drm/xe: Set PTE_DM bit for stolen on MTL

Integrated graphics 1270 and beyond should set the PTE_LM bit in the PTE
when it's stolen memory. Add a new function, xe_bo_is_stolen_devmem(),
and use it when encoding the PTE.

In some places in the spec the PTE bit is called "Local Memory",
abbreviated as LM, and in others it's called "Device Memory" (DM). Since
we moved away from "Local Memory" and preferred the "vram" terminology,
also rename the macros as DM to follow the name of the new function.
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20230726160708.3967790-7-lucas.demarchi@intel.comSigned-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drivers/gpu/drm/xe/xe_bo.c

@@ -80,6 +80,21 @@ bool xe_bo_is_stolen(struct xe_bo *bo)
 	return bo->ttm.resource->mem_type == XE_PL_STOLEN;
 }
 
+/**
+ * xe_bo_is_stolen_devmem - check if BO is of stolen type accessed via PCI BAR
+ * @bo: The BO
+ *
+ * The stolen memory is accessed through the PCI BAR for both DGFX and some
+ * integrated platforms that have a dedicated bit in the PTE for devmem (DM).
+ *
+ * Returns: true if it's stolen memory accessed via PCI BAR, false otherwise.
+ */
+bool xe_bo_is_stolen_devmem(struct xe_bo *bo)
+{
+	return xe_bo_is_stolen(bo) &&
+		GRAPHICS_VERx100(xe_bo_device(bo)) >= 1270;
+}
+
 static bool xe_bo_is_user(struct xe_bo *bo)
 {
 	return bo->flags & XE_BO_CREATE_USER_BIT;

drivers/gpu/drm/xe/xe_bo.h

@@ -67,9 +67,9 @@
 #define XE_PDPE_PS_1G			BIT_ULL(7)
 #define XE_PDE_IPS_64K			BIT_ULL(11)
 
-#define XE_GGTT_PTE_LM			BIT_ULL(1)
+#define XE_GGTT_PTE_DM			BIT_ULL(1)
 #define XE_USM_PPGTT_PTE_AE		BIT_ULL(10)
-#define XE_PPGTT_PTE_LM			BIT_ULL(11)
+#define XE_PPGTT_PTE_DM			BIT_ULL(11)
 #define XE_PDE_64K			BIT_ULL(6)
 #define XE_PTE_PS64			BIT_ULL(8)
 #define XE_PTE_NULL			BIT_ULL(9)
@@ -239,6 +239,7 @@ void xe_bo_vunmap(struct xe_bo *bo);
 bool mem_type_is_vram(u32 mem_type);
 bool xe_bo_is_vram(struct xe_bo *bo);
 bool xe_bo_is_stolen(struct xe_bo *bo);
+bool xe_bo_is_stolen_devmem(struct xe_bo *bo);
 uint64_t vram_region_gpu_offset(struct ttm_resource *res);
 
 bool xe_bo_can_migrate(struct xe_bo *bo, u32 mem_type);

drivers/gpu/drm/xe/xe_ggtt.c

@@ -35,8 +35,8 @@ u64 xe_ggtt_pte_encode(struct xe_bo *bo, u64 bo_offset)
 	pte = xe_bo_addr(bo, bo_offset, XE_PAGE_SIZE);
 	pte |= XE_PAGE_PRESENT;
 
-	if (xe_bo_is_vram(bo))
-		pte |= XE_GGTT_PTE_LM;
+	if (xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo))
+		pte |= XE_GGTT_PTE_DM;
 
 	/* FIXME: vfunc + pass in caching rules */
 	if (xe->info.platform == XE_METEORLAKE) {

drivers/gpu/drm/xe/xe_migrate.c

@@ -258,7 +258,7 @@ static int xe_migrate_prepare_vm(struct xe_tile *tile, struct xe_migrate *m,
 		level = 2;
 		ofs = map_ofs + XE_PAGE_SIZE * level + 256 * 8;
 		flags = XE_PAGE_RW | XE_PAGE_PRESENT | PPAT_CACHED |
-			XE_PPGTT_PTE_LM | XE_PDPE_PS_1G;
+			XE_PPGTT_PTE_DM | XE_PDPE_PS_1G;
 
 		/*
 		 * Use 1GB pages, it shouldn't matter the physical amount of
@@ -463,7 +463,7 @@ static void emit_pte(struct xe_migrate *m,
 				}
 
 				addr += vram_region_gpu_offset(bo->ttm.resource);
-				addr |= XE_PPGTT_PTE_LM;
+				addr |= XE_PPGTT_PTE_DM;
 			}
 			addr |= PPAT_CACHED | XE_PAGE_PRESENT | XE_PAGE_RW;
 			bb->cs[bb->len++] = lower_32_bits(addr);

drivers/gpu/drm/xe/xe_pt.c

@@ -127,8 +127,8 @@ u64 xe_pte_encode(struct xe_bo *bo, u64 offset, enum xe_cache_level cache,
 	u64 pte;
 
 	pte = xe_bo_addr(bo, offset, XE_PAGE_SIZE);
-	if (xe_bo_is_vram(bo))
-		pte |= XE_PPGTT_PTE_LM;
+	if (xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo))
+		pte |= XE_PPGTT_PTE_DM;
 
 	return __pte_encode(pte, cache, NULL, pt_level);
 }
@@ -714,7 +714,8 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
 		 struct xe_vm_pgtable_update *entries, u32 *num_entries)
 {
 	struct xe_bo *bo = xe_vma_bo(vma);
-	bool is_vram = !xe_vma_is_userptr(vma) && bo && xe_bo_is_vram(bo);
+	bool is_devmem = !xe_vma_is_userptr(vma) && bo &&
+		(xe_bo_is_vram(bo) || xe_bo_is_stolen_devmem(bo));
 	struct xe_res_cursor curs;
 	struct xe_pt_stage_bind_walk xe_walk = {
 		.base = {
@@ -728,13 +729,13 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
 		.va_curs_start = xe_vma_start(vma),
 		.vma = vma,
 		.wupd.entries = entries,
-		.needs_64K = (xe_vma_vm(vma)->flags & XE_VM_FLAG_64K) && is_vram,
+		.needs_64K = (xe_vma_vm(vma)->flags & XE_VM_FLAG_64K) && is_devmem,
 	};
 	struct xe_pt *pt = xe_vma_vm(vma)->pt_root[tile->id];
 	int ret;
 
-	if (is_vram) {
-		xe_walk.default_pte = XE_PPGTT_PTE_LM;
+	if (is_devmem) {
+		xe_walk.default_pte = XE_PPGTT_PTE_DM;
 		if (vma && vma->gpuva.flags & XE_VMA_ATOMIC_PTE_BIT)
 			xe_walk.default_pte |= XE_USM_PPGTT_PTE_AE;
 		xe_walk.dma_offset = vram_region_gpu_offset(bo->ttm.resource);