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, 224 insertions, 0 deletions

diff --git a/Documentation/devicetree/of_unittest.rst b/Documentation/devicetree/of_unittest.rst
new file mode 100644
index 000000000..6ed6e3291
--- /dev/null
+++ b/Documentation/devicetree/of_unittest.rst
@@ -0,0 +1,224 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+=================================
+Open Firmware Devicetree Unittest
+=================================
+
+Author: Gaurav Minocha <gaurav.minocha.os@gmail.com>
+
+1. Introduction
+===============
+
+This document explains how the test data required for executing OF unittest
+is attached to the live tree dynamically, independent of the machine's
+architecture.
+
+It is recommended to read the following documents before moving ahead.
+
+(1) Documentation/devicetree/usage-model.rst
+(2) http://www.devicetree.org/Device_Tree_Usage
+
+OF Selftest has been designed to test the interface (include/linux/of.h)
+provided to device driver developers to fetch the device information..etc.
+from the unflattened device tree data structure. This interface is used by
+most of the device drivers in various use cases.
+
+
+2. Verbose Output (EXPECT)
+==========================
+
+If unittest detects a problem it will print a warning or error message to
+the console.  Unittest also triggers warning and error messages from other
+kernel code as a result of intentionally bad unittest data.  This has led
+to confusion as to whether the triggered messages are an expected result
+of a test or whether there is a real problem that is independent of unittest.
+
+'EXPECT \ : text' (begin) and 'EXPECT / : text' (end) messages have been
+added to unittest to report that a warning or error is expected.  The
+begin is printed before triggering the warning or error, and the end is
+printed after triggering the warning or error.
+
+The EXPECT messages result in very noisy console messages that are difficult
+to read.  The script scripts/dtc/of_unittest_expect was created to filter
+this verbosity and highlight mismatches between triggered warnings and
+errors vs expected warnings and errors.  More information is available
+from 'scripts/dtc/of_unittest_expect --help'.
+
+
+3. Test-data
+============
+
+The Device Tree Source file (drivers/of/unittest-data/testcases.dtso) contains
+the test data required for executing the unit tests automated in
+drivers/of/unittest.c. See the content of the folder::
+
+    drivers/of/unittest-data/tests-*.dtsi
+
+for the Device Tree Source Include files (.dtsi) included in testcases.dtso.
+
+When the kernel is built with CONFIG_OF_UNITTEST enabled, then the following make
+rule::
+
+    $(obj)/%.dtbo: $(src)/%.dtso $(DTC) FORCE
+	    $(call if_changed_dep,dtc)
+
+is used to compile the DT source file (testcases.dtso) into a binary blob
+(testcases.dtbo), also referred as flattened DT.
+
+After that, using the following rule the binary blob above is wrapped as an
+assembly file (testcases.dtbo.S)::
+
+    $(obj)/%.dtbo.S: $(obj)/%.dtbo FORCE
+	    $(call if_changed,wrap_S_dtb)
+
+The assembly file is compiled into an object file (testcases.dtbo.o), and is
+linked into the kernel image.
+
+
+3.1. Adding the test data
+-------------------------
+
+Un-flattened device tree structure:
+
+Un-flattened device tree consists of connected device_node(s) in form of a tree
+structure described below::
+
+    // following struct members are used to construct the tree
+    struct device_node {
+	...
+	struct  device_node *parent;
+	struct  device_node *child;
+	struct  device_node *sibling;
+	...
+    };
+
+Figure 1, describes a generic structure of machine's un-flattened device tree
+considering only child and sibling pointers. There exists another pointer,
+``*parent``, that is used to traverse the tree in the reverse direction. So, at
+a particular level the child node and all the sibling nodes will have a parent
+pointer pointing to a common node (e.g. child1, sibling2, sibling3, sibling4's
+parent points to root node)::
+
+    root ('/')
+    |
+    child1 -> sibling2 -> sibling3 -> sibling4 -> null
+    |         |           |           |
+    |         |           |          null
+    |         |           |
+    |         |        child31 -> sibling32 -> null
+    |         |           |          |
+    |         |          null       null
+    |         |
+    |      child21 -> sibling22 -> sibling23 -> null
+    |         |          |            |
+    |        null       null         null
+    |
+    child11 -> sibling12 -> sibling13 -> sibling14 -> null
+    |           |           |            |
+    |           |           |           null
+    |           |           |
+    null        null       child131 -> null
+			    |
+			    null
+
+Figure 1: Generic structure of un-flattened device tree
+
+
+Before executing OF unittest, it is required to attach the test data to
+machine's device tree (if present). So, when selftest_data_add() is called,
+at first it reads the flattened device tree data linked into the kernel image
+via the following kernel symbols::
+
+    __dtb_testcases_begin - address marking the start of test data blob
+    __dtb_testcases_end   - address marking the end of test data blob
+
+Secondly, it calls of_fdt_unflatten_tree() to unflatten the flattened
+blob. And finally, if the machine's device tree (i.e. live tree) is present,
+then it attaches the unflattened test data tree to the live tree, else it
+attaches itself as a live device tree.
+
+attach_node_and_children() uses of_attach_node() to attach the nodes into the
+live tree as explained below. To explain the same, the test data tree described
+in Figure 2 is attached to the live tree described in Figure 1::
+
+    root ('/')
+	|
+    testcase-data
+	|
+    test-child0 -> test-sibling1 -> test-sibling2 -> test-sibling3 -> null
+	|               |                |                |
+    test-child01      null             null             null
+
+
+Figure 2: Example test data tree to be attached to live tree.
+
+According to the scenario above, the live tree is already present so it isn't
+required to attach the root('/') node. All other nodes are attached by calling
+of_attach_node() on each node.
+
+In the function of_attach_node(), the new node is attached as the child of the
+given parent in live tree. But, if parent already has a child then the new node
+replaces the current child and turns it into its sibling. So, when the testcase
+data node is attached to the live tree above (Figure 1), the final structure is
+as shown in Figure 3::
+
+    root ('/')
+    |
+    testcase-data -> child1 -> sibling2 -> sibling3 -> sibling4 -> null
+    |               |          |           |           |
+    (...)             |          |           |          null
+		    |          |         child31 -> sibling32 -> null
+		    |          |           |           |
+		    |          |          null        null
+		    |          |
+		    |        child21 -> sibling22 -> sibling23 -> null
+		    |          |           |            |
+		    |         null        null         null
+		    |
+		    child11 -> sibling12 -> sibling13 -> sibling14 -> null
+		    |          |            |            |
+		    null       null          |           null
+					    |
+					    child131 -> null
+					    |
+					    null
+    -----------------------------------------------------------------------
+
+    root ('/')
+    |
+    testcase-data -> child1 -> sibling2 -> sibling3 -> sibling4 -> null
+    |               |          |           |           |
+    |             (...)      (...)       (...)        null
+    |
+    test-sibling3 -> test-sibling2 -> test-sibling1 -> test-child0 -> null
+    |                |                   |                |
+    null             null                null         test-child01
+
+
+Figure 3: Live device tree structure after attaching the testcase-data.
+
+
+Astute readers would have noticed that test-child0 node becomes the last
+sibling compared to the earlier structure (Figure 2). After attaching first
+test-child0 the test-sibling1 is attached that pushes the child node
+(i.e. test-child0) to become a sibling and makes itself a child node,
+as mentioned above.
+
+If a duplicate node is found (i.e. if a node with same full_name property is
+already present in the live tree), then the node isn't attached rather its
+properties are updated to the live tree's node by calling the function
+update_node_properties().
+
+
+3.2. Removing the test data
+---------------------------
+
+Once the test case execution is complete, selftest_data_remove is called in
+order to remove the device nodes attached initially (first the leaf nodes are
+detached and then moving up the parent nodes are removed, and eventually the
+whole tree). selftest_data_remove() calls detach_node_and_children() that uses
+of_detach_node() to detach the nodes from the live device tree.
+
+To detach a node, of_detach_node() either updates the child pointer of given
+node's parent to its sibling or attaches the previous sibling to the given
+node's sibling, as appropriate. That is it :)