Merge tag 'gvt-fixes-2023-01-05' of https://github.com/intel/gvt-linux into drm-intel-fixes

gvt-fixes-2023-01-05

- Fix one missed unpin in error of intel_vgpu_shadow_mm_pin()
- Fix two debugfs destroy oops issues for vgpu and gvt entries
- Fix one potential double free issue in gtt shadow pt code
- Fix to use atomic bit flag for vgpu status
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
From: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/Y7YWoFpz4plnSLCd@zhen-hp.sh.intel.com

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

@@ -151,6 +151,22 @@ DEFINE_SIMPLE_ATTRIBUTE(vgpu_scan_nonprivbb_fops,
 			vgpu_scan_nonprivbb_get, vgpu_scan_nonprivbb_set,
 			"0x%llx\n");
 
+static int vgpu_status_get(void *data, u64 *val)
+{
+	struct intel_vgpu *vgpu = (struct intel_vgpu *)data;
+
+	*val = 0;
+
+	if (test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
+		*val |= (1 << INTEL_VGPU_STATUS_ATTACHED);
+	if (test_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status))
+		*val |= (1 << INTEL_VGPU_STATUS_ACTIVE);
+
+	return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(vgpu_status_fops, vgpu_status_get, NULL, "0x%llx\n");
+
 /**
  * intel_gvt_debugfs_add_vgpu - register debugfs entries for a vGPU
  * @vgpu: a vGPU
@@ -162,11 +178,12 @@ void intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu)
 	snprintf(name, 16, "vgpu%d", vgpu->id);
 	vgpu->debugfs = debugfs_create_dir(name, vgpu->gvt->debugfs_root);
 
-	debugfs_create_bool("active", 0444, vgpu->debugfs, &vgpu->active);
 	debugfs_create_file("mmio_diff", 0444, vgpu->debugfs, vgpu,
 			    &vgpu_mmio_diff_fops);
 	debugfs_create_file("scan_nonprivbb", 0644, vgpu->debugfs, vgpu,
 			    &vgpu_scan_nonprivbb_fops);
+	debugfs_create_file("status", 0644, vgpu->debugfs, vgpu,
+			    &vgpu_status_fops);
 }
 
 /**
@@ -175,8 +192,13 @@ void intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu)
  */
 void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu)
 {
-	debugfs_remove_recursive(vgpu->debugfs);
-	vgpu->debugfs = NULL;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_minor *minor = gvt->gt->i915->drm.primary;
+
+	if (minor->debugfs_root && gvt->debugfs_root) {
+		debugfs_remove_recursive(vgpu->debugfs);
+		vgpu->debugfs = NULL;
+	}
 }
 
 /**
@@ -199,6 +221,10 @@ void intel_gvt_debugfs_init(struct intel_gvt *gvt)
  */
 void intel_gvt_debugfs_clean(struct intel_gvt *gvt)
 {
-	debugfs_remove_recursive(gvt->debugfs_root);
-	gvt->debugfs_root = NULL;
+	struct drm_minor *minor = gvt->gt->i915->drm.primary;
+
+	if (minor->debugfs_root) {
+		debugfs_remove_recursive(gvt->debugfs_root);
+		gvt->debugfs_root = NULL;
+	}
 }

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

@@ -134,7 +134,8 @@ static void dmabuf_gem_object_free(struct kref *kref)
 	struct list_head *pos;
 	struct intel_vgpu_dmabuf_obj *dmabuf_obj;
 
-	if (vgpu && vgpu->active && !list_empty(&vgpu->dmabuf_obj_list_head)) {
+	if (vgpu && test_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status) &&
+	    !list_empty(&vgpu->dmabuf_obj_list_head)) {
 		list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
 			dmabuf_obj = list_entry(pos, struct intel_vgpu_dmabuf_obj, list);
 			if (dmabuf_obj == obj) {

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

@@ -55,7 +55,7 @@ static bool intel_gvt_is_valid_gfn(struct intel_vgpu *vgpu, unsigned long gfn)
 	int idx;
 	bool ret;
 
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return false;
 
 	idx = srcu_read_lock(&kvm->srcu);
@@ -1178,7 +1178,7 @@ static int is_2MB_gtt_possible(struct intel_vgpu *vgpu,
 	if (!HAS_PAGE_SIZES(vgpu->gvt->gt->i915, I915_GTT_PAGE_SIZE_2M))
 		return 0;
 
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return -EINVAL;
 	pfn = gfn_to_pfn(vgpu->vfio_device.kvm, ops->get_pfn(entry));
 	if (is_error_noslot_pfn(pfn))
@@ -1209,10 +1209,8 @@ static int split_2MB_gtt_entry(struct intel_vgpu *vgpu,
 	for_each_shadow_entry(sub_spt, &sub_se, sub_index) {
 		ret = intel_gvt_dma_map_guest_page(vgpu, start_gfn + sub_index,
 						   PAGE_SIZE, &dma_addr);
-		if (ret) {
-			ppgtt_invalidate_spt(spt);
-			return ret;
-		}
+		if (ret)
+			goto err;
 		sub_se.val64 = se->val64;
 
 		/* Copy the PAT field from PDE. */
@@ -1231,6 +1229,17 @@ static int split_2MB_gtt_entry(struct intel_vgpu *vgpu,
 	ops->set_pfn(se, sub_spt->shadow_page.mfn);
 	ppgtt_set_shadow_entry(spt, se, index);
 	return 0;
+err:
+	/* Cancel the existing addess mappings of DMA addr. */
+	for_each_present_shadow_entry(sub_spt, &sub_se, sub_index) {
+		gvt_vdbg_mm("invalidate 4K entry\n");
+		ppgtt_invalidate_pte(sub_spt, &sub_se);
+	}
+	/* Release the new allocated spt. */
+	trace_spt_change(sub_spt->vgpu->id, "release", sub_spt,
+		sub_spt->guest_page.gfn, sub_spt->shadow_page.type);
+	ppgtt_free_spt(sub_spt);
+	return ret;
 }
 
 static int split_64KB_gtt_entry(struct intel_vgpu *vgpu,

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

@@ -172,13 +172,18 @@ struct intel_vgpu_submission {
 
 #define KVMGT_DEBUGFS_FILENAME		"kvmgt_nr_cache_entries"
 
+enum {
+	INTEL_VGPU_STATUS_ATTACHED = 0,
+	INTEL_VGPU_STATUS_ACTIVE,
+	INTEL_VGPU_STATUS_NR_BITS,
+};
+
 struct intel_vgpu {
 	struct vfio_device vfio_device;
 	struct intel_gvt *gvt;
 	struct mutex vgpu_lock;
 	int id;
-	bool active;
-	bool attached;
+	DECLARE_BITMAP(status, INTEL_VGPU_STATUS_NR_BITS);
 	bool pv_notified;
 	bool failsafe;
 	unsigned int resetting_eng;
@@ -467,7 +472,7 @@ void intel_vgpu_write_fence(struct intel_vgpu *vgpu,
 
 #define for_each_active_vgpu(gvt, vgpu, id) \
 	idr_for_each_entry((&(gvt)->vgpu_idr), (vgpu), (id)) \
-		for_each_if(vgpu->active)
+		for_each_if(test_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status))
 
 static inline void intel_vgpu_write_pci_bar(struct intel_vgpu *vgpu,
 					    u32 offset, u32 val, bool low)
@@ -725,7 +730,7 @@ static inline bool intel_gvt_mmio_is_cmd_write_patch(
 static inline int intel_gvt_read_gpa(struct intel_vgpu *vgpu, unsigned long gpa,
 		void *buf, unsigned long len)
 {
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return -ESRCH;
 	return vfio_dma_rw(&vgpu->vfio_device, gpa, buf, len, false);
 }
@@ -743,7 +748,7 @@ static inline int intel_gvt_read_gpa(struct intel_vgpu *vgpu, unsigned long gpa,
 static inline int intel_gvt_write_gpa(struct intel_vgpu *vgpu,
 		unsigned long gpa, void *buf, unsigned long len)
 {
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return -ESRCH;
 	return vfio_dma_rw(&vgpu->vfio_device, gpa, buf, len, true);
 }

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

@@ -433,7 +433,7 @@ static int inject_virtual_interrupt(struct intel_vgpu *vgpu)
 	 * enabled by guest. so if msi_trigger is null, success is still
 	 * returned and don't inject interrupt into guest.
 	 */
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return -ESRCH;
 	if (vgpu->msi_trigger && eventfd_signal(vgpu->msi_trigger, 1) != 1)
 		return -EFAULT;

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

@@ -638,7 +638,7 @@ static bool __kvmgt_vgpu_exist(struct intel_vgpu *vgpu)
 
 	mutex_lock(&vgpu->gvt->lock);
 	for_each_active_vgpu(vgpu->gvt, itr, id) {
-		if (!itr->attached)
+		if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, itr->status))
 			continue;
 
 		if (vgpu->vfio_device.kvm == itr->vfio_device.kvm) {
@@ -655,9 +655,6 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
 {
 	struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
 
-	if (vgpu->attached)
-		return -EEXIST;
-
 	if (!vgpu->vfio_device.kvm ||
 	    vgpu->vfio_device.kvm->mm != current->mm) {
 		gvt_vgpu_err("KVM is required to use Intel vGPU\n");
@@ -667,14 +664,14 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
 	if (__kvmgt_vgpu_exist(vgpu))
 		return -EEXIST;
 
-	vgpu->attached = true;
-
 	vgpu->track_node.track_write = kvmgt_page_track_write;
 	vgpu->track_node.track_flush_slot = kvmgt_page_track_flush_slot;
 	kvm_get_kvm(vgpu->vfio_device.kvm);
 	kvm_page_track_register_notifier(vgpu->vfio_device.kvm,
 					 &vgpu->track_node);
 
+	set_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status);
+
 	debugfs_create_ulong(KVMGT_DEBUGFS_FILENAME, 0444, vgpu->debugfs,
 			     &vgpu->nr_cache_entries);
 
@@ -698,11 +695,10 @@ static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
 {
 	struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
 
-	if (!vgpu->attached)
-		return;
-
 	intel_gvt_release_vgpu(vgpu);
 
+	clear_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status);
+
 	debugfs_remove(debugfs_lookup(KVMGT_DEBUGFS_FILENAME, vgpu->debugfs));
 
 	kvm_page_track_unregister_notifier(vgpu->vfio_device.kvm,
@@ -718,8 +714,6 @@ static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
 	vgpu->dma_addr_cache = RB_ROOT;
 
 	intel_vgpu_release_msi_eventfd_ctx(vgpu);
-
-	vgpu->attached = false;
 }
 
 static u64 intel_vgpu_get_bar_addr(struct intel_vgpu *vgpu, int bar)
@@ -1512,9 +1506,6 @@ static void intel_vgpu_remove(struct mdev_device *mdev)
 {
 	struct intel_vgpu *vgpu = dev_get_drvdata(&mdev->dev);
 
-	if (WARN_ON_ONCE(vgpu->attached))
-		return;
-
 	vfio_unregister_group_dev(&vgpu->vfio_device);
 	vfio_put_device(&vgpu->vfio_device);
 }
@@ -1559,7 +1550,7 @@ int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn)
 	struct kvm_memory_slot *slot;
 	int idx;
 
-	if (!info->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, info->status))
 		return -ESRCH;
 
 	idx = srcu_read_lock(&kvm->srcu);
@@ -1589,8 +1580,8 @@ int intel_gvt_page_track_remove(struct intel_vgpu *info, u64 gfn)
 	struct kvm_memory_slot *slot;
 	int idx;
 
-	if (!info->attached)
-		return 0;
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, info->status))
+		return -ESRCH;
 
 	idx = srcu_read_lock(&kvm->srcu);
 	slot = gfn_to_memslot(kvm, gfn);
@@ -1668,7 +1659,7 @@ int intel_gvt_dma_map_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
 	struct gvt_dma *entry;
 	int ret;
 
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return -EINVAL;
 
 	mutex_lock(&vgpu->cache_lock);
@@ -1714,8 +1705,8 @@ int intel_gvt_dma_pin_guest_page(struct intel_vgpu *vgpu, dma_addr_t dma_addr)
 	struct gvt_dma *entry;
 	int ret = 0;
 
-	if (!vgpu->attached)
-		return -ENODEV;
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
+		return -EINVAL;
 
 	mutex_lock(&vgpu->cache_lock);
 	entry = __gvt_cache_find_dma_addr(vgpu, dma_addr);
@@ -1742,7 +1733,7 @@ void intel_gvt_dma_unmap_guest_page(struct intel_vgpu *vgpu,
 {
 	struct gvt_dma *entry;
 
-	if (!vgpu->attached)
+	if (!test_bit(INTEL_VGPU_STATUS_ATTACHED, vgpu->status))
 		return;
 
 	mutex_lock(&vgpu->cache_lock);
@@ -1778,7 +1769,7 @@ static void intel_gvt_test_and_emulate_vblank(struct intel_gvt *gvt)
 	idr_for_each_entry((&(gvt)->vgpu_idr), (vgpu), (id)) {
 		if (test_and_clear_bit(INTEL_GVT_REQUEST_EMULATE_VBLANK + id,
 				       (void *)&gvt->service_request)) {
-			if (vgpu->active)
+			if (test_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status))
 				intel_vgpu_emulate_vblank(vgpu);
 		}
 	}

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

@@ -695,6 +695,7 @@ intel_vgpu_shadow_mm_pin(struct intel_vgpu_workload *workload)
 
 	if (workload->shadow_mm->type != INTEL_GVT_MM_PPGTT ||
 	    !workload->shadow_mm->ppgtt_mm.shadowed) {
+		intel_vgpu_unpin_mm(workload->shadow_mm);
 		gvt_vgpu_err("workload shadow ppgtt isn't ready\n");
 		return -EINVAL;
 	}
@@ -865,7 +866,8 @@ pick_next_workload(struct intel_gvt *gvt, struct intel_engine_cs *engine)
 		goto out;
 	}
 
-	if (!scheduler->current_vgpu->active ||
+	if (!test_bit(INTEL_VGPU_STATUS_ACTIVE,
+		      scheduler->current_vgpu->status) ||
 	    list_empty(workload_q_head(scheduler->current_vgpu, engine)))
 		goto out;
 

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

@@ -166,9 +166,7 @@ void intel_gvt_clean_vgpu_types(struct intel_gvt *gvt)
  */
 void intel_gvt_activate_vgpu(struct intel_vgpu *vgpu)
 {
-	mutex_lock(&vgpu->vgpu_lock);
-	vgpu->active = true;
-	mutex_unlock(&vgpu->vgpu_lock);
+	set_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status);
 }
 
 /**
@@ -183,7 +181,7 @@ void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu)
 {
 	mutex_lock(&vgpu->vgpu_lock);
 
-	vgpu->active = false;
+	clear_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status);
 
 	if (atomic_read(&vgpu->submission.running_workload_num)) {
 		mutex_unlock(&vgpu->vgpu_lock);
@@ -228,7 +226,8 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 	struct intel_gvt *gvt = vgpu->gvt;
 	struct drm_i915_private *i915 = gvt->gt->i915;
 
-	drm_WARN(&i915->drm, vgpu->active, "vGPU is still active!\n");
+	drm_WARN(&i915->drm, test_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status),
+		 "vGPU is still active!\n");
 
 	/*
 	 * remove idr first so later clean can judge if need to stop
@@ -285,8 +284,7 @@ struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt)
 	if (ret)
 		goto out_free_vgpu;
 
-	vgpu->active = false;
-
+	clear_bit(INTEL_VGPU_STATUS_ACTIVE, vgpu->status);
 	return vgpu;
 
 out_free_vgpu: