Linux Audio

Check our new training course

Loading...
v4.17
   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 "soc15.h"
  29#include "soc15d.h"
  30#include "soc15_common.h"
  31
  32#include "vcn/vcn_1_0_offset.h"
  33#include "vcn/vcn_1_0_sh_mask.h"
  34#include "hdp/hdp_4_0_offset.h"
  35#include "mmhub/mmhub_9_1_offset.h"
  36#include "mmhub/mmhub_9_1_sh_mask.h"
  37
  38static int vcn_v1_0_start(struct amdgpu_device *adev);
 
 
 
 
 
 
  39static int vcn_v1_0_stop(struct amdgpu_device *adev);
  40static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev);
  41static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev);
 
  42static 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 */
  51static 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 */
  71static 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, SOC15_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, SOC15_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 */
 122static 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 */
 143static 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
 
 
 
 
 
 170done:
 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 */
 184static 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 */
 206static 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 */
 227static 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 */
 248static 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 */
 291static 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 */
 418static 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 */
 490static 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 */
 660static 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
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 689static 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 */
 703static 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 */
 717static 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 */
 731static 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 */
 745static void vcn_v1_0_dec_ring_insert_start(struct amdgpu_ring *ring)
 746{
 747	struct amdgpu_device *adev = ring->adev;
 748
 749	amdgpu_ring_write(ring,
 750		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
 751	amdgpu_ring_write(ring, 0);
 752	amdgpu_ring_write(ring,
 753		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
 754	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_START << 1);
 755}
 756
 757/**
 758 * vcn_v1_0_dec_ring_insert_end - insert a end command
 759 *
 760 * @ring: amdgpu_ring pointer
 761 *
 762 * Write a end command to the ring.
 763 */
 764static void vcn_v1_0_dec_ring_insert_end(struct amdgpu_ring *ring)
 765{
 766	struct amdgpu_device *adev = ring->adev;
 767
 768	amdgpu_ring_write(ring,
 769		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
 770	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_END << 1);
 771}
 772
 773/**
 774 * vcn_v1_0_dec_ring_emit_fence - emit an fence & trap command
 775 *
 776 * @ring: amdgpu_ring pointer
 777 * @fence: fence to emit
 778 *
 779 * Write a fence and a trap command to the ring.
 780 */
 781static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
 782				     unsigned flags)
 783{
 784	struct amdgpu_device *adev = ring->adev;
 785
 786	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
 787
 788	amdgpu_ring_write(ring,
 789		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0));
 790	amdgpu_ring_write(ring, seq);
 791	amdgpu_ring_write(ring,
 792		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
 793	amdgpu_ring_write(ring, addr & 0xffffffff);
 794	amdgpu_ring_write(ring,
 795		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
 796	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
 797	amdgpu_ring_write(ring,
 798		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
 799	amdgpu_ring_write(ring, VCN_DEC_CMD_FENCE << 1);
 800
 801	amdgpu_ring_write(ring,
 802		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
 803	amdgpu_ring_write(ring, 0);
 804	amdgpu_ring_write(ring,
 805		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
 806	amdgpu_ring_write(ring, 0);
 807	amdgpu_ring_write(ring,
 808		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
 809	amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1);
 810}
 811
 812/**
 813 * vcn_v1_0_dec_ring_emit_ib - execute indirect buffer
 814 *
 815 * @ring: amdgpu_ring pointer
 816 * @ib: indirect buffer to execute
 817 *
 818 * Write ring commands to execute the indirect buffer
 819 */
 820static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
 821				  struct amdgpu_ib *ib,
 822				  unsigned vmid, bool ctx_switch)
 
 823{
 824	struct amdgpu_device *adev = ring->adev;
 
 825
 826	amdgpu_ring_write(ring,
 827		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0));
 828	amdgpu_ring_write(ring, vmid);
 829
 830	amdgpu_ring_write(ring,
 831		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0));
 832	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
 833	amdgpu_ring_write(ring,
 834		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0));
 835	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
 836	amdgpu_ring_write(ring,
 837		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_IB_SIZE), 0));
 838	amdgpu_ring_write(ring, ib->length_dw);
 839}
 840
 841static void vcn_v1_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring,
 842					    uint32_t reg, uint32_t val,
 843					    uint32_t mask)
 844{
 845	struct amdgpu_device *adev = ring->adev;
 846
 847	amdgpu_ring_write(ring,
 848		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
 849	amdgpu_ring_write(ring, reg << 2);
 850	amdgpu_ring_write(ring,
 851		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
 852	amdgpu_ring_write(ring, val);
 853	amdgpu_ring_write(ring,
 854		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH8), 0));
 855	amdgpu_ring_write(ring, mask);
 856	amdgpu_ring_write(ring,
 857		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
 858	amdgpu_ring_write(ring, VCN_DEC_CMD_REG_READ_COND_WAIT << 1);
 859}
 860
 861static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
 862					    unsigned vmid, uint64_t pd_addr)
 863{
 864	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 865	uint32_t data0, data1, mask;
 866
 867	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 868
 869	/* wait for register write */
 870	data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2;
 871	data1 = lower_32_bits(pd_addr);
 872	mask = 0xffffffff;
 873	vcn_v1_0_dec_ring_emit_reg_wait(ring, data0, data1, mask);
 874}
 875
 876static void vcn_v1_0_dec_ring_emit_wreg(struct amdgpu_ring *ring,
 877					uint32_t reg, uint32_t val)
 878{
 879	struct amdgpu_device *adev = ring->adev;
 880
 881	amdgpu_ring_write(ring,
 882		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
 883	amdgpu_ring_write(ring, reg << 2);
 884	amdgpu_ring_write(ring,
 885		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
 886	amdgpu_ring_write(ring, val);
 887	amdgpu_ring_write(ring,
 888		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
 889	amdgpu_ring_write(ring, VCN_DEC_CMD_WRITE_REG << 1);
 890}
 891
 892/**
 893 * vcn_v1_0_enc_ring_get_rptr - get enc read pointer
 894 *
 895 * @ring: amdgpu_ring pointer
 896 *
 897 * Returns the current hardware enc read pointer
 898 */
 899static uint64_t vcn_v1_0_enc_ring_get_rptr(struct amdgpu_ring *ring)
 900{
 901	struct amdgpu_device *adev = ring->adev;
 902
 903	if (ring == &adev->vcn.ring_enc[0])
 904		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR);
 905	else
 906		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2);
 907}
 908
 909 /**
 910 * vcn_v1_0_enc_ring_get_wptr - get enc write pointer
 911 *
 912 * @ring: amdgpu_ring pointer
 913 *
 914 * Returns the current hardware enc write pointer
 915 */
 916static uint64_t vcn_v1_0_enc_ring_get_wptr(struct amdgpu_ring *ring)
 917{
 918	struct amdgpu_device *adev = ring->adev;
 919
 920	if (ring == &adev->vcn.ring_enc[0])
 921		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR);
 922	else
 923		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2);
 924}
 925
 926 /**
 927 * vcn_v1_0_enc_ring_set_wptr - set enc write pointer
 928 *
 929 * @ring: amdgpu_ring pointer
 930 *
 931 * Commits the enc write pointer to the hardware
 932 */
 933static void vcn_v1_0_enc_ring_set_wptr(struct amdgpu_ring *ring)
 934{
 935	struct amdgpu_device *adev = ring->adev;
 936
 937	if (ring == &adev->vcn.ring_enc[0])
 938		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR,
 939			lower_32_bits(ring->wptr));
 940	else
 941		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2,
 942			lower_32_bits(ring->wptr));
 943}
 944
 945/**
 946 * vcn_v1_0_enc_ring_emit_fence - emit an enc fence & trap command
 947 *
 948 * @ring: amdgpu_ring pointer
 949 * @fence: fence to emit
 950 *
 951 * Write enc a fence and a trap command to the ring.
 952 */
 953static void vcn_v1_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 954			u64 seq, unsigned flags)
 955{
 956	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
 957
 958	amdgpu_ring_write(ring, VCN_ENC_CMD_FENCE);
 959	amdgpu_ring_write(ring, addr);
 960	amdgpu_ring_write(ring, upper_32_bits(addr));
 961	amdgpu_ring_write(ring, seq);
 962	amdgpu_ring_write(ring, VCN_ENC_CMD_TRAP);
 963}
 964
 965static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring)
 966{
 967	amdgpu_ring_write(ring, VCN_ENC_CMD_END);
 968}
 969
 970/**
 971 * vcn_v1_0_enc_ring_emit_ib - enc execute indirect buffer
 972 *
 973 * @ring: amdgpu_ring pointer
 974 * @ib: indirect buffer to execute
 975 *
 976 * Write enc ring commands to execute the indirect buffer
 977 */
 978static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
 979		struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch)
 
 
 980{
 
 
 981	amdgpu_ring_write(ring, VCN_ENC_CMD_IB);
 982	amdgpu_ring_write(ring, vmid);
 983	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
 984	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
 985	amdgpu_ring_write(ring, ib->length_dw);
 986}
 987
 988static void vcn_v1_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring,
 989					    uint32_t reg, uint32_t val,
 990					    uint32_t mask)
 991{
 992	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT);
 993	amdgpu_ring_write(ring, reg << 2);
 994	amdgpu_ring_write(ring, mask);
 995	amdgpu_ring_write(ring, val);
 996}
 997
 998static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
 999					    unsigned int vmid, uint64_t pd_addr)
1000{
1001	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
1002
1003	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
1004
1005	/* wait for reg writes */
1006	vcn_v1_0_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 + vmid * 2,
1007					lower_32_bits(pd_addr), 0xffffffff);
1008}
1009
1010static void vcn_v1_0_enc_ring_emit_wreg(struct amdgpu_ring *ring,
1011					uint32_t reg, uint32_t val)
1012{
1013	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
1014	amdgpu_ring_write(ring,	reg << 2);
1015	amdgpu_ring_write(ring, val);
1016}
1017
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1018static int vcn_v1_0_set_interrupt_state(struct amdgpu_device *adev,
1019					struct amdgpu_irq_src *source,
1020					unsigned type,
1021					enum amdgpu_interrupt_state state)
1022{
1023	return 0;
1024}
1025
1026static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev,
1027				      struct amdgpu_irq_src *source,
1028				      struct amdgpu_iv_entry *entry)
1029{
1030	DRM_DEBUG("IH: VCN TRAP\n");
1031
1032	switch (entry->src_id) {
1033	case 124:
1034		amdgpu_fence_process(&adev->vcn.ring_dec);
1035		break;
1036	case 119:
1037		amdgpu_fence_process(&adev->vcn.ring_enc[0]);
1038		break;
1039	case 120:
1040		amdgpu_fence_process(&adev->vcn.ring_enc[1]);
 
 
 
1041		break;
1042	default:
1043		DRM_ERROR("Unhandled interrupt: %d %d\n",
1044			  entry->src_id, entry->src_data[0]);
1045		break;
1046	}
1047
1048	return 0;
1049}
1050
1051static void vcn_v1_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
1052{
1053	int i;
1054	struct amdgpu_device *adev = ring->adev;
 
1055
1056	for (i = 0; i < count; i++)
1057		amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0));
1058
 
 
 
 
1059}
1060
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1061
1062static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
1063	.name = "vcn_v1_0",
1064	.early_init = vcn_v1_0_early_init,
1065	.late_init = NULL,
1066	.sw_init = vcn_v1_0_sw_init,
1067	.sw_fini = vcn_v1_0_sw_fini,
1068	.hw_init = vcn_v1_0_hw_init,
1069	.hw_fini = vcn_v1_0_hw_fini,
1070	.suspend = vcn_v1_0_suspend,
1071	.resume = vcn_v1_0_resume,
1072	.is_idle = NULL /* vcn_v1_0_is_idle */,
1073	.wait_for_idle = NULL /* vcn_v1_0_wait_for_idle */,
1074	.check_soft_reset = NULL /* vcn_v1_0_check_soft_reset */,
1075	.pre_soft_reset = NULL /* vcn_v1_0_pre_soft_reset */,
1076	.soft_reset = NULL /* vcn_v1_0_soft_reset */,
1077	.post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */,
1078	.set_clockgating_state = vcn_v1_0_set_clockgating_state,
1079	.set_powergating_state = NULL /* vcn_v1_0_set_powergating_state */,
1080};
1081
1082static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = {
1083	.type = AMDGPU_RING_TYPE_VCN_DEC,
1084	.align_mask = 0xf,
1085	.nop = PACKET0(0x81ff, 0),
1086	.support_64bit_ptrs = false,
1087	.vmhub = AMDGPU_MMHUB,
 
1088	.get_rptr = vcn_v1_0_dec_ring_get_rptr,
1089	.get_wptr = vcn_v1_0_dec_ring_get_wptr,
1090	.set_wptr = vcn_v1_0_dec_ring_set_wptr,
1091	.emit_frame_size =
1092		6 + 6 + /* hdp invalidate / flush */
1093		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
1094		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
1095		8 + /* vcn_v1_0_dec_ring_emit_vm_flush */
1096		14 + 14 + /* vcn_v1_0_dec_ring_emit_fence x2 vm fence */
1097		6,
1098	.emit_ib_size = 8, /* vcn_v1_0_dec_ring_emit_ib */
1099	.emit_ib = vcn_v1_0_dec_ring_emit_ib,
1100	.emit_fence = vcn_v1_0_dec_ring_emit_fence,
1101	.emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush,
1102	.test_ring = amdgpu_vcn_dec_ring_test_ring,
1103	.test_ib = amdgpu_vcn_dec_ring_test_ib,
1104	.insert_nop = vcn_v1_0_ring_insert_nop,
1105	.insert_start = vcn_v1_0_dec_ring_insert_start,
1106	.insert_end = vcn_v1_0_dec_ring_insert_end,
1107	.pad_ib = amdgpu_ring_generic_pad_ib,
1108	.begin_use = amdgpu_vcn_ring_begin_use,
1109	.end_use = amdgpu_vcn_ring_end_use,
1110	.emit_wreg = vcn_v1_0_dec_ring_emit_wreg,
1111	.emit_reg_wait = vcn_v1_0_dec_ring_emit_reg_wait,
 
1112};
1113
1114static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = {
1115	.type = AMDGPU_RING_TYPE_VCN_ENC,
1116	.align_mask = 0x3f,
1117	.nop = VCN_ENC_CMD_NO_OP,
1118	.support_64bit_ptrs = false,
1119	.vmhub = AMDGPU_MMHUB,
 
1120	.get_rptr = vcn_v1_0_enc_ring_get_rptr,
1121	.get_wptr = vcn_v1_0_enc_ring_get_wptr,
1122	.set_wptr = vcn_v1_0_enc_ring_set_wptr,
1123	.emit_frame_size =
1124		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
1125		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
1126		4 + /* vcn_v1_0_enc_ring_emit_vm_flush */
1127		5 + 5 + /* vcn_v1_0_enc_ring_emit_fence x2 vm fence */
1128		1, /* vcn_v1_0_enc_ring_insert_end */
1129	.emit_ib_size = 5, /* vcn_v1_0_enc_ring_emit_ib */
1130	.emit_ib = vcn_v1_0_enc_ring_emit_ib,
1131	.emit_fence = vcn_v1_0_enc_ring_emit_fence,
1132	.emit_vm_flush = vcn_v1_0_enc_ring_emit_vm_flush,
1133	.test_ring = amdgpu_vcn_enc_ring_test_ring,
1134	.test_ib = amdgpu_vcn_enc_ring_test_ib,
1135	.insert_nop = amdgpu_ring_insert_nop,
1136	.insert_end = vcn_v1_0_enc_ring_insert_end,
1137	.pad_ib = amdgpu_ring_generic_pad_ib,
1138	.begin_use = amdgpu_vcn_ring_begin_use,
1139	.end_use = amdgpu_vcn_ring_end_use,
1140	.emit_wreg = vcn_v1_0_enc_ring_emit_wreg,
1141	.emit_reg_wait = vcn_v1_0_enc_ring_emit_reg_wait,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1142};
1143
1144static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev)
1145{
1146	adev->vcn.ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs;
1147	DRM_INFO("VCN decode is enabled in VM mode\n");
1148}
1149
1150static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev)
1151{
1152	int i;
1153
1154	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
1155		adev->vcn.ring_enc[i].funcs = &vcn_v1_0_enc_ring_vm_funcs;
1156
1157	DRM_INFO("VCN encode is enabled in VM mode\n");
1158}
1159
 
 
 
 
 
 
1160static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = {
1161	.set = vcn_v1_0_set_interrupt_state,
1162	.process = vcn_v1_0_process_interrupt,
1163};
1164
1165static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev)
1166{
1167	adev->vcn.irq.num_types = adev->vcn.num_enc_rings + 1;
1168	adev->vcn.irq.funcs = &vcn_v1_0_irq_funcs;
1169}
1170
1171const struct amdgpu_ip_block_version vcn_v1_0_ip_block =
1172{
1173		.type = AMD_IP_BLOCK_TYPE_VCN,
1174		.major = 1,
1175		.minor = 0,
1176		.rev = 0,
1177		.funcs = &vcn_v1_0_ip_funcs,
1178};
v5.4
   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
  26#include "amdgpu.h"
  27#include "amdgpu_vcn.h"
  28#include "soc15.h"
  29#include "soc15d.h"
  30#include "soc15_common.h"
  31
  32#include "vcn/vcn_1_0_offset.h"
  33#include "vcn/vcn_1_0_sh_mask.h"
  34#include "hdp/hdp_4_0_offset.h"
  35#include "mmhub/mmhub_9_1_offset.h"
  36#include "mmhub/mmhub_9_1_sh_mask.h"
  37
  38#include "ivsrcid/vcn/irqsrcs_vcn_1_0.h"
  39
  40#define mmUVD_RBC_XX_IB_REG_CHECK				0x05ab
  41#define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX	1
  42#define mmUVD_REG_XX_MASK							0x05ac
  43#define mmUVD_REG_XX_MASK_BASE_IDX				1
  44
  45static int vcn_v1_0_stop(struct amdgpu_device *adev);
  46static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev);
  47static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev);
  48static void vcn_v1_0_set_jpeg_ring_funcs(struct amdgpu_device *adev);
  49static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev);
  50static void vcn_v1_0_jpeg_ring_set_patch_ring(struct amdgpu_ring *ring, uint32_t ptr);
  51static int vcn_v1_0_set_powergating_state(void *handle, enum amd_powergating_state state);
  52static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
  53				struct dpg_pause_state *new_state);
  54
  55/**
  56 * vcn_v1_0_early_init - set function pointers
  57 *
  58 * @handle: amdgpu_device pointer
  59 *
  60 * Set ring and irq function pointers
  61 */
  62static int vcn_v1_0_early_init(void *handle)
  63{
  64	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
  65
  66	adev->vcn.num_vcn_inst = 1;
  67	adev->vcn.num_enc_rings = 2;
  68
  69	vcn_v1_0_set_dec_ring_funcs(adev);
  70	vcn_v1_0_set_enc_ring_funcs(adev);
  71	vcn_v1_0_set_jpeg_ring_funcs(adev);
  72	vcn_v1_0_set_irq_funcs(adev);
  73
  74	return 0;
  75}
  76
  77/**
  78 * vcn_v1_0_sw_init - sw init for VCN block
  79 *
  80 * @handle: amdgpu_device pointer
  81 *
  82 * Load firmware and sw initialization
  83 */
  84static int vcn_v1_0_sw_init(void *handle)
  85{
  86	struct amdgpu_ring *ring;
  87	int i, r;
  88	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
  89
  90	/* VCN DEC TRAP */
  91	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
  92			VCN_1_0__SRCID__UVD_SYSTEM_MESSAGE_INTERRUPT, &adev->vcn.inst->irq);
  93	if (r)
  94		return r;
  95
  96	/* VCN ENC TRAP */
  97	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
  98		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, i + VCN_1_0__SRCID__UVD_ENC_GENERAL_PURPOSE,
  99					&adev->vcn.inst->irq);
 100		if (r)
 101			return r;
 102	}
 103
 104	/* VCN JPEG TRAP */
 105	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, 126, &adev->vcn.inst->irq);
 106	if (r)
 107		return r;
 108
 109	r = amdgpu_vcn_sw_init(adev);
 110	if (r)
 111		return r;
 112
 113	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 114		const struct common_firmware_header *hdr;
 115		hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
 116		adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].ucode_id = AMDGPU_UCODE_ID_VCN;
 117		adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw;
 118		adev->firmware.fw_size +=
 119			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
 120		DRM_INFO("PSP loading VCN firmware\n");
 121	}
 122
 123	r = amdgpu_vcn_resume(adev);
 124	if (r)
 125		return r;
 126
 127	ring = &adev->vcn.inst->ring_dec;
 128	sprintf(ring->name, "vcn_dec");
 129	r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0);
 130	if (r)
 131		return r;
 132
 133	adev->vcn.internal.scratch9 = adev->vcn.inst->external.scratch9 =
 134		SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9);
 135	adev->vcn.internal.data0 = adev->vcn.inst->external.data0 =
 136		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0);
 137	adev->vcn.internal.data1 = adev->vcn.inst->external.data1 =
 138		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1);
 139	adev->vcn.internal.cmd = adev->vcn.inst->external.cmd =
 140		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD);
 141	adev->vcn.internal.nop = adev->vcn.inst->external.nop =
 142		SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP);
 143
 144	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
 145		ring = &adev->vcn.inst->ring_enc[i];
 146		sprintf(ring->name, "vcn_enc%d", i);
 147		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0);
 148		if (r)
 149			return r;
 150	}
 151
 152	ring = &adev->vcn.inst->ring_jpeg;
 153	sprintf(ring->name, "vcn_jpeg");
 154	r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0);
 155	if (r)
 156		return r;
 157
 158	adev->vcn.pause_dpg_mode = vcn_v1_0_pause_dpg_mode;
 159	adev->vcn.internal.jpeg_pitch = adev->vcn.inst->external.jpeg_pitch =
 160		SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_PITCH);
 161
 162	return 0;
 163}
 164
 165/**
 166 * vcn_v1_0_sw_fini - sw fini for VCN block
 167 *
 168 * @handle: amdgpu_device pointer
 169 *
 170 * VCN suspend and free up sw allocation
 171 */
 172static int vcn_v1_0_sw_fini(void *handle)
 173{
 174	int r;
 175	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 176
 177	r = amdgpu_vcn_suspend(adev);
 178	if (r)
 179		return r;
 180
 181	r = amdgpu_vcn_sw_fini(adev);
 182
 183	return r;
 184}
 185
 186/**
 187 * vcn_v1_0_hw_init - start and test VCN block
 188 *
 189 * @handle: amdgpu_device pointer
 190 *
 191 * Initialize the hardware, boot up the VCPU and do some testing
 192 */
 193static int vcn_v1_0_hw_init(void *handle)
 194{
 195	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 196	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 197	int i, r;
 198
 199	r = amdgpu_ring_test_helper(ring);
 200	if (r)
 201		goto done;
 202
 
 
 
 
 
 
 
 203	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
 204		ring = &adev->vcn.inst->ring_enc[i];
 205		ring->sched.ready = true;
 206		r = amdgpu_ring_test_helper(ring);
 207		if (r)
 
 208			goto done;
 
 209	}
 210
 211	ring = &adev->vcn.inst->ring_jpeg;
 212	r = amdgpu_ring_test_helper(ring);
 213	if (r)
 214		goto done;
 215
 216done:
 217	if (!r)
 218		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
 219			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
 220
 221	return r;
 222}
 223
 224/**
 225 * vcn_v1_0_hw_fini - stop the hardware block
 226 *
 227 * @handle: amdgpu_device pointer
 228 *
 229 * Stop the VCN block, mark ring as not ready any more
 230 */
 231static int vcn_v1_0_hw_fini(void *handle)
 232{
 233	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 234	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 
 235
 236	if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
 237		RREG32_SOC15(VCN, 0, mmUVD_STATUS))
 238		vcn_v1_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
 239
 240	ring->sched.ready = false;
 241
 242	return 0;
 243}
 244
 245/**
 246 * vcn_v1_0_suspend - suspend VCN block
 247 *
 248 * @handle: amdgpu_device pointer
 249 *
 250 * HW fini and suspend VCN block
 251 */
 252static int vcn_v1_0_suspend(void *handle)
 253{
 254	int r;
 255	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 256
 257	r = vcn_v1_0_hw_fini(adev);
 258	if (r)
 259		return r;
 260
 261	r = amdgpu_vcn_suspend(adev);
 262
 263	return r;
 264}
 265
 266/**
 267 * vcn_v1_0_resume - resume VCN block
 268 *
 269 * @handle: amdgpu_device pointer
 270 *
 271 * Resume firmware and hw init VCN block
 272 */
 273static int vcn_v1_0_resume(void *handle)
 274{
 275	int r;
 276	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 277
 278	r = amdgpu_vcn_resume(adev);
 279	if (r)
 280		return r;
 281
 282	r = vcn_v1_0_hw_init(adev);
 283
 284	return r;
 285}
 286
 287/**
 288 * vcn_v1_0_mc_resume_spg_mode - memory controller programming
 289 *
 290 * @adev: amdgpu_device pointer
 291 *
 292 * Let the VCN memory controller know it's offsets
 293 */
 294static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev)
 295{
 296	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
 297	uint32_t offset;
 298
 299	/* cache window 0: fw */
 300	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 301		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
 302			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo));
 303		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
 304			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi));
 305		WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0);
 306		offset = 0;
 307	} else {
 308		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
 309			lower_32_bits(adev->vcn.inst->gpu_addr));
 310		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
 311			upper_32_bits(adev->vcn.inst->gpu_addr));
 312		offset = size;
 313		WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0,
 314			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
 315	}
 316
 
 
 
 
 
 
 317	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size);
 318
 319	/* cache window 1: stack */
 320	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
 321		     lower_32_bits(adev->vcn.inst->gpu_addr + offset));
 322	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
 323		     upper_32_bits(adev->vcn.inst->gpu_addr + offset));
 324	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0);
 325	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
 326
 327	/* cache window 2: context */
 328	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
 329		     lower_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
 330	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
 331		     upper_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
 332	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0);
 333	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
 
 334
 335	WREG32_SOC15(UVD, 0, mmUVD_UDEC_ADDR_CONFIG,
 336			adev->gfx.config.gb_addr_config);
 337	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG,
 338			adev->gfx.config.gb_addr_config);
 339	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG,
 340			adev->gfx.config.gb_addr_config);
 341	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG,
 342			adev->gfx.config.gb_addr_config);
 343	WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG,
 344			adev->gfx.config.gb_addr_config);
 345	WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG,
 346			adev->gfx.config.gb_addr_config);
 347	WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG,
 348			adev->gfx.config.gb_addr_config);
 349	WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG,
 350			adev->gfx.config.gb_addr_config);
 351	WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG,
 352			adev->gfx.config.gb_addr_config);
 353	WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG,
 354			adev->gfx.config.gb_addr_config);
 355	WREG32_SOC15(UVD, 0, mmUVD_JPEG_ADDR_CONFIG,
 356			adev->gfx.config.gb_addr_config);
 357	WREG32_SOC15(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG,
 358			adev->gfx.config.gb_addr_config);
 359}
 360
 361static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev)
 362{
 363	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
 364	uint32_t offset;
 365
 366	/* cache window 0: fw */
 367	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 368		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
 369			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo),
 370			     0xFFFFFFFF, 0);
 371		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
 372			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi),
 373			     0xFFFFFFFF, 0);
 374		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0,
 375			     0xFFFFFFFF, 0);
 376		offset = 0;
 377	} else {
 378		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
 379			lower_32_bits(adev->vcn.inst->gpu_addr), 0xFFFFFFFF, 0);
 380		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
 381			upper_32_bits(adev->vcn.inst->gpu_addr), 0xFFFFFFFF, 0);
 382		offset = size;
 383		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0,
 384			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0xFFFFFFFF, 0);
 385	}
 386
 387	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size, 0xFFFFFFFF, 0);
 388
 389	/* cache window 1: stack */
 390	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
 391		     lower_32_bits(adev->vcn.inst->gpu_addr + offset), 0xFFFFFFFF, 0);
 392	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
 393		     upper_32_bits(adev->vcn.inst->gpu_addr + offset), 0xFFFFFFFF, 0);
 394	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0,
 395			     0xFFFFFFFF, 0);
 396	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE,
 397			     0xFFFFFFFF, 0);
 398
 399	/* cache window 2: context */
 400	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
 401		     lower_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE),
 402			     0xFFFFFFFF, 0);
 403	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
 404		     upper_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE),
 405			     0xFFFFFFFF, 0);
 406	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0, 0xFFFFFFFF, 0);
 407	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE,
 408			     0xFFFFFFFF, 0);
 409
 410	/* VCN global tiling registers */
 411	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_ADDR_CONFIG,
 412			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 413	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG,
 414			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 415	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG,
 416			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 417	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG,
 418		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 419	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG,
 420		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 421	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG,
 422		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 423	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG,
 424		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 425	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG,
 426		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 427	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG,
 428		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 429	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG,
 430		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
 431}
 432
 433/**
 434 * vcn_v1_0_disable_clock_gating - disable VCN clock gating
 435 *
 436 * @adev: amdgpu_device pointer
 437 * @sw: enable SW clock gating
 438 *
 439 * Disable clock gating for VCN block
 440 */
 441static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev)
 442{
 443	uint32_t data;
 444
 445	/* JPEG disable CGC */
 446	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL);
 447
 448	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 449		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 450	else
 451		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE_MASK;
 452
 453	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 454	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
 455	WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data);
 456
 457	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE);
 458	data &= ~(JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK);
 459	WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data);
 460
 461	/* UVD disable CGC */
 462	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
 463	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 464		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 465	else
 466		data &= ~ UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
 467
 468	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 469	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
 470	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
 471
 472	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_GATE);
 473	data &= ~(UVD_CGC_GATE__SYS_MASK
 474		| UVD_CGC_GATE__UDEC_MASK
 475		| UVD_CGC_GATE__MPEG2_MASK
 476		| UVD_CGC_GATE__REGS_MASK
 477		| UVD_CGC_GATE__RBC_MASK
 478		| UVD_CGC_GATE__LMI_MC_MASK
 479		| UVD_CGC_GATE__LMI_UMC_MASK
 480		| UVD_CGC_GATE__IDCT_MASK
 481		| UVD_CGC_GATE__MPRD_MASK
 482		| UVD_CGC_GATE__MPC_MASK
 483		| UVD_CGC_GATE__LBSI_MASK
 484		| UVD_CGC_GATE__LRBBM_MASK
 485		| UVD_CGC_GATE__UDEC_RE_MASK
 486		| UVD_CGC_GATE__UDEC_CM_MASK
 487		| UVD_CGC_GATE__UDEC_IT_MASK
 488		| UVD_CGC_GATE__UDEC_DB_MASK
 489		| UVD_CGC_GATE__UDEC_MP_MASK
 490		| UVD_CGC_GATE__WCB_MASK
 491		| UVD_CGC_GATE__VCPU_MASK
 492		| UVD_CGC_GATE__SCPU_MASK);
 493	WREG32_SOC15(VCN, 0, mmUVD_CGC_GATE, data);
 494
 495	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
 496	data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK
 497		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
 498		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
 499		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
 500		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
 501		| UVD_CGC_CTRL__SYS_MODE_MASK
 502		| UVD_CGC_CTRL__UDEC_MODE_MASK
 503		| UVD_CGC_CTRL__MPEG2_MODE_MASK
 504		| UVD_CGC_CTRL__REGS_MODE_MASK
 505		| UVD_CGC_CTRL__RBC_MODE_MASK
 506		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
 507		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
 508		| UVD_CGC_CTRL__IDCT_MODE_MASK
 509		| UVD_CGC_CTRL__MPRD_MODE_MASK
 510		| UVD_CGC_CTRL__MPC_MODE_MASK
 511		| UVD_CGC_CTRL__LBSI_MODE_MASK
 512		| UVD_CGC_CTRL__LRBBM_MODE_MASK
 513		| UVD_CGC_CTRL__WCB_MODE_MASK
 514		| UVD_CGC_CTRL__VCPU_MODE_MASK
 515		| UVD_CGC_CTRL__SCPU_MODE_MASK);
 516	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
 517
 518	/* turn on */
 519	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE);
 520	data |= (UVD_SUVD_CGC_GATE__SRE_MASK
 521		| UVD_SUVD_CGC_GATE__SIT_MASK
 522		| UVD_SUVD_CGC_GATE__SMP_MASK
 523		| UVD_SUVD_CGC_GATE__SCM_MASK
 524		| UVD_SUVD_CGC_GATE__SDB_MASK
 525		| UVD_SUVD_CGC_GATE__SRE_H264_MASK
 526		| UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
 527		| UVD_SUVD_CGC_GATE__SIT_H264_MASK
 528		| UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
 529		| UVD_SUVD_CGC_GATE__SCM_H264_MASK
 530		| UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
 531		| UVD_SUVD_CGC_GATE__SDB_H264_MASK
 532		| UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
 533		| UVD_SUVD_CGC_GATE__SCLR_MASK
 534		| UVD_SUVD_CGC_GATE__UVD_SC_MASK
 535		| UVD_SUVD_CGC_GATE__ENT_MASK
 536		| UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
 537		| UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK
 538		| UVD_SUVD_CGC_GATE__SITE_MASK
 539		| UVD_SUVD_CGC_GATE__SRE_VP9_MASK
 540		| UVD_SUVD_CGC_GATE__SCM_VP9_MASK
 541		| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
 542		| UVD_SUVD_CGC_GATE__SDB_VP9_MASK
 543		| UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
 544	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE, data);
 545
 546	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL);
 547	data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
 548		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
 549		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
 550		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
 551		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
 552		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
 553		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
 554		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
 555		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
 556		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
 557	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
 558}
 559
 560/**
 561 * vcn_v1_0_enable_clock_gating - enable VCN clock gating
 562 *
 563 * @adev: amdgpu_device pointer
 564 * @sw: enable SW clock gating
 565 *
 566 * Enable clock gating for VCN block
 567 */
 568static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev)
 569{
 570	uint32_t data = 0;
 571
 572	/* enable JPEG CGC */
 573	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL);
 574	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 575		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 576	else
 577		data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 578	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 579	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
 580	WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data);
 581
 582	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE);
 583	data |= (JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK);
 584	WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data);
 585
 586	/* enable UVD CGC */
 587	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
 588	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 589		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 590	else
 591		data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 592	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 593	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
 594	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
 595
 596	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
 597	data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK
 598		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
 599		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
 600		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
 601		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
 602		| UVD_CGC_CTRL__SYS_MODE_MASK
 603		| UVD_CGC_CTRL__UDEC_MODE_MASK
 604		| UVD_CGC_CTRL__MPEG2_MODE_MASK
 605		| UVD_CGC_CTRL__REGS_MODE_MASK
 606		| UVD_CGC_CTRL__RBC_MODE_MASK
 607		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
 608		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
 609		| UVD_CGC_CTRL__IDCT_MODE_MASK
 610		| UVD_CGC_CTRL__MPRD_MODE_MASK
 611		| UVD_CGC_CTRL__MPC_MODE_MASK
 612		| UVD_CGC_CTRL__LBSI_MODE_MASK
 613		| UVD_CGC_CTRL__LRBBM_MODE_MASK
 614		| UVD_CGC_CTRL__WCB_MODE_MASK
 615		| UVD_CGC_CTRL__VCPU_MODE_MASK
 616		| UVD_CGC_CTRL__SCPU_MODE_MASK);
 617	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
 618
 619	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL);
 620	data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
 621		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
 622		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
 623		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
 624		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
 625		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
 626		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
 627		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
 628		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
 629		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
 630	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
 631}
 632
 633static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel)
 634{
 635	uint32_t reg_data = 0;
 636
 637	/* disable JPEG CGC */
 638	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 639		reg_data = 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 640	else
 641		reg_data = 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 642	reg_data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 643	reg_data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
 644	WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
 645
 646	WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_GATE, 0, 0xFFFFFFFF, sram_sel);
 647
 648	/* enable sw clock gating control */
 649	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
 650		reg_data = 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 651	else
 652		reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
 653	reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
 654	reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
 655	reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
 656		 UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
 657		 UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
 658		 UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
 659		 UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
 660		 UVD_CGC_CTRL__SYS_MODE_MASK |
 661		 UVD_CGC_CTRL__UDEC_MODE_MASK |
 662		 UVD_CGC_CTRL__MPEG2_MODE_MASK |
 663		 UVD_CGC_CTRL__REGS_MODE_MASK |
 664		 UVD_CGC_CTRL__RBC_MODE_MASK |
 665		 UVD_CGC_CTRL__LMI_MC_MODE_MASK |
 666		 UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
 667		 UVD_CGC_CTRL__IDCT_MODE_MASK |
 668		 UVD_CGC_CTRL__MPRD_MODE_MASK |
 669		 UVD_CGC_CTRL__MPC_MODE_MASK |
 670		 UVD_CGC_CTRL__LBSI_MODE_MASK |
 671		 UVD_CGC_CTRL__LRBBM_MODE_MASK |
 672		 UVD_CGC_CTRL__WCB_MODE_MASK |
 673		 UVD_CGC_CTRL__VCPU_MODE_MASK |
 674		 UVD_CGC_CTRL__SCPU_MODE_MASK);
 675	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
 676
 677	/* turn off clock gating */
 678	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_GATE, 0, 0xFFFFFFFF, sram_sel);
 679
 680	/* turn on SUVD clock gating */
 681	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_GATE, 1, 0xFFFFFFFF, sram_sel);
 682
 683	/* turn on sw mode in UVD_SUVD_CGC_CTRL */
 684	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_CTRL, 0, 0xFFFFFFFF, sram_sel);
 685}
 686
 687static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev)
 688{
 689	uint32_t data = 0;
 690	int ret;
 691
 692	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
 693		data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT
 694			| 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT
 695			| 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT
 696			| 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT
 697			| 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT
 698			| 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT
 699			| 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT
 700			| 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT
 701			| 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT
 702			| 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT
 703			| 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT);
 704
 705		WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data);
 706		SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON, 0xFFFFFF, ret);
 707	} else {
 708		data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT
 709			| 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT
 710			| 1 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT
 711			| 1 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT
 712			| 1 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT
 713			| 1 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT
 714			| 1 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT
 715			| 1 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT
 716			| 1 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT
 717			| 1 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT
 718			| 1 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT);
 719		WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data);
 720		SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, 0,  0xFFFFFFFF, ret);
 721	}
 722
 723	/* polling UVD_PGFSM_STATUS to confirm UVDM_PWR_STATUS , UVDU_PWR_STATUS are 0 (power on) */
 724
 725	data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS);
 726	data &= ~0x103;
 727	if (adev->pg_flags & AMD_PG_SUPPORT_VCN)
 728		data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON | UVD_POWER_STATUS__UVD_PG_EN_MASK;
 729
 730	WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data);
 731}
 732
 733static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev)
 734{
 735	uint32_t data = 0;
 736	int ret;
 737
 738	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
 739		/* Before power off, this indicator has to be turned on */
 740		data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS);
 741		data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK;
 742		data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
 743		WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data);
 744
 745
 746		data = (2 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT
 747			| 2 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT
 748			| 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT
 749			| 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT
 750			| 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT
 751			| 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT
 752			| 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT
 753			| 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT
 754			| 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT
 755			| 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT
 756			| 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT);
 757
 758		WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data);
 759
 760		data = (2 << UVD_PGFSM_STATUS__UVDM_PWR_STATUS__SHIFT
 761			| 2 << UVD_PGFSM_STATUS__UVDU_PWR_STATUS__SHIFT
 762			| 2 << UVD_PGFSM_STATUS__UVDF_PWR_STATUS__SHIFT
 763			| 2 << UVD_PGFSM_STATUS__UVDC_PWR_STATUS__SHIFT
 764			| 2 << UVD_PGFSM_STATUS__UVDB_PWR_STATUS__SHIFT
 765			| 2 << UVD_PGFSM_STATUS__UVDIL_PWR_STATUS__SHIFT
 766			| 2 << UVD_PGFSM_STATUS__UVDIR_PWR_STATUS__SHIFT
 767			| 2 << UVD_PGFSM_STATUS__UVDTD_PWR_STATUS__SHIFT
 768			| 2 << UVD_PGFSM_STATUS__UVDTE_PWR_STATUS__SHIFT
 769			| 2 << UVD_PGFSM_STATUS__UVDE_PWR_STATUS__SHIFT
 770			| 2 << UVD_PGFSM_STATUS__UVDW_PWR_STATUS__SHIFT);
 771		SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, data, 0xFFFFFFFF, ret);
 772	}
 773}
 774
 775/**
 776 * vcn_v1_0_start - start VCN block
 777 *
 778 * @adev: amdgpu_device pointer
 779 *
 780 * Setup and start the VCN block
 781 */
 782static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
 783{
 784	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 785	uint32_t rb_bufsz, tmp;
 786	uint32_t lmi_swap_cntl;
 787	int i, j, r;
 788
 789	/* disable byte swapping */
 790	lmi_swap_cntl = 0;
 791
 792	vcn_1_0_disable_static_power_gating(adev);
 793
 794	tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
 795	WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp);
 796
 797	/* disable clock gating */
 798	vcn_v1_0_disable_clock_gating(adev);
 799
 800	/* disable interupt */
 801	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0,
 802			~UVD_MASTINT_EN__VCPU_EN_MASK);
 803
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 804	/* initialize VCN memory controller */
 805	tmp = RREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL);
 806	WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL, tmp		|
 807		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK	|
 808		UVD_LMI_CTRL__MASK_MC_URGENT_MASK			|
 809		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK		|
 810		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
 
 811
 812#ifdef __BIG_ENDIAN
 813	/* swap (8 in 32) RB and IB */
 814	lmi_swap_cntl = 0xa;
 815#endif
 816	WREG32_SOC15(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
 817
 818	tmp = RREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL);
 819	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
 820	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
 821	WREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL, tmp);
 822
 823	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0,
 824		((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
 825		(0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
 826		(0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
 827		(0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
 828
 829	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0,
 830		((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
 831		(0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
 832		(0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
 833		(0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
 834
 835	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX,
 836		((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
 837		(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
 838		(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
 839
 840	vcn_v1_0_mc_resume_spg_mode(adev);
 841
 842	WREG32_SOC15(UVD, 0, mmUVD_REG_XX_MASK, 0x10);
 843	WREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK,
 844		RREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK) | 0x3);
 845
 846	/* enable VCPU clock */
 847	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK);
 848
 849	/* boot up the VCPU */
 850	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0,
 851			~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
 852
 853	/* enable UMC */
 854	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0,
 855			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
 856
 857	tmp = RREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET);
 858	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
 859	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
 860	WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, tmp);
 861
 862	for (i = 0; i < 10; ++i) {
 863		uint32_t status;
 864
 865		for (j = 0; j < 100; ++j) {
 866			status = RREG32_SOC15(UVD, 0, mmUVD_STATUS);
 867			if (status & UVD_STATUS__IDLE)
 868				break;
 869			mdelay(10);
 870		}
 871		r = 0;
 872		if (status & UVD_STATUS__IDLE)
 873			break;
 874
 875		DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n");
 876		WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
 877				UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
 878				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
 879		mdelay(10);
 880		WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0,
 881				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
 882		mdelay(10);
 883		r = -1;
 884	}
 885
 886	if (r) {
 887		DRM_ERROR("VCN decode not responding, giving up!!!\n");
 888		return r;
 889	}
 890	/* enable master interrupt */
 891	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN),
 892		UVD_MASTINT_EN__VCPU_EN_MASK, ~UVD_MASTINT_EN__VCPU_EN_MASK);
 
 893
 894	/* enable system interrupt for JRBC, TODO: move to set interrupt*/
 895	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SYS_INT_EN),
 896		UVD_SYS_INT_EN__UVD_JRBC_EN_MASK,
 897		~UVD_SYS_INT_EN__UVD_JRBC_EN_MASK);
 898
 899	/* clear the busy bit of UVD_STATUS */
 900	tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) & ~UVD_STATUS__UVD_BUSY;
 901	WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp);
 902
 903	/* force RBC into idle state */
 904	rb_bufsz = order_base_2(ring->ring_size);
 905	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
 906	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
 907	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
 
 908	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
 909	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
 910	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp);
 911
 912	/* set the write pointer delay */
 913	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0);
 914
 915	/* set the wb address */
 916	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR,
 917			(upper_32_bits(ring->gpu_addr) >> 2));
 918
 919	/* programm the RB_BASE for ring buffer */
 920	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
 921			lower_32_bits(ring->gpu_addr));
 922	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
 923			upper_32_bits(ring->gpu_addr));
 924
 925	/* Initialize the ring buffer's read and write pointers */
 926	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0);
 927
 928	WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0);
 929
 930	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
 931	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
 932			lower_32_bits(ring->wptr));
 933
 934	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0,
 935			~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
 936
 937	ring = &adev->vcn.inst->ring_enc[0];
 938	WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
 939	WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
 940	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr);
 941	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
 942	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4);
 943
 944	ring = &adev->vcn.inst->ring_enc[1];
 945	WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
 946	WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
 947	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr);
 948	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
 949	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
 950
 951	ring = &adev->vcn.inst->ring_jpeg;
 952	WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
 953	WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK |
 954			UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
 955	WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW, lower_32_bits(ring->gpu_addr));
 956	WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH, upper_32_bits(ring->gpu_addr));
 957	WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, 0);
 958	WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, 0);
 959	WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
 960
 961	/* initialize wptr */
 962	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR);
 963
 964	/* copy patch commands to the jpeg ring */
 965	vcn_v1_0_jpeg_ring_set_patch_ring(ring,
 966		(ring->wptr + ring->max_dw * amdgpu_sched_hw_submission));
 967
 968	return 0;
 969}
 970
 971static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
 972{
 973	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
 974	uint32_t rb_bufsz, tmp;
 975	uint32_t lmi_swap_cntl;
 976
 977	/* disable byte swapping */
 978	lmi_swap_cntl = 0;
 979
 980	vcn_1_0_enable_static_power_gating(adev);
 981
 982	/* enable dynamic power gating mode */
 983	tmp = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
 984	tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
 985	tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
 986	WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, tmp);
 987
 988	/* enable clock gating */
 989	vcn_v1_0_clock_gating_dpg_mode(adev, 0);
 990
 991	/* enable VCPU clock */
 992	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
 993	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
 994	tmp |= UVD_VCPU_CNTL__MIF_WR_LOW_THRESHOLD_BP_MASK;
 995	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CNTL, tmp, 0xFFFFFFFF, 0);
 996
 997	/* disable interupt */
 998	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN,
 999			0, UVD_MASTINT_EN__VCPU_EN_MASK, 0);
1000
1001	/* initialize VCN memory controller */
1002	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL,
1003		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
1004		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
1005		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
1006		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
1007		UVD_LMI_CTRL__REQ_MODE_MASK |
1008		UVD_LMI_CTRL__CRC_RESET_MASK |
1009		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
1010		0x00100000L, 0xFFFFFFFF, 0);
1011
1012#ifdef __BIG_ENDIAN
1013	/* swap (8 in 32) RB and IB */
1014	lmi_swap_cntl = 0xa;
1015#endif
1016	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl, 0xFFFFFFFF, 0);
1017
1018	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_CNTL,
1019		0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0xFFFFFFFF, 0);
1020
1021	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXA0,
1022		((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
1023		 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
1024		 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
1025		 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0xFFFFFFFF, 0);
1026
1027	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXB0,
1028		((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
1029		 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
1030		 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
1031		 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0xFFFFFFFF, 0);
1032
1033	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUX,
1034		((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
1035		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
1036		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0xFFFFFFFF, 0);
1037
1038	vcn_v1_0_mc_resume_dpg_mode(adev);
1039
1040	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_REG_XX_MASK, 0x10, 0xFFFFFFFF, 0);
1041	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, 0x3, 0xFFFFFFFF, 0);
1042
1043	/* boot up the VCPU */
1044	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SOFT_RESET, 0, 0xFFFFFFFF, 0);
1045
1046	/* enable UMC */
1047	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL2,
1048		0x1F << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT,
1049		0xFFFFFFFF, 0);
1050
1051	/* enable master interrupt */
1052	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN,
1053			UVD_MASTINT_EN__VCPU_EN_MASK, UVD_MASTINT_EN__VCPU_EN_MASK, 0);
1054
1055	vcn_v1_0_clock_gating_dpg_mode(adev, 1);
1056	/* setup mmUVD_LMI_CTRL */
1057	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL,
1058		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
1059		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
1060		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
1061		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
1062		UVD_LMI_CTRL__REQ_MODE_MASK |
1063		UVD_LMI_CTRL__CRC_RESET_MASK |
1064		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
1065		0x00100000L, 0xFFFFFFFF, 1);
1066
1067	tmp = adev->gfx.config.gb_addr_config;
1068	/* setup VCN global tiling registers */
1069	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1);
1070	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1);
1071
1072	/* enable System Interrupt for JRBC */
1073	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SYS_INT_EN,
1074									UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, 0xFFFFFFFF, 1);
1075
1076	/* force RBC into idle state */
1077	rb_bufsz = order_base_2(ring->ring_size);
1078	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
1079	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
1080	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
1081	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
1082	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
1083	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp);
1084
1085	/* set the write pointer delay */
1086	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0);
1087
1088	/* set the wb address */
1089	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR,
1090								(upper_32_bits(ring->gpu_addr) >> 2));
1091
1092	/* programm the RB_BASE for ring buffer */
1093	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
1094								lower_32_bits(ring->gpu_addr));
1095	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
1096								upper_32_bits(ring->gpu_addr));
1097
1098	/* Initialize the ring buffer's read and write pointers */
1099	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0);
1100
1101	WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0);
1102
1103	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
1104	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
1105								lower_32_bits(ring->wptr));
1106
1107	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0,
1108			~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
1109
1110	/* initialize JPEG wptr */
1111	ring = &adev->vcn.inst->ring_jpeg;
1112	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR);
1113
1114	/* copy patch commands to the jpeg ring */
1115	vcn_v1_0_jpeg_ring_set_patch_ring(ring,
1116		(ring->wptr + ring->max_dw * amdgpu_sched_hw_submission));
1117
1118	return 0;
1119}
1120
1121static int vcn_v1_0_start(struct amdgpu_device *adev)
1122{
1123	int r;
1124
1125	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
1126		r = vcn_v1_0_start_dpg_mode(adev);
1127	else
1128		r = vcn_v1_0_start_spg_mode(adev);
1129	return r;
1130}
1131
1132/**
1133 * vcn_v1_0_stop - stop VCN block
1134 *
1135 * @adev: amdgpu_device pointer
1136 *
1137 * stop the VCN block
1138 */
1139static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev)
1140{
1141	int ret_code, tmp;
 
1142
1143	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7, ret_code);
1144
1145	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
1146		UVD_LMI_STATUS__READ_CLEAN_MASK |
1147		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
1148		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
1149	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp, ret_code);
1150
1151	/* put VCPU into reset */
1152	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
1153		UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
1154		~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
1155
1156	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
1157		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
1158	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp, ret_code);
1159
1160	/* disable VCPU clock */
1161	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL), 0,
1162		~UVD_VCPU_CNTL__CLK_EN_MASK);
1163
1164	/* reset LMI UMC/LMI */
1165	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
1166		UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK,
1167		~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK);
1168
1169	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
1170		UVD_SOFT_RESET__LMI_SOFT_RESET_MASK,
1171		~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK);
1172
1173	WREG32_SOC15(UVD, 0, mmUVD_STATUS, 0);
 
1174
1175	vcn_v1_0_enable_clock_gating(adev);
1176	vcn_1_0_enable_static_power_gating(adev);
1177	return 0;
1178}
1179
1180static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev)
1181{
1182	int ret_code = 0;
1183	uint32_t tmp;
1184
1185	/* Wait for power status to be UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF */
1186	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1187			UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1188			UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
1189
1190	/* wait for read ptr to be equal to write ptr */
1191	tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR);
1192	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR, tmp, 0xFFFFFFFF, ret_code);
1193
1194	tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2);
1195	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR2, tmp, 0xFFFFFFFF, ret_code);
1196
1197	tmp = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR);
1198	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_JRBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code);
1199
1200	tmp = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR) & 0x7FFFFFFF;
1201	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code);
1202
1203	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1204		UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1205		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
1206
1207	/* disable dynamic power gating mode */
1208	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0,
1209			~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
1210
1211	return 0;
1212}
1213
1214static int vcn_v1_0_stop(struct amdgpu_device *adev)
1215{
1216	int r;
1217
1218	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
1219		r = vcn_v1_0_stop_dpg_mode(adev);
1220	else
1221		r = vcn_v1_0_stop_spg_mode(adev);
1222
1223	return r;
1224}
1225
1226static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
1227				struct dpg_pause_state *new_state)
1228{
1229	int ret_code;
1230	uint32_t reg_data = 0;
1231	uint32_t reg_data2 = 0;
1232	struct amdgpu_ring *ring;
1233
1234	/* pause/unpause if state is changed */
1235	if (adev->vcn.pause_state.fw_based != new_state->fw_based) {
1236		DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d",
1237			adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg,
1238			new_state->fw_based, new_state->jpeg);
1239
1240		reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) &
1241			(~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
1242
1243		if (new_state->fw_based == VCN_DPG_STATE__PAUSE) {
1244			ret_code = 0;
1245
1246			if (!(reg_data & UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK))
1247				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1248						   UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1249						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
1250
1251			if (!ret_code) {
1252				/* pause DPG non-jpeg */
1253				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
1254				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1255				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE,
1256						   UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK,
1257						   UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, ret_code);
1258
1259				/* Restore */
1260				ring = &adev->vcn.inst->ring_enc[0];
1261				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr);
1262				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
1263				WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4);
1264				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
1265				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
1266
1267				ring = &adev->vcn.inst->ring_enc[1];
1268				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr);
1269				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
1270				WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
1271				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
1272				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
1273
1274				ring = &adev->vcn.inst->ring_dec;
1275				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
1276						   RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF);
1277				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1278						   UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
1279						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
1280			}
1281		} else {
1282			/* unpause dpg non-jpeg, no need to wait */
1283			reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
1284			WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1285		}
1286		adev->vcn.pause_state.fw_based = new_state->fw_based;
1287	}
1288
1289	/* pause/unpause if state is changed */
1290	if (adev->vcn.pause_state.jpeg != new_state->jpeg) {
1291		DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d",
1292			adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg,
1293			new_state->fw_based, new_state->jpeg);
1294
1295		reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) &
1296			(~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK);
1297
1298		if (new_state->jpeg == VCN_DPG_STATE__PAUSE) {
1299			ret_code = 0;
1300
1301			if (!(reg_data & UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK))
1302				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1303						   UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1304						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
1305
1306			if (!ret_code) {
1307				/* Make sure JPRG Snoop is disabled before sending the pause */
1308				reg_data2 = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
1309				reg_data2 |= UVD_POWER_STATUS__JRBC_SNOOP_DIS_MASK;
1310				WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, reg_data2);
1311
1312				/* pause DPG jpeg */
1313				reg_data |= UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK;
1314				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1315				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE,
1316							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK,
1317							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK, ret_code);
1318
1319				/* Restore */
1320				ring = &adev->vcn.inst->ring_jpeg;
1321				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
1322				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL,
1323							UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK |
1324							UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
1325				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
1326							lower_32_bits(ring->gpu_addr));
1327				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
1328							upper_32_bits(ring->gpu_addr));
1329				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, ring->wptr);
1330				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, ring->wptr);
1331				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL,
1332							UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
1333
1334				ring = &adev->vcn.inst->ring_dec;
1335				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
1336						   RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF);
1337				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1338						   UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
1339						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
1340			}
1341		} else {
1342			/* unpause dpg jpeg, no need to wait */
1343			reg_data &= ~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK;
1344			WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1345		}
1346		adev->vcn.pause_state.jpeg = new_state->jpeg;
1347	}
1348
1349	return 0;
1350}
1351
1352static bool vcn_v1_0_is_idle(void *handle)
1353{
1354	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
1355
1356	return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == UVD_STATUS__IDLE);
1357}
1358
1359static int vcn_v1_0_wait_for_idle(void *handle)
1360{
1361	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
1362	int ret = 0;
1363
1364	SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, UVD_STATUS__IDLE,
1365		UVD_STATUS__IDLE, ret);
1366
1367	return ret;
1368}
1369
1370static int vcn_v1_0_set_clockgating_state(void *handle,
1371					  enum amd_clockgating_state state)
1372{
1373	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
1374	bool enable = (state == AMD_CG_STATE_GATE) ? true : false;
1375
1376	if (enable) {
1377		/* wait for STATUS to clear */
1378		if (vcn_v1_0_is_idle(handle))
1379			return -EBUSY;
1380		vcn_v1_0_enable_clock_gating(adev);
1381	} else {
1382		/* disable HW gating and enable Sw gating */
1383		vcn_v1_0_disable_clock_gating(adev);
1384	}
1385	return 0;
1386}
1387
1388/**
1389 * vcn_v1_0_dec_ring_get_rptr - get read pointer
1390 *
1391 * @ring: amdgpu_ring pointer
1392 *
1393 * Returns the current hardware read pointer
1394 */
1395static uint64_t vcn_v1_0_dec_ring_get_rptr(struct amdgpu_ring *ring)
1396{
1397	struct amdgpu_device *adev = ring->adev;
1398
1399	return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
1400}
1401
1402/**
1403 * vcn_v1_0_dec_ring_get_wptr - get write pointer
1404 *
1405 * @ring: amdgpu_ring pointer
1406 *
1407 * Returns the current hardware write pointer
1408 */
1409static uint64_t vcn_v1_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
1410{
1411	struct amdgpu_device *adev = ring->adev;
1412
1413	return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR);
1414}
1415
1416/**
1417 * vcn_v1_0_dec_ring_set_wptr - set write pointer
1418 *
1419 * @ring: amdgpu_ring pointer
1420 *
1421 * Commits the write pointer to the hardware
1422 */
1423static void vcn_v1_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
1424{
1425	struct amdgpu_device *adev = ring->adev;
1426
1427	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
1428		WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2,
1429			lower_32_bits(ring->wptr) | 0x80000000);
1430
1431	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
1432}
1433
1434/**
1435 * vcn_v1_0_dec_ring_insert_start - insert a start command
1436 *
1437 * @ring: amdgpu_ring pointer
1438 *
1439 * Write a start command to the ring.
1440 */
1441static void vcn_v1_0_dec_ring_insert_start(struct amdgpu_ring *ring)
1442{
1443	struct amdgpu_device *adev = ring->adev;
1444
1445	amdgpu_ring_write(ring,
1446		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1447	amdgpu_ring_write(ring, 0);
1448	amdgpu_ring_write(ring,
1449		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1450	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_START << 1);
1451}
1452
1453/**
1454 * vcn_v1_0_dec_ring_insert_end - insert a end command
1455 *
1456 * @ring: amdgpu_ring pointer
1457 *
1458 * Write a end command to the ring.
1459 */
1460static void vcn_v1_0_dec_ring_insert_end(struct amdgpu_ring *ring)
1461{
1462	struct amdgpu_device *adev = ring->adev;
1463
1464	amdgpu_ring_write(ring,
1465		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1466	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_END << 1);
1467}
1468
1469/**
1470 * vcn_v1_0_dec_ring_emit_fence - emit an fence & trap command
1471 *
1472 * @ring: amdgpu_ring pointer
1473 * @fence: fence to emit
1474 *
1475 * Write a fence and a trap command to the ring.
1476 */
1477static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
1478				     unsigned flags)
1479{
1480	struct amdgpu_device *adev = ring->adev;
1481
1482	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
1483
1484	amdgpu_ring_write(ring,
1485		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0));
1486	amdgpu_ring_write(ring, seq);
1487	amdgpu_ring_write(ring,
1488		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1489	amdgpu_ring_write(ring, addr & 0xffffffff);
1490	amdgpu_ring_write(ring,
1491		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1492	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
1493	amdgpu_ring_write(ring,
1494		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1495	amdgpu_ring_write(ring, VCN_DEC_CMD_FENCE << 1);
1496
1497	amdgpu_ring_write(ring,
1498		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1499	amdgpu_ring_write(ring, 0);
1500	amdgpu_ring_write(ring,
1501		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1502	amdgpu_ring_write(ring, 0);
1503	amdgpu_ring_write(ring,
1504		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1505	amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1);
1506}
1507
1508/**
1509 * vcn_v1_0_dec_ring_emit_ib - execute indirect buffer
1510 *
1511 * @ring: amdgpu_ring pointer
1512 * @ib: indirect buffer to execute
1513 *
1514 * Write ring commands to execute the indirect buffer
1515 */
1516static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
1517					struct amdgpu_job *job,
1518					struct amdgpu_ib *ib,
1519					uint32_t flags)
1520{
1521	struct amdgpu_device *adev = ring->adev;
1522	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
1523
1524	amdgpu_ring_write(ring,
1525		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0));
1526	amdgpu_ring_write(ring, vmid);
1527
1528	amdgpu_ring_write(ring,
1529		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0));
1530	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
1531	amdgpu_ring_write(ring,
1532		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0));
1533	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
1534	amdgpu_ring_write(ring,
1535		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_IB_SIZE), 0));
1536	amdgpu_ring_write(ring, ib->length_dw);
1537}
1538
1539static void vcn_v1_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring,
1540					    uint32_t reg, uint32_t val,
1541					    uint32_t mask)
1542{
1543	struct amdgpu_device *adev = ring->adev;
1544
1545	amdgpu_ring_write(ring,
1546		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1547	amdgpu_ring_write(ring, reg << 2);
1548	amdgpu_ring_write(ring,
1549		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1550	amdgpu_ring_write(ring, val);
1551	amdgpu_ring_write(ring,
1552		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH8), 0));
1553	amdgpu_ring_write(ring, mask);
1554	amdgpu_ring_write(ring,
1555		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1556	amdgpu_ring_write(ring, VCN_DEC_CMD_REG_READ_COND_WAIT << 1);
1557}
1558
1559static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
1560					    unsigned vmid, uint64_t pd_addr)
1561{
1562	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
1563	uint32_t data0, data1, mask;
1564
1565	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
1566
1567	/* wait for register write */
1568	data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2;
1569	data1 = lower_32_bits(pd_addr);
1570	mask = 0xffffffff;
1571	vcn_v1_0_dec_ring_emit_reg_wait(ring, data0, data1, mask);
1572}
1573
1574static void vcn_v1_0_dec_ring_emit_wreg(struct amdgpu_ring *ring,
1575					uint32_t reg, uint32_t val)
1576{
1577	struct amdgpu_device *adev = ring->adev;
1578
1579	amdgpu_ring_write(ring,
1580		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1581	amdgpu_ring_write(ring, reg << 2);
1582	amdgpu_ring_write(ring,
1583		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1584	amdgpu_ring_write(ring, val);
1585	amdgpu_ring_write(ring,
1586		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1587	amdgpu_ring_write(ring, VCN_DEC_CMD_WRITE_REG << 1);
1588}
1589
1590/**
1591 * vcn_v1_0_enc_ring_get_rptr - get enc read pointer
1592 *
1593 * @ring: amdgpu_ring pointer
1594 *
1595 * Returns the current hardware enc read pointer
1596 */
1597static uint64_t vcn_v1_0_enc_ring_get_rptr(struct amdgpu_ring *ring)
1598{
1599	struct amdgpu_device *adev = ring->adev;
1600
1601	if (ring == &adev->vcn.inst->ring_enc[0])
1602		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR);
1603	else
1604		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2);
1605}
1606
1607 /**
1608 * vcn_v1_0_enc_ring_get_wptr - get enc write pointer
1609 *
1610 * @ring: amdgpu_ring pointer
1611 *
1612 * Returns the current hardware enc write pointer
1613 */
1614static uint64_t vcn_v1_0_enc_ring_get_wptr(struct amdgpu_ring *ring)
1615{
1616	struct amdgpu_device *adev = ring->adev;
1617
1618	if (ring == &adev->vcn.inst->ring_enc[0])
1619		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR);
1620	else
1621		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2);
1622}
1623
1624 /**
1625 * vcn_v1_0_enc_ring_set_wptr - set enc write pointer
1626 *
1627 * @ring: amdgpu_ring pointer
1628 *
1629 * Commits the enc write pointer to the hardware
1630 */
1631static void vcn_v1_0_enc_ring_set_wptr(struct amdgpu_ring *ring)
1632{
1633	struct amdgpu_device *adev = ring->adev;
1634
1635	if (ring == &adev->vcn.inst->ring_enc[0])
1636		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR,
1637			lower_32_bits(ring->wptr));
1638	else
1639		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2,
1640			lower_32_bits(ring->wptr));
1641}
1642
1643/**
1644 * vcn_v1_0_enc_ring_emit_fence - emit an enc fence & trap command
1645 *
1646 * @ring: amdgpu_ring pointer
1647 * @fence: fence to emit
1648 *
1649 * Write enc a fence and a trap command to the ring.
1650 */
1651static void vcn_v1_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
1652			u64 seq, unsigned flags)
1653{
1654	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
1655
1656	amdgpu_ring_write(ring, VCN_ENC_CMD_FENCE);
1657	amdgpu_ring_write(ring, addr);
1658	amdgpu_ring_write(ring, upper_32_bits(addr));
1659	amdgpu_ring_write(ring, seq);
1660	amdgpu_ring_write(ring, VCN_ENC_CMD_TRAP);
1661}
1662
1663static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring)
1664{
1665	amdgpu_ring_write(ring, VCN_ENC_CMD_END);
1666}
1667
1668/**
1669 * vcn_v1_0_enc_ring_emit_ib - enc execute indirect buffer
1670 *
1671 * @ring: amdgpu_ring pointer
1672 * @ib: indirect buffer to execute
1673 *
1674 * Write enc ring commands to execute the indirect buffer
1675 */
1676static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
1677					struct amdgpu_job *job,
1678					struct amdgpu_ib *ib,
1679					uint32_t flags)
1680{
1681	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
1682
1683	amdgpu_ring_write(ring, VCN_ENC_CMD_IB);
1684	amdgpu_ring_write(ring, vmid);
1685	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
1686	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
1687	amdgpu_ring_write(ring, ib->length_dw);
1688}
1689
1690static void vcn_v1_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring,
1691					    uint32_t reg, uint32_t val,
1692					    uint32_t mask)
1693{
1694	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT);
1695	amdgpu_ring_write(ring, reg << 2);
1696	amdgpu_ring_write(ring, mask);
1697	amdgpu_ring_write(ring, val);
1698}
1699
1700static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
1701					    unsigned int vmid, uint64_t pd_addr)
1702{
1703	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
1704
1705	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
1706
1707	/* wait for reg writes */
1708	vcn_v1_0_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 + vmid * 2,
1709					lower_32_bits(pd_addr), 0xffffffff);
1710}
1711
1712static void vcn_v1_0_enc_ring_emit_wreg(struct amdgpu_ring *ring,
1713					uint32_t reg, uint32_t val)
1714{
1715	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
1716	amdgpu_ring_write(ring,	reg << 2);
1717	amdgpu_ring_write(ring, val);
1718}
1719
1720
1721/**
1722 * vcn_v1_0_jpeg_ring_get_rptr - get read pointer
1723 *
1724 * @ring: amdgpu_ring pointer
1725 *
1726 * Returns the current hardware read pointer
1727 */
1728static uint64_t vcn_v1_0_jpeg_ring_get_rptr(struct amdgpu_ring *ring)
1729{
1730	struct amdgpu_device *adev = ring->adev;
1731
1732	return RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR);
1733}
1734
1735/**
1736 * vcn_v1_0_jpeg_ring_get_wptr - get write pointer
1737 *
1738 * @ring: amdgpu_ring pointer
1739 *
1740 * Returns the current hardware write pointer
1741 */
1742static uint64_t vcn_v1_0_jpeg_ring_get_wptr(struct amdgpu_ring *ring)
1743{
1744	struct amdgpu_device *adev = ring->adev;
1745
1746	return RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR);
1747}
1748
1749/**
1750 * vcn_v1_0_jpeg_ring_set_wptr - set write pointer
1751 *
1752 * @ring: amdgpu_ring pointer
1753 *
1754 * Commits the write pointer to the hardware
1755 */
1756static void vcn_v1_0_jpeg_ring_set_wptr(struct amdgpu_ring *ring)
1757{
1758	struct amdgpu_device *adev = ring->adev;
1759
1760	WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr));
1761}
1762
1763/**
1764 * vcn_v1_0_jpeg_ring_insert_start - insert a start command
1765 *
1766 * @ring: amdgpu_ring pointer
1767 *
1768 * Write a start command to the ring.
1769 */
1770static void vcn_v1_0_jpeg_ring_insert_start(struct amdgpu_ring *ring)
1771{
1772	struct amdgpu_device *adev = ring->adev;
1773
1774	amdgpu_ring_write(ring,
1775		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0));
1776	amdgpu_ring_write(ring, 0x68e04);
1777
1778	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE0));
1779	amdgpu_ring_write(ring, 0x80010000);
1780}
1781
1782/**
1783 * vcn_v1_0_jpeg_ring_insert_end - insert a end command
1784 *
1785 * @ring: amdgpu_ring pointer
1786 *
1787 * Write a end command to the ring.
1788 */
1789static void vcn_v1_0_jpeg_ring_insert_end(struct amdgpu_ring *ring)
1790{
1791	struct amdgpu_device *adev = ring->adev;
1792
1793	amdgpu_ring_write(ring,
1794		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0));
1795	amdgpu_ring_write(ring, 0x68e04);
1796
1797	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE0));
1798	amdgpu_ring_write(ring, 0x00010000);
1799}
1800
1801/**
1802 * vcn_v1_0_jpeg_ring_emit_fence - emit an fence & trap command
1803 *
1804 * @ring: amdgpu_ring pointer
1805 * @fence: fence to emit
1806 *
1807 * Write a fence and a trap command to the ring.
1808 */
1809static void vcn_v1_0_jpeg_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
1810				     unsigned flags)
1811{
1812	struct amdgpu_device *adev = ring->adev;
1813
1814	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
1815
1816	amdgpu_ring_write(ring,
1817		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_DATA0), 0, 0, PACKETJ_TYPE0));
1818	amdgpu_ring_write(ring, seq);
1819
1820	amdgpu_ring_write(ring,
1821		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_DATA1), 0, 0, PACKETJ_TYPE0));
1822	amdgpu_ring_write(ring, seq);
1823
1824	amdgpu_ring_write(ring,
1825		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0));
1826	amdgpu_ring_write(ring, lower_32_bits(addr));
1827
1828	amdgpu_ring_write(ring,
1829		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0));
1830	amdgpu_ring_write(ring, upper_32_bits(addr));
1831
1832	amdgpu_ring_write(ring,
1833		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_CMD), 0, 0, PACKETJ_TYPE0));
1834	amdgpu_ring_write(ring, 0x8);
1835
1836	amdgpu_ring_write(ring,
1837		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_CMD), 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE4));
1838	amdgpu_ring_write(ring, 0);
1839
1840	amdgpu_ring_write(ring,
1841		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0));
1842	amdgpu_ring_write(ring, 0x01400200);
1843
1844	amdgpu_ring_write(ring,
1845		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0));
1846	amdgpu_ring_write(ring, seq);
1847
1848	amdgpu_ring_write(ring,
1849		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0));
1850	amdgpu_ring_write(ring, lower_32_bits(addr));
1851
1852	amdgpu_ring_write(ring,
1853		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0));
1854	amdgpu_ring_write(ring, upper_32_bits(addr));
1855
1856	amdgpu_ring_write(ring,
1857		PACKETJ(0, 0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE2));
1858	amdgpu_ring_write(ring, 0xffffffff);
1859
1860	amdgpu_ring_write(ring,
1861		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0));
1862	amdgpu_ring_write(ring, 0x3fbc);
1863
1864	amdgpu_ring_write(ring,
1865		PACKETJ(0, 0, 0, PACKETJ_TYPE0));
1866	amdgpu_ring_write(ring, 0x1);
1867
1868	/* emit trap */
1869	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE7));
1870	amdgpu_ring_write(ring, 0);
1871}
1872
1873/**
1874 * vcn_v1_0_jpeg_ring_emit_ib - execute indirect buffer
1875 *
1876 * @ring: amdgpu_ring pointer
1877 * @ib: indirect buffer to execute
1878 *
1879 * Write ring commands to execute the indirect buffer.
1880 */
1881static void vcn_v1_0_jpeg_ring_emit_ib(struct amdgpu_ring *ring,
1882					struct amdgpu_job *job,
1883					struct amdgpu_ib *ib,
1884					uint32_t flags)
1885{
1886	struct amdgpu_device *adev = ring->adev;
1887	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
1888
1889	amdgpu_ring_write(ring,
1890		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0));
1891	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
1892
1893	amdgpu_ring_write(ring,
1894		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JPEG_VMID), 0, 0, PACKETJ_TYPE0));
1895	amdgpu_ring_write(ring, (vmid | (vmid << 4)));
1896
1897	amdgpu_ring_write(ring,
1898		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0));
1899	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
1900
1901	amdgpu_ring_write(ring,
1902		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0));
1903	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
1904
1905	amdgpu_ring_write(ring,
1906		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_IB_SIZE), 0, 0, PACKETJ_TYPE0));
1907	amdgpu_ring_write(ring, ib->length_dw);
1908
1909	amdgpu_ring_write(ring,
1910		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0));
1911	amdgpu_ring_write(ring, lower_32_bits(ring->gpu_addr));
1912
1913	amdgpu_ring_write(ring,
1914		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0));
1915	amdgpu_ring_write(ring, upper_32_bits(ring->gpu_addr));
1916
1917	amdgpu_ring_write(ring,
1918		PACKETJ(0, 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE2));
1919	amdgpu_ring_write(ring, 0);
1920
1921	amdgpu_ring_write(ring,
1922		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0));
1923	amdgpu_ring_write(ring, 0x01400200);
1924
1925	amdgpu_ring_write(ring,
1926		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0));
1927	amdgpu_ring_write(ring, 0x2);
1928
1929	amdgpu_ring_write(ring,
1930		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_STATUS), 0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE3));
1931	amdgpu_ring_write(ring, 0x2);
1932}
1933
1934static void vcn_v1_0_jpeg_ring_emit_reg_wait(struct amdgpu_ring *ring,
1935					    uint32_t reg, uint32_t val,
1936					    uint32_t mask)
1937{
1938	struct amdgpu_device *adev = ring->adev;
1939	uint32_t reg_offset = (reg << 2);
1940
1941	amdgpu_ring_write(ring,
1942		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0));
1943	amdgpu_ring_write(ring, 0x01400200);
1944
1945	amdgpu_ring_write(ring,
1946		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0));
1947	amdgpu_ring_write(ring, val);
1948
1949	amdgpu_ring_write(ring,
1950		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0));
1951	if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) ||
1952		((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) {
1953		amdgpu_ring_write(ring, 0);
1954		amdgpu_ring_write(ring,
1955			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3));
1956	} else {
1957		amdgpu_ring_write(ring, reg_offset);
1958		amdgpu_ring_write(ring,
1959			PACKETJ(0, 0, 0, PACKETJ_TYPE3));
1960	}
1961	amdgpu_ring_write(ring, mask);
1962}
1963
1964static void vcn_v1_0_jpeg_ring_emit_vm_flush(struct amdgpu_ring *ring,
1965		unsigned vmid, uint64_t pd_addr)
1966{
1967	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
1968	uint32_t data0, data1, mask;
1969
1970	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
1971
1972	/* wait for register write */
1973	data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2;
1974	data1 = lower_32_bits(pd_addr);
1975	mask = 0xffffffff;
1976	vcn_v1_0_jpeg_ring_emit_reg_wait(ring, data0, data1, mask);
1977}
1978
1979static void vcn_v1_0_jpeg_ring_emit_wreg(struct amdgpu_ring *ring,
1980					uint32_t reg, uint32_t val)
1981{
1982	struct amdgpu_device *adev = ring->adev;
1983	uint32_t reg_offset = (reg << 2);
1984
1985	amdgpu_ring_write(ring,
1986		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0));
1987	if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) ||
1988			((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) {
1989		amdgpu_ring_write(ring, 0);
1990		amdgpu_ring_write(ring,
1991			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0));
1992	} else {
1993		amdgpu_ring_write(ring, reg_offset);
1994		amdgpu_ring_write(ring,
1995			PACKETJ(0, 0, 0, PACKETJ_TYPE0));
1996	}
1997	amdgpu_ring_write(ring, val);
1998}
1999
2000static void vcn_v1_0_jpeg_ring_nop(struct amdgpu_ring *ring, uint32_t count)
2001{
2002	int i;
2003
2004	WARN_ON(ring->wptr % 2 || count % 2);
2005
2006	for (i = 0; i < count / 2; i++) {
2007		amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
2008		amdgpu_ring_write(ring, 0);
2009	}
2010}
2011
2012static void vcn_v1_0_jpeg_ring_patch_wreg(struct amdgpu_ring *ring, uint32_t *ptr, uint32_t reg_offset, uint32_t val)
2013{
2014	struct amdgpu_device *adev = ring->adev;
2015	ring->ring[(*ptr)++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0);
2016	if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) ||
2017		((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) {
2018		ring->ring[(*ptr)++] = 0;
2019		ring->ring[(*ptr)++] = PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0);
2020	} else {
2021		ring->ring[(*ptr)++] = reg_offset;
2022		ring->ring[(*ptr)++] = PACKETJ(0, 0, 0, PACKETJ_TYPE0);
2023	}
2024	ring->ring[(*ptr)++] = val;
2025}
2026
2027static void vcn_v1_0_jpeg_ring_set_patch_ring(struct amdgpu_ring *ring, uint32_t ptr)
2028{
2029	struct amdgpu_device *adev = ring->adev;
2030
2031	uint32_t reg, reg_offset, val, mask, i;
2032
2033	// 1st: program mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW
2034	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW);
2035	reg_offset = (reg << 2);
2036	val = lower_32_bits(ring->gpu_addr);
2037	vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val);
2038
2039	// 2nd: program mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH
2040	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH);
2041	reg_offset = (reg << 2);
2042	val = upper_32_bits(ring->gpu_addr);
2043	vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val);
2044
2045	// 3rd to 5th: issue MEM_READ commands
2046	for (i = 0; i <= 2; i++) {
2047		ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE2);
2048		ring->ring[ptr++] = 0;
2049	}
2050
2051	// 6th: program mmUVD_JRBC_RB_CNTL register to enable NO_FETCH and RPTR write ability
2052	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL);
2053	reg_offset = (reg << 2);
2054	val = 0x13;
2055	vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val);
2056
2057	// 7th: program mmUVD_JRBC_RB_REF_DATA
2058	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA);
2059	reg_offset = (reg << 2);
2060	val = 0x1;
2061	vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val);
2062
2063	// 8th: issue conditional register read mmUVD_JRBC_RB_CNTL
2064	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL);
2065	reg_offset = (reg << 2);
2066	val = 0x1;
2067	mask = 0x1;
2068
2069	ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0);
2070	ring->ring[ptr++] = 0x01400200;
2071	ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0);
2072	ring->ring[ptr++] = val;
2073	ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0);
2074	if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) ||
2075		((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) {
2076		ring->ring[ptr++] = 0;
2077		ring->ring[ptr++] = PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3);
2078	} else {
2079		ring->ring[ptr++] = reg_offset;
2080		ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE3);
2081	}
2082	ring->ring[ptr++] = mask;
2083
2084	//9th to 21st: insert no-op
2085	for (i = 0; i <= 12; i++) {
2086		ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE6);
2087		ring->ring[ptr++] = 0;
2088	}
2089
2090	//22nd: reset mmUVD_JRBC_RB_RPTR
2091	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_RPTR);
2092	reg_offset = (reg << 2);
2093	val = 0;
2094	vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val);
2095
2096	//23rd: program mmUVD_JRBC_RB_CNTL to disable no_fetch
2097	reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL);
2098	reg_offset = (reg << 2);
2099	val = 0x12;
2100	vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val);
2101}
2102
2103static int vcn_v1_0_set_interrupt_state(struct amdgpu_device *adev,
2104					struct amdgpu_irq_src *source,
2105					unsigned type,
2106					enum amdgpu_interrupt_state state)
2107{
2108	return 0;
2109}
2110
2111static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev,
2112				      struct amdgpu_irq_src *source,
2113				      struct amdgpu_iv_entry *entry)
2114{
2115	DRM_DEBUG("IH: VCN TRAP\n");
2116
2117	switch (entry->src_id) {
2118	case 124:
2119		amdgpu_fence_process(&adev->vcn.inst->ring_dec);
2120		break;
2121	case 119:
2122		amdgpu_fence_process(&adev->vcn.inst->ring_enc[0]);
2123		break;
2124	case 120:
2125		amdgpu_fence_process(&adev->vcn.inst->ring_enc[1]);
2126		break;
2127	case 126:
2128		amdgpu_fence_process(&adev->vcn.inst->ring_jpeg);
2129		break;
2130	default:
2131		DRM_ERROR("Unhandled interrupt: %d %d\n",
2132			  entry->src_id, entry->src_data[0]);
2133		break;
2134	}
2135
2136	return 0;
2137}
2138
2139static void vcn_v1_0_dec_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
2140{
 
2141	struct amdgpu_device *adev = ring->adev;
2142	int i;
2143
2144	WARN_ON(ring->wptr % 2 || count % 2);
 
2145
2146	for (i = 0; i < count / 2; i++) {
2147		amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0));
2148		amdgpu_ring_write(ring, 0);
2149	}
2150}
2151
2152static int vcn_v1_0_set_powergating_state(void *handle,
2153					  enum amd_powergating_state state)
2154{
2155	/* This doesn't actually powergate the VCN block.
2156	 * That's done in the dpm code via the SMC.  This
2157	 * just re-inits the block as necessary.  The actual
2158	 * gating still happens in the dpm code.  We should
2159	 * revisit this when there is a cleaner line between
2160	 * the smc and the hw blocks
2161	 */
2162	int ret;
2163	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
2164
2165	if(state == adev->vcn.cur_state)
2166		return 0;
2167
2168	if (state == AMD_PG_STATE_GATE)
2169		ret = vcn_v1_0_stop(adev);
2170	else
2171		ret = vcn_v1_0_start(adev);
2172
2173	if(!ret)
2174		adev->vcn.cur_state = state;
2175	return ret;
2176}
2177
2178static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
2179	.name = "vcn_v1_0",
2180	.early_init = vcn_v1_0_early_init,
2181	.late_init = NULL,
2182	.sw_init = vcn_v1_0_sw_init,
2183	.sw_fini = vcn_v1_0_sw_fini,
2184	.hw_init = vcn_v1_0_hw_init,
2185	.hw_fini = vcn_v1_0_hw_fini,
2186	.suspend = vcn_v1_0_suspend,
2187	.resume = vcn_v1_0_resume,
2188	.is_idle = vcn_v1_0_is_idle,
2189	.wait_for_idle = vcn_v1_0_wait_for_idle,
2190	.check_soft_reset = NULL /* vcn_v1_0_check_soft_reset */,
2191	.pre_soft_reset = NULL /* vcn_v1_0_pre_soft_reset */,
2192	.soft_reset = NULL /* vcn_v1_0_soft_reset */,
2193	.post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */,
2194	.set_clockgating_state = vcn_v1_0_set_clockgating_state,
2195	.set_powergating_state = vcn_v1_0_set_powergating_state,
2196};
2197
2198static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = {
2199	.type = AMDGPU_RING_TYPE_VCN_DEC,
2200	.align_mask = 0xf,
 
2201	.support_64bit_ptrs = false,
2202	.no_user_fence = true,
2203	.vmhub = AMDGPU_MMHUB_0,
2204	.get_rptr = vcn_v1_0_dec_ring_get_rptr,
2205	.get_wptr = vcn_v1_0_dec_ring_get_wptr,
2206	.set_wptr = vcn_v1_0_dec_ring_set_wptr,
2207	.emit_frame_size =
2208		6 + 6 + /* hdp invalidate / flush */
2209		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
2210		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
2211		8 + /* vcn_v1_0_dec_ring_emit_vm_flush */
2212		14 + 14 + /* vcn_v1_0_dec_ring_emit_fence x2 vm fence */
2213		6,
2214	.emit_ib_size = 8, /* vcn_v1_0_dec_ring_emit_ib */
2215	.emit_ib = vcn_v1_0_dec_ring_emit_ib,
2216	.emit_fence = vcn_v1_0_dec_ring_emit_fence,
2217	.emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush,
2218	.test_ring = amdgpu_vcn_dec_ring_test_ring,
2219	.test_ib = amdgpu_vcn_dec_ring_test_ib,
2220	.insert_nop = vcn_v1_0_dec_ring_insert_nop,
2221	.insert_start = vcn_v1_0_dec_ring_insert_start,
2222	.insert_end = vcn_v1_0_dec_ring_insert_end,
2223	.pad_ib = amdgpu_ring_generic_pad_ib,
2224	.begin_use = amdgpu_vcn_ring_begin_use,
2225	.end_use = amdgpu_vcn_ring_end_use,
2226	.emit_wreg = vcn_v1_0_dec_ring_emit_wreg,
2227	.emit_reg_wait = vcn_v1_0_dec_ring_emit_reg_wait,
2228	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
2229};
2230
2231static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = {
2232	.type = AMDGPU_RING_TYPE_VCN_ENC,
2233	.align_mask = 0x3f,
2234	.nop = VCN_ENC_CMD_NO_OP,
2235	.support_64bit_ptrs = false,
2236	.no_user_fence = true,
2237	.vmhub = AMDGPU_MMHUB_0,
2238	.get_rptr = vcn_v1_0_enc_ring_get_rptr,
2239	.get_wptr = vcn_v1_0_enc_ring_get_wptr,
2240	.set_wptr = vcn_v1_0_enc_ring_set_wptr,
2241	.emit_frame_size =
2242		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
2243		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
2244		4 + /* vcn_v1_0_enc_ring_emit_vm_flush */
2245		5 + 5 + /* vcn_v1_0_enc_ring_emit_fence x2 vm fence */
2246		1, /* vcn_v1_0_enc_ring_insert_end */
2247	.emit_ib_size = 5, /* vcn_v1_0_enc_ring_emit_ib */
2248	.emit_ib = vcn_v1_0_enc_ring_emit_ib,
2249	.emit_fence = vcn_v1_0_enc_ring_emit_fence,
2250	.emit_vm_flush = vcn_v1_0_enc_ring_emit_vm_flush,
2251	.test_ring = amdgpu_vcn_enc_ring_test_ring,
2252	.test_ib = amdgpu_vcn_enc_ring_test_ib,
2253	.insert_nop = amdgpu_ring_insert_nop,
2254	.insert_end = vcn_v1_0_enc_ring_insert_end,
2255	.pad_ib = amdgpu_ring_generic_pad_ib,
2256	.begin_use = amdgpu_vcn_ring_begin_use,
2257	.end_use = amdgpu_vcn_ring_end_use,
2258	.emit_wreg = vcn_v1_0_enc_ring_emit_wreg,
2259	.emit_reg_wait = vcn_v1_0_enc_ring_emit_reg_wait,
2260	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
2261};
2262
2263static const struct amdgpu_ring_funcs vcn_v1_0_jpeg_ring_vm_funcs = {
2264	.type = AMDGPU_RING_TYPE_VCN_JPEG,
2265	.align_mask = 0xf,
2266	.nop = PACKET0(0x81ff, 0),
2267	.support_64bit_ptrs = false,
2268	.no_user_fence = true,
2269	.vmhub = AMDGPU_MMHUB_0,
2270	.extra_dw = 64,
2271	.get_rptr = vcn_v1_0_jpeg_ring_get_rptr,
2272	.get_wptr = vcn_v1_0_jpeg_ring_get_wptr,
2273	.set_wptr = vcn_v1_0_jpeg_ring_set_wptr,
2274	.emit_frame_size =
2275		6 + 6 + /* hdp invalidate / flush */
2276		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
2277		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
2278		8 + /* vcn_v1_0_jpeg_ring_emit_vm_flush */
2279		26 + 26 + /* vcn_v1_0_jpeg_ring_emit_fence x2 vm fence */
2280		6,
2281	.emit_ib_size = 22, /* vcn_v1_0_jpeg_ring_emit_ib */
2282	.emit_ib = vcn_v1_0_jpeg_ring_emit_ib,
2283	.emit_fence = vcn_v1_0_jpeg_ring_emit_fence,
2284	.emit_vm_flush = vcn_v1_0_jpeg_ring_emit_vm_flush,
2285	.test_ring = amdgpu_vcn_jpeg_ring_test_ring,
2286	.test_ib = amdgpu_vcn_jpeg_ring_test_ib,
2287	.insert_nop = vcn_v1_0_jpeg_ring_nop,
2288	.insert_start = vcn_v1_0_jpeg_ring_insert_start,
2289	.insert_end = vcn_v1_0_jpeg_ring_insert_end,
2290	.pad_ib = amdgpu_ring_generic_pad_ib,
2291	.begin_use = amdgpu_vcn_ring_begin_use,
2292	.end_use = amdgpu_vcn_ring_end_use,
2293	.emit_wreg = vcn_v1_0_jpeg_ring_emit_wreg,
2294	.emit_reg_wait = vcn_v1_0_jpeg_ring_emit_reg_wait,
2295	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
2296};
2297
2298static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev)
2299{
2300	adev->vcn.inst->ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs;
2301	DRM_INFO("VCN decode is enabled in VM mode\n");
2302}
2303
2304static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev)
2305{
2306	int i;
2307
2308	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
2309		adev->vcn.inst->ring_enc[i].funcs = &vcn_v1_0_enc_ring_vm_funcs;
2310
2311	DRM_INFO("VCN encode is enabled in VM mode\n");
2312}
2313
2314static void vcn_v1_0_set_jpeg_ring_funcs(struct amdgpu_device *adev)
2315{
2316	adev->vcn.inst->ring_jpeg.funcs = &vcn_v1_0_jpeg_ring_vm_funcs;
2317	DRM_INFO("VCN jpeg decode is enabled in VM mode\n");
2318}
2319
2320static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = {
2321	.set = vcn_v1_0_set_interrupt_state,
2322	.process = vcn_v1_0_process_interrupt,
2323};
2324
2325static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev)
2326{
2327	adev->vcn.inst->irq.num_types = adev->vcn.num_enc_rings + 2;
2328	adev->vcn.inst->irq.funcs = &vcn_v1_0_irq_funcs;
2329}
2330
2331const struct amdgpu_ip_block_version vcn_v1_0_ip_block =
2332{
2333		.type = AMD_IP_BLOCK_TYPE_VCN,
2334		.major = 1,
2335		.minor = 0,
2336		.rev = 0,
2337		.funcs = &vcn_v1_0_ip_funcs,
2338};