1 /* $NetBSD: v3d_drm.h,v 1.2 2021/12/18 23:45:46 riastradh Exp $ */ 2 3 /* 4 * Copyright © 2014-2018 Broadcom 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a 7 * copy of this software and associated documentation files (the "Software"), 8 * to deal in the Software without restriction, including without limitation 9 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 10 * and/or sell copies of the Software, and to permit persons to whom the 11 * Software is furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice (including the next 14 * paragraph) shall be included in all copies or substantial portions of the 15 * Software. 16 * 17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 22 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 23 * IN THE SOFTWARE. 24 */ 25 26 #ifndef _V3D_DRM_H_ 27 #define _V3D_DRM_H_ 28 29 #include "drm.h" 30 31 #if defined(__cplusplus) 32 extern "C" { 33 #endif 34 35 #define DRM_V3D_SUBMIT_CL 0x00 36 #define DRM_V3D_WAIT_BO 0x01 37 #define DRM_V3D_CREATE_BO 0x02 38 #define DRM_V3D_MMAP_BO 0x03 39 #define DRM_V3D_GET_PARAM 0x04 40 #define DRM_V3D_GET_BO_OFFSET 0x05 41 #define DRM_V3D_SUBMIT_TFU 0x06 42 #define DRM_V3D_SUBMIT_CSD 0x07 43 44 #define DRM_IOCTL_V3D_SUBMIT_CL DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CL, struct drm_v3d_submit_cl) 45 #define DRM_IOCTL_V3D_WAIT_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_WAIT_BO, struct drm_v3d_wait_bo) 46 #define DRM_IOCTL_V3D_CREATE_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_CREATE_BO, struct drm_v3d_create_bo) 47 #define DRM_IOCTL_V3D_MMAP_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_MMAP_BO, struct drm_v3d_mmap_bo) 48 #define DRM_IOCTL_V3D_GET_PARAM DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_GET_PARAM, struct drm_v3d_get_param) 49 #define DRM_IOCTL_V3D_GET_BO_OFFSET DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_GET_BO_OFFSET, struct drm_v3d_get_bo_offset) 50 #define DRM_IOCTL_V3D_SUBMIT_TFU DRM_IOW(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_TFU, struct drm_v3d_submit_tfu) 51 #define DRM_IOCTL_V3D_SUBMIT_CSD DRM_IOW(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CSD, struct drm_v3d_submit_csd) 52 53 #define DRM_V3D_SUBMIT_CL_FLUSH_CACHE 0x01 54 55 /** 56 * struct drm_v3d_submit_cl - ioctl argument for submitting commands to the 3D 57 * engine. 58 * 59 * This asks the kernel to have the GPU execute an optional binner 60 * command list, and a render command list. 61 * 62 * The L1T, slice, L2C, L2T, and GCA caches will be flushed before 63 * each CL executes. The VCD cache should be flushed (if necessary) 64 * by the submitted CLs. The TLB writes are guaranteed to have been 65 * flushed by the time the render done IRQ happens, which is the 66 * trigger for out_sync. Any dirtying of cachelines by the job (only 67 * possible using TMU writes) must be flushed by the caller using the 68 * DRM_V3D_SUBMIT_CL_FLUSH_CACHE_FLAG flag. 69 */ 70 struct drm_v3d_submit_cl { 71 /* Pointer to the binner command list. 72 * 73 * This is the first set of commands executed, which runs the 74 * coordinate shader to determine where primitives land on the screen, 75 * then writes out the state updates and draw calls necessary per tile 76 * to the tile allocation BO. 77 * 78 * This BCL will block on any previous BCL submitted on the 79 * same FD, but not on any RCL or BCLs submitted by other 80 * clients -- that is left up to the submitter to control 81 * using in_sync_bcl if necessary. 82 */ 83 __u32 bcl_start; 84 85 /** End address of the BCL (first byte after the BCL) */ 86 __u32 bcl_end; 87 88 /* Offset of the render command list. 89 * 90 * This is the second set of commands executed, which will either 91 * execute the tiles that have been set up by the BCL, or a fixed set 92 * of tiles (in the case of RCL-only blits). 93 * 94 * This RCL will block on this submit's BCL, and any previous 95 * RCL submitted on the same FD, but not on any RCL or BCLs 96 * submitted by other clients -- that is left up to the 97 * submitter to control using in_sync_rcl if necessary. 98 */ 99 __u32 rcl_start; 100 101 /** End address of the RCL (first byte after the RCL) */ 102 __u32 rcl_end; 103 104 /** An optional sync object to wait on before starting the BCL. */ 105 __u32 in_sync_bcl; 106 /** An optional sync object to wait on before starting the RCL. */ 107 __u32 in_sync_rcl; 108 /** An optional sync object to place the completion fence in. */ 109 __u32 out_sync; 110 111 /* Offset of the tile alloc memory 112 * 113 * This is optional on V3D 3.3 (where the CL can set the value) but 114 * required on V3D 4.1. 115 */ 116 __u32 qma; 117 118 /** Size of the tile alloc memory. */ 119 __u32 qms; 120 121 /** Offset of the tile state data array. */ 122 __u32 qts; 123 124 /* Pointer to a u32 array of the BOs that are referenced by the job. 125 */ 126 __u64 bo_handles; 127 128 /* Number of BO handles passed in (size is that times 4). */ 129 __u32 bo_handle_count; 130 131 __u32 flags; 132 }; 133 134 /** 135 * struct drm_v3d_wait_bo - ioctl argument for waiting for 136 * completion of the last DRM_V3D_SUBMIT_CL on a BO. 137 * 138 * This is useful for cases where multiple processes might be 139 * rendering to a BO and you want to wait for all rendering to be 140 * completed. 141 */ 142 struct drm_v3d_wait_bo { 143 __u32 handle; 144 __u32 pad; 145 __u64 timeout_ns; 146 }; 147 148 /** 149 * struct drm_v3d_create_bo - ioctl argument for creating V3D BOs. 150 * 151 * There are currently no values for the flags argument, but it may be 152 * used in a future extension. 153 */ 154 struct drm_v3d_create_bo { 155 __u32 size; 156 __u32 flags; 157 /** Returned GEM handle for the BO. */ 158 __u32 handle; 159 /** 160 * Returned offset for the BO in the V3D address space. This offset 161 * is private to the DRM fd and is valid for the lifetime of the GEM 162 * handle. 163 * 164 * This offset value will always be nonzero, since various HW 165 * units treat 0 specially. 166 */ 167 __u32 offset; 168 }; 169 170 /** 171 * struct drm_v3d_mmap_bo - ioctl argument for mapping V3D BOs. 172 * 173 * This doesn't actually perform an mmap. Instead, it returns the 174 * offset you need to use in an mmap on the DRM device node. This 175 * means that tools like valgrind end up knowing about the mapped 176 * memory. 177 * 178 * There are currently no values for the flags argument, but it may be 179 * used in a future extension. 180 */ 181 struct drm_v3d_mmap_bo { 182 /** Handle for the object being mapped. */ 183 __u32 handle; 184 __u32 flags; 185 /** offset into the drm node to use for subsequent mmap call. */ 186 __u64 offset; 187 }; 188 189 enum drm_v3d_param { 190 DRM_V3D_PARAM_V3D_UIFCFG, 191 DRM_V3D_PARAM_V3D_HUB_IDENT1, 192 DRM_V3D_PARAM_V3D_HUB_IDENT2, 193 DRM_V3D_PARAM_V3D_HUB_IDENT3, 194 DRM_V3D_PARAM_V3D_CORE0_IDENT0, 195 DRM_V3D_PARAM_V3D_CORE0_IDENT1, 196 DRM_V3D_PARAM_V3D_CORE0_IDENT2, 197 DRM_V3D_PARAM_SUPPORTS_TFU, 198 DRM_V3D_PARAM_SUPPORTS_CSD, 199 DRM_V3D_PARAM_SUPPORTS_CACHE_FLUSH, 200 }; 201 202 struct drm_v3d_get_param { 203 __u32 param; 204 __u32 pad; 205 __u64 value; 206 }; 207 208 /** 209 * Returns the offset for the BO in the V3D address space for this DRM fd. 210 * This is the same value returned by drm_v3d_create_bo, if that was called 211 * from this DRM fd. 212 */ 213 struct drm_v3d_get_bo_offset { 214 __u32 handle; 215 __u32 offset; 216 }; 217 218 struct drm_v3d_submit_tfu { 219 __u32 icfg; 220 __u32 iia; 221 __u32 iis; 222 __u32 ica; 223 __u32 iua; 224 __u32 ioa; 225 __u32 ios; 226 __u32 coef[4]; 227 /* First handle is the output BO, following are other inputs. 228 * 0 for unused. 229 */ 230 __u32 bo_handles[4]; 231 /* sync object to block on before running the TFU job. Each TFU 232 * job will execute in the order submitted to its FD. Synchronization 233 * against rendering jobs requires using sync objects. 234 */ 235 __u32 in_sync; 236 /* Sync object to signal when the TFU job is done. */ 237 __u32 out_sync; 238 }; 239 240 /* Submits a compute shader for dispatch. This job will block on any 241 * previous compute shaders submitted on this fd, and any other 242 * synchronization must be performed with in_sync/out_sync. 243 */ 244 struct drm_v3d_submit_csd { 245 __u32 cfg[7]; 246 __u32 coef[4]; 247 248 /* Pointer to a u32 array of the BOs that are referenced by the job. 249 */ 250 __u64 bo_handles; 251 252 /* Number of BO handles passed in (size is that times 4). */ 253 __u32 bo_handle_count; 254 255 /* sync object to block on before running the CSD job. Each 256 * CSD job will execute in the order submitted to its FD. 257 * Synchronization against rendering/TFU jobs or CSD from 258 * other fds requires using sync objects. 259 */ 260 __u32 in_sync; 261 /* Sync object to signal when the CSD job is done. */ 262 __u32 out_sync; 263 }; 264 265 #if defined(__cplusplus) 266 } 267 #endif 268 269 #endif /* _V3D_DRM_H_ */ 270