From 602816c3ee3fdbf4a72cb4d2e5b8b756a5104b0f Mon Sep 17 00:00:00 2001 From: Jonathan Kim Date: Wed, 12 Jul 2023 16:32:29 -0400 Subject: drm/amdkfd: fix trap handling work around for debugging Update the list of devices that require the cwsr trap handling workaround for debugging use cases. Signed-off-by: Jonathan Kim Acked-by: Ruili Ji Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_debug.c | 5 ++--- drivers/gpu/drm/amd/amdkfd/kfd_debug.h | 6 ++++++ drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 6 ++---- 3 files changed, 10 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c index fff3ccc04fa9..9766076e9ec4 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.c @@ -302,8 +302,7 @@ static int kfd_dbg_set_queue_workaround(struct queue *q, bool enable) if (!q) return 0; - if (KFD_GC_VERSION(q->device) < IP_VERSION(11, 0, 0) || - KFD_GC_VERSION(q->device) >= IP_VERSION(12, 0, 0)) + if (!kfd_dbg_has_cwsr_workaround(q->device)) return 0; if (enable && q->properties.is_user_cu_masked) @@ -349,7 +348,7 @@ int kfd_dbg_set_mes_debug_mode(struct kfd_process_device *pdd) { uint32_t spi_dbg_cntl = pdd->spi_dbg_override | pdd->spi_dbg_launch_mode; uint32_t flags = pdd->process->dbg_flags; - bool sq_trap_en = !!spi_dbg_cntl; + bool sq_trap_en = !!spi_dbg_cntl || !kfd_dbg_has_cwsr_workaround(pdd->dev); if (!kfd_dbg_is_per_vmid_supported(pdd->dev)) return 0; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h index a289e59ceb79..662a13a0d582 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_debug.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_debug.h @@ -100,6 +100,12 @@ static inline bool kfd_dbg_is_rlc_restore_supported(struct kfd_node *dev) KFD_GC_VERSION(dev) == IP_VERSION(10, 1, 1)); } +static inline bool kfd_dbg_has_cwsr_workaround(struct kfd_node *dev) +{ + return KFD_GC_VERSION(dev) >= IP_VERSION(11, 0, 0) && + KFD_GC_VERSION(dev) <= IP_VERSION(11, 0, 3); +} + static inline bool kfd_dbg_has_gws_support(struct kfd_node *dev) { if ((KFD_GC_VERSION(dev) == IP_VERSION(9, 0, 1) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index f515cb8f30ca..2df153828ff4 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -226,8 +226,7 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q, queue_input.paging = false; queue_input.tba_addr = qpd->tba_addr; queue_input.tma_addr = qpd->tma_addr; - queue_input.trap_en = KFD_GC_VERSION(q->device) < IP_VERSION(11, 0, 0) || - KFD_GC_VERSION(q->device) > IP_VERSION(11, 0, 3); + queue_input.trap_en = !kfd_dbg_has_cwsr_workaround(q->device); queue_input.skip_process_ctx_clear = qpd->pqm->process->debug_trap_enabled; queue_type = convert_to_mes_queue_type(q->properties.type); @@ -1806,8 +1805,7 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, */ q->properties.is_evicted = !!qpd->evicted; q->properties.is_dbg_wa = qpd->pqm->process->debug_trap_enabled && - KFD_GC_VERSION(q->device) >= IP_VERSION(11, 0, 0) && - KFD_GC_VERSION(q->device) <= IP_VERSION(11, 0, 3); + kfd_dbg_has_cwsr_workaround(q->device); if (qd) mqd_mgr->restore_mqd(mqd_mgr, &q->mqd, q->mqd_mem_obj, &q->gart_mqd_addr, -- cgit v1.2.3 From 9beb223f2a3d1bb2cc4dff71b9750d2b82f01ae5 Mon Sep 17 00:00:00 2001 From: Victor Lu Date: Tue, 18 Jul 2023 14:31:58 -0400 Subject: drm/amdgpu: Fix infinite loop in gfxhub_v1_2_xcc_gart_enable (v2) An instance of for_each_inst() was not changed to match its new behaviour and is causing a loop. v2: remove tmp_mask variable Fixes: b579ea632fca ("drm/amdgpu: Modify for_each_inst macro") Signed-off-by: Victor Lu Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c index 4dabf910334b..d9f14dc55998 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c @@ -402,18 +402,15 @@ static void gfxhub_v1_2_xcc_program_invalidation(struct amdgpu_device *adev, static int gfxhub_v1_2_xcc_gart_enable(struct amdgpu_device *adev, uint32_t xcc_mask) { - uint32_t tmp_mask; int i; - tmp_mask = xcc_mask; /* * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, because they are * VF copy registers so vbios post doesn't program them, for * SRIOV driver need to program them */ if (amdgpu_sriov_vf(adev)) { - for_each_inst(i, tmp_mask) { - i = ffs(tmp_mask) - 1; + for_each_inst(i, xcc_mask) { WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_BASE, adev->gmc.vram_start >> 24); WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_TOP, -- cgit v1.2.3 From c01aebeef3ce45f696ffa0a1303cea9b34babb45 Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Thu, 13 Jul 2023 00:14:59 -0500 Subject: drm/amd: Fix an error handling mistake in psp_sw_init() If the second call to amdgpu_bo_create_kernel() fails, the memory allocated from the first call should be cleared. If the third call fails, the memory from the second call should be cleared. Fixes: b95b5391684b ("drm/amdgpu/psp: move PSP memory alloc from hw_init to sw_init") Signed-off-by: Mario Limonciello Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 6d676bdd1505..78d1ee71f3f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -498,11 +498,11 @@ static int psp_sw_init(void *handle) return 0; failed2: - amdgpu_bo_free_kernel(&psp->fw_pri_bo, - &psp->fw_pri_mc_addr, &psp->fw_pri_buf); -failed1: amdgpu_bo_free_kernel(&psp->fence_buf_bo, &psp->fence_buf_mc_addr, &psp->fence_buf); +failed1: + amdgpu_bo_free_kernel(&psp->fw_pri_bo, + &psp->fw_pri_mc_addr, &psp->fw_pri_buf); return ret; } -- cgit v1.2.3 From 4a37c55b859a69f429bfa7fab4fc43ee470b60ed Mon Sep 17 00:00:00 2001 From: Jane Jian Date: Thu, 20 Jul 2023 18:08:07 +0800 Subject: drm/amd/smu: use AverageGfxclkFrequency* to replace previous GFX Curr Clock Report current GFX clock also from average clock value as the original CurrClock data is not valid/accurate any more as per FW team Signed-off-by: Jane Jian Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c index b9bde5fa8f8f..3d188616ba24 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c @@ -1734,7 +1734,7 @@ static ssize_t smu_v13_0_0_get_gpu_metrics(struct smu_context *smu, gpu_metrics->average_vclk1_frequency = metrics->AverageVclk1Frequency; gpu_metrics->average_dclk1_frequency = metrics->AverageDclk1Frequency; - gpu_metrics->current_gfxclk = metrics->CurrClock[PPCLK_GFXCLK]; + gpu_metrics->current_gfxclk = gpu_metrics->average_gfxclk_frequency; gpu_metrics->current_socclk = metrics->CurrClock[PPCLK_SOCCLK]; gpu_metrics->current_uclk = metrics->CurrClock[PPCLK_UCLK]; gpu_metrics->current_vclk0 = metrics->CurrClock[PPCLK_VCLK_0]; -- cgit v1.2.3 From 25b054c3c89cb6a7106a7982f0f70e83d0797dab Mon Sep 17 00:00:00 2001 From: George Shen Date: Tue, 11 Jul 2023 13:22:36 -0400 Subject: drm/amd/display: Guard DCN31 PHYD32CLK logic against chip family [Why] Current yellow carp B0 PHYD32CLK logic is incorrectly applied to other ASICs. [How] Add guard to check chip family is yellow carp before applying logic. Reviewed-by: Hansen Dsouza Acked-by: Alex Hung Signed-off-by: George Shen Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dcn31/dcn31_dccg.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_dccg.c b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_dccg.c index 65c1d754e2d6..01cc679ae418 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_dccg.c +++ b/drivers/gpu/drm/amd/display/dc/dcn31/dcn31_dccg.c @@ -84,7 +84,8 @@ static enum phyd32clk_clock_source get_phy_mux_symclk( struct dcn_dccg *dccg_dcn, enum phyd32clk_clock_source src) { - if (dccg_dcn->base.ctx->asic_id.hw_internal_rev == YELLOW_CARP_B0) { + if (dccg_dcn->base.ctx->asic_id.chip_family == FAMILY_YELLOW_CARP && + dccg_dcn->base.ctx->asic_id.hw_internal_rev == YELLOW_CARP_B0) { if (src == PHYD32CLKC) src = PHYD32CLKF; if (src == PHYD32CLKD) -- cgit v1.2.3 From 4509e69a07761d08df7c46d4a08c8222522b1933 Mon Sep 17 00:00:00 2001 From: Alvin Lee Date: Wed, 12 Jul 2023 17:22:25 -0400 Subject: drm/amd/display: Don't apply FIFO resync W/A if rdivider = 0 [Description] It is not valid to set the WDIVIDER value to 0, so do not re-write to DISPCLK_WDIVIDER if the current value is 0 (i.e., it is at it's initial value and we have not made any requests to change DISPCLK yet). Reviewed-by: Saaem Rizvi Acked-by: Alex Hung Signed-off-by: Alvin Lee Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dcn32/dcn32_dccg.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_dccg.c b/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_dccg.c index 11e28e056cf7..61ceff6bc0b1 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_dccg.c +++ b/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_dccg.c @@ -49,7 +49,10 @@ static void dccg32_trigger_dio_fifo_resync( uint32_t dispclk_rdivider_value = 0; REG_GET(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_RDIVIDER, &dispclk_rdivider_value); - REG_UPDATE(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_WDIVIDER, dispclk_rdivider_value); + + /* Not valid for the WDIVIDER to be set to 0 */ + if (dispclk_rdivider_value != 0) + REG_UPDATE(DENTIST_DISPCLK_CNTL, DENTIST_DISPCLK_WDIVIDER, dispclk_rdivider_value); } static void dccg32_get_pixel_rate_div( -- cgit v1.2.3 From de612738e9771bd66aeb20044486c457c512f684 Mon Sep 17 00:00:00 2001 From: Leo Chen Date: Wed, 12 Jul 2023 16:50:15 -0400 Subject: drm/amd/display: Exit idle optimizations before attempt to access PHY [Why & How] DMUB may hang when powering down pixel clocks due to no dprefclk. It is fixed by exiting idle optimization before the attempt to access PHY. Reviewed-by: Nicholas Kazlauskas Acked-by: Alex Hung Signed-off-by: Leo Chen Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c index 6c9ca43d1040..20d4d08a6a2f 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c @@ -1792,10 +1792,13 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) hws->funcs.edp_backlight_control(edp_link_with_sink, false); } /*resume from S3, no vbios posting, no need to power down again*/ + clk_mgr_exit_optimized_pwr_state(dc, dc->clk_mgr); + power_down_all_hw_blocks(dc); disable_vga_and_power_gate_all_controllers(dc); if (edp_link_with_sink && !keep_edp_vdd_on) dc->hwss.edp_power_control(edp_link_with_sink, false); + clk_mgr_optimize_pwr_state(dc, dc->clk_mgr); } bios_set_scratch_acc_mode_change(dc->ctx->dc_bios, 1); } -- cgit v1.2.3 From 38ac4e8385ffb275b1837986ca6c16f26ea028c5 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Fri, 21 Jul 2023 17:55:49 +0300 Subject: drm/amd/display: Unlock on error path in dm_handle_mst_sideband_msg_ready_event() This error path needs to unlock the "aconnector->handle_mst_msg_ready" mutex before returning. Fixes: 4f6d9e38c4d2 ("drm/amd/display: Add polling method to handle MST reply packet") Signed-off-by: Dan Carpenter Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index 888e80f498e9..9bc86deac9e8 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -706,7 +706,7 @@ void dm_handle_mst_sideband_msg_ready_event( if (retry == 3) { DRM_ERROR("Failed to ack MST event.\n"); - return; + break; } drm_dp_mst_hpd_irq_send_new_request(&aconnector->mst_mgr); -- cgit v1.2.3 From bc1688fce2ec7726112276650762275392d1bab1 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Tue, 25 Jul 2023 11:00:08 +0530 Subject: drm/amdgpu: Restore HQD persistent state register On GFX v9.4.3, compute queue MQD is populated using the values in HQD persistent state register. Hence don't clear the values on module unload, instead restore it to the default reset value so that MQD is initialized correctly during next module load. In particular, preload flag needs to be set on compute queue MQD, otherwise it could cause uninitialized values being used at device reset state resulting in EDC. Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Reviewed-by: Asad Kamal Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c index 9e3b835bdbb2..4f883b94f98e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c @@ -46,6 +46,7 @@ MODULE_FIRMWARE("amdgpu/gc_9_4_3_rlc.bin"); #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L #define GOLDEN_GB_ADDR_CONFIG 0x2a114042 +#define CP_HQD_PERSISTENT_STATE_DEFAULT 0xbe05301 struct amdgpu_gfx_ras gfx_v9_4_3_ras; @@ -1736,7 +1737,7 @@ static int gfx_v9_4_3_xcc_q_fini_register(struct amdgpu_ring *ring, WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IQ_TIMER, 0); WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IB_CONTROL, 0); - WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE, 0); + WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE, CP_HQD_PERSISTENT_STATE_DEFAULT); WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0x40000000); WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0); WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR, 0); -- cgit v1.2.3