Commit | Line | Data |
---|---|---|
e409b128 CK |
1 | /* |
2 | * Copyright 2013 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 <christian.koenig@amd.com> | |
23 | */ | |
24 | ||
25 | #include <drm/drmP.h> | |
26 | #include "radeon.h" | |
27 | #include "radeon_asic.h" | |
28 | #include "r600d.h" | |
29 | ||
30 | /** | |
31 | * uvd_v1_0_get_rptr - get read pointer | |
32 | * | |
33 | * @rdev: radeon_device pointer | |
34 | * @ring: radeon_ring pointer | |
35 | * | |
36 | * Returns the current hardware read pointer | |
37 | */ | |
38 | uint32_t uvd_v1_0_get_rptr(struct radeon_device *rdev, | |
39 | struct radeon_ring *ring) | |
40 | { | |
41 | return RREG32(UVD_RBC_RB_RPTR); | |
42 | } | |
43 | ||
44 | /** | |
45 | * uvd_v1_0_get_wptr - get write pointer | |
46 | * | |
47 | * @rdev: radeon_device pointer | |
48 | * @ring: radeon_ring pointer | |
49 | * | |
50 | * Returns the current hardware write pointer | |
51 | */ | |
52 | uint32_t uvd_v1_0_get_wptr(struct radeon_device *rdev, | |
53 | struct radeon_ring *ring) | |
54 | { | |
55 | return RREG32(UVD_RBC_RB_WPTR); | |
56 | } | |
57 | ||
58 | /** | |
59 | * uvd_v1_0_set_wptr - set write pointer | |
60 | * | |
61 | * @rdev: radeon_device pointer | |
62 | * @ring: radeon_ring pointer | |
63 | * | |
64 | * Commits the write pointer to the hardware | |
65 | */ | |
66 | void uvd_v1_0_set_wptr(struct radeon_device *rdev, | |
67 | struct radeon_ring *ring) | |
68 | { | |
69 | WREG32(UVD_RBC_RB_WPTR, ring->wptr); | |
70 | } | |
71 | ||
72 | /** | |
73 | * uvd_v1_0_init - start and test UVD block | |
74 | * | |
75 | * @rdev: radeon_device pointer | |
76 | * | |
77 | * Initialize the hardware, boot up the VCPU and do some testing | |
78 | */ | |
79 | int uvd_v1_0_init(struct radeon_device *rdev) | |
80 | { | |
81 | struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; | |
82 | uint32_t tmp; | |
83 | int r; | |
84 | ||
85 | /* raise clocks while booting up the VCPU */ | |
e4518762 CK |
86 | if (rdev->family < CHIP_RV740) |
87 | radeon_set_uvd_clocks(rdev, 10000, 10000); | |
88 | else | |
89 | radeon_set_uvd_clocks(rdev, 53300, 40000); | |
e409b128 | 90 | |
a7f28f0f AD |
91 | r = uvd_v1_0_start(rdev); |
92 | if (r) | |
93 | goto done; | |
e409b128 CK |
94 | |
95 | ring->ready = true; | |
96 | r = radeon_ring_test(rdev, R600_RING_TYPE_UVD_INDEX, ring); | |
97 | if (r) { | |
98 | ring->ready = false; | |
99 | goto done; | |
100 | } | |
101 | ||
102 | r = radeon_ring_lock(rdev, ring, 10); | |
103 | if (r) { | |
104 | DRM_ERROR("radeon: ring failed to lock UVD ring (%d).\n", r); | |
105 | goto done; | |
106 | } | |
107 | ||
108 | tmp = PACKET0(UVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0); | |
109 | radeon_ring_write(ring, tmp); | |
110 | radeon_ring_write(ring, 0xFFFFF); | |
111 | ||
112 | tmp = PACKET0(UVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0); | |
113 | radeon_ring_write(ring, tmp); | |
114 | radeon_ring_write(ring, 0xFFFFF); | |
115 | ||
116 | tmp = PACKET0(UVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0); | |
117 | radeon_ring_write(ring, tmp); | |
118 | radeon_ring_write(ring, 0xFFFFF); | |
119 | ||
120 | /* Clear timeout status bits */ | |
121 | radeon_ring_write(ring, PACKET0(UVD_SEMA_TIMEOUT_STATUS, 0)); | |
122 | radeon_ring_write(ring, 0x8); | |
123 | ||
124 | radeon_ring_write(ring, PACKET0(UVD_SEMA_CNTL, 0)); | |
125 | radeon_ring_write(ring, 3); | |
126 | ||
1538a9e0 | 127 | radeon_ring_unlock_commit(rdev, ring, false); |
e409b128 CK |
128 | |
129 | done: | |
130 | /* lower clocks again */ | |
131 | radeon_set_uvd_clocks(rdev, 0, 0); | |
132 | ||
133 | if (!r) | |
134 | DRM_INFO("UVD initialized successfully.\n"); | |
135 | ||
136 | return r; | |
137 | } | |
138 | ||
139 | /** | |
140 | * uvd_v1_0_fini - stop the hardware block | |
141 | * | |
142 | * @rdev: radeon_device pointer | |
143 | * | |
144 | * Stop the UVD block, mark ring as not ready any more | |
145 | */ | |
146 | void uvd_v1_0_fini(struct radeon_device *rdev) | |
147 | { | |
148 | struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; | |
149 | ||
150 | uvd_v1_0_stop(rdev); | |
151 | ring->ready = false; | |
152 | } | |
153 | ||
154 | /** | |
155 | * uvd_v1_0_start - start UVD block | |
156 | * | |
157 | * @rdev: radeon_device pointer | |
158 | * | |
159 | * Setup and start the UVD block | |
160 | */ | |
161 | int uvd_v1_0_start(struct radeon_device *rdev) | |
162 | { | |
163 | struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; | |
164 | uint32_t rb_bufsz; | |
165 | int i, j, r; | |
166 | ||
167 | /* disable byte swapping */ | |
168 | u32 lmi_swap_cntl = 0; | |
169 | u32 mp_swap_cntl = 0; | |
170 | ||
171 | /* disable clock gating */ | |
172 | WREG32(UVD_CGC_GATE, 0); | |
173 | ||
174 | /* disable interupt */ | |
175 | WREG32_P(UVD_MASTINT_EN, 0, ~(1 << 1)); | |
176 | ||
177 | /* Stall UMC and register bus before resetting VCPU */ | |
178 | WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); | |
179 | WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); | |
180 | mdelay(1); | |
181 | ||
182 | /* put LMI, VCPU, RBC etc... into reset */ | |
183 | WREG32(UVD_SOFT_RESET, LMI_SOFT_RESET | VCPU_SOFT_RESET | | |
184 | LBSI_SOFT_RESET | RBC_SOFT_RESET | CSM_SOFT_RESET | | |
185 | CXW_SOFT_RESET | TAP_SOFT_RESET | LMI_UMC_SOFT_RESET); | |
186 | mdelay(5); | |
187 | ||
188 | /* take UVD block out of reset */ | |
189 | WREG32_P(SRBM_SOFT_RESET, 0, ~SOFT_RESET_UVD); | |
190 | mdelay(5); | |
191 | ||
192 | /* initialize UVD memory controller */ | |
193 | WREG32(UVD_LMI_CTRL, 0x40 | (1 << 8) | (1 << 13) | | |
194 | (1 << 21) | (1 << 9) | (1 << 20)); | |
195 | ||
196 | #ifdef __BIG_ENDIAN | |
197 | /* swap (8 in 32) RB and IB */ | |
198 | lmi_swap_cntl = 0xa; | |
199 | mp_swap_cntl = 0; | |
200 | #endif | |
201 | WREG32(UVD_LMI_SWAP_CNTL, lmi_swap_cntl); | |
202 | WREG32(UVD_MP_SWAP_CNTL, mp_swap_cntl); | |
203 | ||
204 | WREG32(UVD_MPC_SET_MUXA0, 0x40c2040); | |
205 | WREG32(UVD_MPC_SET_MUXA1, 0x0); | |
206 | WREG32(UVD_MPC_SET_MUXB0, 0x40c2040); | |
207 | WREG32(UVD_MPC_SET_MUXB1, 0x0); | |
208 | WREG32(UVD_MPC_SET_ALU, 0); | |
209 | WREG32(UVD_MPC_SET_MUX, 0x88); | |
210 | ||
211 | /* take all subblocks out of reset, except VCPU */ | |
212 | WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); | |
213 | mdelay(5); | |
214 | ||
215 | /* enable VCPU clock */ | |
216 | WREG32(UVD_VCPU_CNTL, 1 << 9); | |
217 | ||
bcf6f1e9 CK |
218 | /* enable UMC */ |
219 | WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); | |
e409b128 CK |
220 | |
221 | /* boot up the VCPU */ | |
222 | WREG32(UVD_SOFT_RESET, 0); | |
223 | mdelay(10); | |
224 | ||
225 | WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); | |
226 | ||
227 | for (i = 0; i < 10; ++i) { | |
228 | uint32_t status; | |
229 | for (j = 0; j < 100; ++j) { | |
230 | status = RREG32(UVD_STATUS); | |
231 | if (status & 2) | |
232 | break; | |
233 | mdelay(10); | |
234 | } | |
235 | r = 0; | |
236 | if (status & 2) | |
237 | break; | |
238 | ||
239 | DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n"); | |
240 | WREG32_P(UVD_SOFT_RESET, VCPU_SOFT_RESET, ~VCPU_SOFT_RESET); | |
241 | mdelay(10); | |
242 | WREG32_P(UVD_SOFT_RESET, 0, ~VCPU_SOFT_RESET); | |
243 | mdelay(10); | |
244 | r = -1; | |
245 | } | |
246 | ||
247 | if (r) { | |
248 | DRM_ERROR("UVD not responding, giving up!!!\n"); | |
249 | return r; | |
250 | } | |
251 | ||
252 | /* enable interupt */ | |
253 | WREG32_P(UVD_MASTINT_EN, 3<<1, ~(3 << 1)); | |
254 | ||
255 | /* force RBC into idle state */ | |
256 | WREG32(UVD_RBC_RB_CNTL, 0x11010101); | |
257 | ||
258 | /* Set the write pointer delay */ | |
259 | WREG32(UVD_RBC_RB_WPTR_CNTL, 0); | |
260 | ||
261 | /* programm the 4GB memory segment for rptr and ring buffer */ | |
262 | WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) | | |
263 | (0x7 << 16) | (0x1 << 31)); | |
264 | ||
265 | /* Initialize the ring buffer's read and write pointers */ | |
266 | WREG32(UVD_RBC_RB_RPTR, 0x0); | |
267 | ||
ff212f25 | 268 | ring->wptr = RREG32(UVD_RBC_RB_RPTR); |
e409b128 CK |
269 | WREG32(UVD_RBC_RB_WPTR, ring->wptr); |
270 | ||
271 | /* set the ring address */ | |
272 | WREG32(UVD_RBC_RB_BASE, ring->gpu_addr); | |
273 | ||
274 | /* Set ring buffer size */ | |
9c725e5b | 275 | rb_bufsz = order_base_2(ring->ring_size); |
e409b128 CK |
276 | rb_bufsz = (0x1 << 8) | rb_bufsz; |
277 | WREG32_P(UVD_RBC_RB_CNTL, rb_bufsz, ~0x11f1f); | |
278 | ||
279 | return 0; | |
280 | } | |
281 | ||
282 | /** | |
283 | * uvd_v1_0_stop - stop UVD block | |
284 | * | |
285 | * @rdev: radeon_device pointer | |
286 | * | |
287 | * stop the UVD block | |
288 | */ | |
289 | void uvd_v1_0_stop(struct radeon_device *rdev) | |
290 | { | |
291 | /* force RBC into idle state */ | |
292 | WREG32(UVD_RBC_RB_CNTL, 0x11010101); | |
293 | ||
294 | /* Stall UMC and register bus before resetting VCPU */ | |
295 | WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); | |
296 | WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); | |
297 | mdelay(1); | |
298 | ||
299 | /* put VCPU into reset */ | |
300 | WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); | |
301 | mdelay(5); | |
302 | ||
303 | /* disable VCPU clock */ | |
304 | WREG32(UVD_VCPU_CNTL, 0x0); | |
305 | ||
306 | /* Unstall UMC and register bus */ | |
307 | WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); | |
308 | WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); | |
309 | } | |
310 | ||
311 | /** | |
312 | * uvd_v1_0_ring_test - register write test | |
313 | * | |
314 | * @rdev: radeon_device pointer | |
315 | * @ring: radeon_ring pointer | |
316 | * | |
317 | * Test if we can successfully write to the context register | |
318 | */ | |
319 | int uvd_v1_0_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) | |
320 | { | |
321 | uint32_t tmp = 0; | |
322 | unsigned i; | |
323 | int r; | |
324 | ||
325 | WREG32(UVD_CONTEXT_ID, 0xCAFEDEAD); | |
326 | r = radeon_ring_lock(rdev, ring, 3); | |
327 | if (r) { | |
328 | DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", | |
329 | ring->idx, r); | |
330 | return r; | |
331 | } | |
332 | radeon_ring_write(ring, PACKET0(UVD_CONTEXT_ID, 0)); | |
333 | radeon_ring_write(ring, 0xDEADBEEF); | |
1538a9e0 | 334 | radeon_ring_unlock_commit(rdev, ring, false); |
e409b128 CK |
335 | for (i = 0; i < rdev->usec_timeout; i++) { |
336 | tmp = RREG32(UVD_CONTEXT_ID); | |
337 | if (tmp == 0xDEADBEEF) | |
338 | break; | |
339 | DRM_UDELAY(1); | |
340 | } | |
341 | ||
342 | if (i < rdev->usec_timeout) { | |
343 | DRM_INFO("ring test on %d succeeded in %d usecs\n", | |
344 | ring->idx, i); | |
345 | } else { | |
346 | DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", | |
347 | ring->idx, tmp); | |
348 | r = -EINVAL; | |
349 | } | |
350 | return r; | |
351 | } | |
352 | ||
353 | /** | |
354 | * uvd_v1_0_semaphore_emit - emit semaphore command | |
355 | * | |
356 | * @rdev: radeon_device pointer | |
357 | * @ring: radeon_ring pointer | |
358 | * @semaphore: semaphore to emit commands for | |
359 | * @emit_wait: true if we should emit a wait command | |
360 | * | |
361 | * Emit a semaphore command (either wait or signal) to the UVD ring. | |
362 | */ | |
1654b817 | 363 | bool uvd_v1_0_semaphore_emit(struct radeon_device *rdev, |
e409b128 CK |
364 | struct radeon_ring *ring, |
365 | struct radeon_semaphore *semaphore, | |
366 | bool emit_wait) | |
367 | { | |
368 | uint64_t addr = semaphore->gpu_addr; | |
369 | ||
370 | radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_LOW, 0)); | |
371 | radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); | |
372 | ||
373 | radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_HIGH, 0)); | |
374 | radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); | |
375 | ||
376 | radeon_ring_write(ring, PACKET0(UVD_SEMA_CMD, 0)); | |
377 | radeon_ring_write(ring, emit_wait ? 1 : 0); | |
1654b817 CK |
378 | |
379 | return true; | |
e409b128 CK |
380 | } |
381 | ||
382 | /** | |
383 | * uvd_v1_0_ib_execute - execute indirect buffer | |
384 | * | |
385 | * @rdev: radeon_device pointer | |
386 | * @ib: indirect buffer to execute | |
387 | * | |
388 | * Write ring commands to execute the indirect buffer | |
389 | */ | |
390 | void uvd_v1_0_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) | |
391 | { | |
392 | struct radeon_ring *ring = &rdev->ring[ib->ring]; | |
393 | ||
394 | radeon_ring_write(ring, PACKET0(UVD_RBC_IB_BASE, 0)); | |
395 | radeon_ring_write(ring, ib->gpu_addr); | |
396 | radeon_ring_write(ring, PACKET0(UVD_RBC_IB_SIZE, 0)); | |
397 | radeon_ring_write(ring, ib->length_dw); | |
398 | } | |
399 | ||
400 | /** | |
401 | * uvd_v1_0_ib_test - test ib execution | |
402 | * | |
403 | * @rdev: radeon_device pointer | |
404 | * @ring: radeon_ring pointer | |
405 | * | |
406 | * Test if we can successfully execute an IB | |
407 | */ | |
408 | int uvd_v1_0_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) | |
409 | { | |
410 | struct radeon_fence *fence = NULL; | |
411 | int r; | |
412 | ||
e4518762 CK |
413 | if (rdev->family < CHIP_RV740) |
414 | r = radeon_set_uvd_clocks(rdev, 10000, 10000); | |
415 | else | |
416 | r = radeon_set_uvd_clocks(rdev, 53300, 40000); | |
e409b128 CK |
417 | if (r) { |
418 | DRM_ERROR("radeon: failed to raise UVD clocks (%d).\n", r); | |
419 | return r; | |
420 | } | |
421 | ||
422 | r = radeon_uvd_get_create_msg(rdev, ring->idx, 1, NULL); | |
423 | if (r) { | |
424 | DRM_ERROR("radeon: failed to get create msg (%d).\n", r); | |
425 | goto error; | |
426 | } | |
427 | ||
428 | r = radeon_uvd_get_destroy_msg(rdev, ring->idx, 1, &fence); | |
429 | if (r) { | |
430 | DRM_ERROR("radeon: failed to get destroy ib (%d).\n", r); | |
431 | goto error; | |
432 | } | |
433 | ||
434 | r = radeon_fence_wait(fence, false); | |
435 | if (r) { | |
436 | DRM_ERROR("radeon: fence wait failed (%d).\n", r); | |
437 | goto error; | |
438 | } | |
439 | DRM_INFO("ib test on ring %d succeeded\n", ring->idx); | |
440 | error: | |
441 | radeon_fence_unref(&fence); | |
442 | radeon_set_uvd_clocks(rdev, 0, 0); | |
443 | return r; | |
444 | } |