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  * Authors: AMD
23  *
24  */
25 
26 
27 #include "dm_services.h"
28 #include "dm_helpers.h"
29 #include "core_types.h"
30 #include "resource.h"
31 #include "dccg.h"
32 #include "dce/dce_hwseq.h"
33 #include "dcn30/dcn30_cm_common.h"
34 #include "reg_helper.h"
35 #include "abm.h"
36 #include "hubp.h"
37 #include "dchubbub.h"
38 #include "timing_generator.h"
39 #include "opp.h"
40 #include "ipp.h"
41 #include "mpc.h"
42 #include "mcif_wb.h"
43 #include "dc_dmub_srv.h"
44 #include "link_hwss.h"
45 #include "dpcd_defs.h"
46 #include "dcn32_hwseq.h"
47 #include "clk_mgr.h"
48 #include "dsc.h"
49 #include "dcn20/dcn20_optc.h"
50 #include "dc_link_dp.h"
51 
52 #define DC_LOGGER_INIT(logger)
53 
54 #define CTX \
55 	hws->ctx
56 #define REG(reg)\
57 	hws->regs->reg
58 #define DC_LOGGER \
59 		dc->ctx->logger
60 
61 
62 #undef FN
63 #define FN(reg_name, field_name) \
64 	hws->shifts->field_name, hws->masks->field_name
65 
66 void dcn32_dsc_pg_control(
67 		struct dce_hwseq *hws,
68 		unsigned int dsc_inst,
69 		bool power_on)
70 {
71 	uint32_t power_gate = power_on ? 0 : 1;
72 	uint32_t pwr_status = power_on ? 0 : 2;
73 	uint32_t org_ip_request_cntl = 0;
74 
75 	if (hws->ctx->dc->debug.disable_dsc_power_gate)
76 		return;
77 
78 	REG_GET(DC_IP_REQUEST_CNTL, IP_REQUEST_EN, &org_ip_request_cntl);
79 	if (org_ip_request_cntl == 0)
80 		REG_SET(DC_IP_REQUEST_CNTL, 0, IP_REQUEST_EN, 1);
81 
82 	switch (dsc_inst) {
83 	case 0: /* DSC0 */
84 		REG_UPDATE(DOMAIN16_PG_CONFIG,
85 				DOMAIN_POWER_GATE, power_gate);
86 
87 		REG_WAIT(DOMAIN16_PG_STATUS,
88 				DOMAIN_PGFSM_PWR_STATUS, pwr_status,
89 				1, 1000);
90 		break;
91 	case 1: /* DSC1 */
92 		REG_UPDATE(DOMAIN17_PG_CONFIG,
93 				DOMAIN_POWER_GATE, power_gate);
94 
95 		REG_WAIT(DOMAIN17_PG_STATUS,
96 				DOMAIN_PGFSM_PWR_STATUS, pwr_status,
97 				1, 1000);
98 		break;
99 	case 2: /* DSC2 */
100 		REG_UPDATE(DOMAIN18_PG_CONFIG,
101 				DOMAIN_POWER_GATE, power_gate);
102 
103 		REG_WAIT(DOMAIN18_PG_STATUS,
104 				DOMAIN_PGFSM_PWR_STATUS, pwr_status,
105 				1, 1000);
106 		break;
107 	case 3: /* DSC3 */
108 		REG_UPDATE(DOMAIN19_PG_CONFIG,
109 				DOMAIN_POWER_GATE, power_gate);
110 
111 		REG_WAIT(DOMAIN19_PG_STATUS,
112 				DOMAIN_PGFSM_PWR_STATUS, pwr_status,
113 				1, 1000);
114 		break;
115 	default:
116 		BREAK_TO_DEBUGGER();
117 		break;
118 	}
119 
120 	if (org_ip_request_cntl == 0)
121 		REG_SET(DC_IP_REQUEST_CNTL, 0, IP_REQUEST_EN, 0);
122 }
123 
124 
125 void dcn32_enable_power_gating_plane(
126 	struct dce_hwseq *hws,
127 	bool enable)
128 {
129 	bool force_on = true; /* disable power gating */
130 
131 	if (enable)
132 		force_on = false;
133 
134 	/* DCHUBP0/1/2/3 */
135 	REG_UPDATE(DOMAIN0_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
136 	REG_UPDATE(DOMAIN1_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
137 	REG_UPDATE(DOMAIN2_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
138 	REG_UPDATE(DOMAIN3_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
139 
140 	/* DCS0/1/2/3 */
141 	REG_UPDATE(DOMAIN16_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
142 	REG_UPDATE(DOMAIN17_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
143 	REG_UPDATE(DOMAIN18_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
144 	REG_UPDATE(DOMAIN19_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
145 }
146 
147 void dcn32_hubp_pg_control(struct dce_hwseq *hws, unsigned int hubp_inst, bool power_on)
148 {
149 	uint32_t power_gate = power_on ? 0 : 1;
150 	uint32_t pwr_status = power_on ? 0 : 2;
151 
152 	if (hws->ctx->dc->debug.disable_hubp_power_gate)
153 		return;
154 
155 	if (REG(DOMAIN0_PG_CONFIG) == 0)
156 		return;
157 
158 	switch (hubp_inst) {
159 	case 0:
160 		REG_SET(DOMAIN0_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
161 		REG_WAIT(DOMAIN0_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
162 		break;
163 	case 1:
164 		REG_SET(DOMAIN1_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
165 		REG_WAIT(DOMAIN1_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
166 		break;
167 	case 2:
168 		REG_SET(DOMAIN2_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
169 		REG_WAIT(DOMAIN2_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
170 		break;
171 	case 3:
172 		REG_SET(DOMAIN3_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
173 		REG_WAIT(DOMAIN3_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
174 		break;
175 	default:
176 		BREAK_TO_DEBUGGER();
177 		break;
178 	}
179 }
180 
181 static bool dcn32_check_no_memory_request_for_cab(struct dc *dc)
182 {
183 	int i;
184 
185     /* First, check no-memory-request case */
186 	for (i = 0; i < dc->current_state->stream_count; i++) {
187 		if (dc->current_state->stream_status[i].plane_count)
188 			/* Fail eligibility on a visible stream */
189 			break;
190 	}
191 
192 	if (i == dc->current_state->stream_count)
193 		return true;
194 
195 	return false;
196 }
197 
198 /* This function takes in the start address and surface size to be cached in CAB
199  * and calculates the total number of cache lines required to store the surface.
200  * The number of cache lines used for each surface is calculated independently of
201  * one another. For example, if there is a primary surface(1), meta surface(2), and
202  * cursor(3), this function should be called 3 times to calculate the number of cache
203  * lines used for each of those surfaces.
204  */
205 static uint32_t dcn32_cache_lines_for_surface(struct dc *dc, uint32_t surface_size, uint64_t start_address)
206 {
207 	uint32_t lines_used = 1;
208 	uint32_t num_cached_bytes = 0;
209 	uint32_t remaining_size = 0;
210 	uint32_t cache_line_size = dc->caps.cache_line_size;
211 
212 	/* 1. Calculate surface size minus the number of bytes stored
213 	 * in the first cache line (all bytes in first cache line might
214 	 * not be fully used).
215 	 */
216 	num_cached_bytes = cache_line_size - (start_address % cache_line_size);
217 	remaining_size = surface_size - num_cached_bytes;
218 
219 	/* 2. Calculate number of cache lines that will be fully used with
220 	 * the remaining number of bytes to be stored.
221 	 */
222 	lines_used += (remaining_size / cache_line_size);
223 
224 	/* 3. Check if we need an extra line due to the remaining size not being
225 	 * a multiple of CACHE_LINE_SIZE.
226 	 */
227 	if (remaining_size % cache_line_size > 0)
228 		lines_used++;
229 
230 	return lines_used;
231 }
232 
233 /* This function loops through every surface that needs to be cached in CAB for SS,
234  * and calculates the total number of ways required to store all surfaces (primary,
235  * meta, cursor).
236  */
237 static uint32_t dcn32_calculate_cab_allocation(struct dc *dc, struct dc_state *ctx)
238 {
239 	uint8_t i, j;
240 	struct dc_stream_state *stream = NULL;
241 	struct dc_plane_state *plane = NULL;
242 	uint32_t surface_size = 0;
243 	uint32_t cursor_size = 0;
244 	uint32_t cache_lines_used = 0;
245 	uint32_t total_lines = 0;
246 	uint32_t lines_per_way = 0;
247 	uint32_t num_ways = 0;
248 
249 	for (i = 0; i < ctx->stream_count; i++) {
250 		stream = ctx->streams[i];
251 
252 		// Don't include PSR surface in the total surface size for CAB allocation
253 		if (stream->link->psr_settings.psr_version != DC_PSR_VERSION_UNSUPPORTED)
254 			continue;
255 
256 		if (ctx->stream_status[i].plane_count == 0)
257 			continue;
258 
259 		// For each stream, loop through each plane to calculate the number of cache
260 		// lines required to store the surface in CAB
261 		for (j = 0; j < ctx->stream_status[i].plane_count; j++) {
262 			plane = ctx->stream_status[i].plane_states[j];
263 
264 			// Calculate total surface size
265 			surface_size = plane->plane_size.surface_pitch *
266 					plane->plane_size.surface_size.height *
267 					(plane->format >= SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616 ? 8 : 4);
268 
269 			// Convert surface size + starting address to number of cache lines required
270 			// (alignment accounted for)
271 			cache_lines_used += dcn32_cache_lines_for_surface(dc, surface_size,
272 					plane->address.grph.addr.quad_part);
273 
274 			if (plane->address.grph.meta_addr.quad_part) {
275 				// Meta surface
276 				cache_lines_used += dcn32_cache_lines_for_surface(dc, surface_size,
277 						plane->address.grph.meta_addr.quad_part);
278 			}
279 		}
280 
281 		// Include cursor size for CAB allocation
282 		if (stream->cursor_position.enable && plane->address.grph.cursor_cache_addr.quad_part) {
283 			cursor_size = dc->caps.max_cursor_size * dc->caps.max_cursor_size;
284 			switch (stream->cursor_attributes.color_format) {
285 			case CURSOR_MODE_MONO:
286 				cursor_size /= 2;
287 				break;
288 			case CURSOR_MODE_COLOR_1BIT_AND:
289 			case CURSOR_MODE_COLOR_PRE_MULTIPLIED_ALPHA:
290 			case CURSOR_MODE_COLOR_UN_PRE_MULTIPLIED_ALPHA:
291 				cursor_size *= 4;
292 				break;
293 
294 			case CURSOR_MODE_COLOR_64BIT_FP_PRE_MULTIPLIED:
295 			case CURSOR_MODE_COLOR_64BIT_FP_UN_PRE_MULTIPLIED:
296 				cursor_size *= 8;
297 				break;
298 			}
299 			cache_lines_used += dcn32_cache_lines_for_surface(dc, surface_size,
300 					plane->address.grph.cursor_cache_addr.quad_part);
301 		}
302 	}
303 
304 	// Convert number of cache lines required to number of ways
305 	total_lines = dc->caps.max_cab_allocation_bytes / dc->caps.cache_line_size;
306 	lines_per_way = total_lines / dc->caps.cache_num_ways;
307 	num_ways = cache_lines_used / lines_per_way;
308 
309 	if (cache_lines_used % lines_per_way > 0)
310 		num_ways++;
311 
312 	return num_ways;
313 }
314 
315 bool dcn32_apply_idle_power_optimizations(struct dc *dc, bool enable)
316 {
317 	union dmub_rb_cmd cmd;
318 	uint8_t ways;
319 
320 	if (!dc->ctx->dmub_srv)
321 		return false;
322 
323 	if (enable) {
324 		if (dc->current_state) {
325 
326 			/* 1. Check no memory request case for CAB.
327 			 * If no memory request case, send CAB_ACTION NO_DF_REQ DMUB message
328 			 */
329 			if (dcn32_check_no_memory_request_for_cab(dc)) {
330 				/* Enable no-memory-requests case */
331 				memset(&cmd, 0, sizeof(cmd));
332 				cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
333 				cmd.cab.header.sub_type = DMUB_CMD__CAB_NO_DCN_REQ;
334 				cmd.cab.header.payload_bytes = sizeof(cmd.cab) - sizeof(cmd.cab.header);
335 
336 				dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
337 				dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
338 
339 				return true;
340 			}
341 
342 			/* 2. Check if all surfaces can fit in CAB.
343 			 * If surfaces can fit into CAB, send CAB_ACTION_ALLOW DMUB message
344 			 * and configure HUBP's to fetch from MALL
345 			 */
346 			ways = dcn32_calculate_cab_allocation(dc, dc->current_state);
347 			if (ways <= dc->caps.cache_num_ways) {
348 				memset(&cmd, 0, sizeof(cmd));
349 				cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
350 				cmd.cab.header.sub_type = DMUB_CMD__CAB_DCN_SS_FIT_IN_CAB;
351 				cmd.cab.header.payload_bytes = sizeof(cmd.cab) - sizeof(cmd.cab.header);
352 				cmd.cab.cab_alloc_ways = ways;
353 
354 				dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
355 				dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
356 
357 				return true;
358 			}
359 
360 		}
361 		return false;
362 	}
363 
364 	/* Disable CAB */
365 	memset(&cmd, 0, sizeof(cmd));
366 	cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
367 	cmd.cab.header.sub_type = DMUB_CMD__CAB_NO_IDLE_OPTIMIZATION;
368 	cmd.cab.header.payload_bytes =
369 			sizeof(cmd.cab) - sizeof(cmd.cab.header);
370 
371 	dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
372 	dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
373 	dc_dmub_srv_wait_idle(dc->ctx->dmub_srv);
374 
375 	return false;
376 }
377 
378 /* Send DMCUB message with SubVP pipe info
379  * - For each pipe in context, populate payload with required SubVP information
380  *   if the pipe is using SubVP for MCLK switch
381  * - This function must be called while the DMUB HW lock is acquired by driver
382  */
383 void dcn32_commit_subvp_config(struct dc *dc, struct dc_state *context)
384 {
385 /*
386 	int i;
387 	bool enable_subvp = false;
388 
389 	if (!dc->ctx || !dc->ctx->dmub_srv)
390 		return;
391 
392 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
393 		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
394 
395 		if (pipe_ctx->stream && pipe_ctx->stream->mall_stream_config.paired_stream &&
396 				pipe_ctx->stream->mall_stream_config.type == SUBVP_MAIN) {
397 			// There is at least 1 SubVP pipe, so enable SubVP
398 			enable_subvp = true;
399 			break;
400 		}
401 	}
402 	dc_dmub_setup_subvp_dmub_command(dc, context, enable_subvp);
403 */
404 }
405 
406 static bool dcn32_set_mpc_shaper_3dlut(
407 	struct pipe_ctx *pipe_ctx, const struct dc_stream_state *stream)
408 {
409 	struct dpp *dpp_base = pipe_ctx->plane_res.dpp;
410 	int mpcc_id = pipe_ctx->plane_res.hubp->inst;
411 	struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc;
412 	bool result = false;
413 
414 	const struct pwl_params *shaper_lut = NULL;
415 	//get the shaper lut params
416 	if (stream->func_shaper) {
417 		if (stream->func_shaper->type == TF_TYPE_HWPWL)
418 			shaper_lut = &stream->func_shaper->pwl;
419 		else if (stream->func_shaper->type == TF_TYPE_DISTRIBUTED_POINTS) {
420 			cm_helper_translate_curve_to_hw_format(
421 					stream->func_shaper,
422 					&dpp_base->shaper_params, true);
423 			shaper_lut = &dpp_base->shaper_params;
424 		}
425 	}
426 
427 	if (stream->lut3d_func &&
428 		stream->lut3d_func->state.bits.initialized == 1) {
429 
430 		result = mpc->funcs->program_3dlut(mpc,
431 								&stream->lut3d_func->lut_3d,
432 								mpcc_id);
433 
434 		result = mpc->funcs->program_shaper(mpc,
435 								shaper_lut,
436 								mpcc_id);
437 	}
438 
439 	return result;
440 }
441 bool dcn32_set_output_transfer_func(struct dc *dc,
442 				struct pipe_ctx *pipe_ctx,
443 				const struct dc_stream_state *stream)
444 {
445 	int mpcc_id = pipe_ctx->plane_res.hubp->inst;
446 	struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc;
447 	struct pwl_params *params = NULL;
448 	bool ret = false;
449 
450 	/* program OGAM or 3DLUT only for the top pipe*/
451 	if (pipe_ctx->top_pipe == NULL) {
452 		/*program shaper and 3dlut in MPC*/
453 		ret = dcn32_set_mpc_shaper_3dlut(pipe_ctx, stream);
454 		if (ret == false && mpc->funcs->set_output_gamma && stream->out_transfer_func) {
455 			if (stream->out_transfer_func->type == TF_TYPE_HWPWL)
456 				params = &stream->out_transfer_func->pwl;
457 			else if (pipe_ctx->stream->out_transfer_func->type ==
458 					TF_TYPE_DISTRIBUTED_POINTS &&
459 					cm3_helper_translate_curve_to_hw_format(
460 					stream->out_transfer_func,
461 					&mpc->blender_params, false))
462 				params = &mpc->blender_params;
463 		 /* there are no ROM LUTs in OUTGAM */
464 		if (stream->out_transfer_func->type == TF_TYPE_PREDEFINED)
465 			BREAK_TO_DEBUGGER();
466 		}
467 	}
468 
469 	mpc->funcs->set_output_gamma(mpc, mpcc_id, params);
470 	return ret;
471 }
472 
473 /* Program P-State force value according to if pipe is using SubVP or not:
474  * 1. Reset P-State force on all pipes first
475  * 2. For each main pipe, force P-State disallow (P-State allow moderated by DMUB)
476  */
477 void dcn32_subvp_update_force_pstate(struct dc *dc, struct dc_state *context)
478 {
479 	int i;
480 	int num_subvp = 0;
481 	/* Unforce p-state for each pipe
482 	 */
483 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
484 		struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
485 		struct hubp *hubp = pipe->plane_res.hubp;
486 
487 		if (hubp && hubp->funcs->hubp_update_force_pstate_disallow)
488 			hubp->funcs->hubp_update_force_pstate_disallow(hubp, false);
489 		if (pipe->stream && pipe->stream->mall_stream_config.type == SUBVP_MAIN)
490 			num_subvp++;
491 	}
492 
493 	if (num_subvp == 0)
494 		return;
495 
496 	/* Loop through each pipe -- for each subvp main pipe force p-state allow equal to false.
497 	 */
498 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
499 		struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
500 
501 		if (pipe->stream && pipe->plane_state && pipe->stream->mall_stream_config.type == SUBVP_MAIN) {
502 			struct hubp *hubp = pipe->plane_res.hubp;
503 
504 			if (hubp && hubp->funcs->hubp_update_force_pstate_disallow)
505 				hubp->funcs->hubp_update_force_pstate_disallow(hubp, true);
506 		}
507 	}
508 }
509 
510 /* Update MALL_SEL register based on if pipe / plane
511  * is a phantom pipe, main pipe, and if using MALL
512  * for SS.
513  */
514 void dcn32_update_mall_sel(struct dc *dc, struct dc_state *context)
515 {
516 	int i;
517 	unsigned int num_ways = dcn32_calculate_cab_allocation(dc, context);
518 
519 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
520 		struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
521 		struct hubp *hubp = pipe->plane_res.hubp;
522 
523 		if (pipe->stream && pipe->plane_state && hubp && hubp->funcs->hubp_update_mall_sel) {
524 			if (pipe->stream->mall_stream_config.type == SUBVP_PHANTOM) {
525 					hubp->funcs->hubp_update_mall_sel(hubp, 1);
526 			} else {
527 				hubp->funcs->hubp_update_mall_sel(hubp,
528 					num_ways <= dc->caps.cache_num_ways &&
529 					pipe->stream->link->psr_settings.psr_version == DC_PSR_VERSION_UNSUPPORTED ? 2 : 0);
530 			}
531 		}
532 	}
533 }
534 
535 /* Program the sub-viewport pipe configuration after the main / phantom pipes
536  * have been programmed in hardware.
537  * 1. Update force P-State for all the main pipes (disallow P-state)
538  * 2. Update MALL_SEL register
539  * 3. Program FORCE_ONE_ROW_FOR_FRAME for main subvp pipes
540  */
541 void dcn32_program_mall_pipe_config(struct dc *dc, struct dc_state *context)
542 {
543 	int i;
544 	struct dce_hwseq *hws = dc->hwseq;
545 	// Update force P-state for each pipe accordingly
546 	if (hws && hws->funcs.subvp_update_force_pstate)
547 		hws->funcs.subvp_update_force_pstate(dc, context);
548 
549 	// Update MALL_SEL register for each pipe
550 	if (hws && hws->funcs.update_mall_sel)
551 		hws->funcs.update_mall_sel(dc, context);
552 
553 	// Program FORCE_ONE_ROW_FOR_FRAME and CURSOR_REQ_MODE for main subvp pipes
554 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
555 		struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
556 		struct hubp *hubp = pipe->plane_res.hubp;
557 
558 		if (pipe->stream && hubp && hubp->funcs->hubp_prepare_subvp_buffering) {
559 			/* TODO - remove setting CURSOR_REQ_MODE to 0 for legacy cases
560 			 *      - need to investigate single pipe MPO + SubVP case to
561 			 *        see if CURSOR_REQ_MODE will be back to 1 for SubVP
562 			 *        when it should be 0 for MPO
563 			 */
564 			if (pipe->stream->mall_stream_config.type == SUBVP_MAIN) {
565 				hubp->funcs->hubp_prepare_subvp_buffering(hubp, true);
566 			}
567 		}
568 	}
569 }
570 
571 void dcn32_init_hw(struct dc *dc)
572 {
573 	struct abm **abms = dc->res_pool->multiple_abms;
574 	struct dce_hwseq *hws = dc->hwseq;
575 	struct dc_bios *dcb = dc->ctx->dc_bios;
576 	struct resource_pool *res_pool = dc->res_pool;
577 	int i;
578 	int edp_num;
579 	uint32_t backlight = MAX_BACKLIGHT_LEVEL;
580 
581 	dc->debug.disable_idle_power_optimizations = true;
582 	if (dc->clk_mgr && dc->clk_mgr->funcs->init_clocks)
583 		dc->clk_mgr->funcs->init_clocks(dc->clk_mgr);
584 
585 	// Initialize the dccg
586 	if (res_pool->dccg->funcs->dccg_init)
587 		res_pool->dccg->funcs->dccg_init(res_pool->dccg);
588 
589 	if (!dcb->funcs->is_accelerated_mode(dcb)) {
590 		hws->funcs.bios_golden_init(dc);
591 		hws->funcs.disable_vga(dc->hwseq);
592 	}
593 
594 	// Set default OPTC memory power states
595 	if (dc->debug.enable_mem_low_power.bits.optc) {
596 		// Shutdown when unassigned and light sleep in VBLANK
597 		REG_SET_2(ODM_MEM_PWR_CTRL3, 0, ODM_MEM_UNASSIGNED_PWR_MODE, 3, ODM_MEM_VBLANK_PWR_MODE, 1);
598 	}
599 
600 	if (dc->debug.enable_mem_low_power.bits.vga) {
601 		// Power down VGA memory
602 		REG_UPDATE(MMHUBBUB_MEM_PWR_CNTL, VGA_MEM_PWR_FORCE, 1);
603 	}
604 
605 	if (dc->ctx->dc_bios->fw_info_valid) {
606 		res_pool->ref_clocks.xtalin_clock_inKhz =
607 				dc->ctx->dc_bios->fw_info.pll_info.crystal_frequency;
608 
609 		if (res_pool->dccg && res_pool->hubbub) {
610 			(res_pool->dccg->funcs->get_dccg_ref_freq)(res_pool->dccg,
611 					dc->ctx->dc_bios->fw_info.pll_info.crystal_frequency,
612 					&res_pool->ref_clocks.dccg_ref_clock_inKhz);
613 
614 			(res_pool->hubbub->funcs->get_dchub_ref_freq)(res_pool->hubbub,
615 					res_pool->ref_clocks.dccg_ref_clock_inKhz,
616 					&res_pool->ref_clocks.dchub_ref_clock_inKhz);
617 		} else {
618 			// Not all ASICs have DCCG sw component
619 			res_pool->ref_clocks.dccg_ref_clock_inKhz =
620 					res_pool->ref_clocks.xtalin_clock_inKhz;
621 			res_pool->ref_clocks.dchub_ref_clock_inKhz =
622 					res_pool->ref_clocks.xtalin_clock_inKhz;
623 		}
624 	} else
625 		ASSERT_CRITICAL(false);
626 
627 	for (i = 0; i < dc->link_count; i++) {
628 		/* Power up AND update implementation according to the
629 		 * required signal (which may be different from the
630 		 * default signal on connector).
631 		 */
632 		struct dc_link *link = dc->links[i];
633 
634 		link->link_enc->funcs->hw_init(link->link_enc);
635 
636 		/* Check for enabled DIG to identify enabled display */
637 		if (link->link_enc->funcs->is_dig_enabled &&
638 			link->link_enc->funcs->is_dig_enabled(link->link_enc)) {
639 			link->link_status.link_active = true;
640 			if (link->link_enc->funcs->fec_is_active &&
641 					link->link_enc->funcs->fec_is_active(link->link_enc))
642 				link->fec_state = dc_link_fec_enabled;
643 		}
644 	}
645 
646 	/* Power gate DSCs */
647 	for (i = 0; i < res_pool->res_cap->num_dsc; i++)
648 		if (hws->funcs.dsc_pg_control != NULL)
649 			hws->funcs.dsc_pg_control(hws, res_pool->dscs[i]->inst, false);
650 
651 	/* we want to turn off all dp displays before doing detection */
652 	dc_link_blank_all_dp_displays(dc);
653 
654 	/* If taking control over from VBIOS, we may want to optimize our first
655 	 * mode set, so we need to skip powering down pipes until we know which
656 	 * pipes we want to use.
657 	 * Otherwise, if taking control is not possible, we need to power
658 	 * everything down.
659 	 */
660 	if (dcb->funcs->is_accelerated_mode(dcb) || !dc->config.seamless_boot_edp_requested) {
661 		hws->funcs.init_pipes(dc, dc->current_state);
662 		if (dc->res_pool->hubbub->funcs->allow_self_refresh_control)
663 			dc->res_pool->hubbub->funcs->allow_self_refresh_control(dc->res_pool->hubbub,
664 					!dc->res_pool->hubbub->ctx->dc->debug.disable_stutter);
665 	}
666 
667 	/* In headless boot cases, DIG may be turned
668 	 * on which causes HW/SW discrepancies.
669 	 * To avoid this, power down hardware on boot
670 	 * if DIG is turned on and seamless boot not enabled
671 	 */
672 	if (!dc->config.seamless_boot_edp_requested) {
673 		struct dc_link *edp_links[MAX_NUM_EDP];
674 		struct dc_link *edp_link;
675 
676 		get_edp_links(dc, edp_links, &edp_num);
677 		if (edp_num) {
678 			for (i = 0; i < edp_num; i++) {
679 				edp_link = edp_links[i];
680 				if (edp_link->link_enc->funcs->is_dig_enabled &&
681 						edp_link->link_enc->funcs->is_dig_enabled(edp_link->link_enc) &&
682 						dc->hwss.edp_backlight_control &&
683 						dc->hwss.power_down &&
684 						dc->hwss.edp_power_control) {
685 					dc->hwss.edp_backlight_control(edp_link, false);
686 					dc->hwss.power_down(dc);
687 					dc->hwss.edp_power_control(edp_link, false);
688 				}
689 			}
690 		} else {
691 			for (i = 0; i < dc->link_count; i++) {
692 				struct dc_link *link = dc->links[i];
693 
694 				if (link->link_enc->funcs->is_dig_enabled &&
695 						link->link_enc->funcs->is_dig_enabled(link->link_enc) &&
696 						dc->hwss.power_down) {
697 					dc->hwss.power_down(dc);
698 					break;
699 				}
700 
701 			}
702 		}
703 	}
704 
705 	for (i = 0; i < res_pool->audio_count; i++) {
706 		struct audio *audio = res_pool->audios[i];
707 
708 		audio->funcs->hw_init(audio);
709 	}
710 
711 	for (i = 0; i < dc->link_count; i++) {
712 		struct dc_link *link = dc->links[i];
713 
714 		if (link->panel_cntl)
715 			backlight = link->panel_cntl->funcs->hw_init(link->panel_cntl);
716 	}
717 
718 	for (i = 0; i < dc->res_pool->pipe_count; i++) {
719 		if (abms[i] != NULL && abms[i]->funcs != NULL)
720 			abms[i]->funcs->abm_init(abms[i], backlight);
721 	}
722 
723 	/* power AFMT HDMI memory TODO: may move to dis/en output save power*/
724 	REG_WRITE(DIO_MEM_PWR_CTRL, 0);
725 
726 	if (!dc->debug.disable_clock_gate) {
727 		/* enable all DCN clock gating */
728 		REG_WRITE(DCCG_GATE_DISABLE_CNTL, 0);
729 
730 		REG_WRITE(DCCG_GATE_DISABLE_CNTL2, 0);
731 
732 		REG_UPDATE(DCFCLK_CNTL, DCFCLK_GATE_DIS, 0);
733 	}
734 	if (hws->funcs.enable_power_gating_plane)
735 		hws->funcs.enable_power_gating_plane(dc->hwseq, true);
736 
737 	if (!dcb->funcs->is_accelerated_mode(dcb) && dc->res_pool->hubbub->funcs->init_watermarks)
738 		dc->res_pool->hubbub->funcs->init_watermarks(dc->res_pool->hubbub);
739 
740 	if (dc->clk_mgr->funcs->notify_wm_ranges)
741 		dc->clk_mgr->funcs->notify_wm_ranges(dc->clk_mgr);
742 
743 	if (dc->clk_mgr->funcs->set_hard_max_memclk)
744 		dc->clk_mgr->funcs->set_hard_max_memclk(dc->clk_mgr);
745 
746 	if (dc->res_pool->hubbub->funcs->force_pstate_change_control)
747 		dc->res_pool->hubbub->funcs->force_pstate_change_control(
748 				dc->res_pool->hubbub, false, false);
749 
750 	if (dc->res_pool->hubbub->funcs->init_crb)
751 		dc->res_pool->hubbub->funcs->init_crb(dc->res_pool->hubbub);
752 
753 	// Get DMCUB capabilities
754     if (dc->ctx->dmub_srv) {
755 	dc_dmub_srv_query_caps_cmd(dc->ctx->dmub_srv->dmub);
756 	dc->caps.dmub_caps.psr = dc->ctx->dmub_srv->dmub->feature_caps.psr;
757     }
758 }
759 
760 static int calc_mpc_flow_ctrl_cnt(const struct dc_stream_state *stream,
761 		int opp_cnt)
762 {
763 	bool hblank_halved = optc2_is_two_pixels_per_containter(&stream->timing);
764 	int flow_ctrl_cnt;
765 
766 	if (opp_cnt >= 2)
767 		hblank_halved = true;
768 
769 	flow_ctrl_cnt = stream->timing.h_total - stream->timing.h_addressable -
770 			stream->timing.h_border_left -
771 			stream->timing.h_border_right;
772 
773 	if (hblank_halved)
774 		flow_ctrl_cnt /= 2;
775 
776 	/* ODM combine 4:1 case */
777 	if (opp_cnt == 4)
778 		flow_ctrl_cnt /= 2;
779 
780 	return flow_ctrl_cnt;
781 }
782 
783 static void update_dsc_on_stream(struct pipe_ctx *pipe_ctx, bool enable)
784 {
785 	struct display_stream_compressor *dsc = pipe_ctx->stream_res.dsc;
786 	struct dc_stream_state *stream = pipe_ctx->stream;
787 	struct pipe_ctx *odm_pipe;
788 	int opp_cnt = 1;
789 
790 	ASSERT(dsc);
791 	for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe)
792 		opp_cnt++;
793 
794 	if (enable) {
795 		struct dsc_config dsc_cfg;
796 		struct dsc_optc_config dsc_optc_cfg;
797 		enum optc_dsc_mode optc_dsc_mode;
798 
799 		/* Enable DSC hw block */
800 		dsc_cfg.pic_width = (stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right) / opp_cnt;
801 		dsc_cfg.pic_height = stream->timing.v_addressable + stream->timing.v_border_top + stream->timing.v_border_bottom;
802 		dsc_cfg.pixel_encoding = stream->timing.pixel_encoding;
803 		dsc_cfg.color_depth = stream->timing.display_color_depth;
804 		dsc_cfg.is_odm = pipe_ctx->next_odm_pipe ? true : false;
805 		dsc_cfg.dc_dsc_cfg = stream->timing.dsc_cfg;
806 		ASSERT(dsc_cfg.dc_dsc_cfg.num_slices_h % opp_cnt == 0);
807 		dsc_cfg.dc_dsc_cfg.num_slices_h /= opp_cnt;
808 
809 		dsc->funcs->dsc_set_config(dsc, &dsc_cfg, &dsc_optc_cfg);
810 		dsc->funcs->dsc_enable(dsc, pipe_ctx->stream_res.opp->inst);
811 		for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
812 			struct display_stream_compressor *odm_dsc = odm_pipe->stream_res.dsc;
813 
814 			ASSERT(odm_dsc);
815 			odm_dsc->funcs->dsc_set_config(odm_dsc, &dsc_cfg, &dsc_optc_cfg);
816 			odm_dsc->funcs->dsc_enable(odm_dsc, odm_pipe->stream_res.opp->inst);
817 		}
818 		dsc_cfg.dc_dsc_cfg.num_slices_h *= opp_cnt;
819 		dsc_cfg.pic_width *= opp_cnt;
820 
821 		optc_dsc_mode = dsc_optc_cfg.is_pixel_format_444 ? OPTC_DSC_ENABLED_444 : OPTC_DSC_ENABLED_NATIVE_SUBSAMPLED;
822 
823 		/* Enable DSC in OPTC */
824 		DC_LOG_DSC("Setting optc DSC config for tg instance %d:", pipe_ctx->stream_res.tg->inst);
825 		pipe_ctx->stream_res.tg->funcs->set_dsc_config(pipe_ctx->stream_res.tg,
826 							optc_dsc_mode,
827 							dsc_optc_cfg.bytes_per_pixel,
828 							dsc_optc_cfg.slice_width);
829 	} else {
830 		/* disable DSC in OPTC */
831 		pipe_ctx->stream_res.tg->funcs->set_dsc_config(
832 				pipe_ctx->stream_res.tg,
833 				OPTC_DSC_DISABLED, 0, 0);
834 
835 		/* disable DSC block */
836 		dsc->funcs->dsc_disable(pipe_ctx->stream_res.dsc);
837 		for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
838 			ASSERT(odm_pipe->stream_res.dsc);
839 			odm_pipe->stream_res.dsc->funcs->dsc_disable(odm_pipe->stream_res.dsc);
840 		}
841 	}
842 }
843 
844 /*
845 * Given any pipe_ctx, return the total ODM combine factor, and optionally return
846 * the OPPids which are used
847 * */
848 static unsigned int get_odm_config(struct pipe_ctx *pipe_ctx, unsigned int *opp_instances)
849 {
850 	unsigned int opp_count = 1;
851 	struct pipe_ctx *odm_pipe;
852 
853 	/* First get to the top pipe */
854 	for (odm_pipe = pipe_ctx; odm_pipe->prev_odm_pipe; odm_pipe = odm_pipe->prev_odm_pipe)
855 		;
856 
857 	/* First pipe is always used */
858 	if (opp_instances)
859 		opp_instances[0] = odm_pipe->stream_res.opp->inst;
860 
861 	/* Find and count odm pipes, if any */
862 	for (odm_pipe = odm_pipe->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
863 		if (opp_instances)
864 			opp_instances[opp_count] = odm_pipe->stream_res.opp->inst;
865 		opp_count++;
866 	}
867 
868 	return opp_count;
869 }
870 
871 void dcn32_update_odm(struct dc *dc, struct dc_state *context, struct pipe_ctx *pipe_ctx)
872 {
873 	struct pipe_ctx *odm_pipe;
874 	int opp_cnt = 0;
875 	int opp_inst[MAX_PIPES] = {0};
876 	bool rate_control_2x_pclk = (pipe_ctx->stream->timing.flags.INTERLACE || optc2_is_two_pixels_per_containter(&pipe_ctx->stream->timing));
877 	struct mpc_dwb_flow_control flow_control;
878 	struct mpc *mpc = dc->res_pool->mpc;
879 	int i;
880 
881 	opp_cnt = get_odm_config(pipe_ctx, opp_inst);
882 
883 	if (opp_cnt > 1)
884 		pipe_ctx->stream_res.tg->funcs->set_odm_combine(
885 				pipe_ctx->stream_res.tg,
886 				opp_inst, opp_cnt,
887 				&pipe_ctx->stream->timing);
888 	else
889 		pipe_ctx->stream_res.tg->funcs->set_odm_bypass(
890 				pipe_ctx->stream_res.tg, &pipe_ctx->stream->timing);
891 
892 	rate_control_2x_pclk = rate_control_2x_pclk || opp_cnt > 1;
893 	flow_control.flow_ctrl_mode = 0;
894 	flow_control.flow_ctrl_cnt0 = 0x80;
895 	flow_control.flow_ctrl_cnt1 = calc_mpc_flow_ctrl_cnt(pipe_ctx->stream, opp_cnt);
896 	if (mpc->funcs->set_out_rate_control) {
897 		for (i = 0; i < opp_cnt; ++i) {
898 			mpc->funcs->set_out_rate_control(
899 					mpc, opp_inst[i],
900 					true,
901 					rate_control_2x_pclk,
902 					&flow_control);
903 		}
904 	}
905 
906 	for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
907 		odm_pipe->stream_res.opp->funcs->opp_pipe_clock_control(
908 				odm_pipe->stream_res.opp,
909 				true);
910 	}
911 
912 	// Don't program pixel clock after link is already enabled
913 /*	if (false == pipe_ctx->clock_source->funcs->program_pix_clk(
914 			pipe_ctx->clock_source,
915 			&pipe_ctx->stream_res.pix_clk_params,
916 			&pipe_ctx->pll_settings)) {
917 		BREAK_TO_DEBUGGER();
918 	}*/
919 
920 	if (pipe_ctx->stream_res.dsc)
921 		update_dsc_on_stream(pipe_ctx, pipe_ctx->stream->timing.flags.DSC);
922 }
923 
924 unsigned int dcn32_calculate_dccg_k1_k2_values(struct pipe_ctx *pipe_ctx, unsigned int *k1_div, unsigned int *k2_div)
925 {
926 	struct dc_stream_state *stream = pipe_ctx->stream;
927 	unsigned int odm_combine_factor = 0;
928 
929 	odm_combine_factor = get_odm_config(pipe_ctx, NULL);
930 
931 	if (is_dp_128b_132b_signal(pipe_ctx)) {
932 		*k2_div = PIXEL_RATE_DIV_BY_1;
933 	} else if (dc_is_hdmi_tmds_signal(pipe_ctx->stream->signal) || dc_is_dvi_signal(pipe_ctx->stream->signal)) {
934 		*k1_div = PIXEL_RATE_DIV_BY_1;
935 		if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420)
936 			*k2_div = PIXEL_RATE_DIV_BY_2;
937 		else
938 			*k2_div = PIXEL_RATE_DIV_BY_4;
939 	} else if (dc_is_dp_signal(pipe_ctx->stream->signal)) {
940 		if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420) {
941 			*k1_div = PIXEL_RATE_DIV_BY_1;
942 			*k2_div = PIXEL_RATE_DIV_BY_2;
943 		} else if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR422) {
944 			*k1_div = PIXEL_RATE_DIV_BY_2;
945 			*k2_div = PIXEL_RATE_DIV_BY_2;
946 		} else {
947 			if (odm_combine_factor == 1)
948 				*k2_div = PIXEL_RATE_DIV_BY_4;
949 			else if (odm_combine_factor == 2)
950 				*k2_div = PIXEL_RATE_DIV_BY_2;
951 		}
952 	}
953 
954 	if ((*k1_div == PIXEL_RATE_DIV_NA) && (*k2_div == PIXEL_RATE_DIV_NA))
955 		ASSERT(false);
956 
957 	return odm_combine_factor;
958 }
959