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, 287 insertions, 0 deletions

diff --git a/arch/sh/lib/mcount.S b/arch/sh/lib/mcount.S
new file mode 100644
index 000000000..c6ca90cc9
--- /dev/null
+++ b/arch/sh/lib/mcount.S
@@ -0,0 +1,287 @@
+/* SPDX-License-Identifier: GPL-2.0
+ *
+ * arch/sh/lib/mcount.S
+ *
+ *  Copyright (C) 2008, 2009  Paul Mundt
+ *  Copyright (C) 2008, 2009  Matt Fleming
+ */
+#include <asm/ftrace.h>
+#include <asm/thread_info.h>
+#include <asm/asm-offsets.h>
+
+#define MCOUNT_ENTER()		\
+	mov.l	r4, @-r15;	\
+	mov.l	r5, @-r15;	\
+	mov.l	r6, @-r15;	\
+	mov.l	r7, @-r15;	\
+	sts.l	pr, @-r15;	\
+				\
+	mov.l	@(20,r15),r4;	\
+	sts	pr, r5
+
+#define MCOUNT_LEAVE()		\
+	lds.l	@r15+, pr;	\
+	mov.l	@r15+, r7;	\
+	mov.l	@r15+, r6;	\
+	mov.l	@r15+, r5;	\
+	rts;			\
+	 mov.l	@r15+, r4
+
+#ifdef CONFIG_STACK_DEBUG
+/*
+ * Perform diagnostic checks on the state of the kernel stack.
+ *
+ * Check for stack overflow. If there is less than 1KB free
+ * then it has overflowed.
+ *
+ * Make sure the stack pointer contains a valid address. Valid
+ * addresses for kernel stacks are anywhere after the bss
+ * (after __bss_stop) and anywhere in init_thread_union (init_stack).
+ */
+#define STACK_CHECK()					\
+	mov	#(THREAD_SIZE >> 10), r0;		\
+	shll8	r0;					\
+	shll2	r0;					\
+							\
+	/* r1 = sp & (THREAD_SIZE - 1) */		\
+	mov	#-1, r1;				\
+	add	r0, r1;					\
+	and	r15, r1;				\
+							\
+	mov	#TI_SIZE, r3;				\
+	mov	#(STACK_WARN >> 8), r2;			\
+	shll8	r2;					\
+	add	r3, r2;					\
+							\
+	/* Is the stack overflowing? */			\
+	cmp/hi	r2, r1;					\
+	bf	stack_panic;				\
+							\
+	/* If sp > __bss_stop then we're OK. */		\
+	mov.l	.L_ebss, r1;				\
+	cmp/hi	r1, r15;				\
+	bt	1f;					\
+							\
+	/* If sp < init_stack, we're not OK. */		\
+	mov.l	.L_init_thread_union, r1;		\
+	cmp/hs	r1, r15;				\
+	bf	stack_panic;				\
+							\
+	/* If sp > init_stack && sp < __bss_stop, not OK. */	\
+	add	r0, r1;					\
+	cmp/hs	r1, r15;				\
+	bt	stack_panic;				\
+1:
+#else
+#define STACK_CHECK()
+#endif /* CONFIG_STACK_DEBUG */
+
+	.align 2
+	.globl	_mcount
+	.type	_mcount,@function
+	.globl	mcount
+	.type	mcount,@function
+_mcount:
+mcount:
+	STACK_CHECK()
+
+#ifndef CONFIG_FUNCTION_TRACER
+	rts
+	 nop
+#else
+	MCOUNT_ENTER()
+
+#ifdef CONFIG_DYNAMIC_FTRACE
+	.globl	mcount_call
+mcount_call:
+	mov.l	.Lftrace_stub, r6
+#else
+	mov.l	.Lftrace_trace_function, r6
+	mov.l	ftrace_stub, r7
+	cmp/eq	r6, r7
+	bt	skip_trace
+	mov.l	@r6, r6
+#endif
+
+	jsr	@r6
+	 nop
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	mov.l   .Lftrace_graph_return, r6
+	mov.l   .Lftrace_stub, r7
+	cmp/eq  r6, r7
+	bt      1f
+
+	mov.l   .Lftrace_graph_caller, r0
+	jmp     @r0
+	 nop
+
+1:
+	mov.l	.Lftrace_graph_entry, r6
+	mov.l	.Lftrace_graph_entry_stub, r7
+	cmp/eq	r6, r7
+	bt	skip_trace
+
+	mov.l   .Lftrace_graph_caller, r0
+	jmp	@r0
+	 nop
+
+	.align 2
+.Lftrace_graph_return:
+	.long   ftrace_graph_return
+.Lftrace_graph_entry:
+	.long   ftrace_graph_entry
+.Lftrace_graph_entry_stub:
+	.long   ftrace_graph_entry_stub
+.Lftrace_graph_caller:
+	.long   ftrace_graph_caller
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+
+	.globl skip_trace
+skip_trace:
+	MCOUNT_LEAVE()
+
+	.align 2
+.Lftrace_trace_function:
+	.long   ftrace_trace_function
+
+#ifdef CONFIG_DYNAMIC_FTRACE
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+/*
+ * NOTE: Do not move either ftrace_graph_call or ftrace_caller
+ * as this will affect the calculation of GRAPH_INSN_OFFSET.
+ */
+	.globl ftrace_graph_call
+ftrace_graph_call:
+	mov.l	.Lskip_trace, r0
+	jmp	@r0
+	 nop
+
+	.align 2
+.Lskip_trace:
+	.long	skip_trace
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+
+	.globl ftrace_caller
+ftrace_caller:
+	MCOUNT_ENTER()
+
+	.globl ftrace_call
+ftrace_call:
+	mov.l	.Lftrace_stub, r6
+	jsr	@r6
+	 nop
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	bra	ftrace_graph_call
+	 nop
+#else
+	MCOUNT_LEAVE()
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+#endif /* CONFIG_DYNAMIC_FTRACE */
+
+	.align 2
+
+/*
+ * NOTE: From here on the locations of the .Lftrace_stub label and
+ * ftrace_stub itself are fixed. Adding additional data here will skew
+ * the displacement for the memory table and break the block replacement.
+ * Place new labels either after the ftrace_stub body, or before
+ * ftrace_caller. You have been warned.
+ */
+.Lftrace_stub:
+	.long	ftrace_stub
+
+	.globl	ftrace_stub
+ftrace_stub:
+	rts
+	 nop
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	.globl	ftrace_graph_caller
+ftrace_graph_caller:
+	mov.l	2f, r1
+	jmp	@r1
+	 nop
+1:
+	/*
+	 * MCOUNT_ENTER() pushed 5 registers onto the stack, so
+	 * the stack address containing our return address is
+	 * r15 + 20.
+	 */
+	mov	#20, r0
+	add	r15, r0
+	mov	r0, r4
+
+	mov.l	.Lprepare_ftrace_return, r0
+	jsr	@r0
+	 nop
+
+	MCOUNT_LEAVE()
+
+	.align 2
+2:	.long	skip_trace
+.Lprepare_ftrace_return:
+	.long	prepare_ftrace_return
+
+	.globl	return_to_handler
+return_to_handler:
+	/*
+	 * Save the return values.
+	 */
+	mov.l	r0, @-r15
+	mov.l	r1, @-r15
+
+	mov	#0, r4
+
+	mov.l	.Lftrace_return_to_handler, r0
+	jsr	@r0
+	 nop
+
+	/*
+	 * The return value from ftrace_return_handler has the real
+	 * address that we should return to.
+	 */
+	lds	r0, pr
+	mov.l	@r15+, r1
+	rts
+	 mov.l	@r15+, r0
+
+
+	.align 2
+.Lftrace_return_to_handler:
+	.long	ftrace_return_to_handler
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+#endif /* CONFIG_FUNCTION_TRACER */
+
+#ifdef CONFIG_STACK_DEBUG
+	.globl	stack_panic
+stack_panic:
+	mov.l	.Ldump_stack, r0
+	jsr	@r0
+	 nop
+
+	mov.l	.Lpanic, r0
+	jsr	@r0
+	 mov.l	.Lpanic_s, r4
+
+	rts
+	 nop
+
+	.align 2
+.L_init_thread_union:
+	.long	init_thread_union
+.L_ebss:
+	.long	__bss_stop
+.Lpanic:
+	.long	panic
+.Lpanic_s:
+	.long	.Lpanic_str
+.Ldump_stack:
+	.long	dump_stack
+
+	.section	.rodata
+	.align 2
+.Lpanic_str:
+	.string "Stack error"
+#endif /* CONFIG_STACK_DEBUG */