Merge tag 'amd-drm-fixes-6.4-2023-05-03' of...

Merge tag 'amd-drm-fixes-6.4-2023-05-03' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-fixes-6.4-2023-05-03:

amdgpu:
- GPU reset fixes
- Doorbell fix when resizing BARs
- Fix spurious warnings in gmc
- Locking fix for AMDGPU_SCHED IOCTL
- SR-IOV fix
- DCN 3.1.4 fix
- DCN 3.2 fix
- Fix job cleanup when CS is aborted
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230504034018.7950-1-alexander.deucher@amd.com

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

@@ -1276,7 +1276,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 		r = drm_sched_job_add_dependency(&leader->base, fence);
 		if (r) {
 			dma_fence_put(fence);
-			goto error_cleanup;
+			return r;
 		}
 	}
 
@@ -1303,7 +1303,8 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	}
 	if (r) {
 		r = -EAGAIN;
-		goto error_unlock;
+		mutex_unlock(&p->adev->notifier_lock);
+		return r;
 	}
 
 	p->fence = dma_fence_get(&leader->base.s_fence->finished);
@@ -1350,14 +1351,6 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	mutex_unlock(&p->adev->notifier_lock);
 	mutex_unlock(&p->bo_list->bo_list_mutex);
 	return 0;
-
-error_unlock:
-	mutex_unlock(&p->adev->notifier_lock);
-
-error_cleanup:
-	for (i = 0; i < p->gang_size; ++i)
-		drm_sched_job_cleanup(&p->jobs[i]->base);
-	return r;
 }
 
 /* Cleanup the parser structure */

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

@@ -3578,6 +3578,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	int r, i;
 	bool px = false;
 	u32 max_MBps;
+	int tmp;
 
 	adev->shutdown = false;
 	adev->flags = flags;
@@ -3799,7 +3800,13 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 				}
 			}
 		} else {
+			tmp = amdgpu_reset_method;
+			/* It should do a default reset when loading or reloading the driver,
+			 * regardless of the module parameter reset_method.
+			 */
+			amdgpu_reset_method = AMD_RESET_METHOD_NONE;
 			r = amdgpu_asic_reset(adev);
+			amdgpu_reset_method = tmp;
 			if (r) {
 				dev_err(adev->dev, "asic reset on init failed\n");
 				goto failed;

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

@@ -38,6 +38,7 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
 {
 	struct fd f = fdget(fd);
 	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_ctx_mgr *mgr;
 	struct amdgpu_ctx *ctx;
 	uint32_t id;
 	int r;
@@ -51,8 +52,11 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
 		return r;
 	}
 
-	idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id)
+	mgr = &fpriv->ctx_mgr;
+	mutex_lock(&mgr->lock);
+	idr_for_each_entry(&mgr->ctx_handles, ctx, id)
 		amdgpu_ctx_priority_override(ctx, priority);
+	mutex_unlock(&mgr->lock);
 
 	fdput(f);
 	return 0;

drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c

@@ -1143,7 +1143,6 @@ static int gmc_v10_0_hw_fini(void *handle)
 		return 0;
 	}
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 
 	return 0;

drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c

@@ -951,7 +951,6 @@ static int gmc_v11_0_hw_fini(void *handle)
 		return 0;
 	}
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v11_0_gart_disable(adev);
 

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c

@@ -1999,7 +1999,6 @@ static int gmc_v9_0_hw_fini(void *handle)
 	if (adev->mmhub.funcs->update_power_gating)
 		adev->mmhub.funcs->update_power_gating(adev, false);
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 
 	return 0;

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

@@ -531,13 +531,6 @@ static void nv_program_aspm(struct amdgpu_device *adev)
 
 }
 
-static void nv_enable_doorbell_aperture(struct amdgpu_device *adev,
-					bool enable)
-{
-	adev->nbio.funcs->enable_doorbell_aperture(adev, enable);
-	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable);
-}
-
 const struct amdgpu_ip_block_version nv_common_ip_block =
 {
 	.type = AMD_IP_BLOCK_TYPE_COMMON,
@@ -999,6 +992,11 @@ static int nv_common_late_init(void *handle)
 		}
 	}
 
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
 	return 0;
 }
 
@@ -1038,7 +1036,7 @@ static int nv_common_hw_init(void *handle)
 	if (adev->nbio.funcs->remap_hdp_registers && !amdgpu_sriov_vf(adev))
 		adev->nbio.funcs->remap_hdp_registers(adev);
 	/* enable the doorbell aperture */
-	nv_enable_doorbell_aperture(adev, true);
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
 
 	return 0;
 }
@@ -1047,8 +1045,13 @@ static int nv_common_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	/* disable the doorbell aperture */
-	nv_enable_doorbell_aperture(adev, false);
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because nv_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
 
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c

@@ -510,9 +510,6 @@ static int sdma_v6_0_gfx_resume(struct amdgpu_device *adev)
 		       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
 
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
-		if (amdgpu_sriov_vf(adev))
-			rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 1);
-		else
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, WPTR_POLL_ENABLE, 0);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_QUEUE0_RB_CNTL, F32_WPTR_POLL_ENABLE, 1);
 

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

@@ -40,7 +40,7 @@ static bool sienna_cichlid_is_mode2_default(struct amdgpu_reset_control *reset_c
 	    adev->pm.fw_version >= 0x3a5500 && !amdgpu_sriov_vf(adev))
 		return true;
 #endif
-	return false;
+	return amdgpu_reset_method == AMD_RESET_METHOD_MODE2;
 }
 
 static struct amdgpu_reset_handler *

drivers/gpu/drm/amd/amdgpu/soc15.c

@@ -619,13 +619,6 @@ static void soc15_program_aspm(struct amdgpu_device *adev)
 		adev->nbio.funcs->program_aspm(adev);
 }
 
-static void soc15_enable_doorbell_aperture(struct amdgpu_device *adev,
-					   bool enable)
-{
-	adev->nbio.funcs->enable_doorbell_aperture(adev, enable);
-	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable);
-}
-
 const struct amdgpu_ip_block_version vega10_common_ip_block =
 {
 	.type = AMD_IP_BLOCK_TYPE_COMMON,
@@ -1125,6 +1118,11 @@ static int soc15_common_late_init(void *handle)
 	if (amdgpu_sriov_vf(adev))
 		xgpu_ai_mailbox_get_irq(adev);
 
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
 	return 0;
 }
 
@@ -1182,7 +1180,8 @@ static int soc15_common_hw_init(void *handle)
 		adev->nbio.funcs->remap_hdp_registers(adev);
 
 	/* enable the doorbell aperture */
-	soc15_enable_doorbell_aperture(adev, true);
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
+
 	/* HW doorbell routing policy: doorbell writing not
 	 * in SDMA/IH/MM/ACV range will be routed to CP. So
 	 * we need to init SDMA doorbell range prior
@@ -1198,8 +1197,14 @@ static int soc15_common_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	/* disable the doorbell aperture */
-	soc15_enable_doorbell_aperture(adev, false);
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because soc15_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
+
 	if (amdgpu_sriov_vf(adev))
 		xgpu_ai_mailbox_put_irq(adev);
 

drivers/gpu/drm/amd/amdgpu/soc21.c

@@ -450,13 +450,6 @@ static void soc21_program_aspm(struct amdgpu_device *adev)
 		adev->nbio.funcs->program_aspm(adev);
 }
 
-static void soc21_enable_doorbell_aperture(struct amdgpu_device *adev,
-					bool enable)
-{
-	adev->nbio.funcs->enable_doorbell_aperture(adev, enable);
-	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, enable);
-}
-
 const struct amdgpu_ip_block_version soc21_common_ip_block =
 {
 	.type = AMD_IP_BLOCK_TYPE_COMMON,
@@ -764,6 +757,11 @@ static int soc21_common_late_init(void *handle)
 			amdgpu_irq_get(adev, &adev->nbio.ras_err_event_athub_irq, 0);
 	}
 
+	/* Enable selfring doorbell aperture late because doorbell BAR
+	 * aperture will change if resize BAR successfully in gmc sw_init.
+	 */
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, true);
+
 	return 0;
 }
 
@@ -797,7 +795,7 @@ static int soc21_common_hw_init(void *handle)
 	if (adev->nbio.funcs->remap_hdp_registers)
 		adev->nbio.funcs->remap_hdp_registers(adev);
 	/* enable the doorbell aperture */
-	soc21_enable_doorbell_aperture(adev, true);
+	adev->nbio.funcs->enable_doorbell_aperture(adev, true);
 
 	return 0;
 }
@@ -806,8 +804,13 @@ static int soc21_common_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	/* disable the doorbell aperture */
-	soc21_enable_doorbell_aperture(adev, false);
+	/* Disable the doorbell aperture and selfring doorbell aperture
+	 * separately in hw_fini because soc21_enable_doorbell_aperture
+	 * has been removed and there is no need to delay disabling
+	 * selfring doorbell.
+	 */
+	adev->nbio.funcs->enable_doorbell_aperture(adev, false);
+	adev->nbio.funcs->enable_doorbell_selfring_aperture(adev, false);
 
 	if (amdgpu_sriov_vf(adev)) {
 		xgpu_nv_mailbox_put_irq(adev);

drivers/gpu/drm/amd/display/dc/dcn32/dcn32_resource.c

@@ -2079,6 +2079,14 @@ static struct resource_funcs dcn32_res_pool_funcs = {
 	.restore_mall_state = dcn32_restore_mall_state,
 };
 
+static uint32_t read_pipe_fuses(struct dc_context *ctx)
+{
+	uint32_t value = REG_READ(CC_DC_PIPE_DIS);
+	/* DCN32 support max 4 pipes */
+	value = value & 0xf;
+	return value;
+}
+
 
 static bool dcn32_resource_construct(
 	uint8_t num_virtual_links,
@@ -2122,7 +2130,7 @@ static bool dcn32_resource_construct(
 	pool->base.res_cap = &res_cap_dcn32;
 	/* max number of pipes for ASIC before checking for pipe fuses */
 	num_pipes  = pool->base.res_cap->num_timing_generator;
-	pipe_fuses = REG_READ(CC_DC_PIPE_DIS);
+	pipe_fuses = read_pipe_fuses(ctx);
 
 	for (i = 0; i < pool->base.res_cap->num_timing_generator; i++)
 		if (pipe_fuses & 1 << i)

drivers/gpu/drm/amd/display/dc/dcn321/dcn321_resource.c

@@ -1632,6 +1632,14 @@ static struct resource_funcs dcn321_res_pool_funcs = {
 	.restore_mall_state = dcn32_restore_mall_state,
 };
 
+static uint32_t read_pipe_fuses(struct dc_context *ctx)
+{
+	uint32_t value = REG_READ(CC_DC_PIPE_DIS);
+	/* DCN321 support max 4 pipes */
+	value = value & 0xf;
+	return value;
+}
+
 
 static bool dcn321_resource_construct(
 	uint8_t num_virtual_links,
@@ -1674,7 +1682,7 @@ static bool dcn321_resource_construct(
 	pool->base.res_cap = &res_cap_dcn321;
 	/* max number of pipes for ASIC before checking for pipe fuses */
 	num_pipes  = pool->base.res_cap->num_timing_generator;
-	pipe_fuses = REG_READ(CC_DC_PIPE_DIS);
+	pipe_fuses = read_pipe_fuses(ctx);
 
 	for (i = 0; i < pool->base.res_cap->num_timing_generator; i++)
 		if (pipe_fuses & 1 << i)

drivers/gpu/drm/amd/display/dc/dml/dcn314/dcn314_fpu.c

@@ -149,8 +149,8 @@ static struct _vcs_dpi_soc_bounding_box_st dcn3_14_soc = {
 	.num_states = 5,
 	.sr_exit_time_us = 16.5,
 	.sr_enter_plus_exit_time_us = 18.5,
-	.sr_exit_z8_time_us = 210.0,
-	.sr_enter_plus_exit_z8_time_us = 310.0,
+	.sr_exit_z8_time_us = 268.0,
+	.sr_enter_plus_exit_z8_time_us = 393.0,
 	.writeback_latency_us = 12.0,
 	.dram_channel_width_bytes = 4,
 	.round_trip_ping_latency_dcfclk_cycles = 106,