diff options
Diffstat (limited to 'drivers/gpu/drm/vc4/vc4_validate.c')
| -rw-r--r-- | drivers/gpu/drm/vc4/vc4_validate.c | 147 |
1 files changed, 77 insertions, 70 deletions
diff --git a/drivers/gpu/drm/vc4/vc4_validate.c b/drivers/gpu/drm/vc4/vc4_validate.c index 814b512c6b9a..545c4c3608f5 100644 --- a/drivers/gpu/drm/vc4/vc4_validate.c +++ b/drivers/gpu/drm/vc4/vc4_validate.c @@ -43,6 +43,8 @@ * to use) happens. */ +#include <drm/drm_print.h> + #include "uapi/drm/vc4_drm.h" #include "vc4_drv.h" #include "vc4_packet.h" @@ -65,7 +67,7 @@ utile_width(int cpp) case 8: return 2; default: - DRM_ERROR("unknown cpp: %d\n", cpp); + pr_err("unknown cpp: %d\n", cpp); return 1; } } @@ -82,7 +84,7 @@ utile_height(int cpp) case 8: return 4; default: - DRM_ERROR("unknown cpp: %d\n", cpp); + pr_err("unknown cpp: %d\n", cpp); return 1; } } @@ -102,22 +104,26 @@ size_is_lt(uint32_t width, uint32_t height, int cpp) height <= 4 * utile_height(cpp)); } -struct drm_gem_cma_object * +struct drm_gem_dma_object * vc4_use_bo(struct vc4_exec_info *exec, uint32_t hindex) { - struct drm_gem_cma_object *obj; + struct vc4_dev *vc4 = exec->dev; + struct drm_gem_dma_object *obj; struct vc4_bo *bo; + if (WARN_ON_ONCE(vc4->gen > VC4_GEN_4)) + return NULL; + if (hindex >= exec->bo_count) { - DRM_ERROR("BO index %d greater than BO count %d\n", + DRM_DEBUG("BO index %d greater than BO count %d\n", hindex, exec->bo_count); return NULL; } - obj = exec->bo[hindex]; + obj = to_drm_gem_dma_obj(exec->bo[hindex]); bo = to_vc4_bo(&obj->base); if (bo->validated_shader) { - DRM_ERROR("Trying to use shader BO as something other than " + DRM_DEBUG("Trying to use shader BO as something other than " "a shader\n"); return NULL; } @@ -125,7 +131,7 @@ vc4_use_bo(struct vc4_exec_info *exec, uint32_t hindex) return obj; } -static struct drm_gem_cma_object * +static struct drm_gem_dma_object * vc4_use_handle(struct vc4_exec_info *exec, uint32_t gem_handles_packet_index) { return vc4_use_bo(exec, exec->bo_index[gem_handles_packet_index]); @@ -156,14 +162,18 @@ gl_shader_rec_size(uint32_t pointer_bits) } bool -vc4_check_tex_size(struct vc4_exec_info *exec, struct drm_gem_cma_object *fbo, +vc4_check_tex_size(struct vc4_exec_info *exec, struct drm_gem_dma_object *fbo, uint32_t offset, uint8_t tiling_format, uint32_t width, uint32_t height, uint8_t cpp) { + struct vc4_dev *vc4 = exec->dev; uint32_t aligned_width, aligned_height, stride, size; uint32_t utile_w = utile_width(cpp); uint32_t utile_h = utile_height(cpp); + if (WARN_ON_ONCE(vc4->gen > VC4_GEN_4)) + return false; + /* The shaded vertex format stores signed 12.4 fixed point * (-2048,2047) offsets from the viewport center, so we should * never have a render target larger than 4096. The texture @@ -172,7 +182,7 @@ vc4_check_tex_size(struct vc4_exec_info *exec, struct drm_gem_cma_object *fbo, * our math. */ if (width > 4096 || height > 4096) { - DRM_ERROR("Surface dimensions (%d,%d) too large", + DRM_DEBUG("Surface dimensions (%d,%d) too large", width, height); return false; } @@ -191,7 +201,7 @@ vc4_check_tex_size(struct vc4_exec_info *exec, struct drm_gem_cma_object *fbo, aligned_height = round_up(height, utile_h); break; default: - DRM_ERROR("buffer tiling %d unsupported\n", tiling_format); + DRM_DEBUG("buffer tiling %d unsupported\n", tiling_format); return false; } @@ -200,7 +210,7 @@ vc4_check_tex_size(struct vc4_exec_info *exec, struct drm_gem_cma_object *fbo, if (size + offset < size || size + offset > fbo->base.size) { - DRM_ERROR("Overflow in %dx%d (%dx%d) fbo size (%d + %d > %zd)\n", + DRM_DEBUG("Overflow in %dx%d (%dx%d) fbo size (%d + %d > %zd)\n", width, height, aligned_width, aligned_height, size, offset, fbo->base.size); @@ -214,7 +224,7 @@ static int validate_flush(VALIDATE_ARGS) { if (!validate_bin_pos(exec, untrusted, exec->args->bin_cl_size - 1)) { - DRM_ERROR("Bin CL must end with VC4_PACKET_FLUSH\n"); + DRM_DEBUG("Bin CL must end with VC4_PACKET_FLUSH\n"); return -EINVAL; } exec->found_flush = true; @@ -226,13 +236,13 @@ static int validate_start_tile_binning(VALIDATE_ARGS) { if (exec->found_start_tile_binning_packet) { - DRM_ERROR("Duplicate VC4_PACKET_START_TILE_BINNING\n"); + DRM_DEBUG("Duplicate VC4_PACKET_START_TILE_BINNING\n"); return -EINVAL; } exec->found_start_tile_binning_packet = true; if (!exec->found_tile_binning_mode_config_packet) { - DRM_ERROR("missing VC4_PACKET_TILE_BINNING_MODE_CONFIG\n"); + DRM_DEBUG("missing VC4_PACKET_TILE_BINNING_MODE_CONFIG\n"); return -EINVAL; } @@ -243,7 +253,7 @@ static int validate_increment_semaphore(VALIDATE_ARGS) { if (!validate_bin_pos(exec, untrusted, exec->args->bin_cl_size - 2)) { - DRM_ERROR("Bin CL must end with " + DRM_DEBUG("Bin CL must end with " "VC4_PACKET_INCREMENT_SEMAPHORE\n"); return -EINVAL; } @@ -255,7 +265,7 @@ validate_increment_semaphore(VALIDATE_ARGS) static int validate_indexed_prim_list(VALIDATE_ARGS) { - struct drm_gem_cma_object *ib; + struct drm_gem_dma_object *ib; uint32_t length = *(uint32_t *)(untrusted + 1); uint32_t offset = *(uint32_t *)(untrusted + 5); uint32_t max_index = *(uint32_t *)(untrusted + 9); @@ -264,7 +274,7 @@ validate_indexed_prim_list(VALIDATE_ARGS) /* Check overflow condition */ if (exec->shader_state_count == 0) { - DRM_ERROR("shader state must precede primitives\n"); + DRM_DEBUG("shader state must precede primitives\n"); return -EINVAL; } shader_state = &exec->shader_state[exec->shader_state_count - 1]; @@ -276,17 +286,14 @@ validate_indexed_prim_list(VALIDATE_ARGS) if (!ib) return -EINVAL; - exec->bin_dep_seqno = max(exec->bin_dep_seqno, - to_vc4_bo(&ib->base)->write_seqno); - if (offset > ib->base.size || (ib->base.size - offset) / index_size < length) { - DRM_ERROR("IB access overflow (%d + %d*%d > %zd)\n", + DRM_DEBUG("IB access overflow (%d + %d*%d > %zd)\n", offset, length, index_size, ib->base.size); return -EINVAL; } - *(uint32_t *)(validated + 5) = ib->paddr + offset; + *(uint32_t *)(validated + 5) = ib->dma_addr + offset; return 0; } @@ -301,13 +308,13 @@ validate_gl_array_primitive(VALIDATE_ARGS) /* Check overflow condition */ if (exec->shader_state_count == 0) { - DRM_ERROR("shader state must precede primitives\n"); + DRM_DEBUG("shader state must precede primitives\n"); return -EINVAL; } shader_state = &exec->shader_state[exec->shader_state_count - 1]; if (length + base_index < length) { - DRM_ERROR("primitive vertex count overflow\n"); + DRM_DEBUG("primitive vertex count overflow\n"); return -EINVAL; } max_index = length + base_index - 1; @@ -324,7 +331,7 @@ validate_gl_shader_state(VALIDATE_ARGS) uint32_t i = exec->shader_state_count++; if (i >= exec->shader_state_size) { - DRM_ERROR("More requests for shader states than declared\n"); + DRM_DEBUG("More requests for shader states than declared\n"); return -EINVAL; } @@ -332,7 +339,7 @@ validate_gl_shader_state(VALIDATE_ARGS) exec->shader_state[i].max_index = 0; if (exec->shader_state[i].addr & ~0xf) { - DRM_ERROR("high bits set in GL shader rec reference\n"); + DRM_DEBUG("high bits set in GL shader rec reference\n"); return -EINVAL; } @@ -356,7 +363,7 @@ validate_tile_binning_config(VALIDATE_ARGS) int bin_slot; if (exec->found_tile_binning_mode_config_packet) { - DRM_ERROR("Duplicate VC4_PACKET_TILE_BINNING_MODE_CONFIG\n"); + DRM_DEBUG("Duplicate VC4_PACKET_TILE_BINNING_MODE_CONFIG\n"); return -EINVAL; } exec->found_tile_binning_mode_config_packet = true; @@ -368,22 +375,22 @@ validate_tile_binning_config(VALIDATE_ARGS) if (exec->bin_tiles_x == 0 || exec->bin_tiles_y == 0) { - DRM_ERROR("Tile binning config of %dx%d too small\n", + DRM_DEBUG("Tile binning config of %dx%d too small\n", exec->bin_tiles_x, exec->bin_tiles_y); return -EINVAL; } if (flags & (VC4_BIN_CONFIG_DB_NON_MS | VC4_BIN_CONFIG_TILE_BUFFER_64BIT)) { - DRM_ERROR("unsupported binning config flags 0x%02x\n", flags); + DRM_DEBUG("unsupported binning config flags 0x%02x\n", flags); return -EINVAL; } bin_slot = vc4_v3d_get_bin_slot(vc4); if (bin_slot < 0) { if (bin_slot != -EINTR && bin_slot != -ERESTARTSYS) { - DRM_ERROR("Failed to allocate binner memory: %d\n", - bin_slot); + drm_err(dev, "Failed to allocate binner memory: %d\n", + bin_slot); } return bin_slot; } @@ -392,7 +399,7 @@ validate_tile_binning_config(VALIDATE_ARGS) * free when the job completes rendering. */ exec->bin_slots |= BIT(bin_slot); - bin_addr = vc4->bin_bo->base.paddr + bin_slot * vc4->bin_alloc_size; + bin_addr = vc4->bin_bo->base.dma_addr + bin_slot * vc4->bin_alloc_size; /* The tile state data array is 48 bytes per tile, and we put it at * the start of a BO containing both it and the tile alloc. @@ -482,10 +489,14 @@ vc4_validate_bin_cl(struct drm_device *dev, void *unvalidated, struct vc4_exec_info *exec) { + struct vc4_dev *vc4 = to_vc4_dev(dev); uint32_t len = exec->args->bin_cl_size; uint32_t dst_offset = 0; uint32_t src_offset = 0; + if (WARN_ON_ONCE(vc4->gen > VC4_GEN_4)) + return -ENODEV; + while (src_offset < len) { void *dst_pkt = validated + dst_offset; void *src_pkt = unvalidated + src_offset; @@ -493,20 +504,20 @@ vc4_validate_bin_cl(struct drm_device *dev, const struct cmd_info *info; if (cmd >= ARRAY_SIZE(cmd_info)) { - DRM_ERROR("0x%08x: packet %d out of bounds\n", + DRM_DEBUG("0x%08x: packet %d out of bounds\n", src_offset, cmd); return -EINVAL; } info = &cmd_info[cmd]; if (!info->name) { - DRM_ERROR("0x%08x: packet %d invalid\n", + DRM_DEBUG("0x%08x: packet %d invalid\n", src_offset, cmd); return -EINVAL; } if (src_offset + info->len > len) { - DRM_ERROR("0x%08x: packet %d (%s) length 0x%08x " + DRM_DEBUG("0x%08x: packet %d (%s) length 0x%08x " "exceeds bounds (0x%08x)\n", src_offset, cmd, info->name, info->len, src_offset + len); @@ -519,7 +530,7 @@ vc4_validate_bin_cl(struct drm_device *dev, if (info->func && info->func(exec, dst_pkt + 1, src_pkt + 1)) { - DRM_ERROR("0x%08x: packet %d (%s) failed to validate\n", + DRM_DEBUG("0x%08x: packet %d (%s) failed to validate\n", src_offset, cmd, info->name); return -EINVAL; } @@ -537,7 +548,7 @@ vc4_validate_bin_cl(struct drm_device *dev, exec->ct0ea = exec->ct0ca + dst_offset; if (!exec->found_start_tile_binning_packet) { - DRM_ERROR("Bin CL missing VC4_PACKET_START_TILE_BINNING\n"); + DRM_DEBUG("Bin CL missing VC4_PACKET_START_TILE_BINNING\n"); return -EINVAL; } @@ -549,7 +560,7 @@ vc4_validate_bin_cl(struct drm_device *dev, * semaphore increment. */ if (!exec->found_increment_semaphore_packet || !exec->found_flush) { - DRM_ERROR("Bin CL missing VC4_PACKET_INCREMENT_SEMAPHORE + " + DRM_DEBUG("Bin CL missing VC4_PACKET_INCREMENT_SEMAPHORE + " "VC4_PACKET_FLUSH\n"); return -EINVAL; } @@ -563,7 +574,7 @@ reloc_tex(struct vc4_exec_info *exec, struct vc4_texture_sample_info *sample, uint32_t texture_handle_index, bool is_cs) { - struct drm_gem_cma_object *tex; + struct drm_gem_dma_object *tex; uint32_t p0 = *(uint32_t *)(uniform_data_u + sample->p_offset[0]); uint32_t p1 = *(uint32_t *)(uniform_data_u + sample->p_offset[1]); uint32_t p2 = (sample->p_offset[2] != ~0 ? @@ -588,15 +599,15 @@ reloc_tex(struct vc4_exec_info *exec, uint32_t remaining_size = tex->base.size - p0; if (p0 > tex->base.size - 4) { - DRM_ERROR("UBO offset greater than UBO size\n"); + DRM_DEBUG("UBO offset greater than UBO size\n"); goto fail; } if (p1 > remaining_size - 4) { - DRM_ERROR("UBO clamp would allow reads " + DRM_DEBUG("UBO clamp would allow reads " "outside of UBO\n"); goto fail; } - *validated_p0 = tex->paddr + p0; + *validated_p0 = tex->dma_addr + p0; return true; } @@ -612,14 +623,14 @@ reloc_tex(struct vc4_exec_info *exec, if (VC4_GET_FIELD(p3, VC4_TEX_P2_PTYPE) == VC4_TEX_P2_PTYPE_CUBE_MAP_STRIDE) { if (cube_map_stride) { - DRM_ERROR("Cube map stride set twice\n"); + DRM_DEBUG("Cube map stride set twice\n"); goto fail; } cube_map_stride = p3 & VC4_TEX_P2_CMST_MASK; } if (!cube_map_stride) { - DRM_ERROR("Cube map stride not set\n"); + DRM_DEBUG("Cube map stride not set\n"); goto fail; } } @@ -660,7 +671,7 @@ reloc_tex(struct vc4_exec_info *exec, case VC4_TEXTURE_TYPE_RGBA64: case VC4_TEXTURE_TYPE_YUV422R: default: - DRM_ERROR("Texture format %d unsupported\n", type); + DRM_DEBUG("Texture format %d unsupported\n", type); goto fail; } utile_w = utile_width(cpp); @@ -713,7 +724,7 @@ reloc_tex(struct vc4_exec_info *exec, level_size = aligned_width * cpp * aligned_height; if (offset < level_size) { - DRM_ERROR("Level %d (%dx%d -> %dx%d) size %db " + DRM_DEBUG("Level %d (%dx%d -> %dx%d) size %db " "overflowed buffer bounds (offset %d)\n", i, level_width, level_height, aligned_width, aligned_height, @@ -724,12 +735,7 @@ reloc_tex(struct vc4_exec_info *exec, offset -= level_size; } - *validated_p0 = tex->paddr + p0; - - if (is_cs) { - exec->bin_dep_seqno = max(exec->bin_dep_seqno, - to_vc4_bo(&tex->base)->write_seqno); - } + *validated_p0 = tex->dma_addr + p0; return true; fail: @@ -753,7 +759,7 @@ validate_gl_shader_rec(struct drm_device *dev, 28, /* cs */ }; uint32_t shader_reloc_count = ARRAY_SIZE(shader_reloc_offsets); - struct drm_gem_cma_object *bo[shader_reloc_count + 8]; + struct drm_gem_dma_object *bo[ARRAY_SIZE(shader_reloc_offsets) + 8]; uint32_t nr_attributes, nr_relocs, packet_size; int i; @@ -764,7 +770,7 @@ validate_gl_shader_rec(struct drm_device *dev, nr_relocs = ARRAY_SIZE(shader_reloc_offsets) + nr_attributes; if (nr_relocs * 4 > exec->shader_rec_size) { - DRM_ERROR("overflowed shader recs reading %d handles " + DRM_DEBUG("overflowed shader recs reading %d handles " "from %d bytes left\n", nr_relocs, exec->shader_rec_size); return -EINVAL; @@ -774,7 +780,7 @@ validate_gl_shader_rec(struct drm_device *dev, exec->shader_rec_size -= nr_relocs * 4; if (packet_size > exec->shader_rec_size) { - DRM_ERROR("overflowed shader recs copying %db packet " + DRM_DEBUG("overflowed shader recs copying %db packet " "from %d bytes left\n", packet_size, exec->shader_rec_size); return -EINVAL; @@ -794,11 +800,11 @@ validate_gl_shader_rec(struct drm_device *dev, for (i = 0; i < shader_reloc_count; i++) { if (src_handles[i] > exec->bo_count) { - DRM_ERROR("Shader handle %d too big\n", src_handles[i]); + DRM_DEBUG("Shader handle %d too big\n", src_handles[i]); return -EINVAL; } - bo[i] = exec->bo[src_handles[i]]; + bo[i] = to_drm_gem_dma_obj(exec->bo[src_handles[i]]); if (!bo[i]) return -EINVAL; } @@ -810,13 +816,13 @@ validate_gl_shader_rec(struct drm_device *dev, if (((*(uint16_t *)pkt_u & VC4_SHADER_FLAG_FS_SINGLE_THREAD) == 0) != to_vc4_bo(&bo[0]->base)->validated_shader->is_threaded) { - DRM_ERROR("Thread mode of CL and FS do not match\n"); + DRM_DEBUG("Thread mode of CL and FS do not match\n"); return -EINVAL; } if (to_vc4_bo(&bo[1]->base)->validated_shader->is_threaded || to_vc4_bo(&bo[2]->base)->validated_shader->is_threaded) { - DRM_ERROR("cs and vs cannot be threaded\n"); + DRM_DEBUG("cs and vs cannot be threaded\n"); return -EINVAL; } @@ -828,10 +834,10 @@ validate_gl_shader_rec(struct drm_device *dev, void *uniform_data_u; uint32_t tex, uni; - *(uint32_t *)(pkt_v + o) = bo[i]->paddr + src_offset; + *(uint32_t *)(pkt_v + o) = bo[i]->dma_addr + src_offset; if (src_offset != 0) { - DRM_ERROR("Shaders must be at offset 0 of " + DRM_DEBUG("Shaders must be at offset 0 of " "the BO.\n"); return -EINVAL; } @@ -842,7 +848,7 @@ validate_gl_shader_rec(struct drm_device *dev, if (validated_shader->uniforms_src_size > exec->uniforms_size) { - DRM_ERROR("Uniforms src buffer overflow\n"); + DRM_DEBUG("Uniforms src buffer overflow\n"); return -EINVAL; } @@ -884,7 +890,7 @@ validate_gl_shader_rec(struct drm_device *dev, } for (i = 0; i < nr_attributes; i++) { - struct drm_gem_cma_object *vbo = + struct drm_gem_dma_object *vbo = bo[ARRAY_SIZE(shader_reloc_offsets) + i]; uint32_t o = 36 + i * 8; uint32_t offset = *(uint32_t *)(pkt_u + o + 0); @@ -892,15 +898,12 @@ validate_gl_shader_rec(struct drm_device *dev, uint32_t stride = *(uint8_t *)(pkt_u + o + 5); uint32_t max_index; - exec->bin_dep_seqno = max(exec->bin_dep_seqno, - to_vc4_bo(&vbo->base)->write_seqno); - if (state->addr & 0x8) stride |= (*(uint32_t *)(pkt_u + 100 + i * 4)) & ~0xff; if (vbo->base.size < offset || vbo->base.size - offset < attr_size) { - DRM_ERROR("BO offset overflow (%d + %d > %zu)\n", + DRM_DEBUG("BO offset overflow (%d + %d > %zu)\n", offset, attr_size, vbo->base.size); return -EINVAL; } @@ -909,14 +912,14 @@ validate_gl_shader_rec(struct drm_device *dev, max_index = ((vbo->base.size - offset - attr_size) / stride); if (state->max_index > max_index) { - DRM_ERROR("primitives use index %d out of " + DRM_DEBUG("primitives use index %d out of " "supplied %d\n", state->max_index, max_index); return -EINVAL; } } - *(uint32_t *)(pkt_v + o) = vbo->paddr + offset; + *(uint32_t *)(pkt_v + o) = vbo->dma_addr + offset; } return 0; @@ -926,9 +929,13 @@ int vc4_validate_shader_recs(struct drm_device *dev, struct vc4_exec_info *exec) { + struct vc4_dev *vc4 = to_vc4_dev(dev); uint32_t i; int ret = 0; + if (WARN_ON_ONCE(vc4->gen > VC4_GEN_4)) + return -ENODEV; + for (i = 0; i < exec->shader_state_count; i++) { ret = validate_gl_shader_rec(dev, exec, &exec->shader_state[i]); if (ret) |
