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, 75 insertions, 0 deletions

diff --git a/drivers/leds/TODO b/drivers/leds/TODO
new file mode 100644
index 000000000..bfa60fa1d
--- /dev/null
+++ b/drivers/leds/TODO
@@ -0,0 +1,75 @@
+-*- org -*-
+
+* On/off LEDs should have max_brightness of 1
+* Get rid of enum led_brightness
+
+It is really an integer, as maximum is configurable. Get rid of it, or
+make it into typedef or something.
+
+* Review atomicity requirements in LED subsystem
+
+Calls that may and that may not block are mixed in same structure, and
+semantics is sometimes non-intuitive. (For example blink callback may
+not sleep.) Review the requirements for any bugs and document them
+clearly.
+
+* LED names are still a mess
+
+No two LEDs have same name, so the names are probably unusable for the
+userland. Nudge authors into creating common LED names for common
+functionality.
+
+? Perhaps check for known LED names during boot, and warn if there are
+LEDs not on the list?
+
+* Split drivers into subdirectories
+
+The number of drivers is getting big, and driver for on/off LED on a
+i/o port is really quite different from camera flash LED, which is
+really different from driver for RGB color LED that can run its own
+microcode. Split the drivers somehow.
+
+* Figure out what to do with RGB leds
+
+Multicolor is a bit too abstract. Yes, we can have
+Green-Magenta-Ultraviolet LED, but so far all the LEDs we support are
+RGB, and not even RGB-White or RGB-Yellow variants emerged.
+
+Multicolor is not a good fit for RGB LED. It does not really know
+about LED color.  In particular, there's no way to make LED "white".
+
+Userspace is interested in knowing "this LED can produce arbitrary
+color", which not all multicolor LEDs can.
+
+	Proposal: let's add "rgb" to led_colors in drivers/leds/led-core.c,
+	add corresponding device tree defines, and use that, instead of
+	multicolor for RGB LEDs.
+
+	We really need to do that now; "white" stuff can wait.
+
+RGB LEDs are quite common, and it would be good to be able to turn LED
+white and to turn it into any arbitrary color. It is essential that
+userspace is able to set arbitrary colors, and it might be good to
+have that ability from kernel, too... to allow full-color triggers.
+
+* Command line utility to manipulate the LEDs?
+
+/sys interface is not really suitable to use by hand, should we have
+an utility to perform LED control?
+
+In particular, LED names are still a mess (see above) and utility
+could help there by presenting both old and new names while we clean
+them up.
+
+In future, I'd like utility to accept both old and new names while we
+clean them up.
+
+It would be also nice to have useful listing mode -- name, type,
+current brightness/trigger...
+
+In future, it would be good to be able to set rgb led to particular
+color.
+
+And probably user-friendly interface to access LEDs for particular
+ethernet interface would be nice.
+