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
42 #define DRM_IOCTL_V3D_SUBMIT_CL DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CL, struct drm_v3d_submit_cl)
43 #define DRM_IOCTL_V3D_WAIT_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_WAIT_BO, struct drm_v3d_wait_bo)
44 #define DRM_IOCTL_V3D_CREATE_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_CREATE_BO, struct drm_v3d_create_bo)
45 #define DRM_IOCTL_V3D_MMAP_BO DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_MMAP_BO, struct drm_v3d_mmap_bo)
46 #define DRM_IOCTL_V3D_GET_PARAM DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_GET_PARAM, struct drm_v3d_get_param)
47 #define DRM_IOCTL_V3D_GET_BO_OFFSET DRM_IOWR(DRM_COMMAND_BASE + DRM_V3D_GET_BO_OFFSET, struct drm_v3d_get_bo_offset)
48 #define DRM_IOCTL_V3D_SUBMIT_TFU DRM_IOW(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_TFU, struct drm_v3d_submit_tfu)
49 #define DRM_IOCTL_V3D_SUBMIT_CSD DRM_IOW(DRM_COMMAND_BASE + DRM_V3D_SUBMIT_CSD, struct drm_v3d_submit_csd)
50
51 /**
52 * struct drm_v3d_submit_cl - ioctl argument for submitting commands to the 3D
53 * engine.
54 *
55 * This asks the kernel to have the GPU execute an optional binner
56 * command list, and a render command list.
57 *
58 * The L1T, slice, L2C, L2T, and GCA caches will be flushed before
59 * each CL executes. The VCD cache should be flushed (if necessary)
60 * by the submitted CLs. The TLB writes are guaranteed to have been
61 * flushed by the time the render done IRQ happens, which is the
62 * trigger for out_sync. Any dirtying of cachelines by the job (only
63 * possible using TMU writes) must be flushed by the caller using the
64 * CL's cache flush commands.
65 */
66 struct drm_v3d_submit_cl {
67 /* Pointer to the binner command list.
68 *
69 * This is the first set of commands executed, which runs the
70 * coordinate shader to determine where primitives land on the screen,
71 * then writes out the state updates and draw calls necessary per tile
72 * to the tile allocation BO.
73 *
74 * This BCL will block on any previous BCL submitted on the
75 * same FD, but not on any RCL or BCLs submitted by other
76 * clients -- that is left up to the submitter to control
77 * using in_sync_bcl if necessary.
78 */
79 __u32 bcl_start;
80
81 /** End address of the BCL (first byte after the BCL) */
82 __u32 bcl_end;
83
84 /* Offset of the render command list.
85 *
86 * This is the second set of commands executed, which will either
87 * execute the tiles that have been set up by the BCL, or a fixed set
88 * of tiles (in the case of RCL-only blits).
89 *
90 * This RCL will block on this submit's BCL, and any previous
91 * RCL submitted on the same FD, but not on any RCL or BCLs
92 * submitted by other clients -- that is left up to the
93 * submitter to control using in_sync_rcl if necessary.
94 */
95 __u32 rcl_start;
96
97 /** End address of the RCL (first byte after the RCL) */
98 __u32 rcl_end;
99
100 /** An optional sync object to wait on before starting the BCL. */
101 __u32 in_sync_bcl;
102 /** An optional sync object to wait on before starting the RCL. */
103 __u32 in_sync_rcl;
104 /** An optional sync object to place the completion fence in. */
105 __u32 out_sync;
106
107 /* Offset of the tile alloc memory
108 *
109 * This is optional on V3D 3.3 (where the CL can set the value) but
110 * required on V3D 4.1.
111 */
112 __u32 qma;
113
114 /** Size of the tile alloc memory. */
115 __u32 qms;
116
117 /** Offset of the tile state data array. */
118 __u32 qts;
119
120 /* Pointer to a u32 array of the BOs that are referenced by the job.
121 */
122 __u64 bo_handles;
123
124 /* Number of BO handles passed in (size is that times 4). */
125 __u32 bo_handle_count;
126
127 /* Pad, must be zero-filled. */
128 __u32 pad;
129 };
130
131 /**
132 * struct drm_v3d_wait_bo - ioctl argument for waiting for
133 * completion of the last DRM_V3D_SUBMIT_CL on a BO.
134 *
135 * This is useful for cases where multiple processes might be
136 * rendering to a BO and you want to wait for all rendering to be
137 * completed.
138 */
139 struct drm_v3d_wait_bo {
140 __u32 handle;
141 __u32 pad;
142 __u64 timeout_ns;
143 };
144
145 /**
146 * struct drm_v3d_create_bo - ioctl argument for creating V3D BOs.
147 *
148 * There are currently no values for the flags argument, but it may be
149 * used in a future extension.
150 */
151 struct drm_v3d_create_bo {
152 __u32 size;
153 __u32 flags;
154 /** Returned GEM handle for the BO. */
155 __u32 handle;
156 /**
157 * Returned offset for the BO in the V3D address space. This offset
158 * is private to the DRM fd and is valid for the lifetime of the GEM
159 * handle.
160 *
161 * This offset value will always be nonzero, since various HW
162 * units treat 0 specially.
163 */
164 __u32 offset;
165 };
166
167 /**
168 * struct drm_v3d_mmap_bo - ioctl argument for mapping V3D BOs.
169 *
170 * This doesn't actually perform an mmap. Instead, it returns the
171 * offset you need to use in an mmap on the DRM device node. This
172 * means that tools like valgrind end up knowing about the mapped
173 * memory.
174 *
175 * There are currently no values for the flags argument, but it may be
176 * used in a future extension.
177 */
178 struct drm_v3d_mmap_bo {
179 /** Handle for the object being mapped. */
180 __u32 handle;
181 __u32 flags;
182 /** offset into the drm node to use for subsequent mmap call. */
183 __u64 offset;
184 };
185
186 enum drm_v3d_param {
187 DRM_V3D_PARAM_V3D_UIFCFG,
188 DRM_V3D_PARAM_V3D_HUB_IDENT1,
189 DRM_V3D_PARAM_V3D_HUB_IDENT2,
190 DRM_V3D_PARAM_V3D_HUB_IDENT3,
191 DRM_V3D_PARAM_V3D_CORE0_IDENT0,
192 DRM_V3D_PARAM_V3D_CORE0_IDENT1,
193 DRM_V3D_PARAM_V3D_CORE0_IDENT2,
194 DRM_V3D_PARAM_SUPPORTS_TFU,
195 DRM_V3D_PARAM_SUPPORTS_CSD,
196 };
197
198 struct drm_v3d_get_param {
199 __u32 param;
200 __u32 pad;
201 __u64 value;
202 };
203
204 /**
205 * Returns the offset for the BO in the V3D address space for this DRM fd.
206 * This is the same value returned by drm_v3d_create_bo, if that was called
207 * from this DRM fd.
208 */
209 struct drm_v3d_get_bo_offset {
210 __u32 handle;
211 __u32 offset;
212 };
213
214 struct drm_v3d_submit_tfu {
215 __u32 icfg;
216 __u32 iia;
217 __u32 iis;
218 __u32 ica;
219 __u32 iua;
220 __u32 ioa;
221 __u32 ios;
222 __u32 coef[4];
223 /* First handle is the output BO, following are other inputs.
224 * 0 for unused.
225 */
226 __u32 bo_handles[4];
227 /* sync object to block on before running the TFU job. Each TFU
228 * job will execute in the order submitted to its FD. Synchronization
229 * against rendering jobs requires using sync objects.
230 */
231 __u32 in_sync;
232 /* Sync object to signal when the TFU job is done. */
233 __u32 out_sync;
234 };
235
236 /* Submits a compute shader for dispatch. This job will block on any
237 * previous compute shaders submitted on this fd, and any other
238 * synchronization must be performed with in_sync/out_sync.
239 */
240 struct drm_v3d_submit_csd {
241 __u32 cfg[7];
242 __u32 coef[4];
243
244 /* Pointer to a u32 array of the BOs that are referenced by the job.
245 */
246 __u64 bo_handles;
247
248 /* Number of BO handles passed in (size is that times 4). */
249 __u32 bo_handle_count;
250
251 /* sync object to block on before running the CSD job. Each
252 * CSD job will execute in the order submitted to its FD.
253 * Synchronization against rendering/TFU jobs or CSD from
254 * other fds requires using sync objects.
255 */
256 __u32 in_sync;
257 /* Sync object to signal when the CSD job is done. */
258 __u32 out_sync;
259 };
260
261 #if defined(__cplusplus)
262 }
263 #endif
264
265 #endif /* _V3D_DRM_H_ */