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, 271 insertions, 0 deletions

diff --git a/drivers/acpi/acpica/nsparse.c b/drivers/acpi/acpica/nsparse.c
new file mode 100644
index 000000000..1e9c70c38
--- /dev/null
+++ b/drivers/acpi/acpica/nsparse.c
@@ -0,0 +1,271 @@
+// SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
+/******************************************************************************
+ *
+ * Module Name: nsparse - namespace interface to AML parser
+ *
+ * Copyright (C) 2000 - 2026, Intel Corp.
+ *
+ *****************************************************************************/
+
+#include <acpi/acpi.h>
+#include "accommon.h"
+#include "acnamesp.h"
+#include "acparser.h"
+#include "acdispat.h"
+#include "actables.h"
+#include "acinterp.h"
+
+#define _COMPONENT          ACPI_NAMESPACE
+ACPI_MODULE_NAME("nsparse")
+
+/*******************************************************************************
+ *
+ * FUNCTION:    ns_execute_table
+ *
+ * PARAMETERS:  table_desc      - An ACPI table descriptor for table to parse
+ *              start_node      - Where to enter the table into the namespace
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Load ACPI/AML table by executing the entire table as a single
+ *              large control method.
+ *
+ * NOTE: The point of this is to execute any module-level code in-place
+ * as the table is parsed. Some AML code depends on this behavior.
+ *
+ * It is a run-time option at this time, but will eventually become
+ * the default.
+ *
+ * Note: This causes the table to only have a single-pass parse.
+ * However, this is compatible with other ACPI implementations.
+ *
+ ******************************************************************************/
+acpi_status
+acpi_ns_execute_table(u32 table_index, struct acpi_namespace_node *start_node)
+{
+	acpi_status status;
+	struct acpi_table_header *table;
+	acpi_owner_id owner_id;
+	struct acpi_evaluate_info *info = NULL;
+	u32 aml_length;
+	u8 *aml_start;
+	union acpi_operand_object *method_obj = NULL;
+
+	ACPI_FUNCTION_TRACE(ns_execute_table);
+
+	status = acpi_get_table_by_index(table_index, &table);
+	if (ACPI_FAILURE(status)) {
+		return_ACPI_STATUS(status);
+	}
+
+	/* Table must consist of at least a complete header */
+
+	if (table->length < sizeof(struct acpi_table_header)) {
+		return_ACPI_STATUS(AE_BAD_HEADER);
+	}
+
+	aml_start = (u8 *)table + sizeof(struct acpi_table_header);
+	aml_length = table->length - sizeof(struct acpi_table_header);
+
+	status = acpi_tb_get_owner_id(table_index, &owner_id);
+	if (ACPI_FAILURE(status)) {
+		return_ACPI_STATUS(status);
+	}
+
+	/* Create, initialize, and link a new temporary method object */
+
+	method_obj = acpi_ut_create_internal_object(ACPI_TYPE_METHOD);
+	if (!method_obj) {
+		return_ACPI_STATUS(AE_NO_MEMORY);
+	}
+
+	/* Allocate the evaluation information block */
+
+	info = ACPI_ALLOCATE_ZEROED(sizeof(struct acpi_evaluate_info));
+	if (!info) {
+		status = AE_NO_MEMORY;
+		goto cleanup;
+	}
+
+	ACPI_DEBUG_PRINT_RAW((ACPI_DB_PARSE,
+			      "%s: Create table pseudo-method for [%4.4s] @%p, method %p\n",
+			      ACPI_GET_FUNCTION_NAME, table->signature, table,
+			      method_obj));
+
+	method_obj->method.aml_start = aml_start;
+	method_obj->method.aml_length = aml_length;
+	method_obj->method.owner_id = owner_id;
+	method_obj->method.info_flags |= ACPI_METHOD_MODULE_LEVEL;
+
+	info->pass_number = ACPI_IMODE_EXECUTE;
+	info->node = start_node;
+	info->obj_desc = method_obj;
+	info->node_flags = info->node->flags;
+	info->full_pathname = acpi_ns_get_normalized_pathname(info->node, TRUE);
+	if (!info->full_pathname) {
+		status = AE_NO_MEMORY;
+		goto cleanup;
+	}
+
+	/* Optional object evaluation log */
+
+	ACPI_DEBUG_PRINT_RAW((ACPI_DB_EVALUATION,
+			      "%-26s:  (Definition Block level)\n",
+			      "Module-level evaluation"));
+
+	status = acpi_ps_execute_table(info);
+
+	/* Optional object evaluation log */
+
+	ACPI_DEBUG_PRINT_RAW((ACPI_DB_EVALUATION,
+			      "%-26s:  (Definition Block level)\n",
+			      "Module-level complete"));
+
+cleanup:
+	if (info) {
+		ACPI_FREE(info->full_pathname);
+		info->full_pathname = NULL;
+	}
+	ACPI_FREE(info);
+	acpi_ut_remove_reference(method_obj);
+	return_ACPI_STATUS(status);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    ns_one_complete_parse
+ *
+ * PARAMETERS:  pass_number             - 1 or 2
+ *              table_desc              - The table to be parsed.
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Perform one complete parse of an ACPI/AML table.
+ *
+ ******************************************************************************/
+
+acpi_status
+acpi_ns_one_complete_parse(u32 pass_number,
+			   u32 table_index,
+			   struct acpi_namespace_node *start_node)
+{
+	union acpi_parse_object *parse_root;
+	acpi_status status;
+	u32 aml_length;
+	u8 *aml_start;
+	struct acpi_walk_state *walk_state;
+	struct acpi_table_header *table;
+	acpi_owner_id owner_id;
+
+	ACPI_FUNCTION_TRACE(ns_one_complete_parse);
+
+	status = acpi_get_table_by_index(table_index, &table);
+	if (ACPI_FAILURE(status)) {
+		return_ACPI_STATUS(status);
+	}
+
+	/* Table must consist of at least a complete header */
+
+	if (table->length < sizeof(struct acpi_table_header)) {
+		return_ACPI_STATUS(AE_BAD_HEADER);
+	}
+
+	aml_start = (u8 *)table + sizeof(struct acpi_table_header);
+	aml_length = table->length - sizeof(struct acpi_table_header);
+
+	status = acpi_tb_get_owner_id(table_index, &owner_id);
+	if (ACPI_FAILURE(status)) {
+		return_ACPI_STATUS(status);
+	}
+
+	/* Create and init a Root Node */
+
+	parse_root = acpi_ps_create_scope_op(aml_start);
+	if (!parse_root) {
+		return_ACPI_STATUS(AE_NO_MEMORY);
+	}
+
+	/* Create and initialize a new walk state */
+
+	walk_state = acpi_ds_create_walk_state(owner_id, NULL, NULL, NULL);
+	if (!walk_state) {
+		acpi_ps_free_op(parse_root);
+		return_ACPI_STATUS(AE_NO_MEMORY);
+	}
+
+	status = acpi_ds_init_aml_walk(walk_state, parse_root, NULL,
+				       aml_start, aml_length, NULL,
+				       (u8)pass_number);
+	if (ACPI_FAILURE(status)) {
+		acpi_ds_delete_walk_state(walk_state);
+		goto cleanup;
+	}
+
+	/* Found OSDT table, enable the namespace override feature */
+
+	if (ACPI_COMPARE_NAMESEG(table->signature, ACPI_SIG_OSDT) &&
+	    pass_number == ACPI_IMODE_LOAD_PASS1) {
+		walk_state->namespace_override = TRUE;
+	}
+
+	/* start_node is the default location to load the table */
+
+	if (start_node && start_node != acpi_gbl_root_node) {
+		status =
+		    acpi_ds_scope_stack_push(start_node, ACPI_TYPE_METHOD,
+					     walk_state);
+		if (ACPI_FAILURE(status)) {
+			acpi_ds_delete_walk_state(walk_state);
+			goto cleanup;
+		}
+	}
+
+	/* Parse the AML */
+
+	ACPI_DEBUG_PRINT((ACPI_DB_PARSE,
+			  "*PARSE* pass %u parse\n", pass_number));
+	acpi_ex_enter_interpreter();
+	status = acpi_ps_parse_aml(walk_state);
+	acpi_ex_exit_interpreter();
+
+cleanup:
+	acpi_ps_delete_parse_tree(parse_root);
+	return_ACPI_STATUS(status);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_ns_parse_table
+ *
+ * PARAMETERS:  table_desc      - An ACPI table descriptor for table to parse
+ *              start_node      - Where to enter the table into the namespace
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Parse AML within an ACPI table and return a tree of ops
+ *
+ ******************************************************************************/
+
+acpi_status
+acpi_ns_parse_table(u32 table_index, struct acpi_namespace_node *start_node)
+{
+	acpi_status status;
+
+	ACPI_FUNCTION_TRACE(ns_parse_table);
+
+	/*
+	 * Executes the AML table as one large control method.
+	 * The point of this is to execute any module-level code in-place
+	 * as the table is parsed. Some AML code depends on this behavior.
+	 *
+	 * Note: This causes the table to only have a single-pass parse.
+	 * However, this is compatible with other ACPI implementations.
+	 */
+	ACPI_DEBUG_PRINT_RAW((ACPI_DB_PARSE,
+			      "%s: **** Start table execution pass\n",
+			      ACPI_GET_FUNCTION_NAME));
+
+	status = acpi_ns_execute_table(table_index, start_node);
+
+	return_ACPI_STATUS(status);
+}