Merge drm/drm-next into drm-misc-next

Let's kick-off this release cycle.

Signed-off-by: Maxime Ripard <maxime@cerno.tech>

1 files changed, 121 insertions, 18 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 62b26f0e37b0..ab8f970b2849 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2365,8 +2365,16 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 		}
 		adev->ip_blocks[i].status.sw = true;
 
-		/* need to do gmc hw init early so we can allocate gpu mem */
-		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) {
+			/* need to do common hw init early so everything is set up for gmc */
+			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
+			if (r) {
+				DRM_ERROR("hw_init %d failed %d\n", i, r);
+				goto init_failed;
+			}
+			adev->ip_blocks[i].status.hw = true;
+		} else if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
+			/* need to do gmc hw init early so we can allocate gpu mem */
 			/* Try to reserve bad pages early */
 			if (amdgpu_sriov_vf(adev))
 				amdgpu_virt_exchange_data(adev);
@@ -2451,19 +2459,21 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	 */
 	if (adev->gmc.xgmi.num_physical_nodes > 1) {
 		if (amdgpu_xgmi_add_device(adev) == 0) {
-			struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+			if (!amdgpu_sriov_vf(adev)) {
+				struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
 
-			if (!hive->reset_domain ||
-			    !amdgpu_reset_get_reset_domain(hive->reset_domain)) {
-				r = -ENOENT;
+				if (!hive->reset_domain ||
+				    !amdgpu_reset_get_reset_domain(hive->reset_domain)) {
+					r = -ENOENT;
+					amdgpu_put_xgmi_hive(hive);
+					goto init_failed;
+				}
+
+				/* Drop the early temporary reset domain we created for device */
+				amdgpu_reset_put_reset_domain(adev->reset_domain);
+				adev->reset_domain = hive->reset_domain;
 				amdgpu_put_xgmi_hive(hive);
-				goto init_failed;
 			}
-
-			/* Drop the early temporary reset domain we created for device */
-			amdgpu_reset_put_reset_domain(adev->reset_domain);
-			adev->reset_domain = hive->reset_domain;
-			amdgpu_put_xgmi_hive(hive);
 		}
 	}
 
@@ -3052,8 +3062,8 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 	int i, r;
 
 	static enum amd_ip_block_type ip_order[] = {
-		AMD_IP_BLOCK_TYPE_GMC,
 		AMD_IP_BLOCK_TYPE_COMMON,
+		AMD_IP_BLOCK_TYPE_GMC,
 		AMD_IP_BLOCK_TYPE_PSP,
 		AMD_IP_BLOCK_TYPE_IH,
 	};
@@ -3144,7 +3154,8 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
-		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) {
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
+		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP && amdgpu_sriov_vf(adev))) {
 
 			r = adev->ip_blocks[i].version->funcs->resume(adev);
 			if (r) {
@@ -3501,6 +3512,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->gmc.gart_size = 512 * 1024 * 1024;
 	adev->accel_working = false;
 	adev->num_rings = 0;
+	RCU_INIT_POINTER(adev->gang_submit, dma_fence_get_stub());
 	adev->mman.buffer_funcs = NULL;
 	adev->mman.buffer_funcs_ring = NULL;
 	adev->vm_manager.vm_pte_funcs = NULL;
@@ -3982,6 +3994,7 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
 	release_firmware(adev->firmware.gpu_info_fw);
 	adev->firmware.gpu_info_fw = NULL;
 	adev->accel_working = false;
+	dma_fence_put(rcu_dereference_protected(adev->gang_submit, true));
 
 	amdgpu_reset_fini(adev);
 
@@ -4057,12 +4070,20 @@ static void amdgpu_device_evict_resources(struct amdgpu_device *adev)
 int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
+	int r = 0;
 
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
 	adev->in_suspend = true;
 
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_fini_data_exchange(adev);
+		r = amdgpu_virt_request_full_gpu(adev, false);
+		if (r)
+			return r;
+	}
+
 	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D3))
 		DRM_WARN("smart shift update failed\n");
 
@@ -4086,6 +4107,9 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
 
 	amdgpu_device_ip_suspend_phase2(adev);
 
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_release_full_gpu(adev, false);
+
 	return 0;
 }
 
@@ -4104,6 +4128,12 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r = 0;
 
+	if (amdgpu_sriov_vf(adev)) {
+		r = amdgpu_virt_request_full_gpu(adev, true);
+		if (r)
+			return r;
+	}
+
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
@@ -4118,6 +4148,13 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 	}
 
 	r = amdgpu_device_ip_resume(adev);
+
+	/* no matter what r is, always need to properly release full GPU */
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_init_data_exchange(adev);
+		amdgpu_virt_release_full_gpu(adev, true);
+	}
+
 	if (r) {
 		dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r);
 		return r;
@@ -4739,6 +4776,7 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 	struct amdgpu_device *tmp_adev = NULL;
 	bool need_full_reset, skip_hw_reset, vram_lost = false;
 	int r = 0;
+	bool gpu_reset_for_dev_remove = 0;
 
 	/* Try reset handler method first */
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
@@ -4758,6 +4796,10 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
 
+	gpu_reset_for_dev_remove =
+		test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
+			test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+
 	/*
 	 * ASIC reset has to be done on all XGMI hive nodes ASAP
 	 * to allow proper links negotiation in FW (within 1 sec)
@@ -4802,6 +4844,18 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 		amdgpu_ras_intr_cleared();
 	}
 
+	/* Since the mode1 reset affects base ip blocks, the
+	 * phase1 ip blocks need to be resumed. Otherwise there
+	 * will be a BIOS signature error and the psp bootloader
+	 * can't load kdb on the next amdgpu install.
+	 */
+	if (gpu_reset_for_dev_remove) {
+		list_for_each_entry(tmp_adev, device_list_handle, reset_list)
+			amdgpu_device_ip_resume_phase1(tmp_adev);
+
+		goto end;
+	}
+
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
 		if (need_full_reset) {
 			/* post card */
@@ -5124,6 +5178,11 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 	int tmp_vram_lost_counter;
+	bool gpu_reset_for_dev_remove = false;
+
+	gpu_reset_for_dev_remove =
+			test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
+				test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
 	/*
 	 * Special case: RAS triggered and full reset isn't supported
@@ -5159,8 +5218,11 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	 */
 	INIT_LIST_HEAD(&device_list);
 	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) {
-		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
 			list_add_tail(&tmp_adev->reset_list, &device_list);
+			if (gpu_reset_for_dev_remove && adev->shutdown)
+				tmp_adev->shutdown = true;
+		}
 		if (!list_is_first(&adev->reset_list, &device_list))
 			list_rotate_to_front(&adev->reset_list, &device_list);
 		device_list_handle = &device_list;
@@ -5243,6 +5305,10 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 
 retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+		if (gpu_reset_for_dev_remove) {
+			/* Workaroud for ASICs need to disable SMC first */
+			amdgpu_device_smu_fini_early(tmp_adev);
+		}
 		r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context);
 		/*TODO Should we stop ?*/
 		if (r) {
@@ -5276,6 +5342,9 @@ retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 			adev->asic_reset_res = 0;
 			goto retry;
 		}
+
+		if (!r && gpu_reset_for_dev_remove)
+			goto recover_end;
 	}
 
 skip_hw_reset:
@@ -5349,6 +5418,7 @@ skip_sched_resume:
 		amdgpu_device_unset_mp1_state(tmp_adev);
 	}
 
+recover_end:
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
 					    reset_list);
 	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
@@ -5531,9 +5601,9 @@ bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
 		~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1);
 	resource_size_t aper_limit =
 		adev->gmc.aper_base + adev->gmc.aper_size - 1;
-	bool p2p_access = !adev->gmc.xgmi.connected_to_cpu &&
-			  !(pci_p2pdma_distance_many(adev->pdev,
-					&peer_adev->dev, 1, true) < 0);
+	bool p2p_access =
+		!adev->gmc.xgmi.connected_to_cpu &&
+		!(pci_p2pdma_distance(adev->pdev, peer_adev->dev, false) < 0);
 
 	return pcie_p2p && p2p_access && (adev->gmc.visible_vram_size &&
 		adev->gmc.real_vram_size == adev->gmc.visible_vram_size &&
@@ -5917,3 +5987,36 @@ void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev,
 	(void)RREG32(data);
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 }
+
+/**
+ * amdgpu_device_switch_gang - switch to a new gang
+ * @adev: amdgpu_device pointer
+ * @gang: the gang to switch to
+ *
+ * Try to switch to a new gang.
+ * Returns: NULL if we switched to the new gang or a reference to the current
+ * gang leader.
+ */
+struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
+					    struct dma_fence *gang)
+{
+	struct dma_fence *old = NULL;
+
+	do {
+		dma_fence_put(old);
+		rcu_read_lock();
+		old = dma_fence_get_rcu_safe(&adev->gang_submit);
+		rcu_read_unlock();
+
+		if (old == gang)
+			break;
+
+		if (!dma_fence_is_signaled(old))
+			return old;
+
+	} while (cmpxchg((struct dma_fence __force **)&adev->gang_submit,
+			 old, gang) != old);
+
+	dma_fence_put(old);
+	return NULL;
+}