xref: /openbmc/linux/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c (revision d2c43ff1)
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  */
23 
24 #include <linux/firmware.h>
25 #include <drm/drmP.h>
26 #include "amdgpu.h"
27 #include "amdgpu_vcn.h"
28 #include "soc15d.h"
29 #include "soc15_common.h"
30 
31 #include "vega10/soc15ip.h"
32 #include "raven1/VCN/vcn_1_0_offset.h"
33 #include "raven1/VCN/vcn_1_0_sh_mask.h"
34 #include "vega10/HDP/hdp_4_0_offset.h"
35 #include "raven1/MMHUB/mmhub_9_1_offset.h"
36 #include "raven1/MMHUB/mmhub_9_1_sh_mask.h"
37 
38 static int vcn_v1_0_start(struct amdgpu_device *adev);
39 static int vcn_v1_0_stop(struct amdgpu_device *adev);
40 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev);
41 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev);
42 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev);
43 
44 /**
45  * vcn_v1_0_early_init - set function pointers
46  *
47  * @handle: amdgpu_device pointer
48  *
49  * Set ring and irq function pointers
50  */
51 static int vcn_v1_0_early_init(void *handle)
52 {
53 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
54 
55 	adev->vcn.num_enc_rings = 2;
56 
57 	vcn_v1_0_set_dec_ring_funcs(adev);
58 	vcn_v1_0_set_enc_ring_funcs(adev);
59 	vcn_v1_0_set_irq_funcs(adev);
60 
61 	return 0;
62 }
63 
64 /**
65  * vcn_v1_0_sw_init - sw init for VCN block
66  *
67  * @handle: amdgpu_device pointer
68  *
69  * Load firmware and sw initialization
70  */
71 static int vcn_v1_0_sw_init(void *handle)
72 {
73 	struct amdgpu_ring *ring;
74 	int i, r;
75 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
76 
77 	/* VCN DEC TRAP */
78 	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_VCN, 124, &adev->vcn.irq);
79 	if (r)
80 		return r;
81 
82 	/* VCN ENC TRAP */
83 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
84 		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_VCN, i + 119,
85 					&adev->vcn.irq);
86 		if (r)
87 			return r;
88 	}
89 
90 	r = amdgpu_vcn_sw_init(adev);
91 	if (r)
92 		return r;
93 
94 	r = amdgpu_vcn_resume(adev);
95 	if (r)
96 		return r;
97 
98 	ring = &adev->vcn.ring_dec;
99 	sprintf(ring->name, "vcn_dec");
100 	r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.irq, 0);
101 	if (r)
102 		return r;
103 
104 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
105 		ring = &adev->vcn.ring_enc[i];
106 		sprintf(ring->name, "vcn_enc%d", i);
107 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.irq, 0);
108 		if (r)
109 			return r;
110 	}
111 
112 	return r;
113 }
114 
115 /**
116  * vcn_v1_0_sw_fini - sw fini for VCN block
117  *
118  * @handle: amdgpu_device pointer
119  *
120  * VCN suspend and free up sw allocation
121  */
122 static int vcn_v1_0_sw_fini(void *handle)
123 {
124 	int r;
125 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
126 
127 	r = amdgpu_vcn_suspend(adev);
128 	if (r)
129 		return r;
130 
131 	r = amdgpu_vcn_sw_fini(adev);
132 
133 	return r;
134 }
135 
136 /**
137  * vcn_v1_0_hw_init - start and test VCN block
138  *
139  * @handle: amdgpu_device pointer
140  *
141  * Initialize the hardware, boot up the VCPU and do some testing
142  */
143 static int vcn_v1_0_hw_init(void *handle)
144 {
145 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
146 	struct amdgpu_ring *ring = &adev->vcn.ring_dec;
147 	int i, r;
148 
149 	r = vcn_v1_0_start(adev);
150 	if (r)
151 		goto done;
152 
153 	ring->ready = true;
154 	r = amdgpu_ring_test_ring(ring);
155 	if (r) {
156 		ring->ready = false;
157 		goto done;
158 	}
159 
160 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
161 		ring = &adev->vcn.ring_enc[i];
162 		ring->ready = true;
163 		r = amdgpu_ring_test_ring(ring);
164 		if (r) {
165 			ring->ready = false;
166 			goto done;
167 		}
168 	}
169 
170 done:
171 	if (!r)
172 		DRM_INFO("VCN decode and encode initialized successfully.\n");
173 
174 	return r;
175 }
176 
177 /**
178  * vcn_v1_0_hw_fini - stop the hardware block
179  *
180  * @handle: amdgpu_device pointer
181  *
182  * Stop the VCN block, mark ring as not ready any more
183  */
184 static int vcn_v1_0_hw_fini(void *handle)
185 {
186 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
187 	struct amdgpu_ring *ring = &adev->vcn.ring_dec;
188 	int r;
189 
190 	r = vcn_v1_0_stop(adev);
191 	if (r)
192 		return r;
193 
194 	ring->ready = false;
195 
196 	return 0;
197 }
198 
199 /**
200  * vcn_v1_0_suspend - suspend VCN block
201  *
202  * @handle: amdgpu_device pointer
203  *
204  * HW fini and suspend VCN block
205  */
206 static int vcn_v1_0_suspend(void *handle)
207 {
208 	int r;
209 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
210 
211 	r = vcn_v1_0_hw_fini(adev);
212 	if (r)
213 		return r;
214 
215 	r = amdgpu_vcn_suspend(adev);
216 
217 	return r;
218 }
219 
220 /**
221  * vcn_v1_0_resume - resume VCN block
222  *
223  * @handle: amdgpu_device pointer
224  *
225  * Resume firmware and hw init VCN block
226  */
227 static int vcn_v1_0_resume(void *handle)
228 {
229 	int r;
230 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
231 
232 	r = amdgpu_vcn_resume(adev);
233 	if (r)
234 		return r;
235 
236 	r = vcn_v1_0_hw_init(adev);
237 
238 	return r;
239 }
240 
241 /**
242  * vcn_v1_0_mc_resume - memory controller programming
243  *
244  * @adev: amdgpu_device pointer
245  *
246  * Let the VCN memory controller know it's offsets
247  */
248 static void vcn_v1_0_mc_resume(struct amdgpu_device *adev)
249 {
250 	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
251 
252 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
253 			lower_32_bits(adev->vcn.gpu_addr));
254 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
255 			upper_32_bits(adev->vcn.gpu_addr));
256 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0,
257 				AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
258 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size);
259 
260 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
261 			lower_32_bits(adev->vcn.gpu_addr + size));
262 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
263 			upper_32_bits(adev->vcn.gpu_addr + size));
264 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0);
265 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_HEAP_SIZE);
266 
267 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
268 			lower_32_bits(adev->vcn.gpu_addr + size + AMDGPU_VCN_HEAP_SIZE));
269 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
270 			upper_32_bits(adev->vcn.gpu_addr + size + AMDGPU_VCN_HEAP_SIZE));
271 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0);
272 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2,
273 			AMDGPU_VCN_STACK_SIZE + (AMDGPU_VCN_SESSION_SIZE * 40));
274 
275 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_ADDR_CONFIG,
276 			adev->gfx.config.gb_addr_config);
277 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG,
278 			adev->gfx.config.gb_addr_config);
279 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG,
280 			adev->gfx.config.gb_addr_config);
281 }
282 
283 /**
284  * vcn_v1_0_disable_clock_gating - disable VCN clock gating
285  *
286  * @adev: amdgpu_device pointer
287  * @sw: enable SW clock gating
288  *
289  * Disable clock gating for VCN block
290  */
291 static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev, bool sw)
292 {
293 	uint32_t data;
294 
295 	/* JPEG disable CGC */
296 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL);
297 
298 	if (sw)
299 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
300 	else
301 		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE_MASK;
302 
303 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
304 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
305 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data);
306 
307 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE);
308 	data &= ~(JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK);
309 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data);
310 
311 	/* UVD disable CGC */
312 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
313 	if (sw)
314 		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
315 	else
316 		data &= ~ UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
317 
318 	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
319 	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
320 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
321 
322 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_GATE);
323 	data &= ~(UVD_CGC_GATE__SYS_MASK
324 		| UVD_CGC_GATE__UDEC_MASK
325 		| UVD_CGC_GATE__MPEG2_MASK
326 		| UVD_CGC_GATE__REGS_MASK
327 		| UVD_CGC_GATE__RBC_MASK
328 		| UVD_CGC_GATE__LMI_MC_MASK
329 		| UVD_CGC_GATE__LMI_UMC_MASK
330 		| UVD_CGC_GATE__IDCT_MASK
331 		| UVD_CGC_GATE__MPRD_MASK
332 		| UVD_CGC_GATE__MPC_MASK
333 		| UVD_CGC_GATE__LBSI_MASK
334 		| UVD_CGC_GATE__LRBBM_MASK
335 		| UVD_CGC_GATE__UDEC_RE_MASK
336 		| UVD_CGC_GATE__UDEC_CM_MASK
337 		| UVD_CGC_GATE__UDEC_IT_MASK
338 		| UVD_CGC_GATE__UDEC_DB_MASK
339 		| UVD_CGC_GATE__UDEC_MP_MASK
340 		| UVD_CGC_GATE__WCB_MASK
341 		| UVD_CGC_GATE__VCPU_MASK
342 		| UVD_CGC_GATE__SCPU_MASK);
343 	WREG32_SOC15(VCN, 0, mmUVD_CGC_GATE, data);
344 
345 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
346 	data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK
347 		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
348 		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
349 		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
350 		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
351 		| UVD_CGC_CTRL__SYS_MODE_MASK
352 		| UVD_CGC_CTRL__UDEC_MODE_MASK
353 		| UVD_CGC_CTRL__MPEG2_MODE_MASK
354 		| UVD_CGC_CTRL__REGS_MODE_MASK
355 		| UVD_CGC_CTRL__RBC_MODE_MASK
356 		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
357 		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
358 		| UVD_CGC_CTRL__IDCT_MODE_MASK
359 		| UVD_CGC_CTRL__MPRD_MODE_MASK
360 		| UVD_CGC_CTRL__MPC_MODE_MASK
361 		| UVD_CGC_CTRL__LBSI_MODE_MASK
362 		| UVD_CGC_CTRL__LRBBM_MODE_MASK
363 		| UVD_CGC_CTRL__WCB_MODE_MASK
364 		| UVD_CGC_CTRL__VCPU_MODE_MASK
365 		| UVD_CGC_CTRL__SCPU_MODE_MASK);
366 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
367 
368 	/* turn on */
369 	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE);
370 	data |= (UVD_SUVD_CGC_GATE__SRE_MASK
371 		| UVD_SUVD_CGC_GATE__SIT_MASK
372 		| UVD_SUVD_CGC_GATE__SMP_MASK
373 		| UVD_SUVD_CGC_GATE__SCM_MASK
374 		| UVD_SUVD_CGC_GATE__SDB_MASK
375 		| UVD_SUVD_CGC_GATE__SRE_H264_MASK
376 		| UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
377 		| UVD_SUVD_CGC_GATE__SIT_H264_MASK
378 		| UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
379 		| UVD_SUVD_CGC_GATE__SCM_H264_MASK
380 		| UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
381 		| UVD_SUVD_CGC_GATE__SDB_H264_MASK
382 		| UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
383 		| UVD_SUVD_CGC_GATE__SCLR_MASK
384 		| UVD_SUVD_CGC_GATE__UVD_SC_MASK
385 		| UVD_SUVD_CGC_GATE__ENT_MASK
386 		| UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
387 		| UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK
388 		| UVD_SUVD_CGC_GATE__SITE_MASK
389 		| UVD_SUVD_CGC_GATE__SRE_VP9_MASK
390 		| UVD_SUVD_CGC_GATE__SCM_VP9_MASK
391 		| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
392 		| UVD_SUVD_CGC_GATE__SDB_VP9_MASK
393 		| UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
394 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE, data);
395 
396 	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL);
397 	data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
398 		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
399 		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
400 		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
401 		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
402 		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
403 		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
404 		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
405 		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
406 		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
407 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
408 }
409 
410 /**
411  * vcn_v1_0_enable_clock_gating - enable VCN clock gating
412  *
413  * @adev: amdgpu_device pointer
414  * @sw: enable SW clock gating
415  *
416  * Enable clock gating for VCN block
417  */
418 static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev, bool sw)
419 {
420 	uint32_t data = 0;
421 
422 	/* enable JPEG CGC */
423 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL);
424 	if (sw)
425 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
426 	else
427 		data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
428 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
429 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
430 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data);
431 
432 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE);
433 	data |= (JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK);
434 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data);
435 
436 	/* enable UVD CGC */
437 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
438 	if (sw)
439 		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
440 	else
441 		data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
442 	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
443 	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
444 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
445 
446 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
447 	data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK
448 		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
449 		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
450 		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
451 		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
452 		| UVD_CGC_CTRL__SYS_MODE_MASK
453 		| UVD_CGC_CTRL__UDEC_MODE_MASK
454 		| UVD_CGC_CTRL__MPEG2_MODE_MASK
455 		| UVD_CGC_CTRL__REGS_MODE_MASK
456 		| UVD_CGC_CTRL__RBC_MODE_MASK
457 		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
458 		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
459 		| UVD_CGC_CTRL__IDCT_MODE_MASK
460 		| UVD_CGC_CTRL__MPRD_MODE_MASK
461 		| UVD_CGC_CTRL__MPC_MODE_MASK
462 		| UVD_CGC_CTRL__LBSI_MODE_MASK
463 		| UVD_CGC_CTRL__LRBBM_MODE_MASK
464 		| UVD_CGC_CTRL__WCB_MODE_MASK
465 		| UVD_CGC_CTRL__VCPU_MODE_MASK
466 		| UVD_CGC_CTRL__SCPU_MODE_MASK);
467 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
468 
469 	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL);
470 	data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
471 		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
472 		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
473 		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
474 		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
475 		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
476 		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
477 		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
478 		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
479 		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
480 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
481 }
482 
483 /**
484  * vcn_v1_0_start - start VCN block
485  *
486  * @adev: amdgpu_device pointer
487  *
488  * Setup and start the VCN block
489  */
490 static int vcn_v1_0_start(struct amdgpu_device *adev)
491 {
492 	struct amdgpu_ring *ring = &adev->vcn.ring_dec;
493 	uint32_t rb_bufsz, tmp;
494 	uint32_t lmi_swap_cntl;
495 	int i, j, r;
496 
497 	/* disable byte swapping */
498 	lmi_swap_cntl = 0;
499 
500 	vcn_v1_0_mc_resume(adev);
501 
502 	/* disable clock gating */
503 	vcn_v1_0_disable_clock_gating(adev, true);
504 
505 	/* disable interupt */
506 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0,
507 			~UVD_MASTINT_EN__VCPU_EN_MASK);
508 
509 	/* stall UMC and register bus before resetting VCPU */
510 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2),
511 			UVD_LMI_CTRL2__STALL_ARB_UMC_MASK,
512 			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
513 	mdelay(1);
514 
515 	/* put LMI, VCPU, RBC etc... into reset */
516 	WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET,
517 		UVD_SOFT_RESET__LMI_SOFT_RESET_MASK |
518 		UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK |
519 		UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK |
520 		UVD_SOFT_RESET__RBC_SOFT_RESET_MASK |
521 		UVD_SOFT_RESET__CSM_SOFT_RESET_MASK |
522 		UVD_SOFT_RESET__CXW_SOFT_RESET_MASK |
523 		UVD_SOFT_RESET__TAP_SOFT_RESET_MASK |
524 		UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK);
525 	mdelay(5);
526 
527 	/* initialize VCN memory controller */
528 	WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL,
529 		(0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
530 		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
531 		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
532 		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
533 		UVD_LMI_CTRL__REQ_MODE_MASK |
534 		0x00100000L);
535 
536 #ifdef __BIG_ENDIAN
537 	/* swap (8 in 32) RB and IB */
538 	lmi_swap_cntl = 0xa;
539 #endif
540 	WREG32_SOC15(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
541 
542 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0, 0x40c2040);
543 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA1, 0x0);
544 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0, 0x40c2040);
545 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB1, 0x0);
546 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_ALU, 0);
547 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX, 0x88);
548 
549 	/* take all subblocks out of reset, except VCPU */
550 	WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET,
551 			UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
552 	mdelay(5);
553 
554 	/* enable VCPU clock */
555 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL,
556 			UVD_VCPU_CNTL__CLK_EN_MASK);
557 
558 	/* enable UMC */
559 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0,
560 			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
561 
562 	/* boot up the VCPU */
563 	WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, 0);
564 	mdelay(10);
565 
566 	for (i = 0; i < 10; ++i) {
567 		uint32_t status;
568 
569 		for (j = 0; j < 100; ++j) {
570 			status = RREG32_SOC15(UVD, 0, mmUVD_STATUS);
571 			if (status & 2)
572 				break;
573 			mdelay(10);
574 		}
575 		r = 0;
576 		if (status & 2)
577 			break;
578 
579 		DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n");
580 		WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
581 				UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
582 				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
583 		mdelay(10);
584 		WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0,
585 				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
586 		mdelay(10);
587 		r = -1;
588 	}
589 
590 	if (r) {
591 		DRM_ERROR("VCN decode not responding, giving up!!!\n");
592 		return r;
593 	}
594 	/* enable master interrupt */
595 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN),
596 		(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK),
597 		~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK));
598 
599 	/* clear the bit 4 of VCN_STATUS */
600 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0,
601 			~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
602 
603 	/* force RBC into idle state */
604 	rb_bufsz = order_base_2(ring->ring_size);
605 	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
606 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
607 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
608 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0);
609 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
610 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
611 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp);
612 
613 	/* set the write pointer delay */
614 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0);
615 
616 	/* set the wb address */
617 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR,
618 			(upper_32_bits(ring->gpu_addr) >> 2));
619 
620 	/* programm the RB_BASE for ring buffer */
621 	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
622 			lower_32_bits(ring->gpu_addr));
623 	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
624 			upper_32_bits(ring->gpu_addr));
625 
626 	/* Initialize the ring buffer's read and write pointers */
627 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0);
628 
629 	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
630 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
631 			lower_32_bits(ring->wptr));
632 
633 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0,
634 			~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
635 
636 	ring = &adev->vcn.ring_enc[0];
637 	WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
638 	WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
639 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr);
640 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
641 	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4);
642 
643 	ring = &adev->vcn.ring_enc[1];
644 	WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
645 	WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
646 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr);
647 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
648 	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
649 
650 	return 0;
651 }
652 
653 /**
654  * vcn_v1_0_stop - stop VCN block
655  *
656  * @adev: amdgpu_device pointer
657  *
658  * stop the VCN block
659  */
660 static int vcn_v1_0_stop(struct amdgpu_device *adev)
661 {
662 	/* force RBC into idle state */
663 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, 0x11010101);
664 
665 	/* Stall UMC and register bus before resetting VCPU */
666 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2),
667 			UVD_LMI_CTRL2__STALL_ARB_UMC_MASK,
668 			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
669 	mdelay(1);
670 
671 	/* put VCPU into reset */
672 	WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET,
673 			UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
674 	mdelay(5);
675 
676 	/* disable VCPU clock */
677 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, 0x0);
678 
679 	/* Unstall UMC and register bus */
680 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0,
681 			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
682 
683 	/* enable clock gating */
684 	vcn_v1_0_enable_clock_gating(adev, true);
685 
686 	return 0;
687 }
688 
689 static int vcn_v1_0_set_clockgating_state(void *handle,
690 					  enum amd_clockgating_state state)
691 {
692 	/* needed for driver unload*/
693 	return 0;
694 }
695 
696 /**
697  * vcn_v1_0_dec_ring_get_rptr - get read pointer
698  *
699  * @ring: amdgpu_ring pointer
700  *
701  * Returns the current hardware read pointer
702  */
703 static uint64_t vcn_v1_0_dec_ring_get_rptr(struct amdgpu_ring *ring)
704 {
705 	struct amdgpu_device *adev = ring->adev;
706 
707 	return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
708 }
709 
710 /**
711  * vcn_v1_0_dec_ring_get_wptr - get write pointer
712  *
713  * @ring: amdgpu_ring pointer
714  *
715  * Returns the current hardware write pointer
716  */
717 static uint64_t vcn_v1_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
718 {
719 	struct amdgpu_device *adev = ring->adev;
720 
721 	return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR);
722 }
723 
724 /**
725  * vcn_v1_0_dec_ring_set_wptr - set write pointer
726  *
727  * @ring: amdgpu_ring pointer
728  *
729  * Commits the write pointer to the hardware
730  */
731 static void vcn_v1_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
732 {
733 	struct amdgpu_device *adev = ring->adev;
734 
735 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
736 }
737 
738 /**
739  * vcn_v1_0_dec_ring_insert_start - insert a start command
740  *
741  * @ring: amdgpu_ring pointer
742  *
743  * Write a start command to the ring.
744  */
745 static void vcn_v1_0_dec_ring_insert_start(struct amdgpu_ring *ring)
746 {
747 	amdgpu_ring_write(ring,
748 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
749 	amdgpu_ring_write(ring, 0);
750 	amdgpu_ring_write(ring,
751 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
752 	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_START << 1);
753 }
754 
755 /**
756  * vcn_v1_0_dec_ring_insert_end - insert a end command
757  *
758  * @ring: amdgpu_ring pointer
759  *
760  * Write a end command to the ring.
761  */
762 static void vcn_v1_0_dec_ring_insert_end(struct amdgpu_ring *ring)
763 {
764 	amdgpu_ring_write(ring,
765 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
766 	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_END << 1);
767 }
768 
769 /**
770  * vcn_v1_0_dec_ring_emit_fence - emit an fence & trap command
771  *
772  * @ring: amdgpu_ring pointer
773  * @fence: fence to emit
774  *
775  * Write a fence and a trap command to the ring.
776  */
777 static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
778 				     unsigned flags)
779 {
780 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
781 
782 	amdgpu_ring_write(ring,
783 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0));
784 	amdgpu_ring_write(ring, seq);
785 	amdgpu_ring_write(ring,
786 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
787 	amdgpu_ring_write(ring, addr & 0xffffffff);
788 	amdgpu_ring_write(ring,
789 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
790 	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
791 	amdgpu_ring_write(ring,
792 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
793 	amdgpu_ring_write(ring, VCN_DEC_CMD_FENCE << 1);
794 
795 	amdgpu_ring_write(ring,
796 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
797 	amdgpu_ring_write(ring, 0);
798 	amdgpu_ring_write(ring,
799 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
800 	amdgpu_ring_write(ring, 0);
801 	amdgpu_ring_write(ring,
802 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
803 	amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1);
804 }
805 
806 /**
807  * vcn_v1_0_dec_ring_hdp_invalidate - emit an hdp invalidate
808  *
809  * @ring: amdgpu_ring pointer
810  *
811  * Emits an hdp invalidate.
812  */
813 static void vcn_v1_0_dec_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
814 {
815 	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0));
816 	amdgpu_ring_write(ring, 1);
817 }
818 
819 /**
820  * vcn_v1_0_dec_ring_emit_ib - execute indirect buffer
821  *
822  * @ring: amdgpu_ring pointer
823  * @ib: indirect buffer to execute
824  *
825  * Write ring commands to execute the indirect buffer
826  */
827 static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
828 				  struct amdgpu_ib *ib,
829 				  unsigned vm_id, bool ctx_switch)
830 {
831 	amdgpu_ring_write(ring,
832 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0));
833 	amdgpu_ring_write(ring, vm_id);
834 
835 	amdgpu_ring_write(ring,
836 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0));
837 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
838 	amdgpu_ring_write(ring,
839 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0));
840 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
841 	amdgpu_ring_write(ring,
842 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_IB_SIZE), 0));
843 	amdgpu_ring_write(ring, ib->length_dw);
844 }
845 
846 static void vcn_v1_0_dec_vm_reg_write(struct amdgpu_ring *ring,
847 				uint32_t data0, uint32_t data1)
848 {
849 	amdgpu_ring_write(ring,
850 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
851 	amdgpu_ring_write(ring, data0);
852 	amdgpu_ring_write(ring,
853 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
854 	amdgpu_ring_write(ring, data1);
855 	amdgpu_ring_write(ring,
856 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
857 	amdgpu_ring_write(ring, VCN_DEC_CMD_WRITE_REG << 1);
858 }
859 
860 static void vcn_v1_0_dec_vm_reg_wait(struct amdgpu_ring *ring,
861 				uint32_t data0, uint32_t data1, uint32_t mask)
862 {
863 	amdgpu_ring_write(ring,
864 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
865 	amdgpu_ring_write(ring, data0);
866 	amdgpu_ring_write(ring,
867 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
868 	amdgpu_ring_write(ring, data1);
869 	amdgpu_ring_write(ring,
870 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH8), 0));
871 	amdgpu_ring_write(ring, mask);
872 	amdgpu_ring_write(ring,
873 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
874 	amdgpu_ring_write(ring, VCN_DEC_CMD_REG_READ_COND_WAIT << 1);
875 }
876 
877 static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
878 					unsigned vm_id, uint64_t pd_addr)
879 {
880 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
881 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
882 	uint32_t data0, data1, mask;
883 	unsigned eng = ring->vm_inv_eng;
884 
885 	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
886 	pd_addr |= AMDGPU_PTE_VALID;
887 
888 	data0 = (hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2;
889 	data1 = upper_32_bits(pd_addr);
890 	vcn_v1_0_dec_vm_reg_write(ring, data0, data1);
891 
892 	data0 = (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2;
893 	data1 = lower_32_bits(pd_addr);
894 	vcn_v1_0_dec_vm_reg_write(ring, data0, data1);
895 
896 	data0 = (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2;
897 	data1 = lower_32_bits(pd_addr);
898 	mask = 0xffffffff;
899 	vcn_v1_0_dec_vm_reg_wait(ring, data0, data1, mask);
900 
901 	/* flush TLB */
902 	data0 = (hub->vm_inv_eng0_req + eng) << 2;
903 	data1 = req;
904 	vcn_v1_0_dec_vm_reg_write(ring, data0, data1);
905 
906 	/* wait for flush */
907 	data0 = (hub->vm_inv_eng0_ack + eng) << 2;
908 	data1 = 1 << vm_id;
909 	mask =  1 << vm_id;
910 	vcn_v1_0_dec_vm_reg_wait(ring, data0, data1, mask);
911 }
912 
913 /**
914  * vcn_v1_0_enc_ring_get_rptr - get enc read pointer
915  *
916  * @ring: amdgpu_ring pointer
917  *
918  * Returns the current hardware enc read pointer
919  */
920 static uint64_t vcn_v1_0_enc_ring_get_rptr(struct amdgpu_ring *ring)
921 {
922 	struct amdgpu_device *adev = ring->adev;
923 
924 	if (ring == &adev->vcn.ring_enc[0])
925 		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR);
926 	else
927 		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2);
928 }
929 
930  /**
931  * vcn_v1_0_enc_ring_get_wptr - get enc write pointer
932  *
933  * @ring: amdgpu_ring pointer
934  *
935  * Returns the current hardware enc write pointer
936  */
937 static uint64_t vcn_v1_0_enc_ring_get_wptr(struct amdgpu_ring *ring)
938 {
939 	struct amdgpu_device *adev = ring->adev;
940 
941 	if (ring == &adev->vcn.ring_enc[0])
942 		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR);
943 	else
944 		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2);
945 }
946 
947  /**
948  * vcn_v1_0_enc_ring_set_wptr - set enc write pointer
949  *
950  * @ring: amdgpu_ring pointer
951  *
952  * Commits the enc write pointer to the hardware
953  */
954 static void vcn_v1_0_enc_ring_set_wptr(struct amdgpu_ring *ring)
955 {
956 	struct amdgpu_device *adev = ring->adev;
957 
958 	if (ring == &adev->vcn.ring_enc[0])
959 		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR,
960 			lower_32_bits(ring->wptr));
961 	else
962 		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2,
963 			lower_32_bits(ring->wptr));
964 }
965 
966 /**
967  * vcn_v1_0_enc_ring_emit_fence - emit an enc fence & trap command
968  *
969  * @ring: amdgpu_ring pointer
970  * @fence: fence to emit
971  *
972  * Write enc a fence and a trap command to the ring.
973  */
974 static void vcn_v1_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
975 			u64 seq, unsigned flags)
976 {
977 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
978 
979 	amdgpu_ring_write(ring, VCN_ENC_CMD_FENCE);
980 	amdgpu_ring_write(ring, addr);
981 	amdgpu_ring_write(ring, upper_32_bits(addr));
982 	amdgpu_ring_write(ring, seq);
983 	amdgpu_ring_write(ring, VCN_ENC_CMD_TRAP);
984 }
985 
986 static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring)
987 {
988 	amdgpu_ring_write(ring, VCN_ENC_CMD_END);
989 }
990 
991 /**
992  * vcn_v1_0_enc_ring_emit_ib - enc execute indirect buffer
993  *
994  * @ring: amdgpu_ring pointer
995  * @ib: indirect buffer to execute
996  *
997  * Write enc ring commands to execute the indirect buffer
998  */
999 static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
1000 		struct amdgpu_ib *ib, unsigned int vm_id, bool ctx_switch)
1001 {
1002 	amdgpu_ring_write(ring, VCN_ENC_CMD_IB);
1003 	amdgpu_ring_write(ring, vm_id);
1004 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
1005 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
1006 	amdgpu_ring_write(ring, ib->length_dw);
1007 }
1008 
1009 static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
1010 			 unsigned int vm_id, uint64_t pd_addr)
1011 {
1012 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
1013 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
1014 	unsigned eng = ring->vm_inv_eng;
1015 
1016 	pd_addr = amdgpu_gart_get_vm_pde(ring->adev, pd_addr);
1017 	pd_addr |= AMDGPU_PTE_VALID;
1018 
1019 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
1020 	amdgpu_ring_write(ring,
1021 			  (hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);
1022 	amdgpu_ring_write(ring, upper_32_bits(pd_addr));
1023 
1024 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
1025 	amdgpu_ring_write(ring,
1026 			  (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
1027 	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
1028 
1029 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT);
1030 	amdgpu_ring_write(ring,
1031 			  (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
1032 	amdgpu_ring_write(ring, 0xffffffff);
1033 	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
1034 
1035 	/* flush TLB */
1036 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
1037 	amdgpu_ring_write(ring,	(hub->vm_inv_eng0_req + eng) << 2);
1038 	amdgpu_ring_write(ring, req);
1039 
1040 	/* wait for flush */
1041 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT);
1042 	amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
1043 	amdgpu_ring_write(ring, 1 << vm_id);
1044 	amdgpu_ring_write(ring, 1 << vm_id);
1045 }
1046 
1047 static int vcn_v1_0_set_interrupt_state(struct amdgpu_device *adev,
1048 					struct amdgpu_irq_src *source,
1049 					unsigned type,
1050 					enum amdgpu_interrupt_state state)
1051 {
1052 	return 0;
1053 }
1054 
1055 static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev,
1056 				      struct amdgpu_irq_src *source,
1057 				      struct amdgpu_iv_entry *entry)
1058 {
1059 	DRM_DEBUG("IH: VCN TRAP\n");
1060 
1061 	switch (entry->src_id) {
1062 	case 124:
1063 		amdgpu_fence_process(&adev->vcn.ring_dec);
1064 		break;
1065 	case 119:
1066 		amdgpu_fence_process(&adev->vcn.ring_enc[0]);
1067 		break;
1068 	case 120:
1069 		amdgpu_fence_process(&adev->vcn.ring_enc[1]);
1070 		break;
1071 	default:
1072 		DRM_ERROR("Unhandled interrupt: %d %d\n",
1073 			  entry->src_id, entry->src_data[0]);
1074 		break;
1075 	}
1076 
1077 	return 0;
1078 }
1079 
1080 static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
1081 	.name = "vcn_v1_0",
1082 	.early_init = vcn_v1_0_early_init,
1083 	.late_init = NULL,
1084 	.sw_init = vcn_v1_0_sw_init,
1085 	.sw_fini = vcn_v1_0_sw_fini,
1086 	.hw_init = vcn_v1_0_hw_init,
1087 	.hw_fini = vcn_v1_0_hw_fini,
1088 	.suspend = vcn_v1_0_suspend,
1089 	.resume = vcn_v1_0_resume,
1090 	.is_idle = NULL /* vcn_v1_0_is_idle */,
1091 	.wait_for_idle = NULL /* vcn_v1_0_wait_for_idle */,
1092 	.check_soft_reset = NULL /* vcn_v1_0_check_soft_reset */,
1093 	.pre_soft_reset = NULL /* vcn_v1_0_pre_soft_reset */,
1094 	.soft_reset = NULL /* vcn_v1_0_soft_reset */,
1095 	.post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */,
1096 	.set_clockgating_state = vcn_v1_0_set_clockgating_state,
1097 	.set_powergating_state = NULL /* vcn_v1_0_set_powergating_state */,
1098 };
1099 
1100 static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = {
1101 	.type = AMDGPU_RING_TYPE_VCN_DEC,
1102 	.align_mask = 0xf,
1103 	.nop = PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0),
1104 	.support_64bit_ptrs = false,
1105 	.vmhub = AMDGPU_MMHUB,
1106 	.get_rptr = vcn_v1_0_dec_ring_get_rptr,
1107 	.get_wptr = vcn_v1_0_dec_ring_get_wptr,
1108 	.set_wptr = vcn_v1_0_dec_ring_set_wptr,
1109 	.emit_frame_size =
1110 		2 + /* vcn_v1_0_dec_ring_emit_hdp_invalidate */
1111 		34 + /* vcn_v1_0_dec_ring_emit_vm_flush */
1112 		14 + 14 + /* vcn_v1_0_dec_ring_emit_fence x2 vm fence */
1113 		6,
1114 	.emit_ib_size = 8, /* vcn_v1_0_dec_ring_emit_ib */
1115 	.emit_ib = vcn_v1_0_dec_ring_emit_ib,
1116 	.emit_fence = vcn_v1_0_dec_ring_emit_fence,
1117 	.emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush,
1118 	.emit_hdp_invalidate = vcn_v1_0_dec_ring_emit_hdp_invalidate,
1119 	.test_ring = amdgpu_vcn_dec_ring_test_ring,
1120 	.test_ib = amdgpu_vcn_dec_ring_test_ib,
1121 	.insert_nop = amdgpu_ring_insert_nop,
1122 	.insert_start = vcn_v1_0_dec_ring_insert_start,
1123 	.insert_end = vcn_v1_0_dec_ring_insert_end,
1124 	.pad_ib = amdgpu_ring_generic_pad_ib,
1125 	.begin_use = amdgpu_vcn_ring_begin_use,
1126 	.end_use = amdgpu_vcn_ring_end_use,
1127 };
1128 
1129 static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = {
1130 	.type = AMDGPU_RING_TYPE_VCN_ENC,
1131 	.align_mask = 0x3f,
1132 	.nop = VCN_ENC_CMD_NO_OP,
1133 	.support_64bit_ptrs = false,
1134 	.vmhub = AMDGPU_MMHUB,
1135 	.get_rptr = vcn_v1_0_enc_ring_get_rptr,
1136 	.get_wptr = vcn_v1_0_enc_ring_get_wptr,
1137 	.set_wptr = vcn_v1_0_enc_ring_set_wptr,
1138 	.emit_frame_size =
1139 		17 + /* vcn_v1_0_enc_ring_emit_vm_flush */
1140 		5 + 5 + /* vcn_v1_0_enc_ring_emit_fence x2 vm fence */
1141 		1, /* vcn_v1_0_enc_ring_insert_end */
1142 	.emit_ib_size = 5, /* vcn_v1_0_enc_ring_emit_ib */
1143 	.emit_ib = vcn_v1_0_enc_ring_emit_ib,
1144 	.emit_fence = vcn_v1_0_enc_ring_emit_fence,
1145 	.emit_vm_flush = vcn_v1_0_enc_ring_emit_vm_flush,
1146 	.test_ring = amdgpu_vcn_enc_ring_test_ring,
1147 	.test_ib = amdgpu_vcn_enc_ring_test_ib,
1148 	.insert_nop = amdgpu_ring_insert_nop,
1149 	.insert_end = vcn_v1_0_enc_ring_insert_end,
1150 	.pad_ib = amdgpu_ring_generic_pad_ib,
1151 	.begin_use = amdgpu_vcn_ring_begin_use,
1152 	.end_use = amdgpu_vcn_ring_end_use,
1153 };
1154 
1155 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev)
1156 {
1157 	adev->vcn.ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs;
1158 	DRM_INFO("VCN decode is enabled in VM mode\n");
1159 }
1160 
1161 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev)
1162 {
1163 	int i;
1164 
1165 	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
1166 		adev->vcn.ring_enc[i].funcs = &vcn_v1_0_enc_ring_vm_funcs;
1167 
1168 	DRM_INFO("VCN encode is enabled in VM mode\n");
1169 }
1170 
1171 static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = {
1172 	.set = vcn_v1_0_set_interrupt_state,
1173 	.process = vcn_v1_0_process_interrupt,
1174 };
1175 
1176 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev)
1177 {
1178 	adev->uvd.irq.num_types = adev->vcn.num_enc_rings + 1;
1179 	adev->vcn.irq.funcs = &vcn_v1_0_irq_funcs;
1180 }
1181 
1182 const struct amdgpu_ip_block_version vcn_v1_0_ip_block =
1183 {
1184 		.type = AMD_IP_BLOCK_TYPE_VCN,
1185 		.major = 1,
1186 		.minor = 0,
1187 		.rev = 0,
1188 		.funcs = &vcn_v1_0_ip_funcs,
1189 };
1190