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, 94 insertions, 0 deletions

diff --git a/arch/openrisc/lib/memset.S b/arch/openrisc/lib/memset.S
new file mode 100644
index 000000000..c3ac2a8b6
--- /dev/null
+++ b/arch/openrisc/lib/memset.S
@@ -0,0 +1,94 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * OpenRISC memset.S
+ *
+ * Hand-optimized assembler version of memset for OpenRISC.
+ * Algorithm inspired by several other arch-specific memset routines
+ * in the kernel tree
+ *
+ * Copyright (C) 2015 Olof Kindgren <olof.kindgren@gmail.com>
+ */
+
+	.global memset
+	.type	memset, @function
+memset:
+	/* arguments:
+	 * r3 = *s
+	 * r4 = c
+	 * r5 = n
+	 * r13, r15, r17, r19 used as temp regs
+	*/
+
+	/* Exit if n == 0 */
+	l.sfeqi		r5, 0
+	l.bf		4f
+
+	/* Truncate c to char */
+	l.andi  	r13, r4, 0xff
+
+	/* Skip word extension if c is 0 */
+	l.sfeqi		r13, 0
+	l.bf		1f
+	/* Check for at least two whole words (8 bytes) */
+	 l.sfleui	r5, 7
+
+	/* Extend char c to 32-bit word cccc in r13 */
+	l.slli		r15, r13, 16  // r13 = 000c, r15 = 0c00
+	l.or		r13, r13, r15 // r13 = 0c0c, r15 = 0c00
+	l.slli		r15, r13, 8   // r13 = 0c0c, r15 = c0c0
+	l.or		r13, r13, r15 // r13 = cccc, r15 = c0c0
+
+1:	l.addi		r19, r3, 0 // Set r19 = src
+	/* Jump to byte copy loop if less than two words */
+	l.bf		3f
+	 l.or		r17, r5, r0 // Set r17 = n
+
+	/* Mask out two LSBs to check alignment */
+	l.andi		r15, r3, 0x3
+
+	/* lsb == 00, jump to word copy loop */
+	l.sfeqi		r15, 0
+	l.bf		2f
+	 l.addi		r19, r3, 0 // Set r19 = src
+
+	/* lsb == 01,10 or 11 */
+	l.sb		0(r3), r13   // *src = c
+	l.addi		r17, r17, -1 // Decrease n
+
+	l.sfeqi		r15, 3
+	l.bf		2f
+	 l.addi		r19, r3, 1  // src += 1
+
+	/* lsb == 01 or 10 */
+	l.sb		1(r3), r13   // *(src+1) = c
+	l.addi		r17, r17, -1 // Decrease n
+
+	l.sfeqi		r15, 2
+	l.bf		2f
+	 l.addi		r19, r3, 2  // src += 2
+
+	/* lsb == 01 */
+	l.sb		2(r3), r13   // *(src+2) = c
+	l.addi		r17, r17, -1 // Decrease n
+	l.addi		r19, r3, 3   // src += 3
+
+	/* Word copy loop */
+2:	l.sw		0(r19), r13  // *src = cccc
+	l.addi		r17, r17, -4 // Decrease n
+	l.sfgeui	r17, 4
+	l.bf		2b
+	 l.addi		r19, r19, 4  // Increase src
+
+	/* When n > 0, copy the remaining bytes, otherwise jump to exit */
+	l.sfeqi		r17, 0
+	l.bf		4f
+
+	/* Byte copy loop */
+3:	l.addi		r17, r17, -1 // Decrease n
+	l.sb		0(r19), r13  // *src = cccc
+	l.sfnei		r17, 0
+	l.bf		3b
+	 l.addi		r19, r19, 1  // Increase src
+
+4:	l.jr		r9
+	 l.ori		r11, r3, 0