Merge tag 'gvt-next-2017-11-16' of https://github.com/intel/gvt-linux into drm-intel-next-queued

gvt-next-2017-11-16

- CSB HWSP update support (Weinan)
- GVT debug helpers, dyndbg and debugfs (Chuanxiao, Shuo)
- full virtualized opregion (Xiaolin)
- VM health check for sane fallback (Fred)
- workload submission code refactor for future enabling (Zhi)
- Updated repo URL in MAINTAINERS (Zhenyu)
- other many misc fixes
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171116092007.ww5bvfx7rf36bjmn@zhen-hp.sh.intel.com

MAINTAINERS

@@ -6928,7 +6928,7 @@ M:	Zhi Wang <zhi.a.wang@intel.com>
 L:	intel-gvt-dev@lists.freedesktop.org
 L:	intel-gfx@lists.freedesktop.org
 W:	https://01.org/igvt-g
-T:	git https://github.com/01org/gvt-linux.git
+T:	git https://github.com/intel/gvt-linux.git
 S:	Supported
 F:	drivers/gpu/drm/i915/gvt/
 

drivers/gpu/drm/i915/gvt/Makefile

 GVT_DIR := gvt
 GVT_SOURCE := gvt.o aperture_gm.o handlers.o vgpu.o trace_points.o firmware.o \
 	interrupt.o gtt.o cfg_space.o opregion.o mmio.o display.o edid.o \
-	execlist.o scheduler.o sched_policy.o render.o cmd_parser.o
+	execlist.o scheduler.o sched_policy.o render.o cmd_parser.o debugfs.o
 
 ccflags-y				+= -I$(src) -I$(src)/$(GVT_DIR)
 i915-y					+= $(addprefix $(GVT_DIR)/, $(GVT_SOURCE))

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

@@ -208,6 +208,20 @@ static int emulate_pci_command_write(struct intel_vgpu *vgpu,
 	return 0;
 }
 
+static int emulate_pci_rom_bar_write(struct intel_vgpu *vgpu,
+	unsigned int offset, void *p_data, unsigned int bytes)
+{
+	u32 *pval = (u32 *)(vgpu_cfg_space(vgpu) + offset);
+	u32 new = *(u32 *)(p_data);
+
+	if ((new & PCI_ROM_ADDRESS_MASK) == PCI_ROM_ADDRESS_MASK)
+		/* We don't have rom, return size of 0. */
+		*pval = 0;
+	else
+		vgpu_pci_cfg_mem_write(vgpu, offset, p_data, bytes);
+	return 0;
+}
+
 static int emulate_pci_bar_write(struct intel_vgpu *vgpu, unsigned int offset,
 	void *p_data, unsigned int bytes)
 {
@@ -300,6 +314,11 @@ int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset,
 	}
 
 	switch (rounddown(offset, 4)) {
+	case PCI_ROM_ADDRESS:
+		if (WARN_ON(!IS_ALIGNED(offset, 4)))
+			return -EINVAL;
+		return emulate_pci_rom_bar_write(vgpu, offset, p_data, bytes);
+
 	case PCI_BASE_ADDRESS_0 ... PCI_BASE_ADDRESS_5:
 		if (WARN_ON(!IS_ALIGNED(offset, 4)))
 			return -EINVAL;
@@ -375,6 +394,8 @@ void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu,
 				pci_resource_len(gvt->dev_priv->drm.pdev, 0);
 	vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].size =
 				pci_resource_len(gvt->dev_priv->drm.pdev, 2);
+
+	memset(vgpu_cfg_space(vgpu) + PCI_ROM_ADDRESS, 0, 4);
 }
 
 /**

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

@@ -25,41 +25,41 @@
 #define __GVT_DEBUG_H__
 
 #define gvt_err(fmt, args...) \
-	DRM_ERROR("gvt: "fmt, ##args)
+	pr_err("gvt: "fmt, ##args)
 
 #define gvt_vgpu_err(fmt, args...)					\
 do {									\
 	if (IS_ERR_OR_NULL(vgpu))					\
-		DRM_DEBUG_DRIVER("gvt: "fmt, ##args);			\
+		pr_err("gvt: "fmt, ##args);			\
 	else								\
-		DRM_DEBUG_DRIVER("gvt: vgpu %d: "fmt, vgpu->id, ##args);\
+		pr_err("gvt: vgpu %d: "fmt, vgpu->id, ##args);\
 } while (0)
 
 #define gvt_dbg_core(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: core: "fmt, ##args)
+	pr_debug("gvt: core: "fmt, ##args)
 
 #define gvt_dbg_irq(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: irq: "fmt, ##args)
+	pr_debug("gvt: irq: "fmt, ##args)
 
 #define gvt_dbg_mm(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: mm: "fmt, ##args)
+	pr_debug("gvt: mm: "fmt, ##args)
 
 #define gvt_dbg_mmio(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: mmio: "fmt, ##args)
+	pr_debug("gvt: mmio: "fmt, ##args)
 
 #define gvt_dbg_dpy(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: dpy: "fmt, ##args)
+	pr_debug("gvt: dpy: "fmt, ##args)
 
 #define gvt_dbg_el(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: el: "fmt, ##args)
+	pr_debug("gvt: el: "fmt, ##args)
 
 #define gvt_dbg_sched(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: sched: "fmt, ##args)
+	pr_debug("gvt: sched: "fmt, ##args)
 
 #define gvt_dbg_render(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: render: "fmt, ##args)
+	pr_debug("gvt: render: "fmt, ##args)
 
 #define gvt_dbg_cmd(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: cmd: "fmt, ##args)
+	pr_debug("gvt: cmd: "fmt, ##args)
 
 #endif

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

+/*
+ * Copyright(c) 2011-2017 Intel Corporation. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/debugfs.h>
+#include <linux/list_sort.h>
+#include "i915_drv.h"
+#include "gvt.h"
+
+struct mmio_diff_param {
+	struct intel_vgpu *vgpu;
+	int total;
+	int diff;
+	struct list_head diff_mmio_list;
+};
+
+struct diff_mmio {
+	struct list_head node;
+	u32 offset;
+	u32 preg;
+	u32 vreg;
+};
+
+/* Compare two diff_mmio items. */
+static int mmio_offset_compare(void *priv,
+	struct list_head *a, struct list_head *b)
+{
+	struct diff_mmio *ma;
+	struct diff_mmio *mb;
+
+	ma = container_of(a, struct diff_mmio, node);
+	mb = container_of(b, struct diff_mmio, node);
+	if (ma->offset < mb->offset)
+		return -1;
+	else if (ma->offset > mb->offset)
+		return 1;
+	return 0;
+}
+
+static inline int mmio_diff_handler(struct intel_gvt *gvt,
+				    u32 offset, void *data)
+{
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct mmio_diff_param *param = data;
+	struct diff_mmio *node;
+	u32 preg, vreg;
+
+	preg = I915_READ_NOTRACE(_MMIO(offset));
+	vreg = vgpu_vreg(param->vgpu, offset);
+
+	if (preg != vreg) {
+		node = kmalloc(sizeof(*node), GFP_KERNEL);
+		if (!node)
+			return -ENOMEM;
+
+		node->offset = offset;
+		node->preg = preg;
+		node->vreg = vreg;
+		list_add(&node->node, &param->diff_mmio_list);
+		param->diff++;
+	}
+	param->total++;
+	return 0;
+}
+
+/* Show the all the different values of tracked mmio. */
+static int vgpu_mmio_diff_show(struct seq_file *s, void *unused)
+{
+	struct intel_vgpu *vgpu = s->private;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct mmio_diff_param param = {
+		.vgpu = vgpu,
+		.total = 0,
+		.diff = 0,
+	};
+	struct diff_mmio *node, *next;
+
+	INIT_LIST_HEAD(&param.diff_mmio_list);
+
+	mutex_lock(&gvt->lock);
+	spin_lock_bh(&gvt->scheduler.mmio_context_lock);
+
+	mmio_hw_access_pre(gvt->dev_priv);
+	/* Recognize all the diff mmios to list. */
+	intel_gvt_for_each_tracked_mmio(gvt, mmio_diff_handler, &param);
+	mmio_hw_access_post(gvt->dev_priv);
+
+	spin_unlock_bh(&gvt->scheduler.mmio_context_lock);
+	mutex_unlock(&gvt->lock);
+
+	/* In an ascending order by mmio offset. */
+	list_sort(NULL, &param.diff_mmio_list, mmio_offset_compare);
+
+	seq_printf(s, "%-8s %-8s %-8s %-8s\n", "Offset", "HW", "vGPU", "Diff");
+	list_for_each_entry_safe(node, next, &param.diff_mmio_list, node) {
+		u32 diff = node->preg ^ node->vreg;
+
+		seq_printf(s, "%08x %08x %08x %*pbl\n",
+			   node->offset, node->preg, node->vreg,
+			   32, &diff);
+		list_del(&node->node);
+		kfree(node);
+	}
+	seq_printf(s, "Total: %d, Diff: %d\n", param.total, param.diff);
+	return 0;
+}
+
+static int vgpu_mmio_diff_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, vgpu_mmio_diff_show, inode->i_private);
+}
+
+static const struct file_operations vgpu_mmio_diff_fops = {
+	.open		= vgpu_mmio_diff_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+/**
+ * intel_gvt_debugfs_add_vgpu - register debugfs entries for a vGPU
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu)
+{
+	struct dentry *ent;
+	char name[10] = "";
+
+	sprintf(name, "vgpu%d", vgpu->id);
+	vgpu->debugfs = debugfs_create_dir(name, vgpu->gvt->debugfs_root);
+	if (!vgpu->debugfs)
+		return -ENOMEM;
+
+	ent = debugfs_create_bool("active", 0444, vgpu->debugfs,
+				  &vgpu->active);
+	if (!ent)
+		return -ENOMEM;
+
+	ent = debugfs_create_file("mmio_diff", 0444, vgpu->debugfs,
+				  vgpu, &vgpu_mmio_diff_fops);
+	if (!ent)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * intel_gvt_debugfs_remove_vgpu - remove debugfs entries of a vGPU
+ * @vgpu: a vGPU
+ */
+void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu)
+{
+	debugfs_remove_recursive(vgpu->debugfs);
+	vgpu->debugfs = NULL;
+}
+
+/**
+ * intel_gvt_debugfs_init - register gvt debugfs root entry
+ * @gvt: GVT device
+ *
+ * Returns:
+ * zero on success, negative if failed.
+ */
+int intel_gvt_debugfs_init(struct intel_gvt *gvt)
+{
+	struct drm_minor *minor = gvt->dev_priv->drm.primary;
+	struct dentry *ent;
+
+	gvt->debugfs_root = debugfs_create_dir("gvt", minor->debugfs_root);
+	if (!gvt->debugfs_root) {
+		gvt_err("Cannot create debugfs dir\n");
+		return -ENOMEM;
+	}
+
+	ent = debugfs_create_ulong("num_tracked_mmio", 0444, gvt->debugfs_root,
+				   &gvt->mmio.num_tracked_mmio);
+	if (!ent)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * intel_gvt_debugfs_clean - remove debugfs entries
+ * @gvt: GVT device
+ */
+void intel_gvt_debugfs_clean(struct intel_gvt *gvt)
+{
+	debugfs_remove_recursive(gvt->debugfs_root);
+	gvt->debugfs_root = NULL;
+}

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

@@ -36,10 +36,6 @@
 #define _GVT_EXECLIST_H_
 
 struct execlist_ctx_descriptor_format {
-	union {
-		u32 udw;
-		u32 context_id;
-	};
 	union {
 		u32 ldw;
 		struct {
@@ -54,6 +50,10 @@ struct execlist_ctx_descriptor_format {
 			u32 lrca                   : 20;
 		};
 	};
+	union {
+		u32 udw;
+		u32 context_id;
+	};
 };
 
 struct execlist_status_format {

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

@@ -66,20 +66,23 @@ static struct bin_attribute firmware_attr = {
 	.mmap = NULL,
 };
 
-static int expose_firmware_sysfs(struct intel_gvt *gvt)
+static int mmio_snapshot_handler(struct intel_gvt *gvt, u32 offset, void *data)
 {
 	struct drm_i915_private *dev_priv = gvt->dev_priv;
+
+	*(u32 *)(data + offset) = I915_READ_NOTRACE(_MMIO(offset));
+	return 0;
+}
+
+static int expose_firmware_sysfs(struct intel_gvt *gvt)
+{
 	struct intel_gvt_device_info *info = &gvt->device_info;
 	struct pci_dev *pdev = gvt->dev_priv->drm.pdev;
-	struct intel_gvt_mmio_info *e;
-	struct gvt_mmio_block *block = gvt->mmio.mmio_block;
-	int num = gvt->mmio.num_mmio_block;
 	struct gvt_firmware_header *h;
 	void *firmware;
 	void *p;
 	unsigned long size, crc32_start;
-	int i, j;
-	int ret;
+	int i, ret;
 
 	size = sizeof(*h) + info->mmio_size + info->cfg_space_size;
 	firmware = vzalloc(size);
@@ -104,15 +107,8 @@ static int expose_firmware_sysfs(struct intel_gvt *gvt)
 
 	p = firmware + h->mmio_offset;
 
-	hash_for_each(gvt->mmio.mmio_info_table, i, e, node)
-		*(u32 *)(p + e->offset) = I915_READ_NOTRACE(_MMIO(e->offset));
-
-	for (i = 0; i < num; i++, block++) {
-		for (j = 0; j < block->size; j += 4)
-			*(u32 *)(p + INTEL_GVT_MMIO_OFFSET(block->offset) + j) =
-				I915_READ_NOTRACE(_MMIO(INTEL_GVT_MMIO_OFFSET(
-							block->offset) + j));
-	}
+	/* Take a snapshot of hw mmio registers. */
+	intel_gvt_for_each_tracked_mmio(gvt, mmio_snapshot_handler, p);
 
 	memcpy(gvt->firmware.mmio, p, info->mmio_size);
 

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

@@ -34,9 +34,8 @@
 #ifndef _GVT_GTT_H_
 #define _GVT_GTT_H_
 
-#define GTT_PAGE_SHIFT		12
-#define GTT_PAGE_SIZE		(1UL << GTT_PAGE_SHIFT)
-#define GTT_PAGE_MASK		(~(GTT_PAGE_SIZE-1))
+#define I915_GTT_PAGE_SHIFT         12
+#define I915_GTT_PAGE_MASK		(~(I915_GTT_PAGE_SIZE - 1))
 
 struct intel_vgpu_mm;
 
@@ -63,6 +62,7 @@ struct intel_gvt_gtt_pte_ops {
 			 struct intel_vgpu *vgpu);
 	bool (*test_present)(struct intel_gvt_gtt_entry *e);
 	void (*clear_present)(struct intel_gvt_gtt_entry *e);
+	void (*set_present)(struct intel_gvt_gtt_entry *e);
 	bool (*test_pse)(struct intel_gvt_gtt_entry *e);
 	void (*set_pfn)(struct intel_gvt_gtt_entry *e, unsigned long pfn);
 	unsigned long (*get_pfn)(struct intel_gvt_gtt_entry *e);
@@ -86,8 +86,8 @@ struct intel_gvt_gtt {
 	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;
+	struct page *scratch_page;
+	unsigned long scratch_mfn;
 };
 
 enum {
@@ -193,18 +193,16 @@ struct intel_vgpu_scratch_pt {
 	unsigned long page_mfn;
 };
 
-
 struct intel_vgpu_gtt {
 	struct intel_vgpu_mm *ggtt_mm;
 	unsigned long active_ppgtt_mm_bitmap;
 	struct list_head mm_list_head;
 	DECLARE_HASHTABLE(shadow_page_hash_table, INTEL_GVT_GTT_HASH_BITS);
-	DECLARE_HASHTABLE(guest_page_hash_table, INTEL_GVT_GTT_HASH_BITS);
-	atomic_t n_write_protected_guest_page;
+	DECLARE_HASHTABLE(tracked_guest_page_hash_table, INTEL_GVT_GTT_HASH_BITS);
+	atomic_t n_tracked_guest_page;
 	struct list_head oos_page_list_head;
 	struct list_head post_shadow_list_head;
 	struct intel_vgpu_scratch_pt scratch_pt[GTT_TYPE_MAX];
-
 };
 
 extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu);
@@ -228,12 +226,16 @@ struct intel_vgpu_shadow_page {
 	unsigned long mfn;
 };
 
-struct intel_vgpu_guest_page {
+struct intel_vgpu_page_track {
 	struct hlist_node node;
-	bool writeprotection;
+	bool tracked;
 	unsigned long gfn;
 	int (*handler)(void *, u64, void *, int);
 	void *data;
+};
+
+struct intel_vgpu_guest_page {
+	struct intel_vgpu_page_track track;
 	unsigned long write_cnt;
 	struct intel_vgpu_oos_page *oos_page;
 };
@@ -243,7 +245,7 @@ struct intel_vgpu_oos_page {
 	struct list_head list;
 	struct list_head vm_list;
 	int id;
-	unsigned char mem[GTT_PAGE_SIZE];
+	unsigned char mem[I915_GTT_PAGE_SIZE];
 };
 
 #define GTT_ENTRY_NUM_IN_ONE_PAGE 512
@@ -258,22 +260,16 @@ struct intel_vgpu_ppgtt_spt {
 	struct list_head post_shadow_list;
 };
 
-int intel_vgpu_init_guest_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page,
+int intel_vgpu_init_page_track(struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t,
 		unsigned long gfn,
 		int (*handler)(void *gp, u64, void *, int),
 		void *data);
 
-void intel_vgpu_clean_guest_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page);
-
-int intel_vgpu_set_guest_page_writeprotection(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page);
-
-void intel_vgpu_clear_guest_page_writeprotection(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page);
+void intel_vgpu_clean_page_track(struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t);
 
-struct intel_vgpu_guest_page *intel_vgpu_find_guest_page(
+struct intel_vgpu_page_track *intel_vgpu_find_tracked_page(
 		struct intel_vgpu *vgpu, unsigned long gfn);
 
 int intel_vgpu_sync_oos_pages(struct intel_vgpu *vgpu);

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

@@ -36,6 +36,8 @@
 
 #include "i915_drv.h"
 #include "gvt.h"
+#include <linux/vfio.h>
+#include <linux/mdev.h>
 
 struct intel_gvt_host intel_gvt_host;
 
@@ -44,6 +46,129 @@ static const char * const supported_hypervisors[] = {
 	[INTEL_GVT_HYPERVISOR_KVM] = "KVM",
 };
 
+static struct intel_vgpu_type *intel_gvt_find_vgpu_type(struct intel_gvt *gvt,
+		const char *name)
+{
+	int i;
+	struct intel_vgpu_type *t;
+	const char *driver_name = dev_driver_string(
+			&gvt->dev_priv->drm.pdev->dev);
+
+	for (i = 0; i < gvt->num_types; i++) {
+		t = &gvt->types[i];
+		if (!strncmp(t->name, name + strlen(driver_name) + 1,
+			sizeof(t->name)))
+			return t;
+	}
+
+	return NULL;
+}
+
+static ssize_t available_instances_show(struct kobject *kobj,
+					struct device *dev, char *buf)
+{
+	struct intel_vgpu_type *type;
+	unsigned int num = 0;
+	void *gvt = kdev_to_i915(dev)->gvt;
+
+	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
+	if (!type)
+		num = 0;
+	else
+		num = type->avail_instance;
+
+	return sprintf(buf, "%u\n", num);
+}
+
+static ssize_t device_api_show(struct kobject *kobj, struct device *dev,
+		char *buf)
+{
+	return sprintf(buf, "%s\n", VFIO_DEVICE_API_PCI_STRING);
+}
+
+static ssize_t description_show(struct kobject *kobj, struct device *dev,
+		char *buf)
+{
+	struct intel_vgpu_type *type;
+	void *gvt = kdev_to_i915(dev)->gvt;
+
+	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
+	if (!type)
+		return 0;
+
+	return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n"
+		       "fence: %d\nresolution: %s\n"
+		       "weight: %d\n",
+		       BYTES_TO_MB(type->low_gm_size),
+		       BYTES_TO_MB(type->high_gm_size),
+		       type->fence, vgpu_edid_str(type->resolution),
+		       type->weight);
+}
+
+static MDEV_TYPE_ATTR_RO(available_instances);
+static MDEV_TYPE_ATTR_RO(device_api);
+static MDEV_TYPE_ATTR_RO(description);
+
+static struct attribute *gvt_type_attrs[] = {
+	&mdev_type_attr_available_instances.attr,
+	&mdev_type_attr_device_api.attr,
+	&mdev_type_attr_description.attr,
+	NULL,
+};
+
+static struct attribute_group *gvt_vgpu_type_groups[] = {
+	[0 ... NR_MAX_INTEL_VGPU_TYPES - 1] = NULL,
+};
+
+static bool intel_get_gvt_attrs(struct attribute ***type_attrs,
+		struct attribute_group ***intel_vgpu_type_groups)
+{
+	*type_attrs = gvt_type_attrs;
+	*intel_vgpu_type_groups = gvt_vgpu_type_groups;
+	return true;
+}
+
+static bool intel_gvt_init_vgpu_type_groups(struct intel_gvt *gvt)
+{
+	int i, j;
+	struct intel_vgpu_type *type;
+	struct attribute_group *group;
+
+	for (i = 0; i < gvt->num_types; i++) {
+		type = &gvt->types[i];
+
+		group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL);
+		if (WARN_ON(!group))
+			goto unwind;
+
+		group->name = type->name;
+		group->attrs = gvt_type_attrs;
+		gvt_vgpu_type_groups[i] = group;
+	}
+
+	return true;
+
+unwind:
+	for (j = 0; j < i; j++) {
+		group = gvt_vgpu_type_groups[j];
+		kfree(group);
+	}
+
+	return false;
+}
+
+static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt)
+{
+	int i;
+	struct attribute_group *group;
+
+	for (i = 0; i < gvt->num_types; i++) {
+		group = gvt_vgpu_type_groups[i];
+		gvt_vgpu_type_groups[i] = NULL;
+		kfree(group);
+	}
+}
+
 static const struct intel_gvt_ops intel_gvt_ops = {
 	.emulate_cfg_read = intel_vgpu_emulate_cfg_read,
 	.emulate_cfg_write = intel_vgpu_emulate_cfg_write,
@@ -54,6 +179,8 @@ static const struct intel_gvt_ops intel_gvt_ops = {
 	.vgpu_reset = intel_gvt_reset_vgpu,
 	.vgpu_activate = intel_gvt_activate_vgpu,
 	.vgpu_deactivate = intel_gvt_deactivate_vgpu,
+	.gvt_find_vgpu_type = intel_gvt_find_vgpu_type,
+	.get_gvt_attrs = intel_get_gvt_attrs,
 };
 
 /**
@@ -191,17 +318,18 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv)
 	if (WARN_ON(!gvt))
 		return;
 
+	intel_gvt_debugfs_clean(gvt);
 	clean_service_thread(gvt);
 	intel_gvt_clean_cmd_parser(gvt);
 	intel_gvt_clean_sched_policy(gvt);
 	intel_gvt_clean_workload_scheduler(gvt);
-	intel_gvt_clean_opregion(gvt);
 	intel_gvt_clean_gtt(gvt);
 	intel_gvt_clean_irq(gvt);
 	intel_gvt_clean_mmio_info(gvt);
 	intel_gvt_free_firmware(gvt);
 
 	intel_gvt_hypervisor_host_exit(&dev_priv->drm.pdev->dev, gvt);
+	intel_gvt_cleanup_vgpu_type_groups(gvt);
 	intel_gvt_clean_vgpu_types(gvt);
 
 	idr_destroy(&gvt->vgpu_idr);
@@ -268,13 +396,9 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	if (ret)
 		goto out_clean_irq;
 
-	ret = intel_gvt_init_opregion(gvt);
-	if (ret)
-		goto out_clean_gtt;
-
 	ret = intel_gvt_init_workload_scheduler(gvt);
 	if (ret)
-		goto out_clean_opregion;
+		goto out_clean_gtt;
 
 	ret = intel_gvt_init_sched_policy(gvt);
 	if (ret)
@@ -292,6 +416,12 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	if (ret)
 		goto out_clean_thread;
 
+	ret = intel_gvt_init_vgpu_type_groups(gvt);
+	if (ret == false) {
+		gvt_err("failed to init vgpu type groups: %d\n", ret);
+		goto out_clean_types;
+	}
+
 	ret = intel_gvt_hypervisor_host_init(&dev_priv->drm.pdev->dev, gvt,
 				&intel_gvt_ops);
 	if (ret) {
@@ -307,6 +437,10 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	}
 	gvt->idle_vgpu = vgpu;
 
+	ret = intel_gvt_debugfs_init(gvt);
+	if (ret)
+		gvt_err("debugfs registeration failed, go on.\n");
+
 	gvt_dbg_core("gvt device initialization is done\n");
 	dev_priv->gvt = gvt;
 	return 0;
@@ -321,8 +455,6 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	intel_gvt_clean_sched_policy(gvt);
 out_clean_workload_scheduler:
 	intel_gvt_clean_workload_scheduler(gvt);
-out_clean_opregion:
-	intel_gvt_clean_opregion(gvt);
 out_clean_gtt:
 	intel_gvt_clean_gtt(gvt);
 out_clean_irq:

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

@@ -125,7 +125,6 @@ struct intel_vgpu_irq {
 struct intel_vgpu_opregion {
 	void *va;
 	u32 gfn[INTEL_GVT_OPREGION_PAGES];
-	struct page *pages[INTEL_GVT_OPREGION_PAGES];
 };
 
 #define vgpu_opregion(vgpu) (&(vgpu->opregion))
@@ -142,6 +141,33 @@ struct vgpu_sched_ctl {
 	int weight;
 };
 
+enum {
+	INTEL_VGPU_EXECLIST_SUBMISSION = 1,
+	INTEL_VGPU_GUC_SUBMISSION,
+};
+
+struct intel_vgpu_submission_ops {
+	const char *name;
+	int (*init)(struct intel_vgpu *vgpu);
+	void (*clean)(struct intel_vgpu *vgpu);
+	void (*reset)(struct intel_vgpu *vgpu, unsigned long engine_mask);
+};
+
+struct intel_vgpu_submission {
+	struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
+	struct list_head workload_q_head[I915_NUM_ENGINES];
+	struct kmem_cache *workloads;
+	atomic_t running_workload_num;
+	struct i915_gem_context *shadow_ctx;
+	DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
+	void *ring_scan_buffer[I915_NUM_ENGINES];
+	int ring_scan_buffer_size[I915_NUM_ENGINES];
+	const struct intel_vgpu_submission_ops *ops;
+	int virtual_submission_interface;
+	bool active;
+};
+
 struct intel_vgpu {
 	struct intel_gvt *gvt;
 	int id;
@@ -161,16 +187,10 @@ struct intel_vgpu {
 	struct intel_vgpu_gtt gtt;
 	struct intel_vgpu_opregion opregion;
 	struct intel_vgpu_display display;
-	struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
-	struct list_head workload_q_head[I915_NUM_ENGINES];
-	struct kmem_cache *workloads;
-	atomic_t running_workload_num;
-	/* 1/2K for each reserve ring buffer */
-	void *reserve_ring_buffer_va[I915_NUM_ENGINES];
-	int reserve_ring_buffer_size[I915_NUM_ENGINES];
-	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
-	struct i915_gem_context *shadow_ctx;
-	DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+	struct intel_vgpu_submission submission;
+	u32 hws_pga[I915_NUM_ENGINES];
+
+	struct dentry *debugfs;
 
 #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
 	struct {
@@ -190,6 +210,10 @@ struct intel_vgpu {
 #endif
 };
 
+/* validating GM healthy status*/
+#define vgpu_is_vm_unhealthy(ret_val) \
+	(((ret_val) == -EBADRQC) || ((ret_val) == -EFAULT))
+
 struct intel_gvt_gm {
 	unsigned long vgpu_allocated_low_gm_size;
 	unsigned long vgpu_allocated_high_gm_size;
@@ -231,7 +255,7 @@ struct intel_gvt_mmio {
 	unsigned int num_mmio_block;
 
 	DECLARE_HASHTABLE(mmio_info_table, INTEL_GVT_MMIO_HASH_BITS);
-	unsigned int num_tracked_mmio;
+	unsigned long num_tracked_mmio;
 };
 
 struct intel_gvt_firmware {
@@ -240,11 +264,6 @@ struct intel_gvt_firmware {
 	bool firmware_loaded;
 };
 
-struct intel_gvt_opregion {
-	void *opregion_va;
-	u32 opregion_pa;
-};
-
 #define NR_MAX_INTEL_VGPU_TYPES 20
 struct intel_vgpu_type {
 	char name[16];
@@ -268,7 +287,6 @@ struct intel_gvt {
 	struct intel_gvt_firmware firmware;
 	struct intel_gvt_irq irq;
 	struct intel_gvt_gtt gtt;
-	struct intel_gvt_opregion opregion;
 	struct intel_gvt_workload_scheduler scheduler;
 	struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES];
 	DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
@@ -279,6 +297,8 @@ struct intel_gvt {
 	struct task_struct *service_thread;
 	wait_queue_head_t service_thread_wq;
 	unsigned long service_request;
+
+	struct dentry *debugfs_root;
 };
 
 static inline struct intel_gvt *to_gvt(struct drm_i915_private *i915)
@@ -484,9 +504,6 @@ static inline u64 intel_vgpu_get_bar_gpa(struct intel_vgpu *vgpu, int bar)
 			PCI_BASE_ADDRESS_MEM_MASK;
 }
 
-void intel_gvt_clean_opregion(struct intel_gvt *gvt);
-int intel_gvt_init_opregion(struct intel_gvt *gvt);
-
 void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu);
 int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa);
 
@@ -494,6 +511,7 @@ int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci);
 void populate_pvinfo_page(struct intel_vgpu *vgpu);
 
 int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload);
+void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason);
 
 struct intel_gvt_ops {
 	int (*emulate_cfg_read)(struct intel_vgpu *, unsigned int, void *,
@@ -510,12 +528,17 @@ struct intel_gvt_ops {
 	void (*vgpu_reset)(struct intel_vgpu *);
 	void (*vgpu_activate)(struct intel_vgpu *);
 	void (*vgpu_deactivate)(struct intel_vgpu *);
+	struct intel_vgpu_type *(*gvt_find_vgpu_type)(struct intel_gvt *gvt,
+			const char *name);
+	bool (*get_gvt_attrs)(struct attribute ***type_attrs,
+			struct attribute_group ***intel_vgpu_type_groups);
 };
 
 
 enum {
 	GVT_FAILSAFE_UNSUPPORTED_GUEST,
 	GVT_FAILSAFE_INSUFFICIENT_RESOURCE,
+	GVT_FAILSAFE_GUEST_ERR,
 };
 
 static inline void mmio_hw_access_pre(struct drm_i915_private *dev_priv)
@@ -591,6 +614,12 @@ static inline bool intel_gvt_mmio_has_mode_mask(
 	return gvt->mmio.mmio_attribute[offset >> 2] & F_MODE_MASK;
 }
 
+int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu);
+int intel_gvt_debugfs_init(struct intel_gvt *gvt);
+void intel_gvt_debugfs_clean(struct intel_gvt *gvt);
+
+
 #include "trace.h"
 #include "mpt.h"
 

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

@@ -137,17 +137,26 @@ static int new_mmio_info(struct intel_gvt *gvt,
 	return 0;
 }
 
-static int render_mmio_to_ring_id(struct intel_gvt *gvt, unsigned int reg)
+/**
+ * intel_gvt_render_mmio_to_ring_id - convert a mmio offset into ring id
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * Ring ID on success, negative error code if failed.
+ */
+int intel_gvt_render_mmio_to_ring_id(struct intel_gvt *gvt,
+		unsigned int offset)
 {
 	enum intel_engine_id id;
 	struct intel_engine_cs *engine;
 
-	reg &= ~GENMASK(11, 0);
+	offset &= ~GENMASK(11, 0);
 	for_each_engine(engine, gvt->dev_priv, id) {
-		if (engine->mmio_base == reg)
+		if (engine->mmio_base == offset)
 			return id;
 	}
-	return -1;
+	return -ENODEV;
 }
 
 #define offset_to_fence_num(offset) \
@@ -157,7 +166,7 @@ static int render_mmio_to_ring_id(struct intel_gvt *gvt, unsigned int reg)
 	(num * 8 + i915_mmio_reg_offset(FENCE_REG_GEN6_LO(0)))
 
 
-static void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
+void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
 {
 	switch (reason) {
 	case GVT_FAILSAFE_UNSUPPORTED_GUEST:
@@ -165,6 +174,8 @@ static void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
 		break;
 	case GVT_FAILSAFE_INSUFFICIENT_RESOURCE:
 		pr_err("Graphics resource is not enough for the guest\n");
+	case GVT_FAILSAFE_GUEST_ERR:
+		pr_err("GVT Internal error  for the guest\n");
 	default:
 		break;
 	}
@@ -1369,6 +1380,34 @@ static int mailbox_write(struct intel_vgpu *vgpu, unsigned int offset,
 	return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes);
 }
 
+static int hws_pga_write(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	u32 value = *(u32 *)p_data;
+	int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset);
+
+	if (!intel_gvt_ggtt_validate_range(vgpu, value, I915_GTT_PAGE_SIZE)) {
+		gvt_vgpu_err("VM(%d) write invalid HWSP address, reg:0x%x, value:0x%x\n",
+			      vgpu->id, offset, value);
+		return -EINVAL;
+	}
+	/*
+	 * Need to emulate all the HWSP register write to ensure host can
+	 * update the VM CSB status correctly. Here listed registers can
+	 * support BDW, SKL or other platforms with same HWSP registers.
+	 */
+	if (unlikely(ring_id < 0 || ring_id > I915_NUM_ENGINES)) {
+		gvt_vgpu_err("VM(%d) access unknown hardware status page register:0x%x\n",
+			     vgpu->id, offset);
+		return -EINVAL;
+	}
+	vgpu->hws_pga[ring_id] = value;
+	gvt_dbg_mmio("VM(%d) write: 0x%x to HWSP: 0x%x\n",
+		     vgpu->id, value, offset);
+
+	return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes);
+}
+
 static int skl_power_well_ctl_write(struct intel_vgpu *vgpu,
 		unsigned int offset, void *p_data, unsigned int bytes)
 {
@@ -1432,18 +1471,36 @@ static int skl_lcpll_write(struct intel_vgpu *vgpu, unsigned int offset,
 static int mmio_read_from_hw(struct intel_vgpu *vgpu,
 		unsigned int offset, void *p_data, unsigned int bytes)
 {
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	int ring_id;
+	u32 ring_base;
 
+	ring_id = intel_gvt_render_mmio_to_ring_id(gvt, offset);
+	/**
+	 * Read HW reg in following case
+	 * a. the offset isn't a ring mmio
+	 * b. the offset's ring is running on hw.
+	 * c. the offset is ring time stamp mmio
+	 */
+	if (ring_id >= 0)
+		ring_base = dev_priv->engine[ring_id]->mmio_base;
+
+	if (ring_id < 0 || vgpu  == gvt->scheduler.engine_owner[ring_id] ||
+	    offset == i915_mmio_reg_offset(RING_TIMESTAMP(ring_base)) ||
+	    offset == i915_mmio_reg_offset(RING_TIMESTAMP_UDW(ring_base))) {
 		mmio_hw_access_pre(dev_priv);
 		vgpu_vreg(vgpu, offset) = I915_READ(_MMIO(offset));
 		mmio_hw_access_post(dev_priv);
+	}
+
 	return intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes);
 }
 
 static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
-	int ring_id = render_mmio_to_ring_id(vgpu->gvt, offset);
+	int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset);
 	struct intel_vgpu_execlist *execlist;
 	u32 data = *(u32 *)p_data;
 	int ret = 0;
@@ -1451,9 +1508,9 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 	if (WARN_ON(ring_id < 0 || ring_id > I915_NUM_ENGINES - 1))
 		return -EINVAL;
 
-	execlist = &vgpu->execlist[ring_id];
+	execlist = &vgpu->submission.execlist[ring_id];
 
-	execlist->elsp_dwords.data[execlist->elsp_dwords.index] = data;
+	execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data;
 	if (execlist->elsp_dwords.index == 3) {
 		ret = intel_vgpu_submit_execlist(vgpu, ring_id);
 		if(ret)
@@ -1469,9 +1526,11 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	u32 data = *(u32 *)p_data;
-	int ring_id = render_mmio_to_ring_id(vgpu->gvt, offset);
+	int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset);
 	bool enable_execlist;
+	int ret;
 
 	write_vreg(vgpu, offset, p_data, bytes);
 
@@ -1493,7 +1552,17 @@ static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 				(enable_execlist ? "enabling" : "disabling"),
 				ring_id);
 
-		if (enable_execlist)
+		if (!enable_execlist)
+			return 0;
+
+		if (s->active)
+			return 0;
+
+		ret = intel_vgpu_select_submission_ops(vgpu,
+				INTEL_VGPU_EXECLIST_SUBMISSION);
+		if (ret)
+			return ret;
+
 		intel_vgpu_start_schedule(vgpu);
 	}
 	return 0;
@@ -1526,7 +1595,7 @@ static int gvt_reg_tlb_control_handler(struct intel_vgpu *vgpu,
 	default:
 		return -EINVAL;
 	}
-	set_bit(id, (void *)vgpu->tlb_handle_pending);
+	set_bit(id, (void *)vgpu->submission.tlb_handle_pending);
 
 	return 0;
 }
@@ -2512,7 +2581,7 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt)
 	MMIO_RING_F(RING_REG, 32, 0, 0, 0, D_BDW_PLUS, NULL, NULL);
 #undef RING_REG
 
-	MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, NULL);
+	MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, hws_pga_write);
 
 	MMIO_DFH(HDC_CHICKEN0, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 
@@ -2914,14 +2983,46 @@ int intel_gvt_setup_mmio_info(struct intel_gvt *gvt)
 	gvt->mmio.mmio_block = mmio_blocks;
 	gvt->mmio.num_mmio_block = ARRAY_SIZE(mmio_blocks);
 
-	gvt_dbg_mmio("traced %u virtual mmio registers\n",
-		     gvt->mmio.num_tracked_mmio);
 	return 0;
 err:
 	intel_gvt_clean_mmio_info(gvt);
 	return ret;
 }
 
+/**
+ * intel_gvt_for_each_tracked_mmio - iterate each tracked mmio
+ * @gvt: a GVT device
+ * @handler: the handler
+ * @data: private data given to handler
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
+	int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
+	void *data)
+{
+	struct gvt_mmio_block *block = gvt->mmio.mmio_block;
+	struct intel_gvt_mmio_info *e;
+	int i, j, ret;
+
+	hash_for_each(gvt->mmio.mmio_info_table, i, e, node) {
+		ret = handler(gvt, e->offset, data);
+		if (ret)
+			return ret;
+	}
+
+	for (i = 0; i < gvt->mmio.num_mmio_block; i++, block++) {
+		for (j = 0; j < block->size; j += 4) {
+			ret = handler(gvt,
+				INTEL_GVT_MMIO_OFFSET(block->offset) + j,
+				data);
+			if (ret)
+				return ret;
+		}
+	}
+	return 0;
+}
 
 /**
  * intel_vgpu_default_mmio_read - default MMIO read handler

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

@@ -248,120 +248,6 @@ static void gvt_cache_destroy(struct intel_vgpu *vgpu)
 	}
 }
 
-static struct intel_vgpu_type *intel_gvt_find_vgpu_type(struct intel_gvt *gvt,
-		const char *name)
-{
-	int i;
-	struct intel_vgpu_type *t;
-	const char *driver_name = dev_driver_string(
-			&gvt->dev_priv->drm.pdev->dev);
-
-	for (i = 0; i < gvt->num_types; i++) {
-		t = &gvt->types[i];
-		if (!strncmp(t->name, name + strlen(driver_name) + 1,
-			sizeof(t->name)))
-			return t;
-	}
-
-	return NULL;
-}
-
-static ssize_t available_instances_show(struct kobject *kobj,
-					struct device *dev, char *buf)
-{
-	struct intel_vgpu_type *type;
-	unsigned int num = 0;
-	void *gvt = kdev_to_i915(dev)->gvt;
-
-	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
-	if (!type)
-		num = 0;
-	else
-		num = type->avail_instance;
-
-	return sprintf(buf, "%u\n", num);
-}
-
-static ssize_t device_api_show(struct kobject *kobj, struct device *dev,
-		char *buf)
-{
-	return sprintf(buf, "%s\n", VFIO_DEVICE_API_PCI_STRING);
-}
-
-static ssize_t description_show(struct kobject *kobj, struct device *dev,
-		char *buf)
-{
-	struct intel_vgpu_type *type;
-	void *gvt = kdev_to_i915(dev)->gvt;
-
-	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
-	if (!type)
-		return 0;
-
-	return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n"
-		       "fence: %d\nresolution: %s\n"
-		       "weight: %d\n",
-		       BYTES_TO_MB(type->low_gm_size),
-		       BYTES_TO_MB(type->high_gm_size),
-		       type->fence, vgpu_edid_str(type->resolution),
-		       type->weight);
-}
-
-static MDEV_TYPE_ATTR_RO(available_instances);
-static MDEV_TYPE_ATTR_RO(device_api);
-static MDEV_TYPE_ATTR_RO(description);
-
-static struct attribute *type_attrs[] = {
-	&mdev_type_attr_available_instances.attr,
-	&mdev_type_attr_device_api.attr,
-	&mdev_type_attr_description.attr,
-	NULL,
-};
-
-static struct attribute_group *intel_vgpu_type_groups[] = {
-	[0 ... NR_MAX_INTEL_VGPU_TYPES - 1] = NULL,
-};
-
-static bool intel_gvt_init_vgpu_type_groups(struct intel_gvt *gvt)
-{
-	int i, j;
-	struct intel_vgpu_type *type;
-	struct attribute_group *group;
-
-	for (i = 0; i < gvt->num_types; i++) {
-		type = &gvt->types[i];
-
-		group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL);
-		if (WARN_ON(!group))
-			goto unwind;
-
-		group->name = type->name;
-		group->attrs = type_attrs;
-		intel_vgpu_type_groups[i] = group;
-	}
-
-	return true;
-
-unwind:
-	for (j = 0; j < i; j++) {
-		group = intel_vgpu_type_groups[j];
-		kfree(group);
-	}
-
-	return false;
-}
-
-static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt)
-{
-	int i;
-	struct attribute_group *group;
-
-	for (i = 0; i < gvt->num_types; i++) {
-		group = intel_vgpu_type_groups[i];
-		kfree(group);
-	}
-}
-
 static void kvmgt_protect_table_init(struct kvmgt_guest_info *info)
 {
 	hash_init(info->ptable);
@@ -441,7 +327,7 @@ static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev)
 	pdev = mdev_parent_dev(mdev);
 	gvt = kdev_to_i915(pdev)->gvt;
 
-	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
+	type = intel_gvt_ops->gvt_find_vgpu_type(gvt, kobject_name(kobj));
 	if (!type) {
 		gvt_vgpu_err("failed to find type %s to create\n",
 						kobject_name(kobj));
@@ -1188,7 +1074,7 @@ hw_id_show(struct device *dev, struct device_attribute *attr,
 		struct intel_vgpu *vgpu = (struct intel_vgpu *)
 			mdev_get_drvdata(mdev);
 		return sprintf(buf, "%u\n",
-			       vgpu->shadow_ctx->hw_id);
+			       vgpu->submission.shadow_ctx->hw_id);
 	}
 	return sprintf(buf, "\n");
 }
@@ -1212,8 +1098,7 @@ static const struct attribute_group *intel_vgpu_groups[] = {
 	NULL,
 };
 
-static const struct mdev_parent_ops intel_vgpu_ops = {
-	.supported_type_groups	= intel_vgpu_type_groups,
+static struct mdev_parent_ops intel_vgpu_ops = {
 	.mdev_attr_groups       = intel_vgpu_groups,
 	.create			= intel_vgpu_create,
 	.remove			= intel_vgpu_remove,
@@ -1229,17 +1114,20 @@ static const struct mdev_parent_ops intel_vgpu_ops = {
 
 static int kvmgt_host_init(struct device *dev, void *gvt, const void *ops)
 {
-	if (!intel_gvt_init_vgpu_type_groups(gvt))
-		return -EFAULT;
+	struct attribute **kvm_type_attrs;
+	struct attribute_group **kvm_vgpu_type_groups;
 
 	intel_gvt_ops = ops;
+	if (!intel_gvt_ops->get_gvt_attrs(&kvm_type_attrs,
+			&kvm_vgpu_type_groups))
+		return -EFAULT;
+	intel_vgpu_ops.supported_type_groups = kvm_vgpu_type_groups;
 
 	return mdev_register_device(dev, &intel_vgpu_ops);
 }
 
 static void kvmgt_host_exit(struct device *dev, void *gvt)
 {
-	intel_gvt_cleanup_vgpu_type_groups(gvt);
 	mdev_unregister_device(dev);
 }
 

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

@@ -117,18 +117,18 @@ static void failsafe_emulate_mmio_rw(struct intel_vgpu *vgpu, uint64_t pa,
 		else
 			memcpy(pt, p_data, bytes);
 
-	} else if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
-		struct intel_vgpu_guest_page *gp;
+	} else if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) {
+		struct intel_vgpu_page_track *t;
 
 		/* Since we enter the failsafe mode early during guest boot,
 		 * guest may not have chance to set up its ppgtt table, so
 		 * there should not be any wp pages for guest. Keep the wp
 		 * related code here in case we need to handle it in furture.
 		 */
-		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
-		if (gp) {
+		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
+		if (t) {
 			/* remove write protection to prevent furture traps */
-			intel_vgpu_clean_guest_page(vgpu, gp);
+			intel_vgpu_clean_page_track(vgpu, t);
 			if (read)
 				intel_gvt_hypervisor_read_gpa(vgpu, pa,
 						p_data, bytes);
@@ -170,17 +170,17 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 		return ret;
 	}
 
-	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
-		struct intel_vgpu_guest_page *gp;
+	if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) {
+		struct intel_vgpu_page_track *t;
 
-		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
-		if (gp) {
+		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
+		if (t) {
 			ret = intel_gvt_hypervisor_read_gpa(vgpu, pa,
 					p_data, bytes);
 			if (ret) {
 				gvt_vgpu_err("guest page read error %d, "
 					"gfn 0x%lx, pa 0x%llx, var 0x%x, len %d\n",
-					ret, gp->gfn, pa, *(u32 *)p_data,
+					ret, t->gfn, pa, *(u32 *)p_data,
 					bytes);
 			}
 			mutex_unlock(&gvt->lock);
@@ -267,17 +267,17 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 		return ret;
 	}
 
-	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
-		struct intel_vgpu_guest_page *gp;
+	if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) {
+		struct intel_vgpu_page_track *t;
 
-		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
-		if (gp) {
-			ret = gp->handler(gp, pa, p_data, bytes);
+		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
+		if (t) {
+			ret = t->handler(t, pa, p_data, bytes);
 			if (ret) {
 				gvt_err("guest page write error %d, "
 					"gfn 0x%lx, pa 0x%llx, "
 					"var 0x%x, len %d\n",
-					ret, gp->gfn, pa,
+					ret, t->gfn, pa,
 					*(u32 *)p_data, bytes);
 			}
 			mutex_unlock(&gvt->lock);

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

@@ -65,11 +65,17 @@ struct intel_gvt_mmio_info {
 	struct hlist_node node;
 };
 
+int intel_gvt_render_mmio_to_ring_id(struct intel_gvt *gvt,
+		unsigned int reg);
 unsigned long intel_gvt_get_device_type(struct intel_gvt *gvt);
 bool intel_gvt_match_device(struct intel_gvt *gvt, unsigned long device);
 
 int intel_gvt_setup_mmio_info(struct intel_gvt *gvt);
 void intel_gvt_clean_mmio_info(struct intel_gvt *gvt);
+int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
+	int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
+	void *data);
+
 
 #define INTEL_GVT_MMIO_OFFSET(reg) ({ \
 	typeof(reg) __reg = reg; \

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

@@ -154,51 +154,53 @@ static inline unsigned long intel_gvt_hypervisor_virt_to_mfn(void *p)
 }
 
 /**
- * intel_gvt_hypervisor_set_wp_page - set a guest page to write-protected
+ * intel_gvt_hypervisor_enable - set a guest page to write-protected
  * @vgpu: a vGPU
- * @p: intel_vgpu_guest_page
+ * @t: page track data structure
  *
  * Returns:
  * Zero on success, negative error code if failed.
  */
-static inline int intel_gvt_hypervisor_set_wp_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *p)
+static inline int intel_gvt_hypervisor_enable_page_track(
+		struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t)
 {
 	int ret;
 
-	if (p->writeprotection)
+	if (t->tracked)
 		return 0;
 
-	ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, p->gfn);
+	ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, t->gfn);
 	if (ret)
 		return ret;
-	p->writeprotection = true;
-	atomic_inc(&vgpu->gtt.n_write_protected_guest_page);
+	t->tracked = true;
+	atomic_inc(&vgpu->gtt.n_tracked_guest_page);
 	return 0;
 }
 
 /**
- * intel_gvt_hypervisor_unset_wp_page - remove the write-protection of a
+ * intel_gvt_hypervisor_disable_page_track - remove the write-protection of a
  * guest page
  * @vgpu: a vGPU
- * @p: intel_vgpu_guest_page
+ * @t: page track data structure
  *
  * Returns:
  * Zero on success, negative error code if failed.
  */
-static inline int intel_gvt_hypervisor_unset_wp_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *p)
+static inline int intel_gvt_hypervisor_disable_page_track(
+		struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t)
 {
 	int ret;
 
-	if (!p->writeprotection)
+	if (!t->tracked)
 		return 0;
 
-	ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, p->gfn);
+	ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, t->gfn);
 	if (ret)
 		return ret;
-	p->writeprotection = false;
-	atomic_dec(&vgpu->gtt.n_write_protected_guest_page);
+	t->tracked = false;
+	atomic_dec(&vgpu->gtt.n_tracked_guest_page);
 	return 0;
 }
 

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

@@ -25,36 +25,247 @@
 #include "i915_drv.h"
 #include "gvt.h"
 
-static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa)
+/*
+ * Note: Only for GVT-g virtual VBT generation, other usage must
+ * not do like this.
+ */
+#define _INTEL_BIOS_PRIVATE
+#include "intel_vbt_defs.h"
+
+#define OPREGION_SIGNATURE "IntelGraphicsMem"
+#define MBOX_VBT      (1<<3)
+
+/* device handle */
+#define DEVICE_TYPE_CRT    0x01
+#define DEVICE_TYPE_EFP1   0x04
+#define DEVICE_TYPE_EFP2   0x40
+#define DEVICE_TYPE_EFP3   0x20
+#define DEVICE_TYPE_EFP4   0x10
+
+#define DEV_SIZE	38
+
+struct opregion_header {
+	u8 signature[16];
+	u32 size;
+	u32 opregion_ver;
+	u8 bios_ver[32];
+	u8 vbios_ver[16];
+	u8 driver_ver[16];
+	u32 mboxes;
+	u32 driver_model;
+	u32 pcon;
+	u8 dver[32];
+	u8 rsvd[124];
+} __packed;
+
+struct bdb_data_header {
+	u8 id;
+	u16 size; /* data size */
+} __packed;
+
+struct efp_child_device_config {
+	u16 handle;
+	u16 device_type;
+	u16 device_class;
+	u8 i2c_speed;
+	u8 dp_onboard_redriver; /* 158 */
+	u8 dp_ondock_redriver; /* 158 */
+	u8 hdmi_level_shifter_value:4; /* 169 */
+	u8 hdmi_max_data_rate:4; /* 204 */
+	u16 dtd_buf_ptr; /* 161 */
+	u8 edidless_efp:1; /* 161 */
+	u8 compression_enable:1; /* 198 */
+	u8 compression_method:1; /* 198 */
+	u8 ganged_edp:1; /* 202 */
+	u8 skip0:4;
+	u8 compression_structure_index:4; /* 198 */
+	u8 skip1:4;
+	u8 slave_port; /*  202 */
+	u8 skip2;
+	u8 dvo_port;
+	u8 i2c_pin; /* for add-in card */
+	u8 slave_addr; /* for add-in card */
+	u8 ddc_pin;
+	u16 edid_ptr;
+	u8 dvo_config;
+	u8 efp_docked_port:1; /* 158 */
+	u8 lane_reversal:1; /* 184 */
+	u8 onboard_lspcon:1; /* 192 */
+	u8 iboost_enable:1; /* 196 */
+	u8 hpd_invert:1; /* BXT 196 */
+	u8 slip3:3;
+	u8 hdmi_compat:1;
+	u8 dp_compat:1;
+	u8 tmds_compat:1;
+	u8 skip4:5;
+	u8 aux_channel;
+	u8 dongle_detect;
+	u8 pipe_cap:2;
+	u8 sdvo_stall:1; /* 158 */
+	u8 hpd_status:2;
+	u8 integrated_encoder:1;
+	u8 skip5:2;
+	u8 dvo_wiring;
+	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 {
+	/* header->bdb_offset point to bdb_header offset */
+	struct vbt_header header;
+	struct bdb_header bdb_header;
+
+	struct bdb_data_header general_features_header;
+	struct bdb_general_features general_features;
+
+	struct bdb_data_header general_definitions_header;
+	struct bdb_general_definitions general_definitions;
+
+	struct efp_child_device_config child0;
+	struct efp_child_device_config child1;
+	struct efp_child_device_config child2;
+	struct efp_child_device_config child3;
+
+	struct bdb_data_header driver_features_header;
+	struct bdb_driver_features driver_features;
+};
+
+static void virt_vbt_generation(struct vbt *v)
 {
-	u8 *buf;
-	int i;
+	int num_child;
+
+	memset(v, 0, sizeof(struct vbt));
+
+	v->header.signature[0] = '$';
+	v->header.signature[1] = 'V';
+	v->header.signature[2] = 'B';
+	v->header.signature[3] = 'T';
+
+	/* there's features depending on version! */
+	v->header.version = 155;
+	v->header.header_size = sizeof(v->header);
+	v->header.vbt_size = sizeof(struct vbt) - sizeof(v->header);
+	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.header_size = sizeof(v->bdb_header);
+
+	v->bdb_header.bdb_size = sizeof(struct vbt) - sizeof(struct vbt_header)
+		- sizeof(struct bdb_header);
+
+	/* general features */
+	v->general_features_header.id = BDB_GENERAL_FEATURES;
+	v->general_features_header.size = sizeof(struct bdb_general_features);
+	v->general_features.int_crt_support = 0;
+	v->general_features.int_tv_support = 0;
+
+	/* child device */
+	num_child = 4; /* each port has one child */
+	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;
+
+	/* portA */
+	v->child0.handle = DEVICE_TYPE_EFP1;
+	v->child0.device_type = DEVICE_TYPE_DP;
+	v->child0.dvo_port = DVO_PORT_DPA;
+	v->child0.aux_channel = DP_AUX_A;
+	v->child0.dp_compat = true;
+	v->child0.integrated_encoder = true;
+
+	/* portB */
+	v->child1.handle = DEVICE_TYPE_EFP2;
+	v->child1.device_type = DEVICE_TYPE_DP;
+	v->child1.dvo_port = DVO_PORT_DPB;
+	v->child1.aux_channel = DP_AUX_B;
+	v->child1.dp_compat = true;
+	v->child1.integrated_encoder = true;
+
+	/* portC */
+	v->child2.handle = DEVICE_TYPE_EFP3;
+	v->child2.device_type = DEVICE_TYPE_DP;
+	v->child2.dvo_port = DVO_PORT_DPC;
+	v->child2.aux_channel = DP_AUX_C;
+	v->child2.dp_compat = true;
+	v->child2.integrated_encoder = true;
+
+	/* portD */
+	v->child3.handle = DEVICE_TYPE_EFP4;
+	v->child3.device_type = DEVICE_TYPE_DP;
+	v->child3.dvo_port = DVO_PORT_DPD;
+	v->child3.aux_channel = DP_AUX_D;
+	v->child3.dp_compat = true;
+	v->child3.integrated_encoder = true;
+
+	/* driver features */
+	v->driver_features_header.id = BDB_DRIVER_FEATURES;
+	v->driver_features_header.size = sizeof(struct bdb_driver_features);
+	v->driver_features.lvds_config = BDB_DRIVER_FEATURE_NO_LVDS;
+}
 
-	if (WARN((vgpu_opregion(vgpu)->va),
-			"vgpu%d: opregion has been initialized already.\n",
-			vgpu->id))
-		return -EINVAL;
+static int alloc_and_init_virt_opregion(struct intel_vgpu *vgpu)
+{
+	u8 *buf;
+	struct opregion_header *header;
+	struct vbt v;
 
+	gvt_dbg_core("init vgpu%d opregion\n", vgpu->id);
 	vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_KERNEL |
 			__GFP_ZERO,
 			get_order(INTEL_GVT_OPREGION_SIZE));
-
-	if (!vgpu_opregion(vgpu)->va)
+	if (!vgpu_opregion(vgpu)->va) {
+		gvt_err("fail to get memory for vgpu virt opregion\n");
 		return -ENOMEM;
+	}
 
-	memcpy(vgpu_opregion(vgpu)->va, vgpu->gvt->opregion.opregion_va,
-	       INTEL_GVT_OPREGION_SIZE);
-
-	for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
-		vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
+	/* emulated opregion with VBT mailbox only */
+	buf = (u8 *)vgpu_opregion(vgpu)->va;
+	header = (struct opregion_header *)buf;
+	memcpy(header->signature, OPREGION_SIGNATURE,
+			sizeof(OPREGION_SIGNATURE));
+	header->size = 0x8;
+	header->opregion_ver = 0x02000000;
+	header->mboxes = MBOX_VBT;
 
 	/* for unknown reason, the value in LID field is incorrect
 	 * which block the windows guest, so workaround it by force
 	 * setting it to "OPEN"
 	 */
-	buf = (u8 *)vgpu_opregion(vgpu)->va;
 	buf[INTEL_GVT_OPREGION_CLID] = 0x3;
 
+	/* emulated vbt from virt vbt generation */
+	virt_vbt_generation(&v);
+	memcpy(buf + INTEL_GVT_OPREGION_VBT_OFFSET, &v, sizeof(struct vbt));
+
+	return 0;
+}
+
+static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa)
+{
+	int i, ret;
+
+	if (WARN((vgpu_opregion(vgpu)->va),
+			"vgpu%d: opregion has been initialized already.\n",
+			vgpu->id))
+		return -EINVAL;
+
+	ret = alloc_and_init_virt_opregion(vgpu);
+	if (ret < 0)
+		return ret;
+
+	for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
+		vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
+
 	return 0;
 }
 
@@ -132,40 +343,6 @@ int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa)
 	return 0;
 }
 
-/**
- * intel_gvt_clean_opregion - clean host opergion related stuffs
- * @gvt: a GVT device
- *
- */
-void intel_gvt_clean_opregion(struct intel_gvt *gvt)
-{
-	memunmap(gvt->opregion.opregion_va);
-	gvt->opregion.opregion_va = NULL;
-}
-
-/**
- * intel_gvt_init_opregion - initialize host opergion related stuffs
- * @gvt: a GVT device
- *
- * Returns:
- * Zero on success, negative error code if failed.
- */
-int intel_gvt_init_opregion(struct intel_gvt *gvt)
-{
-	gvt_dbg_core("init host opregion\n");
-
-	pci_read_config_dword(gvt->dev_priv->drm.pdev, INTEL_GVT_PCI_OPREGION,
-			&gvt->opregion.opregion_pa);
-
-	gvt->opregion.opregion_va = memremap(gvt->opregion.opregion_pa,
-					     INTEL_GVT_OPREGION_SIZE, MEMREMAP_WB);
-	if (!gvt->opregion.opregion_va) {
-		gvt_err("fail to map host opregion\n");
-		return -EFAULT;
-	}
-	return 0;
-}
-
 #define GVT_OPREGION_FUNC(scic)					\
 	({							\
 	 u32 __ret;						\

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

@@ -51,6 +51,9 @@
 
 #define INTEL_GVT_OPREGION_PAGES	2
 #define INTEL_GVT_OPREGION_SIZE		(INTEL_GVT_OPREGION_PAGES * PAGE_SIZE)
+#define INTEL_GVT_OPREGION_VBT_OFFSET	0x400
+#define INTEL_GVT_OPREGION_VBT_SIZE	\
+		(INTEL_GVT_OPREGION_SIZE - INTEL_GVT_OPREGION_VBT_OFFSET)
 
 #define VGT_SPRSTRIDE(pipe)	_PIPE(pipe, _SPRA_STRIDE, _PLANE_STRIDE_2_B)
 
@@ -71,6 +74,7 @@
 #define RB_HEAD_OFF_MASK	((1U << 21) - (1U << 2))
 #define RB_TAIL_OFF_MASK	((1U << 21) - (1U << 3))
 #define RB_TAIL_SIZE_MASK	((1U << 21) - (1U << 12))
-#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + GTT_PAGE_SIZE)
+#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + \
+		I915_GTT_PAGE_SIZE)
 
 #endif

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

@@ -147,6 +147,7 @@ static u32 gen9_render_mocs_L3[32];
 static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	enum forcewake_domains fw;
 	i915_reg_t reg;
 	u32 regs[] = {
@@ -160,7 +161,7 @@ static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
 	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
 		return;
 
-	if (!test_and_clear_bit(ring_id, (void *)vgpu->tlb_handle_pending))
+	if (!test_and_clear_bit(ring_id, (void *)s->tlb_handle_pending))
 		return;
 
 	reg = _MMIO(regs[ring_id]);
@@ -208,7 +209,7 @@ static void load_mocs(struct intel_vgpu *vgpu, int ring_id)
 	offset.reg = regs[ring_id];
 	for (i = 0; i < 64; i++) {
 		gen9_render_mocs[ring_id][i] = I915_READ_FW(offset);
-		I915_WRITE(offset, vgpu_vreg(vgpu, offset));
+		I915_WRITE_FW(offset, vgpu_vreg(vgpu, offset));
 		offset.reg += 4;
 	}
 
@@ -261,14 +262,15 @@ static void restore_mocs(struct intel_vgpu *vgpu, int ring_id)
 static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	struct render_mmio *mmio;
-	u32 v;
-	int i, array_size;
-	u32 *reg_state = vgpu->shadow_ctx->engine[ring_id].lrc_reg_state;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	u32 *reg_state = s->shadow_ctx->engine[ring_id].lrc_reg_state;
 	u32 ctx_ctrl = reg_state[CTX_CONTEXT_CONTROL_VAL];
 	u32 inhibit_mask =
 		_MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
 	i915_reg_t last_reg = _MMIO(0);
+	struct render_mmio *mmio;
+	u32 v;
+	int i, array_size;
 
 	if (IS_SKYLAKE(vgpu->gvt->dev_priv)
 		|| IS_KABYLAKE(vgpu->gvt->dev_priv)) {

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

@@ -112,17 +112,18 @@ struct intel_vgpu_workload {
 	struct intel_shadow_wa_ctx wa_ctx;
 };
 
-/* Intel shadow batch buffer is a i915 gem object */
-struct intel_shadow_bb_entry {
+struct intel_vgpu_shadow_bb {
 	struct list_head list;
 	struct drm_i915_gem_object *obj;
+	struct i915_vma *vma;
 	void *va;
-	unsigned long len;
 	u32 *bb_start_cmd_va;
+	unsigned int clflush;
+	bool accessing;
 };
 
 #define workload_q_head(vgpu, ring_id) \
-	(&(vgpu->workload_q_head[ring_id]))
+	(&(vgpu->submission.workload_q_head[ring_id]))
 
 #define queue_workload(workload) do { \
 	list_add_tail(&workload->list, \
@@ -137,9 +138,23 @@ void intel_gvt_clean_workload_scheduler(struct intel_gvt *gvt);
 
 void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu);
 
-int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu);
+int intel_vgpu_setup_submission(struct intel_vgpu *vgpu);
 
-void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu);
+void intel_vgpu_reset_submission(struct intel_vgpu *vgpu,
+				 unsigned long engine_mask);
+
+void intel_vgpu_clean_submission(struct intel_vgpu *vgpu);
+
+int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+				     unsigned int interface);
+
+extern const struct intel_vgpu_submission_ops
+intel_vgpu_execlist_submission_ops;
+
+struct intel_vgpu_workload *
+intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id,
+			   struct execlist_ctx_descriptor_format *desc);
+
+void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload);
 
-void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx);
 #endif

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

@@ -43,7 +43,10 @@ void populate_pvinfo_page(struct intel_vgpu *vgpu)
 	vgpu_vreg(vgpu, vgtif_reg(version_minor)) = 0;
 	vgpu_vreg(vgpu, vgtif_reg(display_ready)) = 0;
 	vgpu_vreg(vgpu, vgtif_reg(vgt_id)) = vgpu->id;
+
 	vgpu_vreg(vgpu, vgtif_reg(vgt_caps)) = VGT_CAPS_FULL_48BIT_PPGTT;
+	vgpu_vreg(vgpu, vgtif_reg(vgt_caps)) |= VGT_CAPS_HWSP_EMULATION;
+
 	vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.base)) =
 		vgpu_aperture_gmadr_base(vgpu);
 	vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) =
@@ -226,7 +229,7 @@ void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu)
 
 	vgpu->active = false;
 
-	if (atomic_read(&vgpu->running_workload_num)) {
+	if (atomic_read(&vgpu->submission.running_workload_num)) {
 		mutex_unlock(&gvt->lock);
 		intel_gvt_wait_vgpu_idle(vgpu);
 		mutex_lock(&gvt->lock);
@@ -252,10 +255,10 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 
 	WARN(vgpu->active, "vGPU is still active!\n");
 
+	intel_gvt_debugfs_remove_vgpu(vgpu);
 	idr_remove(&gvt->vgpu_idr, vgpu->id);
 	intel_vgpu_clean_sched_policy(vgpu);
-	intel_vgpu_clean_gvt_context(vgpu);
-	intel_vgpu_clean_execlist(vgpu);
+	intel_vgpu_clean_submission(vgpu);
 	intel_vgpu_clean_display(vgpu);
 	intel_vgpu_clean_opregion(vgpu);
 	intel_vgpu_clean_gtt(vgpu);
@@ -293,7 +296,7 @@ struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt)
 	vgpu->gvt = gvt;
 
 	for (i = 0; i < I915_NUM_ENGINES; i++)
-		INIT_LIST_HEAD(&vgpu->workload_q_head[i]);
+		INIT_LIST_HEAD(&vgpu->submission.workload_q_head[i]);
 
 	ret = intel_vgpu_init_sched_policy(vgpu);
 	if (ret)
@@ -346,7 +349,6 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	vgpu->handle = param->handle;
 	vgpu->gvt = gvt;
 	vgpu->sched_ctl.weight = param->weight;
-	bitmap_zero(vgpu->tlb_handle_pending, I915_NUM_ENGINES);
 
 	intel_vgpu_init_cfg_space(vgpu, param->primary);
 
@@ -372,26 +374,26 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	if (ret)
 		goto out_clean_gtt;
 
-	ret = intel_vgpu_init_execlist(vgpu);
+	ret = intel_vgpu_setup_submission(vgpu);
 	if (ret)
 		goto out_clean_display;
 
-	ret = intel_vgpu_init_gvt_context(vgpu);
+	ret = intel_vgpu_init_sched_policy(vgpu);
 	if (ret)
-		goto out_clean_execlist;
+		goto out_clean_submission;
 
-	ret = intel_vgpu_init_sched_policy(vgpu);
+	ret = intel_gvt_debugfs_add_vgpu(vgpu);
 	if (ret)
-		goto out_clean_shadow_ctx;
+		goto out_clean_sched_policy;
 
 	mutex_unlock(&gvt->lock);
 
 	return vgpu;
 
-out_clean_shadow_ctx:
-	intel_vgpu_clean_gvt_context(vgpu);
-out_clean_execlist:
-	intel_vgpu_clean_execlist(vgpu);
+out_clean_sched_policy:
+	intel_vgpu_clean_sched_policy(vgpu);
+out_clean_submission:
+	intel_vgpu_clean_submission(vgpu);
 out_clean_display:
 	intel_vgpu_clean_display(vgpu);
 out_clean_gtt:
@@ -500,10 +502,10 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
 		mutex_lock(&gvt->lock);
 	}
 
-	intel_vgpu_reset_execlist(vgpu, resetting_eng);
-
+	intel_vgpu_reset_submission(vgpu, resetting_eng);
 	/* full GPU reset or device model level reset */
 	if (engine_mask == ALL_ENGINES || dmlr) {
+		intel_vgpu_select_submission_ops(vgpu, 0);
 
 		/*fence will not be reset during virtual reset */
 		if (dmlr) {