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, 191 insertions, 0 deletions

diff --git a/drivers/hid/hid-holtekff.c b/drivers/hid/hid-holtekff.c
new file mode 100644
index 000000000..4834d42b2
--- /dev/null
+++ b/drivers/hid/hid-holtekff.c
@@ -0,0 +1,191 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ *  Force feedback support for Holtek On Line Grip based gamepads
+ *
+ *  These include at least a Brazilian "Clone Joypad Super Power Fire"
+ *  which uses vendor ID 0x1241 and identifies as "HOLTEK On Line Grip".
+ *
+ *  Copyright (c) 2011 Anssi Hannula <anssi.hannula@iki.fi>
+ */
+
+/*
+ */
+
+#include <linux/hid.h>
+#include <linux/input.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#include "hid-ids.h"
+
+#ifdef CONFIG_HOLTEK_FF
+
+/*
+ * These commands and parameters are currently known:
+ *
+ * byte 0: command id:
+ * 	01  set effect parameters
+ * 	02  play specified effect
+ * 	03  stop specified effect
+ * 	04  stop all effects
+ * 	06  stop all effects
+ * 	(the difference between 04 and 06 isn't known; win driver
+ * 	 sends 06,04 on application init, and 06 otherwise)
+ * 
+ * Commands 01 and 02 need to be sent as pairs, i.e. you need to send 01
+ * before each 02.
+ *
+ * The rest of the bytes are parameters. Command 01 takes all of them, and
+ * commands 02,03 take only the effect id.
+ *
+ * byte 1:
+ *	bits 0-3: effect id:
+ * 		1: very strong rumble
+ * 		2: periodic rumble, short intervals
+ * 		3: very strong rumble
+ * 		4: periodic rumble, long intervals
+ * 		5: weak periodic rumble, long intervals
+ * 		6: weak periodic rumble, short intervals
+ * 		7: periodic rumble, short intervals
+ * 		8: strong periodic rumble, short intervals
+ * 		9: very strong rumble
+ * 		a: causes an error
+ * 		b: very strong periodic rumble, very short intervals
+ * 		c-f: nothing
+ *	bit 6: right (weak) motor enabled
+ *	bit 7: left (strong) motor enabled
+ *
+ * bytes 2-3:  time in milliseconds, big-endian
+ * bytes 5-6:  unknown (win driver seems to use at least 10e0 with effect 1
+ * 		       and 0014 with effect 6)
+ * byte 7:
+ *	bits 0-3: effect magnitude
+ */
+
+#define HOLTEKFF_MSG_LENGTH     7
+
+static const u8 start_effect_1[] = { 0x02, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00 };
+static const u8 stop_all4[] =	   { 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 };
+static const u8 stop_all6[] =	   { 0x06, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 };
+
+struct holtekff_device {
+	struct hid_field *field;
+};
+
+static void holtekff_send(struct holtekff_device *holtekff,
+			  struct hid_device *hid,
+			  const u8 data[HOLTEKFF_MSG_LENGTH])
+{
+	int i;
+
+	for (i = 0; i < HOLTEKFF_MSG_LENGTH; i++) {
+		holtekff->field->value[i] = data[i];
+	}
+
+	dbg_hid("sending %7ph\n", data);
+
+	hid_hw_request(hid, holtekff->field->report, HID_REQ_SET_REPORT);
+}
+
+static int holtekff_play(struct input_dev *dev, void *data,
+			 struct ff_effect *effect)
+{
+	struct hid_device *hid = input_get_drvdata(dev);
+	struct holtekff_device *holtekff = data;
+	int left, right;
+	/* effect type 1, length 65535 msec */
+	u8 buf[HOLTEKFF_MSG_LENGTH] =
+		{ 0x01, 0x01, 0xff, 0xff, 0x10, 0xe0, 0x00 };
+
+	left = effect->u.rumble.strong_magnitude;
+	right = effect->u.rumble.weak_magnitude;
+	dbg_hid("called with 0x%04x 0x%04x\n", left, right);
+
+	if (!left && !right) {
+		holtekff_send(holtekff, hid, stop_all6);
+		return 0;
+	}
+
+	if (left)
+		buf[1] |= 0x80;
+	if (right)
+		buf[1] |= 0x40;
+
+	/* The device takes a single magnitude, so we just sum them up. */
+	buf[6] = min(0xf, (left >> 12) + (right >> 12));
+
+	holtekff_send(holtekff, hid, buf);
+	holtekff_send(holtekff, hid, start_effect_1);
+
+	return 0;
+}
+
+static int holtek_input_configured(struct hid_device *hid, struct hid_input *hidinput)
+{
+	struct holtekff_device *holtekff;
+	struct hid_report *report;
+	struct list_head *report_list =
+			&hid->report_enum[HID_OUTPUT_REPORT].report_list;
+	struct input_dev *dev = hidinput->input;
+	int error;
+
+	if (!list_is_first(&hidinput->list, &hid->inputs))
+		return 0;
+
+	report = list_first_entry_or_null(report_list, struct hid_report, list);
+	if (!report) {
+		hid_err(hid, "no output report found\n");
+		return -ENODEV;
+	}
+
+	if (report->maxfield < 1 || report->field[0]->report_count != 7) {
+		hid_err(hid, "unexpected output report layout\n");
+		return -ENODEV;
+	}
+
+	holtekff = kzalloc_obj(*holtekff);
+	if (!holtekff)
+		return -ENOMEM;
+
+	set_bit(FF_RUMBLE, dev->ffbit);
+
+	holtekff->field = report->field[0];
+
+	/* initialize the same way as win driver does */
+	holtekff_send(holtekff, hid, stop_all4);
+	holtekff_send(holtekff, hid, stop_all6);
+
+	error = input_ff_create_memless(dev, holtekff, holtekff_play);
+	if (error) {
+		kfree(holtekff);
+		return error;
+	}
+
+	hid_info(hid, "Force feedback for Holtek On Line Grip based devices by Anssi Hannula <anssi.hannula@iki.fi>\n");
+
+	return 0;
+}
+#else
+static inline int holtek_input_configured(struct hid_device *hid,
+					  struct hid_input *hidinput)
+{
+	return 0;
+}
+#endif
+
+static const struct hid_device_id holtek_devices[] = {
+	{ HID_USB_DEVICE(USB_VENDOR_ID_HOLTEK, USB_DEVICE_ID_HOLTEK_ON_LINE_GRIP) },
+	{ }
+};
+MODULE_DEVICE_TABLE(hid, holtek_devices);
+
+static struct hid_driver holtek_driver = {
+	.name = "holtek",
+	.id_table = holtek_devices,
+	.input_configured = holtek_input_configured,
+};
+module_hid_driver(holtek_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Anssi Hannula <anssi.hannula@iki.fi>");
+MODULE_DESCRIPTION("Force feedback support for Holtek On Line Grip based devices");