libhsakmt: add event age tracking

Keeping last signaled event age to avoid race conditions
for HSA_EVENTTYPE_SIGNAL when event age init value is non-zero.

Change-Id: Ifb9a11a6868e5762a9f92f579e45a0a2c8fa1017
Signed-off-by: James Zhu <James.Zhu@amd.com>


[ROCm/ROCR-Runtime commit: a0cbf90b90]

projects/rocr-runtime/include/hsakmt.h

@@ -243,6 +243,25 @@ hsaKmtWaitOnEvent(
     HSAuint32   Milliseconds    //IN
     );
 
+/**
+  Checks the current state of the event object. If the object's state is
+  nonsignaled, the calling thread enters the wait state. event_age can
+  help avoiding race conditions.
+
+ The function returns when one of the following occurs:
+- The specified event object is in the signaled state.
+- The time-out interval elapses.
+- Tracking event age
+*/
+
+HSAKMT_STATUS
+HSAKMTAPI
+hsaKmtWaitOnEvent_Ext(
+    HsaEvent*   Event,          //IN
+    HSAuint32   Milliseconds,   //IN
+    uint64_t   *event_age       //IN/OUT
+    );
+
 /**
   Checks the current state of multiple event objects.
 
@@ -264,6 +283,30 @@ hsaKmtWaitOnMultipleEvents(
     HSAuint32   Milliseconds    //IN
     );
 
+/**
+  Checks the current state of multiple event objects.
+  event_age can help avoiding race conditions.
+
+ The function returns when one of the following occurs:
+- Either any one or all of the specified objects are in the signaled state
+  - if "WaitOnAll" is "true" the function returns when the state of all
+    objects in array is signaled
+  - if "WaitOnAll" is "false" the function returns when the state of any
+    one of the objects is set to signaled
+- The time-out interval elapses.
+- Tracking event age
+*/
+
+HSAKMT_STATUS
+HSAKMTAPI
+hsaKmtWaitOnMultipleEvents_Ext(
+    HsaEvent*   Events[],       //IN
+    HSAuint32   NumEvents,      //IN
+    bool        WaitOnAll,      //IN
+    HSAuint32   Milliseconds,   //IN
+    uint64_t   *event_age       //IN/OUT
+    );
+
 /**
   new TEMPORARY function definition - to be used only on "Triniti + Southern Islands" platform
   If used on other platforms the function will return HSAKMT_STATUS_ERROR

projects/rocr-runtime/include/linux/kfd_ioctl.h

@@ -39,9 +39,10 @@
  * - 1.11 - Add unified memory for ctx save/restore area
  * - 1.12 - Add DMA buf export ioctl
  * - 1.13 - Add debugger API
+ * - 1.14 - Update kfd_event_data
  */
 #define KFD_IOCTL_MAJOR_VERSION 1
-#define KFD_IOCTL_MINOR_VERSION 13
+#define KFD_IOCTL_MINOR_VERSION 14
 
 /*
  * Debug revision change log
@@ -979,12 +980,20 @@ struct kfd_hsa_hw_exception_data {
 	__u32 gpu_id;
 };
 
+/* hsa signal event data */
+struct kfd_hsa_signal_event_data {
+	__u64 last_event_age;	/* to and from KFD */
+};
+
 /* Event data */
 struct kfd_event_data {
 	union {
+		/* From KFD */
 		struct kfd_hsa_memory_exception_data memory_exception_data;
 		struct kfd_hsa_hw_exception_data hw_exception_data;
-	};				/* From KFD */
+		/* To and From KFD */
+		struct kfd_hsa_signal_event_data signal_event_data;
+	};
 	__u64 kfd_event_data_ext;	/* pointer to an extension structure
 					   for future exception types */
 	__u32 event_id;		/* to KFD */

projects/rocr-runtime/src/events.c

@@ -213,11 +213,17 @@ HSAKMT_STATUS HSAKMTAPI hsaKmtQueryEventState(HsaEvent *Event)
 
 HSAKMT_STATUS HSAKMTAPI hsaKmtWaitOnEvent(HsaEvent *Event,
 		HSAuint32 Milliseconds)
+{
+	return hsaKmtWaitOnEvent_Ext(Event, Milliseconds, NULL);
+}
+
+HSAKMT_STATUS HSAKMTAPI hsaKmtWaitOnEvent_Ext(HsaEvent *Event,
+		HSAuint32 Milliseconds, uint64_t *event_age)
 {
 	if (!Event)
 		return HSAKMT_STATUS_INVALID_HANDLE;
 
-	return hsaKmtWaitOnMultipleEvents(&Event, 1, true, Milliseconds);
+	return hsaKmtWaitOnMultipleEvents_Ext(&Event, 1, true, Milliseconds, event_age);
 }
 
 static HSAKMT_STATUS get_mem_info_svm_api(uint64_t address, uint32_t gpu_id)
@@ -365,6 +371,15 @@ HSAKMT_STATUS HSAKMTAPI hsaKmtWaitOnMultipleEvents(HsaEvent *Events[],
 						   HSAuint32 NumEvents,
 						   bool WaitOnAll,
 						   HSAuint32 Milliseconds)
+{
+	return hsaKmtWaitOnMultipleEvents_Ext(Events, NumEvents, WaitOnAll, Milliseconds, NULL);
+}
+
+HSAKMT_STATUS HSAKMTAPI hsaKmtWaitOnMultipleEvents_Ext(HsaEvent *Events[],
+						   HSAuint32 NumEvents,
+						   bool WaitOnAll,
+						   HSAuint32 Milliseconds,
+						   uint64_t *event_age)
 {
 	CHECK_KFD_OPEN();
 
@@ -376,6 +391,8 @@ HSAKMT_STATUS HSAKMTAPI hsaKmtWaitOnMultipleEvents(HsaEvent *Events[],
 	for (HSAuint32 i = 0; i < NumEvents; i++) {
 		event_data[i].event_id = Events[i]->EventId;
 		event_data[i].kfd_event_data_ext = (uint64_t)(uintptr_t)NULL;
+		if (event_age && Events[i]->EventData.EventType == HSA_EVENTTYPE_SIGNAL)
+			event_data[i].signal_event_data.last_event_age = event_age[i];
 	}
 
 	struct kfd_ioctl_wait_events_args args = {0};
@@ -413,6 +430,12 @@ HSAKMT_STATUS HSAKMTAPI hsaKmtWaitOnMultipleEvents(HsaEvent *Events[],
 		}
 	}
 out:
+
+	for (HSAuint32 i = 0; i < NumEvents; i++) {
+		if (event_age && Events[i]->EventData.EventType == HSA_EVENTTYPE_SIGNAL)
+			event_age[i] = event_data[i].signal_event_data.last_event_age;
+	}
+
 	free(event_data);
 
 	return result;