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, 589 insertions, 0 deletions

diff --git a/drivers/acpi/acpica/dbmethod.c b/drivers/acpi/acpica/dbmethod.c
new file mode 100644
index 000000000..889d13828
--- /dev/null
+++ b/drivers/acpi/acpica/dbmethod.c
@@ -0,0 +1,589 @@
+// SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
+/*******************************************************************************
+ *
+ * Module Name: dbmethod - Debug commands for control methods
+ *
+ ******************************************************************************/
+
+#include <acpi/acpi.h>
+#include "accommon.h"
+#include "acdispat.h"
+#include "acnamesp.h"
+#include "acdebug.h"
+#include "acparser.h"
+#include "acpredef.h"
+
+#define _COMPONENT          ACPI_CA_DEBUGGER
+ACPI_MODULE_NAME("dbmethod")
+
+/* Local prototypes */
+static acpi_status
+acpi_db_walk_for_execute(acpi_handle obj_handle,
+			 u32 nesting_level, void *context, void **return_value);
+
+static acpi_status acpi_db_evaluate_object(struct acpi_namespace_node *node);
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_set_method_breakpoint
+ *
+ * PARAMETERS:  location            - AML offset of breakpoint
+ *              walk_state          - Current walk info
+ *              op                  - Current Op (from parse walk)
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Set a breakpoint in a control method at the specified
+ *              AML offset
+ *
+ ******************************************************************************/
+
+void
+acpi_db_set_method_breakpoint(char *location,
+			      struct acpi_walk_state *walk_state,
+			      union acpi_parse_object *op)
+{
+	u32 address;
+	u32 aml_offset;
+
+	if (!op) {
+		acpi_os_printf("There is no method currently executing\n");
+		return;
+	}
+
+	/* Get and verify the breakpoint address */
+
+	address = strtoul(location, NULL, 16);
+	aml_offset = (u32)ACPI_PTR_DIFF(op->common.aml,
+					walk_state->parser_state.aml_start);
+	if (address <= aml_offset) {
+		acpi_os_printf("Breakpoint %X is beyond current address %X\n",
+			       address, aml_offset);
+	}
+
+	/* Save breakpoint in current walk */
+
+	walk_state->user_breakpoint = address;
+	acpi_os_printf("Breakpoint set at AML offset %X\n", address);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_set_method_call_breakpoint
+ *
+ * PARAMETERS:  op                  - Current Op (from parse walk)
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Set a breakpoint in a control method at the specified
+ *              AML offset
+ *
+ ******************************************************************************/
+
+void acpi_db_set_method_call_breakpoint(union acpi_parse_object *op)
+{
+
+	if (!op) {
+		acpi_os_printf("There is no method currently executing\n");
+		return;
+	}
+
+	acpi_gbl_step_to_next_call = TRUE;
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_set_method_data
+ *
+ * PARAMETERS:  type_arg        - L for local, A for argument
+ *              index_arg       - which one
+ *              value_arg       - Value to set.
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Set a local or argument for the running control method.
+ *              NOTE: only object supported is Number.
+ *
+ ******************************************************************************/
+
+void acpi_db_set_method_data(char *type_arg, char *index_arg, char *value_arg)
+{
+	char type;
+	u32 index;
+	u32 value;
+	struct acpi_walk_state *walk_state;
+	union acpi_operand_object *obj_desc;
+	acpi_status status;
+	struct acpi_namespace_node *node;
+
+	/* Validate type_arg */
+
+	acpi_ut_strupr(type_arg);
+	type = type_arg[0];
+	if ((type != 'L') && (type != 'A') && (type != 'N')) {
+		acpi_os_printf("Invalid SET operand: %s\n", type_arg);
+		return;
+	}
+
+	value = strtoul(value_arg, NULL, 16);
+
+	if (type == 'N') {
+		node = acpi_db_convert_to_node(index_arg);
+		if (!node) {
+			return;
+		}
+
+		if (node->type != ACPI_TYPE_INTEGER) {
+			acpi_os_printf("Can only set Integer nodes\n");
+			return;
+		}
+		obj_desc = node->object;
+		obj_desc->integer.value = value;
+		return;
+	}
+
+	/* Get the index and value */
+
+	index = strtoul(index_arg, NULL, 16);
+
+	walk_state = acpi_ds_get_current_walk_state(acpi_gbl_current_walk_list);
+	if (!walk_state) {
+		acpi_os_printf("There is no method currently executing\n");
+		return;
+	}
+
+	/* Create and initialize the new object */
+
+	obj_desc = acpi_ut_create_integer_object((u64)value);
+	if (!obj_desc) {
+		acpi_os_printf("Could not create an internal object\n");
+		return;
+	}
+
+	/* Store the new object into the target */
+
+	switch (type) {
+	case 'A':
+
+		/* Set a method argument */
+
+		if (index > ACPI_METHOD_MAX_ARG) {
+			acpi_os_printf("Arg%u - Invalid argument name\n",
+				       index);
+			goto cleanup;
+		}
+
+		status = acpi_ds_store_object_to_local(ACPI_REFCLASS_ARG,
+						       index, obj_desc,
+						       walk_state);
+		if (ACPI_FAILURE(status)) {
+			goto cleanup;
+		}
+
+		obj_desc = walk_state->arguments[index].object;
+
+		acpi_os_printf("Arg%u: ", index);
+		acpi_db_display_internal_object(obj_desc, walk_state);
+		break;
+
+	case 'L':
+
+		/* Set a method local */
+
+		if (index > ACPI_METHOD_MAX_LOCAL) {
+			acpi_os_printf
+			    ("Local%u - Invalid local variable name\n", index);
+			goto cleanup;
+		}
+
+		status = acpi_ds_store_object_to_local(ACPI_REFCLASS_LOCAL,
+						       index, obj_desc,
+						       walk_state);
+		if (ACPI_FAILURE(status)) {
+			goto cleanup;
+		}
+
+		obj_desc = walk_state->local_variables[index].object;
+
+		acpi_os_printf("Local%u: ", index);
+		acpi_db_display_internal_object(obj_desc, walk_state);
+		break;
+
+	default:
+
+		break;
+	}
+
+cleanup:
+	acpi_ut_remove_reference(obj_desc);
+}
+
+#ifdef ACPI_DISASSEMBLER
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_disassemble_aml
+ *
+ * PARAMETERS:  statements          - Number of statements to disassemble
+ *              op                  - Current Op (from parse walk)
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Display disassembled AML (ASL) starting from Op for the number
+ *              of statements specified.
+ *
+ ******************************************************************************/
+
+void acpi_db_disassemble_aml(char *statements, union acpi_parse_object *op)
+{
+	u32 num_statements = 8;
+
+	if (!op) {
+		acpi_os_printf("There is no method currently executing\n");
+		return;
+	}
+
+	if (statements) {
+		num_statements = strtoul(statements, NULL, 0);
+	}
+
+	acpi_dm_disassemble(NULL, op, num_statements);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_disassemble_method
+ *
+ * PARAMETERS:  name            - Name of control method
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Display disassembled AML (ASL) starting from Op for the number
+ *              of statements specified.
+ *
+ ******************************************************************************/
+
+acpi_status acpi_db_disassemble_method(char *name)
+{
+	acpi_status status;
+	union acpi_parse_object *op;
+	struct acpi_walk_state *walk_state;
+	union acpi_operand_object *obj_desc;
+	struct acpi_namespace_node *method;
+
+	method = acpi_db_convert_to_node(name);
+	if (!method) {
+		return (AE_BAD_PARAMETER);
+	}
+
+	if (method->type != ACPI_TYPE_METHOD) {
+		ACPI_ERROR((AE_INFO, "%s (%s): Object must be a control method",
+			    name, acpi_ut_get_type_name(method->type)));
+		return (AE_BAD_PARAMETER);
+	}
+
+	obj_desc = method->object;
+
+	op = acpi_ps_create_scope_op(obj_desc->method.aml_start);
+	if (!op) {
+		return (AE_NO_MEMORY);
+	}
+
+	/* Create and initialize a new walk state */
+
+	walk_state = acpi_ds_create_walk_state(0, op, NULL, NULL);
+	if (!walk_state) {
+		return (AE_NO_MEMORY);
+	}
+
+	status = acpi_ds_init_aml_walk(walk_state, op, NULL,
+				       obj_desc->method.aml_start,
+				       obj_desc->method.aml_length, NULL,
+				       ACPI_IMODE_LOAD_PASS1);
+	if (ACPI_FAILURE(status)) {
+		return (status);
+	}
+
+	status = acpi_ut_allocate_owner_id(&obj_desc->method.owner_id);
+	if (ACPI_FAILURE(status)) {
+		return (status);
+	}
+
+	walk_state->owner_id = obj_desc->method.owner_id;
+
+	/* Push start scope on scope stack and make it current */
+
+	status = acpi_ds_scope_stack_push(method, method->type, walk_state);
+	if (ACPI_FAILURE(status)) {
+		return (status);
+	}
+
+	/* Parse the entire method AML including deferred operators */
+
+	walk_state->parse_flags &= ~ACPI_PARSE_DELETE_TREE;
+	walk_state->parse_flags |= ACPI_PARSE_DISASSEMBLE;
+
+	status = acpi_ps_parse_aml(walk_state);
+	if (ACPI_FAILURE(status)) {
+		return (status);
+	}
+
+	(void)acpi_dm_parse_deferred_ops(op);
+
+	/* Now we can disassemble the method */
+
+	acpi_gbl_dm_opt_verbose = FALSE;
+	acpi_dm_disassemble(NULL, op, 0);
+	acpi_gbl_dm_opt_verbose = TRUE;
+
+	acpi_ps_delete_parse_tree(op);
+
+	/* Method cleanup */
+
+	acpi_ns_delete_namespace_subtree(method);
+	acpi_ns_delete_namespace_by_owner(obj_desc->method.owner_id);
+	acpi_ut_release_owner_id(&obj_desc->method.owner_id);
+	return (AE_OK);
+}
+#endif
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_evaluate_object
+ *
+ * PARAMETERS:  node                - Namespace node for the object
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Main execution function for the Evaluate/Execute/All debugger
+ *              commands.
+ *
+ ******************************************************************************/
+
+static acpi_status acpi_db_evaluate_object(struct acpi_namespace_node *node)
+{
+	char *pathname;
+	u32 i;
+	struct acpi_device_info *obj_info;
+	struct acpi_object_list param_objects;
+	union acpi_object params[ACPI_METHOD_NUM_ARGS];
+	struct acpi_buffer return_obj;
+	acpi_status status;
+
+	pathname = acpi_ns_get_external_pathname(node);
+	if (!pathname) {
+		return (AE_OK);
+	}
+
+	/* Get the object info for number of method parameters */
+
+	status = acpi_get_object_info(node, &obj_info);
+	if (ACPI_FAILURE(status)) {
+		ACPI_FREE(pathname);
+		return (status);
+	}
+
+	param_objects.pointer = NULL;
+	param_objects.count = 0;
+
+	if (obj_info->type == ACPI_TYPE_METHOD) {
+
+		/* Setup default parameters */
+
+		for (i = 0; i < obj_info->param_count; i++) {
+			params[i].type = ACPI_TYPE_INTEGER;
+			params[i].integer.value = 1;
+		}
+
+		param_objects.pointer = params;
+		param_objects.count = obj_info->param_count;
+	}
+
+	ACPI_FREE(obj_info);
+	return_obj.pointer = NULL;
+	return_obj.length = ACPI_ALLOCATE_BUFFER;
+
+	/* Do the actual method execution */
+
+	acpi_gbl_method_executing = TRUE;
+
+	status = acpi_evaluate_object(node, NULL, &param_objects, &return_obj);
+	acpi_gbl_method_executing = FALSE;
+
+	acpi_os_printf("%-32s returned %s\n", pathname,
+		       acpi_format_exception(status));
+	if (return_obj.length) {
+		acpi_os_printf("Evaluation of %s returned object %p, "
+			       "external buffer length %X\n",
+			       pathname, return_obj.pointer,
+			       (u32)return_obj.length);
+
+		acpi_db_dump_external_object(return_obj.pointer, 1);
+		acpi_os_printf("\n");
+	}
+
+	ACPI_FREE(pathname);
+
+	/* Ignore status from method execution */
+
+	return (AE_OK);
+
+	/* Update count, check if we have executed enough methods */
+
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_walk_for_execute
+ *
+ * PARAMETERS:  Callback from walk_namespace
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Batch execution function. Evaluates all "predefined" objects --
+ *              the nameseg begins with an underscore.
+ *
+ ******************************************************************************/
+
+static acpi_status
+acpi_db_walk_for_execute(acpi_handle obj_handle,
+			 u32 nesting_level, void *context, void **return_value)
+{
+	struct acpi_namespace_node *node =
+	    (struct acpi_namespace_node *)obj_handle;
+	struct acpi_db_execute_walk *info =
+	    (struct acpi_db_execute_walk *)context;
+	acpi_status status;
+	const union acpi_predefined_info *predefined;
+
+	predefined = acpi_ut_match_predefined_method(node->name.ascii);
+	if (!predefined) {
+		return (AE_OK);
+	}
+
+	if (node->type == ACPI_TYPE_LOCAL_SCOPE) {
+		return (AE_OK);
+	}
+
+	acpi_db_evaluate_object(node);
+
+	/* Ignore status from object evaluation */
+
+	status = AE_OK;
+
+	/* Update count, check if we have executed enough methods */
+
+	info->count++;
+	if (info->count >= info->max_count) {
+		status = AE_CTRL_TERMINATE;
+	}
+
+	return (status);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_walk_for_execute_all
+ *
+ * PARAMETERS:  Callback from walk_namespace
+ *
+ * RETURN:      Status
+ *
+ * DESCRIPTION: Batch execution function. Evaluates all objects whose path ends
+ *              with the nameseg "Info->NameSeg". Used for the "ALL" command.
+ *
+ ******************************************************************************/
+
+static acpi_status
+acpi_db_walk_for_execute_all(acpi_handle obj_handle,
+			     u32 nesting_level,
+			     void *context, void **return_value)
+{
+	struct acpi_namespace_node *node =
+	    (struct acpi_namespace_node *)obj_handle;
+	struct acpi_db_execute_walk *info =
+	    (struct acpi_db_execute_walk *)context;
+	acpi_status status;
+
+	if (!ACPI_COMPARE_NAMESEG(node->name.ascii, info->name_seg)) {
+		return (AE_OK);
+	}
+
+	if (node->type == ACPI_TYPE_LOCAL_SCOPE) {
+		return (AE_OK);
+	}
+
+	/* Now evaluate the input object (node) */
+
+	acpi_db_evaluate_object(node);
+
+	/* Ignore status from method execution */
+
+	status = AE_OK;
+
+	/* Update count of executed methods/objects */
+
+	info->count++;
+	return (status);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_evaluate_predefined_names
+ *
+ * PARAMETERS:  None
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Namespace batch execution. Execute predefined names in the
+ *              namespace, up to the max count, if specified.
+ *
+ ******************************************************************************/
+
+void acpi_db_evaluate_predefined_names(void)
+{
+	struct acpi_db_execute_walk info;
+
+	info.count = 0;
+	info.max_count = ACPI_UINT32_MAX;
+
+	/* Search all nodes in namespace */
+
+	(void)acpi_walk_namespace(ACPI_TYPE_ANY, ACPI_ROOT_OBJECT,
+				  ACPI_UINT32_MAX, acpi_db_walk_for_execute,
+				  NULL, (void *)&info, NULL);
+
+	acpi_os_printf("Evaluated %u predefined names in the namespace\n",
+		       info.count);
+}
+
+/*******************************************************************************
+ *
+ * FUNCTION:    acpi_db_evaluate_all
+ *
+ * PARAMETERS:  none_acpi_gbl_db_method_info
+ *
+ * RETURN:      None
+ *
+ * DESCRIPTION: Namespace batch execution. Implements the "ALL" command.
+ *              Execute all namepaths whose final nameseg matches the
+ *              input nameseg.
+ *
+ ******************************************************************************/
+
+void acpi_db_evaluate_all(char *name_seg)
+{
+	struct acpi_db_execute_walk info;
+
+	info.count = 0;
+	info.max_count = ACPI_UINT32_MAX;
+	ACPI_COPY_NAMESEG(info.name_seg, name_seg);
+	info.name_seg[ACPI_NAMESEG_SIZE] = 0;
+
+	/* Search all nodes in namespace */
+
+	(void)acpi_walk_namespace(ACPI_TYPE_ANY, ACPI_ROOT_OBJECT,
+				  ACPI_UINT32_MAX, acpi_db_walk_for_execute_all,
+				  NULL, (void *)&info, NULL);
+
+	acpi_os_printf("Evaluated %u names in the namespace\n", info.count);
+}