Merge tag 'drm-intel-fixes-2017-01-05' of git://anongit.freedesktop.o…

…rg/git/drm-intel

Pull i915 drm fixes from Jani Nikula:
 "Here's a bunch of drm/i915 fixes for v4.10-rc3. It includes GVT-g
  fixes.

  My new year's resolution is to start using signed tags for pulls. If
  that feels like a déjà vu, it's ((new year's) resolution), not (new
  (year's resolution))"

[ Taking this directly from Jani because Dave Airlie is only partially
  connected right now.  - Linus ]

* tag 'drm-intel-fixes-2017-01-05' of git://anongit.freedesktop.org/git/drm-intel:
  drm/i915: Prevent timeline updates whilst performing reset
  drm/i915: Silence allocation failure during sg_trim()
  drm/i915: Don't clflush before release phys object
  drm/i915: Fix oops in overlay due to frontbuffer tracking
  drm/i915: Fix oopses in the overlay code due to i915_gem_active stuff
  drm/i915: Initialize overlay->last_flip properly
  drm/i915: Move the min_pixclk[] handling to the end of readout
  drm/i915: Force VDD off on the new power seqeuencer before starting to use it
  drm/i915/gvt: fix typo in cfg_space range check
  drm/i915/gvt: fix an issue in emulating cfg space PCI_COMMAND
  drm/i915/gvt/kvmgt: trival: code cleanup
  drm/i915/gvt/kvmgt: prevent double-release of vgpu
  drm/i915/gvt/kvmgt: check returned slot for gfn
  drm/i915/gvt/kvmgt: dereference the pointer within lock
  drm/i915/gvt: reset the GGTT entry when vGPU created
  drm/i915/gvt: fix an error in opregion handling

drivers/gpu/drm/i915/gvt/cfg_space.c

@@ -123,6 +123,7 @@ static int emulate_pci_command_write(struct intel_vgpu *vgpu,
 	u8 changed = old ^ new;
 	int ret;
 
+	memcpy(vgpu_cfg_space(vgpu) + offset, p_data, bytes);
 	if (!(changed & PCI_COMMAND_MEMORY))
 		return 0;
 
@@ -142,7 +143,6 @@ static int emulate_pci_command_write(struct intel_vgpu *vgpu,
 			return ret;
 	}
 
-	memcpy(vgpu_cfg_space(vgpu) + offset, p_data, bytes);
 	return 0;
 }
 
@@ -240,7 +240,7 @@ int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset,
 	if (WARN_ON(bytes > 4))
 		return -EINVAL;
 
-	if (WARN_ON(offset + bytes >= INTEL_GVT_MAX_CFG_SPACE_SZ))
+	if (WARN_ON(offset + bytes > INTEL_GVT_MAX_CFG_SPACE_SZ))
 		return -EINVAL;
 
 	/* First check if it's PCI_COMMAND */

drivers/gpu/drm/i915/gvt/gtt.c

@@ -1998,6 +1998,8 @@ int intel_vgpu_init_gtt(struct intel_vgpu *vgpu)
 	INIT_LIST_HEAD(&gtt->oos_page_list_head);
 	INIT_LIST_HEAD(&gtt->post_shadow_list_head);
 
+	intel_vgpu_reset_ggtt(vgpu);
+
 	ggtt_mm = intel_vgpu_create_mm(vgpu, INTEL_GVT_MM_GGTT,
 			NULL, 1, 0);
 	if (IS_ERR(ggtt_mm)) {
@@ -2206,6 +2208,7 @@ int intel_vgpu_g2v_destroy_ppgtt_mm(struct intel_vgpu *vgpu,
 int intel_gvt_init_gtt(struct intel_gvt *gvt)
 {
 	int ret;
+	void *page_addr;
 
 	gvt_dbg_core("init gtt\n");
 
@@ -2218,6 +2221,23 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt)
 		return -ENODEV;
 	}
 
+	gvt->gtt.scratch_ggtt_page =
+		alloc_page(GFP_KERNEL | GFP_ATOMIC | __GFP_ZERO);
+	if (!gvt->gtt.scratch_ggtt_page) {
+		gvt_err("fail to allocate scratch ggtt page\n");
+		return -ENOMEM;
+	}
+
+	page_addr = page_address(gvt->gtt.scratch_ggtt_page);
+
+	gvt->gtt.scratch_ggtt_mfn =
+		intel_gvt_hypervisor_virt_to_mfn(page_addr);
+	if (gvt->gtt.scratch_ggtt_mfn == INTEL_GVT_INVALID_ADDR) {
+		gvt_err("fail to translate scratch ggtt page\n");
+		__free_page(gvt->gtt.scratch_ggtt_page);
+		return -EFAULT;
+	}
+
 	if (enable_out_of_sync) {
 		ret = setup_spt_oos(gvt);
 		if (ret) {
@@ -2239,6 +2259,41 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt)
  */
 void intel_gvt_clean_gtt(struct intel_gvt *gvt)
 {
+	__free_page(gvt->gtt.scratch_ggtt_page);
+
 	if (enable_out_of_sync)
 		clean_spt_oos(gvt);
 }
+
+/**
+ * intel_vgpu_reset_ggtt - reset the GGTT entry
+ * @vgpu: a vGPU
+ *
+ * This function is called at the vGPU create stage
+ * to reset all the GGTT entries.
+ *
+ */
+void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+	u32 index;
+	u32 offset;
+	u32 num_entries;
+	struct intel_gvt_gtt_entry e;
+
+	memset(&e, 0, sizeof(struct intel_gvt_gtt_entry));
+	e.type = GTT_TYPE_GGTT_PTE;
+	ops->set_pfn(&e, gvt->gtt.scratch_ggtt_mfn);
+	e.val64 |= _PAGE_PRESENT;
+
+	index = vgpu_aperture_gmadr_base(vgpu) >> PAGE_SHIFT;
+	num_entries = vgpu_aperture_sz(vgpu) >> PAGE_SHIFT;
+	for (offset = 0; offset < num_entries; offset++)
+		ops->set_entry(NULL, &e, index + offset, false, 0, vgpu);
+
+	index = vgpu_hidden_gmadr_base(vgpu) >> PAGE_SHIFT;
+	num_entries = vgpu_hidden_sz(vgpu) >> PAGE_SHIFT;
+	for (offset = 0; offset < num_entries; offset++)
+		ops->set_entry(NULL, &e, index + offset, false, 0, vgpu);
+}

drivers/gpu/drm/i915/gvt/gtt.h

@@ -81,6 +81,9 @@ struct intel_gvt_gtt {
 	struct list_head oos_page_use_list_head;
 	struct list_head oos_page_free_list_head;
 	struct list_head mm_lru_list_head;
+
+	struct page *scratch_ggtt_page;
+	unsigned long scratch_ggtt_mfn;
 };
 
 enum {
@@ -202,6 +205,7 @@ struct intel_vgpu_gtt {
 
 extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu);
 extern void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu);
+void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu);
 
 extern int intel_gvt_init_gtt(struct intel_gvt *gvt);
 extern void intel_gvt_clean_gtt(struct intel_gvt *gvt);

drivers/gpu/drm/i915/gvt/gvt.h

@@ -175,6 +175,7 @@ struct intel_vgpu {
 		struct notifier_block group_notifier;
 		struct kvm *kvm;
 		struct work_struct release_work;
+		atomic_t released;
 	} vdev;
 #endif
 };

drivers/gpu/drm/i915/gvt/kvmgt.c

@@ -114,12 +114,15 @@ static struct gvt_dma *__gvt_cache_find(struct intel_vgpu *vgpu, gfn_t gfn)
 static kvm_pfn_t gvt_cache_find(struct intel_vgpu *vgpu, gfn_t gfn)
 {
 	struct gvt_dma *entry;
+	kvm_pfn_t pfn;
 
 	mutex_lock(&vgpu->vdev.cache_lock);
+
 	entry = __gvt_cache_find(vgpu, gfn);
-	mutex_unlock(&vgpu->vdev.cache_lock);
+	pfn = (entry == NULL) ? 0 : entry->pfn;
 
-	return entry == NULL ? 0 : entry->pfn;
+	mutex_unlock(&vgpu->vdev.cache_lock);
+	return pfn;
 }
 
 static void gvt_cache_add(struct intel_vgpu *vgpu, gfn_t gfn, kvm_pfn_t pfn)
@@ -497,7 +500,16 @@ static int intel_vgpu_open(struct mdev_device *mdev)
 		goto undo_iommu;
 	}
 
-	return kvmgt_guest_init(mdev);
+	ret = kvmgt_guest_init(mdev);
+	if (ret)
+		goto undo_group;
+
+	atomic_set(&vgpu->vdev.released, 0);
+	return ret;
+
+undo_group:
+	vfio_unregister_notifier(&mdev->dev, VFIO_GROUP_NOTIFY,
+					&vgpu->vdev.group_notifier);
 
 undo_iommu:
 	vfio_unregister_notifier(&mdev->dev, VFIO_IOMMU_NOTIFY,
@@ -509,17 +521,26 @@ static int intel_vgpu_open(struct mdev_device *mdev)
 static void __intel_vgpu_release(struct intel_vgpu *vgpu)
 {
 	struct kvmgt_guest_info *info;
+	int ret;
 
 	if (!handle_valid(vgpu->handle))
 		return;
 
-	vfio_unregister_notifier(&vgpu->vdev.mdev->dev, VFIO_IOMMU_NOTIFY,
+	if (atomic_cmpxchg(&vgpu->vdev.released, 0, 1))
+		return;
+
+	ret = vfio_unregister_notifier(&vgpu->vdev.mdev->dev, VFIO_IOMMU_NOTIFY,
 					&vgpu->vdev.iommu_notifier);
-	vfio_unregister_notifier(&vgpu->vdev.mdev->dev, VFIO_GROUP_NOTIFY,
+	WARN(ret, "vfio_unregister_notifier for iommu failed: %d\n", ret);
+
+	ret = vfio_unregister_notifier(&vgpu->vdev.mdev->dev, VFIO_GROUP_NOTIFY,
 					&vgpu->vdev.group_notifier);
+	WARN(ret, "vfio_unregister_notifier for group failed: %d\n", ret);
 
 	info = (struct kvmgt_guest_info *)vgpu->handle;
 	kvmgt_guest_exit(info);
+
+	vgpu->vdev.kvm = NULL;
 	vgpu->handle = 0;
 }
 
@@ -534,6 +555,7 @@ static void intel_vgpu_release_work(struct work_struct *work)
 {
 	struct intel_vgpu *vgpu = container_of(work, struct intel_vgpu,
 					vdev.release_work);
+
 	__intel_vgpu_release(vgpu);
 }
 
@@ -1134,6 +1156,10 @@ static int kvmgt_write_protect_add(unsigned long handle, u64 gfn)
 
 	idx = srcu_read_lock(&kvm->srcu);
 	slot = gfn_to_memslot(kvm, gfn);
+	if (!slot) {
+		srcu_read_unlock(&kvm->srcu, idx);
+		return -EINVAL;
+	}
 
 	spin_lock(&kvm->mmu_lock);
 
@@ -1164,6 +1190,10 @@ static int kvmgt_write_protect_remove(unsigned long handle, u64 gfn)
 
 	idx = srcu_read_lock(&kvm->srcu);
 	slot = gfn_to_memslot(kvm, gfn);
+	if (!slot) {
+		srcu_read_unlock(&kvm->srcu, idx);
+		return -EINVAL;
+	}
 
 	spin_lock(&kvm->mmu_lock);
 
@@ -1311,18 +1341,14 @@ static int kvmgt_guest_init(struct mdev_device *mdev)
 
 static bool kvmgt_guest_exit(struct kvmgt_guest_info *info)
 {
-	struct intel_vgpu *vgpu;
-
 	if (!info) {
 		gvt_err("kvmgt_guest_info invalid\n");
 		return false;
 	}
 
-	vgpu = info->vgpu;
-
 	kvm_page_track_unregister_notifier(info->kvm, &info->track_node);
 	kvmgt_protect_table_destroy(info);
-	gvt_cache_destroy(vgpu);
+	gvt_cache_destroy(info->vgpu);
 	vfree(info);
 
 	return true;

drivers/gpu/drm/i915/gvt/opregion.c

@@ -65,7 +65,7 @@ static int map_vgpu_opregion(struct intel_vgpu *vgpu, bool map)
 	int i, ret;
 
 	for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++) {
-		mfn = intel_gvt_hypervisor_virt_to_mfn(vgpu_opregion(vgpu)
+		mfn = intel_gvt_hypervisor_virt_to_mfn(vgpu_opregion(vgpu)->va
 			+ i * PAGE_SIZE);
 		if (mfn == INTEL_GVT_INVALID_ADDR) {
 			gvt_err("fail to get MFN from VA\n");

drivers/gpu/drm/i915/i915_gem.c

@@ -244,14 +244,16 @@ i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj)
 
 static void
 __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
-				struct sg_table *pages)
+				struct sg_table *pages,
+				bool needs_clflush)
 {
 	GEM_BUG_ON(obj->mm.madv == __I915_MADV_PURGED);
 
 	if (obj->mm.madv == I915_MADV_DONTNEED)
 		obj->mm.dirty = false;
 
-	if ((obj->base.read_domains & I915_GEM_DOMAIN_CPU) == 0 &&
+	if (needs_clflush &&
+	    (obj->base.read_domains & I915_GEM_DOMAIN_CPU) == 0 &&
 	    !cpu_cache_is_coherent(obj->base.dev, obj->cache_level))
 		drm_clflush_sg(pages);
 
@@ -263,7 +265,7 @@ static void
 i915_gem_object_put_pages_phys(struct drm_i915_gem_object *obj,
 			       struct sg_table *pages)
 {
-	__i915_gem_object_release_shmem(obj, pages);
+	__i915_gem_object_release_shmem(obj, pages, false);
 
 	if (obj->mm.dirty) {
 		struct address_space *mapping = obj->base.filp->f_mapping;
@@ -2231,7 +2233,7 @@ i915_gem_object_put_pages_gtt(struct drm_i915_gem_object *obj,
 	struct sgt_iter sgt_iter;
 	struct page *page;
 
-	__i915_gem_object_release_shmem(obj, pages);
+	__i915_gem_object_release_shmem(obj, pages, true);
 
 	i915_gem_gtt_finish_pages(obj, pages);
 
@@ -2322,7 +2324,7 @@ static void i915_sg_trim(struct sg_table *orig_st)
 	if (orig_st->nents == orig_st->orig_nents)
 		return;
 
-	if (sg_alloc_table(&new_st, orig_st->nents, GFP_KERNEL))
+	if (sg_alloc_table(&new_st, orig_st->nents, GFP_KERNEL | __GFP_NOWARN))
 		return;
 
 	new_sg = new_st.sgl;
@@ -2728,6 +2730,7 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
 	struct drm_i915_gem_request *request;
 	struct i915_gem_context *incomplete_ctx;
 	struct intel_timeline *timeline;
+	unsigned long flags;
 	bool ring_hung;
 
 	if (engine->irq_seqno_barrier)
@@ -2763,13 +2766,20 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
 	if (i915_gem_context_is_default(incomplete_ctx))
 		return;
 
+	timeline = i915_gem_context_lookup_timeline(incomplete_ctx, engine);
+
+	spin_lock_irqsave(&engine->timeline->lock, flags);
+	spin_lock(&timeline->lock);
+
 	list_for_each_entry_continue(request, &engine->timeline->requests, link)
 		if (request->ctx == incomplete_ctx)
 			reset_request(request);
 
-	timeline = i915_gem_context_lookup_timeline(incomplete_ctx, engine);
 	list_for_each_entry(request, &timeline->requests, link)
 		reset_request(request);
+
+	spin_unlock(&timeline->lock);
+	spin_unlock_irqrestore(&engine->timeline->lock, flags);
 }
 
 void i915_gem_reset(struct drm_i915_private *dev_priv)

drivers/gpu/drm/i915/i915_gem_request.h

@@ -413,6 +413,25 @@ i915_gem_active_set(struct i915_gem_active *active,
 	rcu_assign_pointer(active->request, request);
 }
 
+/**
+ * i915_gem_active_set_retire_fn - updates the retirement callback
+ * @active - the active tracker
+ * @fn - the routine called when the request is retired
+ * @mutex - struct_mutex used to guard retirements
+ *
+ * i915_gem_active_set_retire_fn() updates the function pointer that
+ * is called when the final request associated with the @active tracker
+ * is retired.
+ */
+static inline void
+i915_gem_active_set_retire_fn(struct i915_gem_active *active,
+			      i915_gem_retire_fn fn,
+			      struct mutex *mutex)
+{
+	lockdep_assert_held(mutex);
+	active->retire = fn ?: i915_gem_retire_noop;
+}
+
 static inline struct drm_i915_gem_request *
 __i915_gem_active_peek(const struct i915_gem_active *active)
 {