• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright © 2014-2015 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 _UAPI_VC4_DRM_H_
25 #define _UAPI_VC4_DRM_H_
26 
27 #include "drm.h"
28 
29 #if defined(__cplusplus)
30 extern "C" {
31 #endif
32 
33 #define DRM_VC4_SUBMIT_CL                         0x00
34 #define DRM_VC4_WAIT_SEQNO                        0x01
35 #define DRM_VC4_WAIT_BO                           0x02
36 #define DRM_VC4_CREATE_BO                         0x03
37 #define DRM_VC4_MMAP_BO                           0x04
38 #define DRM_VC4_CREATE_SHADER_BO                  0x05
39 #define DRM_VC4_GET_HANG_STATE                    0x06
40 #define DRM_VC4_GET_PARAM                         0x07
41 
42 #define DRM_IOCTL_VC4_SUBMIT_CL           DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_SUBMIT_CL, struct drm_vc4_submit_cl)
43 #define DRM_IOCTL_VC4_WAIT_SEQNO          DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_WAIT_SEQNO, struct drm_vc4_wait_seqno)
44 #define DRM_IOCTL_VC4_WAIT_BO             DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_WAIT_BO, struct drm_vc4_wait_bo)
45 #define DRM_IOCTL_VC4_CREATE_BO           DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_CREATE_BO, struct drm_vc4_create_bo)
46 #define DRM_IOCTL_VC4_MMAP_BO             DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_MMAP_BO, struct drm_vc4_mmap_bo)
47 #define DRM_IOCTL_VC4_CREATE_SHADER_BO    DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_CREATE_SHADER_BO, struct drm_vc4_create_shader_bo)
48 #define DRM_IOCTL_VC4_GET_HANG_STATE      DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_GET_HANG_STATE, struct drm_vc4_get_hang_state)
49 #define DRM_IOCTL_VC4_GET_PARAM           DRM_IOWR(DRM_COMMAND_BASE + DRM_VC4_GET_PARAM, struct drm_vc4_get_param)
50 
51 struct drm_vc4_submit_rcl_surface {
52 	__u32 hindex; /* Handle index, or ~0 if not present. */
53 	__u32 offset; /* Offset to start of buffer. */
54 	/*
55 	 * Bits for either render config (color_write) or load/store packet.
56 	 * Bits should all be 0 for MSAA load/stores.
57 	 */
58 	__u16 bits;
59 
60 #define VC4_SUBMIT_RCL_SURFACE_READ_IS_FULL_RES		(1 << 0)
61 	__u16 flags;
62 };
63 
64 /**
65  * struct drm_vc4_submit_cl - ioctl argument for submitting commands to the 3D
66  * engine.
67  *
68  * Drivers typically use GPU BOs to store batchbuffers / command lists and
69  * their associated state.  However, because the VC4 lacks an MMU, we have to
70  * do validation of memory accesses by the GPU commands.  If we were to store
71  * our commands in BOs, we'd need to do uncached readback from them to do the
72  * validation process, which is too expensive.  Instead, userspace accumulates
73  * commands and associated state in plain memory, then the kernel copies the
74  * data to its own address space, and then validates and stores it in a GPU
75  * BO.
76  */
77 struct drm_vc4_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 	__u64 bin_cl;
86 
87 	/* Pointer to the shader records.
88 	 *
89 	 * Shader records are the structures read by the hardware that contain
90 	 * pointers to uniforms, shaders, and vertex attributes.  The
91 	 * reference to the shader record has enough information to determine
92 	 * how many pointers are necessary (fixed number for shaders/uniforms,
93 	 * and an attribute count), so those BO indices into bo_handles are
94 	 * just stored as __u32s before each shader record passed in.
95 	 */
96 	__u64 shader_rec;
97 
98 	/* Pointer to uniform data and texture handles for the textures
99 	 * referenced by the shader.
100 	 *
101 	 * For each shader state record, there is a set of uniform data in the
102 	 * order referenced by the record (FS, VS, then CS).  Each set of
103 	 * uniform data has a __u32 index into bo_handles per texture
104 	 * sample operation, in the order the QPU_W_TMUn_S writes appear in
105 	 * the program.  Following the texture BO handle indices is the actual
106 	 * uniform data.
107 	 *
108 	 * The individual uniform state blocks don't have sizes passed in,
109 	 * because the kernel has to determine the sizes anyway during shader
110 	 * code validation.
111 	 */
112 	__u64 uniforms;
113 	__u64 bo_handles;
114 
115 	/* Size in bytes of the binner command list. */
116 	__u32 bin_cl_size;
117 	/* Size in bytes of the set of shader records. */
118 	__u32 shader_rec_size;
119 	/* Number of shader records.
120 	 *
121 	 * This could just be computed from the contents of shader_records and
122 	 * the address bits of references to them from the bin CL, but it
123 	 * keeps the kernel from having to resize some allocations it makes.
124 	 */
125 	__u32 shader_rec_count;
126 	/* Size in bytes of the uniform state. */
127 	__u32 uniforms_size;
128 
129 	/* Number of BO handles passed in (size is that times 4). */
130 	__u32 bo_handle_count;
131 
132 	/* RCL setup: */
133 	__u16 width;
134 	__u16 height;
135 	__u8 min_x_tile;
136 	__u8 min_y_tile;
137 	__u8 max_x_tile;
138 	__u8 max_y_tile;
139 	struct drm_vc4_submit_rcl_surface color_read;
140 	struct drm_vc4_submit_rcl_surface color_write;
141 	struct drm_vc4_submit_rcl_surface zs_read;
142 	struct drm_vc4_submit_rcl_surface zs_write;
143 	struct drm_vc4_submit_rcl_surface msaa_color_write;
144 	struct drm_vc4_submit_rcl_surface msaa_zs_write;
145 	__u32 clear_color[2];
146 	__u32 clear_z;
147 	__u8 clear_s;
148 
149 	__u32 pad:24;
150 
151 #define VC4_SUBMIT_CL_USE_CLEAR_COLOR			(1 << 0)
152 	__u32 flags;
153 
154 	/* Returned value of the seqno of this render job (for the
155 	 * wait ioctl).
156 	 */
157 	__u64 seqno;
158 };
159 
160 /**
161  * struct drm_vc4_wait_seqno - ioctl argument for waiting for
162  * DRM_VC4_SUBMIT_CL completion using its returned seqno.
163  *
164  * timeout_ns is the timeout in nanoseconds, where "0" means "don't
165  * block, just return the status."
166  */
167 struct drm_vc4_wait_seqno {
168 	__u64 seqno;
169 	__u64 timeout_ns;
170 };
171 
172 /**
173  * struct drm_vc4_wait_bo - ioctl argument for waiting for
174  * completion of the last DRM_VC4_SUBMIT_CL on a BO.
175  *
176  * This is useful for cases where multiple processes might be
177  * rendering to a BO and you want to wait for all rendering to be
178  * completed.
179  */
180 struct drm_vc4_wait_bo {
181 	__u32 handle;
182 	__u32 pad;
183 	__u64 timeout_ns;
184 };
185 
186 /**
187  * struct drm_vc4_create_bo - ioctl argument for creating VC4 BOs.
188  *
189  * There are currently no values for the flags argument, but it may be
190  * used in a future extension.
191  */
192 struct drm_vc4_create_bo {
193 	__u32 size;
194 	__u32 flags;
195 	/** Returned GEM handle for the BO. */
196 	__u32 handle;
197 	__u32 pad;
198 };
199 
200 /**
201  * struct drm_vc4_mmap_bo - ioctl argument for mapping VC4 BOs.
202  *
203  * This doesn't actually perform an mmap.  Instead, it returns the
204  * offset you need to use in an mmap on the DRM device node.  This
205  * means that tools like valgrind end up knowing about the mapped
206  * memory.
207  *
208  * There are currently no values for the flags argument, but it may be
209  * used in a future extension.
210  */
211 struct drm_vc4_mmap_bo {
212 	/** Handle for the object being mapped. */
213 	__u32 handle;
214 	__u32 flags;
215 	/** offset into the drm node to use for subsequent mmap call. */
216 	__u64 offset;
217 };
218 
219 /**
220  * struct drm_vc4_create_shader_bo - ioctl argument for creating VC4
221  * shader BOs.
222  *
223  * Since allowing a shader to be overwritten while it's also being
224  * executed from would allow privlege escalation, shaders must be
225  * created using this ioctl, and they can't be mmapped later.
226  */
227 struct drm_vc4_create_shader_bo {
228 	/* Size of the data argument. */
229 	__u32 size;
230 	/* Flags, currently must be 0. */
231 	__u32 flags;
232 
233 	/* Pointer to the data. */
234 	__u64 data;
235 
236 	/** Returned GEM handle for the BO. */
237 	__u32 handle;
238 	/* Pad, must be 0. */
239 	__u32 pad;
240 };
241 
242 struct drm_vc4_get_hang_state_bo {
243 	__u32 handle;
244 	__u32 paddr;
245 	__u32 size;
246 	__u32 pad;
247 };
248 
249 /**
250  * struct drm_vc4_hang_state - ioctl argument for collecting state
251  * from a GPU hang for analysis.
252 */
253 struct drm_vc4_get_hang_state {
254 	/** Pointer to array of struct drm_vc4_get_hang_state_bo. */
255 	__u64 bo;
256 	/**
257 	 * On input, the size of the bo array.  Output is the number
258 	 * of bos to be returned.
259 	 */
260 	__u32 bo_count;
261 
262 	__u32 start_bin, start_render;
263 
264 	__u32 ct0ca, ct0ea;
265 	__u32 ct1ca, ct1ea;
266 	__u32 ct0cs, ct1cs;
267 	__u32 ct0ra0, ct1ra0;
268 
269 	__u32 bpca, bpcs;
270 	__u32 bpoa, bpos;
271 
272 	__u32 vpmbase;
273 
274 	__u32 dbge;
275 	__u32 fdbgo;
276 	__u32 fdbgb;
277 	__u32 fdbgr;
278 	__u32 fdbgs;
279 	__u32 errstat;
280 
281 	/* Pad that we may save more registers into in the future. */
282 	__u32 pad[16];
283 };
284 
285 #define DRM_VC4_PARAM_V3D_IDENT0		0
286 #define DRM_VC4_PARAM_V3D_IDENT1		1
287 #define DRM_VC4_PARAM_V3D_IDENT2		2
288 #define DRM_VC4_PARAM_SUPPORTS_BRANCHES		3
289 #define DRM_VC4_PARAM_SUPPORTS_ETC1		4
290 #define DRM_VC4_PARAM_SUPPORTS_THREADED_FS	5
291 
292 struct drm_vc4_get_param {
293 	__u32 param;
294 	__u32 pad;
295 	__u64 value;
296 };
297 
298 #if defined(__cplusplus)
299 }
300 #endif
301 
302 #endif /* _UAPI_VC4_DRM_H_ */
303