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, 240 insertions, 0 deletions

diff --git a/scripts/config b/scripts/config
new file mode 100755
index 000000000..32428ea90
--- /dev/null
+++ b/scripts/config
@@ -0,0 +1,240 @@
+#!/usr/bin/env bash
+# SPDX-License-Identifier: GPL-2.0
+# Manipulate options in a .config file from the command line
+
+myname=${0##*/}
+
+# If no prefix forced, use the default CONFIG_
+CONFIG_="${CONFIG_-CONFIG_}"
+
+# We use an uncommon delimiter for sed substitutions
+SED_DELIM=$(echo -en "\001")
+
+usage() {
+	cat >&2 <<EOL
+Manipulate options in a .config file from the command line.
+Usage:
+$myname options command ...
+commands:
+	--enable|-e option   Enable option
+	--disable|-d option  Disable option
+	--module|-m option   Turn option into a module
+	--set-str option string
+	                     Set option to "string"
+	--set-val option value
+	                     Set option to value
+	--undefine|-u option Undefine option
+	--state|-s option    Print state of option (n,y,m,undef)
+
+	--enable-after|-E beforeopt option
+                             Enable option directly after other option
+	--disable-after|-D beforeopt option
+                             Disable option directly after other option
+	--module-after|-M beforeopt option
+                             Turn option into module directly after other option
+	--refresh            Refresh the config using old settings
+
+	commands can be repeated multiple times
+
+options:
+	--file config-file   .config file to change (default .config)
+	--keep-case|-k       Keep next symbols' case (don't upper-case it)
+
+$myname doesn't check the validity of the .config file. This is done at next
+make time.
+
+By default, $myname will upper-case the given symbol. Use --keep-case to keep
+the case of all following symbols unchanged.
+
+$myname uses 'CONFIG_' as the default symbol prefix. Set the environment
+variable CONFIG_ to the prefix to use. Eg.: CONFIG_="FOO_" $myname ...
+EOL
+	exit 1
+}
+
+checkarg() {
+	ARG="$1"
+	if [ "$ARG" = "" ] ; then
+		usage
+	fi
+	case "$ARG" in
+	${CONFIG_}*)
+		ARG="${ARG/${CONFIG_}/}"
+		;;
+	esac
+	if [ "$MUNGE_CASE" = "yes" ] ; then
+		ARG="`echo $ARG | tr a-z A-Z`"
+	fi
+}
+
+txt_append() {
+	local anchor="$1"
+	local insert="$2"
+	local infile="$3"
+
+	# sed append cmd: 'a\' + newline + text + newline
+	cmd="$(printf "a\\%b$insert" "\n")"
+
+	# We don't really need a backup file, but in-place editing with backup
+	# skipped is not portable due to different implementations parsing
+	# arguments in incompatible manners.
+	# Create a backup file anyway to ensure portability. The file will be
+	# deleted on exit.
+	sed -E -i.swp -e "/$anchor/$cmd" "$infile"
+	SED_EDITED=1
+}
+
+txt_subst() {
+	local before="$1"
+	local after="$2"
+	local infile="$3"
+
+	sed -E -i.swp -e "s$SED_DELIM$before$SED_DELIM$after$SED_DELIM" "$infile"
+	SED_EDITED=1
+}
+
+txt_delete() {
+	local text="$1"
+	local infile="$2"
+
+	sed -E -i.swp -e "/$text/d" "$infile"
+	SED_EDITED=1
+}
+
+set_var() {
+	local name=$1 new=$2 before=$3
+
+	name_re="^($name=.*|# $name is not set)"
+	before_re="^($before=|# $before is not set)"
+	if test -n "$before" && grep -Eq "$before_re" "$FN"; then
+		txt_append "$before_re" "$new" "$FN"
+	elif grep -Eq "$name_re" "$FN"; then
+		txt_subst "$name_re" "$new" "$FN"
+	else
+		echo "$new" >>"$FN"
+	fi
+}
+
+undef_var() {
+	local name=$1
+
+	txt_delete "^($name=|# $name is not set)" "$FN"
+}
+
+SED_EDITED=0
+on_exit() {
+	if [ "$SED_EDITED" -ge 1 ]; then
+		rm -f "$FN.swp"
+	fi
+}
+trap on_exit EXIT
+
+FN=.config
+CMDS=()
+while [[ $# -gt 0 ]]; do
+	if [ "$1" = "--file" ]; then
+		if [ "$2" = "" ]; then
+			usage
+		fi
+		FN="$2"
+		shift 2
+	else
+		CMDS+=("$1")
+		shift
+	fi
+done
+
+set -- "${CMDS[@]}"
+if [ "$1" = "" ] ; then
+	usage
+fi
+
+MUNGE_CASE=yes
+while [ "$1" != "" ] ; do
+	CMD="$1"
+	shift
+	case "$CMD" in
+	--keep-case|-k)
+		MUNGE_CASE=no
+		continue
+		;;
+	--refresh)
+		;;
+	--*-after|-E|-D|-M)
+		checkarg "$1"
+		A=$ARG
+		checkarg "$2"
+		B=$ARG
+		shift 2
+		;;
+	-*)
+		checkarg "$1"
+		shift
+		;;
+	esac
+	case "$CMD" in
+	--enable|-e)
+		set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=y"
+		;;
+
+	--disable|-d)
+		set_var "${CONFIG_}$ARG" "# ${CONFIG_}$ARG is not set"
+		;;
+
+	--module|-m)
+		set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=m"
+		;;
+
+	--set-str)
+		# sed swallows one level of escaping, so we need double-escaping
+		set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=\"${1//\"/\\\\\"}\""
+		shift
+		;;
+
+	--set-val)
+		set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=$1"
+		shift
+		;;
+	--undefine|-u)
+		undef_var "${CONFIG_}$ARG"
+		;;
+
+	--state|-s)
+		if grep -q "# ${CONFIG_}$ARG is not set" $FN ; then
+			echo n
+		else
+			V="$(grep "^${CONFIG_}$ARG=" $FN)"
+			if [ $? != 0 ] ; then
+				echo undef
+			else
+				V="${V/#${CONFIG_}$ARG=/}"
+				V="${V/#\"/}"
+				V="${V/%\"/}"
+				V="${V//\\\"/\"}"
+				echo "${V}"
+			fi
+		fi
+		;;
+
+	--enable-after|-E)
+		set_var "${CONFIG_}$B" "${CONFIG_}$B=y" "${CONFIG_}$A"
+		;;
+
+	--disable-after|-D)
+		set_var "${CONFIG_}$B" "# ${CONFIG_}$B is not set" "${CONFIG_}$A"
+		;;
+
+	--module-after|-M)
+		set_var "${CONFIG_}$B" "${CONFIG_}$B=m" "${CONFIG_}$A"
+		;;
+
+	--refresh)
+		yes "" | make oldconfig KCONFIG_CONFIG=$FN
+		;;
+
+	*)
+		echo "bad command: $CMD" >&2
+		usage
+		;;
+	esac
+done