1 /*
2  * Copyright 2019 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 "amdgpu.h"
25 #include "amdgpu_jpeg.h"
26 #include "amdgpu_pm.h"
27 #include "soc15.h"
28 #include "soc15d.h"
29 
30 #include "vcn/vcn_2_0_0_offset.h"
31 #include "vcn/vcn_2_0_0_sh_mask.h"
32 #include "ivsrcid/vcn/irqsrcs_vcn_2_0.h"
33 
34 #define mmUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET 			0x1bfff
35 #define mmUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET				0x4029
36 #define mmUVD_JPEG_GPCOM_DATA0_INTERNAL_OFFSET				0x402a
37 #define mmUVD_JPEG_GPCOM_DATA1_INTERNAL_OFFSET				0x402b
38 #define mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW_INTERNAL_OFFSET		0x40ea
39 #define mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH_INTERNAL_OFFSET 	0x40eb
40 #define mmUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET				0x40cf
41 #define mmUVD_LMI_JPEG_VMID_INTERNAL_OFFSET				0x40d1
42 #define mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET 		0x40e8
43 #define mmUVD_LMI_JRBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET		0x40e9
44 #define mmUVD_JRBC_IB_SIZE_INTERNAL_OFFSET				0x4082
45 #define mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW_INTERNAL_OFFSET		0x40ec
46 #define mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH_INTERNAL_OFFSET 	0x40ed
47 #define mmUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET			0x4085
48 #define mmUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET				0x4084
49 #define mmUVD_JRBC_STATUS_INTERNAL_OFFSET				0x4089
50 #define mmUVD_JPEG_PITCH_INTERNAL_OFFSET				0x401f
51 
52 #define JRBC_DEC_EXTERNAL_REG_WRITE_ADDR				0x18000
53 
54 static void jpeg_v2_0_set_dec_ring_funcs(struct amdgpu_device *adev);
55 static void jpeg_v2_0_set_irq_funcs(struct amdgpu_device *adev);
56 static int jpeg_v2_0_set_powergating_state(void *handle,
57 				enum amd_powergating_state state);
58 
59 /**
60  * jpeg_v2_0_early_init - set function pointers
61  *
62  * @handle: amdgpu_device pointer
63  *
64  * Set ring and irq function pointers
65  */
66 static int jpeg_v2_0_early_init(void *handle)
67 {
68 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
69 
70 	adev->jpeg.num_jpeg_inst = 1;
71 
72 	jpeg_v2_0_set_dec_ring_funcs(adev);
73 	jpeg_v2_0_set_irq_funcs(adev);
74 
75 	return 0;
76 }
77 
78 /**
79  * jpeg_v2_0_sw_init - sw init for JPEG block
80  *
81  * @handle: amdgpu_device pointer
82  *
83  * Load firmware and sw initialization
84  */
85 static int jpeg_v2_0_sw_init(void *handle)
86 {
87 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
88 	struct amdgpu_ring *ring;
89 	int r;
90 
91 	/* JPEG TRAP */
92 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
93 		VCN_2_0__SRCID__JPEG_DECODE, &adev->jpeg.inst->irq);
94 	if (r)
95 		return r;
96 
97 	r = amdgpu_jpeg_sw_init(adev);
98 	if (r)
99 		return r;
100 
101 	r = amdgpu_jpeg_resume(adev);
102 	if (r)
103 		return r;
104 
105 	ring = &adev->jpeg.inst->ring_dec;
106 	ring->use_doorbell = true;
107 	ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
108 	sprintf(ring->name, "jpeg_dec");
109 	r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0);
110 	if (r)
111 		return r;
112 
113 	adev->jpeg.internal.jpeg_pitch = mmUVD_JPEG_PITCH_INTERNAL_OFFSET;
114 	adev->jpeg.inst->external.jpeg_pitch = SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_PITCH);
115 
116 	return 0;
117 }
118 
119 /**
120  * jpeg_v2_0_sw_fini - sw fini for JPEG block
121  *
122  * @handle: amdgpu_device pointer
123  *
124  * JPEG suspend and free up sw allocation
125  */
126 static int jpeg_v2_0_sw_fini(void *handle)
127 {
128 	int r;
129 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
130 
131 	r = amdgpu_jpeg_suspend(adev);
132 	if (r)
133 		return r;
134 
135 	r = amdgpu_jpeg_sw_fini(adev);
136 
137 	return r;
138 }
139 
140 /**
141  * jpeg_v2_0_hw_init - start and test JPEG block
142  *
143  * @handle: amdgpu_device pointer
144  *
145  */
146 static int jpeg_v2_0_hw_init(void *handle)
147 {
148 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
149 	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
150 	int r;
151 
152 	adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell,
153 		(adev->doorbell_index.vcn.vcn_ring0_1 << 1), 0);
154 
155 	r = amdgpu_ring_test_helper(ring);
156 	if (!r)
157 		DRM_INFO("JPEG decode initialized successfully.\n");
158 
159 	return r;
160 }
161 
162 /**
163  * jpeg_v2_0_hw_fini - stop the hardware block
164  *
165  * @handle: amdgpu_device pointer
166  *
167  * Stop the JPEG block, mark ring as not ready any more
168  */
169 static int jpeg_v2_0_hw_fini(void *handle)
170 {
171 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
172 	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
173 
174 	if (adev->jpeg.cur_state != AMD_PG_STATE_GATE &&
175 	      RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS))
176 		jpeg_v2_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
177 
178 	ring->sched.ready = false;
179 
180 	return 0;
181 }
182 
183 /**
184  * jpeg_v2_0_suspend - suspend JPEG block
185  *
186  * @handle: amdgpu_device pointer
187  *
188  * HW fini and suspend JPEG block
189  */
190 static int jpeg_v2_0_suspend(void *handle)
191 {
192 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
193 	int r;
194 
195 	r = jpeg_v2_0_hw_fini(adev);
196 	if (r)
197 		return r;
198 
199 	r = amdgpu_jpeg_suspend(adev);
200 
201 	return r;
202 }
203 
204 /**
205  * jpeg_v2_0_resume - resume JPEG block
206  *
207  * @handle: amdgpu_device pointer
208  *
209  * Resume firmware and hw init JPEG block
210  */
211 static int jpeg_v2_0_resume(void *handle)
212 {
213 	int r;
214 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
215 
216 	r = amdgpu_jpeg_resume(adev);
217 	if (r)
218 		return r;
219 
220 	r = jpeg_v2_0_hw_init(adev);
221 
222 	return r;
223 }
224 
225 static int jpeg_v2_0_disable_power_gating(struct amdgpu_device *adev)
226 {
227 	uint32_t data;
228 	int r = 0;
229 
230 	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) {
231 		data = 1 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT;
232 		WREG32(SOC15_REG_OFFSET(JPEG, 0, mmUVD_PGFSM_CONFIG), data);
233 
234 		SOC15_WAIT_ON_RREG(JPEG, 0,
235 			mmUVD_PGFSM_STATUS, UVD_PGFSM_STATUS_UVDJ_PWR_ON,
236 			UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK, r);
237 
238 		if (r) {
239 			DRM_ERROR("amdgpu: JPEG disable power gating failed\n");
240 			return r;
241 		}
242 	}
243 
244 	/* Removing the anti hang mechanism to indicate the UVDJ tile is ON */
245 	data = RREG32(SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_POWER_STATUS)) & ~0x1;
246 	WREG32(SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_POWER_STATUS), data);
247 
248 	return 0;
249 }
250 
251 static int jpeg_v2_0_enable_power_gating(struct amdgpu_device* adev)
252 {
253 	if (adev->pg_flags & AMD_PG_SUPPORT_JPEG) {
254 		uint32_t data;
255 		int r = 0;
256 
257 		data = RREG32(SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_POWER_STATUS));
258 		data &= ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK;
259 		data |=  0x1; //UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_TILES_OFF;
260 		WREG32(SOC15_REG_OFFSET(JPEG, 0, mmUVD_JPEG_POWER_STATUS), data);
261 
262 		data = 2 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT;
263 		WREG32(SOC15_REG_OFFSET(JPEG, 0, mmUVD_PGFSM_CONFIG), data);
264 
265 		SOC15_WAIT_ON_RREG(JPEG, 0, mmUVD_PGFSM_STATUS,
266 			(2 << UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT),
267 			UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK, r);
268 
269 		if (r) {
270 			DRM_ERROR("amdgpu: JPEG enable power gating failed\n");
271 			return r;
272 		}
273 	}
274 
275 	return 0;
276 }
277 
278 static void jpeg_v2_0_disable_clock_gating(struct amdgpu_device* adev)
279 {
280 	uint32_t data;
281 
282 	data = RREG32_SOC15(JPEG, 0, mmJPEG_CGC_CTRL);
283 	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG)
284 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
285 	else
286 		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
287 
288 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
289 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
290 	WREG32_SOC15(JPEG, 0, mmJPEG_CGC_CTRL, data);
291 
292 	data = RREG32_SOC15(JPEG, 0, mmJPEG_CGC_GATE);
293 	data &= ~(JPEG_CGC_GATE__JPEG_DEC_MASK
294 		| JPEG_CGC_GATE__JPEG2_DEC_MASK
295 		| JPEG_CGC_GATE__JPEG_ENC_MASK
296 		| JPEG_CGC_GATE__JMCIF_MASK
297 		| JPEG_CGC_GATE__JRBBM_MASK);
298 	WREG32_SOC15(JPEG, 0, mmJPEG_CGC_GATE, data);
299 }
300 
301 static void jpeg_v2_0_enable_clock_gating(struct amdgpu_device* adev)
302 {
303 	uint32_t data;
304 
305 	data = RREG32_SOC15(JPEG, 0, mmJPEG_CGC_CTRL);
306 	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG)
307 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
308 	else
309 		data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
310 
311 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
312 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
313 	WREG32_SOC15(JPEG, 0, mmJPEG_CGC_CTRL, data);
314 
315 	data = RREG32_SOC15(JPEG, 0, mmJPEG_CGC_GATE);
316 	data |= (JPEG_CGC_GATE__JPEG_DEC_MASK
317 		|JPEG_CGC_GATE__JPEG2_DEC_MASK
318 		|JPEG_CGC_GATE__JPEG_ENC_MASK
319 		|JPEG_CGC_GATE__JMCIF_MASK
320 		|JPEG_CGC_GATE__JRBBM_MASK);
321 	WREG32_SOC15(JPEG, 0, mmJPEG_CGC_GATE, data);
322 }
323 
324 /**
325  * jpeg_v2_0_start - start JPEG block
326  *
327  * @adev: amdgpu_device pointer
328  *
329  * Setup and start the JPEG block
330  */
331 static int jpeg_v2_0_start(struct amdgpu_device *adev)
332 {
333 	struct amdgpu_ring *ring = &adev->jpeg.inst->ring_dec;
334 	int r;
335 
336 	if (adev->pm.dpm_enabled)
337 		amdgpu_dpm_enable_jpeg(adev, true);
338 
339 	/* disable power gating */
340 	r = jpeg_v2_0_disable_power_gating(adev);
341 	if (r)
342 		return r;
343 
344 	/* JPEG disable CGC */
345 	jpeg_v2_0_disable_clock_gating(adev);
346 
347 	WREG32_SOC15(JPEG, 0, mmJPEG_DEC_GFX10_ADDR_CONFIG, adev->gfx.config.gb_addr_config);
348 
349 	/* enable JMI channel */
350 	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, mmUVD_JMI_CNTL), 0,
351 		~UVD_JMI_CNTL__SOFT_RESET_MASK);
352 
353 	/* enable System Interrupt for JRBC */
354 	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, mmJPEG_SYS_INT_EN),
355 		JPEG_SYS_INT_EN__DJRBC_MASK,
356 		~JPEG_SYS_INT_EN__DJRBC_MASK);
357 
358 	WREG32_SOC15(JPEG, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
359 	WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_CNTL, (0x00000001L | 0x00000002L));
360 	WREG32_SOC15(JPEG, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
361 		lower_32_bits(ring->gpu_addr));
362 	WREG32_SOC15(JPEG, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
363 		upper_32_bits(ring->gpu_addr));
364 	WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_RPTR, 0);
365 	WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR, 0);
366 	WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_CNTL, 0x00000002L);
367 	WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_SIZE, ring->ring_size / 4);
368 	ring->wptr = RREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR);
369 
370 	return 0;
371 }
372 
373 /**
374  * jpeg_v2_0_stop - stop JPEG block
375  *
376  * @adev: amdgpu_device pointer
377  *
378  * stop the JPEG block
379  */
380 static int jpeg_v2_0_stop(struct amdgpu_device *adev)
381 {
382 	int r;
383 
384 	/* reset JMI */
385 	WREG32_P(SOC15_REG_OFFSET(JPEG, 0, mmUVD_JMI_CNTL),
386 		UVD_JMI_CNTL__SOFT_RESET_MASK,
387 		~UVD_JMI_CNTL__SOFT_RESET_MASK);
388 
389 	/* enable JPEG CGC */
390 	jpeg_v2_0_enable_clock_gating(adev);
391 
392 	/* enable power gating */
393 	r = jpeg_v2_0_enable_power_gating(adev);
394 	if (r)
395 		return r;
396 
397 	if (adev->pm.dpm_enabled)
398 		amdgpu_dpm_enable_jpeg(adev, false);
399 
400 	return 0;
401 }
402 
403 /**
404  * jpeg_v2_0_dec_ring_get_rptr - get read pointer
405  *
406  * @ring: amdgpu_ring pointer
407  *
408  * Returns the current hardware read pointer
409  */
410 static uint64_t jpeg_v2_0_dec_ring_get_rptr(struct amdgpu_ring *ring)
411 {
412 	struct amdgpu_device *adev = ring->adev;
413 
414 	return RREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_RPTR);
415 }
416 
417 /**
418  * jpeg_v2_0_dec_ring_get_wptr - get write pointer
419  *
420  * @ring: amdgpu_ring pointer
421  *
422  * Returns the current hardware write pointer
423  */
424 static uint64_t jpeg_v2_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
425 {
426 	struct amdgpu_device *adev = ring->adev;
427 
428 	if (ring->use_doorbell)
429 		return adev->wb.wb[ring->wptr_offs];
430 	else
431 		return RREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR);
432 }
433 
434 /**
435  * jpeg_v2_0_dec_ring_set_wptr - set write pointer
436  *
437  * @ring: amdgpu_ring pointer
438  *
439  * Commits the write pointer to the hardware
440  */
441 static void jpeg_v2_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
442 {
443 	struct amdgpu_device *adev = ring->adev;
444 
445 	if (ring->use_doorbell) {
446 		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);
447 		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
448 	} else {
449 		WREG32_SOC15(JPEG, 0, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr));
450 	}
451 }
452 
453 /**
454  * jpeg_v2_0_dec_ring_insert_start - insert a start command
455  *
456  * @ring: amdgpu_ring pointer
457  *
458  * Write a start command to the ring.
459  */
460 void jpeg_v2_0_dec_ring_insert_start(struct amdgpu_ring *ring)
461 {
462 	amdgpu_ring_write(ring, PACKETJ(mmUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
463 		0, 0, PACKETJ_TYPE0));
464 	amdgpu_ring_write(ring, 0x68e04);
465 
466 	amdgpu_ring_write(ring, PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
467 		0, 0, PACKETJ_TYPE0));
468 	amdgpu_ring_write(ring, 0x80010000);
469 }
470 
471 /**
472  * jpeg_v2_0_dec_ring_insert_end - insert a end command
473  *
474  * @ring: amdgpu_ring pointer
475  *
476  * Write a end command to the ring.
477  */
478 void jpeg_v2_0_dec_ring_insert_end(struct amdgpu_ring *ring)
479 {
480 	amdgpu_ring_write(ring, PACKETJ(mmUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
481 		0, 0, PACKETJ_TYPE0));
482 	amdgpu_ring_write(ring, 0x68e04);
483 
484 	amdgpu_ring_write(ring, PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
485 		0, 0, PACKETJ_TYPE0));
486 	amdgpu_ring_write(ring, 0x00010000);
487 }
488 
489 /**
490  * jpeg_v2_0_dec_ring_emit_fence - emit an fence & trap command
491  *
492  * @ring: amdgpu_ring pointer
493  * @fence: fence to emit
494  *
495  * Write a fence and a trap command to the ring.
496  */
497 void jpeg_v2_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
498 				unsigned flags)
499 {
500 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
501 
502 	amdgpu_ring_write(ring, PACKETJ(mmUVD_JPEG_GPCOM_DATA0_INTERNAL_OFFSET,
503 		0, 0, PACKETJ_TYPE0));
504 	amdgpu_ring_write(ring, seq);
505 
506 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JPEG_GPCOM_DATA1_INTERNAL_OFFSET,
507 		0, 0, PACKETJ_TYPE0));
508 	amdgpu_ring_write(ring, seq);
509 
510 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW_INTERNAL_OFFSET,
511 		0, 0, PACKETJ_TYPE0));
512 	amdgpu_ring_write(ring, lower_32_bits(addr));
513 
514 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH_INTERNAL_OFFSET,
515 		0, 0, PACKETJ_TYPE0));
516 	amdgpu_ring_write(ring, upper_32_bits(addr));
517 
518 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET,
519 		0, 0, PACKETJ_TYPE0));
520 	amdgpu_ring_write(ring, 0x8);
521 
522 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET,
523 		0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE4));
524 	amdgpu_ring_write(ring, 0);
525 
526 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
527 		0, 0, PACKETJ_TYPE0));
528 	amdgpu_ring_write(ring, 0x3fbc);
529 
530 	amdgpu_ring_write(ring, PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
531 		0, 0, PACKETJ_TYPE0));
532 	amdgpu_ring_write(ring, 0x1);
533 
534 	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE7));
535 	amdgpu_ring_write(ring, 0);
536 }
537 
538 /**
539  * jpeg_v2_0_dec_ring_emit_ib - execute indirect buffer
540  *
541  * @ring: amdgpu_ring pointer
542  * @ib: indirect buffer to execute
543  *
544  * Write ring commands to execute the indirect buffer.
545  */
546 void jpeg_v2_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
547 				struct amdgpu_job *job,
548 				struct amdgpu_ib *ib,
549 				uint32_t flags)
550 {
551 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
552 
553 	amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET,
554 		0, 0, PACKETJ_TYPE0));
555 	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
556 
557 	amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JPEG_VMID_INTERNAL_OFFSET,
558 		0, 0, PACKETJ_TYPE0));
559 	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
560 
561 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET,
562 		0, 0, PACKETJ_TYPE0));
563 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
564 
565 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET,
566 		0, 0, PACKETJ_TYPE0));
567 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
568 
569 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JRBC_IB_SIZE_INTERNAL_OFFSET,
570 		0, 0, PACKETJ_TYPE0));
571 	amdgpu_ring_write(ring, ib->length_dw);
572 
573 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW_INTERNAL_OFFSET,
574 		0, 0, PACKETJ_TYPE0));
575 	amdgpu_ring_write(ring, lower_32_bits(ring->gpu_addr));
576 
577 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH_INTERNAL_OFFSET,
578 		0, 0, PACKETJ_TYPE0));
579 	amdgpu_ring_write(ring, upper_32_bits(ring->gpu_addr));
580 
581 	amdgpu_ring_write(ring,	PACKETJ(0, 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE2));
582 	amdgpu_ring_write(ring, 0);
583 
584 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET,
585 		0, 0, PACKETJ_TYPE0));
586 	amdgpu_ring_write(ring, 0x01400200);
587 
588 	amdgpu_ring_write(ring, PACKETJ(mmUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET,
589 		0, 0, PACKETJ_TYPE0));
590 	amdgpu_ring_write(ring, 0x2);
591 
592 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JRBC_STATUS_INTERNAL_OFFSET,
593 		0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE3));
594 	amdgpu_ring_write(ring, 0x2);
595 }
596 
597 void jpeg_v2_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
598 				uint32_t val, uint32_t mask)
599 {
600 	uint32_t reg_offset = (reg << 2);
601 
602 	amdgpu_ring_write(ring, PACKETJ(mmUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET,
603 		0, 0, PACKETJ_TYPE0));
604 	amdgpu_ring_write(ring, 0x01400200);
605 
606 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET,
607 		0, 0, PACKETJ_TYPE0));
608 	amdgpu_ring_write(ring, val);
609 
610 	amdgpu_ring_write(ring, PACKETJ(mmUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
611 		0, 0, PACKETJ_TYPE0));
612 	if (reg_offset >= 0x10000 && reg_offset <= 0x105ff) {
613 		amdgpu_ring_write(ring, 0);
614 		amdgpu_ring_write(ring,
615 			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3));
616 	} else {
617 		amdgpu_ring_write(ring, reg_offset);
618 		amdgpu_ring_write(ring,	PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
619 			0, 0, PACKETJ_TYPE3));
620 	}
621 	amdgpu_ring_write(ring, mask);
622 }
623 
624 void jpeg_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
625 				unsigned vmid, uint64_t pd_addr)
626 {
627 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
628 	uint32_t data0, data1, mask;
629 
630 	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
631 
632 	/* wait for register write */
633 	data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2;
634 	data1 = lower_32_bits(pd_addr);
635 	mask = 0xffffffff;
636 	jpeg_v2_0_dec_ring_emit_reg_wait(ring, data0, data1, mask);
637 }
638 
639 void jpeg_v2_0_dec_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val)
640 {
641 	uint32_t reg_offset = (reg << 2);
642 
643 	amdgpu_ring_write(ring,	PACKETJ(mmUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
644 		0, 0, PACKETJ_TYPE0));
645 	if (reg_offset >= 0x10000 && reg_offset <= 0x105ff) {
646 		amdgpu_ring_write(ring, 0);
647 		amdgpu_ring_write(ring,
648 			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0));
649 	} else {
650 		amdgpu_ring_write(ring, reg_offset);
651 		amdgpu_ring_write(ring,	PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
652 			0, 0, PACKETJ_TYPE0));
653 	}
654 	amdgpu_ring_write(ring, val);
655 }
656 
657 void jpeg_v2_0_dec_ring_nop(struct amdgpu_ring *ring, uint32_t count)
658 {
659 	int i;
660 
661 	WARN_ON(ring->wptr % 2 || count % 2);
662 
663 	for (i = 0; i < count / 2; i++) {
664 		amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
665 		amdgpu_ring_write(ring, 0);
666 	}
667 }
668 
669 static bool jpeg_v2_0_is_idle(void *handle)
670 {
671 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
672 
673 	return ((RREG32_SOC15(JPEG, 0, mmUVD_JRBC_STATUS) &
674 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
675 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
676 }
677 
678 static int jpeg_v2_0_wait_for_idle(void *handle)
679 {
680 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
681 	int ret = 0;
682 
683 	SOC15_WAIT_ON_RREG(JPEG, 0, mmUVD_JRBC_STATUS, UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
684 		UVD_JRBC_STATUS__RB_JOB_DONE_MASK, ret);
685 
686 	return ret;
687 }
688 
689 static int jpeg_v2_0_set_clockgating_state(void *handle,
690 					  enum amd_clockgating_state state)
691 {
692 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
693 	bool enable = (state == AMD_CG_STATE_GATE);
694 
695 	if (enable) {
696 		if (jpeg_v2_0_is_idle(handle))
697 			return -EBUSY;
698 		jpeg_v2_0_enable_clock_gating(adev);
699 	} else {
700 		jpeg_v2_0_disable_clock_gating(adev);
701 	}
702 
703 	return 0;
704 }
705 
706 static int jpeg_v2_0_set_powergating_state(void *handle,
707 					enum amd_powergating_state state)
708 {
709 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
710 	int ret;
711 
712 	if (state == adev->jpeg.cur_state)
713 		return 0;
714 
715 	if (state == AMD_PG_STATE_GATE)
716 		ret = jpeg_v2_0_stop(adev);
717 	else
718 		ret = jpeg_v2_0_start(adev);
719 
720 	if (!ret)
721 		adev->jpeg.cur_state = state;
722 
723 	return ret;
724 }
725 
726 static int jpeg_v2_0_set_interrupt_state(struct amdgpu_device *adev,
727 					struct amdgpu_irq_src *source,
728 					unsigned type,
729 					enum amdgpu_interrupt_state state)
730 {
731 	return 0;
732 }
733 
734 static int jpeg_v2_0_process_interrupt(struct amdgpu_device *adev,
735 				      struct amdgpu_irq_src *source,
736 				      struct amdgpu_iv_entry *entry)
737 {
738 	DRM_DEBUG("IH: JPEG TRAP\n");
739 
740 	switch (entry->src_id) {
741 	case VCN_2_0__SRCID__JPEG_DECODE:
742 		amdgpu_fence_process(&adev->jpeg.inst->ring_dec);
743 		break;
744 	default:
745 		DRM_ERROR("Unhandled interrupt: %d %d\n",
746 			  entry->src_id, entry->src_data[0]);
747 		break;
748 	}
749 
750 	return 0;
751 }
752 
753 static const struct amd_ip_funcs jpeg_v2_0_ip_funcs = {
754 	.name = "jpeg_v2_0",
755 	.early_init = jpeg_v2_0_early_init,
756 	.late_init = NULL,
757 	.sw_init = jpeg_v2_0_sw_init,
758 	.sw_fini = jpeg_v2_0_sw_fini,
759 	.hw_init = jpeg_v2_0_hw_init,
760 	.hw_fini = jpeg_v2_0_hw_fini,
761 	.suspend = jpeg_v2_0_suspend,
762 	.resume = jpeg_v2_0_resume,
763 	.is_idle = jpeg_v2_0_is_idle,
764 	.wait_for_idle = jpeg_v2_0_wait_for_idle,
765 	.check_soft_reset = NULL,
766 	.pre_soft_reset = NULL,
767 	.soft_reset = NULL,
768 	.post_soft_reset = NULL,
769 	.set_clockgating_state = jpeg_v2_0_set_clockgating_state,
770 	.set_powergating_state = jpeg_v2_0_set_powergating_state,
771 };
772 
773 static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = {
774 	.type = AMDGPU_RING_TYPE_VCN_JPEG,
775 	.align_mask = 0xf,
776 	.vmhub = AMDGPU_MMHUB_0,
777 	.get_rptr = jpeg_v2_0_dec_ring_get_rptr,
778 	.get_wptr = jpeg_v2_0_dec_ring_get_wptr,
779 	.set_wptr = jpeg_v2_0_dec_ring_set_wptr,
780 	.emit_frame_size =
781 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
782 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
783 		8 + /* jpeg_v2_0_dec_ring_emit_vm_flush */
784 		18 + 18 + /* jpeg_v2_0_dec_ring_emit_fence x2 vm fence */
785 		8 + 16,
786 	.emit_ib_size = 22, /* jpeg_v2_0_dec_ring_emit_ib */
787 	.emit_ib = jpeg_v2_0_dec_ring_emit_ib,
788 	.emit_fence = jpeg_v2_0_dec_ring_emit_fence,
789 	.emit_vm_flush = jpeg_v2_0_dec_ring_emit_vm_flush,
790 	.test_ring = amdgpu_jpeg_dec_ring_test_ring,
791 	.test_ib = amdgpu_jpeg_dec_ring_test_ib,
792 	.insert_nop = jpeg_v2_0_dec_ring_nop,
793 	.insert_start = jpeg_v2_0_dec_ring_insert_start,
794 	.insert_end = jpeg_v2_0_dec_ring_insert_end,
795 	.pad_ib = amdgpu_ring_generic_pad_ib,
796 	.begin_use = amdgpu_jpeg_ring_begin_use,
797 	.end_use = amdgpu_jpeg_ring_end_use,
798 	.emit_wreg = jpeg_v2_0_dec_ring_emit_wreg,
799 	.emit_reg_wait = jpeg_v2_0_dec_ring_emit_reg_wait,
800 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
801 };
802 
803 static void jpeg_v2_0_set_dec_ring_funcs(struct amdgpu_device *adev)
804 {
805 	adev->jpeg.inst->ring_dec.funcs = &jpeg_v2_0_dec_ring_vm_funcs;
806 	DRM_INFO("JPEG decode is enabled in VM mode\n");
807 }
808 
809 static const struct amdgpu_irq_src_funcs jpeg_v2_0_irq_funcs = {
810 	.set = jpeg_v2_0_set_interrupt_state,
811 	.process = jpeg_v2_0_process_interrupt,
812 };
813 
814 static void jpeg_v2_0_set_irq_funcs(struct amdgpu_device *adev)
815 {
816 	adev->jpeg.inst->irq.num_types = 1;
817 	adev->jpeg.inst->irq.funcs = &jpeg_v2_0_irq_funcs;
818 }
819 
820 const struct amdgpu_ip_block_version jpeg_v2_0_ip_block =
821 {
822 		.type = AMD_IP_BLOCK_TYPE_JPEG,
823 		.major = 2,
824 		.minor = 0,
825 		.rev = 0,
826 		.funcs = &jpeg_v2_0_ip_funcs,
827 };
828