drm/amdgpu: Reduce dequeue retry timeout for gfx9 family

Dequeue retry timeout controls the interval between checks for unmet
conditions. On MI series, reduce this from 0x40 to 0x1 (~ 1 uS). The
cost of additional bandwidth consumed by CP when polling memory
shouldn't be substantial.

Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Reviewed-by: Jonathan Kim <jonathan.kim@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

8 files changed, 39 insertions, 37 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index 6e861d08d044..7e9f7a280c1b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -189,7 +189,7 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
 	.set_address_watch = kgd_gfx_aldebaran_set_address_watch,
 	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
 	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
-	.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
 	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
 	.hqd_reset = kgd_gfx_v9_hqd_reset,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index c820418e8ccd..ffbaa8bc5eea 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -415,7 +415,7 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
 	.set_address_watch = kgd_gfx_v9_set_address_watch,
 	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
 	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
-	.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
 	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
 	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
index 0c0998477598..89a45a9218f3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
@@ -541,8 +541,8 @@ const struct kfd2kgd_calls gc_9_4_3_kfd2kgd = {
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
 	.program_trap_handler_settings =
 				kgd_gfx_v9_program_trap_handler_settings,
-	.build_grace_period_packet_info =
-				kgd_gfx_v9_build_grace_period_packet_info,
+	.build_dequeue_wait_counts_packet_info =
+				kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
 	.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
 	.disable_debug_trap = kgd_gfx_v9_4_3_disable_debug_trap,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 2887b6f3eaa2..04ef0ca10541 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -1021,25 +1021,25 @@ void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
 	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
 }
 
-void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
+void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
 						uint32_t wait_times,
-						uint32_t grace_period,
+						uint32_t sch_wave,
+						uint32_t que_sleep,
 						uint32_t *reg_offset,
 						uint32_t *reg_data)
 {
 	*reg_data = wait_times;
 
-	/*
-	 * The CP cannont handle a 0 grace period input and will result in
-	 * an infinite grace period being set so set to 1 to prevent this.
-	 */
-	if (grace_period == 0)
-		grace_period = 1;
-
-	*reg_data = REG_SET_FIELD(*reg_data,
-			CP_IQ_WAIT_TIME2,
-			SCH_WAVE,
-			grace_period);
+	if (sch_wave)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				SCH_WAVE,
+				sch_wave);
+	if (que_sleep)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				QUE_SLEEP,
+				que_sleep);
 
 	*reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
 }
@@ -1115,7 +1115,7 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
 	.set_address_watch = kgd_gfx_v10_set_address_watch,
 	.clear_address_watch = kgd_gfx_v10_clear_address_watch,
 	.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
-	.build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info,
 	.program_trap_handler_settings = program_trap_handler_settings,
 	.hqd_get_pq_addr = kgd_gfx_v10_hqd_get_pq_addr,
 	.hqd_reset = kgd_gfx_v10_hqd_reset,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
index db577c2a847a..a4c607c88178 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -51,9 +51,10 @@ uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
 void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
 				uint32_t *wait_times,
 				uint32_t inst);
-void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev,
+void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
 					       uint32_t wait_times,
-					       uint32_t grace_period,
+					       uint32_t sch_wave,
+					       uint32_t que_sleep,
 					       uint32_t *reg_offset,
 					       uint32_t *reg_data);
 uint64_t kgd_gfx_v10_hqd_get_pq_addr(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index ac9ad505f9d7..6d08bc2781a3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -673,7 +673,7 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
 	.set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3,
 	.program_trap_handler_settings = program_trap_handler_settings_v10_3,
 	.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
-	.build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info,
 	.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
 	.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
 	.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 84135eb90660..088d09cc7a72 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -1077,25 +1077,25 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev,
 				adev->gfx.cu_info.max_waves_per_simd;
 }
 
-void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev,
+void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
 		uint32_t wait_times,
-		uint32_t grace_period,
+		uint32_t sch_wave,
+		uint32_t que_sleep,
 		uint32_t *reg_offset,
 		uint32_t *reg_data)
 {
 	*reg_data = wait_times;
 
-	/*
-	 * The CP cannot handle a 0 grace period input and will result in
-	 * an infinite grace period being set so set to 1 to prevent this.
-	 */
-	if (grace_period == 0)
-		grace_period = 1;
-
-	*reg_data = REG_SET_FIELD(*reg_data,
-			CP_IQ_WAIT_TIME2,
-			SCH_WAVE,
-			grace_period);
+	if (sch_wave)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				SCH_WAVE,
+				sch_wave);
+	if (que_sleep)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				QUE_SLEEP,
+				que_sleep);
 
 	*reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
 }
@@ -1255,7 +1255,7 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
 	.set_address_watch = kgd_gfx_v9_set_address_watch,
 	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
 	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
-	.build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
 	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
 	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index 90c8fa13d519..704452ca62f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -97,9 +97,10 @@ uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
 void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev,
 				uint32_t *wait_times,
 				uint32_t inst);
-void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev,
+void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
 					       uint32_t wait_times,
-					       uint32_t grace_period,
+					       uint32_t sch_wave,
+					       uint32_t que_sleep,
 					       uint32_t *reg_offset,
 					       uint32_t *reg_data);
 uint64_t kgd_gfx_v9_hqd_get_pq_addr(struct amdgpu_device *adev,