drm/amd/display: Fix increment when sampling OTF in DCE
[linux-block.git] / drivers / gpu / drm / amd / amdgpu / amdgpu_ring.h
CommitLineData
78023016
CK
1/*
2 * Copyright 2016 Advanced Micro Devices, Inc.
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 shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 *
22 * Authors: Christian König
23 */
24#ifndef __AMDGPU_RING_H__
25#define __AMDGPU_RING_H__
26
b2ff0e8a 27#include <drm/amdgpu_drm.h>
1b1f42d8 28#include <drm/gpu_scheduler.h>
78023016
CK
29
30/* max number of rings */
f7243053 31#define AMDGPU_MAX_RINGS 18
78023016
CK
32#define AMDGPU_MAX_GFX_RINGS 1
33#define AMDGPU_MAX_COMPUTE_RINGS 8
34#define AMDGPU_MAX_VCE_RINGS 3
f7243053 35#define AMDGPU_MAX_UVD_ENC_RINGS 2
78023016
CK
36
37/* some special values for the owner field */
38#define AMDGPU_FENCE_OWNER_UNDEFINED ((void*)0ul)
39#define AMDGPU_FENCE_OWNER_VM ((void*)1ul)
40
41#define AMDGPU_FENCE_FLAG_64BIT (1 << 0)
42#define AMDGPU_FENCE_FLAG_INT (1 << 1)
43
44enum amdgpu_ring_type {
45 AMDGPU_RING_TYPE_GFX,
46 AMDGPU_RING_TYPE_COMPUTE,
47 AMDGPU_RING_TYPE_SDMA,
48 AMDGPU_RING_TYPE_UVD,
2068751d 49 AMDGPU_RING_TYPE_VCE,
50c3e232 50 AMDGPU_RING_TYPE_KIQ,
cca69fe8 51 AMDGPU_RING_TYPE_UVD_ENC,
8ace845f
LL
52 AMDGPU_RING_TYPE_VCN_DEC,
53 AMDGPU_RING_TYPE_VCN_ENC
78023016
CK
54};
55
56struct amdgpu_device;
57struct amdgpu_ring;
58struct amdgpu_ib;
59struct amdgpu_cs_parser;
b2ff0e8a 60struct amdgpu_job;
78023016
CK
61
62/*
63 * Fences.
64 */
65struct amdgpu_fence_driver {
66 uint64_t gpu_addr;
67 volatile uint32_t *cpu_addr;
68 /* sync_seq is protected by ring emission lock */
69 uint32_t sync_seq;
70 atomic_t last_seq;
71 bool initialized;
72 struct amdgpu_irq_src *irq_src;
73 unsigned irq_type;
74 struct timer_list fallback_timer;
75 unsigned num_fences_mask;
76 spinlock_t lock;
220196b3 77 struct dma_fence **fences;
78023016
CK
78};
79
80int amdgpu_fence_driver_init(struct amdgpu_device *adev);
81void amdgpu_fence_driver_fini(struct amdgpu_device *adev);
2f9d4084 82void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring);
78023016
CK
83
84int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring,
85 unsigned num_hw_submission);
86int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
87 struct amdgpu_irq_src *irq_src,
88 unsigned irq_type);
89void amdgpu_fence_driver_suspend(struct amdgpu_device *adev);
90void amdgpu_fence_driver_resume(struct amdgpu_device *adev);
220196b3 91int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence);
43ca8efa 92int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s);
78023016
CK
93void amdgpu_fence_process(struct amdgpu_ring *ring);
94int amdgpu_fence_wait_empty(struct amdgpu_ring *ring);
43ca8efa 95signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
96 uint32_t wait_seq,
97 signed long timeout);
78023016
CK
98unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring);
99
100/*
101 * Rings.
102 */
103
104/* provided by hw blocks that expose a ring buffer for commands */
105struct amdgpu_ring_funcs {
21cd942e 106 enum amdgpu_ring_type type;
79887142
CK
107 uint32_t align_mask;
108 u32 nop;
536fbf94 109 bool support_64bit_ptrs;
0eeb68b3 110 unsigned vmhub;
21cd942e 111
78023016 112 /* ring read/write ptr handling */
536fbf94
KW
113 u64 (*get_rptr)(struct amdgpu_ring *ring);
114 u64 (*get_wptr)(struct amdgpu_ring *ring);
78023016
CK
115 void (*set_wptr)(struct amdgpu_ring *ring);
116 /* validating and patching of IBs */
117 int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx);
e12f3d7a
CK
118 /* constants to calculate how many DW are needed for an emit */
119 unsigned emit_frame_size;
120 unsigned emit_ib_size;
78023016
CK
121 /* command emit functions */
122 void (*emit_ib)(struct amdgpu_ring *ring,
123 struct amdgpu_ib *ib,
c4f46f22 124 unsigned vmid, bool ctx_switch);
78023016
CK
125 void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr,
126 uint64_t seq, unsigned flags);
127 void (*emit_pipeline_sync)(struct amdgpu_ring *ring);
c4f46f22 128 void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vmid,
5a4633c4 129 unsigned pasid, uint64_t pd_addr);
78023016 130 void (*emit_hdp_flush)(struct amdgpu_ring *ring);
78023016
CK
131 void (*emit_gds_switch)(struct amdgpu_ring *ring, uint32_t vmid,
132 uint32_t gds_base, uint32_t gds_size,
133 uint32_t gws_base, uint32_t gws_size,
134 uint32_t oa_base, uint32_t oa_size);
135 /* testing functions */
136 int (*test_ring)(struct amdgpu_ring *ring);
137 int (*test_ib)(struct amdgpu_ring *ring, long timeout);
138 /* insert NOP packets */
139 void (*insert_nop)(struct amdgpu_ring *ring, uint32_t count);
ef44f854 140 void (*insert_start)(struct amdgpu_ring *ring);
135d4735 141 void (*insert_end)(struct amdgpu_ring *ring);
78023016
CK
142 /* pad the indirect buffer to the necessary number of dw */
143 void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
144 unsigned (*init_cond_exec)(struct amdgpu_ring *ring);
145 void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset);
146 /* note usage for clock and power gating */
147 void (*begin_use)(struct amdgpu_ring *ring);
148 void (*end_use)(struct amdgpu_ring *ring);
149 void (*emit_switch_buffer) (struct amdgpu_ring *ring);
150 void (*emit_cntxcntl) (struct amdgpu_ring *ring, uint32_t flags);
b6091c12
XY
151 void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg);
152 void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val);
c1e877da
CK
153 void (*emit_reg_wait)(struct amdgpu_ring *ring, uint32_t reg,
154 uint32_t val, uint32_t mask);
3b4d68e9 155 void (*emit_tmz)(struct amdgpu_ring *ring, bool start);
b2ff0e8a
AR
156 /* priority functions */
157 void (*set_priority) (struct amdgpu_ring *ring,
1b1f42d8 158 enum drm_sched_priority priority);
78023016
CK
159};
160
161struct amdgpu_ring {
162 struct amdgpu_device *adev;
163 const struct amdgpu_ring_funcs *funcs;
164 struct amdgpu_fence_driver fence_drv;
1b1f42d8 165 struct drm_gpu_scheduler sched;
795f2813 166 struct list_head lru_list;
78023016
CK
167
168 struct amdgpu_bo *ring_obj;
169 volatile uint32_t *ring;
170 unsigned rptr_offs;
536fbf94
KW
171 u64 wptr;
172 u64 wptr_old;
78023016
CK
173 unsigned ring_size;
174 unsigned max_dw;
175 int count_dw;
176 uint64_t gpu_addr;
536fbf94
KW
177 uint64_t ptr_mask;
178 uint32_t buf_mask;
78023016 179 bool ready;
78023016
CK
180 u32 idx;
181 u32 me;
182 u32 pipe;
183 u32 queue;
184 struct amdgpu_bo *mqd_obj;
f3972b53 185 uint64_t mqd_gpu_addr;
59a82d7d 186 void *mqd_ptr;
34534610 187 uint64_t eop_gpu_addr;
78023016
CK
188 u32 doorbell_index;
189 bool use_doorbell;
2ffe31de 190 bool use_pollmem;
78023016
CK
191 unsigned wptr_offs;
192 unsigned fence_offs;
193 uint64_t current_ctx;
78023016
CK
194 char name[16];
195 unsigned cond_exe_offs;
196 u64 cond_exe_gpu_addr;
197 volatile u32 *cond_exe_cpu_addr;
4789c463 198 unsigned vm_inv_eng;
3af81440 199 struct dma_fence *vmid_wait;
dd684d31 200 bool has_compute_vm_bug;
b2ff0e8a 201
1b1f42d8 202 atomic_t num_jobs[DRM_SCHED_PRIORITY_MAX];
b2ff0e8a
AR
203 struct mutex priority_mutex;
204 /* protected by priority_mutex */
205 int priority;
206
78023016
CK
207#if defined(CONFIG_DEBUG_FS)
208 struct dentry *ent;
209#endif
210};
211
212int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw);
213void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
214void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
215void amdgpu_ring_commit(struct amdgpu_ring *ring);
216void amdgpu_ring_undo(struct amdgpu_ring *ring);
b2ff0e8a 217void amdgpu_ring_priority_get(struct amdgpu_ring *ring,
1b1f42d8 218 enum drm_sched_priority priority);
b2ff0e8a 219void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
1b1f42d8 220 enum drm_sched_priority priority);
78023016 221int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
79887142
CK
222 unsigned ring_size, struct amdgpu_irq_src *irq_src,
223 unsigned irq_type);
78023016 224void amdgpu_ring_fini(struct amdgpu_ring *ring);
35161bbc
AR
225int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type,
226 int *blacklist, int num_blacklist,
227 bool lru_pipe_order, struct amdgpu_ring **ring);
795f2813 228void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring);
c79ecfbf
ML
229static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring)
230{
231 int i = 0;
e09706f4 232 while (i <= ring->buf_mask)
c79ecfbf
ML
233 ring->ring[i++] = ring->funcs->nop;
234
235}
78023016 236
e8110b1c
CK
237static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
238{
239 if (ring->count_dw <= 0)
240 DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
241 ring->ring[ring->wptr++ & ring->buf_mask] = v;
242 ring->wptr &= ring->ptr_mask;
243 ring->count_dw--;
244}
245
246static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
247 void *src, int count_dw)
248{
249 unsigned occupied, chunk1, chunk2;
250 void *dst;
251
369421cb 252 if (unlikely(ring->count_dw < count_dw))
e8110b1c 253 DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
e8110b1c
CK
254
255 occupied = ring->wptr & ring->buf_mask;
256 dst = (void *)&ring->ring[occupied];
257 chunk1 = ring->buf_mask + 1 - occupied;
258 chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1;
259 chunk2 = count_dw - chunk1;
260 chunk1 <<= 2;
261 chunk2 <<= 2;
262
263 if (chunk1)
264 memcpy(dst, src, chunk1);
265
266 if (chunk2) {
267 src += chunk1;
268 dst = (void *)ring->ring;
269 memcpy(dst, src, chunk2);
270 }
271
272 ring->wptr += count_dw;
273 ring->wptr &= ring->ptr_mask;
274 ring->count_dw -= count_dw;
275}
276
78023016 277#endif