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, 132 insertions, 0 deletions

diff --git a/Documentation/driver-api/rfkill.rst b/Documentation/driver-api/rfkill.rst
new file mode 100644
index 000000000..7d3684e81
--- /dev/null
+++ b/Documentation/driver-api/rfkill.rst
@@ -0,0 +1,132 @@
+===============================
+rfkill - RF kill switch support
+===============================
+
+
+.. contents::
+   :depth: 2
+
+Introduction
+============
+
+The rfkill subsystem provides a generic interface for disabling any radio
+transmitter in the system. When a transmitter is blocked, it shall not
+radiate any power.
+
+The subsystem also provides the ability to react on button presses and
+disable all transmitters of a certain type (or all). This is intended for
+situations where transmitters need to be turned off, for example on
+aircraft.
+
+The rfkill subsystem has a concept of "hard" and "soft" block, which
+differ little in their meaning (block == transmitters off) but rather in
+whether they can be changed or not:
+
+ - hard block
+	read-only radio block that cannot be overridden by software
+
+ - soft block
+	writable radio block (need not be readable) that is set by
+        the system software.
+
+The rfkill subsystem has two parameters, rfkill.default_state and
+rfkill.master_switch_mode, which are documented in
+admin-guide/kernel-parameters.rst.
+
+
+Implementation details
+======================
+
+The rfkill subsystem is composed of three main components:
+
+ * the rfkill core,
+ * the deprecated rfkill-input module (an input layer handler, being
+   replaced by userspace policy code) and
+ * the rfkill drivers.
+
+The rfkill core provides API for kernel drivers to register their radio
+transmitter with the kernel, methods for turning it on and off, and letting
+the system know about hardware-disabled states that may be implemented on
+the device.
+
+The rfkill core code also notifies userspace of state changes, and provides
+ways for userspace to query the current states. See the "Userspace support"
+section below.
+
+When the device is hard-blocked (either by a call to rfkill_set_hw_state()
+or from query_hw_block), set_block() will be invoked for additional software
+block, but drivers can ignore the method call since they can use the return
+value of the function rfkill_set_hw_state() to sync the software state
+instead of keeping track of calls to set_block(). In fact, drivers should
+use the return value of rfkill_set_hw_state() unless the hardware actually
+keeps track of soft and hard block separately.
+
+
+Kernel API
+==========
+
+Drivers for radio transmitters normally implement an rfkill driver.
+
+Platform drivers might implement input devices if the rfkill button is just
+that, a button. If that button influences the hardware then you need to
+implement an rfkill driver instead. This also applies if the platform provides
+a way to turn on/off the transmitter(s).
+
+For some platforms, it is possible that the hardware state changes during
+suspend/hibernation, in which case it will be necessary to update the rfkill
+core with the current state at resume time.
+
+To create an rfkill driver, driver's Kconfig needs to have::
+
+	depends on RFKILL || !RFKILL
+
+to ensure the driver cannot be built-in when rfkill is modular. The !RFKILL
+case allows the driver to be built when rfkill is not configured, in which
+case all rfkill API can still be used but will be provided by static inlines
+which compile to almost nothing.
+
+Calling rfkill_set_hw_state() when a state change happens is required from
+rfkill drivers that control devices that can be hard-blocked unless they also
+assign the poll_hw_block() callback (then the rfkill core will poll the
+device). Don't do this unless you cannot get the event in any other way.
+
+rfkill provides per-switch LED triggers, which can be used to drive LEDs
+according to the switch state (LED_FULL when blocked, LED_OFF otherwise).
+
+
+Userspace support
+=================
+
+The recommended userspace interface to use is /dev/rfkill, which is a misc
+character device that allows userspace to obtain and set the state of rfkill
+devices and sets of devices. It also notifies userspace about device addition
+and removal. The API is a simple read/write API that is defined in
+linux/rfkill.h, with one ioctl that allows turning off the deprecated input
+handler in the kernel for the transition period.
+
+Except for the one ioctl, communication with the kernel is done via read()
+and write() of instances of 'struct rfkill_event'. In this structure, the
+soft and hard block are properly separated (unlike sysfs, see below) and
+userspace is able to get a consistent snapshot of all rfkill devices in the
+system. Also, it is possible to switch all rfkill drivers (or all drivers of
+a specified type) into a state which also updates the default state for
+hotplugged devices.
+
+After an application opens /dev/rfkill, it can read the current state of all
+devices. Changes can be obtained by either polling the descriptor for
+hotplug or state change events or by listening for uevents emitted by the
+rfkill core framework.
+
+Additionally, each rfkill device is registered in sysfs and emits uevents.
+
+rfkill devices issue uevents (with an action of "change"), with the following
+environment variables set::
+
+	RFKILL_NAME
+	RFKILL_STATE
+	RFKILL_TYPE
+
+The content of these variables corresponds to the "name", "state" and
+"type" sysfs files explained above.
+
+For further details consult Documentation/ABI/stable/sysfs-class-rfkill.