drm/i915/gvt: Add emulation for BAR2 (aperture) with normal file RW approach

For vfio-pci, if the region support MMAP then it should support both
mmap and normal file access. The user-space is free to choose which is
being used. For qemu, we just need add 'x-no-mmap=on' for vfio-pci
option.

Currently GVTg only support MMAP for BAR2. So GVTg will not work when
user turn on x-no-mmap option.

This patch added file style access for BAR2, aka the GPU aperture. We
map the entire aperture partition of active vGPU to kernel space when
guest driver try to enable PCI Memory Space. Then we redirect the file
RW operation from kvmgt to this mapped area.

Link: https://bugzilla.redhat.com/show_bug.cgi?id=1458032
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

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

@@ -110,13 +110,25 @@ int intel_vgpu_emulate_cfg_read(struct intel_vgpu *vgpu, unsigned int offset,
 
 static int map_aperture(struct intel_vgpu *vgpu, bool map)
 {
-	u64 first_gfn, first_mfn;
+	phys_addr_t aperture_pa = vgpu_aperture_pa_base(vgpu);
+	unsigned long aperture_sz = vgpu_aperture_sz(vgpu);
+	u64 first_gfn;
 	u64 val;
 	int ret;
 
 	if (map == vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked)
 		return 0;
 
+	if (map) {
+		vgpu->gm.aperture_va = memremap(aperture_pa, aperture_sz,
+						MEMREMAP_WC);
+		if (!vgpu->gm.aperture_va)
+			return -ENOMEM;
+	} else {
+		memunmap(vgpu->gm.aperture_va);
+		vgpu->gm.aperture_va = NULL;
+	}
+
 	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);
@@ -124,14 +136,16 @@ static int map_aperture(struct intel_vgpu *vgpu, bool map)
 		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);
-	if (ret)
+						  aperture_pa >> PAGE_SHIFT,
+						  aperture_sz >> PAGE_SHIFT,
+						  map);
+	if (ret) {
+		memunmap(vgpu->gm.aperture_va);
+		vgpu->gm.aperture_va = NULL;
 		return ret;
+	}
 
 	vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].tracked = map;
 	return 0;

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

@@ -80,6 +80,7 @@ struct intel_gvt_device_info {
 struct intel_vgpu_gm {
 	u64 aperture_sz;
 	u64 hidden_sz;
+	void *aperture_va;
 	struct drm_mm_node low_gm_node;
 	struct drm_mm_node high_gm_node;
 };
@@ -474,6 +475,13 @@ int intel_vgpu_emulate_cfg_read(struct intel_vgpu *vgpu, unsigned int offset,
 int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes);
 
+static inline u64 intel_vgpu_get_bar_gpa(struct intel_vgpu *vgpu, int bar)
+{
+	/* We are 64bit bar. */
+	return (*(u64 *)(vgpu->cfg_space.virtual_cfg_space + bar)) &
+			PCI_BASE_ADDRESS_MEM_MASK;
+}
+
 void intel_gvt_clean_opregion(struct intel_gvt *gvt);
 int intel_gvt_init_opregion(struct intel_gvt *gvt);
 

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

@@ -609,21 +609,20 @@ static void intel_vgpu_release_work(struct work_struct *work)
 	__intel_vgpu_release(vgpu);
 }
 
-static uint64_t intel_vgpu_get_bar0_addr(struct intel_vgpu *vgpu)
+static uint64_t intel_vgpu_get_bar_addr(struct intel_vgpu *vgpu, int bar)
 {
 	u32 start_lo, start_hi;
 	u32 mem_type;
-	int pos = PCI_BASE_ADDRESS_0;
 
-	start_lo = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space + pos)) &
+	start_lo = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space + bar)) &
 			PCI_BASE_ADDRESS_MEM_MASK;
-	mem_type = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space + pos)) &
+	mem_type = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space + bar)) &
 			PCI_BASE_ADDRESS_MEM_TYPE_MASK;
 
 	switch (mem_type) {
 	case PCI_BASE_ADDRESS_MEM_TYPE_64:
 		start_hi = (*(u32 *)(vgpu->cfg_space.virtual_cfg_space
-						+ pos + 4));
+						+ bar + 4));
 		break;
 	case PCI_BASE_ADDRESS_MEM_TYPE_32:
 	case PCI_BASE_ADDRESS_MEM_TYPE_1M:
@@ -637,6 +636,21 @@ static uint64_t intel_vgpu_get_bar0_addr(struct intel_vgpu *vgpu)
 	return ((u64)start_hi << 32) | start_lo;
 }
 
+static int intel_vgpu_bar_rw(struct intel_vgpu *vgpu, int bar, uint64_t off,
+			     void *buf, unsigned int count, bool is_write)
+{
+	uint64_t bar_start = intel_vgpu_get_bar_addr(vgpu, bar);
+	int ret;
+
+	if (is_write)
+		ret = intel_gvt_ops->emulate_mmio_write(vgpu,
+					bar_start + off, buf, count);
+	else
+		ret = intel_gvt_ops->emulate_mmio_read(vgpu,
+					bar_start + off, buf, count);
+	return ret;
+}
+
 static ssize_t intel_vgpu_rw(struct mdev_device *mdev, char *buf,
 			size_t count, loff_t *ppos, bool is_write)
 {
@@ -661,20 +675,14 @@ static ssize_t intel_vgpu_rw(struct mdev_device *mdev, char *buf,
 						buf, count);
 		break;
 	case VFIO_PCI_BAR0_REGION_INDEX:
-		if (is_write) {
-			uint64_t bar0_start = intel_vgpu_get_bar0_addr(vgpu);
-
-			ret = intel_gvt_ops->emulate_mmio_write(vgpu,
-						bar0_start + pos, buf, count);
-		} else {
-			uint64_t bar0_start = intel_vgpu_get_bar0_addr(vgpu);
-
-			ret = intel_gvt_ops->emulate_mmio_read(vgpu,
-						bar0_start + pos, buf, count);
-		}
+		ret = intel_vgpu_bar_rw(vgpu, PCI_BASE_ADDRESS_0, pos,
+					buf, count, is_write);
 		break;
-	case VFIO_PCI_BAR1_REGION_INDEX:
 	case VFIO_PCI_BAR2_REGION_INDEX:
+		ret = intel_vgpu_bar_rw(vgpu, PCI_BASE_ADDRESS_2, pos,
+					buf, count, is_write);
+		break;
+	case VFIO_PCI_BAR1_REGION_INDEX:
 	case VFIO_PCI_BAR3_REGION_INDEX:
 	case VFIO_PCI_BAR4_REGION_INDEX:
 	case VFIO_PCI_BAR5_REGION_INDEX:

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

@@ -45,8 +45,7 @@
  */
 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);
+	u64 gttmmio_gpa = intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_0);
 	return gpa - gttmmio_gpa;
 }
 
@@ -57,6 +56,38 @@ int intel_vgpu_gpa_to_mmio_offset(struct intel_vgpu *vgpu, u64 gpa)
 	(reg >= gvt->device_info.gtt_start_offset \
 	 && reg < gvt->device_info.gtt_start_offset + gvt_ggtt_sz(gvt))
 
+static bool vgpu_gpa_is_aperture(struct intel_vgpu *vgpu, uint64_t gpa)
+{
+	u64 aperture_gpa = intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_2);
+	u64 aperture_sz = vgpu_aperture_sz(vgpu);
+
+	return gpa >= aperture_gpa && gpa < aperture_gpa + aperture_sz;
+}
+
+static int vgpu_aperture_rw(struct intel_vgpu *vgpu, uint64_t gpa,
+			    void *pdata, unsigned int size, bool is_read)
+{
+	u64 aperture_gpa = intel_vgpu_get_bar_gpa(vgpu, PCI_BASE_ADDRESS_2);
+	u64 offset = gpa - aperture_gpa;
+
+	if (!vgpu_gpa_is_aperture(vgpu, gpa + size - 1)) {
+		gvt_vgpu_err("Aperture rw out of range, offset %llx, size %d\n",
+			     offset, size);
+		return -EINVAL;
+	}
+
+	if (!vgpu->gm.aperture_va) {
+		gvt_vgpu_err("BAR is not enabled\n");
+		return -ENXIO;
+	}
+
+	if (is_read)
+		memcpy(pdata, vgpu->gm.aperture_va + offset, size);
+	else
+		memcpy(vgpu->gm.aperture_va + offset, pdata, size);
+	return 0;
+}
+
 static void failsafe_emulate_mmio_rw(struct intel_vgpu *vgpu, uint64_t pa,
 		void *p_data, unsigned int bytes, bool read)
 {
@@ -133,6 +164,12 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 	}
 	mutex_lock(&gvt->lock);
 
+	if (vgpu_gpa_is_aperture(vgpu, pa)) {
+		ret = vgpu_aperture_rw(vgpu, pa, p_data, bytes, true);
+		mutex_unlock(&gvt->lock);
+		return ret;
+	}
+
 	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
 		struct intel_vgpu_guest_page *gp;
 
@@ -224,6 +261,12 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 
 	mutex_lock(&gvt->lock);
 
+	if (vgpu_gpa_is_aperture(vgpu, pa)) {
+		ret = vgpu_aperture_rw(vgpu, pa, p_data, bytes, false);
+		mutex_unlock(&gvt->lock);
+		return ret;
+	}
+
 	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
 		struct intel_vgpu_guest_page *gp;