1 files changed, 20 insertions, 142 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 78452cf0115d..4ef4d31f5231 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -254,6 +254,7 @@ static const struct soc15_reg_golden golden_settings_sdma_4_3[] = {
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000),
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000),
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0),
+	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xfc000000, 0x00000000)
 };
 
 static u32 sdma_v4_0_get_reg_offset(struct amdgpu_device *adev,
@@ -746,13 +747,13 @@ static void sdma_v4_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	u32 ref_and_mask = 0;
-	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio_funcs->hdp_flush_reg;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
 
 	ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me;
 
 	sdma_v4_0_wait_reg_mem(ring, 0, 1,
-			       adev->nbio_funcs->get_hdp_flush_done_offset(adev),
-			       adev->nbio_funcs->get_hdp_flush_req_offset(adev),
+			       adev->nbio.funcs->get_hdp_flush_done_offset(adev),
+			       adev->nbio.funcs->get_hdp_flush_req_offset(adev),
 			       ref_and_mask, ref_and_mask, 10);
 }
 
@@ -1690,102 +1691,17 @@ static int sdma_v4_0_early_init(void *handle)
 }
 
 static int sdma_v4_0_process_ras_data_cb(struct amdgpu_device *adev,
-		struct ras_err_data *err_data,
+		void *err_data,
 		struct amdgpu_iv_entry *entry);
 
 static int sdma_v4_0_late_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	struct ras_common_if **ras_if = &adev->sdma.ras_if;
 	struct ras_ih_if ih_info = {
 		.cb = sdma_v4_0_process_ras_data_cb,
 	};
-	struct ras_fs_if fs_info = {
-		.sysfs_name = "sdma_err_count",
-		.debugfs_name = "sdma_err_inject",
-	};
-	struct ras_common_if ras_block = {
-		.block = AMDGPU_RAS_BLOCK__SDMA,
-		.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
-		.sub_block_index = 0,
-		.name = "sdma",
-	};
-	int r, i;
-
-	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
-		amdgpu_ras_feature_enable_on_boot(adev, &ras_block, 0);
-		return 0;
-	}
-
-	/* handle resume path. */
-	if (*ras_if) {
-		/* resend ras TA enable cmd during resume.
-		 * prepare to handle failure.
-		 */
-		ih_info.head = **ras_if;
-		r = amdgpu_ras_feature_enable_on_boot(adev, *ras_if, 1);
-		if (r) {
-			if (r == -EAGAIN) {
-				/* request a gpu reset. will run again. */
-				amdgpu_ras_request_reset_on_boot(adev,
-						AMDGPU_RAS_BLOCK__SDMA);
-				return 0;
-			}
-			/* fail to enable ras, cleanup all. */
-			goto irq;
-		}
-		/* enable successfully. continue. */
-		goto resume;
-	}
-
-	*ras_if = kmalloc(sizeof(**ras_if), GFP_KERNEL);
-	if (!*ras_if)
-		return -ENOMEM;
-
-	**ras_if = ras_block;
-
-	r = amdgpu_ras_feature_enable_on_boot(adev, *ras_if, 1);
-	if (r) {
-		if (r == -EAGAIN) {
-			amdgpu_ras_request_reset_on_boot(adev,
-					AMDGPU_RAS_BLOCK__SDMA);
-			r = 0;
-		}
-		goto feature;
-	}
-
-	ih_info.head = **ras_if;
-	fs_info.head = **ras_if;
-
-	r = amdgpu_ras_interrupt_add_handler(adev, &ih_info);
-	if (r)
-		goto interrupt;
-
-	amdgpu_ras_debugfs_create(adev, &fs_info);
 
-	r = amdgpu_ras_sysfs_create(adev, &fs_info);
-	if (r)
-		goto sysfs;
-resume:
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		r = amdgpu_irq_get(adev, &adev->sdma.ecc_irq,
-				   AMDGPU_SDMA_IRQ_INSTANCE0 + i);
-		if (r)
-			goto irq;
-	}
-
-	return 0;
-irq:
-	amdgpu_ras_sysfs_remove(adev, *ras_if);
-sysfs:
-	amdgpu_ras_debugfs_remove(adev, *ras_if);
-	amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
-interrupt:
-	amdgpu_ras_feature_enable(adev, *ras_if, 0);
-feature:
-	kfree(*ras_if);
-	*ras_if = NULL;
-	return r;
+	return amdgpu_sdma_ras_late_init(adev, &ih_info);
 }
 
 static int sdma_v4_0_sw_init(void *handle)
@@ -1857,21 +1773,7 @@ static int sdma_v4_0_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int i;
 
-	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA) &&
-			adev->sdma.ras_if) {
-		struct ras_common_if *ras_if = adev->sdma.ras_if;
-		struct ras_ih_if ih_info = {
-			.head = *ras_if,
-		};
-
-		/*remove fs first*/
-		amdgpu_ras_debugfs_remove(adev, ras_if);
-		amdgpu_ras_sysfs_remove(adev, ras_if);
-		/*remove the IH*/
-		amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
-		amdgpu_ras_feature_enable(adev, ras_if, 0);
-		kfree(ras_if);
-	}
+	amdgpu_sdma_ras_fini(adev);
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
@@ -1891,7 +1793,7 @@ static int sdma_v4_0_hw_init(void *handle)
 
 	if ((adev->asic_type == CHIP_RAVEN && adev->powerplay.pp_funcs &&
 			adev->powerplay.pp_funcs->set_powergating_by_smu) ||
-			adev->asic_type == CHIP_RENOIR)
+			(adev->asic_type == CHIP_RENOIR && !adev->in_gpu_reset))
 		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false);
 
 	if (!amdgpu_sriov_vf(adev))
@@ -2024,52 +1926,28 @@ static int sdma_v4_0_process_trap_irq(struct amdgpu_device *adev,
 }
 
 static int sdma_v4_0_process_ras_data_cb(struct amdgpu_device *adev,
-		struct ras_err_data *err_data,
+		void *err_data,
 		struct amdgpu_iv_entry *entry)
 {
-	uint32_t err_source;
 	int instance;
 
+	/* When “Full RAS” is enabled, the per-IP interrupt sources should
+	 * be disabled and the driver should only look for the aggregated
+	 * interrupt via sync flood
+	 */
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
+		goto out;
+
 	instance = sdma_v4_0_irq_id_to_seq(entry->client_id);
 	if (instance < 0)
-		return 0;
-
-	switch (entry->src_id) {
-	case SDMA0_4_0__SRCID__SDMA_SRAM_ECC:
-		err_source = 0;
-		break;
-	case SDMA0_4_0__SRCID__SDMA_ECC:
-		err_source = 1;
-		break;
-	default:
-		return 0;
-	}
-
-	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+		goto out;
 
-	amdgpu_ras_reset_gpu(adev, 0);
+	amdgpu_sdma_process_ras_data_cb(adev, err_data, entry);
 
+out:
 	return AMDGPU_RAS_SUCCESS;
 }
 
-static int sdma_v4_0_process_ecc_irq(struct amdgpu_device *adev,
-				      struct amdgpu_irq_src *source,
-				      struct amdgpu_iv_entry *entry)
-{
-	struct ras_common_if *ras_if = adev->sdma.ras_if;
-	struct ras_dispatch_if ih_data = {
-		.entry = entry,
-	};
-
-	if (!ras_if)
-		return 0;
-
-	ih_data.head = *ras_if;
-
-	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
-	return 0;
-}
-
 static int sdma_v4_0_process_illegal_inst_irq(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
 					      struct amdgpu_iv_entry *entry)
@@ -2417,7 +2295,7 @@ static const struct amdgpu_irq_src_funcs sdma_v4_0_illegal_inst_irq_funcs = {
 
 static const struct amdgpu_irq_src_funcs sdma_v4_0_ecc_irq_funcs = {
 	.set = sdma_v4_0_set_ecc_irq_state,
-	.process = sdma_v4_0_process_ecc_irq,
+	.process = amdgpu_sdma_process_ecc_irq,
 };