Commit | Line | Data |
---|---|---|
95d0906f LL |
1 | /* |
2 | * Copyright 2016 Advanced Micro Devices, Inc. | |
3 | * All Rights Reserved. | |
4 | * | |
5 | * Permission is hereby granted, free of charge, to any person obtaining a | |
6 | * copy of this software and associated documentation files (the | |
7 | * "Software"), to deal in the Software without restriction, including | |
8 | * without limitation the rights to use, copy, modify, merge, publish, | |
9 | * distribute, sub license, and/or sell copies of the Software, and to | |
10 | * permit persons to whom the Software is furnished to do so, subject to | |
11 | * the following conditions: | |
12 | * | |
13 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
14 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
15 | * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL | |
16 | * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, | |
17 | * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR | |
18 | * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE | |
19 | * USE OR OTHER DEALINGS IN THE SOFTWARE. | |
20 | * | |
21 | * The above copyright notice and this permission notice (including the | |
22 | * next paragraph) shall be included in all copies or substantial portions | |
23 | * of the Software. | |
24 | * | |
25 | */ | |
26 | ||
27 | #include <linux/firmware.h> | |
28 | #include <linux/module.h> | |
fdf2f6c5 SR |
29 | #include <linux/pci.h> |
30 | ||
95d0906f LL |
31 | #include <drm/drm.h> |
32 | ||
33 | #include "amdgpu.h" | |
34 | #include "amdgpu_pm.h" | |
35 | #include "amdgpu_vcn.h" | |
36 | #include "soc15d.h" | |
37 | #include "soc15_common.h" | |
38 | ||
b1ebd7c0 | 39 | #include "vcn/vcn_1_0_offset.h" |
bd5d5180 | 40 | #include "vcn/vcn_1_0_sh_mask.h" |
95d0906f LL |
41 | |
42 | /* 1 second timeout */ | |
43 | #define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) | |
44 | ||
45 | /* Firmware Names */ | |
46 | #define FIRMWARE_RAVEN "amdgpu/raven_vcn.bin" | |
86771d9a | 47 | #define FIRMWARE_PICASSO "amdgpu/picasso_vcn.bin" |
8b47cc9b | 48 | #define FIRMWARE_RAVEN2 "amdgpu/raven2_vcn.bin" |
a8790e24 | 49 | #define FIRMWARE_NAVI10 "amdgpu/navi10_vcn.bin" |
e149a2f6 | 50 | #define FIRMWARE_NAVI14 "amdgpu/navi14_vcn.bin" |
95d0906f LL |
51 | |
52 | MODULE_FIRMWARE(FIRMWARE_RAVEN); | |
86771d9a | 53 | MODULE_FIRMWARE(FIRMWARE_PICASSO); |
8b47cc9b | 54 | MODULE_FIRMWARE(FIRMWARE_RAVEN2); |
a8790e24 | 55 | MODULE_FIRMWARE(FIRMWARE_NAVI10); |
e149a2f6 | 56 | MODULE_FIRMWARE(FIRMWARE_NAVI14); |
95d0906f LL |
57 | |
58 | static void amdgpu_vcn_idle_work_handler(struct work_struct *work); | |
59 | ||
60 | int amdgpu_vcn_sw_init(struct amdgpu_device *adev) | |
61 | { | |
95d0906f LL |
62 | unsigned long bo_size; |
63 | const char *fw_name; | |
64 | const struct common_firmware_header *hdr; | |
62d5b8e3 | 65 | unsigned char fw_check; |
95d0906f LL |
66 | int r; |
67 | ||
68 | INIT_DELAYED_WORK(&adev->vcn.idle_work, amdgpu_vcn_idle_work_handler); | |
69 | ||
70 | switch (adev->asic_type) { | |
71 | case CHIP_RAVEN: | |
741deade | 72 | if (adev->rev_id >= 8) |
8b47cc9b | 73 | fw_name = FIRMWARE_RAVEN2; |
741deade AD |
74 | else if (adev->pdev->device == 0x15d8) |
75 | fw_name = FIRMWARE_PICASSO; | |
8b47cc9b FX |
76 | else |
77 | fw_name = FIRMWARE_RAVEN; | |
95d0906f | 78 | break; |
a8790e24 LL |
79 | case CHIP_NAVI10: |
80 | fw_name = FIRMWARE_NAVI10; | |
450af30c LL |
81 | if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && |
82 | (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) | |
83 | adev->vcn.indirect_sram = true; | |
a8790e24 | 84 | break; |
e149a2f6 JZ |
85 | case CHIP_NAVI14: |
86 | fw_name = FIRMWARE_NAVI14; | |
0377b088 XY |
87 | if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && |
88 | (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) | |
89 | adev->vcn.indirect_sram = true; | |
e149a2f6 | 90 | break; |
95d0906f LL |
91 | default: |
92 | return -EINVAL; | |
93 | } | |
94 | ||
95 | r = request_firmware(&adev->vcn.fw, fw_name, adev->dev); | |
96 | if (r) { | |
97 | dev_err(adev->dev, "amdgpu_vcn: Can't load firmware \"%s\"\n", | |
98 | fw_name); | |
99 | return r; | |
100 | } | |
101 | ||
102 | r = amdgpu_ucode_validate(adev->vcn.fw); | |
103 | if (r) { | |
104 | dev_err(adev->dev, "amdgpu_vcn: Can't validate firmware \"%s\"\n", | |
105 | fw_name); | |
106 | release_firmware(adev->vcn.fw); | |
107 | adev->vcn.fw = NULL; | |
108 | return r; | |
109 | } | |
110 | ||
111 | hdr = (const struct common_firmware_header *)adev->vcn.fw->data; | |
a0b2ac29 | 112 | adev->vcn.fw_version = le32_to_cpu(hdr->ucode_version); |
95d0906f | 113 | |
62d5b8e3 JZ |
114 | /* Bit 20-23, it is encode major and non-zero for new naming convention. |
115 | * This field is part of version minor and DRM_DISABLED_FLAG in old naming | |
116 | * convention. Since the l:wq!atest version minor is 0x5B and DRM_DISABLED_FLAG | |
117 | * is zero in old naming convention, this field is always zero so far. | |
118 | * These four bits are used to tell which naming convention is present. | |
119 | */ | |
120 | fw_check = (le32_to_cpu(hdr->ucode_version) >> 20) & 0xf; | |
121 | if (fw_check) { | |
122 | unsigned int dec_ver, enc_major, enc_minor, vep, fw_rev; | |
123 | ||
124 | fw_rev = le32_to_cpu(hdr->ucode_version) & 0xfff; | |
125 | enc_minor = (le32_to_cpu(hdr->ucode_version) >> 12) & 0xff; | |
126 | enc_major = fw_check; | |
127 | dec_ver = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xf; | |
128 | vep = (le32_to_cpu(hdr->ucode_version) >> 28) & 0xf; | |
129 | DRM_INFO("Found VCN firmware Version ENC: %hu.%hu DEC: %hu VEP: %hu Revision: %hu\n", | |
130 | enc_major, enc_minor, dec_ver, vep, fw_rev); | |
131 | } else { | |
132 | unsigned int version_major, version_minor, family_id; | |
133 | ||
134 | family_id = le32_to_cpu(hdr->ucode_version) & 0xff; | |
135 | version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff; | |
136 | version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff; | |
137 | DRM_INFO("Found VCN firmware Version: %hu.%hu Family ID: %hu\n", | |
138 | version_major, version_minor, family_id); | |
139 | } | |
95d0906f | 140 | |
825da4d9 | 141 | bo_size = AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_CONTEXT_SIZE; |
4d77c0f6 LG |
142 | if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) |
143 | bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); | |
95d0906f LL |
144 | r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, |
145 | AMDGPU_GEM_DOMAIN_VRAM, &adev->vcn.vcpu_bo, | |
146 | &adev->vcn.gpu_addr, &adev->vcn.cpu_addr); | |
147 | if (r) { | |
148 | dev_err(adev->dev, "(%d) failed to allocate vcn bo\n", r); | |
149 | return r; | |
150 | } | |
151 | ||
a77b9fdf LL |
152 | if (adev->vcn.indirect_sram) { |
153 | r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE, | |
154 | AMDGPU_GEM_DOMAIN_VRAM, &adev->vcn.dpg_sram_bo, | |
155 | &adev->vcn.dpg_sram_gpu_addr, &adev->vcn.dpg_sram_cpu_addr); | |
156 | if (r) { | |
157 | dev_err(adev->dev, "(%d) failed to allocate DPG bo\n", r); | |
158 | return r; | |
159 | } | |
160 | } | |
161 | ||
95d0906f LL |
162 | return 0; |
163 | } | |
164 | ||
165 | int amdgpu_vcn_sw_fini(struct amdgpu_device *adev) | |
166 | { | |
101c6fee LL |
167 | int i; |
168 | ||
c9533d1b | 169 | kvfree(adev->vcn.saved_bo); |
95d0906f | 170 | |
a77b9fdf LL |
171 | if (adev->vcn.indirect_sram) { |
172 | amdgpu_bo_free_kernel(&adev->vcn.dpg_sram_bo, | |
173 | &adev->vcn.dpg_sram_gpu_addr, | |
174 | (void **)&adev->vcn.dpg_sram_cpu_addr); | |
175 | } | |
176 | ||
95d0906f LL |
177 | amdgpu_bo_free_kernel(&adev->vcn.vcpu_bo, |
178 | &adev->vcn.gpu_addr, | |
179 | (void **)&adev->vcn.cpu_addr); | |
180 | ||
181 | amdgpu_ring_fini(&adev->vcn.ring_dec); | |
182 | ||
101c6fee LL |
183 | for (i = 0; i < adev->vcn.num_enc_rings; ++i) |
184 | amdgpu_ring_fini(&adev->vcn.ring_enc[i]); | |
185 | ||
0c5e4b3e BZ |
186 | amdgpu_ring_fini(&adev->vcn.ring_jpeg); |
187 | ||
95d0906f LL |
188 | release_firmware(adev->vcn.fw); |
189 | ||
190 | return 0; | |
191 | } | |
192 | ||
193 | int amdgpu_vcn_suspend(struct amdgpu_device *adev) | |
194 | { | |
195 | unsigned size; | |
196 | void *ptr; | |
197 | ||
61ea6f58 RZ |
198 | cancel_delayed_work_sync(&adev->vcn.idle_work); |
199 | ||
95d0906f LL |
200 | if (adev->vcn.vcpu_bo == NULL) |
201 | return 0; | |
202 | ||
95d0906f LL |
203 | size = amdgpu_bo_size(adev->vcn.vcpu_bo); |
204 | ptr = adev->vcn.cpu_addr; | |
205 | ||
c9533d1b | 206 | adev->vcn.saved_bo = kvmalloc(size, GFP_KERNEL); |
95d0906f LL |
207 | if (!adev->vcn.saved_bo) |
208 | return -ENOMEM; | |
209 | ||
210 | memcpy_fromio(adev->vcn.saved_bo, ptr, size); | |
211 | ||
212 | return 0; | |
213 | } | |
214 | ||
215 | int amdgpu_vcn_resume(struct amdgpu_device *adev) | |
216 | { | |
217 | unsigned size; | |
218 | void *ptr; | |
219 | ||
220 | if (adev->vcn.vcpu_bo == NULL) | |
221 | return -EINVAL; | |
222 | ||
223 | size = amdgpu_bo_size(adev->vcn.vcpu_bo); | |
224 | ptr = adev->vcn.cpu_addr; | |
225 | ||
226 | if (adev->vcn.saved_bo != NULL) { | |
227 | memcpy_toio(ptr, adev->vcn.saved_bo, size); | |
c9533d1b | 228 | kvfree(adev->vcn.saved_bo); |
95d0906f LL |
229 | adev->vcn.saved_bo = NULL; |
230 | } else { | |
231 | const struct common_firmware_header *hdr; | |
232 | unsigned offset; | |
233 | ||
234 | hdr = (const struct common_firmware_header *)adev->vcn.fw->data; | |
4d77c0f6 LG |
235 | if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { |
236 | offset = le32_to_cpu(hdr->ucode_array_offset_bytes); | |
237 | memcpy_toio(adev->vcn.cpu_addr, adev->vcn.fw->data + offset, | |
238 | le32_to_cpu(hdr->ucode_size_bytes)); | |
239 | size -= le32_to_cpu(hdr->ucode_size_bytes); | |
240 | ptr += le32_to_cpu(hdr->ucode_size_bytes); | |
241 | } | |
95d0906f LL |
242 | memset_io(ptr, 0, size); |
243 | } | |
244 | ||
245 | return 0; | |
246 | } | |
247 | ||
3e1086cf LL |
248 | static void amdgpu_vcn_idle_work_handler(struct work_struct *work) |
249 | { | |
250 | struct amdgpu_device *adev = | |
251 | container_of(work, struct amdgpu_device, vcn.idle_work.work); | |
bd5d5180 JZ |
252 | unsigned int fences = 0; |
253 | unsigned int i; | |
646e906d AD |
254 | |
255 | for (i = 0; i < adev->vcn.num_enc_rings; ++i) { | |
256 | fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); | |
257 | } | |
3e1086cf | 258 | |
bd5d5180 JZ |
259 | if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { |
260 | struct dpg_pause_state new_state; | |
261 | ||
262 | if (fences) | |
263 | new_state.fw_based = VCN_DPG_STATE__PAUSE; | |
264 | else | |
265 | new_state.fw_based = VCN_DPG_STATE__UNPAUSE; | |
266 | ||
267 | if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg)) | |
268 | new_state.jpeg = VCN_DPG_STATE__PAUSE; | |
269 | else | |
270 | new_state.jpeg = VCN_DPG_STATE__UNPAUSE; | |
271 | ||
9dc7b02a | 272 | adev->vcn.pause_dpg_mode(adev, &new_state); |
bd5d5180 JZ |
273 | } |
274 | ||
7b4e54a9 | 275 | fences += amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg); |
bd5d5180 | 276 | fences += amdgpu_fence_count_emitted(&adev->vcn.ring_dec); |
7b4e54a9 | 277 | |
3e1086cf | 278 | if (fences == 0) { |
3fded222 | 279 | amdgpu_gfx_off_ctrl(adev, true); |
6e4cb4e8 | 280 | if (adev->asic_type < CHIP_NAVI10 && adev->pm.dpm_enabled) |
3e1086cf | 281 | amdgpu_dpm_enable_uvd(adev, false); |
22cc6c5e RZ |
282 | else |
283 | amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, | |
284 | AMD_PG_STATE_GATE); | |
3e1086cf LL |
285 | } else { |
286 | schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT); | |
287 | } | |
288 | } | |
289 | ||
290 | void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) | |
291 | { | |
292 | struct amdgpu_device *adev = ring->adev; | |
293 | bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work); | |
294 | ||
c1ee15b3 | 295 | if (set_clocks) { |
3fded222 | 296 | amdgpu_gfx_off_ctrl(adev, false); |
6e4cb4e8 | 297 | if (adev->asic_type < CHIP_NAVI10 && adev->pm.dpm_enabled) |
22cc6c5e RZ |
298 | amdgpu_dpm_enable_uvd(adev, true); |
299 | else | |
300 | amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, | |
301 | AMD_PG_STATE_UNGATE); | |
3e1086cf | 302 | } |
bd5d5180 JZ |
303 | |
304 | if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { | |
305 | struct dpg_pause_state new_state; | |
12e8b301 JZ |
306 | unsigned int fences = 0; |
307 | unsigned int i; | |
bd5d5180 | 308 | |
12e8b301 JZ |
309 | for (i = 0; i < adev->vcn.num_enc_rings; ++i) { |
310 | fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); | |
311 | } | |
312 | if (fences) | |
bd5d5180 JZ |
313 | new_state.fw_based = VCN_DPG_STATE__PAUSE; |
314 | else | |
12e8b301 | 315 | new_state.fw_based = VCN_DPG_STATE__UNPAUSE; |
bd5d5180 | 316 | |
12e8b301 | 317 | if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg)) |
bd5d5180 JZ |
318 | new_state.jpeg = VCN_DPG_STATE__PAUSE; |
319 | else | |
12e8b301 JZ |
320 | new_state.jpeg = VCN_DPG_STATE__UNPAUSE; |
321 | ||
322 | if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) | |
323 | new_state.fw_based = VCN_DPG_STATE__PAUSE; | |
324 | else if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) | |
325 | new_state.jpeg = VCN_DPG_STATE__PAUSE; | |
bd5d5180 | 326 | |
9dc7b02a | 327 | adev->vcn.pause_dpg_mode(adev, &new_state); |
bd5d5180 | 328 | } |
3e1086cf LL |
329 | } |
330 | ||
331 | void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring) | |
332 | { | |
333 | schedule_delayed_work(&ring->adev->vcn.idle_work, VCN_IDLE_TIMEOUT); | |
334 | } | |
335 | ||
8c303c01 LL |
336 | int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) |
337 | { | |
338 | struct amdgpu_device *adev = ring->adev; | |
339 | uint32_t tmp = 0; | |
340 | unsigned i; | |
341 | int r; | |
342 | ||
45a1a48b | 343 | WREG32(adev->vcn.external.scratch9, 0xCAFEDEAD); |
8c303c01 | 344 | r = amdgpu_ring_alloc(ring, 3); |
dc9eeff8 | 345 | if (r) |
8c303c01 | 346 | return r; |
45a1a48b | 347 | amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.scratch9, 0)); |
8c303c01 LL |
348 | amdgpu_ring_write(ring, 0xDEADBEEF); |
349 | amdgpu_ring_commit(ring); | |
350 | for (i = 0; i < adev->usec_timeout; i++) { | |
45a1a48b | 351 | tmp = RREG32(adev->vcn.external.scratch9); |
8c303c01 LL |
352 | if (tmp == 0xDEADBEEF) |
353 | break; | |
c366be54 | 354 | udelay(1); |
8c303c01 LL |
355 | } |
356 | ||
dc9eeff8 CK |
357 | if (i >= adev->usec_timeout) |
358 | r = -ETIMEDOUT; | |
359 | ||
8c303c01 LL |
360 | return r; |
361 | } | |
362 | ||
add9f9a8 | 363 | static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, |
4c6530fd | 364 | struct amdgpu_bo *bo, |
add9f9a8 | 365 | struct dma_fence **fence) |
95d0906f | 366 | { |
add9f9a8 CK |
367 | struct amdgpu_device *adev = ring->adev; |
368 | struct dma_fence *f = NULL; | |
95d0906f LL |
369 | struct amdgpu_job *job; |
370 | struct amdgpu_ib *ib; | |
95d0906f LL |
371 | uint64_t addr; |
372 | int i, r; | |
373 | ||
95d0906f LL |
374 | r = amdgpu_job_alloc_with_ib(adev, 64, &job); |
375 | if (r) | |
376 | goto err; | |
377 | ||
378 | ib = &job->ibs[0]; | |
379 | addr = amdgpu_bo_gpu_offset(bo); | |
60a2309e | 380 | ib->ptr[0] = PACKET0(adev->vcn.internal.data0, 0); |
95d0906f | 381 | ib->ptr[1] = addr; |
60a2309e | 382 | ib->ptr[2] = PACKET0(adev->vcn.internal.data1, 0); |
95d0906f | 383 | ib->ptr[3] = addr >> 32; |
60a2309e | 384 | ib->ptr[4] = PACKET0(adev->vcn.internal.cmd, 0); |
95d0906f LL |
385 | ib->ptr[5] = 0; |
386 | for (i = 6; i < 16; i += 2) { | |
60a2309e | 387 | ib->ptr[i] = PACKET0(adev->vcn.internal.nop, 0); |
95d0906f LL |
388 | ib->ptr[i+1] = 0; |
389 | } | |
390 | ib->length_dw = 16; | |
391 | ||
ee913fd9 | 392 | r = amdgpu_job_submit_direct(job, ring, &f); |
4c6530fd LL |
393 | if (r) |
394 | goto err_free; | |
95d0906f | 395 | |
add9f9a8 CK |
396 | amdgpu_bo_fence(bo, f, false); |
397 | amdgpu_bo_unreserve(bo); | |
398 | amdgpu_bo_unref(&bo); | |
95d0906f LL |
399 | |
400 | if (fence) | |
401 | *fence = dma_fence_get(f); | |
95d0906f LL |
402 | dma_fence_put(f); |
403 | ||
404 | return 0; | |
405 | ||
406 | err_free: | |
407 | amdgpu_job_free(job); | |
408 | ||
409 | err: | |
add9f9a8 CK |
410 | amdgpu_bo_unreserve(bo); |
411 | amdgpu_bo_unref(&bo); | |
95d0906f LL |
412 | return r; |
413 | } | |
414 | ||
415 | static int amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, | |
416 | struct dma_fence **fence) | |
417 | { | |
418 | struct amdgpu_device *adev = ring->adev; | |
add9f9a8 | 419 | struct amdgpu_bo *bo = NULL; |
95d0906f LL |
420 | uint32_t *msg; |
421 | int r, i; | |
422 | ||
add9f9a8 CK |
423 | r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE, |
424 | AMDGPU_GEM_DOMAIN_VRAM, | |
425 | &bo, NULL, (void **)&msg); | |
95d0906f LL |
426 | if (r) |
427 | return r; | |
428 | ||
2d8a425b | 429 | msg[0] = cpu_to_le32(0x00000028); |
3b8f5ab3 | 430 | msg[1] = cpu_to_le32(0x00000038); |
2d8a425b | 431 | msg[2] = cpu_to_le32(0x00000001); |
95d0906f | 432 | msg[3] = cpu_to_le32(0x00000000); |
2d8a425b | 433 | msg[4] = cpu_to_le32(handle); |
95d0906f | 434 | msg[5] = cpu_to_le32(0x00000000); |
2d8a425b LL |
435 | msg[6] = cpu_to_le32(0x00000001); |
436 | msg[7] = cpu_to_le32(0x00000028); | |
3b8f5ab3 | 437 | msg[8] = cpu_to_le32(0x00000010); |
95d0906f | 438 | msg[9] = cpu_to_le32(0x00000000); |
2d8a425b LL |
439 | msg[10] = cpu_to_le32(0x00000007); |
440 | msg[11] = cpu_to_le32(0x00000000); | |
3b8f5ab3 LL |
441 | msg[12] = cpu_to_le32(0x00000780); |
442 | msg[13] = cpu_to_le32(0x00000440); | |
443 | for (i = 14; i < 1024; ++i) | |
95d0906f LL |
444 | msg[i] = cpu_to_le32(0x0); |
445 | ||
4c6530fd | 446 | return amdgpu_vcn_dec_send_msg(ring, bo, fence); |
95d0906f LL |
447 | } |
448 | ||
449 | static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, | |
4c6530fd | 450 | struct dma_fence **fence) |
95d0906f LL |
451 | { |
452 | struct amdgpu_device *adev = ring->adev; | |
add9f9a8 | 453 | struct amdgpu_bo *bo = NULL; |
95d0906f LL |
454 | uint32_t *msg; |
455 | int r, i; | |
456 | ||
add9f9a8 CK |
457 | r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE, |
458 | AMDGPU_GEM_DOMAIN_VRAM, | |
459 | &bo, NULL, (void **)&msg); | |
95d0906f LL |
460 | if (r) |
461 | return r; | |
462 | ||
2d8a425b LL |
463 | msg[0] = cpu_to_le32(0x00000028); |
464 | msg[1] = cpu_to_le32(0x00000018); | |
465 | msg[2] = cpu_to_le32(0x00000000); | |
466 | msg[3] = cpu_to_le32(0x00000002); | |
467 | msg[4] = cpu_to_le32(handle); | |
468 | msg[5] = cpu_to_le32(0x00000000); | |
469 | for (i = 6; i < 1024; ++i) | |
95d0906f LL |
470 | msg[i] = cpu_to_le32(0x0); |
471 | ||
4c6530fd | 472 | return amdgpu_vcn_dec_send_msg(ring, bo, fence); |
95d0906f LL |
473 | } |
474 | ||
95d0906f LL |
475 | int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout) |
476 | { | |
477 | struct dma_fence *fence; | |
478 | long r; | |
479 | ||
480 | r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL); | |
98079389 | 481 | if (r) |
95d0906f | 482 | goto error; |
95d0906f | 483 | |
4c6530fd | 484 | r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence); |
98079389 | 485 | if (r) |
95d0906f | 486 | goto error; |
95d0906f LL |
487 | |
488 | r = dma_fence_wait_timeout(fence, false, timeout); | |
98079389 | 489 | if (r == 0) |
95d0906f | 490 | r = -ETIMEDOUT; |
98079389 | 491 | else if (r > 0) |
95d0906f | 492 | r = 0; |
95d0906f LL |
493 | |
494 | dma_fence_put(fence); | |
95d0906f LL |
495 | error: |
496 | return r; | |
497 | } | |
2d531d81 | 498 | |
3e1086cf LL |
499 | int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring) |
500 | { | |
501 | struct amdgpu_device *adev = ring->adev; | |
e038b901 | 502 | uint32_t rptr; |
3e1086cf LL |
503 | unsigned i; |
504 | int r; | |
505 | ||
506 | r = amdgpu_ring_alloc(ring, 16); | |
dc9eeff8 | 507 | if (r) |
3e1086cf | 508 | return r; |
dc9eeff8 | 509 | |
e038b901 S |
510 | rptr = amdgpu_ring_get_rptr(ring); |
511 | ||
c3bd3040 | 512 | amdgpu_ring_write(ring, VCN_ENC_CMD_END); |
3e1086cf LL |
513 | amdgpu_ring_commit(ring); |
514 | ||
515 | for (i = 0; i < adev->usec_timeout; i++) { | |
516 | if (amdgpu_ring_get_rptr(ring) != rptr) | |
517 | break; | |
c366be54 | 518 | udelay(1); |
3e1086cf LL |
519 | } |
520 | ||
dc9eeff8 | 521 | if (i >= adev->usec_timeout) |
3e1086cf | 522 | r = -ETIMEDOUT; |
3e1086cf LL |
523 | |
524 | return r; | |
525 | } | |
526 | ||
2d531d81 LL |
527 | static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, |
528 | struct dma_fence **fence) | |
529 | { | |
25547cfd | 530 | const unsigned ib_size_dw = 16; |
2d531d81 LL |
531 | struct amdgpu_job *job; |
532 | struct amdgpu_ib *ib; | |
533 | struct dma_fence *f = NULL; | |
534 | uint64_t dummy; | |
535 | int i, r; | |
536 | ||
537 | r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); | |
538 | if (r) | |
539 | return r; | |
540 | ||
541 | ib = &job->ibs[0]; | |
2d531d81 LL |
542 | dummy = ib->gpu_addr + 1024; |
543 | ||
2d531d81 | 544 | ib->length_dw = 0; |
25547cfd LL |
545 | ib->ptr[ib->length_dw++] = 0x00000018; |
546 | ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ | |
2d531d81 | 547 | ib->ptr[ib->length_dw++] = handle; |
25547cfd LL |
548 | ib->ptr[ib->length_dw++] = upper_32_bits(dummy); |
549 | ib->ptr[ib->length_dw++] = dummy; | |
550 | ib->ptr[ib->length_dw++] = 0x0000000b; | |
2d531d81 | 551 | |
25547cfd LL |
552 | ib->ptr[ib->length_dw++] = 0x00000014; |
553 | ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ | |
554 | ib->ptr[ib->length_dw++] = 0x0000001c; | |
2d531d81 LL |
555 | ib->ptr[ib->length_dw++] = 0x00000000; |
556 | ib->ptr[ib->length_dw++] = 0x00000000; | |
557 | ||
25547cfd LL |
558 | ib->ptr[ib->length_dw++] = 0x00000008; |
559 | ib->ptr[ib->length_dw++] = 0x08000001; /* op initialize */ | |
2d531d81 LL |
560 | |
561 | for (i = ib->length_dw; i < ib_size_dw; ++i) | |
562 | ib->ptr[i] = 0x0; | |
563 | ||
ee913fd9 | 564 | r = amdgpu_job_submit_direct(job, ring, &f); |
2d531d81 LL |
565 | if (r) |
566 | goto err; | |
567 | ||
2d531d81 LL |
568 | if (fence) |
569 | *fence = dma_fence_get(f); | |
570 | dma_fence_put(f); | |
25547cfd | 571 | |
2d531d81 LL |
572 | return 0; |
573 | ||
574 | err: | |
575 | amdgpu_job_free(job); | |
576 | return r; | |
577 | } | |
578 | ||
579 | static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, | |
25547cfd | 580 | struct dma_fence **fence) |
2d531d81 | 581 | { |
25547cfd | 582 | const unsigned ib_size_dw = 16; |
2d531d81 LL |
583 | struct amdgpu_job *job; |
584 | struct amdgpu_ib *ib; | |
585 | struct dma_fence *f = NULL; | |
25547cfd | 586 | uint64_t dummy; |
2d531d81 LL |
587 | int i, r; |
588 | ||
589 | r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); | |
590 | if (r) | |
591 | return r; | |
592 | ||
593 | ib = &job->ibs[0]; | |
25547cfd | 594 | dummy = ib->gpu_addr + 1024; |
2d531d81 | 595 | |
2d531d81 | 596 | ib->length_dw = 0; |
25547cfd LL |
597 | ib->ptr[ib->length_dw++] = 0x00000018; |
598 | ib->ptr[ib->length_dw++] = 0x00000001; | |
2d531d81 | 599 | ib->ptr[ib->length_dw++] = handle; |
25547cfd LL |
600 | ib->ptr[ib->length_dw++] = upper_32_bits(dummy); |
601 | ib->ptr[ib->length_dw++] = dummy; | |
602 | ib->ptr[ib->length_dw++] = 0x0000000b; | |
2d531d81 | 603 | |
25547cfd LL |
604 | ib->ptr[ib->length_dw++] = 0x00000014; |
605 | ib->ptr[ib->length_dw++] = 0x00000002; | |
606 | ib->ptr[ib->length_dw++] = 0x0000001c; | |
2d531d81 | 607 | ib->ptr[ib->length_dw++] = 0x00000000; |
2d531d81 LL |
608 | ib->ptr[ib->length_dw++] = 0x00000000; |
609 | ||
25547cfd LL |
610 | ib->ptr[ib->length_dw++] = 0x00000008; |
611 | ib->ptr[ib->length_dw++] = 0x08000002; /* op close session */ | |
2d531d81 LL |
612 | |
613 | for (i = ib->length_dw; i < ib_size_dw; ++i) | |
614 | ib->ptr[i] = 0x0; | |
615 | ||
ee913fd9 | 616 | r = amdgpu_job_submit_direct(job, ring, &f); |
25547cfd LL |
617 | if (r) |
618 | goto err; | |
2d531d81 LL |
619 | |
620 | if (fence) | |
621 | *fence = dma_fence_get(f); | |
622 | dma_fence_put(f); | |
25547cfd | 623 | |
2d531d81 LL |
624 | return 0; |
625 | ||
626 | err: | |
627 | amdgpu_job_free(job); | |
628 | return r; | |
629 | } | |
630 | ||
2d531d81 LL |
631 | int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) |
632 | { | |
633 | struct dma_fence *fence = NULL; | |
634 | long r; | |
635 | ||
636 | r = amdgpu_vcn_enc_get_create_msg(ring, 1, NULL); | |
98079389 | 637 | if (r) |
2d531d81 | 638 | goto error; |
2d531d81 | 639 | |
25547cfd | 640 | r = amdgpu_vcn_enc_get_destroy_msg(ring, 1, &fence); |
98079389 | 641 | if (r) |
2d531d81 | 642 | goto error; |
2d531d81 LL |
643 | |
644 | r = dma_fence_wait_timeout(fence, false, timeout); | |
98079389 | 645 | if (r == 0) |
2d531d81 | 646 | r = -ETIMEDOUT; |
98079389 | 647 | else if (r > 0) |
2d531d81 | 648 | r = 0; |
98079389 | 649 | |
2d531d81 LL |
650 | error: |
651 | dma_fence_put(fence); | |
652 | return r; | |
653 | } | |
b1d37606 BZ |
654 | |
655 | int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring) | |
656 | { | |
657 | struct amdgpu_device *adev = ring->adev; | |
658 | uint32_t tmp = 0; | |
659 | unsigned i; | |
660 | int r; | |
661 | ||
9085914a | 662 | WREG32(adev->vcn.external.jpeg_pitch, 0xCAFEDEAD); |
b1d37606 | 663 | r = amdgpu_ring_alloc(ring, 3); |
dc9eeff8 | 664 | if (r) |
b1d37606 | 665 | return r; |
b1d37606 | 666 | |
9085914a | 667 | amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.jpeg_pitch, 0)); |
b1d37606 BZ |
668 | amdgpu_ring_write(ring, 0xDEADBEEF); |
669 | amdgpu_ring_commit(ring); | |
670 | ||
671 | for (i = 0; i < adev->usec_timeout; i++) { | |
9085914a | 672 | tmp = RREG32(adev->vcn.external.jpeg_pitch); |
b1d37606 BZ |
673 | if (tmp == 0xDEADBEEF) |
674 | break; | |
c366be54 | 675 | udelay(1); |
b1d37606 BZ |
676 | } |
677 | ||
dc9eeff8 CK |
678 | if (i >= adev->usec_timeout) |
679 | r = -ETIMEDOUT; | |
b1d37606 BZ |
680 | |
681 | return r; | |
682 | } | |
6173040f BZ |
683 | |
684 | static int amdgpu_vcn_jpeg_set_reg(struct amdgpu_ring *ring, uint32_t handle, | |
685 | struct dma_fence **fence) | |
686 | { | |
687 | struct amdgpu_device *adev = ring->adev; | |
688 | struct amdgpu_job *job; | |
689 | struct amdgpu_ib *ib; | |
690 | struct dma_fence *f = NULL; | |
691 | const unsigned ib_size_dw = 16; | |
692 | int i, r; | |
693 | ||
694 | r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); | |
695 | if (r) | |
696 | return r; | |
697 | ||
698 | ib = &job->ibs[0]; | |
699 | ||
54bb93c2 | 700 | ib->ptr[0] = PACKETJ(adev->vcn.internal.jpeg_pitch, 0, 0, PACKETJ_TYPE0); |
6173040f BZ |
701 | ib->ptr[1] = 0xDEADBEEF; |
702 | for (i = 2; i < 16; i += 2) { | |
703 | ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6); | |
704 | ib->ptr[i+1] = 0; | |
705 | } | |
706 | ib->length_dw = 16; | |
707 | ||
ee913fd9 | 708 | r = amdgpu_job_submit_direct(job, ring, &f); |
6173040f BZ |
709 | if (r) |
710 | goto err; | |
711 | ||
6173040f BZ |
712 | if (fence) |
713 | *fence = dma_fence_get(f); | |
714 | dma_fence_put(f); | |
715 | ||
716 | return 0; | |
717 | ||
718 | err: | |
719 | amdgpu_job_free(job); | |
720 | return r; | |
721 | } | |
722 | ||
723 | int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout) | |
724 | { | |
725 | struct amdgpu_device *adev = ring->adev; | |
726 | uint32_t tmp = 0; | |
727 | unsigned i; | |
728 | struct dma_fence *fence = NULL; | |
729 | long r = 0; | |
730 | ||
731 | r = amdgpu_vcn_jpeg_set_reg(ring, 1, &fence); | |
98079389 | 732 | if (r) |
6173040f | 733 | goto error; |
6173040f BZ |
734 | |
735 | r = dma_fence_wait_timeout(fence, false, timeout); | |
736 | if (r == 0) { | |
6173040f BZ |
737 | r = -ETIMEDOUT; |
738 | goto error; | |
739 | } else if (r < 0) { | |
6173040f | 740 | goto error; |
98079389 | 741 | } else { |
6173040f | 742 | r = 0; |
98079389 | 743 | } |
6173040f BZ |
744 | |
745 | for (i = 0; i < adev->usec_timeout; i++) { | |
54bb93c2 | 746 | tmp = RREG32(adev->vcn.external.jpeg_pitch); |
6173040f BZ |
747 | if (tmp == 0xDEADBEEF) |
748 | break; | |
c366be54 | 749 | udelay(1); |
6173040f BZ |
750 | } |
751 | ||
98079389 CK |
752 | if (i >= adev->usec_timeout) |
753 | r = -ETIMEDOUT; | |
6173040f BZ |
754 | |
755 | dma_fence_put(fence); | |
6173040f BZ |
756 | error: |
757 | return r; | |
758 | } |