diff --git a/Makefile b/Makefile index db27cf6..90805dd 100644 --- a/Makefile +++ b/Makefile @@ -42,14 +42,16 @@ endif CPPFLAGS += $(PC_CFLAGS) LDLIBS += $(PC_LIBS) -LIBDIR ?= /usr/lib/ +DESTDIR ?= $(OUT)/install +LIBDIR ?= /usr/lib +INCLUDEDIR ?= /usr/include -GBM_VERSION_MAJOR := 1 -MINIGBM_VERSION := $(GBM_VERSION_MAJOR).0.0 +MINIGBM_VERSION_MAJOR := 1 +MINIGBM_VERSION := $(MINIGBM_VERSION_MAJOR).0.0 MINIGBM_FILENAME := libminigbm.so.$(MINIGBM_VERSION) -CC_LIBRARY($(MINIGBM_FILENAME)): LDFLAGS += -Wl,-soname,libgbm.so.$(GBM_VERSION_MAJOR) -CC_LIBRARY($(MINIGBM_FILENAME)): $(C_OBJECTS) +CC_LIBRARY($(MINIGBM_FILENAME)): LDFLAGS += -Wl,-soname,libgbm.so.$(MINIGBM_VERSION_MAJOR) +CC_LIBRARY($(MINIGBM_FILENAME)): $(C_OBJECTS) pkgconfig CC_STATIC_LIBRARY(libminigbm.pie.a): $(C_OBJECTS) all: CC_LIBRARY($(MINIGBM_FILENAME)) @@ -63,11 +65,33 @@ CXX_BINARY(gbm_unittest): $(UNITTEST_DEPS) clean: CLEAN(gbm_unittest) tests: TEST(CXX_BINARY(gbm_unittest)) +define pkgconfig_contents +prefix=$(DESTDIR) +exec_prefix=$${prefix} +includedir=$${prefix}/$(INCLUDEDIR) +libdir=$${prefix}/$(LIBDIR) + +Name: libgbm +Description: A small gbm implementation +Version: 18.0.0 +Cflags: -I$${includedir} +Libs: -L$${libdir} -lgbm +Requires.private: libdrm >= 2.4.50 +endef + +.PHONY: pkgconfig +pkgconfig: + @echo "generating $(OUT)gbm.pc" + $(file > $(OUT)/gbm.pc,$(pkgconfig_contents)) +clean: + rm -f $(OUT)gbm.pc + install: all mkdir -p $(DESTDIR)/$(LIBDIR) install -D -m 755 $(OUT)/$(MINIGBM_FILENAME) $(DESTDIR)/$(LIBDIR) ln -sf $(MINIGBM_FILENAME) $(DESTDIR)/$(LIBDIR)/libgbm.so - ln -sf $(MINIGBM_FILENAME) $(DESTDIR)/$(LIBDIR)/libgbm.so.$(GBM_VERSION_MAJOR) - install -D -m 0644 $(SRC)/gbm.pc $(DESTDIR)$(LIBDIR)/pkgconfig/gbm.pc - install -D -m 0644 $(SRC)/gbm.h $(DESTDIR)/usr/include/gbm.h - install -D -m 0644 $(SRC)/minigbm_helpers.h $(DESTDIR)/usr/include/minigbm/minigbm_helpers.h + ln -sf $(MINIGBM_FILENAME) $(DESTDIR)/$(LIBDIR)/libgbm.so.$(MINIGBM_VERSION_MAJOR) + ln -sf $(MINIGBM_FILENAME) $(DESTDIR)/$(LIBDIR)/libgbm.so.$(MINIGBM_VERSION) + install -D -m 0644 $(OUT)/gbm.pc $(DESTDIR)/$(LIBDIR)/pkgconfig/gbm.pc + install -D -m 0644 $(SRC)/gbm.h $(DESTDIR)/$(INCLUDEDIR)/gbm.h + install -D -m 0644 $(SRC)/minigbm_helpers.h $(DESTDIR)/$(INCLUDEDIR)/minigbm/minigbm_helpers.h diff --git a/OWNERS b/OWNERS index e47b7b8..e03c24d 100644 --- a/OWNERS +++ b/OWNERS @@ -1,20 +1,21 @@ basni@chromium.org +dawnhan@google.com dbehr@chromium.org ddavenport@chromium.org gurchetansingh@chromium.org hiroh@chromium.org -hoegsberg@chromium.org marcheu@chromium.org mcasas@chromium.org +niwa@chromium.org olv@google.com robdclark@chromium.org tfiga@chromium.org zzyiwei@chromium.org # just reference for backend specific reviews -per-file amdgpu.c = basni@chromium.org, ddavenport@chromium.org -per-file i915.c = chadversary@chromium.org, linyaa@google.com, hoegsberg@chromium.org -per-file mediatek.c = fshao@chromium.org, hsinyi@chromium.org, tzungbi@chromium.org -per-file msm.c = robdclark@chromium.org, hoegsberg@chromium.org +per-file amdgpu.c = basni@chromium.org, ddavenport@chromium.org, olv@google.com +per-file i915.c = linyaa@google.com, msturner@google.com +per-file mediatek.c = fshao@chromium.org, hsinyi@chromium.org +per-file msm.c = robdclark@chromium.org per-file rockchip.c = tfiga@chromium.org -per-file virtgpu* = jbates@chromium.org, natsu@google.com, olv@google.com, zzyiwei@chromium.org +per-file virtgpu* = dawnhan@google.com, natsu@google.com, zzyiwei@chromium.org diff --git a/backend_mock.c b/backend_mock.c index ae0d758..67012ad 100644 --- a/backend_mock.c +++ b/backend_mock.c @@ -6,7 +6,8 @@ #include "drv_priv.h" -static int backend_mock_init(struct driver *drv) { +static int backend_mock_init(struct driver *drv) +{ return 0; } diff --git a/cros_gralloc/cros_gralloc_helpers.cc b/cros_gralloc/cros_gralloc_helpers.cc index 67a53a0..4b40bfd 100644 --- a/cros_gralloc/cros_gralloc_helpers.cc +++ b/cros_gralloc/cros_gralloc_helpers.cc @@ -78,6 +78,16 @@ uint32_t cros_gralloc_convert_format(int format) case HAL_PIXEL_FORMAT_YCBCR_P010: return DRM_FORMAT_P010; #endif + case HAL_PIXEL_FORMAT_DEPTH_16: + return DRM_FORMAT_DEPTH16; + case HAL_PIXEL_FORMAT_DEPTH_24: + return DRM_FORMAT_DEPTH24; + case HAL_PIXEL_FORMAT_DEPTH_24_STENCIL_8: + return DRM_FORMAT_DEPTH24_STENCIL8; + case HAL_PIXEL_FORMAT_DEPTH_32F: + return DRM_FORMAT_DEPTH32; + case HAL_PIXEL_FORMAT_DEPTH_32F_STENCIL_8: + return DRM_FORMAT_DEPTH32_STENCIL8; } return DRM_FORMAT_NONE; diff --git a/cros_gralloc/gralloc0/tests/gralloctest.c b/cros_gralloc/gralloc0/tests/gralloctest.c index 9d8101c..1641b8a 100644 --- a/cros_gralloc/gralloc0/tests/gralloctest.c +++ b/cros_gralloc/gralloc0/tests/gralloctest.c @@ -22,7 +22,7 @@ #include #include -#define ALIGN(A, B) (((A) + (B)-1) / (B) * (B)) +#define ALIGN(A, B) (((A) + (B) - 1) / (B) * (B)) #define ARRAY_SIZE(A) (sizeof(A) / sizeof(*(A))) #define CHECK(cond) \ diff --git a/drv.c b/drv.c index ea3bd64..7fd2250 100644 --- a/drv.c +++ b/drv.c @@ -121,7 +121,8 @@ struct driver *drv_create(int fd) const char *minigbm_debug; minigbm_debug = drv_get_os_option(MINIGBM_DEBUG); - drv->compression = (minigbm_debug == NULL) || (strstr(minigbm_debug, "nocompression") == NULL); + drv->compression = + (minigbm_debug == NULL) || (strstr(minigbm_debug, "nocompression") == NULL); drv->log_bos = (minigbm_debug && strstr(minigbm_debug, "log_bos") != NULL); drv->fd = fd; @@ -732,19 +733,15 @@ void drv_bo_log_info(const struct bo *bo, const char *prefix) { const struct bo_metadata *meta = &bo->meta; - drv_logd("%s %s bo %p: %dx%d '%c%c%c%c' tiling %d plane %zu mod 0x%" PRIx64 " use 0x%" PRIx64 " size %zu\n", - prefix, bo->drv->backend->name, bo, - meta->width, meta->height, - meta->format & 0xff, - (meta->format >> 8) & 0xff, - (meta->format >> 16) & 0xff, - (meta->format >> 24) & 0xff, - meta->tiling, meta->num_planes, meta->format_modifier, + drv_logd("%s %s bo %p: %dx%d '%c%c%c%c' tiling %d plane %zu mod 0x%" PRIx64 + " use 0x%" PRIx64 " size %zu\n", + prefix, bo->drv->backend->name, bo, meta->width, meta->height, meta->format & 0xff, + (meta->format >> 8) & 0xff, (meta->format >> 16) & 0xff, + (meta->format >> 24) & 0xff, meta->tiling, meta->num_planes, meta->format_modifier, meta->use_flags, meta->total_size); for (uint32_t i = 0; i < meta->num_planes; i++) { - drv_logd(" bo %p plane %d: offset %d size %d stride %d\n", - bo, i, meta->offsets[i], meta->sizes[i], - meta->strides[i]); + drv_logd(" bo %p plane %d: offset %d size %d stride %d\n", bo, i, meta->offsets[i], + meta->sizes[i], meta->strides[i]); } } diff --git a/drv.h b/drv.h index b61aedf..bd380b5 100644 --- a/drv.h +++ b/drv.h @@ -64,6 +64,14 @@ extern "C" { #define DRM_FORMAT_FLEX_IMPLEMENTATION_DEFINED fourcc_code('9', '9', '9', '8') #define DRM_FORMAT_FLEX_YCbCr_420_888 fourcc_code('9', '9', '9', '9') +#ifndef DRM_FORMAT_DEPTH16 +#define DRM_FORMAT_DEPTH16 fourcc_code('D', '1', '6', ' ') +#define DRM_FORMAT_DEPTH24 fourcc_code('D', '2', '4', 'X') +#define DRM_FORMAT_DEPTH24_STENCIL8 fourcc_code('D', '2', '4', 'S') +#define DRM_FORMAT_DEPTH32 fourcc_code('D', '3', '2', 'F') +#define DRM_FORMAT_DEPTH32_STENCIL8 fourcc_code('D', 'F', 'S', '8') +#endif + /* This is a 10-bit bayer format for private reprocessing on MediaTek ISP. It's * a private RAW format that other DRM drivers will never support and thus * making it not upstreamable (i.e., defined in official DRM headers). */ @@ -98,6 +106,22 @@ extern "C" { #define I915_FORMAT_MOD_4_TILED_MTL_RC_CCS fourcc_mod_code(INTEL, 13) #endif +#ifndef I915_FORMAT_MOD_4_TILED_MTL_MC_CCS +//TODO: remove this defination once drm_fourcc.h contains it. +/* + * Intel color control surfaces (CCS) for display ver 14 media compression + * + * The main surface is tile4 and at plane index 0, the CCS is linear and + * at index 1. A 64B CCS cache line corresponds to an area of 4x1 tiles in + * main surface. In other words, 4 bits in CCS map to a main surface cache + * line pair. The main surface pitch is required to be a multiple of four + * tile4 widths. For semi-planar formats like NV12, CCS planes follow the + * Y and UV planes i.e., planes 0 and 1 are used for Y and UV surfaces, + * planes 2 and 3 for the respective CCS. + */ +#define I915_FORMAT_MOD_4_TILED_MTL_MC_CCS fourcc_mod_code(INTEL, 14) +#endif + // clang-format on struct driver; struct bo; diff --git a/drv_helpers.c b/drv_helpers.c index c810d8e..12b671a 100644 --- a/drv_helpers.c +++ b/drv_helpers.c @@ -100,6 +100,7 @@ static const struct planar_layout *layout_from_format(uint32_t format) return &packed_1bpp_layout; case DRM_FORMAT_R16: + case DRM_FORMAT_DEPTH16: return &packed_2bpp_layout; case DRM_FORMAT_YVU420: @@ -144,6 +145,9 @@ static const struct planar_layout *layout_from_format(uint32_t format) case DRM_FORMAT_RGB888: return &packed_3bpp_layout; + case DRM_FORMAT_DEPTH24: + case DRM_FORMAT_DEPTH24_STENCIL8: + case DRM_FORMAT_DEPTH32: case DRM_FORMAT_ABGR2101010: case DRM_FORMAT_ABGR8888: case DRM_FORMAT_ARGB2101010: @@ -163,6 +167,7 @@ static const struct planar_layout *layout_from_format(uint32_t format) case DRM_FORMAT_XRGB8888: return &packed_4bpp_layout; + case DRM_FORMAT_DEPTH32_STENCIL8: case DRM_FORMAT_ABGR16161616F: return &packed_8bpp_layout; diff --git a/gbm.h b/gbm.h index 8e32769..197324c 100644 --- a/gbm.h +++ b/gbm.h @@ -413,6 +413,18 @@ gbm_bo_map(struct gbm_bo *bo, uint32_t x, uint32_t y, uint32_t width, uint32_t height, uint32_t flags, uint32_t *stride, void **map_data); +/* Neither gbm_bo_map_cache_mode nor gbm_bo_get_map_info are defined in mesa's gbm.h, or older + * versions of minigbm. For backwards-compatibility, users should first test for availability and + * provide a fallback implementation with: + * + * #if defined(MINIGBM) && defined(MINIGBM_HAS_GBM_BO_GET_MAP_INFO) + * // use gbm_bo_get_map_info() + * #else + * // fallback + * #endif + */ +#define MINIGBM_HAS_GBM_BO_GET_MAP_INFO + /** * Enum to indicate the cache attributes of CPU mapping returned by * gbm_bo_map() diff --git a/gbm.pc b/gbm.pc deleted file mode 100644 index a7509fc..0000000 --- a/gbm.pc +++ /dev/null @@ -1,10 +0,0 @@ -prefix=/usr -exec_prefix=${prefix} -includedir=${prefix}/include -libdir=${exec_prefix}/lib - -Name: libgbm -Description: A small gbm implementation -Version: 18.0.0 -Cflags: -I${includedir} -Libs: -L${libdir} -lgbm diff --git a/gbm_unittest.cc b/gbm_unittest.cc index 9f2a46f..bad568f 100644 --- a/gbm_unittest.cc +++ b/gbm_unittest.cc @@ -14,7 +14,7 @@ class MockDrm { - public: + public: MOCK_METHOD(drmVersionPtr, drmGetVersion, (int fd)); MOCK_METHOD(void, drmFreeVersion, (drmVersionPtr v)); }; @@ -30,7 +30,7 @@ drmVersionPtr drmGetVersion(int fd) // Define a mock version of drmFreeVersion void drmFreeVersion(drmVersionPtr v) { - delete(v); + delete (v); } /* TODO : This is a protocol to add unit tests for the public APIs in minigbm. diff --git a/i915.c b/i915.c index b9e377f..f9b79b1 100644 --- a/i915.c +++ b/i915.c @@ -37,15 +37,16 @@ static const uint32_t texture_only_formats[] = { DRM_FORMAT_R8, DRM_FORMAT_NV12, static const uint64_t gen_modifier_order[] = { I915_FORMAT_MOD_Y_TILED_CCS, I915_FORMAT_MOD_Y_TILED, I915_FORMAT_MOD_X_TILED, DRM_FORMAT_MOD_LINEAR }; -static const uint64_t gen12_modifier_order[] = { - I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS, I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS, - I915_FORMAT_MOD_Y_TILED, I915_FORMAT_MOD_X_TILED, DRM_FORMAT_MOD_LINEAR -}; +static const uint64_t gen12_modifier_order[] = { I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS, + I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS, + I915_FORMAT_MOD_Y_TILED, I915_FORMAT_MOD_X_TILED, + DRM_FORMAT_MOD_LINEAR }; static const uint64_t gen11_modifier_order[] = { I915_FORMAT_MOD_Y_TILED, I915_FORMAT_MOD_X_TILED, DRM_FORMAT_MOD_LINEAR }; static const uint64_t xe_lpdp_modifier_order[] = { I915_FORMAT_MOD_4_TILED_MTL_RC_CCS, + I915_FORMAT_MOD_4_TILED_MTL_MC_CCS, I915_FORMAT_MOD_4_TILED, I915_FORMAT_MOD_X_TILED, DRM_FORMAT_MOD_LINEAR }; @@ -112,12 +113,13 @@ static void i915_info_from_device_id(struct i915_device *i915) 0x46b3, 0x46c0, 0x46c1, 0x46c2, 0x46c3, 0x9A40, 0x9A49, 0x9A59, 0x9A60, 0x9A68, 0x9A70, 0x9A78, 0x9AC0, 0x9AC9, 0x9AD9, 0x9AF8, 0x4905, 0x4906, 0x4907, 0x4908 }; - const uint16_t adlp_ids[] = { 0x46A0, 0x46A1, 0x46A2, 0x46A3, 0x46A6, 0x46A8, 0x46AA, - 0x462A, 0x4626, 0x4628, 0x46B0, 0x46B1, 0x46B2, 0x46B3, - 0x46C0, 0x46C1, 0x46C2, 0x46C3, 0x46D0, 0x46D1, 0x46D2, - 0x46D3, 0x46D4 }; + const uint16_t adlp_ids[] = { 0x46A0, 0x46A1, 0x46A2, 0x46A3, 0x46A6, 0x46A8, + 0x46AA, 0x462A, 0x4626, 0x4628, 0x46B0, 0x46B1, + 0x46B2, 0x46B3, 0x46C0, 0x46C1, 0x46C2, 0x46C3, + 0x46D0, 0x46D1, 0x46D2, 0x46D3, 0x46D4 }; - const uint16_t rplp_ids[] = { 0xA720, 0xA721, 0xA7A0, 0xA7A1, 0xA7A8, 0xA7A9, 0xA7AA, 0xA7AB, 0xA7AC, 0xA7AD }; + const uint16_t rplp_ids[] = { 0xA720, 0xA721, 0xA7A0, 0xA7A1, 0xA7A8, + 0xA7A9, 0xA7AA, 0xA7AB, 0xA7AC, 0xA7AD }; const uint16_t mtl_ids[] = { 0x7D40, 0x7D60, 0x7D45, 0x7D55, 0x7DD5 }; @@ -562,7 +564,8 @@ static size_t i915_num_planes_from_modifier(struct driver *drv, uint32_t format, modifier == I915_FORMAT_MOD_4_TILED_MTL_RC_CCS) { assert(num_planes == 1); return 2; - } else if (modifier == I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS) { + } else if (modifier == I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS || + modifier == I915_FORMAT_MOD_4_TILED_MTL_MC_CCS) { assert(num_planes == 2); return 4; } @@ -570,6 +573,13 @@ static size_t i915_num_planes_from_modifier(struct driver *drv, uint32_t format, return num_planes; } +#define gbm_fls(x) \ + ((x) ? __builtin_choose_expr(sizeof(x) == 8, 64 - __builtin_clzll(x), \ + 32 - __builtin_clz(x)) \ + : 0) + +#define roundup_power_of_two(x) ((x) != 0 ? 1ULL << gbm_fls((x) - 1) : 0) + static int i915_bo_compute_metadata(struct bo *bo, uint32_t width, uint32_t height, uint32_t format, uint64_t use_flags, const uint64_t *modifiers, uint32_t count) { @@ -590,11 +600,13 @@ static int i915_bo_compute_metadata(struct bo *bo, uint32_t width, uint32_t heig * on |use_flags|. Instead the client should request them explicitly through * gbm_bo_create_with_modifiers(). */ - assert(modifier != I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS); + assert(modifier != I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS && + modifier != I915_FORMAT_MOD_4_TILED_MTL_MC_CCS); /* TODO(b/323863689): Account for driver's bandwidth compression in minigbm for * media compressed buffers. */ } - if (modifier == I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS && + if ((modifier == I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS || + modifier == I915_FORMAT_MOD_4_TILED_MTL_MC_CCS) && !(format == DRM_FORMAT_NV12 || format == DRM_FORMAT_P010)) { drv_loge("Media compression is only supported for NV12 and P010\n"); return -EINVAL; @@ -656,6 +668,7 @@ static int i915_bo_compute_metadata(struct bo *bo, uint32_t width, uint32_t heig break; case I915_FORMAT_MOD_4_TILED: case I915_FORMAT_MOD_4_TILED_MTL_RC_CCS: + case I915_FORMAT_MOD_4_TILED_MTL_MC_CCS: bo->meta.tiling = I915_TILING_4; break; } @@ -782,39 +795,68 @@ static int i915_bo_compute_metadata(struct bo *bo, uint32_t width, uint32_t heig /* Total number of planes & sizes */ bo->meta.num_planes = plane + a_plane; bo->meta.total_size = offset; - } else if (modifier == I915_FORMAT_MOD_4_TILED_MTL_RC_CCS) { - - /* - * considering only 128 byte compression and one cache line of - * aux buffer(64B) contains compression status of 4-Y tiles. - * Which is 4 * (128B * 32L). - * line stride(bytes) is 4 * 128B - * and tile stride(lines) is 32L + } else if (modifier == I915_FORMAT_MOD_4_TILED_MTL_RC_CCS || + modifier == I915_FORMAT_MOD_4_TILED_MTL_MC_CCS) { + /* Media compression modifiers should only be possible via the + * gbm_bo_create_with_modifiers() path, i.e., the minigbm client needs to + * explicitly request it. */ - uint32_t stride = ALIGN(drv_stride_from_format(format, width, 0), 512); - stride = ALIGN(stride, 256); + assert(modifier != I915_FORMAT_MOD_4_TILED_MTL_MC_CCS || use_flags == BO_USE_NONE); + assert(modifier != I915_FORMAT_MOD_4_TILED_MTL_MC_CCS || + bo->meta.use_flags == BO_USE_NONE); + assert(modifier != I915_FORMAT_MOD_4_TILED_MTL_MC_CCS || + (!!modifiers && count > 0)); + assert(modifier != I915_FORMAT_MOD_4_TILED_MTL_MC_CCS || + (format == DRM_FORMAT_NV12 || format == DRM_FORMAT_P010 || + format == DRM_FORMAT_XRGB8888 || format == DRM_FORMAT_XBGR8888)); + assert(drv_num_planes_from_format(format) > 0); - height = ALIGN(drv_height_from_format(format, height, 0), 32); + uint32_t offset = 0, stride = 0; + size_t plane = 0; + size_t a_plane = 0; + for (plane = 0; plane < drv_num_planes_from_format(format); plane++) { + uint32_t alignment = 0, val, tmpoffset = 0; + /* + * tile_align = 4 (for width) for CCS and + * tile_width = 128, tile_height = 32 for MC CCS + */ + stride = ALIGN(drv_stride_from_format(format, width, plane), 512); + height = ALIGN(drv_height_from_format(format, height, plane), 32); + bo->meta.strides[plane] = stride; - bo->meta.strides[0] = stride; - /* size calculation and alignment are 64KB aligned - * size as per spec - */ - bo->meta.sizes[0] = ALIGN(stride * height, 65536); - bo->meta.offsets[0] = 0; + /* MTL needs 1MB Alignment */ + bo->meta.sizes[plane] = ALIGN(stride * height, 0x100000); + if (plane == 1 && + (format == DRM_FORMAT_NV12 || format == DRM_FORMAT_P010)) { + alignment = 1 << 20; + offset += alignment - (offset % alignment); + tmpoffset = offset; + val = roundup_power_of_two(stride); + if ((stride * val) > tmpoffset) + offset = stride * val; + } + + bo->meta.offsets[plane] = offset; + offset += bo->meta.sizes[plane]; + } /* Aux buffer is linear and page aligned. It is placed after * other planes and aligned to main buffer stride. */ - bo->meta.strides[1] = bo->meta.strides[0] / 8; + for (a_plane = 0; a_plane < plane; a_plane++) { + stride = bo->meta.strides[a_plane] / 8; + bo->meta.strides[a_plane + plane] = stride; - /* Aligned to page size */ - bo->meta.sizes[1] = ALIGN(bo->meta.sizes[0] / 256, getpagesize()); - bo->meta.offsets[1] = bo->meta.sizes[0]; - /* Total number of planes & sizes */ - bo->meta.num_planes = 2; - bo->meta.total_size = bo->meta.sizes[0] + bo->meta.sizes[1]; + /* Aligned to page size */ + bo->meta.sizes[a_plane + plane] = + ALIGN(bo->meta.sizes[a_plane] / 256, getpagesize()); + bo->meta.offsets[a_plane + plane] = offset; + /* next buffer offset */ + offset += bo->meta.sizes[plane + a_plane]; + } + bo->meta.num_planes = a_plane + plane; + bo->meta.total_size = offset; } else { return i915_bo_from_format(bo, width, height, format); } @@ -880,8 +922,7 @@ static int i915_bo_create_from_metadata(struct bo *bo) } } - bo->meta.cached = (i915->has_llc || i915->is_mtl) && - !(bo->meta.use_flags & BO_USE_SCANOUT); + bo->meta.cached = (i915->has_llc || i915->is_mtl) && !(bo->meta.use_flags & BO_USE_SCANOUT); return 0; } @@ -933,7 +974,8 @@ static void *i915_bo_map(struct bo *bo, struct vma *vma, uint32_t map_flags) (bo->meta.format_modifier == I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS) || (bo->meta.format_modifier == I915_FORMAT_MOD_Y_TILED_GEN12_MC_CCS) || (bo->meta.format_modifier == I915_FORMAT_MOD_4_TILED) || - (bo->meta.format_modifier == I915_FORMAT_MOD_4_TILED_MTL_RC_CCS)) + (bo->meta.format_modifier == I915_FORMAT_MOD_4_TILED_MTL_RC_CCS) || + (bo->meta.format_modifier == I915_FORMAT_MOD_4_TILED_MTL_MC_CCS)) return MAP_FAILED; if (bo->meta.tiling == I915_TILING_NONE) { diff --git a/mediatek.c b/mediatek.c index 6ebde6e..64661fb 100644 --- a/mediatek.c +++ b/mediatek.c @@ -39,7 +39,8 @@ defined(MTK_MT8186) || \ defined(MTK_MT8188G) || \ defined(MTK_MT8192) || \ - defined(MTK_MT8195) + defined(MTK_MT8195) || \ + defined(MTK_MT8196) // clang-format on #define USE_NV12_FOR_HW_VIDEO_DECODING #define SUPPORT_FP16_AND_10BIT_ABGR @@ -324,11 +325,25 @@ static int mediatek_bo_create_with_modifiers(struct bo *bo, uint32_t width, uint #endif } - gem_create.size = bo->meta.total_size; - /* For protected data buffer needs to be allocated from GEM */ - if (bo->meta.use_flags & BO_USE_PROTECTED) + if (bo->meta.use_flags & BO_USE_PROTECTED) { + if (format == DRM_FORMAT_P010) { + /* + * Adjust the size so we don't waste tons of space. This was allocated + * with 16 bpp, but we only need 10 bpp. We can safely divide by 8 because + * we are aligned at a multiple higher than that. + */ + bo->meta.strides[0] = bo->meta.strides[0] * 10 / 16; + bo->meta.strides[1] = bo->meta.strides[1] * 10 / 16; + bo->meta.sizes[0] = bo->meta.sizes[0] * 10 / 16; + bo->meta.sizes[1] = bo->meta.sizes[1] * 10 / 16; + bo->meta.offsets[1] = bo->meta.sizes[0]; + bo->meta.total_size = bo->meta.total_size * 10 / 16; + } gem_create.flags |= DRM_MTK_GEM_CREATE_ENCRYPTED; + } + + gem_create.size = bo->meta.total_size; ret = drmIoctl(bo->drv->fd, DRM_IOCTL_MTK_GEM_CREATE, &gem_create); if (ret) { diff --git a/msm.c b/msm.c index d6f74cc..ee585c7 100644 --- a/msm.c +++ b/msm.c @@ -152,6 +152,13 @@ static void msm_calculate_layout(struct bo *bo) /* Calculate size and assign stride, size, offset to each plane based on format */ drv_bo_from_format(bo, stride, 1, alignh, bo->meta.format); + if (bo->meta.format == DRM_FORMAT_YVU420_ANDROID || + bo->meta.format == DRM_FORMAT_YVU420) { + const uint32_t u_size = + drv_size_from_format(bo->meta.format, bo->meta.strides[2], alignh, 2); + const uint32_t padding = ALIGN(u_size, PLANE_SIZE_ALIGN) - u_size; + bo->meta.total_size += padding; + } /* For all RGB UBWC formats */ if (bo->meta.tiling == MSM_UBWC_TILING) { diff --git a/util.h b/util.h index 8f8bb0d..aa1e79e 100644 --- a/util.h +++ b/util.h @@ -10,9 +10,9 @@ #define MAX(A, B) ((A) > (B) ? (A) : (B)) #define ARRAY_SIZE(A) (sizeof(A) / sizeof(*(A))) #define PUBLIC __attribute__((visibility("default"))) -#define ALIGN(A, B) (((A) + (B)-1) & ~((B)-1)) +#define ALIGN(A, B) (((A) + (B) - 1) & ~((B) - 1)) #define IS_ALIGNED(A, B) (ALIGN((A), (B)) == (A)) -#define DIV_ROUND_UP(n, d) (((n) + (d)-1) / (d)) +#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d)) #define STRINGIZE_NO_EXPANSION(x) #x #define STRINGIZE(x) STRINGIZE_NO_EXPANSION(x) diff --git a/virtgpu_virgl.c b/virtgpu_virgl.c index 222703e..1af4bae 100644 --- a/virtgpu_virgl.c +++ b/virtgpu_virgl.c @@ -52,6 +52,11 @@ static const uint32_t texture_source_formats[] = { DRM_FORMAT_ABGR2101010, DRM_FORMAT_ABGR16161616F }; +static const uint32_t depth_stencil_formats[] = { + DRM_FORMAT_DEPTH16, DRM_FORMAT_DEPTH24, DRM_FORMAT_DEPTH24_STENCIL8, + DRM_FORMAT_DEPTH32, DRM_FORMAT_DEPTH32_STENCIL8 +}; + extern struct virtgpu_param params[]; struct virgl_blob_metadata_cache { @@ -108,6 +113,16 @@ static uint32_t translate_format(uint32_t drm_fourcc) case DRM_FORMAT_YVU420: case DRM_FORMAT_YVU420_ANDROID: return VIRGL_FORMAT_YV12; + case DRM_FORMAT_DEPTH16: + return VIRGL_FORMAT_Z16_UNORM; + case DRM_FORMAT_DEPTH24: + return VIRGL_FORMAT_Z24X8_UNORM; + case DRM_FORMAT_DEPTH24_STENCIL8: + return VIRGL_FORMAT_Z24_UNORM_S8_UINT; + case DRM_FORMAT_DEPTH32: + return VIRGL_FORMAT_Z32_FLOAT; + case DRM_FORMAT_DEPTH32_STENCIL8: + return VIRGL_FORMAT_Z32_FLOAT_S8X24_UINT; default: drv_loge("Unhandled format:%d\n", drm_fourcc); return 0; @@ -639,6 +654,9 @@ static int virgl_init(struct driver *drv) virgl_add_combinations(drv, texture_source_formats, ARRAY_SIZE(texture_source_formats), &LINEAR_METADATA, BO_USE_TEXTURE_MASK); + virgl_add_combinations(drv, depth_stencil_formats, + ARRAY_SIZE(depth_stencil_formats), &LINEAR_METADATA, + BO_USE_RENDER_MASK | BO_USE_TEXTURE_MASK); /* NV12 with scanout must flow through virgl_add_combination, so that the native * support is checked and scanout use_flag can be conditionally stripped. */ virgl_add_combination(drv, DRM_FORMAT_NV12, &LINEAR_METADATA,