cve/published/2021/CVE-2021-46957.mbox - pub/scm/linux/security/vulns - Git at Google

 From bippy-5f407fcff5a0 Mon Sep 17 00:00:00 2001
 From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
 To: <linux-cve-announce@vger.kernel.org>
 Reply-to: <cve@kernel.org>, <linux-kernel@vger.kernel.org>
 Subject: CVE-2021-46957: riscv/kprobe: fix kernel panic when invoking sys_read traced by kprobe

 Description
 ===========

 In the Linux kernel, the following vulnerability has been resolved:

 riscv/kprobe: fix kernel panic when invoking sys_read traced by kprobe

 The execution of sys_read end up hitting a BUG_ON() in __find_get_block
 after installing kprobe at sys_read, the BUG message like the following:

 [   65.708663] ------------[ cut here ]------------
 [   65.709987] kernel BUG at fs/buffer.c:1251!
 [   65.711283] Kernel BUG [#1]
 [   65.712032] Modules linked in:
 [   65.712925] CPU: 0 PID: 51 Comm: sh Not tainted 5.12.0-rc4 #1
 [   65.714407] Hardware name: riscv-virtio,qemu (DT)
 [   65.715696] epc : __find_get_block+0x218/0x2c8
 [   65.716835]  ra : __getblk_gfp+0x1c/0x4a
 [   65.717831] epc : ffffffe00019f11e ra : ffffffe00019f56a sp : ffffffe002437930
 [   65.719553]  gp : ffffffe000f06030 tp : ffffffe0015abc00 t0 : ffffffe00191e038
 [   65.721290]  t1 : ffffffe00191e038 t2 : 000000000000000a s0 : ffffffe002437960
 [   65.723051]  s1 : ffffffe00160ad00 a0 : ffffffe00160ad00 a1 : 000000000000012a
 [   65.724772]  a2 : 0000000000000400 a3 : 0000000000000008 a4 : 0000000000000040
 [   65.726545]  a5 : 0000000000000000 a6 : ffffffe00191e000 a7 : 0000000000000000
 [   65.728308]  s2 : 000000000000012a s3 : 0000000000000400 s4 : 0000000000000008
 [   65.730049]  s5 : 000000000000006c s6 : ffffffe00240f800 s7 : ffffffe000f080a8
 [   65.731802]  s8 : 0000000000000001 s9 : 000000000000012a s10: 0000000000000008
 [   65.733516]  s11: 0000000000000008 t3 : 00000000000003ff t4 : 000000000000000f
 [   65.734434]  t5 : 00000000000003ff t6 : 0000000000040000
 [   65.734613] status: 0000000000000100 badaddr: 0000000000000000 cause: 0000000000000003
 [   65.734901] Call Trace:
 [   65.735076] [<ffffffe00019f11e>] __find_get_block+0x218/0x2c8
 [   65.735417] [<ffffffe00020017a>] __ext4_get_inode_loc+0xb2/0x2f6
 [   65.735618] [<ffffffe000201b6c>] ext4_get_inode_loc+0x3a/0x8a
 [   65.735802] [<ffffffe000203380>] ext4_reserve_inode_write+0x2e/0x8c
 [   65.735999] [<ffffffe00020357a>] __ext4_mark_inode_dirty+0x4c/0x18e
 [   65.736208] [<ffffffe000206bb0>] ext4_dirty_inode+0x46/0x66
 [   65.736387] [<ffffffe000192914>] __mark_inode_dirty+0x12c/0x3da
 [   65.736576] [<ffffffe000180dd2>] touch_atime+0x146/0x150
 [   65.736748] [<ffffffe00010d762>] filemap_read+0x234/0x246
 [   65.736920] [<ffffffe00010d834>] generic_file_read_iter+0xc0/0x114
 [   65.737114] [<ffffffe0001f5d7a>] ext4_file_read_iter+0x42/0xea
 [   65.737310] [<ffffffe000163f2c>] new_sync_read+0xe2/0x15a
 [   65.737483] [<ffffffe000165814>] vfs_read+0xca/0xf2
 [   65.737641] [<ffffffe000165bae>] ksys_read+0x5e/0xc8
 [   65.737816] [<ffffffe000165c26>] sys_read+0xe/0x16
 [   65.737973] [<ffffffe000003972>] ret_from_syscall+0x0/0x2
 [   65.738858] ---[ end trace fe93f985456c935d ]---

 A simple reproducer looks like:
 	echo 'p:myprobe sys_read fd=%a0 buf=%a1 count=%a2' > /sys/kernel/debug/tracing/kprobe_events
 	echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable
 	cat /sys/kernel/debug/tracing/trace

 Here's what happens to hit that BUG_ON():

 1) After installing kprobe at entry of sys_read, the first instruction
    is replaced by 'ebreak' instruction on riscv64 platform.

 2) Once kernel reach the 'ebreak' instruction at the entry of sys_read,
    it trap into the riscv breakpoint handler, where it do something to
    setup for coming single-step of origin instruction, including backup
    the 'sstatus' in pt_regs, followed by disable interrupt during single
    stepping via clear 'SIE' bit of 'sstatus' in pt_regs.

 3) Then kernel restore to the instruction slot contains two instructions,
    one is original instruction at entry of sys_read, the other is 'ebreak'.
    Here it trigger a 'Instruction page fault' exception (value at 'scause'
    is '0xc'), if PF is not filled into PageTabe for that slot yet.

 4) Again kernel trap into page fault exception handler, where it choose
    different policy according to the state of running kprobe. Because
    afte 2) the state is KPROBE_HIT_SS, so kernel reset the current kprobe
    and 'pc' points back to the probe address.

 5) Because 'epc' point back to 'ebreak' instrution at sys_read probe,
    kernel trap into breakpoint handler again, and repeat the operations
    at 2), however 'sstatus' without 'SIE' is keep at 4), it cause the
    real 'sstatus' saved at 2) is overwritten by the one withou 'SIE'.

 6) When kernel cross the probe the 'sstatus' CSR restore with value
    without 'SIE', and reach __find_get_block where it requires the
    interrupt must be enabled.

 Fix this is very trivial, just restore the value of 'sstatus' in pt_regs
 with backup one at 2) when the instruction being single stepped cause a
 page fault.

 The Linux kernel CVE team has assigned CVE-2021-46957 to this issue.


 Affected and fixed versions
 ===========================

 	Issue introduced in 5.12 with commit c22b0bcb1dd024cb9caad9230e3a387d8b061df5 and fixed in 5.12.3 with commit fd0f06590d35c99f98d12c7984897ec4201a6263
 	Issue introduced in 5.12 with commit c22b0bcb1dd024cb9caad9230e3a387d8b061df5 and fixed in 5.13 with commit b1ebaa0e1318494a7637099a26add50509e37964

 Please see https://www.kernel.org for a full list of currently supported
 kernel versions by the kernel community.

 Unaffected versions might change over time as fixes are backported to
 older supported kernel versions.  The official CVE entry at
 	https://cve.org/CVERecord/?id=CVE-2021-46957
 will be updated if fixes are backported, please check that for the most
 up to date information about this issue.


 Affected files
 ==============

 The file(s) affected by this issue are:
 	arch/riscv/kernel/probes/kprobes.c


 Mitigation
 ==========

 The Linux kernel CVE team recommends that you update to the latest
 stable kernel version for this, and many other bugfixes.  Individual
 changes are never tested alone, but rather are part of a larger kernel
 release.  Cherry-picking individual commits is not recommended or
 supported by the Linux kernel community at all.  If however, updating to
 the latest release is impossible, the individual changes to resolve this
 issue can be found at these commits:
 	https://git.kernel.org/stable/c/fd0f06590d35c99f98d12c7984897ec4201a6263
 	https://git.kernel.org/stable/c/b1ebaa0e1318494a7637099a26add50509e37964
	From bippy-5f407fcff5a0 Mon Sep 17 00:00:00 2001
	From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
	To: <linux-cve-announce@vger.kernel.org>
	Reply-to: <cve@kernel.org>, <linux-kernel@vger.kernel.org>
	Subject: CVE-2021-46957: riscv/kprobe: fix kernel panic when invoking sys_read traced by kprobe

	Description
	===========

	In the Linux kernel, the following vulnerability has been resolved:

	riscv/kprobe: fix kernel panic when invoking sys_read traced by kprobe

	The execution of sys_read end up hitting a BUG_ON() in __find_get_block
	after installing kprobe at sys_read, the BUG message like the following:

	[ 65.708663] ------------[ cut here ]------------
	[ 65.709987] kernel BUG at fs/buffer.c:1251!
	[ 65.711283] Kernel BUG [#1]
	[ 65.712032] Modules linked in:
	[ 65.712925] CPU: 0 PID: 51 Comm: sh Not tainted 5.12.0-rc4 #1
	[ 65.714407] Hardware name: riscv-virtio,qemu (DT)
	[ 65.715696] epc : __find_get_block+0x218/0x2c8
	[ 65.716835] ra : __getblk_gfp+0x1c/0x4a
	[ 65.717831] epc : ffffffe00019f11e ra : ffffffe00019f56a sp : ffffffe002437930
	[ 65.719553] gp : ffffffe000f06030 tp : ffffffe0015abc00 t0 : ffffffe00191e038
	[ 65.721290] t1 : ffffffe00191e038 t2 : 000000000000000a s0 : ffffffe002437960
	[ 65.723051] s1 : ffffffe00160ad00 a0 : ffffffe00160ad00 a1 : 000000000000012a
	[ 65.724772] a2 : 0000000000000400 a3 : 0000000000000008 a4 : 0000000000000040
	[ 65.726545] a5 : 0000000000000000 a6 : ffffffe00191e000 a7 : 0000000000000000
	[ 65.728308] s2 : 000000000000012a s3 : 0000000000000400 s4 : 0000000000000008
	[ 65.730049] s5 : 000000000000006c s6 : ffffffe00240f800 s7 : ffffffe000f080a8
	[ 65.731802] s8 : 0000000000000001 s9 : 000000000000012a s10: 0000000000000008
	[ 65.733516] s11: 0000000000000008 t3 : 00000000000003ff t4 : 000000000000000f
	[ 65.734434] t5 : 00000000000003ff t6 : 0000000000040000
	[ 65.734613] status: 0000000000000100 badaddr: 0000000000000000 cause: 0000000000000003
	[ 65.734901] Call Trace:
	[ 65.735076] [<ffffffe00019f11e>] __find_get_block+0x218/0x2c8
	[ 65.735417] [<ffffffe00020017a>] __ext4_get_inode_loc+0xb2/0x2f6
	[ 65.735618] [<ffffffe000201b6c>] ext4_get_inode_loc+0x3a/0x8a
	[ 65.735802] [<ffffffe000203380>] ext4_reserve_inode_write+0x2e/0x8c
	[ 65.735999] [<ffffffe00020357a>] __ext4_mark_inode_dirty+0x4c/0x18e
	[ 65.736208] [<ffffffe000206bb0>] ext4_dirty_inode+0x46/0x66
	[ 65.736387] [<ffffffe000192914>] __mark_inode_dirty+0x12c/0x3da
	[ 65.736576] [<ffffffe000180dd2>] touch_atime+0x146/0x150
	[ 65.736748] [<ffffffe00010d762>] filemap_read+0x234/0x246
	[ 65.736920] [<ffffffe00010d834>] generic_file_read_iter+0xc0/0x114
	[ 65.737114] [<ffffffe0001f5d7a>] ext4_file_read_iter+0x42/0xea
	[ 65.737310] [<ffffffe000163f2c>] new_sync_read+0xe2/0x15a
	[ 65.737483] [<ffffffe000165814>] vfs_read+0xca/0xf2
	[ 65.737641] [<ffffffe000165bae>] ksys_read+0x5e/0xc8
	[ 65.737816] [<ffffffe000165c26>] sys_read+0xe/0x16
	[ 65.737973] [<ffffffe000003972>] ret_from_syscall+0x0/0x2
	[ 65.738858] ---[ end trace fe93f985456c935d ]---

	A simple reproducer looks like:
	echo 'p:myprobe sys_read fd=%a0 buf=%a1 count=%a2' > /sys/kernel/debug/tracing/kprobe_events
	echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable
	cat /sys/kernel/debug/tracing/trace

	Here's what happens to hit that BUG_ON():

	1) After installing kprobe at entry of sys_read, the first instruction
	is replaced by 'ebreak' instruction on riscv64 platform.

	2) Once kernel reach the 'ebreak' instruction at the entry of sys_read,
	it trap into the riscv breakpoint handler, where it do something to
	setup for coming single-step of origin instruction, including backup
	the 'sstatus' in pt_regs, followed by disable interrupt during single
	stepping via clear 'SIE' bit of 'sstatus' in pt_regs.

	3) Then kernel restore to the instruction slot contains two instructions,
	one is original instruction at entry of sys_read, the other is 'ebreak'.
	Here it trigger a 'Instruction page fault' exception (value at 'scause'
	is '0xc'), if PF is not filled into PageTabe for that slot yet.

	4) Again kernel trap into page fault exception handler, where it choose
	different policy according to the state of running kprobe. Because
	afte 2) the state is KPROBE_HIT_SS, so kernel reset the current kprobe
	and 'pc' points back to the probe address.

	5) Because 'epc' point back to 'ebreak' instrution at sys_read probe,
	kernel trap into breakpoint handler again, and repeat the operations
	at 2), however 'sstatus' without 'SIE' is keep at 4), it cause the
	real 'sstatus' saved at 2) is overwritten by the one withou 'SIE'.

	6) When kernel cross the probe the 'sstatus' CSR restore with value
	without 'SIE', and reach __find_get_block where it requires the
	interrupt must be enabled.

	Fix this is very trivial, just restore the value of 'sstatus' in pt_regs
	with backup one at 2) when the instruction being single stepped cause a
	page fault.

	The Linux kernel CVE team has assigned CVE-2021-46957 to this issue.


	Affected and fixed versions
	===========================

	Issue introduced in 5.12 with commit c22b0bcb1dd024cb9caad9230e3a387d8b061df5 and fixed in 5.12.3 with commit fd0f06590d35c99f98d12c7984897ec4201a6263
	Issue introduced in 5.12 with commit c22b0bcb1dd024cb9caad9230e3a387d8b061df5 and fixed in 5.13 with commit b1ebaa0e1318494a7637099a26add50509e37964

	Please see https://www.kernel.org for a full list of currently supported
	kernel versions by the kernel community.

	Unaffected versions might change over time as fixes are backported to
	older supported kernel versions. The official CVE entry at
	https://cve.org/CVERecord/?id=CVE-2021-46957
	will be updated if fixes are backported, please check that for the most
	up to date information about this issue.


	Affected files
	==============

	The file(s) affected by this issue are:
	arch/riscv/kernel/probes/kprobes.c


	Mitigation
	==========

	The Linux kernel CVE team recommends that you update to the latest
	stable kernel version for this, and many other bugfixes. Individual
	changes are never tested alone, but rather are part of a larger kernel
	release. Cherry-picking individual commits is not recommended or
	supported by the Linux kernel community at all. If however, updating to
	the latest release is impossible, the individual changes to resolve this
	issue can be found at these commits:
	https://git.kernel.org/stable/c/fd0f06590d35c99f98d12c7984897ec4201a6263
	https://git.kernel.org/stable/c/b1ebaa0e1318494a7637099a26add50509e37964