Merge tag 'gvt-next-2016-10-14' of https://github.com/01org/gvt-linux into drm-intel-next-queued

Zhenyu Wang writes:

This is first pull request to merge GVT-g device model in i915
which contains core GVT-g device model work to virtualize GPU
resources. This tries to add feature of Intel GVT-g technology
for full GPU virtualization. This version will support KVM based
virtualization solution named as KVMGT.

More background is on official project home: https://01.org/igvt-g

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drivers/gpu/drm/i915/gvt/Makefile

@@ -1,5 +1,7 @@
 GVT_DIR := gvt
-GVT_SOURCE := gvt.o
+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
 
 ccflags-y                      += -I$(src) -I$(src)/$(GVT_DIR) -Wall
 i915-y			       += $(addprefix $(GVT_DIR)/, $(GVT_SOURCE))

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

@@ -0,0 +1,341 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Dexuan Cui
+ *
+ * Contributors:
+ *    Pei Zhang <pei.zhang@intel.com>
+ *    Min He <min.he@intel.com>
+ *    Niu Bing <bing.niu@intel.com>
+ *    Yulei Zhang <yulei.zhang@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+#define MB_TO_BYTES(mb) ((mb) << 20ULL)
+#define BYTES_TO_MB(b) ((b) >> 20ULL)
+
+#define HOST_LOW_GM_SIZE MB_TO_BYTES(128)
+#define HOST_HIGH_GM_SIZE MB_TO_BYTES(384)
+#define HOST_FENCE 4
+
+static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	u32 alloc_flag, search_flag;
+	u64 start, end, size;
+	struct drm_mm_node *node;
+	int retried = 0;
+	int ret;
+
+	if (high_gm) {
+		search_flag = DRM_MM_SEARCH_BELOW;
+		alloc_flag = DRM_MM_CREATE_TOP;
+		node = &vgpu->gm.high_gm_node;
+		size = vgpu_hidden_sz(vgpu);
+		start = gvt_hidden_gmadr_base(gvt);
+		end = gvt_hidden_gmadr_end(gvt);
+	} else {
+		search_flag = DRM_MM_SEARCH_DEFAULT;
+		alloc_flag = DRM_MM_CREATE_DEFAULT;
+		node = &vgpu->gm.low_gm_node;
+		size = vgpu_aperture_sz(vgpu);
+		start = gvt_aperture_gmadr_base(gvt);
+		end = gvt_aperture_gmadr_end(gvt);
+	}
+
+	mutex_lock(&dev_priv->drm.struct_mutex);
+search_again:
+	ret = drm_mm_insert_node_in_range_generic(&dev_priv->ggtt.base.mm,
+						  node, size, 4096, 0,
+						  start, end, search_flag,
+						  alloc_flag);
+	if (ret) {
+		ret = i915_gem_evict_something(&dev_priv->ggtt.base,
+					       size, 4096, 0, start, end, 0);
+		if (ret == 0 && ++retried < 3)
+			goto search_again;
+
+		gvt_err("fail to alloc %s gm space from host, retried %d\n",
+				high_gm ? "high" : "low", retried);
+	}
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+	return ret;
+}
+
+static int alloc_vgpu_gm(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	int ret;
+
+	ret = alloc_gm(vgpu, false);
+	if (ret)
+		return ret;
+
+	ret = alloc_gm(vgpu, true);
+	if (ret)
+		goto out_free_aperture;
+
+	gvt_dbg_core("vgpu%d: alloc low GM start %llx size %llx\n", vgpu->id,
+		     vgpu_aperture_offset(vgpu), vgpu_aperture_sz(vgpu));
+
+	gvt_dbg_core("vgpu%d: alloc high GM start %llx size %llx\n", vgpu->id,
+		     vgpu_hidden_offset(vgpu), vgpu_hidden_sz(vgpu));
+
+	return 0;
+out_free_aperture:
+	mutex_lock(&dev_priv->drm.struct_mutex);
+	drm_mm_remove_node(&vgpu->gm.low_gm_node);
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+	return ret;
+}
+
+static void free_vgpu_gm(struct intel_vgpu *vgpu)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	mutex_lock(&dev_priv->drm.struct_mutex);
+	drm_mm_remove_node(&vgpu->gm.low_gm_node);
+	drm_mm_remove_node(&vgpu->gm.high_gm_node);
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+}
+
+/**
+ * intel_vgpu_write_fence - write fence registers owned by a vGPU
+ * @vgpu: vGPU instance
+ * @fence: vGPU fence register number
+ * @value: Fence register value to be written
+ *
+ * This function is used to write fence registers owned by a vGPU. The vGPU
+ * fence register number will be translated into HW fence register number.
+ *
+ */
+void intel_vgpu_write_fence(struct intel_vgpu *vgpu,
+		u32 fence, u64 value)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct drm_i915_fence_reg *reg;
+	i915_reg_t fence_reg_lo, fence_reg_hi;
+
+	if (WARN_ON(fence > vgpu_fence_sz(vgpu)))
+		return;
+
+	reg = vgpu->fence.regs[fence];
+	if (WARN_ON(!reg))
+		return;
+
+	fence_reg_lo = FENCE_REG_GEN6_LO(reg->id);
+	fence_reg_hi = FENCE_REG_GEN6_HI(reg->id);
+
+	I915_WRITE(fence_reg_lo, 0);
+	POSTING_READ(fence_reg_lo);
+
+	I915_WRITE(fence_reg_hi, upper_32_bits(value));
+	I915_WRITE(fence_reg_lo, lower_32_bits(value));
+	POSTING_READ(fence_reg_lo);
+}
+
+static void free_vgpu_fence(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct drm_i915_fence_reg *reg;
+	u32 i;
+
+	if (WARN_ON(!vgpu_fence_sz(vgpu)))
+		return;
+
+	mutex_lock(&dev_priv->drm.struct_mutex);
+	for (i = 0; i < vgpu_fence_sz(vgpu); i++) {
+		reg = vgpu->fence.regs[i];
+		intel_vgpu_write_fence(vgpu, i, 0);
+		list_add_tail(&reg->link,
+			      &dev_priv->mm.fence_list);
+	}
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+}
+
+static int alloc_vgpu_fence(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct drm_i915_fence_reg *reg;
+	int i;
+	struct list_head *pos, *q;
+
+	/* Request fences from host */
+	mutex_lock(&dev_priv->drm.struct_mutex);
+	i = 0;
+	list_for_each_safe(pos, q, &dev_priv->mm.fence_list) {
+		reg = list_entry(pos, struct drm_i915_fence_reg, link);
+		if (reg->pin_count || reg->vma)
+			continue;
+		list_del(pos);
+		vgpu->fence.regs[i] = reg;
+		intel_vgpu_write_fence(vgpu, i, 0);
+		if (++i == vgpu_fence_sz(vgpu))
+			break;
+	}
+	if (i != vgpu_fence_sz(vgpu))
+		goto out_free_fence;
+
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+	return 0;
+out_free_fence:
+	/* Return fences to host, if fail */
+	for (i = 0; i < vgpu_fence_sz(vgpu); i++) {
+		reg = vgpu->fence.regs[i];
+		if (!reg)
+			continue;
+		list_add_tail(&reg->link,
+			      &dev_priv->mm.fence_list);
+	}
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+	return -ENOSPC;
+}
+
+static void free_resource(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+
+	gvt->gm.vgpu_allocated_low_gm_size -= vgpu_aperture_sz(vgpu);
+	gvt->gm.vgpu_allocated_high_gm_size -= vgpu_hidden_sz(vgpu);
+	gvt->fence.vgpu_allocated_fence_num -= vgpu_fence_sz(vgpu);
+}
+
+static int alloc_resource(struct intel_vgpu *vgpu,
+		struct intel_vgpu_creation_params *param)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	unsigned long request, avail, max, taken;
+	const char *item;
+
+	if (!param->low_gm_sz || !param->high_gm_sz || !param->fence_sz) {
+		gvt_err("Invalid vGPU creation params\n");
+		return -EINVAL;
+	}
+
+	item = "low GM space";
+	max = gvt_aperture_sz(gvt) - HOST_LOW_GM_SIZE;
+	taken = gvt->gm.vgpu_allocated_low_gm_size;
+	avail = max - taken;
+	request = MB_TO_BYTES(param->low_gm_sz);
+
+	if (request > avail)
+		goto no_enough_resource;
+
+	vgpu_aperture_sz(vgpu) = request;
+
+	item = "high GM space";
+	max = gvt_hidden_sz(gvt) - HOST_HIGH_GM_SIZE;
+	taken = gvt->gm.vgpu_allocated_high_gm_size;
+	avail = max - taken;
+	request = MB_TO_BYTES(param->high_gm_sz);
+
+	if (request > avail)
+		goto no_enough_resource;
+
+	vgpu_hidden_sz(vgpu) = request;
+
+	item = "fence";
+	max = gvt_fence_sz(gvt) - HOST_FENCE;
+	taken = gvt->fence.vgpu_allocated_fence_num;
+	avail = max - taken;
+	request = param->fence_sz;
+
+	if (request > avail)
+		goto no_enough_resource;
+
+	vgpu_fence_sz(vgpu) = request;
+
+	gvt->gm.vgpu_allocated_low_gm_size += MB_TO_BYTES(param->low_gm_sz);
+	gvt->gm.vgpu_allocated_high_gm_size += MB_TO_BYTES(param->high_gm_sz);
+	gvt->fence.vgpu_allocated_fence_num += param->fence_sz;
+	return 0;
+
+no_enough_resource:
+	gvt_err("vgpu%d: fail to allocate resource %s\n", vgpu->id, item);
+	gvt_err("vgpu%d: request %luMB avail %luMB max %luMB taken %luMB\n",
+		vgpu->id, BYTES_TO_MB(request), BYTES_TO_MB(avail),
+		BYTES_TO_MB(max), BYTES_TO_MB(taken));
+	return -ENOSPC;
+}
+
+/**
+ * inte_gvt_free_vgpu_resource - free HW resource owned by a vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is used to free the HW resource owned by a vGPU.
+ *
+ */
+void intel_vgpu_free_resource(struct intel_vgpu *vgpu)
+{
+	free_vgpu_gm(vgpu);
+	free_vgpu_fence(vgpu);
+	free_resource(vgpu);
+}
+
+/**
+ * intel_alloc_vgpu_resource - allocate HW resource for a vGPU
+ * @vgpu: vGPU
+ * @param: vGPU creation params
+ *
+ * This function is used to allocate HW resource for a vGPU. User specifies
+ * the resource configuration through the creation params.
+ *
+ * Returns:
+ * zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_alloc_resource(struct intel_vgpu *vgpu,
+		struct intel_vgpu_creation_params *param)
+{
+	int ret;
+
+	ret = alloc_resource(vgpu, param);
+	if (ret)
+		return ret;
+
+	ret = alloc_vgpu_gm(vgpu);
+	if (ret)
+		goto out_free_resource;
+
+	ret = alloc_vgpu_fence(vgpu);
+	if (ret)
+		goto out_free_vgpu_gm;
+
+	return 0;
+
+out_free_vgpu_gm:
+	free_vgpu_gm(vgpu);
+out_free_resource:
+	free_resource(vgpu);
+	return ret;
+}

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

@@ -0,0 +1,287 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+enum {
+	INTEL_GVT_PCI_BAR_GTTMMIO = 0,
+	INTEL_GVT_PCI_BAR_APERTURE,
+	INTEL_GVT_PCI_BAR_PIO,
+	INTEL_GVT_PCI_BAR_MAX,
+};
+
+/**
+ * intel_vgpu_emulate_cfg_read - emulate vGPU configuration space read
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_emulate_cfg_read(void *__vgpu, unsigned int offset,
+	void *p_data, unsigned int bytes)
+{
+	struct intel_vgpu *vgpu = __vgpu;
+
+	if (WARN_ON(bytes > 4))
+		return -EINVAL;
+
+	if (WARN_ON(offset + bytes > INTEL_GVT_MAX_CFG_SPACE_SZ))
+		return -EINVAL;
+
+	memcpy(p_data, vgpu_cfg_space(vgpu) + offset, bytes);
+	return 0;
+}
+
+static int map_aperture(struct intel_vgpu *vgpu, bool map)
+{
+	u64 first_gfn, first_mfn;
+	u64 val;
+	int ret;
+
+	if (map == vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked)
+		return 0;
+
+	val = vgpu_cfg_space(vgpu)[PCI_BASE_ADDRESS_2];
+	if (val & PCI_BASE_ADDRESS_MEM_TYPE_64)
+		val = *(u64 *)(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_2);
+	else
+		val = *(u32 *)(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_2);
+
+	first_gfn = (val + vgpu_aperture_offset(vgpu)) >> PAGE_SHIFT;
+	first_mfn = vgpu_aperture_pa_base(vgpu) >> PAGE_SHIFT;
+
+	ret = intel_gvt_hypervisor_map_gfn_to_mfn(vgpu, first_gfn,
+						  first_mfn,
+						  vgpu_aperture_sz(vgpu)
+						  >> PAGE_SHIFT, map,
+						  GVT_MAP_APERTURE);
+	if (ret)
+		return ret;
+
+	vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked = map;
+	return 0;
+}
+
+static int trap_gttmmio(struct intel_vgpu *vgpu, bool trap)
+{
+	u64 start, end;
+	u64 val;
+	int ret;
+
+	if (trap == vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_GTTMMIO].tracked)
+		return 0;
+
+	val = vgpu_cfg_space(vgpu)[PCI_BASE_ADDRESS_0];
+	if (val & PCI_BASE_ADDRESS_MEM_TYPE_64)
+		start = *(u64 *)(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_0);
+	else
+		start = *(u32 *)(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_0);
+
+	start &= ~GENMASK(3, 0);
+	end = start + vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_GTTMMIO].size - 1;
+
+	ret = intel_gvt_hypervisor_set_trap_area(vgpu, start, end, trap);
+	if (ret)
+		return ret;
+
+	vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_GTTMMIO].tracked = trap;
+	return 0;
+}
+
+static int emulate_pci_command_write(struct intel_vgpu *vgpu,
+	unsigned int offset, void *p_data, unsigned int bytes)
+{
+	u8 old = vgpu_cfg_space(vgpu)[offset];
+	u8 new = *(u8 *)p_data;
+	u8 changed = old ^ new;
+	int ret;
+
+	if (!(changed & PCI_COMMAND_MEMORY))
+		return 0;
+
+	if (old & PCI_COMMAND_MEMORY) {
+		ret = trap_gttmmio(vgpu, false);
+		if (ret)
+			return ret;
+		ret = map_aperture(vgpu, false);
+		if (ret)
+			return ret;
+	} else {
+		ret = trap_gttmmio(vgpu, true);
+		if (ret)
+			return ret;
+		ret = map_aperture(vgpu, true);
+		if (ret)
+			return ret;
+	}
+
+	memcpy(vgpu_cfg_space(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)
+{
+	unsigned int bar_index =
+		(rounddown(offset, 8) % PCI_BASE_ADDRESS_0) / 8;
+	u32 new = *(u32 *)(p_data);
+	bool lo = IS_ALIGNED(offset, 8);
+	u64 size;
+	int ret = 0;
+	bool mmio_enabled =
+		vgpu_cfg_space(vgpu)[PCI_COMMAND] & PCI_COMMAND_MEMORY;
+
+	if (WARN_ON(bar_index >= INTEL_GVT_PCI_BAR_MAX))
+		return -EINVAL;
+
+	if (new == 0xffffffff) {
+		/*
+		 * Power-up software can determine how much address
+		 * space the device requires by writing a value of
+		 * all 1's to the register and then reading the value
+		 * back. The device will return 0's in all don't-care
+		 * address bits.
+		 */
+		size = vgpu->cfg_space.bar[bar_index].size;
+		if (lo) {
+			new = rounddown(new, size);
+		} else {
+			u32 val = vgpu_cfg_space(vgpu)[rounddown(offset, 8)];
+			/* for 32bit mode bar it returns all-0 in upper 32
+			 * bit, for 64bit mode bar it will calculate the
+			 * size with lower 32bit and return the corresponding
+			 * value
+			 */
+			if (val & PCI_BASE_ADDRESS_MEM_TYPE_64)
+				new &= (~(size-1)) >> 32;
+			else
+				new = 0;
+		}
+		/*
+		 * Unmapp & untrap the BAR, since guest hasn't configured a
+		 * valid GPA
+		 */
+		switch (bar_index) {
+		case INTEL_GVT_PCI_BAR_GTTMMIO:
+			ret = trap_gttmmio(vgpu, false);
+			break;
+		case INTEL_GVT_PCI_BAR_APERTURE:
+			ret = map_aperture(vgpu, false);
+			break;
+		}
+		intel_vgpu_write_pci_bar(vgpu, offset, new, lo);
+	} else {
+		/*
+		 * Unmapp & untrap the old BAR first, since guest has
+		 * re-configured the BAR
+		 */
+		switch (bar_index) {
+		case INTEL_GVT_PCI_BAR_GTTMMIO:
+			ret = trap_gttmmio(vgpu, false);
+			break;
+		case INTEL_GVT_PCI_BAR_APERTURE:
+			ret = map_aperture(vgpu, false);
+			break;
+		}
+		intel_vgpu_write_pci_bar(vgpu, offset, new, lo);
+		/* Track the new BAR */
+		if (mmio_enabled) {
+			switch (bar_index) {
+			case INTEL_GVT_PCI_BAR_GTTMMIO:
+				ret = trap_gttmmio(vgpu, true);
+				break;
+			case INTEL_GVT_PCI_BAR_APERTURE:
+				ret = map_aperture(vgpu, true);
+				break;
+			}
+		}
+	}
+	return ret;
+}
+
+/**
+ * intel_vgpu_emulate_cfg_read - emulate vGPU configuration space write
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_emulate_cfg_write(void *__vgpu, unsigned int offset,
+	void *p_data, unsigned int bytes)
+{
+	struct intel_vgpu *vgpu = __vgpu;
+	int ret;
+
+	if (WARN_ON(bytes > 4))
+		return -EINVAL;
+
+	if (WARN_ON(offset + bytes >= INTEL_GVT_MAX_CFG_SPACE_SZ))
+		return -EINVAL;
+
+	/* First check if it's PCI_COMMAND */
+	if (IS_ALIGNED(offset, 2) && offset == PCI_COMMAND) {
+		if (WARN_ON(bytes > 2))
+			return -EINVAL;
+		return emulate_pci_command_write(vgpu, offset, p_data, bytes);
+	}
+
+	switch (rounddown(offset, 4)) {
+	case PCI_BASE_ADDRESS_0:
+	case PCI_BASE_ADDRESS_1:
+	case PCI_BASE_ADDRESS_2:
+	case PCI_BASE_ADDRESS_3:
+		if (WARN_ON(!IS_ALIGNED(offset, 4)))
+			return -EINVAL;
+		return emulate_pci_bar_write(vgpu, offset, p_data, bytes);
+
+	case INTEL_GVT_PCI_SWSCI:
+		if (WARN_ON(!IS_ALIGNED(offset, 4)))
+			return -EINVAL;
+		ret = intel_vgpu_emulate_opregion_request(vgpu, *(u32 *)p_data);
+		if (ret)
+			return ret;
+		break;
+
+	case INTEL_GVT_PCI_OPREGION:
+		if (WARN_ON(!IS_ALIGNED(offset, 4)))
+			return -EINVAL;
+		ret = intel_vgpu_init_opregion(vgpu, *(u32 *)p_data);
+		if (ret)
+			return ret;
+
+		memcpy(vgpu_cfg_space(vgpu) + offset, p_data, bytes);
+		break;
+	default:
+		memcpy(vgpu_cfg_space(vgpu) + offset, p_data, bytes);
+		break;
+	}
+	return 0;
+}

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

@@ -0,0 +1,49 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Min He <min.he@intel.com>
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Yulei Zhang <yulei.zhang@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+#ifndef _GVT_CMD_PARSER_H_
+#define _GVT_CMD_PARSER_H_
+
+#define GVT_CMD_HASH_BITS 7
+
+void intel_gvt_clean_cmd_parser(struct intel_gvt *gvt);
+
+int intel_gvt_init_cmd_parser(struct intel_gvt *gvt);
+
+int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload);
+
+int intel_gvt_scan_and_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx);
+
+#endif

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

@@ -24,11 +24,34 @@
 #ifndef __GVT_DEBUG_H__
 #define __GVT_DEBUG_H__
 
+#define gvt_err(fmt, args...) \
+	DRM_ERROR("gvt: "fmt, ##args)
+
 #define gvt_dbg_core(fmt, args...) \
 	DRM_DEBUG_DRIVER("gvt: core: "fmt, ##args)
 
-/*
- * Other GVT debug stuff will be introduced in the GVT device model patches.
- */
+#define gvt_dbg_irq(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: irq: "fmt, ##args)
+
+#define gvt_dbg_mm(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: mm: "fmt, ##args)
+
+#define gvt_dbg_mmio(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: mmio: "fmt, ##args)
+
+#define gvt_dbg_dpy(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: dpy: "fmt, ##args)
+
+#define gvt_dbg_el(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: el: "fmt, ##args)
+
+#define gvt_dbg_sched(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: sched: "fmt, ##args)
+
+#define gvt_dbg_render(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: render: "fmt, ##args)
+
+#define gvt_dbg_cmd(fmt, args...) \
+	DRM_DEBUG_DRIVER("gvt: cmd: "fmt, ##args)
 
 #endif

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

@@ -0,0 +1,329 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Terrence Xu <terrence.xu@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+static int get_edp_pipe(struct intel_vgpu *vgpu)
+{
+	u32 data = vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP);
+	int pipe = -1;
+
+	switch (data & TRANS_DDI_EDP_INPUT_MASK) {
+	case TRANS_DDI_EDP_INPUT_A_ON:
+	case TRANS_DDI_EDP_INPUT_A_ONOFF:
+		pipe = PIPE_A;
+		break;
+	case TRANS_DDI_EDP_INPUT_B_ONOFF:
+		pipe = PIPE_B;
+		break;
+	case TRANS_DDI_EDP_INPUT_C_ONOFF:
+		pipe = PIPE_C;
+		break;
+	}
+	return pipe;
+}
+
+static int edp_pipe_is_enabled(struct intel_vgpu *vgpu)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	if (!(vgpu_vreg(vgpu, PIPECONF(_PIPE_EDP)) & PIPECONF_ENABLE))
+		return 0;
+
+	if (!(vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP) & TRANS_DDI_FUNC_ENABLE))
+		return 0;
+	return 1;
+}
+
+static int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	if (WARN_ON(pipe < PIPE_A || pipe >= I915_MAX_PIPES))
+		return -EINVAL;
+
+	if (vgpu_vreg(vgpu, PIPECONF(pipe)) & PIPECONF_ENABLE)
+		return 1;
+
+	if (edp_pipe_is_enabled(vgpu) &&
+			get_edp_pipe(vgpu) == pipe)
+		return 1;
+	return 0;
+}
+
+/* EDID with 1024x768 as its resolution */
+static unsigned char virtual_dp_monitor_edid[] = {
+	/*Header*/
+	0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x00,
+	/* Vendor & Product Identification */
+	0x22, 0xf0, 0x54, 0x29, 0x00, 0x00, 0x00, 0x00, 0x04, 0x17,
+	/* Version & Revision */
+	0x01, 0x04,
+	/* Basic Display Parameters & Features */
+	0xa5, 0x34, 0x20, 0x78, 0x23,
+	/* Color Characteristics */
+	0xfc, 0x81, 0xa4, 0x55, 0x4d, 0x9d, 0x25, 0x12, 0x50, 0x54,
+	/* Established Timings: maximum resolution is 1024x768 */
+	0x21, 0x08, 0x00,
+	/* Standard Timings. All invalid */
+	0x00, 0xc0, 0x00, 0xc0, 0x00, 0x40, 0x00, 0x80, 0x00, 0x00,
+	0x00, 0x40, 0x00, 0x00, 0x00, 0x01,
+	/* 18 Byte Data Blocks 1: invalid */
+	0x00, 0x00, 0x80, 0xa0, 0x70, 0xb0,
+	0x23, 0x40, 0x30, 0x20, 0x36, 0x00, 0x06, 0x44, 0x21, 0x00, 0x00, 0x1a,
+	/* 18 Byte Data Blocks 2: invalid */
+	0x00, 0x00, 0x00, 0xfd, 0x00, 0x18, 0x3c, 0x18, 0x50, 0x11, 0x00, 0x0a,
+	0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+	/* 18 Byte Data Blocks 3: invalid */
+	0x00, 0x00, 0x00, 0xfc, 0x00, 0x48,
+	0x50, 0x20, 0x5a, 0x52, 0x32, 0x34, 0x34, 0x30, 0x77, 0x0a, 0x20, 0x20,
+	/* 18 Byte Data Blocks 4: invalid */
+	0x00, 0x00, 0x00, 0xff, 0x00, 0x43, 0x4e, 0x34, 0x33, 0x30, 0x34, 0x30,
+	0x44, 0x58, 0x51, 0x0a, 0x20, 0x20,
+	/* Extension Block Count */
+	0x00,
+	/* Checksum */
+	0xef,
+};
+
+#define DPCD_HEADER_SIZE        0xb
+
+u8 dpcd_fix_data[DPCD_HEADER_SIZE] = {
+	0x11, 0x0a, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+};
+
+static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	vgpu_vreg(vgpu, SDEISR) &= ~(SDE_PORTB_HOTPLUG_CPT |
+			SDE_PORTC_HOTPLUG_CPT |
+			SDE_PORTD_HOTPLUG_CPT);
+
+	if (IS_SKYLAKE(dev_priv))
+		vgpu_vreg(vgpu, SDEISR) &= ~(SDE_PORTA_HOTPLUG_SPT |
+				SDE_PORTE_HOTPLUG_SPT);
+
+	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B))
+		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTB_HOTPLUG_CPT;
+
+	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C))
+		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTC_HOTPLUG_CPT;
+
+	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_D))
+		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
+
+	if (IS_SKYLAKE(dev_priv) &&
+			intel_vgpu_has_monitor_on_port(vgpu, PORT_E)) {
+		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTE_HOTPLUG_SPT;
+	}
+
+	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
+		if (IS_BROADWELL(dev_priv))
+			vgpu_vreg(vgpu, GEN8_DE_PORT_ISR) |=
+				GEN8_PORT_DP_A_HOTPLUG;
+		else
+			vgpu_vreg(vgpu, SDEISR) |= SDE_PORTA_HOTPLUG_SPT;
+	}
+}
+
+static void clean_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num)
+{
+	struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
+
+	kfree(port->edid);
+	port->edid = NULL;
+
+	kfree(port->dpcd);
+	port->dpcd = NULL;
+}
+
+static int setup_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num,
+		int type)
+{
+	struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
+
+	port->edid = kzalloc(sizeof(*(port->edid)), GFP_KERNEL);
+	if (!port->edid)
+		return -ENOMEM;
+
+	port->dpcd = kzalloc(sizeof(*(port->dpcd)), GFP_KERNEL);
+	if (!port->dpcd) {
+		kfree(port->edid);
+		return -ENOMEM;
+	}
+
+	memcpy(port->edid->edid_block, virtual_dp_monitor_edid,
+			EDID_SIZE);
+	port->edid->data_valid = true;
+
+	memcpy(port->dpcd->data, dpcd_fix_data, DPCD_HEADER_SIZE);
+	port->dpcd->data_valid = true;
+	port->dpcd->data[DPCD_SINK_COUNT] = 0x1;
+	port->type = type;
+
+	emulate_monitor_status_change(vgpu);
+	return 0;
+}
+
+/**
+ * intel_gvt_check_vblank_emulation - check if vblank emulation timer should
+ * be turned on/off when a virtual pipe is enabled/disabled.
+ * @gvt: a GVT device
+ *
+ * This function is used to turn on/off vblank timer according to currently
+ * enabled/disabled virtual pipes.
+ *
+ */
+void intel_gvt_check_vblank_emulation(struct intel_gvt *gvt)
+{
+	struct intel_gvt_irq *irq = &gvt->irq;
+	struct intel_vgpu *vgpu;
+	bool have_enabled_pipe = false;
+	int pipe, id;
+
+	if (WARN_ON(!mutex_is_locked(&gvt->lock)))
+		return;
+
+	hrtimer_cancel(&irq->vblank_timer.timer);
+
+	for_each_active_vgpu(gvt, vgpu, id) {
+		for (pipe = 0; pipe < I915_MAX_PIPES; pipe++) {
+			have_enabled_pipe =
+				pipe_is_enabled(vgpu, pipe);
+			if (have_enabled_pipe)
+				break;
+		}
+	}
+
+	if (have_enabled_pipe)
+		hrtimer_start(&irq->vblank_timer.timer,
+			ktime_add_ns(ktime_get(), irq->vblank_timer.period),
+			HRTIMER_MODE_ABS);
+}
+
+static void emulate_vblank_on_pipe(struct intel_vgpu *vgpu, int pipe)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_vgpu_irq *irq = &vgpu->irq;
+	int vblank_event[] = {
+		[PIPE_A] = PIPE_A_VBLANK,
+		[PIPE_B] = PIPE_B_VBLANK,
+		[PIPE_C] = PIPE_C_VBLANK,
+	};
+	int event;
+
+	if (pipe < PIPE_A || pipe > PIPE_C)
+		return;
+
+	for_each_set_bit(event, irq->flip_done_event[pipe],
+			INTEL_GVT_EVENT_MAX) {
+		clear_bit(event, irq->flip_done_event[pipe]);
+		if (!pipe_is_enabled(vgpu, pipe))
+			continue;
+
+		vgpu_vreg(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
+		intel_vgpu_trigger_virtual_event(vgpu, event);
+	}
+
+	if (pipe_is_enabled(vgpu, pipe)) {
+		vgpu_vreg(vgpu, PIPE_FRMCOUNT_G4X(pipe))++;
+		intel_vgpu_trigger_virtual_event(vgpu, vblank_event[pipe]);
+	}
+}
+
+static void emulate_vblank(struct intel_vgpu *vgpu)
+{
+	int pipe;
+
+	for_each_pipe(vgpu->gvt->dev_priv, pipe)
+		emulate_vblank_on_pipe(vgpu, pipe);
+}
+
+/**
+ * intel_gvt_emulate_vblank - trigger vblank events for vGPUs on GVT device
+ * @gvt: a GVT device
+ *
+ * This function is used to trigger vblank interrupts for vGPUs on GVT device
+ *
+ */
+void intel_gvt_emulate_vblank(struct intel_gvt *gvt)
+{
+	struct intel_vgpu *vgpu;
+	int id;
+
+	if (WARN_ON(!mutex_is_locked(&gvt->lock)))
+		return;
+
+	for_each_active_vgpu(gvt, vgpu, id)
+		emulate_vblank(vgpu);
+}
+
+/**
+ * intel_vgpu_clean_display - clean vGPU virtual display emulation
+ * @vgpu: a vGPU
+ *
+ * This function is used to clean vGPU virtual display emulation stuffs
+ *
+ */
+void intel_vgpu_clean_display(struct intel_vgpu *vgpu)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	if (IS_SKYLAKE(dev_priv))
+		clean_virtual_dp_monitor(vgpu, PORT_D);
+	else
+		clean_virtual_dp_monitor(vgpu, PORT_B);
+}
+
+/**
+ * intel_vgpu_init_display- initialize vGPU virtual display emulation
+ * @vgpu: a vGPU
+ *
+ * This function is used to initialize vGPU virtual display emulation stuffs
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_init_display(struct intel_vgpu *vgpu)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	intel_vgpu_init_i2c_edid(vgpu);
+
+	if (IS_SKYLAKE(dev_priv))
+		return setup_virtual_dp_monitor(vgpu, PORT_D, GVT_DP_D);
+	else
+		return setup_virtual_dp_monitor(vgpu, PORT_B, GVT_DP_B);
+}

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

@@ -0,0 +1,163 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Terrence Xu <terrence.xu@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_DISPLAY_H_
+#define _GVT_DISPLAY_H_
+
+#define SBI_REG_MAX	20
+#define DPCD_SIZE	0x700
+
+#define intel_vgpu_port(vgpu, port) \
+	(&(vgpu->display.ports[port]))
+
+#define intel_vgpu_has_monitor_on_port(vgpu, port) \
+	(intel_vgpu_port(vgpu, port)->edid && \
+		intel_vgpu_port(vgpu, port)->edid->data_valid)
+
+#define intel_vgpu_port_is_dp(vgpu, port) \
+	((intel_vgpu_port(vgpu, port)->type == GVT_DP_A) || \
+	(intel_vgpu_port(vgpu, port)->type == GVT_DP_B) || \
+	(intel_vgpu_port(vgpu, port)->type == GVT_DP_C) || \
+	(intel_vgpu_port(vgpu, port)->type == GVT_DP_D))
+
+#define INTEL_GVT_MAX_UEVENT_VARS	3
+
+/* DPCD start */
+#define DPCD_SIZE	0x700
+
+/* DPCD */
+#define DP_SET_POWER            0x600
+#define DP_SET_POWER_D0         0x1
+#define AUX_NATIVE_WRITE        0x8
+#define AUX_NATIVE_READ         0x9
+
+#define AUX_NATIVE_REPLY_MASK   (0x3 << 4)
+#define AUX_NATIVE_REPLY_ACK    (0x0 << 4)
+#define AUX_NATIVE_REPLY_NAK    (0x1 << 4)
+#define AUX_NATIVE_REPLY_DEFER  (0x2 << 4)
+
+#define AUX_BURST_SIZE          16
+
+/* DPCD addresses */
+#define DPCD_REV			0x000
+#define DPCD_MAX_LINK_RATE		0x001
+#define DPCD_MAX_LANE_COUNT		0x002
+
+#define DPCD_TRAINING_PATTERN_SET	0x102
+#define	DPCD_SINK_COUNT			0x200
+#define DPCD_LANE0_1_STATUS		0x202
+#define DPCD_LANE2_3_STATUS		0x203
+#define DPCD_LANE_ALIGN_STATUS_UPDATED	0x204
+#define DPCD_SINK_STATUS		0x205
+
+/* link training */
+#define DPCD_TRAINING_PATTERN_SET_MASK	0x03
+#define DPCD_LINK_TRAINING_DISABLED	0x00
+#define DPCD_TRAINING_PATTERN_1		0x01
+#define DPCD_TRAINING_PATTERN_2		0x02
+
+#define DPCD_CP_READY_MASK		(1 << 6)
+
+/* lane status */
+#define DPCD_LANES_CR_DONE		0x11
+#define DPCD_LANES_EQ_DONE		0x22
+#define DPCD_SYMBOL_LOCKED		0x44
+
+#define DPCD_INTERLANE_ALIGN_DONE	0x01
+
+#define DPCD_SINK_IN_SYNC		0x03
+/* DPCD end */
+
+#define SBI_RESPONSE_MASK               0x3
+#define SBI_RESPONSE_SHIFT              0x1
+#define SBI_STAT_MASK                   0x1
+#define SBI_STAT_SHIFT                  0x0
+#define SBI_OPCODE_SHIFT                8
+#define SBI_OPCODE_MASK			(0xff << SBI_OPCODE_SHIFT)
+#define SBI_CMD_IORD                    2
+#define SBI_CMD_IOWR                    3
+#define SBI_CMD_CRRD                    6
+#define SBI_CMD_CRWR                    7
+#define SBI_ADDR_OFFSET_SHIFT           16
+#define SBI_ADDR_OFFSET_MASK            (0xffff << SBI_ADDR_OFFSET_SHIFT)
+
+struct intel_vgpu_sbi_register {
+	unsigned int offset;
+	u32 value;
+};
+
+struct intel_vgpu_sbi {
+	int number;
+	struct intel_vgpu_sbi_register registers[SBI_REG_MAX];
+};
+
+enum intel_gvt_plane_type {
+	PRIMARY_PLANE = 0,
+	CURSOR_PLANE,
+	SPRITE_PLANE,
+	MAX_PLANE
+};
+
+struct intel_vgpu_dpcd_data {
+	bool data_valid;
+	u8 data[DPCD_SIZE];
+};
+
+enum intel_vgpu_port_type {
+	GVT_CRT = 0,
+	GVT_DP_A,
+	GVT_DP_B,
+	GVT_DP_C,
+	GVT_DP_D,
+	GVT_HDMI_B,
+	GVT_HDMI_C,
+	GVT_HDMI_D,
+	GVT_PORT_MAX
+};
+
+struct intel_vgpu_port {
+	/* per display EDID information */
+	struct intel_vgpu_edid_data *edid;
+	/* per display DPCD information */
+	struct intel_vgpu_dpcd_data *dpcd;
+	int type;
+};
+
+void intel_gvt_emulate_vblank(struct intel_gvt *gvt);
+void intel_gvt_check_vblank_emulation(struct intel_gvt *gvt);
+
+int intel_vgpu_init_display(struct intel_vgpu *vgpu);
+void intel_vgpu_clean_display(struct intel_vgpu *vgpu);
+
+#endif

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

@@ -0,0 +1,531 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Terrence Xu <terrence.xu@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+#define GMBUS1_TOTAL_BYTES_SHIFT 16
+#define GMBUS1_TOTAL_BYTES_MASK 0x1ff
+#define gmbus1_total_byte_count(v) (((v) >> \
+	GMBUS1_TOTAL_BYTES_SHIFT) & GMBUS1_TOTAL_BYTES_MASK)
+#define gmbus1_slave_addr(v) (((v) & 0xff) >> 1)
+#define gmbus1_slave_index(v) (((v) >> 8) & 0xff)
+#define gmbus1_bus_cycle(v) (((v) >> 25) & 0x7)
+
+/* GMBUS0 bits definitions */
+#define _GMBUS_PIN_SEL_MASK     (0x7)
+
+static unsigned char edid_get_byte(struct intel_vgpu *vgpu)
+{
+	struct intel_vgpu_i2c_edid *edid = &vgpu->display.i2c_edid;
+	unsigned char chr = 0;
+
+	if (edid->state == I2C_NOT_SPECIFIED || !edid->slave_selected) {
+		gvt_err("Driver tries to read EDID without proper sequence!\n");
+		return 0;
+	}
+	if (edid->current_edid_read >= EDID_SIZE) {
+		gvt_err("edid_get_byte() exceeds the size of EDID!\n");
+		return 0;
+	}
+
+	if (!edid->edid_available) {
+		gvt_err("Reading EDID but EDID is not available!\n");
+		return 0;
+	}
+
+	if (intel_vgpu_has_monitor_on_port(vgpu, edid->port)) {
+		struct intel_vgpu_edid_data *edid_data =
+			intel_vgpu_port(vgpu, edid->port)->edid;
+
+		chr = edid_data->edid_block[edid->current_edid_read];
+		edid->current_edid_read++;
+	} else {
+		gvt_err("No EDID available during the reading?\n");
+	}
+	return chr;
+}
+
+static inline int get_port_from_gmbus0(u32 gmbus0)
+{
+	int port_select = gmbus0 & _GMBUS_PIN_SEL_MASK;
+	int port = -EINVAL;
+
+	if (port_select == 2)
+		port = PORT_E;
+	else if (port_select == 4)
+		port = PORT_C;
+	else if (port_select == 5)
+		port = PORT_B;
+	else if (port_select == 6)
+		port = PORT_D;
+	return port;
+}
+
+static void reset_gmbus_controller(struct intel_vgpu *vgpu)
+{
+	vgpu_vreg(vgpu, PCH_GMBUS2) = GMBUS_HW_RDY;
+	if (!vgpu->display.i2c_edid.edid_available)
+		vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
+	vgpu->display.i2c_edid.gmbus.phase = GMBUS_IDLE_PHASE;
+}
+
+/* GMBUS0 */
+static int gmbus0_mmio_write(struct intel_vgpu *vgpu,
+			unsigned int offset, void *p_data, unsigned int bytes)
+{
+	int port, pin_select;
+
+	memcpy(&vgpu_vreg(vgpu, offset), p_data, bytes);
+
+	pin_select = vgpu_vreg(vgpu, offset) & _GMBUS_PIN_SEL_MASK;
+
+	intel_vgpu_init_i2c_edid(vgpu);
+
+	if (pin_select == 0)
+		return 0;
+
+	port = get_port_from_gmbus0(pin_select);
+	if (WARN_ON(port < 0))
+		return 0;
+
+	vgpu->display.i2c_edid.state = I2C_GMBUS;
+	vgpu->display.i2c_edid.gmbus.phase = GMBUS_IDLE_PHASE;
+
+	vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
+	vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY | GMBUS_HW_WAIT_PHASE;
+
+	if (intel_vgpu_has_monitor_on_port(vgpu, port) &&
+			!intel_vgpu_port_is_dp(vgpu, port)) {
+		vgpu->display.i2c_edid.port = port;
+		vgpu->display.i2c_edid.edid_available = true;
+		vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_SATOER;
+	} else
+		vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
+	return 0;
+}
+
+static int gmbus1_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	struct intel_vgpu_i2c_edid *i2c_edid = &vgpu->display.i2c_edid;
+	u32 slave_addr;
+	u32 wvalue = *(u32 *)p_data;
+
+	if (vgpu_vreg(vgpu, offset) & GMBUS_SW_CLR_INT) {
+		if (!(wvalue & GMBUS_SW_CLR_INT)) {
+			vgpu_vreg(vgpu, offset) &= ~GMBUS_SW_CLR_INT;
+			reset_gmbus_controller(vgpu);
+		}
+		/*
+		 * TODO: "This bit is cleared to zero when an event
+		 * causes the HW_RDY bit transition to occur "
+		 */
+	} else {
+		/*
+		 * per bspec setting this bit can cause:
+		 * 1) INT status bit cleared
+		 * 2) HW_RDY bit asserted
+		 */
+		if (wvalue & GMBUS_SW_CLR_INT) {
+			vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_INT;
+			vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY;
+		}
+
+		/* For virtualization, we suppose that HW is always ready,
+		 * so GMBUS_SW_RDY should always be cleared
+		 */
+		if (wvalue & GMBUS_SW_RDY)
+			wvalue &= ~GMBUS_SW_RDY;
+
+		i2c_edid->gmbus.total_byte_count =
+			gmbus1_total_byte_count(wvalue);
+		slave_addr = gmbus1_slave_addr(wvalue);
+
+		/* vgpu gmbus only support EDID */
+		if (slave_addr == EDID_ADDR) {
+			i2c_edid->slave_selected = true;
+		} else if (slave_addr != 0) {
+			gvt_dbg_dpy(
+				"vgpu%d: unsupported gmbus slave addr(0x%x)\n"
+				"	gmbus operations will be ignored.\n",
+					vgpu->id, slave_addr);
+		}
+
+		if (wvalue & GMBUS_CYCLE_INDEX)
+			i2c_edid->current_edid_read =
+				gmbus1_slave_index(wvalue);
+
+		i2c_edid->gmbus.cycle_type = gmbus1_bus_cycle(wvalue);
+		switch (gmbus1_bus_cycle(wvalue)) {
+		case GMBUS_NOCYCLE:
+			break;
+		case GMBUS_STOP:
+			/* From spec:
+			 * This can only cause a STOP to be generated
+			 * if a GMBUS cycle is generated, the GMBUS is
+			 * currently in a data/wait/idle phase, or it is in a
+			 * WAIT phase
+			 */
+			if (gmbus1_bus_cycle(vgpu_vreg(vgpu, offset))
+				!= GMBUS_NOCYCLE) {
+				intel_vgpu_init_i2c_edid(vgpu);
+				/* After the 'stop' cycle, hw state would become
+				 * 'stop phase' and then 'idle phase' after a
+				 * few milliseconds. In emulation, we just set
+				 * it as 'idle phase' ('stop phase' is not
+				 * visible in gmbus interface)
+				 */
+				i2c_edid->gmbus.phase = GMBUS_IDLE_PHASE;
+				vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
+			}
+			break;
+		case NIDX_NS_W:
+		case IDX_NS_W:
+		case NIDX_STOP:
+		case IDX_STOP:
+			/* From hw spec the GMBUS phase
+			 * transition like this:
+			 * START (-->INDEX) -->DATA
+			 */
+			i2c_edid->gmbus.phase = GMBUS_DATA_PHASE;
+			vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_ACTIVE;
+			break;
+		default:
+			gvt_err("Unknown/reserved GMBUS cycle detected!\n");
+			break;
+		}
+		/*
+		 * From hw spec the WAIT state will be
+		 * cleared:
+		 * (1) in a new GMBUS cycle
+		 * (2) by generating a stop
+		 */
+		vgpu_vreg(vgpu, offset) = wvalue;
+	}
+	return 0;
+}
+
+static int gmbus3_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+	void *p_data, unsigned int bytes)
+{
+	WARN_ON(1);
+	return 0;
+}
+
+static int gmbus3_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	int i;
+	unsigned char byte_data;
+	struct intel_vgpu_i2c_edid *i2c_edid = &vgpu->display.i2c_edid;
+	int byte_left = i2c_edid->gmbus.total_byte_count -
+				i2c_edid->current_edid_read;
+	int byte_count = byte_left;
+	u32 reg_data = 0;
+
+	/* Data can only be recevied if previous settings correct */
+	if (vgpu_vreg(vgpu, PCH_GMBUS1) & GMBUS_SLAVE_READ) {
+		if (byte_left <= 0) {
+			memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+			return 0;
+		}
+
+		if (byte_count > 4)
+			byte_count = 4;
+		for (i = 0; i < byte_count; i++) {
+			byte_data = edid_get_byte(vgpu);
+			reg_data |= (byte_data << (i << 3));
+		}
+
+		memcpy(&vgpu_vreg(vgpu, offset), &reg_data, byte_count);
+		memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+
+		if (byte_left <= 4) {
+			switch (i2c_edid->gmbus.cycle_type) {
+			case NIDX_STOP:
+			case IDX_STOP:
+				i2c_edid->gmbus.phase = GMBUS_IDLE_PHASE;
+				break;
+			case NIDX_NS_W:
+			case IDX_NS_W:
+			default:
+				i2c_edid->gmbus.phase = GMBUS_WAIT_PHASE;
+				break;
+			}
+			intel_vgpu_init_i2c_edid(vgpu);
+		}
+		/*
+		 * Read GMBUS3 during send operation,
+		 * return the latest written value
+		 */
+	} else {
+		memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+		gvt_err("vgpu%d: warning: gmbus3 read with nothing returned\n",
+				vgpu->id);
+	}
+	return 0;
+}
+
+static int gmbus2_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	u32 value = vgpu_vreg(vgpu, offset);
+
+	if (!(vgpu_vreg(vgpu, offset) & GMBUS_INUSE))
+		vgpu_vreg(vgpu, offset) |= GMBUS_INUSE;
+	memcpy(p_data, (void *)&value, bytes);
+	return 0;
+}
+
+static int gmbus2_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	u32 wvalue = *(u32 *)p_data;
+
+	if (wvalue & GMBUS_INUSE)
+		vgpu_vreg(vgpu, offset) &= ~GMBUS_INUSE;
+	/* All other bits are read-only */
+	return 0;
+}
+
+/**
+ * intel_gvt_i2c_handle_gmbus_read - emulate gmbus register mmio read
+ * @vgpu: a vGPU
+ *
+ * This function is used to emulate gmbus register mmio read
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_gvt_i2c_handle_gmbus_read(struct intel_vgpu *vgpu,
+	unsigned int offset, void *p_data, unsigned int bytes)
+{
+	if (WARN_ON(bytes > 8 && (offset & (bytes - 1))))
+		return -EINVAL;
+
+	if (offset == i915_mmio_reg_offset(PCH_GMBUS2))
+		return gmbus2_mmio_read(vgpu, offset, p_data, bytes);
+	else if (offset == i915_mmio_reg_offset(PCH_GMBUS3))
+		return gmbus3_mmio_read(vgpu, offset, p_data, bytes);
+
+	memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
+	return 0;
+}
+
+/**
+ * intel_gvt_i2c_handle_gmbus_write - emulate gmbus register mmio write
+ * @vgpu: a vGPU
+ *
+ * This function is used to emulate gmbus register mmio write
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_gvt_i2c_handle_gmbus_write(struct intel_vgpu *vgpu,
+		unsigned int offset, void *p_data, unsigned int bytes)
+{
+	if (WARN_ON(bytes > 8 && (offset & (bytes - 1))))
+		return -EINVAL;
+
+	if (offset == i915_mmio_reg_offset(PCH_GMBUS0))
+		return gmbus0_mmio_write(vgpu, offset, p_data, bytes);
+	else if (offset == i915_mmio_reg_offset(PCH_GMBUS1))
+		return gmbus1_mmio_write(vgpu, offset, p_data, bytes);
+	else if (offset == i915_mmio_reg_offset(PCH_GMBUS2))
+		return gmbus2_mmio_write(vgpu, offset, p_data, bytes);
+	else if (offset == i915_mmio_reg_offset(PCH_GMBUS3))
+		return gmbus3_mmio_write(vgpu, offset, p_data, bytes);
+
+	memcpy(&vgpu_vreg(vgpu, offset), p_data, bytes);
+	return 0;
+}
+
+enum {
+	AUX_CH_CTL = 0,
+	AUX_CH_DATA1,
+	AUX_CH_DATA2,
+	AUX_CH_DATA3,
+	AUX_CH_DATA4,
+	AUX_CH_DATA5
+};
+
+static inline int get_aux_ch_reg(unsigned int offset)
+{
+	int reg;
+
+	switch (offset & 0xff) {
+	case 0x10:
+		reg = AUX_CH_CTL;
+		break;
+	case 0x14:
+		reg = AUX_CH_DATA1;
+		break;
+	case 0x18:
+		reg = AUX_CH_DATA2;
+		break;
+	case 0x1c:
+		reg = AUX_CH_DATA3;
+		break;
+	case 0x20:
+		reg = AUX_CH_DATA4;
+		break;
+	case 0x24:
+		reg = AUX_CH_DATA5;
+		break;
+	default:
+		reg = -1;
+		break;
+	}
+	return reg;
+}
+
+#define AUX_CTL_MSG_LENGTH(reg) \
+	((reg & DP_AUX_CH_CTL_MESSAGE_SIZE_MASK) >> \
+		DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT)
+
+/**
+ * intel_gvt_i2c_handle_aux_ch_write - emulate AUX channel register write
+ * @vgpu: a vGPU
+ *
+ * This function is used to emulate AUX channel register write
+ *
+ */
+void intel_gvt_i2c_handle_aux_ch_write(struct intel_vgpu *vgpu,
+				int port_idx,
+				unsigned int offset,
+				void *p_data)
+{
+	struct intel_vgpu_i2c_edid *i2c_edid = &vgpu->display.i2c_edid;
+	int msg_length, ret_msg_size;
+	int msg, addr, ctrl, op;
+	u32 value = *(u32 *)p_data;
+	int aux_data_for_write = 0;
+	int reg = get_aux_ch_reg(offset);
+
+	if (reg != AUX_CH_CTL) {
+		vgpu_vreg(vgpu, offset) = value;
+		return;
+	}
+
+	msg_length = AUX_CTL_MSG_LENGTH(value);
+	// check the msg in DATA register.
+	msg = vgpu_vreg(vgpu, offset + 4);
+	addr = (msg >> 8) & 0xffff;
+	ctrl = (msg >> 24) & 0xff;
+	op = ctrl >> 4;
+	if (!(value & DP_AUX_CH_CTL_SEND_BUSY)) {
+		/* The ctl write to clear some states */
+		return;
+	}
+
+	/* Always set the wanted value for vms. */
+	ret_msg_size = (((op & 0x1) == GVT_AUX_I2C_READ) ? 2 : 1);
+	vgpu_vreg(vgpu, offset) =
+		DP_AUX_CH_CTL_DONE |
+		((ret_msg_size << DP_AUX_CH_CTL_MESSAGE_SIZE_SHIFT) &
+		DP_AUX_CH_CTL_MESSAGE_SIZE_MASK);
+
+	if (msg_length == 3) {
+		if (!(op & GVT_AUX_I2C_MOT)) {
+			/* stop */
+			intel_vgpu_init_i2c_edid(vgpu);
+		} else {
+			/* start or restart */
+			i2c_edid->aux_ch.i2c_over_aux_ch = true;
+			i2c_edid->aux_ch.aux_ch_mot = true;
+			if (addr == 0) {
+				/* reset the address */
+				intel_vgpu_init_i2c_edid(vgpu);
+			} else if (addr == EDID_ADDR) {
+				i2c_edid->state = I2C_AUX_CH;
+				i2c_edid->port = port_idx;
+				i2c_edid->slave_selected = true;
+				if (intel_vgpu_has_monitor_on_port(vgpu,
+					port_idx) &&
+					intel_vgpu_port_is_dp(vgpu, port_idx))
+					i2c_edid->edid_available = true;
+			}
+		}
+	} else if ((op & 0x1) == GVT_AUX_I2C_WRITE) {
+		/* TODO
+		 * We only support EDID reading from I2C_over_AUX. And
+		 * we do not expect the index mode to be used. Right now
+		 * the WRITE operation is ignored. It is good enough to
+		 * support the gfx driver to do EDID access.
+		 */
+	} else {
+		if (WARN_ON((op & 0x1) != GVT_AUX_I2C_READ))
+			return;
+		if (WARN_ON(msg_length != 4))
+			return;
+		if (i2c_edid->edid_available && i2c_edid->slave_selected) {
+			unsigned char val = edid_get_byte(vgpu);
+
+			aux_data_for_write = (val << 16);
+		}
+	}
+	/* write the return value in AUX_CH_DATA reg which includes:
+	 * ACK of I2C_WRITE
+	 * returned byte if it is READ
+	 */
+
+	aux_data_for_write |= (GVT_AUX_I2C_REPLY_ACK & 0xff) << 24;
+	vgpu_vreg(vgpu, offset + 4) = aux_data_for_write;
+}
+
+/**
+ * intel_vgpu_init_i2c_edid - initialize vGPU i2c edid emulation
+ * @vgpu: a vGPU
+ *
+ * This function is used to initialize vGPU i2c edid emulation stuffs
+ *
+ */
+void intel_vgpu_init_i2c_edid(struct intel_vgpu *vgpu)
+{
+	struct intel_vgpu_i2c_edid *edid = &vgpu->display.i2c_edid;
+
+	edid->state = I2C_NOT_SPECIFIED;
+
+	edid->port = -1;
+	edid->slave_selected = false;
+	edid->edid_available = false;
+	edid->current_edid_read = 0;
+
+	memset(&edid->gmbus, 0, sizeof(struct intel_vgpu_i2c_gmbus));
+
+	edid->aux_ch.i2c_over_aux_ch = false;
+	edid->aux_ch.aux_ch_mot = false;
+}

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

@@ -0,0 +1,150 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Terrence Xu <terrence.xu@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_EDID_H_
+#define _GVT_EDID_H_
+
+#define EDID_SIZE		128
+#define EDID_ADDR		0x50 /* Linux hvm EDID addr */
+
+#define GVT_AUX_NATIVE_WRITE			0x8
+#define GVT_AUX_NATIVE_READ			0x9
+#define GVT_AUX_I2C_WRITE			0x0
+#define GVT_AUX_I2C_READ			0x1
+#define GVT_AUX_I2C_STATUS			0x2
+#define GVT_AUX_I2C_MOT				0x4
+#define GVT_AUX_I2C_REPLY_ACK			(0x0 << 6)
+
+struct intel_vgpu_edid_data {
+	bool data_valid;
+	unsigned char edid_block[EDID_SIZE];
+};
+
+enum gmbus_cycle_type {
+	GMBUS_NOCYCLE	= 0x0,
+	NIDX_NS_W	= 0x1,
+	IDX_NS_W	= 0x3,
+	GMBUS_STOP	= 0x4,
+	NIDX_STOP	= 0x5,
+	IDX_STOP	= 0x7
+};
+
+/*
+ * States of GMBUS
+ *
+ * GMBUS0-3 could be related to the EDID virtualization. Another two GMBUS
+ * registers, GMBUS4 (interrupt mask) and GMBUS5 (2 byte indes register), are
+ * not considered here. Below describes the usage of GMBUS registers that are
+ * cared by the EDID virtualization
+ *
+ * GMBUS0:
+ *      R/W
+ *      port selection. value of bit0 - bit2 corresponds to the GPIO registers.
+ *
+ * GMBUS1:
+ *      R/W Protect
+ *      Command and Status.
+ *      bit0 is the direction bit: 1 is read; 0 is write.
+ *      bit1 - bit7 is slave 7-bit address.
+ *      bit16 - bit24 total byte count (ignore?)
+ *
+ * GMBUS2:
+ *      Most of bits are read only except bit 15 (IN_USE)
+ *      Status register
+ *      bit0 - bit8 current byte count
+ *      bit 11: hardware ready;
+ *
+ * GMBUS3:
+ *      Read/Write
+ *      Data for transfer
+ */
+
+/* From hw specs, Other phases like START, ADDRESS, INDEX
+ * are invisible to GMBUS MMIO interface. So no definitions
+ * in below enum types
+ */
+enum gvt_gmbus_phase {
+	GMBUS_IDLE_PHASE = 0,
+	GMBUS_DATA_PHASE,
+	GMBUS_WAIT_PHASE,
+	//GMBUS_STOP_PHASE,
+	GMBUS_MAX_PHASE
+};
+
+struct intel_vgpu_i2c_gmbus {
+	unsigned int total_byte_count; /* from GMBUS1 */
+	enum gmbus_cycle_type cycle_type;
+	enum gvt_gmbus_phase phase;
+};
+
+struct intel_vgpu_i2c_aux_ch {
+	bool i2c_over_aux_ch;
+	bool aux_ch_mot;
+};
+
+enum i2c_state {
+	I2C_NOT_SPECIFIED = 0,
+	I2C_GMBUS = 1,
+	I2C_AUX_CH = 2
+};
+
+/* I2C sequences cannot interleave.
+ * GMBUS and AUX_CH sequences cannot interleave.
+ */
+struct intel_vgpu_i2c_edid {
+	enum i2c_state state;
+
+	unsigned int port;
+	bool slave_selected;
+	bool edid_available;
+	unsigned int current_edid_read;
+
+	struct intel_vgpu_i2c_gmbus gmbus;
+	struct intel_vgpu_i2c_aux_ch aux_ch;
+};
+
+void intel_vgpu_init_i2c_edid(struct intel_vgpu *vgpu);
+
+int intel_gvt_i2c_handle_gmbus_read(struct intel_vgpu *vgpu,
+		unsigned int offset, void *p_data, unsigned int bytes);
+
+int intel_gvt_i2c_handle_gmbus_write(struct intel_vgpu *vgpu,
+		unsigned int offset, void *p_data, unsigned int bytes);
+
+void intel_gvt_i2c_handle_aux_ch_write(struct intel_vgpu *vgpu,
+		int port_idx,
+		unsigned int offset,
+		void *p_data);
+
+#endif /*_GVT_EDID_H_*/

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

@@ -0,0 +1,852 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+#define _EL_OFFSET_STATUS       0x234
+#define _EL_OFFSET_STATUS_BUF   0x370
+#define _EL_OFFSET_STATUS_PTR   0x3A0
+
+#define execlist_ring_mmio(gvt, ring_id, offset) \
+	(gvt->dev_priv->engine[ring_id].mmio_base + (offset))
+
+#define valid_context(ctx) ((ctx)->valid)
+#define same_context(a, b) (((a)->context_id == (b)->context_id) && \
+		((a)->lrca == (b)->lrca))
+
+static int context_switch_events[] = {
+	[RCS] = RCS_AS_CONTEXT_SWITCH,
+	[BCS] = BCS_AS_CONTEXT_SWITCH,
+	[VCS] = VCS_AS_CONTEXT_SWITCH,
+	[VCS2] = VCS2_AS_CONTEXT_SWITCH,
+	[VECS] = VECS_AS_CONTEXT_SWITCH,
+};
+
+static int ring_id_to_context_switch_event(int ring_id)
+{
+	if (WARN_ON(ring_id < RCS && ring_id >
+				ARRAY_SIZE(context_switch_events)))
+		return -EINVAL;
+
+	return context_switch_events[ring_id];
+}
+
+static void switch_virtual_execlist_slot(struct intel_vgpu_execlist *execlist)
+{
+	gvt_dbg_el("[before] running slot %d/context %x pending slot %d\n",
+			execlist->running_slot ?
+			execlist->running_slot->index : -1,
+			execlist->running_context ?
+			execlist->running_context->context_id : 0,
+			execlist->pending_slot ?
+			execlist->pending_slot->index : -1);
+
+	execlist->running_slot = execlist->pending_slot;
+	execlist->pending_slot = NULL;
+	execlist->running_context = execlist->running_context ?
+		&execlist->running_slot->ctx[0] : NULL;
+
+	gvt_dbg_el("[after] running slot %d/context %x pending slot %d\n",
+			execlist->running_slot ?
+			execlist->running_slot->index : -1,
+			execlist->running_context ?
+			execlist->running_context->context_id : 0,
+			execlist->pending_slot ?
+			execlist->pending_slot->index : -1);
+}
+
+static void emulate_execlist_status(struct intel_vgpu_execlist *execlist)
+{
+	struct intel_vgpu_execlist_slot *running = execlist->running_slot;
+	struct intel_vgpu_execlist_slot *pending = execlist->pending_slot;
+	struct execlist_ctx_descriptor_format *desc = execlist->running_context;
+	struct intel_vgpu *vgpu = execlist->vgpu;
+	struct execlist_status_format status;
+	int ring_id = execlist->ring_id;
+	u32 status_reg = execlist_ring_mmio(vgpu->gvt,
+			ring_id, _EL_OFFSET_STATUS);
+
+	status.ldw = vgpu_vreg(vgpu, status_reg);
+	status.udw = vgpu_vreg(vgpu, status_reg + 4);
+
+	if (running) {
+		status.current_execlist_pointer = !!running->index;
+		status.execlist_write_pointer = !!!running->index;
+		status.execlist_0_active = status.execlist_0_valid =
+			!!!(running->index);
+		status.execlist_1_active = status.execlist_1_valid =
+			!!(running->index);
+	} else {
+		status.context_id = 0;
+		status.execlist_0_active = status.execlist_0_valid = 0;
+		status.execlist_1_active = status.execlist_1_valid = 0;
+	}
+
+	status.context_id = desc ? desc->context_id : 0;
+	status.execlist_queue_full = !!(pending);
+
+	vgpu_vreg(vgpu, status_reg) = status.ldw;
+	vgpu_vreg(vgpu, status_reg + 4) = status.udw;
+
+	gvt_dbg_el("vgpu%d: status reg offset %x ldw %x udw %x\n",
+		vgpu->id, status_reg, status.ldw, status.udw);
+}
+
+static void emulate_csb_update(struct intel_vgpu_execlist *execlist,
+		struct execlist_context_status_format *status,
+		bool trigger_interrupt_later)
+{
+	struct intel_vgpu *vgpu = execlist->vgpu;
+	int ring_id = execlist->ring_id;
+	struct execlist_context_status_pointer_format ctx_status_ptr;
+	u32 write_pointer;
+	u32 ctx_status_ptr_reg, ctx_status_buf_reg, offset;
+
+	ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
+			_EL_OFFSET_STATUS_PTR);
+	ctx_status_buf_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
+			_EL_OFFSET_STATUS_BUF);
+
+	ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);
+
+	write_pointer = ctx_status_ptr.write_ptr;
+
+	if (write_pointer == 0x7)
+		write_pointer = 0;
+	else {
+		++write_pointer;
+		write_pointer %= 0x6;
+	}
+
+	offset = ctx_status_buf_reg + write_pointer * 8;
+
+	vgpu_vreg(vgpu, offset) = status->ldw;
+	vgpu_vreg(vgpu, offset + 4) = status->udw;
+
+	ctx_status_ptr.write_ptr = write_pointer;
+	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
+
+	gvt_dbg_el("vgpu%d: w pointer %u reg %x csb l %x csb h %x\n",
+		vgpu->id, write_pointer, offset, status->ldw, status->udw);
+
+	if (trigger_interrupt_later)
+		return;
+
+	intel_vgpu_trigger_virtual_event(vgpu,
+			ring_id_to_context_switch_event(execlist->ring_id));
+}
+
+static int emulate_execlist_ctx_schedule_out(
+		struct intel_vgpu_execlist *execlist,
+		struct execlist_ctx_descriptor_format *ctx)
+{
+	struct intel_vgpu_execlist_slot *running = execlist->running_slot;
+	struct intel_vgpu_execlist_slot *pending = execlist->pending_slot;
+	struct execlist_ctx_descriptor_format *ctx0 = &running->ctx[0];
+	struct execlist_ctx_descriptor_format *ctx1 = &running->ctx[1];
+	struct execlist_context_status_format status;
+
+	memset(&status, 0, sizeof(status));
+
+	gvt_dbg_el("schedule out context id %x\n", ctx->context_id);
+
+	if (WARN_ON(!same_context(ctx, execlist->running_context))) {
+		gvt_err("schedule out context is not running context,"
+				"ctx id %x running ctx id %x\n",
+				ctx->context_id,
+				execlist->running_context->context_id);
+		return -EINVAL;
+	}
+
+	/* ctx1 is valid, ctx0/ctx is scheduled-out -> element switch */
+	if (valid_context(ctx1) && same_context(ctx0, ctx)) {
+		gvt_dbg_el("ctx 1 valid, ctx/ctx 0 is scheduled-out\n");
+
+		execlist->running_context = ctx1;
+
+		emulate_execlist_status(execlist);
+
+		status.context_complete = status.element_switch = 1;
+		status.context_id = ctx->context_id;
+
+		emulate_csb_update(execlist, &status, false);
+		/*
+		 * ctx1 is not valid, ctx == ctx0
+		 * ctx1 is valid, ctx1 == ctx
+		 *	--> last element is finished
+		 * emulate:
+		 *	active-to-idle if there is *no* pending execlist
+		 *	context-complete if there *is* pending execlist
+		 */
+	} else if ((!valid_context(ctx1) && same_context(ctx0, ctx))
+			|| (valid_context(ctx1) && same_context(ctx1, ctx))) {
+		gvt_dbg_el("need to switch virtual execlist slot\n");
+
+		switch_virtual_execlist_slot(execlist);
+
+		emulate_execlist_status(execlist);
+
+		status.context_complete = status.active_to_idle = 1;
+		status.context_id = ctx->context_id;
+
+		if (!pending) {
+			emulate_csb_update(execlist, &status, false);
+		} else {
+			emulate_csb_update(execlist, &status, true);
+
+			memset(&status, 0, sizeof(status));
+
+			status.idle_to_active = 1;
+			status.context_id = 0;
+
+			emulate_csb_update(execlist, &status, false);
+		}
+	} else {
+		WARN_ON(1);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static struct intel_vgpu_execlist_slot *get_next_execlist_slot(
+		struct intel_vgpu_execlist *execlist)
+{
+	struct intel_vgpu *vgpu = execlist->vgpu;
+	int ring_id = execlist->ring_id;
+	u32 status_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
+			_EL_OFFSET_STATUS);
+	struct execlist_status_format status;
+
+	status.ldw = vgpu_vreg(vgpu, status_reg);
+	status.udw = vgpu_vreg(vgpu, status_reg + 4);
+
+	if (status.execlist_queue_full) {
+		gvt_err("virtual execlist slots are full\n");
+		return NULL;
+	}
+
+	return &execlist->slot[status.execlist_write_pointer];
+}
+
+static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist,
+		struct execlist_ctx_descriptor_format ctx[2])
+{
+	struct intel_vgpu_execlist_slot *running = execlist->running_slot;
+	struct intel_vgpu_execlist_slot *slot =
+		get_next_execlist_slot(execlist);
+
+	struct execlist_ctx_descriptor_format *ctx0, *ctx1;
+	struct execlist_context_status_format status;
+
+	gvt_dbg_el("emulate schedule-in\n");
+
+	if (!slot) {
+		gvt_err("no available execlist slot\n");
+		return -EINVAL;
+	}
+
+	memset(&status, 0, sizeof(status));
+	memset(slot->ctx, 0, sizeof(slot->ctx));
+
+	slot->ctx[0] = ctx[0];
+	slot->ctx[1] = ctx[1];
+
+	gvt_dbg_el("alloc slot index %d ctx 0 %x ctx 1 %x\n",
+			slot->index, ctx[0].context_id,
+			ctx[1].context_id);
+
+	/*
+	 * no running execlist, make this write bundle as running execlist
+	 * -> idle-to-active
+	 */
+	if (!running) {
+		gvt_dbg_el("no current running execlist\n");
+
+		execlist->running_slot = slot;
+		execlist->pending_slot = NULL;
+		execlist->running_context = &slot->ctx[0];
+
+		gvt_dbg_el("running slot index %d running context %x\n",
+				execlist->running_slot->index,
+				execlist->running_context->context_id);
+
+		emulate_execlist_status(execlist);
+
+		status.idle_to_active = 1;
+		status.context_id = 0;
+
+		emulate_csb_update(execlist, &status, false);
+		return 0;
+	}
+
+	ctx0 = &running->ctx[0];
+	ctx1 = &running->ctx[1];
+
+	gvt_dbg_el("current running slot index %d ctx 0 %x ctx 1 %x\n",
+		running->index, ctx0->context_id, ctx1->context_id);
+
+	/*
+	 * already has an running execlist
+	 *	a. running ctx1 is valid,
+	 *	   ctx0 is finished, and running ctx1 == new execlist ctx[0]
+	 *	b. running ctx1 is not valid,
+	 *	   ctx0 == new execlist ctx[0]
+	 * ----> lite-restore + preempted
+	 */
+	if ((valid_context(ctx1) && same_context(ctx1, &slot->ctx[0]) &&
+		/* condition a */
+		(!same_context(ctx0, execlist->running_context))) ||
+			(!valid_context(ctx1) &&
+			 same_context(ctx0, &slot->ctx[0]))) { /* condition b */
+		gvt_dbg_el("need to switch virtual execlist slot\n");
+
+		execlist->pending_slot = slot;
+		switch_virtual_execlist_slot(execlist);
+
+		emulate_execlist_status(execlist);
+
+		status.lite_restore = status.preempted = 1;
+		status.context_id = ctx[0].context_id;
+
+		emulate_csb_update(execlist, &status, false);
+	} else {
+		gvt_dbg_el("emulate as pending slot\n");
+		/*
+		 * otherwise
+		 * --> emulate pending execlist exist + but no preemption case
+		 */
+		execlist->pending_slot = slot;
+		emulate_execlist_status(execlist);
+	}
+	return 0;
+}
+
+static void free_workload(struct intel_vgpu_workload *workload)
+{
+	intel_vgpu_unpin_mm(workload->shadow_mm);
+	intel_gvt_mm_unreference(workload->shadow_mm);
+	kmem_cache_free(workload->vgpu->workloads, workload);
+}
+
+#define get_desc_from_elsp_dwords(ed, i) \
+	((struct execlist_ctx_descriptor_format *)&((ed)->data[i * 2]))
+
+
+#define BATCH_BUFFER_ADDR_MASK ((1UL << 32) - (1U << 2))
+#define BATCH_BUFFER_ADDR_HIGH_MASK ((1UL << 16) - (1U))
+static int set_gma_to_bb_cmd(struct intel_shadow_bb_entry *entry_obj,
+			     unsigned long add, int gmadr_bytes)
+{
+	if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8))
+		return -1;
+
+	*((u32 *)(entry_obj->bb_start_cmd_va + (1 << 2))) = add &
+		BATCH_BUFFER_ADDR_MASK;
+	if (gmadr_bytes == 8) {
+		*((u32 *)(entry_obj->bb_start_cmd_va + (2 << 2))) =
+			add & BATCH_BUFFER_ADDR_HIGH_MASK;
+	}
+
+	return 0;
+}
+
+static void prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
+{
+	int gmadr_bytes = workload->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+	struct i915_vma *vma;
+	unsigned long gma;
+
+	/* pin the gem object to ggtt */
+	if (!list_empty(&workload->shadow_bb)) {
+		struct intel_shadow_bb_entry *entry_obj =
+			list_first_entry(&workload->shadow_bb,
+					 struct intel_shadow_bb_entry,
+					 list);
+		struct intel_shadow_bb_entry *temp;
+
+		list_for_each_entry_safe(entry_obj, temp, &workload->shadow_bb,
+				list) {
+			vma = i915_gem_object_ggtt_pin(entry_obj->obj, NULL, 0,
+					0, 0);
+			if (IS_ERR(vma)) {
+				gvt_err("Cannot pin\n");
+				return;
+			}
+			i915_gem_object_unpin_pages(entry_obj->obj);
+
+			/* update the relocate gma with shadow batch buffer*/
+			gma = i915_gem_object_ggtt_offset(entry_obj->obj, NULL);
+			WARN_ON(!IS_ALIGNED(gma, 4));
+			set_gma_to_bb_cmd(entry_obj, gma, gmadr_bytes);
+		}
+	}
+}
+
+static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+	int ring_id = wa_ctx->workload->ring_id;
+	struct i915_gem_context *shadow_ctx =
+		wa_ctx->workload->vgpu->shadow_ctx;
+	struct drm_i915_gem_object *ctx_obj =
+		shadow_ctx->engine[ring_id].state->obj;
+	struct execlist_ring_context *shadow_ring_context;
+	struct page *page;
+
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
+	shadow_ring_context = kmap_atomic(page);
+
+	shadow_ring_context->bb_per_ctx_ptr.val =
+		(shadow_ring_context->bb_per_ctx_ptr.val &
+		(~PER_CTX_ADDR_MASK)) | wa_ctx->per_ctx.shadow_gma;
+	shadow_ring_context->rcs_indirect_ctx.val =
+		(shadow_ring_context->rcs_indirect_ctx.val &
+		(~INDIRECT_CTX_ADDR_MASK)) | wa_ctx->indirect_ctx.shadow_gma;
+
+	kunmap_atomic(shadow_ring_context);
+	return 0;
+}
+
+static void prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+	struct i915_vma *vma;
+	unsigned long gma;
+	unsigned char *per_ctx_va =
+		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
+		wa_ctx->indirect_ctx.size;
+
+	if (wa_ctx->indirect_ctx.size == 0)
+		return;
+
+	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL, 0, 0, 0);
+	if (IS_ERR(vma)) {
+		gvt_err("Cannot pin indirect ctx obj\n");
+		return;
+	}
+	i915_gem_object_unpin_pages(wa_ctx->indirect_ctx.obj);
+
+	gma = i915_gem_object_ggtt_offset(wa_ctx->indirect_ctx.obj, NULL);
+	WARN_ON(!IS_ALIGNED(gma, CACHELINE_BYTES));
+	wa_ctx->indirect_ctx.shadow_gma = gma;
+
+	wa_ctx->per_ctx.shadow_gma = *((unsigned int *)per_ctx_va + 1);
+	memset(per_ctx_va, 0, CACHELINE_BYTES);
+
+	update_wa_ctx_2_shadow_ctx(wa_ctx);
+}
+
+static int prepare_execlist_workload(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct execlist_ctx_descriptor_format ctx[2];
+	int ring_id = workload->ring_id;
+
+	intel_vgpu_pin_mm(workload->shadow_mm);
+	intel_vgpu_sync_oos_pages(workload->vgpu);
+	intel_vgpu_flush_post_shadow(workload->vgpu);
+	prepare_shadow_batch_buffer(workload);
+	prepare_shadow_wa_ctx(&workload->wa_ctx);
+	if (!workload->emulate_schedule_in)
+		return 0;
+
+	ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1);
+	ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0);
+
+	return emulate_execlist_schedule_in(&vgpu->execlist[ring_id], ctx);
+}
+
+static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
+{
+	/* release all the shadow batch buffer */
+	if (!list_empty(&workload->shadow_bb)) {
+		struct intel_shadow_bb_entry *entry_obj =
+			list_first_entry(&workload->shadow_bb,
+					 struct intel_shadow_bb_entry,
+					 list);
+		struct intel_shadow_bb_entry *temp;
+
+		list_for_each_entry_safe(entry_obj, temp, &workload->shadow_bb,
+					 list) {
+			drm_gem_object_unreference(&(entry_obj->obj->base));
+			kvfree(entry_obj->va);
+			list_del(&entry_obj->list);
+			kfree(entry_obj);
+		}
+	}
+}
+
+static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+	if (wa_ctx->indirect_ctx.size == 0)
+		return;
+
+	drm_gem_object_unreference(&(wa_ctx->indirect_ctx.obj->base));
+	kvfree(wa_ctx->indirect_ctx.shadow_va);
+}
+
+static int complete_execlist_workload(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_vgpu_execlist *execlist =
+		&vgpu->execlist[workload->ring_id];
+	struct intel_vgpu_workload *next_workload;
+	struct list_head *next = workload_q_head(vgpu, workload->ring_id)->next;
+	bool lite_restore = false;
+	int ret;
+
+	gvt_dbg_el("complete workload %p status %d\n", workload,
+			workload->status);
+
+	release_shadow_batch_buffer(workload);
+	release_shadow_wa_ctx(&workload->wa_ctx);
+
+	if (workload->status || vgpu->resetting)
+		goto out;
+
+	if (!list_empty(workload_q_head(vgpu, workload->ring_id))) {
+		struct execlist_ctx_descriptor_format *this_desc, *next_desc;
+
+		next_workload = container_of(next,
+				struct intel_vgpu_workload, list);
+		this_desc = &workload->ctx_desc;
+		next_desc = &next_workload->ctx_desc;
+
+		lite_restore = same_context(this_desc, next_desc);
+	}
+
+	if (lite_restore) {
+		gvt_dbg_el("next context == current - no schedule-out\n");
+		free_workload(workload);
+		return 0;
+	}
+
+	ret = emulate_execlist_ctx_schedule_out(execlist, &workload->ctx_desc);
+	if (ret)
+		goto err;
+out:
+	free_workload(workload);
+	return 0;
+err:
+	free_workload(workload);
+	return ret;
+}
+
+#define RING_CTX_OFF(x) \
+	offsetof(struct execlist_ring_context, x)
+
+static void read_guest_pdps(struct intel_vgpu *vgpu,
+		u64 ring_context_gpa, u32 pdp[8])
+{
+	u64 gpa;
+	int i;
+
+	gpa = ring_context_gpa + RING_CTX_OFF(pdp3_UDW.val);
+
+	for (i = 0; i < 8; i++)
+		intel_gvt_hypervisor_read_gpa(vgpu,
+				gpa + i * 8, &pdp[7 - i], 4);
+}
+
+static int prepare_mm(struct intel_vgpu_workload *workload)
+{
+	struct execlist_ctx_descriptor_format *desc = &workload->ctx_desc;
+	struct intel_vgpu_mm *mm;
+	int page_table_level;
+	u32 pdp[8];
+
+	if (desc->addressing_mode == 1) { /* legacy 32-bit */
+		page_table_level = 3;
+	} else if (desc->addressing_mode == 3) { /* legacy 64 bit */
+		page_table_level = 4;
+	} else {
+		gvt_err("Advanced Context mode(SVM) is not supported!\n");
+		return -EINVAL;
+	}
+
+	read_guest_pdps(workload->vgpu, workload->ring_context_gpa, pdp);
+
+	mm = intel_vgpu_find_ppgtt_mm(workload->vgpu, page_table_level, pdp);
+	if (mm) {
+		intel_gvt_mm_reference(mm);
+	} else {
+
+		mm = intel_vgpu_create_mm(workload->vgpu, INTEL_GVT_MM_PPGTT,
+				pdp, page_table_level, 0);
+		if (IS_ERR(mm)) {
+			gvt_err("fail to create mm object.\n");
+			return PTR_ERR(mm);
+		}
+	}
+	workload->shadow_mm = mm;
+	return 0;
+}
+
+#define get_last_workload(q) \
+	(list_empty(q) ? NULL : container_of(q->prev, \
+	struct intel_vgpu_workload, list))
+
+bool submit_context(struct intel_vgpu *vgpu, int ring_id,
+		struct execlist_ctx_descriptor_format *desc,
+		bool emulate_schedule_in)
+{
+	struct list_head *q = workload_q_head(vgpu, ring_id);
+	struct intel_vgpu_workload *last_workload = get_last_workload(q);
+	struct intel_vgpu_workload *workload = NULL;
+	u64 ring_context_gpa;
+	u32 head, tail, start, ctl, ctx_ctl, per_ctx, indirect_ctx;
+	int ret;
+
+	ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+			(u32)((desc->lrca + 1) << GTT_PAGE_SHIFT));
+	if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) {
+		gvt_err("invalid guest context LRCA: %x\n", desc->lrca);
+		return -EINVAL;
+	}
+
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(ring_header.val), &head, 4);
+
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(ring_tail.val), &tail, 4);
+
+	head &= RB_HEAD_OFF_MASK;
+	tail &= RB_TAIL_OFF_MASK;
+
+	if (last_workload && same_context(&last_workload->ctx_desc, desc)) {
+		gvt_dbg_el("ring id %d cur workload == last\n", ring_id);
+		gvt_dbg_el("ctx head %x real head %lx\n", head,
+				last_workload->rb_tail);
+		/*
+		 * cannot use guest context head pointer here,
+		 * as it might not be updated at this time
+		 */
+		head = last_workload->rb_tail;
+	}
+
+	gvt_dbg_el("ring id %d begin a new workload\n", ring_id);
+
+	workload = kmem_cache_zalloc(vgpu->workloads, GFP_KERNEL);
+	if (!workload)
+		return -ENOMEM;
+
+	/* record some ring buffer register values for scan and shadow */
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(rb_start.val), &start, 4);
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(rb_ctrl.val), &ctl, 4);
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4);
+
+	INIT_LIST_HEAD(&workload->list);
+	INIT_LIST_HEAD(&workload->shadow_bb);
+
+	init_waitqueue_head(&workload->shadow_ctx_status_wq);
+	atomic_set(&workload->shadow_ctx_active, 0);
+
+	workload->vgpu = vgpu;
+	workload->ring_id = ring_id;
+	workload->ctx_desc = *desc;
+	workload->ring_context_gpa = ring_context_gpa;
+	workload->rb_head = head;
+	workload->rb_tail = tail;
+	workload->rb_start = start;
+	workload->rb_ctl = ctl;
+	workload->prepare = prepare_execlist_workload;
+	workload->complete = complete_execlist_workload;
+	workload->status = -EINPROGRESS;
+	workload->emulate_schedule_in = emulate_schedule_in;
+
+	if (ring_id == RCS) {
+		intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(bb_per_ctx_ptr.val), &per_ctx, 4);
+		intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(rcs_indirect_ctx.val), &indirect_ctx, 4);
+
+		workload->wa_ctx.indirect_ctx.guest_gma =
+			indirect_ctx & INDIRECT_CTX_ADDR_MASK;
+		workload->wa_ctx.indirect_ctx.size =
+			(indirect_ctx & INDIRECT_CTX_SIZE_MASK) *
+			CACHELINE_BYTES;
+		workload->wa_ctx.per_ctx.guest_gma =
+			per_ctx & PER_CTX_ADDR_MASK;
+		workload->wa_ctx.workload = workload;
+
+		WARN_ON(workload->wa_ctx.indirect_ctx.size && !(per_ctx & 0x1));
+	}
+
+	if (emulate_schedule_in)
+		memcpy(&workload->elsp_dwords,
+				&vgpu->execlist[ring_id].elsp_dwords,
+				sizeof(workload->elsp_dwords));
+
+	gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n",
+			workload, ring_id, head, tail, start, ctl);
+
+	gvt_dbg_el("workload %p emulate schedule_in %d\n", workload,
+			emulate_schedule_in);
+
+	ret = prepare_mm(workload);
+	if (ret) {
+		kmem_cache_free(vgpu->workloads, workload);
+		return ret;
+	}
+
+	queue_workload(workload);
+	return 0;
+}
+
+int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id)
+{
+	struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+	struct execlist_ctx_descriptor_format *desc[2], valid_desc[2];
+	unsigned long valid_desc_bitmap = 0;
+	bool emulate_schedule_in = true;
+	int ret;
+	int i;
+
+	memset(valid_desc, 0, sizeof(valid_desc));
+
+	desc[0] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1);
+	desc[1] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0);
+
+	for (i = 0; i < 2; i++) {
+		if (!desc[i]->valid)
+			continue;
+
+		if (!desc[i]->privilege_access) {
+			gvt_err("vgpu%d: unexpected GGTT elsp submission\n",
+					vgpu->id);
+			return -EINVAL;
+		}
+
+		/* TODO: add another guest context checks here. */
+		set_bit(i, &valid_desc_bitmap);
+		valid_desc[i] = *desc[i];
+	}
+
+	if (!valid_desc_bitmap) {
+		gvt_err("vgpu%d: no valid desc in a elsp submission\n",
+				vgpu->id);
+		return -EINVAL;
+	}
+
+	if (!test_bit(0, (void *)&valid_desc_bitmap) &&
+			test_bit(1, (void *)&valid_desc_bitmap)) {
+		gvt_err("vgpu%d: weird elsp submission, desc 0 is not valid\n",
+				vgpu->id);
+		return -EINVAL;
+	}
+
+	/* submit workload */
+	for_each_set_bit(i, (void *)&valid_desc_bitmap, 2) {
+		ret = submit_context(vgpu, ring_id, &valid_desc[i],
+				emulate_schedule_in);
+		if (ret) {
+			gvt_err("vgpu%d: fail to schedule workload\n",
+					vgpu->id);
+			return ret;
+		}
+		emulate_schedule_in = false;
+	}
+	return 0;
+}
+
+static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id)
+{
+	struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+	struct execlist_context_status_pointer_format ctx_status_ptr;
+	u32 ctx_status_ptr_reg;
+
+	memset(execlist, 0, sizeof(*execlist));
+
+	execlist->vgpu = vgpu;
+	execlist->ring_id = ring_id;
+	execlist->slot[0].index = 0;
+	execlist->slot[1].index = 1;
+
+	ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
+			_EL_OFFSET_STATUS_PTR);
+
+	ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);
+	ctx_status_ptr.read_ptr = ctx_status_ptr.write_ptr = 0x7;
+	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
+}
+
+void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu)
+{
+	kmem_cache_destroy(vgpu->workloads);
+}
+
+int intel_vgpu_init_execlist(struct intel_vgpu *vgpu)
+{
+	int i;
+
+	/* each ring has a virtual execlist engine */
+	for (i = 0; i < I915_NUM_ENGINES; i++) {
+		init_vgpu_execlist(vgpu, i);
+		INIT_LIST_HEAD(&vgpu->workload_q_head[i]);
+	}
+
+	vgpu->workloads = kmem_cache_create("gvt-g vgpu workload",
+			sizeof(struct intel_vgpu_workload), 0,
+			SLAB_HWCACHE_ALIGN,
+			NULL);
+
+	if (!vgpu->workloads)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu,
+		unsigned long ring_bitmap)
+{
+	int bit;
+	struct list_head *pos, *n;
+	struct intel_vgpu_workload *workload = NULL;
+
+	for_each_set_bit(bit, &ring_bitmap, sizeof(ring_bitmap) * 8) {
+		if (bit >= I915_NUM_ENGINES)
+			break;
+		/* free the unsubmited workload in the queue */
+		list_for_each_safe(pos, n, &vgpu->workload_q_head[bit]) {
+			workload = container_of(pos,
+					struct intel_vgpu_workload, list);
+			list_del_init(&workload->list);
+			free_workload(workload);
+		}
+
+		init_vgpu_execlist(vgpu, bit);
+	}
+}

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

@@ -0,0 +1,188 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#ifndef _GVT_EXECLIST_H_
+#define _GVT_EXECLIST_H_
+
+struct execlist_ctx_descriptor_format {
+	union {
+		u32 udw;
+		u32 context_id;
+	};
+	union {
+		u32 ldw;
+		struct {
+			u32 valid                  : 1;
+			u32 force_pd_restore       : 1;
+			u32 force_restore          : 1;
+			u32 addressing_mode        : 2;
+			u32 llc_coherency          : 1;
+			u32 fault_handling         : 2;
+			u32 privilege_access       : 1;
+			u32 reserved               : 3;
+			u32 lrca                   : 20;
+		};
+	};
+};
+
+struct execlist_status_format {
+	union {
+		u32 ldw;
+		struct {
+			u32 current_execlist_pointer       :1;
+			u32 execlist_write_pointer         :1;
+			u32 execlist_queue_full            :1;
+			u32 execlist_1_valid               :1;
+			u32 execlist_0_valid               :1;
+			u32 last_ctx_switch_reason         :9;
+			u32 current_active_elm_status      :2;
+			u32 arbitration_enable             :1;
+			u32 execlist_1_active              :1;
+			u32 execlist_0_active              :1;
+			u32 reserved                       :13;
+		};
+	};
+	union {
+		u32 udw;
+		u32 context_id;
+	};
+};
+
+struct execlist_context_status_pointer_format {
+	union {
+		u32 dw;
+		struct {
+			u32 write_ptr              :3;
+			u32 reserved               :5;
+			u32 read_ptr               :3;
+			u32 reserved2              :5;
+			u32 mask                   :16;
+		};
+	};
+};
+
+struct execlist_context_status_format {
+	union {
+		u32 ldw;
+		struct {
+			u32 idle_to_active         :1;
+			u32 preempted              :1;
+			u32 element_switch         :1;
+			u32 active_to_idle         :1;
+			u32 context_complete       :1;
+			u32 wait_on_sync_flip      :1;
+			u32 wait_on_vblank         :1;
+			u32 wait_on_semaphore      :1;
+			u32 wait_on_scanline       :1;
+			u32 reserved               :2;
+			u32 semaphore_wait_mode    :1;
+			u32 display_plane          :3;
+			u32 lite_restore           :1;
+			u32 reserved_2             :16;
+		};
+	};
+	union {
+		u32 udw;
+		u32 context_id;
+	};
+};
+
+struct execlist_mmio_pair {
+	u32 addr;
+	u32 val;
+};
+
+/* The first 52 dwords in register state context */
+struct execlist_ring_context {
+	u32 nop1;
+	u32 lri_cmd_1;
+	struct execlist_mmio_pair ctx_ctrl;
+	struct execlist_mmio_pair ring_header;
+	struct execlist_mmio_pair ring_tail;
+	struct execlist_mmio_pair rb_start;
+	struct execlist_mmio_pair rb_ctrl;
+	struct execlist_mmio_pair bb_cur_head_UDW;
+	struct execlist_mmio_pair bb_cur_head_LDW;
+	struct execlist_mmio_pair bb_state;
+	struct execlist_mmio_pair second_bb_addr_UDW;
+	struct execlist_mmio_pair second_bb_addr_LDW;
+	struct execlist_mmio_pair second_bb_state;
+	struct execlist_mmio_pair bb_per_ctx_ptr;
+	struct execlist_mmio_pair rcs_indirect_ctx;
+	struct execlist_mmio_pair rcs_indirect_ctx_offset;
+	u32 nop2;
+	u32 nop3;
+	u32 nop4;
+	u32 lri_cmd_2;
+	struct execlist_mmio_pair ctx_timestamp;
+	struct execlist_mmio_pair pdp3_UDW;
+	struct execlist_mmio_pair pdp3_LDW;
+	struct execlist_mmio_pair pdp2_UDW;
+	struct execlist_mmio_pair pdp2_LDW;
+	struct execlist_mmio_pair pdp1_UDW;
+	struct execlist_mmio_pair pdp1_LDW;
+	struct execlist_mmio_pair pdp0_UDW;
+	struct execlist_mmio_pair pdp0_LDW;
+};
+
+struct intel_vgpu_elsp_dwords {
+	u32 data[4];
+	u32 index;
+};
+
+struct intel_vgpu_execlist_slot {
+	struct execlist_ctx_descriptor_format ctx[2];
+	u32 index;
+};
+
+struct intel_vgpu_execlist {
+	struct intel_vgpu_execlist_slot slot[2];
+	struct intel_vgpu_execlist_slot *running_slot;
+	struct intel_vgpu_execlist_slot *pending_slot;
+	struct execlist_ctx_descriptor_format *running_context;
+	int ring_id;
+	struct intel_vgpu *vgpu;
+	struct intel_vgpu_elsp_dwords elsp_dwords;
+};
+
+void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu);
+
+int intel_vgpu_init_execlist(struct intel_vgpu *vgpu);
+
+int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id);
+
+void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu,
+		unsigned long ring_bitmap);
+
+#endif /*_GVT_EXECLIST_H_*/

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

@@ -0,0 +1,308 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Changbin Du <changbin.du@intel.com>
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/crc32.h>
+
+#include "i915_drv.h"
+
+#define FIRMWARE_VERSION (0x0)
+
+struct gvt_firmware_header {
+	u64 magic;
+	u32 crc32;		/* protect the data after this field */
+	u32 version;
+	u64 cfg_space_size;
+	u64 cfg_space_offset;	/* offset in the file */
+	u64 mmio_size;
+	u64 mmio_offset;	/* offset in the file */
+	unsigned char data[1];
+};
+
+#define RD(offset) (readl(mmio + offset.reg))
+#define WR(v, offset) (writel(v, mmio + offset.reg))
+
+static void bdw_forcewake_get(void *mmio)
+{
+	WR(_MASKED_BIT_DISABLE(0xffff), FORCEWAKE_MT);
+
+	RD(ECOBUS);
+
+	if (wait_for((RD(FORCEWAKE_ACK_HSW) & FORCEWAKE_KERNEL) == 0, 50))
+		gvt_err("fail to wait forcewake idle\n");
+
+	WR(_MASKED_BIT_ENABLE(FORCEWAKE_KERNEL), FORCEWAKE_MT);
+
+	if (wait_for((RD(FORCEWAKE_ACK_HSW) & FORCEWAKE_KERNEL), 50))
+		gvt_err("fail to wait forcewake ack\n");
+
+	if (wait_for((RD(GEN6_GT_THREAD_STATUS_REG) &
+		      GEN6_GT_THREAD_STATUS_CORE_MASK) == 0, 50))
+		gvt_err("fail to wait c0 wake up\n");
+}
+
+#undef RD
+#undef WR
+
+#define dev_to_drm_minor(d) dev_get_drvdata((d))
+
+static ssize_t
+gvt_firmware_read(struct file *filp, struct kobject *kobj,
+	     struct bin_attribute *attr, char *buf,
+	     loff_t offset, size_t count)
+{
+	memcpy(buf, attr->private + offset, count);
+	return count;
+}
+
+static struct bin_attribute firmware_attr = {
+	.attr = {.name = "gvt_firmware", .mode = (S_IRUSR)},
+	.read = gvt_firmware_read,
+	.write = NULL,
+	.mmap = NULL,
+};
+
+static int expose_firmware_sysfs(struct intel_gvt *gvt, void *mmio)
+{
+	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_firmware_header *h;
+	void *firmware;
+	void *p;
+	unsigned long size;
+	int i;
+	int ret;
+
+	size = sizeof(*h) + info->mmio_size + info->cfg_space_size - 1;
+	firmware = vmalloc(size);
+	if (!firmware)
+		return -ENOMEM;
+
+	h = firmware;
+
+	h->magic = VGT_MAGIC;
+	h->version = FIRMWARE_VERSION;
+	h->cfg_space_size = info->cfg_space_size;
+	h->cfg_space_offset = offsetof(struct gvt_firmware_header, data);
+	h->mmio_size = info->mmio_size;
+	h->mmio_offset = h->cfg_space_offset + h->cfg_space_size;
+
+	p = firmware + h->cfg_space_offset;
+
+	for (i = 0; i < h->cfg_space_size; i += 4)
+		pci_read_config_dword(pdev, i, p + i);
+
+	memcpy(gvt->firmware.cfg_space, p, info->cfg_space_size);
+
+	p = firmware + h->mmio_offset;
+
+	hash_for_each(gvt->mmio.mmio_info_table, i, e, node) {
+		int j;
+
+		for (j = 0; j < e->length; j += 4)
+			*(u32 *)(p + e->offset + j) =
+				readl(mmio + e->offset + j);
+	}
+
+	memcpy(gvt->firmware.mmio, p, info->mmio_size);
+
+	firmware_attr.size = size;
+	firmware_attr.private = firmware;
+
+	ret = device_create_bin_file(&pdev->dev, &firmware_attr);
+	if (ret) {
+		vfree(firmware);
+		return ret;
+	}
+	return 0;
+}
+
+static void clean_firmware_sysfs(struct intel_gvt *gvt)
+{
+	struct pci_dev *pdev = gvt->dev_priv->drm.pdev;
+
+	device_remove_bin_file(&pdev->dev, &firmware_attr);
+	vfree(firmware_attr.private);
+}
+
+/**
+ * intel_gvt_free_firmware - free GVT firmware
+ * @gvt: intel gvt device
+ *
+ */
+void intel_gvt_free_firmware(struct intel_gvt *gvt)
+{
+	if (!gvt->firmware.firmware_loaded)
+		clean_firmware_sysfs(gvt);
+
+	kfree(gvt->firmware.cfg_space);
+	kfree(gvt->firmware.mmio);
+}
+
+static int verify_firmware(struct intel_gvt *gvt,
+			   const struct firmware *fw)
+{
+	struct intel_gvt_device_info *info = &gvt->device_info;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct pci_dev *pdev = dev_priv->drm.pdev;
+	struct gvt_firmware_header *h;
+	unsigned long id, crc32_start;
+	const void *mem;
+	const char *item;
+	u64 file, request;
+
+	h = (struct gvt_firmware_header *)fw->data;
+
+	crc32_start = offsetof(struct gvt_firmware_header, crc32) + 4;
+	mem = fw->data + crc32_start;
+
+#define VERIFY(s, a, b) do { \
+	item = (s); file = (u64)(a); request = (u64)(b); \
+	if ((a) != (b)) \
+		goto invalid_firmware; \
+} while (0)
+
+	VERIFY("magic number", h->magic, VGT_MAGIC);
+	VERIFY("version", h->version, FIRMWARE_VERSION);
+	VERIFY("crc32", h->crc32, crc32_le(0, mem, fw->size - crc32_start));
+	VERIFY("cfg space size", h->cfg_space_size, info->cfg_space_size);
+	VERIFY("mmio size", h->mmio_size, info->mmio_size);
+
+	mem = (fw->data + h->cfg_space_offset);
+
+	id = *(u16 *)(mem + PCI_VENDOR_ID);
+	VERIFY("vender id", id, pdev->vendor);
+
+	id = *(u16 *)(mem + PCI_DEVICE_ID);
+	VERIFY("device id", id, pdev->device);
+
+	id = *(u8 *)(mem + PCI_REVISION_ID);
+	VERIFY("revision id", id, pdev->revision);
+
+#undef VERIFY
+	return 0;
+
+invalid_firmware:
+	gvt_dbg_core("Invalid firmware: %s [file] 0x%llx [request] 0x%llx\n",
+		     item, file, request);
+	return -EINVAL;
+}
+
+#define GVT_FIRMWARE_PATH "i915/gvt"
+
+/**
+ * intel_gvt_load_firmware - load GVT firmware
+ * @gvt: intel gvt device
+ *
+ */
+int intel_gvt_load_firmware(struct intel_gvt *gvt)
+{
+	struct intel_gvt_device_info *info = &gvt->device_info;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct pci_dev *pdev = dev_priv->drm.pdev;
+	struct intel_gvt_firmware *firmware = &gvt->firmware;
+	struct gvt_firmware_header *h;
+	const struct firmware *fw;
+	char *path;
+	void *mmio, *mem;
+	int ret;
+
+	path = kmalloc(PATH_MAX, GFP_KERNEL);
+	if (!path)
+		return -ENOMEM;
+
+	mem = kmalloc(info->cfg_space_size, GFP_KERNEL);
+	if (!mem) {
+		kfree(path);
+		return -ENOMEM;
+	}
+
+	firmware->cfg_space = mem;
+
+	mem = kmalloc(info->mmio_size, GFP_KERNEL);
+	if (!mem) {
+		kfree(path);
+		kfree(firmware->cfg_space);
+		return -ENOMEM;
+	}
+
+	firmware->mmio = mem;
+
+	mmio = pci_iomap(pdev, info->mmio_bar, info->mmio_size);
+	if (!mmio) {
+		kfree(path);
+		kfree(firmware->cfg_space);
+		kfree(firmware->mmio);
+		return -EINVAL;
+	}
+
+	if (IS_BROADWELL(gvt->dev_priv) || IS_SKYLAKE(gvt->dev_priv))
+		bdw_forcewake_get(mmio);
+
+	sprintf(path, "%s/vid_0x%04x_did_0x%04x_rid_0x%04x.golden_hw_state",
+		 GVT_FIRMWARE_PATH, pdev->vendor, pdev->device,
+		 pdev->revision);
+
+	gvt_dbg_core("request hw state firmware %s...\n", path);
+
+	ret = request_firmware(&fw, path, &dev_priv->drm.pdev->dev);
+	kfree(path);
+
+	if (ret)
+		goto expose_firmware;
+
+	gvt_dbg_core("success.\n");
+
+	ret = verify_firmware(gvt, fw);
+	if (ret)
+		goto out_free_fw;
+
+	gvt_dbg_core("verified.\n");
+
+	h = (struct gvt_firmware_header *)fw->data;
+
+	memcpy(firmware->cfg_space, fw->data + h->cfg_space_offset,
+	       h->cfg_space_size);
+	memcpy(firmware->mmio, fw->data + h->mmio_offset,
+	       h->mmio_size);
+
+	release_firmware(fw);
+	firmware->firmware_loaded = true;
+	pci_iounmap(pdev, mmio);
+	return 0;
+
+out_free_fw:
+	release_firmware(fw);
+expose_firmware:
+	expose_firmware_sysfs(gvt, mmio);
+	pci_iounmap(pdev, mmio);
+	return 0;
+}

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

@@ -0,0 +1,270 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *    Xiao Zheng <xiao.zheng@intel.com>
+ *
+ * Contributors:
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#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))
+
+struct intel_vgpu_mm;
+
+#define INTEL_GVT_GTT_HASH_BITS 8
+#define INTEL_GVT_INVALID_ADDR (~0UL)
+
+struct intel_gvt_gtt_entry {
+	u64 val64;
+	int type;
+};
+
+struct intel_gvt_gtt_pte_ops {
+	struct intel_gvt_gtt_entry *(*get_entry)(void *pt,
+		struct intel_gvt_gtt_entry *e,
+		unsigned long index, bool hypervisor_access, unsigned long gpa,
+		struct intel_vgpu *vgpu);
+	struct intel_gvt_gtt_entry *(*set_entry)(void *pt,
+		struct intel_gvt_gtt_entry *e,
+		unsigned long index, bool hypervisor_access, unsigned long gpa,
+		struct intel_vgpu *vgpu);
+	bool (*test_present)(struct intel_gvt_gtt_entry *e);
+	void (*clear_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);
+};
+
+struct intel_gvt_gtt_gma_ops {
+	unsigned long (*gma_to_ggtt_pte_index)(unsigned long gma);
+	unsigned long (*gma_to_pte_index)(unsigned long gma);
+	unsigned long (*gma_to_pde_index)(unsigned long gma);
+	unsigned long (*gma_to_l3_pdp_index)(unsigned long gma);
+	unsigned long (*gma_to_l4_pdp_index)(unsigned long gma);
+	unsigned long (*gma_to_pml4_index)(unsigned long gma);
+};
+
+struct intel_gvt_gtt {
+	struct intel_gvt_gtt_pte_ops *pte_ops;
+	struct intel_gvt_gtt_gma_ops *gma_ops;
+	int (*mm_alloc_page_table)(struct intel_vgpu_mm *mm);
+	void (*mm_free_page_table)(struct intel_vgpu_mm *mm);
+	struct list_head oos_page_use_list_head;
+	struct list_head oos_page_free_list_head;
+	struct list_head mm_lru_list_head;
+};
+
+enum {
+	INTEL_GVT_MM_GGTT = 0,
+	INTEL_GVT_MM_PPGTT,
+};
+
+struct intel_vgpu_mm {
+	int type;
+	bool initialized;
+	bool shadowed;
+
+	int page_table_entry_type;
+	u32 page_table_entry_size;
+	u32 page_table_entry_cnt;
+	void *virtual_page_table;
+	void *shadow_page_table;
+
+	int page_table_level;
+	bool has_shadow_page_table;
+	u32 pde_base_index;
+
+	struct list_head list;
+	struct kref ref;
+	atomic_t pincount;
+	struct list_head lru_list;
+	struct intel_vgpu *vgpu;
+};
+
+extern struct intel_gvt_gtt_entry *intel_vgpu_mm_get_entry(
+		struct intel_vgpu_mm *mm,
+		void *page_table, struct intel_gvt_gtt_entry *e,
+		unsigned long index);
+
+extern struct intel_gvt_gtt_entry *intel_vgpu_mm_set_entry(
+		struct intel_vgpu_mm *mm,
+		void *page_table, struct intel_gvt_gtt_entry *e,
+		unsigned long index);
+
+#define ggtt_get_guest_entry(mm, e, index) \
+	intel_vgpu_mm_get_entry(mm, mm->virtual_page_table, e, index)
+
+#define ggtt_set_guest_entry(mm, e, index) \
+	intel_vgpu_mm_set_entry(mm, mm->virtual_page_table, e, index)
+
+#define ggtt_get_shadow_entry(mm, e, index) \
+	intel_vgpu_mm_get_entry(mm, mm->shadow_page_table, e, index)
+
+#define ggtt_set_shadow_entry(mm, e, index) \
+	intel_vgpu_mm_set_entry(mm, mm->shadow_page_table, e, index)
+
+#define ppgtt_get_guest_root_entry(mm, e, index) \
+	intel_vgpu_mm_get_entry(mm, mm->virtual_page_table, e, index)
+
+#define ppgtt_set_guest_root_entry(mm, e, index) \
+	intel_vgpu_mm_set_entry(mm, mm->virtual_page_table, e, index)
+
+#define ppgtt_get_shadow_root_entry(mm, e, index) \
+	intel_vgpu_mm_get_entry(mm, mm->shadow_page_table, e, index)
+
+#define ppgtt_set_shadow_root_entry(mm, e, index) \
+	intel_vgpu_mm_set_entry(mm, mm->shadow_page_table, e, index)
+
+extern struct intel_vgpu_mm *intel_vgpu_create_mm(struct intel_vgpu *vgpu,
+		int mm_type, void *virtual_page_table, int page_table_level,
+		u32 pde_base_index);
+extern void intel_vgpu_destroy_mm(struct kref *mm_ref);
+
+struct intel_vgpu_guest_page;
+
+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;
+	struct list_head oos_page_list_head;
+	struct list_head post_shadow_list_head;
+	struct page *scratch_page;
+	unsigned long scratch_page_mfn;
+};
+
+extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu);
+extern void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu);
+
+extern int intel_gvt_init_gtt(struct intel_gvt *gvt);
+extern void intel_gvt_clean_gtt(struct intel_gvt *gvt);
+
+extern struct intel_vgpu_mm *intel_gvt_find_ppgtt_mm(struct intel_vgpu *vgpu,
+		int page_table_level, void *root_entry);
+
+struct intel_vgpu_oos_page;
+
+struct intel_vgpu_shadow_page {
+	void *vaddr;
+	struct page *page;
+	int type;
+	struct hlist_node node;
+	unsigned long mfn;
+};
+
+struct intel_vgpu_guest_page {
+	struct hlist_node node;
+	bool writeprotection;
+	unsigned long gfn;
+	int (*handler)(void *, u64, void *, int);
+	void *data;
+	unsigned long write_cnt;
+	struct intel_vgpu_oos_page *oos_page;
+};
+
+struct intel_vgpu_oos_page {
+	struct intel_vgpu_guest_page *guest_page;
+	struct list_head list;
+	struct list_head vm_list;
+	int id;
+	unsigned char mem[GTT_PAGE_SIZE];
+};
+
+#define GTT_ENTRY_NUM_IN_ONE_PAGE 512
+
+struct intel_vgpu_ppgtt_spt {
+	struct intel_vgpu_shadow_page shadow_page;
+	struct intel_vgpu_guest_page guest_page;
+	int guest_page_type;
+	atomic_t refcount;
+	struct intel_vgpu *vgpu;
+	DECLARE_BITMAP(post_shadow_bitmap, GTT_ENTRY_NUM_IN_ONE_PAGE);
+	struct list_head post_shadow_list;
+};
+
+int intel_vgpu_init_guest_page(struct intel_vgpu *vgpu,
+		struct intel_vgpu_guest_page *guest_page,
+		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);
+
+struct intel_vgpu_guest_page *intel_vgpu_find_guest_page(
+		struct intel_vgpu *vgpu, unsigned long gfn);
+
+int intel_vgpu_sync_oos_pages(struct intel_vgpu *vgpu);
+
+int intel_vgpu_flush_post_shadow(struct intel_vgpu *vgpu);
+
+static inline void intel_gvt_mm_reference(struct intel_vgpu_mm *mm)
+{
+	kref_get(&mm->ref);
+}
+
+static inline void intel_gvt_mm_unreference(struct intel_vgpu_mm *mm)
+{
+	kref_put(&mm->ref, intel_vgpu_destroy_mm);
+}
+
+int intel_vgpu_pin_mm(struct intel_vgpu_mm *mm);
+
+void intel_vgpu_unpin_mm(struct intel_vgpu_mm *mm);
+
+unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm,
+		unsigned long gma);
+
+struct intel_vgpu_mm *intel_vgpu_find_ppgtt_mm(struct intel_vgpu *vgpu,
+		int page_table_level, void *root_entry);
+
+int intel_vgpu_g2v_create_ppgtt_mm(struct intel_vgpu *vgpu,
+		int page_table_level);
+
+int intel_vgpu_g2v_destroy_ppgtt_mm(struct intel_vgpu *vgpu,
+		int page_table_level);
+
+int intel_vgpu_emulate_gtt_mmio_read(struct intel_vgpu *vgpu,
+	unsigned int off, void *p_data, unsigned int bytes);
+
+int intel_vgpu_emulate_gtt_mmio_write(struct intel_vgpu *vgpu,
+	unsigned int off, void *p_data, unsigned int bytes);
+
+#endif /* _GVT_GTT_H_ */

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

@@ -19,10 +19,20 @@
  * 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.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Eddie Dong <eddie.dong@intel.com>
+ *
+ * Contributors:
+ *    Niu Bing <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
  */
 
 #include <linux/types.h>
 #include <xen/xen.h>
+#include <linux/kthread.h>
 
 #include "i915_drv.h"
 
@@ -33,6 +43,13 @@ static const char * const supported_hypervisors[] = {
 	[INTEL_GVT_HYPERVISOR_KVM] = "KVM",
 };
 
+struct intel_gvt_io_emulation_ops intel_gvt_io_emulation_ops = {
+	.emulate_cfg_read = intel_vgpu_emulate_cfg_read,
+	.emulate_cfg_write = intel_vgpu_emulate_cfg_write,
+	.emulate_mmio_read = intel_vgpu_emulate_mmio_read,
+	.emulate_mmio_write = intel_vgpu_emulate_mmio_write,
+};
+
 /**
  * intel_gvt_init_host - Load MPT modules and detect if we're running in host
  * @gvt: intel gvt device
@@ -84,9 +101,66 @@ int intel_gvt_init_host(void)
 
 static void init_device_info(struct intel_gvt *gvt)
 {
-	if (IS_BROADWELL(gvt->dev_priv))
-		gvt->device_info.max_support_vgpus = 8;
-	/* This function will grow large in GVT device model patches. */
+	struct intel_gvt_device_info *info = &gvt->device_info;
+
+	if (IS_BROADWELL(gvt->dev_priv) || IS_SKYLAKE(gvt->dev_priv)) {
+		info->max_support_vgpus = 8;
+		info->cfg_space_size = 256;
+		info->mmio_size = 2 * 1024 * 1024;
+		info->mmio_bar = 0;
+		info->msi_cap_offset = IS_SKYLAKE(gvt->dev_priv) ? 0xac : 0x90;
+		info->gtt_start_offset = 8 * 1024 * 1024;
+		info->gtt_entry_size = 8;
+		info->gtt_entry_size_shift = 3;
+		info->gmadr_bytes_in_cmd = 8;
+		info->max_surface_size = 36 * 1024 * 1024;
+	}
+}
+
+static int gvt_service_thread(void *data)
+{
+	struct intel_gvt *gvt = (struct intel_gvt *)data;
+	int ret;
+
+	gvt_dbg_core("service thread start\n");
+
+	while (!kthread_should_stop()) {
+		ret = wait_event_interruptible(gvt->service_thread_wq,
+				kthread_should_stop() || gvt->service_request);
+
+		if (kthread_should_stop())
+			break;
+
+		if (WARN_ONCE(ret, "service thread is waken up by signal.\n"))
+			continue;
+
+		if (test_and_clear_bit(INTEL_GVT_REQUEST_EMULATE_VBLANK,
+					(void *)&gvt->service_request)) {
+			mutex_lock(&gvt->lock);
+			intel_gvt_emulate_vblank(gvt);
+			mutex_unlock(&gvt->lock);
+		}
+	}
+
+	return 0;
+}
+
+static void clean_service_thread(struct intel_gvt *gvt)
+{
+	kthread_stop(gvt->service_thread);
+}
+
+static int init_service_thread(struct intel_gvt *gvt)
+{
+	init_waitqueue_head(&gvt->service_thread_wq);
+
+	gvt->service_thread = kthread_run(gvt_service_thread,
+			gvt, "gvt_service_thread");
+	if (IS_ERR(gvt->service_thread)) {
+		gvt_err("fail to start service thread.\n");
+		return PTR_ERR(gvt->service_thread);
+	}
+	return 0;
 }
 
 /**
@@ -104,7 +178,15 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv)
 	if (WARN_ON(!gvt->initialized))
 		return;
 
-	/* Other de-initialization of GVT components will be introduced. */
+	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);
 
 	gvt->initialized = false;
 }
@@ -123,6 +205,8 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv)
 int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 {
 	struct intel_gvt *gvt = &dev_priv->gvt;
+	int ret;
+
 	/*
 	 * Cannot initialize GVT device without intel_gvt_host gets
 	 * initialized first.
@@ -135,11 +219,66 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 
 	gvt_dbg_core("init gvt device\n");
 
+	mutex_init(&gvt->lock);
+	gvt->dev_priv = dev_priv;
+
 	init_device_info(gvt);
-	/*
-	 * Other initialization of GVT components will be introduce here.
-	 */
+
+	ret = intel_gvt_setup_mmio_info(gvt);
+	if (ret)
+		return ret;
+
+	ret = intel_gvt_load_firmware(gvt);
+	if (ret)
+		goto out_clean_mmio_info;
+
+	ret = intel_gvt_init_irq(gvt);
+	if (ret)
+		goto out_free_firmware;
+
+	ret = intel_gvt_init_gtt(gvt);
+	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;
+
+	ret = intel_gvt_init_sched_policy(gvt);
+	if (ret)
+		goto out_clean_workload_scheduler;
+
+	ret = intel_gvt_init_cmd_parser(gvt);
+	if (ret)
+		goto out_clean_sched_policy;
+
+	ret = init_service_thread(gvt);
+	if (ret)
+		goto out_clean_cmd_parser;
+
 	gvt_dbg_core("gvt device creation is done\n");
 	gvt->initialized = true;
 	return 0;
+
+out_clean_cmd_parser:
+	intel_gvt_clean_cmd_parser(gvt);
+out_clean_sched_policy:
+	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:
+	intel_gvt_clean_irq(gvt);
+out_free_firmware:
+	intel_gvt_free_firmware(gvt);
+out_clean_mmio_info:
+	intel_gvt_clean_mmio_info(gvt);
+	return ret;
 }

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

@@ -19,6 +19,15 @@
  * 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.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Eddie Dong <eddie.dong@intel.com>
+ *
+ * Contributors:
+ *    Niu Bing <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
  */
 
 #ifndef _GVT_H_
@@ -26,6 +35,17 @@
 
 #include "debug.h"
 #include "hypercall.h"
+#include "mmio.h"
+#include "reg.h"
+#include "interrupt.h"
+#include "gtt.h"
+#include "display.h"
+#include "edid.h"
+#include "execlist.h"
+#include "scheduler.h"
+#include "sched_policy.h"
+#include "render.h"
+#include "cmd_parser.h"
 
 #define GVT_MAX_VGPU 8
 
@@ -45,13 +65,129 @@ extern struct intel_gvt_host intel_gvt_host;
 /* Describe per-platform limitations. */
 struct intel_gvt_device_info {
 	u32 max_support_vgpus;
-	/* This data structure will grow bigger in GVT device model patches */
+	u32 cfg_space_size;
+	u32 mmio_size;
+	u32 mmio_bar;
+	unsigned long msi_cap_offset;
+	u32 gtt_start_offset;
+	u32 gtt_entry_size;
+	u32 gtt_entry_size_shift;
+	int gmadr_bytes_in_cmd;
+	u32 max_surface_size;
+};
+
+/* GM resources owned by a vGPU */
+struct intel_vgpu_gm {
+	u64 aperture_sz;
+	u64 hidden_sz;
+	struct drm_mm_node low_gm_node;
+	struct drm_mm_node high_gm_node;
+};
+
+#define INTEL_GVT_MAX_NUM_FENCES 32
+
+/* Fences owned by a vGPU */
+struct intel_vgpu_fence {
+	struct drm_i915_fence_reg *regs[INTEL_GVT_MAX_NUM_FENCES];
+	u32 base;
+	u32 size;
+};
+
+struct intel_vgpu_mmio {
+	void *vreg;
+	void *sreg;
+	bool disable_warn_untrack;
+};
+
+#define INTEL_GVT_MAX_CFG_SPACE_SZ 256
+#define INTEL_GVT_MAX_BAR_NUM 4
+
+struct intel_vgpu_pci_bar {
+	u64 size;
+	bool tracked;
+};
+
+struct intel_vgpu_cfg_space {
+	unsigned char virtual_cfg_space[INTEL_GVT_MAX_CFG_SPACE_SZ];
+	struct intel_vgpu_pci_bar bar[INTEL_GVT_MAX_BAR_NUM];
+};
+
+#define vgpu_cfg_space(vgpu) ((vgpu)->cfg_space.virtual_cfg_space)
+
+#define INTEL_GVT_MAX_PIPE 4
+
+struct intel_vgpu_irq {
+	bool irq_warn_once[INTEL_GVT_EVENT_MAX];
+	DECLARE_BITMAP(flip_done_event[INTEL_GVT_MAX_PIPE],
+		       INTEL_GVT_EVENT_MAX);
+};
+
+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))
+
+#define INTEL_GVT_MAX_PORT 5
+
+struct intel_vgpu_display {
+	struct intel_vgpu_i2c_edid i2c_edid;
+	struct intel_vgpu_port ports[INTEL_GVT_MAX_PORT];
+	struct intel_vgpu_sbi sbi;
 };
 
 struct intel_vgpu {
 	struct intel_gvt *gvt;
 	int id;
 	unsigned long handle; /* vGPU handle used by hypervisor MPT modules */
+	bool active;
+	bool resetting;
+	void *sched_data;
+
+	struct intel_vgpu_fence fence;
+	struct intel_vgpu_gm gm;
+	struct intel_vgpu_cfg_space cfg_space;
+	struct intel_vgpu_mmio mmio;
+	struct intel_vgpu_irq irq;
+	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;
+	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
+	struct i915_gem_context *shadow_ctx;
+	struct notifier_block shadow_ctx_notifier_block;
+};
+
+struct intel_gvt_gm {
+	unsigned long vgpu_allocated_low_gm_size;
+	unsigned long vgpu_allocated_high_gm_size;
+};
+
+struct intel_gvt_fence {
+	unsigned long vgpu_allocated_fence_num;
+};
+
+#define INTEL_GVT_MMIO_HASH_BITS 9
+
+struct intel_gvt_mmio {
+	u32 *mmio_attribute;
+	DECLARE_HASHTABLE(mmio_info_table, INTEL_GVT_MMIO_HASH_BITS);
+};
+
+struct intel_gvt_firmware {
+	void *cfg_space;
+	void *mmio;
+	bool firmware_loaded;
+};
+
+struct intel_gvt_opregion {
+	void *opregion_va;
+	u32 opregion_pa;
 };
 
 struct intel_gvt {
@@ -62,8 +198,188 @@ struct intel_gvt {
 	struct idr vgpu_idr;	/* vGPU IDR pool */
 
 	struct intel_gvt_device_info device_info;
+	struct intel_gvt_gm gm;
+	struct intel_gvt_fence fence;
+	struct intel_gvt_mmio mmio;
+	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;
+	DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
+
+	struct task_struct *service_thread;
+	wait_queue_head_t service_thread_wq;
+	unsigned long service_request;
 };
 
+enum {
+	INTEL_GVT_REQUEST_EMULATE_VBLANK = 0,
+};
+
+static inline void intel_gvt_request_service(struct intel_gvt *gvt,
+		int service)
+{
+	set_bit(service, (void *)&gvt->service_request);
+	wake_up(&gvt->service_thread_wq);
+}
+
+void intel_gvt_free_firmware(struct intel_gvt *gvt);
+int intel_gvt_load_firmware(struct intel_gvt *gvt);
+
+/* Aperture/GM space definitions for GVT device */
+#define gvt_aperture_sz(gvt)	  (gvt->dev_priv->ggtt.mappable_end)
+#define gvt_aperture_pa_base(gvt) (gvt->dev_priv->ggtt.mappable_base)
+
+#define gvt_ggtt_gm_sz(gvt)	  (gvt->dev_priv->ggtt.base.total)
+#define gvt_ggtt_sz(gvt) \
+	((gvt->dev_priv->ggtt.base.total >> PAGE_SHIFT) << 3)
+#define gvt_hidden_sz(gvt)	  (gvt_ggtt_gm_sz(gvt) - gvt_aperture_sz(gvt))
+
+#define gvt_aperture_gmadr_base(gvt) (0)
+#define gvt_aperture_gmadr_end(gvt) (gvt_aperture_gmadr_base(gvt) \
+				     + gvt_aperture_sz(gvt) - 1)
+
+#define gvt_hidden_gmadr_base(gvt) (gvt_aperture_gmadr_base(gvt) \
+				    + gvt_aperture_sz(gvt))
+#define gvt_hidden_gmadr_end(gvt) (gvt_hidden_gmadr_base(gvt) \
+				   + gvt_hidden_sz(gvt) - 1)
+
+#define gvt_fence_sz(gvt) (gvt->dev_priv->num_fence_regs)
+
+/* Aperture/GM space definitions for vGPU */
+#define vgpu_aperture_offset(vgpu)	((vgpu)->gm.low_gm_node.start)
+#define vgpu_hidden_offset(vgpu)	((vgpu)->gm.high_gm_node.start)
+#define vgpu_aperture_sz(vgpu)		((vgpu)->gm.aperture_sz)
+#define vgpu_hidden_sz(vgpu)		((vgpu)->gm.hidden_sz)
+
+#define vgpu_aperture_pa_base(vgpu) \
+	(gvt_aperture_pa_base(vgpu->gvt) + vgpu_aperture_offset(vgpu))
+
+#define vgpu_ggtt_gm_sz(vgpu) ((vgpu)->gm.aperture_sz + (vgpu)->gm.hidden_sz)
+
+#define vgpu_aperture_pa_end(vgpu) \
+	(vgpu_aperture_pa_base(vgpu) + vgpu_aperture_sz(vgpu) - 1)
+
+#define vgpu_aperture_gmadr_base(vgpu) (vgpu_aperture_offset(vgpu))
+#define vgpu_aperture_gmadr_end(vgpu) \
+	(vgpu_aperture_gmadr_base(vgpu) + vgpu_aperture_sz(vgpu) - 1)
+
+#define vgpu_hidden_gmadr_base(vgpu) (vgpu_hidden_offset(vgpu))
+#define vgpu_hidden_gmadr_end(vgpu) \
+	(vgpu_hidden_gmadr_base(vgpu) + vgpu_hidden_sz(vgpu) - 1)
+
+#define vgpu_fence_base(vgpu) (vgpu->fence.base)
+#define vgpu_fence_sz(vgpu) (vgpu->fence.size)
+
+struct intel_vgpu_creation_params {
+	__u64 handle;
+	__u64 low_gm_sz;  /* in MB */
+	__u64 high_gm_sz; /* in MB */
+	__u64 fence_sz;
+	__s32 primary;
+	__u64 vgpu_id;
+};
+
+int intel_vgpu_alloc_resource(struct intel_vgpu *vgpu,
+			      struct intel_vgpu_creation_params *param);
+void intel_vgpu_free_resource(struct intel_vgpu *vgpu);
+void intel_vgpu_write_fence(struct intel_vgpu *vgpu,
+	u32 fence, u64 value);
+
+/* Macros for easily accessing vGPU virtual/shadow register */
+#define vgpu_vreg(vgpu, reg) \
+	(*(u32 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_vreg8(vgpu, reg) \
+	(*(u8 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_vreg16(vgpu, reg) \
+	(*(u16 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_vreg64(vgpu, reg) \
+	(*(u64 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_sreg(vgpu, reg) \
+	(*(u32 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_sreg8(vgpu, reg) \
+	(*(u8 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_sreg16(vgpu, reg) \
+	(*(u16 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
+#define vgpu_sreg64(vgpu, reg) \
+	(*(u64 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
+
+#define for_each_active_vgpu(gvt, vgpu, id) \
+	idr_for_each_entry((&(gvt)->vgpu_idr), (vgpu), (id)) \
+		for_each_if(vgpu->active)
+
+static inline void intel_vgpu_write_pci_bar(struct intel_vgpu *vgpu,
+					    u32 offset, u32 val, bool low)
+{
+	u32 *pval;
+
+	/* BAR offset should be 32 bits algiend */
+	offset = rounddown(offset, 4);
+	pval = (u32 *)(vgpu_cfg_space(vgpu) + offset);
+
+	if (low) {
+		/*
+		 * only update bit 31 - bit 4,
+		 * leave the bit 3 - bit 0 unchanged.
+		 */
+		*pval = (val & GENMASK(31, 4)) | (*pval & GENMASK(3, 0));
+	}
+}
+
+struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
+					 struct intel_vgpu_creation_params *
+					 param);
+
+void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu);
+
+/* validating GM functions */
+#define vgpu_gmadr_is_aperture(vgpu, gmadr) \
+	((gmadr >= vgpu_aperture_gmadr_base(vgpu)) && \
+	 (gmadr <= vgpu_aperture_gmadr_end(vgpu)))
+
+#define vgpu_gmadr_is_hidden(vgpu, gmadr) \
+	((gmadr >= vgpu_hidden_gmadr_base(vgpu)) && \
+	 (gmadr <= vgpu_hidden_gmadr_end(vgpu)))
+
+#define vgpu_gmadr_is_valid(vgpu, gmadr) \
+	 ((vgpu_gmadr_is_aperture(vgpu, gmadr) || \
+	  (vgpu_gmadr_is_hidden(vgpu, gmadr))))
+
+#define gvt_gmadr_is_aperture(gvt, gmadr) \
+	 ((gmadr >= gvt_aperture_gmadr_base(gvt)) && \
+	  (gmadr <= gvt_aperture_gmadr_end(gvt)))
+
+#define gvt_gmadr_is_hidden(gvt, gmadr) \
+	  ((gmadr >= gvt_hidden_gmadr_base(gvt)) && \
+	   (gmadr <= gvt_hidden_gmadr_end(gvt)))
+
+#define gvt_gmadr_is_valid(gvt, gmadr) \
+	  (gvt_gmadr_is_aperture(gvt, gmadr) || \
+	    gvt_gmadr_is_hidden(gvt, gmadr))
+
+bool intel_gvt_ggtt_validate_range(struct intel_vgpu *vgpu, u64 addr, u32 size);
+int intel_gvt_ggtt_gmadr_g2h(struct intel_vgpu *vgpu, u64 g_addr, u64 *h_addr);
+int intel_gvt_ggtt_gmadr_h2g(struct intel_vgpu *vgpu, u64 h_addr, u64 *g_addr);
+int intel_gvt_ggtt_index_g2h(struct intel_vgpu *vgpu, unsigned long g_index,
+			     unsigned long *h_index);
+int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index,
+			     unsigned long *g_index);
+
+int intel_vgpu_emulate_cfg_read(void *__vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes);
+
+int intel_vgpu_emulate_cfg_write(void *__vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes);
+
+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);
+
+int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci);
+
 #include "mpt.h"
 
 #endif

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

@@ -19,17 +19,51 @@
  * 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.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Dexuan Cui
+ *    Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
  */
 
 #ifndef _GVT_HYPERCALL_H_
 #define _GVT_HYPERCALL_H_
 
+struct intel_gvt_io_emulation_ops {
+	int (*emulate_cfg_read)(void *, unsigned int, void *, unsigned int);
+	int (*emulate_cfg_write)(void *, unsigned int, void *, unsigned int);
+	int (*emulate_mmio_read)(void *, u64, void *, unsigned int);
+	int (*emulate_mmio_write)(void *, u64, void *, unsigned int);
+};
+
+extern struct intel_gvt_io_emulation_ops intel_gvt_io_emulation_ops;
+
 /*
  * Specific GVT-g MPT modules function collections. Currently GVT-g supports
  * both Xen and KVM by providing dedicated hypervisor-related MPT modules.
  */
 struct intel_gvt_mpt {
 	int (*detect_host)(void);
+	int (*attach_vgpu)(void *vgpu, unsigned long *handle);
+	void (*detach_vgpu)(unsigned long handle);
+	int (*inject_msi)(unsigned long handle, u32 addr, u16 data);
+	unsigned long (*from_virt_to_mfn)(void *p);
+	int (*set_wp_page)(unsigned long handle, u64 gfn);
+	int (*unset_wp_page)(unsigned long handle, u64 gfn);
+	int (*read_gpa)(unsigned long handle, unsigned long gpa, void *buf,
+			unsigned long len);
+	int (*write_gpa)(unsigned long handle, unsigned long gpa, void *buf,
+			 unsigned long len);
+	unsigned long (*gfn_to_mfn)(unsigned long handle, unsigned long gfn);
+	int (*map_gfn_to_mfn)(unsigned long handle, unsigned long gfn,
+			      unsigned long mfn, unsigned int nr, bool map,
+			      int type);
+	int (*set_trap_area)(unsigned long handle, u64 start, u64 end,
+			     bool map);
 };
 
 extern struct intel_gvt_mpt xengt_mpt;

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

@@ -0,0 +1,740 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Min he <min.he@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+/* common offset among interrupt control registers */
+#define regbase_to_isr(base)	(base)
+#define regbase_to_imr(base)	(base + 0x4)
+#define regbase_to_iir(base)	(base + 0x8)
+#define regbase_to_ier(base)	(base + 0xC)
+
+#define iir_to_regbase(iir)    (iir - 0x8)
+#define ier_to_regbase(ier)    (ier - 0xC)
+
+#define get_event_virt_handler(irq, e)	(irq->events[e].v_handler)
+#define get_irq_info(irq, e)		(irq->events[e].info)
+
+#define irq_to_gvt(irq) \
+	container_of(irq, struct intel_gvt, irq)
+
+static void update_upstream_irq(struct intel_vgpu *vgpu,
+		struct intel_gvt_irq_info *info);
+
+const char * const irq_name[INTEL_GVT_EVENT_MAX] = {
+	[RCS_MI_USER_INTERRUPT] = "Render CS MI USER INTERRUPT",
+	[RCS_DEBUG] = "Render EU debug from SVG",
+	[RCS_MMIO_SYNC_FLUSH] = "Render MMIO sync flush status",
+	[RCS_CMD_STREAMER_ERR] = "Render CS error interrupt",
+	[RCS_PIPE_CONTROL] = "Render PIPE CONTROL notify",
+	[RCS_WATCHDOG_EXCEEDED] = "Render CS Watchdog counter exceeded",
+	[RCS_PAGE_DIRECTORY_FAULT] = "Render page directory faults",
+	[RCS_AS_CONTEXT_SWITCH] = "Render AS Context Switch Interrupt",
+
+	[VCS_MI_USER_INTERRUPT] = "Video CS MI USER INTERRUPT",
+	[VCS_MMIO_SYNC_FLUSH] = "Video MMIO sync flush status",
+	[VCS_CMD_STREAMER_ERR] = "Video CS error interrupt",
+	[VCS_MI_FLUSH_DW] = "Video MI FLUSH DW notify",
+	[VCS_WATCHDOG_EXCEEDED] = "Video CS Watchdog counter exceeded",
+	[VCS_PAGE_DIRECTORY_FAULT] = "Video page directory faults",
+	[VCS_AS_CONTEXT_SWITCH] = "Video AS Context Switch Interrupt",
+	[VCS2_MI_USER_INTERRUPT] = "VCS2 Video CS MI USER INTERRUPT",
+	[VCS2_MI_FLUSH_DW] = "VCS2 Video MI FLUSH DW notify",
+	[VCS2_AS_CONTEXT_SWITCH] = "VCS2 Context Switch Interrupt",
+
+	[BCS_MI_USER_INTERRUPT] = "Blitter CS MI USER INTERRUPT",
+	[BCS_MMIO_SYNC_FLUSH] = "Billter MMIO sync flush status",
+	[BCS_CMD_STREAMER_ERR] = "Blitter CS error interrupt",
+	[BCS_MI_FLUSH_DW] = "Blitter MI FLUSH DW notify",
+	[BCS_PAGE_DIRECTORY_FAULT] = "Blitter page directory faults",
+	[BCS_AS_CONTEXT_SWITCH] = "Blitter AS Context Switch Interrupt",
+
+	[VECS_MI_FLUSH_DW] = "Video Enhanced Streamer MI FLUSH DW notify",
+	[VECS_AS_CONTEXT_SWITCH] = "VECS Context Switch Interrupt",
+
+	[PIPE_A_FIFO_UNDERRUN] = "Pipe A FIFO underrun",
+	[PIPE_A_CRC_ERR] = "Pipe A CRC error",
+	[PIPE_A_CRC_DONE] = "Pipe A CRC done",
+	[PIPE_A_VSYNC] = "Pipe A vsync",
+	[PIPE_A_LINE_COMPARE] = "Pipe A line compare",
+	[PIPE_A_ODD_FIELD] = "Pipe A odd field",
+	[PIPE_A_EVEN_FIELD] = "Pipe A even field",
+	[PIPE_A_VBLANK] = "Pipe A vblank",
+	[PIPE_B_FIFO_UNDERRUN] = "Pipe B FIFO underrun",
+	[PIPE_B_CRC_ERR] = "Pipe B CRC error",
+	[PIPE_B_CRC_DONE] = "Pipe B CRC done",
+	[PIPE_B_VSYNC] = "Pipe B vsync",
+	[PIPE_B_LINE_COMPARE] = "Pipe B line compare",
+	[PIPE_B_ODD_FIELD] = "Pipe B odd field",
+	[PIPE_B_EVEN_FIELD] = "Pipe B even field",
+	[PIPE_B_VBLANK] = "Pipe B vblank",
+	[PIPE_C_VBLANK] = "Pipe C vblank",
+	[DPST_PHASE_IN] = "DPST phase in event",
+	[DPST_HISTOGRAM] = "DPST histogram event",
+	[GSE] = "GSE",
+	[DP_A_HOTPLUG] = "DP A Hotplug",
+	[AUX_CHANNEL_A] = "AUX Channel A",
+	[PERF_COUNTER] = "Performance counter",
+	[POISON] = "Poison",
+	[GTT_FAULT] = "GTT fault",
+	[PRIMARY_A_FLIP_DONE] = "Primary Plane A flip done",
+	[PRIMARY_B_FLIP_DONE] = "Primary Plane B flip done",
+	[PRIMARY_C_FLIP_DONE] = "Primary Plane C flip done",
+	[SPRITE_A_FLIP_DONE] = "Sprite Plane A flip done",
+	[SPRITE_B_FLIP_DONE] = "Sprite Plane B flip done",
+	[SPRITE_C_FLIP_DONE] = "Sprite Plane C flip done",
+
+	[PCU_THERMAL] = "PCU Thermal Event",
+	[PCU_PCODE2DRIVER_MAILBOX] = "PCU pcode2driver mailbox event",
+
+	[FDI_RX_INTERRUPTS_TRANSCODER_A] = "FDI RX Interrupts Combined A",
+	[AUDIO_CP_CHANGE_TRANSCODER_A] = "Audio CP Change Transcoder A",
+	[AUDIO_CP_REQUEST_TRANSCODER_A] = "Audio CP Request Transcoder A",
+	[FDI_RX_INTERRUPTS_TRANSCODER_B] = "FDI RX Interrupts Combined B",
+	[AUDIO_CP_CHANGE_TRANSCODER_B] = "Audio CP Change Transcoder B",
+	[AUDIO_CP_REQUEST_TRANSCODER_B] = "Audio CP Request Transcoder B",
+	[FDI_RX_INTERRUPTS_TRANSCODER_C] = "FDI RX Interrupts Combined C",
+	[AUDIO_CP_CHANGE_TRANSCODER_C] = "Audio CP Change Transcoder C",
+	[AUDIO_CP_REQUEST_TRANSCODER_C] = "Audio CP Request Transcoder C",
+	[ERR_AND_DBG] = "South Error and Debug Interupts Combined",
+	[GMBUS] = "Gmbus",
+	[SDVO_B_HOTPLUG] = "SDVO B hotplug",
+	[CRT_HOTPLUG] = "CRT Hotplug",
+	[DP_B_HOTPLUG] = "DisplayPort/HDMI/DVI B Hotplug",
+	[DP_C_HOTPLUG] = "DisplayPort/HDMI/DVI C Hotplug",
+	[DP_D_HOTPLUG] = "DisplayPort/HDMI/DVI D Hotplug",
+	[AUX_CHANNEL_B] = "AUX Channel B",
+	[AUX_CHANNEL_C] = "AUX Channel C",
+	[AUX_CHANNEL_D] = "AUX Channel D",
+	[AUDIO_POWER_STATE_CHANGE_B] = "Audio Power State change Port B",
+	[AUDIO_POWER_STATE_CHANGE_C] = "Audio Power State change Port C",
+	[AUDIO_POWER_STATE_CHANGE_D] = "Audio Power State change Port D",
+
+	[INTEL_GVT_EVENT_RESERVED] = "RESERVED EVENTS!!!",
+};
+
+static inline struct intel_gvt_irq_info *regbase_to_irq_info(
+		struct intel_gvt *gvt,
+		unsigned int reg)
+{
+	struct intel_gvt_irq *irq = &gvt->irq;
+	int i;
+
+	for_each_set_bit(i, irq->irq_info_bitmap, INTEL_GVT_IRQ_INFO_MAX) {
+		if (i915_mmio_reg_offset(irq->info[i]->reg_base) == reg)
+			return irq->info[i];
+	}
+
+	return NULL;
+}
+
+/**
+ * intel_vgpu_reg_imr_handler - Generic IMR register emulation write handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the generic IMR register bit change
+ * behavior.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_imr_handler(struct intel_vgpu *vgpu,
+	unsigned int reg, void *p_data, unsigned int bytes)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+	u32 changed, masked, unmasked;
+	u32 imr = *(u32 *)p_data;
+
+	gvt_dbg_irq("write IMR %x with val %x\n",
+		reg, imr);
+
+	gvt_dbg_irq("old vIMR %x\n", vgpu_vreg(vgpu, reg));
+
+	/* figure out newly masked/unmasked bits */
+	changed = vgpu_vreg(vgpu, reg) ^ imr;
+	masked = (vgpu_vreg(vgpu, reg) & changed) ^ changed;
+	unmasked = masked ^ changed;
+
+	gvt_dbg_irq("changed %x, masked %x, unmasked %x\n",
+		changed, masked, unmasked);
+
+	vgpu_vreg(vgpu, reg) = imr;
+
+	ops->check_pending_irq(vgpu);
+	gvt_dbg_irq("IRQ: new vIMR %x\n", vgpu_vreg(vgpu, reg));
+	return 0;
+}
+
+/**
+ * intel_vgpu_reg_master_irq_handler - master IRQ write emulation handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the master IRQ register on gen8+.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_master_irq_handler(struct intel_vgpu *vgpu,
+	unsigned int reg, void *p_data, unsigned int bytes)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+	u32 changed, enabled, disabled;
+	u32 ier = *(u32 *)p_data;
+	u32 virtual_ier = vgpu_vreg(vgpu, reg);
+
+	gvt_dbg_irq("write master irq reg %x with val %x\n",
+		reg, ier);
+
+	gvt_dbg_irq("old vreg %x\n", vgpu_vreg(vgpu, reg));
+
+	/*
+	 * GEN8_MASTER_IRQ is a special irq register,
+	 * only bit 31 is allowed to be modified
+	 * and treated as an IER bit.
+	 */
+	ier &= GEN8_MASTER_IRQ_CONTROL;
+	virtual_ier &= GEN8_MASTER_IRQ_CONTROL;
+	vgpu_vreg(vgpu, reg) &= ~GEN8_MASTER_IRQ_CONTROL;
+	vgpu_vreg(vgpu, reg) |= ier;
+
+	/* figure out newly enabled/disable bits */
+	changed = virtual_ier ^ ier;
+	enabled = (virtual_ier & changed) ^ changed;
+	disabled = enabled ^ changed;
+
+	gvt_dbg_irq("changed %x, enabled %x, disabled %x\n",
+			changed, enabled, disabled);
+
+	ops->check_pending_irq(vgpu);
+	gvt_dbg_irq("new vreg %x\n", vgpu_vreg(vgpu, reg));
+	return 0;
+}
+
+/**
+ * intel_vgpu_reg_ier_handler - Generic IER write emulation handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the generic IER register behavior.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_ier_handler(struct intel_vgpu *vgpu,
+	unsigned int reg, void *p_data, unsigned int bytes)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+	struct intel_gvt_irq_info *info;
+	u32 changed, enabled, disabled;
+	u32 ier = *(u32 *)p_data;
+
+	gvt_dbg_irq("write IER %x with val %x\n",
+		reg, ier);
+
+	gvt_dbg_irq("old vIER %x\n", vgpu_vreg(vgpu, reg));
+
+	/* figure out newly enabled/disable bits */
+	changed = vgpu_vreg(vgpu, reg) ^ ier;
+	enabled = (vgpu_vreg(vgpu, reg) & changed) ^ changed;
+	disabled = enabled ^ changed;
+
+	gvt_dbg_irq("changed %x, enabled %x, disabled %x\n",
+			changed, enabled, disabled);
+	vgpu_vreg(vgpu, reg) = ier;
+
+	info = regbase_to_irq_info(gvt, ier_to_regbase(reg));
+	if (WARN_ON(!info))
+		return -EINVAL;
+
+	if (info->has_upstream_irq)
+		update_upstream_irq(vgpu, info);
+
+	ops->check_pending_irq(vgpu);
+	gvt_dbg_irq("new vIER %x\n", vgpu_vreg(vgpu, reg));
+	return 0;
+}
+
+/**
+ * intel_vgpu_reg_iir_handler - Generic IIR write emulation handler
+ * @vgpu: a vGPU
+ * @reg: register offset written by guest
+ * @p_data: register data written by guest
+ * @bytes: register data length
+ *
+ * This function is used to emulate the generic IIR register behavior.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_reg_iir_handler(struct intel_vgpu *vgpu, unsigned int reg,
+	void *p_data, unsigned int bytes)
+{
+	struct intel_gvt_irq_info *info = regbase_to_irq_info(vgpu->gvt,
+		iir_to_regbase(reg));
+	u32 iir = *(u32 *)p_data;
+
+	gvt_dbg_irq("write IIR %x with val %x\n", reg, iir);
+
+	if (WARN_ON(!info))
+		return -EINVAL;
+
+	vgpu_vreg(vgpu, reg) &= ~iir;
+
+	if (info->has_upstream_irq)
+		update_upstream_irq(vgpu, info);
+	return 0;
+}
+
+static struct intel_gvt_irq_map gen8_irq_map[] = {
+	{ INTEL_GVT_IRQ_INFO_MASTER, 0, INTEL_GVT_IRQ_INFO_GT0, 0xffff },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 1, INTEL_GVT_IRQ_INFO_GT0, 0xffff0000 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 2, INTEL_GVT_IRQ_INFO_GT1, 0xffff },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 3, INTEL_GVT_IRQ_INFO_GT1, 0xffff0000 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 4, INTEL_GVT_IRQ_INFO_GT2, 0xffff },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 6, INTEL_GVT_IRQ_INFO_GT3, 0xffff },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 16, INTEL_GVT_IRQ_INFO_DE_PIPE_A, ~0 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 17, INTEL_GVT_IRQ_INFO_DE_PIPE_B, ~0 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 18, INTEL_GVT_IRQ_INFO_DE_PIPE_C, ~0 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 20, INTEL_GVT_IRQ_INFO_DE_PORT, ~0 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 22, INTEL_GVT_IRQ_INFO_DE_MISC, ~0 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 23, INTEL_GVT_IRQ_INFO_PCH, ~0 },
+	{ INTEL_GVT_IRQ_INFO_MASTER, 30, INTEL_GVT_IRQ_INFO_PCU, ~0 },
+	{ -1, -1, ~0 },
+};
+
+static void update_upstream_irq(struct intel_vgpu *vgpu,
+		struct intel_gvt_irq_info *info)
+{
+	struct intel_gvt_irq *irq = &vgpu->gvt->irq;
+	struct intel_gvt_irq_map *map = irq->irq_map;
+	struct intel_gvt_irq_info *up_irq_info = NULL;
+	u32 set_bits = 0;
+	u32 clear_bits = 0;
+	int bit;
+	u32 val = vgpu_vreg(vgpu,
+			regbase_to_iir(i915_mmio_reg_offset(info->reg_base)))
+		& vgpu_vreg(vgpu,
+			regbase_to_ier(i915_mmio_reg_offset(info->reg_base)));
+
+	if (!info->has_upstream_irq)
+		return;
+
+	for (map = irq->irq_map; map->up_irq_bit != -1; map++) {
+		if (info->group != map->down_irq_group)
+			continue;
+
+		if (!up_irq_info)
+			up_irq_info = irq->info[map->up_irq_group];
+		else
+			WARN_ON(up_irq_info != irq->info[map->up_irq_group]);
+
+		bit = map->up_irq_bit;
+
+		if (val & map->down_irq_bitmask)
+			set_bits |= (1 << bit);
+		else
+			clear_bits |= (1 << bit);
+	}
+
+	WARN_ON(!up_irq_info);
+
+	if (up_irq_info->group == INTEL_GVT_IRQ_INFO_MASTER) {
+		u32 isr = i915_mmio_reg_offset(up_irq_info->reg_base);
+
+		vgpu_vreg(vgpu, isr) &= ~clear_bits;
+		vgpu_vreg(vgpu, isr) |= set_bits;
+	} else {
+		u32 iir = regbase_to_iir(
+			i915_mmio_reg_offset(up_irq_info->reg_base));
+		u32 imr = regbase_to_imr(
+			i915_mmio_reg_offset(up_irq_info->reg_base));
+
+		vgpu_vreg(vgpu, iir) |= (set_bits & ~vgpu_vreg(vgpu, imr));
+	}
+
+	if (up_irq_info->has_upstream_irq)
+		update_upstream_irq(vgpu, up_irq_info);
+}
+
+static void init_irq_map(struct intel_gvt_irq *irq)
+{
+	struct intel_gvt_irq_map *map;
+	struct intel_gvt_irq_info *up_info, *down_info;
+	int up_bit;
+
+	for (map = irq->irq_map; map->up_irq_bit != -1; map++) {
+		up_info = irq->info[map->up_irq_group];
+		up_bit = map->up_irq_bit;
+		down_info = irq->info[map->down_irq_group];
+
+		set_bit(up_bit, up_info->downstream_irq_bitmap);
+		down_info->has_upstream_irq = true;
+
+		gvt_dbg_irq("[up] grp %d bit %d -> [down] grp %d bitmask %x\n",
+			up_info->group, up_bit,
+			down_info->group, map->down_irq_bitmask);
+	}
+}
+
+/* =======================vEvent injection===================== */
+static int inject_virtual_interrupt(struct intel_vgpu *vgpu)
+{
+	return intel_gvt_hypervisor_inject_msi(vgpu);
+}
+
+static void propagate_event(struct intel_gvt_irq *irq,
+	enum intel_gvt_event_type event, struct intel_vgpu *vgpu)
+{
+	struct intel_gvt_irq_info *info;
+	unsigned int reg_base;
+	int bit;
+
+	info = get_irq_info(irq, event);
+	if (WARN_ON(!info))
+		return;
+
+	reg_base = i915_mmio_reg_offset(info->reg_base);
+	bit = irq->events[event].bit;
+
+	if (!test_bit(bit, (void *)&vgpu_vreg(vgpu,
+					regbase_to_imr(reg_base)))) {
+		gvt_dbg_irq("set bit (%d) for (%s) for vgpu (%d)\n",
+				bit, irq_name[event], vgpu->id);
+		set_bit(bit, (void *)&vgpu_vreg(vgpu,
+					regbase_to_iir(reg_base)));
+	}
+}
+
+/* =======================vEvent Handlers===================== */
+static void handle_default_event_virt(struct intel_gvt_irq *irq,
+	enum intel_gvt_event_type event, struct intel_vgpu *vgpu)
+{
+	if (!vgpu->irq.irq_warn_once[event]) {
+		gvt_dbg_core("vgpu%d: IRQ receive event %d (%s)\n",
+			vgpu->id, event, irq_name[event]);
+		vgpu->irq.irq_warn_once[event] = true;
+	}
+	propagate_event(irq, event, vgpu);
+}
+
+/* =====================GEN specific logic======================= */
+/* GEN8 interrupt routines. */
+
+#define DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(regname, regbase) \
+static struct intel_gvt_irq_info gen8_##regname##_info = { \
+	.name = #regname"-IRQ", \
+	.reg_base = (regbase), \
+	.bit_to_event = {[0 ... INTEL_GVT_IRQ_BITWIDTH-1] = \
+		INTEL_GVT_EVENT_RESERVED}, \
+}
+
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt0, GEN8_GT_ISR(0));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt1, GEN8_GT_ISR(1));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt2, GEN8_GT_ISR(2));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(gt3, GEN8_GT_ISR(3));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_pipe_a, GEN8_DE_PIPE_ISR(PIPE_A));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_pipe_b, GEN8_DE_PIPE_ISR(PIPE_B));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_pipe_c, GEN8_DE_PIPE_ISR(PIPE_C));
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_port, GEN8_DE_PORT_ISR);
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(de_misc, GEN8_DE_MISC_ISR);
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(pcu, GEN8_PCU_ISR);
+DEFINE_GVT_GEN8_INTEL_GVT_IRQ_INFO(master, GEN8_MASTER_IRQ);
+
+static struct intel_gvt_irq_info gvt_base_pch_info = {
+	.name = "PCH-IRQ",
+	.reg_base = SDEISR,
+	.bit_to_event = {[0 ... INTEL_GVT_IRQ_BITWIDTH-1] =
+		INTEL_GVT_EVENT_RESERVED},
+};
+
+static void gen8_check_pending_irq(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt_irq *irq = &vgpu->gvt->irq;
+	int i;
+
+	if (!(vgpu_vreg(vgpu, i915_mmio_reg_offset(GEN8_MASTER_IRQ)) &
+				GEN8_MASTER_IRQ_CONTROL))
+		return;
+
+	for_each_set_bit(i, irq->irq_info_bitmap, INTEL_GVT_IRQ_INFO_MAX) {
+		struct intel_gvt_irq_info *info = irq->info[i];
+		u32 reg_base;
+
+		if (!info->has_upstream_irq)
+			continue;
+
+		reg_base = i915_mmio_reg_offset(info->reg_base);
+		if ((vgpu_vreg(vgpu, regbase_to_iir(reg_base))
+				& vgpu_vreg(vgpu, regbase_to_ier(reg_base))))
+			update_upstream_irq(vgpu, info);
+	}
+
+	if (vgpu_vreg(vgpu, i915_mmio_reg_offset(GEN8_MASTER_IRQ))
+			& ~GEN8_MASTER_IRQ_CONTROL)
+		inject_virtual_interrupt(vgpu);
+}
+
+static void gen8_init_irq(
+		struct intel_gvt_irq *irq)
+{
+	struct intel_gvt *gvt = irq_to_gvt(irq);
+
+#define SET_BIT_INFO(s, b, e, i)		\
+	do {					\
+		s->events[e].bit = b;		\
+		s->events[e].info = s->info[i];	\
+		s->info[i]->bit_to_event[b] = e;\
+	} while (0)
+
+#define SET_IRQ_GROUP(s, g, i) \
+	do { \
+		s->info[g] = i; \
+		(i)->group = g; \
+		set_bit(g, s->irq_info_bitmap); \
+	} while (0)
+
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_MASTER, &gen8_master_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT0, &gen8_gt0_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT1, &gen8_gt1_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT2, &gen8_gt2_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_GT3, &gen8_gt3_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PIPE_A, &gen8_de_pipe_a_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PIPE_B, &gen8_de_pipe_b_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PIPE_C, &gen8_de_pipe_c_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_PORT, &gen8_de_port_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_DE_MISC, &gen8_de_misc_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_PCU, &gen8_pcu_info);
+	SET_IRQ_GROUP(irq, INTEL_GVT_IRQ_INFO_PCH, &gvt_base_pch_info);
+
+	/* GEN8 level 2 interrupts. */
+
+	/* GEN8 interrupt GT0 events */
+	SET_BIT_INFO(irq, 0, RCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT0);
+	SET_BIT_INFO(irq, 4, RCS_PIPE_CONTROL, INTEL_GVT_IRQ_INFO_GT0);
+	SET_BIT_INFO(irq, 8, RCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT0);
+
+	SET_BIT_INFO(irq, 16, BCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT0);
+	SET_BIT_INFO(irq, 20, BCS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT0);
+	SET_BIT_INFO(irq, 24, BCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT0);
+
+	/* GEN8 interrupt GT1 events */
+	SET_BIT_INFO(irq, 0, VCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT1);
+	SET_BIT_INFO(irq, 4, VCS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT1);
+	SET_BIT_INFO(irq, 8, VCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT1);
+
+	if (HAS_BSD2(gvt->dev_priv)) {
+		SET_BIT_INFO(irq, 16, VCS2_MI_USER_INTERRUPT,
+			INTEL_GVT_IRQ_INFO_GT1);
+		SET_BIT_INFO(irq, 20, VCS2_MI_FLUSH_DW,
+			INTEL_GVT_IRQ_INFO_GT1);
+		SET_BIT_INFO(irq, 24, VCS2_AS_CONTEXT_SWITCH,
+			INTEL_GVT_IRQ_INFO_GT1);
+	}
+
+	/* GEN8 interrupt GT3 events */
+	SET_BIT_INFO(irq, 0, VECS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT3);
+	SET_BIT_INFO(irq, 4, VECS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT3);
+	SET_BIT_INFO(irq, 8, VECS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT3);
+
+	SET_BIT_INFO(irq, 0, PIPE_A_VBLANK, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+	SET_BIT_INFO(irq, 0, PIPE_B_VBLANK, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+	SET_BIT_INFO(irq, 0, PIPE_C_VBLANK, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+
+	/* GEN8 interrupt DE PORT events */
+	SET_BIT_INFO(irq, 0, AUX_CHANNEL_A, INTEL_GVT_IRQ_INFO_DE_PORT);
+	SET_BIT_INFO(irq, 3, DP_A_HOTPLUG, INTEL_GVT_IRQ_INFO_DE_PORT);
+
+	/* GEN8 interrupt DE MISC events */
+	SET_BIT_INFO(irq, 0, GSE, INTEL_GVT_IRQ_INFO_DE_MISC);
+
+	/* PCH events */
+	SET_BIT_INFO(irq, 17, GMBUS, INTEL_GVT_IRQ_INFO_PCH);
+	SET_BIT_INFO(irq, 19, CRT_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+	SET_BIT_INFO(irq, 21, DP_B_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+	SET_BIT_INFO(irq, 22, DP_C_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+	SET_BIT_INFO(irq, 23, DP_D_HOTPLUG, INTEL_GVT_IRQ_INFO_PCH);
+
+	if (IS_BROADWELL(gvt->dev_priv)) {
+		SET_BIT_INFO(irq, 25, AUX_CHANNEL_B, INTEL_GVT_IRQ_INFO_PCH);
+		SET_BIT_INFO(irq, 26, AUX_CHANNEL_C, INTEL_GVT_IRQ_INFO_PCH);
+		SET_BIT_INFO(irq, 27, AUX_CHANNEL_D, INTEL_GVT_IRQ_INFO_PCH);
+
+		SET_BIT_INFO(irq, 4, PRIMARY_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+		SET_BIT_INFO(irq, 5, SPRITE_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+
+		SET_BIT_INFO(irq, 4, PRIMARY_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+		SET_BIT_INFO(irq, 5, SPRITE_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+
+		SET_BIT_INFO(irq, 4, PRIMARY_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+		SET_BIT_INFO(irq, 5, SPRITE_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+	} else if (IS_SKYLAKE(gvt->dev_priv)) {
+		SET_BIT_INFO(irq, 25, AUX_CHANNEL_B, INTEL_GVT_IRQ_INFO_DE_PORT);
+		SET_BIT_INFO(irq, 26, AUX_CHANNEL_C, INTEL_GVT_IRQ_INFO_DE_PORT);
+		SET_BIT_INFO(irq, 27, AUX_CHANNEL_D, INTEL_GVT_IRQ_INFO_DE_PORT);
+
+		SET_BIT_INFO(irq, 3, PRIMARY_A_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_A);
+		SET_BIT_INFO(irq, 3, PRIMARY_B_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_B);
+		SET_BIT_INFO(irq, 3, PRIMARY_C_FLIP_DONE, INTEL_GVT_IRQ_INFO_DE_PIPE_C);
+	}
+
+	/* GEN8 interrupt PCU events */
+	SET_BIT_INFO(irq, 24, PCU_THERMAL, INTEL_GVT_IRQ_INFO_PCU);
+	SET_BIT_INFO(irq, 25, PCU_PCODE2DRIVER_MAILBOX, INTEL_GVT_IRQ_INFO_PCU);
+}
+
+static struct intel_gvt_irq_ops gen8_irq_ops = {
+	.init_irq = gen8_init_irq,
+	.check_pending_irq = gen8_check_pending_irq,
+};
+
+/**
+ * intel_vgpu_trigger_virtual_event - Trigger a virtual event for a vGPU
+ * @vgpu: a vGPU
+ * @event: interrupt event
+ *
+ * This function is used to trigger a virtual interrupt event for vGPU.
+ * The caller provides the event to be triggered, the framework itself
+ * will emulate the IRQ register bit change.
+ *
+ */
+void intel_vgpu_trigger_virtual_event(struct intel_vgpu *vgpu,
+	enum intel_gvt_event_type event)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_irq *irq = &gvt->irq;
+	gvt_event_virt_handler_t handler;
+	struct intel_gvt_irq_ops *ops = gvt->irq.ops;
+
+	handler = get_event_virt_handler(irq, event);
+	WARN_ON(!handler);
+
+	handler(irq, event, vgpu);
+
+	ops->check_pending_irq(vgpu);
+}
+
+static void init_events(
+	struct intel_gvt_irq *irq)
+{
+	int i;
+
+	for (i = 0; i < INTEL_GVT_EVENT_MAX; i++) {
+		irq->events[i].info = NULL;
+		irq->events[i].v_handler = handle_default_event_virt;
+	}
+}
+
+static enum hrtimer_restart vblank_timer_fn(struct hrtimer *data)
+{
+	struct intel_gvt_vblank_timer *vblank_timer;
+	struct intel_gvt_irq *irq;
+	struct intel_gvt *gvt;
+
+	vblank_timer = container_of(data, struct intel_gvt_vblank_timer, timer);
+	irq = container_of(vblank_timer, struct intel_gvt_irq, vblank_timer);
+	gvt = container_of(irq, struct intel_gvt, irq);
+
+	intel_gvt_request_service(gvt, INTEL_GVT_REQUEST_EMULATE_VBLANK);
+	hrtimer_add_expires_ns(&vblank_timer->timer, vblank_timer->period);
+	return HRTIMER_RESTART;
+}
+
+/**
+ * intel_gvt_clean_irq - clean up GVT-g IRQ emulation subsystem
+ * @gvt: a GVT device
+ *
+ * This function is called at driver unloading stage, to clean up GVT-g IRQ
+ * emulation subsystem.
+ *
+ */
+void intel_gvt_clean_irq(struct intel_gvt *gvt)
+{
+	struct intel_gvt_irq *irq = &gvt->irq;
+
+	hrtimer_cancel(&irq->vblank_timer.timer);
+}
+
+#define VBLNAK_TIMER_PERIOD 16000000
+
+/**
+ * intel_gvt_init_irq - initialize GVT-g IRQ emulation subsystem
+ * @gvt: a GVT device
+ *
+ * This function is called at driver loading stage, to initialize the GVT-g IRQ
+ * emulation subsystem.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_init_irq(struct intel_gvt *gvt)
+{
+	struct intel_gvt_irq *irq = &gvt->irq;
+	struct intel_gvt_vblank_timer *vblank_timer = &irq->vblank_timer;
+
+	gvt_dbg_core("init irq framework\n");
+
+	if (IS_BROADWELL(gvt->dev_priv) || IS_SKYLAKE(gvt->dev_priv)) {
+		irq->ops = &gen8_irq_ops;
+		irq->irq_map = gen8_irq_map;
+	} else {
+		WARN_ON(1);
+		return -ENODEV;
+	}
+
+	/* common event initialization */
+	init_events(irq);
+
+	/* gen specific initialization */
+	irq->ops->init_irq(irq);
+
+	init_irq_map(irq);
+
+	hrtimer_init(&vblank_timer->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
+	vblank_timer->timer.function = vblank_timer_fn;
+	vblank_timer->period = VBLNAK_TIMER_PERIOD;
+
+	return 0;
+}

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

@@ -0,0 +1,233 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Min he <min.he@intel.com>
+ *
+ */
+
+#ifndef _GVT_INTERRUPT_H_
+#define _GVT_INTERRUPT_H_
+
+enum intel_gvt_event_type {
+	RCS_MI_USER_INTERRUPT = 0,
+	RCS_DEBUG,
+	RCS_MMIO_SYNC_FLUSH,
+	RCS_CMD_STREAMER_ERR,
+	RCS_PIPE_CONTROL,
+	RCS_L3_PARITY_ERR,
+	RCS_WATCHDOG_EXCEEDED,
+	RCS_PAGE_DIRECTORY_FAULT,
+	RCS_AS_CONTEXT_SWITCH,
+	RCS_MONITOR_BUFF_HALF_FULL,
+
+	VCS_MI_USER_INTERRUPT,
+	VCS_MMIO_SYNC_FLUSH,
+	VCS_CMD_STREAMER_ERR,
+	VCS_MI_FLUSH_DW,
+	VCS_WATCHDOG_EXCEEDED,
+	VCS_PAGE_DIRECTORY_FAULT,
+	VCS_AS_CONTEXT_SWITCH,
+
+	VCS2_MI_USER_INTERRUPT,
+	VCS2_MI_FLUSH_DW,
+	VCS2_AS_CONTEXT_SWITCH,
+
+	BCS_MI_USER_INTERRUPT,
+	BCS_MMIO_SYNC_FLUSH,
+	BCS_CMD_STREAMER_ERR,
+	BCS_MI_FLUSH_DW,
+	BCS_PAGE_DIRECTORY_FAULT,
+	BCS_AS_CONTEXT_SWITCH,
+
+	VECS_MI_USER_INTERRUPT,
+	VECS_MI_FLUSH_DW,
+	VECS_AS_CONTEXT_SWITCH,
+
+	PIPE_A_FIFO_UNDERRUN,
+	PIPE_B_FIFO_UNDERRUN,
+	PIPE_A_CRC_ERR,
+	PIPE_B_CRC_ERR,
+	PIPE_A_CRC_DONE,
+	PIPE_B_CRC_DONE,
+	PIPE_A_ODD_FIELD,
+	PIPE_B_ODD_FIELD,
+	PIPE_A_EVEN_FIELD,
+	PIPE_B_EVEN_FIELD,
+	PIPE_A_LINE_COMPARE,
+	PIPE_B_LINE_COMPARE,
+	PIPE_C_LINE_COMPARE,
+	PIPE_A_VBLANK,
+	PIPE_B_VBLANK,
+	PIPE_C_VBLANK,
+	PIPE_A_VSYNC,
+	PIPE_B_VSYNC,
+	PIPE_C_VSYNC,
+	PRIMARY_A_FLIP_DONE,
+	PRIMARY_B_FLIP_DONE,
+	PRIMARY_C_FLIP_DONE,
+	SPRITE_A_FLIP_DONE,
+	SPRITE_B_FLIP_DONE,
+	SPRITE_C_FLIP_DONE,
+
+	PCU_THERMAL,
+	PCU_PCODE2DRIVER_MAILBOX,
+
+	DPST_PHASE_IN,
+	DPST_HISTOGRAM,
+	GSE,
+	DP_A_HOTPLUG,
+	AUX_CHANNEL_A,
+	PERF_COUNTER,
+	POISON,
+	GTT_FAULT,
+	ERROR_INTERRUPT_COMBINED,
+
+	FDI_RX_INTERRUPTS_TRANSCODER_A,
+	AUDIO_CP_CHANGE_TRANSCODER_A,
+	AUDIO_CP_REQUEST_TRANSCODER_A,
+	FDI_RX_INTERRUPTS_TRANSCODER_B,
+	AUDIO_CP_CHANGE_TRANSCODER_B,
+	AUDIO_CP_REQUEST_TRANSCODER_B,
+	FDI_RX_INTERRUPTS_TRANSCODER_C,
+	AUDIO_CP_CHANGE_TRANSCODER_C,
+	AUDIO_CP_REQUEST_TRANSCODER_C,
+	ERR_AND_DBG,
+	GMBUS,
+	SDVO_B_HOTPLUG,
+	CRT_HOTPLUG,
+	DP_B_HOTPLUG,
+	DP_C_HOTPLUG,
+	DP_D_HOTPLUG,
+	AUX_CHANNEL_B,
+	AUX_CHANNEL_C,
+	AUX_CHANNEL_D,
+	AUDIO_POWER_STATE_CHANGE_B,
+	AUDIO_POWER_STATE_CHANGE_C,
+	AUDIO_POWER_STATE_CHANGE_D,
+
+	INTEL_GVT_EVENT_RESERVED,
+	INTEL_GVT_EVENT_MAX,
+};
+
+struct intel_gvt_irq;
+struct intel_gvt;
+
+typedef void (*gvt_event_virt_handler_t)(struct intel_gvt_irq *irq,
+	enum intel_gvt_event_type event, struct intel_vgpu *vgpu);
+
+struct intel_gvt_irq_ops {
+	void (*init_irq)(struct intel_gvt_irq *irq);
+	void (*check_pending_irq)(struct intel_vgpu *vgpu);
+};
+
+/* the list of physical interrupt control register groups */
+enum intel_gvt_irq_type {
+	INTEL_GVT_IRQ_INFO_GT,
+	INTEL_GVT_IRQ_INFO_DPY,
+	INTEL_GVT_IRQ_INFO_PCH,
+	INTEL_GVT_IRQ_INFO_PM,
+
+	INTEL_GVT_IRQ_INFO_MASTER,
+	INTEL_GVT_IRQ_INFO_GT0,
+	INTEL_GVT_IRQ_INFO_GT1,
+	INTEL_GVT_IRQ_INFO_GT2,
+	INTEL_GVT_IRQ_INFO_GT3,
+	INTEL_GVT_IRQ_INFO_DE_PIPE_A,
+	INTEL_GVT_IRQ_INFO_DE_PIPE_B,
+	INTEL_GVT_IRQ_INFO_DE_PIPE_C,
+	INTEL_GVT_IRQ_INFO_DE_PORT,
+	INTEL_GVT_IRQ_INFO_DE_MISC,
+	INTEL_GVT_IRQ_INFO_AUD,
+	INTEL_GVT_IRQ_INFO_PCU,
+
+	INTEL_GVT_IRQ_INFO_MAX,
+};
+
+#define INTEL_GVT_IRQ_BITWIDTH	32
+
+/* device specific interrupt bit definitions */
+struct intel_gvt_irq_info {
+	char *name;
+	i915_reg_t reg_base;
+	enum intel_gvt_event_type bit_to_event[INTEL_GVT_IRQ_BITWIDTH];
+	unsigned long warned;
+	int group;
+	DECLARE_BITMAP(downstream_irq_bitmap, INTEL_GVT_IRQ_BITWIDTH);
+	bool has_upstream_irq;
+};
+
+/* per-event information */
+struct intel_gvt_event_info {
+	int bit;				/* map to register bit */
+	int policy;				/* forwarding policy */
+	struct intel_gvt_irq_info *info;	/* register info */
+	gvt_event_virt_handler_t v_handler;	/* for v_event */
+};
+
+struct intel_gvt_irq_map {
+	int up_irq_group;
+	int up_irq_bit;
+	int down_irq_group;
+	u32 down_irq_bitmask;
+};
+
+struct intel_gvt_vblank_timer {
+	struct hrtimer timer;
+	u64 period;
+};
+
+/* structure containing device specific IRQ state */
+struct intel_gvt_irq {
+	struct intel_gvt_irq_ops *ops;
+	struct intel_gvt_irq_info *info[INTEL_GVT_IRQ_INFO_MAX];
+	DECLARE_BITMAP(irq_info_bitmap, INTEL_GVT_IRQ_INFO_MAX);
+	struct intel_gvt_event_info events[INTEL_GVT_EVENT_MAX];
+	DECLARE_BITMAP(pending_events, INTEL_GVT_EVENT_MAX);
+	struct intel_gvt_irq_map *irq_map;
+	struct intel_gvt_vblank_timer vblank_timer;
+};
+
+int intel_gvt_init_irq(struct intel_gvt *gvt);
+void intel_gvt_clean_irq(struct intel_gvt *gvt);
+
+void intel_vgpu_trigger_virtual_event(struct intel_vgpu *vgpu,
+	enum intel_gvt_event_type event);
+
+int intel_vgpu_reg_iir_handler(struct intel_vgpu *vgpu, unsigned int reg,
+	void *p_data, unsigned int bytes);
+int intel_vgpu_reg_ier_handler(struct intel_vgpu *vgpu,
+	unsigned int reg, void *p_data, unsigned int bytes);
+int intel_vgpu_reg_master_irq_handler(struct intel_vgpu *vgpu,
+	unsigned int reg, void *p_data, unsigned int bytes);
+int intel_vgpu_reg_imr_handler(struct intel_vgpu *vgpu,
+	unsigned int reg, void *p_data, unsigned int bytes);
+
+int gvt_ring_id_to_pipe_control_notify_event(int ring_id);
+int gvt_ring_id_to_mi_flush_dw_event(int ring_id);
+int gvt_ring_id_to_mi_user_interrupt_event(int ring_id);
+
+#endif /* _GVT_INTERRUPT_H_ */

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

@@ -0,0 +1,305 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Dexuan Cui
+ *
+ * Contributors:
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Min He <min.he@intel.com>
+ *    Niu Bing <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+/**
+ * intel_vgpu_gpa_to_mmio_offset - translate a GPA to MMIO offset
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa)
+{
+	u64 gttmmio_gpa = *(u64 *)(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_0) &
+			  ~GENMASK(3, 0);
+	return gpa - gttmmio_gpa;
+}
+
+#define reg_is_mmio(gvt, reg)  \
+	(reg >= 0 && reg < gvt->device_info.mmio_size)
+
+#define reg_is_gtt(gvt, reg)   \
+	(reg >= gvt->device_info.gtt_start_offset \
+	 && reg < gvt->device_info.gtt_start_offset + gvt_ggtt_sz(gvt))
+
+/**
+ * intel_vgpu_emulate_mmio_read - emulate MMIO read
+ * @vgpu: a vGPU
+ * @pa: guest physical address
+ * @p_data: data return buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_emulate_mmio_read(void *__vgpu, uint64_t pa,
+		void *p_data, unsigned int bytes)
+{
+	struct intel_vgpu *vgpu = __vgpu;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_mmio_info *mmio;
+	unsigned int offset = 0;
+	int ret = -EINVAL;
+
+	mutex_lock(&gvt->lock);
+
+	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
+		struct intel_vgpu_guest_page *gp;
+
+		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
+		if (gp) {
+			ret = intel_gvt_hypervisor_read_gpa(vgpu, pa,
+					p_data, bytes);
+			if (ret) {
+				gvt_err("vgpu%d: guest page read error %d, "
+					"gfn 0x%lx, pa 0x%llx, var 0x%x, len %d\n",
+					vgpu->id, ret,
+					gp->gfn, pa, *(u32 *)p_data, bytes);
+			}
+			mutex_unlock(&gvt->lock);
+			return ret;
+		}
+	}
+
+	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
+
+	if (WARN_ON(bytes > 8))
+		goto err;
+
+	if (reg_is_gtt(gvt, offset)) {
+		if (WARN_ON(!IS_ALIGNED(offset, 4) && !IS_ALIGNED(offset, 8)))
+			goto err;
+		if (WARN_ON(bytes != 4 && bytes != 8))
+			goto err;
+		if (WARN_ON(!reg_is_gtt(gvt, offset + bytes - 1)))
+			goto err;
+
+		ret = intel_vgpu_emulate_gtt_mmio_read(vgpu, offset,
+				p_data, bytes);
+		if (ret)
+			goto err;
+		mutex_unlock(&gvt->lock);
+		return ret;
+	}
+
+	if (WARN_ON_ONCE(!reg_is_mmio(gvt, offset))) {
+		ret = intel_gvt_hypervisor_read_gpa(vgpu, pa, p_data, bytes);
+		mutex_unlock(&gvt->lock);
+		return ret;
+	}
+
+	if (WARN_ON(!reg_is_mmio(gvt, offset + bytes - 1)))
+		goto err;
+
+	mmio = intel_gvt_find_mmio_info(gvt, rounddown(offset, 4));
+	if (!mmio && !vgpu->mmio.disable_warn_untrack) {
+		gvt_err("vgpu%d: read untracked MMIO %x len %d val %x\n",
+				vgpu->id, offset, bytes, *(u32 *)p_data);
+
+		if (offset == 0x206c) {
+			gvt_err("------------------------------------------\n");
+			gvt_err("vgpu%d: likely triggers a gfx reset\n",
+			vgpu->id);
+			gvt_err("------------------------------------------\n");
+			vgpu->mmio.disable_warn_untrack = true;
+		}
+	}
+
+	if (!intel_gvt_mmio_is_unalign(gvt, offset)) {
+		if (WARN_ON(!IS_ALIGNED(offset, bytes)))
+			goto err;
+	}
+
+	if (mmio) {
+		if (!intel_gvt_mmio_is_unalign(gvt, mmio->offset)) {
+			if (WARN_ON(offset + bytes > mmio->offset + mmio->size))
+				goto err;
+			if (WARN_ON(mmio->offset != offset))
+				goto err;
+		}
+		ret = mmio->read(vgpu, offset, p_data, bytes);
+	} else
+		ret = intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes);
+
+	if (ret)
+		goto err;
+
+	intel_gvt_mmio_set_accessed(gvt, offset);
+	mutex_unlock(&gvt->lock);
+	return 0;
+err:
+	gvt_err("vgpu%d: fail to emulate MMIO read %08x len %d\n",
+			vgpu->id, offset, bytes);
+	mutex_unlock(&gvt->lock);
+	return ret;
+}
+
+/**
+ * intel_vgpu_emulate_mmio_write - emulate MMIO write
+ * @vgpu: a vGPU
+ * @pa: guest physical address
+ * @p_data: write data buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_emulate_mmio_write(void *__vgpu, uint64_t pa,
+		void *p_data, unsigned int bytes)
+{
+	struct intel_vgpu *vgpu = __vgpu;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_mmio_info *mmio;
+	unsigned int offset = 0;
+	u32 old_vreg = 0, old_sreg = 0;
+	int ret = -EINVAL;
+
+	mutex_lock(&gvt->lock);
+
+	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
+		struct intel_vgpu_guest_page *gp;
+
+		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
+		if (gp) {
+			ret = gp->handler(gp, pa, p_data, bytes);
+			if (ret) {
+				gvt_err("vgpu%d: guest page write error %d, "
+					"gfn 0x%lx, pa 0x%llx, var 0x%x, len %d\n",
+					vgpu->id, ret,
+					gp->gfn, pa, *(u32 *)p_data, bytes);
+			}
+			mutex_unlock(&gvt->lock);
+			return ret;
+		}
+	}
+
+	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
+
+	if (WARN_ON(bytes > 8))
+		goto err;
+
+	if (reg_is_gtt(gvt, offset)) {
+		if (WARN_ON(!IS_ALIGNED(offset, 4) && !IS_ALIGNED(offset, 8)))
+			goto err;
+		if (WARN_ON(bytes != 4 && bytes != 8))
+			goto err;
+		if (WARN_ON(!reg_is_gtt(gvt, offset + bytes - 1)))
+			goto err;
+
+		ret = intel_vgpu_emulate_gtt_mmio_write(vgpu, offset,
+				p_data, bytes);
+		if (ret)
+			goto err;
+		mutex_unlock(&gvt->lock);
+		return ret;
+	}
+
+	if (WARN_ON_ONCE(!reg_is_mmio(gvt, offset))) {
+		ret = intel_gvt_hypervisor_write_gpa(vgpu, pa, p_data, bytes);
+		mutex_unlock(&gvt->lock);
+		return ret;
+	}
+
+	mmio = intel_gvt_find_mmio_info(gvt, rounddown(offset, 4));
+	if (!mmio && !vgpu->mmio.disable_warn_untrack)
+		gvt_err("vgpu%d: write untracked MMIO %x len %d val %x\n",
+				vgpu->id, offset, bytes, *(u32 *)p_data);
+
+	if (!intel_gvt_mmio_is_unalign(gvt, offset)) {
+		if (WARN_ON(!IS_ALIGNED(offset, bytes)))
+			goto err;
+	}
+
+	if (mmio) {
+		u64 ro_mask = mmio->ro_mask;
+
+		if (!intel_gvt_mmio_is_unalign(gvt, mmio->offset)) {
+			if (WARN_ON(offset + bytes > mmio->offset + mmio->size))
+				goto err;
+			if (WARN_ON(mmio->offset != offset))
+				goto err;
+		}
+
+		if (intel_gvt_mmio_has_mode_mask(gvt, mmio->offset)) {
+			old_vreg = vgpu_vreg(vgpu, offset);
+			old_sreg = vgpu_sreg(vgpu, offset);
+		}
+
+		if (!ro_mask) {
+			ret = mmio->write(vgpu, offset, p_data, bytes);
+		} else {
+			/* Protect RO bits like HW */
+			u64 data = 0;
+
+			/* all register bits are RO. */
+			if (ro_mask == ~(u64)0) {
+				gvt_err("vgpu%d: try to write RO reg %x\n",
+						vgpu->id, offset);
+				ret = 0;
+				goto out;
+			}
+			/* keep the RO bits in the virtual register */
+			memcpy(&data, p_data, bytes);
+			data &= ~mmio->ro_mask;
+			data |= vgpu_vreg(vgpu, offset) & mmio->ro_mask;
+			ret = mmio->write(vgpu, offset, &data, bytes);
+		}
+
+		/* higher 16bits of mode ctl regs are mask bits for change */
+		if (intel_gvt_mmio_has_mode_mask(gvt, mmio->offset)) {
+			u32 mask = vgpu_vreg(vgpu, offset) >> 16;
+
+			vgpu_vreg(vgpu, offset) = (old_vreg & ~mask)
+				| (vgpu_vreg(vgpu, offset) & mask);
+			vgpu_sreg(vgpu, offset) = (old_sreg & ~mask)
+				| (vgpu_sreg(vgpu, offset) & mask);
+		}
+	} else
+		ret = intel_vgpu_default_mmio_write(vgpu, offset, p_data,
+				bytes);
+	if (ret)
+		goto err;
+out:
+	intel_gvt_mmio_set_accessed(gvt, offset);
+	mutex_unlock(&gvt->lock);
+	return 0;
+err:
+	gvt_err("vgpu%d: fail to emulate MMIO write %08x len %d\n",
+			vgpu->id, offset, bytes);
+	mutex_unlock(&gvt->lock);
+	return ret;
+}

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

@@ -0,0 +1,105 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Ke Yu
+ *    Kevin Tian <kevin.tian@intel.com>
+ *    Dexuan Cui
+ *
+ * Contributors:
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Min He <min.he@intel.com>
+ *    Niu Bing <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef _GVT_MMIO_H_
+#define _GVT_MMIO_H_
+
+struct intel_gvt;
+struct intel_vgpu;
+
+#define D_SNB   (1 << 0)
+#define D_IVB   (1 << 1)
+#define D_HSW   (1 << 2)
+#define D_BDW   (1 << 3)
+#define D_SKL	(1 << 4)
+
+#define D_GEN9PLUS	(D_SKL)
+#define D_GEN8PLUS	(D_BDW | D_SKL)
+#define D_GEN75PLUS	(D_HSW | D_BDW | D_SKL)
+#define D_GEN7PLUS	(D_IVB | D_HSW | D_BDW | D_SKL)
+
+#define D_SKL_PLUS	(D_SKL)
+#define D_BDW_PLUS	(D_BDW | D_SKL)
+#define D_HSW_PLUS	(D_HSW | D_BDW | D_SKL)
+#define D_IVB_PLUS	(D_IVB | D_HSW | D_BDW | D_SKL)
+
+#define D_PRE_BDW	(D_SNB | D_IVB | D_HSW)
+#define D_PRE_SKL	(D_SNB | D_IVB | D_HSW | D_BDW)
+#define D_ALL		(D_SNB | D_IVB | D_HSW | D_BDW | D_SKL)
+
+struct intel_gvt_mmio_info {
+	u32 offset;
+	u32 size;
+	u32 length;
+	u32 addr_mask;
+	u64 ro_mask;
+	u32 device;
+	int (*read)(struct intel_vgpu *, unsigned int, void *, unsigned int);
+	int (*write)(struct intel_vgpu *, unsigned int, void *, unsigned int);
+	u32 addr_range;
+	struct hlist_node node;
+};
+
+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);
+
+struct intel_gvt_mmio_info *intel_gvt_find_mmio_info(struct intel_gvt *gvt,
+						     unsigned int offset);
+#define INTEL_GVT_MMIO_OFFSET(reg) ({ \
+	typeof(reg) __reg = reg; \
+	u32 *offset = (u32 *)&__reg; \
+	*offset; \
+})
+
+int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa);
+int intel_vgpu_emulate_mmio_read(void *__vgpu, u64 pa, void *p_data,
+				 unsigned int bytes);
+int intel_vgpu_emulate_mmio_write(void *__vgpu, u64 pa, void *p_data,
+				  unsigned int bytes);
+bool intel_gvt_mmio_is_cmd_access(struct intel_gvt *gvt,
+				  unsigned int offset);
+bool intel_gvt_mmio_is_unalign(struct intel_gvt *gvt, unsigned int offset);
+void intel_gvt_mmio_set_accessed(struct intel_gvt *gvt, unsigned int offset);
+void intel_gvt_mmio_set_cmd_accessed(struct intel_gvt *gvt,
+				     unsigned int offset);
+bool intel_gvt_mmio_has_mode_mask(struct intel_gvt *gvt, unsigned int offset);
+int intel_vgpu_default_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
+				 void *p_data, unsigned int bytes);
+int intel_vgpu_default_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+				  void *p_data, unsigned int bytes);
+#endif

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

@@ -19,6 +19,15 @@
  * 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.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Dexuan Cui
+ *    Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
  */
 
 #ifndef _GVT_MPT_H_
@@ -46,4 +55,215 @@ static inline int intel_gvt_hypervisor_detect_host(void)
 	return intel_gvt_host.mpt->detect_host();
 }
 
+/**
+ * intel_gvt_hypervisor_attach_vgpu - call hypervisor to initialize vGPU
+ * related stuffs inside hypervisor.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_attach_vgpu(struct intel_vgpu *vgpu)
+{
+	return intel_gvt_host.mpt->attach_vgpu(vgpu, &vgpu->handle);
+}
+
+/**
+ * intel_gvt_hypervisor_detach_vgpu - call hypervisor to release vGPU
+ * related stuffs inside hypervisor.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline void intel_gvt_hypervisor_detach_vgpu(struct intel_vgpu *vgpu)
+{
+	intel_gvt_host.mpt->detach_vgpu(vgpu->handle);
+}
+
+#define MSI_CAP_CONTROL(offset) (offset + 2)
+#define MSI_CAP_ADDRESS(offset) (offset + 4)
+#define MSI_CAP_DATA(offset) (offset + 8)
+#define MSI_CAP_EN 0x1
+
+/**
+ * intel_gvt_hypervisor_inject_msi - inject a MSI interrupt into vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_inject_msi(struct intel_vgpu *vgpu)
+{
+	unsigned long offset = vgpu->gvt->device_info.msi_cap_offset;
+	u16 control, data;
+	u32 addr;
+	int ret;
+
+	control = *(u16 *)(vgpu_cfg_space(vgpu) + MSI_CAP_CONTROL(offset));
+	addr = *(u32 *)(vgpu_cfg_space(vgpu) + MSI_CAP_ADDRESS(offset));
+	data = *(u16 *)(vgpu_cfg_space(vgpu) + MSI_CAP_DATA(offset));
+
+	/* Do not generate MSI if MSIEN is disable */
+	if (!(control & MSI_CAP_EN))
+		return 0;
+
+	if (WARN(control & GENMASK(15, 1), "only support one MSI format\n"))
+		return -EINVAL;
+
+	gvt_dbg_irq("vgpu%d: inject msi address %x data%x\n", vgpu->id, addr,
+		    data);
+
+	ret = intel_gvt_host.mpt->inject_msi(vgpu->handle, addr, data);
+	if (ret)
+		return ret;
+	return 0;
+}
+
+/**
+ * intel_gvt_hypervisor_set_wp_page - translate a host VA into MFN
+ * @p: host kernel virtual address
+ *
+ * Returns:
+ * MFN on success, INTEL_GVT_INVALID_ADDR if failed.
+ */
+static inline unsigned long intel_gvt_hypervisor_virt_to_mfn(void *p)
+{
+	return intel_gvt_host.mpt->from_virt_to_mfn(p);
+}
+
+/**
+ * intel_gvt_hypervisor_set_wp_page - set a guest page to write-protected
+ * @vgpu: a vGPU
+ * @p: intel_vgpu_guest_page
+ *
+ * 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)
+{
+	int ret;
+
+	if (p->writeprotection)
+		return 0;
+
+	ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, p->gfn);
+	if (ret)
+		return ret;
+	p->writeprotection = true;
+	atomic_inc(&vgpu->gtt.n_write_protected_guest_page);
+	return 0;
+}
+
+/**
+ * intel_gvt_hypervisor_unset_wp_page - remove the write-protection of a
+ * guest page
+ * @vgpu: a vGPU
+ * @p: intel_vgpu_guest_page
+ *
+ * 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)
+{
+	int ret;
+
+	if (!p->writeprotection)
+		return 0;
+
+	ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, p->gfn);
+	if (ret)
+		return ret;
+	p->writeprotection = false;
+	atomic_dec(&vgpu->gtt.n_write_protected_guest_page);
+	return 0;
+}
+
+/**
+ * intel_gvt_hypervisor_read_gpa - copy data from GPA to host data buffer
+ * @vgpu: a vGPU
+ * @gpa: guest physical address
+ * @buf: host data buffer
+ * @len: data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_read_gpa(struct intel_vgpu *vgpu,
+		unsigned long gpa, void *buf, unsigned long len)
+{
+	return intel_gvt_host.mpt->read_gpa(vgpu->handle, gpa, buf, len);
+}
+
+/**
+ * intel_gvt_hypervisor_write_gpa - copy data from host data buffer to GPA
+ * @vgpu: a vGPU
+ * @gpa: guest physical address
+ * @buf: host data buffer
+ * @len: data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_write_gpa(struct intel_vgpu *vgpu,
+		unsigned long gpa, void *buf, unsigned long len)
+{
+	return intel_gvt_host.mpt->write_gpa(vgpu->handle, gpa, buf, len);
+}
+
+/**
+ * intel_gvt_hypervisor_gfn_to_mfn - translate a GFN to MFN
+ * @vgpu: a vGPU
+ * @gpfn: guest pfn
+ *
+ * Returns:
+ * MFN on success, INTEL_GVT_INVALID_ADDR if failed.
+ */
+static inline unsigned long intel_gvt_hypervisor_gfn_to_mfn(
+		struct intel_vgpu *vgpu, unsigned long gfn)
+{
+	return intel_gvt_host.mpt->gfn_to_mfn(vgpu->handle, gfn);
+}
+
+enum {
+	GVT_MAP_APERTURE = 0,
+	GVT_MAP_OPREGION,
+};
+
+/**
+ * intel_gvt_hypervisor_map_gfn_to_mfn - map a GFN region to MFN
+ * @vgpu: a vGPU
+ * @gfn: guest PFN
+ * @mfn: host PFN
+ * @nr: amount of PFNs
+ * @map: map or unmap
+ * @type: map type
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_map_gfn_to_mfn(
+		struct intel_vgpu *vgpu, unsigned long gfn,
+		unsigned long mfn, unsigned int nr,
+		bool map, int type)
+{
+	return intel_gvt_host.mpt->map_gfn_to_mfn(vgpu->handle, gfn, mfn, nr,
+						  map, type);
+}
+
+/**
+ * intel_gvt_hypervisor_set_trap_area - Trap a guest PA region
+ * @vgpu: a vGPU
+ * @start: the beginning of the guest physical address region
+ * @end: the end of the guest physical address region
+ * @map: map or unmap
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_set_trap_area(
+		struct intel_vgpu *vgpu, u64 start, u64 end, bool map)
+{
+	return intel_gvt_host.mpt->set_trap_area(vgpu->handle, start, end, map);
+}
+
 #endif /* _GVT_MPT_H_ */

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

@@ -0,0 +1,343 @@
+/*
+ * Copyright(c) 2011-2016 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/acpi.h>
+#include "i915_drv.h"
+
+static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa)
+{
+	void *host_va = vgpu->gvt->opregion.opregion_va;
+	u8 *buf;
+	int i;
+
+	if (WARN((vgpu_opregion(vgpu)->va),
+			"vgpu%d: opregion has been initialized already.\n",
+			vgpu->id))
+		return -EINVAL;
+
+	vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_ATOMIC |
+			GFP_DMA32 | __GFP_ZERO,
+			INTEL_GVT_OPREGION_PORDER);
+
+	if (!vgpu_opregion(vgpu)->va)
+		return -ENOMEM;
+
+	memcpy_fromio(vgpu_opregion(vgpu)->va, host_va,
+			INTEL_GVT_OPREGION_SIZE);
+
+	for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
+		vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
+
+	/* 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;
+
+	return 0;
+}
+
+static int map_vgpu_opregion(struct intel_vgpu *vgpu, bool map)
+{
+	u64 mfn;
+	int i, ret;
+
+	for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++) {
+		mfn = intel_gvt_hypervisor_virt_to_mfn(vgpu_opregion(vgpu)
+			+ i * PAGE_SIZE);
+		if (mfn == INTEL_GVT_INVALID_ADDR) {
+			gvt_err("fail to get MFN from VA\n");
+			return -EINVAL;
+		}
+		ret = intel_gvt_hypervisor_map_gfn_to_mfn(vgpu,
+				vgpu_opregion(vgpu)->gfn[i],
+				mfn, 1, map, GVT_MAP_OPREGION);
+		if (ret) {
+			gvt_err("fail to map GFN to MFN, errno: %d\n", ret);
+			return ret;
+		}
+	}
+	return 0;
+}
+
+/**
+ * intel_vgpu_clean_opregion - clean the stuff used to emulate opregion
+ * @vgpu: a vGPU
+ *
+ */
+void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu)
+{
+	int i;
+
+	gvt_dbg_core("vgpu%d: clean vgpu opregion\n", vgpu->id);
+
+	if (!vgpu_opregion(vgpu)->va)
+		return;
+
+	if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_KVM) {
+		vunmap(vgpu_opregion(vgpu)->va);
+		for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++) {
+			if (vgpu_opregion(vgpu)->pages[i]) {
+				put_page(vgpu_opregion(vgpu)->pages[i]);
+				vgpu_opregion(vgpu)->pages[i] = NULL;
+			}
+		}
+	} else {
+		map_vgpu_opregion(vgpu, false);
+		free_pages((unsigned long)vgpu_opregion(vgpu)->va,
+				INTEL_GVT_OPREGION_PORDER);
+	}
+
+	vgpu_opregion(vgpu)->va = NULL;
+}
+
+/**
+ * intel_vgpu_init_opregion - initialize the stuff used to emulate opregion
+ * @vgpu: a vGPU
+ * @gpa: guest physical address of opregion
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa)
+{
+	int ret;
+
+	gvt_dbg_core("vgpu%d: init vgpu opregion\n", vgpu->id);
+
+	if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_XEN) {
+		gvt_dbg_core("emulate opregion from kernel\n");
+
+		ret = init_vgpu_opregion(vgpu, gpa);
+		if (ret)
+			return ret;
+
+		ret = map_vgpu_opregion(vgpu, true);
+		if (ret)
+			return ret;
+	} else {
+		gvt_dbg_core("emulate opregion from userspace\n");
+
+		/*
+		 * If opregion pages are not allocated from host kenrel,
+		 * most of the params are meaningless
+		 */
+		ret = intel_gvt_hypervisor_map_gfn_to_mfn(vgpu,
+				0, /* not used */
+				0, /* not used */
+				2, /* not used */
+				1,
+				GVT_MAP_OPREGION);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
+/**
+ * intel_gvt_clean_opregion - clean host opergion related stuffs
+ * @gvt: a GVT device
+ *
+ */
+void intel_gvt_clean_opregion(struct intel_gvt *gvt)
+{
+	iounmap(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 = acpi_os_ioremap(gvt->opregion.opregion_pa,
+			INTEL_GVT_OPREGION_SIZE);
+	if (!gvt->opregion.opregion_va) {
+		gvt_err("fail to map host opregion\n");
+		return -EFAULT;
+	}
+	return 0;
+}
+
+#define GVT_OPREGION_FUNC(scic)					\
+	({							\
+	 u32 __ret;						\
+	 __ret = (scic & OPREGION_SCIC_FUNC_MASK) >>		\
+	 OPREGION_SCIC_FUNC_SHIFT;				\
+	 __ret;							\
+	 })
+
+#define GVT_OPREGION_SUBFUNC(scic)				\
+	({							\
+	 u32 __ret;						\
+	 __ret = (scic & OPREGION_SCIC_SUBFUNC_MASK) >>		\
+	 OPREGION_SCIC_SUBFUNC_SHIFT;				\
+	 __ret;							\
+	 })
+
+static const char *opregion_func_name(u32 func)
+{
+	const char *name = NULL;
+
+	switch (func) {
+	case 0 ... 3:
+	case 5:
+	case 7 ... 15:
+		name = "Reserved";
+		break;
+
+	case 4:
+		name = "Get BIOS Data";
+		break;
+
+	case 6:
+		name = "System BIOS Callbacks";
+		break;
+
+	default:
+		name = "Unknown";
+		break;
+	}
+	return name;
+}
+
+static const char *opregion_subfunc_name(u32 subfunc)
+{
+	const char *name = NULL;
+
+	switch (subfunc) {
+	case 0:
+		name = "Supported Calls";
+		break;
+
+	case 1:
+		name = "Requested Callbacks";
+		break;
+
+	case 2 ... 3:
+	case 8 ... 9:
+		name = "Reserved";
+		break;
+
+	case 5:
+		name = "Boot Display";
+		break;
+
+	case 6:
+		name = "TV-Standard/Video-Connector";
+		break;
+
+	case 7:
+		name = "Internal Graphics";
+		break;
+
+	case 10:
+		name = "Spread Spectrum Clocks";
+		break;
+
+	case 11:
+		name = "Get AKSV";
+		break;
+
+	default:
+		name = "Unknown";
+		break;
+	}
+	return name;
+};
+
+static bool querying_capabilities(u32 scic)
+{
+	u32 func, subfunc;
+
+	func = GVT_OPREGION_FUNC(scic);
+	subfunc = GVT_OPREGION_SUBFUNC(scic);
+
+	if ((func == INTEL_GVT_OPREGION_SCIC_F_GETBIOSDATA &&
+		subfunc == INTEL_GVT_OPREGION_SCIC_SF_SUPPRTEDCALLS)
+		|| (func == INTEL_GVT_OPREGION_SCIC_F_GETBIOSDATA &&
+		 subfunc == INTEL_GVT_OPREGION_SCIC_SF_REQEUSTEDCALLBACKS)
+		|| (func == INTEL_GVT_OPREGION_SCIC_F_GETBIOSCALLBACKS &&
+		 subfunc == INTEL_GVT_OPREGION_SCIC_SF_SUPPRTEDCALLS)) {
+		return true;
+	}
+	return false;
+}
+
+/**
+ * intel_vgpu_emulate_opregion_request - emulating OpRegion request
+ * @vgpu: a vGPU
+ * @swsci: SWSCI request
+ *
+ * Returns:
+ * Zero on success, negative error code if failed
+ */
+int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci)
+{
+	u32 *scic, *parm;
+	u32 func, subfunc;
+
+	scic = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_SCIC;
+	parm = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_PARM;
+
+	if (!(swsci & SWSCI_SCI_SELECT)) {
+		gvt_err("vgpu%d: requesting SMI service\n", vgpu->id);
+		return 0;
+	}
+	/* ignore non 0->1 trasitions */
+	if ((vgpu_cfg_space(vgpu)[INTEL_GVT_PCI_SWSCI]
+				& SWSCI_SCI_TRIGGER) ||
+			!(swsci & SWSCI_SCI_TRIGGER)) {
+		return 0;
+	}
+
+	func = GVT_OPREGION_FUNC(*scic);
+	subfunc = GVT_OPREGION_SUBFUNC(*scic);
+	if (!querying_capabilities(*scic)) {
+		gvt_err("vgpu%d: requesting runtime service: func \"%s\","
+				" subfunc \"%s\"\n",
+				vgpu->id,
+				opregion_func_name(func),
+				opregion_subfunc_name(subfunc));
+		/*
+		 * emulate exit status of function call, '0' means
+		 * "failure, generic, unsupported or unknown cause"
+		 */
+		*scic &= ~OPREGION_SCIC_EXIT_MASK;
+		return 0;
+	}
+
+	*scic = 0;
+	*parm = 0;
+	return 0;
+}

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

@@ -0,0 +1,80 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ */
+
+#ifndef _GVT_REG_H
+#define _GVT_REG_H
+
+#define INTEL_GVT_PCI_CLASS_VGA_OTHER   0x80
+
+#define INTEL_GVT_PCI_GMCH_CONTROL	0x50
+#define   BDW_GMCH_GMS_SHIFT		8
+#define   BDW_GMCH_GMS_MASK		0xff
+
+#define INTEL_GVT_PCI_SWSCI		0xe8
+#define   SWSCI_SCI_SELECT		(1 << 15)
+#define   SWSCI_SCI_TRIGGER		1
+
+#define INTEL_GVT_PCI_OPREGION		0xfc
+
+#define INTEL_GVT_OPREGION_CLID		0x1AC
+#define INTEL_GVT_OPREGION_SCIC		0x200
+#define   OPREGION_SCIC_FUNC_MASK	0x1E
+#define   OPREGION_SCIC_FUNC_SHIFT	1
+#define   OPREGION_SCIC_SUBFUNC_MASK	0xFF00
+#define   OPREGION_SCIC_SUBFUNC_SHIFT	8
+#define   OPREGION_SCIC_EXIT_MASK	0xE0
+#define INTEL_GVT_OPREGION_SCIC_F_GETBIOSDATA         4
+#define INTEL_GVT_OPREGION_SCIC_F_GETBIOSCALLBACKS    6
+#define INTEL_GVT_OPREGION_SCIC_SF_SUPPRTEDCALLS      0
+#define INTEL_GVT_OPREGION_SCIC_SF_REQEUSTEDCALLBACKS 1
+#define INTEL_GVT_OPREGION_PARM                   0x204
+
+#define INTEL_GVT_OPREGION_PAGES	2
+#define INTEL_GVT_OPREGION_PORDER	1
+#define INTEL_GVT_OPREGION_SIZE		(2 * 4096)
+
+#define VGT_SPRSTRIDE(pipe)	_PIPE(pipe, _SPRA_STRIDE, _PLANE_STRIDE_2_B)
+
+#define _REG_VECS_EXCC		0x1A028
+#define _REG_VCS2_EXCC		0x1c028
+
+#define _REG_701C0(pipe, plane) (0x701c0 + pipe * 0x1000 + (plane - 1) * 0x100)
+#define _REG_701C4(pipe, plane) (0x701c4 + pipe * 0x1000 + (plane - 1) * 0x100)
+
+#define GFX_MODE_BIT_SET_IN_MASK(val, bit) \
+		((((bit) & 0xffff0000) == 0) && !!((val) & (((bit) << 16))))
+
+#define FORCEWAKE_RENDER_GEN9_REG 0xa278
+#define FORCEWAKE_ACK_RENDER_GEN9_REG 0x0D84
+#define FORCEWAKE_BLITTER_GEN9_REG 0xa188
+#define FORCEWAKE_ACK_BLITTER_GEN9_REG 0x130044
+#define FORCEWAKE_MEDIA_GEN9_REG 0xa270
+#define FORCEWAKE_ACK_MEDIA_GEN9_REG 0x0D88
+#define FORCEWAKE_ACK_HSW_REG 0x130044
+
+#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)
+
+#endif

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

@@ -0,0 +1,290 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+struct render_mmio {
+	int ring_id;
+	i915_reg_t reg;
+	u32 mask;
+	bool in_context;
+	u32 value;
+};
+
+static struct render_mmio gen8_render_mmio_list[] = {
+	{RCS, _MMIO(0x229c), 0xffff, false},
+	{RCS, _MMIO(0x2248), 0x0, false},
+	{RCS, _MMIO(0x2098), 0x0, false},
+	{RCS, _MMIO(0x20c0), 0xffff, true},
+	{RCS, _MMIO(0x24d0), 0, false},
+	{RCS, _MMIO(0x24d4), 0, false},
+	{RCS, _MMIO(0x24d8), 0, false},
+	{RCS, _MMIO(0x24dc), 0, false},
+	{RCS, _MMIO(0x7004), 0xffff, true},
+	{RCS, _MMIO(0x7008), 0xffff, true},
+	{RCS, _MMIO(0x7000), 0xffff, true},
+	{RCS, _MMIO(0x7010), 0xffff, true},
+	{RCS, _MMIO(0x7300), 0xffff, true},
+	{RCS, _MMIO(0x83a4), 0xffff, true},
+
+	{BCS, _MMIO(0x2229c), 0xffff, false},
+	{BCS, _MMIO(0x2209c), 0xffff, false},
+	{BCS, _MMIO(0x220c0), 0xffff, false},
+	{BCS, _MMIO(0x22098), 0x0, false},
+	{BCS, _MMIO(0x22028), 0x0, false},
+};
+
+static struct render_mmio gen9_render_mmio_list[] = {
+	{RCS, _MMIO(0x229c), 0xffff, false},
+	{RCS, _MMIO(0x2248), 0x0, false},
+	{RCS, _MMIO(0x2098), 0x0, false},
+	{RCS, _MMIO(0x20c0), 0xffff, true},
+	{RCS, _MMIO(0x24d0), 0, false},
+	{RCS, _MMIO(0x24d4), 0, false},
+	{RCS, _MMIO(0x24d8), 0, false},
+	{RCS, _MMIO(0x24dc), 0, false},
+	{RCS, _MMIO(0x7004), 0xffff, true},
+	{RCS, _MMIO(0x7008), 0xffff, true},
+	{RCS, _MMIO(0x7000), 0xffff, true},
+	{RCS, _MMIO(0x7010), 0xffff, true},
+	{RCS, _MMIO(0x7300), 0xffff, true},
+	{RCS, _MMIO(0x83a4), 0xffff, true},
+
+	{RCS, _MMIO(0x40e0), 0, false},
+	{RCS, _MMIO(0x40e4), 0, false},
+	{RCS, _MMIO(0x2580), 0xffff, true},
+	{RCS, _MMIO(0x7014), 0xffff, true},
+	{RCS, _MMIO(0x20ec), 0xffff, false},
+	{RCS, _MMIO(0xb118), 0, false},
+	{RCS, _MMIO(0xe100), 0xffff, true},
+	{RCS, _MMIO(0xe180), 0xffff, true},
+	{RCS, _MMIO(0xe184), 0xffff, true},
+	{RCS, _MMIO(0xe188), 0xffff, true},
+	{RCS, _MMIO(0xe194), 0xffff, true},
+	{RCS, _MMIO(0x4de0), 0, false},
+	{RCS, _MMIO(0x4de4), 0, false},
+	{RCS, _MMIO(0x4de8), 0, false},
+	{RCS, _MMIO(0x4dec), 0, false},
+	{RCS, _MMIO(0x4df0), 0, false},
+	{RCS, _MMIO(0x4df4), 0, false},
+
+	{BCS, _MMIO(0x2229c), 0xffff, false},
+	{BCS, _MMIO(0x2209c), 0xffff, false},
+	{BCS, _MMIO(0x220c0), 0xffff, false},
+	{BCS, _MMIO(0x22098), 0x0, false},
+	{BCS, _MMIO(0x22028), 0x0, false},
+
+	{VCS2, _MMIO(0x1c028), 0xffff, false},
+
+	{VECS, _MMIO(0x1a028), 0xffff, false},
+};
+
+static u32 gen9_render_mocs[I915_NUM_ENGINES][64];
+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;
+	i915_reg_t reg;
+	u32 regs[] = {
+		[RCS] = 0x4260,
+		[VCS] = 0x4264,
+		[VCS2] = 0x4268,
+		[BCS] = 0x426c,
+		[VECS] = 0x4270,
+	};
+
+	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
+		return;
+
+	if (!test_and_clear_bit(ring_id, (void *)vgpu->tlb_handle_pending))
+		return;
+
+	reg = _MMIO(regs[ring_id]);
+
+	I915_WRITE(reg, 0x1);
+
+	if (wait_for_atomic((I915_READ(reg) == 0), 50))
+		gvt_err("timeout in invalidate ring (%d) tlb\n", ring_id);
+
+	gvt_dbg_core("invalidate TLB for ring %d\n", ring_id);
+}
+
+static void load_mocs(struct intel_vgpu *vgpu, int ring_id)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	i915_reg_t offset, l3_offset;
+	u32 regs[] = {
+		[RCS] = 0xc800,
+		[VCS] = 0xc900,
+		[VCS2] = 0xca00,
+		[BCS] = 0xcc00,
+		[VECS] = 0xcb00,
+	};
+	int i;
+
+	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
+		return;
+
+	if (!IS_SKYLAKE(dev_priv))
+		return;
+
+	for (i = 0; i < 64; i++) {
+		gen9_render_mocs[ring_id][i] = I915_READ(offset);
+		I915_WRITE(offset, vgpu_vreg(vgpu, offset));
+		POSTING_READ(offset);
+		offset.reg += 4;
+	}
+
+	if (ring_id == RCS) {
+		l3_offset.reg = 0xb020;
+		for (i = 0; i < 32; i++) {
+			gen9_render_mocs_L3[i] = I915_READ(l3_offset);
+			I915_WRITE(l3_offset, vgpu_vreg(vgpu, offset));
+			POSTING_READ(l3_offset);
+			l3_offset.reg += 4;
+		}
+	}
+}
+
+static void restore_mocs(struct intel_vgpu *vgpu, int ring_id)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	i915_reg_t offset, l3_offset;
+	u32 regs[] = {
+		[RCS] = 0xc800,
+		[VCS] = 0xc900,
+		[VCS2] = 0xca00,
+		[BCS] = 0xcc00,
+		[VECS] = 0xcb00,
+	};
+	int i;
+
+	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
+		return;
+
+	if (!IS_SKYLAKE(dev_priv))
+		return;
+
+	for (i = 0; i < 64; i++) {
+		vgpu_vreg(vgpu, offset) = I915_READ(offset);
+		I915_WRITE(offset, gen9_render_mocs[ring_id][i]);
+		POSTING_READ(offset);
+		offset.reg += 4;
+	}
+
+	if (ring_id == RCS) {
+		l3_offset.reg = 0xb020;
+		for (i = 0; i < 32; i++) {
+			vgpu_vreg(vgpu, l3_offset) = I915_READ(l3_offset);
+			I915_WRITE(l3_offset, gen9_render_mocs_L3[i]);
+			POSTING_READ(l3_offset);
+			l3_offset.reg += 4;
+		}
+	}
+}
+
+void intel_gvt_load_render_mmio(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;
+
+	if (IS_SKYLAKE(vgpu->gvt->dev_priv)) {
+		mmio = gen9_render_mmio_list;
+		array_size = ARRAY_SIZE(gen9_render_mmio_list);
+		load_mocs(vgpu, ring_id);
+	} else {
+		mmio = gen8_render_mmio_list;
+		array_size = ARRAY_SIZE(gen8_render_mmio_list);
+	}
+
+	for (i = 0; i < array_size; i++, mmio++) {
+		if (mmio->ring_id != ring_id)
+			continue;
+
+		mmio->value = I915_READ(mmio->reg);
+		if (mmio->mask)
+			v = vgpu_vreg(vgpu, mmio->reg) | (mmio->mask << 16);
+		else
+			v = vgpu_vreg(vgpu, mmio->reg);
+
+		I915_WRITE(mmio->reg, v);
+		POSTING_READ(mmio->reg);
+
+		gvt_dbg_render("load reg %x old %x new %x\n",
+				i915_mmio_reg_offset(mmio->reg),
+				mmio->value, v);
+	}
+	handle_tlb_pending_event(vgpu, ring_id);
+}
+
+void intel_gvt_restore_render_mmio(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;
+
+	if (IS_SKYLAKE(dev_priv)) {
+		mmio = gen9_render_mmio_list;
+		array_size = ARRAY_SIZE(gen9_render_mmio_list);
+		restore_mocs(vgpu, ring_id);
+	} else {
+		mmio = gen8_render_mmio_list;
+		array_size = ARRAY_SIZE(gen8_render_mmio_list);
+	}
+
+	for (i = 0; i < array_size; i++, mmio++) {
+		if (mmio->ring_id != ring_id)
+			continue;
+
+		vgpu_vreg(vgpu, mmio->reg) = I915_READ(mmio->reg);
+
+		if (mmio->mask) {
+			vgpu_vreg(vgpu, mmio->reg) &= ~(mmio->mask << 16);
+			v = mmio->value | (mmio->mask << 16);
+		} else
+			v = mmio->value;
+
+		I915_WRITE(mmio->reg, v);
+		POSTING_READ(mmio->reg);
+
+		gvt_dbg_render("restore reg %x old %x new %x\n",
+				i915_mmio_reg_offset(mmio->reg),
+				mmio->value, v);
+	}
+}

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

@@ -0,0 +1,43 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#ifndef __GVT_RENDER_H__
+#define __GVT_RENDER_H__
+
+void intel_gvt_load_render_mmio(struct intel_vgpu *vgpu, int ring_id);
+
+void intel_gvt_restore_render_mmio(struct intel_vgpu *vgpu, int ring_id);
+
+#endif

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

@@ -0,0 +1,291 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Anhua Xu
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+static bool vgpu_has_pending_workload(struct intel_vgpu *vgpu)
+{
+	struct intel_vgpu_execlist *execlist;
+	int i;
+
+	for (i = 0; i < I915_NUM_ENGINES; i++) {
+		execlist = &vgpu->execlist[i];
+		if (!list_empty(workload_q_head(vgpu, i)))
+			return true;
+	}
+
+	return false;
+}
+
+static void try_to_schedule_next_vgpu(struct intel_gvt *gvt)
+{
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+	int i;
+
+	/* no target to schedule */
+	if (!scheduler->next_vgpu)
+		return;
+
+	gvt_dbg_sched("try to schedule next vgpu %d\n",
+			scheduler->next_vgpu->id);
+
+	/*
+	 * after the flag is set, workload dispatch thread will
+	 * stop dispatching workload for current vgpu
+	 */
+	scheduler->need_reschedule = true;
+
+	/* still have uncompleted workload? */
+	for (i = 0; i < I915_NUM_ENGINES; i++) {
+		if (scheduler->current_workload[i]) {
+			gvt_dbg_sched("still have running workload\n");
+			return;
+		}
+	}
+
+	gvt_dbg_sched("switch to next vgpu %d\n",
+			scheduler->next_vgpu->id);
+
+	/* switch current vgpu */
+	scheduler->current_vgpu = scheduler->next_vgpu;
+	scheduler->next_vgpu = NULL;
+
+	scheduler->need_reschedule = false;
+
+	/* wake up workload dispatch thread */
+	for (i = 0; i < I915_NUM_ENGINES; i++)
+		wake_up(&scheduler->waitq[i]);
+}
+
+struct tbs_vgpu_data {
+	struct list_head list;
+	struct intel_vgpu *vgpu;
+	/* put some per-vgpu sched stats here */
+};
+
+struct tbs_sched_data {
+	struct intel_gvt *gvt;
+	struct delayed_work work;
+	unsigned long period;
+	struct list_head runq_head;
+};
+
+#define GVT_DEFAULT_TIME_SLICE (1 * HZ / 1000)
+
+static void tbs_sched_func(struct work_struct *work)
+{
+	struct tbs_sched_data *sched_data = container_of(work,
+			struct tbs_sched_data, work.work);
+	struct tbs_vgpu_data *vgpu_data;
+
+	struct intel_gvt *gvt = sched_data->gvt;
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+
+	struct intel_vgpu *vgpu = NULL;
+	struct list_head *pos, *head;
+
+	mutex_lock(&gvt->lock);
+
+	/* no vgpu or has already had a target */
+	if (list_empty(&sched_data->runq_head) || scheduler->next_vgpu)
+		goto out;
+
+	if (scheduler->current_vgpu) {
+		vgpu_data = scheduler->current_vgpu->sched_data;
+		head = &vgpu_data->list;
+	} else {
+		gvt_dbg_sched("no current vgpu search from q head\n");
+		head = &sched_data->runq_head;
+	}
+
+	/* search a vgpu with pending workload */
+	list_for_each(pos, head) {
+		if (pos == &sched_data->runq_head)
+			continue;
+
+		vgpu_data = container_of(pos, struct tbs_vgpu_data, list);
+		if (!vgpu_has_pending_workload(vgpu_data->vgpu))
+			continue;
+
+		vgpu = vgpu_data->vgpu;
+		break;
+	}
+
+	if (vgpu) {
+		scheduler->next_vgpu = vgpu;
+		gvt_dbg_sched("pick next vgpu %d\n", vgpu->id);
+	}
+out:
+	if (scheduler->next_vgpu) {
+		gvt_dbg_sched("try to schedule next vgpu %d\n",
+				scheduler->next_vgpu->id);
+		try_to_schedule_next_vgpu(gvt);
+	}
+
+	/*
+	 * still have vgpu on runq
+	 * or last schedule haven't finished due to running workload
+	 */
+	if (!list_empty(&sched_data->runq_head) || scheduler->next_vgpu)
+		schedule_delayed_work(&sched_data->work, sched_data->period);
+
+	mutex_unlock(&gvt->lock);
+}
+
+static int tbs_sched_init(struct intel_gvt *gvt)
+{
+	struct intel_gvt_workload_scheduler *scheduler =
+		&gvt->scheduler;
+
+	struct tbs_sched_data *data;
+
+	data = kzalloc(sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&data->runq_head);
+	INIT_DELAYED_WORK(&data->work, tbs_sched_func);
+	data->period = GVT_DEFAULT_TIME_SLICE;
+	data->gvt = gvt;
+
+	scheduler->sched_data = data;
+	return 0;
+}
+
+static void tbs_sched_clean(struct intel_gvt *gvt)
+{
+	struct intel_gvt_workload_scheduler *scheduler =
+		&gvt->scheduler;
+	struct tbs_sched_data *data = scheduler->sched_data;
+
+	cancel_delayed_work(&data->work);
+	kfree(data);
+	scheduler->sched_data = NULL;
+}
+
+static int tbs_sched_init_vgpu(struct intel_vgpu *vgpu)
+{
+	struct tbs_vgpu_data *data;
+
+	data = kzalloc(sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	data->vgpu = vgpu;
+	INIT_LIST_HEAD(&data->list);
+
+	vgpu->sched_data = data;
+	return 0;
+}
+
+static void tbs_sched_clean_vgpu(struct intel_vgpu *vgpu)
+{
+	kfree(vgpu->sched_data);
+	vgpu->sched_data = NULL;
+}
+
+static void tbs_sched_start_schedule(struct intel_vgpu *vgpu)
+{
+	struct tbs_sched_data *sched_data = vgpu->gvt->scheduler.sched_data;
+	struct tbs_vgpu_data *vgpu_data = vgpu->sched_data;
+
+	if (!list_empty(&vgpu_data->list))
+		return;
+
+	list_add_tail(&vgpu_data->list, &sched_data->runq_head);
+	schedule_delayed_work(&sched_data->work, sched_data->period);
+}
+
+static void tbs_sched_stop_schedule(struct intel_vgpu *vgpu)
+{
+	struct tbs_vgpu_data *vgpu_data = vgpu->sched_data;
+
+	list_del_init(&vgpu_data->list);
+}
+
+struct intel_gvt_sched_policy_ops tbs_schedule_ops = {
+	.init = tbs_sched_init,
+	.clean = tbs_sched_clean,
+	.init_vgpu = tbs_sched_init_vgpu,
+	.clean_vgpu = tbs_sched_clean_vgpu,
+	.start_schedule = tbs_sched_start_schedule,
+	.stop_schedule = tbs_sched_stop_schedule,
+};
+
+int intel_gvt_init_sched_policy(struct intel_gvt *gvt)
+{
+	gvt->scheduler.sched_ops = &tbs_schedule_ops;
+
+	return gvt->scheduler.sched_ops->init(gvt);
+}
+
+void intel_gvt_clean_sched_policy(struct intel_gvt *gvt)
+{
+	gvt->scheduler.sched_ops->clean(gvt);
+}
+
+int intel_vgpu_init_sched_policy(struct intel_vgpu *vgpu)
+{
+	return vgpu->gvt->scheduler.sched_ops->init_vgpu(vgpu);
+}
+
+void intel_vgpu_clean_sched_policy(struct intel_vgpu *vgpu)
+{
+	vgpu->gvt->scheduler.sched_ops->clean_vgpu(vgpu);
+}
+
+void intel_vgpu_start_schedule(struct intel_vgpu *vgpu)
+{
+	gvt_dbg_core("vgpu%d: start schedule\n", vgpu->id);
+
+	vgpu->gvt->scheduler.sched_ops->start_schedule(vgpu);
+}
+
+void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt_workload_scheduler *scheduler =
+		&vgpu->gvt->scheduler;
+
+	gvt_dbg_core("vgpu%d: stop schedule\n", vgpu->id);
+
+	scheduler->sched_ops->stop_schedule(vgpu);
+
+	if (scheduler->next_vgpu == vgpu)
+		scheduler->next_vgpu = NULL;
+
+	if (scheduler->current_vgpu == vgpu) {
+		/* stop workload dispatching */
+		scheduler->need_reschedule = true;
+		scheduler->current_vgpu = NULL;
+	}
+}

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

@@ -0,0 +1,58 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Anhua Xu
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#ifndef __GVT_SCHED_POLICY__
+#define __GVT_SCHED_POLICY__
+
+struct intel_gvt_sched_policy_ops {
+	int (*init)(struct intel_gvt *gvt);
+	void (*clean)(struct intel_gvt *gvt);
+	int (*init_vgpu)(struct intel_vgpu *vgpu);
+	void (*clean_vgpu)(struct intel_vgpu *vgpu);
+	void (*start_schedule)(struct intel_vgpu *vgpu);
+	void (*stop_schedule)(struct intel_vgpu *vgpu);
+};
+
+int intel_gvt_init_sched_policy(struct intel_gvt *gvt);
+
+void intel_gvt_clean_sched_policy(struct intel_gvt *gvt);
+
+int intel_vgpu_init_sched_policy(struct intel_vgpu *vgpu);
+
+void intel_vgpu_clean_sched_policy(struct intel_vgpu *vgpu);
+
+void intel_vgpu_start_schedule(struct intel_vgpu *vgpu);
+
+void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu);
+
+#endif

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

@@ -0,0 +1,572 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Chanbin Du <changbin.du@intel.com>
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+#include <linux/kthread.h>
+
+#define RING_CTX_OFF(x) \
+	offsetof(struct execlist_ring_context, x)
+
+void set_context_pdp_root_pointer(struct execlist_ring_context *ring_context,
+		u32 pdp[8])
+{
+	struct execlist_mmio_pair *pdp_pair = &ring_context->pdp3_UDW;
+	int i;
+
+	for (i = 0; i < 8; i++)
+		pdp_pair[i].val = pdp[7 - i];
+}
+
+static int populate_shadow_context(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_gvt *gvt = vgpu->gvt;
+	int ring_id = workload->ring_id;
+	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
+	struct drm_i915_gem_object *ctx_obj =
+		shadow_ctx->engine[ring_id].state->obj;
+	struct execlist_ring_context *shadow_ring_context;
+	struct page *page;
+	void *dst;
+	unsigned long context_gpa, context_page_num;
+	int i;
+
+	gvt_dbg_sched("ring id %d workload lrca %x", ring_id,
+			workload->ctx_desc.lrca);
+
+	context_page_num = intel_lr_context_size(
+			&gvt->dev_priv->engine[ring_id]);
+
+	context_page_num = context_page_num >> PAGE_SHIFT;
+
+	if (IS_BROADWELL(gvt->dev_priv) && ring_id == RCS)
+		context_page_num = 19;
+
+	i = 2;
+
+	while (i < context_page_num) {
+		context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+				(u32)((workload->ctx_desc.lrca + i) <<
+				GTT_PAGE_SHIFT));
+		if (context_gpa == INTEL_GVT_INVALID_ADDR) {
+			gvt_err("Invalid guest context descriptor\n");
+			return -EINVAL;
+		}
+
+		page = i915_gem_object_get_page(ctx_obj, LRC_PPHWSP_PN + i);
+		dst = kmap_atomic(page);
+		intel_gvt_hypervisor_read_gpa(vgpu, context_gpa, dst,
+				GTT_PAGE_SIZE);
+		kunmap_atomic(dst);
+		i++;
+	}
+
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
+	shadow_ring_context = kmap_atomic(page);
+
+#define COPY_REG(name) \
+	intel_gvt_hypervisor_read_gpa(vgpu, workload->ring_context_gpa \
+		+ RING_CTX_OFF(name.val), &shadow_ring_context->name.val, 4)
+
+	COPY_REG(ctx_ctrl);
+	COPY_REG(ctx_timestamp);
+
+	if (ring_id == RCS) {
+		COPY_REG(bb_per_ctx_ptr);
+		COPY_REG(rcs_indirect_ctx);
+		COPY_REG(rcs_indirect_ctx_offset);
+	}
+#undef COPY_REG
+
+	set_context_pdp_root_pointer(shadow_ring_context,
+				     workload->shadow_mm->shadow_page_table);
+
+	intel_gvt_hypervisor_read_gpa(vgpu,
+			workload->ring_context_gpa +
+			sizeof(*shadow_ring_context),
+			(void *)shadow_ring_context +
+			sizeof(*shadow_ring_context),
+			GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
+
+	kunmap_atomic(shadow_ring_context);
+	return 0;
+}
+
+static int shadow_context_status_change(struct notifier_block *nb,
+		unsigned long action, void *data)
+{
+	struct intel_vgpu *vgpu = container_of(nb,
+			struct intel_vgpu, shadow_ctx_notifier_block);
+	struct drm_i915_gem_request *req =
+		(struct drm_i915_gem_request *)data;
+	struct intel_gvt_workload_scheduler *scheduler =
+		&vgpu->gvt->scheduler;
+	struct intel_vgpu_workload *workload =
+		scheduler->current_workload[req->engine->id];
+
+	switch (action) {
+	case INTEL_CONTEXT_SCHEDULE_IN:
+		intel_gvt_load_render_mmio(workload->vgpu,
+					   workload->ring_id);
+		atomic_set(&workload->shadow_ctx_active, 1);
+		break;
+	case INTEL_CONTEXT_SCHEDULE_OUT:
+		intel_gvt_restore_render_mmio(workload->vgpu,
+					      workload->ring_id);
+		atomic_set(&workload->shadow_ctx_active, 0);
+		break;
+	default:
+		WARN_ON(1);
+		return NOTIFY_OK;
+	}
+	wake_up(&workload->shadow_ctx_status_wq);
+	return NOTIFY_OK;
+}
+
+static int dispatch_workload(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_gvt *gvt = vgpu->gvt;
+	int ring_id = workload->ring_id;
+	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
+	struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv;
+	int ret;
+
+	gvt_dbg_sched("ring id %d prepare to dispatch workload %p\n",
+		ring_id, workload);
+
+	shadow_ctx->desc_template = workload->ctx_desc.addressing_mode <<
+				    GEN8_CTX_ADDRESSING_MODE_SHIFT;
+
+	workload->req = i915_gem_request_alloc(&dev_priv->engine[ring_id],
+					       shadow_ctx);
+	if (IS_ERR_OR_NULL(workload->req)) {
+		gvt_err("fail to allocate gem request\n");
+		workload->status = PTR_ERR(workload->req);
+		workload->req = NULL;
+		return workload->status;
+	}
+
+	gvt_dbg_sched("ring id %d get i915 gem request %p\n",
+			ring_id, workload->req);
+
+	mutex_lock(&gvt->lock);
+
+	ret = intel_gvt_scan_and_shadow_workload(workload);
+	if (ret)
+		goto err;
+
+	ret = intel_gvt_scan_and_shadow_wa_ctx(&workload->wa_ctx);
+	if (ret)
+		goto err;
+
+	ret = populate_shadow_context(workload);
+	if (ret)
+		goto err;
+
+	if (workload->prepare) {
+		ret = workload->prepare(workload);
+		if (ret)
+			goto err;
+	}
+
+	mutex_unlock(&gvt->lock);
+
+	gvt_dbg_sched("ring id %d submit workload to i915 %p\n",
+			ring_id, workload->req);
+
+	i915_add_request_no_flush(workload->req);
+
+	workload->dispatched = true;
+	return 0;
+err:
+	workload->status = ret;
+	if (workload->req)
+		workload->req = NULL;
+
+	mutex_unlock(&gvt->lock);
+	return ret;
+}
+
+static struct intel_vgpu_workload *pick_next_workload(
+		struct intel_gvt *gvt, int ring_id)
+{
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+	struct intel_vgpu_workload *workload = NULL;
+
+	mutex_lock(&gvt->lock);
+
+	/*
+	 * no current vgpu / will be scheduled out / no workload
+	 * bail out
+	 */
+	if (!scheduler->current_vgpu) {
+		gvt_dbg_sched("ring id %d stop - no current vgpu\n", ring_id);
+		goto out;
+	}
+
+	if (scheduler->need_reschedule) {
+		gvt_dbg_sched("ring id %d stop - will reschedule\n", ring_id);
+		goto out;
+	}
+
+	if (list_empty(workload_q_head(scheduler->current_vgpu, ring_id))) {
+		gvt_dbg_sched("ring id %d stop - no available workload\n",
+				ring_id);
+		goto out;
+	}
+
+	/*
+	 * still have current workload, maybe the workload disptacher
+	 * fail to submit it for some reason, resubmit it.
+	 */
+	if (scheduler->current_workload[ring_id]) {
+		workload = scheduler->current_workload[ring_id];
+		gvt_dbg_sched("ring id %d still have current workload %p\n",
+				ring_id, workload);
+		goto out;
+	}
+
+	/*
+	 * pick a workload as current workload
+	 * once current workload is set, schedule policy routines
+	 * will wait the current workload is finished when trying to
+	 * schedule out a vgpu.
+	 */
+	scheduler->current_workload[ring_id] = container_of(
+			workload_q_head(scheduler->current_vgpu, ring_id)->next,
+			struct intel_vgpu_workload, list);
+
+	workload = scheduler->current_workload[ring_id];
+
+	gvt_dbg_sched("ring id %d pick new workload %p\n", ring_id, workload);
+
+	atomic_inc(&workload->vgpu->running_workload_num);
+out:
+	mutex_unlock(&gvt->lock);
+	return workload;
+}
+
+static void update_guest_context(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_gvt *gvt = vgpu->gvt;
+	int ring_id = workload->ring_id;
+	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
+	struct drm_i915_gem_object *ctx_obj =
+		shadow_ctx->engine[ring_id].state->obj;
+	struct execlist_ring_context *shadow_ring_context;
+	struct page *page;
+	void *src;
+	unsigned long context_gpa, context_page_num;
+	int i;
+
+	gvt_dbg_sched("ring id %d workload lrca %x\n", ring_id,
+			workload->ctx_desc.lrca);
+
+	context_page_num = intel_lr_context_size(
+			&gvt->dev_priv->engine[ring_id]);
+
+	context_page_num = context_page_num >> PAGE_SHIFT;
+
+	if (IS_BROADWELL(gvt->dev_priv) && ring_id == RCS)
+		context_page_num = 19;
+
+	i = 2;
+
+	while (i < context_page_num) {
+		context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+				(u32)((workload->ctx_desc.lrca + i) <<
+					GTT_PAGE_SHIFT));
+		if (context_gpa == INTEL_GVT_INVALID_ADDR) {
+			gvt_err("invalid guest context descriptor\n");
+			return;
+		}
+
+		page = i915_gem_object_get_page(ctx_obj, LRC_PPHWSP_PN + i);
+		src = kmap_atomic(page);
+		intel_gvt_hypervisor_write_gpa(vgpu, context_gpa, src,
+				GTT_PAGE_SIZE);
+		kunmap_atomic(src);
+		i++;
+	}
+
+	intel_gvt_hypervisor_write_gpa(vgpu, workload->ring_context_gpa +
+		RING_CTX_OFF(ring_header.val), &workload->rb_tail, 4);
+
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
+	shadow_ring_context = kmap_atomic(page);
+
+#define COPY_REG(name) \
+	intel_gvt_hypervisor_write_gpa(vgpu, workload->ring_context_gpa + \
+		RING_CTX_OFF(name.val), &shadow_ring_context->name.val, 4)
+
+	COPY_REG(ctx_ctrl);
+	COPY_REG(ctx_timestamp);
+
+#undef COPY_REG
+
+	intel_gvt_hypervisor_write_gpa(vgpu,
+			workload->ring_context_gpa +
+			sizeof(*shadow_ring_context),
+			(void *)shadow_ring_context +
+			sizeof(*shadow_ring_context),
+			GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
+
+	kunmap_atomic(shadow_ring_context);
+}
+
+static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
+{
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+	struct intel_vgpu_workload *workload;
+	int event;
+
+	mutex_lock(&gvt->lock);
+
+	workload = scheduler->current_workload[ring_id];
+
+	if (!workload->status && !workload->vgpu->resetting) {
+		wait_event(workload->shadow_ctx_status_wq,
+			   !atomic_read(&workload->shadow_ctx_active));
+
+		update_guest_context(workload);
+
+		for_each_set_bit(event, workload->pending_events,
+				 INTEL_GVT_EVENT_MAX)
+			intel_vgpu_trigger_virtual_event(workload->vgpu,
+					event);
+	}
+
+	gvt_dbg_sched("ring id %d complete workload %p status %d\n",
+			ring_id, workload, workload->status);
+
+	scheduler->current_workload[ring_id] = NULL;
+
+	atomic_dec(&workload->vgpu->running_workload_num);
+
+	list_del_init(&workload->list);
+	workload->complete(workload);
+
+	wake_up(&scheduler->workload_complete_wq);
+	mutex_unlock(&gvt->lock);
+}
+
+struct workload_thread_param {
+	struct intel_gvt *gvt;
+	int ring_id;
+};
+
+static int workload_thread(void *priv)
+{
+	struct workload_thread_param *p = (struct workload_thread_param *)priv;
+	struct intel_gvt *gvt = p->gvt;
+	int ring_id = p->ring_id;
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+	struct intel_vgpu_workload *workload = NULL;
+	int ret;
+	bool need_force_wake = IS_SKYLAKE(gvt->dev_priv);
+
+	kfree(p);
+
+	gvt_dbg_core("workload thread for ring %d started\n", ring_id);
+
+	while (!kthread_should_stop()) {
+		ret = wait_event_interruptible(scheduler->waitq[ring_id],
+				kthread_should_stop() ||
+				(workload = pick_next_workload(gvt, ring_id)));
+
+		WARN_ON_ONCE(ret);
+
+		if (kthread_should_stop())
+			break;
+
+		gvt_dbg_sched("ring id %d next workload %p vgpu %d\n",
+				workload->ring_id, workload,
+				workload->vgpu->id);
+
+		intel_runtime_pm_get(gvt->dev_priv);
+
+		/*
+		 * Always take i915 big lock first
+		 */
+		ret = i915_mutex_lock_interruptible(&gvt->dev_priv->drm);
+		if (ret < 0) {
+			gvt_err("i915 submission is not available, retry\n");
+			schedule_timeout(1);
+			continue;
+		}
+
+		gvt_dbg_sched("ring id %d will dispatch workload %p\n",
+				workload->ring_id, workload);
+
+		if (need_force_wake)
+			intel_uncore_forcewake_get(gvt->dev_priv,
+					FORCEWAKE_ALL);
+
+		ret = dispatch_workload(workload);
+		if (ret) {
+			gvt_err("fail to dispatch workload, skip\n");
+			goto complete;
+		}
+
+		gvt_dbg_sched("ring id %d wait workload %p\n",
+				workload->ring_id, workload);
+
+		workload->status = i915_wait_request(workload->req,
+						     I915_WAIT_INTERRUPTIBLE | I915_WAIT_LOCKED,
+						     NULL, NULL);
+		if (workload->status != 0)
+			gvt_err("fail to wait workload, skip\n");
+
+complete:
+		gvt_dbg_sched("will complete workload %p\n, status: %d\n",
+				workload, workload->status);
+
+		complete_current_workload(gvt, ring_id);
+
+		if (need_force_wake)
+			intel_uncore_forcewake_put(gvt->dev_priv,
+					FORCEWAKE_ALL);
+
+		mutex_unlock(&gvt->dev_priv->drm.struct_mutex);
+
+		intel_runtime_pm_put(gvt->dev_priv);
+	}
+	return 0;
+}
+
+void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+
+	if (atomic_read(&vgpu->running_workload_num)) {
+		gvt_dbg_sched("wait vgpu idle\n");
+
+		wait_event(scheduler->workload_complete_wq,
+				!atomic_read(&vgpu->running_workload_num));
+	}
+}
+
+void intel_gvt_clean_workload_scheduler(struct intel_gvt *gvt)
+{
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+	int i;
+
+	gvt_dbg_core("clean workload scheduler\n");
+
+	for (i = 0; i < I915_NUM_ENGINES; i++) {
+		if (scheduler->thread[i]) {
+			kthread_stop(scheduler->thread[i]);
+			scheduler->thread[i] = NULL;
+		}
+	}
+}
+
+int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt)
+{
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
+	struct workload_thread_param *param = NULL;
+	int ret;
+	int i;
+
+	gvt_dbg_core("init workload scheduler\n");
+
+	init_waitqueue_head(&scheduler->workload_complete_wq);
+
+	for (i = 0; i < I915_NUM_ENGINES; i++) {
+		init_waitqueue_head(&scheduler->waitq[i]);
+
+		param = kzalloc(sizeof(*param), GFP_KERNEL);
+		if (!param) {
+			ret = -ENOMEM;
+			goto err;
+		}
+
+		param->gvt = gvt;
+		param->ring_id = i;
+
+		scheduler->thread[i] = kthread_run(workload_thread, param,
+			"gvt workload %d", i);
+		if (IS_ERR(scheduler->thread[i])) {
+			gvt_err("fail to create workload thread\n");
+			ret = PTR_ERR(scheduler->thread[i]);
+			goto err;
+		}
+	}
+	return 0;
+err:
+	intel_gvt_clean_workload_scheduler(gvt);
+	kfree(param);
+	param = NULL;
+	return ret;
+}
+
+void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	atomic_notifier_chain_unregister(&vgpu->shadow_ctx->status_notifier,
+			&vgpu->shadow_ctx_notifier_block);
+
+	mutex_lock(&dev_priv->drm.struct_mutex);
+
+	/* a little hacky to mark as ctx closed */
+	vgpu->shadow_ctx->closed = true;
+	i915_gem_context_put(vgpu->shadow_ctx);
+
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+}
+
+int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu)
+{
+	atomic_set(&vgpu->running_workload_num, 0);
+
+	vgpu->shadow_ctx = i915_gem_context_create_gvt(
+			&vgpu->gvt->dev_priv->drm);
+	if (IS_ERR(vgpu->shadow_ctx))
+		return PTR_ERR(vgpu->shadow_ctx);
+
+	vgpu->shadow_ctx->engine[RCS].initialised = true;
+
+	vgpu->shadow_ctx_notifier_block.notifier_call =
+		shadow_context_status_change;
+
+	atomic_notifier_chain_register(&vgpu->shadow_ctx->status_notifier,
+				       &vgpu->shadow_ctx_notifier_block);
+	return 0;
+}

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

@@ -0,0 +1,139 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ * Contributors:
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Chanbin Du <changbin.du@intel.com>
+ *    Min He <min.he@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *
+ */
+
+#ifndef _GVT_SCHEDULER_H_
+#define _GVT_SCHEDULER_H_
+
+struct intel_gvt_workload_scheduler {
+	struct intel_vgpu *current_vgpu;
+	struct intel_vgpu *next_vgpu;
+	struct intel_vgpu_workload *current_workload[I915_NUM_ENGINES];
+	bool need_reschedule;
+
+	wait_queue_head_t workload_complete_wq;
+	struct task_struct *thread[I915_NUM_ENGINES];
+	wait_queue_head_t waitq[I915_NUM_ENGINES];
+
+	void *sched_data;
+	struct intel_gvt_sched_policy_ops *sched_ops;
+};
+
+#define INDIRECT_CTX_ADDR_MASK 0xffffffc0
+#define INDIRECT_CTX_SIZE_MASK 0x3f
+struct shadow_indirect_ctx {
+	struct drm_i915_gem_object *obj;
+	unsigned long guest_gma;
+	unsigned long shadow_gma;
+	void *shadow_va;
+	uint32_t size;
+};
+
+#define PER_CTX_ADDR_MASK 0xfffff000
+struct shadow_per_ctx {
+	unsigned long guest_gma;
+	unsigned long shadow_gma;
+};
+
+struct intel_shadow_wa_ctx {
+	struct intel_vgpu_workload *workload;
+	struct shadow_indirect_ctx indirect_ctx;
+	struct shadow_per_ctx per_ctx;
+
+};
+
+struct intel_vgpu_workload {
+	struct intel_vgpu *vgpu;
+	int ring_id;
+	struct drm_i915_gem_request *req;
+	/* if this workload has been dispatched to i915? */
+	bool dispatched;
+	int status;
+
+	struct intel_vgpu_mm *shadow_mm;
+
+	/* different submission model may need different handler */
+	int (*prepare)(struct intel_vgpu_workload *);
+	int (*complete)(struct intel_vgpu_workload *);
+	struct list_head list;
+
+	DECLARE_BITMAP(pending_events, INTEL_GVT_EVENT_MAX);
+	void *shadow_ring_buffer_va;
+
+	/* execlist context information */
+	struct execlist_ctx_descriptor_format ctx_desc;
+	struct execlist_ring_context *ring_context;
+	unsigned long rb_head, rb_tail, rb_ctl, rb_start, rb_len;
+	bool restore_inhibit;
+	struct intel_vgpu_elsp_dwords elsp_dwords;
+	bool emulate_schedule_in;
+	atomic_t shadow_ctx_active;
+	wait_queue_head_t shadow_ctx_status_wq;
+	u64 ring_context_gpa;
+
+	/* shadow batch buffer */
+	struct list_head shadow_bb;
+	struct intel_shadow_wa_ctx wa_ctx;
+};
+
+/* Intel shadow batch buffer is a i915 gem object */
+struct intel_shadow_bb_entry {
+	struct list_head list;
+	struct drm_i915_gem_object *obj;
+	void *va;
+	unsigned long len;
+	void *bb_start_cmd_va;
+};
+
+#define workload_q_head(vgpu, ring_id) \
+	(&(vgpu->workload_q_head[ring_id]))
+
+#define queue_workload(workload) do { \
+	list_add_tail(&workload->list, \
+	workload_q_head(workload->vgpu, workload->ring_id)); \
+	wake_up(&workload->vgpu->gvt-> \
+	scheduler.waitq[workload->ring_id]); \
+} while (0)
+
+int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt);
+
+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);
+
+void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu);
+
+#endif

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

@@ -0,0 +1,286 @@
+/*
+ * Copyright © 2011-2016 Intel Corporation
+ *
+ * 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.
+ *
+ * Authors:
+ *    Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#if !defined(_GVT_TRACE_H_) || defined(TRACE_HEADER_MULTI_READ)
+#define _GVT_TRACE_H_
+
+#include <linux/types.h>
+#include <linux/stringify.h>
+#include <linux/tracepoint.h>
+#include <asm/tsc.h>
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM gvt
+
+TRACE_EVENT(spt_alloc,
+	TP_PROTO(int id, void *spt, int type, unsigned long mfn,
+		unsigned long gpt_gfn),
+
+	TP_ARGS(id, spt, type, mfn, gpt_gfn),
+
+	TP_STRUCT__entry(
+		__field(int, id)
+		__field(void *, spt)
+		__field(int, type)
+		__field(unsigned long, mfn)
+		__field(unsigned long, gpt_gfn)
+		),
+
+	TP_fast_assign(
+		__entry->id = id;
+		__entry->spt = spt;
+		__entry->type = type;
+		__entry->mfn = mfn;
+		__entry->gpt_gfn = gpt_gfn;
+	),
+
+	TP_printk("VM%d [alloc] spt %p type %d mfn 0x%lx gfn 0x%lx\n",
+		__entry->id,
+		__entry->spt,
+		__entry->type,
+		__entry->mfn,
+		__entry->gpt_gfn)
+);
+
+TRACE_EVENT(spt_free,
+	TP_PROTO(int id, void *spt, int type),
+
+	TP_ARGS(id, spt, type),
+
+	TP_STRUCT__entry(
+		__field(int, id)
+		__field(void *, spt)
+		__field(int, type)
+		),
+
+	TP_fast_assign(
+		__entry->id = id;
+		__entry->spt = spt;
+		__entry->type = type;
+	),
+
+	TP_printk("VM%u [free] spt %p type %d\n",
+		__entry->id,
+		__entry->spt,
+		__entry->type)
+);
+
+#define MAX_BUF_LEN 256
+
+TRACE_EVENT(gma_index,
+	TP_PROTO(const char *prefix, unsigned long gma,
+		unsigned long index),
+
+	TP_ARGS(prefix, gma, index),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+		snprintf(__entry->buf, MAX_BUF_LEN,
+			"%s gma 0x%lx index 0x%lx\n", prefix, gma, index);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(gma_translate,
+	TP_PROTO(int id, char *type, int ring_id, int pt_level,
+		unsigned long gma, unsigned long gpa),
+
+	TP_ARGS(id, type, ring_id, pt_level, gma, gpa),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+		snprintf(__entry->buf, MAX_BUF_LEN,
+			"VM%d %s ring %d pt_level %d gma 0x%lx -> gpa 0x%lx\n",
+				id, type, ring_id, pt_level, gma, gpa);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(spt_refcount,
+	TP_PROTO(int id, char *action, void *spt, int before, int after),
+
+	TP_ARGS(id, action, spt, before, after),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+		snprintf(__entry->buf, MAX_BUF_LEN,
+			"VM%d [%s] spt %p before %d -> after %d\n",
+				id, action, spt, before, after);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(spt_change,
+	TP_PROTO(int id, char *action, void *spt, unsigned long gfn,
+		int type),
+
+	TP_ARGS(id, action, spt, gfn, type),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+		snprintf(__entry->buf, MAX_BUF_LEN,
+			"VM%d [%s] spt %p gfn 0x%lx type %d\n",
+				id, action, spt, gfn, type);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(gpt_change,
+	TP_PROTO(int id, const char *tag, void *spt, int type, u64 v,
+		unsigned long index),
+
+	TP_ARGS(id, tag, spt, type, v, index),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+		snprintf(__entry->buf, MAX_BUF_LEN,
+		"VM%d [%s] spt %p type %d entry 0x%llx index 0x%lx\n",
+			id, tag, spt, type, v, index);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(oos_change,
+	TP_PROTO(int id, const char *tag, int page_id, void *gpt, int type),
+
+	TP_ARGS(id, tag, page_id, gpt, type),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+		snprintf(__entry->buf, MAX_BUF_LEN,
+		"VM%d [oos %s] page id %d gpt %p type %d\n",
+			id, tag, page_id, gpt, type);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+TRACE_EVENT(oos_sync,
+	TP_PROTO(int id, int page_id, void *gpt, int type, u64 v,
+		unsigned long index),
+
+	TP_ARGS(id, page_id, gpt, type, v, index),
+
+	TP_STRUCT__entry(
+		__array(char, buf, MAX_BUF_LEN)
+	),
+
+	TP_fast_assign(
+	snprintf(__entry->buf, MAX_BUF_LEN,
+	"VM%d [oos sync] page id %d gpt %p type %d entry 0x%llx index 0x%lx\n",
+				id, page_id, gpt, type, v, index);
+	),
+
+	TP_printk("%s", __entry->buf)
+);
+
+#define MAX_CMD_STR_LEN	256
+TRACE_EVENT(gvt_command,
+		TP_PROTO(u8 vm_id, u8 ring_id, u32 ip_gma, u32 *cmd_va, u32 cmd_len, bool ring_buffer_cmd, cycles_t cost_pre_cmd_handler, cycles_t cost_cmd_handler),
+
+		TP_ARGS(vm_id, ring_id, ip_gma, cmd_va, cmd_len, ring_buffer_cmd, cost_pre_cmd_handler, cost_cmd_handler),
+
+		TP_STRUCT__entry(
+			__field(u8, vm_id)
+			__field(u8, ring_id)
+			__field(int, i)
+			__array(char, tmp_buf, MAX_CMD_STR_LEN)
+			__array(char, cmd_str, MAX_CMD_STR_LEN)
+			),
+
+		TP_fast_assign(
+			__entry->vm_id = vm_id;
+			__entry->ring_id = ring_id;
+			__entry->cmd_str[0] = '\0';
+			snprintf(__entry->tmp_buf, MAX_CMD_STR_LEN, "VM(%d) Ring(%d): %s ip(%08x) pre handler cost (%llu), handler cost (%llu) ", vm_id, ring_id, ring_buffer_cmd ? "RB":"BB", ip_gma, cost_pre_cmd_handler, cost_cmd_handler);
+			strcat(__entry->cmd_str, __entry->tmp_buf);
+			entry->i = 0;
+			while (cmd_len > 0) {
+				if (cmd_len >= 8) {
+					snprintf(__entry->tmp_buf, MAX_CMD_STR_LEN, "%08x %08x %08x %08x %08x %08x %08x %08x ",
+						cmd_va[__entry->i], cmd_va[__entry->i+1], cmd_va[__entry->i+2], cmd_va[__entry->i+3],
+						cmd_va[__entry->i+4], cmd_va[__entry->i+5], cmd_va[__entry->i+6], cmd_va[__entry->i+7]);
+					__entry->i += 8;
+					cmd_len -= 8;
+					strcat(__entry->cmd_str, __entry->tmp_buf);
+				} else if (cmd_len >= 4) {
+					snprintf(__entry->tmp_buf, MAX_CMD_STR_LEN, "%08x %08x %08x %08x ",
+						cmd_va[__entry->i], cmd_va[__entry->i+1], cmd_va[__entry->i+2], cmd_va[__entry->i+3]);
+					__entry->i += 4;
+					cmd_len -= 4;
+					strcat(__entry->cmd_str, __entry->tmp_buf);
+				} else if (cmd_len >= 2) {
+					snprintf(__entry->tmp_buf, MAX_CMD_STR_LEN, "%08x %08x ", cmd_va[__entry->i], cmd_va[__entry->i+1]);
+					__entry->i += 2;
+					cmd_len -= 2;
+					strcat(__entry->cmd_str, __entry->tmp_buf);
+				} else if (cmd_len == 1) {
+					snprintf(__entry->tmp_buf, MAX_CMD_STR_LEN, "%08x ", cmd_va[__entry->i]);
+					__entry->i += 1;
+					cmd_len -= 1;
+					strcat(__entry->cmd_str, __entry->tmp_buf);
+				}
+			}
+			strcat(__entry->cmd_str, "\n");
+		),
+
+		TP_printk("%s", __entry->cmd_str)
+);
+#endif /* _GVT_TRACE_H_ */
+
+/* This part must be out of protection */
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH .
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_FILE trace
+#include <trace/define_trace.h>

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

@@ -0,0 +1,36 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Jike Song <jike.song@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *
+ */
+
+#include "trace.h"
+
+#ifndef __CHECKER__
+#define CREATE_TRACE_POINTS
+#include "trace.h"
+#endif

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

@@ -0,0 +1,272 @@
+/*
+ * Copyright(c) 2011-2016 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.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+
+static void clean_vgpu_mmio(struct intel_vgpu *vgpu)
+{
+	vfree(vgpu->mmio.vreg);
+	vgpu->mmio.vreg = vgpu->mmio.sreg = NULL;
+}
+
+static int setup_vgpu_mmio(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	const struct intel_gvt_device_info *info = &gvt->device_info;
+
+	vgpu->mmio.vreg = vzalloc(info->mmio_size * 2);
+	if (!vgpu->mmio.vreg)
+		return -ENOMEM;
+
+	vgpu->mmio.sreg = vgpu->mmio.vreg + info->mmio_size;
+
+	memcpy(vgpu->mmio.vreg, gvt->firmware.mmio, info->mmio_size);
+	memcpy(vgpu->mmio.sreg, gvt->firmware.mmio, info->mmio_size);
+
+	vgpu_vreg(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0;
+
+	/* set the bit 0:2(Core C-State ) to C0 */
+	vgpu_vreg(vgpu, GEN6_GT_CORE_STATUS) = 0;
+	return 0;
+}
+
+static void setup_vgpu_cfg_space(struct intel_vgpu *vgpu,
+	struct intel_vgpu_creation_params *param)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	const struct intel_gvt_device_info *info = &gvt->device_info;
+	u16 *gmch_ctl;
+	int i;
+
+	memcpy(vgpu_cfg_space(vgpu), gvt->firmware.cfg_space,
+	       info->cfg_space_size);
+
+	if (!param->primary) {
+		vgpu_cfg_space(vgpu)[PCI_CLASS_DEVICE] =
+			INTEL_GVT_PCI_CLASS_VGA_OTHER;
+		vgpu_cfg_space(vgpu)[PCI_CLASS_PROG] =
+			INTEL_GVT_PCI_CLASS_VGA_OTHER;
+	}
+
+	/* Show guest that there isn't any stolen memory.*/
+	gmch_ctl = (u16 *)(vgpu_cfg_space(vgpu) + INTEL_GVT_PCI_GMCH_CONTROL);
+	*gmch_ctl &= ~(BDW_GMCH_GMS_MASK << BDW_GMCH_GMS_SHIFT);
+
+	intel_vgpu_write_pci_bar(vgpu, PCI_BASE_ADDRESS_2,
+				 gvt_aperture_pa_base(gvt), true);
+
+	vgpu_cfg_space(vgpu)[PCI_COMMAND] &= ~(PCI_COMMAND_IO
+					     | PCI_COMMAND_MEMORY
+					     | PCI_COMMAND_MASTER);
+	/*
+	 * Clear the bar upper 32bit and let guest to assign the new value
+	 */
+	memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_1, 0, 4);
+	memset(vgpu_cfg_space(vgpu) + PCI_BASE_ADDRESS_3, 0, 4);
+
+	for (i = 0; i < INTEL_GVT_MAX_BAR_NUM; i++) {
+		vgpu->cfg_space.bar[i].size = pci_resource_len(
+					      gvt->dev_priv->drm.pdev, i * 2);
+		vgpu->cfg_space.bar[i].tracked = false;
+	}
+}
+
+static void populate_pvinfo_page(struct intel_vgpu *vgpu)
+{
+	/* setup the ballooning information */
+	vgpu_vreg64(vgpu, vgtif_reg(magic)) = VGT_MAGIC;
+	vgpu_vreg(vgpu, vgtif_reg(version_major)) = 1;
+	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(avail_rs.mappable_gmadr.base)) =
+		vgpu_aperture_gmadr_base(vgpu);
+	vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) =
+		vgpu_aperture_sz(vgpu);
+	vgpu_vreg(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.base)) =
+		vgpu_hidden_gmadr_base(vgpu);
+	vgpu_vreg(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.size)) =
+		vgpu_hidden_sz(vgpu);
+
+	vgpu_vreg(vgpu, vgtif_reg(avail_rs.fence_num)) = vgpu_fence_sz(vgpu);
+
+	gvt_dbg_core("Populate PVINFO PAGE for vGPU %d\n", vgpu->id);
+	gvt_dbg_core("aperture base [GMADR] 0x%llx size 0x%llx\n",
+		vgpu_aperture_gmadr_base(vgpu), vgpu_aperture_sz(vgpu));
+	gvt_dbg_core("hidden base [GMADR] 0x%llx size=0x%llx\n",
+		vgpu_hidden_gmadr_base(vgpu), vgpu_hidden_sz(vgpu));
+	gvt_dbg_core("fence size %d\n", vgpu_fence_sz(vgpu));
+
+	WARN_ON(sizeof(struct vgt_if) != VGT_PVINFO_SIZE);
+}
+
+/**
+ * intel_gvt_destroy_vgpu - destroy a virtual GPU
+ * @vgpu: virtual GPU
+ *
+ * This function is called when user wants to destroy a virtual GPU.
+ *
+ */
+void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+
+	mutex_lock(&gvt->lock);
+
+	vgpu->active = false;
+	idr_remove(&gvt->vgpu_idr, vgpu->id);
+
+	if (atomic_read(&vgpu->running_workload_num)) {
+		mutex_unlock(&gvt->lock);
+		intel_gvt_wait_vgpu_idle(vgpu);
+		mutex_lock(&gvt->lock);
+	}
+
+	intel_vgpu_stop_schedule(vgpu);
+	intel_vgpu_clean_sched_policy(vgpu);
+	intel_vgpu_clean_gvt_context(vgpu);
+	intel_vgpu_clean_execlist(vgpu);
+	intel_vgpu_clean_display(vgpu);
+	intel_vgpu_clean_opregion(vgpu);
+	intel_vgpu_clean_gtt(vgpu);
+	intel_gvt_hypervisor_detach_vgpu(vgpu);
+	intel_vgpu_free_resource(vgpu);
+	clean_vgpu_mmio(vgpu);
+	vfree(vgpu);
+
+	mutex_unlock(&gvt->lock);
+}
+
+/**
+ * intel_gvt_create_vgpu - create a virtual GPU
+ * @gvt: GVT device
+ * @param: vGPU creation parameters
+ *
+ * This function is called when user wants to create a virtual GPU.
+ *
+ * Returns:
+ * pointer to intel_vgpu, error pointer if failed.
+ */
+struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
+		struct intel_vgpu_creation_params *param)
+{
+	struct intel_vgpu *vgpu;
+	int ret;
+
+	gvt_dbg_core("handle %llu low %llu MB high %llu MB fence %llu\n",
+			param->handle, param->low_gm_sz, param->high_gm_sz,
+			param->fence_sz);
+
+	vgpu = vzalloc(sizeof(*vgpu));
+	if (!vgpu)
+		return ERR_PTR(-ENOMEM);
+
+	mutex_lock(&gvt->lock);
+
+	ret = idr_alloc(&gvt->vgpu_idr, vgpu, 1, GVT_MAX_VGPU, GFP_KERNEL);
+	if (ret < 0)
+		goto out_free_vgpu;
+
+	vgpu->id = ret;
+	vgpu->handle = param->handle;
+	vgpu->gvt = gvt;
+	bitmap_zero(vgpu->tlb_handle_pending, I915_NUM_ENGINES);
+
+	setup_vgpu_cfg_space(vgpu, param);
+
+	ret = setup_vgpu_mmio(vgpu);
+	if (ret)
+		goto out_free_vgpu;
+
+	ret = intel_vgpu_alloc_resource(vgpu, param);
+	if (ret)
+		goto out_clean_vgpu_mmio;
+
+	populate_pvinfo_page(vgpu);
+
+	ret = intel_gvt_hypervisor_attach_vgpu(vgpu);
+	if (ret)
+		goto out_clean_vgpu_resource;
+
+	ret = intel_vgpu_init_gtt(vgpu);
+	if (ret)
+		goto out_detach_hypervisor_vgpu;
+
+	if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_KVM) {
+		ret = intel_vgpu_init_opregion(vgpu, 0);
+		if (ret)
+			goto out_clean_gtt;
+	}
+
+	ret = intel_vgpu_init_display(vgpu);
+	if (ret)
+		goto out_clean_opregion;
+
+	ret = intel_vgpu_init_execlist(vgpu);
+	if (ret)
+		goto out_clean_display;
+
+	ret = intel_vgpu_init_gvt_context(vgpu);
+	if (ret)
+		goto out_clean_execlist;
+
+	ret = intel_vgpu_init_sched_policy(vgpu);
+	if (ret)
+		goto out_clean_shadow_ctx;
+
+	vgpu->active = true;
+	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_display:
+	intel_vgpu_clean_display(vgpu);
+out_clean_opregion:
+	intel_vgpu_clean_opregion(vgpu);
+out_clean_gtt:
+	intel_vgpu_clean_gtt(vgpu);
+out_detach_hypervisor_vgpu:
+	intel_gvt_hypervisor_detach_vgpu(vgpu);
+out_clean_vgpu_resource:
+	intel_vgpu_free_resource(vgpu);
+out_clean_vgpu_mmio:
+	clean_vgpu_mmio(vgpu);
+out_free_vgpu:
+	vfree(vgpu);
+	mutex_unlock(&gvt->lock);
+	return ERR_PTR(ret);
+}

drivers/gpu/drm/i915/intel_gvt.c

@@ -39,6 +39,8 @@ static bool is_supported_device(struct drm_i915_private *dev_priv)
 {
 	if (IS_BROADWELL(dev_priv))
 		return true;
+	if (IS_SKYLAKE(dev_priv))
+		return true;
 	return false;
 }
 

drivers/gpu/drm/i915/intel_gvt.h

@@ -24,6 +24,7 @@
 #ifndef _INTEL_GVT_H_
 #define _INTEL_GVT_H_
 
+#include "i915_pvinfo.h"
 #include "gvt/gvt.h"
 
 #ifdef CONFIG_DRM_I915_GVT