drm/nouveau/fb: handle sysmem flush page from common code

- also executes pre-DEVINIT, so early boot is able to DMA sysmem

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>

drivers/gpu/drm/nouveau/include/nvkm/subdev/fb.h

@@ -35,6 +35,11 @@ struct nvkm_fb {
 
 	struct nvkm_blob vpr_scrubber;
 
+	struct {
+		struct page *flush_page;
+		dma_addr_t flush_page_addr;
+	} sysmem;
+
 	struct nvkm_ram *ram;
 
 	struct {

drivers/gpu/drm/nouveau/nvkm/subdev/fb/base.c

@@ -57,6 +57,15 @@ nvkm_fb_tile_prog(struct nvkm_fb *fb, int region, struct nvkm_fb_tile *tile)
 	}
 }
 
+static void
+nvkm_fb_sysmem_flush_page_init(struct nvkm_device *device)
+{
+	struct nvkm_fb *fb = device->fb;
+
+	if (fb->func->sysmem.flush_page_init)
+		fb->func->sysmem.flush_page_init(fb);
+}
+
 int
 nvkm_fb_bios_memtype(struct nvkm_bios *bios)
 {
@@ -168,6 +177,8 @@ nvkm_fb_init(struct nvkm_subdev *subdev)
 	for (i = 0; i < fb->tile.regions; i++)
 		fb->func->tile.prog(fb, i, &fb->tile.region[i]);
 
+	nvkm_fb_sysmem_flush_page_init(subdev->device);
+
 	if (fb->func->init)
 		fb->func->init(fb);
 
@@ -193,6 +204,13 @@ nvkm_fb_init(struct nvkm_subdev *subdev)
 	return 0;
 }
 
+static int
+nvkm_fb_preinit(struct nvkm_subdev *subdev)
+{
+	nvkm_fb_sysmem_flush_page_init(subdev->device);
+	return 0;
+}
+
 static void *
 nvkm_fb_dtor(struct nvkm_subdev *subdev)
 {
@@ -212,20 +230,28 @@ nvkm_fb_dtor(struct nvkm_subdev *subdev)
 
 	nvkm_blob_dtor(&fb->vpr_scrubber);
 
+	if (fb->sysmem.flush_page) {
+		dma_unmap_page(subdev->device->dev, fb->sysmem.flush_page_addr,
+			       PAGE_SIZE, DMA_BIDIRECTIONAL);
+		__free_page(fb->sysmem.flush_page);
+	}
+
 	if (fb->func->dtor)
 		return fb->func->dtor(fb);
+
 	return fb;
 }
 
 static const struct nvkm_subdev_func
 nvkm_fb = {
 	.dtor = nvkm_fb_dtor,
+	.preinit = nvkm_fb_preinit,
 	.oneinit = nvkm_fb_oneinit,
 	.init = nvkm_fb_init,
 	.intr = nvkm_fb_intr,
 };
 
-void
+int
 nvkm_fb_ctor(const struct nvkm_fb_func *func, struct nvkm_device *device,
 	     enum nvkm_subdev_type type, int inst, struct nvkm_fb *fb)
 {
@@ -234,6 +260,19 @@ nvkm_fb_ctor(const struct nvkm_fb_func *func, struct nvkm_device *device,
 	fb->tile.regions = fb->func->tile.regions;
 	fb->page = nvkm_longopt(device->cfgopt, "NvFbBigPage", fb->func->default_bigpage);
 	mutex_init(&fb->tags.mutex);
+
+	if (func->sysmem.flush_page_init) {
+		fb->sysmem.flush_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
+		if (!fb->sysmem.flush_page)
+			return -ENOMEM;
+
+		fb->sysmem.flush_page_addr = dma_map_page(device->dev, fb->sysmem.flush_page,
+							  0, PAGE_SIZE, DMA_BIDIRECTIONAL);
+		if (dma_mapping_error(device->dev, fb->sysmem.flush_page_addr))
+			return -EFAULT;
+	}
+
+	return 0;
 }
 
 int
@@ -242,6 +281,5 @@ nvkm_fb_new_(const struct nvkm_fb_func *func, struct nvkm_device *device,
 {
 	if (!(*pfb = kzalloc(sizeof(**pfb), GFP_KERNEL)))
 		return -ENOMEM;
-	nvkm_fb_ctor(func, device, type, inst, *pfb);
-	return 0;
+	return nvkm_fb_ctor(func, device, type, inst, *pfb);
 }

drivers/gpu/drm/nouveau/nvkm/subdev/fb/ga100.c

@@ -26,9 +26,10 @@ static const struct nvkm_fb_func
 ga100_fb = {
 	.dtor = gf100_fb_dtor,
 	.oneinit = gf100_fb_oneinit,
-	.init = gp100_fb_init,
+	.init = gm200_fb_init,
 	.init_page = gv100_fb_init_page,
 	.init_unkn = gp100_fb_init_unkn,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.ram_new = gp100_ram_new,
 	.default_bigpage = 16,
 };

drivers/gpu/drm/nouveau/nvkm/subdev/fb/ga102.c

@@ -26,9 +26,10 @@ static const struct nvkm_fb_func
 ga102_fb = {
 	.dtor = gf100_fb_dtor,
 	.oneinit = gf100_fb_oneinit,
-	.init = gp100_fb_init,
+	.init = gm200_fb_init,
 	.init_page = gv100_fb_init_page,
 	.init_unkn = gp100_fb_init_unkn,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.ram_new = ga102_ram_new,
 	.default_bigpage = 16,
 };

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.c

@@ -61,14 +61,6 @@ gf100_fb_oneinit(struct nvkm_fb *base)
 	if (ret)
 		return ret;
 
-	fb->r100c10_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (fb->r100c10_page) {
-		fb->r100c10 = dma_map_page(device->dev, fb->r100c10_page, 0,
-					   PAGE_SIZE, DMA_BIDIRECTIONAL);
-		if (dma_mapping_error(device->dev, fb->r100c10))
-			return -EFAULT;
-	}
-
 	return 0;
 }
 
@@ -85,15 +77,18 @@ gf100_fb_init_page(struct nvkm_fb *fb)
 	return 0;
 }
 
+void
+gf100_fb_sysmem_flush_page_init(struct nvkm_fb *fb)
+{
+	nvkm_wr32(fb->subdev.device, 0x100c10, fb->sysmem.flush_page_addr >> 8);
+}
+
 void
 gf100_fb_init(struct nvkm_fb *base)
 {
 	struct gf100_fb *fb = gf100_fb(base);
 	struct nvkm_device *device = fb->base.subdev.device;
 
-	if (fb->r100c10_page)
-		nvkm_wr32(device, 0x100c10, fb->r100c10 >> 8);
-
 	if (base->func->clkgate_pack) {
 		nvkm_therm_clkgate_init(device->therm,
 					base->func->clkgate_pack);
@@ -104,13 +99,6 @@ void *
 gf100_fb_dtor(struct nvkm_fb *base)
 {
 	struct gf100_fb *fb = gf100_fb(base);
-	struct nvkm_device *device = fb->base.subdev.device;
-
-	if (fb->r100c10_page) {
-		dma_unmap_page(device->dev, fb->r100c10, PAGE_SIZE,
-			       DMA_BIDIRECTIONAL);
-		__free_page(fb->r100c10_page);
-	}
 
 	return fb;
 }
@@ -136,6 +124,7 @@ gf100_fb = {
 	.init = gf100_fb_init,
 	.init_page = gf100_fb_init_page,
 	.intr = gf100_fb_intr,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.ram_new = gf100_ram_new,
 	.default_bigpage = 17,
 };

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.h

@@ -6,8 +6,6 @@
 
 struct gf100_fb {
 	struct nvkm_fb base;
-	struct page *r100c10_page;
-	dma_addr_t r100c10;
 };
 
 int gf100_fb_new_(const struct nvkm_fb_func *, struct nvkm_device *, enum nvkm_subdev_type, int,
@@ -16,7 +14,5 @@ void *gf100_fb_dtor(struct nvkm_fb *);
 void gf100_fb_init(struct nvkm_fb *);
 void gf100_fb_intr(struct nvkm_fb *);
 
-void gp100_fb_init(struct nvkm_fb *);
-
 void gm200_fb_init(struct nvkm_fb *base);
 #endif

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm200.c

@@ -46,9 +46,6 @@ gm200_fb_init(struct nvkm_fb *base)
 	struct gf100_fb *fb = gf100_fb(base);
 	struct nvkm_device *device = fb->base.subdev.device;
 
-	if (fb->r100c10_page)
-		nvkm_wr32(device, 0x100c10, fb->r100c10 >> 8);
-
 	nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(fb->base.mmu_wr) >> 8);
 	nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(fb->base.mmu_rd) >> 8);
 	nvkm_mask(device, 0x100cc4, 0x00060000,
@@ -62,6 +59,7 @@ gm200_fb = {
 	.init = gm200_fb_init,
 	.init_page = gm200_fb_init_page,
 	.intr = gf100_fb_intr,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.ram_new = gm200_ram_new,
 	.default_bigpage = 0 /* per-instance. */,
 };

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm20b.c

@@ -30,6 +30,7 @@ gm20b_fb = {
 	.init = gm200_fb_init,
 	.init_page = gm200_fb_init_page,
 	.intr = gf100_fb_intr,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.default_bigpage = 0 /* per-instance. */,
 };
 

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp100.c

@@ -44,29 +44,15 @@ gp100_fb_init_remapper(struct nvkm_fb *fb)
 	nvkm_mask(device, 0x100c14, 0x00040000, 0x00000000);
 }
 
-void
-gp100_fb_init(struct nvkm_fb *base)
-{
-	struct gf100_fb *fb = gf100_fb(base);
-	struct nvkm_device *device = fb->base.subdev.device;
-
-	if (fb->r100c10_page)
-		nvkm_wr32(device, 0x100c10, fb->r100c10 >> 8);
-
-	nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(fb->base.mmu_wr) >> 8);
-	nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(fb->base.mmu_rd) >> 8);
-	nvkm_mask(device, 0x100cc4, 0x00060000,
-		  min(nvkm_memory_size(fb->base.mmu_rd) >> 16, (u64)2) << 17);
-}
-
 static const struct nvkm_fb_func
 gp100_fb = {
 	.dtor = gf100_fb_dtor,
 	.oneinit = gf100_fb_oneinit,
-	.init = gp100_fb_init,
+	.init = gm200_fb_init,
 	.init_remapper = gp100_fb_init_remapper,
 	.init_page = gm200_fb_init_page,
 	.init_unkn = gp100_fb_init_unkn,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.ram_new = gp100_ram_new,
 };
 

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp102.c

@@ -55,9 +55,10 @@ static const struct nvkm_fb_func
 gp102_fb = {
 	.dtor = gf100_fb_dtor,
 	.oneinit = gf100_fb_oneinit,
-	.init = gp100_fb_init,
+	.init = gm200_fb_init,
 	.init_remapper = gp100_fb_init_remapper,
 	.init_page = gm200_fb_init_page,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.vpr.scrub_required = gp102_fb_vpr_scrub_required,
 	.vpr.scrub = gp102_fb_vpr_scrub,
 	.ram_new = gp100_ram_new,

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gp10b.c

@@ -28,6 +28,7 @@ gp10b_fb = {
 	.init = gm200_fb_init,
 	.init_page = gm200_fb_init_page,
 	.intr = gf100_fb_intr,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 };
 
 int

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gv100.c

@@ -32,9 +32,10 @@ static const struct nvkm_fb_func
 gv100_fb = {
 	.dtor = gf100_fb_dtor,
 	.oneinit = gf100_fb_oneinit,
-	.init = gp100_fb_init,
+	.init = gm200_fb_init,
 	.init_page = gv100_fb_init_page,
 	.init_unkn = gp100_fb_init_unkn,
+	.sysmem.flush_page_init = gf100_fb_sysmem_flush_page_init,
 	.vpr.scrub_required = gp102_fb_vpr_scrub_required,
 	.vpr.scrub = gp102_fb_vpr_scrub,
 	.ram_new = gp100_ram_new,

drivers/gpu/drm/nouveau/nvkm/subdev/fb/nv50.c

@@ -191,35 +191,12 @@ nv50_fb_intr(struct nvkm_fb *base)
 	nvkm_chan_put(&chan, flags);
 }
 
-static int
-nv50_fb_oneinit(struct nvkm_fb *base)
-{
-	struct nv50_fb *fb = nv50_fb(base);
-	struct nvkm_device *device = fb->base.subdev.device;
-
-	fb->r100c08_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (fb->r100c08_page) {
-		fb->r100c08 = dma_map_page(device->dev, fb->r100c08_page, 0,
-					   PAGE_SIZE, DMA_BIDIRECTIONAL);
-		if (dma_mapping_error(device->dev, fb->r100c08))
-			return -EFAULT;
-	}
-
-	return 0;
-}
-
 static void
 nv50_fb_init(struct nvkm_fb *base)
 {
 	struct nv50_fb *fb = nv50_fb(base);
 	struct nvkm_device *device = fb->base.subdev.device;
 
-	/* Not a clue what this is exactly.  Without pointing it at a
-	 * scratch page, VRAM->GART blits with M2MF (as in DDX DFS)
-	 * cause IOMMU "read from address 0" errors (rh#561267)
-	 */
-	nvkm_wr32(device, 0x100c08, fb->r100c08 >> 8);
-
 	/* This is needed to get meaningful information from 100c90
 	 * on traps. No idea what these values mean exactly. */
 	nvkm_wr32(device, 0x100c90, fb->func->trap);
@@ -234,17 +211,16 @@ nv50_fb_tags(struct nvkm_fb *base)
 	return 0;
 }
 
+static void
+nv50_fb_sysmem_flush_page_init(struct nvkm_fb *fb)
+{
+	nvkm_wr32(fb->subdev.device, 0x100c08, fb->sysmem.flush_page_addr >> 8);
+}
+
 static void *
 nv50_fb_dtor(struct nvkm_fb *base)
 {
 	struct nv50_fb *fb = nv50_fb(base);
-	struct nvkm_device *device = fb->base.subdev.device;
-
-	if (fb->r100c08_page) {
-		dma_unmap_page(device->dev, fb->r100c08, PAGE_SIZE,
-			       DMA_BIDIRECTIONAL);
-		__free_page(fb->r100c08_page);
-	}
 
 	return fb;
 }
@@ -253,9 +229,9 @@ static const struct nvkm_fb_func
 nv50_fb_ = {
 	.dtor = nv50_fb_dtor,
 	.tags = nv50_fb_tags,
-	.oneinit = nv50_fb_oneinit,
 	.init = nv50_fb_init,
 	.intr = nv50_fb_intr,
+	.sysmem.flush_page_init = nv50_fb_sysmem_flush_page_init,
 	.ram_new = nv50_fb_ram_new,
 };
 

drivers/gpu/drm/nouveau/nvkm/subdev/fb/nv50.h

@@ -7,8 +7,6 @@
 struct nv50_fb {
 	const struct nv50_fb_func *func;
 	struct nvkm_fb base;
-	struct page *r100c08_page;
-	dma_addr_t r100c08;
 };
 
 struct nv50_fb_func {