sys.c source code [linux/kernel/sys.c]

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/ksm.h>
19	#include <linux/perf_event.h>
20	#include <linux/resource.h>
21	#include <linux/kernel.h>
22	#include <linux/workqueue.h>
23	#include <linux/capability.h>
24	#include <linux/device.h>
25	#include <linux/key.h>
26	#include <linux/times.h>
27	#include <linux/posix-timers.h>
28	#include <linux/security.h>
29	#include <linux/random.h>
30	#include <linux/suspend.h>
31	#include <linux/tty.h>
32	#include <linux/signal.h>
33	#include <linux/cn_proc.h>
34	#include <linux/getcpu.h>
35	#include <linux/task_io_accounting_ops.h>
36	#include <linux/seccomp.h>
37	#include <linux/cpu.h>
38	#include <linux/personality.h>
39	#include <linux/ptrace.h>
40	#include <linux/fs_struct.h>
41	#include <linux/file.h>
42	#include <linux/mount.h>
43	#include <linux/gfp.h>
44	#include <linux/syscore_ops.h>
45	#include <linux/version.h>
46	#include <linux/ctype.h>
47	#include <linux/syscall_user_dispatch.h>
48
49	#include <linux/compat.h>
50	#include <linux/syscalls.h>
51	#include <linux/kprobes.h>
52	#include <linux/user_namespace.h>
53	#include <linux/time_namespace.h>
54	#include <linux/binfmts.h>
55
56	#include <linux/sched.h>
57	#include <linux/sched/autogroup.h>
58	#include <linux/sched/loadavg.h>
59	#include <linux/sched/stat.h>
60	#include <linux/sched/mm.h>
61	#include <linux/sched/coredump.h>
62	#include <linux/sched/task.h>
63	#include <linux/sched/cputime.h>
64	#include <linux/rcupdate.h>
65	#include <linux/uidgid.h>
66	#include <linux/cred.h>
67
68	#include <linux/nospec.h>
69
70	#include <linux/kmsg_dump.h>
71	/ Move somewhere else to avoid recompiling? /
72	#include <generated/utsrelease.h>
73
74	#include <linux/uaccess.h>
75	#include <asm/io.h>
76	#include <asm/unistd.h>
77
78	#include "uid16.h"
79
80	#ifndef SET_UNALIGN_CTL
81	# define SET_UNALIGN_CTL(a, b) (-EINVAL)
82	#endif
83	#ifndef GET_UNALIGN_CTL
84	# define GET_UNALIGN_CTL(a, b) (-EINVAL)
85	#endif
86	#ifndef SET_FPEMU_CTL
87	# define SET_FPEMU_CTL(a, b) (-EINVAL)
88	#endif
89	#ifndef GET_FPEMU_CTL
90	# define GET_FPEMU_CTL(a, b) (-EINVAL)
91	#endif
92	#ifndef SET_FPEXC_CTL
93	# define SET_FPEXC_CTL(a, b) (-EINVAL)
94	#endif
95	#ifndef GET_FPEXC_CTL
96	# define GET_FPEXC_CTL(a, b) (-EINVAL)
97	#endif
98	#ifndef GET_ENDIAN
99	# define GET_ENDIAN(a, b) (-EINVAL)
100	#endif
101	#ifndef SET_ENDIAN
102	# define SET_ENDIAN(a, b) (-EINVAL)
103	#endif
104	#ifndef GET_TSC_CTL
105	# define GET_TSC_CTL(a) (-EINVAL)
106	#endif
107	#ifndef SET_TSC_CTL
108	# define SET_TSC_CTL(a) (-EINVAL)
109	#endif
110	#ifndef GET_FP_MODE
111	# define GET_FP_MODE(a) (-EINVAL)
112	#endif
113	#ifndef SET_FP_MODE
114	# define SET_FP_MODE(a,b) (-EINVAL)
115	#endif
116	#ifndef SVE_SET_VL
117	# define SVE_SET_VL(a) (-EINVAL)
118	#endif
119	#ifndef SVE_GET_VL
120	# define SVE_GET_VL() (-EINVAL)
121	#endif
122	#ifndef SME_SET_VL
123	# define SME_SET_VL(a) (-EINVAL)
124	#endif
125	#ifndef SME_GET_VL
126	# define SME_GET_VL() (-EINVAL)
127	#endif
128	#ifndef PAC_RESET_KEYS
129	# define PAC_RESET_KEYS(a, b) (-EINVAL)
130	#endif
131	#ifndef PAC_SET_ENABLED_KEYS
132	# define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL)
133	#endif
134	#ifndef PAC_GET_ENABLED_KEYS
135	# define PAC_GET_ENABLED_KEYS(a) (-EINVAL)
136	#endif
137	#ifndef SET_TAGGED_ADDR_CTRL
138	# define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
139	#endif
140	#ifndef GET_TAGGED_ADDR_CTRL
141	# define GET_TAGGED_ADDR_CTRL() (-EINVAL)
142	#endif
143	#ifndef RISCV_V_SET_CONTROL
144	# define RISCV_V_SET_CONTROL(a) (-EINVAL)
145	#endif
146	#ifndef RISCV_V_GET_CONTROL
147	# define RISCV_V_GET_CONTROL() (-EINVAL)
148	#endif
149
150	/*
151	* this is where the system-wide overflow UID and GID are defined, for
152	* architectures that now have 32-bit UID/GID but didn't in the past
153	*/
154
155	int overflowuid = DEFAULT_OVERFLOWUID;
156	int overflowgid = DEFAULT_OVERFLOWGID;
157
158	EXPORT_SYMBOL(overflowuid);
159	EXPORT_SYMBOL(overflowgid);
160
161	/*
162	* the same as above, but for filesystems which can only store a 16-bit
163	* UID and GID. as such, this is needed on all architectures
164	*/
165
166	int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
167	int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
168
169	EXPORT_SYMBOL(fs_overflowuid);
170	EXPORT_SYMBOL(fs_overflowgid);
171
172	/*
173	* Returns true if current's euid is same as p's uid or euid,
174	* or has CAP_SYS_NICE to p's user_ns.
175	*
176	* Called with rcu_read_lock, creds are safe
177	*/
178	static bool set_one_prio_perm(struct task_struct *p)
179	{
180	const struct cred cred = current_cred(), pcred = __task_cred(p);
181
182	if (uid_eq(left: pcred->uid, right: cred->euid) \|\|
183	uid_eq(left: pcred->euid, right: cred->euid))
184	return true;
185	if (ns_capable(ns: pcred->user_ns, CAP_SYS_NICE))
186	return true;
187	return false;
188	}
189
190	/*
191	* set the priority of a task
192	* - the caller must hold the RCU read lock
193	*/
194	static int set_one_prio(struct task_struct p, int* niceval, int error)
195	{
196	int no_nice;
197
198	if (!set_one_prio_perm(p)) {
199	error = -EPERM;
200	goto out;
201	}
202	if (niceval < task_nice(p) && !can_nice(p, nice: niceval)) {
203	error = -EACCES;
204	goto out;
205	}
206	no_nice = security_task_setnice(p, nice: niceval);
207	if (no_nice) {
208	error = no_nice;
209	goto out;
210	}
211	if (error == -ESRCH)
212	error = `0`;
213	set_user_nice(p, nice: niceval);
214	out:
215	return error;
216	}
217
218	SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
219	{
220	struct task_struct g, p;
221	struct user_struct *user;
222	const struct cred *cred = current_cred();
223	int error = -EINVAL;
224	struct pid *pgrp;
225	kuid_t uid;
226
227	if (which > PRIO_USER \|\| which < PRIO_PROCESS)
228	goto out;
229
230	/ normalize: avoid signed division (rounding problems) /
231	error = -ESRCH;
232	if (niceval < MIN_NICE)
233	niceval = MIN_NICE;
234	if (niceval > MAX_NICE)
235	niceval = MAX_NICE;
236
237	rcu_read_lock();
238	switch (which) {
239	case PRIO_PROCESS:
240	if (who)
241	p = find_task_by_vpid(nr: who);
242	else
243	p = current;
244	if (p)
245	error = set_one_prio(p, niceval, error);
246	break;
247	case PRIO_PGRP:
248	if (who)
249	pgrp = find_vpid(nr: who);
250	else
251	pgrp = task_pgrp(current);
252	read_lock(&tasklist_lock);
253	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
254	error = set_one_prio(p, niceval, error);
255	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
256	read_unlock(&tasklist_lock);
257	break;
258	case PRIO_USER:
259	uid = make_kuid(from: cred->user_ns, uid: who);
260	user = cred->user;
261	if (!who)
262	uid = cred->uid;
263	else if (!uid_eq(left: uid, right: cred->uid)) {
264	user = find_user(uid);
265	if (!user)
266	goto out_unlock; / No processes for this user /
267	}
268	for_each_process_thread(g, p) {
269	if (uid_eq(task_uid(p), right: uid) && task_pid_vnr(tsk: p))
270	error = set_one_prio(p, niceval, error);
271	}
272	if (!uid_eq(left: uid, right: cred->uid))
273	free_uid(user); / For find_user() /
274	break;
275	}
276	out_unlock:
277	rcu_read_unlock();
278	out:
279	return error;
280	}
281
282	/*
283	* Ugh. To avoid negative return values, "getpriority()" will
284	* not return the normal nice-value, but a negated value that
285	* has been offset by 20 (ie it returns 40..1 instead of -20..19)
286	* to stay compatible.
287	*/
288	SYSCALL_DEFINE2(getpriority, int, which, int, who)
289	{
290	struct task_struct g, p;
291	struct user_struct *user;
292	const struct cred *cred = current_cred();
293	long niceval, retval = -ESRCH;
294	struct pid *pgrp;
295	kuid_t uid;
296
297	if (which > PRIO_USER \|\| which < PRIO_PROCESS)
298	return -EINVAL;
299
300	rcu_read_lock();
301	switch (which) {
302	case PRIO_PROCESS:
303	if (who)
304	p = find_task_by_vpid(nr: who);
305	else
306	p = current;
307	if (p) {
308	niceval = nice_to_rlimit(nice: task_nice(p));
309	if (niceval > retval)
310	retval = niceval;
311	}
312	break;
313	case PRIO_PGRP:
314	if (who)
315	pgrp = find_vpid(nr: who);
316	else
317	pgrp = task_pgrp(current);
318	read_lock(&tasklist_lock);
319	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
320	niceval = nice_to_rlimit(nice: task_nice(p));
321	if (niceval > retval)
322	retval = niceval;
323	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
324	read_unlock(&tasklist_lock);
325	break;
326	case PRIO_USER:
327	uid = make_kuid(from: cred->user_ns, uid: who);
328	user = cred->user;
329	if (!who)
330	uid = cred->uid;
331	else if (!uid_eq(left: uid, right: cred->uid)) {
332	user = find_user(uid);
333	if (!user)
334	goto out_unlock; / No processes for this user /
335	}
336	for_each_process_thread(g, p) {
337	if (uid_eq(task_uid(p), right: uid) && task_pid_vnr(tsk: p)) {
338	niceval = nice_to_rlimit(nice: task_nice(p));
339	if (niceval > retval)
340	retval = niceval;
341	}
342	}
343	if (!uid_eq(left: uid, right: cred->uid))
344	free_uid(user); / for find_user() /
345	break;
346	}
347	out_unlock:
348	rcu_read_unlock();
349
350	return retval;
351	}
352
353	/*
354	* Unprivileged users may change the real gid to the effective gid
355	* or vice versa. (BSD-style)
356	*
357	* If you set the real gid at all, or set the effective gid to a value not
358	* equal to the real gid, then the saved gid is set to the new effective gid.
359	*
360	* This makes it possible for a setgid program to completely drop its
361	* privileges, which is often a useful assertion to make when you are doing
362	* a security audit over a program.
363	*
364	* The general idea is that a program which uses just setregid() will be
365	* 100% compatible with BSD. A program which uses just setgid() will be
366	* 100% compatible with POSIX with saved IDs.
367	*
368	* SMP: There are not races, the GIDs are checked only by filesystem
369	* operations (as far as semantic preservation is concerned).
370	*/
371	#ifdef CONFIG_MULTIUSER
372	long __sys_setregid(gid_t rgid, gid_t egid)
373	{
374	struct user_namespace *ns = current_user_ns();
375	const struct cred *old;
376	struct cred *new;
377	int retval;
378	kgid_t krgid, kegid;
379
380	krgid = make_kgid(from: ns, gid: rgid);
381	kegid = make_kgid(from: ns, gid: egid);
382
383	if ((rgid != (gid_t) -`1`) && !gid_valid(gid: krgid))
384	return -EINVAL;
385	if ((egid != (gid_t) -`1`) && !gid_valid(gid: kegid))
386	return -EINVAL;
387
388	new = prepare_creds();
389	if (!new)
390	return -ENOMEM;
391	old = current_cred();
392
393	retval = -EPERM;
394	if (rgid != (gid_t) -`1`) {
395	if (gid_eq(left: old->gid, right: krgid) \|\|
396	gid_eq(left: old->egid, right: krgid) \|\|
397	ns_capable_setid(ns: old->user_ns, CAP_SETGID))
398	new->gid = krgid;
399	else
400	goto error;
401	}
402	if (egid != (gid_t) -`1`) {
403	if (gid_eq(left: old->gid, right: kegid) \|\|
404	gid_eq(left: old->egid, right: kegid) \|\|
405	gid_eq(left: old->sgid, right: kegid) \|\|
406	ns_capable_setid(ns: old->user_ns, CAP_SETGID))
407	new->egid = kegid;
408	else
409	goto error;
410	}
411
412	if (rgid != (gid_t) -`1` \|\|
413	(egid != (gid_t) -`1` && !gid_eq(left: kegid, right: old->gid)))
414	new->sgid = new->egid;
415	new->fsgid = new->egid;
416
417	retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
418	if (retval < `0`)
419	goto error;
420
421	return commit_creds(new);
422
423	error:
424	abort_creds(new);
425	return retval;
426	}
427
428	SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
429	{
430	return __sys_setregid(rgid, egid);
431	}
432
433	/*
434	* setgid() is implemented like SysV w/ SAVED_IDS
435	*
436	* SMP: Same implicit races as above.
437	*/
438	long __sys_setgid(gid_t gid)
439	{
440	struct user_namespace *ns = current_user_ns();
441	const struct cred *old;
442	struct cred *new;
443	int retval;
444	kgid_t kgid;
445
446	kgid = make_kgid(from: ns, gid);
447	if (!gid_valid(gid: kgid))
448	return -EINVAL;
449
450	new = prepare_creds();
451	if (!new)
452	return -ENOMEM;
453	old = current_cred();
454
455	retval = -EPERM;
456	if (ns_capable_setid(ns: old->user_ns, CAP_SETGID))
457	new->gid = new->egid = new->sgid = new->fsgid = kgid;
458	else if (gid_eq(left: kgid, right: old->gid) \|\| gid_eq(left: kgid, right: old->sgid))
459	new->egid = new->fsgid = kgid;
460	else
461	goto error;
462
463	retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
464	if (retval < `0`)
465	goto error;
466
467	return commit_creds(new);
468
469	error:
470	abort_creds(new);
471	return retval;
472	}
473
474	SYSCALL_DEFINE1(setgid, gid_t, gid)
475	{
476	return __sys_setgid(gid);
477	}
478
479	/*
480	* change the user struct in a credentials set to match the new UID
481	*/
482	static int set_user(struct cred *new)
483	{
484	struct user_struct *new_user;
485
486	new_user = alloc_uid(new->uid);
487	if (!new_user)
488	return -EAGAIN;
489
490	free_uid(new->user);
491	new->user = new_user;
492	return `0`;
493	}
494
495	static void flag_nproc_exceeded(struct cred *new)
496	{
497	if (new->ucounts == current_ucounts())
498	return;
499
500	/*
501	* We don't fail in case of NPROC limit excess here because too many
502	* poorly written programs don't check set*uid() return code, assuming
503	* it never fails if called by root. We may still enforce NPROC limit
504	* for programs doing set*uid()+execve() by harmlessly deferring the
505	* failure to the execve() stage.
506	*/
507	if (is_rlimit_overlimit(ucounts: new->ucounts, type: UCOUNT_RLIMIT_NPROC, max: rlimit(RLIMIT_NPROC)) &&
508	new->user != INIT_USER)
509	current->flags \|= PF_NPROC_EXCEEDED;
510	else
511	current->flags &= ~PF_NPROC_EXCEEDED;
512	}
513
514	/*
515	* Unprivileged users may change the real uid to the effective uid
516	* or vice versa. (BSD-style)
517	*
518	* If you set the real uid at all, or set the effective uid to a value not
519	* equal to the real uid, then the saved uid is set to the new effective uid.
520	*
521	* This makes it possible for a setuid program to completely drop its
522	* privileges, which is often a useful assertion to make when you are doing
523	* a security audit over a program.
524	*
525	* The general idea is that a program which uses just setreuid() will be
526	* 100% compatible with BSD. A program which uses just setuid() will be
527	* 100% compatible with POSIX with saved IDs.
528	*/
529	long __sys_setreuid(uid_t ruid, uid_t euid)
530	{
531	struct user_namespace *ns = current_user_ns();
532	const struct cred *old;
533	struct cred *new;
534	int retval;
535	kuid_t kruid, keuid;
536
537	kruid = make_kuid(from: ns, uid: ruid);
538	keuid = make_kuid(from: ns, uid: euid);
539
540	if ((ruid != (uid_t) -`1`) && !uid_valid(uid: kruid))
541	return -EINVAL;
542	if ((euid != (uid_t) -`1`) && !uid_valid(uid: keuid))
543	return -EINVAL;
544
545	new = prepare_creds();
546	if (!new)
547	return -ENOMEM;
548	old = current_cred();
549
550	retval = -EPERM;
551	if (ruid != (uid_t) -`1`) {
552	new->uid = kruid;
553	if (!uid_eq(left: old->uid, right: kruid) &&
554	!uid_eq(left: old->euid, right: kruid) &&
555	!ns_capable_setid(ns: old->user_ns, CAP_SETUID))
556	goto error;
557	}
558
559	if (euid != (uid_t) -`1`) {
560	new->euid = keuid;
561	if (!uid_eq(left: old->uid, right: keuid) &&
562	!uid_eq(left: old->euid, right: keuid) &&
563	!uid_eq(left: old->suid, right: keuid) &&
564	!ns_capable_setid(ns: old->user_ns, CAP_SETUID))
565	goto error;
566	}
567
568	if (!uid_eq(left: new->uid, right: old->uid)) {
569	retval = set_user(new);
570	if (retval < `0`)
571	goto error;
572	}
573	if (ruid != (uid_t) -`1` \|\|
574	(euid != (uid_t) -`1` && !uid_eq(left: keuid, right: old->uid)))
575	new->suid = new->euid;
576	new->fsuid = new->euid;
577
578	retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
579	if (retval < `0`)
580	goto error;
581
582	retval = set_cred_ucounts(new);
583	if (retval < `0`)
584	goto error;
585
586	flag_nproc_exceeded(new);
587	return commit_creds(new);
588
589	error:
590	abort_creds(new);
591	return retval;
592	}
593
594	SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
595	{
596	return __sys_setreuid(ruid, euid);
597	}
598
599	/*
600	* setuid() is implemented like SysV with SAVED_IDS
601	*
602	* Note that SAVED_ID's is deficient in that a setuid root program
603	* like sendmail, for example, cannot set its uid to be a normal
604	* user and then switch back, because if you're root, setuid() sets
605	* the saved uid too. If you don't like this, blame the bright people
606	* in the POSIX committee and/or USG. Note that the BSD-style setreuid()
607	* will allow a root program to temporarily drop privileges and be able to
608	* regain them by swapping the real and effective uid.
609	*/
610	long __sys_setuid(uid_t uid)
611	{
612	struct user_namespace *ns = current_user_ns();
613	const struct cred *old;
614	struct cred *new;
615	int retval;
616	kuid_t kuid;
617
618	kuid = make_kuid(from: ns, uid);
619	if (!uid_valid(uid: kuid))
620	return -EINVAL;
621
622	new = prepare_creds();
623	if (!new)
624	return -ENOMEM;
625	old = current_cred();
626
627	retval = -EPERM;
628	if (ns_capable_setid(ns: old->user_ns, CAP_SETUID)) {
629	new->suid = new->uid = kuid;
630	if (!uid_eq(left: kuid, right: old->uid)) {
631	retval = set_user(new);
632	if (retval < `0`)
633	goto error;
634	}
635	} else if (!uid_eq(left: kuid, right: old->uid) && !uid_eq(left: kuid, right: new->suid)) {
636	goto error;
637	}
638
639	new->fsuid = new->euid = kuid;
640
641	retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
642	if (retval < `0`)
643	goto error;
644
645	retval = set_cred_ucounts(new);
646	if (retval < `0`)
647	goto error;
648
649	flag_nproc_exceeded(new);
650	return commit_creds(new);
651
652	error:
653	abort_creds(new);
654	return retval;
655	}
656
657	SYSCALL_DEFINE1(setuid, uid_t, uid)
658	{
659	return __sys_setuid(uid);
660	}
661
662
663	/*
664	* This function implements a generic ability to update ruid, euid,
665	* and suid. This allows you to implement the 4.4 compatible seteuid().
666	*/
667	long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
668	{
669	struct user_namespace *ns = current_user_ns();
670	const struct cred *old;
671	struct cred *new;
672	int retval;
673	kuid_t kruid, keuid, ksuid;
674	bool ruid_new, euid_new, suid_new;
675
676	kruid = make_kuid(from: ns, uid: ruid);
677	keuid = make_kuid(from: ns, uid: euid);
678	ksuid = make_kuid(from: ns, uid: suid);
679
680	if ((ruid != (uid_t) -`1`) && !uid_valid(uid: kruid))
681	return -EINVAL;
682
683	if ((euid != (uid_t) -`1`) && !uid_valid(uid: keuid))
684	return -EINVAL;
685
686	if ((suid != (uid_t) -`1`) && !uid_valid(uid: ksuid))
687	return -EINVAL;
688
689	old = current_cred();
690
691	/ check for no-op /
692	if ((ruid == (uid_t) -`1` \|\| uid_eq(left: kruid, right: old->uid)) &&
693	(euid == (uid_t) -`1` \|\| (uid_eq(left: keuid, right: old->euid) &&
694	uid_eq(left: keuid, right: old->fsuid))) &&
695	(suid == (uid_t) -`1` \|\| uid_eq(left: ksuid, right: old->suid)))
696	return `0`;
697
698	ruid_new = ruid != (uid_t) -`1` && !uid_eq(left: kruid, right: old->uid) &&
699	!uid_eq(left: kruid, right: old->euid) && !uid_eq(left: kruid, right: old->suid);
700	euid_new = euid != (uid_t) -`1` && !uid_eq(left: keuid, right: old->uid) &&
701	!uid_eq(left: keuid, right: old->euid) && !uid_eq(left: keuid, right: old->suid);
702	suid_new = suid != (uid_t) -`1` && !uid_eq(left: ksuid, right: old->uid) &&
703	!uid_eq(left: ksuid, right: old->euid) && !uid_eq(left: ksuid, right: old->suid);
704	if ((ruid_new \|\| euid_new \|\| suid_new) &&
705	!ns_capable_setid(ns: old->user_ns, CAP_SETUID))
706	return -EPERM;
707
708	new = prepare_creds();
709	if (!new)
710	return -ENOMEM;
711
712	if (ruid != (uid_t) -`1`) {
713	new->uid = kruid;
714	if (!uid_eq(left: kruid, right: old->uid)) {
715	retval = set_user(new);
716	if (retval < `0`)
717	goto error;
718	}
719	}
720	if (euid != (uid_t) -`1`)
721	new->euid = keuid;
722	if (suid != (uid_t) -`1`)
723	new->suid = ksuid;
724	new->fsuid = new->euid;
725
726	retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
727	if (retval < `0`)
728	goto error;
729
730	retval = set_cred_ucounts(new);
731	if (retval < `0`)
732	goto error;
733
734	flag_nproc_exceeded(new);
735	return commit_creds(new);
736
737	error:
738	abort_creds(new);
739	return retval;
740	}
741
742	SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
743	{
744	return __sys_setresuid(ruid, euid, suid);
745	}
746
747	SYSCALL_DEFINE3(getresuid, uid_t __user , ruidp, uid_t __user , euidp, uid_t __user *, suidp)
748	{
749	const struct cred *cred = current_cred();
750	int retval;
751	uid_t ruid, euid, suid;
752
753	ruid = from_kuid_munged(to: cred->user_ns, uid: cred->uid);
754	euid = from_kuid_munged(to: cred->user_ns, uid: cred->euid);
755	suid = from_kuid_munged(to: cred->user_ns, uid: cred->suid);
756
757	retval = put_user(ruid, ruidp);
758	if (!retval) {
759	retval = put_user(euid, euidp);
760	if (!retval)
761	return put_user(suid, suidp);
762	}
763	return retval;
764	}
765
766	/*
767	* Same as above, but for rgid, egid, sgid.
768	*/
769	long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
770	{
771	struct user_namespace *ns = current_user_ns();
772	const struct cred *old;
773	struct cred *new;
774	int retval;
775	kgid_t krgid, kegid, ksgid;
776	bool rgid_new, egid_new, sgid_new;
777
778	krgid = make_kgid(from: ns, gid: rgid);
779	kegid = make_kgid(from: ns, gid: egid);
780	ksgid = make_kgid(from: ns, gid: sgid);
781
782	if ((rgid != (gid_t) -`1`) && !gid_valid(gid: krgid))
783	return -EINVAL;
784	if ((egid != (gid_t) -`1`) && !gid_valid(gid: kegid))
785	return -EINVAL;
786	if ((sgid != (gid_t) -`1`) && !gid_valid(gid: ksgid))
787	return -EINVAL;
788
789	old = current_cred();
790
791	/ check for no-op /
792	if ((rgid == (gid_t) -`1` \|\| gid_eq(left: krgid, right: old->gid)) &&
793	(egid == (gid_t) -`1` \|\| (gid_eq(left: kegid, right: old->egid) &&
794	gid_eq(left: kegid, right: old->fsgid))) &&
795	(sgid == (gid_t) -`1` \|\| gid_eq(left: ksgid, right: old->sgid)))
796	return `0`;
797
798	rgid_new = rgid != (gid_t) -`1` && !gid_eq(left: krgid, right: old->gid) &&
799	!gid_eq(left: krgid, right: old->egid) && !gid_eq(left: krgid, right: old->sgid);
800	egid_new = egid != (gid_t) -`1` && !gid_eq(left: kegid, right: old->gid) &&
801	!gid_eq(left: kegid, right: old->egid) && !gid_eq(left: kegid, right: old->sgid);
802	sgid_new = sgid != (gid_t) -`1` && !gid_eq(left: ksgid, right: old->gid) &&
803	!gid_eq(left: ksgid, right: old->egid) && !gid_eq(left: ksgid, right: old->sgid);
804	if ((rgid_new \|\| egid_new \|\| sgid_new) &&
805	!ns_capable_setid(ns: old->user_ns, CAP_SETGID))
806	return -EPERM;
807
808	new = prepare_creds();
809	if (!new)
810	return -ENOMEM;
811
812	if (rgid != (gid_t) -`1`)
813	new->gid = krgid;
814	if (egid != (gid_t) -`1`)
815	new->egid = kegid;
816	if (sgid != (gid_t) -`1`)
817	new->sgid = ksgid;
818	new->fsgid = new->egid;
819
820	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
821	if (retval < `0`)
822	goto error;
823
824	return commit_creds(new);
825
826	error:
827	abort_creds(new);
828	return retval;
829	}
830
831	SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
832	{
833	return __sys_setresgid(rgid, egid, sgid);
834	}
835
836	SYSCALL_DEFINE3(getresgid, gid_t __user , rgidp, gid_t __user , egidp, gid_t __user *, sgidp)
837	{
838	const struct cred *cred = current_cred();
839	int retval;
840	gid_t rgid, egid, sgid;
841
842	rgid = from_kgid_munged(to: cred->user_ns, gid: cred->gid);
843	egid = from_kgid_munged(to: cred->user_ns, gid: cred->egid);
844	sgid = from_kgid_munged(to: cred->user_ns, gid: cred->sgid);
845
846	retval = put_user(rgid, rgidp);
847	if (!retval) {
848	retval = put_user(egid, egidp);
849	if (!retval)
850	retval = put_user(sgid, sgidp);
851	}
852
853	return retval;
854	}
855
856
857	/*
858	* "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
859	* is used for "access()" and for the NFS daemon (letting nfsd stay at
860	* whatever uid it wants to). It normally shadows "euid", except when
861	* explicitly set by setfsuid() or for access..
862	*/
863	long __sys_setfsuid(uid_t uid)
864	{
865	const struct cred *old;
866	struct cred *new;
867	uid_t old_fsuid;
868	kuid_t kuid;
869
870	old = current_cred();
871	old_fsuid = from_kuid_munged(to: old->user_ns, uid: old->fsuid);
872
873	kuid = make_kuid(from: old->user_ns, uid);
874	if (!uid_valid(uid: kuid))
875	return old_fsuid;
876
877	new = prepare_creds();
878	if (!new)
879	return old_fsuid;
880
881	if (uid_eq(left: kuid, right: old->uid) \|\| uid_eq(left: kuid, right: old->euid) \|\|
882	uid_eq(left: kuid, right: old->suid) \|\| uid_eq(left: kuid, right: old->fsuid) \|\|
883	ns_capable_setid(ns: old->user_ns, CAP_SETUID)) {
884	if (!uid_eq(left: kuid, right: old->fsuid)) {
885	new->fsuid = kuid;
886	if (security_task_fix_setuid(new, old, LSM_SETID_FS) == `0`)
887	goto change_okay;
888	}
889	}
890
891	abort_creds(new);
892	return old_fsuid;
893
894	change_okay:
895	commit_creds(new);
896	return old_fsuid;
897	}
898
899	SYSCALL_DEFINE1(setfsuid, uid_t, uid)
900	{
901	return __sys_setfsuid(uid);
902	}
903
904	/*
905	* Samma på svenska..
906	*/
907	long __sys_setfsgid(gid_t gid)
908	{
909	const struct cred *old;
910	struct cred *new;
911	gid_t old_fsgid;
912	kgid_t kgid;
913
914	old = current_cred();
915	old_fsgid = from_kgid_munged(to: old->user_ns, gid: old->fsgid);
916
917	kgid = make_kgid(from: old->user_ns, gid);
918	if (!gid_valid(gid: kgid))
919	return old_fsgid;
920
921	new = prepare_creds();
922	if (!new)
923	return old_fsgid;
924
925	if (gid_eq(left: kgid, right: old->gid) \|\| gid_eq(left: kgid, right: old->egid) \|\|
926	gid_eq(left: kgid, right: old->sgid) \|\| gid_eq(left: kgid, right: old->fsgid) \|\|
927	ns_capable_setid(ns: old->user_ns, CAP_SETGID)) {
928	if (!gid_eq(left: kgid, right: old->fsgid)) {
929	new->fsgid = kgid;
930	if (security_task_fix_setgid(new,old,LSM_SETID_FS) == `0`)
931	goto change_okay;
932	}
933	}
934
935	abort_creds(new);
936	return old_fsgid;
937
938	change_okay:
939	commit_creds(new);
940	return old_fsgid;
941	}
942
943	SYSCALL_DEFINE1(setfsgid, gid_t, gid)
944	{
945	return __sys_setfsgid(gid);
946	}
947	#endif /* CONFIG_MULTIUSER */
948
949	/**
950	* sys_getpid - return the thread group id of the current process
951	*
952	* Note, despite the name, this returns the tgid not the pid. The tgid and
953	* the pid are identical unless CLONE_THREAD was specified on clone() in
954	* which case the tgid is the same in all threads of the same group.
955	*
956	* This is SMP safe as current->tgid does not change.
957	*/
958	SYSCALL_DEFINE0(getpid)
959	{
960	return task_tgid_vnr(current);
961	}
962
963	/ Thread ID - the internal kernel "pid" /
964	SYSCALL_DEFINE0(gettid)
965	{
966	return task_pid_vnr(current);
967	}
968
969	/*
970	* Accessing ->real_parent is not SMP-safe, it could
971	* change from under us. However, we can use a stale
972	* value of ->real_parent under rcu_read_lock(), see
973	* release_task()->call_rcu(delayed_put_task_struct).
974	*/
975	SYSCALL_DEFINE0(getppid)
976	{
977	int pid;
978
979	rcu_read_lock();
980	pid = task_tgid_vnr(rcu_dereference(current->real_parent));
981	rcu_read_unlock();
982
983	return pid;
984	}
985
986	SYSCALL_DEFINE0(getuid)
987	{
988	/ Only we change this so SMP safe /
989	return from_kuid_munged(current_user_ns(), current_uid());
990	}
991
992	SYSCALL_DEFINE0(geteuid)
993	{
994	/ Only we change this so SMP safe /
995	return from_kuid_munged(current_user_ns(), current_euid());
996	}
997
998	SYSCALL_DEFINE0(getgid)
999	{
1000	/ Only we change this so SMP safe /
1001	return from_kgid_munged(current_user_ns(), current_gid());
1002	}
1003
1004	SYSCALL_DEFINE0(getegid)
1005	{
1006	/ Only we change this so SMP safe /
1007	return from_kgid_munged(current_user_ns(), current_egid());
1008	}
1009
1010	static void do_sys_times(struct tms *tms)
1011	{
1012	u64 tgutime, tgstime, cutime, cstime;
1013
1014	thread_group_cputime_adjusted(current, ut: &tgutime, st: &tgstime);
1015	cutime = current->signal->cutime;
1016	cstime = current->signal->cstime;
1017	tms->tms_utime = nsec_to_clock_t(x: tgutime);
1018	tms->tms_stime = nsec_to_clock_t(x: tgstime);
1019	tms->tms_cutime = nsec_to_clock_t(x: cutime);
1020	tms->tms_cstime = nsec_to_clock_t(x: cstime);
1021	}
1022
1023	SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1024	{
1025	if (tbuf) {
1026	struct tms tmp;
1027
1028	do_sys_times(tms: &tmp);
1029	if (copy_to_user(to: tbuf, from: &tmp, n: sizeof(struct tms)))
1030	return -EFAULT;
1031	}
1032	force_successful_syscall_return();
1033	return (long) jiffies_64_to_clock_t(x: get_jiffies_64());
1034	}
1035
1036	#ifdef CONFIG_COMPAT
1037	static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1038	{
1039	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1040	}
1041
1042	COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1043	{
1044	if (tbuf) {
1045	struct tms tms;
1046	struct compat_tms tmp;
1047
1048	do_sys_times(tms: &tms);
1049	/ Convert our struct tms to the compat version. /
1050	tmp.tms_utime = clock_t_to_compat_clock_t(x: tms.tms_utime);
1051	tmp.tms_stime = clock_t_to_compat_clock_t(x: tms.tms_stime);
1052	tmp.tms_cutime = clock_t_to_compat_clock_t(x: tms.tms_cutime);
1053	tmp.tms_cstime = clock_t_to_compat_clock_t(x: tms.tms_cstime);
1054	if (copy_to_user(to: tbuf, from: &tmp, n: sizeof(tmp)))
1055	return -EFAULT;
1056	}
1057	force_successful_syscall_return();
1058	return compat_jiffies_to_clock_t(jiffies);
1059	}
1060	#endif
1061
1062	/*
1063	* This needs some heavy checking ...
1064	* I just haven't the stomach for it. I also don't fully
1065	* understand sessions/pgrp etc. Let somebody who does explain it.
1066	*
1067	* OK, I think I have the protection semantics right.... this is really
1068	* only important on a multi-user system anyway, to make sure one user
1069	* can't send a signal to a process owned by another. -TYT, 12/12/91
1070	*
1071	* !PF_FORKNOEXEC check to conform completely to POSIX.
1072	*/
1073	SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1074	{
1075	struct task_struct *p;
1076	struct task_struct *group_leader = current->group_leader;
1077	struct pid *pgrp;
1078	int err;
1079
1080	if (!pid)
1081	pid = task_pid_vnr(tsk: group_leader);
1082	if (!pgid)
1083	pgid = pid;
1084	if (pgid < `0`)
1085	return -EINVAL;
1086	rcu_read_lock();
1087
1088	/ From this point forward we keep holding onto the tasklist lock*
1089	* so that our parent does not change from under us. -DaveM
1090	*/
1091	write_lock_irq(&tasklist_lock);
1092
1093	err = -ESRCH;
1094	p = find_task_by_vpid(nr: pid);
1095	if (!p)
1096	goto out;
1097
1098	err = -EINVAL;
1099	if (!thread_group_leader(p))
1100	goto out;
1101
1102	if (same_thread_group(p1: p->real_parent, p2: group_leader)) {
1103	err = -EPERM;
1104	if (task_session(task: p) != task_session(task: group_leader))
1105	goto out;
1106	err = -EACCES;
1107	if (!(p->flags & PF_FORKNOEXEC))
1108	goto out;
1109	} else {
1110	err = -ESRCH;
1111	if (p != group_leader)
1112	goto out;
1113	}
1114
1115	err = -EPERM;
1116	if (p->signal->leader)
1117	goto out;
1118
1119	pgrp = task_pid(task: p);
1120	if (pgid != pid) {
1121	struct task_struct *g;
1122
1123	pgrp = find_vpid(nr: pgid);
1124	g = pid_task(pid: pgrp, PIDTYPE_PGID);
1125	if (!g \|\| task_session(task: g) != task_session(task: group_leader))
1126	goto out;
1127	}
1128
1129	err = security_task_setpgid(p, pgid);
1130	if (err)
1131	goto out;
1132
1133	if (task_pgrp(task: p) != pgrp)
1134	change_pid(task: p, PIDTYPE_PGID, pid: pgrp);
1135
1136	err = `0`;
1137	out:
1138	/ All paths lead to here, thus we are safe. -DaveM /
1139	write_unlock_irq(&tasklist_lock);
1140	rcu_read_unlock();
1141	return err;
1142	}
1143
1144	static int do_getpgid(pid_t pid)
1145	{
1146	struct task_struct *p;
1147	struct pid *grp;
1148	int retval;
1149
1150	rcu_read_lock();
1151	if (!pid)
1152	grp = task_pgrp(current);
1153	else {
1154	retval = -ESRCH;
1155	p = find_task_by_vpid(nr: pid);
1156	if (!p)
1157	goto out;
1158	grp = task_pgrp(task: p);
1159	if (!grp)
1160	goto out;
1161
1162	retval = security_task_getpgid(p);
1163	if (retval)
1164	goto out;
1165	}
1166	retval = pid_vnr(pid: grp);
1167	out:
1168	rcu_read_unlock();
1169	return retval;
1170	}
1171
1172	SYSCALL_DEFINE1(getpgid, pid_t, pid)
1173	{
1174	return do_getpgid(pid);
1175	}
1176
1177	#ifdef __ARCH_WANT_SYS_GETPGRP
1178
1179	SYSCALL_DEFINE0(getpgrp)
1180	{
1181	return do_getpgid(pid: `0`);
1182	}
1183
1184	#endif
1185
1186	SYSCALL_DEFINE1(getsid, pid_t, pid)
1187	{
1188	struct task_struct *p;
1189	struct pid *sid;
1190	int retval;
1191
1192	rcu_read_lock();
1193	if (!pid)
1194	sid = task_session(current);
1195	else {
1196	retval = -ESRCH;
1197	p = find_task_by_vpid(nr: pid);
1198	if (!p)
1199	goto out;
1200	sid = task_session(task: p);
1201	if (!sid)
1202	goto out;
1203
1204	retval = security_task_getsid(p);
1205	if (retval)
1206	goto out;
1207	}
1208	retval = pid_vnr(pid: sid);
1209	out:
1210	rcu_read_unlock();
1211	return retval;
1212	}
1213
1214	static void set_special_pids(struct pid *pid)
1215	{
1216	struct task_struct *curr = current->group_leader;
1217
1218	if (task_session(task: curr) != pid)
1219	change_pid(task: curr, PIDTYPE_SID, pid);
1220
1221	if (task_pgrp(task: curr) != pid)
1222	change_pid(task: curr, PIDTYPE_PGID, pid);
1223	}
1224
1225	int ksys_setsid(void)
1226	{
1227	struct task_struct *group_leader = current->group_leader;
1228	struct pid *sid = task_pid(task: group_leader);
1229	pid_t session = pid_vnr(pid: sid);
1230	int err = -EPERM;
1231
1232	write_lock_irq(&tasklist_lock);
1233	/ Fail if I am already a session leader /
1234	if (group_leader->signal->leader)
1235	goto out;
1236
1237	/ Fail if a process group id already exists that equals the*
1238	* proposed session id.
1239	*/
1240	if (pid_task(pid: sid, PIDTYPE_PGID))
1241	goto out;
1242
1243	group_leader->signal->leader = `1`;
1244	set_special_pids(sid);
1245
1246	proc_clear_tty(p: group_leader);
1247
1248	err = session;
1249	out:
1250	write_unlock_irq(&tasklist_lock);
1251	if (err > `0`) {
1252	proc_sid_connector(task: group_leader);
1253	sched_autogroup_create_attach(p: group_leader);
1254	}
1255	return err;
1256	}
1257
1258	SYSCALL_DEFINE0(setsid)
1259	{
1260	return ksys_setsid();
1261	}
1262
1263	DECLARE_RWSEM(uts_sem);
1264
1265	#ifdef COMPAT_UTS_MACHINE
1266	#define override_architecture(name) \
1267	(personality(current->personality) == PER_LINUX32 && \
1268	copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1269	sizeof(COMPAT_UTS_MACHINE)))
1270	#else
1271	#define override_architecture(name) 0
1272	#endif
1273
1274	/*
1275	* Work around broken programs that cannot handle "Linux 3.0".
1276	* Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1277	* And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1278	* 2.6.60.
1279	*/
1280	static int override_release(char __user *release, size_t len)
1281	{
1282	int ret = `0`;
1283
1284	if (current->personality & UNAME26) {
1285	const char *rest = UTS_RELEASE;
1286	char buf[`65`] = { `0` };
1287	int ndots = `0`;
1288	unsigned v;
1289	size_t copy;
1290
1291	while (*rest) {
1292	if (*rest == `'.'` && ++ndots >= `3`)
1293	break;
1294	if (!isdigit(c: rest) && rest != `'.'`)
1295	break;
1296	rest++;
1297	}
1298	v = LINUX_VERSION_PATCHLEVEL + `60`;
1299	copy = clamp_t(size_t, len, `1`, sizeof(buf));
1300	copy = scnprintf(buf, size: copy, fmt: "2.6.%u%s", v, rest);
1301	ret = copy_to_user(to: release, from: buf, n: copy + `1`);
1302	}
1303	return ret;
1304	}
1305
1306	SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1307	{
1308	struct new_utsname tmp;
1309
1310	down_read(sem: &uts_sem);
1311	memcpy(&tmp, utsname(), sizeof(tmp));
1312	up_read(sem: &uts_sem);
1313	if (copy_to_user(to: name, from: &tmp, n: sizeof(tmp)))
1314	return -EFAULT;
1315
1316	if (override_release(release: name->release, len: sizeof(name->release)))
1317	return -EFAULT;
1318	if (override_architecture(name))
1319	return -EFAULT;
1320	return `0`;
1321	}
1322
1323	#ifdef __ARCH_WANT_SYS_OLD_UNAME
1324	/*
1325	* Old cruft
1326	*/
1327	SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1328	{
1329	struct old_utsname tmp;
1330
1331	if (!name)
1332	return -EFAULT;
1333
1334	down_read(sem: &uts_sem);
1335	memcpy(&tmp, utsname(), sizeof(tmp));
1336	up_read(sem: &uts_sem);
1337	if (copy_to_user(to: name, from: &tmp, n: sizeof(tmp)))
1338	return -EFAULT;
1339
1340	if (override_release(release: name->release, len: sizeof(name->release)))
1341	return -EFAULT;
1342	if (override_architecture(name))
1343	return -EFAULT;
1344	return `0`;
1345	}
1346
1347	SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1348	{
1349	struct oldold_utsname tmp;
1350
1351	if (!name)
1352	return -EFAULT;
1353
1354	memset(&tmp, `0`, sizeof(tmp));
1355
1356	down_read(sem: &uts_sem);
1357	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1358	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1359	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1360	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1361	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1362	up_read(sem: &uts_sem);
1363	if (copy_to_user(to: name, from: &tmp, n: sizeof(tmp)))
1364	return -EFAULT;
1365
1366	if (override_architecture(name))
1367	return -EFAULT;
1368	if (override_release(release: name->release, len: sizeof(name->release)))
1369	return -EFAULT;
1370	return `0`;
1371	}
1372	#endif
1373
1374	SYSCALL_DEFINE2(sethostname, char __user , name, int*, len)
1375	{
1376	int errno;
1377	char tmp[__NEW_UTS_LEN];
1378
1379	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1380	return -EPERM;
1381
1382	if (len < `0` \|\| len > __NEW_UTS_LEN)
1383	return -EINVAL;
1384	errno = -EFAULT;
1385	if (!copy_from_user(to: tmp, from: name, n: len)) {
1386	struct new_utsname *u;
1387
1388	add_device_randomness(buf: tmp, len);
1389	down_write(sem: &uts_sem);
1390	u = utsname();
1391	memcpy(u->nodename, tmp, len);
1392	memset(u->nodename + len, `0`, sizeof(u->nodename) - len);
1393	errno = `0`;
1394	uts_proc_notify(proc: UTS_PROC_HOSTNAME);
1395	up_write(sem: &uts_sem);
1396	}
1397	return errno;
1398	}
1399
1400	#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1401
1402	SYSCALL_DEFINE2(gethostname, char __user , name, int*, len)
1403	{
1404	int i;
1405	struct new_utsname *u;
1406	char tmp[__NEW_UTS_LEN + `1`];
1407
1408	if (len < `0`)
1409	return -EINVAL;
1410	down_read(sem: &uts_sem);
1411	u = utsname();
1412	i = `1` + strlen(u->nodename);
1413	if (i > len)
1414	i = len;
1415	memcpy(tmp, u->nodename, i);
1416	up_read(sem: &uts_sem);
1417	if (copy_to_user(to: name, from: tmp, n: i))
1418	return -EFAULT;
1419	return `0`;
1420	}
1421
1422	#endif
1423
1424	/*
1425	* Only setdomainname; getdomainname can be implemented by calling
1426	* uname()
1427	*/
1428	SYSCALL_DEFINE2(setdomainname, char __user , name, int*, len)
1429	{
1430	int errno;
1431	char tmp[__NEW_UTS_LEN];
1432
1433	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1434	return -EPERM;
1435	if (len < `0` \|\| len > __NEW_UTS_LEN)
1436	return -EINVAL;
1437
1438	errno = -EFAULT;
1439	if (!copy_from_user(to: tmp, from: name, n: len)) {
1440	struct new_utsname *u;
1441
1442	add_device_randomness(buf: tmp, len);
1443	down_write(sem: &uts_sem);
1444	u = utsname();
1445	memcpy(u->domainname, tmp, len);
1446	memset(u->domainname + len, `0`, sizeof(u->domainname) - len);
1447	errno = `0`;
1448	uts_proc_notify(proc: UTS_PROC_DOMAINNAME);
1449	up_write(sem: &uts_sem);
1450	}
1451	return errno;
1452	}
1453
1454	/ make sure you are allowed to change @tsk limits before calling this /
1455	static int do_prlimit(struct task_struct tsk, unsigned* int resource,
1456	struct rlimit new_rlim, struct* rlimit *old_rlim)
1457	{
1458	struct rlimit *rlim;
1459	int retval = `0`;
1460
1461	if (resource >= RLIM_NLIMITS)
1462	return -EINVAL;
1463	resource = array_index_nospec(resource, RLIM_NLIMITS);
1464
1465	if (new_rlim) {
1466	if (new_rlim->rlim_cur > new_rlim->rlim_max)
1467	return -EINVAL;
1468	if (resource == RLIMIT_NOFILE &&
1469	new_rlim->rlim_max > sysctl_nr_open)
1470	return -EPERM;
1471	}
1472
1473	/ Holding a refcount on tsk protects tsk->signal from disappearing. /
1474	rlim = tsk->signal->rlim + resource;
1475	task_lock(p: tsk->group_leader);
1476	if (new_rlim) {
1477	/*
1478	* Keep the capable check against init_user_ns until cgroups can
1479	* contain all limits.
1480	*/
1481	if (new_rlim->rlim_max > rlim->rlim_max &&
1482	!capable(CAP_SYS_RESOURCE))
1483	retval = -EPERM;
1484	if (!retval)
1485	retval = security_task_setrlimit(p: tsk, resource, new_rlim);
1486	}
1487	if (!retval) {
1488	if (old_rlim)
1489	old_rlim = rlim;
1490	if (new_rlim)
1491	rlim = new_rlim;
1492	}
1493	task_unlock(p: tsk->group_leader);
1494
1495	/*
1496	* RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1497	* infinite. In case of RLIM_INFINITY the posix CPU timer code
1498	* ignores the rlimit.
1499	*/
1500	if (!retval && new_rlim && resource == RLIMIT_CPU &&
1501	new_rlim->rlim_cur != RLIM_INFINITY &&
1502	IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1503	/*
1504	* update_rlimit_cpu can fail if the task is exiting, but there
1505	* may be other tasks in the thread group that are not exiting,
1506	* and they need their cpu timers adjusted.
1507	*
1508	* The group_leader is the last task to be released, so if we
1509	* cannot update_rlimit_cpu on it, then the entire process is
1510	* exiting and we do not need to update at all.
1511	*/
1512	update_rlimit_cpu(task: tsk->group_leader, rlim_new: new_rlim->rlim_cur);
1513	}
1514
1515	return retval;
1516	}
1517
1518	SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1519	{
1520	struct rlimit value;
1521	int ret;
1522
1523	ret = do_prlimit(current, resource, NULL, old_rlim: &value);
1524	if (!ret)
1525	ret = copy_to_user(to: rlim, from: &value, n: sizeof(*rlim)) ? -EFAULT : `0`;
1526
1527	return ret;
1528	}
1529
1530	#ifdef CONFIG_COMPAT
1531
1532	COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1533	struct compat_rlimit __user *, rlim)
1534	{
1535	struct rlimit r;
1536	struct compat_rlimit r32;
1537
1538	if (copy_from_user(to: &r32, from: rlim, n: sizeof(struct compat_rlimit)))
1539	return -EFAULT;
1540
1541	if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1542	r.rlim_cur = RLIM_INFINITY;
1543	else
1544	r.rlim_cur = r32.rlim_cur;
1545	if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1546	r.rlim_max = RLIM_INFINITY;
1547	else
1548	r.rlim_max = r32.rlim_max;
1549	return do_prlimit(current, resource, new_rlim: &r, NULL);
1550	}
1551
1552	COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1553	struct compat_rlimit __user *, rlim)
1554	{
1555	struct rlimit r;
1556	int ret;
1557
1558	ret = do_prlimit(current, resource, NULL, old_rlim: &r);
1559	if (!ret) {
1560	struct compat_rlimit r32;
1561	if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1562	r32.rlim_cur = COMPAT_RLIM_INFINITY;
1563	else
1564	r32.rlim_cur = r.rlim_cur;
1565	if (r.rlim_max > COMPAT_RLIM_INFINITY)
1566	r32.rlim_max = COMPAT_RLIM_INFINITY;
1567	else
1568	r32.rlim_max = r.rlim_max;
1569
1570	if (copy_to_user(to: rlim, from: &r32, n: sizeof(struct compat_rlimit)))
1571	return -EFAULT;
1572	}
1573	return ret;
1574	}
1575
1576	#endif
1577
1578	#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1579
1580	/*
1581	* Back compatibility for getrlimit. Needed for some apps.
1582	*/
1583	SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1584	struct rlimit __user *, rlim)
1585	{
1586	struct rlimit x;
1587	if (resource >= RLIM_NLIMITS)
1588	return -EINVAL;
1589
1590	resource = array_index_nospec(resource, RLIM_NLIMITS);
1591	task_lock(current->group_leader);
1592	x = current->signal->rlim[resource];
1593	task_unlock(current->group_leader);
1594	if (x.rlim_cur > `0x7FFFFFFF`)
1595	x.rlim_cur = `0x7FFFFFFF`;
1596	if (x.rlim_max > `0x7FFFFFFF`)
1597	x.rlim_max = `0x7FFFFFFF`;
1598	return copy_to_user(to: rlim, from: &x, n: sizeof(x)) ? -EFAULT : `0`;
1599	}
1600
1601	#ifdef CONFIG_COMPAT
1602	COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1603	struct compat_rlimit __user *, rlim)
1604	{
1605	struct rlimit r;
1606
1607	if (resource >= RLIM_NLIMITS)
1608	return -EINVAL;
1609
1610	resource = array_index_nospec(resource, RLIM_NLIMITS);
1611	task_lock(current->group_leader);
1612	r = current->signal->rlim[resource];
1613	task_unlock(current->group_leader);
1614	if (r.rlim_cur > `0x7FFFFFFF`)
1615	r.rlim_cur = `0x7FFFFFFF`;
1616	if (r.rlim_max > `0x7FFFFFFF`)
1617	r.rlim_max = `0x7FFFFFFF`;
1618
1619	if (put_user(r.rlim_cur, &rlim->rlim_cur) \|\|
1620	put_user(r.rlim_max, &rlim->rlim_max))
1621	return -EFAULT;
1622	return `0`;
1623	}
1624	#endif
1625
1626	#endif
1627
1628	static inline bool rlim64_is_infinity(__u64 rlim64)
1629	{
1630	#if BITS_PER_LONG < 64
1631	return rlim64 >= ULONG_MAX;
1632	#else
1633	return rlim64 == RLIM64_INFINITY;
1634	#endif
1635	}
1636
1637	static void rlim_to_rlim64(const struct rlimit rlim, struct* rlimit64 *rlim64)
1638	{
1639	if (rlim->rlim_cur == RLIM_INFINITY)
1640	rlim64->rlim_cur = RLIM64_INFINITY;
1641	else
1642	rlim64->rlim_cur = rlim->rlim_cur;
1643	if (rlim->rlim_max == RLIM_INFINITY)
1644	rlim64->rlim_max = RLIM64_INFINITY;
1645	else
1646	rlim64->rlim_max = rlim->rlim_max;
1647	}
1648
1649	static void rlim64_to_rlim(const struct rlimit64 rlim64, struct* rlimit *rlim)
1650	{
1651	if (rlim64_is_infinity(rlim64: rlim64->rlim_cur))
1652	rlim->rlim_cur = RLIM_INFINITY;
1653	else
1654	rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1655	if (rlim64_is_infinity(rlim64: rlim64->rlim_max))
1656	rlim->rlim_max = RLIM_INFINITY;
1657	else
1658	rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1659	}
1660
1661	/ rcu lock must be held /
1662	static int check_prlimit_permission(struct task_struct *task,
1663	unsigned int flags)
1664	{
1665	const struct cred cred = current_cred(), tcred;
1666	bool id_match;
1667
1668	if (current == task)
1669	return `0`;
1670
1671	tcred = __task_cred(task);
1672	id_match = (uid_eq(left: cred->uid, right: tcred->euid) &&
1673	uid_eq(left: cred->uid, right: tcred->suid) &&
1674	uid_eq(left: cred->uid, right: tcred->uid) &&
1675	gid_eq(left: cred->gid, right: tcred->egid) &&
1676	gid_eq(left: cred->gid, right: tcred->sgid) &&
1677	gid_eq(left: cred->gid, right: tcred->gid));
1678	if (!id_match && !ns_capable(ns: tcred->user_ns, CAP_SYS_RESOURCE))
1679	return -EPERM;
1680
1681	return security_task_prlimit(cred, tcred, flags);
1682	}
1683
1684	SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1685	const struct rlimit64 __user *, new_rlim,
1686	struct rlimit64 __user *, old_rlim)
1687	{
1688	struct rlimit64 old64, new64;
1689	struct rlimit old, new;
1690	struct task_struct *tsk;
1691	unsigned int checkflags = `0`;
1692	int ret;
1693
1694	if (old_rlim)
1695	checkflags \|= LSM_PRLIMIT_READ;
1696
1697	if (new_rlim) {
1698	if (copy_from_user(to: &new64, from: new_rlim, n: sizeof(new64)))
1699	return -EFAULT;
1700	rlim64_to_rlim(rlim64: &new64, rlim: &new);
1701	checkflags \|= LSM_PRLIMIT_WRITE;
1702	}
1703
1704	rcu_read_lock();
1705	tsk = pid ? find_task_by_vpid(nr: pid) : current;
1706	if (!tsk) {
1707	rcu_read_unlock();
1708	return -ESRCH;
1709	}
1710	ret = check_prlimit_permission(task: tsk, flags: checkflags);
1711	if (ret) {
1712	rcu_read_unlock();
1713	return ret;
1714	}
1715	get_task_struct(t: tsk);
1716	rcu_read_unlock();
1717
1718	ret = do_prlimit(tsk, resource, new_rlim: new_rlim ? &new : NULL,
1719	old_rlim: old_rlim ? &old : NULL);
1720
1721	if (!ret && old_rlim) {
1722	rlim_to_rlim64(rlim: &old, rlim64: &old64);
1723	if (copy_to_user(to: old_rlim, from: &old64, n: sizeof(old64)))
1724	ret = -EFAULT;
1725	}
1726
1727	put_task_struct(t: tsk);
1728	return ret;
1729	}
1730
1731	SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1732	{
1733	struct rlimit new_rlim;
1734
1735	if (copy_from_user(to: &new_rlim, from: rlim, n: sizeof(*rlim)))
1736	return -EFAULT;
1737	return do_prlimit(current, resource, new_rlim: &new_rlim, NULL);
1738	}
1739
1740	/*
1741	* It would make sense to put struct rusage in the task_struct,
1742	* except that would make the task_struct be really big. After
1743	* task_struct gets moved into malloc'ed memory, it would
1744	* make sense to do this. It will make moving the rest of the information
1745	* a lot simpler! (Which we're not doing right now because we're not
1746	* measuring them yet).
1747	*
1748	* When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1749	* races with threads incrementing their own counters. But since word
1750	* reads are atomic, we either get new values or old values and we don't
1751	* care which for the sums. We always take the siglock to protect reading
1752	* the c* fields from p->signal from races with exit.c updating those
1753	* fields when reaping, so a sample either gets all the additions of a
1754	* given child after it's reaped, or none so this sample is before reaping.
1755	*
1756	* Locking:
1757	* We need to take the siglock for CHILDEREN, SELF and BOTH
1758	* for the cases current multithreaded, non-current single threaded
1759	* non-current multithreaded. Thread traversal is now safe with
1760	* the siglock held.
1761	* Strictly speaking, we donot need to take the siglock if we are current and
1762	* single threaded, as no one else can take our signal_struct away, no one
1763	* else can reap the children to update signal->c* counters, and no one else
1764	* can race with the signal-> fields. If we do not take any lock, the
1765	* signal-> fields could be read out of order while another thread was just
1766	* exiting. So we should place a read memory barrier when we avoid the lock.
1767	* On the writer side, write memory barrier is implied in __exit_signal
1768	* as __exit_signal releases the siglock spinlock after updating the signal->
1769	* fields. But we don't do this yet to keep things simple.
1770	*
1771	*/
1772
1773	static void accumulate_thread_rusage(struct task_struct t, struct* rusage *r)
1774	{
1775	r->ru_nvcsw += t->nvcsw;
1776	r->ru_nivcsw += t->nivcsw;
1777	r->ru_minflt += t->min_flt;
1778	r->ru_majflt += t->maj_flt;
1779	r->ru_inblock += task_io_get_inblock(p: t);
1780	r->ru_oublock += task_io_get_oublock(p: t);
1781	}
1782
1783	void getrusage(struct task_struct p, int* who, struct rusage *r)
1784	{
1785	struct task_struct *t;
1786	unsigned long flags;
1787	u64 tgutime, tgstime, utime, stime;
1788	unsigned long maxrss = `0`;
1789	struct signal_struct *sig = p->signal;
1790
1791	memset((char )r, `0`, sizeof* (*r));
1792	utime = stime = `0`;
1793
1794	if (who == RUSAGE_THREAD) {
1795	task_cputime_adjusted(current, ut: &utime, st: &stime);
1796	accumulate_thread_rusage(t: p, r);
1797	maxrss = sig->maxrss;
1798	goto out;
1799	}
1800
1801	if (!lock_task_sighand(task: p, flags: &flags))
1802	return;
1803
1804	switch (who) {
1805	case RUSAGE_BOTH:
1806	case RUSAGE_CHILDREN:
1807	utime = sig->cutime;
1808	stime = sig->cstime;
1809	r->ru_nvcsw = sig->cnvcsw;
1810	r->ru_nivcsw = sig->cnivcsw;
1811	r->ru_minflt = sig->cmin_flt;
1812	r->ru_majflt = sig->cmaj_flt;
1813	r->ru_inblock = sig->cinblock;
1814	r->ru_oublock = sig->coublock;
1815	maxrss = sig->cmaxrss;
1816
1817	if (who == RUSAGE_CHILDREN)
1818	break;
1819	fallthrough;
1820
1821	case RUSAGE_SELF:
1822	thread_group_cputime_adjusted(p, ut: &tgutime, st: &tgstime);
1823	utime += tgutime;
1824	stime += tgstime;
1825	r->ru_nvcsw += sig->nvcsw;
1826	r->ru_nivcsw += sig->nivcsw;
1827	r->ru_minflt += sig->min_flt;
1828	r->ru_majflt += sig->maj_flt;
1829	r->ru_inblock += sig->inblock;
1830	r->ru_oublock += sig->oublock;
1831	if (maxrss < sig->maxrss)
1832	maxrss = sig->maxrss;
1833	__for_each_thread(sig, t)
1834	accumulate_thread_rusage(t, r);
1835	break;
1836
1837	default:
1838	BUG();
1839	}
1840	unlock_task_sighand(task: p, flags: &flags);
1841
1842	out:
1843	r->ru_utime = ns_to_kernel_old_timeval(nsec: utime);
1844	r->ru_stime = ns_to_kernel_old_timeval(nsec: stime);
1845
1846	if (who != RUSAGE_CHILDREN) {
1847	struct mm_struct *mm = get_task_mm(task: p);
1848
1849	if (mm) {
1850	setmax_mm_hiwater_rss(maxrss: &maxrss, mm);
1851	mmput(mm);
1852	}
1853	}
1854	r->ru_maxrss = maxrss * (PAGE_SIZE / `1024`); / convert pages to KBs /
1855	}
1856
1857	SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1858	{
1859	struct rusage r;
1860
1861	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1862	who != RUSAGE_THREAD)
1863	return -EINVAL;
1864
1865	getrusage(current, who, r: &r);
1866	return copy_to_user(to: ru, from: &r, n: sizeof(r)) ? -EFAULT : `0`;
1867	}
1868
1869	#ifdef CONFIG_COMPAT
1870	COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1871	{
1872	struct rusage r;
1873
1874	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1875	who != RUSAGE_THREAD)
1876	return -EINVAL;
1877
1878	getrusage(current, who, r: &r);
1879	return put_compat_rusage(&r, ru);
1880	}
1881	#endif
1882
1883	SYSCALL_DEFINE1(umask, int, mask)
1884	{
1885	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1886	return mask;
1887	}
1888
1889	static int prctl_set_mm_exe_file(struct mm_struct mm, unsigned* int fd)
1890	{
1891	struct fd exe;
1892	struct inode *inode;
1893	int err;
1894
1895	exe = fdget(fd);
1896	if (!exe.file)
1897	return -EBADF;
1898
1899	inode = file_inode(f: exe.file);
1900
1901	/*
1902	* Because the original mm->exe_file points to executable file, make
1903	* sure that this one is executable as well, to avoid breaking an
1904	* overall picture.
1905	*/
1906	err = -EACCES;
1907	if (!S_ISREG(inode->i_mode) \|\| path_noexec(path: &exe.file->f_path))
1908	goto exit;
1909
1910	err = file_permission(file: exe.file, MAY_EXEC);
1911	if (err)
1912	goto exit;
1913
1914	err = replace_mm_exe_file(mm, new_exe_file: exe.file);
1915	exit:
1916	fdput(fd: exe);
1917	return err;
1918	}
1919
1920	/*
1921	* Check arithmetic relations of passed addresses.
1922	*
1923	* WARNING: we don't require any capability here so be very careful
1924	* in what is allowed for modification from userspace.
1925	*/
1926	static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1927	{
1928	unsigned long mmap_max_addr = TASK_SIZE;
1929	int error = -EINVAL, i;
1930
1931	static const unsigned char offsets[] = {
1932	offsetof(struct prctl_mm_map, start_code),
1933	offsetof(struct prctl_mm_map, end_code),
1934	offsetof(struct prctl_mm_map, start_data),
1935	offsetof(struct prctl_mm_map, end_data),
1936	offsetof(struct prctl_mm_map, start_brk),
1937	offsetof(struct prctl_mm_map, brk),
1938	offsetof(struct prctl_mm_map, start_stack),
1939	offsetof(struct prctl_mm_map, arg_start),
1940	offsetof(struct prctl_mm_map, arg_end),
1941	offsetof(struct prctl_mm_map, env_start),
1942	offsetof(struct prctl_mm_map, env_end),
1943	};
1944
1945	/*
1946	* Make sure the members are not somewhere outside
1947	* of allowed address space.
1948	*/
1949	for (i = `0`; i < ARRAY_SIZE(offsets); i++) {
1950	u64 val = (u64 )((char *)prctl_map + offsets[i]);
1951
1952	if ((unsigned long)val >= mmap_max_addr \|\|
1953	(unsigned long)val < mmap_min_addr)
1954	goto out;
1955	}
1956
1957	/*
1958	* Make sure the pairs are ordered.
1959	*/
1960	#define __prctl_check_order(__m1, __op, __m2) \
1961	((unsigned long)prctl_map->__m1 __op \
1962	(unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1963	error = __prctl_check_order(start_code, <, end_code);
1964	error \|= __prctl_check_order(start_data,<=, end_data);
1965	error \|= __prctl_check_order(start_brk, <=, brk);
1966	error \|= __prctl_check_order(arg_start, <=, arg_end);
1967	error \|= __prctl_check_order(env_start, <=, env_end);
1968	if (error)
1969	goto out;
1970	#undef __prctl_check_order
1971
1972	error = -EINVAL;
1973
1974	/*
1975	* Neither we should allow to override limits if they set.
1976	*/
1977	if (check_data_rlimit(rlim: rlimit(RLIMIT_DATA), new: prctl_map->brk,
1978	start: prctl_map->start_brk, end_data: prctl_map->end_data,
1979	start_data: prctl_map->start_data))
1980	goto out;
1981
1982	error = `0`;
1983	out:
1984	return error;
1985	}
1986
1987	#ifdef CONFIG_CHECKPOINT_RESTORE
1988	static int prctl_set_mm_map(int opt, const void __user addr, unsigned* long data_size)
1989	{
1990	struct prctl_mm_map prctl_map = { .exe_fd = (u32)-`1`, };
1991	unsigned long user_auxv[AT_VECTOR_SIZE];
1992	struct mm_struct *mm = current->mm;
1993	int error;
1994
1995	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1996	BUILD_BUG_ON(sizeof(struct prctl_mm_map) > `256`);
1997
1998	if (opt == PR_SET_MM_MAP_SIZE)
1999	return put_user((unsigned int)sizeof(prctl_map),
2000	(unsigned int __user *)addr);
2001
2002	if (data_size != sizeof(prctl_map))
2003	return -EINVAL;
2004
2005	if (copy_from_user(to: &prctl_map, from: addr, n: sizeof(prctl_map)))
2006	return -EFAULT;
2007
2008	error = validate_prctl_map_addr(prctl_map: &prctl_map);
2009	if (error)
2010	return error;
2011
2012	if (prctl_map.auxv_size) {
2013	/*
2014	* Someone is trying to cheat the auxv vector.
2015	*/
2016	if (!prctl_map.auxv \|\|
2017	prctl_map.auxv_size > sizeof(mm->saved_auxv))
2018	return -EINVAL;
2019
2020	memset(user_auxv, `0`, sizeof(user_auxv));
2021	if (copy_from_user(to: user_auxv,
2022	from: (const void __user *)prctl_map.auxv,
2023	n: prctl_map.auxv_size))
2024	return -EFAULT;
2025
2026	/ Last entry must be AT_NULL as specification requires /
2027	user_auxv[AT_VECTOR_SIZE - `2`] = AT_NULL;
2028	user_auxv[AT_VECTOR_SIZE - `1`] = AT_NULL;
2029	}
2030
2031	if (prctl_map.exe_fd != (u32)-`1`) {
2032	/*
2033	* Check if the current user is checkpoint/restore capable.
2034	* At the time of this writing, it checks for CAP_SYS_ADMIN
2035	* or CAP_CHECKPOINT_RESTORE.
2036	* Note that a user with access to ptrace can masquerade an
2037	* arbitrary program as any executable, even setuid ones.
2038	* This may have implications in the tomoyo subsystem.
2039	*/
2040	if (!checkpoint_restore_ns_capable(current_user_ns()))
2041	return -EPERM;
2042
2043	error = prctl_set_mm_exe_file(mm, fd: prctl_map.exe_fd);
2044	if (error)
2045	return error;
2046	}
2047
2048	/*
2049	* arg_lock protects concurrent updates but we still need mmap_lock for
2050	* read to exclude races with sys_brk.
2051	*/
2052	mmap_read_lock(mm);
2053
2054	/*
2055	* We don't validate if these members are pointing to
2056	* real present VMAs because application may have correspond
2057	* VMAs already unmapped and kernel uses these members for statistics
2058	* output in procfs mostly, except
2059	*
2060	* - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2061	* for VMAs when updating these members so anything wrong written
2062	* here cause kernel to swear at userspace program but won't lead
2063	* to any problem in kernel itself
2064	*/
2065
2066	spin_lock(lock: &mm->arg_lock);
2067	mm->start_code = prctl_map.start_code;
2068	mm->end_code = prctl_map.end_code;
2069	mm->start_data = prctl_map.start_data;
2070	mm->end_data = prctl_map.end_data;
2071	mm->start_brk = prctl_map.start_brk;
2072	mm->brk = prctl_map.brk;
2073	mm->start_stack = prctl_map.start_stack;
2074	mm->arg_start = prctl_map.arg_start;
2075	mm->arg_end = prctl_map.arg_end;
2076	mm->env_start = prctl_map.env_start;
2077	mm->env_end = prctl_map.env_end;
2078	spin_unlock(lock: &mm->arg_lock);
2079
2080	/*
2081	* Note this update of @saved_auxv is lockless thus
2082	* if someone reads this member in procfs while we're
2083	* updating -- it may get partly updated results. It's
2084	* known and acceptable trade off: we leave it as is to
2085	* not introduce additional locks here making the kernel
2086	* more complex.
2087	*/
2088	if (prctl_map.auxv_size)
2089	memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2090
2091	mmap_read_unlock(mm);
2092	return `0`;
2093	}
2094	#endif /* CONFIG_CHECKPOINT_RESTORE */
2095
2096	static int prctl_set_auxv(struct mm_struct mm, unsigned* long addr,
2097	unsigned long len)
2098	{
2099	/*
2100	* This doesn't move the auxiliary vector itself since it's pinned to
2101	* mm_struct, but it permits filling the vector with new values. It's
2102	* up to the caller to provide sane values here, otherwise userspace
2103	* tools which use this vector might be unhappy.
2104	*/
2105	unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2106
2107	if (len > sizeof(user_auxv))
2108	return -EINVAL;
2109
2110	if (copy_from_user(to: user_auxv, from: (const void __user *)addr, n: len))
2111	return -EFAULT;
2112
2113	/ Make sure the last entry is always AT_NULL /
2114	user_auxv[AT_VECTOR_SIZE - `2`] = `0`;
2115	user_auxv[AT_VECTOR_SIZE - `1`] = `0`;
2116
2117	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2118
2119	task_lock(current);
2120	memcpy(mm->saved_auxv, user_auxv, len);
2121	task_unlock(current);
2122
2123	return `0`;
2124	}
2125
2126	static int prctl_set_mm(int opt, unsigned long addr,
2127	unsigned long arg4, unsigned long arg5)
2128	{
2129	struct mm_struct *mm = current->mm;
2130	struct prctl_mm_map prctl_map = {
2131	.auxv = NULL,
2132	.auxv_size = `0`,
2133	.exe_fd = -`1`,
2134	};
2135	struct vm_area_struct *vma;
2136	int error;
2137
2138	if (arg5 \|\| (arg4 && (opt != PR_SET_MM_AUXV &&
2139	opt != PR_SET_MM_MAP &&
2140	opt != PR_SET_MM_MAP_SIZE)))
2141	return -EINVAL;
2142
2143	#ifdef CONFIG_CHECKPOINT_RESTORE
2144	if (opt == PR_SET_MM_MAP \|\| opt == PR_SET_MM_MAP_SIZE)
2145	return prctl_set_mm_map(opt, addr: (const void __user *)addr, data_size: arg4);
2146	#endif
2147
2148	if (!capable(CAP_SYS_RESOURCE))
2149	return -EPERM;
2150
2151	if (opt == PR_SET_MM_EXE_FILE)
2152	return prctl_set_mm_exe_file(mm, fd: (unsigned int)addr);
2153
2154	if (opt == PR_SET_MM_AUXV)
2155	return prctl_set_auxv(mm, addr, len: arg4);
2156
2157	if (addr >= TASK_SIZE \|\| addr < mmap_min_addr)
2158	return -EINVAL;
2159
2160	error = -EINVAL;
2161
2162	/*
2163	* arg_lock protects concurrent updates of arg boundaries, we need
2164	* mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2165	* validation.
2166	*/
2167	mmap_read_lock(mm);
2168	vma = find_vma(mm, addr);
2169
2170	spin_lock(lock: &mm->arg_lock);
2171	prctl_map.start_code = mm->start_code;
2172	prctl_map.end_code = mm->end_code;
2173	prctl_map.start_data = mm->start_data;
2174	prctl_map.end_data = mm->end_data;
2175	prctl_map.start_brk = mm->start_brk;
2176	prctl_map.brk = mm->brk;
2177	prctl_map.start_stack = mm->start_stack;
2178	prctl_map.arg_start = mm->arg_start;
2179	prctl_map.arg_end = mm->arg_end;
2180	prctl_map.env_start = mm->env_start;
2181	prctl_map.env_end = mm->env_end;
2182
2183	switch (opt) {
2184	case PR_SET_MM_START_CODE:
2185	prctl_map.start_code = addr;
2186	break;
2187	case PR_SET_MM_END_CODE:
2188	prctl_map.end_code = addr;
2189	break;
2190	case PR_SET_MM_START_DATA:
2191	prctl_map.start_data = addr;
2192	break;
2193	case PR_SET_MM_END_DATA:
2194	prctl_map.end_data = addr;
2195	break;
2196	case PR_SET_MM_START_STACK:
2197	prctl_map.start_stack = addr;
2198	break;
2199	case PR_SET_MM_START_BRK:
2200	prctl_map.start_brk = addr;
2201	break;
2202	case PR_SET_MM_BRK:
2203	prctl_map.brk = addr;
2204	break;
2205	case PR_SET_MM_ARG_START:
2206	prctl_map.arg_start = addr;
2207	break;
2208	case PR_SET_MM_ARG_END:
2209	prctl_map.arg_end = addr;
2210	break;
2211	case PR_SET_MM_ENV_START:
2212	prctl_map.env_start = addr;
2213	break;
2214	case PR_SET_MM_ENV_END:
2215	prctl_map.env_end = addr;
2216	break;
2217	default:
2218	goto out;
2219	}
2220
2221	error = validate_prctl_map_addr(prctl_map: &prctl_map);
2222	if (error)
2223	goto out;
2224
2225	switch (opt) {
2226	/*
2227	* If command line arguments and environment
2228	* are placed somewhere else on stack, we can
2229	* set them up here, ARG_START/END to setup
2230	* command line arguments and ENV_START/END
2231	* for environment.
2232	*/
2233	case PR_SET_MM_START_STACK:
2234	case PR_SET_MM_ARG_START:
2235	case PR_SET_MM_ARG_END:
2236	case PR_SET_MM_ENV_START:
2237	case PR_SET_MM_ENV_END:
2238	if (!vma) {
2239	error = -EFAULT;
2240	goto out;
2241	}
2242	}
2243
2244	mm->start_code = prctl_map.start_code;
2245	mm->end_code = prctl_map.end_code;
2246	mm->start_data = prctl_map.start_data;
2247	mm->end_data = prctl_map.end_data;
2248	mm->start_brk = prctl_map.start_brk;
2249	mm->brk = prctl_map.brk;
2250	mm->start_stack = prctl_map.start_stack;
2251	mm->arg_start = prctl_map.arg_start;
2252	mm->arg_end = prctl_map.arg_end;
2253	mm->env_start = prctl_map.env_start;
2254	mm->env_end = prctl_map.env_end;
2255
2256	error = `0`;
2257	out:
2258	spin_unlock(lock: &mm->arg_lock);
2259	mmap_read_unlock(mm);
2260	return error;
2261	}
2262
2263	#ifdef CONFIG_CHECKPOINT_RESTORE
2264	static int prctl_get_tid_address(struct task_struct me, int* __user * __user *tid_addr)
2265	{
2266	return put_user(me->clear_child_tid, tid_addr);
2267	}
2268	#else
2269	static int prctl_get_tid_address(struct task_struct me, int* __user * __user *tid_addr)
2270	{
2271	return -EINVAL;
2272	}
2273	#endif
2274
2275	static int propagate_has_child_subreaper(struct task_struct p, void* *data)
2276	{
2277	/*
2278	* If task has has_child_subreaper - all its descendants
2279	* already have these flag too and new descendants will
2280	* inherit it on fork, skip them.
2281	*
2282	* If we've found child_reaper - skip descendants in
2283	* it's subtree as they will never get out pidns.
2284	*/
2285	if (p->signal->has_child_subreaper \|\|
2286	is_child_reaper(pid: task_pid(task: p)))
2287	return `0`;
2288
2289	p->signal->has_child_subreaper = `1`;
2290	return `1`;
2291	}
2292
2293	int __weak arch_prctl_spec_ctrl_get(struct task_struct t, unsigned* long which)
2294	{
2295	return -EINVAL;
2296	}
2297
2298	int __weak arch_prctl_spec_ctrl_set(struct task_struct t, unsigned* long which,
2299	unsigned long ctrl)
2300	{
2301	return -EINVAL;
2302	}
2303
2304	#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO \| PF_LOCAL_THROTTLE)
2305
2306	#ifdef CONFIG_ANON_VMA_NAME
2307
2308	#define ANON_VMA_NAME_MAX_LEN 80
2309	#define ANON_VMA_NAME_INVALID_CHARS "\\`$[]"
2310
2311	static inline bool is_valid_name_char(char ch)
2312	{
2313	/ printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS /
2314	return ch > `0x1f` && ch < `0x7f` &&
2315	!strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2316	}
2317
2318	static int prctl_set_vma(unsigned long opt, unsigned long addr,
2319	unsigned long size, unsigned long arg)
2320	{
2321	struct mm_struct *mm = current->mm;
2322	const char __user *uname;
2323	struct anon_vma_name *anon_name = NULL;
2324	int error;
2325
2326	switch (opt) {
2327	case PR_SET_VMA_ANON_NAME:
2328	uname = (const char __user *)arg;
2329	if (uname) {
2330	char name, pch;
2331
2332	name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2333	if (IS_ERR(ptr: name))
2334	return PTR_ERR(ptr: name);
2335
2336	for (pch = name; *pch != `'\0'`; pch++) {
2337	if (!is_valid_name_char(ch: *pch)) {
2338	kfree(objp: name);
2339	return -EINVAL;
2340	}
2341	}
2342	/ anon_vma has its own copy /
2343	anon_name = anon_vma_name_alloc(name);
2344	kfree(objp: name);
2345	if (!anon_name)
2346	return -ENOMEM;
2347
2348	}
2349
2350	mmap_write_lock(mm);
2351	error = madvise_set_anon_name(mm, start: addr, len_in: size, anon_name);
2352	mmap_write_unlock(mm);
2353	anon_vma_name_put(anon_name);
2354	break;
2355	default:
2356	error = -EINVAL;
2357	}
2358
2359	return error;
2360	}
2361
2362	#else /* CONFIG_ANON_VMA_NAME */
2363	static int prctl_set_vma(unsigned long opt, unsigned long start,
2364	unsigned long size, unsigned long arg)
2365	{
2366	return -EINVAL;
2367	}
2368	#endif /* CONFIG_ANON_VMA_NAME */
2369
2370	static inline unsigned long get_current_mdwe(void)
2371	{
2372	unsigned long ret = `0`;
2373
2374	if (test_bit(MMF_HAS_MDWE, &current->mm->flags))
2375	ret \|= PR_MDWE_REFUSE_EXEC_GAIN;
2376	if (test_bit(MMF_HAS_MDWE_NO_INHERIT, &current->mm->flags))
2377	ret \|= PR_MDWE_NO_INHERIT;
2378
2379	return ret;
2380	}
2381
2382	static inline int prctl_set_mdwe(unsigned long bits, unsigned long arg3,
2383	unsigned long arg4, unsigned long arg5)
2384	{
2385	unsigned long current_bits;
2386
2387	if (arg3 \|\| arg4 \|\| arg5)
2388	return -EINVAL;
2389
2390	if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN \| PR_MDWE_NO_INHERIT))
2391	return -EINVAL;
2392
2393	/ NO_INHERIT only makes sense with REFUSE_EXEC_GAIN /
2394	if (bits & PR_MDWE_NO_INHERIT && !(bits & PR_MDWE_REFUSE_EXEC_GAIN))
2395	return -EINVAL;
2396
2397	current_bits = get_current_mdwe();
2398	if (current_bits && current_bits != bits)
2399	return -EPERM; / Cannot unset the flags /
2400
2401	if (bits & PR_MDWE_NO_INHERIT)
2402	set_bit(MMF_HAS_MDWE_NO_INHERIT, addr: &current->mm->flags);
2403	if (bits & PR_MDWE_REFUSE_EXEC_GAIN)
2404	set_bit(MMF_HAS_MDWE, addr: &current->mm->flags);
2405
2406	return `0`;
2407	}
2408
2409	static inline int prctl_get_mdwe(unsigned long arg2, unsigned long arg3,
2410	unsigned long arg4, unsigned long arg5)
2411	{
2412	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2413	return -EINVAL;
2414	return get_current_mdwe();
2415	}
2416
2417	static int prctl_get_auxv(void __user addr, unsigned* long len)
2418	{
2419	struct mm_struct *mm = current->mm;
2420	unsigned long size = min_t(unsigned long, sizeof(mm->saved_auxv), len);
2421
2422	if (size && copy_to_user(to: addr, from: mm->saved_auxv, n: size))
2423	return -EFAULT;
2424	return sizeof(mm->saved_auxv);
2425	}
2426
2427	SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2428	unsigned long, arg4, unsigned long, arg5)
2429	{
2430	struct task_struct *me = current;
2431	unsigned char comm[sizeof(me->comm)];
2432	long error;
2433
2434	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2435	if (error != -ENOSYS)
2436	return error;
2437
2438	error = `0`;
2439	switch (option) {
2440	case PR_SET_PDEATHSIG:
2441	if (!valid_signal(sig: arg2)) {
2442	error = -EINVAL;
2443	break;
2444	}
2445	me->pdeath_signal = arg2;
2446	break;
2447	case PR_GET_PDEATHSIG:
2448	error = put_user(me->pdeath_signal, (int __user *)arg2);
2449	break;
2450	case PR_GET_DUMPABLE:
2451	error = get_dumpable(mm: me->mm);
2452	break;
2453	case PR_SET_DUMPABLE:
2454	if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2455	error = -EINVAL;
2456	break;
2457	}
2458	set_dumpable(mm: me->mm, value: arg2);
2459	break;
2460
2461	case PR_SET_UNALIGN:
2462	error = SET_UNALIGN_CTL(me, arg2);
2463	break;
2464	case PR_GET_UNALIGN:
2465	error = GET_UNALIGN_CTL(me, arg2);
2466	break;
2467	case PR_SET_FPEMU:
2468	error = SET_FPEMU_CTL(me, arg2);
2469	break;
2470	case PR_GET_FPEMU:
2471	error = GET_FPEMU_CTL(me, arg2);
2472	break;
2473	case PR_SET_FPEXC:
2474	error = SET_FPEXC_CTL(me, arg2);
2475	break;
2476	case PR_GET_FPEXC:
2477	error = GET_FPEXC_CTL(me, arg2);
2478	break;
2479	case PR_GET_TIMING:
2480	error = PR_TIMING_STATISTICAL;
2481	break;
2482	case PR_SET_TIMING:
2483	if (arg2 != PR_TIMING_STATISTICAL)
2484	error = -EINVAL;
2485	break;
2486	case PR_SET_NAME:
2487	comm[sizeof(me->comm) - `1`] = `0`;
2488	if (strncpy_from_user(dst: comm, src: (char __user *)arg2,
2489	count: sizeof(me->comm) - `1`) < `0`)
2490	return -EFAULT;
2491	set_task_comm(tsk: me, from: comm);
2492	proc_comm_connector(task: me);
2493	break;
2494	case PR_GET_NAME:
2495	get_task_comm(comm, me);
2496	if (copy_to_user(to: (char __user )arg2, from: comm, n: sizeof*(comm)))
2497	return -EFAULT;
2498	break;
2499	case PR_GET_ENDIAN:
2500	error = GET_ENDIAN(me, arg2);
2501	break;
2502	case PR_SET_ENDIAN:
2503	error = SET_ENDIAN(me, arg2);
2504	break;
2505	case PR_GET_SECCOMP:
2506	error = prctl_get_seccomp();
2507	break;
2508	case PR_SET_SECCOMP:
2509	error = prctl_set_seccomp(arg2, (char __user *)arg3);
2510	break;
2511	case PR_GET_TSC:
2512	error = GET_TSC_CTL(arg2);
2513	break;
2514	case PR_SET_TSC:
2515	error = SET_TSC_CTL(arg2);
2516	break;
2517	case PR_TASK_PERF_EVENTS_DISABLE:
2518	error = perf_event_task_disable();
2519	break;
2520	case PR_TASK_PERF_EVENTS_ENABLE:
2521	error = perf_event_task_enable();
2522	break;
2523	case PR_GET_TIMERSLACK:
2524	if (current->timer_slack_ns > ULONG_MAX)
2525	error = ULONG_MAX;
2526	else
2527	error = current->timer_slack_ns;
2528	break;
2529	case PR_SET_TIMERSLACK:
2530	if (arg2 <= `0`)
2531	current->timer_slack_ns =
2532	current->default_timer_slack_ns;
2533	else
2534	current->timer_slack_ns = arg2;
2535	break;
2536	case PR_MCE_KILL:
2537	if (arg4 \| arg5)
2538	return -EINVAL;
2539	switch (arg2) {
2540	case PR_MCE_KILL_CLEAR:
2541	if (arg3 != `0`)
2542	return -EINVAL;
2543	current->flags &= ~PF_MCE_PROCESS;
2544	break;
2545	case PR_MCE_KILL_SET:
2546	current->flags \|= PF_MCE_PROCESS;
2547	if (arg3 == PR_MCE_KILL_EARLY)
2548	current->flags \|= PF_MCE_EARLY;
2549	else if (arg3 == PR_MCE_KILL_LATE)
2550	current->flags &= ~PF_MCE_EARLY;
2551	else if (arg3 == PR_MCE_KILL_DEFAULT)
2552	current->flags &=
2553	~(PF_MCE_EARLY\|PF_MCE_PROCESS);
2554	else
2555	return -EINVAL;
2556	break;
2557	default:
2558	return -EINVAL;
2559	}
2560	break;
2561	case PR_MCE_KILL_GET:
2562	if (arg2 \| arg3 \| arg4 \| arg5)
2563	return -EINVAL;
2564	if (current->flags & PF_MCE_PROCESS)
2565	error = (current->flags & PF_MCE_EARLY) ?
2566	PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2567	else
2568	error = PR_MCE_KILL_DEFAULT;
2569	break;
2570	case PR_SET_MM:
2571	error = prctl_set_mm(opt: arg2, addr: arg3, arg4, arg5);
2572	break;
2573	case PR_GET_TID_ADDRESS:
2574	error = prctl_get_tid_address(me, tid_addr: (int __user * __user *)arg2);
2575	break;
2576	case PR_SET_CHILD_SUBREAPER:
2577	me->signal->is_child_subreaper = !!arg2;
2578	if (!arg2)
2579	break;
2580
2581	walk_process_tree(top: me, propagate_has_child_subreaper, NULL);
2582	break;
2583	case PR_GET_CHILD_SUBREAPER:
2584	error = put_user(me->signal->is_child_subreaper,
2585	(int __user *)arg2);
2586	break;
2587	case PR_SET_NO_NEW_PRIVS:
2588	if (arg2 != `1` \|\| arg3 \|\| arg4 \|\| arg5)
2589	return -EINVAL;
2590
2591	task_set_no_new_privs(current);
2592	break;
2593	case PR_GET_NO_NEW_PRIVS:
2594	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2595	return -EINVAL;
2596	return task_no_new_privs(current) ? `1` : `0`;
2597	case PR_GET_THP_DISABLE:
2598	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2599	return -EINVAL;
2600	error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2601	break;
2602	case PR_SET_THP_DISABLE:
2603	if (arg3 \|\| arg4 \|\| arg5)
2604	return -EINVAL;
2605	if (mmap_write_lock_killable(mm: me->mm))
2606	return -EINTR;
2607	if (arg2)
2608	set_bit(MMF_DISABLE_THP, addr: &me->mm->flags);
2609	else
2610	clear_bit(MMF_DISABLE_THP, addr: &me->mm->flags);
2611	mmap_write_unlock(mm: me->mm);
2612	break;
2613	case PR_MPX_ENABLE_MANAGEMENT:
2614	case PR_MPX_DISABLE_MANAGEMENT:
2615	/ No longer implemented: /
2616	return -EINVAL;
2617	case PR_SET_FP_MODE:
2618	error = SET_FP_MODE(me, arg2);
2619	break;
2620	case PR_GET_FP_MODE:
2621	error = GET_FP_MODE(me);
2622	break;
2623	case PR_SVE_SET_VL:
2624	error = SVE_SET_VL(arg2);
2625	break;
2626	case PR_SVE_GET_VL:
2627	error = SVE_GET_VL();
2628	break;
2629	case PR_SME_SET_VL:
2630	error = SME_SET_VL(arg2);
2631	break;
2632	case PR_SME_GET_VL:
2633	error = SME_GET_VL();
2634	break;
2635	case PR_GET_SPECULATION_CTRL:
2636	if (arg3 \|\| arg4 \|\| arg5)
2637	return -EINVAL;
2638	error = arch_prctl_spec_ctrl_get(t: me, which: arg2);
2639	break;
2640	case PR_SET_SPECULATION_CTRL:
2641	if (arg4 \|\| arg5)
2642	return -EINVAL;
2643	error = arch_prctl_spec_ctrl_set(t: me, which: arg2, ctrl: arg3);
2644	break;
2645	case PR_PAC_RESET_KEYS:
2646	if (arg3 \|\| arg4 \|\| arg5)
2647	return -EINVAL;
2648	error = PAC_RESET_KEYS(me, arg2);
2649	break;
2650	case PR_PAC_SET_ENABLED_KEYS:
2651	if (arg4 \|\| arg5)
2652	return -EINVAL;
2653	error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2654	break;
2655	case PR_PAC_GET_ENABLED_KEYS:
2656	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2657	return -EINVAL;
2658	error = PAC_GET_ENABLED_KEYS(me);
2659	break;
2660	case PR_SET_TAGGED_ADDR_CTRL:
2661	if (arg3 \|\| arg4 \|\| arg5)
2662	return -EINVAL;
2663	error = SET_TAGGED_ADDR_CTRL(arg2);
2664	break;
2665	case PR_GET_TAGGED_ADDR_CTRL:
2666	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2667	return -EINVAL;
2668	error = GET_TAGGED_ADDR_CTRL();
2669	break;
2670	case PR_SET_IO_FLUSHER:
2671	if (!capable(CAP_SYS_RESOURCE))
2672	return -EPERM;
2673
2674	if (arg3 \|\| arg4 \|\| arg5)
2675	return -EINVAL;
2676
2677	if (arg2 == `1`)
2678	current->flags \|= PR_IO_FLUSHER;
2679	else if (!arg2)
2680	current->flags &= ~PR_IO_FLUSHER;
2681	else
2682	return -EINVAL;
2683	break;
2684	case PR_GET_IO_FLUSHER:
2685	if (!capable(CAP_SYS_RESOURCE))
2686	return -EPERM;
2687
2688	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2689	return -EINVAL;
2690
2691	error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2692	break;
2693	case PR_SET_SYSCALL_USER_DISPATCH:
2694	error = set_syscall_user_dispatch(mode: arg2, offset: arg3, len: arg4,
2695	selector: (char __user *) arg5);
2696	break;
2697	#ifdef CONFIG_SCHED_CORE
2698	case PR_SCHED_CORE:
2699	error = sched_core_share_pid(cmd: arg2, pid: arg3, type: arg4, uaddr: arg5);
2700	break;
2701	#endif
2702	case PR_SET_MDWE:
2703	error = prctl_set_mdwe(bits: arg2, arg3, arg4, arg5);
2704	break;
2705	case PR_GET_MDWE:
2706	error = prctl_get_mdwe(arg2, arg3, arg4, arg5);
2707	break;
2708	case PR_SET_VMA:
2709	error = prctl_set_vma(opt: arg2, addr: arg3, size: arg4, arg: arg5);
2710	break;
2711	case PR_GET_AUXV:
2712	if (arg4 \|\| arg5)
2713	return -EINVAL;
2714	error = prctl_get_auxv(addr: (void __user *)arg2, len: arg3);
2715	break;
2716	#ifdef CONFIG_KSM
2717	case PR_SET_MEMORY_MERGE:
2718	if (arg3 \|\| arg4 \|\| arg5)
2719	return -EINVAL;
2720	if (mmap_write_lock_killable(mm: me->mm))
2721	return -EINTR;
2722
2723	if (arg2)
2724	error = ksm_enable_merge_any(mm: me->mm);
2725	else
2726	error = ksm_disable_merge_any(mm: me->mm);
2727	mmap_write_unlock(mm: me->mm);
2728	break;
2729	case PR_GET_MEMORY_MERGE:
2730	if (arg2 \|\| arg3 \|\| arg4 \|\| arg5)
2731	return -EINVAL;
2732
2733	error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
2734	break;
2735	#endif
2736	case PR_RISCV_V_SET_CONTROL:
2737	error = RISCV_V_SET_CONTROL(arg2);
2738	break;
2739	case PR_RISCV_V_GET_CONTROL:
2740	error = RISCV_V_GET_CONTROL();
2741	break;
2742	default:
2743	error = -EINVAL;
2744	break;
2745	}
2746	return error;
2747	}
2748
2749	SYSCALL_DEFINE3(getcpu, unsigned __user , cpup, unsigned* __user *, nodep,
2750	struct getcpu_cache __user *, unused)
2751	{
2752	int err = `0`;
2753	int cpu = raw_smp_processor_id();
2754
2755	if (cpup)
2756	err \|= put_user(cpu, cpup);
2757	if (nodep)
2758	err \|= put_user(cpu_to_node(cpu), nodep);
2759	return err ? -EFAULT : `0`;
2760	}
2761
2762	/**
2763	* do_sysinfo - fill in sysinfo struct
2764	* @info: pointer to buffer to fill
2765	*/
2766	static int do_sysinfo(struct sysinfo *info)
2767	{
2768	unsigned long mem_total, sav_total;
2769	unsigned int mem_unit, bitcount;
2770	struct timespec64 tp;
2771
2772	memset(info, `0`, sizeof(struct sysinfo));
2773
2774	ktime_get_boottime_ts64(ts: &tp);
2775	timens_add_boottime(ts: &tp);
2776	info->uptime = tp.tv_sec + (tp.tv_nsec ? `1` : `0`);
2777
2778	get_avenrun(loads: info->loads, offset: `0`, SI_LOAD_SHIFT - FSHIFT);
2779
2780	info->procs = nr_threads;
2781
2782	si_meminfo(val: info);
2783	si_swapinfo(info);
2784
2785	/*
2786	* If the sum of all the available memory (i.e. ram + swap)
2787	* is less than can be stored in a 32 bit unsigned long then
2788	* we can be binary compatible with 2.2.x kernels. If not,
2789	* well, in that case 2.2.x was broken anyways...
2790	*
2791	* -Erik Andersen <andersee@debian.org>
2792	*/
2793
2794	mem_total = info->totalram + info->totalswap;
2795	if (mem_total < info->totalram \|\| mem_total < info->totalswap)
2796	goto out;
2797	bitcount = `0`;
2798	mem_unit = info->mem_unit;
2799	while (mem_unit > `1`) {
2800	bitcount++;
2801	mem_unit >>= `1`;
2802	sav_total = mem_total;
2803	mem_total <<= `1`;
2804	if (mem_total < sav_total)
2805	goto out;
2806	}
2807
2808	/*
2809	* If mem_total did not overflow, multiply all memory values by
2810	* info->mem_unit and set it to 1. This leaves things compatible
2811	* with 2.2.x, and also retains compatibility with earlier 2.4.x
2812	* kernels...
2813	*/
2814
2815	info->mem_unit = `1`;
2816	info->totalram <<= bitcount;
2817	info->freeram <<= bitcount;
2818	info->sharedram <<= bitcount;
2819	info->bufferram <<= bitcount;
2820	info->totalswap <<= bitcount;
2821	info->freeswap <<= bitcount;
2822	info->totalhigh <<= bitcount;
2823	info->freehigh <<= bitcount;
2824
2825	out:
2826	return `0`;
2827	}
2828
2829	SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2830	{
2831	struct sysinfo val;
2832
2833	do_sysinfo(info: &val);
2834
2835	if (copy_to_user(to: info, from: &val, n: sizeof(struct sysinfo)))
2836	return -EFAULT;
2837
2838	return `0`;
2839	}
2840
2841	#ifdef CONFIG_COMPAT
2842	struct compat_sysinfo {
2843	s32 uptime;
2844	u32 loads[`3`];
2845	u32 totalram;
2846	u32 freeram;
2847	u32 sharedram;
2848	u32 bufferram;
2849	u32 totalswap;
2850	u32 freeswap;
2851	u16 procs;
2852	u16 pad;
2853	u32 totalhigh;
2854	u32 freehigh;
2855	u32 mem_unit;
2856	char _f[`20`-`2`*sizeof(u32)-sizeof(int)];
2857	};
2858
2859	COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2860	{
2861	struct sysinfo s;
2862	struct compat_sysinfo s_32;
2863
2864	do_sysinfo(info: &s);
2865
2866	/ Check to see if any memory value is too large for 32-bit and scale*
2867	* down if needed
2868	*/
2869	if (upper_32_bits(s.totalram) \|\| upper_32_bits(s.totalswap)) {
2870	int bitcount = `0`;
2871
2872	while (s.mem_unit < PAGE_SIZE) {
2873	s.mem_unit <<= `1`;
2874	bitcount++;
2875	}
2876
2877	s.totalram >>= bitcount;
2878	s.freeram >>= bitcount;
2879	s.sharedram >>= bitcount;
2880	s.bufferram >>= bitcount;
2881	s.totalswap >>= bitcount;
2882	s.freeswap >>= bitcount;
2883	s.totalhigh >>= bitcount;
2884	s.freehigh >>= bitcount;
2885	}
2886
2887	memset(&s_32, `0`, sizeof(s_32));
2888	s_32.uptime = s.uptime;
2889	s_32.loads[`0`] = s.loads[`0`];
2890	s_32.loads[`1`] = s.loads[`1`];
2891	s_32.loads[`2`] = s.loads[`2`];
2892	s_32.totalram = s.totalram;
2893	s_32.freeram = s.freeram;
2894	s_32.sharedram = s.sharedram;
2895	s_32.bufferram = s.bufferram;
2896	s_32.totalswap = s.totalswap;
2897	s_32.freeswap = s.freeswap;
2898	s_32.procs = s.procs;
2899	s_32.totalhigh = s.totalhigh;
2900	s_32.freehigh = s.freehigh;
2901	s_32.mem_unit = s.mem_unit;
2902	if (copy_to_user(to: info, from: &s_32, n: sizeof(s_32)))
2903	return -EFAULT;
2904	return `0`;
2905	}
2906	#endif /* CONFIG_COMPAT */
2907

source code of linux/kernel/sys.c