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, 343 insertions, 0 deletions

diff --git a/net/rfkill/input.c b/net/rfkill/input.c
new file mode 100644
index 000000000..f7cb24ce7
--- /dev/null
+++ b/net/rfkill/input.c
@@ -0,0 +1,343 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Input layer to RF Kill interface connector
+ *
+ * Copyright (c) 2007 Dmitry Torokhov
+ * Copyright 2009 Johannes Berg <johannes@sipsolutions.net>
+ *
+ * If you ever run into a situation in which you have a SW_ type rfkill
+ * input device, then you can revive code that was removed in the patch
+ * "rfkill-input: remove unused code".
+ */
+
+#include <linux/input.h>
+#include <linux/slab.h>
+#include <linux/moduleparam.h>
+#include <linux/workqueue.h>
+#include <linux/init.h>
+#include <linux/rfkill.h>
+#include <linux/sched.h>
+
+#include "rfkill.h"
+
+enum rfkill_input_master_mode {
+	RFKILL_INPUT_MASTER_UNLOCK = 0,
+	RFKILL_INPUT_MASTER_RESTORE = 1,
+	RFKILL_INPUT_MASTER_UNBLOCKALL = 2,
+	NUM_RFKILL_INPUT_MASTER_MODES
+};
+
+/* Delay (in ms) between consecutive switch ops */
+#define RFKILL_OPS_DELAY 200
+
+static enum rfkill_input_master_mode rfkill_master_switch_mode =
+					RFKILL_INPUT_MASTER_UNBLOCKALL;
+module_param_named(master_switch_mode, rfkill_master_switch_mode, uint, 0);
+MODULE_PARM_DESC(master_switch_mode,
+	"SW_RFKILL_ALL ON should: 0=do nothing (only unlock); 1=restore; 2=unblock all");
+
+static DEFINE_SPINLOCK(rfkill_op_lock);
+static bool rfkill_op_pending;
+static unsigned long rfkill_sw_pending[BITS_TO_LONGS(NUM_RFKILL_TYPES)];
+static unsigned long rfkill_sw_state[BITS_TO_LONGS(NUM_RFKILL_TYPES)];
+
+enum rfkill_sched_op {
+	RFKILL_GLOBAL_OP_EPO = 0,
+	RFKILL_GLOBAL_OP_RESTORE,
+	RFKILL_GLOBAL_OP_UNLOCK,
+	RFKILL_GLOBAL_OP_UNBLOCK,
+};
+
+static enum rfkill_sched_op rfkill_master_switch_op;
+static enum rfkill_sched_op rfkill_op;
+
+static void __rfkill_handle_global_op(enum rfkill_sched_op op)
+{
+	unsigned int i;
+
+	switch (op) {
+	case RFKILL_GLOBAL_OP_EPO:
+		rfkill_epo();
+		break;
+	case RFKILL_GLOBAL_OP_RESTORE:
+		rfkill_restore_states();
+		break;
+	case RFKILL_GLOBAL_OP_UNLOCK:
+		rfkill_remove_epo_lock();
+		break;
+	case RFKILL_GLOBAL_OP_UNBLOCK:
+		rfkill_remove_epo_lock();
+		for (i = 0; i < NUM_RFKILL_TYPES; i++)
+			rfkill_switch_all(i, false);
+		break;
+	default:
+		/* memory corruption or bug, fail safely */
+		rfkill_epo();
+		WARN(1, "Unknown requested operation %d! "
+			"rfkill Emergency Power Off activated\n",
+			op);
+	}
+}
+
+static void __rfkill_handle_normal_op(const enum rfkill_type type,
+				      const bool complement)
+{
+	bool blocked;
+
+	blocked = rfkill_get_global_sw_state(type);
+	if (complement)
+		blocked = !blocked;
+
+	rfkill_switch_all(type, blocked);
+}
+
+static void rfkill_op_handler(struct work_struct *work)
+{
+	unsigned int i;
+	bool c;
+
+	spin_lock_irq(&rfkill_op_lock);
+	do {
+		if (rfkill_op_pending) {
+			enum rfkill_sched_op op = rfkill_op;
+			rfkill_op_pending = false;
+			memset(rfkill_sw_pending, 0,
+				sizeof(rfkill_sw_pending));
+			spin_unlock_irq(&rfkill_op_lock);
+
+			__rfkill_handle_global_op(op);
+
+			spin_lock_irq(&rfkill_op_lock);
+
+			/*
+			 * handle global ops first -- during unlocked period
+			 * we might have gotten a new global op.
+			 */
+			if (rfkill_op_pending)
+				continue;
+		}
+
+		if (rfkill_is_epo_lock_active())
+			continue;
+
+		for (i = 0; i < NUM_RFKILL_TYPES; i++) {
+			if (__test_and_clear_bit(i, rfkill_sw_pending)) {
+				c = __test_and_clear_bit(i, rfkill_sw_state);
+				spin_unlock_irq(&rfkill_op_lock);
+
+				__rfkill_handle_normal_op(i, c);
+
+				spin_lock_irq(&rfkill_op_lock);
+			}
+		}
+	} while (rfkill_op_pending);
+	spin_unlock_irq(&rfkill_op_lock);
+}
+
+static DECLARE_DELAYED_WORK(rfkill_op_work, rfkill_op_handler);
+static unsigned long rfkill_last_scheduled;
+
+static unsigned long rfkill_ratelimit(const unsigned long last)
+{
+	const unsigned long delay = msecs_to_jiffies(RFKILL_OPS_DELAY);
+	return time_after(jiffies, last + delay) ? 0 : delay;
+}
+
+static void rfkill_schedule_ratelimited(void)
+{
+	if (schedule_delayed_work(&rfkill_op_work,
+				  rfkill_ratelimit(rfkill_last_scheduled)))
+		rfkill_last_scheduled = jiffies;
+}
+
+static void rfkill_schedule_global_op(enum rfkill_sched_op op)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&rfkill_op_lock, flags);
+	rfkill_op = op;
+	rfkill_op_pending = true;
+	if (op == RFKILL_GLOBAL_OP_EPO && !rfkill_is_epo_lock_active()) {
+		/* bypass the limiter for EPO */
+		mod_delayed_work(system_percpu_wq, &rfkill_op_work, 0);
+		rfkill_last_scheduled = jiffies;
+	} else
+		rfkill_schedule_ratelimited();
+	spin_unlock_irqrestore(&rfkill_op_lock, flags);
+}
+
+static void rfkill_schedule_toggle(enum rfkill_type type)
+{
+	unsigned long flags;
+
+	if (rfkill_is_epo_lock_active())
+		return;
+
+	spin_lock_irqsave(&rfkill_op_lock, flags);
+	if (!rfkill_op_pending) {
+		__set_bit(type, rfkill_sw_pending);
+		__change_bit(type, rfkill_sw_state);
+		rfkill_schedule_ratelimited();
+	}
+	spin_unlock_irqrestore(&rfkill_op_lock, flags);
+}
+
+static void rfkill_schedule_evsw_rfkillall(int state)
+{
+	if (state)
+		rfkill_schedule_global_op(rfkill_master_switch_op);
+	else
+		rfkill_schedule_global_op(RFKILL_GLOBAL_OP_EPO);
+}
+
+static void rfkill_event(struct input_handle *handle, unsigned int type,
+			unsigned int code, int data)
+{
+	if (type == EV_KEY && data == 1) {
+		switch (code) {
+		case KEY_WLAN:
+			rfkill_schedule_toggle(RFKILL_TYPE_WLAN);
+			break;
+		case KEY_BLUETOOTH:
+			rfkill_schedule_toggle(RFKILL_TYPE_BLUETOOTH);
+			break;
+		case KEY_UWB:
+			rfkill_schedule_toggle(RFKILL_TYPE_UWB);
+			break;
+		case KEY_WIMAX:
+			rfkill_schedule_toggle(RFKILL_TYPE_WIMAX);
+			break;
+		case KEY_RFKILL:
+			rfkill_schedule_toggle(RFKILL_TYPE_ALL);
+			break;
+		}
+	} else if (type == EV_SW && code == SW_RFKILL_ALL)
+		rfkill_schedule_evsw_rfkillall(data);
+}
+
+static int rfkill_connect(struct input_handler *handler, struct input_dev *dev,
+			  const struct input_device_id *id)
+{
+	struct input_handle *handle;
+	int error;
+
+	handle = kzalloc_obj(struct input_handle);
+	if (!handle)
+		return -ENOMEM;
+
+	handle->dev = dev;
+	handle->handler = handler;
+	handle->name = "rfkill";
+
+	/* causes rfkill_start() to be called */
+	error = input_register_handle(handle);
+	if (error)
+		goto err_free_handle;
+
+	error = input_open_device(handle);
+	if (error)
+		goto err_unregister_handle;
+
+	return 0;
+
+ err_unregister_handle:
+	input_unregister_handle(handle);
+ err_free_handle:
+	kfree(handle);
+	return error;
+}
+
+static void rfkill_start(struct input_handle *handle)
+{
+	/*
+	 * Take event_lock to guard against configuration changes, we
+	 * should be able to deal with concurrency with rfkill_event()
+	 * just fine (which event_lock will also avoid).
+	 */
+	spin_lock_irq(&handle->dev->event_lock);
+
+	if (test_bit(EV_SW, handle->dev->evbit) &&
+	    test_bit(SW_RFKILL_ALL, handle->dev->swbit))
+		rfkill_schedule_evsw_rfkillall(test_bit(SW_RFKILL_ALL,
+							handle->dev->sw));
+
+	spin_unlock_irq(&handle->dev->event_lock);
+}
+
+static void rfkill_disconnect(struct input_handle *handle)
+{
+	input_close_device(handle);
+	input_unregister_handle(handle);
+	kfree(handle);
+}
+
+static const struct input_device_id rfkill_ids[] = {
+	{
+		.flags = INPUT_DEVICE_ID_MATCH_EVBIT | INPUT_DEVICE_ID_MATCH_KEYBIT,
+		.evbit = { BIT_MASK(EV_KEY) },
+		.keybit = { [BIT_WORD(KEY_WLAN)] = BIT_MASK(KEY_WLAN) },
+	},
+	{
+		.flags = INPUT_DEVICE_ID_MATCH_EVBIT | INPUT_DEVICE_ID_MATCH_KEYBIT,
+		.evbit = { BIT_MASK(EV_KEY) },
+		.keybit = { [BIT_WORD(KEY_BLUETOOTH)] = BIT_MASK(KEY_BLUETOOTH) },
+	},
+	{
+		.flags = INPUT_DEVICE_ID_MATCH_EVBIT | INPUT_DEVICE_ID_MATCH_KEYBIT,
+		.evbit = { BIT_MASK(EV_KEY) },
+		.keybit = { [BIT_WORD(KEY_UWB)] = BIT_MASK(KEY_UWB) },
+	},
+	{
+		.flags = INPUT_DEVICE_ID_MATCH_EVBIT | INPUT_DEVICE_ID_MATCH_KEYBIT,
+		.evbit = { BIT_MASK(EV_KEY) },
+		.keybit = { [BIT_WORD(KEY_WIMAX)] = BIT_MASK(KEY_WIMAX) },
+	},
+	{
+		.flags = INPUT_DEVICE_ID_MATCH_EVBIT | INPUT_DEVICE_ID_MATCH_KEYBIT,
+		.evbit = { BIT_MASK(EV_KEY) },
+		.keybit = { [BIT_WORD(KEY_RFKILL)] = BIT_MASK(KEY_RFKILL) },
+	},
+	{
+		.flags = INPUT_DEVICE_ID_MATCH_EVBIT | INPUT_DEVICE_ID_MATCH_SWBIT,
+		.evbit = { BIT(EV_SW) },
+		.swbit = { [BIT_WORD(SW_RFKILL_ALL)] = BIT_MASK(SW_RFKILL_ALL) },
+	},
+	{ }
+};
+
+static struct input_handler rfkill_handler = {
+	.name =	"rfkill",
+	.event = rfkill_event,
+	.connect = rfkill_connect,
+	.start = rfkill_start,
+	.disconnect = rfkill_disconnect,
+	.id_table = rfkill_ids,
+};
+
+int __init rfkill_handler_init(void)
+{
+	switch (rfkill_master_switch_mode) {
+	case RFKILL_INPUT_MASTER_UNBLOCKALL:
+		rfkill_master_switch_op = RFKILL_GLOBAL_OP_UNBLOCK;
+		break;
+	case RFKILL_INPUT_MASTER_RESTORE:
+		rfkill_master_switch_op = RFKILL_GLOBAL_OP_RESTORE;
+		break;
+	case RFKILL_INPUT_MASTER_UNLOCK:
+		rfkill_master_switch_op = RFKILL_GLOBAL_OP_UNLOCK;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Avoid delay at first schedule */
+	rfkill_last_scheduled =
+			jiffies - msecs_to_jiffies(RFKILL_OPS_DELAY) - 1;
+	return input_register_handler(&rfkill_handler);
+}
+
+void __exit rfkill_handler_exit(void)
+{
+	input_unregister_handler(&rfkill_handler);
+	cancel_delayed_work_sync(&rfkill_op_work);
+}