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