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

5 files changed, 185 insertions, 0 deletions

diff --git a/arch/arm/boot/bootp/Makefile b/arch/arm/boot/bootp/Makefile
new file mode 100644
index 000000000..f3443f7d7
--- /dev/null
+++ b/arch/arm/boot/bootp/Makefile
@@ -0,0 +1,59 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# linux/arch/arm/boot/bootp/Makefile
+#
+# This file is included by the global makefile so that you can add your own
+# architecture-specific flags and dependencies.
+#
+
+ifdef PHYS_OFFSET
+add_hex = $(shell printf 0x%x $$(( $(1) + $(2) )) )
+
+# If PHYS_OFFSET is set, INITRD_PHYS and PARAMS_PHYS can be derived,
+# otherwise they must be passed on the command line.
+#
+# Note: the following conditions must always be true:
+#   PARAMS_PHYS must be within 4MB of ZRELADDR
+#   INITRD_PHYS must be in RAM
+
+PARAMS_PHYS := $(call add_hex, $(PHYS_OFFSET), 0x100)
+
+# guess an initrd location if possible
+initrd_offset-$(CONFIG_ARCH_FOOTBRIDGE)	+= 0x00800000
+initrd_offset-$(CONFIG_ARCH_SA1100)	+= 0x00800000
+initrd_offset-$(CONFIG_ARCH_RPC)	+= 0x08000000
+INITRD_OFFSET := $(initrd_offset-y)
+ifdef INITRD_OFFSET
+INITRD_PHYS := $(call add_hex, $(PHYS_OFFSET), $(INITRD_OFFSET))
+endif
+
+endif
+
+PHONY += initrd
+initrd:
+	@test "$(PARAMS_PHYS)" != "" || \
+	(echo bootpImage: You must specify PHYS_OFFSET of PARAMS_PHYS ; exit -1)
+	@test "$(INITRD_PHYS)" != "" || \
+	(echo bootpImage: You must specify INITRD_OFFSET or INITRD_PHYS ; exit -1)
+	@test "$(INITRD)" != "" || \
+	(echo bootpImage: You must specify INITRD; exit -1)
+
+LDFLAGS_bootp	:= --no-undefined -X \
+		 --defsym initrd_phys=$(INITRD_PHYS) \
+		 --defsym params_phys=$(PARAMS_PHYS) -T
+AFLAGS_initrd.o :=-DINITRD=\"$(INITRD)\"
+
+targets	:= bootp init.o kernel.o initrd.o
+
+# Note that bootp.lds picks up kernel.o and initrd.o
+$(obj)/bootp:	$(src)/bootp.lds $(addprefix $(obj)/,init.o kernel.o initrd.o) FORCE
+	$(call if_changed,ld)
+
+# kernel.o and initrd.o includes a binary image using
+# .incbin, a dependency which is not tracked automatically
+
+$(obj)/kernel.o: arch/arm/boot/zImage FORCE
+
+$(obj)/initrd.o: initrd $(INITRD) FORCE
+
+PHONY += $(INITRD)
diff --git a/arch/arm/boot/bootp/bootp.lds b/arch/arm/boot/bootp/bootp.lds
new file mode 100644
index 000000000..160128186
--- /dev/null
+++ b/arch/arm/boot/bootp/bootp.lds
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ *  linux/arch/arm/boot/bootp/bootp.lds
+ *
+ *  Copyright (C) 2000-2002 Russell King
+ */
+OUTPUT_ARCH(arm)
+ENTRY(_start)
+SECTIONS
+{
+  . = 0;
+  .text : {
+   _stext = .;
+   *(.start)
+   *(.text)
+   initrd_size = initrd_end - initrd_start;
+   _etext = .;
+  }
+
+  .stab 0 : { *(.stab) }
+  .stabstr 0 : { *(.stabstr) }
+  .stab.excl 0 : { *(.stab.excl) }
+  .stab.exclstr 0 : { *(.stab.exclstr) }
+  .stab.index 0 : { *(.stab.index) }
+  .stab.indexstr 0 : { *(.stab.indexstr) }
+  .comment 0 : { *(.comment) }
+}
diff --git a/arch/arm/boot/bootp/init.S b/arch/arm/boot/bootp/init.S
new file mode 100644
index 000000000..b562da2f7
--- /dev/null
+++ b/arch/arm/boot/bootp/init.S
@@ -0,0 +1,85 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ *  linux/arch/arm/boot/bootp/init.S
+ *
+ *  Copyright (C) 2000-2003 Russell King.
+ *
+ *  "Header" file for splitting kernel + initrd.  Note that we pass
+ *  r0 through to r3 straight through.
+ *
+ *  This demonstrates how to append code to the start of the kernel
+ *  zImage, and boot the kernel without copying it around.  This
+ *  example would be simpler; if we didn't have an object of unknown
+ *  size immediately following the kernel, we could build this into
+ *  a binary blob, and concatenate the zImage using the cat command.
+ */
+		.section .start, "ax"
+		.type	_start, #function
+		.globl	_start
+
+_start:		add	lr, pc, #-0x8		@ lr = current load addr
+		adr	r13, data
+		ldmia	r13!, {r4-r6}		@ r5 = dest, r6 = length
+		add	r4, r4, lr		@ r4 = initrd_start + load addr
+		bl	move			@ move the initrd
+
+/*
+ * Setup the initrd parameters to pass to the kernel.  This can only be
+ * passed in via the tagged list.
+ */
+		ldmia	r13, {r5-r9}		@ get size and addr of initrd
+						@ r5 = ATAG_CORE
+						@ r6 = ATAG_INITRD2
+						@ r7 = initrd start
+						@ r8 = initrd end
+						@ r9 = param_struct address
+
+		ldr	r10, [r9, #4]		@ get first tag
+		teq	r10, r5			@ is it ATAG_CORE?
+/*
+ * If we didn't find a valid tag list, create a dummy ATAG_CORE entry.
+ */
+		movne	r10, #0			@ terminator
+		movne	r4, #2			@ Size of this entry (2 words)
+		stmiane	r9, {r4, r5, r10}	@ Size, ATAG_CORE, terminator
+
+/*
+ * find the end of the tag list, and then add an INITRD tag on the end.
+ * If there is already an INITRD tag, then we ignore it; the last INITRD
+ * tag takes precedence.
+ */
+taglist:	ldr	r10, [r9, #0]		@ tag length
+		teq	r10, #0			@ last tag (zero length)?
+		addne	r9, r9, r10, lsl #2
+		bne	taglist
+
+		mov	r5, #4			@ Size of initrd tag (4 words)
+		stmia	r9, {r5, r6, r7, r8, r10}
+		b	kernel_start		@ call kernel
+
+/*
+ * Move the block of memory length r6 from address r4 to address r5
+ */
+move:		ldmia	r4!, {r7 - r10}		@ move 32-bytes at a time
+		stmia	r5!, {r7 - r10}
+		ldmia	r4!, {r7 - r10}
+		stmia	r5!, {r7 - r10}
+		subs	r6, r6, #8 * 4
+		bcs	move
+		mov	pc, lr
+
+		.size	_start, . - _start
+
+		.align
+
+		.type	data,#object
+data:		.word	initrd_start		@ source initrd address
+		.word	initrd_phys		@ destination initrd address
+		.word	initrd_size		@ initrd size
+
+		.word	0x54410001		@ r5 = ATAG_CORE
+		.word	0x54420005		@ r6 = ATAG_INITRD2
+		.word	initrd_phys		@ r7
+		.word	initrd_size		@ r8
+		.word	params_phys		@ r9
+		.size	data, . - data
diff --git a/arch/arm/boot/bootp/initrd.S b/arch/arm/boot/bootp/initrd.S
new file mode 100644
index 000000000..dd3d04971
--- /dev/null
+++ b/arch/arm/boot/bootp/initrd.S
@@ -0,0 +1,7 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+	.type	initrd_start,#object
+	.globl	initrd_start
+initrd_start:
+	.incbin	INITRD
+	.globl	initrd_end
+initrd_end:
diff --git a/arch/arm/boot/bootp/kernel.S b/arch/arm/boot/bootp/kernel.S
new file mode 100644
index 000000000..dc6236c17
--- /dev/null
+++ b/arch/arm/boot/bootp/kernel.S
@@ -0,0 +1,7 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+	.globl	kernel_start
+kernel_start:
+	.incbin	"arch/arm/boot/zImage"
+	.globl	kernel_end
+kernel_end:
+	.align	2