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, 217 insertions, 0 deletions

diff --git a/arch/sh/lib/movmem.S b/arch/sh/lib/movmem.S
new file mode 100644
index 000000000..8ac54d6b3
--- /dev/null
+++ b/arch/sh/lib/movmem.S
@@ -0,0 +1,217 @@
+/* SPDX-License-Identifier: GPL-2.0+ WITH GCC-exception-2.0
+
+   Copyright (C) 1994, 1995, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
+   2004, 2005, 2006
+   Free Software Foundation, Inc.
+*/
+
+!! libgcc routines for the Renesas / SuperH SH CPUs.
+!! Contributed by Steve Chamberlain.
+!! sac@cygnus.com
+
+!! ashiftrt_r4_x, ___ashrsi3, ___ashlsi3, ___lshrsi3 routines
+!! recoded in assembly by Toshiyasu Morita
+!! tm@netcom.com
+
+/* SH2 optimizations for ___ashrsi3, ___ashlsi3, ___lshrsi3 and
+   ELF local label prefixes by J"orn Rennecke
+   amylaar@cygnus.com  */
+
+	.text
+	.balign	4
+	.global	__movmem
+	.global __movstr
+	.set __movstr, __movmem	
+	/* This would be a lot simpler if r6 contained the byte count
+	   minus 64, and we wouldn't be called here for a byte count of 64.  */
+__movmem:
+	sts.l	pr,@-r15
+	shll2	r6
+	bsr	__movmemSI52+2
+	mov.l	@(48,r5),r0
+	.balign	4
+movmem_loop: /* Reached with rts */
+	mov.l	@(60,r5),r0
+	add	#-64,r6
+	mov.l	r0,@(60,r4)
+	tst	r6,r6
+	mov.l	@(56,r5),r0
+	bt	movmem_done
+	mov.l	r0,@(56,r4)
+	cmp/pl	r6
+	mov.l	@(52,r5),r0
+	add	#64,r5
+	mov.l	r0,@(52,r4)
+	add	#64,r4
+	bt	__movmemSI52
+! done all the large groups, do the remainder
+! jump to movmem+
+	mova	__movmemSI4+4,r0
+	add	r6,r0
+	jmp	@r0
+movmem_done: ! share slot insn, works out aligned.
+	lds.l	@r15+,pr
+	mov.l	r0,@(56,r4)
+	mov.l	@(52,r5),r0
+	rts
+	mov.l	r0,@(52,r4)
+	.balign	4
+
+	.global	__movmemSI64
+	.global __movstrSI64
+	.set	__movstrSI64, __movmemSI64
+__movmemSI64:
+	mov.l	@(60,r5),r0
+	mov.l	r0,@(60,r4)
+	.global	__movmemSI60
+	.global __movstrSI60
+	.set	__movstrSI60, __movmemSI60
+__movmemSI60:
+	mov.l	@(56,r5),r0
+	mov.l	r0,@(56,r4)
+	.global	__movmemSI56
+	.global __movstrSI56
+	.set	__movstrSI56, __movmemSI56
+__movmemSI56:
+	mov.l	@(52,r5),r0
+	mov.l	r0,@(52,r4)
+	.global	__movmemSI52
+	.global __movstrSI52
+	.set	__movstrSI52, __movmemSI52
+__movmemSI52:
+	mov.l	@(48,r5),r0
+	mov.l	r0,@(48,r4)
+	.global	__movmemSI48
+	.global	__movstrSI48
+	.set	__movstrSI48, __movmemSI48
+__movmemSI48:
+	mov.l	@(44,r5),r0
+	mov.l	r0,@(44,r4)
+	.global	__movmemSI44
+	.global	__movstrSI44
+	.set	__movstrSI44, __movmemSI44
+__movmemSI44:
+	mov.l	@(40,r5),r0
+	mov.l	r0,@(40,r4)
+	.global	__movmemSI40
+	.global __movstrSI40
+	.set	__movstrSI40, __movmemSI40
+__movmemSI40:
+	mov.l	@(36,r5),r0
+	mov.l	r0,@(36,r4)
+	.global	__movmemSI36
+	.global	__movstrSI36
+	.set	__movstrSI36, __movmemSI36
+__movmemSI36:
+	mov.l	@(32,r5),r0
+	mov.l	r0,@(32,r4)
+	.global	__movmemSI32
+	.global	__movstrSI32
+	.set	__movstrSI32, __movmemSI32
+__movmemSI32:
+	mov.l	@(28,r5),r0
+	mov.l	r0,@(28,r4)
+	.global	__movmemSI28
+	.global	__movstrSI28
+	.set	__movstrSI28, __movmemSI28
+__movmemSI28:
+	mov.l	@(24,r5),r0
+	mov.l	r0,@(24,r4)
+	.global	__movmemSI24
+	.global	__movstrSI24
+	.set	__movstrSI24, __movmemSI24
+__movmemSI24:
+	mov.l	@(20,r5),r0
+	mov.l	r0,@(20,r4)
+	.global	__movmemSI20
+	.global	__movstrSI20
+	.set	__movstrSI20, __movmemSI20
+__movmemSI20:
+	mov.l	@(16,r5),r0
+	mov.l	r0,@(16,r4)
+	.global	__movmemSI16
+	.global	__movstrSI16
+	.set	__movstrSI16, __movmemSI16
+__movmemSI16:
+	mov.l	@(12,r5),r0
+	mov.l	r0,@(12,r4)
+	.global	__movmemSI12
+	.global	__movstrSI12
+	.set	__movstrSI12, __movmemSI12
+__movmemSI12:
+	mov.l	@(8,r5),r0
+	mov.l	r0,@(8,r4)
+	.global	__movmemSI8
+	.global	__movstrSI8
+	.set	__movstrSI8, __movmemSI8
+__movmemSI8:
+	mov.l	@(4,r5),r0
+	mov.l	r0,@(4,r4)
+	.global	__movmemSI4
+	.global	__movstrSI4
+	.set	__movstrSI4, __movmemSI4
+__movmemSI4:
+	mov.l	@(0,r5),r0
+	rts
+	mov.l	r0,@(0,r4)
+
+	.global	__movmem_i4_even
+	.global	__movstr_i4_even
+	.set	__movstr_i4_even, __movmem_i4_even
+
+	.global	__movmem_i4_odd
+	.global	__movstr_i4_odd
+	.set	__movstr_i4_odd, __movmem_i4_odd
+
+	.global	__movmemSI12_i4
+	.global	__movstrSI12_i4
+	.set	__movstrSI12_i4, __movmemSI12_i4
+
+	.p2align	5
+L_movmem_2mod4_end:
+	mov.l	r0,@(16,r4)
+	rts
+	mov.l	r1,@(20,r4)
+
+	.p2align	2
+
+__movmem_i4_even:
+	mov.l	@r5+,r0
+	bra	L_movmem_start_even
+	mov.l	@r5+,r1
+
+__movmem_i4_odd:
+	mov.l	@r5+,r1
+	add	#-4,r4
+	mov.l	@r5+,r2
+	mov.l	@r5+,r3
+	mov.l	r1,@(4,r4)
+	mov.l	r2,@(8,r4)
+
+L_movmem_loop:
+	mov.l	r3,@(12,r4)
+	dt	r6
+	mov.l	@r5+,r0
+	bt/s	L_movmem_2mod4_end
+	mov.l	@r5+,r1
+	add	#16,r4
+L_movmem_start_even:
+	mov.l	@r5+,r2
+	mov.l	@r5+,r3
+	mov.l	r0,@r4
+	dt	r6
+	mov.l	r1,@(4,r4)
+	bf/s	L_movmem_loop
+	mov.l	r2,@(8,r4)
+	rts
+	mov.l	r3,@(12,r4)
+
+	.p2align	4
+__movmemSI12_i4:
+	mov.l	@r5,r0
+	mov.l	@(4,r5),r1
+	mov.l	@(8,r5),r2
+	mov.l	r0,@r4
+	mov.l	r1,@(4,r4)
+	rts
+	mov.l	r2,@(8,r4)