Merge tag 'amd-drm-fixes-6.12-2024-10-08' of...

Merge tag 'amd-drm-fixes-6.12-2024-10-08' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.12-2024-10-08:

amdgpu:
- Fix invalid UBSAN warnings
- Fix artifacts in MPO transitions
- Hibernation fix

amdkfd:
- Fix an eviction fence leak

radeon:
- Add late register for connectors
- Always set GEM function pointers
Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241008142831.3739244-1-alexander.deucher@amd.com

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c

@@ -1439,8 +1439,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	list_add_tail(&vm->vm_list_node,
 			&(vm->process_info->vm_list_head));
 	vm->process_info->n_vms++;
-
-	*ef = dma_fence_get(&vm->process_info->eviction_fence->base);
+	if (ef)
+		*ef = dma_fence_get(&vm->process_info->eviction_fence->base);
 	mutex_unlock(&vm->process_info->lock);
 
 	return 0;

drivers/gpu/drm/amd/amdkfd/kfd_process.c

@@ -1702,12 +1702,15 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 
 	ret = amdgpu_amdkfd_gpuvm_acquire_process_vm(dev->adev, avm,
 						     &p->kgd_process_info,
-						     &ef);
+						     p->ef ? NULL : &ef);
 	if (ret) {
 		dev_err(dev->adev->dev, "Failed to create process VM object\n");
 		return ret;
 	}
-	RCU_INIT_POINTER(p->ef, ef);
+
+	if (!p->ef)
+		RCU_INIT_POINTER(p->ef, ef);
+
 	pdd->drm_priv = drm_file->private_data;
 
 	ret = kfd_process_device_reserve_ib_mem(pdd);

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c

@@ -2972,10 +2972,11 @@ static int dm_suspend(void *handle)
 
 	hpd_rx_irq_work_suspend(dm);
 
-	if (adev->dm.dc->caps.ips_support)
-		dc_allow_idle_optimizations(adev->dm.dc, true);
-
 	dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D3);
+
+	if (dm->dc->caps.ips_support && adev->in_s0ix)
+		dc_allow_idle_optimizations(dm->dc, true);
+
 	dc_dmub_srv_set_power_state(dm->dc->ctx->dmub_srv, DC_ACPI_CM_POWER_STATE_D3);
 
 	return 0;

drivers/gpu/drm/amd/display/dc/core/dc.c

@@ -5065,11 +5065,26 @@ static bool update_planes_and_stream_v3(struct dc *dc,
 	return true;
 }
 
+static void clear_update_flags(struct dc_surface_update *srf_updates,
+	int surface_count, struct dc_stream_state *stream)
+{
+	int i;
+
+	if (stream)
+		stream->update_flags.raw = 0;
+
+	for (i = 0; i < surface_count; i++)
+		if (srf_updates[i].surface)
+			srf_updates[i].surface->update_flags.raw = 0;
+}
+
 bool dc_update_planes_and_stream(struct dc *dc,
 		struct dc_surface_update *srf_updates, int surface_count,
 		struct dc_stream_state *stream,
 		struct dc_stream_update *stream_update)
 {
+	bool ret = false;
+
 	dc_exit_ips_for_hw_access(dc);
 	/*
 	 * update planes and stream version 3 separates FULL and FAST updates
@@ -5086,10 +5101,16 @@ bool dc_update_planes_and_stream(struct dc *dc,
 	 * features as they are now transparent to the new sequence.
 	 */
 	if (dc->ctx->dce_version >= DCN_VERSION_4_01)
-		return update_planes_and_stream_v3(dc, srf_updates,
+		ret = update_planes_and_stream_v3(dc, srf_updates,
 				surface_count, stream, stream_update);
-	return update_planes_and_stream_v2(dc, srf_updates,
+	else
+		ret = update_planes_and_stream_v2(dc, srf_updates,
 			surface_count, stream, stream_update);
+
+	if (ret)
+		clear_update_flags(srf_updates, surface_count, stream);
+
+	return ret;
 }
 
 void dc_commit_updates_for_stream(struct dc *dc,
@@ -5099,6 +5120,8 @@ void dc_commit_updates_for_stream(struct dc *dc,
 		struct dc_stream_update *stream_update,
 		struct dc_state *state)
 {
+	bool ret = false;
+
 	dc_exit_ips_for_hw_access(dc);
 	/* TODO: Since change commit sequence can have a huge impact,
 	 * we decided to only enable it for DCN3x. However, as soon as
@@ -5106,17 +5129,17 @@ void dc_commit_updates_for_stream(struct dc *dc,
 	 * the new sequence for all ASICs.
 	 */
 	if (dc->ctx->dce_version >= DCN_VERSION_4_01) {
-		update_planes_and_stream_v3(dc, srf_updates, surface_count,
+		ret = update_planes_and_stream_v3(dc, srf_updates, surface_count,
 				stream, stream_update);
-		return;
-	}
-	if (dc->ctx->dce_version >= DCN_VERSION_3_2) {
-		update_planes_and_stream_v2(dc, srf_updates, surface_count,
+	} else if (dc->ctx->dce_version >= DCN_VERSION_3_2) {
+		ret = update_planes_and_stream_v2(dc, srf_updates, surface_count,
 				stream, stream_update);
-		return;
-	}
-	update_planes_and_stream_v1(dc, srf_updates, surface_count, stream,
-			stream_update, state);
+	} else
+		ret = update_planes_and_stream_v1(dc, srf_updates, surface_count, stream,
+				stream_update, state);
+
+	if (ret)
+		clear_update_flags(srf_updates, surface_count, stream);
 }
 
 uint8_t dc_get_current_stream_count(struct dc *dc)

drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h

@@ -60,7 +60,7 @@ struct vi_dpm_level {
 
 struct vi_dpm_table {
 	uint32_t count;
-	struct vi_dpm_level dpm_level[] __counted_by(count);
+	struct vi_dpm_level dpm_level[];
 };
 
 #define PCIE_PERF_REQ_REMOVE_REGISTRY   0
@@ -91,7 +91,7 @@ struct phm_set_power_state_input {
 
 struct phm_clock_array {
 	uint32_t count;
-	uint32_t values[] __counted_by(count);
+	uint32_t values[];
 };
 
 struct phm_clock_voltage_dependency_record {
@@ -123,7 +123,7 @@ struct phm_acpclock_voltage_dependency_record {
 
 struct phm_clock_voltage_dependency_table {
 	uint32_t count;
-	struct phm_clock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_clock_voltage_dependency_record entries[];
 };
 
 struct phm_phase_shedding_limits_record {
@@ -140,7 +140,7 @@ struct phm_uvd_clock_voltage_dependency_record {
 
 struct phm_uvd_clock_voltage_dependency_table {
 	uint8_t count;
-	struct phm_uvd_clock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_uvd_clock_voltage_dependency_record entries[];
 };
 
 struct phm_acp_clock_voltage_dependency_record {
@@ -150,7 +150,7 @@ struct phm_acp_clock_voltage_dependency_record {
 
 struct phm_acp_clock_voltage_dependency_table {
 	uint32_t count;
-	struct phm_acp_clock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_acp_clock_voltage_dependency_record entries[];
 };
 
 struct phm_vce_clock_voltage_dependency_record {
@@ -161,32 +161,32 @@ struct phm_vce_clock_voltage_dependency_record {
 
 struct phm_phase_shedding_limits_table {
 	uint32_t count;
-	struct phm_phase_shedding_limits_record  entries[] __counted_by(count);
+	struct phm_phase_shedding_limits_record  entries[];
 };
 
 struct phm_vceclock_voltage_dependency_table {
 	uint8_t count;
-	struct phm_vceclock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_vceclock_voltage_dependency_record entries[];
 };
 
 struct phm_uvdclock_voltage_dependency_table {
 	uint8_t count;
-	struct phm_uvdclock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_uvdclock_voltage_dependency_record entries[];
 };
 
 struct phm_samuclock_voltage_dependency_table {
 	uint8_t count;
-	struct phm_samuclock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_samuclock_voltage_dependency_record entries[];
 };
 
 struct phm_acpclock_voltage_dependency_table {
 	uint32_t count;
-	struct phm_acpclock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_acpclock_voltage_dependency_record entries[];
 };
 
 struct phm_vce_clock_voltage_dependency_table {
 	uint8_t count;
-	struct phm_vce_clock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_vce_clock_voltage_dependency_record entries[];
 };
 
 
@@ -393,7 +393,7 @@ union phm_cac_leakage_record {
 
 struct phm_cac_leakage_table {
 	uint32_t count;
-	union phm_cac_leakage_record entries[] __counted_by(count);
+	union phm_cac_leakage_record entries[];
 };
 
 struct phm_samu_clock_voltage_dependency_record {
@@ -404,7 +404,7 @@ struct phm_samu_clock_voltage_dependency_record {
 
 struct phm_samu_clock_voltage_dependency_table {
 	uint8_t count;
-	struct phm_samu_clock_voltage_dependency_record entries[] __counted_by(count);
+	struct phm_samu_clock_voltage_dependency_record entries[];
 };
 
 struct phm_cac_tdp_table {

drivers/gpu/drm/radeon/atombios_dp.c

@@ -228,10 +228,8 @@ void radeon_dp_aux_init(struct radeon_connector *radeon_connector)
 {
 	struct drm_device *dev = radeon_connector->base.dev;
 	struct radeon_device *rdev = dev->dev_private;
-	int ret;
 
 	radeon_connector->ddc_bus->rec.hpd = radeon_connector->hpd.hpd;
-	radeon_connector->ddc_bus->aux.dev = radeon_connector->base.kdev;
 	radeon_connector->ddc_bus->aux.drm_dev = radeon_connector->base.dev;
 	if (ASIC_IS_DCE5(rdev)) {
 		if (radeon_auxch)
@@ -242,11 +240,8 @@ void radeon_dp_aux_init(struct radeon_connector *radeon_connector)
 		radeon_connector->ddc_bus->aux.transfer = radeon_dp_aux_transfer_atom;
 	}
 
-	ret = drm_dp_aux_register(&radeon_connector->ddc_bus->aux);
-	if (!ret)
-		radeon_connector->ddc_bus->has_aux = true;
-
-	WARN(ret, "drm_dp_aux_register() failed with error %d\n", ret);
+	drm_dp_aux_init(&radeon_connector->ddc_bus->aux);
+	radeon_connector->ddc_bus->has_aux = true;
 }
 
 /***** general DP utility functions *****/

drivers/gpu/drm/radeon/radeon_connectors.c

@@ -1786,6 +1786,20 @@ static enum drm_mode_status radeon_dp_mode_valid(struct drm_connector *connector
 	return MODE_OK;
 }
 
+static int
+radeon_connector_late_register(struct drm_connector *connector)
+{
+	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
+	int r = 0;
+
+	if (radeon_connector->ddc_bus->has_aux) {
+		radeon_connector->ddc_bus->aux.dev = radeon_connector->base.kdev;
+		r = drm_dp_aux_register(&radeon_connector->ddc_bus->aux);
+	}
+
+	return r;
+}
+
 static const struct drm_connector_helper_funcs radeon_dp_connector_helper_funcs = {
 	.get_modes = radeon_dp_get_modes,
 	.mode_valid = radeon_dp_mode_valid,
@@ -1800,6 +1814,7 @@ static const struct drm_connector_funcs radeon_dp_connector_funcs = {
 	.early_unregister = radeon_connector_unregister,
 	.destroy = radeon_connector_destroy,
 	.force = radeon_dvi_force,
+	.late_register = radeon_connector_late_register,
 };
 
 static const struct drm_connector_funcs radeon_edp_connector_funcs = {
@@ -1810,6 +1825,7 @@ static const struct drm_connector_funcs radeon_edp_connector_funcs = {
 	.early_unregister = radeon_connector_unregister,
 	.destroy = radeon_connector_destroy,
 	.force = radeon_dvi_force,
+	.late_register = radeon_connector_late_register,
 };
 
 static const struct drm_connector_funcs radeon_lvds_bridge_connector_funcs = {
@@ -1820,6 +1836,7 @@ static const struct drm_connector_funcs radeon_lvds_bridge_connector_funcs = {
 	.early_unregister = radeon_connector_unregister,
 	.destroy = radeon_connector_destroy,
 	.force = radeon_dvi_force,
+	.late_register = radeon_connector_late_register,
 };
 
 void

drivers/gpu/drm/radeon/radeon_gem.c

@@ -44,8 +44,6 @@ struct sg_table *radeon_gem_prime_get_sg_table(struct drm_gem_object *obj);
 int radeon_gem_prime_pin(struct drm_gem_object *obj);
 void radeon_gem_prime_unpin(struct drm_gem_object *obj);
 
-const struct drm_gem_object_funcs radeon_gem_object_funcs;
-
 static vm_fault_t radeon_gem_fault(struct vm_fault *vmf)
 {
 	struct ttm_buffer_object *bo = vmf->vma->vm_private_data;
@@ -132,7 +130,6 @@ int radeon_gem_object_create(struct radeon_device *rdev, unsigned long size,
 		return r;
 	}
 	*obj = &robj->tbo.base;
-	(*obj)->funcs = &radeon_gem_object_funcs;
 	robj->pid = task_pid_nr(current);
 
 	mutex_lock(&rdev->gem.mutex);

drivers/gpu/drm/radeon/radeon_object.c

@@ -151,6 +151,7 @@ int radeon_bo_create(struct radeon_device *rdev,
 	if (bo == NULL)
 		return -ENOMEM;
 	drm_gem_private_object_init(rdev_to_drm(rdev), &bo->tbo.base, size);
+	bo->tbo.base.funcs = &radeon_gem_object_funcs;
 	bo->rdev = rdev;
 	bo->surface_reg = -1;
 	INIT_LIST_HEAD(&bo->list);