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, 171 insertions, 0 deletions

diff --git a/drivers/acpi/wakeup.c b/drivers/acpi/wakeup.c
new file mode 100644
index 000000000..0cbee0f06
--- /dev/null
+++ b/drivers/acpi/wakeup.c
@@ -0,0 +1,171 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * wakeup.c - support wakeup devices
+ * Copyright (C) 2004 Li Shaohua <shaohua.li@intel.com>
+ */
+
+#include <linux/init.h>
+#include <linux/acpi.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+
+#include "internal.h"
+#include "sleep.h"
+
+struct acpi_wakeup_handler {
+	struct list_head list_node;
+	bool (*wakeup)(void *context);
+	void *context;
+};
+
+static LIST_HEAD(acpi_wakeup_handler_head);
+static DEFINE_MUTEX(acpi_wakeup_handler_mutex);
+
+/*
+ * We didn't lock acpi_device_lock in the file, because it invokes oops in
+ * suspend/resume and isn't really required as this is called in S-state. At
+ * that time, there is no device hotplug
+ **/
+
+/**
+ * acpi_enable_wakeup_devices - Enable wake-up device GPEs.
+ * @sleep_state: ACPI system sleep state.
+ *
+ * Enable wakeup device power of devices with the state.enable flag set and set
+ * the wakeup enable mask bits in the GPE registers that correspond to wakeup
+ * devices.
+ */
+void acpi_enable_wakeup_devices(u8 sleep_state)
+{
+	struct acpi_device *dev, *tmp;
+
+	list_for_each_entry_safe(dev, tmp, &acpi_wakeup_device_list,
+				 wakeup_list) {
+		if (!dev->wakeup.flags.valid
+		    || sleep_state > dev->wakeup.sleep_state
+		    || !(device_may_wakeup(&dev->dev)
+			 || dev->wakeup.prepare_count))
+			continue;
+
+		if (device_may_wakeup(&dev->dev))
+			acpi_enable_wakeup_device_power(dev, sleep_state);
+
+		/* The wake-up power should have been enabled already. */
+		acpi_set_gpe_wake_mask(dev->wakeup.gpe_device, dev->wakeup.gpe_number,
+				ACPI_GPE_ENABLE);
+	}
+}
+
+/**
+ * acpi_disable_wakeup_devices - Disable devices' wakeup capability.
+ * @sleep_state: ACPI system sleep state.
+ */
+void acpi_disable_wakeup_devices(u8 sleep_state)
+{
+	struct acpi_device *dev, *tmp;
+
+	list_for_each_entry_safe(dev, tmp, &acpi_wakeup_device_list,
+				 wakeup_list) {
+		if (!dev->wakeup.flags.valid
+		    || sleep_state > dev->wakeup.sleep_state
+		    || !(device_may_wakeup(&dev->dev)
+			 || dev->wakeup.prepare_count))
+			continue;
+
+		acpi_set_gpe_wake_mask(dev->wakeup.gpe_device, dev->wakeup.gpe_number,
+				ACPI_GPE_DISABLE);
+
+		if (device_may_wakeup(&dev->dev))
+			acpi_disable_wakeup_device_power(dev);
+	}
+}
+
+int __init acpi_wakeup_device_init(void)
+{
+	struct acpi_device *dev, *tmp;
+
+	mutex_lock(&acpi_device_lock);
+	list_for_each_entry_safe(dev, tmp, &acpi_wakeup_device_list,
+				 wakeup_list) {
+		if (device_can_wakeup(&dev->dev)) {
+			/* Button GPEs are supposed to be always enabled. */
+			acpi_enable_gpe(dev->wakeup.gpe_device,
+					dev->wakeup.gpe_number);
+			device_set_wakeup_enable(&dev->dev, true);
+		}
+	}
+	mutex_unlock(&acpi_device_lock);
+	return 0;
+}
+
+/**
+ * acpi_register_wakeup_handler - Register wakeup handler
+ * @wake_irq: The IRQ through which the device may receive wakeups
+ * @wakeup:   Wakeup-handler to call when the SCI has triggered a wakeup
+ * @context:  Context to pass to the handler when calling it
+ *
+ * Drivers which may share an IRQ with the SCI can use this to register
+ * a handler which returns true when the device they are managing wants
+ * to trigger a wakeup.
+ */
+int acpi_register_wakeup_handler(int wake_irq, bool (*wakeup)(void *context),
+				 void *context)
+{
+	struct acpi_wakeup_handler *handler;
+
+	/*
+	 * If the device is not sharing its IRQ with the SCI, there is no
+	 * need to register the handler.
+	 */
+	if (!acpi_sci_irq_valid() || wake_irq != acpi_sci_irq)
+		return 0;
+
+	handler = kmalloc_obj(*handler);
+	if (!handler)
+		return -ENOMEM;
+
+	handler->wakeup = wakeup;
+	handler->context = context;
+
+	mutex_lock(&acpi_wakeup_handler_mutex);
+	list_add(&handler->list_node, &acpi_wakeup_handler_head);
+	mutex_unlock(&acpi_wakeup_handler_mutex);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(acpi_register_wakeup_handler);
+
+/**
+ * acpi_unregister_wakeup_handler - Unregister wakeup handler
+ * @wakeup:   Wakeup-handler passed to acpi_register_wakeup_handler()
+ * @context:  Context passed to acpi_register_wakeup_handler()
+ */
+void acpi_unregister_wakeup_handler(bool (*wakeup)(void *context),
+				    void *context)
+{
+	struct acpi_wakeup_handler *handler;
+
+	mutex_lock(&acpi_wakeup_handler_mutex);
+	list_for_each_entry(handler, &acpi_wakeup_handler_head, list_node) {
+		if (handler->wakeup == wakeup && handler->context == context) {
+			list_del(&handler->list_node);
+			kfree(handler);
+			break;
+		}
+	}
+	mutex_unlock(&acpi_wakeup_handler_mutex);
+}
+EXPORT_SYMBOL_GPL(acpi_unregister_wakeup_handler);
+
+bool acpi_check_wakeup_handlers(void)
+{
+	struct acpi_wakeup_handler *handler;
+
+	/* No need to lock, nothing else is running when we're called. */
+	list_for_each_entry(handler, &acpi_wakeup_handler_head, list_node) {
+		if (handler->wakeup(handler->context))
+			return true;
+	}
+
+	return false;
+}