From a55737dab6ba63eb4241e9c6547629058af31e12 Mon Sep 17 00:00:00 2001 From: "Jesse.Zhang" Date: Thu, 29 May 2025 11:27:37 +0800 Subject: [PATCH 01/20] drm/amdkfd: move SDMA queue reset capability check to node_show Relocate the per-SDMA queue reset capability check from kfd_topology_set_capabilities() to node_show() to ensure we read the latest value of sdma.supported_reset after all IP blocks are initialized. Fixes: ceb7114c961b ("drm/amdkfd: flag per-sdma queue reset supported to user space") Reviewed-by: Jonathan Kim Signed-off-by: Jesse Zhang Signed-off-by: Alex Deucher (cherry picked from commit e17df7b086cf908cedd919f448da9e00028419bb) --- drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index baa2374acdeb..4ec73f33535e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -510,6 +510,10 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr, dev->node_props.capability |= HSA_CAP_AQL_QUEUE_DOUBLE_MAP; + if (KFD_GC_VERSION(dev->gpu) < IP_VERSION(10, 0, 0) && + (dev->gpu->adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE)) + dev->node_props.capability2 |= HSA_CAP2_PER_SDMA_QUEUE_RESET_SUPPORTED; + sysfs_show_32bit_prop(buffer, offs, "max_engine_clk_fcompute", dev->node_props.max_engine_clk_fcompute); @@ -2008,8 +2012,6 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev) if (!amdgpu_sriov_vf(dev->gpu->adev)) dev->node_props.capability |= HSA_CAP_PER_QUEUE_RESET_SUPPORTED; - if (dev->gpu->adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE) - dev->node_props.capability2 |= HSA_CAP2_PER_SDMA_QUEUE_RESET_SUPPORTED; } else { dev->node_props.debug_prop |= HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10 | HSA_DBG_WATCH_ADDR_MASK_HI_BIT; From 3251b69b7efb82eba24c9e168a6142a3078de72f Mon Sep 17 00:00:00 2001 From: Peichen Huang Date: Mon, 26 May 2025 16:04:10 +0800 Subject: [PATCH 02/20] drm/amd/display: Add dc cap for dp tunneling [WHAT] 1. add dc cap for dp tunneling 2. add function to get index of host router Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Cruise Hung Signed-off-by: Peichen Huang Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit 29e178d13979cf6fdb42c5fe2dfec2da2306c4ad) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/display/dc/core/dc.c | 33 +++++++++++++++++++ drivers/gpu/drm/amd/display/dc/dc.h | 8 ++++- .../dc/resource/dcn31/dcn31_resource.c | 3 ++ .../dc/resource/dcn314/dcn314_resource.c | 3 ++ .../dc/resource/dcn35/dcn35_resource.c | 3 ++ .../dc/resource/dcn351/dcn351_resource.c | 3 ++ .../dc/resource/dcn36/dcn36_resource.c | 3 ++ 7 files changed, 55 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 56d011a1323c..b34b5b52236d 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -241,6 +241,7 @@ static bool create_links( DC_LOG_DC("BIOS object table - end"); /* Create a link for each usb4 dpia port */ + dc->lowest_dpia_link_index = MAX_LINKS; for (i = 0; i < dc->res_pool->usb4_dpia_count; i++) { struct link_init_data link_init_params = {0}; struct dc_link *link; @@ -253,6 +254,9 @@ static bool create_links( link = dc->link_srv->create_link(&link_init_params); if (link) { + if (dc->lowest_dpia_link_index > dc->link_count) + dc->lowest_dpia_link_index = dc->link_count; + dc->links[dc->link_count] = link; link->dc = dc; ++dc->link_count; @@ -6376,6 +6380,35 @@ unsigned int dc_get_det_buffer_size_from_state(const struct dc_state *context) else return 0; } +/** + *********************************************************************************************** + * dc_get_host_router_index: Get index of host router from a dpia link + * + * This function return a host router index of the target link. If the target link is dpia link. + * + * @param [in] link: target link + * @param [out] host_router_index: host router index of the target link + * + * @return: true if the host router index is found and valid. + * + *********************************************************************************************** + */ +bool dc_get_host_router_index(const struct dc_link *link, unsigned int *host_router_index) +{ + struct dc *dc = link->ctx->dc; + + if (link->ep_type != DISPLAY_ENDPOINT_USB4_DPIA) + return false; + + if (link->link_index < dc->lowest_dpia_link_index) + return false; + + *host_router_index = (link->link_index - dc->lowest_dpia_link_index) / dc->caps.num_of_dpias_per_host_router; + if (*host_router_index < dc->caps.num_of_host_routers) + return true; + else + return false; +} bool dc_is_cursor_limit_pending(struct dc *dc) { diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 1d917be36fc4..f41073c0147e 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -66,7 +66,8 @@ struct dmub_notification; #define MAX_STREAMS 6 #define MIN_VIEWPORT_SIZE 12 #define MAX_NUM_EDP 2 -#define MAX_HOST_ROUTERS_NUM 2 +#define MAX_HOST_ROUTERS_NUM 3 +#define MAX_DPIA_PER_HOST_ROUTER 2 /* Display Core Interfaces */ struct dc_versions { @@ -305,6 +306,8 @@ struct dc_caps { /* Conservative limit for DCC cases which require ODM4:1 to support*/ uint32_t dcc_plane_width_limit; struct dc_scl_caps scl_caps; + uint8_t num_of_host_routers; + uint8_t num_of_dpias_per_host_router; }; struct dc_bug_wa { @@ -1603,6 +1606,7 @@ struct dc { uint8_t link_count; struct dc_link *links[MAX_LINKS]; + uint8_t lowest_dpia_link_index; struct link_service *link_srv; struct dc_state *current_state; @@ -2595,6 +2599,8 @@ struct dc_power_profile dc_get_power_profile_for_dc_state(const struct dc_state unsigned int dc_get_det_buffer_size_from_state(const struct dc_state *context); +bool dc_get_host_router_index(const struct dc_link *link, unsigned int *host_router_index); + /* DSC Interfaces */ #include "dc_dsc.h" diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn31/dcn31_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn31/dcn31_resource.c index 7e0af5297dc4..51ca0b2959fc 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn31/dcn31_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn31/dcn31_resource.c @@ -1954,6 +1954,9 @@ static bool dcn31_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + dc->caps.num_of_host_routers = 2; + dc->caps.num_of_dpias_per_host_router = 2; + /* Use pipe context based otg sync logic */ dc->config.use_pipe_ctx_sync_logic = true; dc->config.disable_hbr_audio_dp2 = true; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn314/dcn314_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn314/dcn314_resource.c index d96bc6cb73ad..8383e2e59be5 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn314/dcn314_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn314/dcn314_resource.c @@ -1885,6 +1885,9 @@ static bool dcn314_resource_construct( dc->caps.max_disp_clock_khz_at_vmin = 650000; + dc->caps.num_of_host_routers = 2; + dc->caps.num_of_dpias_per_host_router = 2; + /* Use pipe context based otg sync logic */ dc->config.use_pipe_ctx_sync_logic = true; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c index 72c6cf047db0..e01aa2f2e13e 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c @@ -1894,6 +1894,9 @@ static bool dcn35_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + dc->caps.num_of_host_routers = 2; + dc->caps.num_of_dpias_per_host_router = 2; + /* max_disp_clock_khz_at_vmin is slightly lower than the STA value in order * to provide some margin. * It's expected for furture ASIC to have equal or higher value, in order to diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c index 989a270f7dea..4ebe4e00a4f8 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c @@ -1866,6 +1866,9 @@ static bool dcn351_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + dc->caps.num_of_host_routers = 2; + dc->caps.num_of_dpias_per_host_router = 2; + /* max_disp_clock_khz_at_vmin is slightly lower than the STA value in order * to provide some margin. * It's expected for furture ASIC to have equal or higher value, in order to diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn36/dcn36_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn36/dcn36_resource.c index 48e1f234185f..db36b8f9ce65 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn36/dcn36_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn36/dcn36_resource.c @@ -1867,6 +1867,9 @@ static bool dcn36_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + dc->caps.num_of_host_routers = 2; + dc->caps.num_of_dpias_per_host_router = 2; + /* max_disp_clock_khz_at_vmin is slightly lower than the STA value in order * to provide some margin. * It's expected for furture ASIC to have equal or higher value, in order to From d358a51444c88bcc995e471dc8cc840f19e4b374 Mon Sep 17 00:00:00 2001 From: Michael Strauss Date: Wed, 26 Feb 2025 10:03:48 -0500 Subject: [PATCH 03/20] drm/amd/display: Get LTTPR IEEE OUI/Device ID From Closest LTTPR To Host [WHY] These fields are read for the explicit purpose of detecting embedded LTTPRs (i.e. between host ASIC and the user-facing port), and thus need to calculate the correct DPCD address offset based on LTTPR count to target the appropriate LTTPR's DPCD register space with these queries. [HOW] Cascaded LTTPRs in a link each snoop and increment LTTPR count when queried via DPCD read, so an LTTPR embedded in a source device (e.g. USB4 port on a laptop) will always be addressible using the max LTTPR count seen by the host. Therefore we simply need to use a recently added helper function to calculate the correct DPCD address to target potentially embedded LTTPRs based on the received LTTPR count. Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Wenjing Liu Signed-off-by: Michael Strauss Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit 791897f5c77a2a65d0e500be4743af2ddf6eb061) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/display/dc/dc_dp_types.h | 4 +- .../dc/link/protocols/link_dp_capability.c | 38 +++++++++++++++---- 2 files changed, 33 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h index 0bad8304ccf6..d346f8ae1634 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -1172,8 +1172,8 @@ struct dc_lttpr_caps { union dp_128b_132b_supported_lttpr_link_rates supported_128b_132b_rates; union dp_alpm_lttpr_cap alpm; uint8_t aux_rd_interval[MAX_REPEATER_CNT - 1]; - uint8_t lttpr_ieee_oui[3]; - uint8_t lttpr_device_id[6]; + uint8_t lttpr_ieee_oui[3]; // Always read from closest LTTPR to host + uint8_t lttpr_device_id[6]; // Always read from closest LTTPR to host }; struct dc_dongle_dfp_cap_ext { diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c index a5127c2d47ef..0f965380a9b4 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c @@ -385,9 +385,15 @@ bool dp_is_128b_132b_signal(struct pipe_ctx *pipe_ctx) bool dp_is_lttpr_present(struct dc_link *link) { /* Some sink devices report invalid LTTPR revision, so don't validate against that cap */ - return (dp_parse_lttpr_repeater_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt) != 0 && + uint32_t lttpr_count = dp_parse_lttpr_repeater_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt); + bool is_lttpr_present = (lttpr_count > 0 && link->dpcd_caps.lttpr_caps.max_lane_count > 0 && link->dpcd_caps.lttpr_caps.max_lane_count <= 4); + + if (lttpr_count > 0 && !is_lttpr_present) + DC_LOG_ERROR("LTTPR count is nonzero but invalid lane count reported. Assuming no LTTPR present.\n"); + + return is_lttpr_present; } /* in DP compliance test, DPR-120 may have @@ -1551,6 +1557,8 @@ enum dc_status dp_retrieve_lttpr_cap(struct dc_link *link) uint8_t lttpr_dpcd_data[10] = {0}; enum dc_status status; bool is_lttpr_present; + uint32_t lttpr_count; + uint32_t closest_lttpr_offset; /* Logic to determine LTTPR support*/ bool vbios_lttpr_interop = link->dc->caps.vbios_lttpr_aware; @@ -1602,20 +1610,22 @@ enum dc_status dp_retrieve_lttpr_cap(struct dc_link *link) lttpr_dpcd_data[DP_LTTPR_ALPM_CAPABILITIES - DP_LT_TUNABLE_PHY_REPEATER_FIELD_DATA_STRUCTURE_REV]; + lttpr_count = dp_parse_lttpr_repeater_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt); + /* If this chip cap is set, at least one retimer must exist in the chain * Override count to 1 if we receive a known bad count (0 or an invalid value) */ if (((link->chip_caps & AMD_EXT_DISPLAY_PATH_CAPS__EXT_CHIP_MASK) == AMD_EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN) && - (dp_parse_lttpr_repeater_count(link->dpcd_caps.lttpr_caps.phy_repeater_cnt) == 0)) { + lttpr_count == 0) { /* If you see this message consistently, either the host platform has FIXED_VS flag * incorrectly configured or the sink device is returning an invalid count. */ DC_LOG_ERROR("lttpr_caps phy_repeater_cnt is 0x%x, forcing it to 0x80.", link->dpcd_caps.lttpr_caps.phy_repeater_cnt); link->dpcd_caps.lttpr_caps.phy_repeater_cnt = 0x80; + lttpr_count = 1; DC_LOG_DC("lttpr_caps forced phy_repeater_cnt = %d\n", link->dpcd_caps.lttpr_caps.phy_repeater_cnt); } - /* Attempt to train in LTTPR transparent mode if repeater count exceeds 8. */ is_lttpr_present = dp_is_lttpr_present(link); DC_LOG_DC("is_lttpr_present = %d\n", is_lttpr_present); @@ -1623,11 +1633,25 @@ enum dc_status dp_retrieve_lttpr_cap(struct dc_link *link) if (is_lttpr_present) { CONN_DATA_DETECT(link, lttpr_dpcd_data, sizeof(lttpr_dpcd_data), "LTTPR Caps: "); - core_link_read_dpcd(link, DP_LTTPR_IEEE_OUI, link->dpcd_caps.lttpr_caps.lttpr_ieee_oui, sizeof(link->dpcd_caps.lttpr_caps.lttpr_ieee_oui)); - CONN_DATA_DETECT(link, link->dpcd_caps.lttpr_caps.lttpr_ieee_oui, sizeof(link->dpcd_caps.lttpr_caps.lttpr_ieee_oui), "LTTPR IEEE OUI: "); + // Identify closest LTTPR to determine if workarounds required for known embedded LTTPR + closest_lttpr_offset = dp_get_closest_lttpr_offset(lttpr_count); - core_link_read_dpcd(link, DP_LTTPR_DEVICE_ID, link->dpcd_caps.lttpr_caps.lttpr_device_id, sizeof(link->dpcd_caps.lttpr_caps.lttpr_device_id)); - CONN_DATA_DETECT(link, link->dpcd_caps.lttpr_caps.lttpr_device_id, sizeof(link->dpcd_caps.lttpr_caps.lttpr_device_id), "LTTPR Device ID: "); + core_link_read_dpcd(link, (DP_LTTPR_IEEE_OUI + closest_lttpr_offset), + link->dpcd_caps.lttpr_caps.lttpr_ieee_oui, sizeof(link->dpcd_caps.lttpr_caps.lttpr_ieee_oui)); + core_link_read_dpcd(link, (DP_LTTPR_DEVICE_ID + closest_lttpr_offset), + link->dpcd_caps.lttpr_caps.lttpr_device_id, sizeof(link->dpcd_caps.lttpr_caps.lttpr_device_id)); + + if (lttpr_count > 1) { + CONN_DATA_DETECT(link, link->dpcd_caps.lttpr_caps.lttpr_ieee_oui, sizeof(link->dpcd_caps.lttpr_caps.lttpr_ieee_oui), + "Closest LTTPR To Host's IEEE OUI: "); + CONN_DATA_DETECT(link, link->dpcd_caps.lttpr_caps.lttpr_device_id, sizeof(link->dpcd_caps.lttpr_caps.lttpr_device_id), + "Closest LTTPR To Host's LTTPR Device ID: "); + } else { + CONN_DATA_DETECT(link, link->dpcd_caps.lttpr_caps.lttpr_ieee_oui, sizeof(link->dpcd_caps.lttpr_caps.lttpr_ieee_oui), + "LTTPR IEEE OUI: "); + CONN_DATA_DETECT(link, link->dpcd_caps.lttpr_caps.lttpr_device_id, sizeof(link->dpcd_caps.lttpr_caps.lttpr_device_id), + "LTTPR Device ID: "); + } } return status; From 0d57dd1765d311111d9885346108c4deeae1deb4 Mon Sep 17 00:00:00 2001 From: Nicholas Kazlauskas Date: Wed, 21 May 2025 16:40:25 -0400 Subject: [PATCH 04/20] drm/amd/display: Add more checks for DSC / HUBP ONO guarantees [WHY] For non-zero DSC instances it's possible that the HUBP domain required to drive it for sequential ONO ASICs isn't met, potentially causing the logic to the tile to enter an undefined state leading to a system hang. [HOW] Add more checks to ensure that the HUBP domain matching the DSC instance is appropriately powered. Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Duncan Ma Signed-off-by: Nicholas Kazlauskas Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit da63df07112e5a9857a8d2aaa04255c4206754ec) Cc: stable@vger.kernel.org --- .../amd/display/dc/hwss/dcn35/dcn35_hwseq.c | 28 +++++++++++++++++++ 1 file changed, 28 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c index c814d957305a..a267f574b619 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c @@ -1047,6 +1047,15 @@ void dcn35_calc_blocks_to_gate(struct dc *dc, struct dc_state *context, if (dc->caps.sequential_ono) { update_state->pg_pipe_res_update[PG_HUBP][pipe_ctx->stream_res.dsc->inst] = false; update_state->pg_pipe_res_update[PG_DPP][pipe_ctx->stream_res.dsc->inst] = false; + + /* All HUBP/DPP instances must be powered if the DSC inst != HUBP inst */ + if (!pipe_ctx->top_pipe && pipe_ctx->plane_res.hubp && + pipe_ctx->plane_res.hubp->inst != pipe_ctx->stream_res.dsc->inst) { + for (j = 0; j < dc->res_pool->pipe_count; ++j) { + update_state->pg_pipe_res_update[PG_HUBP][j] = false; + update_state->pg_pipe_res_update[PG_DPP][j] = false; + } + } } } @@ -1193,6 +1202,25 @@ void dcn35_calc_blocks_to_ungate(struct dc *dc, struct dc_state *context, update_state->pg_pipe_res_update[PG_HDMISTREAM][0] = true; if (dc->caps.sequential_ono) { + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *new_pipe = &context->res_ctx.pipe_ctx[i]; + + if (new_pipe->stream_res.dsc && !new_pipe->top_pipe && + update_state->pg_pipe_res_update[PG_DSC][new_pipe->stream_res.dsc->inst]) { + update_state->pg_pipe_res_update[PG_HUBP][new_pipe->stream_res.dsc->inst] = true; + update_state->pg_pipe_res_update[PG_DPP][new_pipe->stream_res.dsc->inst] = true; + + /* All HUBP/DPP instances must be powered if the DSC inst != HUBP inst */ + if (new_pipe->plane_res.hubp && + new_pipe->plane_res.hubp->inst != new_pipe->stream_res.dsc->inst) { + for (j = 0; j < dc->res_pool->pipe_count; ++j) { + update_state->pg_pipe_res_update[PG_HUBP][j] = true; + update_state->pg_pipe_res_update[PG_DPP][j] = true; + } + } + } + } + for (i = dc->res_pool->pipe_count - 1; i >= 0; i--) { if (update_state->pg_pipe_res_update[PG_HUBP][i] && update_state->pg_pipe_res_update[PG_DPP][i]) { From 8724a5380c4390eed81e271d22f34ff06453ded9 Mon Sep 17 00:00:00 2001 From: Alex Hung Date: Thu, 29 May 2025 10:59:19 -0600 Subject: [PATCH 05/20] drm/amd/display: Fix mpv playback corruption on weston [WHAT] Severe video playback corruption is observed in the following setup: weston 14.0.90 (built from source) + mpv v0.40.0 with command: mpv bbb_sunflower_1080p_60fps_normal.mp4 --vo=gpu [HOW] ABGR16161616 needs to be included in dml2/2.1 translation. Cc: Mario Limonciello Cc: Alex Deucher Acked-by: Aurabindo Pillai Reviewed-by: Harry Wentland Reviewed-by: Austin Zheng Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit d023de809f85307ca819a9dbbceee6ae1f50e2ad) Cc: stable@vger.kernel.org --- .../gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c | 1 + drivers/gpu/drm/amd/display/dc/dml2/dml2_translation_helper.c | 1 + 2 files changed, 2 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c index d47cacfdb695..2aa6d44bb359 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c @@ -788,6 +788,7 @@ static void populate_dml21_plane_config_from_plane_state(struct dml2_context *dm plane->pixel_format = dml2_420_10; break; case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616: + case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616: case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616F: case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F: plane->pixel_format = dml2_444_64; diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml2_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2/dml2_translation_helper.c index 5de775fd8fce..208630754c8a 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2/dml2_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml2_translation_helper.c @@ -953,6 +953,7 @@ static void populate_dml_surface_cfg_from_plane_state(enum dml_project_id dml2_p out->SourcePixelFormat[location] = dml_420_10; break; case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616: + case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616: case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616F: case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F: out->SourcePixelFormat[location] = dml_444_64; From 158f9944ac05dafd2d3a23d0688e6cf40ef68b90 Mon Sep 17 00:00:00 2001 From: Yihan Zhu Date: Tue, 27 May 2025 16:47:40 -0400 Subject: [PATCH 06/20] drm/amd/display: Fix RMCM programming seq errors [WHY & HOW] Fix RMCM programming sequence errors and mapping issues to pass the RMCM test. Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Dmytro Laktyushkin Signed-off-by: Yihan Zhu Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit 11baa4975025033547f45f5894087a0dda6efbb8) Cc: stable@vger.kernel.org --- .../dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c index c4dad7164d31..5b62cd19d979 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c @@ -4685,7 +4685,10 @@ static void calculate_tdlut_setting( //the tdlut is fetched during the 2 row times of prefetch. if (p->setup_for_tdlut) { *p->tdlut_groups_per_2row_ub = (unsigned int)math_ceil2((double) *p->tdlut_bytes_per_frame / *p->tdlut_bytes_per_group, 1); - *p->tdlut_opt_time = (*p->tdlut_bytes_per_frame - p->cursor_buffer_size * 1024) / tdlut_drain_rate; + if (*p->tdlut_bytes_per_frame > p->cursor_buffer_size * 1024) + *p->tdlut_opt_time = (*p->tdlut_bytes_per_frame - p->cursor_buffer_size * 1024) / tdlut_drain_rate; + else + *p->tdlut_opt_time = 0; *p->tdlut_drain_time = p->cursor_buffer_size * 1024 / tdlut_drain_rate; *p->tdlut_bytes_to_deliver = (unsigned int) (p->cursor_buffer_size * 1024.0); } From ffcaed1d7ecef31198000dfbbea791f30f7ca437 Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Thu, 29 May 2025 11:33:44 -0500 Subject: [PATCH 07/20] drm/amd/display: Only read ACPI backlight caps once [WHY] Backlight caps are read already in amdgpu_dm_update_backlight_caps(). They may be updated by update_connector_ext_caps(). Reading again when registering backlight device may cause wrong values to be used. [HOW] Use backlight caps already registered to the dm. Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Roman Li Signed-off-by: Mario Limonciello Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit 148144f6d2f14b02eaaa39b86bbe023cbff350bd) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index d3100f641ac6..0b5d5ab14a69 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -4908,7 +4908,7 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector) struct drm_device *drm = aconnector->base.dev; struct amdgpu_display_manager *dm = &drm_to_adev(drm)->dm; struct backlight_properties props = { 0 }; - struct amdgpu_dm_backlight_caps caps = { 0 }; + struct amdgpu_dm_backlight_caps *caps; char bl_name[16]; int min, max; @@ -4922,20 +4922,20 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector) return; } - amdgpu_acpi_get_backlight_caps(&caps); - if (caps.caps_valid && get_brightness_range(&caps, &min, &max)) { + caps = &dm->backlight_caps[aconnector->bl_idx]; + if (get_brightness_range(caps, &min, &max)) { if (power_supply_is_system_supplied() > 0) - props.brightness = (max - min) * DIV_ROUND_CLOSEST(caps.ac_level, 100); + props.brightness = (max - min) * DIV_ROUND_CLOSEST(caps->ac_level, 100); else - props.brightness = (max - min) * DIV_ROUND_CLOSEST(caps.dc_level, 100); + props.brightness = (max - min) * DIV_ROUND_CLOSEST(caps->dc_level, 100); /* min is zero, so max needs to be adjusted */ props.max_brightness = max - min; drm_dbg(drm, "Backlight caps: min: %d, max: %d, ac %d, dc %d\n", min, max, - caps.ac_level, caps.dc_level); + caps->ac_level, caps->dc_level); } else props.brightness = AMDGPU_MAX_BL_LEVEL; - if (caps.data_points && !(amdgpu_dc_debug_mask & DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE)) + if (caps->data_points && !(amdgpu_dc_debug_mask & DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE)) drm_info(drm, "Using custom brightness curve\n"); props.max_brightness = AMDGPU_MAX_BL_LEVEL; props.type = BACKLIGHT_RAW; From 16dc8bc27c2aa3c93905d3e885e27f1e3535f09a Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Thu, 29 May 2025 09:46:32 -0500 Subject: [PATCH 08/20] drm/amd/display: Export full brightness range to userspace [WHY] Userspace currently is offered a range from 0-0xFF but the PWM is programmed from 0-0xFFFF. This can be limiting to some software that wants to apply greater granularity. [HOW] Convert internally to firmware values only when mapping custom brightness curves because these are in 0-0xFF range. Advertise full PWM range to userspace. Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Roman Li Signed-off-by: Mario Limonciello Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher (cherry picked from commit 8dbd72cb790058ce52279af38a43c2b302fdd3e5) Cc: stable@vger.kernel.org --- .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 43 ++++++++++++------- 1 file changed, 28 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 0b5d5ab14a69..bc4cd11bfc79 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -4718,9 +4718,23 @@ static int get_brightness_range(const struct amdgpu_dm_backlight_caps *caps, return 1; } -static void convert_custom_brightness(const struct amdgpu_dm_backlight_caps *caps, - uint32_t *brightness) +/* Rescale from [min..max] to [0..AMDGPU_MAX_BL_LEVEL] */ +static inline u32 scale_input_to_fw(int min, int max, u64 input) { + return DIV_ROUND_CLOSEST_ULL(input * AMDGPU_MAX_BL_LEVEL, max - min); +} + +/* Rescale from [0..AMDGPU_MAX_BL_LEVEL] to [min..max] */ +static inline u32 scale_fw_to_input(int min, int max, u64 input) +{ + return min + DIV_ROUND_CLOSEST_ULL(input * (max - min), AMDGPU_MAX_BL_LEVEL); +} + +static void convert_custom_brightness(const struct amdgpu_dm_backlight_caps *caps, + unsigned int min, unsigned int max, + uint32_t *user_brightness) +{ + u32 brightness = scale_input_to_fw(min, max, *user_brightness); u8 prev_signal = 0, prev_lum = 0; int i = 0; @@ -4731,7 +4745,7 @@ static void convert_custom_brightness(const struct amdgpu_dm_backlight_caps *cap return; /* choose start to run less interpolation steps */ - if (caps->luminance_data[caps->data_points/2].input_signal > *brightness) + if (caps->luminance_data[caps->data_points/2].input_signal > brightness) i = caps->data_points/2; do { u8 signal = caps->luminance_data[i].input_signal; @@ -4742,17 +4756,18 @@ static void convert_custom_brightness(const struct amdgpu_dm_backlight_caps *cap * brightness < signal: interpolate between previous and current luminance numerator * brightness > signal: find next data point */ - if (*brightness > signal) { + if (brightness > signal) { prev_signal = signal; prev_lum = lum; i++; continue; } - if (*brightness < signal) + if (brightness < signal) lum = prev_lum + DIV_ROUND_CLOSEST((lum - prev_lum) * - (*brightness - prev_signal), + (brightness - prev_signal), signal - prev_signal); - *brightness = DIV_ROUND_CLOSEST(lum * *brightness, 101); + *user_brightness = scale_fw_to_input(min, max, + DIV_ROUND_CLOSEST(lum * brightness, 101)); return; } while (i < caps->data_points); } @@ -4765,11 +4780,10 @@ static u32 convert_brightness_from_user(const struct amdgpu_dm_backlight_caps *c if (!get_brightness_range(caps, &min, &max)) return brightness; - convert_custom_brightness(caps, &brightness); + convert_custom_brightness(caps, min, max, &brightness); - // Rescale 0..255 to min..max - return min + DIV_ROUND_CLOSEST((max - min) * brightness, - AMDGPU_MAX_BL_LEVEL); + // Rescale 0..max to min..max + return min + DIV_ROUND_CLOSEST_ULL((u64)(max - min) * brightness, max); } static u32 convert_brightness_to_user(const struct amdgpu_dm_backlight_caps *caps, @@ -4782,8 +4796,8 @@ static u32 convert_brightness_to_user(const struct amdgpu_dm_backlight_caps *cap if (brightness < min) return 0; - // Rescale min..max to 0..255 - return DIV_ROUND_CLOSEST(AMDGPU_MAX_BL_LEVEL * (brightness - min), + // Rescale min..max to 0..max + return DIV_ROUND_CLOSEST_ULL((u64)max * (brightness - min), max - min); } @@ -4933,11 +4947,10 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector) drm_dbg(drm, "Backlight caps: min: %d, max: %d, ac %d, dc %d\n", min, max, caps->ac_level, caps->dc_level); } else - props.brightness = AMDGPU_MAX_BL_LEVEL; + props.brightness = props.max_brightness = AMDGPU_MAX_BL_LEVEL; if (caps->data_points && !(amdgpu_dc_debug_mask & DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE)) drm_info(drm, "Using custom brightness curve\n"); - props.max_brightness = AMDGPU_MAX_BL_LEVEL; props.type = BACKLIGHT_RAW; snprintf(bl_name, sizeof(bl_name), "amdgpu_bl%d", From 0bbf5fd86c585d437b75003f11365b324360a5d6 Mon Sep 17 00:00:00 2001 From: Frank Min Date: Wed, 4 Jun 2025 21:00:44 +0800 Subject: [PATCH 09/20] drm/amdgpu: Add kicker device detection 1. add kicker device list 2. add kicker device checking helper function Signed-off-by: Frank Min Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher (cherry picked from commit 09aa2b408f4ab689c3541d22b0968de0392ee406) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 17 +++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 6 ++++++ 2 files changed, 23 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 2505c46a9c3d..eaddc441c51a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -30,6 +30,10 @@ #define AMDGPU_UCODE_NAME_MAX (128) +static const struct kicker_device kicker_device_list[] = { + {0x744B, 0x00}, +}; + static void amdgpu_ucode_print_common_hdr(const struct common_firmware_header *hdr) { DRM_DEBUG("size_bytes: %u\n", le32_to_cpu(hdr->size_bytes)); @@ -1387,6 +1391,19 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl return NULL; } +bool amdgpu_is_kicker_fw(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(kicker_device_list); i++) { + if (adev->pdev->device == kicker_device_list[i].device && + adev->pdev->revision == kicker_device_list[i].revision) + return true; + } + + return false; +} + void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len) { int maj, min, rev; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index 9e89c3487be5..6349aad6da35 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -605,6 +605,11 @@ struct amdgpu_firmware { uint32_t pldm_version; }; +struct kicker_device{ + unsigned short device; + u8 revision; +}; + void amdgpu_ucode_print_mc_hdr(const struct common_firmware_header *hdr); void amdgpu_ucode_print_smc_hdr(const struct common_firmware_header *hdr); void amdgpu_ucode_print_imu_hdr(const struct common_firmware_header *hdr); @@ -632,5 +637,6 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type); const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id); void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len); +bool amdgpu_is_kicker_fw(struct amdgpu_device *adev); #endif From 854171405e7f093532b33d8ed0875b9e34fc55b4 Mon Sep 17 00:00:00 2001 From: Frank Min Date: Wed, 4 Jun 2025 21:17:05 +0800 Subject: [PATCH 10/20] drm/amdgpu: add kicker fws loading for gfx11/smu13/psp13 1. Add kicker firmwares loading for gfx11/smu13/psp13 2. Register additional MODULE_FIRMWARE entries for kicker fws - gc_11_0_0_rlc_kicker.bin - gc_11_0_0_imu_kicker.bin - psp_13_0_0_sos_kicker.bin - psp_13_0_0_ta_kicker.bin - smu_13_0_0_kicker.bin Signed-off-by: Frank Min Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher (cherry picked from commit fb5ec2174d70a8989bc207d257db90ffeca3b163) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 16 ++++++++++++---- drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 5 +++++ drivers/gpu/drm/amd/amdgpu/imu_v11_0.c | 9 +++++++-- drivers/gpu/drm/amd/amdgpu/psp_v13_0.c | 2 ++ drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c | 12 +++++++++--- 5 files changed, 35 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index e6f0b035e20b..c14f63cefe67 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -3522,8 +3522,12 @@ int psp_init_sos_microcode(struct psp_context *psp, const char *chip_name) uint8_t *ucode_array_start_addr; int err = 0; - err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED, - "amdgpu/%s_sos.bin", chip_name); + if (amdgpu_is_kicker_fw(adev)) + err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_sos_kicker.bin", chip_name); + else + err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_sos.bin", chip_name); if (err) goto out; @@ -3799,8 +3803,12 @@ int psp_init_ta_microcode(struct psp_context *psp, const char *chip_name) struct amdgpu_device *adev = psp->adev; int err; - err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED, - "amdgpu/%s_ta.bin", chip_name); + if (amdgpu_is_kicker_fw(adev)) + err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_ta_kicker.bin", chip_name); + else + err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_ta.bin", chip_name); if (err) return err; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index afd6d59164bf..ec9b84f92d46 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -85,6 +85,7 @@ MODULE_FIRMWARE("amdgpu/gc_11_0_0_pfp.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_me.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_mec.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_kicker.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_1.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_toc.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_1_pfp.bin"); @@ -759,6 +760,10 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, AMDGPU_UCODE_REQUIRED, "amdgpu/gc_11_0_0_rlc_1.bin"); + else if (amdgpu_is_kicker_fw(adev)) + err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_rlc_kicker.bin", ucode_prefix); else err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, AMDGPU_UCODE_REQUIRED, diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c index cfa91d709d49..cc626036ed9c 100644 --- a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c @@ -32,6 +32,7 @@ #include "gc/gc_11_0_0_sh_mask.h" MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu_kicker.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_1_imu.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_2_imu.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_3_imu.bin"); @@ -51,8 +52,12 @@ static int imu_v11_0_init_microcode(struct amdgpu_device *adev) DRM_DEBUG("\n"); amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); - err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED, - "amdgpu/%s_imu.bin", ucode_prefix); + if (amdgpu_is_kicker_fw(adev)) + err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_imu_kicker.bin", ucode_prefix); + else + err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_imu.bin", ucode_prefix); if (err) goto out; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c index df612fd9cc50..ead616c11705 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c @@ -42,7 +42,9 @@ MODULE_FIRMWARE("amdgpu/psp_13_0_5_ta.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_8_toc.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_8_ta.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_0_sos.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_0_sos_kicker.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_0_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_0_ta_kicker.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_7_sos.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_7_ta.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_10_sos.bin"); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c index a7167668d189..1c7235935d14 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c @@ -58,6 +58,7 @@ MODULE_FIRMWARE("amdgpu/aldebaran_smc.bin"); MODULE_FIRMWARE("amdgpu/smu_13_0_0.bin"); +MODULE_FIRMWARE("amdgpu/smu_13_0_0_kicker.bin"); MODULE_FIRMWARE("amdgpu/smu_13_0_7.bin"); MODULE_FIRMWARE("amdgpu/smu_13_0_10.bin"); @@ -92,7 +93,7 @@ const int pmfw_decoded_link_width[7] = {0, 1, 2, 4, 8, 12, 16}; int smu_v13_0_init_microcode(struct smu_context *smu) { struct amdgpu_device *adev = smu->adev; - char ucode_prefix[15]; + char ucode_prefix[30]; int err = 0; const struct smc_firmware_header_v1_0 *hdr; const struct common_firmware_header *header; @@ -103,8 +104,13 @@ int smu_v13_0_init_microcode(struct smu_context *smu) return 0; amdgpu_ucode_ip_version_decode(adev, MP1_HWIP, ucode_prefix, sizeof(ucode_prefix)); - err = amdgpu_ucode_request(adev, &adev->pm.fw, AMDGPU_UCODE_REQUIRED, - "amdgpu/%s.bin", ucode_prefix); + + if (amdgpu_is_kicker_fw(adev)) + err = amdgpu_ucode_request(adev, &adev->pm.fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_kicker.bin", ucode_prefix); + else + err = amdgpu_ucode_request(adev, &adev->pm.fw, AMDGPU_UCODE_REQUIRED, + "amdgpu/%s.bin", ucode_prefix); if (err) goto out; From 09b585592fa481384597c81388733aed4a04dd05 Mon Sep 17 00:00:00 2001 From: Jesse Zhang Date: Wed, 11 Jun 2025 15:02:09 +0800 Subject: [PATCH 11/20] drm/amdgpu: Fix SDMA engine reset with logical instance ID This commit makes the following improvements to SDMA engine reset handling: 1. Clarifies in the function documentation that instance_id refers to a logical ID 2. Adds conversion from logical to physical instance ID before performing reset using GET_INST(SDMA0, instance_id) macro 3. Improves error messaging to indicate when a logical instance reset fails 4. Adds better code organization with blank lines for readability The change ensures proper SDMA engine reset by using the correct physical instance ID while maintaining the logical ID interface for callers. V2: Remove harvest_config check and convert directly to physical instance (Lijo) Suggested-by: Jonathan Kim Signed-off-by: Jesse Zhang Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher (cherry picked from commit 5efa6217c239ed1ceec0f0414f9b6f6927387dfc) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c index 6716ac281c49..9b54a1ece447 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c @@ -540,8 +540,10 @@ static int amdgpu_sdma_soft_reset(struct amdgpu_device *adev, u32 instance_id) case IP_VERSION(4, 4, 2): case IP_VERSION(4, 4, 4): case IP_VERSION(4, 4, 5): - /* For SDMA 4.x, use the existing DPM interface for backward compatibility */ - r = amdgpu_dpm_reset_sdma(adev, 1 << instance_id); + /* For SDMA 4.x, use the existing DPM interface for backward compatibility, + * we need to convert the logical instance ID to physical instance ID before reset. + */ + r = amdgpu_dpm_reset_sdma(adev, 1 << GET_INST(SDMA0, instance_id)); break; case IP_VERSION(5, 0, 0): case IP_VERSION(5, 0, 1): @@ -568,7 +570,7 @@ static int amdgpu_sdma_soft_reset(struct amdgpu_device *adev, u32 instance_id) /** * amdgpu_sdma_reset_engine - Reset a specific SDMA engine * @adev: Pointer to the AMDGPU device - * @instance_id: ID of the SDMA engine instance to reset + * @instance_id: Logical ID of the SDMA engine instance to reset * * Returns: 0 on success, or a negative error code on failure. */ @@ -601,7 +603,7 @@ int amdgpu_sdma_reset_engine(struct amdgpu_device *adev, uint32_t instance_id) /* Perform the SDMA reset for the specified instance */ ret = amdgpu_sdma_soft_reset(adev, instance_id); if (ret) { - dev_err(adev->dev, "Failed to reset SDMA instance %u\n", instance_id); + dev_err(adev->dev, "Failed to reset SDMA logical instance %u\n", instance_id); goto exit; } From caade9d69f2e2b76d2e2d47089736a99135b8772 Mon Sep 17 00:00:00 2001 From: Jesse Zhang Date: Wed, 11 Jun 2025 15:07:11 +0800 Subject: [PATCH 12/20] drm/amdgpu: Use logical instance ID for SDMA v4_4_2 queue operations Simplify SDMA v4_4_2 queue reset and stop operations by: 1. Removing GET_INST(SDMA0) conversion for ring->me 2. Using the logical instance ID (ring->me) directly 3. Maintaining consistent behavior with other SDMA queue operations This change aligns with the existing queue handling logic where ring->me already represents the correct instance identifier. Signed-off-by: Lijo Lazar Signed-off-by: Jesse Zhang Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher (cherry picked from commit 3bab282dfe25dff7a55add432f56898505a6cc6c) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c index 9c169112a5e7..3de125062ee3 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c @@ -1670,7 +1670,7 @@ static bool sdma_v4_4_2_page_ring_is_guilty(struct amdgpu_ring *ring) static int sdma_v4_4_2_reset_queue(struct amdgpu_ring *ring, unsigned int vmid) { struct amdgpu_device *adev = ring->adev; - u32 id = GET_INST(SDMA0, ring->me); + u32 id = ring->me; int r; if (!(adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE)) @@ -1686,7 +1686,7 @@ static int sdma_v4_4_2_reset_queue(struct amdgpu_ring *ring, unsigned int vmid) static int sdma_v4_4_2_stop_queue(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - u32 instance_id = GET_INST(SDMA0, ring->me); + u32 instance_id = ring->me; u32 inst_mask; uint64_t rptr; From 46e15197b513e60786a44107759d6ca293d6288c Mon Sep 17 00:00:00 2001 From: Sonny Jiang Date: Thu, 12 Jun 2025 11:01:08 -0400 Subject: [PATCH 13/20] drm/amdgpu: VCN v5_0_1 to prevent FW checking RB during DPG pause Add a protection to ensure programming are all complete prior VCPU starting. This is a WA for an unintended VCPU running. Signed-off-by: Sonny Jiang Acked-by: Leo Liu Reviewed-by: Ruijing Dong Signed-off-by: Alex Deucher (cherry picked from commit c29521b529fa5e225feaf709d863a636ca0cbbfa) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c index 338cf43c45fe..cdefd7fcb0da 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c @@ -669,6 +669,9 @@ static int vcn_v5_0_1_start_dpg_mode(struct amdgpu_vcn_inst *vinst, if (indirect) amdgpu_vcn_psp_update_sram(adev, inst_idx, AMDGPU_UCODE_ID_VCN0_RAM); + /* resetting ring, fw should not check RB ring */ + fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET; + /* Pause dpg */ vcn_v5_0_1_pause_dpg_mode(vinst, &state); @@ -681,7 +684,7 @@ static int vcn_v5_0_1_start_dpg_mode(struct amdgpu_vcn_inst *vinst, tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE); tmp &= ~(VCN_RB_ENABLE__RB1_EN_MASK); WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp); - fw_shared->sq.queue_mode |= FW_QUEUE_RING_RESET; + WREG32_SOC15(VCN, vcn_inst, regUVD_RB_RPTR, 0); WREG32_SOC15(VCN, vcn_inst, regUVD_RB_WPTR, 0); @@ -692,6 +695,7 @@ static int vcn_v5_0_1_start_dpg_mode(struct amdgpu_vcn_inst *vinst, tmp = RREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE); tmp |= VCN_RB_ENABLE__RB1_EN_MASK; WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp); + /* resetting done, fw can check RB ring */ fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); WREG32_SOC15(VCN, vcn_inst, regVCN_RB1_DB_CTRL, From b669507b637eb6b1aaecf347f193efccc65d756e Mon Sep 17 00:00:00 2001 From: Alex Hung Date: Tue, 3 Jun 2025 18:30:55 -0600 Subject: [PATCH 14/20] drm/amd/display: Check dce_hwseq before dereferencing it [WHAT] hws was checked for null earlier in dce110_blank_stream, indicating hws can be null, and should be checked whenever it is used. Cc: Mario Limonciello Cc: Alex Deucher Reviewed-by: Aurabindo Pillai Signed-off-by: Alex Hung Signed-off-by: Aurabindo Pillai Signed-off-by: Alex Deucher (cherry picked from commit 79db43611ff61280b6de58ce1305e0b2ecf675ad) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c index e8730cc40edb..38e17b1796e1 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c @@ -1225,7 +1225,7 @@ void dce110_blank_stream(struct pipe_ctx *pipe_ctx) return; if (link->local_sink && link->local_sink->sink_signal == SIGNAL_TYPE_EDP) { - if (!link->skip_implict_edp_power_control) + if (!link->skip_implict_edp_power_control && hws) hws->funcs.edp_backlight_control(link, false); link->dc->hwss.set_abm_immediate_disable(pipe_ctx); } From 785c536c31c0bb057252fddedb30d79ae4979f4b Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Fri, 6 Jun 2025 10:29:28 +0530 Subject: [PATCH 15/20] drm/amdgpu: Release reset locks during failures Make sure to release reset domain lock in case of failures. Signed-off-by: Lijo Lazar Signed-off-by: Ce Sun Reviewed-by: Hawking Zhang Fixes: 11bb33766f66 ("drm/amdgpu: refactor amdgpu_device_gpu_recover") Signed-off-by: Alex Deucher (cherry picked from commit 1ab11a82681eb33a66f423216cb063e7f40c6f85) --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 80 +++++++++++++++------- 1 file changed, 55 insertions(+), 25 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e1bab6a96cb6..500bf85b3ee8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -6019,16 +6019,12 @@ static int amdgpu_device_health_check(struct list_head *device_list_handle) return ret; } -static int amdgpu_device_halt_activities(struct amdgpu_device *adev, - struct amdgpu_job *job, - struct amdgpu_reset_context *reset_context, - struct list_head *device_list, - struct amdgpu_hive_info *hive, - bool need_emergency_restart) +static int amdgpu_device_recovery_prepare(struct amdgpu_device *adev, + struct list_head *device_list, + struct amdgpu_hive_info *hive) { - struct list_head *device_list_handle = NULL; struct amdgpu_device *tmp_adev = NULL; - int i, r = 0; + int r; /* * Build list of devices to reset. @@ -6045,26 +6041,54 @@ static int amdgpu_device_halt_activities(struct amdgpu_device *adev, } if (!list_is_first(&adev->reset_list, device_list)) list_rotate_to_front(&adev->reset_list, device_list); - device_list_handle = device_list; } else { list_add_tail(&adev->reset_list, device_list); - device_list_handle = device_list; } if (!amdgpu_sriov_vf(adev) && (!adev->pcie_reset_ctx.occurs_dpc)) { - r = amdgpu_device_health_check(device_list_handle); + r = amdgpu_device_health_check(device_list); if (r) return r; } - /* We need to lock reset domain only once both for XGMI and single device */ - tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device, - reset_list); + return 0; +} + +static void amdgpu_device_recovery_get_reset_lock(struct amdgpu_device *adev, + struct list_head *device_list) +{ + struct amdgpu_device *tmp_adev = NULL; + + if (list_empty(device_list)) + return; + tmp_adev = + list_first_entry(device_list, struct amdgpu_device, reset_list); amdgpu_device_lock_reset_domain(tmp_adev->reset_domain); +} + +static void amdgpu_device_recovery_put_reset_lock(struct amdgpu_device *adev, + struct list_head *device_list) +{ + struct amdgpu_device *tmp_adev = NULL; + + if (list_empty(device_list)) + return; + tmp_adev = + list_first_entry(device_list, struct amdgpu_device, reset_list); + amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain); +} + +static int amdgpu_device_halt_activities( + struct amdgpu_device *adev, struct amdgpu_job *job, + struct amdgpu_reset_context *reset_context, + struct list_head *device_list, struct amdgpu_hive_info *hive, + bool need_emergency_restart) +{ + struct amdgpu_device *tmp_adev = NULL; + int i, r = 0; /* block all schedulers and reset given job's ring */ - list_for_each_entry(tmp_adev, device_list_handle, reset_list) { - + list_for_each_entry(tmp_adev, device_list, reset_list) { amdgpu_device_set_mp1_state(tmp_adev); /* @@ -6252,11 +6276,6 @@ static void amdgpu_device_gpu_resume(struct amdgpu_device *adev, amdgpu_ras_set_error_query_ready(tmp_adev, true); } - - tmp_adev = list_first_entry(device_list, struct amdgpu_device, - reset_list); - amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain); - } @@ -6324,10 +6343,16 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, reset_context->hive = hive; INIT_LIST_HEAD(&device_list); + if (amdgpu_device_recovery_prepare(adev, &device_list, hive)) + goto end_reset; + + /* We need to lock reset domain only once both for XGMI and single device */ + amdgpu_device_recovery_get_reset_lock(adev, &device_list); + r = amdgpu_device_halt_activities(adev, job, reset_context, &device_list, hive, need_emergency_restart); if (r) - goto end_reset; + goto reset_unlock; if (need_emergency_restart) goto skip_sched_resume; @@ -6345,13 +6370,15 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, r = amdgpu_device_asic_reset(adev, &device_list, reset_context); if (r) - goto end_reset; + goto reset_unlock; skip_hw_reset: r = amdgpu_device_sched_resume(&device_list, reset_context, job_signaled); if (r) - goto end_reset; + goto reset_unlock; skip_sched_resume: amdgpu_device_gpu_resume(adev, &device_list, need_emergency_restart); +reset_unlock: + amdgpu_device_recovery_put_reset_lock(adev, &device_list); end_reset: if (hive) { mutex_unlock(&hive->hive_lock); @@ -6763,6 +6790,8 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta memset(&reset_context, 0, sizeof(reset_context)); INIT_LIST_HEAD(&device_list); + amdgpu_device_recovery_prepare(adev, &device_list, hive); + amdgpu_device_recovery_get_reset_lock(adev, &device_list); r = amdgpu_device_halt_activities(adev, NULL, &reset_context, &device_list, hive, false); if (hive) { @@ -6880,8 +6909,8 @@ out: if (hive) { list_for_each_entry(tmp_adev, &device_list, reset_list) amdgpu_device_unset_mp1_state(tmp_adev); - amdgpu_device_unlock_reset_domain(adev->reset_domain); } + amdgpu_device_recovery_put_reset_lock(adev, &device_list); } if (hive) { @@ -6927,6 +6956,7 @@ void amdgpu_pci_resume(struct pci_dev *pdev) amdgpu_device_sched_resume(&device_list, NULL, NULL); amdgpu_device_gpu_resume(adev, &device_list, false); + amdgpu_device_recovery_put_reset_lock(adev, &device_list); adev->pcie_reset_ctx.occurs_dpc = false; if (hive) { From 7f3b16f3f229e37cc3e02e9e4e7106c523b119e9 Mon Sep 17 00:00:00 2001 From: Jesse Zhang Date: Mon, 16 Jun 2025 19:21:41 +0800 Subject: [PATCH 16/20] drm/amdgpu: Fix SDMA UTC_L1 handling during start/stop sequences This commit makes two key fixes to SDMA v4.4.2 handling: 1. disable UTC_L1 in sdma_cntl register when stopping SDMA engines by reading the current value before modifying UTC_L1_ENABLE bit. 2. Ensure UTC_L1_ENABLE is consistently managed by: - Adding the missing register write when enabling UTC_L1 during start - Keeping UTC_L1 enabled by default as per hardware requirements v2: Correct SDMA_CNTL setting (Philip) Suggested-by: Jonathan Kim Signed-off-by: Jesse Zhang Acked-by: Alex Deucher Signed-off-by: Alex Deucher (cherry picked from commit 375bf564654e85a7b1b0657b191645b3edca1bda) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c index 3de125062ee3..cef68df4c663 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c @@ -490,7 +490,7 @@ static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev, { struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES]; u32 doorbell_offset, doorbell; - u32 rb_cntl, ib_cntl; + u32 rb_cntl, ib_cntl, sdma_cntl; int i; for_each_inst(i, inst_mask) { @@ -502,6 +502,9 @@ static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev, ib_cntl = RREG32_SDMA(i, regSDMA_GFX_IB_CNTL); ib_cntl = REG_SET_FIELD(ib_cntl, SDMA_GFX_IB_CNTL, IB_ENABLE, 0); WREG32_SDMA(i, regSDMA_GFX_IB_CNTL, ib_cntl); + sdma_cntl = RREG32_SDMA(i, regSDMA_CNTL); + sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA_CNTL, UTC_L1_ENABLE, 0); + WREG32_SDMA(i, regSDMA_CNTL, sdma_cntl); if (sdma[i]->use_doorbell) { doorbell = RREG32_SDMA(i, regSDMA_GFX_DOORBELL); @@ -995,6 +998,7 @@ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev, /* set utc l1 enable flag always to 1 */ temp = RREG32_SDMA(i, regSDMA_CNTL); temp = REG_SET_FIELD(temp, SDMA_CNTL, UTC_L1_ENABLE, 1); + WREG32_SDMA(i, regSDMA_CNTL, temp); if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) < IP_VERSION(4, 4, 5)) { /* enable context empty interrupt during initialization */ From ebe43542702c3d15d1a1d95e8e13b1b54076f05a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 2 Jun 2025 11:31:52 -0400 Subject: [PATCH 17/20] drm/amdgpu: switch job hw_fence to amdgpu_fence MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use the amdgpu fence container so we can store additional data in the fence. This also fixes the start_time handling for MCBP since we were casting the fence to an amdgpu_fence and it wasn't. Fixes: 3f4c175d62d8 ("drm/amdgpu: MCBP based on DRM scheduler (v9)") Reviewed-by: Christian König Signed-off-by: Alex Deucher (cherry picked from commit bf1cd14f9e2e1fdf981eed273ddd595863f5288c) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c | 30 +++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 12 ++++----- drivers/gpu/drm/amd/amdgpu/amdgpu_job.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 16 +++++++++++ 6 files changed, 32 insertions(+), 32 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index 8e626f50b362..f81608330a3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -1902,7 +1902,7 @@ no_preempt: continue; } job = to_amdgpu_job(s_job); - if (preempted && (&job->hw_fence) == fence) + if (preempted && (&job->hw_fence.base) == fence) /* mark the job as preempted */ job->preemption_status |= AMDGPU_IB_PREEMPTED; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 500bf85b3ee8..78f8755996f0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -6362,7 +6362,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, * * job->base holds a reference to parent fence */ - if (job && dma_fence_is_signaled(&job->hw_fence)) { + if (job && dma_fence_is_signaled(&job->hw_fence.base)) { job_signaled = true; dev_info(adev->dev, "Guilty job already signaled, skipping HW reset"); goto skip_hw_reset; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 8cecf25996ed..5fec808d7f54 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -41,22 +41,6 @@ #include "amdgpu_trace.h" #include "amdgpu_reset.h" -/* - * Fences mark an event in the GPUs pipeline and are used - * for GPU/CPU synchronization. When the fence is written, - * it is expected that all buffers associated with that fence - * are no longer in use by the associated ring on the GPU and - * that the relevant GPU caches have been flushed. - */ - -struct amdgpu_fence { - struct dma_fence base; - - /* RB, DMA, etc. */ - struct amdgpu_ring *ring; - ktime_t start_timestamp; -}; - static struct kmem_cache *amdgpu_fence_slab; int amdgpu_fence_slab_init(void) @@ -151,12 +135,12 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd am_fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_ATOMIC); if (am_fence == NULL) return -ENOMEM; - fence = &am_fence->base; - am_fence->ring = ring; } else { /* take use of job-embedded fence */ - fence = &job->hw_fence; + am_fence = &job->hw_fence; } + fence = &am_fence->base; + am_fence->ring = ring; seq = ++ring->fence_drv.sync_seq; if (job && job->job_run_counter) { @@ -718,7 +702,7 @@ void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring) * it right here or we won't be able to track them in fence_drv * and they will remain unsignaled during sa_bo free. */ - job = container_of(old, struct amdgpu_job, hw_fence); + job = container_of(old, struct amdgpu_job, hw_fence.base); if (!job->base.s_fence && !dma_fence_is_signaled(old)) dma_fence_signal(old); RCU_INIT_POINTER(*ptr, NULL); @@ -780,7 +764,7 @@ static const char *amdgpu_fence_get_timeline_name(struct dma_fence *f) static const char *amdgpu_job_fence_get_timeline_name(struct dma_fence *f) { - struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence); + struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence.base); return (const char *)to_amdgpu_ring(job->base.sched)->name; } @@ -810,7 +794,7 @@ static bool amdgpu_fence_enable_signaling(struct dma_fence *f) */ static bool amdgpu_job_fence_enable_signaling(struct dma_fence *f) { - struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence); + struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence.base); if (!timer_pending(&to_amdgpu_ring(job->base.sched)->fence_drv.fallback_timer)) amdgpu_fence_schedule_fallback(to_amdgpu_ring(job->base.sched)); @@ -845,7 +829,7 @@ static void amdgpu_job_fence_free(struct rcu_head *rcu) struct dma_fence *f = container_of(rcu, struct dma_fence, rcu); /* free job if fence has a parent job */ - kfree(container_of(f, struct amdgpu_job, hw_fence)); + kfree(container_of(f, struct amdgpu_job, hw_fence.base)); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index acb21fc8b3ce..ddb9d3269357 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -272,8 +272,8 @@ void amdgpu_job_free_resources(struct amdgpu_job *job) /* Check if any fences where initialized */ if (job->base.s_fence && job->base.s_fence->finished.ops) f = &job->base.s_fence->finished; - else if (job->hw_fence.ops) - f = &job->hw_fence; + else if (job->hw_fence.base.ops) + f = &job->hw_fence.base; else f = NULL; @@ -290,10 +290,10 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job) amdgpu_sync_free(&job->explicit_sync); /* only put the hw fence if has embedded fence */ - if (!job->hw_fence.ops) + if (!job->hw_fence.base.ops) kfree(job); else - dma_fence_put(&job->hw_fence); + dma_fence_put(&job->hw_fence.base); } void amdgpu_job_set_gang_leader(struct amdgpu_job *job, @@ -322,10 +322,10 @@ void amdgpu_job_free(struct amdgpu_job *job) if (job->gang_submit != &job->base.s_fence->scheduled) dma_fence_put(job->gang_submit); - if (!job->hw_fence.ops) + if (!job->hw_fence.base.ops) kfree(job); else - dma_fence_put(&job->hw_fence); + dma_fence_put(&job->hw_fence.base); } struct dma_fence *amdgpu_job_submit(struct amdgpu_job *job) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h index f2c049129661..931fed8892cc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h @@ -48,7 +48,7 @@ struct amdgpu_job { struct drm_sched_job base; struct amdgpu_vm *vm; struct amdgpu_sync explicit_sync; - struct dma_fence hw_fence; + struct amdgpu_fence hw_fence; struct dma_fence *gang_submit; uint32_t preamble_status; uint32_t preemption_status; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index b95b47110769..e1f25218943a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -127,6 +127,22 @@ struct amdgpu_fence_driver { struct dma_fence **fences; }; +/* + * Fences mark an event in the GPUs pipeline and are used + * for GPU/CPU synchronization. When the fence is written, + * it is expected that all buffers associated with that fence + * are no longer in use by the associated ring on the GPU and + * that the relevant GPU caches have been flushed. + */ + +struct amdgpu_fence { + struct dma_fence base; + + /* RB, DMA, etc. */ + struct amdgpu_ring *ring; + ktime_t start_timestamp; +}; + extern const struct drm_sched_backend_ops amdgpu_sched_ops; void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring); From 49cc5beeabd5b9b6a6660be8ea2e95de30ec0a07 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 16 Jun 2025 15:43:55 -0400 Subject: [PATCH 18/20] drm/amdgpu/sdma5: init engine reset mutex MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Missing the mutex init. Fixes: e56d4bf57fab ("drm/amdgpu/: drm/amdgpu: Register the new sdma function pointers for sdma_v5_0") Reviewed-by: Jesse Zhang Reviewed-by: Christian König Signed-off-by: Alex Deucher (cherry picked from commit 3f4caf092f02f0de169c6122639af481c7edc8f9) --- drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c index 9505ae96fbec..1813c3ed0aa6 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c @@ -1399,6 +1399,7 @@ static int sdma_v5_0_sw_init(struct amdgpu_ip_block *ip_block) return r; for (i = 0; i < adev->sdma.num_instances; i++) { + mutex_init(&adev->sdma.instance[i].engine_reset_mutex); adev->sdma.instance[i].funcs = &sdma_v5_0_sdma_funcs; ring = &adev->sdma.instance[i].ring; ring->ring_obj = NULL; From cfb05257ae168a0496c7637e1d9e3ab8a25cbffe Mon Sep 17 00:00:00 2001 From: Jay Cornwall Date: Wed, 11 Jun 2025 09:52:14 -0500 Subject: [PATCH 19/20] drm/amdkfd: Fix race in GWS queue scheduling q->gws is not updated atomically with qpd->mapped_gws_queue. If a runlist is created between pqm_set_gws and update_queue it will contain a queue which uses GWS in a process with no GWS allocated. This will result in a scheduler hang. Use q->properties.is_gws which is changed while holding the DQM lock. Signed-off-by: Jay Cornwall Reviewed-by: Harish Kasiviswanathan Signed-off-by: Alex Deucher (cherry picked from commit b98370220eb3110e82248e3354e16a489a492cfb) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c index 8fa6489b6f5d..505036968a77 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c @@ -240,7 +240,7 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer, packet->bitfields2.engine_sel = engine_sel__mes_map_queues__compute_vi; - packet->bitfields2.gws_control_queue = q->gws ? 1 : 0; + packet->bitfields2.gws_control_queue = q->properties.is_gws ? 1 : 0; packet->bitfields2.extended_engine_sel = extended_engine_sel__mes_map_queues__legacy_engine_sel; packet->bitfields2.queue_type = From fe79ef3530d3c681ef2d5644a9d1d1a67b21426a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 16 Jun 2025 15:47:13 -0400 Subject: [PATCH 20/20] drm/amdgpu/sdma5.2: init engine reset mutex Missing the mutex init. Fixes: 47454f2dc0bf ("drm/amdgpu: Register the new sdma function pointers for sdma_v5_2") Reviewed-by: Jesse Zhang Signed-off-by: Alex Deucher (cherry picked from commit ea685ff30a51a25dd9be90786933ada49a088f65) --- drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index a6e612b4a892..23f97da62808 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -1318,6 +1318,7 @@ static int sdma_v5_2_sw_init(struct amdgpu_ip_block *ip_block) } for (i = 0; i < adev->sdma.num_instances; i++) { + mutex_init(&adev->sdma.instance[i].engine_reset_mutex); adev->sdma.instance[i].funcs = &sdma_v5_2_sdma_funcs; ring = &adev->sdma.instance[i].ring; ring->ring_obj = NULL;