x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 95 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-devices-soc b/Documentation/ABI/testing/sysfs-devices-soc
new file mode 100644
index 000000000..cb6776a4a
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-devices-soc
@@ -0,0 +1,95 @@
+What:		/sys/devices/socX
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		The /sys/devices/ directory contains a sub-directory for each
+		System-on-Chip (SoC) device on a running platform. Information
+		regarding each SoC can be obtained by reading sysfs files. This
+		functionality is only available if implemented by the platform.
+
+		The directory created for each SoC will also house information
+		about devices which are commonly contained in /sys/devices/platform.
+		It has been agreed that if an SoC device exists, its supported
+		devices would be better suited to appear as children of that SoC.
+
+What:		/sys/devices/socX/machine
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		Read-only attribute common to all SoCs. Contains the SoC machine
+		name (e.g. DB8500).
+
+What:		/sys/devices/socX/family
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		Read-only attribute common to all SoCs. Contains SoC family name
+		(e.g. ux500).
+
+		On many of ARM based silicon with SMCCC v1.2+ compliant firmware
+		this will contain the JEDEC JEP106 manufacturer’s identification
+		code. The format is "jep106:XXYY" where XX is identity code and
+		YY is continuation code.
+
+		This manufacturer’s identification code is defined by one
+		or more eight (8) bit fields, each consisting of seven (7)
+		data bits plus one (1) odd parity bit. It is a single field,
+		limiting the possible number of vendors to 126. To expand
+		the maximum number of identification codes, a continuation
+		scheme has been defined.
+
+		The specified mechanism is that an identity code of 0x7F
+		represents the "continuation code" and implies the presence
+		of an additional identity code field, and this mechanism
+		may be extended to multiple continuation codes followed
+		by the manufacturer's identity code.
+
+		For example, ARM has identity code 0x7F 0x7F 0x7F 0x7F 0x3B,
+		which is code 0x3B on the fifth 'page'. This is shortened
+		as JEP106 identity code of 0x3B and a continuation code of
+		0x4 to represent the four continuation codes preceding the
+		identity code.
+
+What:		/sys/devices/socX/serial_number
+Date:		January 2019
+contact:	Bjorn Andersson <bjorn.andersson@linaro.org>
+Description:
+		Read-only attribute supported by most SoCs. Contains the SoC's
+		serial number, if available.
+
+What:		/sys/devices/socX/soc_id
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		Read-only attribute supported by most SoCs. In the case of
+		ST-Ericsson's chips this contains the SoC serial number.
+
+		On many of ARM based silicon with SMCCC v1.2+ compliant firmware
+		this will contain the SOC ID appended to the family attribute
+		to ensure there is no conflict in this namespace across various
+		vendors. The format is "jep106:XXYY:ZZZZ" where XX is identity
+		code, YY is continuation code and ZZZZ is the SOC ID.
+
+What:		/sys/devices/socX/revision
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		Read-only attribute supported by most SoCs. Contains the SoC's
+		manufacturing revision number.
+
+What:		/sys/devices/socX/process
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		Read-only attribute supported ST-Ericsson's silicon. Contains the
+		the process by which the silicon chip was manufactured.
+
+What:		/sys/bus/soc
+Date:		January 2012
+contact:	Lee Jones <lee@kernel.org>
+Description:
+		The /sys/bus/soc/ directory contains the usual sub-folders
+		expected under most buses. /sys/bus/soc/devices is of particular
+		interest, as it contains a symlink for each SoC device found on
+		the system. Each symlink points back into the aforementioned
+		/sys/devices/socX devices.