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, 96 insertions, 0 deletions

diff --git a/arch/alpha/lib/ev67-strncat.S b/arch/alpha/lib/ev67-strncat.S
new file mode 100644
index 000000000..8f313233e
--- /dev/null
+++ b/arch/alpha/lib/ev67-strncat.S
@@ -0,0 +1,96 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * arch/alpha/lib/ev67-strncat.S
+ * 21264 version contributed by Rick Gorton <rick.gorton@api-networks.com>
+ *
+ * Append no more than COUNT characters from the null-terminated string SRC
+ * to the null-terminated string DST.  Always null-terminate the new DST.
+ *
+ * This differs slightly from the semantics in libc in that we never write
+ * past count, whereas libc may write to count+1.  This follows the generic
+ * implementation in lib/string.c and is, IMHO, more sensible.
+ *
+ * Much of the information about 21264 scheduling/coding comes from:
+ *	Compiler Writer's Guide for the Alpha 21264
+ *	abbreviated as 'CWG' in other comments here
+ *	ftp.digital.com/pub/Digital/info/semiconductor/literature/dsc-library.html
+ * Scheduling notation:
+ *	E	- either cluster
+ *	U	- upper subcluster; U0 - subcluster U0; U1 - subcluster U1
+ *	L	- lower subcluster; L0 - subcluster L0; L1 - subcluster L1
+ * Try not to change the actual algorithm if possible for consistency.
+ */
+
+#include <linux/export.h>
+	.text
+
+	.align 4
+	.globl strncat
+	.ent strncat
+strncat:
+	.frame $30, 0, $26
+	.prologue 0
+
+	mov	$16, $0		# set up return value
+	beq	$18, $zerocount	# U :
+	/* Find the end of the string.  */
+	ldq_u   $1, 0($16)	# L : load first quadword ($16 may be misaligned)
+	lda     $2, -1($31)	# E :
+
+	insqh   $2, $0, $2	# U :
+	andnot  $16, 7, $16	# E :
+	nop			# E :
+	or      $2, $1, $1	# E :
+
+	nop			# E :
+	nop			# E :
+	cmpbge  $31, $1, $2	# E : bits set iff byte == 0
+	bne     $2, $found	# U :
+
+$loop:	ldq     $1, 8($16)	# L :
+	addq    $16, 8, $16	# E :
+	cmpbge  $31, $1, $2	# E :
+	beq     $2, $loop	# U :
+
+$found:	cttz	$2, $3		# U0 :
+	addq	$16, $3, $16	# E :
+	nop			# E :
+	bsr	$23, __stxncpy	# L0 :/* Now do the append.  */
+
+	/* Worry about the null termination.  */
+
+	zapnot	$1, $27, $2	# U : was last byte a null?
+	cmplt	$27, $24, $5	# E : did we fill the buffer completely?
+	bne	$2, 0f		# U :
+	ret			# L0 :
+
+0:	or	$5, $18, $2	# E :
+	nop
+	bne	$2, 2f		# U :
+	and	$24, 0x80, $3	# E : no zero next byte
+
+	nop			# E :
+	bne	$3, 1f		# U :
+	/* Here there are bytes left in the current word.  Clear one.  */
+	addq	$24, $24, $24	# E : end-of-count bit <<= 1
+	nop			# E :
+
+2:	zap	$1, $24, $1	# U :
+	nop			# E :
+	stq_u	$1, 0($16)	# L :
+	ret			# L0 :
+
+1:	/* Here we must clear the first byte of the next DST word */
+	stb	$31, 8($16)	# L :
+	nop			# E :
+	nop			# E :
+	ret			# L0 :
+
+$zerocount:
+	nop			# E :
+	nop			# E :
+	nop			# E :
+	ret			# L0 :
+
+	.end strncat
+	EXPORT_SYMBOL(strncat)