2 * Copyright 2016 Advanced Micro Devices, Inc.
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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.
27 #include "dm_services.h"
28 #include "dm_helpers.h"
29 #include "core_types.h"
32 #include "dce/dce_hwseq.h"
33 #include "dcn30/dcn30_cm_common.h"
34 #include "reg_helper.h"
38 #include "timing_generator.h"
43 #include "dc_dmub_srv.h"
44 #include "link_hwss.h"
45 #include "dpcd_defs.h"
46 #include "dcn32_hwseq.h"
49 #include "dcn20/dcn20_optc.h"
50 #include "dmub_subvp_state.h"
51 #include "dce/dmub_hw_lock_mgr.h"
52 #include "dc_link_dp.h"
53 #include "dmub/inc/dmub_subvp_state.h"
55 #define DC_LOGGER_INIT(logger)
66 #define FN(reg_name, field_name) \
67 hws->shifts->field_name, hws->masks->field_name
69 void dcn32_dsc_pg_control(
70 struct dce_hwseq *hws,
71 unsigned int dsc_inst,
74 uint32_t power_gate = power_on ? 0 : 1;
75 uint32_t pwr_status = power_on ? 0 : 2;
76 uint32_t org_ip_request_cntl = 0;
78 if (hws->ctx->dc->debug.disable_dsc_power_gate)
81 REG_GET(DC_IP_REQUEST_CNTL, IP_REQUEST_EN, &org_ip_request_cntl);
82 if (org_ip_request_cntl == 0)
83 REG_SET(DC_IP_REQUEST_CNTL, 0, IP_REQUEST_EN, 1);
87 REG_UPDATE(DOMAIN16_PG_CONFIG,
88 DOMAIN_POWER_GATE, power_gate);
90 REG_WAIT(DOMAIN16_PG_STATUS,
91 DOMAIN_PGFSM_PWR_STATUS, pwr_status,
95 REG_UPDATE(DOMAIN17_PG_CONFIG,
96 DOMAIN_POWER_GATE, power_gate);
98 REG_WAIT(DOMAIN17_PG_STATUS,
99 DOMAIN_PGFSM_PWR_STATUS, pwr_status,
103 REG_UPDATE(DOMAIN18_PG_CONFIG,
104 DOMAIN_POWER_GATE, power_gate);
106 REG_WAIT(DOMAIN18_PG_STATUS,
107 DOMAIN_PGFSM_PWR_STATUS, pwr_status,
111 REG_UPDATE(DOMAIN19_PG_CONFIG,
112 DOMAIN_POWER_GATE, power_gate);
114 REG_WAIT(DOMAIN19_PG_STATUS,
115 DOMAIN_PGFSM_PWR_STATUS, pwr_status,
123 if (org_ip_request_cntl == 0)
124 REG_SET(DC_IP_REQUEST_CNTL, 0, IP_REQUEST_EN, 0);
128 void dcn32_enable_power_gating_plane(
129 struct dce_hwseq *hws,
132 bool force_on = true; /* disable power gating */
138 REG_UPDATE(DOMAIN0_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
139 REG_UPDATE(DOMAIN1_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
140 REG_UPDATE(DOMAIN2_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
141 REG_UPDATE(DOMAIN3_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
144 REG_UPDATE(DOMAIN16_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
145 REG_UPDATE(DOMAIN17_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
146 REG_UPDATE(DOMAIN18_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
147 REG_UPDATE(DOMAIN19_PG_CONFIG, DOMAIN_POWER_FORCEON, force_on);
150 void dcn32_hubp_pg_control(struct dce_hwseq *hws, unsigned int hubp_inst, bool power_on)
152 uint32_t power_gate = power_on ? 0 : 1;
153 uint32_t pwr_status = power_on ? 0 : 2;
155 if (hws->ctx->dc->debug.disable_hubp_power_gate)
158 if (REG(DOMAIN0_PG_CONFIG) == 0)
163 REG_SET(DOMAIN0_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
164 REG_WAIT(DOMAIN0_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
167 REG_SET(DOMAIN1_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
168 REG_WAIT(DOMAIN1_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
171 REG_SET(DOMAIN2_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
172 REG_WAIT(DOMAIN2_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
175 REG_SET(DOMAIN3_PG_CONFIG, 0, DOMAIN_POWER_GATE, power_gate);
176 REG_WAIT(DOMAIN3_PG_STATUS, DOMAIN_PGFSM_PWR_STATUS, pwr_status, 1, 1000);
184 static bool dcn32_check_no_memory_request_for_cab(struct dc *dc)
188 /* First, check no-memory-request case */
189 for (i = 0; i < dc->current_state->stream_count; i++) {
190 if (dc->current_state->stream_status[i].plane_count)
191 /* Fail eligibility on a visible stream */
195 if (i == dc->current_state->stream_count)
201 /* This function takes in the start address and surface size to be cached in CAB
202 * and calculates the total number of cache lines required to store the surface.
203 * The number of cache lines used for each surface is calculated independently of
204 * one another. For example, if there is a primary surface(1), meta surface(2), and
205 * cursor(3), this function should be called 3 times to calculate the number of cache
206 * lines used for each of those surfaces.
208 static uint32_t dcn32_cache_lines_for_surface(struct dc *dc, uint32_t surface_size, uint64_t start_address)
210 uint32_t lines_used = 1;
211 uint32_t num_cached_bytes = 0;
212 uint32_t remaining_size = 0;
213 uint32_t cache_line_size = dc->caps.cache_line_size;
214 uint32_t remainder = 0;
216 /* 1. Calculate surface size minus the number of bytes stored
217 * in the first cache line (all bytes in first cache line might
218 * not be fully used).
220 div_u64_rem(start_address, cache_line_size, &remainder);
221 num_cached_bytes = cache_line_size - remainder;
222 remaining_size = surface_size - num_cached_bytes;
224 /* 2. Calculate number of cache lines that will be fully used with
225 * the remaining number of bytes to be stored.
227 lines_used += (remaining_size / cache_line_size);
229 /* 3. Check if we need an extra line due to the remaining size not being
230 * a multiple of CACHE_LINE_SIZE.
232 if (remaining_size % cache_line_size > 0)
238 /* This function loops through every surface that needs to be cached in CAB for SS,
239 * and calculates the total number of ways required to store all surfaces (primary,
242 static uint32_t dcn32_calculate_cab_allocation(struct dc *dc, struct dc_state *ctx)
245 struct dc_stream_state *stream = NULL;
246 struct dc_plane_state *plane = NULL;
247 uint32_t surface_size = 0;
248 uint32_t cursor_size = 0;
249 uint32_t cache_lines_used = 0;
250 uint32_t total_lines = 0;
251 uint32_t lines_per_way = 0;
252 uint32_t num_ways = 0;
254 for (i = 0; i < ctx->stream_count; i++) {
255 stream = ctx->streams[i];
257 // Don't include PSR surface in the total surface size for CAB allocation
258 if (stream->link->psr_settings.psr_version != DC_PSR_VERSION_UNSUPPORTED)
261 if (ctx->stream_status[i].plane_count == 0)
264 // For each stream, loop through each plane to calculate the number of cache
265 // lines required to store the surface in CAB
266 for (j = 0; j < ctx->stream_status[i].plane_count; j++) {
267 plane = ctx->stream_status[i].plane_states[j];
269 // Calculate total surface size
270 surface_size = plane->plane_size.surface_pitch *
271 plane->plane_size.surface_size.height *
272 (plane->format >= SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616 ? 8 : 4);
274 // Convert surface size + starting address to number of cache lines required
275 // (alignment accounted for)
276 cache_lines_used += dcn32_cache_lines_for_surface(dc, surface_size,
277 plane->address.grph.addr.quad_part);
279 if (plane->address.grph.meta_addr.quad_part) {
281 cache_lines_used += dcn32_cache_lines_for_surface(dc, surface_size,
282 plane->address.grph.meta_addr.quad_part);
286 // Include cursor size for CAB allocation
287 if (stream->cursor_position.enable && plane->address.grph.cursor_cache_addr.quad_part) {
288 cursor_size = dc->caps.max_cursor_size * dc->caps.max_cursor_size;
289 switch (stream->cursor_attributes.color_format) {
290 case CURSOR_MODE_MONO:
293 case CURSOR_MODE_COLOR_1BIT_AND:
294 case CURSOR_MODE_COLOR_PRE_MULTIPLIED_ALPHA:
295 case CURSOR_MODE_COLOR_UN_PRE_MULTIPLIED_ALPHA:
299 case CURSOR_MODE_COLOR_64BIT_FP_PRE_MULTIPLIED:
300 case CURSOR_MODE_COLOR_64BIT_FP_UN_PRE_MULTIPLIED:
304 cache_lines_used += dcn32_cache_lines_for_surface(dc, surface_size,
305 plane->address.grph.cursor_cache_addr.quad_part);
309 // Convert number of cache lines required to number of ways
310 total_lines = dc->caps.max_cab_allocation_bytes / dc->caps.cache_line_size;
311 lines_per_way = total_lines / dc->caps.cache_num_ways;
312 num_ways = cache_lines_used / lines_per_way;
314 if (cache_lines_used % lines_per_way > 0)
320 bool dcn32_apply_idle_power_optimizations(struct dc *dc, bool enable)
322 union dmub_rb_cmd cmd;
325 if (!dc->ctx->dmub_srv)
329 if (dc->current_state) {
331 /* 1. Check no memory request case for CAB.
332 * If no memory request case, send CAB_ACTION NO_DF_REQ DMUB message
334 if (dcn32_check_no_memory_request_for_cab(dc)) {
335 /* Enable no-memory-requests case */
336 memset(&cmd, 0, sizeof(cmd));
337 cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
338 cmd.cab.header.sub_type = DMUB_CMD__CAB_NO_DCN_REQ;
339 cmd.cab.header.payload_bytes = sizeof(cmd.cab) - sizeof(cmd.cab.header);
341 dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
342 dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
347 /* 2. Check if all surfaces can fit in CAB.
348 * If surfaces can fit into CAB, send CAB_ACTION_ALLOW DMUB message
349 * and configure HUBP's to fetch from MALL
351 ways = dcn32_calculate_cab_allocation(dc, dc->current_state);
352 if (ways <= dc->caps.cache_num_ways) {
353 memset(&cmd, 0, sizeof(cmd));
354 cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
355 cmd.cab.header.sub_type = DMUB_CMD__CAB_DCN_SS_FIT_IN_CAB;
356 cmd.cab.header.payload_bytes = sizeof(cmd.cab) - sizeof(cmd.cab.header);
357 cmd.cab.cab_alloc_ways = ways;
359 dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
360 dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
370 memset(&cmd, 0, sizeof(cmd));
371 cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
372 cmd.cab.header.sub_type = DMUB_CMD__CAB_NO_IDLE_OPTIMIZATION;
373 cmd.cab.header.payload_bytes =
374 sizeof(cmd.cab) - sizeof(cmd.cab.header);
376 dc_dmub_srv_cmd_queue(dc->ctx->dmub_srv, &cmd);
377 dc_dmub_srv_cmd_execute(dc->ctx->dmub_srv);
378 dc_dmub_srv_wait_idle(dc->ctx->dmub_srv);
383 /* Send DMCUB message with SubVP pipe info
384 * - For each pipe in context, populate payload with required SubVP information
385 * if the pipe is using SubVP for MCLK switch
386 * - This function must be called while the DMUB HW lock is acquired by driver
388 void dcn32_commit_subvp_config(struct dc *dc, struct dc_state *context)
392 bool enable_subvp = false;
394 if (!dc->ctx || !dc->ctx->dmub_srv)
397 for (i = 0; i < dc->res_pool->pipe_count; i++) {
398 struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
400 if (pipe_ctx->stream && pipe_ctx->stream->mall_stream_config.paired_stream &&
401 pipe_ctx->stream->mall_stream_config.type == SUBVP_MAIN) {
402 // There is at least 1 SubVP pipe, so enable SubVP
407 dc_dmub_setup_subvp_dmub_command(dc, context, enable_subvp);
411 /* Sub-Viewport DMUB lock needs to be acquired by driver whenever SubVP is active and:
412 * 1. Any full update for any SubVP main pipe
413 * 2. Any immediate flip for any SubVP pipe
414 * 3. Any flip for DRR pipe
415 * 4. If SubVP was previously in use (i.e. in old context)
417 void dcn32_subvp_pipe_control_lock(struct dc *dc,
418 struct dc_state *context,
420 bool should_lock_all_pipes,
421 struct pipe_ctx *top_pipe_to_program,
425 bool subvp_immediate_flip = false;
426 bool subvp_in_use = false;
427 struct pipe_ctx *pipe;
429 for (i = 0; i < dc->res_pool->pipe_count; i++) {
430 pipe = &context->res_ctx.pipe_ctx[i];
432 if (pipe->stream && pipe->plane_state && pipe->stream->mall_stream_config.type == SUBVP_MAIN) {
438 if (top_pipe_to_program && top_pipe_to_program->stream && top_pipe_to_program->plane_state) {
439 if (top_pipe_to_program->stream->mall_stream_config.type == SUBVP_MAIN &&
440 top_pipe_to_program->plane_state->flip_immediate)
441 subvp_immediate_flip = true;
444 // Don't need to lock for DRR VSYNC flips -- FW will wait for DRR pending update cleared.
445 if ((subvp_in_use && (should_lock_all_pipes || subvp_immediate_flip)) || (!subvp_in_use && subvp_prev_use)) {
446 union dmub_inbox0_cmd_lock_hw hw_lock_cmd = { 0 };
449 for (i = 0; i < dc->res_pool->pipe_count; i++) {
450 pipe = &context->res_ctx.pipe_ctx[i];
451 if (pipe->stream && pipe->plane_state && pipe->stream->mall_stream_config.type == SUBVP_MAIN &&
452 should_lock_all_pipes)
453 pipe->stream_res.tg->funcs->wait_for_state(pipe->stream_res.tg, CRTC_STATE_VBLANK);
457 hw_lock_cmd.bits.command_code = DMUB_INBOX0_CMD__HW_LOCK;
458 hw_lock_cmd.bits.hw_lock_client = HW_LOCK_CLIENT_DRIVER;
459 hw_lock_cmd.bits.lock = lock;
460 hw_lock_cmd.bits.should_release = !lock;
461 dmub_hw_lock_mgr_inbox0_cmd(dc->ctx->dmub_srv, hw_lock_cmd);
466 static bool dcn32_set_mpc_shaper_3dlut(
467 struct pipe_ctx *pipe_ctx, const struct dc_stream_state *stream)
469 struct dpp *dpp_base = pipe_ctx->plane_res.dpp;
470 int mpcc_id = pipe_ctx->plane_res.hubp->inst;
471 struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc;
474 const struct pwl_params *shaper_lut = NULL;
475 //get the shaper lut params
476 if (stream->func_shaper) {
477 if (stream->func_shaper->type == TF_TYPE_HWPWL)
478 shaper_lut = &stream->func_shaper->pwl;
479 else if (stream->func_shaper->type == TF_TYPE_DISTRIBUTED_POINTS) {
480 cm_helper_translate_curve_to_hw_format(
482 &dpp_base->shaper_params, true);
483 shaper_lut = &dpp_base->shaper_params;
487 if (stream->lut3d_func &&
488 stream->lut3d_func->state.bits.initialized == 1) {
490 result = mpc->funcs->program_3dlut(mpc,
491 &stream->lut3d_func->lut_3d,
494 result = mpc->funcs->program_shaper(mpc,
502 bool dcn32_set_mcm_luts(
503 struct pipe_ctx *pipe_ctx, const struct dc_plane_state *plane_state)
505 struct dpp *dpp_base = pipe_ctx->plane_res.dpp;
506 int mpcc_id = pipe_ctx->plane_res.hubp->inst;
507 struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc;
509 struct pwl_params *lut_params = NULL;
512 if (plane_state->blend_tf) {
513 if (plane_state->blend_tf->type == TF_TYPE_HWPWL)
514 lut_params = &plane_state->blend_tf->pwl;
515 else if (plane_state->blend_tf->type == TF_TYPE_DISTRIBUTED_POINTS) {
516 cm_helper_translate_curve_to_hw_format(
517 plane_state->blend_tf,
518 &dpp_base->regamma_params, false);
519 lut_params = &dpp_base->regamma_params;
522 result = mpc->funcs->program_1dlut(mpc, lut_params, mpcc_id);
525 if (plane_state->in_shaper_func) {
526 if (plane_state->in_shaper_func->type == TF_TYPE_HWPWL)
527 lut_params = &plane_state->in_shaper_func->pwl;
528 else if (plane_state->in_shaper_func->type == TF_TYPE_DISTRIBUTED_POINTS) {
529 // TODO: dpp_base replace
531 cm_helper_translate_curve_to_hw_format(
532 plane_state->in_shaper_func,
533 &dpp_base->shaper_params, true);
534 lut_params = &dpp_base->shaper_params;
538 result = mpc->funcs->program_shaper(mpc, lut_params, mpcc_id);
541 if (plane_state->lut3d_func && plane_state->lut3d_func->state.bits.initialized == 1)
542 result = mpc->funcs->program_3dlut(mpc, &plane_state->lut3d_func->lut_3d, mpcc_id);
544 result = mpc->funcs->program_3dlut(mpc, NULL, mpcc_id);
549 bool dcn32_set_input_transfer_func(struct dc *dc,
550 struct pipe_ctx *pipe_ctx,
551 const struct dc_plane_state *plane_state)
553 struct dce_hwseq *hws = dc->hwseq;
554 struct mpc *mpc = dc->res_pool->mpc;
555 struct dpp *dpp_base = pipe_ctx->plane_res.dpp;
557 enum dc_transfer_func_predefined tf;
559 struct pwl_params *params = NULL;
561 if (mpc == NULL || plane_state == NULL)
564 tf = TRANSFER_FUNCTION_UNITY;
566 if (plane_state->in_transfer_func &&
567 plane_state->in_transfer_func->type == TF_TYPE_PREDEFINED)
568 tf = plane_state->in_transfer_func->tf;
570 dpp_base->funcs->dpp_set_pre_degam(dpp_base, tf);
572 if (plane_state->in_transfer_func) {
573 if (plane_state->in_transfer_func->type == TF_TYPE_HWPWL)
574 params = &plane_state->in_transfer_func->pwl;
575 else if (plane_state->in_transfer_func->type == TF_TYPE_DISTRIBUTED_POINTS &&
576 cm3_helper_translate_curve_to_hw_format(plane_state->in_transfer_func,
577 &dpp_base->degamma_params, false))
578 params = &dpp_base->degamma_params;
581 result = dpp_base->funcs->dpp_program_gamcor_lut(dpp_base, params);
584 pipe_ctx->stream_res.opp &&
585 pipe_ctx->stream_res.opp->ctx &&
586 hws->funcs.set_mcm_luts)
587 result = hws->funcs.set_mcm_luts(pipe_ctx, plane_state);
592 bool dcn32_set_output_transfer_func(struct dc *dc,
593 struct pipe_ctx *pipe_ctx,
594 const struct dc_stream_state *stream)
596 int mpcc_id = pipe_ctx->plane_res.hubp->inst;
597 struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc;
598 struct pwl_params *params = NULL;
601 /* program OGAM or 3DLUT only for the top pipe*/
602 if (pipe_ctx->top_pipe == NULL) {
603 /*program shaper and 3dlut in MPC*/
604 ret = dcn32_set_mpc_shaper_3dlut(pipe_ctx, stream);
605 if (ret == false && mpc->funcs->set_output_gamma && stream->out_transfer_func) {
606 if (stream->out_transfer_func->type == TF_TYPE_HWPWL)
607 params = &stream->out_transfer_func->pwl;
608 else if (pipe_ctx->stream->out_transfer_func->type ==
609 TF_TYPE_DISTRIBUTED_POINTS &&
610 cm3_helper_translate_curve_to_hw_format(
611 stream->out_transfer_func,
612 &mpc->blender_params, false))
613 params = &mpc->blender_params;
614 /* there are no ROM LUTs in OUTGAM */
615 if (stream->out_transfer_func->type == TF_TYPE_PREDEFINED)
620 mpc->funcs->set_output_gamma(mpc, mpcc_id, params);
624 /* Program P-State force value according to if pipe is using SubVP or not:
625 * 1. Reset P-State force on all pipes first
626 * 2. For each main pipe, force P-State disallow (P-State allow moderated by DMUB)
628 void dcn32_subvp_update_force_pstate(struct dc *dc, struct dc_state *context)
632 /* Unforce p-state for each pipe
634 for (i = 0; i < dc->res_pool->pipe_count; i++) {
635 struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
636 struct hubp *hubp = pipe->plane_res.hubp;
638 if (hubp && hubp->funcs->hubp_update_force_pstate_disallow)
639 hubp->funcs->hubp_update_force_pstate_disallow(hubp, false);
640 if (pipe->stream && pipe->stream->mall_stream_config.type == SUBVP_MAIN)
647 /* Loop through each pipe -- for each subvp main pipe force p-state allow equal to false.
649 for (i = 0; i < dc->res_pool->pipe_count; i++) {
650 struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
652 // For SubVP + DRR, also force disallow on the DRR pipe
653 // (We will force allow in the DMUB sequence -- some DRR timings by default won't allow P-State so we have
654 // to force once the vblank is stretched).
655 if (pipe->stream && pipe->plane_state && (pipe->stream->mall_stream_config.type == SUBVP_MAIN ||
656 (pipe->stream->mall_stream_config.type == SUBVP_NONE && pipe->stream->ignore_msa_timing_param))) {
657 struct hubp *hubp = pipe->plane_res.hubp;
659 if (hubp && hubp->funcs->hubp_update_force_pstate_disallow)
660 hubp->funcs->hubp_update_force_pstate_disallow(hubp, true);
665 /* Update MALL_SEL register based on if pipe / plane
666 * is a phantom pipe, main pipe, and if using MALL
669 void dcn32_update_mall_sel(struct dc *dc, struct dc_state *context)
672 unsigned int num_ways = dcn32_calculate_cab_allocation(dc, context);
673 bool cache_cursor = false;
675 for (i = 0; i < dc->res_pool->pipe_count; i++) {
676 struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
677 struct hubp *hubp = pipe->plane_res.hubp;
679 if (pipe->stream && pipe->plane_state && hubp && hubp->funcs->hubp_update_mall_sel) {
680 if (hubp->curs_attr.width * hubp->curs_attr.height * 4 > 16384)
683 if (pipe->stream->mall_stream_config.type == SUBVP_PHANTOM) {
684 hubp->funcs->hubp_update_mall_sel(hubp, 1, false);
686 hubp->funcs->hubp_update_mall_sel(hubp,
687 num_ways <= dc->caps.cache_num_ways &&
688 pipe->stream->link->psr_settings.psr_version == DC_PSR_VERSION_UNSUPPORTED ? 2 : 0,
695 /* Program the sub-viewport pipe configuration after the main / phantom pipes
696 * have been programmed in hardware.
697 * 1. Update force P-State for all the main pipes (disallow P-state)
698 * 2. Update MALL_SEL register
699 * 3. Program FORCE_ONE_ROW_FOR_FRAME for main subvp pipes
701 void dcn32_program_mall_pipe_config(struct dc *dc, struct dc_state *context)
704 struct dce_hwseq *hws = dc->hwseq;
706 // Don't force p-state disallow -- can't block dummy p-state
708 // Update MALL_SEL register for each pipe
709 if (hws && hws->funcs.update_mall_sel)
710 hws->funcs.update_mall_sel(dc, context);
712 // Program FORCE_ONE_ROW_FOR_FRAME and CURSOR_REQ_MODE for main subvp pipes
713 for (i = 0; i < dc->res_pool->pipe_count; i++) {
714 struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i];
715 struct hubp *hubp = pipe->plane_res.hubp;
717 if (pipe->stream && hubp && hubp->funcs->hubp_prepare_subvp_buffering) {
718 /* TODO - remove setting CURSOR_REQ_MODE to 0 for legacy cases
719 * - need to investigate single pipe MPO + SubVP case to
720 * see if CURSOR_REQ_MODE will be back to 1 for SubVP
721 * when it should be 0 for MPO
723 if (pipe->stream->mall_stream_config.type == SUBVP_MAIN) {
724 hubp->funcs->hubp_prepare_subvp_buffering(hubp, true);
730 void dcn32_init_hw(struct dc *dc)
732 struct abm **abms = dc->res_pool->multiple_abms;
733 struct dce_hwseq *hws = dc->hwseq;
734 struct dc_bios *dcb = dc->ctx->dc_bios;
735 struct resource_pool *res_pool = dc->res_pool;
738 uint32_t backlight = MAX_BACKLIGHT_LEVEL;
740 if (dc->clk_mgr && dc->clk_mgr->funcs->init_clocks)
741 dc->clk_mgr->funcs->init_clocks(dc->clk_mgr);
743 // Initialize the dccg
744 if (res_pool->dccg->funcs->dccg_init)
745 res_pool->dccg->funcs->dccg_init(res_pool->dccg);
747 if (!dcb->funcs->is_accelerated_mode(dcb)) {
748 hws->funcs.bios_golden_init(dc);
749 hws->funcs.disable_vga(dc->hwseq);
752 // Set default OPTC memory power states
753 if (dc->debug.enable_mem_low_power.bits.optc) {
754 // Shutdown when unassigned and light sleep in VBLANK
755 REG_SET_2(ODM_MEM_PWR_CTRL3, 0, ODM_MEM_UNASSIGNED_PWR_MODE, 3, ODM_MEM_VBLANK_PWR_MODE, 1);
758 if (dc->debug.enable_mem_low_power.bits.vga) {
759 // Power down VGA memory
760 REG_UPDATE(MMHUBBUB_MEM_PWR_CNTL, VGA_MEM_PWR_FORCE, 1);
763 if (dc->ctx->dc_bios->fw_info_valid) {
764 res_pool->ref_clocks.xtalin_clock_inKhz =
765 dc->ctx->dc_bios->fw_info.pll_info.crystal_frequency;
767 if (res_pool->dccg && res_pool->hubbub) {
768 (res_pool->dccg->funcs->get_dccg_ref_freq)(res_pool->dccg,
769 dc->ctx->dc_bios->fw_info.pll_info.crystal_frequency,
770 &res_pool->ref_clocks.dccg_ref_clock_inKhz);
772 (res_pool->hubbub->funcs->get_dchub_ref_freq)(res_pool->hubbub,
773 res_pool->ref_clocks.dccg_ref_clock_inKhz,
774 &res_pool->ref_clocks.dchub_ref_clock_inKhz);
776 // Not all ASICs have DCCG sw component
777 res_pool->ref_clocks.dccg_ref_clock_inKhz =
778 res_pool->ref_clocks.xtalin_clock_inKhz;
779 res_pool->ref_clocks.dchub_ref_clock_inKhz =
780 res_pool->ref_clocks.xtalin_clock_inKhz;
783 ASSERT_CRITICAL(false);
785 for (i = 0; i < dc->link_count; i++) {
786 /* Power up AND update implementation according to the
787 * required signal (which may be different from the
788 * default signal on connector).
790 struct dc_link *link = dc->links[i];
792 link->link_enc->funcs->hw_init(link->link_enc);
794 /* Check for enabled DIG to identify enabled display */
795 if (link->link_enc->funcs->is_dig_enabled &&
796 link->link_enc->funcs->is_dig_enabled(link->link_enc)) {
797 link->link_status.link_active = true;
798 if (link->link_enc->funcs->fec_is_active &&
799 link->link_enc->funcs->fec_is_active(link->link_enc))
800 link->fec_state = dc_link_fec_enabled;
804 /* Power gate DSCs */
805 for (i = 0; i < res_pool->res_cap->num_dsc; i++)
806 if (hws->funcs.dsc_pg_control != NULL)
807 hws->funcs.dsc_pg_control(hws, res_pool->dscs[i]->inst, false);
809 /* we want to turn off all dp displays before doing detection */
810 dc_link_blank_all_dp_displays(dc);
812 /* If taking control over from VBIOS, we may want to optimize our first
813 * mode set, so we need to skip powering down pipes until we know which
814 * pipes we want to use.
815 * Otherwise, if taking control is not possible, we need to power
818 if (dcb->funcs->is_accelerated_mode(dcb) || !dc->config.seamless_boot_edp_requested) {
819 hws->funcs.init_pipes(dc, dc->current_state);
820 if (dc->res_pool->hubbub->funcs->allow_self_refresh_control)
821 dc->res_pool->hubbub->funcs->allow_self_refresh_control(dc->res_pool->hubbub,
822 !dc->res_pool->hubbub->ctx->dc->debug.disable_stutter);
825 /* In headless boot cases, DIG may be turned
826 * on which causes HW/SW discrepancies.
827 * To avoid this, power down hardware on boot
828 * if DIG is turned on and seamless boot not enabled
830 if (!dc->config.seamless_boot_edp_requested) {
831 struct dc_link *edp_links[MAX_NUM_EDP];
832 struct dc_link *edp_link;
834 get_edp_links(dc, edp_links, &edp_num);
836 for (i = 0; i < edp_num; i++) {
837 edp_link = edp_links[i];
838 if (edp_link->link_enc->funcs->is_dig_enabled &&
839 edp_link->link_enc->funcs->is_dig_enabled(edp_link->link_enc) &&
840 dc->hwss.edp_backlight_control &&
841 dc->hwss.power_down &&
842 dc->hwss.edp_power_control) {
843 dc->hwss.edp_backlight_control(edp_link, false);
844 dc->hwss.power_down(dc);
845 dc->hwss.edp_power_control(edp_link, false);
849 for (i = 0; i < dc->link_count; i++) {
850 struct dc_link *link = dc->links[i];
852 if (link->link_enc->funcs->is_dig_enabled &&
853 link->link_enc->funcs->is_dig_enabled(link->link_enc) &&
854 dc->hwss.power_down) {
855 dc->hwss.power_down(dc);
863 for (i = 0; i < res_pool->audio_count; i++) {
864 struct audio *audio = res_pool->audios[i];
866 audio->funcs->hw_init(audio);
869 for (i = 0; i < dc->link_count; i++) {
870 struct dc_link *link = dc->links[i];
872 if (link->panel_cntl)
873 backlight = link->panel_cntl->funcs->hw_init(link->panel_cntl);
876 for (i = 0; i < dc->res_pool->pipe_count; i++) {
877 if (abms[i] != NULL && abms[i]->funcs != NULL)
878 abms[i]->funcs->abm_init(abms[i], backlight);
881 /* power AFMT HDMI memory TODO: may move to dis/en output save power*/
882 REG_WRITE(DIO_MEM_PWR_CTRL, 0);
884 if (!dc->debug.disable_clock_gate) {
885 /* enable all DCN clock gating */
886 REG_WRITE(DCCG_GATE_DISABLE_CNTL, 0);
888 REG_WRITE(DCCG_GATE_DISABLE_CNTL2, 0);
890 REG_UPDATE(DCFCLK_CNTL, DCFCLK_GATE_DIS, 0);
892 if (hws->funcs.enable_power_gating_plane)
893 hws->funcs.enable_power_gating_plane(dc->hwseq, true);
895 if (!dcb->funcs->is_accelerated_mode(dcb) && dc->res_pool->hubbub->funcs->init_watermarks)
896 dc->res_pool->hubbub->funcs->init_watermarks(dc->res_pool->hubbub);
898 if (dc->clk_mgr->funcs->notify_wm_ranges)
899 dc->clk_mgr->funcs->notify_wm_ranges(dc->clk_mgr);
901 if (dc->clk_mgr->funcs->set_hard_max_memclk)
902 dc->clk_mgr->funcs->set_hard_max_memclk(dc->clk_mgr);
904 if (dc->res_pool->hubbub->funcs->force_pstate_change_control)
905 dc->res_pool->hubbub->funcs->force_pstate_change_control(
906 dc->res_pool->hubbub, false, false);
908 if (dc->res_pool->hubbub->funcs->init_crb)
909 dc->res_pool->hubbub->funcs->init_crb(dc->res_pool->hubbub);
911 // Get DMCUB capabilities
912 if (dc->ctx->dmub_srv) {
913 dc_dmub_srv_query_caps_cmd(dc->ctx->dmub_srv->dmub);
914 dc->caps.dmub_caps.psr = dc->ctx->dmub_srv->dmub->feature_caps.psr;
918 static int calc_mpc_flow_ctrl_cnt(const struct dc_stream_state *stream,
921 bool hblank_halved = optc2_is_two_pixels_per_containter(&stream->timing);
925 hblank_halved = true;
927 flow_ctrl_cnt = stream->timing.h_total - stream->timing.h_addressable -
928 stream->timing.h_border_left -
929 stream->timing.h_border_right;
934 /* ODM combine 4:1 case */
938 return flow_ctrl_cnt;
941 static void update_dsc_on_stream(struct pipe_ctx *pipe_ctx, bool enable)
943 struct display_stream_compressor *dsc = pipe_ctx->stream_res.dsc;
944 struct dc_stream_state *stream = pipe_ctx->stream;
945 struct pipe_ctx *odm_pipe;
949 for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe)
953 struct dsc_config dsc_cfg;
954 struct dsc_optc_config dsc_optc_cfg;
955 enum optc_dsc_mode optc_dsc_mode;
957 /* Enable DSC hw block */
958 dsc_cfg.pic_width = (stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right) / opp_cnt;
959 dsc_cfg.pic_height = stream->timing.v_addressable + stream->timing.v_border_top + stream->timing.v_border_bottom;
960 dsc_cfg.pixel_encoding = stream->timing.pixel_encoding;
961 dsc_cfg.color_depth = stream->timing.display_color_depth;
962 dsc_cfg.is_odm = pipe_ctx->next_odm_pipe ? true : false;
963 dsc_cfg.dc_dsc_cfg = stream->timing.dsc_cfg;
964 ASSERT(dsc_cfg.dc_dsc_cfg.num_slices_h % opp_cnt == 0);
965 dsc_cfg.dc_dsc_cfg.num_slices_h /= opp_cnt;
967 dsc->funcs->dsc_set_config(dsc, &dsc_cfg, &dsc_optc_cfg);
968 dsc->funcs->dsc_enable(dsc, pipe_ctx->stream_res.opp->inst);
969 for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
970 struct display_stream_compressor *odm_dsc = odm_pipe->stream_res.dsc;
973 odm_dsc->funcs->dsc_set_config(odm_dsc, &dsc_cfg, &dsc_optc_cfg);
974 odm_dsc->funcs->dsc_enable(odm_dsc, odm_pipe->stream_res.opp->inst);
976 dsc_cfg.dc_dsc_cfg.num_slices_h *= opp_cnt;
977 dsc_cfg.pic_width *= opp_cnt;
979 optc_dsc_mode = dsc_optc_cfg.is_pixel_format_444 ? OPTC_DSC_ENABLED_444 : OPTC_DSC_ENABLED_NATIVE_SUBSAMPLED;
981 /* Enable DSC in OPTC */
982 DC_LOG_DSC("Setting optc DSC config for tg instance %d:", pipe_ctx->stream_res.tg->inst);
983 pipe_ctx->stream_res.tg->funcs->set_dsc_config(pipe_ctx->stream_res.tg,
985 dsc_optc_cfg.bytes_per_pixel,
986 dsc_optc_cfg.slice_width);
988 /* disable DSC in OPTC */
989 pipe_ctx->stream_res.tg->funcs->set_dsc_config(
990 pipe_ctx->stream_res.tg,
991 OPTC_DSC_DISABLED, 0, 0);
993 /* disable DSC block */
994 dsc->funcs->dsc_disable(pipe_ctx->stream_res.dsc);
995 for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
996 ASSERT(odm_pipe->stream_res.dsc);
997 odm_pipe->stream_res.dsc->funcs->dsc_disable(odm_pipe->stream_res.dsc);
1003 * Given any pipe_ctx, return the total ODM combine factor, and optionally return
1004 * the OPPids which are used
1006 static unsigned int get_odm_config(struct pipe_ctx *pipe_ctx, unsigned int *opp_instances)
1008 unsigned int opp_count = 1;
1009 struct pipe_ctx *odm_pipe;
1011 /* First get to the top pipe */
1012 for (odm_pipe = pipe_ctx; odm_pipe->prev_odm_pipe; odm_pipe = odm_pipe->prev_odm_pipe)
1015 /* First pipe is always used */
1017 opp_instances[0] = odm_pipe->stream_res.opp->inst;
1019 /* Find and count odm pipes, if any */
1020 for (odm_pipe = odm_pipe->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
1022 opp_instances[opp_count] = odm_pipe->stream_res.opp->inst;
1029 void dcn32_update_odm(struct dc *dc, struct dc_state *context, struct pipe_ctx *pipe_ctx)
1031 struct pipe_ctx *odm_pipe;
1033 int opp_inst[MAX_PIPES] = {0};
1034 bool rate_control_2x_pclk = (pipe_ctx->stream->timing.flags.INTERLACE || optc2_is_two_pixels_per_containter(&pipe_ctx->stream->timing));
1035 struct mpc_dwb_flow_control flow_control;
1036 struct mpc *mpc = dc->res_pool->mpc;
1039 opp_cnt = get_odm_config(pipe_ctx, opp_inst);
1042 pipe_ctx->stream_res.tg->funcs->set_odm_combine(
1043 pipe_ctx->stream_res.tg,
1045 &pipe_ctx->stream->timing);
1047 pipe_ctx->stream_res.tg->funcs->set_odm_bypass(
1048 pipe_ctx->stream_res.tg, &pipe_ctx->stream->timing);
1050 rate_control_2x_pclk = rate_control_2x_pclk || opp_cnt > 1;
1051 flow_control.flow_ctrl_mode = 0;
1052 flow_control.flow_ctrl_cnt0 = 0x80;
1053 flow_control.flow_ctrl_cnt1 = calc_mpc_flow_ctrl_cnt(pipe_ctx->stream, opp_cnt);
1054 if (mpc->funcs->set_out_rate_control) {
1055 for (i = 0; i < opp_cnt; ++i) {
1056 mpc->funcs->set_out_rate_control(
1059 rate_control_2x_pclk,
1064 for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) {
1065 odm_pipe->stream_res.opp->funcs->opp_pipe_clock_control(
1066 odm_pipe->stream_res.opp,
1070 // Don't program pixel clock after link is already enabled
1071 /* if (false == pipe_ctx->clock_source->funcs->program_pix_clk(
1072 pipe_ctx->clock_source,
1073 &pipe_ctx->stream_res.pix_clk_params,
1074 &pipe_ctx->pll_settings)) {
1075 BREAK_TO_DEBUGGER();
1078 if (pipe_ctx->stream_res.dsc)
1079 update_dsc_on_stream(pipe_ctx, pipe_ctx->stream->timing.flags.DSC);
1082 unsigned int dcn32_calculate_dccg_k1_k2_values(struct pipe_ctx *pipe_ctx, unsigned int *k1_div, unsigned int *k2_div)
1084 struct dc_stream_state *stream = pipe_ctx->stream;
1085 unsigned int odm_combine_factor = 0;
1086 struct dc *dc = pipe_ctx->stream->ctx->dc;
1087 bool two_pix_per_container = false;
1089 // For phantom pipes, use the same programming as the main pipes
1090 if (pipe_ctx->stream->mall_stream_config.type == SUBVP_PHANTOM) {
1091 stream = pipe_ctx->stream->mall_stream_config.paired_stream;
1093 two_pix_per_container = optc2_is_two_pixels_per_containter(&stream->timing);
1094 odm_combine_factor = get_odm_config(pipe_ctx, NULL);
1096 if (is_dp_128b_132b_signal(pipe_ctx)) {
1097 *k2_div = PIXEL_RATE_DIV_BY_1;
1098 } else if (dc_is_hdmi_tmds_signal(pipe_ctx->stream->signal) || dc_is_dvi_signal(pipe_ctx->stream->signal)) {
1099 *k1_div = PIXEL_RATE_DIV_BY_1;
1100 if (stream->timing.pixel_encoding == PIXEL_ENCODING_YCBCR420)
1101 *k2_div = PIXEL_RATE_DIV_BY_2;
1103 *k2_div = PIXEL_RATE_DIV_BY_4;
1104 } else if (dc_is_dp_signal(pipe_ctx->stream->signal)) {
1105 if (two_pix_per_container) {
1106 *k1_div = PIXEL_RATE_DIV_BY_1;
1107 *k2_div = PIXEL_RATE_DIV_BY_2;
1109 *k1_div = PIXEL_RATE_DIV_BY_1;
1110 *k2_div = PIXEL_RATE_DIV_BY_4;
1111 if ((odm_combine_factor == 2) || dc->debug.enable_dp_dig_pixel_rate_div_policy)
1112 *k2_div = PIXEL_RATE_DIV_BY_2;
1116 if ((*k1_div == PIXEL_RATE_DIV_NA) && (*k2_div == PIXEL_RATE_DIV_NA))
1119 return odm_combine_factor;
1122 void dcn32_set_pixels_per_cycle(struct pipe_ctx *pipe_ctx)
1124 uint32_t pix_per_cycle = 1;
1125 uint32_t odm_combine_factor = 1;
1127 if (!pipe_ctx || !pipe_ctx->stream || !pipe_ctx->stream_res.stream_enc)
1130 odm_combine_factor = get_odm_config(pipe_ctx, NULL);
1131 if (optc2_is_two_pixels_per_containter(&pipe_ctx->stream->timing) || odm_combine_factor > 1
1132 || dcn32_is_dp_dig_pixel_rate_div_policy(pipe_ctx))
1135 if (pipe_ctx->stream_res.stream_enc->funcs->set_input_mode)
1136 pipe_ctx->stream_res.stream_enc->funcs->set_input_mode(pipe_ctx->stream_res.stream_enc,
1140 void dcn32_unblank_stream(struct pipe_ctx *pipe_ctx,
1141 struct dc_link_settings *link_settings)
1143 struct encoder_unblank_param params = {0};
1144 struct dc_stream_state *stream = pipe_ctx->stream;
1145 struct dc_link *link = stream->link;
1146 struct dce_hwseq *hws = link->dc->hwseq;
1147 struct pipe_ctx *odm_pipe;
1148 struct dc *dc = pipe_ctx->stream->ctx->dc;
1149 uint32_t pix_per_cycle = 1;
1152 for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe)
1155 /* only 3 items below are used by unblank */
1156 params.timing = pipe_ctx->stream->timing;
1158 params.link_settings.link_rate = link_settings->link_rate;
1160 if (is_dp_128b_132b_signal(pipe_ctx)) {
1161 /* TODO - DP2.0 HW: Set ODM mode in dp hpo encoder here */
1162 pipe_ctx->stream_res.hpo_dp_stream_enc->funcs->dp_unblank(
1163 pipe_ctx->stream_res.hpo_dp_stream_enc,
1164 pipe_ctx->stream_res.tg->inst);
1165 } else if (dc_is_dp_signal(pipe_ctx->stream->signal)) {
1166 if (optc2_is_two_pixels_per_containter(&stream->timing) || params.opp_cnt > 1
1167 || dc->debug.enable_dp_dig_pixel_rate_div_policy) {
1168 params.timing.pix_clk_100hz /= 2;
1171 pipe_ctx->stream_res.stream_enc->funcs->dp_set_odm_combine(
1172 pipe_ctx->stream_res.stream_enc, pix_per_cycle > 1);
1173 pipe_ctx->stream_res.stream_enc->funcs->dp_unblank(link, pipe_ctx->stream_res.stream_enc, ¶ms);
1176 if (link->local_sink && link->local_sink->sink_signal == SIGNAL_TYPE_EDP)
1177 hws->funcs.edp_backlight_control(link, true);
1180 bool dcn32_is_dp_dig_pixel_rate_div_policy(struct pipe_ctx *pipe_ctx)
1182 struct dc *dc = pipe_ctx->stream->ctx->dc;
1184 if (dc_is_dp_signal(pipe_ctx->stream->signal) && !is_dp_128b_132b_signal(pipe_ctx) &&
1185 dc->debug.enable_dp_dig_pixel_rate_div_policy)