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, 74 insertions, 0 deletions

diff --git a/Documentation/crypto/intro.rst b/Documentation/crypto/intro.rst
new file mode 100644
index 000000000..9aa89ebbf
--- /dev/null
+++ b/Documentation/crypto/intro.rst
@@ -0,0 +1,74 @@
+Kernel Crypto API Interface Specification
+=========================================
+
+Introduction
+------------
+
+The kernel crypto API offers a rich set of cryptographic ciphers as well
+as other data transformation mechanisms and methods to invoke these.
+This document contains a description of the API and provides example
+code.
+
+To understand and properly use the kernel crypto API a brief explanation
+of its structure is given. Based on the architecture, the API can be
+separated into different components. Following the architecture
+specification, hints to developers of ciphers are provided. Pointers to
+the API function call documentation are given at the end.
+
+The kernel crypto API refers to all algorithms as "transformations".
+Therefore, a cipher handle variable usually has the name "tfm". Besides
+cryptographic operations, the kernel crypto API also knows compression
+transformations and handles them the same way as ciphers.
+
+The kernel crypto API serves the following entity types:
+
+-  consumers requesting cryptographic services
+
+-  data transformation implementations (typically ciphers) that can be
+   called by consumers using the kernel crypto API
+
+This specification is intended for consumers of the kernel crypto API as
+well as for developers implementing ciphers. This API specification,
+however, does not discuss all API calls available to data transformation
+implementations (i.e. implementations of ciphers and other
+transformations (such as CRC or even compression algorithms) that can
+register with the kernel crypto API).
+
+Note: The terms "transformation" and cipher algorithm are used
+interchangeably.
+
+Terminology
+-----------
+
+The transformation implementation is an actual code or interface to
+hardware which implements a certain transformation with precisely
+defined behavior.
+
+The transformation object (TFM) is an instance of a transformation
+implementation. There can be multiple transformation objects associated
+with a single transformation implementation. Each of those
+transformation objects is held by a crypto API consumer or another
+transformation. Transformation object is allocated when a crypto API
+consumer requests a transformation implementation. The consumer is then
+provided with a structure, which contains a transformation object (TFM).
+
+The structure that contains transformation objects may also be referred
+to as a "cipher handle". Such a cipher handle is always subject to the
+following phases that are reflected in the API calls applicable to such
+a cipher handle:
+
+1. Initialization of a cipher handle.
+
+2. Execution of all intended cipher operations applicable for the handle
+   where the cipher handle must be furnished to every API call.
+
+3. Destruction of a cipher handle.
+
+When using the initialization API calls, a cipher handle is created and
+returned to the consumer. Therefore, please refer to all initialization
+API calls that refer to the data structure type a consumer is expected
+to receive and subsequently to use. The initialization API calls have
+all the same naming conventions of crypto_alloc\*.
+
+The transformation context is private data associated with the
+transformation object.