drm/amdkfd: Extend PM4 packets to support 8 SDMA

Extend map_queue and unmap_queue PM4 packets to support 8
SDMA engines. The new format is backward compatible.

Signed-off-by: Oak Zeng <Oak.Zeng@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

2 files changed, 34 insertions, 6 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c
index 2d5ddf199bd0..91da72d0d405 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c
@@ -161,6 +161,8 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 	packet->bitfields2.engine_sel =
 		engine_sel__mes_map_queues__compute_vi;
 	packet->bitfields2.gws_control_queue = q->gws ? 1 : 0;
+	packet->bitfields2.extended_engine_sel =
+		extended_engine_sel__mes_map_queues__legacy_engine_sel;
 	packet->bitfields2.queue_type =
 		queue_type__mes_map_queues__normal_compute_vi;
 
@@ -176,9 +178,15 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 		break;
 	case KFD_QUEUE_TYPE_SDMA:
 	case KFD_QUEUE_TYPE_SDMA_XGMI:
-		packet->bitfields2.engine_sel = q->properties.sdma_engine_id +
-				engine_sel__mes_map_queues__sdma0_vi;
 		use_static = false; /* no static queues under SDMA */
+		if (q->properties.sdma_engine_id < 2)
+			packet->bitfields2.engine_sel = q->properties.sdma_engine_id +
+				engine_sel__mes_map_queues__sdma0_vi;
+		else {
+			packet->bitfields2.extended_engine_sel =
+				extended_engine_sel__mes_map_queues__sdma0_to_7_sel;
+			packet->bitfields2.engine_sel = q->properties.sdma_engine_id;
+		}
 		break;
 	default:
 		WARN(1, "queue type %d", q->properties.type);
@@ -218,13 +226,23 @@ static int pm_unmap_queues_v9(struct packet_manager *pm, uint32_t *buffer,
 	switch (type) {
 	case KFD_QUEUE_TYPE_COMPUTE:
 	case KFD_QUEUE_TYPE_DIQ:
+		packet->bitfields2.extended_engine_sel =
+			extended_engine_sel__mes_unmap_queues__legacy_engine_sel;
 		packet->bitfields2.engine_sel =
 			engine_sel__mes_unmap_queues__compute;
 		break;
 	case KFD_QUEUE_TYPE_SDMA:
 	case KFD_QUEUE_TYPE_SDMA_XGMI:
-		packet->bitfields2.engine_sel =
-			engine_sel__mes_unmap_queues__sdma0 + sdma_engine;
+		if (sdma_engine < 2) {
+			packet->bitfields2.extended_engine_sel =
+				extended_engine_sel__mes_unmap_queues__legacy_engine_sel;
+			packet->bitfields2.engine_sel =
+				engine_sel__mes_unmap_queues__sdma0 + sdma_engine;
+		} else {
+			packet->bitfields2.extended_engine_sel =
+				extended_engine_sel__mes_unmap_queues__sdma0_to_7_sel;
+			packet->bitfields2.engine_sel = sdma_engine;
+		}
 		break;
 	default:
 		WARN(1, "queue type %d", type);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
index e3e21404cfa0..44ed94239513 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
@@ -260,6 +260,10 @@ enum mes_map_queues_engine_sel_enum {
 	engine_sel__mes_map_queues__sdma1_vi = 3
 };
 
+enum mes_map_queues_extended_engine_sel_enum {
+	extended_engine_sel__mes_map_queues__legacy_engine_sel = 0,
+	extended_engine_sel__mes_map_queues__sdma0_to_7_sel = 1
+};
 
 struct pm4_mes_map_queues {
 	union {
@@ -269,7 +273,8 @@ struct pm4_mes_map_queues {
 
 	union {
 		struct {
-			uint32_t reserved1:4;
+			uint32_t reserved1:2;
+			enum mes_map_queues_extended_engine_sel_enum extended_engine_sel:2;
 			enum mes_map_queues_queue_sel_enum queue_sel:2;
 			uint32_t reserved5:6;
 			uint32_t gws_control_queue:1;
@@ -382,6 +387,11 @@ enum mes_unmap_queues_engine_sel_enum {
 	engine_sel__mes_unmap_queues__sdmal = 3
 };
 
+enum mes_unmap_queues_extended_engine_sel_enum {
+	extended_engine_sel__mes_unmap_queues__legacy_engine_sel = 0,
+	extended_engine_sel__mes_unmap_queues__sdma0_to_7_sel = 1
+};
+
 struct pm4_mes_unmap_queues {
 	union {
 		union PM4_MES_TYPE_3_HEADER   header;            /* header */
@@ -391,7 +401,7 @@ struct pm4_mes_unmap_queues {
 	union {
 		struct {
 			enum mes_unmap_queues_action_enum action:2;
-			uint32_t reserved1:2;
+			enum mes_unmap_queues_extended_engine_sel_enum extended_engine_sel:2;
 			enum mes_unmap_queues_queue_sel_enum queue_sel:2;
 			uint32_t reserved2:20;
 			enum mes_unmap_queues_engine_sel_enum engine_sel:3;