xref: /linux/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c (revision 7f71507851fc7764b36a3221839607d3a45c2025)
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_cs.h"
28 #include "amdgpu_vcn.h"
29 #include "amdgpu_pm.h"
30 #include "soc15.h"
31 #include "soc15d.h"
32 #include "soc15_common.h"
33 
34 #include "vcn/vcn_1_0_offset.h"
35 #include "vcn/vcn_1_0_sh_mask.h"
36 #include "mmhub/mmhub_9_1_offset.h"
37 #include "mmhub/mmhub_9_1_sh_mask.h"
38 
39 #include "ivsrcid/vcn/irqsrcs_vcn_1_0.h"
40 #include "jpeg_v1_0.h"
41 #include "vcn_v1_0.h"
42 
43 #define mmUVD_RBC_XX_IB_REG_CHECK_1_0		0x05ab
44 #define mmUVD_RBC_XX_IB_REG_CHECK_1_0_BASE_IDX	1
45 #define mmUVD_REG_XX_MASK_1_0			0x05ac
46 #define mmUVD_REG_XX_MASK_1_0_BASE_IDX		1
47 
48 static const struct amdgpu_hwip_reg_entry vcn_reg_list_1_0[] = {
49 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_POWER_STATUS),
50 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_STATUS),
51 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID),
52 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_CONTEXT_ID2),
53 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA0),
54 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_DATA1),
55 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_GPCOM_VCPU_CMD),
56 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI),
57 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO),
58 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI2),
59 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO2),
60 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI3),
61 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO3),
62 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_HI4),
63 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_BASE_LO4),
64 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR),
65 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR),
66 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR2),
67 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR2),
68 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR3),
69 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR3),
70 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_RPTR4),
71 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_WPTR4),
72 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE),
73 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE2),
74 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE3),
75 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_RB_SIZE4),
76 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_CONFIG),
77 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_PGFSM_STATUS),
78 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_CTL),
79 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_DATA),
80 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_LMA_MASK),
81 	SOC15_REG_ENTRY_STR(VCN, 0, mmUVD_DPG_PAUSE)
82 };
83 
84 static int vcn_v1_0_stop(struct amdgpu_device *adev);
85 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev);
86 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev);
87 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev);
88 static int vcn_v1_0_set_powergating_state(void *handle, enum amd_powergating_state state);
89 static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
90 				int inst_idx, struct dpg_pause_state *new_state);
91 
92 static void vcn_v1_0_idle_work_handler(struct work_struct *work);
93 static void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring);
94 
95 /**
96  * vcn_v1_0_early_init - set function pointers and load microcode
97  *
98  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
99  *
100  * Set ring and irq function pointers
101  * Load microcode from filesystem
102  */
103 static int vcn_v1_0_early_init(struct amdgpu_ip_block *ip_block)
104 {
105 	struct amdgpu_device *adev = ip_block->adev;
106 
107 	adev->vcn.num_enc_rings = 2;
108 
109 	vcn_v1_0_set_dec_ring_funcs(adev);
110 	vcn_v1_0_set_enc_ring_funcs(adev);
111 	vcn_v1_0_set_irq_funcs(adev);
112 
113 	jpeg_v1_0_early_init(ip_block);
114 
115 	return amdgpu_vcn_early_init(adev);
116 }
117 
118 /**
119  * vcn_v1_0_sw_init - sw init for VCN block
120  *
121  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
122  *
123  * Load firmware and sw initialization
124  */
125 static int vcn_v1_0_sw_init(struct amdgpu_ip_block *ip_block)
126 {
127 	struct amdgpu_ring *ring;
128 	int i, r;
129 	uint32_t reg_count = ARRAY_SIZE(vcn_reg_list_1_0);
130 	uint32_t *ptr;
131 	struct amdgpu_device *adev = ip_block->adev;
132 
133 	/* VCN DEC TRAP */
134 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
135 			VCN_1_0__SRCID__UVD_SYSTEM_MESSAGE_INTERRUPT, &adev->vcn.inst->irq);
136 	if (r)
137 		return r;
138 
139 	/* VCN ENC TRAP */
140 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
141 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, i + VCN_1_0__SRCID__UVD_ENC_GENERAL_PURPOSE,
142 					&adev->vcn.inst->irq);
143 		if (r)
144 			return r;
145 	}
146 
147 	r = amdgpu_vcn_sw_init(adev);
148 	if (r)
149 		return r;
150 
151 	/* Override the work func */
152 	adev->vcn.idle_work.work.func = vcn_v1_0_idle_work_handler;
153 
154 	amdgpu_vcn_setup_ucode(adev);
155 
156 	r = amdgpu_vcn_resume(adev);
157 	if (r)
158 		return r;
159 
160 	ring = &adev->vcn.inst->ring_dec;
161 	ring->vm_hub = AMDGPU_MMHUB0(0);
162 	sprintf(ring->name, "vcn_dec");
163 	r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
164 			     AMDGPU_RING_PRIO_DEFAULT, NULL);
165 	if (r)
166 		return r;
167 
168 	adev->vcn.internal.scratch9 = adev->vcn.inst->external.scratch9 =
169 		SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9);
170 	adev->vcn.internal.data0 = adev->vcn.inst->external.data0 =
171 		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0);
172 	adev->vcn.internal.data1 = adev->vcn.inst->external.data1 =
173 		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1);
174 	adev->vcn.internal.cmd = adev->vcn.inst->external.cmd =
175 		SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD);
176 	adev->vcn.internal.nop = adev->vcn.inst->external.nop =
177 		SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP);
178 
179 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
180 		enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
181 
182 		ring = &adev->vcn.inst->ring_enc[i];
183 		ring->vm_hub = AMDGPU_MMHUB0(0);
184 		sprintf(ring->name, "vcn_enc%d", i);
185 		r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
186 				     hw_prio, NULL);
187 		if (r)
188 			return r;
189 	}
190 
191 	adev->vcn.pause_dpg_mode = vcn_v1_0_pause_dpg_mode;
192 
193 	if (amdgpu_vcnfw_log) {
194 		volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst->fw_shared.cpu_addr;
195 
196 		fw_shared->present_flag_0 = 0;
197 		amdgpu_vcn_fwlog_init(adev->vcn.inst);
198 	}
199 
200 	r = jpeg_v1_0_sw_init(ip_block);
201 
202 	/* Allocate memory for VCN IP Dump buffer */
203 	ptr = kcalloc(adev->vcn.num_vcn_inst * reg_count, sizeof(uint32_t), GFP_KERNEL);
204 	if (!ptr) {
205 		DRM_ERROR("Failed to allocate memory for VCN IP Dump\n");
206 		adev->vcn.ip_dump = NULL;
207 	} else {
208 		adev->vcn.ip_dump = ptr;
209 	}
210 	return r;
211 }
212 
213 /**
214  * vcn_v1_0_sw_fini - sw fini for VCN block
215  *
216  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
217  *
218  * VCN suspend and free up sw allocation
219  */
220 static int vcn_v1_0_sw_fini(struct amdgpu_ip_block *ip_block)
221 {
222 	int r;
223 	struct amdgpu_device *adev = ip_block->adev;
224 
225 	r = amdgpu_vcn_suspend(adev);
226 	if (r)
227 		return r;
228 
229 	jpeg_v1_0_sw_fini(ip_block);
230 
231 	r = amdgpu_vcn_sw_fini(adev);
232 
233 	kfree(adev->vcn.ip_dump);
234 
235 	return r;
236 }
237 
238 /**
239  * vcn_v1_0_hw_init - start and test VCN block
240  *
241  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
242  *
243  * Initialize the hardware, boot up the VCPU and do some testing
244  */
245 static int vcn_v1_0_hw_init(struct amdgpu_ip_block *ip_block)
246 {
247 	struct amdgpu_device *adev = ip_block->adev;
248 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
249 	int i, r;
250 
251 	r = amdgpu_ring_test_helper(ring);
252 	if (r)
253 		return r;
254 
255 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
256 		ring = &adev->vcn.inst->ring_enc[i];
257 		r = amdgpu_ring_test_helper(ring);
258 		if (r)
259 			return r;
260 	}
261 
262 	ring = adev->jpeg.inst->ring_dec;
263 	r = amdgpu_ring_test_helper(ring);
264 
265 	return r;
266 }
267 
268 /**
269  * vcn_v1_0_hw_fini - stop the hardware block
270  *
271  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
272  *
273  * Stop the VCN block, mark ring as not ready any more
274  */
275 static int vcn_v1_0_hw_fini(struct amdgpu_ip_block *ip_block)
276 {
277 	struct amdgpu_device *adev = ip_block->adev;
278 
279 	cancel_delayed_work_sync(&adev->vcn.idle_work);
280 
281 	if ((adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ||
282 		(adev->vcn.cur_state != AMD_PG_STATE_GATE &&
283 		 RREG32_SOC15(VCN, 0, mmUVD_STATUS))) {
284 		vcn_v1_0_set_powergating_state(adev, AMD_PG_STATE_GATE);
285 	}
286 
287 	return 0;
288 }
289 
290 /**
291  * vcn_v1_0_suspend - suspend VCN block
292  *
293  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
294  *
295  * HW fini and suspend VCN block
296  */
297 static int vcn_v1_0_suspend(struct amdgpu_ip_block *ip_block)
298 {
299 	int r;
300 	struct amdgpu_device *adev = ip_block->adev;
301 	bool idle_work_unexecuted;
302 
303 	idle_work_unexecuted = cancel_delayed_work_sync(&adev->vcn.idle_work);
304 	if (idle_work_unexecuted) {
305 		if (adev->pm.dpm_enabled)
306 			amdgpu_dpm_enable_uvd(adev, false);
307 	}
308 
309 	r = vcn_v1_0_hw_fini(ip_block);
310 	if (r)
311 		return r;
312 
313 	r = amdgpu_vcn_suspend(adev);
314 
315 	return r;
316 }
317 
318 /**
319  * vcn_v1_0_resume - resume VCN block
320  *
321  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
322  *
323  * Resume firmware and hw init VCN block
324  */
325 static int vcn_v1_0_resume(struct amdgpu_ip_block *ip_block)
326 {
327 	int r;
328 
329 	r = amdgpu_vcn_resume(ip_block->adev);
330 	if (r)
331 		return r;
332 
333 	r = vcn_v1_0_hw_init(ip_block);
334 
335 	return r;
336 }
337 
338 /**
339  * vcn_v1_0_mc_resume_spg_mode - memory controller programming
340  *
341  * @adev: amdgpu_device pointer
342  *
343  * Let the VCN memory controller know it's offsets
344  */
345 static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev)
346 {
347 	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
348 	uint32_t offset;
349 
350 	/* cache window 0: fw */
351 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
352 		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
353 			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo));
354 		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
355 			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi));
356 		WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0);
357 		offset = 0;
358 	} else {
359 		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
360 			lower_32_bits(adev->vcn.inst->gpu_addr));
361 		WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
362 			upper_32_bits(adev->vcn.inst->gpu_addr));
363 		offset = size;
364 		WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0,
365 			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
366 	}
367 
368 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size);
369 
370 	/* cache window 1: stack */
371 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
372 		     lower_32_bits(adev->vcn.inst->gpu_addr + offset));
373 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
374 		     upper_32_bits(adev->vcn.inst->gpu_addr + offset));
375 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0);
376 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE);
377 
378 	/* cache window 2: context */
379 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
380 		     lower_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
381 	WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
382 		     upper_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE));
383 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0);
384 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE);
385 
386 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_ADDR_CONFIG,
387 			adev->gfx.config.gb_addr_config);
388 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG,
389 			adev->gfx.config.gb_addr_config);
390 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG,
391 			adev->gfx.config.gb_addr_config);
392 	WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG,
393 			adev->gfx.config.gb_addr_config);
394 	WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG,
395 			adev->gfx.config.gb_addr_config);
396 	WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG,
397 			adev->gfx.config.gb_addr_config);
398 	WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG,
399 			adev->gfx.config.gb_addr_config);
400 	WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG,
401 			adev->gfx.config.gb_addr_config);
402 	WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG,
403 			adev->gfx.config.gb_addr_config);
404 	WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG,
405 			adev->gfx.config.gb_addr_config);
406 	WREG32_SOC15(UVD, 0, mmUVD_JPEG_ADDR_CONFIG,
407 			adev->gfx.config.gb_addr_config);
408 	WREG32_SOC15(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG,
409 			adev->gfx.config.gb_addr_config);
410 }
411 
412 static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev)
413 {
414 	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
415 	uint32_t offset;
416 
417 	/* cache window 0: fw */
418 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
419 		WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
420 			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo),
421 			     0xFFFFFFFF, 0);
422 		WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
423 			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi),
424 			     0xFFFFFFFF, 0);
425 		WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0,
426 			     0xFFFFFFFF, 0);
427 		offset = 0;
428 	} else {
429 		WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
430 			lower_32_bits(adev->vcn.inst->gpu_addr), 0xFFFFFFFF, 0);
431 		WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
432 			upper_32_bits(adev->vcn.inst->gpu_addr), 0xFFFFFFFF, 0);
433 		offset = size;
434 		WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0,
435 			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0xFFFFFFFF, 0);
436 	}
437 
438 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size, 0xFFFFFFFF, 0);
439 
440 	/* cache window 1: stack */
441 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
442 		     lower_32_bits(adev->vcn.inst->gpu_addr + offset), 0xFFFFFFFF, 0);
443 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
444 		     upper_32_bits(adev->vcn.inst->gpu_addr + offset), 0xFFFFFFFF, 0);
445 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0,
446 			     0xFFFFFFFF, 0);
447 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE,
448 			     0xFFFFFFFF, 0);
449 
450 	/* cache window 2: context */
451 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
452 		     lower_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE),
453 			     0xFFFFFFFF, 0);
454 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
455 		     upper_32_bits(adev->vcn.inst->gpu_addr + offset + AMDGPU_VCN_STACK_SIZE),
456 			     0xFFFFFFFF, 0);
457 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0, 0xFFFFFFFF, 0);
458 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE,
459 			     0xFFFFFFFF, 0);
460 
461 	/* VCN global tiling registers */
462 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_UDEC_ADDR_CONFIG,
463 			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
464 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG,
465 			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
466 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG,
467 			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
468 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG,
469 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
470 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG,
471 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
472 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG,
473 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
474 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG,
475 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
476 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG,
477 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
478 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG,
479 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
480 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG,
481 		adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
482 }
483 
484 /**
485  * vcn_v1_0_disable_clock_gating - disable VCN clock gating
486  *
487  * @adev: amdgpu_device pointer
488  *
489  * Disable clock gating for VCN block
490  */
491 static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev)
492 {
493 	uint32_t data;
494 
495 	/* JPEG disable CGC */
496 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL);
497 
498 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
499 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
500 	else
501 		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE_MASK;
502 
503 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
504 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
505 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data);
506 
507 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE);
508 	data &= ~(JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK);
509 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data);
510 
511 	/* UVD disable CGC */
512 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
513 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
514 		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
515 	else
516 		data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
517 
518 	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
519 	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
520 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
521 
522 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_GATE);
523 	data &= ~(UVD_CGC_GATE__SYS_MASK
524 		| UVD_CGC_GATE__UDEC_MASK
525 		| UVD_CGC_GATE__MPEG2_MASK
526 		| UVD_CGC_GATE__REGS_MASK
527 		| UVD_CGC_GATE__RBC_MASK
528 		| UVD_CGC_GATE__LMI_MC_MASK
529 		| UVD_CGC_GATE__LMI_UMC_MASK
530 		| UVD_CGC_GATE__IDCT_MASK
531 		| UVD_CGC_GATE__MPRD_MASK
532 		| UVD_CGC_GATE__MPC_MASK
533 		| UVD_CGC_GATE__LBSI_MASK
534 		| UVD_CGC_GATE__LRBBM_MASK
535 		| UVD_CGC_GATE__UDEC_RE_MASK
536 		| UVD_CGC_GATE__UDEC_CM_MASK
537 		| UVD_CGC_GATE__UDEC_IT_MASK
538 		| UVD_CGC_GATE__UDEC_DB_MASK
539 		| UVD_CGC_GATE__UDEC_MP_MASK
540 		| UVD_CGC_GATE__WCB_MASK
541 		| UVD_CGC_GATE__VCPU_MASK
542 		| UVD_CGC_GATE__SCPU_MASK);
543 	WREG32_SOC15(VCN, 0, mmUVD_CGC_GATE, data);
544 
545 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
546 	data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK
547 		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
548 		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
549 		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
550 		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
551 		| UVD_CGC_CTRL__SYS_MODE_MASK
552 		| UVD_CGC_CTRL__UDEC_MODE_MASK
553 		| UVD_CGC_CTRL__MPEG2_MODE_MASK
554 		| UVD_CGC_CTRL__REGS_MODE_MASK
555 		| UVD_CGC_CTRL__RBC_MODE_MASK
556 		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
557 		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
558 		| UVD_CGC_CTRL__IDCT_MODE_MASK
559 		| UVD_CGC_CTRL__MPRD_MODE_MASK
560 		| UVD_CGC_CTRL__MPC_MODE_MASK
561 		| UVD_CGC_CTRL__LBSI_MODE_MASK
562 		| UVD_CGC_CTRL__LRBBM_MODE_MASK
563 		| UVD_CGC_CTRL__WCB_MODE_MASK
564 		| UVD_CGC_CTRL__VCPU_MODE_MASK
565 		| UVD_CGC_CTRL__SCPU_MODE_MASK);
566 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
567 
568 	/* turn on */
569 	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE);
570 	data |= (UVD_SUVD_CGC_GATE__SRE_MASK
571 		| UVD_SUVD_CGC_GATE__SIT_MASK
572 		| UVD_SUVD_CGC_GATE__SMP_MASK
573 		| UVD_SUVD_CGC_GATE__SCM_MASK
574 		| UVD_SUVD_CGC_GATE__SDB_MASK
575 		| UVD_SUVD_CGC_GATE__SRE_H264_MASK
576 		| UVD_SUVD_CGC_GATE__SRE_HEVC_MASK
577 		| UVD_SUVD_CGC_GATE__SIT_H264_MASK
578 		| UVD_SUVD_CGC_GATE__SIT_HEVC_MASK
579 		| UVD_SUVD_CGC_GATE__SCM_H264_MASK
580 		| UVD_SUVD_CGC_GATE__SCM_HEVC_MASK
581 		| UVD_SUVD_CGC_GATE__SDB_H264_MASK
582 		| UVD_SUVD_CGC_GATE__SDB_HEVC_MASK
583 		| UVD_SUVD_CGC_GATE__SCLR_MASK
584 		| UVD_SUVD_CGC_GATE__UVD_SC_MASK
585 		| UVD_SUVD_CGC_GATE__ENT_MASK
586 		| UVD_SUVD_CGC_GATE__SIT_HEVC_DEC_MASK
587 		| UVD_SUVD_CGC_GATE__SIT_HEVC_ENC_MASK
588 		| UVD_SUVD_CGC_GATE__SITE_MASK
589 		| UVD_SUVD_CGC_GATE__SRE_VP9_MASK
590 		| UVD_SUVD_CGC_GATE__SCM_VP9_MASK
591 		| UVD_SUVD_CGC_GATE__SIT_VP9_DEC_MASK
592 		| UVD_SUVD_CGC_GATE__SDB_VP9_MASK
593 		| UVD_SUVD_CGC_GATE__IME_HEVC_MASK);
594 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_GATE, data);
595 
596 	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL);
597 	data &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
598 		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
599 		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
600 		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
601 		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
602 		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
603 		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
604 		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
605 		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
606 		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
607 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
608 }
609 
610 /**
611  * vcn_v1_0_enable_clock_gating - enable VCN clock gating
612  *
613  * @adev: amdgpu_device pointer
614  *
615  * Enable clock gating for VCN block
616  */
617 static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev)
618 {
619 	uint32_t data = 0;
620 
621 	/* enable JPEG CGC */
622 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL);
623 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
624 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
625 	else
626 		data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
627 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
628 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
629 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL, data);
630 
631 	data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE);
632 	data |= (JPEG_CGC_GATE__JPEG_MASK | JPEG_CGC_GATE__JPEG2_MASK);
633 	WREG32_SOC15(VCN, 0, mmJPEG_CGC_GATE, data);
634 
635 	/* enable UVD CGC */
636 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
637 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
638 		data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
639 	else
640 		data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
641 	data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
642 	data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
643 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
644 
645 	data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL);
646 	data |= (UVD_CGC_CTRL__UDEC_RE_MODE_MASK
647 		| UVD_CGC_CTRL__UDEC_CM_MODE_MASK
648 		| UVD_CGC_CTRL__UDEC_IT_MODE_MASK
649 		| UVD_CGC_CTRL__UDEC_DB_MODE_MASK
650 		| UVD_CGC_CTRL__UDEC_MP_MODE_MASK
651 		| UVD_CGC_CTRL__SYS_MODE_MASK
652 		| UVD_CGC_CTRL__UDEC_MODE_MASK
653 		| UVD_CGC_CTRL__MPEG2_MODE_MASK
654 		| UVD_CGC_CTRL__REGS_MODE_MASK
655 		| UVD_CGC_CTRL__RBC_MODE_MASK
656 		| UVD_CGC_CTRL__LMI_MC_MODE_MASK
657 		| UVD_CGC_CTRL__LMI_UMC_MODE_MASK
658 		| UVD_CGC_CTRL__IDCT_MODE_MASK
659 		| UVD_CGC_CTRL__MPRD_MODE_MASK
660 		| UVD_CGC_CTRL__MPC_MODE_MASK
661 		| UVD_CGC_CTRL__LBSI_MODE_MASK
662 		| UVD_CGC_CTRL__LRBBM_MODE_MASK
663 		| UVD_CGC_CTRL__WCB_MODE_MASK
664 		| UVD_CGC_CTRL__VCPU_MODE_MASK
665 		| UVD_CGC_CTRL__SCPU_MODE_MASK);
666 	WREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL, data);
667 
668 	data = RREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL);
669 	data |= (UVD_SUVD_CGC_CTRL__SRE_MODE_MASK
670 		| UVD_SUVD_CGC_CTRL__SIT_MODE_MASK
671 		| UVD_SUVD_CGC_CTRL__SMP_MODE_MASK
672 		| UVD_SUVD_CGC_CTRL__SCM_MODE_MASK
673 		| UVD_SUVD_CGC_CTRL__SDB_MODE_MASK
674 		| UVD_SUVD_CGC_CTRL__SCLR_MODE_MASK
675 		| UVD_SUVD_CGC_CTRL__UVD_SC_MODE_MASK
676 		| UVD_SUVD_CGC_CTRL__ENT_MODE_MASK
677 		| UVD_SUVD_CGC_CTRL__IME_MODE_MASK
678 		| UVD_SUVD_CGC_CTRL__SITE_MODE_MASK);
679 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
680 }
681 
682 static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel)
683 {
684 	uint32_t reg_data = 0;
685 
686 	/* disable JPEG CGC */
687 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
688 		reg_data = 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
689 	else
690 		reg_data = 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
691 	reg_data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
692 	reg_data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
693 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmJPEG_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
694 
695 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmJPEG_CGC_GATE, 0, 0xFFFFFFFF, sram_sel);
696 
697 	/* enable sw clock gating control */
698 	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
699 		reg_data = 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
700 	else
701 		reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
702 	reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
703 	reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
704 	reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
705 		 UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
706 		 UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
707 		 UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
708 		 UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
709 		 UVD_CGC_CTRL__SYS_MODE_MASK |
710 		 UVD_CGC_CTRL__UDEC_MODE_MASK |
711 		 UVD_CGC_CTRL__MPEG2_MODE_MASK |
712 		 UVD_CGC_CTRL__REGS_MODE_MASK |
713 		 UVD_CGC_CTRL__RBC_MODE_MASK |
714 		 UVD_CGC_CTRL__LMI_MC_MODE_MASK |
715 		 UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
716 		 UVD_CGC_CTRL__IDCT_MODE_MASK |
717 		 UVD_CGC_CTRL__MPRD_MODE_MASK |
718 		 UVD_CGC_CTRL__MPC_MODE_MASK |
719 		 UVD_CGC_CTRL__LBSI_MODE_MASK |
720 		 UVD_CGC_CTRL__LRBBM_MODE_MASK |
721 		 UVD_CGC_CTRL__WCB_MODE_MASK |
722 		 UVD_CGC_CTRL__VCPU_MODE_MASK |
723 		 UVD_CGC_CTRL__SCPU_MODE_MASK);
724 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
725 
726 	/* turn off clock gating */
727 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_CGC_GATE, 0, 0xFFFFFFFF, sram_sel);
728 
729 	/* turn on SUVD clock gating */
730 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_SUVD_CGC_GATE, 1, 0xFFFFFFFF, sram_sel);
731 
732 	/* turn on sw mode in UVD_SUVD_CGC_CTRL */
733 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_SUVD_CGC_CTRL, 0, 0xFFFFFFFF, sram_sel);
734 }
735 
736 static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev)
737 {
738 	uint32_t data = 0;
739 
740 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
741 		data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT
742 			| 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT
743 			| 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT
744 			| 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT
745 			| 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT
746 			| 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT
747 			| 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT
748 			| 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT
749 			| 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT
750 			| 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT
751 			| 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT);
752 
753 		WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data);
754 		SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON, 0xFFFFFF);
755 	} else {
756 		data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT
757 			| 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT
758 			| 1 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT
759 			| 1 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT
760 			| 1 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT
761 			| 1 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT
762 			| 1 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT
763 			| 1 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT
764 			| 1 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT
765 			| 1 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT
766 			| 1 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT);
767 		WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data);
768 		SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, 0,  0xFFFFFFFF);
769 	}
770 
771 	/* polling UVD_PGFSM_STATUS to confirm UVDM_PWR_STATUS , UVDU_PWR_STATUS are 0 (power on) */
772 
773 	data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS);
774 	data &= ~0x103;
775 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN)
776 		data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON | UVD_POWER_STATUS__UVD_PG_EN_MASK;
777 
778 	WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data);
779 }
780 
781 static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev)
782 {
783 	uint32_t data = 0;
784 
785 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN) {
786 		/* Before power off, this indicator has to be turned on */
787 		data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS);
788 		data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK;
789 		data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
790 		WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data);
791 
792 
793 		data = (2 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT
794 			| 2 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT
795 			| 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT
796 			| 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT
797 			| 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT
798 			| 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT
799 			| 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT
800 			| 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT
801 			| 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT
802 			| 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT
803 			| 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT);
804 
805 		WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data);
806 
807 		data = (2 << UVD_PGFSM_STATUS__UVDM_PWR_STATUS__SHIFT
808 			| 2 << UVD_PGFSM_STATUS__UVDU_PWR_STATUS__SHIFT
809 			| 2 << UVD_PGFSM_STATUS__UVDF_PWR_STATUS__SHIFT
810 			| 2 << UVD_PGFSM_STATUS__UVDC_PWR_STATUS__SHIFT
811 			| 2 << UVD_PGFSM_STATUS__UVDB_PWR_STATUS__SHIFT
812 			| 2 << UVD_PGFSM_STATUS__UVDIL_PWR_STATUS__SHIFT
813 			| 2 << UVD_PGFSM_STATUS__UVDIR_PWR_STATUS__SHIFT
814 			| 2 << UVD_PGFSM_STATUS__UVDTD_PWR_STATUS__SHIFT
815 			| 2 << UVD_PGFSM_STATUS__UVDTE_PWR_STATUS__SHIFT
816 			| 2 << UVD_PGFSM_STATUS__UVDE_PWR_STATUS__SHIFT
817 			| 2 << UVD_PGFSM_STATUS__UVDW_PWR_STATUS__SHIFT);
818 		SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, data, 0xFFFFFFFF);
819 	}
820 }
821 
822 /**
823  * vcn_v1_0_start_spg_mode - start VCN block
824  *
825  * @adev: amdgpu_device pointer
826  *
827  * Setup and start the VCN block
828  */
829 static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
830 {
831 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
832 	uint32_t rb_bufsz, tmp;
833 	uint32_t lmi_swap_cntl;
834 	int i, j, r;
835 
836 	/* disable byte swapping */
837 	lmi_swap_cntl = 0;
838 
839 	vcn_1_0_disable_static_power_gating(adev);
840 
841 	tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY;
842 	WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp);
843 
844 	/* disable clock gating */
845 	vcn_v1_0_disable_clock_gating(adev);
846 
847 	/* disable interupt */
848 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0,
849 			~UVD_MASTINT_EN__VCPU_EN_MASK);
850 
851 	/* initialize VCN memory controller */
852 	tmp = RREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL);
853 	WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL, tmp		|
854 		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK	|
855 		UVD_LMI_CTRL__MASK_MC_URGENT_MASK			|
856 		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK		|
857 		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK);
858 
859 #ifdef __BIG_ENDIAN
860 	/* swap (8 in 32) RB and IB */
861 	lmi_swap_cntl = 0xa;
862 #endif
863 	WREG32_SOC15(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
864 
865 	tmp = RREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL);
866 	tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK;
867 	tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT;
868 	WREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL, tmp);
869 
870 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0,
871 		((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
872 		(0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
873 		(0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
874 		(0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)));
875 
876 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0,
877 		((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
878 		(0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
879 		(0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
880 		(0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)));
881 
882 	WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX,
883 		((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
884 		(0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
885 		(0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)));
886 
887 	vcn_v1_0_mc_resume_spg_mode(adev);
888 
889 	WREG32_SOC15(UVD, 0, mmUVD_REG_XX_MASK_1_0, 0x10);
890 	WREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK_1_0,
891 		RREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK_1_0) | 0x3);
892 
893 	/* enable VCPU clock */
894 	WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK);
895 
896 	/* boot up the VCPU */
897 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0,
898 			~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
899 
900 	/* enable UMC */
901 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0,
902 			~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
903 
904 	tmp = RREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET);
905 	tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK;
906 	tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK;
907 	WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, tmp);
908 
909 	for (i = 0; i < 10; ++i) {
910 		uint32_t status;
911 
912 		for (j = 0; j < 100; ++j) {
913 			status = RREG32_SOC15(UVD, 0, mmUVD_STATUS);
914 			if (status & UVD_STATUS__IDLE)
915 				break;
916 			mdelay(10);
917 		}
918 		r = 0;
919 		if (status & UVD_STATUS__IDLE)
920 			break;
921 
922 		DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n");
923 		WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
924 				UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
925 				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
926 		mdelay(10);
927 		WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0,
928 				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
929 		mdelay(10);
930 		r = -1;
931 	}
932 
933 	if (r) {
934 		DRM_ERROR("VCN decode not responding, giving up!!!\n");
935 		return r;
936 	}
937 	/* enable master interrupt */
938 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN),
939 		UVD_MASTINT_EN__VCPU_EN_MASK, ~UVD_MASTINT_EN__VCPU_EN_MASK);
940 
941 	/* enable system interrupt for JRBC, TODO: move to set interrupt*/
942 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SYS_INT_EN),
943 		UVD_SYS_INT_EN__UVD_JRBC_EN_MASK,
944 		~UVD_SYS_INT_EN__UVD_JRBC_EN_MASK);
945 
946 	/* clear the busy bit of UVD_STATUS */
947 	tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) & ~UVD_STATUS__UVD_BUSY;
948 	WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp);
949 
950 	/* force RBC into idle state */
951 	rb_bufsz = order_base_2(ring->ring_size);
952 	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
953 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
954 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
955 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
956 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
957 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp);
958 
959 	/* set the write pointer delay */
960 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0);
961 
962 	/* set the wb address */
963 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR,
964 			(upper_32_bits(ring->gpu_addr) >> 2));
965 
966 	/* program the RB_BASE for ring buffer */
967 	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
968 			lower_32_bits(ring->gpu_addr));
969 	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
970 			upper_32_bits(ring->gpu_addr));
971 
972 	/* Initialize the ring buffer's read and write pointers */
973 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0);
974 
975 	WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0);
976 
977 	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
978 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
979 			lower_32_bits(ring->wptr));
980 
981 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0,
982 			~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
983 
984 	ring = &adev->vcn.inst->ring_enc[0];
985 	WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
986 	WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
987 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr);
988 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
989 	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4);
990 
991 	ring = &adev->vcn.inst->ring_enc[1];
992 	WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
993 	WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
994 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr);
995 	WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
996 	WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
997 
998 	jpeg_v1_0_start(adev, 0);
999 
1000 	return 0;
1001 }
1002 
1003 static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
1004 {
1005 	struct amdgpu_ring *ring = &adev->vcn.inst->ring_dec;
1006 	uint32_t rb_bufsz, tmp;
1007 	uint32_t lmi_swap_cntl;
1008 
1009 	/* disable byte swapping */
1010 	lmi_swap_cntl = 0;
1011 
1012 	vcn_1_0_enable_static_power_gating(adev);
1013 
1014 	/* enable dynamic power gating mode */
1015 	tmp = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
1016 	tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
1017 	tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
1018 	WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, tmp);
1019 
1020 	/* enable clock gating */
1021 	vcn_v1_0_clock_gating_dpg_mode(adev, 0);
1022 
1023 	/* enable VCPU clock */
1024 	tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
1025 	tmp |= UVD_VCPU_CNTL__CLK_EN_MASK;
1026 	tmp |= UVD_VCPU_CNTL__MIF_WR_LOW_THRESHOLD_BP_MASK;
1027 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_VCPU_CNTL, tmp, 0xFFFFFFFF, 0);
1028 
1029 	/* disable interupt */
1030 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MASTINT_EN,
1031 			0, UVD_MASTINT_EN__VCPU_EN_MASK, 0);
1032 
1033 	/* initialize VCN memory controller */
1034 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_CTRL,
1035 		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
1036 		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
1037 		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
1038 		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
1039 		UVD_LMI_CTRL__REQ_MODE_MASK |
1040 		UVD_LMI_CTRL__CRC_RESET_MASK |
1041 		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
1042 		0x00100000L, 0xFFFFFFFF, 0);
1043 
1044 #ifdef __BIG_ENDIAN
1045 	/* swap (8 in 32) RB and IB */
1046 	lmi_swap_cntl = 0xa;
1047 #endif
1048 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl, 0xFFFFFFFF, 0);
1049 
1050 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MPC_CNTL,
1051 		0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0xFFFFFFFF, 0);
1052 
1053 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MPC_SET_MUXA0,
1054 		((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) |
1055 		 (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) |
1056 		 (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) |
1057 		 (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0xFFFFFFFF, 0);
1058 
1059 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MPC_SET_MUXB0,
1060 		((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) |
1061 		 (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) |
1062 		 (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) |
1063 		 (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0xFFFFFFFF, 0);
1064 
1065 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MPC_SET_MUX,
1066 		((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) |
1067 		 (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) |
1068 		 (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0xFFFFFFFF, 0);
1069 
1070 	vcn_v1_0_mc_resume_dpg_mode(adev);
1071 
1072 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_REG_XX_MASK, 0x10, 0xFFFFFFFF, 0);
1073 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, 0x3, 0xFFFFFFFF, 0);
1074 
1075 	/* boot up the VCPU */
1076 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_SOFT_RESET, 0, 0xFFFFFFFF, 0);
1077 
1078 	/* enable UMC */
1079 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_CTRL2,
1080 		0x1F << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT,
1081 		0xFFFFFFFF, 0);
1082 
1083 	/* enable master interrupt */
1084 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_MASTINT_EN,
1085 			UVD_MASTINT_EN__VCPU_EN_MASK, UVD_MASTINT_EN__VCPU_EN_MASK, 0);
1086 
1087 	vcn_v1_0_clock_gating_dpg_mode(adev, 1);
1088 	/* setup mmUVD_LMI_CTRL */
1089 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_LMI_CTRL,
1090 		(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
1091 		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
1092 		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
1093 		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
1094 		UVD_LMI_CTRL__REQ_MODE_MASK |
1095 		UVD_LMI_CTRL__CRC_RESET_MASK |
1096 		UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
1097 		0x00100000L, 0xFFFFFFFF, 1);
1098 
1099 	tmp = adev->gfx.config.gb_addr_config;
1100 	/* setup VCN global tiling registers */
1101 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1);
1102 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1);
1103 
1104 	/* enable System Interrupt for JRBC */
1105 	WREG32_SOC15_DPG_MODE_1_0(UVD, 0, mmUVD_SYS_INT_EN,
1106 									UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, 0xFFFFFFFF, 1);
1107 
1108 	/* force RBC into idle state */
1109 	rb_bufsz = order_base_2(ring->ring_size);
1110 	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
1111 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
1112 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
1113 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
1114 	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
1115 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp);
1116 
1117 	/* set the write pointer delay */
1118 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0);
1119 
1120 	/* set the wb address */
1121 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR,
1122 								(upper_32_bits(ring->gpu_addr) >> 2));
1123 
1124 	/* program the RB_BASE for ring buffer */
1125 	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
1126 								lower_32_bits(ring->gpu_addr));
1127 	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
1128 								upper_32_bits(ring->gpu_addr));
1129 
1130 	/* Initialize the ring buffer's read and write pointers */
1131 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0);
1132 
1133 	WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0);
1134 
1135 	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
1136 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
1137 								lower_32_bits(ring->wptr));
1138 
1139 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0,
1140 			~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
1141 
1142 	jpeg_v1_0_start(adev, 1);
1143 
1144 	return 0;
1145 }
1146 
1147 static int vcn_v1_0_start(struct amdgpu_device *adev)
1148 {
1149 	return (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) ?
1150 		vcn_v1_0_start_dpg_mode(adev) : vcn_v1_0_start_spg_mode(adev);
1151 }
1152 
1153 /**
1154  * vcn_v1_0_stop_spg_mode - stop VCN block
1155  *
1156  * @adev: amdgpu_device pointer
1157  *
1158  * stop the VCN block
1159  */
1160 static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev)
1161 {
1162 	int tmp;
1163 
1164 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7);
1165 
1166 	tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK |
1167 		UVD_LMI_STATUS__READ_CLEAN_MASK |
1168 		UVD_LMI_STATUS__WRITE_CLEAN_MASK |
1169 		UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK;
1170 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp);
1171 
1172 	/* stall UMC channel */
1173 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2),
1174 		UVD_LMI_CTRL2__STALL_ARB_UMC_MASK,
1175 		~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
1176 
1177 	tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK |
1178 		UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK;
1179 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp);
1180 
1181 	/* disable VCPU clock */
1182 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL), 0,
1183 		~UVD_VCPU_CNTL__CLK_EN_MASK);
1184 
1185 	/* reset LMI UMC/LMI */
1186 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
1187 		UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK,
1188 		~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK);
1189 
1190 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
1191 		UVD_SOFT_RESET__LMI_SOFT_RESET_MASK,
1192 		~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK);
1193 
1194 	/* put VCPU into reset */
1195 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
1196 		UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
1197 		~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
1198 
1199 	WREG32_SOC15(UVD, 0, mmUVD_STATUS, 0);
1200 
1201 	vcn_v1_0_enable_clock_gating(adev);
1202 	vcn_1_0_enable_static_power_gating(adev);
1203 	return 0;
1204 }
1205 
1206 static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev)
1207 {
1208 	uint32_t tmp;
1209 
1210 	/* Wait for power status to be UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF */
1211 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1212 			UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1213 			UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
1214 
1215 	/* wait for read ptr to be equal to write ptr */
1216 	tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR);
1217 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR, tmp, 0xFFFFFFFF);
1218 
1219 	tmp = RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2);
1220 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RB_RPTR2, tmp, 0xFFFFFFFF);
1221 
1222 	tmp = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR);
1223 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_JRBC_RB_RPTR, tmp, 0xFFFFFFFF);
1224 
1225 	tmp = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR) & 0x7FFFFFFF;
1226 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RBC_RB_RPTR, tmp, 0xFFFFFFFF);
1227 
1228 	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1229 		UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1230 		UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
1231 
1232 	/* disable dynamic power gating mode */
1233 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0,
1234 			~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
1235 
1236 	return 0;
1237 }
1238 
1239 static int vcn_v1_0_stop(struct amdgpu_device *adev)
1240 {
1241 	int r;
1242 
1243 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
1244 		r = vcn_v1_0_stop_dpg_mode(adev);
1245 	else
1246 		r = vcn_v1_0_stop_spg_mode(adev);
1247 
1248 	return r;
1249 }
1250 
1251 static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev,
1252 				int inst_idx, struct dpg_pause_state *new_state)
1253 {
1254 	int ret_code;
1255 	uint32_t reg_data = 0;
1256 	uint32_t reg_data2 = 0;
1257 	struct amdgpu_ring *ring;
1258 
1259 	/* pause/unpause if state is changed */
1260 	if (adev->vcn.inst[inst_idx].pause_state.fw_based != new_state->fw_based) {
1261 		DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d",
1262 			adev->vcn.inst[inst_idx].pause_state.fw_based,
1263 			adev->vcn.inst[inst_idx].pause_state.jpeg,
1264 			new_state->fw_based, new_state->jpeg);
1265 
1266 		reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) &
1267 			(~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
1268 
1269 		if (new_state->fw_based == VCN_DPG_STATE__PAUSE) {
1270 			ret_code = 0;
1271 
1272 			if (!(reg_data & UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK))
1273 				ret_code = SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1274 						   UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1275 						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
1276 
1277 			if (!ret_code) {
1278 				/* pause DPG non-jpeg */
1279 				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
1280 				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1281 				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE,
1282 						   UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK,
1283 						   UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
1284 
1285 				/* Restore */
1286 				ring = &adev->vcn.inst->ring_enc[0];
1287 				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr);
1288 				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
1289 				WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4);
1290 				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
1291 				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
1292 
1293 				ring = &adev->vcn.inst->ring_enc[1];
1294 				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr);
1295 				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
1296 				WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
1297 				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
1298 				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
1299 
1300 				ring = &adev->vcn.inst->ring_dec;
1301 				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
1302 						   RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF);
1303 				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1304 						   UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
1305 						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
1306 			}
1307 		} else {
1308 			/* unpause dpg non-jpeg, no need to wait */
1309 			reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
1310 			WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1311 		}
1312 		adev->vcn.inst[inst_idx].pause_state.fw_based = new_state->fw_based;
1313 	}
1314 
1315 	/* pause/unpause if state is changed */
1316 	if (adev->vcn.inst[inst_idx].pause_state.jpeg != new_state->jpeg) {
1317 		DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d",
1318 			adev->vcn.inst[inst_idx].pause_state.fw_based,
1319 			adev->vcn.inst[inst_idx].pause_state.jpeg,
1320 			new_state->fw_based, new_state->jpeg);
1321 
1322 		reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) &
1323 			(~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK);
1324 
1325 		if (new_state->jpeg == VCN_DPG_STATE__PAUSE) {
1326 			ret_code = 0;
1327 
1328 			if (!(reg_data & UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK))
1329 				ret_code = SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1330 						   UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
1331 						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
1332 
1333 			if (!ret_code) {
1334 				/* Make sure JPRG Snoop is disabled before sending the pause */
1335 				reg_data2 = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
1336 				reg_data2 |= UVD_POWER_STATUS__JRBC_SNOOP_DIS_MASK;
1337 				WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, reg_data2);
1338 
1339 				/* pause DPG jpeg */
1340 				reg_data |= UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK;
1341 				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1342 				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE,
1343 							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK,
1344 							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK);
1345 
1346 				/* Restore */
1347 				ring = adev->jpeg.inst->ring_dec;
1348 				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
1349 				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL,
1350 							UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK |
1351 							UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
1352 				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
1353 							lower_32_bits(ring->gpu_addr));
1354 				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
1355 							upper_32_bits(ring->gpu_addr));
1356 				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, ring->wptr);
1357 				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, ring->wptr);
1358 				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL,
1359 							UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK);
1360 
1361 				ring = &adev->vcn.inst->ring_dec;
1362 				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
1363 						   RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF);
1364 				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
1365 						   UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
1366 						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK);
1367 			}
1368 		} else {
1369 			/* unpause dpg jpeg, no need to wait */
1370 			reg_data &= ~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK;
1371 			WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
1372 		}
1373 		adev->vcn.inst[inst_idx].pause_state.jpeg = new_state->jpeg;
1374 	}
1375 
1376 	return 0;
1377 }
1378 
1379 static bool vcn_v1_0_is_idle(void *handle)
1380 {
1381 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
1382 
1383 	return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == UVD_STATUS__IDLE);
1384 }
1385 
1386 static int vcn_v1_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
1387 {
1388 	struct amdgpu_device *adev = ip_block->adev;
1389 	int ret;
1390 
1391 	ret = SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, UVD_STATUS__IDLE,
1392 		UVD_STATUS__IDLE);
1393 
1394 	return ret;
1395 }
1396 
1397 static int vcn_v1_0_set_clockgating_state(void *handle,
1398 					  enum amd_clockgating_state state)
1399 {
1400 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
1401 	bool enable = (state == AMD_CG_STATE_GATE);
1402 
1403 	if (enable) {
1404 		/* wait for STATUS to clear */
1405 		if (!vcn_v1_0_is_idle(handle))
1406 			return -EBUSY;
1407 		vcn_v1_0_enable_clock_gating(adev);
1408 	} else {
1409 		/* disable HW gating and enable Sw gating */
1410 		vcn_v1_0_disable_clock_gating(adev);
1411 	}
1412 	return 0;
1413 }
1414 
1415 /**
1416  * vcn_v1_0_dec_ring_get_rptr - get read pointer
1417  *
1418  * @ring: amdgpu_ring pointer
1419  *
1420  * Returns the current hardware read pointer
1421  */
1422 static uint64_t vcn_v1_0_dec_ring_get_rptr(struct amdgpu_ring *ring)
1423 {
1424 	struct amdgpu_device *adev = ring->adev;
1425 
1426 	return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
1427 }
1428 
1429 /**
1430  * vcn_v1_0_dec_ring_get_wptr - get write pointer
1431  *
1432  * @ring: amdgpu_ring pointer
1433  *
1434  * Returns the current hardware write pointer
1435  */
1436 static uint64_t vcn_v1_0_dec_ring_get_wptr(struct amdgpu_ring *ring)
1437 {
1438 	struct amdgpu_device *adev = ring->adev;
1439 
1440 	return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR);
1441 }
1442 
1443 /**
1444  * vcn_v1_0_dec_ring_set_wptr - set write pointer
1445  *
1446  * @ring: amdgpu_ring pointer
1447  *
1448  * Commits the write pointer to the hardware
1449  */
1450 static void vcn_v1_0_dec_ring_set_wptr(struct amdgpu_ring *ring)
1451 {
1452 	struct amdgpu_device *adev = ring->adev;
1453 
1454 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
1455 		WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2,
1456 			lower_32_bits(ring->wptr) | 0x80000000);
1457 
1458 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
1459 }
1460 
1461 /**
1462  * vcn_v1_0_dec_ring_insert_start - insert a start command
1463  *
1464  * @ring: amdgpu_ring pointer
1465  *
1466  * Write a start command to the ring.
1467  */
1468 static void vcn_v1_0_dec_ring_insert_start(struct amdgpu_ring *ring)
1469 {
1470 	struct amdgpu_device *adev = ring->adev;
1471 
1472 	amdgpu_ring_write(ring,
1473 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1474 	amdgpu_ring_write(ring, 0);
1475 	amdgpu_ring_write(ring,
1476 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1477 	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_START << 1);
1478 }
1479 
1480 /**
1481  * vcn_v1_0_dec_ring_insert_end - insert a end command
1482  *
1483  * @ring: amdgpu_ring pointer
1484  *
1485  * Write a end command to the ring.
1486  */
1487 static void vcn_v1_0_dec_ring_insert_end(struct amdgpu_ring *ring)
1488 {
1489 	struct amdgpu_device *adev = ring->adev;
1490 
1491 	amdgpu_ring_write(ring,
1492 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1493 	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_END << 1);
1494 }
1495 
1496 /**
1497  * vcn_v1_0_dec_ring_emit_fence - emit an fence & trap command
1498  *
1499  * @ring: amdgpu_ring pointer
1500  * @addr: address
1501  * @seq: sequence number
1502  * @flags: fence related flags
1503  *
1504  * Write a fence and a trap command to the ring.
1505  */
1506 static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
1507 				     unsigned flags)
1508 {
1509 	struct amdgpu_device *adev = ring->adev;
1510 
1511 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
1512 
1513 	amdgpu_ring_write(ring,
1514 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0));
1515 	amdgpu_ring_write(ring, seq);
1516 	amdgpu_ring_write(ring,
1517 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1518 	amdgpu_ring_write(ring, addr & 0xffffffff);
1519 	amdgpu_ring_write(ring,
1520 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1521 	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
1522 	amdgpu_ring_write(ring,
1523 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1524 	amdgpu_ring_write(ring, VCN_DEC_CMD_FENCE << 1);
1525 
1526 	amdgpu_ring_write(ring,
1527 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1528 	amdgpu_ring_write(ring, 0);
1529 	amdgpu_ring_write(ring,
1530 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1531 	amdgpu_ring_write(ring, 0);
1532 	amdgpu_ring_write(ring,
1533 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1534 	amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1);
1535 }
1536 
1537 /**
1538  * vcn_v1_0_dec_ring_emit_ib - execute indirect buffer
1539  *
1540  * @ring: amdgpu_ring pointer
1541  * @job: job to retrieve vmid from
1542  * @ib: indirect buffer to execute
1543  * @flags: unused
1544  *
1545  * Write ring commands to execute the indirect buffer
1546  */
1547 static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
1548 					struct amdgpu_job *job,
1549 					struct amdgpu_ib *ib,
1550 					uint32_t flags)
1551 {
1552 	struct amdgpu_device *adev = ring->adev;
1553 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
1554 
1555 	amdgpu_ring_write(ring,
1556 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0));
1557 	amdgpu_ring_write(ring, vmid);
1558 
1559 	amdgpu_ring_write(ring,
1560 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0));
1561 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
1562 	amdgpu_ring_write(ring,
1563 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0));
1564 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
1565 	amdgpu_ring_write(ring,
1566 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_IB_SIZE), 0));
1567 	amdgpu_ring_write(ring, ib->length_dw);
1568 }
1569 
1570 static void vcn_v1_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring,
1571 					    uint32_t reg, uint32_t val,
1572 					    uint32_t mask)
1573 {
1574 	struct amdgpu_device *adev = ring->adev;
1575 
1576 	amdgpu_ring_write(ring,
1577 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1578 	amdgpu_ring_write(ring, reg << 2);
1579 	amdgpu_ring_write(ring,
1580 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1581 	amdgpu_ring_write(ring, val);
1582 	amdgpu_ring_write(ring,
1583 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH8), 0));
1584 	amdgpu_ring_write(ring, mask);
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_REG_READ_COND_WAIT << 1);
1588 }
1589 
1590 static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
1591 					    unsigned vmid, uint64_t pd_addr)
1592 {
1593 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
1594 	uint32_t data0, data1, mask;
1595 
1596 	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
1597 
1598 	/* wait for register write */
1599 	data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance;
1600 	data1 = lower_32_bits(pd_addr);
1601 	mask = 0xffffffff;
1602 	vcn_v1_0_dec_ring_emit_reg_wait(ring, data0, data1, mask);
1603 }
1604 
1605 static void vcn_v1_0_dec_ring_emit_wreg(struct amdgpu_ring *ring,
1606 					uint32_t reg, uint32_t val)
1607 {
1608 	struct amdgpu_device *adev = ring->adev;
1609 
1610 	amdgpu_ring_write(ring,
1611 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0));
1612 	amdgpu_ring_write(ring, reg << 2);
1613 	amdgpu_ring_write(ring,
1614 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0));
1615 	amdgpu_ring_write(ring, val);
1616 	amdgpu_ring_write(ring,
1617 		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0));
1618 	amdgpu_ring_write(ring, VCN_DEC_CMD_WRITE_REG << 1);
1619 }
1620 
1621 /**
1622  * vcn_v1_0_enc_ring_get_rptr - get enc read pointer
1623  *
1624  * @ring: amdgpu_ring pointer
1625  *
1626  * Returns the current hardware enc read pointer
1627  */
1628 static uint64_t vcn_v1_0_enc_ring_get_rptr(struct amdgpu_ring *ring)
1629 {
1630 	struct amdgpu_device *adev = ring->adev;
1631 
1632 	if (ring == &adev->vcn.inst->ring_enc[0])
1633 		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR);
1634 	else
1635 		return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2);
1636 }
1637 
1638  /**
1639  * vcn_v1_0_enc_ring_get_wptr - get enc write pointer
1640  *
1641  * @ring: amdgpu_ring pointer
1642  *
1643  * Returns the current hardware enc write pointer
1644  */
1645 static uint64_t vcn_v1_0_enc_ring_get_wptr(struct amdgpu_ring *ring)
1646 {
1647 	struct amdgpu_device *adev = ring->adev;
1648 
1649 	if (ring == &adev->vcn.inst->ring_enc[0])
1650 		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR);
1651 	else
1652 		return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2);
1653 }
1654 
1655  /**
1656  * vcn_v1_0_enc_ring_set_wptr - set enc write pointer
1657  *
1658  * @ring: amdgpu_ring pointer
1659  *
1660  * Commits the enc write pointer to the hardware
1661  */
1662 static void vcn_v1_0_enc_ring_set_wptr(struct amdgpu_ring *ring)
1663 {
1664 	struct amdgpu_device *adev = ring->adev;
1665 
1666 	if (ring == &adev->vcn.inst->ring_enc[0])
1667 		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR,
1668 			lower_32_bits(ring->wptr));
1669 	else
1670 		WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2,
1671 			lower_32_bits(ring->wptr));
1672 }
1673 
1674 /**
1675  * vcn_v1_0_enc_ring_emit_fence - emit an enc fence & trap command
1676  *
1677  * @ring: amdgpu_ring pointer
1678  * @addr: address
1679  * @seq: sequence number
1680  * @flags: fence related flags
1681  *
1682  * Write enc a fence and a trap command to the ring.
1683  */
1684 static void vcn_v1_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
1685 			u64 seq, unsigned flags)
1686 {
1687 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
1688 
1689 	amdgpu_ring_write(ring, VCN_ENC_CMD_FENCE);
1690 	amdgpu_ring_write(ring, addr);
1691 	amdgpu_ring_write(ring, upper_32_bits(addr));
1692 	amdgpu_ring_write(ring, seq);
1693 	amdgpu_ring_write(ring, VCN_ENC_CMD_TRAP);
1694 }
1695 
1696 static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring)
1697 {
1698 	amdgpu_ring_write(ring, VCN_ENC_CMD_END);
1699 }
1700 
1701 /**
1702  * vcn_v1_0_enc_ring_emit_ib - enc execute indirect buffer
1703  *
1704  * @ring: amdgpu_ring pointer
1705  * @job: job to retrive vmid from
1706  * @ib: indirect buffer to execute
1707  * @flags: unused
1708  *
1709  * Write enc ring commands to execute the indirect buffer
1710  */
1711 static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
1712 					struct amdgpu_job *job,
1713 					struct amdgpu_ib *ib,
1714 					uint32_t flags)
1715 {
1716 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
1717 
1718 	amdgpu_ring_write(ring, VCN_ENC_CMD_IB);
1719 	amdgpu_ring_write(ring, vmid);
1720 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
1721 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
1722 	amdgpu_ring_write(ring, ib->length_dw);
1723 }
1724 
1725 static void vcn_v1_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring,
1726 					    uint32_t reg, uint32_t val,
1727 					    uint32_t mask)
1728 {
1729 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WAIT);
1730 	amdgpu_ring_write(ring, reg << 2);
1731 	amdgpu_ring_write(ring, mask);
1732 	amdgpu_ring_write(ring, val);
1733 }
1734 
1735 static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
1736 					    unsigned int vmid, uint64_t pd_addr)
1737 {
1738 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
1739 
1740 	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
1741 
1742 	/* wait for reg writes */
1743 	vcn_v1_0_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 +
1744 					vmid * hub->ctx_addr_distance,
1745 					lower_32_bits(pd_addr), 0xffffffff);
1746 }
1747 
1748 static void vcn_v1_0_enc_ring_emit_wreg(struct amdgpu_ring *ring,
1749 					uint32_t reg, uint32_t val)
1750 {
1751 	amdgpu_ring_write(ring, VCN_ENC_CMD_REG_WRITE);
1752 	amdgpu_ring_write(ring,	reg << 2);
1753 	amdgpu_ring_write(ring, val);
1754 }
1755 
1756 static int vcn_v1_0_set_interrupt_state(struct amdgpu_device *adev,
1757 					struct amdgpu_irq_src *source,
1758 					unsigned type,
1759 					enum amdgpu_interrupt_state state)
1760 {
1761 	return 0;
1762 }
1763 
1764 static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev,
1765 				      struct amdgpu_irq_src *source,
1766 				      struct amdgpu_iv_entry *entry)
1767 {
1768 	DRM_DEBUG("IH: VCN TRAP\n");
1769 
1770 	switch (entry->src_id) {
1771 	case 124:
1772 		amdgpu_fence_process(&adev->vcn.inst->ring_dec);
1773 		break;
1774 	case 119:
1775 		amdgpu_fence_process(&adev->vcn.inst->ring_enc[0]);
1776 		break;
1777 	case 120:
1778 		amdgpu_fence_process(&adev->vcn.inst->ring_enc[1]);
1779 		break;
1780 	default:
1781 		DRM_ERROR("Unhandled interrupt: %d %d\n",
1782 			  entry->src_id, entry->src_data[0]);
1783 		break;
1784 	}
1785 
1786 	return 0;
1787 }
1788 
1789 static void vcn_v1_0_dec_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
1790 {
1791 	struct amdgpu_device *adev = ring->adev;
1792 	int i;
1793 
1794 	WARN_ON(ring->wptr % 2 || count % 2);
1795 
1796 	for (i = 0; i < count / 2; i++) {
1797 		amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0));
1798 		amdgpu_ring_write(ring, 0);
1799 	}
1800 }
1801 
1802 static int vcn_v1_0_set_powergating_state(void *handle,
1803 					  enum amd_powergating_state state)
1804 {
1805 	/* This doesn't actually powergate the VCN block.
1806 	 * That's done in the dpm code via the SMC.  This
1807 	 * just re-inits the block as necessary.  The actual
1808 	 * gating still happens in the dpm code.  We should
1809 	 * revisit this when there is a cleaner line between
1810 	 * the smc and the hw blocks
1811 	 */
1812 	int ret;
1813 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
1814 
1815 	if (state == adev->vcn.cur_state)
1816 		return 0;
1817 
1818 	if (state == AMD_PG_STATE_GATE)
1819 		ret = vcn_v1_0_stop(adev);
1820 	else
1821 		ret = vcn_v1_0_start(adev);
1822 
1823 	if (!ret)
1824 		adev->vcn.cur_state = state;
1825 	return ret;
1826 }
1827 
1828 static void vcn_v1_0_idle_work_handler(struct work_struct *work)
1829 {
1830 	struct amdgpu_device *adev =
1831 		container_of(work, struct amdgpu_device, vcn.idle_work.work);
1832 	unsigned int fences = 0, i;
1833 
1834 	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
1835 		fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_enc[i]);
1836 
1837 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
1838 		struct dpg_pause_state new_state;
1839 
1840 		if (fences)
1841 			new_state.fw_based = VCN_DPG_STATE__PAUSE;
1842 		else
1843 			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
1844 
1845 		if (amdgpu_fence_count_emitted(adev->jpeg.inst->ring_dec))
1846 			new_state.jpeg = VCN_DPG_STATE__PAUSE;
1847 		else
1848 			new_state.jpeg = VCN_DPG_STATE__UNPAUSE;
1849 
1850 		adev->vcn.pause_dpg_mode(adev, 0, &new_state);
1851 	}
1852 
1853 	fences += amdgpu_fence_count_emitted(adev->jpeg.inst->ring_dec);
1854 	fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_dec);
1855 
1856 	if (fences == 0) {
1857 		amdgpu_gfx_off_ctrl(adev, true);
1858 		if (adev->pm.dpm_enabled)
1859 			amdgpu_dpm_enable_uvd(adev, false);
1860 		else
1861 			amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
1862 			       AMD_PG_STATE_GATE);
1863 	} else {
1864 		schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
1865 	}
1866 }
1867 
1868 static void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring)
1869 {
1870 	struct	amdgpu_device *adev = ring->adev;
1871 	bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work);
1872 
1873 	mutex_lock(&adev->vcn.vcn1_jpeg1_workaround);
1874 
1875 	if (amdgpu_fence_wait_empty(ring->adev->jpeg.inst->ring_dec))
1876 		DRM_ERROR("VCN dec: jpeg dec ring may not be empty\n");
1877 
1878 	vcn_v1_0_set_pg_for_begin_use(ring, set_clocks);
1879 
1880 }
1881 
1882 void vcn_v1_0_set_pg_for_begin_use(struct amdgpu_ring *ring, bool set_clocks)
1883 {
1884 	struct amdgpu_device *adev = ring->adev;
1885 
1886 	if (set_clocks) {
1887 		amdgpu_gfx_off_ctrl(adev, false);
1888 		if (adev->pm.dpm_enabled)
1889 			amdgpu_dpm_enable_uvd(adev, true);
1890 		else
1891 			amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
1892 			       AMD_PG_STATE_UNGATE);
1893 	}
1894 
1895 	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) {
1896 		struct dpg_pause_state new_state;
1897 		unsigned int fences = 0, i;
1898 
1899 		for (i = 0; i < adev->vcn.num_enc_rings; ++i)
1900 			fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_enc[i]);
1901 
1902 		if (fences)
1903 			new_state.fw_based = VCN_DPG_STATE__PAUSE;
1904 		else
1905 			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
1906 
1907 		if (amdgpu_fence_count_emitted(adev->jpeg.inst->ring_dec))
1908 			new_state.jpeg = VCN_DPG_STATE__PAUSE;
1909 		else
1910 			new_state.jpeg = VCN_DPG_STATE__UNPAUSE;
1911 
1912 		if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC)
1913 			new_state.fw_based = VCN_DPG_STATE__PAUSE;
1914 		else if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG)
1915 			new_state.jpeg = VCN_DPG_STATE__PAUSE;
1916 
1917 		adev->vcn.pause_dpg_mode(adev, 0, &new_state);
1918 	}
1919 }
1920 
1921 void vcn_v1_0_ring_end_use(struct amdgpu_ring *ring)
1922 {
1923 	schedule_delayed_work(&ring->adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
1924 	mutex_unlock(&ring->adev->vcn.vcn1_jpeg1_workaround);
1925 }
1926 
1927 static void vcn_v1_0_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
1928 {
1929 	struct amdgpu_device *adev = ip_block->adev;
1930 	int i, j;
1931 	uint32_t reg_count = ARRAY_SIZE(vcn_reg_list_1_0);
1932 	uint32_t inst_off, is_powered;
1933 
1934 	if (!adev->vcn.ip_dump)
1935 		return;
1936 
1937 	drm_printf(p, "num_instances:%d\n", adev->vcn.num_vcn_inst);
1938 	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
1939 		if (adev->vcn.harvest_config & (1 << i)) {
1940 			drm_printf(p, "\nHarvested Instance:VCN%d Skipping dump\n", i);
1941 			continue;
1942 		}
1943 
1944 		inst_off = i * reg_count;
1945 		is_powered = (adev->vcn.ip_dump[inst_off] &
1946 				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK) != 1;
1947 
1948 		if (is_powered) {
1949 			drm_printf(p, "\nActive Instance:VCN%d\n", i);
1950 			for (j = 0; j < reg_count; j++)
1951 				drm_printf(p, "%-50s \t 0x%08x\n", vcn_reg_list_1_0[j].reg_name,
1952 					   adev->vcn.ip_dump[inst_off + j]);
1953 		} else {
1954 			drm_printf(p, "\nInactive Instance:VCN%d\n", i);
1955 		}
1956 	}
1957 }
1958 
1959 static void vcn_v1_0_dump_ip_state(struct amdgpu_ip_block *ip_block)
1960 {
1961 	struct amdgpu_device *adev = ip_block->adev;
1962 	int i, j;
1963 	bool is_powered;
1964 	uint32_t inst_off;
1965 	uint32_t reg_count = ARRAY_SIZE(vcn_reg_list_1_0);
1966 
1967 	if (!adev->vcn.ip_dump)
1968 		return;
1969 
1970 	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
1971 		if (adev->vcn.harvest_config & (1 << i))
1972 			continue;
1973 
1974 		inst_off = i * reg_count;
1975 		/* mmUVD_POWER_STATUS is always readable and is first element of the array */
1976 		adev->vcn.ip_dump[inst_off] = RREG32_SOC15(VCN, i, mmUVD_POWER_STATUS);
1977 		is_powered = (adev->vcn.ip_dump[inst_off] &
1978 				UVD_POWER_STATUS__UVD_POWER_STATUS_MASK) != 1;
1979 
1980 		if (is_powered)
1981 			for (j = 1; j < reg_count; j++)
1982 				adev->vcn.ip_dump[inst_off + j] =
1983 					RREG32(SOC15_REG_ENTRY_OFFSET_INST(vcn_reg_list_1_0[j], i));
1984 	}
1985 }
1986 
1987 static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
1988 	.name = "vcn_v1_0",
1989 	.early_init = vcn_v1_0_early_init,
1990 	.sw_init = vcn_v1_0_sw_init,
1991 	.sw_fini = vcn_v1_0_sw_fini,
1992 	.hw_init = vcn_v1_0_hw_init,
1993 	.hw_fini = vcn_v1_0_hw_fini,
1994 	.suspend = vcn_v1_0_suspend,
1995 	.resume = vcn_v1_0_resume,
1996 	.is_idle = vcn_v1_0_is_idle,
1997 	.wait_for_idle = vcn_v1_0_wait_for_idle,
1998 	.set_clockgating_state = vcn_v1_0_set_clockgating_state,
1999 	.set_powergating_state = vcn_v1_0_set_powergating_state,
2000 	.dump_ip_state = vcn_v1_0_dump_ip_state,
2001 	.print_ip_state = vcn_v1_0_print_ip_state,
2002 };
2003 
2004 /*
2005  * It is a hardware issue that VCN can't handle a GTT TMZ buffer on
2006  * CHIP_RAVEN series ASIC. Move such a GTT TMZ buffer to VRAM domain
2007  * before command submission as a workaround.
2008  */
2009 static int vcn_v1_0_validate_bo(struct amdgpu_cs_parser *parser,
2010 				struct amdgpu_job *job,
2011 				uint64_t addr)
2012 {
2013 	struct ttm_operation_ctx ctx = { false, false };
2014 	struct amdgpu_fpriv *fpriv = parser->filp->driver_priv;
2015 	struct amdgpu_vm *vm = &fpriv->vm;
2016 	struct amdgpu_bo_va_mapping *mapping;
2017 	struct amdgpu_bo *bo;
2018 	int r;
2019 
2020 	addr &= AMDGPU_GMC_HOLE_MASK;
2021 	if (addr & 0x7) {
2022 		DRM_ERROR("VCN messages must be 8 byte aligned!\n");
2023 		return -EINVAL;
2024 	}
2025 
2026 	mapping = amdgpu_vm_bo_lookup_mapping(vm, addr/AMDGPU_GPU_PAGE_SIZE);
2027 	if (!mapping || !mapping->bo_va || !mapping->bo_va->base.bo)
2028 		return -EINVAL;
2029 
2030 	bo = mapping->bo_va->base.bo;
2031 	if (!(bo->flags & AMDGPU_GEM_CREATE_ENCRYPTED))
2032 		return 0;
2033 
2034 	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM);
2035 	r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
2036 	if (r) {
2037 		DRM_ERROR("Failed to validate the VCN message BO (%d)!\n", r);
2038 		return r;
2039 	}
2040 
2041 	return r;
2042 }
2043 
2044 static int vcn_v1_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
2045 					   struct amdgpu_job *job,
2046 					   struct amdgpu_ib *ib)
2047 {
2048 	uint32_t msg_lo = 0, msg_hi = 0;
2049 	int i, r;
2050 
2051 	if (!(ib->flags & AMDGPU_IB_FLAGS_SECURE))
2052 		return 0;
2053 
2054 	for (i = 0; i < ib->length_dw; i += 2) {
2055 		uint32_t reg = amdgpu_ib_get_value(ib, i);
2056 		uint32_t val = amdgpu_ib_get_value(ib, i + 1);
2057 
2058 		if (reg == PACKET0(p->adev->vcn.internal.data0, 0)) {
2059 			msg_lo = val;
2060 		} else if (reg == PACKET0(p->adev->vcn.internal.data1, 0)) {
2061 			msg_hi = val;
2062 		} else if (reg == PACKET0(p->adev->vcn.internal.cmd, 0)) {
2063 			r = vcn_v1_0_validate_bo(p, job,
2064 						 ((u64)msg_hi) << 32 | msg_lo);
2065 			if (r)
2066 				return r;
2067 		}
2068 	}
2069 
2070 	return 0;
2071 }
2072 
2073 static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = {
2074 	.type = AMDGPU_RING_TYPE_VCN_DEC,
2075 	.align_mask = 0xf,
2076 	.support_64bit_ptrs = false,
2077 	.no_user_fence = true,
2078 	.secure_submission_supported = true,
2079 	.get_rptr = vcn_v1_0_dec_ring_get_rptr,
2080 	.get_wptr = vcn_v1_0_dec_ring_get_wptr,
2081 	.set_wptr = vcn_v1_0_dec_ring_set_wptr,
2082 	.patch_cs_in_place = vcn_v1_0_ring_patch_cs_in_place,
2083 	.emit_frame_size =
2084 		6 + 6 + /* hdp invalidate / flush */
2085 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
2086 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
2087 		8 + /* vcn_v1_0_dec_ring_emit_vm_flush */
2088 		14 + 14 + /* vcn_v1_0_dec_ring_emit_fence x2 vm fence */
2089 		6,
2090 	.emit_ib_size = 8, /* vcn_v1_0_dec_ring_emit_ib */
2091 	.emit_ib = vcn_v1_0_dec_ring_emit_ib,
2092 	.emit_fence = vcn_v1_0_dec_ring_emit_fence,
2093 	.emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush,
2094 	.test_ring = amdgpu_vcn_dec_ring_test_ring,
2095 	.test_ib = amdgpu_vcn_dec_ring_test_ib,
2096 	.insert_nop = vcn_v1_0_dec_ring_insert_nop,
2097 	.insert_start = vcn_v1_0_dec_ring_insert_start,
2098 	.insert_end = vcn_v1_0_dec_ring_insert_end,
2099 	.pad_ib = amdgpu_ring_generic_pad_ib,
2100 	.begin_use = vcn_v1_0_ring_begin_use,
2101 	.end_use = vcn_v1_0_ring_end_use,
2102 	.emit_wreg = vcn_v1_0_dec_ring_emit_wreg,
2103 	.emit_reg_wait = vcn_v1_0_dec_ring_emit_reg_wait,
2104 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
2105 };
2106 
2107 static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = {
2108 	.type = AMDGPU_RING_TYPE_VCN_ENC,
2109 	.align_mask = 0x3f,
2110 	.nop = VCN_ENC_CMD_NO_OP,
2111 	.support_64bit_ptrs = false,
2112 	.no_user_fence = true,
2113 	.get_rptr = vcn_v1_0_enc_ring_get_rptr,
2114 	.get_wptr = vcn_v1_0_enc_ring_get_wptr,
2115 	.set_wptr = vcn_v1_0_enc_ring_set_wptr,
2116 	.emit_frame_size =
2117 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
2118 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
2119 		4 + /* vcn_v1_0_enc_ring_emit_vm_flush */
2120 		5 + 5 + /* vcn_v1_0_enc_ring_emit_fence x2 vm fence */
2121 		1, /* vcn_v1_0_enc_ring_insert_end */
2122 	.emit_ib_size = 5, /* vcn_v1_0_enc_ring_emit_ib */
2123 	.emit_ib = vcn_v1_0_enc_ring_emit_ib,
2124 	.emit_fence = vcn_v1_0_enc_ring_emit_fence,
2125 	.emit_vm_flush = vcn_v1_0_enc_ring_emit_vm_flush,
2126 	.test_ring = amdgpu_vcn_enc_ring_test_ring,
2127 	.test_ib = amdgpu_vcn_enc_ring_test_ib,
2128 	.insert_nop = amdgpu_ring_insert_nop,
2129 	.insert_end = vcn_v1_0_enc_ring_insert_end,
2130 	.pad_ib = amdgpu_ring_generic_pad_ib,
2131 	.begin_use = vcn_v1_0_ring_begin_use,
2132 	.end_use = vcn_v1_0_ring_end_use,
2133 	.emit_wreg = vcn_v1_0_enc_ring_emit_wreg,
2134 	.emit_reg_wait = vcn_v1_0_enc_ring_emit_reg_wait,
2135 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
2136 };
2137 
2138 static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev)
2139 {
2140 	adev->vcn.inst->ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs;
2141 }
2142 
2143 static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev)
2144 {
2145 	int i;
2146 
2147 	for (i = 0; i < adev->vcn.num_enc_rings; ++i)
2148 		adev->vcn.inst->ring_enc[i].funcs = &vcn_v1_0_enc_ring_vm_funcs;
2149 }
2150 
2151 static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = {
2152 	.set = vcn_v1_0_set_interrupt_state,
2153 	.process = vcn_v1_0_process_interrupt,
2154 };
2155 
2156 static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev)
2157 {
2158 	adev->vcn.inst->irq.num_types = adev->vcn.num_enc_rings + 2;
2159 	adev->vcn.inst->irq.funcs = &vcn_v1_0_irq_funcs;
2160 }
2161 
2162 const struct amdgpu_ip_block_version vcn_v1_0_ip_block = {
2163 		.type = AMD_IP_BLOCK_TYPE_VCN,
2164 		.major = 1,
2165 		.minor = 0,
2166 		.rev = 0,
2167 		.funcs = &vcn_v1_0_ip_funcs,
2168 };
2169