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, 267 insertions, 0 deletions

diff --git a/drivers/media/rc/ir-imon-decoder.c b/drivers/media/rc/ir-imon-decoder.c
new file mode 100644
index 000000000..dc68f64e7
--- /dev/null
+++ b/drivers/media/rc/ir-imon-decoder.c
@@ -0,0 +1,267 @@
+// SPDX-License-Identifier: GPL-2.0+
+// ir-imon-decoder.c - handle iMon protocol
+//
+// Copyright (C) 2018 by Sean Young <sean@mess.org>
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include "rc-core-priv.h"
+
+#define IMON_UNIT		416 /* us */
+#define IMON_BITS		30
+#define IMON_CHKBITS		(BIT(30) | BIT(25) | BIT(24) | BIT(22) | \
+				 BIT(21) | BIT(20) | BIT(19) | BIT(18) | \
+				 BIT(17) | BIT(16) | BIT(14) | BIT(13) | \
+				 BIT(12) | BIT(11) | BIT(10) | BIT(9))
+
+/*
+ * This protocol has 30 bits. The format is one IMON_UNIT header pulse,
+ * followed by 30 bits. Each bit is one IMON_UNIT check field, and then
+ * one IMON_UNIT field with the actual bit (1=space, 0=pulse).
+ * The check field is always space for some bits, for others it is pulse if
+ * both the preceding and current bit are zero, else space. IMON_CHKBITS
+ * defines which bits are of type check.
+ *
+ * There is no way to distinguish an incomplete message from one where
+ * the lower bits are all set, iow. the last pulse is for the lowest
+ * bit which is 0.
+ */
+enum imon_state {
+	STATE_INACTIVE,
+	STATE_BIT_CHK,
+	STATE_BIT_START,
+	STATE_FINISHED,
+	STATE_ERROR,
+};
+
+static void ir_imon_decode_scancode(struct rc_dev *dev)
+{
+	struct imon_dec *imon = &dev->raw->imon;
+
+	/* Keyboard/Mouse toggle */
+	if (imon->bits == 0x299115b7)
+		imon->stick_keyboard = !imon->stick_keyboard;
+
+	if ((imon->bits & 0xfc0000ff) == 0x680000b7) {
+		int rel_x, rel_y;
+		u8 buf;
+
+		buf = imon->bits >> 16;
+		rel_x = (buf & 0x08) | (buf & 0x10) >> 2 |
+			(buf & 0x20) >> 4 | (buf & 0x40) >> 6;
+		if (imon->bits & 0x02000000)
+			rel_x |= ~0x0f;
+		buf = imon->bits >> 8;
+		rel_y = (buf & 0x08) | (buf & 0x10) >> 2 |
+			(buf & 0x20) >> 4 | (buf & 0x40) >> 6;
+		if (imon->bits & 0x01000000)
+			rel_y |= ~0x0f;
+
+		if (rel_x && rel_y && imon->stick_keyboard) {
+			if (abs(rel_y) > abs(rel_x))
+				imon->bits = rel_y > 0 ?
+					0x289515b7 : /* KEY_DOWN */
+					0x2aa515b7;  /* KEY_UP */
+			else
+				imon->bits = rel_x > 0 ?
+					0x2ba515b7 : /* KEY_RIGHT */
+					0x29a515b7;  /* KEY_LEFT */
+		}
+
+		if (!imon->stick_keyboard) {
+			input_report_rel(dev->input_dev, REL_X, rel_x);
+			input_report_rel(dev->input_dev, REL_Y, rel_y);
+
+			input_report_key(dev->input_dev, BTN_LEFT,
+					 (imon->bits & 0x00010000) != 0);
+			input_report_key(dev->input_dev, BTN_RIGHT,
+					 (imon->bits & 0x00040000) != 0);
+		}
+	}
+
+	rc_keydown(dev, RC_PROTO_IMON, imon->bits, 0);
+}
+
+/**
+ * ir_imon_decode() - Decode one iMON pulse or space
+ * @dev:	the struct rc_dev descriptor of the device
+ * @ev:		the struct ir_raw_event descriptor of the pulse/space
+ *
+ * This function returns -EINVAL if the pulse violates the state machine
+ */
+static int ir_imon_decode(struct rc_dev *dev, struct ir_raw_event ev)
+{
+	struct imon_dec *data = &dev->raw->imon;
+
+	if (!is_timing_event(ev)) {
+		if (ev.overflow)
+			data->state = STATE_INACTIVE;
+		return 0;
+	}
+
+	dev_dbg(&dev->dev,
+		"iMON decode started at state %d bitno %d (%uus %s)\n",
+		data->state, data->count, ev.duration, TO_STR(ev.pulse));
+
+	/*
+	 * Since iMON protocol is a series of bits, if at any point
+	 * we encounter an error, make sure that any remaining bits
+	 * aren't parsed as a scancode made up of less bits.
+	 *
+	 * Note that if the stick is held, then the remote repeats
+	 * the scancode with about 12ms between them. So, make sure
+	 * we have at least 10ms of space after an error. That way,
+	 * we're at a new scancode.
+	 */
+	if (data->state == STATE_ERROR) {
+		if (!ev.pulse && ev.duration > MS_TO_US(10))
+			data->state = STATE_INACTIVE;
+		return 0;
+	}
+
+	for (;;) {
+		if (!geq_margin(ev.duration, IMON_UNIT, IMON_UNIT / 2))
+			return 0;
+
+		decrease_duration(&ev, IMON_UNIT);
+
+		switch (data->state) {
+		case STATE_INACTIVE:
+			if (ev.pulse) {
+				data->state = STATE_BIT_CHK;
+				data->bits = 0;
+				data->count = IMON_BITS;
+			}
+			break;
+		case STATE_BIT_CHK:
+			if (IMON_CHKBITS & BIT(data->count))
+				data->last_chk = ev.pulse;
+			else if (ev.pulse)
+				goto err_out;
+			data->state = STATE_BIT_START;
+			break;
+		case STATE_BIT_START:
+			data->bits <<= 1;
+			if (!ev.pulse)
+				data->bits |= 1;
+
+			if (IMON_CHKBITS & BIT(data->count)) {
+				if (data->last_chk != !(data->bits & 3))
+					goto err_out;
+			}
+
+			if (!data->count--)
+				data->state = STATE_FINISHED;
+			else
+				data->state = STATE_BIT_CHK;
+			break;
+		case STATE_FINISHED:
+			if (ev.pulse)
+				goto err_out;
+			ir_imon_decode_scancode(dev);
+			data->state = STATE_INACTIVE;
+			break;
+		}
+	}
+
+err_out:
+	dev_dbg(&dev->dev,
+		"iMON decode failed at state %d bitno %d (%uus %s)\n",
+		data->state, data->count, ev.duration, TO_STR(ev.pulse));
+
+	data->state = STATE_ERROR;
+
+	return -EINVAL;
+}
+
+/**
+ * ir_imon_encode() - Encode a scancode as a stream of raw events
+ *
+ * @protocol:	protocol to encode
+ * @scancode:	scancode to encode
+ * @events:	array of raw ir events to write into
+ * @max:	maximum size of @events
+ *
+ * Returns:	The number of events written.
+ *		-ENOBUFS if there isn't enough space in the array to fit the
+ *		encoding. In this case all @max events will have been written.
+ */
+static int ir_imon_encode(enum rc_proto protocol, u32 scancode,
+			  struct ir_raw_event *events, unsigned int max)
+{
+	struct ir_raw_event *e = events;
+	int i, pulse;
+
+	if (!max--)
+		return -ENOBUFS;
+	init_ir_raw_event_duration(e, 1, IMON_UNIT);
+
+	for (i = IMON_BITS; i >= 0; i--) {
+		if (BIT(i) & IMON_CHKBITS)
+			pulse = !(scancode & (BIT(i) | BIT(i + 1)));
+		else
+			pulse = 0;
+
+		if (pulse == e->pulse) {
+			e->duration += IMON_UNIT;
+		} else {
+			if (!max--)
+				return -ENOBUFS;
+			init_ir_raw_event_duration(++e, pulse, IMON_UNIT);
+		}
+
+		pulse = !(scancode & BIT(i));
+
+		if (pulse == e->pulse) {
+			e->duration += IMON_UNIT;
+		} else {
+			if (!max--)
+				return -ENOBUFS;
+			init_ir_raw_event_duration(++e, pulse, IMON_UNIT);
+		}
+	}
+
+	if (e->pulse)
+		e++;
+
+	return e - events;
+}
+
+static int ir_imon_register(struct rc_dev *dev)
+{
+	struct imon_dec *imon = &dev->raw->imon;
+
+	imon->stick_keyboard = false;
+
+	return 0;
+}
+
+static struct ir_raw_handler imon_handler = {
+	.protocols	= RC_PROTO_BIT_IMON,
+	.decode		= ir_imon_decode,
+	.encode		= ir_imon_encode,
+	.carrier	= 38000,
+	.raw_register	= ir_imon_register,
+	.min_timeout	= IMON_UNIT * IMON_BITS * 2,
+};
+
+static int __init ir_imon_decode_init(void)
+{
+	ir_raw_handler_register(&imon_handler);
+
+	pr_info("IR iMON protocol handler initialized\n");
+	return 0;
+}
+
+static void __exit ir_imon_decode_exit(void)
+{
+	ir_raw_handler_unregister(&imon_handler);
+}
+
+module_init(ir_imon_decode_init);
+module_exit(ir_imon_decode_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Sean Young <sean@mess.org>");
+MODULE_DESCRIPTION("iMON IR protocol decoder");