From d5b1a78a772f1e31a94f8babfa964152ec5e9aa5 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Mon, 30 Nov 2015 12:13:37 -0800 Subject: drm/vc4: Add support for drawing 3D frames. The user submission is basically a pointer to a command list and a pointer to uniforms. We copy those in to the kernel, validate and relocate them, and store the result in a GPU BO which we queue for execution. v2: Drop support for NV shader recs (not necessary for GL), simplify vc4_use_bo(), improve bin flush/semaphore checks, use __u32 style types. Signed-off-by: Eric Anholt --- include/uapi/drm/vc4_drm.h | 141 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 141 insertions(+) (limited to 'include/uapi') diff --git a/include/uapi/drm/vc4_drm.h b/include/uapi/drm/vc4_drm.h index 74de184..fe4161b 100644 --- a/include/uapi/drm/vc4_drm.h +++ b/include/uapi/drm/vc4_drm.h @@ -26,14 +26,155 @@ #include "drm.h" +#define DRM_VC4_SUBMIT_CL 0x00 +#define DRM_VC4_WAIT_SEQNO 0x01 +#define DRM_VC4_WAIT_BO 0x02 #define DRM_VC4_CREATE_BO 0x03 #define DRM_VC4_MMAP_BO 0x04 #define DRM_VC4_CREATE_SHADER_BO 0x05 +#define DRM_IOCTL_VC4_SUBMIT_CL DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_SUBMIT_CL, struct drm_vc4_submit_cl) +#define DRM_IOCTL_VC4_WAIT_SEQNO DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_WAIT_SEQNO, struct drm_vc4_wait_seqno) +#define DRM_IOCTL_VC4_WAIT_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_WAIT_BO, struct drm_vc4_wait_bo) #define DRM_IOCTL_VC4_CREATE_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_CREATE_BO, struct drm_vc4_create_bo) #define DRM_IOCTL_VC4_MMAP_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_MMAP_BO, struct drm_vc4_mmap_bo) #define DRM_IOCTL_VC4_CREATE_SHADER_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_CREATE_SHADER_BO, struct drm_vc4_create_shader_bo) +struct drm_vc4_submit_rcl_surface { + __u32 hindex; /* Handle index, or ~0 if not present. */ + __u32 offset; /* Offset to start of buffer. */ + /* + * Bits for either render config (color_write) or load/store packet. + * Bits should all be 0 for MSAA load/stores. + */ + __u16 bits; + +#define VC4_SUBMIT_RCL_SURFACE_READ_IS_FULL_RES (1 << 0) + __u16 flags; +}; + +/** + * struct drm_vc4_submit_cl - ioctl argument for submitting commands to the 3D + * engine. + * + * Drivers typically use GPU BOs to store batchbuffers / command lists and + * their associated state. However, because the VC4 lacks an MMU, we have to + * do validation of memory accesses by the GPU commands. If we were to store + * our commands in BOs, we'd need to do uncached readback from them to do the + * validation process, which is too expensive. Instead, userspace accumulates + * commands and associated state in plain memory, then the kernel copies the + * data to its own address space, and then validates and stores it in a GPU + * BO. + */ +struct drm_vc4_submit_cl { + /* Pointer to the binner command list. + * + * This is the first set of commands executed, which runs the + * coordinate shader to determine where primitives land on the screen, + * then writes out the state updates and draw calls necessary per tile + * to the tile allocation BO. + */ + __u64 bin_cl; + + /* Pointer to the shader records. + * + * Shader records are the structures read by the hardware that contain + * pointers to uniforms, shaders, and vertex attributes. The + * reference to the shader record has enough information to determine + * how many pointers are necessary (fixed number for shaders/uniforms, + * and an attribute count), so those BO indices into bo_handles are + * just stored as __u32s before each shader record passed in. + */ + __u64 shader_rec; + + /* Pointer to uniform data and texture handles for the textures + * referenced by the shader. + * + * For each shader state record, there is a set of uniform data in the + * order referenced by the record (FS, VS, then CS). Each set of + * uniform data has a __u32 index into bo_handles per texture + * sample operation, in the order the QPU_W_TMUn_S writes appear in + * the program. Following the texture BO handle indices is the actual + * uniform data. + * + * The individual uniform state blocks don't have sizes passed in, + * because the kernel has to determine the sizes anyway during shader + * code validation. + */ + __u64 uniforms; + __u64 bo_handles; + + /* Size in bytes of the binner command list. */ + __u32 bin_cl_size; + /* Size in bytes of the set of shader records. */ + __u32 shader_rec_size; + /* Number of shader records. + * + * This could just be computed from the contents of shader_records and + * the address bits of references to them from the bin CL, but it + * keeps the kernel from having to resize some allocations it makes. + */ + __u32 shader_rec_count; + /* Size in bytes of the uniform state. */ + __u32 uniforms_size; + + /* Number of BO handles passed in (size is that times 4). */ + __u32 bo_handle_count; + + /* RCL setup: */ + __u16 width; + __u16 height; + __u8 min_x_tile; + __u8 min_y_tile; + __u8 max_x_tile; + __u8 max_y_tile; + struct drm_vc4_submit_rcl_surface color_read; + struct drm_vc4_submit_rcl_surface color_write; + struct drm_vc4_submit_rcl_surface zs_read; + struct drm_vc4_submit_rcl_surface zs_write; + struct drm_vc4_submit_rcl_surface msaa_color_write; + struct drm_vc4_submit_rcl_surface msaa_zs_write; + __u32 clear_color[2]; + __u32 clear_z; + __u8 clear_s; + + __u32 pad:24; + +#define VC4_SUBMIT_CL_USE_CLEAR_COLOR (1 << 0) + __u32 flags; + + /* Returned value of the seqno of this render job (for the + * wait ioctl). + */ + __u64 seqno; +}; + +/** + * struct drm_vc4_wait_seqno - ioctl argument for waiting for + * DRM_VC4_SUBMIT_CL completion using its returned seqno. + * + * timeout_ns is the timeout in nanoseconds, where "0" means "don't + * block, just return the status." + */ +struct drm_vc4_wait_seqno { + __u64 seqno; + __u64 timeout_ns; +}; + +/** + * struct drm_vc4_wait_bo - ioctl argument for waiting for + * completion of the last DRM_VC4_SUBMIT_CL on a BO. + * + * This is useful for cases where multiple processes might be + * rendering to a BO and you want to wait for all rendering to be + * completed. + */ +struct drm_vc4_wait_bo { + __u32 handle; + __u32 pad; + __u64 timeout_ns; +}; + /** * struct drm_vc4_create_bo - ioctl argument for creating VC4 BOs. * -- cgit v1.1