Add memory exception notification

Signed-off-by: Alexey Skidanov <Alexey.Skidanov@amd.com>
Signed-off-by: Oded Gabbay <oded.gabbay@amd.com>

3 files changed, 34 insertions, 5 deletions

diff --git a/src/events.c b/src/events.c
index be58253..5d6835e 100644
--- a/src/events.c
+++ b/src/events.c
@@ -89,6 +89,7 @@ hsaKmtCreateEvent(
 		e->EventData.HWData2 = (HSAuint64)&events_page[args.event_slot_index];
 
 	e->EventId = args.event_id;
+	e->EventData.EventType = EventDesc->EventType;
 	e->EventData.HWData1 = args.event_id;
 
 	e->EventData.HWData3 = args.event_trigger_data;
@@ -224,9 +225,10 @@ hsaKmtWaitOnMultipleEvents(
 	if (!Events)
 		return HSAKMT_STATUS_INVALID_HANDLE;
 
-	uint32_t *event_ids = malloc(NumEvents * sizeof(uint32_t));
+	struct kfd_event_data *event_data = malloc(NumEvents * sizeof(struct kfd_event_data));
 	for (HSAuint32 i = 0; i < NumEvents; i++) {
-		event_ids[i] = Events[i]->EventId;
+		event_data[i].event_id = Events[i]->EventId;
+		event_data[i].kfd_event_data_ext = (uint64_t)(uintptr_t)NULL;
 	}
 
 	struct kfd_ioctl_wait_events_args args;
@@ -235,7 +237,7 @@ hsaKmtWaitOnMultipleEvents(
 	args.wait_for_all = WaitOnAll;
 	args.timeout = Milliseconds;
 	args.num_events = NumEvents;
-	args.events_ptr = (uint64_t)(uintptr_t)event_ids;
+	args.events_ptr = (uint64_t)(uintptr_t)event_data;
 
 	HSAKMT_STATUS result;
 
@@ -247,9 +249,21 @@ hsaKmtWaitOnMultipleEvents(
 	}
 	else {
 		result = HSAKMT_STATUS_SUCCESS;
+		for (HSAuint32 i = 0; i < NumEvents; i++) {
+			if (Events[i]->EventData.EventType == HSA_EVENTTYPE_MEMORY) {
+				Events[i]->EventData.EventData.MemoryAccessFault.VirtualAddress = event_data[i].memory_exception_data.va;
+				result = gpuid_to_nodeid(event_data[i].memory_exception_data.gpu_id, &Events[i]->EventData.EventData.MemoryAccessFault.NodeId);
+				if (result != HSAKMT_STATUS_SUCCESS)
+					goto out;
+				Events[i]->EventData.EventData.MemoryAccessFault.Failure.NotPresent = event_data[i].memory_exception_data.failure.NotPresent;
+				Events[i]->EventData.EventData.MemoryAccessFault.Failure.ReadOnly = event_data[i].memory_exception_data.failure.ReadOnly;
+				Events[i]->EventData.EventData.MemoryAccessFault.Failure.NoExecute = event_data[i].memory_exception_data.failure.NoExecute;
+				Events[i]->EventData.EventData.MemoryAccessFault.Flags = HSA_EVENTID_MEMORY_FATAL_PROCESS;
+			}
+		}
 	}
-
-	free(event_ids);
+out:
+	free(event_data);
 
 	return result;
 }
diff --git a/src/libhsakmt.h b/src/libhsakmt.h
index fa8dc0e..0d73c8f 100644
--- a/src/libhsakmt.h
+++ b/src/libhsakmt.h
@@ -62,6 +62,7 @@ extern pthread_mutex_t hsakmt_mutex;
 #define MAX_NODES 8
 
 HSAKMT_STATUS validate_nodeid(uint32_t nodeid, uint32_t *gpu_id);
+HSAKMT_STATUS gpuid_to_nodeid(uint32_t gpu_id, uint32_t* node_id);
 uint16_t get_device_id_by_node(HSAuint32 node_id);
 
 extern int kmtIoctl(int fd, unsigned long request, void *arg);
diff --git a/src/topology.c b/src/topology.c
index 864be2b..903b6f7 100644
--- a/src/topology.c
+++ b/src/topology.c
@@ -698,6 +698,20 @@ validate_nodeid(uint32_t nodeid, uint32_t *gpu_id)
 }
 
 HSAKMT_STATUS
+gpuid_to_nodeid(uint32_t gpu_id, uint32_t* node_id){
+	uint64_t node_idx;
+	for(node_idx = 0; node_idx < system->NumNodes; node_idx++){
+		if (node[node_idx].gpu_id == gpu_id){
+			*node_id = node_idx;
+			return HSAKMT_STATUS_SUCCESS;
+		}
+	}
+
+	return HSAKMT_STATUS_INVALID_NODE_UNIT;
+
+}
+
+HSAKMT_STATUS
 HSAKMTAPI
 hsaKmtAcquireSystemProperties(
     HsaSystemProperties*  SystemProperties    //OUT