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, 48 insertions, 0 deletions

diff --git a/Documentation/admin-guide/init.rst b/Documentation/admin-guide/init.rst
new file mode 100644
index 000000000..41f06a091
--- /dev/null
+++ b/Documentation/admin-guide/init.rst
@@ -0,0 +1,48 @@
+Explaining the "No working init found." boot hang message
+=========================================================
+:Authors: Andreas Mohr <andi at lisas period de>
+          Cristian Souza <cristianmsbr at gmail period com>
+
+This document provides some high-level reasons for failure
+(listed roughly in order of execution) to load the init binary.
+
+1) **Unable to mount root FS**: Set "debug" kernel parameter (in bootloader
+   config file or CONFIG_CMDLINE) to get more detailed kernel messages.
+
+2) **init binary doesn't exist on rootfs**: Make sure you have the correct
+   root FS type (and ``root=`` kernel parameter points to the correct
+   partition), required drivers such as storage hardware (such as SCSI or
+   USB!) and filesystem (ext3, jffs2, etc.) are builtin (alternatively as
+   modules, to be pre-loaded by an initrd).
+
+3) **Broken console device**: Possibly a conflict in ``console= setup``
+   --> initial console unavailable. E.g. some serial consoles are unreliable
+   due to serial IRQ issues (e.g. missing interrupt-based configuration).
+   Try using a different ``console= device`` or e.g. ``netconsole=``.
+
+4) **Binary exists but dependencies not available**: E.g. required library
+   dependencies of the init binary such as ``/lib/ld-linux.so.2`` missing or
+   broken. Use ``readelf -d <INIT>|grep NEEDED`` to find out which libraries
+   are required.
+
+5) **Binary cannot be loaded**: Make sure the binary's architecture matches
+   your hardware. E.g. i386 vs. x86_64 mismatch, or trying to load x86 on ARM
+   hardware. In case you tried loading a non-binary file here (shell script?),
+   you should make sure that the script specifies an interpreter in its
+   shebang header line (``#!/...``) that is fully working (including its
+   library dependencies). And before tackling scripts, better first test a
+   simple non-script binary such as ``/bin/sh`` and confirm its successful
+   execution. To find out more, add code ``to init/main.c`` to display
+   kernel_execve()s return values.
+
+Please extend this explanation whenever you find new failure causes
+(after all loading the init binary is a CRITICAL and hard transition step
+which needs to be made as painless as possible), then submit a patch to LKML.
+Further TODOs:
+
+- Implement the various ``run_init_process()`` invocations via a struct array
+  which can then store the ``kernel_execve()`` result value and on failure
+  log it all by iterating over **all** results (very important usability fix).
+- Try to make the implementation itself more helpful in general, e.g. by
+  providing additional error messages at affected places.
+