Merge tag 'drm-fixes-2018-09-14' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "This is the general drm fixes pull for rc4.

  i915:
   - Two GVT fixes (one for the mm reference issue you pointed out)
   - Gen 2 video playback fix
   - IPS timeout error suppression on Broadwell

  amdgpu:
   - Small memory leak
   - SR-IOV reset
   - locking fix
   - updated SDMA golden registers

  nouveau:
   - Remove some leftover debugging"

* tag 'drm-fixes-2018-09-14' of git://anongit.freedesktop.org/drm/drm:
  drm/nouveau/devinit: fix warning when PMU/PRE_OS is missing
  drm/amdgpu: fix error handling in amdgpu_cs_user_fence_chunk
  drm/i915/overlay: Allocate physical registers from stolen
  drm/amdgpu: move PSP init prior to IH in gpu reset
  drm/amdgpu: Fix SDMA hang in prt mode v2
  drm/amdgpu: fix amdgpu_mn_unlock() in the CS error path
  drm/i915/bdw: Increase IPS disable timeout to 100ms
  drm/i915/gvt: Fix the incorrect length of child_device_config issue
  drm/i915/gvt: Fix life cycle reference on KVM mm

8 files changed, 131 insertions, 190 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index b6e9df11115d..b31d121a876b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -39,6 +39,7 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p,
 {
 	struct drm_gem_object *gobj;
 	unsigned long size;
+	int r;
 
 	gobj = drm_gem_object_lookup(p->filp, data->handle);
 	if (gobj == NULL)
@@ -50,20 +51,26 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p,
 	p->uf_entry.tv.shared = true;
 	p->uf_entry.user_pages = NULL;
 
-	size = amdgpu_bo_size(p->uf_entry.robj);
-	if (size != PAGE_SIZE || (data->offset + 8) > size)
-		return -EINVAL;
-
-	*offset = data->offset;
-
 	drm_gem_object_put_unlocked(gobj);
 
+	size = amdgpu_bo_size(p->uf_entry.robj);
+	if (size != PAGE_SIZE || (data->offset + 8) > size) {
+		r = -EINVAL;
+		goto error_unref;
+	}
+
 	if (amdgpu_ttm_tt_get_usermm(p->uf_entry.robj->tbo.ttm)) {
-		amdgpu_bo_unref(&p->uf_entry.robj);
-		return -EINVAL;
+		r = -EINVAL;
+		goto error_unref;
 	}
 
+	*offset = data->offset;
+
 	return 0;
+
+error_unref:
+	amdgpu_bo_unref(&p->uf_entry.robj);
+	return r;
 }
 
 static int amdgpu_cs_bo_handles_chunk(struct amdgpu_cs_parser *p,
@@ -1262,10 +1269,10 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 error_abort:
 	dma_fence_put(&job->base.s_fence->finished);
 	job->base.s_fence = NULL;
+	amdgpu_mn_unlock(p->mn);
 
 error_unlock:
 	amdgpu_job_free(job);
-	amdgpu_mn_unlock(p->mn);
 	return r;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 8ab5ccbc14ac..39bf2ce548c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2063,6 +2063,7 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 	static enum amd_ip_block_type ip_order[] = {
 		AMD_IP_BLOCK_TYPE_GMC,
 		AMD_IP_BLOCK_TYPE_COMMON,
+		AMD_IP_BLOCK_TYPE_PSP,
 		AMD_IP_BLOCK_TYPE_IH,
 	};
 
@@ -2093,7 +2094,6 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
 
 	static enum amd_ip_block_type ip_order[] = {
 		AMD_IP_BLOCK_TYPE_SMC,
-		AMD_IP_BLOCK_TYPE_PSP,
 		AMD_IP_BLOCK_TYPE_DCE,
 		AMD_IP_BLOCK_TYPE_GFX,
 		AMD_IP_BLOCK_TYPE_SDMA,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index e7ca4623cfb9..7c3b634d8d5f 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -70,6 +70,7 @@ static const struct soc15_reg_golden golden_settings_sdma_4[] = {
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_IB_CNTL, 0x800f0100, 0x00000100),
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000),
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0),
+	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xfc000000, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831f07),
 	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CLK_CTRL, 0xffffffff, 0x3f000100),
 	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_IB_CNTL, 0x800f0100, 0x00000100),
@@ -81,7 +82,8 @@ static const struct soc15_reg_golden golden_settings_sdma_4[] = {
 	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC0_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000),
 	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_IB_CNTL, 0x800f0100, 0x00000100),
 	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000),
-	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0)
+	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0),
+	SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_WATERMK, 0xfc000000, 0x00000000)
 };
 
 static const struct soc15_reg_golden golden_settings_sdma_vg10[] = {
@@ -109,7 +111,8 @@ static const struct soc15_reg_golden golden_settings_sdma_4_1[] =
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000),
 	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_IB_CNTL, 0x800f0111, 0x00000100),
 	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_PAGE, 0x000003ff, 0x000003c0),
+	SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xfc000000, 0x00000000)
 };
 
 static const struct soc15_reg_golden golden_settings_sdma_4_2[] =
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index a45f46d8537f..c7afee37b2b8 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -32,6 +32,7 @@
 #include <linux/device.h>
 #include <linux/mm.h>
 #include <linux/mmu_context.h>
+#include <linux/sched/mm.h>
 #include <linux/types.h>
 #include <linux/list.h>
 #include <linux/rbtree.h>
@@ -1792,16 +1793,21 @@ static int kvmgt_rw_gpa(unsigned long handle, unsigned long gpa,
 	info = (struct kvmgt_guest_info *)handle;
 	kvm = info->kvm;
 
-	if (kthread)
+	if (kthread) {
+		if (!mmget_not_zero(kvm->mm))
+			return -EFAULT;
 		use_mm(kvm->mm);
+	}
 
 	idx = srcu_read_lock(&kvm->srcu);
 	ret = write ? kvm_write_guest(kvm, gpa, buf, len) :
 		      kvm_read_guest(kvm, gpa, buf, len);
 	srcu_read_unlock(&kvm->srcu, idx);
 
-	if (kthread)
+	if (kthread) {
 		unuse_mm(kvm->mm);
+		mmput(kvm->mm);
+	}
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c
index fa75a2eead90..b0d3a43ccd03 100644
--- a/drivers/gpu/drm/i915/gvt/opregion.c
+++ b/drivers/gpu/drm/i915/gvt/opregion.c
@@ -42,8 +42,6 @@
 #define DEVICE_TYPE_EFP3   0x20
 #define DEVICE_TYPE_EFP4   0x10
 
-#define DEV_SIZE	38
-
 struct opregion_header {
 	u8 signature[16];
 	u32 size;
@@ -63,6 +61,10 @@ struct bdb_data_header {
 	u16 size; /* data size */
 } __packed;
 
+/* For supporting windows guest with opregion, here hardcode the emulated
+ * bdb header version as '186', and the corresponding child_device_config
+ * length should be '33' but not '38'.
+ */
 struct efp_child_device_config {
 	u16 handle;
 	u16 device_type;
@@ -109,12 +111,6 @@ struct efp_child_device_config {
 	u8 mipi_bridge_type; /* 171 */
 	u16 device_class_ext;
 	u8 dvo_function;
-	u8 dp_usb_type_c:1; /* 195 */
-	u8 skip6:7;
-	u8 dp_usb_type_c_2x_gpio_index; /* 195 */
-	u16 dp_usb_type_c_2x_gpio_pin; /* 195 */
-	u8 iboost_dp:4; /* 196 */
-	u8 iboost_hdmi:4; /* 196 */
 } __packed;
 
 struct vbt {
@@ -155,7 +151,7 @@ static void virt_vbt_generation(struct vbt *v)
 	v->header.bdb_offset = offsetof(struct vbt, bdb_header);
 
 	strcpy(&v->bdb_header.signature[0], "BIOS_DATA_BLOCK");
-	v->bdb_header.version = 186; /* child_dev_size = 38 */
+	v->bdb_header.version = 186; /* child_dev_size = 33 */
 	v->bdb_header.header_size = sizeof(v->bdb_header);
 
 	v->bdb_header.bdb_size = sizeof(struct vbt) - sizeof(struct vbt_header)
@@ -169,11 +165,13 @@ static void virt_vbt_generation(struct vbt *v)
 
 	/* child device */
 	num_child = 4; /* each port has one child */
+	v->general_definitions.child_dev_size =
+		sizeof(struct efp_child_device_config);
 	v->general_definitions_header.id = BDB_GENERAL_DEFINITIONS;
 	/* size will include child devices */
 	v->general_definitions_header.size =
-		sizeof(struct bdb_general_definitions) + num_child * DEV_SIZE;
-	v->general_definitions.child_dev_size = DEV_SIZE;
+		sizeof(struct bdb_general_definitions) +
+			num_child * v->general_definitions.child_dev_size;
 
 	/* portA */
 	v->child0.handle = DEVICE_TYPE_EFP1;
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 4a3c8ee9a973..d2951096bca0 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -5079,10 +5079,14 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state)
 		mutex_lock(&dev_priv->pcu_lock);
 		WARN_ON(sandybridge_pcode_write(dev_priv, DISPLAY_IPS_CONTROL, 0));
 		mutex_unlock(&dev_priv->pcu_lock);
-		/* wait for pcode to finish disabling IPS, which may take up to 42ms */
+		/*
+		 * Wait for PCODE to finish disabling IPS. The BSpec specified
+		 * 42ms timeout value leads to occasional timeouts so use 100ms
+		 * instead.
+		 */
 		if (intel_wait_for_register(dev_priv,
 					    IPS_CTL, IPS_ENABLE, 0,
-					    42))
+					    100))
 			DRM_ERROR("Timed out waiting for IPS disable\n");
 	} else {
 		I915_WRITE(IPS_CTL, 0);
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index c2f10d899329..443dfaefd7a6 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -181,8 +181,9 @@ struct intel_overlay {
 	u32 brightness, contrast, saturation;
 	u32 old_xscale, old_yscale;
 	/* register access */
-	u32 flip_addr;
 	struct drm_i915_gem_object *reg_bo;
+	struct overlay_registers __iomem *regs;
+	u32 flip_addr;
 	/* flip handling */
 	struct i915_gem_active last_flip;
 };
@@ -210,29 +211,6 @@ static void i830_overlay_clock_gating(struct drm_i915_private *dev_priv,
 				  PCI_DEVFN(0, 0), I830_CLOCK_GATE, val);
 }
 
-static struct overlay_registers __iomem *
-intel_overlay_map_regs(struct intel_overlay *overlay)
-{
-	struct drm_i915_private *dev_priv = overlay->i915;
-	struct overlay_registers __iomem *regs;
-
-	if (OVERLAY_NEEDS_PHYSICAL(dev_priv))
-		regs = (struct overlay_registers __iomem *)overlay->reg_bo->phys_handle->vaddr;
-	else
-		regs = io_mapping_map_wc(&dev_priv->ggtt.iomap,
-					 overlay->flip_addr,
-					 PAGE_SIZE);
-
-	return regs;
-}
-
-static void intel_overlay_unmap_regs(struct intel_overlay *overlay,
-				     struct overlay_registers __iomem *regs)
-{
-	if (!OVERLAY_NEEDS_PHYSICAL(overlay->i915))
-		io_mapping_unmap(regs);
-}
-
 static void intel_overlay_submit_request(struct intel_overlay *overlay,
 					 struct i915_request *rq,
 					 i915_gem_retire_fn retire)
@@ -784,13 +762,13 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
 				      struct drm_i915_gem_object *new_bo,
 				      struct put_image_params *params)
 {
-	int ret, tmp_width;
-	struct overlay_registers __iomem *regs;
-	bool scale_changed = false;
+	struct overlay_registers __iomem *regs = overlay->regs;
 	struct drm_i915_private *dev_priv = overlay->i915;
 	u32 swidth, swidthsw, sheight, ostride;
 	enum pipe pipe = overlay->crtc->pipe;
+	bool scale_changed = false;
 	struct i915_vma *vma;
+	int ret, tmp_width;
 
 	lockdep_assert_held(&dev_priv->drm.struct_mutex);
 	WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
@@ -815,30 +793,19 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
 
 	if (!overlay->active) {
 		u32 oconfig;
-		regs = intel_overlay_map_regs(overlay);
-		if (!regs) {
-			ret = -ENOMEM;
-			goto out_unpin;
-		}
+
 		oconfig = OCONF_CC_OUT_8BIT;
 		if (IS_GEN4(dev_priv))
 			oconfig |= OCONF_CSC_MODE_BT709;
 		oconfig |= pipe == 0 ?
 			OCONF_PIPE_A : OCONF_PIPE_B;
 		iowrite32(oconfig, &regs->OCONFIG);
-		intel_overlay_unmap_regs(overlay, regs);
 
 		ret = intel_overlay_on(overlay);
 		if (ret != 0)
 			goto out_unpin;
 	}
 
-	regs = intel_overlay_map_regs(overlay);
-	if (!regs) {
-		ret = -ENOMEM;
-		goto out_unpin;
-	}
-
 	iowrite32((params->dst_y << 16) | params->dst_x, &regs->DWINPOS);
 	iowrite32((params->dst_h << 16) | params->dst_w, &regs->DWINSZ);
 
@@ -882,8 +849,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
 
 	iowrite32(overlay_cmd_reg(params), &regs->OCMD);
 
-	intel_overlay_unmap_regs(overlay, regs);
-
 	ret = intel_overlay_continue(overlay, vma, scale_changed);
 	if (ret)
 		goto out_unpin;
@@ -901,7 +866,6 @@ out_pin_section:
 int intel_overlay_switch_off(struct intel_overlay *overlay)
 {
 	struct drm_i915_private *dev_priv = overlay->i915;
-	struct overlay_registers __iomem *regs;
 	int ret;
 
 	lockdep_assert_held(&dev_priv->drm.struct_mutex);
@@ -918,9 +882,7 @@ int intel_overlay_switch_off(struct intel_overlay *overlay)
 	if (ret != 0)
 		return ret;
 
-	regs = intel_overlay_map_regs(overlay);
-	iowrite32(0, &regs->OCMD);
-	intel_overlay_unmap_regs(overlay, regs);
+	iowrite32(0, &overlay->regs->OCMD);
 
 	return intel_overlay_off(overlay);
 }
@@ -1305,7 +1267,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
 	struct drm_intel_overlay_attrs *attrs = data;
 	struct drm_i915_private *dev_priv = to_i915(dev);
 	struct intel_overlay *overlay;
-	struct overlay_registers __iomem *regs;
 	int ret;
 
 	overlay = dev_priv->overlay;
@@ -1345,15 +1306,7 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
 		overlay->contrast   = attrs->contrast;
 		overlay->saturation = attrs->saturation;
 
-		regs = intel_overlay_map_regs(overlay);
-		if (!regs) {
-			ret = -ENOMEM;
-			goto out_unlock;
-		}
-
-		update_reg_attrs(overlay, regs);
-
-		intel_overlay_unmap_regs(overlay, regs);
+		update_reg_attrs(overlay, overlay->regs);
 
 		if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
 			if (IS_GEN2(dev_priv))
@@ -1386,12 +1339,47 @@ out_unlock:
 	return ret;
 }
 
+static int get_registers(struct intel_overlay *overlay, bool use_phys)
+{
+	struct drm_i915_gem_object *obj;
+	struct i915_vma *vma;
+	int err;
+
+	obj = i915_gem_object_create_stolen(overlay->i915, PAGE_SIZE);
+	if (obj == NULL)
+		obj = i915_gem_object_create_internal(overlay->i915, PAGE_SIZE);
+	if (IS_ERR(obj))
+		return PTR_ERR(obj);
+
+	vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE);
+	if (IS_ERR(vma)) {
+		err = PTR_ERR(vma);
+		goto err_put_bo;
+	}
+
+	if (use_phys)
+		overlay->flip_addr = sg_dma_address(obj->mm.pages->sgl);
+	else
+		overlay->flip_addr = i915_ggtt_offset(vma);
+	overlay->regs = i915_vma_pin_iomap(vma);
+	i915_vma_unpin(vma);
+
+	if (IS_ERR(overlay->regs)) {
+		err = PTR_ERR(overlay->regs);
+		goto err_put_bo;
+	}
+
+	overlay->reg_bo = obj;
+	return 0;
+
+err_put_bo:
+	i915_gem_object_put(obj);
+	return err;
+}
+
 void intel_setup_overlay(struct drm_i915_private *dev_priv)
 {
 	struct intel_overlay *overlay;
-	struct drm_i915_gem_object *reg_bo;
-	struct overlay_registers __iomem *regs;
-	struct i915_vma *vma = NULL;
 	int ret;
 
 	if (!HAS_OVERLAY(dev_priv))
@@ -1401,46 +1389,8 @@ void intel_setup_overlay(struct drm_i915_private *dev_priv)
 	if (!overlay)
 		return;
 
-	mutex_lock(&dev_priv->drm.struct_mutex);
-	if (WARN_ON(dev_priv->overlay))
-		goto out_free;
-
 	overlay->i915 = dev_priv;
 
-	reg_bo = NULL;
-	if (!OVERLAY_NEEDS_PHYSICAL(dev_priv))
-		reg_bo = i915_gem_object_create_stolen(dev_priv, PAGE_SIZE);
-	if (reg_bo == NULL)
-		reg_bo = i915_gem_object_create(dev_priv, PAGE_SIZE);
-	if (IS_ERR(reg_bo))
-		goto out_free;
-	overlay->reg_bo = reg_bo;
-
-	if (OVERLAY_NEEDS_PHYSICAL(dev_priv)) {
-		ret = i915_gem_object_attach_phys(reg_bo, PAGE_SIZE);
-		if (ret) {
-			DRM_ERROR("failed to attach phys overlay regs\n");
-			goto out_free_bo;
-		}
-		overlay->flip_addr = reg_bo->phys_handle->busaddr;
-	} else {
-		vma = i915_gem_object_ggtt_pin(reg_bo, NULL,
-					       0, PAGE_SIZE, PIN_MAPPABLE);
-		if (IS_ERR(vma)) {
-			DRM_ERROR("failed to pin overlay register bo\n");
-			ret = PTR_ERR(vma);
-			goto out_free_bo;
-		}
-		overlay->flip_addr = i915_ggtt_offset(vma);
-
-		ret = i915_gem_object_set_to_gtt_domain(reg_bo, true);
-		if (ret) {
-			DRM_ERROR("failed to move overlay register bo into the GTT\n");
-			goto out_unpin_bo;
-		}
-	}
-
-	/* init all values */
 	overlay->color_key = 0x0101fe;
 	overlay->color_key_enabled = true;
 	overlay->brightness = -19;
@@ -1449,44 +1399,51 @@ void intel_setup_overlay(struct drm_i915_private *dev_priv)
 
 	init_request_active(&overlay->last_flip, NULL);
 
-	regs = intel_overlay_map_regs(overlay);
-	if (!regs)
-		goto out_unpin_bo;
+	mutex_lock(&dev_priv->drm.struct_mutex);
+
+	ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv));
+	if (ret)
+		goto out_free;
+
+	ret = i915_gem_object_set_to_gtt_domain(overlay->reg_bo, true);
+	if (ret)
+		goto out_reg_bo;
 
-	memset_io(regs, 0, sizeof(struct overlay_registers));
-	update_polyphase_filter(regs);
-	update_reg_attrs(overlay, regs);
+	mutex_unlock(&dev_priv->drm.struct_mutex);
 
-	intel_overlay_unmap_regs(overlay, regs);
+	memset_io(overlay->regs, 0, sizeof(struct overlay_registers));
+	update_polyphase_filter(overlay->regs);
+	update_reg_attrs(overlay, overlay->regs);
 
 	dev_priv->overlay = overlay;
-	mutex_unlock(&dev_priv->drm.struct_mutex);
-	DRM_INFO("initialized overlay support\n");
+	DRM_INFO("Initialized overlay support.\n");
 	return;
 
-out_unpin_bo:
-	if (vma)
-		i915_vma_unpin(vma);
-out_free_bo:
-	i915_gem_object_put(reg_bo);
+out_reg_bo:
+	i915_gem_object_put(overlay->reg_bo);
 out_free:
 	mutex_unlock(&dev_priv->drm.struct_mutex);
 	kfree(overlay);
-	return;
 }
 
 void intel_cleanup_overlay(struct drm_i915_private *dev_priv)
 {
-	if (!dev_priv->overlay)
+	struct intel_overlay *overlay;
+
+	overlay = fetch_and_zero(&dev_priv->overlay);
+	if (!overlay)
 		return;
 
-	/* The bo's should be free'd by the generic code already.
+	/*
+	 * The bo's should be free'd by the generic code already.
 	 * Furthermore modesetting teardown happens beforehand so the
-	 * hardware should be off already */
-	WARN_ON(dev_priv->overlay->active);
+	 * hardware should be off already.
+	 */
+	WARN_ON(overlay->active);
+
+	i915_gem_object_put(overlay->reg_bo);
 
-	i915_gem_object_put(dev_priv->overlay->reg_bo);
-	kfree(dev_priv->overlay);
+	kfree(overlay);
 }
 
 #if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR)
@@ -1498,37 +1455,11 @@ struct intel_overlay_error_state {
 	u32 isr;
 };
 
-static struct overlay_registers __iomem *
-intel_overlay_map_regs_atomic(struct intel_overlay *overlay)
-{
-	struct drm_i915_private *dev_priv = overlay->i915;
-	struct overlay_registers __iomem *regs;
-
-	if (OVERLAY_NEEDS_PHYSICAL(dev_priv))
-		/* Cast to make sparse happy, but it's wc memory anyway, so
-		 * equivalent to the wc io mapping on X86. */
-		regs = (struct overlay_registers __iomem *)
-			overlay->reg_bo->phys_handle->vaddr;
-	else
-		regs = io_mapping_map_atomic_wc(&dev_priv->ggtt.iomap,
-						overlay->flip_addr);
-
-	return regs;
-}
-
-static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay,
-					struct overlay_registers __iomem *regs)
-{
-	if (!OVERLAY_NEEDS_PHYSICAL(overlay->i915))
-		io_mapping_unmap_atomic(regs);
-}
-
 struct intel_overlay_error_state *
 intel_overlay_capture_error_state(struct drm_i915_private *dev_priv)
 {
 	struct intel_overlay *overlay = dev_priv->overlay;
 	struct intel_overlay_error_state *error;
-	struct overlay_registers __iomem *regs;
 
 	if (!overlay || !overlay->active)
 		return NULL;
@@ -1541,18 +1472,9 @@ intel_overlay_capture_error_state(struct drm_i915_private *dev_priv)
 	error->isr = I915_READ(ISR);
 	error->base = overlay->flip_addr;
 
-	regs = intel_overlay_map_regs_atomic(overlay);
-	if (!regs)
-		goto err;
-
-	memcpy_fromio(&error->regs, regs, sizeof(struct overlay_registers));
-	intel_overlay_unmap_regs_atomic(overlay, regs);
+	memcpy_fromio(&error->regs, overlay->regs, sizeof(error->regs));
 
 	return error;
-
-err:
-	kfree(error);
-	return NULL;
 }
 
 void
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c
index d65959ef0564..17235e940ca9 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/devinit/gm200.c
@@ -86,10 +86,8 @@ pmu_load(struct nv50_devinit *init, u8 type, bool post,
 	struct nvkm_bios *bios = subdev->device->bios;
 	struct nvbios_pmuR pmu;
 
-	if (!nvbios_pmuRm(bios, type, &pmu)) {
-		nvkm_error(subdev, "VBIOS PMU fuc %02x not found\n", type);
+	if (!nvbios_pmuRm(bios, type, &pmu))
 		return -EINVAL;
-	}
 
 	if (!post)
 		return 0;
@@ -124,29 +122,30 @@ gm200_devinit_post(struct nvkm_devinit *base, bool post)
 		return -EINVAL;
 	}
 
+	/* Upload DEVINIT application from VBIOS onto PMU. */
 	ret = pmu_load(init, 0x04, post, &exec, &args);
-	if (ret)
+	if (ret) {
+		nvkm_error(subdev, "VBIOS PMU/DEVINIT not found\n");
 		return ret;
+	}
 
-	/* upload first chunk of init data */
+	/* Upload tables required by opcodes in boot scripts. */
 	if (post) {
-		// devinit tables
 		u32 pmu = pmu_args(init, args + 0x08, 0x08);
 		u32 img = nvbios_rd16(bios, bit_I.offset + 0x14);
 		u32 len = nvbios_rd16(bios, bit_I.offset + 0x16);
 		pmu_data(init, pmu, img, len);
 	}
 
-	/* upload second chunk of init data */
+	/* Upload boot scripts. */
 	if (post) {
-		// devinit boot scripts
 		u32 pmu = pmu_args(init, args + 0x08, 0x10);
 		u32 img = nvbios_rd16(bios, bit_I.offset + 0x18);
 		u32 len = nvbios_rd16(bios, bit_I.offset + 0x1a);
 		pmu_data(init, pmu, img, len);
 	}
 
-	/* execute init tables */
+	/* Execute DEVINIT. */
 	if (post) {
 		nvkm_wr32(device, 0x10a040, 0x00005000);
 		pmu_exec(init, exec);
@@ -157,7 +156,9 @@ gm200_devinit_post(struct nvkm_devinit *base, bool post)
 			return -ETIMEDOUT;
 	}
 
-	/* load and execute some other ucode image (bios therm?) */
+	/* Optional: Execute PRE_OS application on PMU, which should at
+	 * least take care of fans until a full PMU has been loaded.
+	 */
 	pmu_load(init, 0x01, post, NULL, NULL);
 	return 0;
 }