xref: /openbmc/linux/include/uapi/drm/v3d_drm.h (revision 349f631d)
1 /*
2  * Copyright © 2014-2018 Broadcom
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining a
5  * copy of this software and associated documentation files (the "Software"),
6  * to deal in the Software without restriction, including without limitation
7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8  * and/or sell copies of the Software, and to permit persons to whom the
9  * Software is furnished to do so, subject to the following conditions:
10  *
11  * The above copyright notice and this permission notice (including the next
12  * paragraph) shall be included in all copies or substantial portions of the
13  * Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21  * IN THE SOFTWARE.
22  */
23 
24 #ifndef _V3D_DRM_H_
25 #define _V3D_DRM_H_
26 
27 #include "drm.h"
28 
29 #if defined(__cplusplus)
30 extern "C" {
31 #endif
32 
33 #define DRM_V3D_SUBMIT_CL                         0x00
34 #define DRM_V3D_WAIT_BO                           0x01
35 #define DRM_V3D_CREATE_BO                         0x02
36 #define DRM_V3D_MMAP_BO                           0x03
37 #define DRM_V3D_GET_PARAM                         0x04
38 #define DRM_V3D_GET_BO_OFFSET                     0x05
39 #define DRM_V3D_SUBMIT_TFU                        0x06
40 #define DRM_V3D_SUBMIT_CSD                        0x07
41 #define DRM_V3D_PERFMON_CREATE                    0x08
42 #define DRM_V3D_PERFMON_DESTROY                   0x09
43 #define DRM_V3D_PERFMON_GET_VALUES                0x0a
44 
45 #define DRM_IOCTL_V3D_SUBMIT_CL           DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CL, struct drm_v3d_submit_cl)
46 #define DRM_IOCTL_V3D_WAIT_BO             DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_WAIT_BO, struct drm_v3d_wait_bo)
47 #define DRM_IOCTL_V3D_CREATE_BO           DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_CREATE_BO, struct drm_v3d_create_bo)
48 #define DRM_IOCTL_V3D_MMAP_BO             DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_MMAP_BO, struct drm_v3d_mmap_bo)
49 #define DRM_IOCTL_V3D_GET_PARAM           DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_GET_PARAM, struct drm_v3d_get_param)
50 #define DRM_IOCTL_V3D_GET_BO_OFFSET       DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_GET_BO_OFFSET, struct drm_v3d_get_bo_offset)
51 #define DRM_IOCTL_V3D_SUBMIT_TFU          DRM_IOW(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_TFU, struct drm_v3d_submit_tfu)
52 #define DRM_IOCTL_V3D_SUBMIT_CSD          DRM_IOW(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CSD, struct drm_v3d_submit_csd)
53 #define DRM_IOCTL_V3D_PERFMON_CREATE      DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_PERFMON_CREATE, \
54 						   struct drm_v3d_perfmon_create)
55 #define DRM_IOCTL_V3D_PERFMON_DESTROY     DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_PERFMON_DESTROY, \
56 						   struct drm_v3d_perfmon_destroy)
57 #define DRM_IOCTL_V3D_PERFMON_GET_VALUES  DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_PERFMON_GET_VALUES, \
58 						   struct drm_v3d_perfmon_get_values)
59 
60 #define DRM_V3D_SUBMIT_CL_FLUSH_CACHE             0x01
61 
62 /**
63  * struct drm_v3d_submit_cl - ioctl argument for submitting commands to the 3D
64  * engine.
65  *
66  * This asks the kernel to have the GPU execute an optional binner
67  * command list, and a render command list.
68  *
69  * The L1T, slice, L2C, L2T, and GCA caches will be flushed before
70  * each CL executes.  The VCD cache should be flushed (if necessary)
71  * by the submitted CLs.  The TLB writes are guaranteed to have been
72  * flushed by the time the render done IRQ happens, which is the
73  * trigger for out_sync.  Any dirtying of cachelines by the job (only
74  * possible using TMU writes) must be flushed by the caller using the
75  * DRM_V3D_SUBMIT_CL_FLUSH_CACHE_FLAG flag.
76  */
77 struct drm_v3d_submit_cl {
78 	/* Pointer to the binner command list.
79 	 *
80 	 * This is the first set of commands executed, which runs the
81 	 * coordinate shader to determine where primitives land on the screen,
82 	 * then writes out the state updates and draw calls necessary per tile
83 	 * to the tile allocation BO.
84 	 *
85 	 * This BCL will block on any previous BCL submitted on the
86 	 * same FD, but not on any RCL or BCLs submitted by other
87 	 * clients -- that is left up to the submitter to control
88 	 * using in_sync_bcl if necessary.
89 	 */
90 	__u32 bcl_start;
91 
92 	/** End address of the BCL (first byte after the BCL) */
93 	__u32 bcl_end;
94 
95 	/* Offset of the render command list.
96 	 *
97 	 * This is the second set of commands executed, which will either
98 	 * execute the tiles that have been set up by the BCL, or a fixed set
99 	 * of tiles (in the case of RCL-only blits).
100 	 *
101 	 * This RCL will block on this submit's BCL, and any previous
102 	 * RCL submitted on the same FD, but not on any RCL or BCLs
103 	 * submitted by other clients -- that is left up to the
104 	 * submitter to control using in_sync_rcl if necessary.
105 	 */
106 	__u32 rcl_start;
107 
108 	/** End address of the RCL (first byte after the RCL) */
109 	__u32 rcl_end;
110 
111 	/** An optional sync object to wait on before starting the BCL. */
112 	__u32 in_sync_bcl;
113 	/** An optional sync object to wait on before starting the RCL. */
114 	__u32 in_sync_rcl;
115 	/** An optional sync object to place the completion fence in. */
116 	__u32 out_sync;
117 
118 	/* Offset of the tile alloc memory
119 	 *
120 	 * This is optional on V3D 3.3 (where the CL can set the value) but
121 	 * required on V3D 4.1.
122 	 */
123 	__u32 qma;
124 
125 	/** Size of the tile alloc memory. */
126 	__u32 qms;
127 
128 	/** Offset of the tile state data array. */
129 	__u32 qts;
130 
131 	/* Pointer to a u32 array of the BOs that are referenced by the job.
132 	 */
133 	__u64 bo_handles;
134 
135 	/* Number of BO handles passed in (size is that times 4). */
136 	__u32 bo_handle_count;
137 
138 	__u32 flags;
139 
140 	/* ID of the perfmon to attach to this job. 0 means no perfmon. */
141 	__u32 perfmon_id;
142 
143 	__u32 pad;
144 };
145 
146 /**
147  * struct drm_v3d_wait_bo - ioctl argument for waiting for
148  * completion of the last DRM_V3D_SUBMIT_CL on a BO.
149  *
150  * This is useful for cases where multiple processes might be
151  * rendering to a BO and you want to wait for all rendering to be
152  * completed.
153  */
154 struct drm_v3d_wait_bo {
155 	__u32 handle;
156 	__u32 pad;
157 	__u64 timeout_ns;
158 };
159 
160 /**
161  * struct drm_v3d_create_bo - ioctl argument for creating V3D BOs.
162  *
163  * There are currently no values for the flags argument, but it may be
164  * used in a future extension.
165  */
166 struct drm_v3d_create_bo {
167 	__u32 size;
168 	__u32 flags;
169 	/** Returned GEM handle for the BO. */
170 	__u32 handle;
171 	/**
172 	 * Returned offset for the BO in the V3D address space.  This offset
173 	 * is private to the DRM fd and is valid for the lifetime of the GEM
174 	 * handle.
175 	 *
176 	 * This offset value will always be nonzero, since various HW
177 	 * units treat 0 specially.
178 	 */
179 	__u32 offset;
180 };
181 
182 /**
183  * struct drm_v3d_mmap_bo - ioctl argument for mapping V3D BOs.
184  *
185  * This doesn't actually perform an mmap.  Instead, it returns the
186  * offset you need to use in an mmap on the DRM device node.  This
187  * means that tools like valgrind end up knowing about the mapped
188  * memory.
189  *
190  * There are currently no values for the flags argument, but it may be
191  * used in a future extension.
192  */
193 struct drm_v3d_mmap_bo {
194 	/** Handle for the object being mapped. */
195 	__u32 handle;
196 	__u32 flags;
197 	/** offset into the drm node to use for subsequent mmap call. */
198 	__u64 offset;
199 };
200 
201 enum drm_v3d_param {
202 	DRM_V3D_PARAM_V3D_UIFCFG,
203 	DRM_V3D_PARAM_V3D_HUB_IDENT1,
204 	DRM_V3D_PARAM_V3D_HUB_IDENT2,
205 	DRM_V3D_PARAM_V3D_HUB_IDENT3,
206 	DRM_V3D_PARAM_V3D_CORE0_IDENT0,
207 	DRM_V3D_PARAM_V3D_CORE0_IDENT1,
208 	DRM_V3D_PARAM_V3D_CORE0_IDENT2,
209 	DRM_V3D_PARAM_SUPPORTS_TFU,
210 	DRM_V3D_PARAM_SUPPORTS_CSD,
211 	DRM_V3D_PARAM_SUPPORTS_CACHE_FLUSH,
212 	DRM_V3D_PARAM_SUPPORTS_PERFMON,
213 };
214 
215 struct drm_v3d_get_param {
216 	__u32 param;
217 	__u32 pad;
218 	__u64 value;
219 };
220 
221 /**
222  * Returns the offset for the BO in the V3D address space for this DRM fd.
223  * This is the same value returned by drm_v3d_create_bo, if that was called
224  * from this DRM fd.
225  */
226 struct drm_v3d_get_bo_offset {
227 	__u32 handle;
228 	__u32 offset;
229 };
230 
231 struct drm_v3d_submit_tfu {
232 	__u32 icfg;
233 	__u32 iia;
234 	__u32 iis;
235 	__u32 ica;
236 	__u32 iua;
237 	__u32 ioa;
238 	__u32 ios;
239 	__u32 coef[4];
240 	/* First handle is the output BO, following are other inputs.
241 	 * 0 for unused.
242 	 */
243 	__u32 bo_handles[4];
244 	/* sync object to block on before running the TFU job.  Each TFU
245 	 * job will execute in the order submitted to its FD.  Synchronization
246 	 * against rendering jobs requires using sync objects.
247 	 */
248 	__u32 in_sync;
249 	/* Sync object to signal when the TFU job is done. */
250 	__u32 out_sync;
251 };
252 
253 /* Submits a compute shader for dispatch.  This job will block on any
254  * previous compute shaders submitted on this fd, and any other
255  * synchronization must be performed with in_sync/out_sync.
256  */
257 struct drm_v3d_submit_csd {
258 	__u32 cfg[7];
259 	__u32 coef[4];
260 
261 	/* Pointer to a u32 array of the BOs that are referenced by the job.
262 	 */
263 	__u64 bo_handles;
264 
265 	/* Number of BO handles passed in (size is that times 4). */
266 	__u32 bo_handle_count;
267 
268 	/* sync object to block on before running the CSD job.  Each
269 	 * CSD job will execute in the order submitted to its FD.
270 	 * Synchronization against rendering/TFU jobs or CSD from
271 	 * other fds requires using sync objects.
272 	 */
273 	__u32 in_sync;
274 	/* Sync object to signal when the CSD job is done. */
275 	__u32 out_sync;
276 
277 	/* ID of the perfmon to attach to this job. 0 means no perfmon. */
278 	__u32 perfmon_id;
279 };
280 
281 enum {
282 	V3D_PERFCNT_FEP_VALID_PRIMTS_NO_PIXELS,
283 	V3D_PERFCNT_FEP_VALID_PRIMS,
284 	V3D_PERFCNT_FEP_EZ_NFCLIP_QUADS,
285 	V3D_PERFCNT_FEP_VALID_QUADS,
286 	V3D_PERFCNT_TLB_QUADS_STENCIL_FAIL,
287 	V3D_PERFCNT_TLB_QUADS_STENCILZ_FAIL,
288 	V3D_PERFCNT_TLB_QUADS_STENCILZ_PASS,
289 	V3D_PERFCNT_TLB_QUADS_ZERO_COV,
290 	V3D_PERFCNT_TLB_QUADS_NONZERO_COV,
291 	V3D_PERFCNT_TLB_QUADS_WRITTEN,
292 	V3D_PERFCNT_PTB_PRIM_VIEWPOINT_DISCARD,
293 	V3D_PERFCNT_PTB_PRIM_CLIP,
294 	V3D_PERFCNT_PTB_PRIM_REV,
295 	V3D_PERFCNT_QPU_IDLE_CYCLES,
296 	V3D_PERFCNT_QPU_ACTIVE_CYCLES_VERTEX_COORD_USER,
297 	V3D_PERFCNT_QPU_ACTIVE_CYCLES_FRAG,
298 	V3D_PERFCNT_QPU_CYCLES_VALID_INSTR,
299 	V3D_PERFCNT_QPU_CYCLES_TMU_STALL,
300 	V3D_PERFCNT_QPU_CYCLES_SCOREBOARD_STALL,
301 	V3D_PERFCNT_QPU_CYCLES_VARYINGS_STALL,
302 	V3D_PERFCNT_QPU_IC_HIT,
303 	V3D_PERFCNT_QPU_IC_MISS,
304 	V3D_PERFCNT_QPU_UC_HIT,
305 	V3D_PERFCNT_QPU_UC_MISS,
306 	V3D_PERFCNT_TMU_TCACHE_ACCESS,
307 	V3D_PERFCNT_TMU_TCACHE_MISS,
308 	V3D_PERFCNT_VPM_VDW_STALL,
309 	V3D_PERFCNT_VPM_VCD_STALL,
310 	V3D_PERFCNT_BIN_ACTIVE,
311 	V3D_PERFCNT_RDR_ACTIVE,
312 	V3D_PERFCNT_L2T_HITS,
313 	V3D_PERFCNT_L2T_MISSES,
314 	V3D_PERFCNT_CYCLE_COUNT,
315 	V3D_PERFCNT_QPU_CYCLES_STALLED_VERTEX_COORD_USER,
316 	V3D_PERFCNT_QPU_CYCLES_STALLED_FRAGMENT,
317 	V3D_PERFCNT_PTB_PRIMS_BINNED,
318 	V3D_PERFCNT_AXI_WRITES_WATCH_0,
319 	V3D_PERFCNT_AXI_READS_WATCH_0,
320 	V3D_PERFCNT_AXI_WRITE_STALLS_WATCH_0,
321 	V3D_PERFCNT_AXI_READ_STALLS_WATCH_0,
322 	V3D_PERFCNT_AXI_WRITE_BYTES_WATCH_0,
323 	V3D_PERFCNT_AXI_READ_BYTES_WATCH_0,
324 	V3D_PERFCNT_AXI_WRITES_WATCH_1,
325 	V3D_PERFCNT_AXI_READS_WATCH_1,
326 	V3D_PERFCNT_AXI_WRITE_STALLS_WATCH_1,
327 	V3D_PERFCNT_AXI_READ_STALLS_WATCH_1,
328 	V3D_PERFCNT_AXI_WRITE_BYTES_WATCH_1,
329 	V3D_PERFCNT_AXI_READ_BYTES_WATCH_1,
330 	V3D_PERFCNT_TLB_PARTIAL_QUADS,
331 	V3D_PERFCNT_TMU_CONFIG_ACCESSES,
332 	V3D_PERFCNT_L2T_NO_ID_STALL,
333 	V3D_PERFCNT_L2T_COM_QUE_STALL,
334 	V3D_PERFCNT_L2T_TMU_WRITES,
335 	V3D_PERFCNT_TMU_ACTIVE_CYCLES,
336 	V3D_PERFCNT_TMU_STALLED_CYCLES,
337 	V3D_PERFCNT_CLE_ACTIVE,
338 	V3D_PERFCNT_L2T_TMU_READS,
339 	V3D_PERFCNT_L2T_CLE_READS,
340 	V3D_PERFCNT_L2T_VCD_READS,
341 	V3D_PERFCNT_L2T_TMUCFG_READS,
342 	V3D_PERFCNT_L2T_SLC0_READS,
343 	V3D_PERFCNT_L2T_SLC1_READS,
344 	V3D_PERFCNT_L2T_SLC2_READS,
345 	V3D_PERFCNT_L2T_TMU_W_MISSES,
346 	V3D_PERFCNT_L2T_TMU_R_MISSES,
347 	V3D_PERFCNT_L2T_CLE_MISSES,
348 	V3D_PERFCNT_L2T_VCD_MISSES,
349 	V3D_PERFCNT_L2T_TMUCFG_MISSES,
350 	V3D_PERFCNT_L2T_SLC0_MISSES,
351 	V3D_PERFCNT_L2T_SLC1_MISSES,
352 	V3D_PERFCNT_L2T_SLC2_MISSES,
353 	V3D_PERFCNT_CORE_MEM_WRITES,
354 	V3D_PERFCNT_L2T_MEM_WRITES,
355 	V3D_PERFCNT_PTB_MEM_WRITES,
356 	V3D_PERFCNT_TLB_MEM_WRITES,
357 	V3D_PERFCNT_CORE_MEM_READS,
358 	V3D_PERFCNT_L2T_MEM_READS,
359 	V3D_PERFCNT_PTB_MEM_READS,
360 	V3D_PERFCNT_PSE_MEM_READS,
361 	V3D_PERFCNT_TLB_MEM_READS,
362 	V3D_PERFCNT_GMP_MEM_READS,
363 	V3D_PERFCNT_PTB_W_MEM_WORDS,
364 	V3D_PERFCNT_TLB_W_MEM_WORDS,
365 	V3D_PERFCNT_PSE_R_MEM_WORDS,
366 	V3D_PERFCNT_TLB_R_MEM_WORDS,
367 	V3D_PERFCNT_TMU_MRU_HITS,
368 	V3D_PERFCNT_COMPUTE_ACTIVE,
369 	V3D_PERFCNT_NUM,
370 };
371 
372 #define DRM_V3D_MAX_PERF_COUNTERS                 32
373 
374 struct drm_v3d_perfmon_create {
375 	__u32 id;
376 	__u32 ncounters;
377 	__u8 counters[DRM_V3D_MAX_PERF_COUNTERS];
378 };
379 
380 struct drm_v3d_perfmon_destroy {
381 	__u32 id;
382 };
383 
384 /*
385  * Returns the values of the performance counters tracked by this
386  * perfmon (as an array of ncounters u64 values).
387  *
388  * No implicit synchronization is performed, so the user has to
389  * guarantee that any jobs using this perfmon have already been
390  * completed  (probably by blocking on the seqno returned by the
391  * last exec that used the perfmon).
392  */
393 struct drm_v3d_perfmon_get_values {
394 	__u32 id;
395 	__u32 pad;
396 	__u64 values_ptr;
397 };
398 
399 #if defined(__cplusplus)
400 }
401 #endif
402 
403 #endif /* _V3D_DRM_H_ */
404