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, 262 insertions, 0 deletions

diff --git a/tools/bootconfig/scripts/ftrace2bconf.sh b/tools/bootconfig/scripts/ftrace2bconf.sh
new file mode 100755
index 000000000..8eed445c2
--- /dev/null
+++ b/tools/bootconfig/scripts/ftrace2bconf.sh
@@ -0,0 +1,262 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0-only
+
+usage() {
+	echo "Dump boot-time tracing bootconfig from ftrace"
+	echo "Usage: $0 [--debug] [ > BOOTCONFIG-FILE]"
+	exit 1
+}
+
+DEBUG=
+while [ x"$1" != x ]; do
+	case "$1" in
+	"--debug")
+		DEBUG=$1;;
+	-*)
+		usage
+		;;
+	esac
+	shift 1
+done
+
+if [ x"$DEBUG" != x ]; then
+	set -x
+fi
+
+TRACEFS=`grep -m 1 -w tracefs /proc/mounts | cut -f 2 -d " "`
+if [ -z "$TRACEFS" ]; then
+	if ! grep -wq debugfs /proc/mounts; then
+		echo "Error: No tracefs/debugfs was mounted."
+		exit 1
+	fi
+	TRACEFS=`grep -m 1 -w debugfs /proc/mounts | cut -f 2 -d " "`/tracing
+	if [ ! -d $TRACEFS ]; then
+		echo "Error: ftrace is not enabled on this kernel." 1>&2
+		exit 1
+	fi
+fi
+
+######## main #########
+
+set -e
+
+emit_kv() { # key =|+= value
+	echo "$@"
+}
+
+global_options() {
+	val=`cat $TRACEFS/max_graph_depth`
+	[ $val != 0 ] && emit_kv kernel.fgraph_max_depth = $val
+	if grep -qv "^#" $TRACEFS/set_graph_function $TRACEFS/set_graph_notrace ; then
+		cat 1>&2 << EOF
+# WARN: kernel.fgraph_filters and kernel.fgraph_notrace are not supported, since the wild card expression was expanded and lost from memory.
+EOF
+	fi
+}
+
+kprobe_event_options() {
+	cat $TRACEFS/kprobe_events | while read p args; do
+		case $p in
+		\#*)
+		continue;;
+		r*)
+		cat 1>&2 << EOF
+# WARN: A return probe found but it is not supported by bootconfig. Skip it.
+EOF
+		continue;;
+		esac
+		p=${p#*:}
+		event=${p#*/}
+		group=${p%/*}
+		if [ $group != "kprobes" ]; then
+			cat 1>&2 << EOF
+# WARN: kprobes group name $group is changed to "kprobes" for bootconfig.
+EOF
+		fi
+		emit_kv $PREFIX.event.kprobes.$event.probes += $args
+	done
+}
+
+synth_event_options() {
+	cat $TRACEFS/synthetic_events | while read event fields; do
+		emit_kv $PREFIX.event.synthetic.$event.fields = `echo $fields | sed "s/;/,/g"`
+	done
+}
+
+# Variables resolver
+DEFINED_VARS=
+UNRESOLVED_EVENTS=
+
+defined_vars() { # event-dir
+	grep "^hist" $1/trigger | grep -o ':[a-zA-Z0-9]*='
+}
+referred_vars() {
+	grep "^hist" $1/trigger | grep -o '$[a-zA-Z0-9]*'
+}
+
+event_is_enabled() { # enable-file
+	test -f $1 && grep -q "1" $1
+}
+
+per_event_options() { # event-dir
+	evdir=$1
+	# Check the special event which has no filter and no trigger
+	[ ! -f $evdir/filter ] && return
+
+	if grep -q "^hist:" $evdir/trigger; then
+		# hist action can refer the undefined variables
+		__vars=`defined_vars $evdir`
+		for v in `referred_vars $evdir`; do
+			if echo $DEFINED_VARS $__vars | grep -vqw ${v#$}; then
+				# $v is not defined yet, defer it
+				UNRESOLVED_EVENTS="$UNRESOLVED_EVENTS $evdir"
+				return;
+			fi
+		done
+		DEFINED_VARS="$DEFINED_VARS "`defined_vars $evdir`
+	fi
+	grep -v "^#" $evdir/trigger | while read action active; do
+		emit_kv $PREFIX.event.$group.$event.actions += \'$action\'
+	done
+
+	if [ $GROUP_ENABLED -eq 0 ] && event_is_enabled $evdir/enable; then
+		emit_kv $PREFIX.event.$group.$event.enable
+	fi
+	val=`cat $evdir/filter`
+	if [ "$val" != "none" ]; then
+		emit_kv $PREFIX.event.$group.$event.filter = "$val"
+	fi
+}
+
+retry_unresolved() {
+	unresolved=$UNRESOLVED_EVENTS
+	UNRESOLVED_EVENTS=
+	for evdir in $unresolved; do
+		event=${evdir##*/}
+		group=${evdir%/*}; group=${group##*/}
+		per_event_options $evdir
+	done
+}
+
+event_options() {
+	# PREFIX and INSTANCE must be set
+	if [ $PREFIX = "ftrace" ]; then
+		# define the dynamic events
+		kprobe_event_options
+		synth_event_options
+	fi
+	ALL_ENABLED=0
+	if event_is_enabled $INSTANCE/events/enable; then
+		emit_kv $PREFIX.event.enable
+		ALL_ENABLED=1
+	fi
+	for group in `ls $INSTANCE/events/` ; do
+		[ ! -d $INSTANCE/events/$group ] && continue
+		GROUP_ENABLED=$ALL_ENABLED
+		if [ $ALL_ENABLED -eq 0 ] && \
+		   event_is_enabled $INSTANCE/events/$group/enable ;then
+			emit_kv $PREFIX.event.$group.enable
+			GROUP_ENABLED=1
+		fi
+		for event in `ls $INSTANCE/events/$group/` ;do
+			[ ! -d $INSTANCE/events/$group/$event ] && continue
+			per_event_options $INSTANCE/events/$group/$event
+		done
+	done
+	retry=0
+	while [ $retry -lt 3 ]; do
+		retry_unresolved
+		retry=$((retry + 1))
+	done
+	if [ "$UNRESOLVED_EVENTS" ]; then
+		cat 1>&2 << EOF
+! ERROR: hist triggers in $UNRESOLVED_EVENTS use some undefined variables.
+EOF
+	fi
+}
+
+is_default_trace_option() { # option
+grep -qw $1 << EOF
+print-parent
+nosym-offset
+nosym-addr
+noverbose
+noraw
+nohex
+nobin
+noblock
+trace_printk
+annotate
+nouserstacktrace
+nosym-userobj
+noprintk-msg-only
+context-info
+nolatency-format
+record-cmd
+norecord-tgid
+overwrite
+nodisable_on_free
+irq-info
+markers
+noevent-fork
+nopause-on-trace
+function-trace
+nofunction-fork
+nodisplay-graph
+nostacktrace
+notest_nop_accept
+notest_nop_refuse
+EOF
+}
+
+instance_options() { # [instance-name]
+	if [ $# -eq 0 ]; then
+		PREFIX="ftrace"
+		INSTANCE=$TRACEFS
+	else
+		PREFIX="ftrace.instance.$1"
+		INSTANCE=$TRACEFS/instances/$1
+	fi
+	val=
+	for i in `cat $INSTANCE/trace_options`; do
+		is_default_trace_option $i && continue
+		val="$val, $i"
+	done
+	[ "$val" ] && emit_kv $PREFIX.options = "${val#,}"
+	val="local"
+	for i in `cat $INSTANCE/trace_clock` ; do
+		[ "${i#*]}" ] && continue
+		i=${i%]}; val=${i#[}
+	done
+	[ $val != "local" ] && emit_kv $PREFIX.trace_clock = $val
+	val=`cat $INSTANCE/buffer_size_kb`
+	if echo $val | grep -vq "expanded" ; then
+		emit_kv $PREFIX.buffer_size = $val"KB"
+	fi
+	if grep -q "is allocated" $INSTANCE/snapshot ; then
+		emit_kv $PREFIX.alloc_snapshot
+	fi
+	val=`cat $INSTANCE/tracing_cpumask`
+	if [ `echo $val | sed -e s/f//g`x != x ]; then
+		emit_kv $PREFIX.cpumask = $val
+	fi
+	val=`cat $INSTANCE/tracing_on`
+	if [ "$val" = "0" ]; then
+		emit_kv $PREFIX.tracing_on = 0
+	fi
+
+	val=`cat $INSTANCE/current_tracer`
+	[ $val != nop ] && emit_kv $PREFIX.tracer = $val
+	if grep -qv "^#" $INSTANCE/set_ftrace_filter $INSTANCE/set_ftrace_notrace; then
+		cat 1>&2 << EOF
+# WARN: kernel.ftrace.filters and kernel.ftrace.notrace are not supported, since the wild card expression was expanded and lost from memory.
+EOF
+	fi
+	event_options
+}
+
+global_options
+instance_options
+for i in `ls $TRACEFS/instances` ; do
+	instance_options $i
+done