forked from ~ljy/RK356X_SDK_RELEASE

blame | history | patch | commit | commitdiff
hc
2024-05-10 cde9070d9970eef1f7ec2360586c802a16230ad8 
 kernel/kernel/sys.c
....@@ -49,6 +49,7 @@
4949 #include <linux/syscalls.h>
5050 #include <linux/kprobes.h>
5151 #include <linux/user_namespace.h>
52
+#include <linux/time_namespace.h>
5253 #include <linux/binfmts.h>
5354 
5455 #include <linux/sched.h>
....@@ -74,6 +75,8 @@
7475 #include <asm/unistd.h>
7576 
7677 #include "uid16.h"
78
+
79
+#include <trace/hooks/sys.h>
7780 
7881 #ifndef SET_UNALIGN_CTL
7982 # define SET_UNALIGN_CTL(a, b)	(-EINVAL)
....@@ -105,12 +108,6 @@
105108 #ifndef SET_TSC_CTL
106109 # define SET_TSC_CTL(a)		(-EINVAL)
107110 #endif
108
-#ifndef MPX_ENABLE_MANAGEMENT
109
-# define MPX_ENABLE_MANAGEMENT()	(-EINVAL)
110
-#endif
111
-#ifndef MPX_DISABLE_MANAGEMENT
112
-# define MPX_DISABLE_MANAGEMENT()	(-EINVAL)
113
-#endif
114111 #ifndef GET_FP_MODE
115112 # define GET_FP_MODE(a)		(-EINVAL)
116113 #endif
....@@ -122,6 +119,15 @@
122119 #endif
123120 #ifndef SVE_GET_VL
124121 # define SVE_GET_VL()		(-EINVAL)
122
+#endif
123
+#ifndef PAC_RESET_KEYS
124
+# define PAC_RESET_KEYS(a, b)	(-EINVAL)
125
+#endif
126
+#ifndef PAC_SET_ENABLED_KEYS
127
+# define PAC_SET_ENABLED_KEYS(a, b, c)	(-EINVAL)
128
+#endif
129
+#ifndef PAC_GET_ENABLED_KEYS
130
+# define PAC_GET_ENABLED_KEYS(a)	(-EINVAL)
125131 #endif
126132 #ifndef SET_TAGGED_ADDR_CTRL
127133 # define SET_TAGGED_ADDR_CTRL(a)	(-EINVAL)
....@@ -377,7 +383,7 @@
377383 	if (rgid != (gid_t) -1) {
378384 		if (gid_eq(old->gid, krgid) ||
379385 		    gid_eq(old->egid, krgid) ||
380
-		    ns_capable(old->user_ns, CAP_SETGID))
386
+		    ns_capable_setid(old->user_ns, CAP_SETGID))
381387 			new->gid = krgid;
382388 		else
383389 			goto error;
....@@ -386,7 +392,7 @@
386392 		if (gid_eq(old->gid, kegid) ||
387393 		    gid_eq(old->egid, kegid) ||
388394 		    gid_eq(old->sgid, kegid) ||
389
-		    ns_capable(old->user_ns, CAP_SETGID))
395
+		    ns_capable_setid(old->user_ns, CAP_SETGID))
390396 			new->egid = kegid;
391397 		else
392398 			goto error;
....@@ -396,6 +402,10 @@
396402 	    (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
397403 		new->sgid = new->egid;
398404 	new->fsgid = new->egid;
405
+
406
+	retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
407
+	if (retval < 0)
408
+		goto error;
399409 
400410 	return commit_creds(new);
401411 
....@@ -432,11 +442,15 @@
432442 	old = current_cred();
433443 
434444 	retval = -EPERM;
435
-	if (ns_capable(old->user_ns, CAP_SETGID))
445
+	if (ns_capable_setid(old->user_ns, CAP_SETGID))
436446 		new->gid = new->egid = new->sgid = new->fsgid = kgid;
437447 	else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
438448 		new->egid = new->fsgid = kgid;
439449 	else
450
+		goto error;
451
+
452
+	retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
453
+	if (retval < 0)
440454 		goto error;
441455 
442456 	return commit_creds(new);
....@@ -521,7 +535,7 @@
521535 		new->uid = kruid;
522536 		if (!uid_eq(old->uid, kruid) &&
523537 		    !uid_eq(old->euid, kruid) &&
524
-		    !ns_capable(old->user_ns, CAP_SETUID))
538
+		    !ns_capable_setid(old->user_ns, CAP_SETUID))
525539 			goto error;
526540 	}
527541 
....@@ -530,7 +544,7 @@
530544 		if (!uid_eq(old->uid, keuid) &&
531545 		    !uid_eq(old->euid, keuid) &&
532546 		    !uid_eq(old->suid, keuid) &&
533
-		    !ns_capable(old->user_ns, CAP_SETUID))
547
+		    !ns_capable_setid(old->user_ns, CAP_SETUID))
534548 			goto error;
535549 	}
536550 
....@@ -589,7 +603,7 @@
589603 	old = current_cred();
590604 
591605 	retval = -EPERM;
592
-	if (ns_capable(old->user_ns, CAP_SETUID)) {
606
+	if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
593607 		new->suid = new->uid = kuid;
594608 		if (!uid_eq(kuid, old->uid)) {
595609 			retval = set_user(new);
....@@ -630,6 +644,7 @@
630644 	struct cred *new;
631645 	int retval;
632646 	kuid_t kruid, keuid, ksuid;
647
+	bool ruid_new, euid_new, suid_new;
633648 
634649 	kruid = make_kuid(ns, ruid);
635650 	keuid = make_kuid(ns, euid);
....@@ -644,24 +659,28 @@
644659 	if ((suid != (uid_t) -1) && !uid_valid(ksuid))
645660 		return -EINVAL;
646661 
662
+	old = current_cred();
663
+
664
+	/* check for no-op */
665
+	if ((ruid == (uid_t) -1 || uid_eq(kruid, old->uid)) &&
666
+	    (euid == (uid_t) -1 || (uid_eq(keuid, old->euid) &&
667
+				    uid_eq(keuid, old->fsuid))) &&
668
+	    (suid == (uid_t) -1 || uid_eq(ksuid, old->suid)))
669
+		return 0;
670
+
671
+	ruid_new = ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
672
+		   !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid);
673
+	euid_new = euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
674
+		   !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid);
675
+	suid_new = suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
676
+		   !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid);
677
+	if ((ruid_new || euid_new || suid_new) &&
678
+	    !ns_capable_setid(old->user_ns, CAP_SETUID))
679
+		return -EPERM;
680
+
647681 	new = prepare_creds();
648682 	if (!new)
649683 		return -ENOMEM;
650
-
651
-	old = current_cred();
652
-
653
-	retval = -EPERM;
654
-	if (!ns_capable(old->user_ns, CAP_SETUID)) {
655
-		if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
656
-		    !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
657
-			goto error;
658
-		if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
659
-		    !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
660
-			goto error;
661
-		if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
662
-		    !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
663
-			goto error;
664
-	}
665684 
666685 	if (ruid != (uid_t) -1) {
667686 		new->uid = kruid;
....@@ -722,6 +741,7 @@
722741 	struct cred *new;
723742 	int retval;
724743 	kgid_t krgid, kegid, ksgid;
744
+	bool rgid_new, egid_new, sgid_new;
725745 
726746 	krgid = make_kgid(ns, rgid);
727747 	kegid = make_kgid(ns, egid);
....@@ -734,23 +754,28 @@
734754 	if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
735755 		return -EINVAL;
736756 
757
+	old = current_cred();
758
+
759
+	/* check for no-op */
760
+	if ((rgid == (gid_t) -1 || gid_eq(krgid, old->gid)) &&
761
+	    (egid == (gid_t) -1 || (gid_eq(kegid, old->egid) &&
762
+				    gid_eq(kegid, old->fsgid))) &&
763
+	    (sgid == (gid_t) -1 || gid_eq(ksgid, old->sgid)))
764
+		return 0;
765
+
766
+	rgid_new = rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
767
+		   !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid);
768
+	egid_new = egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
769
+		   !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid);
770
+	sgid_new = sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
771
+		   !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid);
772
+	if ((rgid_new || egid_new || sgid_new) &&
773
+	    !ns_capable_setid(old->user_ns, CAP_SETGID))
774
+		return -EPERM;
775
+
737776 	new = prepare_creds();
738777 	if (!new)
739778 		return -ENOMEM;
740
-	old = current_cred();
741
-
742
-	retval = -EPERM;
743
-	if (!ns_capable(old->user_ns, CAP_SETGID)) {
744
-		if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
745
-		    !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
746
-			goto error;
747
-		if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
748
-		    !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
749
-			goto error;
750
-		if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
751
-		    !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
752
-			goto error;
753
-	}
754779 
755780 	if (rgid != (gid_t) -1)
756781 		new->gid = krgid;
....@@ -759,6 +784,10 @@
759784 	if (sgid != (gid_t) -1)
760785 		new->sgid = ksgid;
761786 	new->fsgid = new->egid;
787
+
788
+	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
789
+	if (retval < 0)
790
+		goto error;
762791 
763792 	return commit_creds(new);
764793 
....@@ -819,7 +848,7 @@
819848 
820849 	if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
821850 	    uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
822
-	    ns_capable(old->user_ns, CAP_SETUID)) {
851
+	    ns_capable_setid(old->user_ns, CAP_SETUID)) {
823852 		if (!uid_eq(kuid, old->fsuid)) {
824853 			new->fsuid = kuid;
825854 			if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
....@@ -863,10 +892,11 @@
863892 
864893 	if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
865894 	    gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
866
-	    ns_capable(old->user_ns, CAP_SETGID)) {
895
+	    ns_capable_setid(old->user_ns, CAP_SETGID)) {
867896 		if (!gid_eq(kgid, old->fsgid)) {
868897 			new->fsgid = kgid;
869
-			goto change_okay;
898
+			if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
899
+				goto change_okay;
870900 		}
871901 	}
872902 
....@@ -1212,7 +1242,8 @@
12121242 /*
12131243  * Work around broken programs that cannot handle "Linux 3.0".
12141244  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1215
- * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
1245
+ * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1246
+ * 2.6.60.
12161247  */
12171248 static int override_release(char __user *release, size_t len)
12181249 {
....@@ -1538,6 +1569,8 @@
15381569 
15391570 	if (resource >= RLIM_NLIMITS)
15401571 		return -EINVAL;
1572
+	resource = array_index_nospec(resource, RLIM_NLIMITS);
1573
+
15411574 	if (new_rlim) {
15421575 		if (new_rlim->rlim_cur > new_rlim->rlim_max)
15431576 			return -EINVAL;
....@@ -1563,15 +1596,6 @@
15631596 			retval = -EPERM;
15641597 		if (!retval)
15651598 			retval = security_task_setrlimit(tsk, resource, new_rlim);
1566
-		if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1567
-			/*
1568
-			 * The caller is asking for an immediate RLIMIT_CPU
1569
-			 * expiry.  But we use the zero value to mean "it was
1570
-			 * never set".  So let's cheat and make it one second
1571
-			 * instead
1572
-			 */
1573
-			new_rlim->rlim_cur = 1;
1574
-		}
15751599 	}
15761600 	if (!retval) {
15771601 		if (old_rlim)
....@@ -1582,10 +1606,9 @@
15821606 	task_unlock(tsk->group_leader);
15831607 
15841608 	/*
1585
-	 * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1586
-	 * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1587
-	 * very long-standing error, and fixing it now risks breakage of
1588
-	 * applications, so we live with it
1609
+	 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1610
+	 * infite. In case of RLIM_INFINITY the posix CPU timer code
1611
+	 * ignores the rlimit.
15891612 	 */
15901613 	 if (!retval && new_rlim && resource == RLIMIT_CPU &&
15911614 	     new_rlim->rlim_cur != RLIM_INFINITY &&
....@@ -1753,6 +1776,7 @@
17531776 
17541777 		if (who == RUSAGE_CHILDREN)
17551778 			break;
1779
+		fallthrough;
17561780 
17571781 	case RUSAGE_SELF:
17581782 		thread_group_cputime_adjusted(p, &tgutime, &tgstime);
....@@ -1778,8 +1802,8 @@
17781802 	unlock_task_sighand(p, &flags);
17791803 
17801804 out:
1781
-	r->ru_utime = ns_to_timeval(utime);
1782
-	r->ru_stime = ns_to_timeval(stime);
1805
+	r->ru_utime = ns_to_kernel_old_timeval(utime);
1806
+	r->ru_stime = ns_to_kernel_old_timeval(stime);
17831807 
17841808 	if (who != RUSAGE_CHILDREN) {
17851809 		struct mm_struct *mm = get_task_mm(p);
....@@ -1858,7 +1882,7 @@
18581882 	if (exe_file) {
18591883 		struct vm_area_struct *vma;
18601884 
1861
-		down_read(&mm->mmap_sem);
1885
+		mmap_read_lock(mm);
18621886 		for (vma = mm->mmap; vma; vma = vma->vm_next) {
18631887 			if (!vma->vm_file)
18641888 				continue;
....@@ -1867,7 +1891,7 @@
18671891 				goto exit_err;
18681892 		}
18691893 
1870
-		up_read(&mm->mmap_sem);
1894
+		mmap_read_unlock(mm);
18711895 		fput(exe_file);
18721896 	}
18731897 
....@@ -1881,19 +1905,20 @@
18811905 	fdput(exe);
18821906 	return err;
18831907 exit_err:
1884
-	up_read(&mm->mmap_sem);
1908
+	mmap_read_unlock(mm);
18851909 	fput(exe_file);
18861910 	goto exit;
18871911 }
18881912 
18891913 /*
1914
+ * Check arithmetic relations of passed addresses.
1915
+ *
18901916  * WARNING: we don't require any capability here so be very careful
18911917  * in what is allowed for modification from userspace.
18921918  */
1893
-static int validate_prctl_map(struct prctl_mm_map *prctl_map)
1919
+static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
18941920 {
18951921 	unsigned long mmap_max_addr = TASK_SIZE;
1896
-	struct mm_struct *mm = current->mm;
18971922 	int error = -EINVAL, i;
18981923 
18991924 	static const unsigned char offsets[] = {
....@@ -1947,24 +1972,6 @@
19471972 			      prctl_map->start_data))
19481973 			goto out;
19491974 
1950
-	/*
1951
-	 * Someone is trying to cheat the auxv vector.
1952
-	 */
1953
-	if (prctl_map->auxv_size) {
1954
-		if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv))
1955
-			goto out;
1956
-	}
1957
-
1958
-	/*
1959
-	 * Finally, make sure the caller has the rights to
1960
-	 * change /proc/pid/exe link: only local sys admin should
1961
-	 * be allowed to.
1962
-	 */
1963
-	if (prctl_map->exe_fd != (u32)-1) {
1964
-		if (!ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1965
-			goto out;
1966
-	}
1967
-
19681975 	error = 0;
19691976 out:
19701977 	return error;
....@@ -1991,11 +1998,18 @@
19911998 	if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
19921999 		return -EFAULT;
19932000 
1994
-	error = validate_prctl_map(&prctl_map);
2001
+	error = validate_prctl_map_addr(&prctl_map);
19952002 	if (error)
19962003 		return error;
19972004 
19982005 	if (prctl_map.auxv_size) {
2006
+		/*
2007
+		 * Someone is trying to cheat the auxv vector.
2008
+		 */
2009
+		if (!prctl_map.auxv ||
2010
+				prctl_map.auxv_size > sizeof(mm->saved_auxv))
2011
+			return -EINVAL;
2012
+
19992013 		memset(user_auxv, 0, sizeof(user_auxv));
20002014 		if (copy_from_user(user_auxv,
20012015 				   (const void __user *)prctl_map.auxv,
....@@ -2008,16 +2022,27 @@
20082022 	}
20092023 
20102024 	if (prctl_map.exe_fd != (u32)-1) {
2025
+		/*
2026
+		 * Check if the current user is checkpoint/restore capable.
2027
+		 * At the time of this writing, it checks for CAP_SYS_ADMIN
2028
+		 * or CAP_CHECKPOINT_RESTORE.
2029
+		 * Note that a user with access to ptrace can masquerade an
2030
+		 * arbitrary program as any executable, even setuid ones.
2031
+		 * This may have implications in the tomoyo subsystem.
2032
+		 */
2033
+		if (!checkpoint_restore_ns_capable(current_user_ns()))
2034
+			return -EPERM;
2035
+
20112036 		error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
20122037 		if (error)
20132038 			return error;
20142039 	}
20152040 
20162041 	/*
2017
-	 * arg_lock protects concurent updates but we still need mmap_sem for
2042
+	 * arg_lock protects concurent updates but we still need mmap_lock for
20182043 	 * read to exclude races with sys_brk.
20192044 	 */
2020
-	down_read(&mm->mmap_sem);
2045
+	mmap_read_lock(mm);
20212046 
20222047 	/*
20232048 	 * We don't validate if these members are pointing to
....@@ -2025,7 +2050,7 @@
20252050 	 * VMAs already unmapped and kernel uses these members for statistics
20262051 	 * output in procfs mostly, except
20272052 	 *
2028
-	 *  - @start_brk/@brk which are used in do_brk but kernel lookups
2053
+	 *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
20292054 	 *    for VMAs when updating these memvers so anything wrong written
20302055 	 *    here cause kernel to swear at userspace program but won't lead
20312056 	 *    to any problem in kernel itself
....@@ -2056,7 +2081,7 @@
20562081 	if (prctl_map.auxv_size)
20572082 		memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
20582083 
2059
-	up_read(&mm->mmap_sem);
2084
+	mmap_read_unlock(mm);
20602085 	return 0;
20612086 }
20622087 #endif /* CONFIG_CHECKPOINT_RESTORE */
....@@ -2095,7 +2120,11 @@
20952120 			unsigned long arg4, unsigned long arg5)
20962121 {
20972122 	struct mm_struct *mm = current->mm;
2098
-	struct prctl_mm_map prctl_map;
2123
+	struct prctl_mm_map prctl_map = {
2124
+		.auxv = NULL,
2125
+		.auxv_size = 0,
2126
+		.exe_fd = -1,
2127
+	};
20992128 	struct vm_area_struct *vma;
21002129 	int error;
21012130 
....@@ -2123,9 +2152,15 @@
21232152 
21242153 	error = -EINVAL;
21252154 
2126
-	down_write(&mm->mmap_sem);
2155
+	/*
2156
+	 * arg_lock protects concurent updates of arg boundaries, we need
2157
+	 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2158
+	 * validation.
2159
+	 */
2160
+	mmap_read_lock(mm);
21272161 	vma = find_vma(mm, addr);
21282162 
2163
+	spin_lock(&mm->arg_lock);
21292164 	prctl_map.start_code	= mm->start_code;
21302165 	prctl_map.end_code	= mm->end_code;
21312166 	prctl_map.start_data	= mm->start_data;
....@@ -2137,9 +2172,6 @@
21372172 	prctl_map.arg_end	= mm->arg_end;
21382173 	prctl_map.env_start	= mm->env_start;
21392174 	prctl_map.env_end	= mm->env_end;
2140
-	prctl_map.auxv		= NULL;
2141
-	prctl_map.auxv_size	= 0;
2142
-	prctl_map.exe_fd	= -1;
21432175 
21442176 	switch (opt) {
21452177 	case PR_SET_MM_START_CODE:
....@@ -2179,7 +2211,7 @@
21792211 		goto out;
21802212 	}
21812213 
2182
-	error = validate_prctl_map(&prctl_map);
2214
+	error = validate_prctl_map_addr(&prctl_map);
21832215 	if (error)
21842216 		goto out;
21852217 
....@@ -2216,17 +2248,18 @@
22162248 
22172249 	error = 0;
22182250 out:
2219
-	up_write(&mm->mmap_sem);
2251
+	spin_unlock(&mm->arg_lock);
2252
+	mmap_read_unlock(mm);
22202253 	return error;
22212254 }
22222255 
22232256 #ifdef CONFIG_CHECKPOINT_RESTORE
2224
-static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2257
+static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
22252258 {
22262259 	return put_user(me->clear_child_tid, tid_addr);
22272260 }
22282261 #else
2229
-static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2262
+static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
22302263 {
22312264 	return -EINVAL;
22322265 }
....@@ -2358,7 +2391,7 @@
23582391 			return error;
23592392 		if (prev)
23602393 			vma = prev->vm_next;
2361
-		else	/* madvise_remove dropped mmap_sem */
2394
+		else	/* madvise_remove dropped mmap_lock */
23622395 			vma = find_vma(current->mm, start);
23632396 	}
23642397 }
....@@ -2386,7 +2419,7 @@
23862419 	if (end == start)
23872420 		return 0;
23882421 
2389
-	down_write(&mm->mmap_sem);
2422
+	mmap_write_lock(mm);
23902423 
23912424 	switch (opt) {
23922425 	case PR_SET_VMA_ANON_NAME:
....@@ -2396,7 +2429,7 @@
23962429 		error = -EINVAL;
23972430 	}
23982431 
2399
-	up_write(&mm->mmap_sem);
2432
+	mmap_write_unlock(mm);
24002433 
24012434 	return error;
24022435 }
....@@ -2407,6 +2440,8 @@
24072440 	return -EINVAL;
24082441 }
24092442 #endif
2443
+
2444
+#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
24102445 
24112446 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
24122447 		unsigned long, arg4, unsigned long, arg5)
....@@ -2555,7 +2590,7 @@
25552590 		error = prctl_set_mm(arg2, arg3, arg4, arg5);
25562591 		break;
25572592 	case PR_GET_TID_ADDRESS:
2558
-		error = prctl_get_tid_address(me, (int __user **)arg2);
2593
+		error = prctl_get_tid_address(me, (int __user * __user *)arg2);
25592594 		break;
25602595 	case PR_SET_CHILD_SUBREAPER:
25612596 		me->signal->is_child_subreaper = !!arg2;
....@@ -2586,24 +2621,18 @@
25862621 	case PR_SET_THP_DISABLE:
25872622 		if (arg3 || arg4 || arg5)
25882623 			return -EINVAL;
2589
-		if (down_write_killable(&me->mm->mmap_sem))
2624
+		if (mmap_write_lock_killable(me->mm))
25902625 			return -EINTR;
25912626 		if (arg2)
25922627 			set_bit(MMF_DISABLE_THP, &me->mm->flags);
25932628 		else
25942629 			clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2595
-		up_write(&me->mm->mmap_sem);
2630
+		mmap_write_unlock(me->mm);
25962631 		break;
25972632 	case PR_MPX_ENABLE_MANAGEMENT:
2598
-		if (arg2 || arg3 || arg4 || arg5)
2599
-			return -EINVAL;
2600
-		error = MPX_ENABLE_MANAGEMENT();
2601
-		break;
26022633 	case PR_MPX_DISABLE_MANAGEMENT:
2603
-		if (arg2 || arg3 || arg4 || arg5)
2604
-			return -EINVAL;
2605
-		error = MPX_DISABLE_MANAGEMENT();
2606
-		break;
2634
+		/* No longer implemented: */
2635
+		return -EINVAL;
26072636 	case PR_SET_FP_MODE:
26082637 		error = SET_FP_MODE(me, arg2);
26092638 		break;
....@@ -2629,6 +2658,21 @@
26292658 	case PR_SET_VMA:
26302659 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
26312660 		break;
2661
+	case PR_PAC_RESET_KEYS:
2662
+		if (arg3 || arg4 || arg5)
2663
+			return -EINVAL;
2664
+		error = PAC_RESET_KEYS(me, arg2);
2665
+		break;
2666
+	case PR_PAC_SET_ENABLED_KEYS:
2667
+		if (arg4 || arg5)
2668
+			return -EINVAL;
2669
+		error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2670
+		break;
2671
+	case PR_PAC_GET_ENABLED_KEYS:
2672
+		if (arg2 || arg3 || arg4 || arg5)
2673
+			return -EINVAL;
2674
+		error = PAC_GET_ENABLED_KEYS(me);
2675
+		break;
26322676 	case PR_SET_TAGGED_ADDR_CTRL:
26332677 		if (arg3 || arg4 || arg5)
26342678 			return -EINVAL;
....@@ -2639,10 +2683,34 @@
26392683 			return -EINVAL;
26402684 		error = GET_TAGGED_ADDR_CTRL();
26412685 		break;
2686
+	case PR_SET_IO_FLUSHER:
2687
+		if (!capable(CAP_SYS_RESOURCE))
2688
+			return -EPERM;
2689
+
2690
+		if (arg3 || arg4 || arg5)
2691
+			return -EINVAL;
2692
+
2693
+		if (arg2 == 1)
2694
+			current->flags |= PR_IO_FLUSHER;
2695
+		else if (!arg2)
2696
+			current->flags &= ~PR_IO_FLUSHER;
2697
+		else
2698
+			return -EINVAL;
2699
+		break;
2700
+	case PR_GET_IO_FLUSHER:
2701
+		if (!capable(CAP_SYS_RESOURCE))
2702
+			return -EPERM;
2703
+
2704
+		if (arg2 || arg3 || arg4 || arg5)
2705
+			return -EINVAL;
2706
+
2707
+		error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2708
+		break;
26422709 	default:
26432710 		error = -EINVAL;
26442711 		break;
26452712 	}
2713
+	trace_android_vh_syscall_prctl_finished(option, me);
26462714 	return error;
26472715 }
26482716 
....@@ -2672,6 +2740,7 @@
26722740 	memset(info, 0, sizeof(struct sysinfo));
26732741 
26742742 	ktime_get_boottime_ts64(&tp);
2743
+	timens_add_boottime(&tp);
26752744 	info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
26762745 
26772746 	get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
....@@ -2758,6 +2827,7 @@
27582827 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
27592828 {
27602829 	struct sysinfo s;
2830
+	struct compat_sysinfo s_32;
27612831 
27622832 	do_sysinfo(&s);
27632833 
....@@ -2782,23 +2852,23 @@
27822852 		s.freehigh >>= bitcount;
27832853 	}
27842854 
2785
-	if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2786
-	    __put_user(s.uptime, &info->uptime) ||
2787
-	    __put_user(s.loads[0], &info->loads[0]) ||
2788
-	    __put_user(s.loads[1], &info->loads[1]) ||
2789
-	    __put_user(s.loads[2], &info->loads[2]) ||
2790
-	    __put_user(s.totalram, &info->totalram) ||
2791
-	    __put_user(s.freeram, &info->freeram) ||
2792
-	    __put_user(s.sharedram, &info->sharedram) ||
2793
-	    __put_user(s.bufferram, &info->bufferram) ||
2794
-	    __put_user(s.totalswap, &info->totalswap) ||
2795
-	    __put_user(s.freeswap, &info->freeswap) ||
2796
-	    __put_user(s.procs, &info->procs) ||
2797
-	    __put_user(s.totalhigh, &info->totalhigh) ||
2798
-	    __put_user(s.freehigh, &info->freehigh) ||
2799
-	    __put_user(s.mem_unit, &info->mem_unit))
2855
+	memset(&s_32, 0, sizeof(s_32));
2856
+	s_32.uptime = s.uptime;
2857
+	s_32.loads[0] = s.loads[0];
2858
+	s_32.loads[1] = s.loads[1];
2859
+	s_32.loads[2] = s.loads[2];
2860
+	s_32.totalram = s.totalram;
2861
+	s_32.freeram = s.freeram;
2862
+	s_32.sharedram = s.sharedram;
2863
+	s_32.bufferram = s.bufferram;
2864
+	s_32.totalswap = s.totalswap;
2865
+	s_32.freeswap = s.freeswap;
2866
+	s_32.procs = s.procs;
2867
+	s_32.totalhigh = s.totalhigh;
2868
+	s_32.freehigh = s.freehigh;
2869
+	s_32.mem_unit = s.mem_unit;
2870
+	if (copy_to_user(info, &s_32, sizeof(s_32)))
28002871 		return -EFAULT;
2801
-
28022872 	return 0;
28032873 }
28042874 #endif /* CONFIG_COMPAT */