL;
2498 		}
2499 		break;
2500 	case PR_MCE_KILL_GET:
2501 		if (arg2 | arg3 | arg4 | arg5)
2502 			return -EINVAL;
2503 		if (current->flags & PF_MCE_PROCESS)
2504 			error = (current->flags & PF_MCE_EARLY) ?
2505 				PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2506 		else
2507 			error = PR_MCE_KILL_DEFAULT;
2508 		break;
2509 	case PR_SET_MM:
2510 		error = prctl_set_mm(arg2, arg3, arg4, arg5);
2511 		break;
2512 	case PR_GET_TID_ADDRESS:
2513 		error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2514 		break;
2515 	case PR_SET_CHILD_SUBREAPER:
2516 		me->signal->is_child_subreaper = !!arg2;
2517 		if (!arg2)
2518 			break;
2519 
2520 		walk_process_tree(me, propagate_has_child_subreaper, NULL);
2521 		break;
2522 	case PR_GET_CHILD_SUBREAPER:
2523 		error = put_user(me->signal->is_child_subreaper,
2524 				 (int __user *)arg2);
2525 		break;
2526 	case PR_SET_NO_NEW_PRIVS:
2527 		if (arg2 != 1 || arg3 || arg4 || arg5)
2528 			return -EINVAL;
2529 
2530 		task_set_no_new_privs(current);
2531 		break;
2532 	case PR_GET_NO_NEW_PRIVS:
2533 		if (arg2 || arg3 || arg4 || arg5)
2534 			return -EINVAL;
2535 		return task_no_new_privs(current) ? 1 : 0;
2536 	case PR_GET_THP_DISABLE:
2537 		if (arg2 || arg3 || arg4 || arg5)
2538 			return -EINVAL;
2539 		error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2540 		break;
2541 	case PR_SET_THP_DISABLE:
2542 		if (arg3 || arg4 || arg5)
2543 			return -EINVAL;
2544 		if (mmap_write_lock_killable(me->mm))
2545 			return -EINTR;
2546 		if (arg2)
2547 			set_bit(MMF_DISABLE_THP, &me->mm->flags);
2548 		else
2549 			clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2550 		mmap_write_unlock(me->mm);
2551 		break;
2552 	case PR_MPX_ENABLE_MANAGEMENT:
2553 	case PR_MPX_DISABLE_MANAGEMENT:
2554 		/* No longer implemented: */
2555 		return -EINVAL;
2556 	case PR_SET_FP_MODE:
2557 		error = SET_FP_MODE(me, arg2);
2558 		break;
2559 	case PR_GET_FP_MODE:
2560 		error = GET_FP_MODE(me);
2561 		break;
2562 	case PR_SVE_SET_VL:
2563 		error = SVE_SET_VL(arg2);
2564 		break;
2565 	case PR_SVE_GET_VL:
2566 		error = SVE_GET_VL();
2567 		break;
2568 	case PR_SME_SET_VL:
2569 		error = SME_SET_VL(arg2);
2570 		break;
2571 	case PR_SME_GET_VL:
2572 		error = SME_GET_VL();
2573 		break;
2574 	case PR_GET_SPECULATION_CTRL:
2575 		if (arg3 || arg4 || arg5)
2576 			return -EINVAL;
2577 		error = arch_prctl_spec_ctrl_get(me, arg2);
2578 		break;
2579 	case PR_SET_SPECULATION_CTRL:
2580 		if (arg4 || arg5)
2581 			return -EINVAL;
2582 		error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2583 		break;
2584 	case PR_PAC_RESET_KEYS:
2585 		if (arg3 || arg4 || arg5)
2586 			return -EINVAL;
2587 		error = PAC_RESET_KEYS(me, arg2);
2588 		break;
2589 	case PR_PAC_SET_ENABLED_KEYS:
2590 		if (arg4 || arg5)
2591 			return -EINVAL;
2592 		error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2593 		break;
2594 	case PR_PAC_GET_ENABLED_KEYS:
2595 		if (arg2 || arg3 || arg4 || arg5)
2596 			return -EINVAL;
2597 		error = PAC_GET_ENABLED_KEYS(me);
2598 		break;
2599 	case PR_SET_TAGGED_ADDR_CTRL:
2600 		if (arg3 || arg4 || arg5)
2601 			return -EINVAL;
2602 		error = SET_TAGGED_ADDR_CTRL(arg2);
2603 		break;
2604 	case PR_GET_TAGGED_ADDR_CTRL:
2605 		if (arg2 || arg3 || arg4 || arg5)
2606 			return -EINVAL;
2607 		error = GET_TAGGED_ADDR_CTRL();
2608 		break;
2609 	case PR_SET_IO_FLUSHER:
2610 		if (!capable(CAP_SYS_RESOURCE))
2611 			return -EPERM;
2612 
2613 		if (arg3 || arg4 || arg5)
2614 			return -EINVAL;
2615 
2616 		if (arg2 == 1)
2617 			current->flags |= PR_IO_FLUSHER;
2618 		else if (!arg2)
2619 			current->flags &= ~PR_IO_FLUSHER;
2620 		else
2621 			return -EINVAL;
2622 		break;
2623 	case PR_GET_IO_FLUSHER:
2624 		if (!capable(CAP_SYS_RESOURCE))
2625 			return -EPERM;
2626 
2627 		if (arg2 || arg3 || arg4 || arg5)
2628 			return -EINVAL;
2629 
2630 		error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2631 		break;
2632 	case PR_SET_SYSCALL_USER_DISPATCH:
2633 		error = set_syscall_user_dispatch(arg2, arg3, arg4,
2634 						  (char __user *) arg5);
2635 		break;
2636 #ifdef CONFIG_SCHED_CORE
2637 	case PR_SCHED_CORE:
2638 		error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2639 		break;
2640 #endif
2641 	case PR_SET_VMA:
2642 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
2643 		break;
2644 	default:
2645 		error = -EINVAL;
2646 		break;
2647 	}
2648 	trace_android_vh_syscall_prctl_finished(option, me);
2649 	return error;
2650 }
2651 
SYSCALL_DEFINE3(getcpu,unsigned __user *,cpup,unsigned __user *,nodep,struct getcpu_cache __user *,unused)2652 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2653 		struct getcpu_cache __user *, unused)
2654 {
2655 	int err = 0;
2656 	int cpu = raw_smp_processor_id();
2657 
2658 	if (cpup)
2659 		err |= put_user(cpu, cpup);
2660 	if (nodep)
2661 		err |= put_user(cpu_to_node(cpu), nodep);
2662 	return err ? -EFAULT : 0;
2663 }
2664 
2665 /**
2666  * do_sysinfo - fill in sysinfo struct
2667  * @info: pointer to buffer to fill
2668  */
do_sysinfo(struct sysinfo * info)2669 static int do_sysinfo(struct sysinfo *info)
2670 {
2671 	unsigned long mem_total, sav_total;
2672 	unsigned int mem_unit, bitcount;
2673 	struct timespec64 tp;
2674 
2675 	memset(info, 0, sizeof(struct sysinfo));
2676 
2677 	ktime_get_boottime_ts64(&tp);
2678 	timens_add_boottime(&tp);
2679 	info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2680 
2681 	get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2682 
2683 	info->procs = nr_threads;
2684 
2685 	si_meminfo(info);
2686 	si_swapinfo(info);
2687 
2688 	/*
2689 	 * If the sum of all the available memory (i.e. ram + swap)
2690 	 * is less than can be stored in a 32 bit unsigned long then
2691 	 * we can be binary compatible with 2.2.x kernels.  If not,
2692 	 * well, in that case 2.2.x was broken anyways...
2693 	 *
2694 	 *  -Erik Andersen <andersee@debian.org>
2695 	 */
2696 
2697 	mem_total = info->totalram + info->totalswap;
2698 	if (mem_total < info->totalram || mem_total < info->totalswap)
2699 		goto out;
2700 	bitcount = 0;
2701 	mem_unit = info->mem_unit;
2702 	while (mem_unit > 1) {
2703 		bitcount++;
2704 		mem_unit >>= 1;
2705 		sav_total = mem_total;
2706 		mem_total <<= 1;
2707 		if (mem_total < sav_total)
2708 			goto out;
2709 	}
2710 
2711 	/*
2712 	 * If mem_total did not overflow, multiply all memory values by
2713 	 * info->mem_unit and set it to 1.  This leaves things compatible
2714 	 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2715 	 * kernels...
2716 	 */
2717 
2718 	info->mem_unit = 1;
2719 	info->totalram <<= bitcount;
2720 	info->freeram <<= bitcount;
2721 	info->sharedram <<= bitcount;
2722 	info->bufferram <<= bitcount;
2723 	info->totalswap <<= bitcount;
2724 	info->freeswap <<= bitcount;
2725 	info->totalhigh <<= bitcount;
2726 	info->freehigh <<= bitcount;
2727 
2728 out:
2729 	return 0;
2730 }
2731 
SYSCALL_DEFINE1(sysinfo,struct sysinfo __user *,info)2732 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2733 {
2734 	struct sysinfo val;
2735 
2736 	do_sysinfo(&val);
2737 
2738 	if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2739 		return -EFAULT;
2740 
2741 	return 0;
2742 }
2743 
2744 #ifdef CONFIG_COMPAT
2745 struct compat_sysinfo {
2746 	s32 uptime;
2747 	u32 loads[3];
2748 	u32 totalram;
2749 	u32 freeram;
2750 	u32 sharedram;
2751 	u32 bufferram;
2752 	u32 totalswap;
2753 	u32 freeswap;
2754 	u16 procs;
2755 	u16 pad;
2756 	u32 totalhigh;
2757 	u32 freehigh;
2758 	u32 mem_unit;
2759 	char _f[20-2*sizeof(u32)-sizeof(int)];
2760 };
2761 
COMPAT_SYSCALL_DEFINE1(sysinfo,struct compat_sysinfo __user *,info)2762 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2763 {
2764 	struct sysinfo s;
2765 	struct compat_sysinfo s_32;
2766 
2767 	do_sysinfo(&s);
2768 
2769 	/* Check to see if any memory value is too large for 32-bit and scale
2770 	 *  down if needed
2771 	 */
2772 	if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2773 		int bitcount = 0;
2774 
2775 		while (s.mem_unit < PAGE_SIZE) {
2776 			s.mem_unit <<= 1;
2777 			bitcount++;
2778 		}
2779 
2780 		s.totalram >>= bitcount;
2781 		s.freeram >>= bitcount;
2782 		s.sharedram >>= bitcount;
2783 		s.bufferram >>= bitcount;
2784 		s.totalswap >>= bitcount;
2785 		s.freeswap >>= bitcount;
2786 		s.totalhigh >>= bitcount;
2787 		s.freehigh >>= bitcount;
2788 	}
2789 
2790 	memset(&s_32, 0, sizeof(s_32));
2791 	s_32.uptime = s.uptime;
2792 	s_32.loads[0] = s.loads[0];
2793 	s_32.loads[1] = s.loads[1];
2794 	s_32.loads[2] = s.loads[2];
2795 	s_32.totalram = s.totalram;
2796 	s_32.freeram = s.freeram;
2797 	s_32.sharedram = s.sharedram;
2798 	s_32.bufferram = s.bufferram;
2799 	s_32.totalswap = s.totalswap;
2800 	s_32.freeswap = s.freeswap;
2801 	s_32.procs = s.procs;
2802 	s_32.totalhigh = s.totalhigh;
2803 	s_32.freehigh = s.freehigh;
2804 	s_32.mem_unit = s.mem_unit;
2805 	if (copy_to_user(info, &s_32, sizeof(s_32)))
2806 		return -EFAULT;
2807 	return 0;
2808 }
2809 #endif /* CONFIG_COMPAT */
2810