diff options
Diffstat (limited to 'drivers/gpu/drm/amd/display/dc/dcn10')
6 files changed, 174 insertions, 40 deletions
diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c index 1008ac8a0f2a..0c987b5d68e2 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c @@ -82,7 +82,7 @@ void print_microsec(struct dc_context *dc_ctx, us_x10 % frac); } -static void dcn10_lock_all_pipes(struct dc *dc, +void dcn10_lock_all_pipes(struct dc *dc, struct dc_state *context, bool lock) { @@ -93,6 +93,7 @@ static void dcn10_lock_all_pipes(struct dc *dc, for (i = 0; i < dc->res_pool->pipe_count; i++) { pipe_ctx = &context->res_ctx.pipe_ctx[i]; tg = pipe_ctx->stream_res.tg; + /* * Only lock the top pipe's tg to prevent redundant * (un)locking. Also skip if pipe is disabled. @@ -103,9 +104,9 @@ static void dcn10_lock_all_pipes(struct dc *dc, continue; if (lock) - tg->funcs->lock(tg); + dc->hwss.pipe_control_lock(dc, pipe_ctx, true); else - tg->funcs->unlock(tg); + dc->hwss.pipe_control_lock(dc, pipe_ctx, false); } } @@ -1576,7 +1577,7 @@ void dcn10_pipe_control_lock( /* use TG master update lock to lock everything on the TG * therefore only top pipe need to lock */ - if (pipe->top_pipe) + if (!pipe || pipe->top_pipe) return; if (dc->debug.sanity_checks) @@ -1591,6 +1592,85 @@ void dcn10_pipe_control_lock( hws->funcs.verify_allow_pstate_change_high(dc); } +/** + * delay_cursor_until_vupdate() - Delay cursor update if too close to VUPDATE. + * + * Software keepout workaround to prevent cursor update locking from stalling + * out cursor updates indefinitely or from old values from being retained in + * the case where the viewport changes in the same frame as the cursor. + * + * The idea is to calculate the remaining time from VPOS to VUPDATE. If it's + * too close to VUPDATE, then stall out until VUPDATE finishes. + * + * TODO: Optimize cursor programming to be once per frame before VUPDATE + * to avoid the need for this workaround. + */ +static void delay_cursor_until_vupdate(struct dc *dc, struct pipe_ctx *pipe_ctx) +{ + struct dc_stream_state *stream = pipe_ctx->stream; + struct crtc_position position; + uint32_t vupdate_start, vupdate_end; + unsigned int lines_to_vupdate, us_to_vupdate, vpos; + unsigned int us_per_line, us_vupdate; + + if (!dc->hwss.calc_vupdate_position || !dc->hwss.get_position) + return; + + if (!pipe_ctx->stream_res.stream_enc || !pipe_ctx->stream_res.tg) + return; + + dc->hwss.calc_vupdate_position(dc, pipe_ctx, &vupdate_start, + &vupdate_end); + + dc->hwss.get_position(&pipe_ctx, 1, &position); + vpos = position.vertical_count; + + /* Avoid wraparound calculation issues */ + vupdate_start += stream->timing.v_total; + vupdate_end += stream->timing.v_total; + vpos += stream->timing.v_total; + + if (vpos <= vupdate_start) { + /* VPOS is in VACTIVE or back porch. */ + lines_to_vupdate = vupdate_start - vpos; + } else if (vpos > vupdate_end) { + /* VPOS is in the front porch. */ + return; + } else { + /* VPOS is in VUPDATE. */ + lines_to_vupdate = 0; + } + + /* Calculate time until VUPDATE in microseconds. */ + us_per_line = + stream->timing.h_total * 10000u / stream->timing.pix_clk_100hz; + us_to_vupdate = lines_to_vupdate * us_per_line; + + /* 70 us is a conservative estimate of cursor update time*/ + if (us_to_vupdate > 70) + return; + + /* Stall out until the cursor update completes. */ + if (vupdate_end < vupdate_start) + vupdate_end += stream->timing.v_total; + us_vupdate = (vupdate_end - vupdate_start + 1) * us_per_line; + udelay(us_to_vupdate + us_vupdate); +} + +void dcn10_cursor_lock(struct dc *dc, struct pipe_ctx *pipe, bool lock) +{ + /* cursor lock is per MPCC tree, so only need to lock one pipe per stream */ + if (!pipe || pipe->top_pipe) + return; + + /* Prevent cursor lock from stalling out cursor updates. */ + if (lock) + delay_cursor_until_vupdate(dc, pipe); + + dc->res_pool->mpc->funcs->cursor_lock(dc->res_pool->mpc, + pipe->stream_res.opp->inst, lock); +} + static bool wait_for_reset_trigger_to_occur( struct dc_context *dc_ctx, struct timing_generator *tg) @@ -2512,7 +2592,6 @@ void dcn10_apply_ctx_for_surface( int i; struct timing_generator *tg; uint32_t underflow_check_delay_us; - bool removed_pipe[4] = { false }; bool interdependent_update = false; struct pipe_ctx *top_pipe_to_program = dcn10_find_top_pipe_for_stream(dc, context, stream); @@ -2531,11 +2610,6 @@ void dcn10_apply_ctx_for_surface( if (underflow_check_delay_us != 0xFFFFFFFF && hws->funcs.did_underflow_occur) ASSERT(hws->funcs.did_underflow_occur(dc, top_pipe_to_program)); - if (interdependent_update) - dcn10_lock_all_pipes(dc, context, true); - else - dcn10_pipe_control_lock(dc, top_pipe_to_program, true); - if (underflow_check_delay_us != 0xFFFFFFFF) udelay(underflow_check_delay_us); @@ -2552,18 +2626,8 @@ void dcn10_apply_ctx_for_surface( struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; struct pipe_ctx *old_pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i]; - /* - * Powergate reused pipes that are not powergated - * fairly hacky right now, using opp_id as indicator - * TODO: After move dc_post to dc_update, this will - * be removed. - */ - if (pipe_ctx->plane_state && !old_pipe_ctx->plane_state) { - if (old_pipe_ctx->stream_res.tg == tg && - old_pipe_ctx->plane_res.hubp && - old_pipe_ctx->plane_res.hubp->opp_id != OPP_ID_INVALID) - dc->hwss.disable_plane(dc, old_pipe_ctx); - } + + pipe_ctx->update_flags.raw = 0; if ((!pipe_ctx->plane_state || pipe_ctx->stream_res.tg != old_pipe_ctx->stream_res.tg) && @@ -2571,7 +2635,7 @@ void dcn10_apply_ctx_for_surface( old_pipe_ctx->stream_res.tg == tg) { hws->funcs.plane_atomic_disconnect(dc, old_pipe_ctx); - removed_pipe[i] = true; + pipe_ctx->update_flags.bits.disable = 1; DC_LOG_DC("Reset mpcc for pipe %d\n", old_pipe_ctx->pipe_idx); @@ -2597,21 +2661,41 @@ void dcn10_apply_ctx_for_surface( &pipe_ctx->dlg_regs, &pipe_ctx->ttu_regs); } +} - if (interdependent_update) - dcn10_lock_all_pipes(dc, context, false); - else - dcn10_pipe_control_lock(dc, top_pipe_to_program, false); +void dcn10_post_unlock_program_front_end( + struct dc *dc, + struct dc_state *context) +{ + int i, j; - if (num_planes == 0) - false_optc_underflow_wa(dc, stream, tg); + DC_LOGGER_INIT(dc->ctx->logger); + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + + if (!pipe_ctx->top_pipe && + !pipe_ctx->prev_odm_pipe && + pipe_ctx->stream) { + struct dc_stream_status *stream_status = NULL; + struct timing_generator *tg = pipe_ctx->stream_res.tg; + + for (j = 0; j < context->stream_count; j++) { + if (pipe_ctx->stream == context->streams[j]) + stream_status = &context->stream_status[j]; + } + + if (context->stream_status[i].plane_count == 0) + false_optc_underflow_wa(dc, pipe_ctx->stream, tg); + } + } for (i = 0; i < dc->res_pool->pipe_count; i++) - if (removed_pipe[i]) + if (context->res_ctx.pipe_ctx[i].update_flags.bits.disable) dc->hwss.disable_plane(dc, &dc->current_state->res_ctx.pipe_ctx[i]); for (i = 0; i < dc->res_pool->pipe_count; i++) - if (removed_pipe[i]) { + if (context->res_ctx.pipe_ctx[i].update_flags.bits.disable) { dc->hwss.optimize_bandwidth(dc, context); break; } @@ -3127,7 +3211,7 @@ int dcn10_get_vupdate_offset_from_vsync(struct pipe_ctx *pipe_ctx) return vertical_line_start; } -static void dcn10_calc_vupdate_position( +void dcn10_calc_vupdate_position( struct dc *dc, struct pipe_ctx *pipe_ctx, uint32_t *start_line, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h index 4d20f6586bb5..42b6e016d71e 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h @@ -34,6 +34,11 @@ struct dc; void dcn10_hw_sequencer_construct(struct dc *dc); int dcn10_get_vupdate_offset_from_vsync(struct pipe_ctx *pipe_ctx); +void dcn10_calc_vupdate_position( + struct dc *dc, + struct pipe_ctx *pipe_ctx, + uint32_t *start_line, + uint32_t *end_line); void dcn10_setup_vupdate_interrupt(struct dc *dc, struct pipe_ctx *pipe_ctx); enum dc_status dcn10_enable_stream_timing( struct pipe_ctx *pipe_ctx, @@ -49,6 +54,7 @@ void dcn10_pipe_control_lock( struct dc *dc, struct pipe_ctx *pipe, bool lock); +void dcn10_cursor_lock(struct dc *dc, struct pipe_ctx *pipe, bool lock); void dcn10_blank_pixel_data( struct dc *dc, struct pipe_ctx *pipe_ctx, @@ -70,11 +76,18 @@ void dcn10_reset_hw_ctx_wrap( struct dc *dc, struct dc_state *context); void dcn10_disable_plane(struct dc *dc, struct pipe_ctx *pipe_ctx); +void dcn10_lock_all_pipes( + struct dc *dc, + struct dc_state *context, + bool lock); void dcn10_apply_ctx_for_surface( struct dc *dc, const struct dc_stream_state *stream, int num_planes, struct dc_state *context); +void dcn10_post_unlock_program_front_end( + struct dc *dc, + struct dc_state *context); void dcn10_hubp_pg_control( struct dce_hwseq *hws, unsigned int hubp_inst, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c index e7e5352ec424..0900c861204f 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_init.c @@ -32,6 +32,7 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .init_hw = dcn10_init_hw, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, .apply_ctx_for_surface = dcn10_apply_ctx_for_surface, + .post_unlock_program_front_end = dcn10_post_unlock_program_front_end, .update_plane_addr = dcn10_update_plane_addr, .update_dchub = dcn10_update_dchub, .update_pending_status = dcn10_update_pending_status, @@ -49,6 +50,8 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .disable_audio_stream = dce110_disable_audio_stream, .disable_plane = dcn10_disable_plane, .pipe_control_lock = dcn10_pipe_control_lock, + .cursor_lock = dcn10_cursor_lock, + .interdependent_update_lock = dcn10_lock_all_pipes, .prepare_bandwidth = dcn10_prepare_bandwidth, .optimize_bandwidth = dcn10_optimize_bandwidth, .set_drr = dcn10_set_drr, @@ -69,6 +72,7 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .set_clock = dcn10_set_clock, .get_clock = dcn10_get_clock, .get_vupdate_offset_from_vsync = dcn10_get_vupdate_offset_from_vsync, + .calc_vupdate_position = dcn10_calc_vupdate_position, }; static const struct hwseq_private_funcs dcn10_private_funcs = { diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c index 04f863499cfb..3fcd408e9103 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c @@ -223,6 +223,9 @@ struct mpcc *mpc1_insert_plane( REG_SET(MPCC_TOP_SEL[mpcc_id], 0, MPCC_TOP_SEL, dpp_id); REG_SET(MPCC_OPP_ID[mpcc_id], 0, MPCC_OPP_ID, tree->opp_id); + /* Configure VUPDATE lock set for this MPCC to map to the OPP */ + REG_SET(MPCC_UPDATE_LOCK_SEL[mpcc_id], 0, MPCC_UPDATE_LOCK_SEL, tree->opp_id); + /* update mpc tree mux setting */ if (tree->opp_list == insert_above_mpcc) { /* insert the toppest mpcc */ @@ -318,6 +321,7 @@ void mpc1_remove_mpcc( REG_SET(MPCC_TOP_SEL[mpcc_id], 0, MPCC_TOP_SEL, 0xf); REG_SET(MPCC_BOT_SEL[mpcc_id], 0, MPCC_BOT_SEL, 0xf); REG_SET(MPCC_OPP_ID[mpcc_id], 0, MPCC_OPP_ID, 0xf); + REG_SET(MPCC_UPDATE_LOCK_SEL[mpcc_id], 0, MPCC_UPDATE_LOCK_SEL, 0xf); /* mark this mpcc as not in use */ mpc10->mpcc_in_use_mask &= ~(1 << mpcc_id); @@ -328,6 +332,7 @@ void mpc1_remove_mpcc( REG_SET(MPCC_TOP_SEL[mpcc_id], 0, MPCC_TOP_SEL, 0xf); REG_SET(MPCC_BOT_SEL[mpcc_id], 0, MPCC_BOT_SEL, 0xf); REG_SET(MPCC_OPP_ID[mpcc_id], 0, MPCC_OPP_ID, 0xf); + REG_SET(MPCC_UPDATE_LOCK_SEL[mpcc_id], 0, MPCC_UPDATE_LOCK_SEL, 0xf); } } @@ -361,6 +366,7 @@ void mpc1_mpc_init(struct mpc *mpc) REG_SET(MPCC_TOP_SEL[mpcc_id], 0, MPCC_TOP_SEL, 0xf); REG_SET(MPCC_BOT_SEL[mpcc_id], 0, MPCC_BOT_SEL, 0xf); REG_SET(MPCC_OPP_ID[mpcc_id], 0, MPCC_OPP_ID, 0xf); + REG_SET(MPCC_UPDATE_LOCK_SEL[mpcc_id], 0, MPCC_UPDATE_LOCK_SEL, 0xf); mpc1_init_mpcc(&(mpc->mpcc_array[mpcc_id]), mpcc_id); } @@ -381,6 +387,7 @@ void mpc1_mpc_init_single_inst(struct mpc *mpc, unsigned int mpcc_id) REG_SET(MPCC_TOP_SEL[mpcc_id], 0, MPCC_TOP_SEL, 0xf); REG_SET(MPCC_BOT_SEL[mpcc_id], 0, MPCC_BOT_SEL, 0xf); REG_SET(MPCC_OPP_ID[mpcc_id], 0, MPCC_OPP_ID, 0xf); + REG_SET(MPCC_UPDATE_LOCK_SEL[mpcc_id], 0, MPCC_UPDATE_LOCK_SEL, 0xf); mpc1_init_mpcc(&(mpc->mpcc_array[mpcc_id]), mpcc_id); @@ -453,6 +460,13 @@ void mpc1_read_mpcc_state( MPCC_BUSY, &s->busy); } +void mpc1_cursor_lock(struct mpc *mpc, int opp_id, bool lock) +{ + struct dcn10_mpc *mpc10 = TO_DCN10_MPC(mpc); + + REG_SET(CUR[opp_id], 0, CUR_VUPDATE_LOCK_SET, lock ? 1 : 0); +} + static const struct mpc_funcs dcn10_mpc_funcs = { .read_mpcc_state = mpc1_read_mpcc_state, .insert_plane = mpc1_insert_plane, @@ -464,6 +478,7 @@ static const struct mpc_funcs dcn10_mpc_funcs = { .assert_mpcc_idle_before_connect = mpc1_assert_mpcc_idle_before_connect, .init_mpcc_list_from_hw = mpc1_init_mpcc_list_from_hw, .update_blending = mpc1_update_blending, + .cursor_lock = mpc1_cursor_lock, .set_denorm = NULL, .set_denorm_clamp = NULL, .set_output_csc = NULL, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h index 962a68e322ee..66a4719c22a0 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h @@ -39,11 +39,12 @@ SRII(MPCC_BG_G_Y, MPCC, inst),\ SRII(MPCC_BG_R_CR, MPCC, inst),\ SRII(MPCC_BG_B_CB, MPCC, inst),\ - SRII(MPCC_BG_B_CB, MPCC, inst),\ - SRII(MPCC_SM_CONTROL, MPCC, inst) + SRII(MPCC_SM_CONTROL, MPCC, inst),\ + SRII(MPCC_UPDATE_LOCK_SEL, MPCC, inst) #define MPC_OUT_MUX_COMMON_REG_LIST_DCN1_0(inst) \ - SRII(MUX, MPC_OUT, inst) + SRII(MUX, MPC_OUT, inst),\ + VUPDATE_SRII(CUR, VUPDATE_LOCK_SET, inst) #define MPC_COMMON_REG_VARIABLE_LIST \ uint32_t MPCC_TOP_SEL[MAX_MPCC]; \ @@ -55,7 +56,9 @@ uint32_t MPCC_BG_R_CR[MAX_MPCC]; \ uint32_t MPCC_BG_B_CB[MAX_MPCC]; \ uint32_t MPCC_SM_CONTROL[MAX_MPCC]; \ - uint32_t MUX[MAX_OPP]; + uint32_t MUX[MAX_OPP]; \ + uint32_t MPCC_UPDATE_LOCK_SEL[MAX_MPCC]; \ + uint32_t CUR[MAX_OPP]; #define MPC_COMMON_MASK_SH_LIST_DCN1_0(mask_sh)\ SF(MPCC0_MPCC_TOP_SEL, MPCC_TOP_SEL, mask_sh),\ @@ -78,7 +81,8 @@ SF(MPCC0_MPCC_SM_CONTROL, MPCC_SM_FIELD_ALT, mask_sh),\ SF(MPCC0_MPCC_SM_CONTROL, MPCC_SM_FORCE_NEXT_FRAME_POL, mask_sh),\ SF(MPCC0_MPCC_SM_CONTROL, MPCC_SM_FORCE_NEXT_TOP_POL, mask_sh),\ - SF(MPC_OUT0_MUX, MPC_OUT_MUX, mask_sh) + SF(MPC_OUT0_MUX, MPC_OUT_MUX, mask_sh),\ + SF(MPCC0_MPCC_UPDATE_LOCK_SEL, MPCC_UPDATE_LOCK_SEL, mask_sh) #define MPC_REG_FIELD_LIST(type) \ type MPCC_TOP_SEL;\ @@ -101,7 +105,9 @@ type MPCC_SM_FIELD_ALT;\ type MPCC_SM_FORCE_NEXT_FRAME_POL;\ type MPCC_SM_FORCE_NEXT_TOP_POL;\ - type MPC_OUT_MUX; + type MPC_OUT_MUX;\ + type MPCC_UPDATE_LOCK_SEL;\ + type CUR_VUPDATE_LOCK_SET; struct dcn_mpc_registers { MPC_COMMON_REG_VARIABLE_LIST @@ -192,4 +198,6 @@ void mpc1_read_mpcc_state( int mpcc_inst, struct mpcc_state *s); +void mpc1_cursor_lock(struct mpc *mpc, int opp_id, bool lock); + #endif diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c index 3b71898e859e..e3c4c06ac191 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c @@ -181,6 +181,14 @@ enum dcn10_clk_src_array_id { .reg_name[id] = BASE(mm ## block ## id ## _ ## reg_name ## _BASE_IDX) + \ mm ## block ## id ## _ ## reg_name +#define VUPDATE_SRII(reg_name, block, id)\ + .reg_name[id] = BASE(mm ## reg_name ## 0 ## _ ## block ## id ## _BASE_IDX) + \ + mm ## reg_name ## 0 ## _ ## block ## id + +/* set field/register/bitfield name */ +#define SFRB(field_name, reg_name, bitfield, post_fix)\ + .field_name = reg_name ## __ ## bitfield ## post_fix + /* NBIO */ #define NBIO_BASE_INNER(seg) \ NBIF_BASE__INST0_SEG ## seg @@ -419,11 +427,13 @@ static const struct dcn_mpc_registers mpc_regs = { }; static const struct dcn_mpc_shift mpc_shift = { - MPC_COMMON_MASK_SH_LIST_DCN1_0(__SHIFT) + MPC_COMMON_MASK_SH_LIST_DCN1_0(__SHIFT),\ + SFRB(CUR_VUPDATE_LOCK_SET, CUR0_VUPDATE_LOCK_SET0, CUR0_VUPDATE_LOCK_SET, __SHIFT) }; static const struct dcn_mpc_mask mpc_mask = { - MPC_COMMON_MASK_SH_LIST_DCN1_0(_MASK), + MPC_COMMON_MASK_SH_LIST_DCN1_0(_MASK),\ + SFRB(CUR_VUPDATE_LOCK_SET, CUR0_VUPDATE_LOCK_SET0, CUR0_VUPDATE_LOCK_SET, _MASK) }; #define tg_regs(id)\ |