xref: /linux/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c (revision 5488bec96bccbd87335921338f8dc38b87db7d2c)
1 /*
2  * Copyright 2022 Advanced Micro Devices, Inc.
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining a
5  * copy of this software and associated documentation files (the "Software"),
6  * to deal in the Software without restriction, including without limitation
7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8  * and/or sell copies of the Software, and to permit persons to whom the
9  * Software is furnished to do so, subject to the following conditions:
10  *
11  * The above copyright notice and this permission notice shall be included in
12  * all copies or substantial portions of the Software.
13  *
14  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
17  * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20  * OTHER DEALINGS IN THE SOFTWARE.
21  *
22  */
23 
24 #include "amdgpu.h"
25 #include "amdgpu_jpeg.h"
26 #include "soc15.h"
27 #include "soc15d.h"
28 #include "jpeg_v2_0.h"
29 #include "jpeg_v4_0_3.h"
30 #include "mmsch_v4_0_3.h"
31 
32 #include "vcn/vcn_4_0_3_offset.h"
33 #include "vcn/vcn_4_0_3_sh_mask.h"
34 #include "ivsrcid/vcn/irqsrcs_vcn_4_0.h"
35 
36 #define NORMALIZE_JPEG_REG_OFFSET(offset) \
37 		(offset & 0x1FFFF)
38 
39 enum jpeg_engin_status {
40 	UVD_PGFSM_STATUS__UVDJ_PWR_ON  = 0,
41 	UVD_PGFSM_STATUS__UVDJ_PWR_OFF = 2,
42 };
43 
44 static void jpeg_v4_0_3_set_dec_ring_funcs(struct amdgpu_device *adev);
45 static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev);
46 static int jpeg_v4_0_3_set_powergating_state(struct amdgpu_ip_block *ip_block,
47 				enum amd_powergating_state state);
48 static void jpeg_v4_0_3_set_ras_funcs(struct amdgpu_device *adev);
49 static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring);
50 
51 static int amdgpu_ih_srcid_jpeg[] = {
52 	VCN_4_0__SRCID__JPEG_DECODE,
53 	VCN_4_0__SRCID__JPEG1_DECODE,
54 	VCN_4_0__SRCID__JPEG2_DECODE,
55 	VCN_4_0__SRCID__JPEG3_DECODE,
56 	VCN_4_0__SRCID__JPEG4_DECODE,
57 	VCN_4_0__SRCID__JPEG5_DECODE,
58 	VCN_4_0__SRCID__JPEG6_DECODE,
59 	VCN_4_0__SRCID__JPEG7_DECODE
60 };
61 
62 static const struct amdgpu_hwip_reg_entry jpeg_reg_list_4_0_3[] = {
63 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_POWER_STATUS),
64 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_INT_STAT),
65 	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_SYS_INT_STATUS),
66 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_RPTR),
67 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_WPTR),
68 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC0_UVD_JRBC_STATUS),
69 	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_ADDR_MODE),
70 	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_GFX10_ADDR_CONFIG),
71 	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_Y_GFX10_TILING_SURFACE),
72 	SOC15_REG_ENTRY_STR(JPEG, 0, regJPEG_DEC_UV_GFX10_TILING_SURFACE),
73 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_PITCH),
74 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JPEG_UV_PITCH),
75 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_RB_RPTR),
76 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_RB_WPTR),
77 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC1_UVD_JRBC_STATUS),
78 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_RB_RPTR),
79 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_RB_WPTR),
80 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC2_UVD_JRBC_STATUS),
81 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_RB_RPTR),
82 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_RB_WPTR),
83 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC3_UVD_JRBC_STATUS),
84 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_RB_RPTR),
85 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_RB_WPTR),
86 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC4_UVD_JRBC_STATUS),
87 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_RB_RPTR),
88 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_RB_WPTR),
89 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC5_UVD_JRBC_STATUS),
90 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_RB_RPTR),
91 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_RB_WPTR),
92 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC6_UVD_JRBC_STATUS),
93 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_RB_RPTR),
94 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_RB_WPTR),
95 	SOC15_REG_ENTRY_STR(JPEG, 0, regUVD_JRBC7_UVD_JRBC_STATUS),
96 };
97 
98 static inline bool jpeg_v4_0_3_normalizn_reqd(struct amdgpu_device *adev)
99 {
100 	return (adev->jpeg.caps & AMDGPU_JPEG_CAPS(RRMT_ENABLED)) == 0;
101 }
102 
103 static inline int jpeg_v4_0_3_core_reg_offset(u32 pipe)
104 {
105 	if (pipe)
106 		return ((0x40 * pipe) - 0xc80);
107 	else
108 		return 0;
109 }
110 
111 /**
112  * jpeg_v4_0_3_early_init - set function pointers
113  *
114  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
115  *
116  * Set ring and irq function pointers
117  */
118 static int jpeg_v4_0_3_early_init(struct amdgpu_ip_block *ip_block)
119 {
120 	struct amdgpu_device *adev = ip_block->adev;
121 
122 	adev->jpeg.num_jpeg_rings = AMDGPU_MAX_JPEG_RINGS_4_0_3;
123 
124 	jpeg_v4_0_3_set_dec_ring_funcs(adev);
125 	jpeg_v4_0_3_set_irq_funcs(adev);
126 	jpeg_v4_0_3_set_ras_funcs(adev);
127 
128 	return 0;
129 }
130 
131 /**
132  * jpeg_v4_0_3_sw_init - sw init for JPEG block
133  *
134  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
135  *
136  * Load firmware and sw initialization
137  */
138 static int jpeg_v4_0_3_sw_init(struct amdgpu_ip_block *ip_block)
139 {
140 	struct amdgpu_device *adev = ip_block->adev;
141 	struct amdgpu_ring *ring;
142 	int i, j, r, jpeg_inst;
143 
144 	for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
145 		/* JPEG TRAP */
146 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN,
147 				amdgpu_ih_srcid_jpeg[j], &adev->jpeg.inst->irq);
148 		if (r)
149 			return r;
150 	}
151 
152 	r = amdgpu_jpeg_sw_init(adev);
153 	if (r)
154 		return r;
155 
156 	r = amdgpu_jpeg_resume(adev);
157 	if (r)
158 		return r;
159 
160 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
161 		jpeg_inst = GET_INST(JPEG, i);
162 
163 		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
164 			ring = &adev->jpeg.inst[i].ring_dec[j];
165 			ring->use_doorbell = true;
166 			ring->vm_hub = AMDGPU_MMHUB0(adev->jpeg.inst[i].aid_id);
167 			if (!amdgpu_sriov_vf(adev)) {
168 				ring->doorbell_index =
169 					(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
170 					1 + j + 9 * jpeg_inst;
171 			} else {
172 				if (j < 4)
173 					ring->doorbell_index =
174 						(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
175 						4 + j + 32 * jpeg_inst;
176 				else
177 					ring->doorbell_index =
178 						(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
179 						8 + j + 32 * jpeg_inst;
180 			}
181 			sprintf(ring->name, "jpeg_dec_%d.%d", adev->jpeg.inst[i].aid_id, j);
182 			r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
183 						AMDGPU_RING_PRIO_DEFAULT, NULL);
184 			if (r)
185 				return r;
186 
187 			adev->jpeg.internal.jpeg_pitch[j] =
188 				regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET;
189 			adev->jpeg.inst[i].external.jpeg_pitch[j] =
190 				SOC15_REG_OFFSET1(JPEG, jpeg_inst, regUVD_JRBC0_UVD_JRBC_SCRATCH0,
191 						  jpeg_v4_0_3_core_reg_offset(j));
192 		}
193 	}
194 
195 	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
196 		r = amdgpu_jpeg_ras_sw_init(adev);
197 		if (r) {
198 			dev_err(adev->dev, "Failed to initialize jpeg ras block!\n");
199 			return r;
200 		}
201 	}
202 
203 	r = amdgpu_jpeg_reg_dump_init(adev, jpeg_reg_list_4_0_3, ARRAY_SIZE(jpeg_reg_list_4_0_3));
204 	if (r)
205 		return r;
206 
207 	if (!amdgpu_sriov_vf(adev)) {
208 		adev->jpeg.supported_reset = AMDGPU_RESET_TYPE_PER_QUEUE;
209 		r = amdgpu_jpeg_sysfs_reset_mask_init(adev);
210 		if (r)
211 			return r;
212 	}
213 
214 	return 0;
215 }
216 
217 /**
218  * jpeg_v4_0_3_sw_fini - sw fini for JPEG block
219  *
220  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
221  *
222  * JPEG suspend and free up sw allocation
223  */
224 static int jpeg_v4_0_3_sw_fini(struct amdgpu_ip_block *ip_block)
225 {
226 	struct amdgpu_device *adev = ip_block->adev;
227 	int r;
228 
229 	r = amdgpu_jpeg_suspend(adev);
230 	if (r)
231 		return r;
232 
233 	if (!amdgpu_sriov_vf(adev))
234 		amdgpu_jpeg_sysfs_reset_mask_fini(adev);
235 
236 	r = amdgpu_jpeg_sw_fini(adev);
237 
238 	return r;
239 }
240 
241 static int jpeg_v4_0_3_start_sriov(struct amdgpu_device *adev)
242 {
243 	struct amdgpu_ring *ring;
244 	uint64_t ctx_addr;
245 	uint32_t param, resp, expected;
246 	uint32_t tmp, timeout;
247 
248 	struct amdgpu_mm_table *table = &adev->virt.mm_table;
249 	uint32_t *table_loc;
250 	uint32_t table_size;
251 	uint32_t size, size_dw, item_offset;
252 	uint32_t init_status;
253 	int i, j, jpeg_inst;
254 
255 	struct mmsch_v4_0_cmd_direct_write
256 		direct_wt = { {0} };
257 	struct mmsch_v4_0_cmd_end end = { {0} };
258 	struct mmsch_v4_0_3_init_header header;
259 
260 	direct_wt.cmd_header.command_type =
261 		MMSCH_COMMAND__DIRECT_REG_WRITE;
262 	end.cmd_header.command_type =
263 		MMSCH_COMMAND__END;
264 
265 	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
266 		jpeg_inst = GET_INST(JPEG, i);
267 
268 		memset(&header, 0, sizeof(struct mmsch_v4_0_3_init_header));
269 		header.version = MMSCH_VERSION;
270 		header.total_size = sizeof(struct mmsch_v4_0_3_init_header) >> 2;
271 
272 		table_loc = (uint32_t *)table->cpu_addr;
273 		table_loc += header.total_size;
274 
275 		item_offset = header.total_size;
276 
277 		for (j = 0; j < adev->jpeg.num_jpeg_rings; j++) {
278 			ring = &adev->jpeg.inst[i].ring_dec[j];
279 			table_size = 0;
280 
281 			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW);
282 			MMSCH_V4_0_INSERT_DIRECT_WT(tmp, lower_32_bits(ring->gpu_addr));
283 			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH);
284 			MMSCH_V4_0_INSERT_DIRECT_WT(tmp, upper_32_bits(ring->gpu_addr));
285 			tmp = SOC15_REG_OFFSET(JPEG, 0, regUVD_JRBC0_UVD_JRBC_RB_SIZE);
286 			MMSCH_V4_0_INSERT_DIRECT_WT(tmp, ring->ring_size / 4);
287 
288 			if (j <= 3) {
289 				header.mjpegdec0[j].table_offset = item_offset;
290 				header.mjpegdec0[j].init_status = 0;
291 				header.mjpegdec0[j].table_size = table_size;
292 			} else {
293 				header.mjpegdec1[j - 4].table_offset = item_offset;
294 				header.mjpegdec1[j - 4].init_status = 0;
295 				header.mjpegdec1[j - 4].table_size = table_size;
296 			}
297 			header.total_size += table_size;
298 			item_offset += table_size;
299 		}
300 
301 		MMSCH_V4_0_INSERT_END();
302 
303 		/* send init table to MMSCH */
304 		size = sizeof(struct mmsch_v4_0_3_init_header);
305 		table_loc = (uint32_t *)table->cpu_addr;
306 		memcpy((void *)table_loc, &header, size);
307 
308 		ctx_addr = table->gpu_addr;
309 		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_LO, lower_32_bits(ctx_addr));
310 		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_ADDR_HI, upper_32_bits(ctx_addr));
311 
312 		tmp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID);
313 		tmp &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK;
314 		tmp |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT);
315 		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_VMID, tmp);
316 
317 		size = header.total_size;
318 		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_CTX_SIZE, size);
319 
320 		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP, 0);
321 
322 		param = 0x00000001;
323 		WREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_HOST, param);
324 		tmp = 0;
325 		timeout = 1000;
326 		resp = 0;
327 		expected = MMSCH_VF_MAILBOX_RESP__OK;
328 		init_status =
329 			((struct mmsch_v4_0_3_init_header *)(table_loc))->mjpegdec0[i].init_status;
330 		while (resp != expected) {
331 			resp = RREG32_SOC15(VCN, jpeg_inst, regMMSCH_VF_MAILBOX_RESP);
332 
333 			if (resp != 0)
334 				break;
335 			udelay(10);
336 			tmp = tmp + 10;
337 			if (tmp >= timeout) {
338 				DRM_ERROR("failed to init MMSCH. TIME-OUT after %d usec"\
339 					" waiting for regMMSCH_VF_MAILBOX_RESP "\
340 					"(expected=0x%08x, readback=0x%08x)\n",
341 					tmp, expected, resp);
342 				return -EBUSY;
343 			}
344 		}
345 		if (resp != expected && resp != MMSCH_VF_MAILBOX_RESP__INCOMPLETE &&
346 				init_status != MMSCH_VF_ENGINE_STATUS__PASS)
347 			DRM_ERROR("MMSCH init status is incorrect! readback=0x%08x, header init status for jpeg: %x\n",
348 					resp, init_status);
349 
350 	}
351 	return 0;
352 }
353 
354 /**
355  * jpeg_v4_0_3_hw_init - start and test JPEG block
356  *
357  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
358  *
359  */
360 static int jpeg_v4_0_3_hw_init(struct amdgpu_ip_block *ip_block)
361 {
362 	struct amdgpu_device *adev = ip_block->adev;
363 	struct amdgpu_ring *ring;
364 	int i, j, r, jpeg_inst;
365 
366 	if (amdgpu_sriov_vf(adev)) {
367 		r = jpeg_v4_0_3_start_sriov(adev);
368 		if (r)
369 			return r;
370 
371 		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
372 			for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
373 				ring = &adev->jpeg.inst[i].ring_dec[j];
374 				ring->wptr = 0;
375 				ring->wptr_old = 0;
376 				jpeg_v4_0_3_dec_ring_set_wptr(ring);
377 				ring->sched.ready = true;
378 			}
379 		}
380 	} else {
381 		/* This flag is not set for VF, assumed to be disabled always */
382 		if (RREG32_SOC15(VCN, GET_INST(VCN, 0), regVCN_RRMT_CNTL) &
383 		    0x100)
384 			adev->jpeg.caps |= AMDGPU_JPEG_CAPS(RRMT_ENABLED);
385 
386 		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
387 			jpeg_inst = GET_INST(JPEG, i);
388 
389 			ring = adev->jpeg.inst[i].ring_dec;
390 
391 			if (ring->use_doorbell)
392 				adev->nbio.funcs->vcn_doorbell_range(
393 					adev, ring->use_doorbell,
394 					(adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
395 						9 * jpeg_inst,
396 					adev->jpeg.inst[i].aid_id);
397 
398 			for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
399 				ring = &adev->jpeg.inst[i].ring_dec[j];
400 				if (ring->use_doorbell)
401 					WREG32_SOC15_OFFSET(
402 						VCN, GET_INST(VCN, i),
403 						regVCN_JPEG_DB_CTRL,
404 						(ring->pipe ? (ring->pipe - 0x15) : 0),
405 						ring->doorbell_index
406 							<< VCN_JPEG_DB_CTRL__OFFSET__SHIFT |
407 							VCN_JPEG_DB_CTRL__EN_MASK);
408 				r = amdgpu_ring_test_helper(ring);
409 				if (r)
410 					return r;
411 			}
412 		}
413 	}
414 
415 	return 0;
416 }
417 
418 /**
419  * jpeg_v4_0_3_hw_fini - stop the hardware block
420  *
421  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
422  *
423  * Stop the JPEG block, mark ring as not ready any more
424  */
425 static int jpeg_v4_0_3_hw_fini(struct amdgpu_ip_block *ip_block)
426 {
427 	struct amdgpu_device *adev = ip_block->adev;
428 	int ret = 0;
429 
430 	cancel_delayed_work_sync(&adev->jpeg.idle_work);
431 
432 	if (!amdgpu_sriov_vf(adev)) {
433 		if (adev->jpeg.cur_state != AMD_PG_STATE_GATE)
434 			ret = jpeg_v4_0_3_set_powergating_state(ip_block, AMD_PG_STATE_GATE);
435 	}
436 
437 	return ret;
438 }
439 
440 /**
441  * jpeg_v4_0_3_suspend - suspend JPEG block
442  *
443  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
444  *
445  * HW fini and suspend JPEG block
446  */
447 static int jpeg_v4_0_3_suspend(struct amdgpu_ip_block *ip_block)
448 {
449 	int r;
450 
451 	r = jpeg_v4_0_3_hw_fini(ip_block);
452 	if (r)
453 		return r;
454 
455 	r = amdgpu_jpeg_suspend(ip_block->adev);
456 
457 	return r;
458 }
459 
460 /**
461  * jpeg_v4_0_3_resume - resume JPEG block
462  *
463  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
464  *
465  * Resume firmware and hw init JPEG block
466  */
467 static int jpeg_v4_0_3_resume(struct amdgpu_ip_block *ip_block)
468 {
469 	int r;
470 
471 	r = amdgpu_jpeg_resume(ip_block->adev);
472 	if (r)
473 		return r;
474 
475 	r = jpeg_v4_0_3_hw_init(ip_block);
476 
477 	return r;
478 }
479 
480 static void jpeg_v4_0_3_disable_clock_gating(struct amdgpu_device *adev, int inst_idx)
481 {
482 	int i, jpeg_inst;
483 	uint32_t data;
484 
485 	jpeg_inst = GET_INST(JPEG, inst_idx);
486 	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL);
487 	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
488 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
489 		data &= (~(JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1));
490 	} else {
491 		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
492 	}
493 
494 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
495 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
496 	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL, data);
497 
498 	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE);
499 	data &= ~(JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK);
500 	for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i)
501 		data &= ~(JPEG_CGC_GATE__JPEG0_DEC_MASK << i);
502 	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE, data);
503 }
504 
505 static void jpeg_v4_0_3_enable_clock_gating(struct amdgpu_device *adev, int inst_idx)
506 {
507 	int i, jpeg_inst;
508 	uint32_t data;
509 
510 	jpeg_inst = GET_INST(JPEG, inst_idx);
511 	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL);
512 	if (adev->cg_flags & AMD_CG_SUPPORT_JPEG_MGCG) {
513 		data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
514 		data |= (JPEG_CGC_CTRL__JPEG0_DEC_MODE_MASK << 1);
515 	} else {
516 		data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
517 	}
518 
519 	data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
520 	data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
521 	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_CTRL, data);
522 
523 	data = RREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE);
524 	data |= (JPEG_CGC_GATE__JMCIF_MASK | JPEG_CGC_GATE__JRBBM_MASK);
525 	for (i = 0; i < adev->jpeg.num_jpeg_rings; ++i)
526 		data |= (JPEG_CGC_GATE__JPEG0_DEC_MASK << i);
527 	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_CGC_GATE, data);
528 }
529 
530 static void jpeg_v4_0_3_start_inst(struct amdgpu_device *adev, int inst)
531 {
532 	int jpeg_inst = GET_INST(JPEG, inst);
533 
534 	WREG32_SOC15(JPEG, jpeg_inst, regUVD_PGFSM_CONFIG,
535 		     1 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT);
536 	SOC15_WAIT_ON_RREG(JPEG, jpeg_inst, regUVD_PGFSM_STATUS,
537 			   UVD_PGFSM_STATUS__UVDJ_PWR_ON <<
538 			   UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT,
539 			   UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK);
540 
541 	/* disable anti hang mechanism */
542 	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS),
543 		 0, ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
544 
545 	/* JPEG disable CGC */
546 	jpeg_v4_0_3_disable_clock_gating(adev, inst);
547 
548 	/* MJPEG global tiling registers */
549 	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_DEC_GFX8_ADDR_CONFIG,
550 		     adev->gfx.config.gb_addr_config);
551 	WREG32_SOC15(JPEG, jpeg_inst, regJPEG_DEC_GFX10_ADDR_CONFIG,
552 		     adev->gfx.config.gb_addr_config);
553 
554 	/* enable JMI channel */
555 	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL), 0,
556 		 ~UVD_JMI_CNTL__SOFT_RESET_MASK);
557 }
558 
559 static void jpeg_v4_0_3_start_jrbc(struct amdgpu_ring *ring)
560 {
561 	struct amdgpu_device *adev = ring->adev;
562 	int jpeg_inst = GET_INST(JPEG, ring->me);
563 	int reg_offset = jpeg_v4_0_3_core_reg_offset(ring->pipe);
564 
565 	/* enable System Interrupt for JRBC */
566 	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regJPEG_SYS_INT_EN),
567 		 JPEG_SYS_INT_EN__DJRBC0_MASK << ring->pipe,
568 		 ~(JPEG_SYS_INT_EN__DJRBC0_MASK << ring->pipe));
569 
570 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
571 			    regUVD_JMI0_UVD_LMI_JRBC_RB_VMID,
572 			    reg_offset, 0);
573 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
574 			    regUVD_JRBC0_UVD_JRBC_RB_CNTL,
575 			    reg_offset,
576 			    (0x00000001L | 0x00000002L));
577 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
578 			    regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_LOW,
579 			    reg_offset, lower_32_bits(ring->gpu_addr));
580 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
581 			    regUVD_JMI0_UVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
582 			    reg_offset, upper_32_bits(ring->gpu_addr));
583 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
584 			    regUVD_JRBC0_UVD_JRBC_RB_RPTR,
585 			    reg_offset, 0);
586 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
587 			    regUVD_JRBC0_UVD_JRBC_RB_WPTR,
588 			    reg_offset, 0);
589 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
590 			    regUVD_JRBC0_UVD_JRBC_RB_CNTL,
591 			    reg_offset, 0x00000002L);
592 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
593 			    regUVD_JRBC0_UVD_JRBC_RB_SIZE,
594 			    reg_offset, ring->ring_size / 4);
595 	ring->wptr = RREG32_SOC15_OFFSET(JPEG, jpeg_inst, regUVD_JRBC0_UVD_JRBC_RB_WPTR,
596 					 reg_offset);
597 }
598 
599 /**
600  * jpeg_v4_0_3_start - start JPEG block
601  *
602  * @adev: amdgpu_device pointer
603  *
604  * Setup and start the JPEG block
605  */
606 static int jpeg_v4_0_3_start(struct amdgpu_device *adev)
607 {
608 	struct amdgpu_ring *ring;
609 	int i, j;
610 
611 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
612 		jpeg_v4_0_3_start_inst(adev, i);
613 		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
614 			ring = &adev->jpeg.inst[i].ring_dec[j];
615 			jpeg_v4_0_3_start_jrbc(ring);
616 		}
617 	}
618 
619 	return 0;
620 }
621 
622 static void jpeg_v4_0_3_stop_inst(struct amdgpu_device *adev, int inst)
623 {
624 	int jpeg_inst = GET_INST(JPEG, inst);
625 	/* reset JMI */
626 	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JMI_CNTL),
627 		 UVD_JMI_CNTL__SOFT_RESET_MASK,
628 		 ~UVD_JMI_CNTL__SOFT_RESET_MASK);
629 
630 	jpeg_v4_0_3_enable_clock_gating(adev, inst);
631 
632 	/* enable anti hang mechanism */
633 	WREG32_P(SOC15_REG_OFFSET(JPEG, jpeg_inst, regUVD_JPEG_POWER_STATUS),
634 		 UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK,
635 		 ~UVD_JPEG_POWER_STATUS__JPEG_POWER_STATUS_MASK);
636 
637 	WREG32_SOC15(JPEG, jpeg_inst, regUVD_PGFSM_CONFIG,
638 		     2 << UVD_PGFSM_CONFIG__UVDJ_PWR_CONFIG__SHIFT);
639 	SOC15_WAIT_ON_RREG(JPEG, jpeg_inst, regUVD_PGFSM_STATUS,
640 			   UVD_PGFSM_STATUS__UVDJ_PWR_OFF <<
641 			   UVD_PGFSM_STATUS__UVDJ_PWR_STATUS__SHIFT,
642 			   UVD_PGFSM_STATUS__UVDJ_PWR_STATUS_MASK);
643 }
644 
645 /**
646  * jpeg_v4_0_3_stop - stop JPEG block
647  *
648  * @adev: amdgpu_device pointer
649  *
650  * stop the JPEG block
651  */
652 static int jpeg_v4_0_3_stop(struct amdgpu_device *adev)
653 {
654 	int i;
655 
656 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i)
657 		jpeg_v4_0_3_stop_inst(adev, i);
658 
659 	return 0;
660 }
661 
662 /**
663  * jpeg_v4_0_3_dec_ring_get_rptr - get read pointer
664  *
665  * @ring: amdgpu_ring pointer
666  *
667  * Returns the current hardware read pointer
668  */
669 static uint64_t jpeg_v4_0_3_dec_ring_get_rptr(struct amdgpu_ring *ring)
670 {
671 	struct amdgpu_device *adev = ring->adev;
672 
673 	return RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_RPTR,
674 				   jpeg_v4_0_3_core_reg_offset(ring->pipe));
675 }
676 
677 /**
678  * jpeg_v4_0_3_dec_ring_get_wptr - get write pointer
679  *
680  * @ring: amdgpu_ring pointer
681  *
682  * Returns the current hardware write pointer
683  */
684 static uint64_t jpeg_v4_0_3_dec_ring_get_wptr(struct amdgpu_ring *ring)
685 {
686 	struct amdgpu_device *adev = ring->adev;
687 
688 	if (ring->use_doorbell)
689 		return adev->wb.wb[ring->wptr_offs];
690 
691 	return RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_WPTR,
692 				   jpeg_v4_0_3_core_reg_offset(ring->pipe));
693 }
694 
695 static void jpeg_v4_0_3_ring_emit_hdp_flush(struct amdgpu_ring *ring)
696 {
697 	/* JPEG engine access for HDP flush doesn't work when RRMT is enabled.
698 	 * This is a workaround to avoid any HDP flush through JPEG ring.
699 	 */
700 }
701 
702 /**
703  * jpeg_v4_0_3_dec_ring_set_wptr - set write pointer
704  *
705  * @ring: amdgpu_ring pointer
706  *
707  * Commits the write pointer to the hardware
708  */
709 static void jpeg_v4_0_3_dec_ring_set_wptr(struct amdgpu_ring *ring)
710 {
711 	struct amdgpu_device *adev = ring->adev;
712 
713 	if (ring->use_doorbell) {
714 		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);
715 		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
716 	} else {
717 		WREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, ring->me), regUVD_JRBC0_UVD_JRBC_RB_WPTR,
718 				    jpeg_v4_0_3_core_reg_offset(ring->pipe),
719 				    lower_32_bits(ring->wptr));
720 	}
721 }
722 
723 /**
724  * jpeg_v4_0_3_dec_ring_insert_start - insert a start command
725  *
726  * @ring: amdgpu_ring pointer
727  *
728  * Write a start command to the ring.
729  */
730 void jpeg_v4_0_3_dec_ring_insert_start(struct amdgpu_ring *ring)
731 {
732 	if (!amdgpu_sriov_vf(ring->adev)) {
733 		amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
734 			0, 0, PACKETJ_TYPE0));
735 		amdgpu_ring_write(ring, 0x62a04); /* PCTL0_MMHUB_DEEPSLEEP_IB */
736 
737 		amdgpu_ring_write(ring,
738 				  PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR, 0,
739 					  0, PACKETJ_TYPE0));
740 		amdgpu_ring_write(ring, 0x80004000);
741 	}
742 }
743 
744 /**
745  * jpeg_v4_0_3_dec_ring_insert_end - insert a end command
746  *
747  * @ring: amdgpu_ring pointer
748  *
749  * Write a end command to the ring.
750  */
751 void jpeg_v4_0_3_dec_ring_insert_end(struct amdgpu_ring *ring)
752 {
753 	if (!amdgpu_sriov_vf(ring->adev)) {
754 		amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
755 			0, 0, PACKETJ_TYPE0));
756 		amdgpu_ring_write(ring, 0x62a04);
757 
758 		amdgpu_ring_write(ring,
759 				  PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR, 0,
760 					  0, PACKETJ_TYPE0));
761 		amdgpu_ring_write(ring, 0x00004000);
762 	}
763 }
764 
765 /**
766  * jpeg_v4_0_3_dec_ring_emit_fence - emit an fence & trap command
767  *
768  * @ring: amdgpu_ring pointer
769  * @addr: address
770  * @seq: sequence number
771  * @flags: fence related flags
772  *
773  * Write a fence and a trap command to the ring.
774  */
775 void jpeg_v4_0_3_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
776 				unsigned int flags)
777 {
778 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
779 
780 	amdgpu_ring_write(ring, PACKETJ(regUVD_JPEG_GPCOM_DATA0_INTERNAL_OFFSET,
781 		0, 0, PACKETJ_TYPE0));
782 	amdgpu_ring_write(ring, seq);
783 
784 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JPEG_GPCOM_DATA1_INTERNAL_OFFSET,
785 		0, 0, PACKETJ_TYPE0));
786 	amdgpu_ring_write(ring, seq);
787 
788 	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW_INTERNAL_OFFSET,
789 		0, 0, PACKETJ_TYPE0));
790 	amdgpu_ring_write(ring, lower_32_bits(addr));
791 
792 	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH_INTERNAL_OFFSET,
793 		0, 0, PACKETJ_TYPE0));
794 	amdgpu_ring_write(ring, upper_32_bits(addr));
795 
796 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET,
797 		0, 0, PACKETJ_TYPE0));
798 	amdgpu_ring_write(ring, 0x8);
799 
800 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JPEG_GPCOM_CMD_INTERNAL_OFFSET,
801 		0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE4));
802 	amdgpu_ring_write(ring, 0);
803 
804 	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
805 	amdgpu_ring_write(ring, 0);
806 
807 	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
808 	amdgpu_ring_write(ring, 0);
809 
810 	amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE7));
811 	amdgpu_ring_write(ring, 0);
812 }
813 
814 /**
815  * jpeg_v4_0_3_dec_ring_emit_ib - execute indirect buffer
816  *
817  * @ring: amdgpu_ring pointer
818  * @job: job to retrieve vmid from
819  * @ib: indirect buffer to execute
820  * @flags: unused
821  *
822  * Write ring commands to execute the indirect buffer.
823  */
824 void jpeg_v4_0_3_dec_ring_emit_ib(struct amdgpu_ring *ring,
825 				struct amdgpu_job *job,
826 				struct amdgpu_ib *ib,
827 				uint32_t flags)
828 {
829 	unsigned int vmid = AMDGPU_JOB_GET_VMID(job);
830 
831 	amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET,
832 		0, 0, PACKETJ_TYPE0));
833 
834 	if (ring->funcs->parse_cs)
835 		amdgpu_ring_write(ring, 0);
836 	else
837 		amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8)));
838 
839 	amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JPEG_VMID_INTERNAL_OFFSET,
840 		0, 0, PACKETJ_TYPE0));
841 	amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8)));
842 
843 	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET,
844 		0, 0, PACKETJ_TYPE0));
845 	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
846 
847 	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_IB_64BIT_BAR_HIGH_INTERNAL_OFFSET,
848 		0, 0, PACKETJ_TYPE0));
849 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
850 
851 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_IB_SIZE_INTERNAL_OFFSET,
852 		0, 0, PACKETJ_TYPE0));
853 	amdgpu_ring_write(ring, ib->length_dw);
854 
855 	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW_INTERNAL_OFFSET,
856 		0, 0, PACKETJ_TYPE0));
857 	amdgpu_ring_write(ring, lower_32_bits(ring->gpu_addr));
858 
859 	amdgpu_ring_write(ring,	PACKETJ(regUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH_INTERNAL_OFFSET,
860 		0, 0, PACKETJ_TYPE0));
861 	amdgpu_ring_write(ring, upper_32_bits(ring->gpu_addr));
862 
863 	amdgpu_ring_write(ring,	PACKETJ(0, 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE2));
864 	amdgpu_ring_write(ring, 0);
865 
866 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET,
867 		0, 0, PACKETJ_TYPE0));
868 	amdgpu_ring_write(ring, 0x01400200);
869 
870 	amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET,
871 		0, 0, PACKETJ_TYPE0));
872 	amdgpu_ring_write(ring, 0x2);
873 
874 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_STATUS_INTERNAL_OFFSET,
875 		0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE3));
876 	amdgpu_ring_write(ring, 0x2);
877 }
878 
879 void jpeg_v4_0_3_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
880 				uint32_t val, uint32_t mask)
881 {
882 	uint32_t reg_offset;
883 
884 	/* Use normalized offsets if required */
885 	if (jpeg_v4_0_3_normalizn_reqd(ring->adev))
886 		reg = NORMALIZE_JPEG_REG_OFFSET(reg);
887 
888 	reg_offset = (reg << 2);
889 
890 	amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_RB_COND_RD_TIMER_INTERNAL_OFFSET,
891 		0, 0, PACKETJ_TYPE0));
892 	amdgpu_ring_write(ring, 0x01400200);
893 
894 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_RB_REF_DATA_INTERNAL_OFFSET,
895 		0, 0, PACKETJ_TYPE0));
896 	amdgpu_ring_write(ring, val);
897 
898 	amdgpu_ring_write(ring, PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
899 		0, 0, PACKETJ_TYPE0));
900 	if (reg_offset >= 0x10000 && reg_offset <= 0x105ff) {
901 		amdgpu_ring_write(ring, 0);
902 		amdgpu_ring_write(ring,
903 			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3));
904 	} else {
905 		amdgpu_ring_write(ring, reg_offset);
906 		amdgpu_ring_write(ring,	PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
907 			0, 0, PACKETJ_TYPE3));
908 	}
909 	amdgpu_ring_write(ring, mask);
910 }
911 
912 void jpeg_v4_0_3_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
913 				unsigned int vmid, uint64_t pd_addr)
914 {
915 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
916 	uint32_t data0, data1, mask;
917 
918 	pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
919 
920 	/* wait for register write */
921 	data0 = hub->ctx0_ptb_addr_lo32 + vmid * hub->ctx_addr_distance;
922 	data1 = lower_32_bits(pd_addr);
923 	mask = 0xffffffff;
924 	jpeg_v4_0_3_dec_ring_emit_reg_wait(ring, data0, data1, mask);
925 }
926 
927 void jpeg_v4_0_3_dec_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val)
928 {
929 	uint32_t reg_offset;
930 
931 	/* Use normalized offsets if required */
932 	if (jpeg_v4_0_3_normalizn_reqd(ring->adev))
933 		reg = NORMALIZE_JPEG_REG_OFFSET(reg);
934 
935 	reg_offset = (reg << 2);
936 
937 	amdgpu_ring_write(ring,	PACKETJ(regUVD_JRBC_EXTERNAL_REG_INTERNAL_OFFSET,
938 		0, 0, PACKETJ_TYPE0));
939 	if (reg_offset >= 0x10000 && reg_offset <= 0x105ff) {
940 		amdgpu_ring_write(ring, 0);
941 		amdgpu_ring_write(ring,
942 			PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0));
943 	} else {
944 		amdgpu_ring_write(ring, reg_offset);
945 		amdgpu_ring_write(ring,	PACKETJ(JRBC_DEC_EXTERNAL_REG_WRITE_ADDR,
946 			0, 0, PACKETJ_TYPE0));
947 	}
948 	amdgpu_ring_write(ring, val);
949 }
950 
951 void jpeg_v4_0_3_dec_ring_nop(struct amdgpu_ring *ring, uint32_t count)
952 {
953 	int i;
954 
955 	WARN_ON(ring->wptr % 2 || count % 2);
956 
957 	for (i = 0; i < count / 2; i++) {
958 		amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6));
959 		amdgpu_ring_write(ring, 0);
960 	}
961 }
962 
963 static bool jpeg_v4_0_3_is_idle(void *handle)
964 {
965 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
966 	bool ret = false;
967 	int i, j;
968 
969 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
970 		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
971 			ret &= ((RREG32_SOC15_OFFSET(JPEG, GET_INST(JPEG, i),
972 				regUVD_JRBC0_UVD_JRBC_STATUS, jpeg_v4_0_3_core_reg_offset(j)) &
973 				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK) ==
974 				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK);
975 		}
976 	}
977 
978 	return ret;
979 }
980 
981 static int jpeg_v4_0_3_wait_for_idle(struct amdgpu_ip_block *ip_block)
982 {
983 	struct amdgpu_device *adev = ip_block->adev;
984 	int ret = 0;
985 	int i, j;
986 
987 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
988 		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
989 			ret &= (SOC15_WAIT_ON_RREG_OFFSET(JPEG, GET_INST(JPEG, i),
990 				regUVD_JRBC0_UVD_JRBC_STATUS, jpeg_v4_0_3_core_reg_offset(j),
991 				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK,
992 				UVD_JRBC0_UVD_JRBC_STATUS__RB_JOB_DONE_MASK));
993 		}
994 	}
995 	return ret;
996 }
997 
998 static int jpeg_v4_0_3_set_clockgating_state(struct amdgpu_ip_block *ip_block,
999 					  enum amd_clockgating_state state)
1000 {
1001 	struct amdgpu_device *adev = ip_block->adev;
1002 	bool enable = state == AMD_CG_STATE_GATE;
1003 	int i;
1004 
1005 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
1006 		if (enable) {
1007 			if (!jpeg_v4_0_3_is_idle(adev))
1008 				return -EBUSY;
1009 			jpeg_v4_0_3_enable_clock_gating(adev, i);
1010 		} else {
1011 			jpeg_v4_0_3_disable_clock_gating(adev, i);
1012 		}
1013 	}
1014 	return 0;
1015 }
1016 
1017 static int jpeg_v4_0_3_set_powergating_state(struct amdgpu_ip_block *ip_block,
1018 					  enum amd_powergating_state state)
1019 {
1020 	struct amdgpu_device *adev = ip_block->adev;
1021 	int ret;
1022 
1023 	if (amdgpu_sriov_vf(adev)) {
1024 		adev->jpeg.cur_state = AMD_PG_STATE_UNGATE;
1025 		return 0;
1026 	}
1027 
1028 	if (state == adev->jpeg.cur_state)
1029 		return 0;
1030 
1031 	if (state == AMD_PG_STATE_GATE)
1032 		ret = jpeg_v4_0_3_stop(adev);
1033 	else
1034 		ret = jpeg_v4_0_3_start(adev);
1035 
1036 	if (!ret)
1037 		adev->jpeg.cur_state = state;
1038 
1039 	return ret;
1040 }
1041 
1042 static int jpeg_v4_0_3_set_interrupt_state(struct amdgpu_device *adev,
1043 					struct amdgpu_irq_src *source,
1044 					unsigned int type,
1045 					enum amdgpu_interrupt_state state)
1046 {
1047 	return 0;
1048 }
1049 
1050 static int jpeg_v4_0_3_process_interrupt(struct amdgpu_device *adev,
1051 				      struct amdgpu_irq_src *source,
1052 				      struct amdgpu_iv_entry *entry)
1053 {
1054 	uint32_t i, inst;
1055 
1056 	i = node_id_to_phys_map[entry->node_id];
1057 	DRM_DEV_DEBUG(adev->dev, "IH: JPEG TRAP\n");
1058 
1059 	for (inst = 0; inst < adev->jpeg.num_jpeg_inst; ++inst)
1060 		if (adev->jpeg.inst[inst].aid_id == i)
1061 			break;
1062 
1063 	if (inst >= adev->jpeg.num_jpeg_inst) {
1064 		dev_WARN_ONCE(adev->dev, 1,
1065 			      "Interrupt received for unknown JPEG instance %d",
1066 			      entry->node_id);
1067 		return 0;
1068 	}
1069 
1070 	switch (entry->src_id) {
1071 	case VCN_4_0__SRCID__JPEG_DECODE:
1072 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[0]);
1073 		break;
1074 	case VCN_4_0__SRCID__JPEG1_DECODE:
1075 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[1]);
1076 		break;
1077 	case VCN_4_0__SRCID__JPEG2_DECODE:
1078 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[2]);
1079 		break;
1080 	case VCN_4_0__SRCID__JPEG3_DECODE:
1081 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[3]);
1082 		break;
1083 	case VCN_4_0__SRCID__JPEG4_DECODE:
1084 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[4]);
1085 		break;
1086 	case VCN_4_0__SRCID__JPEG5_DECODE:
1087 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[5]);
1088 		break;
1089 	case VCN_4_0__SRCID__JPEG6_DECODE:
1090 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[6]);
1091 		break;
1092 	case VCN_4_0__SRCID__JPEG7_DECODE:
1093 		amdgpu_fence_process(&adev->jpeg.inst[inst].ring_dec[7]);
1094 		break;
1095 	default:
1096 		DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n",
1097 			  entry->src_id, entry->src_data[0]);
1098 		break;
1099 	}
1100 
1101 	return 0;
1102 }
1103 
1104 static void jpeg_v4_0_3_core_stall_reset(struct amdgpu_ring *ring)
1105 {
1106 	struct amdgpu_device *adev = ring->adev;
1107 	int jpeg_inst = GET_INST(JPEG, ring->me);
1108 	int reg_offset = jpeg_v4_0_3_core_reg_offset(ring->pipe);
1109 
1110 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
1111 			    regUVD_JMI0_UVD_JMI_CLIENT_STALL,
1112 			    reg_offset, 0x1F);
1113 	SOC15_WAIT_ON_RREG(JPEG, jpeg_inst,
1114 			   regUVD_JMI0_UVD_JMI_CLIENT_CLEAN_STATUS,
1115 			   0x1F, 0x1F);
1116 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
1117 			    regUVD_JMI0_JPEG_LMI_DROP,
1118 			    reg_offset, 0x1F);
1119 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
1120 			    regJPEG_CORE_RST_CTRL,
1121 			    reg_offset, 1 << ring->pipe);
1122 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
1123 			    regUVD_JMI0_UVD_JMI_CLIENT_STALL,
1124 			    reg_offset, 0x00);
1125 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
1126 			    regUVD_JMI0_JPEG_LMI_DROP,
1127 			    reg_offset, 0x00);
1128 	WREG32_SOC15_OFFSET(JPEG, jpeg_inst,
1129 			    regJPEG_CORE_RST_CTRL,
1130 			    reg_offset, 0x00);
1131 }
1132 
1133 static int jpeg_v4_0_3_ring_reset(struct amdgpu_ring *ring, unsigned int vmid)
1134 {
1135 	if (amdgpu_sriov_vf(ring->adev))
1136 		return -EOPNOTSUPP;
1137 
1138 	jpeg_v4_0_3_core_stall_reset(ring);
1139 	jpeg_v4_0_3_start_jrbc(ring);
1140 	return amdgpu_ring_test_helper(ring);
1141 }
1142 
1143 static const struct amd_ip_funcs jpeg_v4_0_3_ip_funcs = {
1144 	.name = "jpeg_v4_0_3",
1145 	.early_init = jpeg_v4_0_3_early_init,
1146 	.sw_init = jpeg_v4_0_3_sw_init,
1147 	.sw_fini = jpeg_v4_0_3_sw_fini,
1148 	.hw_init = jpeg_v4_0_3_hw_init,
1149 	.hw_fini = jpeg_v4_0_3_hw_fini,
1150 	.suspend = jpeg_v4_0_3_suspend,
1151 	.resume = jpeg_v4_0_3_resume,
1152 	.is_idle = jpeg_v4_0_3_is_idle,
1153 	.wait_for_idle = jpeg_v4_0_3_wait_for_idle,
1154 	.set_clockgating_state = jpeg_v4_0_3_set_clockgating_state,
1155 	.set_powergating_state = jpeg_v4_0_3_set_powergating_state,
1156 	.dump_ip_state = amdgpu_jpeg_dump_ip_state,
1157 	.print_ip_state = amdgpu_jpeg_print_ip_state,
1158 };
1159 
1160 static const struct amdgpu_ring_funcs jpeg_v4_0_3_dec_ring_vm_funcs = {
1161 	.type = AMDGPU_RING_TYPE_VCN_JPEG,
1162 	.align_mask = 0xf,
1163 	.get_rptr = jpeg_v4_0_3_dec_ring_get_rptr,
1164 	.get_wptr = jpeg_v4_0_3_dec_ring_get_wptr,
1165 	.set_wptr = jpeg_v4_0_3_dec_ring_set_wptr,
1166 	.parse_cs = jpeg_v2_dec_ring_parse_cs,
1167 	.emit_frame_size =
1168 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
1169 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
1170 		8 + /* jpeg_v4_0_3_dec_ring_emit_vm_flush */
1171 		18 + 18 + /* jpeg_v4_0_3_dec_ring_emit_fence x2 vm fence */
1172 		8 + 16,
1173 	.emit_ib_size = 22, /* jpeg_v4_0_3_dec_ring_emit_ib */
1174 	.emit_ib = jpeg_v4_0_3_dec_ring_emit_ib,
1175 	.emit_fence = jpeg_v4_0_3_dec_ring_emit_fence,
1176 	.emit_vm_flush = jpeg_v4_0_3_dec_ring_emit_vm_flush,
1177 	.emit_hdp_flush = jpeg_v4_0_3_ring_emit_hdp_flush,
1178 	.test_ring = amdgpu_jpeg_dec_ring_test_ring,
1179 	.test_ib = amdgpu_jpeg_dec_ring_test_ib,
1180 	.insert_nop = jpeg_v4_0_3_dec_ring_nop,
1181 	.insert_start = jpeg_v4_0_3_dec_ring_insert_start,
1182 	.insert_end = jpeg_v4_0_3_dec_ring_insert_end,
1183 	.pad_ib = amdgpu_ring_generic_pad_ib,
1184 	.begin_use = amdgpu_jpeg_ring_begin_use,
1185 	.end_use = amdgpu_jpeg_ring_end_use,
1186 	.emit_wreg = jpeg_v4_0_3_dec_ring_emit_wreg,
1187 	.emit_reg_wait = jpeg_v4_0_3_dec_ring_emit_reg_wait,
1188 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
1189 	.reset = jpeg_v4_0_3_ring_reset,
1190 };
1191 
1192 static void jpeg_v4_0_3_set_dec_ring_funcs(struct amdgpu_device *adev)
1193 {
1194 	int i, j, jpeg_inst;
1195 
1196 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
1197 		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
1198 			adev->jpeg.inst[i].ring_dec[j].funcs = &jpeg_v4_0_3_dec_ring_vm_funcs;
1199 			adev->jpeg.inst[i].ring_dec[j].me = i;
1200 			adev->jpeg.inst[i].ring_dec[j].pipe = j;
1201 		}
1202 		jpeg_inst = GET_INST(JPEG, i);
1203 		adev->jpeg.inst[i].aid_id =
1204 			jpeg_inst / adev->jpeg.num_inst_per_aid;
1205 	}
1206 }
1207 
1208 static const struct amdgpu_irq_src_funcs jpeg_v4_0_3_irq_funcs = {
1209 	.set = jpeg_v4_0_3_set_interrupt_state,
1210 	.process = jpeg_v4_0_3_process_interrupt,
1211 };
1212 
1213 static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev)
1214 {
1215 	int i;
1216 
1217 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
1218 		adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings;
1219 	}
1220 	adev->jpeg.inst->irq.funcs = &jpeg_v4_0_3_irq_funcs;
1221 }
1222 
1223 const struct amdgpu_ip_block_version jpeg_v4_0_3_ip_block = {
1224 	.type = AMD_IP_BLOCK_TYPE_JPEG,
1225 	.major = 4,
1226 	.minor = 0,
1227 	.rev = 3,
1228 	.funcs = &jpeg_v4_0_3_ip_funcs,
1229 };
1230 
1231 static const struct amdgpu_ras_err_status_reg_entry jpeg_v4_0_3_ue_reg_list[] = {
1232 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG0S, regVCN_UE_ERR_STATUS_HI_JPEG0S),
1233 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG0S"},
1234 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG0D, regVCN_UE_ERR_STATUS_HI_JPEG0D),
1235 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG0D"},
1236 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG1S, regVCN_UE_ERR_STATUS_HI_JPEG1S),
1237 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG1S"},
1238 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG1D, regVCN_UE_ERR_STATUS_HI_JPEG1D),
1239 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG1D"},
1240 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG2S, regVCN_UE_ERR_STATUS_HI_JPEG2S),
1241 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG2S"},
1242 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG2D, regVCN_UE_ERR_STATUS_HI_JPEG2D),
1243 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG2D"},
1244 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG3S, regVCN_UE_ERR_STATUS_HI_JPEG3S),
1245 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG3S"},
1246 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG3D, regVCN_UE_ERR_STATUS_HI_JPEG3D),
1247 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG3D"},
1248 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG4S, regVCN_UE_ERR_STATUS_HI_JPEG4S),
1249 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG4S"},
1250 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG4D, regVCN_UE_ERR_STATUS_HI_JPEG4D),
1251 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG4D"},
1252 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG5S, regVCN_UE_ERR_STATUS_HI_JPEG5S),
1253 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG5S"},
1254 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG5D, regVCN_UE_ERR_STATUS_HI_JPEG5D),
1255 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG5D"},
1256 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG6S, regVCN_UE_ERR_STATUS_HI_JPEG6S),
1257 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG6S"},
1258 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG6D, regVCN_UE_ERR_STATUS_HI_JPEG6D),
1259 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG6D"},
1260 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG7S, regVCN_UE_ERR_STATUS_HI_JPEG7S),
1261 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG7S"},
1262 	{AMDGPU_RAS_REG_ENTRY(JPEG, 0, regVCN_UE_ERR_STATUS_LO_JPEG7D, regVCN_UE_ERR_STATUS_HI_JPEG7D),
1263 	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "JPEG7D"},
1264 };
1265 
1266 static void jpeg_v4_0_3_inst_query_ras_error_count(struct amdgpu_device *adev,
1267 						   uint32_t jpeg_inst,
1268 						   void *ras_err_status)
1269 {
1270 	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
1271 
1272 	/* jpeg v4_0_3 only support uncorrectable errors */
1273 	amdgpu_ras_inst_query_ras_error_count(adev,
1274 			jpeg_v4_0_3_ue_reg_list,
1275 			ARRAY_SIZE(jpeg_v4_0_3_ue_reg_list),
1276 			NULL, 0, GET_INST(VCN, jpeg_inst),
1277 			AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
1278 			&err_data->ue_count);
1279 }
1280 
1281 static void jpeg_v4_0_3_query_ras_error_count(struct amdgpu_device *adev,
1282 					      void *ras_err_status)
1283 {
1284 	uint32_t i;
1285 
1286 	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
1287 		dev_warn(adev->dev, "JPEG RAS is not supported\n");
1288 		return;
1289 	}
1290 
1291 	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++)
1292 		jpeg_v4_0_3_inst_query_ras_error_count(adev, i, ras_err_status);
1293 }
1294 
1295 static void jpeg_v4_0_3_inst_reset_ras_error_count(struct amdgpu_device *adev,
1296 						   uint32_t jpeg_inst)
1297 {
1298 	amdgpu_ras_inst_reset_ras_error_count(adev,
1299 			jpeg_v4_0_3_ue_reg_list,
1300 			ARRAY_SIZE(jpeg_v4_0_3_ue_reg_list),
1301 			GET_INST(VCN, jpeg_inst));
1302 }
1303 
1304 static void jpeg_v4_0_3_reset_ras_error_count(struct amdgpu_device *adev)
1305 {
1306 	uint32_t i;
1307 
1308 	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) {
1309 		dev_warn(adev->dev, "JPEG RAS is not supported\n");
1310 		return;
1311 	}
1312 
1313 	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++)
1314 		jpeg_v4_0_3_inst_reset_ras_error_count(adev, i);
1315 }
1316 
1317 static const struct amdgpu_ras_block_hw_ops jpeg_v4_0_3_ras_hw_ops = {
1318 	.query_ras_error_count = jpeg_v4_0_3_query_ras_error_count,
1319 	.reset_ras_error_count = jpeg_v4_0_3_reset_ras_error_count,
1320 };
1321 
1322 static int jpeg_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
1323 				      enum aca_smu_type type, void *data)
1324 {
1325 	struct aca_bank_info info;
1326 	u64 misc0;
1327 	int ret;
1328 
1329 	ret = aca_bank_info_decode(bank, &info);
1330 	if (ret)
1331 		return ret;
1332 
1333 	misc0 = bank->regs[ACA_REG_IDX_MISC0];
1334 	switch (type) {
1335 	case ACA_SMU_TYPE_UE:
1336 		bank->aca_err_type = ACA_ERROR_TYPE_UE;
1337 		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE,
1338 						     1ULL);
1339 		break;
1340 	case ACA_SMU_TYPE_CE:
1341 		bank->aca_err_type = ACA_ERROR_TYPE_CE;
1342 		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_CE,
1343 						     ACA_REG__MISC0__ERRCNT(misc0));
1344 		break;
1345 	default:
1346 		return -EINVAL;
1347 	}
1348 
1349 	return ret;
1350 }
1351 
1352 /* reference to smu driver if header file */
1353 static int jpeg_v4_0_3_err_codes[] = {
1354 	16, 17, 18, 19, 20, 21, 22, 23, /* JPEG[0-7][S|D] */
1355 	24, 25, 26, 27, 28, 29, 30, 31
1356 };
1357 
1358 static bool jpeg_v4_0_3_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
1359 					 enum aca_smu_type type, void *data)
1360 {
1361 	u32 instlo;
1362 
1363 	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
1364 	instlo &= GENMASK(31, 1);
1365 
1366 	if (instlo != mmSMNAID_AID0_MCA_SMU)
1367 		return false;
1368 
1369 	if (aca_bank_check_error_codes(handle->adev, bank,
1370 				       jpeg_v4_0_3_err_codes,
1371 				       ARRAY_SIZE(jpeg_v4_0_3_err_codes)))
1372 		return false;
1373 
1374 	return true;
1375 }
1376 
1377 static const struct aca_bank_ops jpeg_v4_0_3_aca_bank_ops = {
1378 	.aca_bank_parser = jpeg_v4_0_3_aca_bank_parser,
1379 	.aca_bank_is_valid = jpeg_v4_0_3_aca_bank_is_valid,
1380 };
1381 
1382 static const struct aca_info jpeg_v4_0_3_aca_info = {
1383 	.hwip = ACA_HWIP_TYPE_SMU,
1384 	.mask = ACA_ERROR_UE_MASK,
1385 	.bank_ops = &jpeg_v4_0_3_aca_bank_ops,
1386 };
1387 
1388 static int jpeg_v4_0_3_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
1389 {
1390 	int r;
1391 
1392 	r = amdgpu_ras_block_late_init(adev, ras_block);
1393 	if (r)
1394 		return r;
1395 
1396 	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__JPEG,
1397 				&jpeg_v4_0_3_aca_info, NULL);
1398 	if (r)
1399 		goto late_fini;
1400 
1401 	return 0;
1402 
1403 late_fini:
1404 	amdgpu_ras_block_late_fini(adev, ras_block);
1405 
1406 	return r;
1407 }
1408 
1409 static struct amdgpu_jpeg_ras jpeg_v4_0_3_ras = {
1410 	.ras_block = {
1411 		.hw_ops = &jpeg_v4_0_3_ras_hw_ops,
1412 		.ras_late_init = jpeg_v4_0_3_ras_late_init,
1413 	},
1414 };
1415 
1416 static void jpeg_v4_0_3_set_ras_funcs(struct amdgpu_device *adev)
1417 {
1418 	adev->jpeg.ras = &jpeg_v4_0_3_ras;
1419 }
1420