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

10 files changed, 390 insertions, 0 deletions

diff --git a/arch/powerpc/tools/.gitignore b/arch/powerpc/tools/.gitignore
new file mode 100644
index 000000000..ec380a14a
--- /dev/null
+++ b/arch/powerpc/tools/.gitignore
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+/vmlinux.arch.S
diff --git a/arch/powerpc/tools/Makefile b/arch/powerpc/tools/Makefile
new file mode 100644
index 000000000..e1f7afcd9
--- /dev/null
+++ b/arch/powerpc/tools/Makefile
@@ -0,0 +1,10 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+quiet_cmd_gen_ftrace_ool_stubs = GEN     $@
+	cmd_gen_ftrace_ool_stubs = $< "$(CONFIG_PPC_FTRACE_OUT_OF_LINE_NUM_RESERVE)" "$(CONFIG_64BIT)" \
+				   "$(OBJDUMP)" vmlinux.o $@
+
+$(obj)/vmlinux.arch.S: $(src)/ftrace-gen-ool-stubs.sh vmlinux.o FORCE
+	$(call if_changed,gen_ftrace_ool_stubs)
+
+targets += vmlinux.arch.S
diff --git a/arch/powerpc/tools/check-fpatchable-function-entry.sh b/arch/powerpc/tools/check-fpatchable-function-entry.sh
new file mode 100755
index 000000000..baed467a0
--- /dev/null
+++ b/arch/powerpc/tools/check-fpatchable-function-entry.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+set -e
+
+# To debug, uncomment the following line
+# set -x
+
+# Output from -fpatchable-function-entry can only vary on ppc64 elfv2, so this
+# should not be invoked for other targets. Therefore we can pass in -m64 and
+# -mabi explicitly, to take care of toolchains defaulting to other targets.
+
+# Test whether the compile option -fpatchable-function-entry exists and
+# generates appropriate code
+echo "int func() { return 0; }" | \
+    $* -m64 -mabi=elfv2 -S -x c -O2 -fpatchable-function-entry=2 - -o - 2> /dev/null | \
+    grep -q "__patchable_function_entries"
+
+# Test whether nops are generated after the local entry point
+echo "int x; int func() { return x; }" | \
+    $* -m64 -mabi=elfv2 -S -x c -O2 -fpatchable-function-entry=2 - -o - 2> /dev/null | \
+    awk 'BEGIN { RS = ";" } /\.localentry.*nop.*\n[[:space:]]*nop/ { print $0 }' | \
+    grep -q "func:"
+
+exit 0
diff --git a/arch/powerpc/tools/checkpatch.sh b/arch/powerpc/tools/checkpatch.sh
new file mode 100755
index 000000000..91c04802e
--- /dev/null
+++ b/arch/powerpc/tools/checkpatch.sh
@@ -0,0 +1,22 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright 2018, Michael Ellerman, IBM Corporation.
+#
+# Wrapper around checkpatch that uses our preferred settings
+
+script_base=$(realpath $(dirname $0))
+
+exec $script_base/../../../scripts/checkpatch.pl \
+	--subjective \
+	--no-summary \
+	--show-types \
+	--ignore ARCH_INCLUDE_LINUX \
+	--ignore BIT_MACRO \
+	--ignore COMPARISON_TO_NULL \
+	--ignore EMAIL_SUBJECT \
+	--ignore FILE_PATH_CHANGES \
+	--ignore GLOBAL_INITIALISERS \
+	--ignore LINE_SPACING \
+	--ignore MULTIPLE_ASSIGNMENTS \
+	--ignore DT_SPLIT_BINDING_PATCH \
+	$@
diff --git a/arch/powerpc/tools/ftrace-gen-ool-stubs.sh b/arch/powerpc/tools/ftrace-gen-ool-stubs.sh
new file mode 100755
index 000000000..9218d43ae
--- /dev/null
+++ b/arch/powerpc/tools/ftrace-gen-ool-stubs.sh
@@ -0,0 +1,52 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+# Error out on error
+set -e
+
+num_ool_stubs_text_builtin="$1"
+is_64bit="$2"
+objdump="$3"
+vmlinux_o="$4"
+arch_vmlinux_S="$5"
+
+RELOCATION=R_PPC64_ADDR64
+if [ -z "$is_64bit" ]; then
+	RELOCATION=R_PPC_ADDR32
+fi
+
+num_ool_stubs_total=$($objdump -r -j __patchable_function_entries -d "$vmlinux_o" |
+		      grep -c "$RELOCATION")
+num_ool_stubs_inittext=$($objdump -r -j __patchable_function_entries -d "$vmlinux_o" |
+			 grep -e ".init.text" -e ".text.startup" | grep -c "$RELOCATION")
+num_ool_stubs_text=$((num_ool_stubs_total - num_ool_stubs_inittext))
+
+if [ "$num_ool_stubs_text" -gt "$num_ool_stubs_text_builtin" ]; then
+	num_ool_stubs_text_end=$((num_ool_stubs_text - num_ool_stubs_text_builtin))
+else
+	num_ool_stubs_text_end=0
+fi
+
+cat > "$arch_vmlinux_S" <<EOF
+#include <asm/asm-offsets.h>
+#include <asm/ppc_asm.h>
+#include <linux/linkage.h>
+
+.pushsection .tramp.ftrace.text,"aw"
+SYM_DATA(ftrace_ool_stub_text_end_count, .long $num_ool_stubs_text_end)
+
+SYM_START(ftrace_ool_stub_text_end, SYM_L_GLOBAL, .balign SZL)
+#if $num_ool_stubs_text_end
+	.space $num_ool_stubs_text_end * FTRACE_OOL_STUB_SIZE
+#endif
+SYM_CODE_END(ftrace_ool_stub_text_end)
+.popsection
+
+.pushsection .tramp.ftrace.init,"aw"
+SYM_DATA(ftrace_ool_stub_inittext_count, .long $num_ool_stubs_inittext)
+
+SYM_START(ftrace_ool_stub_inittext, SYM_L_GLOBAL, .balign SZL)
+	.space $num_ool_stubs_inittext * FTRACE_OOL_STUB_SIZE
+SYM_CODE_END(ftrace_ool_stub_inittext)
+.popsection
+EOF
diff --git a/arch/powerpc/tools/ftrace_check.sh b/arch/powerpc/tools/ftrace_check.sh
new file mode 100755
index 000000000..405e7e306
--- /dev/null
+++ b/arch/powerpc/tools/ftrace_check.sh
@@ -0,0 +1,50 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0-or-later
+#
+# This script checks vmlinux to ensure that all functions can call ftrace_caller() either directly,
+# or through the stub, ftrace_tramp_text, at the end of kernel text.
+
+# Error out if any command fails
+set -e
+
+# Allow for verbose output
+if [ "$V" = "1" ]; then
+	set -x
+fi
+
+if [ $# -lt 2 ]; then
+	echo "$0 [path to nm] [path to vmlinux]" 1>&2
+	exit 1
+fi
+
+# Have Kbuild supply the path to nm so we handle cross compilation.
+nm="$1"
+vmlinux="$2"
+
+stext_addr=$($nm "$vmlinux" | grep -e " [TA] _stext$" | \
+	cut -d' ' -f1 | tr '[:lower:]' '[:upper:]')
+ftrace_caller_addr=$($nm "$vmlinux" | grep -e " T ftrace_caller$" | \
+	cut -d' ' -f1 | tr '[:lower:]' '[:upper:]')
+ftrace_tramp_addr=$($nm "$vmlinux" | grep -e " T ftrace_tramp_text$" | \
+	cut -d' ' -f1 | tr '[:lower:]' '[:upper:]')
+
+ftrace_caller_offset=$(echo "ibase=16;$ftrace_caller_addr - $stext_addr" | bc)
+ftrace_tramp_offset=$(echo "ibase=16;$ftrace_tramp_addr - $ftrace_caller_addr" | bc)
+sz_32m=$(printf "%d" 0x2000000)
+sz_64m=$(printf "%d" 0x4000000)
+
+# ftrace_caller - _stext < 32M
+if [ "$ftrace_caller_offset" -ge "$sz_32m" ]; then
+	echo "ERROR: ftrace_caller (0x$ftrace_caller_addr) is beyond 32MiB of _stext" 1>&2
+	echo "ERROR: consider disabling CONFIG_FUNCTION_TRACER, or reducing the size \
+		of kernel text" 1>&2
+	exit 1
+fi
+
+# ftrace_tramp_text - ftrace_caller < 64M
+if [ "$ftrace_tramp_offset" -ge "$sz_64m" ]; then
+	echo "ERROR: kernel text extends beyond 64MiB from ftrace_caller" 1>&2
+	echo "ERROR: consider disabling CONFIG_FUNCTION_TRACER, or reducing the size \
+		of kernel text" 1>&2
+	exit 1
+fi
diff --git a/arch/powerpc/tools/gcc-check-mprofile-kernel.sh b/arch/powerpc/tools/gcc-check-mprofile-kernel.sh
new file mode 100755
index 000000000..6193b0ed0
--- /dev/null
+++ b/arch/powerpc/tools/gcc-check-mprofile-kernel.sh
@@ -0,0 +1,26 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+set -e
+
+# To debug, uncomment the following line
+# set -x
+
+# -mprofile-kernel is only supported on 64-bit with ELFv2, so this should not
+# be invoked for other targets. Therefore we can pass in -m64 and -mabi
+# explicitly, to take care of toolchains defaulting to other targets.
+
+# Test whether the compile option -mprofile-kernel exists and generates
+# profiling code (ie. a call to _mcount()).
+echo "int func() { return 0; }" | \
+    $* -m64 -mabi=elfv2 -S -x c -O2 -p -mprofile-kernel - -o - \
+    2> /dev/null | grep -q "_mcount"
+
+# Test whether the notrace attribute correctly suppresses calls to _mcount().
+
+echo -e "#include <linux/compiler.h>\nnotrace int func() { return 0; }" | \
+    $* -m64 -mabi=elfv2 -S -x c -O2 -p -mprofile-kernel - -o - \
+    2> /dev/null | grep -q "_mcount" && \
+    exit 1
+
+exit 0
diff --git a/arch/powerpc/tools/head_check.sh b/arch/powerpc/tools/head_check.sh
new file mode 100644
index 000000000..a9cd06958
--- /dev/null
+++ b/arch/powerpc/tools/head_check.sh
@@ -0,0 +1,81 @@
+#!/bin/sh
+# Copyright © 2016 IBM Corporation
+
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version
+# 2 of the License, or (at your option) any later version.
+
+# This script checks the head of a vmlinux for linker stubs that
+# break our placement of fixed-location code for 64-bit.
+
+# based on relocs_check.pl
+# Copyright © 2009 IBM Corporation
+
+# NOTE!
+#
+# If the build dies here, it's likely code in head_64.S/exception-64*.S or
+# nearby, is branching to labels it can't reach directly, which results in the
+# linker inserting branch stubs. This can move code around in ways that break
+# the fixed section calculations (head-64.h). To debug this, disassemble the
+# vmlinux and look for branch stubs (long_branch, plt_branch, etc.) in the
+# fixed section region (0 - 0x8000ish). Check what code is calling those stubs,
+# and perhaps change so a direct branch can reach.
+#
+# A ".linker_stub_catch" section is used to catch some stubs generated by
+# early .text code, which tend to get placed at the start of the section.
+# If there are too many such stubs, they can overflow this section. Expanding
+# it may help (or reducing the number of stub branches).
+#
+# Linker stubs use the TOC pointer, so even if fixed section code could
+# tolerate them being inserted into head code, they can't be allowed in low
+# level entry code (boot, interrupt vectors, etc) until r2 is set up. This
+# could cause the kernel to die in early boot.
+
+# Allow for verbose output
+if [ "$V" = "1" ]; then
+	set -x
+fi
+
+if [ $# -lt 2 ]; then
+	echo "$0 [path to nm] [path to vmlinux]" 1>&2
+	exit 1
+fi
+
+# Have Kbuild supply the path to nm so we handle cross compilation.
+nm="$1"
+vmlinux="$2"
+
+# gcc-4.6-era toolchain make _stext an A (absolute) symbol rather than T
+$nm "$vmlinux" | grep -e " [TA] _stext$" -e " t start_first_256B$" -e " a text_start$" -e " t start_text$" > .tmp_symbols.txt
+
+
+vma=$(grep -e " [TA] _stext$" .tmp_symbols.txt | cut -d' ' -f1)
+
+expected_start_head_addr="$vma"
+
+start_head_addr=$(grep " t start_first_256B$" .tmp_symbols.txt | cut -d' ' -f1)
+
+if [ "$start_head_addr" != "$expected_start_head_addr" ]; then
+	echo "ERROR: head code starts at $start_head_addr, should be $expected_start_head_addr" 1>&2
+	echo "ERROR: try to enable LD_HEAD_STUB_CATCH config option" 1>&2
+	echo "ERROR: see comments in arch/powerpc/tools/head_check.sh" 1>&2
+
+	exit 1
+fi
+
+top_vma=$(echo "$vma" | cut -d'0' -f1)
+
+expected_start_text_addr=$(grep " a text_start$" .tmp_symbols.txt | cut -d' ' -f1 | sed "s/^0/$top_vma/")
+
+start_text_addr=$(grep " t start_text$" .tmp_symbols.txt | cut -d' ' -f1)
+
+if [ "$start_text_addr" != "$expected_start_text_addr" ]; then
+	echo "ERROR: start_text address is $start_text_addr, should be $expected_start_text_addr" 1>&2
+	echo "ERROR: try to enable LD_HEAD_STUB_CATCH config option" 1>&2
+	echo "ERROR: see comments in arch/powerpc/tools/head_check.sh" 1>&2
+
+	exit 1
+fi
+
+rm -f .tmp_symbols.txt
diff --git a/arch/powerpc/tools/relocs_check.sh b/arch/powerpc/tools/relocs_check.sh
new file mode 100755
index 000000000..6b350e750
--- /dev/null
+++ b/arch/powerpc/tools/relocs_check.sh
@@ -0,0 +1,43 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+# Copyright © 2015 IBM Corporation
+
+
+# This script checks the relocations of a vmlinux for "suspicious"
+# relocations.
+
+# based on relocs_check.pl
+# Copyright © 2009 IBM Corporation
+
+if [ $# -lt 3 ]; then
+	echo "$0 [path to objdump] [path to nm] [path to vmlinux]" 1>&2
+	exit 1
+fi
+
+bad_relocs=$(
+${srctree}/scripts/relocs_check.sh "$@" |
+	# These relocations are okay
+	# On PPC64:
+	#	R_PPC64_RELATIVE, R_PPC64_NONE
+	# On PPC:
+	#	R_PPC_RELATIVE, R_PPC_ADDR16_HI,
+	#	R_PPC_ADDR16_HA,R_PPC_ADDR16_LO,
+	#	R_PPC_NONE
+	grep -F -w -v 'R_PPC64_RELATIVE
+R_PPC64_NONE
+R_PPC64_UADDR64
+R_PPC_ADDR16_LO
+R_PPC_ADDR16_HI
+R_PPC_ADDR16_HA
+R_PPC_RELATIVE
+R_PPC_NONE'
+)
+
+if [ -z "$bad_relocs" ]; then
+	exit 0
+fi
+
+num_bad=$(echo "$bad_relocs" | wc -l)
+echo "WARNING: $num_bad bad relocations"
+echo "$bad_relocs"
diff --git a/arch/powerpc/tools/unrel_branch_check.sh b/arch/powerpc/tools/unrel_branch_check.sh
new file mode 100755
index 000000000..8301efee1
--- /dev/null
+++ b/arch/powerpc/tools/unrel_branch_check.sh
@@ -0,0 +1,79 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright © 2016,2020 IBM Corporation
+#
+# This script checks the unrelocated code of a vmlinux for "suspicious"
+# branches to relocated code (head_64.S code).
+
+# Have Kbuild supply the path to objdump and nm so we handle cross compilation.
+objdump="$1"
+nm="$2"
+vmlinux="$3"
+
+kstart=0xc000000000000000
+
+end_intr=0x$($nm -p "$vmlinux" |
+	sed -E -n '/\s+[[:alpha:]]\s+__end_interrupts\s*$/{s///p;q}')
+if [ "$end_intr" = "0x" ]; then
+	exit 0
+fi
+
+# we know that there is a correct branch to
+# __start_initialization_multiplatform, so find its address
+# so we can exclude it.
+sim=0x$($nm -p "$vmlinux" |
+	sed -E -n '/\s+[[:alpha:]]\s+__start_initialization_multiplatform\s*$/{s///p;q}')
+
+$objdump -D --no-show-raw-insn --start-address="$kstart" --stop-address="$end_intr" "$vmlinux" |
+sed -E -n '
+# match lines that start with a kernel address
+/^c[0-9a-f]*:\s*b/ {
+	# drop branches via ctr or lr
+	/\<b.?.?(ct|l)r/d
+	# cope with some differences between Clang and GNU objdumps
+	s/\<bt.?\s*[[:digit:]]+,/beq/
+	s/\<bf.?\s*[[:digit:]]+,/bne/
+	# tidy up
+	s/\s0x/ /
+	s/://
+	# format for the loop below
+	s/^(\S+)\s+(\S+)\s+(\S+)\s*(\S*).*$/\1:\2:\3:\4/
+	# strip out condition registers
+	s/:cr[0-7],/:/
+	p
+}' | {
+
+all_good=true
+while IFS=: read -r from branch to sym; do
+	case "$to" in
+	c*)	to="0x$to"
+		;;
+	.+*)
+		to=${to#.+}
+		if [ "$branch" = 'b' ]; then
+			if (( to >= 0x2000000 )); then
+				to=$(( to - 0x4000000 ))
+			fi
+		elif (( to >= 0x8000 )); then
+			to=$(( to - 0x10000 ))
+		fi
+		printf -v to '0x%x' $(( "0x$from" + to ))
+		;;
+	*)	printf 'Unkown branch format\n'
+		;;
+	esac
+	if [ "$to" = "$sim" ]; then
+		continue
+	fi
+	if (( to > end_intr )); then
+		if $all_good; then
+			printf '%s\n' 'WARNING: Unrelocated relative branches'
+			all_good=false
+		fi
+		printf '%s %s-> %s %s\n' "$from" "$branch" "$to" "$sym"
+	fi
+done
+
+$all_good
+
+}