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