/* * Copyright 2014 The Chromium OS Authors. All rights reserved. * Use of this source code is governed by a BSD-style license that can be * found in the LICENSE file. */ #ifdef DRV_TEGRA #include <assert.h> #include <stdio.h> #include <string.h> #include <sys/mman.h> #include <tegra_drm.h> #include <xf86drm.h> #include "drv_priv.h" #include "helpers.h" #include "util.h" /* * GOB (Group Of Bytes) is the basic unit of the blocklinear layout. * GOBs are arranged to blocks, where the height of the block (measured * in GOBs) is configurable. */ #define NV_BLOCKLINEAR_GOB_HEIGHT 8 #define NV_BLOCKLINEAR_GOB_WIDTH 64 #define NV_DEFAULT_BLOCK_HEIGHT_LOG2 4 #define NV_PREFERRED_PAGE_SIZE (128 * 1024) // clang-format off enum nv_mem_kind { NV_MEM_KIND_PITCH = 0, NV_MEM_KIND_C32_2CRA = 0xdb, NV_MEM_KIND_GENERIC_16Bx2 = 0xfe, }; enum tegra_map_type { TEGRA_READ_TILED_BUFFER = 0, TEGRA_WRITE_TILED_BUFFER = 1, }; // clang-format on struct tegra_private_map_data { void *tiled; void *untiled; }; static const uint32_t render_target_formats[] = { DRM_FORMAT_ARGB8888, DRM_FORMAT_XRGB8888 }; static int compute_block_height_log2(int height) { int block_height_log2 = NV_DEFAULT_BLOCK_HEIGHT_LOG2; if (block_height_log2 > 0) { /* Shrink, if a smaller block height could cover the whole * surface height. */ int proposed = NV_BLOCKLINEAR_GOB_HEIGHT << (block_height_log2 - 1); while (proposed >= height) { block_height_log2--; if (block_height_log2 == 0) break; proposed /= 2; } } return block_height_log2; } static void compute_layout_blocklinear(int width, int height, int format, enum nv_mem_kind *kind, uint32_t *block_height_log2, uint32_t *stride, uint32_t *size) { int pitch = drv_stride_from_format(format, width, 0); /* Align to blocklinear blocks. */ pitch = ALIGN(pitch, NV_BLOCKLINEAR_GOB_WIDTH); /* Compute padded height. */ *block_height_log2 = compute_block_height_log2(height); int block_height = 1 << *block_height_log2; int padded_height = ALIGN(height, NV_BLOCKLINEAR_GOB_HEIGHT * block_height); int bytes = pitch * padded_height; /* Pad the allocation to the preferred page size. * This will reduce the required page table size (see discussion in NV * bug 1321091), and also acts as a WAR for NV bug 1325421. */ bytes = ALIGN(bytes, NV_PREFERRED_PAGE_SIZE); *kind = NV_MEM_KIND_C32_2CRA; *stride = pitch; *size = bytes; } static void compute_layout_linear(int width, int height, int format, uint32_t *stride, uint32_t *size) { *stride = ALIGN(drv_stride_from_format(format, width, 0), 64); *size = *stride * height; } static void transfer_tile(struct bo *bo, uint8_t *tiled, uint8_t *untiled, enum tegra_map_type type, uint32_t bytes_per_pixel, uint32_t gob_top, uint32_t gob_left, uint32_t gob_size_pixels, uint8_t *tiled_last) { uint8_t *tmp; uint32_t x, y, k; for (k = 0; k < gob_size_pixels; k++) { /* * Given the kth pixel starting from the tile specified by * gob_top and gob_left, unswizzle to get the standard (x, y) * representation. */ x = gob_left + (((k >> 3) & 8) | ((k >> 1) & 4) | (k & 3)); y = gob_top + ((k >> 7 << 3) | ((k >> 3) & 6) | ((k >> 2) & 1)); if (tiled >= tiled_last) return; if (x >= bo->width || y >= bo->height) { tiled += bytes_per_pixel; continue; } tmp = untiled + y * bo->strides[0] + x * bytes_per_pixel; if (type == TEGRA_READ_TILED_BUFFER) memcpy(tmp, tiled, bytes_per_pixel); else if (type == TEGRA_WRITE_TILED_BUFFER) memcpy(tiled, tmp, bytes_per_pixel); /* Move on to next pixel. */ tiled += bytes_per_pixel; } } static void transfer_tiled_memory(struct bo *bo, uint8_t *tiled, uint8_t *untiled, enum tegra_map_type type) { uint32_t gob_width, gob_height, gob_size_bytes, gob_size_pixels, gob_count_x, gob_count_y, gob_top, gob_left; uint32_t i, j, offset; uint8_t *tmp, *tiled_last; uint32_t bytes_per_pixel = drv_stride_from_format(bo->format, 1, 0); /* * The blocklinear format consists of 8*(2^n) x 64 byte sized tiles, * where 0 <= n <= 4. */ gob_width = DIV_ROUND_UP(NV_BLOCKLINEAR_GOB_WIDTH, bytes_per_pixel); gob_height = NV_BLOCKLINEAR_GOB_HEIGHT * (1 << NV_DEFAULT_BLOCK_HEIGHT_LOG2); /* Calculate the height from maximum possible gob height */ while (gob_height > NV_BLOCKLINEAR_GOB_HEIGHT && gob_height >= 2 * bo->height) gob_height /= 2; gob_size_bytes = gob_height * NV_BLOCKLINEAR_GOB_WIDTH; gob_size_pixels = gob_height * gob_width; gob_count_x = DIV_ROUND_UP(bo->strides[0], NV_BLOCKLINEAR_GOB_WIDTH); gob_count_y = DIV_ROUND_UP(bo->height, gob_height); tiled_last = tiled + bo->total_size; offset = 0; for (j = 0; j < gob_count_y; j++) { gob_top = j * gob_height; for (i = 0; i < gob_count_x; i++) { tmp = tiled + offset; gob_left = i * gob_width; transfer_tile(bo, tmp, untiled, type, bytes_per_pixel, gob_top, gob_left, gob_size_pixels, tiled_last); offset += gob_size_bytes; } } } static int tegra_init(struct driver *drv) { struct format_metadata metadata; uint64_t use_flags = BO_USE_RENDER_MASK; metadata.tiling = NV_MEM_KIND_PITCH; metadata.priority = 1; metadata.modifier = DRM_FORMAT_MOD_LINEAR; drv_add_combinations(drv, render_target_formats, ARRAY_SIZE(render_target_formats), &metadata, use_flags); drv_modify_combination(drv, DRM_FORMAT_XRGB8888, &metadata, BO_USE_CURSOR | BO_USE_SCANOUT); drv_modify_combination(drv, DRM_FORMAT_ARGB8888, &metadata, BO_USE_CURSOR | BO_USE_SCANOUT); use_flags &= ~BO_USE_SW_WRITE_OFTEN; use_flags &= ~BO_USE_SW_READ_OFTEN; use_flags &= ~BO_USE_LINEAR; metadata.tiling = NV_MEM_KIND_C32_2CRA; metadata.priority = 2; drv_add_combinations(drv, render_target_formats, ARRAY_SIZE(render_target_formats), &metadata, use_flags); drv_modify_combination(drv, DRM_FORMAT_XRGB8888, &metadata, BO_USE_SCANOUT); drv_modify_combination(drv, DRM_FORMAT_ARGB8888, &metadata, BO_USE_SCANOUT); return 0; } static int tegra_bo_create(struct bo *bo, uint32_t width, uint32_t height, uint32_t format, uint64_t use_flags) { uint32_t size, stride, block_height_log2 = 0; enum nv_mem_kind kind = NV_MEM_KIND_PITCH; struct drm_tegra_gem_create gem_create; int ret; if (use_flags & (BO_USE_CURSOR | BO_USE_LINEAR | BO_USE_SW_READ_OFTEN | BO_USE_SW_WRITE_OFTEN)) compute_layout_linear(width, height, format, &stride, &size); else compute_layout_blocklinear(width, height, format, &kind, &block_height_log2, &stride, &size); memset(&gem_create, 0, sizeof(gem_create)); gem_create.size = size; gem_create.flags = 0; ret = drmIoctl(bo->drv->fd, DRM_IOCTL_TEGRA_GEM_CREATE, &gem_create); if (ret) { drv_log("DRM_IOCTL_TEGRA_GEM_CREATE failed (size=%zu)\n", size); return ret; } bo->handles[0].u32 = gem_create.handle; bo->offsets[0] = 0; bo->total_size = bo->sizes[0] = size; bo->strides[0] = stride; if (kind != NV_MEM_KIND_PITCH) { struct drm_tegra_gem_set_tiling gem_tile; memset(&gem_tile, 0, sizeof(gem_tile)); gem_tile.handle = bo->handles[0].u32; gem_tile.mode = DRM_TEGRA_GEM_TILING_MODE_BLOCK; gem_tile.value = block_height_log2; ret = drmCommandWriteRead(bo->drv->fd, DRM_TEGRA_GEM_SET_TILING, &gem_tile, sizeof(gem_tile)); if (ret < 0) { drv_gem_bo_destroy(bo); return ret; } /* Encode blocklinear parameters for EGLImage creation. */ bo->tiling = (kind & 0xff) | ((block_height_log2 & 0xf) << 8); bo->format_modifiers[0] = fourcc_mod_code(NV, bo->tiling); } return 0; } static int tegra_bo_import(struct bo *bo, struct drv_import_fd_data *data) { int ret; struct drm_tegra_gem_get_tiling gem_get_tiling; ret = drv_prime_bo_import(bo, data); if (ret) return ret; /* TODO(gsingh): export modifiers and get rid of backdoor tiling. */ memset(&gem_get_tiling, 0, sizeof(gem_get_tiling)); gem_get_tiling.handle = bo->handles[0].u32; ret = drmIoctl(bo->drv->fd, DRM_IOCTL_TEGRA_GEM_GET_TILING, &gem_get_tiling); if (ret) { drv_gem_bo_destroy(bo); return ret; } /* NOTE(djmk): we only know about one tiled format, so if our drmIoctl call tells us we are tiled, assume it is this format (NV_MEM_KIND_C32_2CRA) otherwise linear (KIND_PITCH). */ if (gem_get_tiling.mode == DRM_TEGRA_GEM_TILING_MODE_PITCH) { bo->tiling = NV_MEM_KIND_PITCH; } else if (gem_get_tiling.mode == DRM_TEGRA_GEM_TILING_MODE_BLOCK) { bo->tiling = NV_MEM_KIND_C32_2CRA; } else { drv_log("%s: unknown tile format %d\n", __func__, gem_get_tiling.mode); drv_gem_bo_destroy(bo); assert(0); } bo->format_modifiers[0] = fourcc_mod_code(NV, bo->tiling); return 0; } static void *tegra_bo_map(struct bo *bo, struct vma *vma, size_t plane, uint32_t map_flags) { int ret; struct drm_tegra_gem_mmap gem_map; struct tegra_private_map_data *priv; memset(&gem_map, 0, sizeof(gem_map)); gem_map.handle = bo->handles[0].u32; ret = drmCommandWriteRead(bo->drv->fd, DRM_TEGRA_GEM_MMAP, &gem_map, sizeof(gem_map)); if (ret < 0) { drv_log("DRM_TEGRA_GEM_MMAP failed\n"); return MAP_FAILED; } void *addr = mmap(0, bo->total_size, drv_get_prot(map_flags), MAP_SHARED, bo->drv->fd, gem_map.offset); vma->length = bo->total_size; if ((bo->tiling & 0xFF) == NV_MEM_KIND_C32_2CRA && addr != MAP_FAILED) { priv = calloc(1, sizeof(*priv)); priv->untiled = calloc(1, bo->total_size); priv->tiled = addr; vma->priv = priv; transfer_tiled_memory(bo, priv->tiled, priv->untiled, TEGRA_READ_TILED_BUFFER); addr = priv->untiled; } return addr; } static int tegra_bo_unmap(struct bo *bo, struct vma *vma) { if (vma->priv) { struct tegra_private_map_data *priv = vma->priv; vma->addr = priv->tiled; free(priv->untiled); free(priv); vma->priv = NULL; } return munmap(vma->addr, vma->length); } static int tegra_bo_flush(struct bo *bo, struct mapping *mapping) { struct tegra_private_map_data *priv = mapping->vma->priv; if (priv && (mapping->vma->map_flags & BO_MAP_WRITE)) transfer_tiled_memory(bo, priv->tiled, priv->untiled, TEGRA_WRITE_TILED_BUFFER); return 0; } const struct backend backend_tegra = { .name = "tegra", .init = tegra_init, .bo_create = tegra_bo_create, .bo_destroy = drv_gem_bo_destroy, .bo_import = tegra_bo_import, .bo_map = tegra_bo_map, .bo_unmap = tegra_bo_unmap, .bo_flush = tegra_bo_flush, }; #endif