From d566fc42c0d33fea79bd8896ae6e77df67ae1ac8 Mon Sep 17 00:00:00 2001 From: Iswara Nagulendran Date: Wed, 11 Dec 2024 17:19:42 -0500 Subject: [PATCH 01/16] drm/amd/display: Add VC for VESA Aux Backlight Control [WHY] There is no way to distinguish the static backlight control type being used and the VABC support without the use of a debugger or reading DPCD registers. [HOW] Add Visual Confirm support for VESA Aux-based Backlight Control. Reviewed-by: Harry Vanzylldejong Signed-off-by: Iswara Nagulendran Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 2 + .../drm/amd/display/dc/core/dc_hw_sequencer.c | 38 +++++++++++++++++++ drivers/gpu/drm/amd/display/dc/dc.h | 1 + .../drm/amd/display/dc/hwss/hw_sequencer.h | 6 +-- 4 files changed, 44 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 08d35d9da443..76c36374de67 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -1236,6 +1236,8 @@ static void dc_update_visual_confirm_color(struct dc *dc, struct dc_state *conte get_mclk_switch_visual_confirm_color(pipe_ctx, &(pipe_ctx->visual_confirm_color)); else if (dc->debug.visual_confirm == VISUAL_CONFIRM_FAMS2) get_fams2_visual_confirm_color(dc, context, pipe_ctx, &(pipe_ctx->visual_confirm_color)); + else if (dc->debug.visual_confirm == VISUAL_CONFIRM_VABC) + get_vabc_visual_confirm_color(pipe_ctx, &(pipe_ctx->visual_confirm_color)); } } } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c index 252af83e34a5..6eb9bae3af91 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c @@ -425,6 +425,44 @@ void get_hdr_visual_confirm_color( } } +/* Visual Confirm color definition for VABC */ +void get_vabc_visual_confirm_color( + struct pipe_ctx *pipe_ctx, + struct tg_color *color) +{ + uint32_t color_value = MAX_TG_COLOR_VALUE; + struct dc_link *edp_link = NULL; + + if (pipe_ctx && pipe_ctx->stream && pipe_ctx->stream->link) { + if (pipe_ctx->stream->link->connector_signal == SIGNAL_TYPE_EDP) + edp_link = pipe_ctx->stream->link; + } + + if (edp_link) { + switch (edp_link->backlight_control_type) { + case BACKLIGHT_CONTROL_PWM: + color->color_r_cr = color_value; + color->color_g_y = 0; + color->color_b_cb = 0; + break; + case BACKLIGHT_CONTROL_AMD_AUX: + color->color_r_cr = 0; + color->color_g_y = color_value; + color->color_b_cb = 0; + break; + case BACKLIGHT_CONTROL_VESA_AUX: + color->color_r_cr = 0; + color->color_g_y = 0; + color->color_b_cb = color_value; + break; + } + } else { + color->color_r_cr = 0; + color->color_g_y = 0; + color->color_b_cb = 0; + } +} + void get_subvp_visual_confirm_color( struct pipe_ctx *pipe_ctx, struct tg_color *color) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 2e4200d5145f..5b32c447da98 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -489,6 +489,7 @@ enum visual_confirm { VISUAL_CONFIRM_MCLK_SWITCH = 16, VISUAL_CONFIRM_FAMS2 = 19, VISUAL_CONFIRM_HW_CURSOR = 20, + VISUAL_CONFIRM_VABC = 21, }; enum dc_psr_power_opts { diff --git a/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer.h index a5bb10d7b160..98d85c7ab3fa 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer.h @@ -194,7 +194,6 @@ enum block_sequence_func { DMUB_SUBVP_SAVE_SURF_ADDR, HUBP_WAIT_FOR_DCC_META_PROP, DMUB_FAMS2_GLOBAL_CONTROL_LOCK_FAST, - }; struct block_sequence { @@ -485,11 +484,12 @@ void get_hdr_visual_confirm_color( void get_mpctree_visual_confirm_color( struct pipe_ctx *pipe_ctx, struct tg_color *color); - +void get_vabc_visual_confirm_color( + struct pipe_ctx *pipe_ctx, + struct tg_color *color); void get_subvp_visual_confirm_color( struct pipe_ctx *pipe_ctx, struct tg_color *color); - void get_fams2_visual_confirm_color( struct dc *dc, struct dc_state *context, -- 2.50.1 From 1e36774f1664b758f7094095a91fd9ca2d32cf96 Mon Sep 17 00:00:00 2001 From: Wayne Lin Date: Wed, 25 Sep 2024 14:24:43 +0800 Subject: [PATCH 02/16] drm/amd/display: Extend capability to get multiple ROI CRCs [Why & How] We already extend our dm, dc and dmub to support setting of multiple CRC instances, now extend the capability to return back the ROI/CRC pair result from psp by specifying activated ROI instances. Reviewed-by: HaoPing Liu Signed-off-by: Wayne Lin Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c | 36 +++++++++++++++---- 1 file changed, 29 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c index a13cc6e67006..7c15082c92ed 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c @@ -343,7 +343,11 @@ static void amdgpu_dm_crtc_notify_ta_to_read(struct work_struct *work) struct amdgpu_dm_connector *aconnector; uint8_t phy_inst; struct amdgpu_display_manager *dm; + struct crc_data crc_cpy[MAX_CRC_WINDOW_NUM]; + unsigned long flags; + uint8_t roi_idx = 0; int ret; + int i; crtc_ctx = container_of(work, struct secure_display_crtc_context, notify_ta_work); crtc = crtc_ctx->crtc; @@ -372,18 +376,36 @@ static void amdgpu_dm_crtc_notify_ta_to_read(struct work_struct *work) } mutex_unlock(&crtc->dev->mode_config.mutex); + spin_lock_irqsave(&crtc->dev->event_lock, flags); + memcpy(crc_cpy, crtc_ctx->crc_info.crc, sizeof(struct crc_data) * MAX_CRC_WINDOW_NUM); + spin_unlock_irqrestore(&crtc->dev->event_lock, flags); + /* need lock for multiple crtcs to use the command buffer */ mutex_lock(&psp->securedisplay_context.mutex); - - psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd, - TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC); - - securedisplay_cmd->securedisplay_in_message.send_roi_crc.phy_id = phy_inst; - /* PSP TA is expected to finish data transmission over I2C within current frame, * even there are up to 4 crtcs request to send in this frame. */ - ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC); + if (dm->secure_display_ctx.support_mul_roi) { + psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd, + TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2); + + securedisplay_cmd->securedisplay_in_message.send_roi_crc_v2.phy_id = phy_inst; + + for (i = 0; i < MAX_CRC_WINDOW_NUM; i++) { + if (crc_cpy[i].crc_ready) + roi_idx |= 1 << i; + } + securedisplay_cmd->securedisplay_in_message.send_roi_crc_v2.roi_idx = roi_idx; + + ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2); + } else { + psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd, + TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC); + + securedisplay_cmd->securedisplay_in_message.send_roi_crc.phy_id = phy_inst; + + ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC); + } if (!ret) { if (securedisplay_cmd->status != TA_SECUREDISPLAY_STATUS__SUCCESS) -- 2.50.1 From da968c3ce459442e83b021d73417b9402c8b14c5 Mon Sep 17 00:00:00 2001 From: Martin Leung Date: Mon, 16 Dec 2024 09:08:04 -0500 Subject: [PATCH 03/16] drm/amd/display: Promote DC to 3.2.315 This version brings along the following: - Add Interface to Dump DSC Caps from dm - Add DP required HBlank size calc to link interface - Add 6bpc RGB case for dcn32 output bpp calculations - Add VC for VESA Aux Backlight Control - Add support for setting multiple CRC windows in dc - Clean up SPL code and outdated interfaces in dcn401_clk_mgr - Disable replay and psr while VRR is enabled - Fix PSR-SU not support but still call the amdgpu_dm_psr_enable - Implement Replay Low Hz Visual Confirm - Extend dc_stream_get_crc to support 2nd crc engine - Update power gating logic for DCN35 hw Reviewed-by: Roman Li Signed-off-by: Martin Leung Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 5b32c447da98..5fb15c6315a4 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -55,7 +55,7 @@ struct aux_payload; struct set_config_cmd_payload; struct dmub_notification; -#define DC_VER "3.2.314" +#define DC_VER "3.2.315" #define MAX_SURFACES 4 #define MAX_PLANES 6 -- 2.50.1 From 62bf9fe6fab01a3f7355931fd9472b3d3ef6a3b2 Mon Sep 17 00:00:00 2001 From: yfeng1 Date: Wed, 25 Dec 2024 23:36:09 -0500 Subject: [PATCH 04/16] drm/amdgpu: Fix for MEC SJT FW Load Fail on VF Users might switch to ROCM build does not include MEC SJT FW and driver needs to consider this case.w Signed-off-by: yfeng1 Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) 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 a5bdcaf7a081..2ba185875baa 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c @@ -579,11 +579,16 @@ static int gfx_v9_4_3_init_cp_compute_microcode(struct amdgpu_device *adev, { int err; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, AMDGPU_UCODE_REQUIRED, "amdgpu/%s_sjt_mec.bin", chip_name); - else + + if (err) + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_mec.bin", chip_name); + } else err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, AMDGPU_UCODE_REQUIRED, "amdgpu/%s_mec.bin", chip_name); -- 2.50.1 From 2774ef7625adb5fb9e9265c26a59dca7b8fd171e Mon Sep 17 00:00:00 2001 From: Zhu Lingshan Date: Wed, 11 Dec 2024 11:51:13 +0800 Subject: [PATCH 05/16] drm/amdkfd: wq_release signals dma_fence only when available kfd_process_wq_release() signals eviction fence by dma_fence_signal() which wanrs if dma_fence is NULL. kfd_process->ef is initialized by kfd_process_device_init_vm() through ioctl. That means the fence is NULL for a new created kfd_process, and close a kfd_process right after open it will trigger the warning. This commit conditionally signals the eviction fence in kfd_process_wq_release() only when it is available. [ 503.660882] WARNING: CPU: 0 PID: 9 at drivers/dma-buf/dma-fence.c:467 dma_fence_signal+0x74/0xa0 [ 503.782940] Workqueue: kfd_process_wq kfd_process_wq_release [amdgpu] [ 503.789640] RIP: 0010:dma_fence_signal+0x74/0xa0 [ 503.877620] Call Trace: [ 503.880066] [ 503.882168] ? __warn+0xcd/0x260 [ 503.885407] ? dma_fence_signal+0x74/0xa0 [ 503.889416] ? report_bug+0x288/0x2d0 [ 503.893089] ? handle_bug+0x53/0xa0 [ 503.896587] ? exc_invalid_op+0x14/0x50 [ 503.900424] ? asm_exc_invalid_op+0x16/0x20 [ 503.904616] ? dma_fence_signal+0x74/0xa0 [ 503.908626] kfd_process_wq_release+0x6b/0x370 [amdgpu] [ 503.914081] process_one_work+0x654/0x10a0 [ 503.918186] worker_thread+0x6c3/0xe70 [ 503.921943] ? srso_alias_return_thunk+0x5/0xfbef5 [ 503.926735] ? srso_alias_return_thunk+0x5/0xfbef5 [ 503.931527] ? __kthread_parkme+0x82/0x140 [ 503.935631] ? __pfx_worker_thread+0x10/0x10 [ 503.939904] kthread+0x2a8/0x380 [ 503.943132] ? __pfx_kthread+0x10/0x10 [ 503.946882] ret_from_fork+0x2d/0x70 [ 503.950458] ? __pfx_kthread+0x10/0x10 [ 503.954210] ret_from_fork_asm+0x1a/0x30 [ 503.958142] [ 503.960328] ---[ end trace 0000000000000000 ]--- Fixes: 967d226eaae8 ("dma-buf: add WARN_ON() illegal dma-fence signaling") Signed-off-by: Zhu Lingshan Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 0976b5b0e8e8..083f83c94531 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -1160,7 +1160,8 @@ static void kfd_process_wq_release(struct work_struct *work) */ synchronize_rcu(); ef = rcu_access_pointer(p->ef); - dma_fence_signal(ef); + if (ef) + dma_fence_signal(ef); kfd_process_remove_sysfs(p); -- 2.50.1 From 9095567bc31bd404be54b0616bdb705011ee2cd9 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Fri, 13 Dec 2024 16:46:42 +0530 Subject: [PATCH 06/16] drm/amdgpu: Fix error handling in amdgpu_ras_add_bad_pages MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit It ensures that appropriate error codes are returned when an error condition is detected Fixes the below; drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2849 amdgpu_ras_add_bad_pages() warn: missing error code here? 'amdgpu_umc_pages_in_a_row()' failed. drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2884 amdgpu_ras_add_bad_pages() warn: missing error code here? 'amdgpu_ras_mca2pa()' failed. v2: s/-EIO/-EINVAL, retained the use of -EINVAL from amdgpu_umc_pages_in_a_row & and amdgpu_ras_mca2pa_by_idx, when the RAS context is not initialized or the convert_ras_err_addr function is unavailable. (Thomas) V3: Returning 0 as the absence of eh_data is acceptable. (Tao) Fixes: a8d133e625ce ("drm/amdgpu: parse legacy RAS bad page mixed with new data in various NPS modes") Reported-by: Dan Carpenter Cc: YiPeng Chai Cc: Tao Zhou Cc: Hawking Zhang Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 01c947066a2e..f0924aa3f4e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2832,8 +2832,10 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, mutex_lock(&con->recovery_lock); data = con->eh_data; - if (!data) + if (!data) { + /* Returning 0 as the absence of eh_data is acceptable */ goto free; + } for (i = 0; i < pages; i++) { if (from_rom && @@ -2845,26 +2847,34 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, * one row */ if (amdgpu_umc_pages_in_a_row(adev, &err_data, - bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT)) + bps[i].retired_page << + AMDGPU_GPU_PAGE_SHIFT)) { + ret = -EINVAL; goto free; - else + } else { find_pages_per_pa = true; + } } else { /* unsupported cases */ + ret = -EOPNOTSUPP; goto free; } } } else { if (amdgpu_umc_pages_in_a_row(adev, &err_data, - bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT)) + bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT)) { + ret = -EINVAL; goto free; + } } } else { if (from_rom && !find_pages_per_pa) { if (bps[i].retired_page & UMC_CHANNEL_IDX_V2) { /* bad page in any NPS mode in eeprom */ - if (amdgpu_ras_mca2pa_by_idx(adev, &bps[i], &err_data)) + if (amdgpu_ras_mca2pa_by_idx(adev, &bps[i], &err_data)) { + ret = -EINVAL; goto free; + } } else { /* legacy bad page in eeprom, generated only in * NPS1 mode @@ -2881,6 +2891,7 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, /* non-nps1 mode, old RAS TA * can't support it */ + ret = -EOPNOTSUPP; goto free; } } -- 2.50.1 From 03cc84b102d1a832e8dfc59344346dedcebcdf42 Mon Sep 17 00:00:00 2001 From: Kun Liu Date: Fri, 27 Dec 2024 11:43:22 +0800 Subject: [PATCH 07/16] drm/amd/pm: fix BUG: scheduling while atomic atomic scheduling will be triggered in interrupt handler for AC/DC mode switch as following backtrace. Call Trace: dump_stack_lvl __schedule_bug __schedule schedule schedule_preempt_disabled __mutex_lock smu_cmn_send_smc_msg_with_param smu_v13_0_irq_process amdgpu_irq_dispatch amdgpu_ih_process amdgpu_irq_handler __handle_irq_event_percpu handle_irq_event handle_edge_irq __common_interrupt common_interrupt asm_common_interrupt Reviewed-by: Lijo Lazar Reviewed-by: Kenneth Feng Signed-off-by: Kun Liu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h | 2 ++ drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c | 12 ++++++------ drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 1 + drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 1 + 4 files changed, 10 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h index ae3563d71fa0..356d9422b411 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h @@ -303,5 +303,7 @@ int smu_v13_0_set_wbrf_exclusion_ranges(struct smu_context *smu, int smu_v13_0_get_boot_freq_by_index(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *value); + +void smu_v13_0_interrupt_work(struct smu_context *smu); #endif #endif 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 7bb45ff6d5c8..fbbdfa54f6a2 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 @@ -1321,11 +1321,11 @@ static int smu_v13_0_set_irq_state(struct amdgpu_device *adev, return 0; } -static int smu_v13_0_ack_ac_dc_interrupt(struct smu_context *smu) +void smu_v13_0_interrupt_work(struct smu_context *smu) { - return smu_cmn_send_smc_msg(smu, - SMU_MSG_ReenableAcDcInterrupt, - NULL); + smu_cmn_send_smc_msg(smu, + SMU_MSG_ReenableAcDcInterrupt, + NULL); } #define THM_11_0__SRCID__THM_DIG_THERM_L2H 0 /* ASIC_TEMP > CG_THERMAL_INT.DIG_THERM_INTH */ @@ -1378,12 +1378,12 @@ static int smu_v13_0_irq_process(struct amdgpu_device *adev, switch (ctxid) { case SMU_IH_INTERRUPT_CONTEXT_ID_AC: dev_dbg(adev->dev, "Switched to AC mode!\n"); - smu_v13_0_ack_ac_dc_interrupt(smu); + schedule_work(&smu->interrupt_work); adev->pm.ac_power = true; break; case SMU_IH_INTERRUPT_CONTEXT_ID_DC: dev_dbg(adev->dev, "Switched to DC mode!\n"); - smu_v13_0_ack_ac_dc_interrupt(smu); + schedule_work(&smu->interrupt_work); adev->pm.ac_power = false; break; case SMU_IH_INTERRUPT_CONTEXT_ID_THERMAL_THROTTLING: 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 3aa705aae4c0..2a09b27788e8 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 @@ -3219,6 +3219,7 @@ static const struct pptable_funcs smu_v13_0_0_ppt_funcs = { .is_asic_wbrf_supported = smu_v13_0_0_wbrf_support_check, .enable_uclk_shadow = smu_v13_0_enable_uclk_shadow, .set_wbrf_exclusion_ranges = smu_v13_0_set_wbrf_exclusion_ranges, + .interrupt_work = smu_v13_0_interrupt_work, }; void smu_v13_0_0_set_ppt_funcs(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c index aabb94796005..55ef18517b0f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c @@ -2797,6 +2797,7 @@ static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .is_asic_wbrf_supported = smu_v13_0_7_wbrf_support_check, .enable_uclk_shadow = smu_v13_0_enable_uclk_shadow, .set_wbrf_exclusion_ranges = smu_v13_0_set_wbrf_exclusion_ranges, + .interrupt_work = smu_v13_0_interrupt_work, }; void smu_v13_0_7_set_ppt_funcs(struct smu_context *smu) -- 2.50.1 From 8b248b90450776c2b59c4a372cb9520e25ff010b Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Wed, 6 Nov 2024 07:49:00 +0530 Subject: [PATCH 08/16] drm/amdgpu/gfx10: Add cleaner shader for GFX10.3.0 MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit This commit adds the cleaner shader microcode for GFX10.3.0 GPUs. The cleaner shader is a piece of GPU code that is used to clear or initialize certain GPU resources, such as Local Data Share (LDS), Vector General Purpose Registers (VGPRs), and Scalar General Purpose Registers (SGPRs). Clearing these resources is important for ensuring data isolation between different workloads running on the GPU. Without the cleaner shader, residual data from a previous workload could potentially be accessed by a subsequent workload, leading to data leaks and incorrect computation results. The cleaner shader microcode is represented as an array of 32-bit words (`gfx_10_3_0_cleaner_shader_hex`). This array is the binary representation of the cleaner shader code, which is written in a low-level GPU instruction set. When the cleaner shader feature is enabled, the AMDGPU driver loads this array into a specific location in the GPU memory. The GPU then reads this memory location to fetch and execute the cleaner shader instructions. The cleaner shader is executed automatically by the GPU at the end of each workload, before the next workload starts. This ensures that all GPU resources are in a clean state before the start of each workload. This addition is part of the cleaner shader feature implementation. The cleaner shader feature helps resource utilization by cleaning up GPU resources after they are used. It also enhances security and reliability by preventing data leaks between workloads. Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 15 +++ .../drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h | 56 ++++++++ .../amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm | 124 ++++++++++++++++++ 3 files changed, 195 insertions(+) create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h create mode 100644 drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index 003522c2d902..88393c2c08e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -45,6 +45,7 @@ #include "clearstate_gfx10.h" #include "v10_structs.h" #include "gfx_v10_0.h" +#include "gfx_v10_0_cleaner_shader.h" #include "nbio_v2_3.h" /* @@ -4738,6 +4739,20 @@ static int gfx_v10_0_sw_init(struct amdgpu_ip_block *ip_block) break; } switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(10, 3, 0): + adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex; + adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex); + if (adev->gfx.me_fw_version >= 64 && + adev->gfx.pfp_fw_version >= 100 && + adev->gfx.mec_fw_version >= 122) { + adev->gfx.enable_cleaner_shader = true; + r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size); + if (r) { + adev->gfx.enable_cleaner_shader = false; + dev_err(adev->dev, "Failed to initialize cleaner shader\n"); + } + } + break; default: adev->gfx.enable_cleaner_shader = false; break; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h new file mode 100644 index 000000000000..663c2572d440 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h @@ -0,0 +1,56 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright 2025 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Define the cleaner shader gfx_10_3_0 */ +static const u32 gfx_10_3_0_cleaner_shader_hex[] = { + 0xb0804004, 0xbf8a0000, + 0xbe8203b8, 0xbefc0380, + 0x7e008480, 0x7e028480, + 0x7e048480, 0x7e068480, + 0x7e088480, 0x7e0a8480, + 0x7e0c8480, 0x7e0e8480, + 0xbefc0302, 0x80828802, + 0xbf84fff5, 0xbe8203ff, + 0x80000000, 0x87020002, + 0xbf840012, 0xbefe03c1, + 0xbeff03c1, 0xd7650001, + 0x0001007f, 0xd7660001, + 0x0002027e, 0x16020288, + 0xbe8203bf, 0xbefc03c1, + 0xd9382000, 0x00020201, + 0xd9386040, 0x00040401, + 0xd70f6a01, 0x000202ff, + 0x00000400, 0x80828102, + 0xbf84fff7, 0xbefc03ff, + 0x00000068, 0xbe803080, + 0xbe813080, 0xbe823080, + 0xbe833080, 0x80fc847c, + 0xbf84fffa, 0xbeea0480, + 0xbeec0480, 0xbeee0480, + 0xbef00480, 0xbef20480, + 0xbef40480, 0xbef60480, + 0xbef80480, 0xbefa0480, + 0xbf810000, 0xbf9f0000, + 0xbf9f0000, 0xbf9f0000, + 0xbf9f0000, 0xbf9f0000, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm new file mode 100644 index 000000000000..0e1c246166c0 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm @@ -0,0 +1,124 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright 2025 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +// This shader is to clean LDS, SGPRs and VGPRs. It is first 64 Dwords or 256 bytes of 192 Dwords cleaner shader. +//To turn this shader program on for complitaion change this to main and lower shader main to main_1 + +// GFX10.3 : Clear SGPRs, VGPRs and LDS +// Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot +// Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD +// Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS) +// It takes 2 workgroups to use all of LDS: one on each CU of the WGP +// Each wave clears SGPRs 0 - 107 +// Each wave clears VGPRs 0 - 63 +// The first wave of the workgroup clears its 64KB of LDS +// The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup +// before any wave in the workgroup could end. Without this, it is possible not all SGPRs get cleared. + + +shader main + asic(GFX10) + type(CS) + wave_size(32) +// Note: original source code from SQ team + +// +// Create 32 waves in a threadgroup (CS waves) +// Each allocates 64 VGPRs +// The workgroup allocates all of LDS (64kbytes) +// +// Takes about 2500 clocks to run. +// (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks) +// + S_BARRIER + s_mov_b32 s2, 0x00000038 // Loop 64/8=8 times (loop unrolled for performance) + s_mov_b32 m0, 0 + // + // CLEAR VGPRs + // +label_0005: + v_movreld_b32 v0, 0 + v_movreld_b32 v1, 0 + v_movreld_b32 v2, 0 + v_movreld_b32 v3, 0 + v_movreld_b32 v4, 0 + v_movreld_b32 v5, 0 + v_movreld_b32 v6, 0 + v_movreld_b32 v7, 0 + s_mov_b32 m0, s2 + s_sub_u32 s2, s2, 8 + s_cbranch_scc0 label_0005 + // + s_mov_b32 s2, 0x80000000 // Bit31 is first_wave + s_and_b32 s2, s2, s0 // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set + s_cbranch_scc0 label_0023 // Clean LDS if its first wave of ThreadGroup/WorkGroup + // CLEAR LDS + // + s_mov_b32 exec_lo, 0xffffffff + s_mov_b32 exec_hi, 0xffffffff + v_mbcnt_lo_u32_b32 v1, exec_hi, 0 // Set V1 to thread-ID (0..63) + v_mbcnt_hi_u32_b32 v1, exec_lo, v1 // Set V1 to thread-ID (0..63) + v_mul_u32_u24 v1, 0x00000008, v1 // * 8, so each thread is a double-dword address (8byte) + s_mov_b32 s2, 0x00000003f // 64 loop iterations + s_mov_b32 m0, 0xffffffff + // Clear all of LDS space + // Each FirstWave of WorkGroup clears 64kbyte block + +label_001F: + ds_write2_b64 v1, v[2:3], v[2:3] offset1:32 + ds_write2_b64 v1, v[4:5], v[4:5] offset0:64 offset1:96 + v_add_co_u32 v1, vcc, 0x00000400, v1 + s_sub_u32 s2, s2, 1 + s_cbranch_scc0 label_001F + + // + // CLEAR SGPRs + // +label_0023: + s_mov_b32 m0, 0x00000068 // Loop 108/4=27 times (loop unrolled for performance) +label_sgpr_loop: + s_movreld_b32 s0, 0 + s_movreld_b32 s1, 0 + s_movreld_b32 s2, 0 + s_movreld_b32 s3, 0 + s_sub_u32 m0, m0, 4 + s_cbranch_scc0 label_sgpr_loop + + //clear vcc + s_mov_b32 flat_scratch_lo, 0 //clear flat scratch lo SGPR + s_mov_b32 flat_scratch_hi, 0 //clear flat scratch hi SGPR + s_mov_b64 vcc, 0 //clear vcc + s_mov_b64 ttmp0, 0 //Clear ttmp0 and ttmp1 + s_mov_b64 ttmp2, 0 //Clear ttmp2 and ttmp3 + s_mov_b64 ttmp4, 0 //Clear ttmp4 and ttmp5 + s_mov_b64 ttmp6, 0 //Clear ttmp6 and ttmp7 + s_mov_b64 ttmp8, 0 //Clear ttmp8 and ttmp9 + s_mov_b64 ttmp10, 0 //Clear ttmp10 and ttmp11 + s_mov_b64 ttmp12, 0 //Clear ttmp12 and ttmp13 + s_mov_b64 ttmp14, 0 //Clear ttmp14 and ttmp15 + + s_endpgm + +end + + -- 2.50.1 From a5d258a00b41143d9c64880eed35799d093c4782 Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Thu, 2 Jan 2025 13:12:24 -0500 Subject: [PATCH 09/16] Revert "drm/amd/display: Optimize cursor position updates" This reverts commit 88c7c56d07c108ed4de319c8dba44aa4b8a38dd1. SW and HW state are not always matching in some cases causing cursor to be disabled. Signed-off-by: Aurabindo Pillai Reviewed-by: Leo Li Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c | 7 +++---- .../gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_cm.c | 6 ++---- drivers/gpu/drm/amd/display/dc/hubp/dcn20/dcn20_hubp.c | 8 +++----- .../gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c | 10 ++++------ 4 files changed, 12 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c index 8f6529a98f31..e1da48b05d00 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c @@ -480,11 +480,10 @@ void dpp1_set_cursor_position( if (src_y_offset + cursor_height <= 0) cur_en = 0; /* not visible beyond top edge*/ - if (dpp_base->pos.cur0_ctl.bits.cur0_enable != cur_en) { - REG_UPDATE(CURSOR0_CONTROL, CUR0_ENABLE, cur_en); + REG_UPDATE(CURSOR0_CONTROL, + CUR0_ENABLE, cur_en); - dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en; - } + dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en; } void dpp1_cnv_set_optional_cursor_attributes( diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_cm.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_cm.c index 1236e0f9a256..3b6ca7974e18 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_cm.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_cm.c @@ -154,11 +154,9 @@ void dpp401_set_cursor_position( struct dcn401_dpp *dpp = TO_DCN401_DPP(dpp_base); uint32_t cur_en = pos->enable ? 1 : 0; - if (dpp_base->pos.cur0_ctl.bits.cur0_enable != cur_en) { - REG_UPDATE(CURSOR0_CONTROL, CUR0_ENABLE, cur_en); + REG_UPDATE(CURSOR0_CONTROL, CUR0_ENABLE, cur_en); - dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en; - } + dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en; } void dpp401_set_optional_cursor_attributes( diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn20/dcn20_hubp.c b/drivers/gpu/drm/amd/display/dc/hubp/dcn20/dcn20_hubp.c index d537d0c53cf0..c74f6a3313a2 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn20/dcn20_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn20/dcn20_hubp.c @@ -1058,13 +1058,11 @@ void hubp2_cursor_set_position( if (src_y_offset + cursor_height <= 0) cur_en = 0; /* not visible beyond top edge*/ - if (hubp->pos.cur_ctl.bits.cur_enable != cur_en) { - if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0) - hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr); + if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0) + hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr); - REG_UPDATE(CURSOR_CONTROL, + REG_UPDATE(CURSOR_CONTROL, CURSOR_ENABLE, cur_en); - } REG_SET_2(CURSOR_POSITION, 0, CURSOR_X_POSITION, pos->x, diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c b/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c index 3595c74a3a2f..d38e3f3a1107 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c @@ -730,13 +730,11 @@ void hubp401_cursor_set_position( dc_fixpt_from_int(dst_x_offset), param->h_scale_ratio)); - if (hubp->pos.cur_ctl.bits.cur_enable != cur_en) { - if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0) - hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr); + if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0) + hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr); - REG_UPDATE(CURSOR_CONTROL, - CURSOR_ENABLE, cur_en); - } + REG_UPDATE(CURSOR_CONTROL, + CURSOR_ENABLE, cur_en); REG_SET_2(CURSOR_POSITION, 0, CURSOR_X_POSITION, x_pos, -- 2.50.1 From 2b11179e1878fafcc5c63de726d1b6b8b281e8e9 Mon Sep 17 00:00:00 2001 From: Prike Liang Date: Mon, 16 Dec 2024 16:37:58 +0800 Subject: [PATCH 10/16] drm/amdgpu: reduce RLC safe mode request for gfx clock gating The driver can only request one time for the power safe mode instead of polling and disabling the power feature each time prior to program the GFX clock gating control registers. This update will reduce the latency on the GFX clock gating entry. Signed-off-by: Prike Liang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 12 ++++-------- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 14 ++------------ 2 files changed, 6 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index af73f85527b7..6a025438f9d0 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -5639,8 +5639,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev { uint32_t temp, data; - amdgpu_gfx_rlc_enter_safe_mode(adev, 0); - /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) { @@ -5734,8 +5732,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* 7- wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */ gfx_v8_0_wait_for_rlc_serdes(adev); } - - amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -5745,8 +5741,6 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL); - amdgpu_gfx_rlc_enter_safe_mode(adev, 0); - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { temp1 = data1 = RREG32(mmRLC_CGTT_MGCG_OVERRIDE); data1 &= ~RLC_CGTT_MGCG_OVERRIDE__CGCG_MASK; @@ -5827,12 +5821,12 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev } gfx_v8_0_wait_for_rlc_serdes(adev); - - amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev, bool enable) { + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); + if (enable) { /* CGCG/CGLS should be enabled after MGCG/MGLS/TS(CG/LS) * === MGCG + MGLS + TS(CG/LS) === @@ -5846,6 +5840,8 @@ static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev, gfx_v8_0_update_coarse_grain_clock_gating(adev, enable); gfx_v8_0_update_medium_grain_clock_gating(adev, enable); } + + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 4b5006dc3d34..fa572b40989e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -4964,8 +4964,6 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev { uint32_t data, def; - amdgpu_gfx_rlc_enter_safe_mode(adev, 0); - /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { /* 1 - RLC_CGTT_MGCG_OVERRIDE */ @@ -5030,8 +5028,6 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data); } } - - amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, @@ -5042,8 +5038,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, if (!adev->gfx.num_gfx_rings) return; - amdgpu_gfx_rlc_enter_safe_mode(adev, 0); - /* Enable 3D CGCG/CGLS */ if (enable) { /* write cmd to clear cgcg/cgls ov */ @@ -5085,8 +5079,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data); } - - amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev, @@ -5094,8 +5086,6 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev { uint32_t def, data; - amdgpu_gfx_rlc_enter_safe_mode(adev, 0); - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) { def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); /* unset CGCG override */ @@ -5137,13 +5127,12 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data); } - - amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, bool enable) { + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); if (enable) { /* CGCG/CGLS should be enabled after MGCG/MGLS * === MGCG + MGLS === @@ -5163,6 +5152,7 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev, /* === MGCG + MGLS === */ gfx_v9_0_update_medium_grain_clock_gating(adev, enable); } + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } -- 2.50.1 From c901693f368f99a858bd410576427adc6a518612 Mon Sep 17 00:00:00 2001 From: Zhu Lingshan Date: Mon, 29 Jul 2024 19:28:36 +0800 Subject: [PATCH 11/16] drm/amdkfd: always include uapi header in priv.h The header usr/linux/kfd_ioctl.h is a duplicate of uapi/linux/kfd_ioctl.h. And it is actually not a file in the source code tree. Ideally, the usr version should be updated whenever the source code is recompiled. However, I have noticed a discrepancy between the two headers even after rebuilding the kernel. This commit modifies kfd_priv.h to always include the header from uapi to ensure the latest changes are reflected. We should always include the source code header other than the duplication. Signed-off-by: Zhu Lingshan Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index c32b255c0eb2..d8cd913aa772 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -32,7 +32,7 @@ #include #include #include -#include +#include #include #include #include -- 2.50.1 From 3318ba94e56b9183d0304577c74b33b6b01ce516 Mon Sep 17 00:00:00 2001 From: Arunpravin Paneer Selvam Date: Tue, 10 Dec 2024 12:50:08 +0530 Subject: [PATCH 12/16] drm/amdgpu: Add a lock when accessing the buddy trim function MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit When running YouTube videos and Steam games simultaneously, the tester found a system hang / race condition issue with the multi-display configuration setting. Adding a lock to the buddy allocator's trim function would be the solution. [ 7197.250436] general protection fault, probably for non-canonical address 0xdead000000000108 [ 7197.250447] RIP: 0010:__alloc_range+0x8b/0x340 [amddrm_buddy] [ 7197.250470] Call Trace: [ 7197.250472] [ 7197.250475] ? show_regs+0x6d/0x80 [ 7197.250481] ? die_addr+0x37/0xa0 [ 7197.250483] ? exc_general_protection+0x1db/0x480 [ 7197.250488] ? drm_suballoc_new+0x13c/0x93d [drm_suballoc_helper] [ 7197.250493] ? asm_exc_general_protection+0x27/0x30 [ 7197.250498] ? __alloc_range+0x8b/0x340 [amddrm_buddy] [ 7197.250501] ? __alloc_range+0x109/0x340 [amddrm_buddy] [ 7197.250506] amddrm_buddy_block_trim+0x1b5/0x260 [amddrm_buddy] [ 7197.250511] amdgpu_vram_mgr_new+0x4f5/0x590 [amdgpu] [ 7197.250682] amdttm_resource_alloc+0x46/0xb0 [amdttm] [ 7197.250689] ttm_bo_alloc_resource+0xe4/0x370 [amdttm] [ 7197.250696] amdttm_bo_validate+0x9d/0x180 [amdttm] [ 7197.250701] amdgpu_bo_pin+0x15a/0x2f0 [amdgpu] [ 7197.250831] amdgpu_dm_plane_helper_prepare_fb+0xb2/0x360 [amdgpu] [ 7197.251025] ? try_wait_for_completion+0x59/0x70 [ 7197.251030] drm_atomic_helper_prepare_planes.part.0+0x2f/0x1e0 [ 7197.251035] drm_atomic_helper_prepare_planes+0x5d/0x70 [ 7197.251037] drm_atomic_helper_commit+0x84/0x160 [ 7197.251040] drm_atomic_nonblocking_commit+0x59/0x70 [ 7197.251043] drm_mode_atomic_ioctl+0x720/0x850 [ 7197.251047] ? __pfx_drm_mode_atomic_ioctl+0x10/0x10 [ 7197.251049] drm_ioctl_kernel+0xb9/0x120 [ 7197.251053] ? srso_alias_return_thunk+0x5/0xfbef5 [ 7197.251056] drm_ioctl+0x2d4/0x550 [ 7197.251058] ? __pfx_drm_mode_atomic_ioctl+0x10/0x10 [ 7197.251063] amdgpu_drm_ioctl+0x4e/0x90 [amdgpu] [ 7197.251186] __x64_sys_ioctl+0xa0/0xf0 [ 7197.251190] x64_sys_call+0x143b/0x25c0 [ 7197.251193] do_syscall_64+0x7f/0x180 [ 7197.251197] ? srso_alias_return_thunk+0x5/0xfbef5 [ 7197.251199] ? amdgpu_display_user_framebuffer_create+0x215/0x320 [amdgpu] [ 7197.251329] ? drm_internal_framebuffer_create+0xb7/0x1a0 [ 7197.251332] ? srso_alias_return_thunk+0x5/0xfbef5 Signed-off-by: Arunpravin Paneer Selvam Fixes: 4a5ad08f5377 ("drm/amdgpu: Add address alignment support to DCC buffers") Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index 7d26a962f811..ff5e52025266 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -567,7 +567,6 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, else remaining_size -= size; } - mutex_unlock(&mgr->lock); if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) { struct drm_buddy_block *dcc_block; @@ -584,6 +583,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, (u64)vres->base.size, &vres->blocks); } + mutex_unlock(&mgr->lock); vres->base.start = 0; size = max_t(u64, amdgpu_vram_mgr_blocks_size(&vres->blocks), -- 2.50.1 From 6caf95b771047dac1cf0f520e0e93aa0e84f59ed Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 22 Nov 2024 16:50:56 +0800 Subject: [PATCH 13/16] drm/amd/pm: Update SMUv13.0.6 PMFW headers Update pmfw headers for smuv13.0.6 to pmfw version 85.121 Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h index 0f96b8c59a0e..274b3e1cc4fb 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h @@ -34,6 +34,8 @@ #define NUM_PCIE_BITRATES 4 #define NUM_XGMI_BITRATES 4 #define NUM_XGMI_WIDTHS 3 +#define NUM_SOC_P2S_TABLES 3 +#define NUM_TDP_GROUPS 4 typedef enum { /*0*/ FEATURE_DATA_CALCULATION = 0, @@ -80,8 +82,10 @@ typedef enum { /*41*/ FEATURE_CXL_QOS = 41, /*42*/ FEATURE_SOC_DC_RTC = 42, /*43*/ FEATURE_GFX_DC_RTC = 43, +/*44*/ FEATURE_DVM_MIN_PSM = 44, +/*45*/ FEATURE_PRC = 45, -/*44*/ NUM_FEATURES = 44 +/*46*/ NUM_FEATURES = 46 } FEATURE_LIST_e; //enum for MPIO PCIe gen speed msgs @@ -123,7 +127,7 @@ typedef enum { VOLTAGE_GUARDBAND_COUNT } GFX_GUARDBAND_e; -#define SMU_METRICS_TABLE_VERSION 0xE +#define SMU_METRICS_TABLE_VERSION 0xF typedef struct __attribute__((packed, aligned(4))) { uint32_t AccumulationCounter; @@ -234,6 +238,9 @@ typedef struct __attribute__((packed, aligned(4))) { //PCIE BW Data and error count uint32_t PCIeOtherEndRecoveryAcc; // The Pcie counter itself is accumulated + + //Total App Clock Counter + uint64_t GfxclkBelowHostLimitAcc[8]; } MetricsTableX_t; typedef struct __attribute__((packed, aligned(4))) { @@ -328,13 +335,14 @@ typedef struct __attribute__((packed, aligned(4))) { uint32_t JpegBusy[32]; } MetricsTableA_t; -#define SMU_VF_METRICS_TABLE_VERSION 0x3 +#define SMU_VF_METRICS_TABLE_VERSION 0x5 typedef struct __attribute__((packed, aligned(4))) { uint32_t AccumulationCounter; uint32_t InstGfxclk_TargFreq; uint64_t AccGfxclk_TargFreq; uint64_t AccGfxRsmuDpm_Busy; + uint64_t AccGfxclkBelowHostLimit; } VfMetricsTable_t; #endif -- 2.50.1 From 24a1b66752bb646f2905b6d952a712e61155c9ca Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 22 Nov 2024 18:00:19 +0800 Subject: [PATCH 14/16] drm/amd/pm: Fill max mem bw & total app clk counter Fill max memory bandwidth and total app clock counter to metrics v1_7 v2: Remove unnecessary check v3: Add app clock counter support for apu Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 23 +++++++++++++++++++ 1 file changed, 23 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 5b86df0c8536..650aa9d0548a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -119,6 +119,21 @@ static inline bool smu_v13_0_6_is_other_end_count_available(struct smu_context * } } +static inline bool smu_v13_0_6_is_blw_host_limit_available(struct smu_context *smu) +{ + if (smu->adev->flags & AMD_IS_APU) + return smu->smc_fw_version >= 0x04556F00; + + switch (amdgpu_ip_version(smu->adev, MP1_HWIP, 0)) { + case IP_VERSION(13, 0, 6): + return smu->smc_fw_version >= 0x557900; + case IP_VERSION(13, 0, 14): + return smu->smc_fw_version >= 0x05551000; + default: + return false; + } +} + struct mca_bank_ipid { enum amdgpu_mca_ip ip; uint16_t hwid; @@ -2358,6 +2373,9 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table gpu_metrics->average_umc_activity = SMUQ10_ROUND(GET_METRIC_FIELD(DramBandwidthUtilization, flag)); + gpu_metrics->mem_max_bandwidth = + SMUQ10_ROUND(GET_METRIC_FIELD(MaxDramBandwidth, flag)); + gpu_metrics->curr_socket_power = SMUQ10_ROUND(GET_METRIC_FIELD(SocketPower, flag)); /* Energy counter reported in 15.259uJ (2^-16) units */ @@ -2496,6 +2514,11 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table SMUQ10_ROUND(metrics_x->GfxBusy[inst]); gpu_metrics->xcp_stats[i].gfx_busy_acc[idx] = SMUQ10_ROUND(metrics_x->GfxBusyAcc[inst]); + + if (smu_v13_0_6_is_blw_host_limit_available(smu)) + gpu_metrics->xcp_stats[i].gfx_below_host_limit_acc[idx] = + SMUQ10_ROUND(metrics_x->GfxclkBelowHostLimitAcc + [inst]); idx++; } } -- 2.50.1 From 6c9c97387be1bf4a5881fe1a6cd00acd4322f283 Mon Sep 17 00:00:00 2001 From: Kent Russell Date: Thu, 2 Jan 2025 09:21:46 -0500 Subject: [PATCH 15/16] drm/amdgpu: Remove unnecessary NULL check container_of cannot return NULL, so it is unnecessary to check for NULL after gem_to_amdgpu_bo, which is just a container_of call Signed-off-by: Kent Russell Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 1a5df8b94661..8dba7425501c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -87,10 +87,8 @@ static void amdgpu_gem_object_free(struct drm_gem_object *gobj) { struct amdgpu_bo *aobj = gem_to_amdgpu_bo(gobj); - if (aobj) { - amdgpu_hmm_unregister(aobj); - ttm_bo_put(&aobj->tbo); - } + amdgpu_hmm_unregister(aobj); + ttm_bo_put(&aobj->tbo); } int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, -- 2.50.1 From 39b0fa29f60db2a8b1563bcd147e46548e54292c Mon Sep 17 00:00:00 2001 From: Jesse Zhang Date: Fri, 13 Dec 2024 19:19:45 +0800 Subject: [PATCH 16/16] drm/amdgpu/sdma4.4.2: add apu support in sdma queue reset Remove apu check in sdma queue reset. Signed-off-by: Jesse Zhang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 +- 1 file changed, 1 insertion(+), 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 56507ae919b0..48537eba225d 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c @@ -1602,7 +1602,7 @@ static int sdma_v4_4_2_reset_queue(struct amdgpu_ring *ring, unsigned int vmid) int i, r; u32 inst_mask; - if ((adev->flags & AMD_IS_APU) || amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) return -EINVAL; /* stop queue */ -- 2.50.1