ACPICA: fix AML mutex re-entrancy
ACPI AML supports "serialized" methods which are protected
by an implicit mutex. The mutex is re-entrant for that AML thread
to allow recursion.
However, Linux implements notify() by creating a new AML thread.
So for systems where notify() re-enters a serialized method,
deadlock results.
The fix is to use the Linux thread_id as the key to allowing
re-entrancy, not the AML thread pointer.
http://bugzilla.kernel.org/show_bug.cgi?id=5534
Signed-off-by: Alexey Starikovskiy <alexey.y.starikovskiy@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
diff --git a/include/acpi/acinterp.h b/include/acpi/acinterp.h
index ce7c9d6..73967c8 100644
--- a/include/acpi/acinterp.h
+++ b/include/acpi/acinterp.h
@@ -253,7 +253,8 @@
void acpi_ex_release_all_mutexes(struct acpi_thread_state *thread);
-void acpi_ex_unlink_mutex(union acpi_operand_object *obj_desc);
+void acpi_ex_unlink_mutex(union acpi_operand_object *obj_desc,
+ struct acpi_thread_state *thread);
/*
* exprep - ACPI AML execution - prep utilities
diff --git a/include/acpi/acobject.h b/include/acpi/acobject.h
index 04e9735..5206d61 100644
--- a/include/acpi/acobject.h
+++ b/include/acpi/acobject.h
@@ -155,7 +155,7 @@
struct acpi_object_mutex {
ACPI_OBJECT_COMMON_HEADER u8 sync_level; /* 0-15, specified in Mutex() call */
u16 acquisition_depth; /* Allow multiple Acquires, same thread */
- struct acpi_thread_state *owner_thread; /* Current owner of the mutex */
+ acpi_thread_id owner_thread_id; /* Current owner of the mutex */
acpi_mutex os_mutex; /* Actual OS synchronization object */
union acpi_operand_object *prev; /* Link for list of acquired mutexes */
union acpi_operand_object *next; /* Link for list of acquired mutexes */