From 79b38858193474d4c35eb17cc238ea8f230f1671 Mon Sep 17 00:00:00 2001 From: "Jesse.Zhang" Date: Fri, 13 Mar 2026 14:17:10 +0800 Subject: [PATCH 01/99] drm/amdgpu: replace WARN with DRM_ERROR for invalid sched priority MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit amdgpu_sched_ioctl() currently uses WARN(1, ...) when userspace passes an out-of-range context priority value. WARN(1, ...) is unconditional and produces a full stack trace, which is disproportionate for a simple input validation failure -- the invalid value is already rejected with -EINVAL on the next line. Replace WARN(1, ...) with DRM_ERROR() to log the invalid value at an appropriate level without generating a stack dump. The -EINVAL return to userspace is unchanged. No functional change for well-formed userspace callers. v2: - Reworked commit message to focus on appropriate log level for parameter validation - Clarified that -EINVAL behavior is preserved (Vitaly) v3: completely drop that warning. Invalid parameters should never clutter the system log. (Christian) Reviewed-by: Vitaly Prosyak Reviewed-by: Christian König Signed-off-by: Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c index 341beec59537..0eecfaa3a94c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c @@ -103,10 +103,8 @@ int amdgpu_sched_ioctl(struct drm_device *dev, void *data, return -EINVAL; } - if (!amdgpu_ctx_priority_is_valid(args->in.priority)) { - WARN(1, "Invalid context priority %d\n", args->in.priority); + if (!amdgpu_ctx_priority_is_valid(args->in.priority)) return -EINVAL; - } switch (args->in.op) { case AMDGPU_SCHED_OP_PROCESS_PRIORITY_OVERRIDE: From 1cc16538402a69d9d028e0cdb7b3bbe19e64019a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 29 Jan 2026 12:58:10 +0100 Subject: [PATCH 02/99] drm/amdgpu: make amdgpu_user_wait_ioctl more resilent v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When the memory allocated by userspace isn't sufficient for all the fences then just wait on them instead of returning an error. v2: use correct variable as pointed out by Sunil Signed-off-by: Christian König Reviewed-by: Sunil Khatri Signed-off-by: Alex Deucher --- .../gpu/drm/amd/amdgpu/amdgpu_userq_fence.c | 48 +++++++++++-------- 1 file changed, 27 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c index 781896c9fd26..f77fc210cb93 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c @@ -705,7 +705,7 @@ amdgpu_userq_wait_count_fences(struct drm_file *filp, num_fences++; } - wait_info->num_fences = num_fences; + wait_info->num_fences = min(num_fences, USHRT_MAX); r = 0; error_unlock: @@ -714,6 +714,19 @@ amdgpu_userq_wait_count_fences(struct drm_file *filp, return r; } +static int +amdgpu_userq_wait_add_fence(struct drm_amdgpu_userq_wait *wait_info, + struct dma_fence **fences, unsigned int *num_fences, + struct dma_fence *fence) +{ + /* As fallback shouldn't userspace allocate enough space */ + if (*num_fences >= wait_info->num_fences) + return dma_fence_wait(fence, true); + + fences[(*num_fences)++] = dma_fence_get(fence); + return 0; +} + static int amdgpu_userq_wait_return_fence_info(struct drm_file *filp, struct drm_amdgpu_userq_wait *wait_info, @@ -757,13 +770,12 @@ amdgpu_userq_wait_return_fence_info(struct drm_file *filp, goto free_fences; dma_fence_unwrap_for_each(f, &iter, fence) { - if (num_fences >= wait_info->num_fences) { - r = -EINVAL; + r = amdgpu_userq_wait_add_fence(wait_info, fences, + &num_fences, f); + if (r) { dma_fence_put(fence); goto free_fences; } - - fences[num_fences++] = dma_fence_get(f); } dma_fence_put(fence); @@ -780,14 +792,12 @@ amdgpu_userq_wait_return_fence_info(struct drm_file *filp, if (r) goto free_fences; - if (num_fences >= wait_info->num_fences) { - dma_fence_put(fence); - r = -EINVAL; + r = amdgpu_userq_wait_add_fence(wait_info, fences, + &num_fences, fence); + dma_fence_put(fence); + if (r) goto free_fences; - } - /* Give the reference to the fence array */ - fences[num_fences++] = fence; } /* Lock all the GEM objects */ @@ -817,12 +827,10 @@ amdgpu_userq_wait_return_fence_info(struct drm_file *filp, dma_resv_for_each_fence(&resv_cursor, gobj_read[i]->resv, DMA_RESV_USAGE_READ, fence) { - if (num_fences >= wait_info->num_fences) { - r = -EINVAL; + r = amdgpu_userq_wait_add_fence(wait_info, fences, + &num_fences, fence); + if (r) goto error_unlock; - } - - fences[num_fences++] = dma_fence_get(fence); } } @@ -833,12 +841,10 @@ amdgpu_userq_wait_return_fence_info(struct drm_file *filp, dma_resv_for_each_fence(&resv_cursor, gobj_write[i]->resv, DMA_RESV_USAGE_WRITE, fence) { - if (num_fences >= wait_info->num_fences) { - r = -EINVAL; + r = amdgpu_userq_wait_add_fence(wait_info, fences, + &num_fences, fence); + if (r) goto error_unlock; - } - - fences[num_fences++] = dma_fence_get(fence); } } From 5396edca845bbbc0b9e46d278fe288b5aacde846 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sun, 2 Nov 2025 18:55:16 +0800 Subject: [PATCH 03/99] drm/amd/pm: Add smu v15_0_8 driver interface header Add smu v15_0_8 driver interface header v2: squash in updates (Alex) Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- .../inc/pmfw_if/smu15_driver_if_v15_0_8.h | 295 ++++++++++++++++++ 1 file changed, 295 insertions(+) create mode 100644 drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu15_driver_if_v15_0_8.h diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu15_driver_if_v15_0_8.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu15_driver_if_v15_0_8.h new file mode 100644 index 000000000000..6993d866183d --- /dev/null +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu15_driver_if_v15_0_8.h @@ -0,0 +1,295 @@ +/* + * 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. + * + */ +#ifndef SMU_15_0_8_DRIVER_IF_H +#define SMU_15_0_8_DRIVER_IF_H + +//I2C Interface +#define NUM_I2C_CONTROLLERS 8 +#define I2C_CONTROLLER_ENABLED 1 +#define I2C_CONTROLLER_DISABLED 0 + +#define MAX_SW_I2C_COMMANDS 24 + +typedef enum { + I2C_CONTROLLER_PORT_0, + I2C_CONTROLLER_PORT_COUNT, +} I2cControllerPort_e; + +typedef enum { + /* 50 Kbits/s not supported anymore! */ + UNSUPPORTED_1, + /* 100 Kbits/s */ + I2C_SPEED_STANDARD_100K, + /* 400 Kbits/s */ + I2C_SPEED_FAST_400K, + /* 1 Mbits/s (in fast mode) */ + I2C_SPEED_FAST_PLUS_1M, + /* 1 Mbits/s (in high speed mode) not supported anymore!*/ + UNSUPPORTED_2, + /* 2.3 Mbits/s not supported anymore! */ + UNSUPPORTED_3, + I2C_SPEED_COUNT, +} I2cSpeed_e; + +typedef enum { + I2C_CMD_READ, + I2C_CMD_WRITE, + I2C_CMD_COUNT, +} I2cCmdType_e; + +#define CMDCONFIG_STOP_BIT 0 +#define CMDCONFIG_RESTART_BIT 1 +/* bit should be 0 for read, 1 for write */ +#define CMDCONFIG_READWRITE_BIT 2 + +#define CMDCONFIG_STOP_MASK (1 << CMDCONFIG_STOP_BIT) +#define CMDCONFIG_RESTART_MASK (1 << CMDCONFIG_RESTART_BIT) +#define CMDCONFIG_READWRITE_MASK (1 << CMDCONFIG_READWRITE_BIT) + +/* 64 Bit register offsets for PPSMC_MSG_McaBankDumpDW, PPSMC_MSG_McaBankCeDumpDW messages + * eg to read MCA_BANK_OFFSET_SYND for CE index, call PPSMC_MSG_McaBankCeDumpDW twice, + * (index << 16 + MCA_BANK_OFFSET_SYND*8) argument for 1st DWORD, and + * ((index << 16 ) + MCA_BANK_OFFSET_SYND*8 + 4) argument for 2nd DWORD */ +typedef enum { + MCA_BANK_OFFSET_CTL = 0, + MCA_BANK_OFFSET_STATUS = 1, + MCA_BANK_OFFSET_ADDR = 2, + MCA_BANK_OFFSET_MISC = 3, + MCA_BANK_OFFSET_IPID = 5, + MCA_BANK_OFFSET_SYND = 6, + MCA_BANK_OFFSET_MAX = 16, +} MCA_BANK_OFFSET_e; + +/* Firmware MP1 AID MCA Error Codes stored in MCA_MP_MP1:MCMP1_SYNDT0 errorinformation */ +typedef enum { + /* MMHUB */ + CODE_DAGB0 = 0, + CODE_DAGB1 = 1, + CODE_DAGB2 = 2, + CODE_DAGB3 = 3, + CODE_DAGB4 = 4, + CODE_EA0 = 5, + CODE_EA1 = 6, + CODE_EA2 = 7, + CODE_EA3 = 8, + CODE_EA4 = 9, + CODE_UTCL2_ROUTER = 10, + CODE_VML2 = 11, + CODE_VML2_WALKER = 12, + CODE_MMCANE = 13, + + /* VCN VCPU */ + CODE_VIDD = 14, + CODE_VIDV = 15, + /* VCN JPEG */ + CODE_JPEG0S = 16, + CODE_JPEG0D = 17, + CODE_JPEG1S = 18, + CODE_JPEG1D = 19, + CODE_JPEG2S = 20, + CODE_JPEG2D = 21, + CODE_JPEG3S = 22, + CODE_JPEG3D = 23, + CODE_JPEG4S = 24, + CODE_JPEG4D = 25, + CODE_JPEG5S = 26, + CODE_JPEG5D = 27, + CODE_JPEG6S = 28, + CODE_JPEG6D = 29, + CODE_JPEG7S = 30, + CODE_JPEG7D = 31, + /* VCN MMSCH */ + CODE_MMSCHD = 32, + + /* SDMA */ + CODE_SDMA0 = 33, + CODE_SDMA1 = 34, + CODE_SDMA2 = 35, + CODE_SDMA3 = 36, + + /* SOC */ + CODE_HDP = 37, + CODE_ATHUB = 38, + CODE_IH = 39, + CODE_XHUB_POISON = 40, + CODE_SMN_SLVERR = 41, + CODE_WDT = 42, + + CODE_UNKNOWN = 43, + CODE_DMA = 44, + CODE_COUNT = 45, +} ERR_CODE_e; + +/* Firmware MP5 XCD MCA Error Codes stored in MCA_MP_MP5:MCMP5_SYNDT0 errorinformation */ +typedef enum { + /* SH POISON FED */ + SH_FED_CODE = 0, + /* GCEA Pin UE_ERR regs */ + GCEA_CODE = 1, + SQ_CODE = 2, + LDS_CODE = 3, + GDS_CODE = 4, + SP0_CODE = 5, + SP1_CODE = 6, + TCC_CODE = 7, + TCA_CODE = 8, + TCX_CODE = 9, + CPC_CODE = 10, + CPF_CODE = 11, + CPG_CODE = 12, + SPI_CODE = 13, + RLC_CODE = 14, + /* GCEA Pin, UE_EDC regs */ + SQC_CODE = 15, + TA_CODE = 16, + TD_CODE = 17, + TCP_CODE = 18, + TCI_CODE = 19, + /* GC Router */ + GC_ROUTER_CODE = 20, + VML2_CODE = 21, + VML2_WALKER_CODE = 22, + ATCL2_CODE = 23, + GC_CANE_CODE = 24, + + /* SOC error codes 41-43 are common with ERR_CODE_e */ + MP5_CODE_SMN_SLVERR = CODE_SMN_SLVERR, + MP5_CODE_UNKNOWN = CODE_UNKNOWN, +} GC_ERROR_CODE_e; + +/* SW I2C Command Table */ +typedef struct { + /* Return data for read. Data to send for write*/ + uint8_t ReadWriteData; + /* Includes whether associated command should have a stop or restart command, + * and is a read or write */ + uint8_t CmdConfig; +} SwI2cCmd_t; + +/* SW I2C Request Table */ +typedef struct { + /* CKSVII2C0(0) or //CKSVII2C1(1) */ + uint8_t I2CcontrollerPort; + /* Use I2cSpeed_e to indicate speed to select */ + uint8_t I2CSpeed; + /* Slave address of device */ + uint8_t SlaveAddress; + /* Number of commands */ + uint8_t NumCmds; + SwI2cCmd_t SwI2cCmds[MAX_SW_I2C_COMMANDS]; +} SwI2cRequest_t; + +typedef struct { + SwI2cRequest_t SwI2cRequest; + uint32_t Spare[8]; + /* SMU internal use */ + uint32_t MmHubPadding[8]; +} SwI2cRequestExternal_t; + +typedef enum { + PPCLK_UCLK, + PPCLK_COUNT, +} PPCLK_e; + +typedef enum { + GPIO_INT_POLARITY_ACTIVE_LOW, + GPIO_INT_POLARITY_ACTIVE_HIGH, +} GpioIntPolarity_e; + +/* TODO confirm if this is used in MI300 PPSMC_MSG_SetUclkDpmMode */ +typedef enum { + UCLK_DPM_MODE_BANDWIDTH, + UCLK_DPM_MODE_LATENCY, +} UCLK_DPM_MODE_e; + +typedef struct { + /* 2 AVFS.PSM chains */ + uint16_t AvgPsmCount_Chain0[13]; + uint16_t AvgPsmCount_Chain1[15]; + uint16_t MinPsmCount_Chain0[13]; + uint16_t MinPsmCount_Chain1[15]; + float MaxTemperature; + + /* For voltage conversions, these are the array indexes + * 0:SOCIO + * 1:065_UCIE + * 2:075_UCIE + * 3:11_GTA + * 4:075_GTA */ + float MinPsmVoltage[5]; + float AvgPsmVoltage[5]; +} AvfsDebugTableMid_t; + +typedef struct { + /* 7 AVFS.PSM chains - not including TRO */ + uint16_t AvgPsmCount_Chain0[15]; + uint16_t AvgPsmCount_Chain1[15]; + uint16_t AvgPsmCount_Chain2[13]; + uint16_t AvgPsmCount_Chain3[13]; + uint16_t AvgPsmCount_Chain4[15]; + uint16_t AvgPsmCount_Chain5[15]; + uint16_t AvgPsmCount_Chain6[5]; + uint16_t MinPsmCount_Chain0[15]; + uint16_t MinPsmCount_Chain1[15]; + uint16_t MinPsmCount_Chain2[13]; + uint16_t MinPsmCount_Chain3[13]; + uint16_t MinPsmCount_Chain4[15]; + uint16_t MinPsmCount_Chain5[15]; + uint16_t MinPsmCount_Chain6[5]; + float MaxTemperature; + + /* For voltage conversions, these are the array indexes + * 0:VDDX */ + float MinPsmVoltage; + float AvgPsmVoltage; +} AvfsDebugTableAid_t; + +typedef struct { + /* 0-27 GFX, 28-29 SOC */ + uint16_t avgPsmCount[30]; + uint16_t minPsmCount[30]; + float avgPsmVoltage[30]; + float minPsmVoltage[30]; +} AvfsDebugTableXcd_t; + +/* Defines used for IH-based thermal interrupts to GFX driver - A/X only */ +#define IH_INTERRUPT_ID_TO_DRIVER 0xFE +#define IH_INTERRUPT_CONTEXT_ID_THERMAL_THROTTLING 0x7 +#define IH_INTERRUPT_VFFLR_INT 0xA + +/* thermal over-temp mask defines for IH interrup to host */ +#define THROTTLER_PROCHOT_BIT 0 +#define THROTTLER_RESERVED 1 +/* AID, XCD, CCD throttling */ +#define THROTTLER_THERMAL_SOCKET_BIT 2 +/* VRHOT */ +#define THROTTLER_THERMAL_VR_BIT 3 +#define THROTTLER_THERMAL_HBM_BIT 4 +/* UEs are always reported, set flag to 0 to prevent clearing of UEs */ +#define ClearMcaOnRead_UE_FLAG_MASK 0x1 +/* Enable CE logging and clearing to driver */ +#define ClearMcaOnRead_CE_POLL_MASK 0x2 +/* AID MMHUB client IP CE Logging and clearing */ +#define ClearMcaOnRead_MMHUB_POLL_MASK 0x4 + +#endif From ad830273db629b7b42e23ffba253cdd5540bb178 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sun, 2 Nov 2025 18:56:52 +0800 Subject: [PATCH 04/99] drm/amd/pm: Add smu v15_0_8 message header Add smu v15_0_8 message header v2: squash in updates (Alex) Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- .../pm/swsmu/inc/pmfw_if/smu_v15_0_8_ppsmc.h | 100 ++++++++++++++++++ 1 file changed, 100 insertions(+) create mode 100644 drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_ppsmc.h diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_ppsmc.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_ppsmc.h new file mode 100644 index 000000000000..7ffb445f4c0c --- /dev/null +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_ppsmc.h @@ -0,0 +1,100 @@ +/* + * 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. + * + */ +#ifndef SMU_15_0_8_PPSMC_H +#define SMU_15_0_8_PPSMC_H + +/* SMU Response Codes */ +#define PPSMC_Result_OK 0x1 +#define PPSMC_Result_Failed 0xFF +#define PPSMC_Result_UnknownCmd 0xFE +#define PPSMC_Result_CmdRejectedPrereq 0xFD +#define PPSMC_Result_CmdRejectedBusy 0xFC + +/* Message Definitions */ +#define PPSMC_MSG_TestMessage 0x1 +#define PPSMC_MSG_GetSmuVersion 0x2 +#define PPSMC_MSG_GfxDriverReset 0x3 +#define PPSMC_MSG_GetDriverIfVersion 0x4 +#define PPSMC_MSG_EnableAllSmuFeatures 0x5 +#define PPSMC_MSG_GetMetricsVersion 0x6 +#define PPSMC_MSG_GetMetricsTable 0x7 +#define PPSMC_MSG_GetEnabledSmuFeatures 0x8 +#define PPSMC_MSG_SetDriverDramAddr 0x9 //ARG0: low address, ARG1: high address +#define PPSMC_MSG_SetToolsDramAddr 0xA //ARG0: low address, ARG1: high address +//#define PPSMC_MSG_SetSystemVirtualDramAddr 0xB +#define PPSMC_MSG_SetSoftMaxByFreq 0xC +#define PPSMC_MSG_SetPptLimit 0xD +#define PPSMC_MSG_GetPptLimit 0xE +#define PPSMC_MSG_DramLogSetDramAddr 0xF //ARG0: low address, ARG1: high address, ARG2: size +#define PPSMC_MSG_HeavySBR 0x10 +#define PPSMC_MSG_DFCstateControl 0x11 +#define PPSMC_MSG_GfxDriverResetRecovery 0x12 +#define PPSMC_MSG_TriggerVFFLR 0x13 +#define PPSMC_MSG_SetSoftMinGfxClk 0x14 +#define PPSMC_MSG_SetSoftMaxGfxClk 0x15 +#define PPSMC_MSG_PrepareForDriverUnload 0x16 +#define PPSMC_MSG_QueryValidMcaCount 0x17 +#define PPSMC_MSG_McaBankDumpDW 0x18 +#define PPSMC_MSG_ClearMcaOnRead 0x19 +#define PPSMC_MSG_QueryValidMcaCeCount 0x1A +#define PPSMC_MSG_McaBankCeDumpDW 0x1B +#define PPSMC_MSG_SelectPLPDMode 0x1C +#define PPSMC_MSG_SetThrottlingPolicy 0x1D +#define PPSMC_MSG_ResetSDMA 0x1E +#define PPSMC_MSG_GetRasTableVersion 0x1F +#define PPSMC_MSG_GetRmaStatus 0x20 +#define PPSMC_MSG_GetBadPageCount 0x21 +#define PPSMC_MSG_GetBadPageMcaAddress 0x22 +#define PPSMC_MSG_GetBadPagePaAddress 0x23 +#define PPSMC_MSG_SetTimestamp 0x24 +#define PPSMC_MSG_GetTimestamp 0x25 +#define PPSMC_MSG_GetRasPolicy 0x26 +#define PPSMC_MSG_GetBadPageIpIdLoHi 0x27 +#define PPSMC_MSG_EraseRasTable 0x28 +#define PPSMC_MSG_GetStaticMetricsTable 0x29 +#define PPSMC_MSG_ResetVfArbitersByIndex 0x2A +#define PPSMC_MSG_GetBadPageSeverity 0x2B +#define PPSMC_MSG_GetSystemMetricsTable 0x2C +#define PPSMC_MSG_GetSystemMetricsVersion 0x2D +#define PPSMC_MSG_ResetVCN 0x2E +#define PPSMC_MSG_SetFastPptLimit 0x2F +#define PPSMC_MSG_GetFastPptLimit 0x30 +#define PPSMC_MSG_SetSoftMinGl2clk 0x31 +#define PPSMC_MSG_SetSoftMaxGl2clk 0x32 +#define PPSMC_MSG_SetSoftMinFclk 0x33 +#define PPSMC_MSG_SetSoftMaxFclk 0x34 +#define PPSMC_Message_Count 0x35 + +/* PSMC Reset Types for driver msg argument */ +#define PPSMC_RESET_TYPE_DRIVER_MODE_1_RESET 0x1 +#define PPSMC_RESET_TYPE_DRIVER_MODE_2_RESET 0x2 +#define PPSMC_RESET_TYPE_DRIVER_MODE_3_RESET 0x3 + +/* PLPD modes */ +#define PPSMC_PLPD_MODE_DEFAULT 0x1 +#define PPSMC_PLPD_MODE_OPTIMIZED 0x2 + +typedef uint32_t PPSMC_Result; +typedef uint32_t PPSMC_MSG; + +#endif From a3175e57235bc4c9accc3a4e845a7ec9972c9370 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sun, 2 Nov 2025 18:58:46 +0800 Subject: [PATCH 05/99] drm/amd/pm: Add smu v15_0_8 pmfw header Add smu v15_0_8 pmfw header v2: squash in updates (Alex) Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- .../pm/swsmu/inc/pmfw_if/smu_v15_0_8_pmfw.h | 427 ++++++++++++++++++ 1 file changed, 427 insertions(+) create mode 100644 drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_pmfw.h diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_pmfw.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_pmfw.h new file mode 100644 index 000000000000..a3401c4cc20b --- /dev/null +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v15_0_8_pmfw.h @@ -0,0 +1,427 @@ +/* + * 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. + * + */ +#ifndef SMU_15_0_8_PMFW_H +#define SMU_15_0_8_PMFW_H + +#define NUM_VCLK_DPM_LEVELS 4 +#define NUM_DCLK_DPM_LEVELS 4 +#define NUM_SOCCLK_DPM_LEVELS 4 +#define NUM_LCLK_DPM_LEVELS 4 +#define NUM_UCLK_DPM_LEVELS 4 +#define NUM_FCLK_DPM_LEVELS 4 +#define NUM_XGMI_DPM_LEVELS 2 +#define NUM_PCIE_BITRATES 4 +#define NUM_XGMI_BITRATES 4 +#define NUM_XGMI_WIDTHS 3 +#define NUM_GFX_P2S_TABLES 8 +#define NUM_PSM_DIDT_THRESHOLDS 3 +#define NUM_XCD_XVMIN_VMIN_THRESHOLDS 3 + +#define PRODUCT_MODEL_NUMBER_LEN 20 +#define PRODUCT_NAME_LEN 64 +#define PRODUCT_SERIAL_LEN 20 +#define PRODUCT_MANUFACTURER_NAME_LEN 32 +#define PRODUCT_FRU_ID_LEN 32 + +//Feature ID list +#define FEATURE_ID_DATA_CALCULATION 1 +#define FEATURE_ID_DPM_FCLK 2 +#define FEATURE_ID_DPM_GFXCLK 3 +#define FEATURE_ID_DPM_SPARE_4 4 +#define FEATURE_ID_DPM_SPARE_5 5 +#define FEATURE_ID_DPM_UCLK 6 +#define FEATURE_ID_DPM_SPARE_7 7 +#define FEATURE_ID_DPM_XGMI 8 +#define FEATURE_ID_DS_FCLK 9 +#define FEATURE_ID_DS_GFXCLK 10 +#define FEATURE_ID_DS_LCLK 11 +#define FEATURE_ID_DS_MP0CLK 12 +#define FEATURE_ID_DS_MP1CLK 13 +#define FEATURE_ID_DS_MPIOCLK 14 +#define FEATURE_ID_DS_SOCCLK 15 +#define FEATURE_ID_DS_VCN 16 +#define FEATURE_ID_PPT 17 +#define FEATURE_ID_TDC 18 +#define FEATURE_ID_THERMAL 19 +#define FEATURE_ID_SOC_PCC 20 +#define FEATURE_ID_PROCHOT 21 +#define FEATURE_ID_XVMIN0_VMIN_AID 22 +#define FEATURE_ID_XVMIN1_DD_AID 23 +#define FEATURE_ID_XVMIN0_VMIN_XCD 24 +#define FEATURE_ID_XVMIN1_DD_XCD 25 +#define FEATURE_ID_FW_CTF 26 +#define FEATURE_ID_MGCG 27 +#define FEATURE_ID_PSI7 28 +#define FEATURE_ID_XGMI_PER_LINK_PWR_DOWN 29 +#define FEATURE_ID_SOC_DC_RTC 30 +#define FEATURE_ID_GFX_DC_RTC 31 +#define FEATURE_ID_DVM_MIN_PSM 32 +#define FEATURE_ID_PRC 33 +#define FEATURE_ID_PSM_DIDT 34 +#define FEATURE_ID_PIT 35 +#define FEATURE_ID_DVO 36 +#define FEATURE_ID_XVMIN_CLKSTOP_DS 37 +#define FEATURE_ID_HBM_THROTTLE_CTRL 38 +#define FEATURE_ID_DPM_GL2CLK 39 +#define FEATURE_ID_GC_CAC_EDC 40 +#define FEATURE_ID_DS_DMABECLK 41 +#define FEATURE_ID_DS_MPIFOECLK 42 +#define FEATURE_ID_DS_MPRASCLK 43 +#define FEATURE_ID_DS_MPNHTCLK 44 +#define FEATURE_ID_DS_FIOCLK 45 +#define FEATURE_ID_DS_DXIOCLK 46 +#define FEATURE_ID_PCC 47 +#define FEATURE_ID_OCP 48 +#define FEATURE_ID_TRO 49 +#define FEATURE_ID_GL2_CAC_EDC 50 +#define FEATURE_ID_SPARE_51 51 +#define FEATURE_ID_GL2_CGCG 52 +#define FEATURE_ID_XCAC 53 +#define FEATURE_ID_DS_GL2CLK 54 +#define FEATURE_ID_FCS_VIN_PCC 55 +#define FEATURE_ID_FCS_VDDX_OCP_WARN 56 +#define FEATURE_ID_FCS_PWRBRK 57 +#define FEATURE_ID_DF_CSTATE 58 +#define FEATURE_ID_ARO 59 +#define FEATURE_ID_PS_PsPowerLimit 60 +#define FEATURE_ID_PS_PsPowerFloor 61 +#define FEATURE_ID_OCPWARNRC 62 +#define FEATURE_ID_XGMI_FOLDING 63 +#define FEATURE_ID_SMU_CG 64 +#define NUM_FEATURES 65 + +//MGCG Feature ID List +#define WAFL_CG 0 +#define SMU_FUSE_CG_DEEPSLEEP 1 +#define SMUIO_CG 2 +#define RSMU_MGCG 3 +#define SMU_CLK_MGCG 4 +#define MP5_CG 5 +#define UMC_CG 6 +#define WAFL0_CLK 7 +#define WAFL1_CLK 8 +#define VCN_MGCG 9 +#define GL2_MGCG 10 +#define MGCG_NUM_FEATURES 11 + +/* enum for MPIO PCIe gen speed msgs */ +typedef enum { + PCIE_LINK_SPEED_INDEX_TABLE_GEN1, + PCIE_LINK_SPEED_INDEX_TABLE_GEN2, + PCIE_LINK_SPEED_INDEX_TABLE_GEN3, + PCIE_LINK_SPEED_INDEX_TABLE_GEN4, + PCIE_LINK_SPEED_INDEX_TABLE_GEN5, + PCIE_LINK_SPEED_INDEX_TABLE_GEN6, + PCIE_LINK_SPEED_INDEX_TABLE_GEN6_ESM, + PCIE_LINK_SPEED_INDEX_TABLE_COUNT +} PCIE_LINK_SPEED_INDEX_TABLE_e; + +typedef enum { + GFX_GUARDBAND_OFFSET_0, + GFX_GUARDBAND_OFFSET_1, + GFX_GUARDBAND_OFFSET_2, + GFX_GUARDBAND_OFFSET_3, + GFX_GUARDBAND_OFFSET_4, + GFX_GUARDBAND_OFFSET_5, + GFX_GUARDBAND_OFFSET_6, + GFX_GUARDBAND_OFFSET_7, + GFX_GUARDBAND_OFFSET_COUNT +} GFX_GUARDBAND_OFFSET_e; + +typedef enum { + GFX_DVM_MARGINHI_0, + GFX_DVM_MARGINHI_1, + GFX_DVM_MARGINHI_2, + GFX_DVM_MARGINHI_3, + GFX_DVM_MARGINHI_4, + GFX_DVM_MARGINHI_5, + GFX_DVM_MARGINHI_6, + GFX_DVM_MARGINHI_7, + GFX_DVM_MARGINLO_0, + GFX_DVM_MARGINLO_1, + GFX_DVM_MARGINLO_2, + GFX_DVM_MARGINLO_3, + GFX_DVM_MARGINLO_4, + GFX_DVM_MARGINLO_5, + GFX_DVM_MARGINLO_6, + GFX_DVM_MARGINLO_7, + GFX_DVM_MARGIN_COUNT +} GFX_DVM_MARGIN_e; + +typedef enum{ + SYSTEM_TEMP_UBB_FPGA, + SYSTEM_TEMP_UBB_FRONT, + SYSTEM_TEMP_UBB_BACK, + SYSTEM_TEMP_UBB_OAM7, + SYSTEM_TEMP_UBB_IBC, + SYSTEM_TEMP_UBB_UFPGA, + SYSTEM_TEMP_UBB_OAM1, + SYSTEM_TEMP_OAM_0_1_HSC, + SYSTEM_TEMP_OAM_2_3_HSC, + SYSTEM_TEMP_OAM_4_5_HSC, + SYSTEM_TEMP_OAM_6_7_HSC, + SYSTEM_TEMP_UBB_FPGA_0V72_VR, + SYSTEM_TEMP_UBB_FPGA_3V3_VR, + SYSTEM_TEMP_RETIMER_0_1_2_3_1V2_VR, + SYSTEM_TEMP_RETIMER_4_5_6_7_1V2_VR, + SYSTEM_TEMP_RETIMER_0_1_0V9_VR, + SYSTEM_TEMP_RETIMER_4_5_0V9_VR, + SYSTEM_TEMP_RETIMER_2_3_0V9_VR, + SYSTEM_TEMP_RETIMER_6_7_0V9_VR, + SYSTEM_TEMP_OAM_0_1_2_3_3V3_VR, + SYSTEM_TEMP_OAM_4_5_6_7_3V3_VR, + SYSTEM_TEMP_IBC_HSC, + SYSTEM_TEMP_IBC, + SYSTEM_TEMP_MAX_ENTRIES = 32 +} SYSTEM_TEMP_e; + +typedef enum{ + NODE_TEMP_RETIMER, + NODE_TEMP_IBC_TEMP, + NODE_TEMP_IBC_2_TEMP, + NODE_TEMP_VDD18_VR_TEMP, + NODE_TEMP_04_HBM_B_VR_TEMP, + NODE_TEMP_04_HBM_D_VR_TEMP, + NODE_TEMP_MAX_TEMP_ENTRIES = 12 +} NODE_TEMP_e; + +typedef enum { + SVI_PLANE_VDDCR_X0_TEMP, + SVI_PLANE_VDDCR_X1_TEMP, + + SVI_PLANE_VDDIO_HBM_B_TEMP, + SVI_PLANE_VDDIO_HBM_D_TEMP, + SVI_PLANE_VDDIO_04_HBM_B_TEMP, + SVI_PLANE_VDDIO_04_HBM_D_TEMP, + SVI_PLANE_VDDCR_HBM_B_TEMP, + SVI_PLANE_VDDCR_HBM_D_TEMP, + SVI_PLANE_VDDCR_075_HBM_B_TEMP, + SVI_PLANE_VDDCR_075_HBM_D_TEMP, + + SVI_PLANE_VDDIO_11_GTA_A_TEMP, + SVI_PLANE_VDDIO_11_GTA_C_TEMP, + SVI_PLANE_VDDAN_075_GTA_A_TEMP, + SVI_PLANE_VDDAN_075_GTA_C_TEMP, + + SVI_PLANE_VDDCR_075_UCIE_TEMP, + SVI_PLANE_VDDIO_065_UCIEAA_TEMP, + SVI_PLANE_VDDIO_065_UCIEAM_A_TEMP, + SVI_PLANE_VDDIO_065_UCIEAM_C_TEMP, + + SVI_PLANE_VDDCR_SOCIO_A_TEMP, + SVI_PLANE_VDDCR_SOCIO_C_TEMP, + + SVI_PLANE_VDDAN_075_TEMP, + SVI_MAX_TEMP_ENTRIES, //22 +} SVI_TEMP_e; + +typedef enum{ + SYSTEM_POWER_UBB_POWER, + SYSTEM_POWER_UBB_POWER_THRESHOLD, + SYSTEM_POWER_MAX_ENTRIES_WO_RESERVED, + SYSTEM_POWER_MAX_ENTRIES = 4 +} SYSTEM_POWER_e; + +#define SMU_METRICS_TABLE_VERSION 0xF + +typedef struct __attribute__((packed, aligned(4))) { + uint64_t AccumulationCounter; + + //TEMPERATURE + uint32_t MaxSocketTemperature; + uint32_t MaxVrTemperature; + uint32_t HbmTemperature[12]; + uint64_t MaxSocketTemperatureAcc; + uint64_t MaxVrTemperatureAcc; + uint64_t HbmTemperatureAcc[12]; + uint32_t MidTemperature[2]; + uint32_t AidTemperature[2]; + uint32_t XcdTemperature[8]; + + //POWER + uint32_t SocketPowerLimit; + uint32_t SocketPower; + + //ENERGY + uint64_t Timestamp; + uint64_t SocketEnergyAcc; + uint64_t HbmEnergyAcc; + + //FREQUENCY + uint32_t GfxclkFrequencyLimit; + uint32_t FclkFrequency[2]; + uint32_t UclkFrequency[2]; + uint64_t GfxclkFrequencyAcc[8]; + uint32_t GfxclkFrequency[8]; + uint32_t SocclkFrequency[2]; + uint32_t VclkFrequency[4]; + uint32_t DclkFrequency[4]; + uint32_t LclkFrequency[2]; + + //XGMI: + uint32_t XgmiWidth; + uint32_t XgmiBitrate; + uint64_t XgmiReadBandwidthAcc; + uint64_t XgmiWriteBandwidthAcc; + + //ACTIVITY: + uint32_t SocketGfxBusy; + uint32_t DramBandwidthUtilization; + uint64_t SocketGfxBusyAcc; + uint64_t DramBandwidthAcc; + uint32_t MaxDramBandwidth; + uint64_t DramBandwidthUtilizationAcc; + uint64_t PcieBandwidthAcc[2]; + + //THROTTLERS + uint64_t ProchotResidencyAcc; + uint64_t PptResidencyAcc; + uint64_t SocketThmResidencyAcc; + uint64_t VrThmResidencyAcc; + uint64_t HbmThmResidencyAcc; + + //PCIE BW Data and error count + uint32_t PcieBandwidth[2]; + uint64_t PCIeL0ToRecoveryCountAcc; + uint64_t PCIenReplayAAcc; + uint64_t PCIenReplayARolloverCountAcc; + uint64_t PCIeNAKSentCountAcc; + uint64_t PCIeNAKReceivedCountAcc; + uint64_t PCIeOtherEndRecoveryAcc; // The Pcie counter itself is accumulated + + // VCN/JPEG ACTIVITY + uint32_t VcnBusy[4]; + uint32_t JpegBusy[40]; + + // PCIE LINK Speed and width + uint32_t PCIeLinkSpeed; + uint32_t PCIeLinkWidth; + + // PER XCD ACTIVITY + uint32_t GfxBusy[8]; + uint64_t GfxBusyAcc[8]; + + //NVML-Parity: Total App Clock Counter + uint64_t GfxclkBelowHostLimitPptAcc[8]; + uint64_t GfxclkBelowHostLimitThmAcc[8]; + uint64_t GfxclkBelowHostLimitTotalAcc[8]; + uint64_t GfxclkLowUtilizationAcc[8]; +} MetricsTable_t; + +#define SMU_SYSTEM_METRICS_TABLE_VERSION 0x1 + +#pragma pack(push, 4) +typedef struct { + uint64_t AccumulationCounter; // Last update timestamp + uint16_t LabelVersion; //Defaults to 0. + uint16_t NodeIdentifier; + int16_t SystemTemperatures[SYSTEM_TEMP_MAX_ENTRIES]; // Signed integer temperature value in Celsius, unused fields are set to 0xFFFF + int16_t NodeTemperatures[NODE_TEMP_MAX_TEMP_ENTRIES]; // Signed integer temperature value in Celsius, unused fields are set to 0xFFFF + int16_t VrTemperatures[SVI_MAX_TEMP_ENTRIES]; // Signed integer temperature value in Celsius, 13 entries, + int16_t spare[7]; + + //NPM: NODE POWER MANAGEMENT + uint32_t NodePowerLimit; + uint32_t NodePower; + uint32_t GlobalPPTResidencyAcc; + + uint16_t SystemPower[SYSTEM_POWER_MAX_ENTRIES]; // UBB Current Power and Power Threshold +} SystemMetricsTable_t; +#pragma pack(pop) + +#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; + +/* FRU product information */ +typedef struct __attribute__((aligned(4))) { + uint8_t ModelNumber[PRODUCT_MODEL_NUMBER_LEN]; + uint8_t Name[PRODUCT_NAME_LEN]; + uint8_t Serial[PRODUCT_SERIAL_LEN]; + uint8_t ManufacturerName[PRODUCT_MANUFACTURER_NAME_LEN]; + uint8_t FruId[PRODUCT_FRU_ID_LEN]; +} FRUProductInfo_t; + +#define SMU_STATIC_METRICS_TABLE_VERSION 0x1 + +#pragma pack(push, 4) +typedef struct { + //FRU PRODUCT INFO + FRUProductInfo_t ProductInfo; //from i2c + + //POWER + uint32_t MaxSocketPowerLimit; + + //FREQUENCY RANGE + uint32_t MaxGfxclkFrequency; + uint32_t MinGfxclkFrequency; + uint32_t MaxFclkFrequency; + uint32_t MinFclkFrequency; + uint32_t MaxGl2clkFrequency; + uint32_t MinGl2clkFrequency; + uint32_t UclkFrequencyTable[4]; + uint32_t SocclkFrequency; + uint32_t LclkFrequency; + uint32_t VclkFrequency; + uint32_t DclkFrequency; + + //CTF limits + uint32_t CTFLimit_MID; + uint32_t CTFLimit_AID; + uint32_t CTFLimit_XCD; + uint32_t CTFLimit_HBM; + + //Thermal Throttling limits + uint32_t ThermalLimit_MID; + uint32_t ThermalLimit_AID; + uint32_t ThermalLimit_XCD; + uint32_t ThermalLimit_HBM; + + //PSNs + uint64_t PublicSerialNumber_MID[2]; + uint64_t PublicSerialNumber_AID[2]; + uint64_t PublicSerialNumber_XCD[8]; + + //XGMI + uint32_t MaxXgmiWidth; + uint32_t MaxXgmiBitrate; + + // Telemetry + uint32_t InputTelemetryVoltageInmV; + + // General info + uint32_t pldmVersion[2]; + + uint32_t PPT1Max; + uint32_t PPT1Min; + uint32_t PPT1Default; +} StaticMetricsTable_t; +#pragma pack(pop) + +#endif From fd490bb9e1054705e1b35e6f321cdc713e0c7348 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Sun, 15 Mar 2026 11:53:31 +0530 Subject: [PATCH 06/99] drm/amd/ras: Remove redundant NULL check in pending bad-bank list iteration MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ras_umc_log_pending_bad_bank() walks through a list of pending ECC bad-bank entries. These entries are saved when a bad-bank error cannot be processed immediately, for example during a GPU reset. Later, this function iterates over the pending list and retries logging each bad-bank error. If logging succeeds, the entry is removed from the list and the memory for that node is freed. The loop uses list_for_each_entry_safe(), which already guarantees that ecc_node points to a valid list entry while the loop body is executing. Checking "ecc_node &&" inside the loop is therefore unnecessary and redundant. Fixes the below: drivers/gpu/drm/amd/amdgpu/../ras/rascore/ras_umc.c:225 ras_umc_log_pending_bad_bank() warn: variable dereferenced before check 'ecc_node' (see line 223) Fixes: 7a3f9c0992c4 ("drm/amd/ras: Add umc common ras functions") Cc: Dan Carpenter Cc: YiPeng Chai Cc: Tao Zhou Cc: Hawking Zhang Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: YiPeng Chai Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/ras/rascore/ras_umc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/ras/rascore/ras_umc.c b/drivers/gpu/drm/amd/ras/rascore/ras_umc.c index 23118f41eb96..c7ae005ea705 100644 --- a/drivers/gpu/drm/amd/ras/rascore/ras_umc.c +++ b/drivers/gpu/drm/amd/ras/rascore/ras_umc.c @@ -222,7 +222,7 @@ int ras_umc_log_pending_bad_bank(struct ras_core_context *ras_core) mutex_lock(&ras_umc->pending_ecc_lock); list_for_each_entry_safe(ecc_node, tmp, &ras_umc->pending_ecc_list, node){ - if (ecc_node && !ras_umc_log_bad_bank(ras_core, &ecc_node->ecc)) { + if (!ras_umc_log_bad_bank(ras_core, &ecc_node->ecc)) { list_del(&ecc_node->node); kfree(ecc_node); } From cc7c7121ae082b7b82891baa7280f1ff2608f22b Mon Sep 17 00:00:00 2001 From: Yussuf Khalil Date: Fri, 6 Mar 2026 12:06:35 +0000 Subject: [PATCH 07/99] drm/amd/display: Do not skip unrelated mode changes in DSC validation Starting with commit 17ce8a6907f7 ("drm/amd/display: Add dsc pre-validation in atomic check"), amdgpu resets the CRTC state mode_changed flag to false when recomputing the DSC configuration results in no timing change for a particular stream. However, this is incorrect in scenarios where a change in MST/DSC configuration happens in the same KMS commit as another (unrelated) mode change. For example, the integrated panel of a laptop may be configured differently (e.g., HDR enabled/disabled) depending on whether external screens are attached. In this case, plugging in external DP-MST screens may result in the mode_changed flag being dropped incorrectly for the integrated panel if its DSC configuration did not change during precomputation in pre_validate_dsc(). At this point, however, dm_update_crtc_state() has already created new streams for CRTCs with DSC-independent mode changes. In turn, amdgpu_dm_commit_streams() will never release the old stream, resulting in a memory leak. amdgpu_dm_atomic_commit_tail() will never acquire a reference to the new stream either, which manifests as a use-after-free when the stream gets disabled later on: BUG: KASAN: use-after-free in dc_stream_release+0x25/0x90 [amdgpu] Write of size 4 at addr ffff88813d836524 by task kworker/9:9/29977 Workqueue: events drm_mode_rmfb_work_fn Call Trace: dump_stack_lvl+0x6e/0xa0 print_address_description.constprop.0+0x88/0x320 ? dc_stream_release+0x25/0x90 [amdgpu] print_report+0xfc/0x1ff ? srso_alias_return_thunk+0x5/0xfbef5 ? __virt_addr_valid+0x225/0x4e0 ? dc_stream_release+0x25/0x90 [amdgpu] kasan_report+0xe1/0x180 ? dc_stream_release+0x25/0x90 [amdgpu] kasan_check_range+0x125/0x200 dc_stream_release+0x25/0x90 [amdgpu] dc_state_destruct+0x14d/0x5c0 [amdgpu] dc_state_release.part.0+0x4e/0x130 [amdgpu] dm_atomic_destroy_state+0x3f/0x70 [amdgpu] drm_atomic_state_default_clear+0x8ee/0xf30 ? drm_mode_object_put.part.0+0xb1/0x130 __drm_atomic_state_free+0x15c/0x2d0 atomic_remove_fb+0x67e/0x980 Since there is no reliable way of figuring out whether a CRTC has unrelated mode changes pending at the time of DSC validation, remember the value of the mode_changed flag from before the point where a CRTC was marked as potentially affected by a change in DSC configuration. Reset the mode_changed flag to this earlier value instead in pre_validate_dsc(). Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5004 Fixes: 17ce8a6907f7 ("drm/amd/display: Add dsc pre-validation in atomic check") Signed-off-by: Yussuf Khalil Reviewed-by: Harry Wentland Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 5 +++++ drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h | 1 + drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 4 +++- 3 files changed, 9 insertions(+), 1 deletion(-) 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 6fb8c4f447a9..1e010d095f06 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -12528,6 +12528,11 @@ static int amdgpu_dm_atomic_check(struct drm_device *dev, } if (dc_resource_is_dsc_encoding_supported(dc)) { + for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { + dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); + dm_new_crtc_state->mode_changed_independent_from_dsc = new_crtc_state->mode_changed; + } + for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { if (drm_atomic_crtc_needs_modeset(new_crtc_state)) { ret = add_affected_mst_dsc_crtcs(state, crtc); diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h index 83fefd902355..d1a14e0c12bd 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -1005,6 +1005,7 @@ struct dm_crtc_state { bool freesync_vrr_info_changed; + bool mode_changed_independent_from_dsc; bool dsc_force_changed; bool vrr_supported; struct mod_freesync_config freesync_config; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index 7be50e8c0636..5d8c4c7020b1 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -1744,9 +1744,11 @@ int pre_validate_dsc(struct drm_atomic_state *state, int ind = find_crtc_index_in_state_by_stream(state, stream); if (ind >= 0) { + struct dm_crtc_state *dm_new_crtc_state = to_dm_crtc_state(state->crtcs[ind].new_state); + DRM_INFO_ONCE("%s:%d MST_DSC no mode changed for stream 0x%p\n", __func__, __LINE__, stream); - state->crtcs[ind].new_state->mode_changed = 0; + dm_new_crtc_state->base.mode_changed = dm_new_crtc_state->mode_changed_independent_from_dsc; } } } From 809773e06692b9feef372c56ca0747cce9bbf860 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Wed, 18 Mar 2026 10:15:09 +0530 Subject: [PATCH 08/99] drm/amdgpu: Avoid NULL dereference in discovery topology coredump path v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When a GPU fault or timeout happens, the driver creates a devcoredump to collect debug information. During this, amdgpu_devcoredump_format() calls amdgpu_discovery_dump() to print IP discovery data. amdgpu_discovery_dump() uses: adev->discovery.ip_top and then accesses: ip_top->die_kset amdgpu_discovery_dump() uses adev->discovery.ip_top. However, ip_top may be NULL if the discovery topology was never initialized. The current code does not check for this before using ip_top. As a result, when ip_top is NULL, the coredump worker crashes while taking the spinlock for ip_top->die_kset. Fix this by checking for a missing ip_top before walking the discovery topology. If it is unavailable, print a short message in the dump and return safely. - If ip_top is NULL, print a message and skip the dump - Also add the same check in the cleanup path This makes the coredump and cleanup paths safe even when the discovery topology is not available. KASAN trace: [ 522.228252] [IGT] amd_deadlock: starting subtest amdgpu-deadlock-sdma [ 522.240681] [IGT] amd_deadlock: starting dynamic subtest amdgpu-deadlock-sdma ... [ 522.952317] Write of size 4 at addr 0000000000000050 by task kworker/u129:5/5434 [ 522.937526] BUG: KASAN: null-ptr-deref in _raw_spin_lock+0x66/0xc0 [ 522.967659] Workqueue: events_unbound amdgpu_devcoredump_deferred_work [amdgpu] ... [ 522.969445] Call Trace: [ 522.969508] _raw_spin_lock+0x66/0xc0 [ 522.969518] ? __pfx__raw_spin_lock+0x10/0x10 [ 522.969534] amdgpu_discovery_dump+0x61/0x530 [amdgpu] [ 522.971346] ? pick_next_task_fair+0x3f6/0x1c60 [ 522.971363] amdgpu_devcoredump_format+0x84f/0x26f0 [amdgpu] [ 522.973188] ? __pfx_amdgpu_devcoredump_format+0x10/0x10 [amdgpu] [ 522.975012] ? psi_task_switch+0x2b5/0x9b0 [ 522.975027] ? __pfx___drm_printfn_coredump+0x10/0x10 [drm] [ 522.975198] ? __pfx___drm_puts_coredump+0x10/0x10 [drm] [ 522.975366] ? __schedule+0x113c/0x38d0 [ 522.975381] amdgpu_devcoredump_deferred_work+0x4c/0x1f0 [amdgpu] v2: Updated commit message - Clarified that ip_top is not freed, it can just be NULL if discovery was not initialized. (Christian/Lijo) v3: Removed the extra drm_warn() for sysfs init failure as sysfs already reports errors. (Christian) Fixes: e81eff80aad6 ("drm/amdgpu: include ip discovery data in devcoredump") Cc: Pierre-Eric Pelloux-Prayer Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 0eb0c62d2f4f..d42a9ff2f78e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -1394,6 +1394,9 @@ static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev) struct list_head *el, *tmp; struct kset *die_kset; + if (!ip_top) + return; + die_kset = &ip_top->die_kset; spin_lock(&die_kset->list_lock); list_for_each_prev_safe(el, tmp, &die_kset->list) { @@ -1418,9 +1421,15 @@ void amdgpu_discovery_dump(struct amdgpu_device *adev, struct drm_printer *p) struct ip_hw_instance *ip_inst; int i = 0, j; + drm_printf(p, "\nHW IP Discovery\n"); + + if (!ip_top) { + drm_printf(p, "ip discovery topology unavailable\n"); + return; + } + die_kset = &ip_top->die_kset; - drm_printf(p, "\nHW IP Discovery\n"); spin_lock(&die_kset->list_lock); list_for_each(el_die, &die_kset->list) { drm_printf(p, "die %d\n", i++); From fc61df1516178958d7e61ec52c05257d92e9f2e8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Mon, 2 Feb 2026 13:06:25 +0100 Subject: [PATCH 09/99] drm/amdgpu: annotate eviction fence signaling path MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Make sure lockdep sees the dependencies here. Signed-off-by: Christian König Reviewed-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c index 55e8b9cc2f9f..9ca21a32bc98 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c @@ -64,8 +64,17 @@ amdgpu_eviction_fence_suspend_worker(struct work_struct *work) container_of(evf_mgr, struct amdgpu_fpriv, evf_mgr); struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr; struct dma_fence *ev_fence; + bool cookie; mutex_lock(&uq_mgr->userq_mutex); + + /* + * This is intentionally after taking the userq_mutex since we do + * allocate memory while holding this lock, but only after ensuring that + * the eviction fence is signaled. + */ + cookie = dma_fence_begin_signalling(); + ev_fence = amdgpu_evf_mgr_get_fence(evf_mgr); amdgpu_userq_evict(uq_mgr, !evf_mgr->shutdown); @@ -75,6 +84,7 @@ amdgpu_eviction_fence_suspend_worker(struct work_struct *work) * next fence. */ dma_fence_signal(ev_fence); + dma_fence_end_signalling(cookie); dma_fence_put(ev_fence); mutex_unlock(&uq_mgr->userq_mutex); } From e495c688edf44e140db001dcdb12548093291b84 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Thu, 19 Mar 2026 11:10:10 +0530 Subject: [PATCH 10/99] drm/amd/ras: Add NULL checks for ras_core sys_fn callbacks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Some ras core helper functions access ras_core and its callback table (sys_fn) without validating them first. Cc: Tao Zhou Cc: YiPeng Chai Cc: Dan Carpenter Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: YiPeng Chai Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/ras/rascore/ras_core.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/drivers/gpu/drm/amd/ras/rascore/ras_core.c b/drivers/gpu/drm/amd/ras/rascore/ras_core.c index e889baec378b..d6229f5f9a31 100644 --- a/drivers/gpu/drm/amd/ras/rascore/ras_core.c +++ b/drivers/gpu/drm/amd/ras/rascore/ras_core.c @@ -116,6 +116,9 @@ bool ras_core_gpu_in_reset(struct ras_core_context *ras_core) { uint32_t status = 0; + if (!ras_core) + return false; + if (ras_core->sys_fn && ras_core->sys_fn->check_gpu_status) ras_core->sys_fn->check_gpu_status(ras_core, &status); @@ -127,6 +130,9 @@ bool ras_core_gpu_is_vf(struct ras_core_context *ras_core) { uint32_t status = 0; + if (!ras_core) + return false; + if (ras_core->sys_fn && ras_core->sys_fn->check_gpu_status) ras_core->sys_fn->check_gpu_status(ras_core, &status); @@ -479,6 +485,9 @@ int ras_core_handle_fatal_error(struct ras_core_context *ras_core) uint32_t ras_core_get_curr_nps_mode(struct ras_core_context *ras_core) { + if (!ras_core) + return 0; + if (ras_core->ras_nbio.ip_func && ras_core->ras_nbio.ip_func->get_memory_partition_mode) return ras_core->ras_nbio.ip_func->get_memory_partition_mode(ras_core); @@ -562,6 +571,8 @@ bool ras_core_ras_interrupt_detected(struct ras_core_context *ras_core) int ras_core_get_gpu_mem(struct ras_core_context *ras_core, enum gpu_mem_type mem_type, struct gpu_mem_block *gpu_mem) { + if (!ras_core || !gpu_mem) + return -EINVAL; if (ras_core->sys_fn && ras_core->sys_fn->get_gpu_mem) return ras_core->sys_fn->get_gpu_mem(ras_core, mem_type, gpu_mem); @@ -572,6 +583,8 @@ int ras_core_get_gpu_mem(struct ras_core_context *ras_core, int ras_core_put_gpu_mem(struct ras_core_context *ras_core, enum gpu_mem_type mem_type, struct gpu_mem_block *gpu_mem) { + if (!ras_core || !gpu_mem) + return -EINVAL; if (ras_core->sys_fn && ras_core->sys_fn->put_gpu_mem) return ras_core->sys_fn->put_gpu_mem(ras_core, mem_type, gpu_mem); From 2ccdefeec7e333e583d99c7619e8fa1d97dde013 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Thu, 19 Mar 2026 11:00:00 +0530 Subject: [PATCH 11/99] drm/amd/ras: Add input pointer validation in ras core helpers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add NULL checks for helper input/output pointers that are directly dereferenced, such as tm, seqno, dev_info and init_config. Cc: Tao Zhou Cc: YiPeng Chai Cc: Dan Carpenter Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: YiPeng Chai Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/ras/rascore/ras_core.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/drivers/gpu/drm/amd/ras/rascore/ras_core.c b/drivers/gpu/drm/amd/ras/rascore/ras_core.c index d6229f5f9a31..49b3aa7489ff 100644 --- a/drivers/gpu/drm/amd/ras/rascore/ras_core.c +++ b/drivers/gpu/drm/amd/ras/rascore/ras_core.c @@ -69,6 +69,9 @@ int ras_core_convert_timestamp_to_time(struct ras_core_context *ras_core, int seconds_per_minute = 60; int days, remaining_seconds; + if (!tm) + return -EINVAL; + days = div64_u64_rem(timestamp, seconds_per_day, &remainder); /* remainder will always be less than seconds_per_day. */ remaining_seconds = remainder; @@ -277,6 +280,9 @@ struct ras_core_context *ras_core_create(struct ras_core_config *init_config) struct ras_core_context *ras_core; struct ras_core_config *config; + if (!init_config) + return NULL; + ras_core = kzalloc_obj(*ras_core); if (!ras_core) return NULL; @@ -638,6 +644,9 @@ int ras_core_event_notify(struct ras_core_context *ras_core, int ras_core_get_device_system_info(struct ras_core_context *ras_core, struct device_system_info *dev_info) { + if (!dev_info) + return -EINVAL; + if (ras_core && ras_core->sys_fn && ras_core->sys_fn->get_device_system_info) return ras_core->sys_fn->get_device_system_info(ras_core, dev_info); From 197b69a5677a32b24abb641bf02eec5314ef7732 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Thu, 19 Mar 2026 16:27:19 +0530 Subject: [PATCH 12/99] drm/amdgpu: Skip discovery dump when topology is unavailable MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When generating a devcoredump, amdgpu_discovery_dump() prints the IP discovery topology. The function already needs to handle the case where adev->discovery.ip_top is NULL to avoid a crash. Currently, the code prints a section header and an additional message when the topology is unavailable. However, for platforms where discovery is not used, this section is not expected to be present. Printing an extra message adds unnecessary output. Simplify this by skipping the entire section when ip_top is NULL. The NULL check is kept to avoid a crash, but no output is generated when the discovery topology is unavailable. Cc: Pierre-Eric Pelloux-Prayer Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index d42a9ff2f78e..97fd66c578a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -1421,15 +1421,13 @@ void amdgpu_discovery_dump(struct amdgpu_device *adev, struct drm_printer *p) struct ip_hw_instance *ip_inst; int i = 0, j; - drm_printf(p, "\nHW IP Discovery\n"); - - if (!ip_top) { - drm_printf(p, "ip discovery topology unavailable\n"); + if (!ip_top) return; - } die_kset = &ip_top->die_kset; + drm_printf(p, "\nHW IP Discovery\n"); + spin_lock(&die_kset->list_lock); list_for_each(el_die, &die_kset->list) { drm_printf(p, "die %d\n", i++); From e471627d56272a791972f25e467348b611c31713 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 17 Mar 2026 16:34:41 -0400 Subject: [PATCH 13/99] drm/amdgpu/pm: drop SMU driver if version not matched messages It just leads to user confusion. Cc: Yang Wang Cc: Lijo Lazar Reviewed-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c | 1 - drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c | 1 - drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c | 1 - drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 4 ---- 4 files changed, 7 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c index 12b052d920f5..7ca8fdd23206 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c @@ -262,7 +262,6 @@ int smu_v11_0_check_fw_version(struct smu_context *smu) "smu fw program = %d, version = 0x%08x (%d.%d.%d)\n", smu->smc_driver_if_version, if_version, smu_program, smu_version, smu_major, smu_minor, smu_debug); - dev_info(smu->adev->dev, "SMU driver if version not matched\n"); } return ret; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c index 2c20624caca4..ac5e44dff6c9 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c @@ -101,7 +101,6 @@ int smu_v12_0_check_fw_version(struct smu_context *smu) "smu fw program = %d, smu fw version = 0x%08x (%d.%d.%d)\n", smu->smc_driver_if_version, if_version, smu_program, smu_version, smu_major, smu_minor, smu_debug); - dev_info(smu->adev->dev, "SMU driver if version not matched\n"); } return ret; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c index cec2df1ad0af..e38354c694c9 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c @@ -284,7 +284,6 @@ int smu_v14_0_check_fw_version(struct smu_context *smu) "smu fw program = %d, smu fw version = 0x%08x (%d.%d.%d)\n", smu->smc_driver_if_version, if_version, smu_program, smu_version, smu_major, smu_minor, smu_debug); - dev_info(adev->dev, "SMU driver if version not matched\n"); } return ret; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index a644579903f4..5ec2b4e6f3f3 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -1057,10 +1057,6 @@ int smu_cmn_check_fw_version(struct smu_context *smu) smu->smc_driver_if_version, if_version, smu_program, smu_version, smu_major, smu_minor, smu_debug); - if (smu->smc_driver_if_version != SMU_IGNORE_IF_VERSION && - if_version != smu->smc_driver_if_version) - dev_info(adev->dev, "SMU driver if version not matched\n"); - return 0; } From 355d96cdec5c61fd83f7eb54f1a28e38809645d6 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 19 Feb 2026 18:18:28 -0500 Subject: [PATCH 14/99] drm/amdgpu/gfx10: look at the right prop for gfx queue priority MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Look at hqd_queue_priority rather than hqd_pipe_priority. In practice, it didn't matter as both were always set for kernel queues, but that will change in the future. Fixes: b07d1d73b09e ("drm/amd/amdgpu: Enable high priority gfx queue") Reviewed-by：Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index 1893ceeeb26c..8b60299b73ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -6752,7 +6752,7 @@ static void gfx_v10_0_gfx_mqd_set_priority(struct amdgpu_device *adev, /* set up default queue priority level * 0x0 = low priority, 0x1 = high priority */ - if (prop->hqd_pipe_priority == AMDGPU_GFX_PIPE_PRIO_HIGH) + if (prop->hqd_queue_priority == AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM) priority = 1; tmp = RREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY); From f9a4e81bcbd04e6f967d851f9fe69d8bb3cc08b3 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 19 Feb 2026 18:20:27 -0500 Subject: [PATCH 15/99] drm/amdgpu/gfx11: look at the right prop for gfx queue priority MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Look at hqd_queue_priority rather than hqd_pipe_priority. In practice, it didn't matter as both were always set for kernel queues, but that will change in the future. Fixes: 2e216b1e6ba2 ("drm/amdgpu/gfx11: handle priority setup for gfx pipe1") Reviewed-by：Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index b1a1b8a10a08..78d1f3eb522e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -4088,7 +4088,7 @@ static void gfx_v11_0_gfx_mqd_set_priority(struct amdgpu_device *adev, /* set up default queue priority level * 0x0 = low priority, 0x1 = high priority */ - if (prop->hqd_pipe_priority == AMDGPU_GFX_PIPE_PRIO_HIGH) + if (prop->hqd_queue_priority == AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM) priority = 1; tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT; From 94d79f51efecb74be1d88dde66bdc8bfcca17935 Mon Sep 17 00:00:00 2001 From: Ruijing Dong Date: Tue, 17 Mar 2026 13:54:11 -0400 Subject: [PATCH 16/99] drm/amdgpu: fix strsep() corrupting lockup_timeout on multi-GPU (v3) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit amdgpu_device_get_job_timeout_settings() passes a pointer directly to the global amdgpu_lockup_timeout[] buffer into strsep(). strsep() destructively replaces delimiter characters with '\0' in-place. On multi-GPU systems, this function is called once per device. When a multi-value setting like "0,0,0,-1" is used, the first GPU's call transforms the global buffer into "0\00\00\0-1". The second GPU then sees only "0" (terminated at the first '\0'), parses a single value, hits the single-value fallthrough (index == 1), and applies timeout=0 to all rings — causing immediate false job timeouts. Fix this by copying into a stack-local array before calling strsep(), so the global module parameter buffer remains intact across calls. The buffer is AMDGPU_MAX_TIMEOUT_PARAM_LENGTH (256) bytes, which is safe for the stack. v2: wrap commit message to 72 columns, add Assisted-by tag. v3: use stack array with strscpy() instead of kstrdup()/kfree() to avoid unnecessary heap allocation (Christian). This patch was developed with assistance from Claude (claude-opus-4-6). Assisted-by: Claude:claude-opus-4-6 Reviewed-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Ruijing Dong Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index ac5769d9e75c..a7038f039b10 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3498,7 +3498,8 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work) static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev) { - char *input = amdgpu_lockup_timeout; + char buf[AMDGPU_MAX_TIMEOUT_PARAM_LENGTH]; + char *input = buf; char *timeout_setting = NULL; int index = 0; long timeout; @@ -3508,9 +3509,17 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev) adev->gfx_timeout = adev->compute_timeout = adev->sdma_timeout = adev->video_timeout = msecs_to_jiffies(2000); - if (!strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) + if (!strnlen(amdgpu_lockup_timeout, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) return 0; + /* + * strsep() destructively modifies its input by replacing delimiters + * with '\0'. Use a stack copy so the global module parameter buffer + * remains intact for multi-GPU systems where this function is called + * once per device. + */ + strscpy(buf, amdgpu_lockup_timeout, sizeof(buf)); + while ((timeout_setting = strsep(&input, ",")) && strnlen(timeout_setting, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) { ret = kstrtol(timeout_setting, 0, &timeout); From 8f1de51f49be692de137c8525106e0fce2d1912d Mon Sep 17 00:00:00 2001 From: Eric Huang Date: Mon, 16 Mar 2026 11:01:30 -0400 Subject: [PATCH 17/99] drm/amdgpu: prevent immediate PASID reuse case MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit PASID resue could cause interrupt issue when process immediately runs into hw state left by previous process exited with the same PASID, it's possible that page faults are still pending in the IH ring buffer when the process exits and frees up its PASID. To prevent the case, it uses idr cyclic allocator same as kernel pid's. Signed-off-by: Eric Huang Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 45 ++++++++++++++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 1 + 3 files changed, 34 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c index 64c519cd7395..d88523568b62 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c @@ -35,10 +35,13 @@ * PASIDs are global address space identifiers that can be shared * between the GPU, an IOMMU and the driver. VMs on different devices * may use the same PASID if they share the same address - * space. Therefore PASIDs are allocated using a global IDA. VMs are - * looked up from the PASID per amdgpu_device. + * space. Therefore PASIDs are allocated using IDR cyclic allocator + * (similar to kernel PID allocation) which naturally delays reuse. + * VMs are looked up from the PASID per amdgpu_device. */ -static DEFINE_IDA(amdgpu_pasid_ida); + +static DEFINE_IDR(amdgpu_pasid_idr); +static DEFINE_SPINLOCK(amdgpu_pasid_idr_lock); /* Helper to free pasid from a fence callback */ struct amdgpu_pasid_cb { @@ -50,8 +53,8 @@ struct amdgpu_pasid_cb { * amdgpu_pasid_alloc - Allocate a PASID * @bits: Maximum width of the PASID in bits, must be at least 1 * - * Allocates a PASID of the given width while keeping smaller PASIDs - * available if possible. + * Uses kernel's IDR cyclic allocator (same as PID allocation). + * Allocates sequentially with automatic wrap-around. * * Returns a positive integer on success. Returns %-EINVAL if bits==0. * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on @@ -59,14 +62,15 @@ struct amdgpu_pasid_cb { */ int amdgpu_pasid_alloc(unsigned int bits) { - int pasid = -EINVAL; + int pasid; - for (bits = min(bits, 31U); bits > 0; bits--) { - pasid = ida_alloc_range(&amdgpu_pasid_ida, 1U << (bits - 1), - (1U << bits) - 1, GFP_KERNEL); - if (pasid != -ENOSPC) - break; - } + if (bits == 0) + return -EINVAL; + + spin_lock(&amdgpu_pasid_idr_lock); + pasid = idr_alloc_cyclic(&amdgpu_pasid_idr, NULL, 1, + 1U << bits, GFP_KERNEL); + spin_unlock(&amdgpu_pasid_idr_lock); if (pasid >= 0) trace_amdgpu_pasid_allocated(pasid); @@ -81,7 +85,10 @@ int amdgpu_pasid_alloc(unsigned int bits) void amdgpu_pasid_free(u32 pasid) { trace_amdgpu_pasid_freed(pasid); - ida_free(&amdgpu_pasid_ida, pasid); + + spin_lock(&amdgpu_pasid_idr_lock); + idr_remove(&amdgpu_pasid_idr, pasid); + spin_unlock(&amdgpu_pasid_idr_lock); } static void amdgpu_pasid_free_cb(struct dma_fence *fence, @@ -616,3 +623,15 @@ void amdgpu_vmid_mgr_fini(struct amdgpu_device *adev) } } } + +/** + * amdgpu_pasid_mgr_cleanup - cleanup PASID manager + * + * Cleanup the IDR allocator. + */ +void amdgpu_pasid_mgr_cleanup(void) +{ + spin_lock(&amdgpu_pasid_idr_lock); + idr_destroy(&amdgpu_pasid_idr); + spin_unlock(&amdgpu_pasid_idr_lock); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h index b3649cd3af56..a57919478d3b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h @@ -74,6 +74,7 @@ int amdgpu_pasid_alloc(unsigned int bits); void amdgpu_pasid_free(u32 pasid); void amdgpu_pasid_free_delayed(struct dma_resv *resv, u32 pasid); +void amdgpu_pasid_mgr_cleanup(void); bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev, struct amdgpu_vmid *id); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 3f5712fc7216..ce5794d1f518 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2916,6 +2916,7 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) xa_destroy(&adev->vm_manager.pasids); amdgpu_vmid_mgr_fini(adev); + amdgpu_pasid_mgr_cleanup(); } /** From 71488869d933b49d4a7d4b2b20448b6f285327bb Mon Sep 17 00:00:00 2001 From: Meenakshikumar Somasundaram Date: Thu, 22 Jan 2026 11:33:10 -0500 Subject: [PATCH 18/99] drm/amd/display: Update dpia supported configuration [Why & How] Init a flag to track if dpia enabled previously and update that to boot options. Reviewed-by: Ovidiu Bunea Signed-off-by: Meenakshikumar Somasundaram Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dmub/src/dmub_dcn42.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn42.c b/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn42.c index 4d274b7034e8..7833a4fb7fbf 100644 --- a/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn42.c +++ b/drivers/gpu/drm/amd/display/dmub/src/dmub_dcn42.c @@ -39,13 +39,14 @@ void dmub_srv_dcn42_regs_init(struct dmub_srv *dmub, struct dc_context *ctx) void dmub_dcn42_enable_dmub_boot_options(struct dmub_srv *dmub, const struct dmub_srv_hw_params *params) { union dmub_fw_boot_options boot_options = {0}; - union dmub_fw_boot_options cur_boot_options = {0}; - cur_boot_options = dmub_dcn42_get_fw_boot_option(dmub); + if (!dmub->dpia_supported) { + dmub->dpia_supported = dmub_dcn42_get_fw_boot_option(dmub).bits.enable_dpia; + } boot_options.bits.z10_disable = params->disable_z10; boot_options.bits.dpia_supported = params->dpia_supported; - boot_options.bits.enable_dpia = cur_boot_options.bits.enable_dpia && !params->disable_dpia; + boot_options.bits.enable_dpia = dmub->dpia_supported && !params->disable_dpia; boot_options.bits.usb4_cm_version = params->usb4_cm_version; boot_options.bits.dpia_hpd_int_enable_supported = params->dpia_hpd_int_enable_supported; boot_options.bits.power_optimization = params->power_optimization; From 31ea7b0ef35e6d986cbff2a59e8f3d2cd2ae3810 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Mon, 2 Mar 2026 15:45:41 -0500 Subject: [PATCH 19/99] drm/amd/display: dcn42 don't round up disclk and dppclk [why] dml2 based on num_enabled clock != 2 to do clock ramming to dpm. apu has 8 levels dispclk/dppclk/dcfclk/fclk, but only 4 levels of memclk. to avoid mapping dispclk/dppclk to DPM clock, based on arch review, force dispclk/dppclk num_level as 2. Reviewed-by: Dillon Varone Signed-off-by: Charlene Liu Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../dcn42/dcn42_soc_and_ip_translator.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c index fd9c24b5df53..1c04171b296c 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c @@ -93,6 +93,10 @@ static void dcn42_convert_dc_clock_table_to_soc_bb_clock_table( } } vmin_limit->dispclk_khz = min(dc_clk_table->entries[0].dispclk_mhz * 1000, vmin_limit->dispclk_khz); + /* dispclk is always fine-grain */ + dml_clk_table->dispclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dispclk_levels >= 2 ? 2 : 1; + dml_clk_table->dispclk.clk_values_khz[0] = 0; + dml_clk_table->dispclk.clk_values_khz[1] = dc_clk_table->entries[dc_clk_table->num_entries_per_clk.num_dispclk_levels - 1].dispclk_mhz * 1000; } /* dppclk */ @@ -105,6 +109,10 @@ static void dcn42_convert_dc_clock_table_to_soc_bb_clock_table( dml_clk_table->dppclk.clk_values_khz[i] = 0; } } + /* dppclk is always fine-grain */ + dml_clk_table->dppclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dppclk_levels >= 2 ? 2 : 1; + dml_clk_table->dppclk.clk_values_khz[0] = 0; + dml_clk_table->dppclk.clk_values_khz[1] = dc_clk_table->entries[dc_clk_table->num_entries_per_clk.num_dppclk_levels - 1].dppclk_mhz * 1000; } /* dtbclk */ From 24ccc041f0a8100cc7aae822a3328b276039fe59 Mon Sep 17 00:00:00 2001 From: Nicholas Kazlauskas Date: Mon, 2 Mar 2026 15:02:33 -0500 Subject: [PATCH 20/99] drm/amd/display: Add MRQ programming for DCN42 [Why] DCN401 didn't have a MRQ present so these fields didn't exist. They are still present on DCN42 so we need to continue programming them like we did on DCN35 or we can block have poor meta requesting efficiency which blocks p-state. [How] Add `hubp42_program_requestor` which takes DML21 input and programs the registers like DCN35 and prior. Reviewed-by: Leo Chen Signed-off-by: Nicholas Kazlauskas Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../amd/display/dc/hubp/dcn42/dcn42_hubp.c | 35 ++++++++++++++++++- .../amd/display/dc/hubp/dcn42/dcn42_hubp.h | 6 ++++ 2 files changed, 40 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c index 0e33c739f459..d85a4ab957a4 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c @@ -245,6 +245,39 @@ static void hubp42_program_deadline( REFCYC_PER_VM_DMDATA, dlg_attr->refcyc_per_vm_dmdata); } +void hubp42_program_requestor( + struct hubp *hubp, + struct dml2_display_rq_regs *rq_regs) +{ + struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + + REG_UPDATE(HUBPRET_CONTROL, + DET_BUF_PLANE1_BASE_ADDRESS, rq_regs->plane1_base_address); + REG_SET_4(DCN_EXPANSION_MODE, 0, + DRQ_EXPANSION_MODE, rq_regs->drq_expansion_mode, + PRQ_EXPANSION_MODE, rq_regs->prq_expansion_mode, + MRQ_EXPANSION_MODE, rq_regs->mrq_expansion_mode, + CRQ_EXPANSION_MODE, rq_regs->crq_expansion_mode); + REG_SET_8(DCHUBP_REQ_SIZE_CONFIG, 0, + CHUNK_SIZE, rq_regs->rq_regs_l.chunk_size, + MIN_CHUNK_SIZE, rq_regs->rq_regs_l.min_chunk_size, + META_CHUNK_SIZE, rq_regs->rq_regs_l.meta_chunk_size, + MIN_META_CHUNK_SIZE, rq_regs->rq_regs_l.min_meta_chunk_size, + DPTE_GROUP_SIZE, rq_regs->rq_regs_l.dpte_group_size, + VM_GROUP_SIZE, rq_regs->rq_regs_l.mpte_group_size, + SWATH_HEIGHT, rq_regs->rq_regs_l.swath_height, + PTE_ROW_HEIGHT_LINEAR, rq_regs->rq_regs_l.pte_row_height_linear); + REG_SET_7(DCHUBP_REQ_SIZE_CONFIG_C, 0, + CHUNK_SIZE_C, rq_regs->rq_regs_c.chunk_size, + MIN_CHUNK_SIZE_C, rq_regs->rq_regs_c.min_chunk_size, + META_CHUNK_SIZE_C, rq_regs->rq_regs_c.meta_chunk_size, + MIN_META_CHUNK_SIZE_C, rq_regs->rq_regs_c.min_meta_chunk_size, + DPTE_GROUP_SIZE_C, rq_regs->rq_regs_c.dpte_group_size, + SWATH_HEIGHT_C, rq_regs->rq_regs_c.swath_height, + PTE_ROW_HEIGHT_LINEAR_C, rq_regs->rq_regs_c.pte_row_height_linear); +} + + void hubp42_setup( struct hubp *hubp, struct dml2_dchub_per_pipe_register_set *pipe_regs, @@ -255,7 +288,7 @@ void hubp42_setup( * disable the requestors is not needed */ hubp401_vready_at_or_After_vsync(hubp, pipe_global_sync, timing); - hubp401_program_requestor(hubp, &pipe_regs->rq_regs); + hubp42_program_requestor(hubp, &pipe_regs->rq_regs); hubp42_program_deadline(hubp, &pipe_regs->dlg_regs, &pipe_regs->ttu_regs); } static void hubp42_program_surface_config( diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h index 486c8907413a..88bb1337ab9d 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h @@ -48,6 +48,8 @@ HUBP_SF(CURSOR0_0_HUBP_3DLUT_ADDRESS_LOW, HUBP_3DLUT_ADDRESS_LOW, mask_sh),\ HUBP_SF(CURSOR0_0_HUBP_3DLUT_DLG_PARAM, REFCYC_PER_3DLUT_GROUP, mask_sh) +struct dml2_display_rq_regs; + bool hubp42_construct( struct dcn20_hubp *hubp2, struct dc_context *ctx, @@ -64,6 +66,10 @@ void hubp42_program_3dlut_fl_config(struct hubp *hubp, void hubp42_read_state(struct hubp *hubp); +void hubp42_program_requestor( + struct hubp *hubp, + struct dml2_display_rq_regs *rq_regs); + void hubp42_setup( struct hubp *hubp, struct dml2_dchub_per_pipe_register_set *pipe_regs, From 058cb97b730f2b47f168d843cd82832bca6babf1 Mon Sep 17 00:00:00 2001 From: Ivan Lipski Date: Tue, 3 Mar 2026 20:07:58 -0500 Subject: [PATCH 21/99] drm/amd/display: Move DPM clk read to clk_mgr_construct in DCN42 [Why&How] The DPM clocks on DCN42 are currently read on every dm_resume, which can cause in gpu memory freeing while the device is still in suspend. Move the DPM clock read functionality to clk_mgr_construct() so it completes once on driver enablement. Reviewed-by: Charlene Liu Signed-off-by: Ivan Lipski Signed-off-by: Dmytro Laktyushkin Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 250 +++++++++--------- 1 file changed, 126 insertions(+), 124 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index 97c9f0ce13e4..df904b6fafe9 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -569,7 +569,6 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) { struct clk_mgr_internal *clk_mgr_int = TO_CLK_MGR_INTERNAL(clk_mgr_base); struct clk_mgr_dcn42 *clk_mgr = TO_CLK_MGR_DCN42(clk_mgr_int); - struct dcn42_smu_dpm_clks smu_dpm_clks = { 0 }; DC_LOGGER_INIT(clk_mgr_base->ctx->logger); (void)dc_logger; @@ -591,127 +590,6 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) /*dtbclk enabled based on*/ clk_mgr_base->clks.dtbclk_en = true; } - - if (clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels != 0) { - /*skip to get clock table and notify pmfw watermark range again*/ - DC_LOG_SMU("skip to get dpm_clks from pmfw from resume and acr\n"); - return; - } - - smu_dpm_clks.dpm_clks = (DpmClocks_t_dcn42 *)dm_helpers_allocate_gpu_mem( - clk_mgr_base->ctx, - DC_MEM_ALLOC_TYPE_GART, - sizeof(DpmClocks_t_dcn42), - &smu_dpm_clks.mc_address.quad_part); - - ASSERT(smu_dpm_clks.dpm_clks); - if (clk_mgr_base->ctx->dc->debug.pstate_enabled && clk_mgr_int->smu_present && smu_dpm_clks.mc_address.quad_part != 0) { - int i; - DpmClocks_t_dcn42 *dpm_clks = smu_dpm_clks.dpm_clks; - - dcn42_get_dpm_table_from_smu(clk_mgr_int, &smu_dpm_clks); - DC_LOG_SMU("NumDcfClkLevelsEnabled: %d\n" - "NumDispClkLevelsEnabled: %d\n" - "NumSocClkLevelsEnabled: %d\n" - "VcnClkLevelsEnabled: %d\n" - "FClkLevelsEnabled: %d\n" - "NumMemPstatesEnabled: %d\n" - "MinGfxClk: %d\n" - "MaxGfxClk: %d\n", - dpm_clks->NumDcfClkLevelsEnabled, - dpm_clks->NumDispClkLevelsEnabled, - dpm_clks->NumSocClkLevelsEnabled, - dpm_clks->VcnClkLevelsEnabled, - dpm_clks->NumFclkLevelsEnabled, - dpm_clks->NumMemPstatesEnabled, - dpm_clks->MinGfxClk, - dpm_clks->MaxGfxClk); - - for (i = 0; i < NUM_DCFCLK_DPM_LEVELS; i++) { - DC_LOG_SMU("dpm_clks->DcfClocks[%d] = %d\n", - i, - dpm_clks->DcfClocks[i]); - } - for (i = 0; i < NUM_DISPCLK_DPM_LEVELS; i++) { - DC_LOG_SMU("dpm_clks->DispClocks[%d] = %d\n", - i, dpm_clks->DispClocks[i]); - } - for (i = 0; i < NUM_SOCCLK_DPM_LEVELS; i++) { - DC_LOG_SMU("dpm_clks->SocClocks[%d] = %d\n", - i, dpm_clks->SocClocks[i]); - } - for (i = 0; i < NUM_FCLK_DPM_LEVELS; i++) { - DC_LOG_SMU("dpm_clks->FclkClocks_Freq[%d] = %d\n", - i, dpm_clks->FclkClocks_Freq[i]); - DC_LOG_SMU("dpm_clks->FclkClocks_Voltage[%d] = %d\n", - i, dpm_clks->FclkClocks_Voltage[i]); - } - for (i = 0; i < NUM_SOCCLK_DPM_LEVELS; i++) - DC_LOG_SMU("dpm_clks->SocVoltage[%d] = %d\n", - i, dpm_clks->SocVoltage[i]); - - for (i = 0; i < NUM_MEM_PSTATE_LEVELS; i++) { - DC_LOG_SMU("dpm_clks.MemPstateTable[%d].UClk = %d\n" - "dpm_clks->MemPstateTable[%d].MemClk= %d\n" - "dpm_clks->MemPstateTable[%d].Voltage = %d\n", - i, dpm_clks->MemPstateTable[i].UClk, - i, dpm_clks->MemPstateTable[i].MemClk, - i, dpm_clks->MemPstateTable[i].Voltage); - } - - if (clk_mgr_base->ctx->dc_bios->integrated_info && clk_mgr_base->ctx->dc->config.use_default_clock_table == false) { - /* DCFCLK */ - dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].dcfclk_mhz, - dpm_clks->DcfClocks, - dpm_clks->NumDcfClkLevelsEnabled); - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels = dpm_clks->NumDcfClkLevelsEnabled; - - /* SOCCLK */ - dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].socclk_mhz, - dpm_clks->SocClocks, - dpm_clks->NumSocClkLevelsEnabled); - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_socclk_levels = dpm_clks->NumSocClkLevelsEnabled; - - /* DISPCLK */ - dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].dispclk_mhz, - dpm_clks->DispClocks, - dpm_clks->NumDispClkLevelsEnabled); - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dispclk_levels = dpm_clks->NumDispClkLevelsEnabled; - - /* DPPCLK */ - dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].dppclk_mhz, - dpm_clks->DppClocks, - dpm_clks->NumDispClkLevelsEnabled); - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dppclk_levels = dpm_clks->NumDispClkLevelsEnabled; - - /* FCLK */ - dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].fclk_mhz, - dpm_clks->FclkClocks_Freq, - NUM_FCLK_DPM_LEVELS); - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_fclk_levels = dpm_clks->NumFclkLevelsEnabled; - clk_mgr_base->bw_params->clk_table.num_entries = dpm_clks->NumFclkLevelsEnabled; - - /* Memory Pstate table is in reverse order*/ - ASSERT(dpm_clks->NumMemPstatesEnabled <= NUM_MEM_PSTATE_LEVELS); - if (dpm_clks->NumMemPstatesEnabled > NUM_MEM_PSTATE_LEVELS) - dpm_clks->NumMemPstatesEnabled = NUM_MEM_PSTATE_LEVELS; - for (i = 0; i < dpm_clks->NumMemPstatesEnabled; i++) { - clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].memclk_mhz = dpm_clks->MemPstateTable[i].UClk; - clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].wck_ratio = dcn42_convert_wck_ratio(dpm_clks->MemPstateTable[i].WckRatio) ; - } - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels = dpm_clks->NumMemPstatesEnabled; - - /* DTBCLK*/ - clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz = clk_mgr_base->clks.ref_dtbclk_khz / 1000; - clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dtbclk_levels = 1; - /* Refresh bounding box */ - clk_mgr_base->ctx->dc->res_pool->funcs->update_bw_bounding_box( - clk_mgr_base->ctx->dc, clk_mgr_base->bw_params); - } - } - if (smu_dpm_clks.dpm_clks && smu_dpm_clks.mc_address.quad_part != 0) - dm_helpers_free_gpu_mem(clk_mgr_base->ctx, DC_MEM_ALLOC_TYPE_GART, - smu_dpm_clks.dpm_clks); } static struct clk_bw_params dcn42_bw_params = { @@ -1071,6 +949,127 @@ bool dcn42_is_smu_present(struct clk_mgr *clk_mgr_base) return clk_mgr->smu_present; } +static void dcn42_get_smu_clocks(struct clk_mgr_internal *clk_mgr_int) +{ + struct clk_mgr *clk_mgr_base = &clk_mgr_int->base; + struct dcn42_smu_dpm_clks smu_dpm_clks = { 0 }; + + DC_LOGGER_INIT(clk_mgr_base->ctx->logger); + (void)dc_logger; + + smu_dpm_clks.dpm_clks = (DpmClocks_t_dcn42 *)dm_helpers_allocate_gpu_mem( + clk_mgr_base->ctx, + DC_MEM_ALLOC_TYPE_GART, + sizeof(DpmClocks_t_dcn42), + &smu_dpm_clks.mc_address.quad_part); + + ASSERT(smu_dpm_clks.dpm_clks); + if (clk_mgr_base->ctx->dc->debug.pstate_enabled && smu_dpm_clks.mc_address.quad_part != 0) { + int i; + DpmClocks_t_dcn42 *dpm_clks = smu_dpm_clks.dpm_clks; + + dcn42_get_dpm_table_from_smu(clk_mgr_int, &smu_dpm_clks); + DC_LOG_SMU("NumDcfClkLevelsEnabled: %d\n" + "NumDispClkLevelsEnabled: %d\n" + "NumSocClkLevelsEnabled: %d\n" + "VcnClkLevelsEnabled: %d\n" + "FClkLevelsEnabled: %d\n" + "NumMemPstatesEnabled: %d\n" + "MinGfxClk: %d\n" + "MaxGfxClk: %d\n", + dpm_clks->NumDcfClkLevelsEnabled, + dpm_clks->NumDispClkLevelsEnabled, + dpm_clks->NumSocClkLevelsEnabled, + dpm_clks->VcnClkLevelsEnabled, + dpm_clks->NumFclkLevelsEnabled, + dpm_clks->NumMemPstatesEnabled, + dpm_clks->MinGfxClk, + dpm_clks->MaxGfxClk); + + for (i = 0; i < NUM_DCFCLK_DPM_LEVELS; i++) { + DC_LOG_SMU("dpm_clks->DcfClocks[%d] = %d\n", + i, + dpm_clks->DcfClocks[i]); + } + for (i = 0; i < NUM_DISPCLK_DPM_LEVELS; i++) { + DC_LOG_SMU("dpm_clks->DispClocks[%d] = %d\n", + i, dpm_clks->DispClocks[i]); + } + for (i = 0; i < NUM_SOCCLK_DPM_LEVELS; i++) { + DC_LOG_SMU("dpm_clks->SocClocks[%d] = %d\n", + i, dpm_clks->SocClocks[i]); + } + for (i = 0; i < NUM_FCLK_DPM_LEVELS; i++) { + DC_LOG_SMU("dpm_clks->FclkClocks_Freq[%d] = %d\n", + i, dpm_clks->FclkClocks_Freq[i]); + DC_LOG_SMU("dpm_clks->FclkClocks_Voltage[%d] = %d\n", + i, dpm_clks->FclkClocks_Voltage[i]); + } + for (i = 0; i < NUM_SOCCLK_DPM_LEVELS; i++) + DC_LOG_SMU("dpm_clks->SocVoltage[%d] = %d\n", + i, dpm_clks->SocVoltage[i]); + + for (i = 0; i < NUM_MEM_PSTATE_LEVELS; i++) { + DC_LOG_SMU("dpm_clks.MemPstateTable[%d].UClk = %d\n" + "dpm_clks->MemPstateTable[%d].MemClk= %d\n" + "dpm_clks->MemPstateTable[%d].Voltage = %d\n", + i, dpm_clks->MemPstateTable[i].UClk, + i, dpm_clks->MemPstateTable[i].MemClk, + i, dpm_clks->MemPstateTable[i].Voltage); + } + + if (clk_mgr_base->ctx->dc_bios->integrated_info && clk_mgr_base->ctx->dc->config.use_default_clock_table == false) { + /* DCFCLK */ + dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].dcfclk_mhz, + dpm_clks->DcfClocks, + dpm_clks->NumDcfClkLevelsEnabled); + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels = dpm_clks->NumDcfClkLevelsEnabled; + + /* SOCCLK */ + dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].socclk_mhz, + dpm_clks->SocClocks, + dpm_clks->NumSocClkLevelsEnabled); + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_socclk_levels = dpm_clks->NumSocClkLevelsEnabled; + + /* DISPCLK */ + dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].dispclk_mhz, + dpm_clks->DispClocks, + dpm_clks->NumDispClkLevelsEnabled); + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dispclk_levels = dpm_clks->NumDispClkLevelsEnabled; + + /* DPPCLK */ + dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].dppclk_mhz, + dpm_clks->DppClocks, + dpm_clks->NumDispClkLevelsEnabled); + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dppclk_levels = dpm_clks->NumDispClkLevelsEnabled; + + /* FCLK */ + dcn42_init_single_clock(&clk_mgr_base->bw_params->clk_table.entries[0].fclk_mhz, + dpm_clks->FclkClocks_Freq, + NUM_FCLK_DPM_LEVELS); + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_fclk_levels = dpm_clks->NumFclkLevelsEnabled; + clk_mgr_base->bw_params->clk_table.num_entries = dpm_clks->NumFclkLevelsEnabled; + + /* Memory Pstate table is in reverse order*/ + ASSERT(dpm_clks->NumMemPstatesEnabled <= NUM_MEM_PSTATE_LEVELS); + if (dpm_clks->NumMemPstatesEnabled > NUM_MEM_PSTATE_LEVELS) + dpm_clks->NumMemPstatesEnabled = NUM_MEM_PSTATE_LEVELS; + for (i = 0; i < dpm_clks->NumMemPstatesEnabled; i++) { + clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].memclk_mhz = dpm_clks->MemPstateTable[i].UClk; + clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].wck_ratio = dcn42_convert_wck_ratio(dpm_clks->MemPstateTable[i].WckRatio) ; + } + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels = dpm_clks->NumMemPstatesEnabled; + + /* DTBCLK*/ + clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz = clk_mgr_base->clks.ref_dtbclk_khz / 1000; + clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dtbclk_levels = 1; + } + } + if (smu_dpm_clks.dpm_clks && smu_dpm_clks.mc_address.quad_part != 0) + dm_helpers_free_gpu_mem(clk_mgr_base->ctx, DC_MEM_ALLOC_TYPE_GART, + smu_dpm_clks.dpm_clks); +} + static struct clk_mgr_funcs dcn42_funcs = { .get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz, .get_dtb_ref_clk_frequency = dcn31_get_dtb_ref_freq_khz, @@ -1139,6 +1138,11 @@ void dcn42_clk_mgr_construct( dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 1; clk_mgr->base.base.dprefclk_khz = dcn42_smu_get_dprefclk(&clk_mgr->base); clk_mgr->base.base.clks.ref_dtbclk_khz = dcn42_smu_get_dtbclk(&clk_mgr->base); + + clk_mgr->base.base.bw_params = &dcn42_bw_params; + + if (clk_mgr->base.smu_present) + dcn42_get_smu_clocks(&clk_mgr->base); } /* in case we don't get a value from the BIOS, use default */ if (clk_mgr->base.base.dentist_vco_freq_khz == 0) @@ -1151,8 +1155,6 @@ void dcn42_clk_mgr_construct( /*when clk src is from FCH, it could have ss, same clock src as DPREF clk*/ dcn42_read_ss_info_from_lut(&clk_mgr->base); - - clk_mgr->base.base.bw_params = &dcn42_bw_params; } void dcn42_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int) From 6c36d8b0c93f18c97d26f92d13945ddc56e4fd86 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Thu, 5 Mar 2026 10:14:39 -0500 Subject: [PATCH 22/99] drm/amd/display: System Hang When System enters to S0i3 w/ iGPU [why] System Hang when system enters to S0i3 w/ iGPU some link_enc are NULL due to BIOS integration info table not correct, but driver should have enough null pointer protection. Reviewed-by: Leo Chen Signed-off-by: Charlene Liu Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c index 8e12dc1297c4..7bf94eae30a6 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c @@ -203,7 +203,8 @@ void dcn42_init_hw(struct dc *dc) for (i = 0; i < dc->link_count; i++) { struct dc_link *link = dc->links[i]; - if (link->link_enc->funcs->is_dig_enabled && + if (link && link->link_enc && + link->link_enc->funcs->is_dig_enabled && link->link_enc->funcs->is_dig_enabled(link->link_enc) && hws->funcs.power_down) { hws->funcs.power_down(dc); From 418e0eaf72caed1938ae4507494f239b8941a8ef Mon Sep 17 00:00:00 2001 From: Dmytro Laktyushkin Date: Thu, 5 Mar 2026 14:47:03 -0500 Subject: [PATCH 23/99] drm/amd/display: move dcn42 bw_params init Move it out of smu present block for cases where it isn't Reviewed-by: Ivan Lipski Signed-off-by: Dmytro Laktyushkin Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index df904b6fafe9..a0296d5f0102 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -1138,11 +1138,6 @@ void dcn42_clk_mgr_construct( dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 1; clk_mgr->base.base.dprefclk_khz = dcn42_smu_get_dprefclk(&clk_mgr->base); clk_mgr->base.base.clks.ref_dtbclk_khz = dcn42_smu_get_dtbclk(&clk_mgr->base); - - clk_mgr->base.base.bw_params = &dcn42_bw_params; - - if (clk_mgr->base.smu_present) - dcn42_get_smu_clocks(&clk_mgr->base); } /* in case we don't get a value from the BIOS, use default */ if (clk_mgr->base.base.dentist_vco_freq_khz == 0) @@ -1155,6 +1150,10 @@ void dcn42_clk_mgr_construct( /*when clk src is from FCH, it could have ss, same clock src as DPREF clk*/ dcn42_read_ss_info_from_lut(&clk_mgr->base); + + clk_mgr->base.base.bw_params = &dcn42_bw_params; + if (clk_mgr->base.smu_present) + dcn42_get_smu_clocks(&clk_mgr->base); } void dcn42_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int) From b6c0783ff278671e38fed978fefb732101ac8836 Mon Sep 17 00:00:00 2001 From: Roman Li Date: Fri, 13 Mar 2026 19:34:02 -0400 Subject: [PATCH 24/99] drm/amd/display: Add get_default_tiling_info for dcn42 Add DCN42 portion that was stripped during previously. Fixes: 8333f22e44a9 ("drm/amd/display: Query DC for gfx handling when setting linear tiling") Reviewed-by: Alex Hung Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c index aef187bcf5c3..7b451c7db02c 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c @@ -1792,6 +1792,7 @@ static struct resource_funcs dcn42_res_pool_funcs = { .build_pipe_pix_clk_params = dcn42_build_pipe_pix_clk_params, .get_vstartup_for_pipe = dcn401_get_vstartup_for_pipe, .get_max_hw_cursor_size = dcn42_get_max_hw_cursor_size, + .get_default_tiling_info = dcn10_get_default_tiling_info }; static uint32_t read_pipe_fuses(struct dc_context *ctx) From c446748aa3f233451955bf63ac9887047fe13e73 Mon Sep 17 00:00:00 2001 From: Roman Li Date: Fri, 13 Mar 2026 20:27:54 -0400 Subject: [PATCH 25/99] drm/amd/display: Add missing dcn42 hubbub function pointers This aligning commit combines: - fix dcn42 det programming) - fix missing dcn42 pointers - fix SDPIF_Request_Rate_Limit programming value V2: Add back dchvm_init for DCN42 Reviewed-by: Alex Hung Reviewed-by: Leo Chen Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../display/dc/hubbub/dcn42/dcn42_hubbub.c | 22 ++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c b/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c index d6e6fbaa041b..9e4d526b6d52 100644 --- a/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c +++ b/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c @@ -6,6 +6,7 @@ #include "dcn31/dcn31_hubbub.h" #include "dcn32/dcn32_hubbub.h" #include "dcn35/dcn35_hubbub.h" +#include "dcn401/dcn401_hubbub.h" #include "dcn42/dcn42_hubbub.h" #include "dm_services.h" #include "reg_helper.h" @@ -494,6 +495,14 @@ static bool hubbub42_program_watermarks( return wm_pending; } +static void hubbub42_set_request_limit(struct hubbub *hubbub, int memory_channel_count, int words_per_channel) +{ + struct dcn20_hubbub *hubbub2 = TO_DCN20_HUBBUB(hubbub); + uint32_t request_limit = 96; //MAX(12 * memory_channel_count, 96); + + REG_UPDATE(SDPIF_REQUEST_RATE_LIMIT, SDPIF_REQUEST_RATE_LIMIT, request_limit); +} + static const struct hubbub_funcs hubbub42_funcs = { .update_dchub = hubbub2_update_dchub, .init_dchub_sys_ctx = hubbub31_init_dchub_sys_ctx, @@ -509,13 +518,16 @@ static const struct hubbub_funcs hubbub42_funcs = { .force_wm_propagate_to_pipes = hubbub32_force_wm_propagate_to_pipes, .force_pstate_change_control = hubbub3_force_pstate_change_control, .init_watermarks = hubbub35_init_watermarks, - .program_det_size = dcn32_program_det_size, - .program_compbuf_size = dcn35_program_compbuf_size, - .init_crb = dcn35_init_crb, + .init_crb = dcn401_init_crb, + .dchvm_init = dcn35_dchvm_init, .hubbub_read_state = hubbub2_read_state, .force_usr_retraining_allow = hubbub32_force_usr_retraining_allow, - .dchubbub_init = hubbub35_init, - .dchvm_init = dcn35_dchvm_init, + .set_request_limit = hubbub42_set_request_limit, + .program_det_segments = dcn401_program_det_segments, + .program_compbuf_segments = dcn401_program_compbuf_segments, + .wait_for_det_update = dcn401_wait_for_det_update, + .program_arbiter = dcn401_program_arbiter, + .hubbub_read_reg_state = hubbub3_read_reg_state }; void hubbub42_construct(struct dcn20_hubbub *hubbub2, From cb50faeb0d7585fed52231627d632462b9f927b0 Mon Sep 17 00:00:00 2001 From: Nicholas Kazlauskas Date: Mon, 2 Mar 2026 10:47:34 -0500 Subject: [PATCH 26/99] drm/amd/display: Split arbiter programming for DCN42 [Why] We don't want to update the timeout threshold for stall recovery in firmware dynamically for DCN42 as we're not using FAMS. Firmware should own programming of this register since the recovery can be broken if driver updates the value to 0. [How] Split program_arbiter for dcn42 and skip the part that updates the timeout threshold. Reviewed-by: Leo Chen Signed-off-by: Nicholas Kazlauskas Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../display/dc/hubbub/dcn42/dcn42_hubbub.c | 34 ++++++++++++++++++- 1 file changed, 33 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c b/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c index 9e4d526b6d52..8582dcab1f22 100644 --- a/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c +++ b/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c @@ -503,6 +503,38 @@ static void hubbub42_set_request_limit(struct hubbub *hubbub, int memory_channel REG_UPDATE(SDPIF_REQUEST_RATE_LIMIT, SDPIF_REQUEST_RATE_LIMIT, request_limit); } +static bool dcn42_program_arbiter(struct hubbub *hubbub, struct dml2_display_arb_regs *arb_regs, + bool safe_to_lower) +{ + struct dcn20_hubbub *hubbub2 = TO_DCN20_HUBBUB(hubbub); + + bool wm_pending = false; + uint32_t temp; + + /* request backpressure and outstanding return threshold (unused)*/ + //REG_UPDATE(DCHUBBUB_TIMEOUT_DETECTION_CTRL1, DCHUBBUB_TIMEOUT_REQ_STALL_THRESHOLD, arb_regs->req_stall_threshold); + + /* 401 delta: do not update P-State stall threshold (handled by fw) */ + // REG_UPDATE(DCHUBBUB_TIMEOUT_DETECTION_CTRL2, DCHUBBUB_TIMEOUT_PSTATE_STALL_THRESHOLD, arb_regs->pstate_stall_threshold); + + if (safe_to_lower || arb_regs->allow_sdpif_rate_limit_when_cstate_req > hubbub2->allow_sdpif_rate_limit_when_cstate_req) { + hubbub2->allow_sdpif_rate_limit_when_cstate_req = arb_regs->allow_sdpif_rate_limit_when_cstate_req; + + /* only update the required bits */ + REG_GET(DCHUBBUB_CTRL_STATUS, DCHUBBUB_HW_DEBUG, &temp); + if (hubbub2->allow_sdpif_rate_limit_when_cstate_req) { + temp |= (1 << 5); + } else { + temp &= ~(1 << 5); + } + REG_UPDATE(DCHUBBUB_CTRL_STATUS, DCHUBBUB_HW_DEBUG, temp); + } else { + wm_pending = true; + } + + return wm_pending; +} + static const struct hubbub_funcs hubbub42_funcs = { .update_dchub = hubbub2_update_dchub, .init_dchub_sys_ctx = hubbub31_init_dchub_sys_ctx, @@ -526,7 +558,7 @@ static const struct hubbub_funcs hubbub42_funcs = { .program_det_segments = dcn401_program_det_segments, .program_compbuf_segments = dcn401_program_compbuf_segments, .wait_for_det_update = dcn401_wait_for_det_update, - .program_arbiter = dcn401_program_arbiter, + .program_arbiter = dcn42_program_arbiter, .hubbub_read_reg_state = hubbub3_read_reg_state }; From 5a90bb2160394fc2fb835ad8142891aea194fe3c Mon Sep 17 00:00:00 2001 From: Roman Li Date: Fri, 13 Mar 2026 20:34:48 -0400 Subject: [PATCH 27/99] drm/amd/display: Clamp min DS DCFCLK value to DCN limit [why & how] DCN has a global limit for minimum DS DCFCLK during any operation. Adhere to that limit and add a debug flag. Reviewed-by: Charlene Liu Signed-off-by: Ovidiu Bunea Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 5 +++++ drivers/gpu/drm/amd/display/dc/dc.h | 1 + .../gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c | 1 + 3 files changed, 7 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index a0296d5f0102..bc11510b63a1 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -291,6 +291,11 @@ void dcn42_update_clocks(struct clk_mgr *clk_mgr_base, if (should_set_clock(safe_to_lower, new_clocks->dcfclk_deep_sleep_khz, clk_mgr_base->clks.dcfclk_deep_sleep_khz)) { clk_mgr_base->clks.dcfclk_deep_sleep_khz = new_clocks->dcfclk_deep_sleep_khz; + + /* Clamp the requested clock to PMFW based on DCN limit. */ + if (dc->debug.min_deep_sleep_dcfclk_khz > 0 && clk_mgr_base->clks.dcfclk_deep_sleep_khz < dc->debug.min_deep_sleep_dcfclk_khz) + clk_mgr_base->clks.dcfclk_deep_sleep_khz = dc->debug.min_deep_sleep_dcfclk_khz; + dcn42_smu_set_min_deep_sleep_dcfclk(clk_mgr, clk_mgr_base->clks.dcfclk_deep_sleep_khz); } diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 0b48feaba131..4d15d97ed7f1 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -1215,6 +1215,7 @@ struct dc_debug_options { bool enable_dmu_recovery; unsigned int force_vmin_threshold; bool enable_otg_frame_sync_pwa; + unsigned int min_deep_sleep_dcfclk_khz; }; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c index 7b451c7db02c..8175109a66b0 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c @@ -760,6 +760,7 @@ static const struct dc_debug_options debug_defaults_drv = { .disable_z10 = false, .ignore_pg = true, .disable_stutter_for_wm_program = true, + .min_deep_sleep_dcfclk_khz = 8000, }; static const struct dc_check_config config_defaults = { From 68bd4f6b8310f309eb63b41e15088690c9cec0a9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Tue, 3 Feb 2026 17:30:57 +0100 Subject: [PATCH 28/99] drm/amdgpu: fix some more bug in amdgpu_gem_va_ioctl MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Some illegal combination of input flags were not checked and we need to take the PDEs into account when returning the fence as well. Signed-off-by: Christian König Acked-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 78 +++++++++++-------------- 1 file changed, 35 insertions(+), 43 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 01dc73309d73..4ebb99e22ff2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -30,6 +30,7 @@ #include #include #include +#include #include #include @@ -741,11 +742,10 @@ amdgpu_gem_va_update_vm(struct amdgpu_device *adev, struct dma_fence *fence; int r = 0; - /* Always start from the VM's existing last update fence. */ - fence = dma_fence_get(vm->last_update); - + /* If the VM is not ready return only a stub. */ if (!amdgpu_vm_ready(vm)) - return fence; + return dma_fence_get_stub(); + /* * First clean up any freed mappings in the VM. @@ -754,7 +754,7 @@ amdgpu_gem_va_update_vm(struct amdgpu_device *adev, * schedules GPU work. If nothing needs clearing, @fence can remain as * the original vm->last_update. */ - r = amdgpu_vm_clear_freed(adev, vm, &fence); + r = amdgpu_vm_clear_freed(adev, vm, &vm->last_update); if (r) goto error; @@ -771,47 +771,34 @@ amdgpu_gem_va_update_vm(struct amdgpu_device *adev, if (r) goto error; - /* - * Decide which fence best represents the last update: - * - * MAP/REPLACE: - * - For always-valid mappings, use vm->last_update. - * - Otherwise, export bo_va->last_pt_update. - * - * UNMAP/CLEAR: - * Keep the fence returned by amdgpu_vm_clear_freed(). If no work was - * needed, it can remain as vm->last_pt_update. - * - * The VM and BO update fences are always initialized to a valid value. - * vm->last_update and bo_va->last_pt_update always start as valid fences. - * and are never expected to be NULL. - */ - switch (operation) { - case AMDGPU_VA_OP_MAP: - case AMDGPU_VA_OP_REPLACE: - /* - * For MAP/REPLACE, return the page table update fence for the - * mapping we just modified. bo_va is expected to be valid here. - */ - dma_fence_put(fence); + if ((operation == AMDGPU_VA_OP_MAP || + operation == AMDGPU_VA_OP_REPLACE) && + !amdgpu_vm_is_bo_always_valid(vm, bo_va->base.bo)) { - if (amdgpu_vm_is_bo_always_valid(vm, bo_va->base.bo)) - fence = dma_fence_get(vm->last_update); - else - fence = dma_fence_get(bo_va->last_pt_update); - break; - case AMDGPU_VA_OP_UNMAP: - case AMDGPU_VA_OP_CLEAR: - default: - /* keep @fence as returned by amdgpu_vm_clear_freed() */ - break; + /* + * For MAP/REPLACE of non per-VM BOs we need to sync to both the + * bo_va->last_pt_update and vm->last_update or otherwise we + * potentially miss the PDE updates. + */ + fence = dma_fence_unwrap_merge(vm->last_update, + bo_va->last_pt_update); + if (!fence) { + /* As fallback in OOM situations */ + dma_fence_wait(vm->last_update, false); + dma_fence_wait(bo_va->last_pt_update, false); + fence = dma_fence_get_stub(); + } + } else { + fence = dma_fence_get(vm->last_update); } + return fence; + error: if (r && r != -ERESTARTSYS) DRM_ERROR("Couldn't update BO_VA (%d)\n", r); - return fence; + return dma_fence_get(vm->last_update); } int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, @@ -832,7 +819,6 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, struct amdgpu_bo_va *bo_va; struct drm_syncobj *timeline_syncobj = NULL; struct dma_fence_chain *timeline_chain = NULL; - struct dma_fence *fence; struct drm_exec exec; uint64_t vm_size; int r = 0; @@ -884,6 +870,10 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return -EINVAL; } + if (args->flags & AMDGPU_VM_DELAY_UPDATE && + args->vm_timeline_syncobj_out) + return -EINVAL; + if ((args->operation != AMDGPU_VA_OP_CLEAR) && !(args->flags & AMDGPU_VM_PAGE_PRT)) { gobj = drm_gem_object_lookup(filp, args->handle); @@ -973,11 +963,13 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, * that represents the last relevant update for this mapping. This * fence can then be exported to the user-visible VM timeline. */ - if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !adev->debug_vm) { + if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && + (!adev->debug_vm || timeline_syncobj)) { + struct dma_fence *fence; + fence = amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va, args->operation); - - if (timeline_syncobj && fence) { + if (timeline_syncobj) { if (!args->vm_timeline_point) { /* Replace the existing fence when no point is given. */ drm_syncobj_replace_fence(timeline_syncobj, From c096932fd4f72f89fed0b80df473f3af8217d818 Mon Sep 17 00:00:00 2001 From: Roman Li Date: Mon, 16 Mar 2026 20:17:57 -0400 Subject: [PATCH 29/99] drm/amd/display: Update underflow detection for DCN42 [Why] The DCN42 underflow detection functions in dcn42_optc.c use OPTC_RSMU_UNDERFLOW register but the register offset definitions were missing from dcn_4_2_0_offset.h and dcn42_resource.h. [How] Add missing register definitions. Fixes: e56e3cff2a1b ("drm/amd/display: Sync dcn42 with DC 3.2.373") Reviewed-by: Alex Hung Signed-off-by: Roman Li Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h | 2 ++ drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h | 6 ++++++ 2 files changed, 8 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h index a9b26df14520..8e7a09402de5 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h @@ -481,6 +481,8 @@ SRI_ARR(OPTC_INPUT_CLOCK_CONTROL, ODM, inst), \ SRI_ARR(OPTC_DATA_SOURCE_SELECT, ODM, inst), \ SRI_ARR(OPTC_INPUT_GLOBAL_CONTROL, ODM, inst), \ + SRI_ARR(OPTC_RSMU_UNDERFLOW, ODM, inst), \ + SRI_ARR(OPTC_UNDERFLOW_THRESHOLD, ODM, inst), \ SRI_ARR(CONTROL, VTG, inst), \ SRI_ARR(OTG_VERT_SYNC_CONTROL, OTG, inst), \ SRI_ARR(OTG_GSL_CONTROL, OTG, inst), \ diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h index 52fbf2dc1899..3755a984681a 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h @@ -9036,6 +9036,8 @@ // base address: 0x40 #define regODM1_OPTC_INPUT_GLOBAL_CONTROL 0x1ada #define regODM1_OPTC_INPUT_GLOBAL_CONTROL_BASE_IDX 2 +#define regODM1_OPTC_RSMU_UNDERFLOW 0x1adb +#define regODM1_OPTC_RSMU_UNDERFLOW_BASE_IDX 2 #define regODM1_OPTC_UNDERFLOW_THRESHOLD 0x1adc #define regODM1_OPTC_UNDERFLOW_THRESHOLD_BASE_IDX 2 #define regODM1_OPTC_DATA_SOURCE_SELECT 0x1add @@ -9060,6 +9062,8 @@ // base address: 0x80 #define regODM2_OPTC_INPUT_GLOBAL_CONTROL 0x1aea #define regODM2_OPTC_INPUT_GLOBAL_CONTROL_BASE_IDX 2 +#define regODM2_OPTC_RSMU_UNDERFLOW 0x1aeb +#define regODM2_OPTC_RSMU_UNDERFLOW_BASE_IDX 2 #define regODM2_OPTC_UNDERFLOW_THRESHOLD 0x1aec #define regODM2_OPTC_UNDERFLOW_THRESHOLD_BASE_IDX 2 #define regODM2_OPTC_DATA_SOURCE_SELECT 0x1aed @@ -9084,6 +9088,8 @@ // base address: 0xc0 #define regODM3_OPTC_INPUT_GLOBAL_CONTROL 0x1afa #define regODM3_OPTC_INPUT_GLOBAL_CONTROL_BASE_IDX 2 +#define regODM3_OPTC_RSMU_UNDERFLOW 0x1afb +#define regODM3_OPTC_RSMU_UNDERFLOW_BASE_IDX 2 #define regODM3_OPTC_UNDERFLOW_THRESHOLD 0x1afc #define regODM3_OPTC_UNDERFLOW_THRESHOLD_BASE_IDX 2 #define regODM3_OPTC_DATA_SOURCE_SELECT 0x1afd From deab056486f4e9945d403ea0ef812a593b7f438e Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Thu, 5 Mar 2026 20:23:25 -0500 Subject: [PATCH 30/99] drm/amd/display: Add Extra SMU Log for dtbclk [why] need to check dtbclk in log for confirmation Reviewed-by: Gabe Teeger Signed-off-by: Charlene Liu Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index bc11510b63a1..7134d8998efc 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -474,6 +474,9 @@ static void dcn42_dump_clk_registers(struct clk_state_registers_and_bypass *regs DC_LOG_SMU("CLK1_CLK1_CURRENT_CNT,%d,dppclk\n", internal.CLK8_CLK1_CURRENT_CNT); + DC_LOG_SMU("CLK1_CLK4_CURRENT_CNT,%d,dtbclk\n", + internal.CLK8_CLK4_CURRENT_CNT); + DC_LOG_SMU("CLK1_CLK3_BYPASS_CNTL,%d,dcfclk_bypass\n", internal.CLK8_CLK3_BYPASS_CNTL); From 6c006fac2c17797510691c1848320a1659ff58b4 Mon Sep 17 00:00:00 2001 From: Alexander Chechik Date: Mon, 9 Mar 2026 13:15:24 -0400 Subject: [PATCH 31/99] drm/amd/display: Fix DCN42 memory clock table using MemClk instead of UClk [Why] DCN42 was using UClk values instead of MemClk from MemPstateTable, causing DML to see half the actual DRAM bandwidth on DDR5 systems and reject high refresh rate modes. [How] Change dcn42_init_clocks() to use MemPstateTable[i].MemClk instead of MemPstateTable[i].UClk for memclk_mhz initialization. Reviewed-by: Charlene Liu Signed-off-by: Alexander Chechik Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index 7134d8998efc..24834f89711d 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -1063,7 +1063,7 @@ static void dcn42_get_smu_clocks(struct clk_mgr_internal *clk_mgr_int) if (dpm_clks->NumMemPstatesEnabled > NUM_MEM_PSTATE_LEVELS) dpm_clks->NumMemPstatesEnabled = NUM_MEM_PSTATE_LEVELS; for (i = 0; i < dpm_clks->NumMemPstatesEnabled; i++) { - clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].memclk_mhz = dpm_clks->MemPstateTable[i].UClk; + clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].memclk_mhz = dpm_clks->MemPstateTable[i].MemClk; clk_mgr_base->bw_params->clk_table.entries[dpm_clks->NumMemPstatesEnabled - 1 - i].wck_ratio = dcn42_convert_wck_ratio(dpm_clks->MemPstateTable[i].WckRatio) ; } clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels = dpm_clks->NumMemPstatesEnabled; From 52024a94e7111366141cfc5d888b2ef011f879e5 Mon Sep 17 00:00:00 2001 From: Alex Hung Date: Mon, 9 Mar 2026 11:16:08 -0600 Subject: [PATCH 32/99] drm/amd/display: Fix drm_edid leak in amdgpu_dm [WHAT] When a sink is connected, aconnector->drm_edid was overwritten without freeing the previous allocation, causing a memory leak on resume. [HOW] Free the previous drm_edid before updating it. Reviewed-by: Roman Li Signed-off-by: Alex Hung Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) 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 1e010d095f06..35c8666b23dc 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -3994,8 +3994,9 @@ void amdgpu_dm_update_connector_after_detect( aconnector->dc_sink = sink; dc_sink_retain(aconnector->dc_sink); + drm_edid_free(aconnector->drm_edid); + aconnector->drm_edid = NULL; if (sink->dc_edid.length == 0) { - aconnector->drm_edid = NULL; hdmi_cec_unset_edid(aconnector); if (aconnector->dc_link->aux_mode) { drm_dp_cec_unset_edid(&aconnector->dm_dp_aux.aux); From c24bb00cc6cfef4afe71de8b9bb5c809a49888f2 Mon Sep 17 00:00:00 2001 From: Dillon Varone Date: Thu, 5 Mar 2026 16:42:29 -0500 Subject: [PATCH 33/99] drm/amd/display: Refactor DC update checks [WHY&HOW] DC currently has fragmented definitions of update types. This changes consolidates them into a single interface, and adds expanded functionality to accommodate all use cases. - adds `dc_check_update_state_and_surfaces_for_stream` to determine update type including state, surface, and stream changes. - adds missing surface/stream update checks to `dc_check_update_surfaces_for_stream` - adds new update type `UPDATE_TYPE_ADDR_ONLY` to accomodate flows where further distinction from `UPDATE_TYPE_FAST` was needed - removes caller reliance on `enable_legacy_fast_update` to determine which commit function to use, instead embedding it in the update type Reviewed-by: Nicholas Kazlauskas Signed-off-by: Dillon Varone Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 6 +- .../amd/display/amdgpu_dm/amdgpu_dm_crtc.c | 2 +- drivers/gpu/drm/amd/display/dc/core/dc.c | 472 ++++++++---------- drivers/gpu/drm/amd/display/dc/dc.h | 20 +- 4 files changed, 201 insertions(+), 299 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 35c8666b23dc..fa88eb548404 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -9886,7 +9886,7 @@ static void amdgpu_dm_enable_self_refresh(struct amdgpu_crtc *acrtc_attach, } /* Decrement skip count when SR is enabled and we're doing fast updates. */ - if (acrtc_state->update_type == UPDATE_TYPE_FAST && + if (acrtc_state->update_type <= UPDATE_TYPE_FAST && (psr->psr_feature_enabled || pr->config.replay_supported)) { if (aconn->sr_skip_count > 0) aconn->sr_skip_count--; @@ -10094,7 +10094,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, * fast updates. */ if (crtc->state->async_flip && - (acrtc_state->update_type != UPDATE_TYPE_FAST || + (acrtc_state->update_type > UPDATE_TYPE_FAST || get_mem_type(old_plane_state->fb) != get_mem_type(fb))) drm_warn_once(state->dev, "[PLANE:%d:%s] async flip with non-fast update\n", @@ -10102,7 +10102,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, bundle->flip_addrs[planes_count].flip_immediate = crtc->state->async_flip && - acrtc_state->update_type == UPDATE_TYPE_FAST && + acrtc_state->update_type <= UPDATE_TYPE_FAST && get_mem_type(old_plane_state->fb) == get_mem_type(fb); timestamp_ns = ktime_get_ns(); diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c index 304437c2284d..82727f6ec469 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c @@ -685,7 +685,7 @@ static int amdgpu_dm_crtc_helper_atomic_check(struct drm_crtc *crtc, * pitch, the DCC state, rotation, etc. */ if (crtc_state->async_flip && - dm_crtc_state->update_type != UPDATE_TYPE_FAST) { + dm_crtc_state->update_type > UPDATE_TYPE_FAST) { drm_dbg_atomic(crtc->dev, "[CRTC:%d:%s] async flips are only supported for fast updates\n", crtc->base.id, crtc->name); diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 8b9c686eefd2..ac2d85e8b00a 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -2695,7 +2695,7 @@ static bool is_surface_in_context( static struct surface_update_descriptor get_plane_info_update_type(const struct dc_surface_update *u) { union surface_update_flags *update_flags = &u->surface->update_flags; - struct surface_update_descriptor update_type = { UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_NONE }; + struct surface_update_descriptor update_type = { UPDATE_TYPE_ADDR_ONLY, LOCK_DESCRIPTOR_NONE }; if (!u->plane_info) return update_type; @@ -2803,7 +2803,7 @@ static struct surface_update_descriptor get_scaling_info_update_type( const struct dc_surface_update *u) { union surface_update_flags *update_flags = &u->surface->update_flags; - struct surface_update_descriptor update_type = { UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_NONE }; + struct surface_update_descriptor update_type = { UPDATE_TYPE_ADDR_ONLY, LOCK_DESCRIPTOR_NONE }; if (!u->scaling_info) return update_type; @@ -2854,11 +2854,11 @@ static struct surface_update_descriptor get_scaling_info_update_type( return update_type; } -static struct surface_update_descriptor det_surface_update( +static struct surface_update_descriptor check_update_surface( const struct dc_check_config *check_config, struct dc_surface_update *u) { - struct surface_update_descriptor overall_type = { UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_NONE }; + struct surface_update_descriptor overall_type = { UPDATE_TYPE_ADDR_ONLY, LOCK_DESCRIPTOR_NONE }; union surface_update_flags *update_flags = &u->surface->update_flags; if (u->surface->force_full_update) { @@ -2878,7 +2878,7 @@ static struct surface_update_descriptor det_surface_update( if (u->flip_addr) { update_flags->bits.addr_update = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, UPDATE_TYPE_ADDR_ONLY, LOCK_DESCRIPTOR_STREAM); if (u->flip_addr->address.tmz_surface != u->surface->address.tmz_surface) { update_flags->bits.tmz_changed = 1; @@ -2892,27 +2892,43 @@ static struct surface_update_descriptor det_surface_update( if (u->input_csc_color_matrix) { update_flags->bits.input_csc_change = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); + } + + if (u->cursor_csc_color_matrix) { + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); } if (u->coeff_reduction_factor) { update_flags->bits.coeff_reduction_change = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); } if (u->gamut_remap_matrix) { update_flags->bits.gamut_remap_change = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); } if (u->cm || (u->gamma && dce_use_lut(u->plane_info ? u->plane_info->format : u->surface->format))) { update_flags->bits.gamma_change = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); } if (u->cm && (u->cm->flags.bits.lut3d_enable || u->surface->cm.flags.bits.lut3d_enable)) { update_flags->bits.lut_3d = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); } if (u->cm && u->cm->flags.bits.lut3d_dma_enable != u->surface->cm.flags.bits.lut3d_dma_enable && @@ -2928,9 +2944,10 @@ static struct surface_update_descriptor det_surface_update( if (u->hdr_mult.value) if (u->hdr_mult.value != u->surface->hdr_mult.value) { - // TODO: Should be fast? update_flags->bits.hdr_mult = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); + elevate_update_type(&overall_type, + check_config->enable_legacy_fast_update ? UPDATE_TYPE_MED : UPDATE_TYPE_FAST, + LOCK_DESCRIPTOR_STREAM); } if (u->sdr_white_level_nits) @@ -2984,7 +3001,7 @@ static struct surface_update_descriptor check_update_surfaces_for_stream( int surface_count, struct dc_stream_update *stream_update) { - struct surface_update_descriptor overall_type = { UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_NONE }; + struct surface_update_descriptor overall_type = { UPDATE_TYPE_ADDR_ONLY, LOCK_DESCRIPTOR_NONE }; /* When countdown finishes, promote this flip to full to trigger deferred final transition */ if (check_config->deferred_transition_state && !check_config->transition_countdown_to_steady_state) { @@ -3051,7 +3068,18 @@ static struct surface_update_descriptor check_update_surfaces_for_stream( if (su_flags->raw) elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); - // Non-global cases + /* Non-global cases */ + if (stream_update->hdr_static_metadata || + stream_update->vrr_infopacket || + stream_update->vsc_infopacket || + stream_update->vsp_infopacket || + stream_update->hfvsif_infopacket || + stream_update->adaptive_sync_infopacket || + stream_update->vtem_infopacket || + stream_update->avi_infopacket) { + elevate_update_type(&overall_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); + } + if (stream_update->output_csc_transform) { su_flags->bits.out_csc = 1; elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); @@ -3061,11 +3089,32 @@ static struct surface_update_descriptor check_update_surfaces_for_stream( su_flags->bits.out_tf = 1; elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); } + + if (stream_update->periodic_interrupt) { + elevate_update_type(&overall_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); + } + + if (stream_update->dither_option) { + elevate_update_type(&overall_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); + } + + if (stream_update->cursor_position || stream_update->cursor_attributes) { + elevate_update_type(&overall_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); + } + + /* TODO - cleanup post blend CM */ + if (stream_update->func_shaper || stream_update->lut3d_func) { + elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); + } + + if (stream_update->pending_test_pattern) { + elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); + } } for (int i = 0 ; i < surface_count; i++) { struct surface_update_descriptor inner_type = - det_surface_update(check_config, &updates[i]); + check_update_surface(check_config, &updates[i]); elevate_update_type(&overall_type, inner_type.update_type, inner_type.lock_descriptor); } @@ -3092,6 +3141,81 @@ struct surface_update_descriptor dc_check_update_surfaces_for_stream( return check_update_surfaces_for_stream(check_config, updates, surface_count, stream_update); } +/* + * check_update_state_and_surfaces_for_stream() - Determine update type (fast, med, or full) + * + * This function performs checks on the DC global state, and is therefore not re-entrant. It + * should not be called from DM. + * + * See :c:type:`enum surface_update_type ` for explanation of update types + */ +static struct surface_update_descriptor check_update_state_and_surfaces_for_stream( + const struct dc *dc, + const struct dc_check_config *check_config, + const struct dc_stream_state *stream, + const struct dc_surface_update *updates, + const int surface_count, + const struct dc_stream_update *stream_update) +{ + const struct dc_state *context = dc->current_state; + + struct surface_update_descriptor overall_type = { UPDATE_TYPE_ADDR_ONLY, LOCK_DESCRIPTOR_NONE}; + + if (updates) + for (int i = 0; i < surface_count; i++) + if (!is_surface_in_context(context, updates[i].surface)) + elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); + + if (stream) { + const struct dc_stream_status *stream_status = dc_stream_get_status_const(stream); + if (stream_status == NULL || stream_status->plane_count != surface_count) + elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); + } + if (dc->idle_optimizations_allowed) + elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); + + if (dc_can_clear_cursor_limit(dc)) + elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); + + return overall_type; +} + +/* + * dc_check_update_state_and_surfaces_for_stream() - Determine update type (fast, med, or full) + * + * This function performs checks on the DC global state, stream and surface update, and is + * therefore not re-entrant. It should not be called from DM. + * + * See :c:type:`enum surface_update_type ` for explanation of update types + */ +static struct surface_update_descriptor dc_check_update_state_and_surfaces_for_stream( + const struct dc *dc, + const struct dc_check_config *check_config, + struct dc_stream_state *stream, + struct dc_surface_update *updates, + int surface_count, + struct dc_stream_update *stream_update) +{ + /* check updates against the entire DC state (global) first */ + struct surface_update_descriptor overall_update_type = check_update_state_and_surfaces_for_stream( + dc, + check_config, + stream, + updates, + surface_count, + stream_update); + + /* check updates for stream and plane */ + struct surface_update_descriptor stream_update_type = dc_check_update_surfaces_for_stream( + check_config, + updates, + surface_count, + stream_update); + elevate_update_type(&overall_update_type, stream_update_type.update_type, stream_update_type.lock_descriptor); + + return overall_update_type; +} + static struct dc_stream_status *stream_get_status( struct dc_state *ctx, struct dc_stream_state *stream) @@ -3448,13 +3572,6 @@ static void update_seamless_boot_flags(struct dc *dc, } } -static bool full_update_required_weak( - const struct dc *dc, - const struct dc_surface_update *srf_updates, - int surface_count, - const struct dc_stream_update *stream_update, - const struct dc_stream_state *stream); - struct pipe_split_policy_backup { bool dynamic_odm_policy; bool subvp_policy; @@ -3524,12 +3641,11 @@ static bool update_planes_and_stream_state(struct dc *dc, struct dc_surface_update *srf_updates, int surface_count, struct dc_stream_state *stream, struct dc_stream_update *stream_update, - enum surface_update_type *new_update_type, + struct surface_update_descriptor *update_descriptor, struct dc_state **new_context) { struct dc_state *context; int i, j; - enum surface_update_type update_type; const struct dc_stream_status *stream_status; struct dc_context *dc_ctx = dc->ctx; @@ -3543,17 +3659,20 @@ static bool update_planes_and_stream_state(struct dc *dc, } context = dc->current_state; - update_type = dc_check_update_surfaces_for_stream( - &dc->check_config, srf_updates, surface_count, stream_update).update_type; - if (full_update_required_weak(dc, srf_updates, surface_count, stream_update, stream)) - update_type = UPDATE_TYPE_FULL; + *update_descriptor = dc_check_update_state_and_surfaces_for_stream( + dc, + &dc->check_config, + stream, + srf_updates, + surface_count, + stream_update); /* It is possible to receive a flip for one plane while there are multiple flip_immediate planes in the same stream. * E.g. Desktop and MPO plane are flip_immediate but only the MPO plane received a flip * Force the other flip_immediate planes to flip so GSL doesn't wait for a flip that won't come. */ force_immediate_gsl_plane_flip(dc, srf_updates, surface_count); - if (update_type == UPDATE_TYPE_FULL) + if (update_descriptor->update_type == UPDATE_TYPE_FULL) backup_planes_and_stream_state(&dc->scratch.current_state, stream); /* update current stream with the new updates */ @@ -3579,7 +3698,7 @@ static bool update_planes_and_stream_state(struct dc *dc, } } - if (update_type == UPDATE_TYPE_FULL) { + if (update_descriptor->update_type == UPDATE_TYPE_FULL) { if (stream_update) { uint32_t dsc_changed = stream_update->stream->update_flags.bits.dsc_changed; stream_update->stream->update_flags.raw = 0xFFFFFFFF; @@ -3589,13 +3708,13 @@ static bool update_planes_and_stream_state(struct dc *dc, srf_updates[i].surface->update_flags.raw = 0xFFFFFFFF; } - if (update_type >= update_surface_trace_level) + if (update_descriptor->update_type >= update_surface_trace_level) update_surface_trace(dc, srf_updates, surface_count); for (i = 0; i < surface_count; i++) copy_surface_update_to_plane(srf_updates[i].surface, &srf_updates[i]); - if (update_type >= UPDATE_TYPE_FULL) { + if (update_descriptor->update_type >= UPDATE_TYPE_FULL) { struct dc_plane_state *new_planes[MAX_SURFACES] = {0}; for (i = 0; i < surface_count; i++) @@ -3633,7 +3752,7 @@ static bool update_planes_and_stream_state(struct dc *dc, for (i = 0; i < surface_count; i++) { struct dc_plane_state *surface = srf_updates[i].surface; - if (update_type != UPDATE_TYPE_MED) + if (update_descriptor->update_type != UPDATE_TYPE_MED) continue; if (surface->update_flags.bits.position_change) { for (j = 0; j < dc->res_pool->pipe_count; j++) { @@ -3647,7 +3766,7 @@ static bool update_planes_and_stream_state(struct dc *dc, } } - if (update_type == UPDATE_TYPE_FULL) { + if (update_descriptor->update_type == UPDATE_TYPE_FULL) { struct pipe_split_policy_backup policy; bool minimize = false; @@ -3676,8 +3795,7 @@ static bool update_planes_and_stream_state(struct dc *dc, update_seamless_boot_flags(dc, context, surface_count, stream); *new_context = context; - *new_update_type = update_type; - if (update_type == UPDATE_TYPE_FULL) + if (update_descriptor->update_type == UPDATE_TYPE_FULL) backup_planes_and_stream_state(&dc->scratch.new_state, stream); return true; @@ -3757,7 +3875,7 @@ static void commit_planes_do_stream_update(struct dc *dc, program_cursor_position(dc, stream); /* Full fe update*/ - if (update_type == UPDATE_TYPE_FAST) + if (update_type <= UPDATE_TYPE_FAST) continue; if (stream_update->dsc_config) @@ -4066,7 +4184,7 @@ static void commit_planes_for_stream_fast(struct dc *dc, struct pipe_ctx *top_pipe_to_program = NULL; struct dc_stream_status *stream_status = NULL; bool should_offload_fams2_flip = false; - bool should_lock_all_pipes = (update_type != UPDATE_TYPE_FAST); + bool should_lock_all_pipes = (update_type > UPDATE_TYPE_FAST); if (should_lock_all_pipes) determine_pipe_unlock_order(dc, context); @@ -4126,7 +4244,7 @@ static void commit_planes_for_stream_fast(struct dc *dc, continue; pipe_ctx->plane_state->triplebuffer_flips = false; - if (update_type == UPDATE_TYPE_FAST && + if (update_type <= UPDATE_TYPE_FAST && dc->hwss.program_triplebuffer != NULL && !pipe_ctx->plane_state->flip_immediate && dc->debug.enable_tri_buf) { /*triple buffer for VUpdate only*/ @@ -4183,7 +4301,7 @@ static void commit_planes_for_stream(struct dc *dc, { int i, j; struct pipe_ctx *top_pipe_to_program = NULL; - bool should_lock_all_pipes = (update_type != UPDATE_TYPE_FAST); + bool should_lock_all_pipes = (update_type > UPDATE_TYPE_FAST); bool subvp_prev_use = false; bool subvp_curr_use = false; uint8_t current_stream_mask = 0; @@ -4200,7 +4318,7 @@ static void commit_planes_for_stream(struct dc *dc, if (update_type == UPDATE_TYPE_FULL && dc->optimized_required) hwss_process_outstanding_hw_updates(dc, dc->current_state); - if (update_type != UPDATE_TYPE_FAST && dc->res_pool->funcs->prepare_mcache_programming) + if (update_type > UPDATE_TYPE_FAST && dc->res_pool->funcs->prepare_mcache_programming) dc->res_pool->funcs->prepare_mcache_programming(dc, context); for (i = 0; i < dc->res_pool->pipe_count; i++) { @@ -4262,7 +4380,7 @@ static void commit_planes_for_stream(struct dc *dc, odm_pipe->ttu_regs.min_ttu_vblank = MAX_TTU; } - if ((update_type != UPDATE_TYPE_FAST) && stream->update_flags.bits.dsc_changed) + if ((update_type > UPDATE_TYPE_FAST) && stream->update_flags.bits.dsc_changed) if (top_pipe_to_program && top_pipe_to_program->stream_res.tg->funcs->lock_doublebuffer_enable) { if (should_use_dmub_inbox1_lock(dc, stream->link)) { @@ -4333,7 +4451,7 @@ static void commit_planes_for_stream(struct dc *dc, } dc->hwss.post_unlock_program_front_end(dc, context); - if (update_type != UPDATE_TYPE_FAST) + if (update_type > UPDATE_TYPE_FAST) if (dc->hwss.commit_subvp_config) dc->hwss.commit_subvp_config(dc, context); @@ -4349,7 +4467,7 @@ static void commit_planes_for_stream(struct dc *dc, return; } - if (update_type != UPDATE_TYPE_FAST) { + if (update_type > UPDATE_TYPE_FAST) { for (j = 0; j < dc->res_pool->pipe_count; j++) { struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[j]; @@ -4377,7 +4495,7 @@ static void commit_planes_for_stream(struct dc *dc, if (!should_update_pipe_for_plane(context, pipe_ctx, plane_state)) continue; pipe_ctx->plane_state->triplebuffer_flips = false; - if (update_type == UPDATE_TYPE_FAST && + if (update_type <= UPDATE_TYPE_FAST && dc->hwss.program_triplebuffer != NULL && !pipe_ctx->plane_state->flip_immediate && dc->debug.enable_tri_buf) { /*triple buffer for VUpdate only*/ @@ -4404,7 +4522,7 @@ static void commit_planes_for_stream(struct dc *dc, continue; /* Full fe update*/ - if (update_type == UPDATE_TYPE_FAST) + if (update_type <= UPDATE_TYPE_FAST) continue; stream_status = @@ -4423,7 +4541,7 @@ static void commit_planes_for_stream(struct dc *dc, continue; /* Full fe update*/ - if (update_type == UPDATE_TYPE_FAST) + if (update_type <= UPDATE_TYPE_FAST) continue; ASSERT(!pipe_ctx->plane_state->triplebuffer_flips); @@ -4434,7 +4552,7 @@ static void commit_planes_for_stream(struct dc *dc, } } - if (dc->hwss.program_front_end_for_ctx && update_type != UPDATE_TYPE_FAST) { + if (dc->hwss.program_front_end_for_ctx && update_type > UPDATE_TYPE_FAST) { dc->hwss.program_front_end_for_ctx(dc, context); //Pipe busy until some frame and line # @@ -4462,7 +4580,7 @@ static void commit_planes_for_stream(struct dc *dc, } // Update Type FAST, Surface updates - if (update_type == UPDATE_TYPE_FAST) { + if (update_type <= UPDATE_TYPE_FAST) { if (dc->hwss.set_flip_control_gsl) for (i = 0; i < surface_count; i++) { struct dc_plane_state *plane_state = srf_updates[i].surface; @@ -4519,7 +4637,7 @@ static void commit_planes_for_stream(struct dc *dc, dc->hwss.pipe_control_lock(dc, top_pipe_to_program, false); } - if ((update_type != UPDATE_TYPE_FAST) && stream->update_flags.bits.dsc_changed) + if ((update_type > UPDATE_TYPE_FAST) && stream->update_flags.bits.dsc_changed) if (top_pipe_to_program && top_pipe_to_program->stream_res.tg->funcs->lock_doublebuffer_enable) { top_pipe_to_program->stream_res.tg->funcs->wait_for_state( @@ -4552,13 +4670,13 @@ static void commit_planes_for_stream(struct dc *dc, /* If enabling subvp or transitioning from subvp->subvp, enable the * phantom streams before we program front end for the phantom pipes. */ - if (update_type != UPDATE_TYPE_FAST) { + if (update_type > UPDATE_TYPE_FAST) { if (dc->hwss.enable_phantom_streams) dc->hwss.enable_phantom_streams(dc, context); } } - if (update_type != UPDATE_TYPE_FAST) + if (update_type > UPDATE_TYPE_FAST) dc->hwss.post_unlock_program_front_end(dc, context); if (subvp_prev_use && !subvp_curr_use) { @@ -4571,7 +4689,7 @@ static void commit_planes_for_stream(struct dc *dc, dc->hwss.disable_phantom_streams(dc, context); } - if (update_type != UPDATE_TYPE_FAST) + if (update_type > UPDATE_TYPE_FAST) if (dc->hwss.commit_subvp_config) dc->hwss.commit_subvp_config(dc, context); /* Since phantom pipe programming is moved to post_unlock_program_front_end, @@ -5043,191 +5161,12 @@ static bool commit_minimal_transition_state(struct dc *dc, return true; } -void populate_fast_updates(struct dc_fast_update *fast_update, - struct dc_surface_update *srf_updates, - int surface_count, - struct dc_stream_update *stream_update) -{ - int i = 0; - - if (stream_update) { - fast_update[0].out_transfer_func = stream_update->out_transfer_func; - fast_update[0].output_csc_transform = stream_update->output_csc_transform; - } else { - fast_update[0].out_transfer_func = NULL; - fast_update[0].output_csc_transform = NULL; - } - - for (i = 0; i < surface_count; i++) { - fast_update[i].flip_addr = srf_updates[i].flip_addr; - fast_update[i].gamma = srf_updates[i].gamma; - fast_update[i].gamut_remap_matrix = srf_updates[i].gamut_remap_matrix; - fast_update[i].input_csc_color_matrix = srf_updates[i].input_csc_color_matrix; - fast_update[i].coeff_reduction_factor = srf_updates[i].coeff_reduction_factor; - fast_update[i].cursor_csc_color_matrix = srf_updates[i].cursor_csc_color_matrix; - fast_update[i].cm_hist_control = srf_updates[i].cm_hist_control; - } -} - -static bool fast_updates_exist(const struct dc_fast_update *fast_update, int surface_count) -{ - int i; - - if (fast_update[0].out_transfer_func || - fast_update[0].output_csc_transform) - return true; - - for (i = 0; i < surface_count; i++) { - if (fast_update[i].flip_addr || - fast_update[i].gamma || - fast_update[i].gamut_remap_matrix || - fast_update[i].input_csc_color_matrix || - fast_update[i].cursor_csc_color_matrix || - fast_update[i].cm_hist_control || - fast_update[i].coeff_reduction_factor) - return true; - } - - return false; -} - -bool fast_nonaddr_updates_exist(struct dc_fast_update *fast_update, int surface_count) -{ - int i; - - if (fast_update[0].out_transfer_func || - fast_update[0].output_csc_transform) - return true; - - for (i = 0; i < surface_count; i++) { - if (fast_update[i].input_csc_color_matrix || - fast_update[i].gamma || - fast_update[i].gamut_remap_matrix || - fast_update[i].coeff_reduction_factor || - fast_update[i].cm_hist_control || - fast_update[i].cursor_csc_color_matrix) - return true; - } - - return false; -} - -static bool full_update_required_weak( - const struct dc *dc, - const struct dc_surface_update *srf_updates, - int surface_count, - const struct dc_stream_update *stream_update, - const struct dc_stream_state *stream) -{ - const struct dc_state *context = dc->current_state; - if (srf_updates) - for (int i = 0; i < surface_count; i++) - if (!is_surface_in_context(context, srf_updates[i].surface)) - return true; - - if (stream) { - const struct dc_stream_status *stream_status = dc_stream_get_status_const(stream); - if (stream_status == NULL || stream_status->plane_count != surface_count) - return true; - } - if (dc->idle_optimizations_allowed) - return true; - - if (dc_can_clear_cursor_limit(dc)) - return true; - - return false; -} - -static bool full_update_required( - const struct dc *dc, - const struct dc_surface_update *srf_updates, - int surface_count, - const struct dc_stream_update *stream_update, - const struct dc_stream_state *stream) -{ - const union dc_plane_cm_flags blend_only_flags = { - .bits = { - .blend_enable = 1, - } - }; - - if (full_update_required_weak(dc, srf_updates, surface_count, stream_update, stream)) - return true; - - for (int i = 0; i < surface_count; i++) { - if (srf_updates && - (srf_updates[i].plane_info || - srf_updates[i].scaling_info || - (srf_updates[i].hdr_mult.value && - srf_updates[i].hdr_mult.value != srf_updates->surface->hdr_mult.value) || - (srf_updates[i].sdr_white_level_nits && - srf_updates[i].sdr_white_level_nits != srf_updates->surface->sdr_white_level_nits) || - srf_updates[i].in_transfer_func || - srf_updates[i].surface->force_full_update || - (srf_updates[i].flip_addr && - srf_updates[i].flip_addr->address.tmz_surface != srf_updates[i].surface->address.tmz_surface) || - (srf_updates[i].cm && - ((srf_updates[i].cm->flags.all != blend_only_flags.all && srf_updates[i].cm->flags.all != 0) || - (srf_updates[i].surface->cm.flags.all != blend_only_flags.all && srf_updates[i].surface->cm.flags.all != 0))))) - return true; - } - - if (stream_update && - (((stream_update->src.height != 0 && stream_update->src.width != 0) || - (stream_update->dst.height != 0 && stream_update->dst.width != 0) || - stream_update->integer_scaling_update) || - stream_update->hdr_static_metadata || - stream_update->abm_level || - stream_update->periodic_interrupt || - stream_update->vrr_infopacket || - stream_update->vsc_infopacket || - stream_update->vsp_infopacket || - stream_update->hfvsif_infopacket || - stream_update->vtem_infopacket || - stream_update->adaptive_sync_infopacket || - stream_update->avi_infopacket || - stream_update->dpms_off || - stream_update->allow_freesync || - stream_update->vrr_active_variable || - stream_update->vrr_active_fixed || - stream_update->gamut_remap || - stream_update->output_color_space || - stream_update->dither_option || - stream_update->wb_update || - stream_update->dsc_config || - stream_update->mst_bw_update || - stream_update->func_shaper || - stream_update->lut3d_func || - stream_update->pending_test_pattern || - stream_update->crtc_timing_adjust || - stream_update->scaler_sharpener_update || - stream_update->hw_cursor_req)) - return true; - - return false; -} - -static bool fast_update_only( - const struct dc *dc, - const struct dc_fast_update *fast_update, - const struct dc_surface_update *srf_updates, - int surface_count, - const struct dc_stream_update *stream_update, - const struct dc_stream_state *stream) -{ - return fast_updates_exist(fast_update, surface_count) - && !full_update_required(dc, srf_updates, surface_count, stream_update, stream); -} - static bool update_planes_and_stream_v2(struct dc *dc, struct dc_surface_update *srf_updates, int surface_count, struct dc_stream_state *stream, struct dc_stream_update *stream_update) { struct dc_state *context; - enum surface_update_type update_type; - struct dc_fast_update fast_update[MAX_SURFACES] = {0}; /* In cases where MPO and split or ODM are used transitions can * cause underflow. Apply stream configuration with minimal pipe @@ -5235,11 +5174,9 @@ static bool update_planes_and_stream_v2(struct dc *dc, */ bool force_minimal_pipe_splitting = 0; bool is_plane_addition = 0; - bool is_fast_update_only; - populate_fast_updates(fast_update, srf_updates, surface_count, stream_update); - is_fast_update_only = fast_update_only(dc, fast_update, srf_updates, - surface_count, stream_update, stream); + struct surface_update_descriptor update_descriptor = {0}; + force_minimal_pipe_splitting = could_mpcc_tree_change_for_active_pipes( dc, stream, @@ -5258,7 +5195,7 @@ static bool update_planes_and_stream_v2(struct dc *dc, surface_count, stream, stream_update, - &update_type, + &update_descriptor, &context)) return false; @@ -5268,7 +5205,7 @@ static bool update_planes_and_stream_v2(struct dc *dc, dc_state_release(context); return false; } - update_type = UPDATE_TYPE_FULL; + elevate_update_type(&update_descriptor, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); } if (dc->hwss.is_pipe_topology_transition_seamless && @@ -5277,13 +5214,13 @@ static bool update_planes_and_stream_v2(struct dc *dc, commit_minimal_transition_state_in_dc_update(dc, context, stream, srf_updates, surface_count); - if (is_fast_update_only && !dc->check_config.enable_legacy_fast_update) { + if (update_descriptor.update_type <= UPDATE_TYPE_FAST) { commit_planes_for_stream_fast(dc, srf_updates, surface_count, stream, stream_update, - update_type, + update_descriptor.update_type, context); } else { if (!stream_update && @@ -5299,7 +5236,7 @@ static bool update_planes_and_stream_v2(struct dc *dc, surface_count, stream, stream_update, - update_type, + update_descriptor.update_type, context); } if (dc->current_state != context) @@ -5313,14 +5250,8 @@ static void commit_planes_and_stream_update_on_current_context(struct dc *dc, struct dc_stream_update *stream_update, enum surface_update_type update_type) { - struct dc_fast_update fast_update[MAX_SURFACES] = {0}; - ASSERT(update_type < UPDATE_TYPE_FULL); - populate_fast_updates(fast_update, srf_updates, surface_count, - stream_update); - if (fast_update_only(dc, fast_update, srf_updates, surface_count, - stream_update, stream) && - !dc->check_config.enable_legacy_fast_update) + if (update_type <= UPDATE_TYPE_FAST) commit_planes_for_stream_fast(dc, srf_updates, surface_count, @@ -5411,7 +5342,7 @@ static bool update_planes_and_stream_v3(struct dc *dc, struct dc_stream_update *stream_update) { struct dc_state *new_context; - enum surface_update_type update_type; + struct surface_update_descriptor update_descriptor = {0}; /* * When this function returns true and new_context is not equal to @@ -5423,22 +5354,26 @@ static bool update_planes_and_stream_v3(struct dc *dc, * replaced by a newer context. Refer to the use of * swap_and_free_current_context below. */ - if (!update_planes_and_stream_state(dc, srf_updates, surface_count, - stream, stream_update, &update_type, + if (!update_planes_and_stream_state(dc, + srf_updates, + surface_count, + stream, + stream_update, + &update_descriptor, &new_context)) return false; if (new_context == dc->current_state) { commit_planes_and_stream_update_on_current_context(dc, srf_updates, surface_count, stream, - stream_update, update_type); + stream_update, update_descriptor.update_type); if (dc->check_config.transition_countdown_to_steady_state) dc->check_config.transition_countdown_to_steady_state--; } else { commit_planes_and_stream_update_with_new_context(dc, srf_updates, surface_count, stream, - stream_update, update_type, new_context); + stream_update, update_descriptor.update_type, new_context); } return true; @@ -7212,7 +7147,7 @@ struct dc_update_scratch_space { struct dc_stream_update *stream_update; bool update_v3; bool do_clear_update_flags; - enum surface_update_type update_type; + struct surface_update_descriptor update_descriptor; struct dc_state *new_context; enum update_v3_flow flow; struct dc_state *backup_context; @@ -7295,45 +7230,28 @@ static bool update_planes_and_stream_prepare_v3( ASSERT(scratch->flow == UPDATE_V3_FLOW_INVALID); dc_exit_ips_for_hw_access(scratch->dc); - /* HWSS path determination needs to be done prior to updating the surface and stream states. */ - struct dc_fast_update fast_update[MAX_SURFACES] = { 0 }; - - populate_fast_updates(fast_update, - scratch->surface_updates, - scratch->surface_count, - scratch->stream_update); - - const bool is_hwss_fast_path_only = - fast_update_only(scratch->dc, - fast_update, - scratch->surface_updates, - scratch->surface_count, - scratch->stream_update, - scratch->stream) && - !scratch->dc->check_config.enable_legacy_fast_update; - if (!update_planes_and_stream_state( scratch->dc, scratch->surface_updates, scratch->surface_count, scratch->stream, scratch->stream_update, - &scratch->update_type, + &scratch->update_descriptor, &scratch->new_context )) { return false; } if (scratch->new_context == scratch->dc->current_state) { - ASSERT(scratch->update_type < UPDATE_TYPE_FULL); + ASSERT(scratch->update_descriptor.update_type < UPDATE_TYPE_FULL); - scratch->flow = is_hwss_fast_path_only + scratch->flow = scratch->update_descriptor.update_type <= UPDATE_TYPE_FAST ? UPDATE_V3_FLOW_NO_NEW_CONTEXT_CONTEXT_FAST : UPDATE_V3_FLOW_NO_NEW_CONTEXT_CONTEXT_FULL; return true; } - ASSERT(scratch->update_type >= UPDATE_TYPE_FULL); + ASSERT(scratch->update_descriptor.update_type >= UPDATE_TYPE_FULL); const bool seamless = scratch->dc->hwss.is_pipe_topology_transition_seamless( scratch->dc, @@ -7406,7 +7324,7 @@ static void update_planes_and_stream_execute_v3_commit( intermediate_update ? scratch->intermediate_count : scratch->surface_count, scratch->stream, use_stream_update ? scratch->stream_update : NULL, - intermediate_context ? UPDATE_TYPE_FULL : scratch->update_type, + intermediate_context ? UPDATE_TYPE_FULL : scratch->update_descriptor.update_type, // `dc->current_state` only used in `NO_NEW_CONTEXT`, where it is equal to `new_context` intermediate_context ? scratch->intermediate_context : scratch->new_context ); @@ -7424,7 +7342,7 @@ static void update_planes_and_stream_execute_v3( scratch->surface_count, scratch->stream, scratch->stream_update, - scratch->update_type, + scratch->update_descriptor.update_type, scratch->new_context ); break; diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 4d15d97ed7f1..1dd0a2cf786c 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -467,6 +467,7 @@ struct dc_static_screen_params { */ enum surface_update_type { + UPDATE_TYPE_ADDR_ONLY, /* only surface address is being updated, no other programming needed */ UPDATE_TYPE_FAST, /* super fast, safe to execute in isr */ UPDATE_TYPE_MED, /* ISR safe, most of programming needed, no bw/clk change*/ UPDATE_TYPE_FULL, /* may need to shuffle resources */ @@ -1880,18 +1881,6 @@ struct dc_scaling_info { struct scaling_taps scaling_quality; }; -struct dc_fast_update { - const struct dc_flip_addrs *flip_addr; - const struct dc_gamma *gamma; - const struct colorspace_transform *gamut_remap_matrix; - const struct dc_csc_transform *input_csc_color_matrix; - const struct fixed31_32 *coeff_reduction_factor; - struct dc_transfer_func *out_transfer_func; - struct dc_csc_transform *output_csc_transform; - const struct dc_csc_transform *cursor_csc_color_matrix; - struct cm_hist_control *cm_hist_control; -}; - struct dc_surface_update { struct dc_plane_state *surface; @@ -2030,12 +2019,7 @@ bool dc_resource_is_dsc_encoding_supported(const struct dc *dc); void get_audio_check(struct audio_info *aud_modes, struct audio_check *aud_chk); -bool fast_nonaddr_updates_exist(struct dc_fast_update *fast_update, int surface_count); -void populate_fast_updates(struct dc_fast_update *fast_update, - struct dc_surface_update *srf_updates, - int surface_count, - struct dc_stream_update *stream_update); -/* + /* * Set up streams and links associated to drive sinks * The streams parameter is an absolute set of all active streams. * From 3eb8ae187d35e75bbdcb5c4410b33a70ef6b1355 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Tue, 10 Mar 2026 13:33:55 -0400 Subject: [PATCH 34/99] drm/amd/display: remove disable_sutter touch pstate debug code [why] diags is using disable_stutter, this will cause issue when pstate switch enabled Reviewed-by: Ilya Bakoulin Signed-off-by: Charlene Liu Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c | 9 --------- 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c b/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c index 8582dcab1f22..a436fa71d4b4 100644 --- a/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c +++ b/drivers/gpu/drm/amd/display/dc/hubbub/dcn42/dcn42_hubbub.c @@ -430,15 +430,6 @@ static void hubbub42_allow_self_refresh_control(struct hubbub *hubbub, bool allo REG_UPDATE_2(DCHUBBUB_ARB_DRAM_STATE_CNTL, DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_VALUE, 0, DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_ENABLE, !allow); - - if (!allow && hubbub->ctx->dc->debug.disable_stutter) {/*controlled by registry key*/ - REG_UPDATE_2(DCHUBBUB_ARB_DRAM_STATE_CNTL, - DCHUBBUB_ARB_ALLOW_DCFCLK_DEEP_SLEEP_FORCE_VALUE, 0, - DCHUBBUB_ARB_ALLOW_DCFCLK_DEEP_SLEEP_FORCE_ENABLE, 1); - REG_UPDATE_2(DCHUBBUB_ARB_DRAM_STATE_CNTL, - DCHUBBUB_ARB_ALLOW_PSTATE_CHANGE_FORCE_VALUE, 0, - DCHUBBUB_ARB_ALLOW_PSTATE_CHANGE_FORCE_ENABLE, 1); - } } static void hubbub42_set_sdp_control(struct hubbub *hubbub, bool dc_control) { From fc2591175507709191c2010a7eb466837496750d Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Tue, 10 Mar 2026 10:53:08 -0400 Subject: [PATCH 35/99] drm/amd/display: pg cntl update based on previous asic. [why] switch to well tested sequence. Reviewed-by: Leo Chen Signed-off-by: Charlene Liu Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/pg/dcn42/dcn42_pg_cntl.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/pg/dcn42/dcn42_pg_cntl.c b/drivers/gpu/drm/amd/display/dc/pg/dcn42/dcn42_pg_cntl.c index 3685080ce9dc..96290538a889 100644 --- a/drivers/gpu/drm/amd/display/dc/pg/dcn42/dcn42_pg_cntl.c +++ b/drivers/gpu/drm/amd/display/dc/pg/dcn42/dcn42_pg_cntl.c @@ -176,11 +176,12 @@ void pg_cntl42_hubp_dpp_pg_control(struct pg_cntl *pg_cntl, unsigned int hubp_dp uint32_t pwr_status = power_on ? 0 : 2; uint32_t org_ip_request_cntl; bool block_enabled; + bool skip_pg = pg_cntl->ctx->dc->debug.ignore_pg || + pg_cntl->ctx->dc->debug.disable_hubp_power_gate || + pg_cntl->ctx->dc->debug.disable_dpp_power_gate || + pg_cntl->ctx->dc->idle_optimizations_allowed; - if (pg_cntl->ctx->dc->debug.ignore_pg || - pg_cntl->ctx->dc->debug.disable_hubp_power_gate || - pg_cntl->ctx->dc->debug.disable_dpp_power_gate || - pg_cntl->ctx->dc->idle_optimizations_allowed) + if (skip_pg && !power_on) return; block_enabled = pg_cntl42_hubp_dpp_pg_status(pg_cntl, hubp_dpp_inst); From a3c7ab163ba9ca4f7d9f85d271c7b8dd47e5f540 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Fri, 6 Mar 2026 10:40:07 -0500 Subject: [PATCH 36/99] drm/amd/display: export get_power_profile interface for later use [why] export dcn401 get_power_profile for later asic. Reviewed-by: Roman Li Signed-off-by: Charlene Liu Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c | 2 +- .../gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.h | 1 + drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c | 1 + drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h | 1 + 4 files changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c index 60a0e7c95c73..491860cc8378 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c @@ -1790,7 +1790,7 @@ static void dcn401_build_pipe_pix_clk_params(struct pipe_ctx *pipe_ctx) } } -static int dcn401_get_power_profile(const struct dc_state *context) +int dcn401_get_power_profile(const struct dc_state *context) { int uclk_mhz = context->bw_ctx.bw.dcn.clk.dramclk_khz / 1000; int dpm_level = 0; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.h b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.h index 5f3b0319cb5b..47f82b818262 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.h +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.h @@ -32,6 +32,7 @@ void dcn401_get_default_tiling_info(struct dc_tiling_info *tiling_info); unsigned int dcn401_get_vstartup_for_pipe(struct pipe_ctx *pipe_ctx); +int dcn401_get_power_profile(const struct dc_state *context); /* Following are definitions for run time init of reg offsets */ /* HUBP */ diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c index 8175109a66b0..fa5d0558192c 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c @@ -1791,6 +1791,7 @@ static struct resource_funcs dcn42_res_pool_funcs = { .calculate_mall_ways_from_bytes = dcn32_calculate_mall_ways_from_bytes, .prepare_mcache_programming = dcn42_prepare_mcache_programming, .build_pipe_pix_clk_params = dcn42_build_pipe_pix_clk_params, + .get_power_profile = dcn401_get_power_profile, .get_vstartup_for_pipe = dcn401_get_vstartup_for_pipe, .get_max_hw_cursor_size = dcn42_get_max_hw_cursor_size, .get_default_tiling_info = dcn10_get_default_tiling_info diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h index 8e7a09402de5..60acf0e423d9 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.h @@ -586,5 +586,6 @@ enum dc_status dcn42_validate_bandwidth(struct dc *dc, enum dc_validate_mode validate_mode); void dcn42_prepare_mcache_programming(struct dc *dc, struct dc_state *context); +int dcn42_get_power_profile(const struct dc_state *context); #endif /* _DCN42_RESOURCE_H_ */ From 60e8ffaf96267e8c520ffa9411f4893ace7a0a11 Mon Sep 17 00:00:00 2001 From: Gaghik Khachatrian Date: Sat, 7 Mar 2026 18:57:57 -0500 Subject: [PATCH 37/99] drm/amd/display: Fix Silence warnings Also affects: freesync, hdcp, info_packet, power [Why] Resolve compiler warnings by marking unused parameters explicitly. [How] In .c/.h keep parameter names in signatures and add a line with `(void)param;` inside the function body Preserved function signatures and avoids breaking code paths that may reference the parameter under conditional compilation. Reviewed-by: Aric Cyr Signed-off-by: Gaghik Khachatrian Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/modules/color/color_gamma.c | 2 ++ drivers/gpu/drm/amd/display/modules/freesync/freesync.c | 6 ++++++ drivers/gpu/drm/amd/display/modules/hdcp/hdcp.h | 1 + 3 files changed, 9 insertions(+) diff --git a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c index 2639163b8ba2..1f225d0d6c44 100644 --- a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c +++ b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c @@ -896,6 +896,7 @@ static void build_de_pq(struct pwl_float_data_ex *de_pq, uint32_t hw_points_num, const struct hw_x_point *coordinate_x) { + (void)coordinate_x; uint32_t i; struct fixed31_32 output; struct fixed31_32 *de_pq_table = mod_color_get_table(type_de_pq_table); @@ -1339,6 +1340,7 @@ static void scale_gamma_dx(struct pwl_float_data *pwl_rgb, const struct dc_gamma *ramp, struct dividers dividers) { + (void)dividers; uint32_t i; struct fixed31_32 min = dc_fixpt_zero; struct fixed31_32 max = dc_fixpt_one; diff --git a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c index 19de72173052..c9150019aab0 100644 --- a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c +++ b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c @@ -114,6 +114,7 @@ static unsigned int calc_duration_in_us_from_v_total( const struct mod_vrr_params *in_vrr, unsigned int v_total) { + (void)in_vrr; unsigned int duration_in_us = (unsigned int)(div64_u64(((unsigned long long)(v_total) * 10000) * stream->timing.h_total, @@ -218,6 +219,7 @@ static void update_v_total_for_static_ramp( const struct dc_stream_state *stream, struct mod_vrr_params *in_out_vrr) { + (void)core_freesync; unsigned int v_total = 0; unsigned int current_duration_in_us = calc_duration_in_us_from_v_total( @@ -292,6 +294,7 @@ static void apply_below_the_range(struct core_freesync *core_freesync, unsigned int last_render_time_in_us, struct mod_vrr_params *in_out_vrr) { + (void)core_freesync; unsigned int inserted_frame_duration_in_us = 0; unsigned int mid_point_frames_ceil = 0; unsigned int mid_point_frames_floor = 0; @@ -447,6 +450,7 @@ static void apply_fixed_refresh(struct core_freesync *core_freesync, unsigned int last_render_time_in_us, struct mod_vrr_params *in_out_vrr) { + (void)core_freesync; bool update = false; unsigned int max_render_time_in_us = in_out_vrr->max_duration_in_us; @@ -545,6 +549,7 @@ static bool vrr_settings_require_update(struct core_freesync *core_freesync, unsigned int max_refresh_in_uhz, struct mod_vrr_params *in_vrr) { + (void)core_freesync; if (in_vrr->state != in_config->state) { return true; } else if (in_vrr->state == VRR_STATE_ACTIVE_FIXED && @@ -946,6 +951,7 @@ void mod_freesync_build_vrr_infopacket(struct mod_freesync *mod_freesync, struct dc_info_packet *infopacket, bool pack_sdp_v1_3) { + (void)mod_freesync; /* SPD info packet for FreeSync * VTEM info packet for HdmiVRR * Check if Freesync is supported. Return if false. If true, diff --git a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.h b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.h index 26a351a184f3..d07387a961dd 100644 --- a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.h +++ b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp.h @@ -501,6 +501,7 @@ static inline void callback_in_ms(uint16_t time, struct mod_hdcp_output *output) static inline void set_watchdog_in_ms(struct mod_hdcp *hdcp, uint16_t time, struct mod_hdcp_output *output) { + (void)hdcp; output->watchdog_timer_needed = 1; output->watchdog_timer_delay = time; } From 7d59465de38e8cbebf55f122b0514de10f52852c Mon Sep 17 00:00:00 2001 From: Dillon Varone Date: Sat, 7 Mar 2026 05:53:03 +0000 Subject: [PATCH 38/99] drm/amd/display: Add 3DLUT DMA broadcast support [WHY&HOW] A single HUBP can be used to fetch 3DLUT and broadcast to a single HUBP. Add logic to select the top pipe for a given plane and use it's HUBP as the broadcast source for multiple MPC's. Reviewed-by: Ilya Bakoulin Signed-off-by: Dillon Varone Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 2 +- .../amd/display/dc/hwss/dcn401/dcn401_hwseq.c | 106 +++++++++++------- .../amd/display/dc/hwss/dcn401/dcn401_hwseq.h | 3 +- .../drm/amd/display/dc/hwss/hw_sequencer.h | 2 +- 4 files changed, 66 insertions(+), 47 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index ac2d85e8b00a..1c57bfd1fd2d 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -4617,7 +4617,7 @@ static void commit_planes_for_stream(struct dc *dc, srf_updates[i].cm->flags.bits.lut3d_enable && srf_updates[i].cm->flags.bits.lut3d_dma_enable && dc->hwss.trigger_3dlut_dma_load) - dc->hwss.trigger_3dlut_dma_load(dc, pipe_ctx); + dc->hwss.trigger_3dlut_dma_load(pipe_ctx); /*program triple buffer after lock based on flip type*/ if (dc->hwss.program_triplebuffer != NULL && dc->debug.enable_tri_buf) { diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c index 357899116ecd..a72284c3fa1c 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c @@ -369,12 +369,14 @@ void dcn401_init_hw(struct dc *dc) } } -void dcn401_trigger_3dlut_dma_load(struct dc *dc, struct pipe_ctx *pipe_ctx) +void dcn401_trigger_3dlut_dma_load(struct pipe_ctx *pipe_ctx) { - struct hubp *hubp = pipe_ctx->plane_res.hubp; + const struct pipe_ctx *primary_dpp_pipe_ctx = resource_get_primary_dpp_pipe(pipe_ctx); + struct hubp *primary_hubp = primary_dpp_pipe_ctx ? + primary_dpp_pipe_ctx->plane_res.hubp : NULL; - if (hubp->funcs->hubp_enable_3dlut_fl) { - hubp->funcs->hubp_enable_3dlut_fl(hubp, true); + if (primary_hubp && primary_hubp->funcs->hubp_enable_3dlut_fl) { + primary_hubp->funcs->hubp_enable_3dlut_fl(primary_hubp, true); } } @@ -382,8 +384,11 @@ bool dcn401_set_mcm_luts(struct pipe_ctx *pipe_ctx, const struct dc_plane_state *plane_state) { struct dc *dc = pipe_ctx->plane_res.hubp->ctx->dc; + const struct pipe_ctx *primary_dpp_pipe_ctx = resource_get_primary_dpp_pipe(pipe_ctx); struct dpp *dpp_base = pipe_ctx->plane_res.dpp; struct hubp *hubp = pipe_ctx->plane_res.hubp; + struct hubp *primary_hubp = primary_dpp_pipe_ctx ? + primary_dpp_pipe_ctx->plane_res.hubp : NULL; const struct dc_plane_cm *cm = &plane_state->cm; int mpcc_id = hubp->inst; struct mpc *mpc = dc->res_pool->mpc; @@ -481,25 +486,41 @@ bool dcn401_set_mcm_luts(struct pipe_ctx *pipe_ctx, mpc->funcs->program_lut_read_write_control(mpc, MCM_LUT_3DLUT, lut_bank_a, 12, mpcc_id); if (mpc->funcs->update_3dlut_fast_load_select) - mpc->funcs->update_3dlut_fast_load_select(mpc, mpcc_id, hubp->inst); + mpc->funcs->update_3dlut_fast_load_select(mpc, mpcc_id, primary_hubp->inst); /* HUBP */ - if (hubp->funcs->hubp_program_3dlut_fl_config) - hubp->funcs->hubp_program_3dlut_fl_config(hubp, &cm->lut3d_dma); + if (primary_hubp->inst == hubp->inst) { + /* only program if this is the primary dpp pipe for the given plane */ + if (hubp->funcs->hubp_program_3dlut_fl_config) + hubp->funcs->hubp_program_3dlut_fl_config(hubp, &cm->lut3d_dma); - if (hubp->funcs->hubp_program_3dlut_fl_crossbar) - hubp->funcs->hubp_program_3dlut_fl_crossbar(hubp, cm->lut3d_dma.format); + if (hubp->funcs->hubp_program_3dlut_fl_crossbar) + hubp->funcs->hubp_program_3dlut_fl_crossbar(hubp, cm->lut3d_dma.format); - if (hubp->funcs->hubp_program_3dlut_fl_addr) - hubp->funcs->hubp_program_3dlut_fl_addr(hubp, &cm->lut3d_dma.addr); + if (hubp->funcs->hubp_program_3dlut_fl_addr) + hubp->funcs->hubp_program_3dlut_fl_addr(hubp, &cm->lut3d_dma.addr); - if (hubp->funcs->hubp_enable_3dlut_fl) { - hubp->funcs->hubp_enable_3dlut_fl(hubp, true); + if (hubp->funcs->hubp_enable_3dlut_fl) { + hubp->funcs->hubp_enable_3dlut_fl(hubp, true); + } else { + /* GPU memory only supports fast load path */ + BREAK_TO_DEBUGGER(); + lut_enable = false; + result = false; + } } else { - /* GPU memory only supports fast load path */ - BREAK_TO_DEBUGGER(); - lut_enable = false; - result = false; + /* re-trigger priamry HUBP to load 3DLUT */ + if (primary_hubp->funcs->hubp_enable_3dlut_fl) { + primary_hubp->funcs->hubp_enable_3dlut_fl(primary_hubp, true); + } + + /* clear FL setup on this pipe's HUBP */ + memset(&lut3d_dma, 0, sizeof(lut3d_dma)); + if (hubp->funcs->hubp_program_3dlut_fl_config) + hubp->funcs->hubp_program_3dlut_fl_config(hubp, &lut3d_dma); + + if (hubp->funcs->hubp_enable_3dlut_fl) + hubp->funcs->hubp_enable_3dlut_fl(hubp, false); } } else { /* Legacy (Host) Load Mode */ @@ -1809,42 +1830,41 @@ void dcn401_perform_3dlut_wa_unlock(struct pipe_ctx *pipe_ctx) * This is meant to work around a known HW issue where VREADY will cancel the pending 3DLUT_ENABLE signal regardless * of whether OTG lock is currently being held or not. */ - struct pipe_ctx *wa_pipes[MAX_PIPES] = { NULL }; - struct pipe_ctx *odm_pipe, *mpc_pipe; - int i, wa_pipe_ct = 0; + const struct pipe_ctx *otg_master_pipe_ctx = resource_get_otg_master(pipe_ctx); + struct timing_generator *tg = otg_master_pipe_ctx ? + otg_master_pipe_ctx->stream_res.tg : NULL; + const struct pipe_ctx *primary_dpp_pipe_ctx = resource_is_pipe_type(pipe_ctx, DPP_PIPE) ? + resource_get_primary_dpp_pipe(pipe_ctx) : pipe_ctx; + struct hubp *primary_hubp = primary_dpp_pipe_ctx ? + primary_dpp_pipe_ctx->plane_res.hubp : NULL; - for (odm_pipe = pipe_ctx; odm_pipe != NULL; odm_pipe = odm_pipe->next_odm_pipe) { - for (mpc_pipe = odm_pipe; mpc_pipe != NULL; mpc_pipe = mpc_pipe->bottom_pipe) { - if (mpc_pipe->plane_state && - mpc_pipe->plane_state->cm.flags.bits.lut3d_enable && - mpc_pipe->plane_state->cm.flags.bits.lut3d_dma_enable) { - wa_pipes[wa_pipe_ct++] = mpc_pipe; - } - } + if (!otg_master_pipe_ctx && !tg) { + return; } - if (wa_pipe_ct > 0) { - if (pipe_ctx->stream_res.tg->funcs->set_vupdate_keepout) - pipe_ctx->stream_res.tg->funcs->set_vupdate_keepout(pipe_ctx->stream_res.tg, true); + if (primary_dpp_pipe_ctx && + primary_dpp_pipe_ctx->plane_state && + primary_dpp_pipe_ctx->plane_state->cm.flags.bits.lut3d_enable && + primary_dpp_pipe_ctx->plane_state->cm.flags.bits.lut3d_dma_enable) { + if (tg->funcs->set_vupdate_keepout) + tg->funcs->set_vupdate_keepout(tg, true); - for (i = 0; i < wa_pipe_ct; ++i) { - if (wa_pipes[i]->plane_res.hubp->funcs->hubp_enable_3dlut_fl) - wa_pipes[i]->plane_res.hubp->funcs->hubp_enable_3dlut_fl(wa_pipes[i]->plane_res.hubp, true); + if (primary_hubp->funcs->hubp_enable_3dlut_fl) { + primary_hubp->funcs->hubp_enable_3dlut_fl(primary_hubp, true); } - pipe_ctx->stream_res.tg->funcs->unlock(pipe_ctx->stream_res.tg); - if (pipe_ctx->stream_res.tg->funcs->wait_update_lock_status) - pipe_ctx->stream_res.tg->funcs->wait_update_lock_status(pipe_ctx->stream_res.tg, false); + tg->funcs->unlock(tg); + if (tg->funcs->wait_update_lock_status) + tg->funcs->wait_update_lock_status(tg, false); - for (i = 0; i < wa_pipe_ct; ++i) { - if (wa_pipes[i]->plane_res.hubp->funcs->hubp_enable_3dlut_fl) - wa_pipes[i]->plane_res.hubp->funcs->hubp_enable_3dlut_fl(wa_pipes[i]->plane_res.hubp, true); + if (primary_hubp->funcs->hubp_enable_3dlut_fl) { + primary_hubp->funcs->hubp_enable_3dlut_fl(primary_hubp, true); } - if (pipe_ctx->stream_res.tg->funcs->set_vupdate_keepout) - pipe_ctx->stream_res.tg->funcs->set_vupdate_keepout(pipe_ctx->stream_res.tg, false); + if (tg->funcs->set_vupdate_keepout) + tg->funcs->set_vupdate_keepout(tg, false); } else { - pipe_ctx->stream_res.tg->funcs->unlock(pipe_ctx->stream_res.tg); + tg->funcs->unlock(tg); } } diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.h b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.h index f78162ab859b..b9a03ffa2717 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.h +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.h @@ -41,8 +41,7 @@ bool dcn401_set_mcm_luts(struct pipe_ctx *pipe_ctx, bool dcn401_set_output_transfer_func(struct dc *dc, struct pipe_ctx *pipe_ctx, const struct dc_stream_state *stream); -void dcn401_trigger_3dlut_dma_load(struct dc *dc, - struct pipe_ctx *pipe_ctx); +void dcn401_trigger_3dlut_dma_load(struct pipe_ctx *pipe_ctx); void dcn401_calculate_dccg_tmds_div_value(struct pipe_ctx *pipe_ctx, unsigned int *tmds_div); enum dc_status dcn401_enable_stream_timing( 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 d1dba7ffcd9b..98abe0d2d30f 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer.h @@ -1120,7 +1120,7 @@ struct hw_sequencer_funcs { void (*program_output_csc)(struct dc *dc, struct pipe_ctx *pipe_ctx, enum dc_color_space colorspace, uint16_t *matrix, int opp_id); - void (*trigger_3dlut_dma_load)(struct dc *dc, struct pipe_ctx *pipe_ctx); + void (*trigger_3dlut_dma_load)(struct pipe_ctx *pipe_ctx); /* VM Related */ int (*init_sys_ctx)(struct dce_hwseq *hws, From 3cf496080f539b69f069027cece82642b46a6cb5 Mon Sep 17 00:00:00 2001 From: Nicholas Kazlauskas Date: Tue, 10 Mar 2026 16:33:44 -0400 Subject: [PATCH 39/99] drm/amd/display: Revert inbox0 lock for cursor due to deadlock [Why] A deadlock occurs when using inbox0 lock for cursor operations on PSR-SU and Replays that does not when using the inbox1 locking path. This is because of a priority inversion issue where inbox1 work cannot be serviced while holding the HW lock from driver and sending cursor notifications to DMUB. Typically the lower priority of inbox1 for the lock command would allow the PSR and Replay FSMs to complete their transition prior to giving driver the lock but this is no longer the case with inbox0 having the highest priority in servicing. [How] This will reintroduce any synchronization bugs that were there with Replay or PSR-SU touching the cursor at the same time as driver. Reviewed-by: Charlene Liu Signed-off-by: Nicholas Kazlauskas Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/core/dc_stream.c | 30 +++---------------- 1 file changed, 4 insertions(+), 26 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c index 908f79b02102..daa7ab362239 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c @@ -33,7 +33,6 @@ #include "dc_dmub_srv.h" #include "dc_state_priv.h" #include "dc_stream_priv.h" -#include "dce/dmub_hw_lock_mgr.h" #define DC_LOGGER dc->ctx->logger #ifndef MIN @@ -259,7 +258,6 @@ void program_cursor_attributes( struct resource_context *res_ctx; struct pipe_ctx *pipe_to_program = NULL; bool enable_cursor_offload = dc_dmub_srv_is_cursor_offload_enabled(dc); - bool unlock_dmub = false; if (!stream) return; @@ -278,12 +276,6 @@ void program_cursor_attributes( if (enable_cursor_offload && dc->hwss.begin_cursor_offload_update) { dc->hwss.begin_cursor_offload_update(dc, pipe_ctx); } else { - if (dc->hwss.dmub_hw_control_lock && pipe_ctx->stream && - should_use_dmub_inbox0_lock_for_link(dc, pipe_ctx->stream->link)) { - dc->hwss.dmub_hw_control_lock(dc, dc->current_state, true); - unlock_dmub = true; - } - dc->hwss.cursor_lock(dc, pipe_to_program, true); if (pipe_to_program->next_odm_pipe) dc->hwss.cursor_lock(dc, pipe_to_program->next_odm_pipe, true); @@ -306,9 +298,6 @@ void program_cursor_attributes( dc->hwss.cursor_lock(dc, pipe_to_program, false); if (pipe_to_program->next_odm_pipe) dc->hwss.cursor_lock(dc, pipe_to_program->next_odm_pipe, false); - - if (unlock_dmub) - dc->hwss.dmub_hw_control_lock(dc, dc->current_state, false); } } } @@ -416,7 +405,6 @@ void program_cursor_position( struct resource_context *res_ctx; struct pipe_ctx *pipe_to_program = NULL; bool enable_cursor_offload = dc_dmub_srv_is_cursor_offload_enabled(dc); - bool unlock_dmub = false; if (!stream) return; @@ -436,16 +424,10 @@ void program_cursor_position( if (!pipe_to_program) { pipe_to_program = pipe_ctx; - if (enable_cursor_offload && dc->hwss.begin_cursor_offload_update) { + if (enable_cursor_offload && dc->hwss.begin_cursor_offload_update) dc->hwss.begin_cursor_offload_update(dc, pipe_ctx); - } else { - if (dc->hwss.dmub_hw_control_lock && pipe_ctx->stream && - should_use_dmub_inbox0_lock_for_link(dc, pipe_ctx->stream->link)) { - dc->hwss.dmub_hw_control_lock(dc, dc->current_state, true); - unlock_dmub = true; - } + else dc->hwss.cursor_lock(dc, pipe_to_program, true); - } } dc->hwss.set_cursor_position(pipe_ctx); @@ -457,14 +439,10 @@ void program_cursor_position( } if (pipe_to_program) { - if (enable_cursor_offload && dc->hwss.commit_cursor_offload_update) { + if (enable_cursor_offload && dc->hwss.commit_cursor_offload_update) dc->hwss.commit_cursor_offload_update(dc, pipe_to_program); - } else { + else dc->hwss.cursor_lock(dc, pipe_to_program, false); - - if (unlock_dmub) - dc->hwss.dmub_hw_control_lock(dc, dc->current_state, false); - } } } From 05112f1fc363d19105ecfce3eec637e63b50cde8 Mon Sep 17 00:00:00 2001 From: Ivan Lipski Date: Tue, 3 Mar 2026 20:07:58 -0500 Subject: [PATCH 40/99] drm/amd/display: Move DPM clk read to clk_mgr_construct in DCN42 [Why&How] The DPM clocks on DCN42 are currently read on every dm_resume, which can cause in gpu memory freeing while the device is still in suspend. Move the DPM clock read functionality to clk_mgr_construct() so it completes once on driver enablement. Reviewed-by: Charlene Liu Signed-off-by: Ivan Lipski Signed-off-by: Dmytro Laktyushkin Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index 24834f89711d..5671fe481d15 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -1146,6 +1146,11 @@ void dcn42_clk_mgr_construct( dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 1; clk_mgr->base.base.dprefclk_khz = dcn42_smu_get_dprefclk(&clk_mgr->base); clk_mgr->base.base.clks.ref_dtbclk_khz = dcn42_smu_get_dtbclk(&clk_mgr->base); + + clk_mgr->base.base.bw_params = &dcn42_bw_params; + + if (clk_mgr->base.smu_present) + dcn42_get_smu_clocks(&clk_mgr->base); } /* in case we don't get a value from the BIOS, use default */ if (clk_mgr->base.base.dentist_vco_freq_khz == 0) From 64fd3f93a85051d765d04efa59c13f03fdc9d286 Mon Sep 17 00:00:00 2001 From: Matthew Stewart Date: Wed, 11 Mar 2026 15:16:00 -0400 Subject: [PATCH 41/99] drm/amd/display: Hardcode dtbclk value in bw_params [why&how] dtbclk should always be 600MHz. Previous logic was to get the real value from SMU, but this returns 0 when dtbclk is off. Not a problem during boot when pre-OS enables dtbclk, but PnP was broken due to this. Reviewed-by: Charlene Liu Reviewed-by: Nicholas Kazlauskas Signed-off-by: Matthew Stewart Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index 5671fe481d15..b4c6522e922c 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -594,10 +594,7 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) dcn42_dump_clk_registers(&clk_mgr_base->boot_snapshot, clk_mgr); clk_mgr_base->clks.ref_dtbclk_khz = clk_mgr_base->boot_snapshot.dtbclk * 10; - if (clk_mgr_base->boot_snapshot.dtbclk > 59000) { - /*dtbclk enabled based on*/ - clk_mgr_base->clks.dtbclk_en = true; - } + clk_mgr_base->clks.dtbclk_en = clk_mgr_base->boot_snapshot.dtbclk > 59000; } static struct clk_bw_params dcn42_bw_params = { @@ -1069,7 +1066,7 @@ static void dcn42_get_smu_clocks(struct clk_mgr_internal *clk_mgr_int) clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_memclk_levels = dpm_clks->NumMemPstatesEnabled; /* DTBCLK*/ - clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz = clk_mgr_base->clks.ref_dtbclk_khz / 1000; + clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz = 600; /* Fixed on platform */ clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dtbclk_levels = 1; } } From 370b132a3ed5c80093b668bb518984bfff6638fb Mon Sep 17 00:00:00 2001 From: Gaghik Khachatrian Date: Thu, 12 Mar 2026 10:21:43 -0400 Subject: [PATCH 42/99] drm/amd/display: Fix Signed/Unsigned Int Usage Compiler Warning [Why] Compiler generates compiler warnings when signed enum constants or literal -1 are implicitly converted to unsigned integer types, cluttering build output and masking genuine issues. [How] Use UINT_MAX as the invalid sentinel for unsigned IDs and align loop/index types to unsigned where appropriate to remove implicit signed-to-unsigned conversions, with no functional behavior change. Reviewed-by: Nicholas Kazlauskas Signed-off-by: Gaghik Khachatrian Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dml2_0/dml2_dc_resource_mgmt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_dc_resource_mgmt.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_dc_resource_mgmt.c index 74812a7d5e28..fd3c61509f1b 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_dc_resource_mgmt.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_dc_resource_mgmt.c @@ -143,7 +143,7 @@ static unsigned int find_pipes_assigned_to_plane(struct dml2_context *ctx, { int i; unsigned int num_found = 0; - unsigned int plane_id_assigned_to_pipe = -1; + unsigned int plane_id_assigned_to_pipe = UINT_MAX; for (i = 0; i < ctx->config.dcn_pipe_count; i++) { struct pipe_ctx *pipe = &state->res_ctx.pipe_ctx[i]; From a1d840a51aa3c30c025a581af10058be2ada301c Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Wed, 11 Mar 2026 17:05:19 -0400 Subject: [PATCH 43/99] drm/amd/display: remove dc_clock_limit for apu [why] current apu pmfw does not support dc_clock_limit Reviewed-by: Roman Li Signed-off-by: Charlene Liu Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c index fa5d0558192c..6328b3dc35f9 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c @@ -2018,7 +2018,7 @@ static bool dcn42_resource_construct( dc->config.dcn_override_sharpness_range.hdr_rgb_mid = 1500; dc->config.use_pipe_ctx_sync_logic = true; - dc->config.dc_mode_clk_limit_support = true; + dc->config.dc_mode_clk_limit_support = false; dc->config.enable_windowed_mpo_odm = true; /* Use psp mailbox to enable assr */ dc->config.use_assr_psp_message = true; From 0e4e84de43d6bc8ec04be80d851f9007348702de Mon Sep 17 00:00:00 2001 From: ChunTao Tso Date: Wed, 4 Mar 2026 16:37:58 +0800 Subject: [PATCH 44/99] drm/amd/display: Agument live + capture with CVT case. 1. Add LIVE_CAPTURE_WITH_CVT bit (bit[2]) in union replay_optimization to control this feature via DalRegKey_ReplayOptimization. 2. Check the bit in mod_power_set_live_capture_with_cvt_activate function before enabling live capture with CVT. 3. Use LIVE_CAPTURE_WITH_CVT to control if Replay want to send CVT in live + capture or not. Reviewed-by: Leon Huang Signed-off-by: ChunTao Tso Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc_types.h | 8 ++++++-- drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h | 1 + 2 files changed, 7 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index e224077c8902..fd8ec1660312 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -1144,10 +1144,12 @@ union replay_low_refresh_rate_enable_options { union replay_optimization { struct { - //BIT[0-3]: Replay Teams Optimization + //BIT[0-1]: Replay Teams Optimization unsigned int TEAMS_OPTIMIZATION_VER_1 :1; unsigned int TEAMS_OPTIMIZATION_VER_2 :1; - unsigned int RESERVED_2_3 :2; + //BIT[2]: Replay Live Capture with CVT + unsigned int LIVE_CAPTURE_WITH_CVT :1; + unsigned int RESERVED_3 :1; } bits; unsigned int raw; @@ -1196,6 +1198,8 @@ struct replay_config { bool frame_skip_supported; /* Replay Received Frame Skipping Error HPD. */ bool received_frame_skipping_error_hpd; + /* Live capture with CVT is activated */ + bool live_capture_with_cvt_activated; }; /* Replay feature flags*/ diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index e11e32afac6b..8fbd179a4c87 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -4437,6 +4437,7 @@ enum dmub_cmd_replay_general_subtype { REPLAY_GENERAL_CMD_VIDEO_CONFERENCING, REPLAY_GENERAL_CMD_SET_CONTINUOUSLY_RESYNC, REPLAY_GENERAL_CMD_SET_COASTING_VTOTAL_WITHOUT_FRAME_UPDATE, + REPLAY_GENERAL_CMD_LIVE_CAPTURE_WITH_CVT, }; struct dmub_alpm_auxless_data { From 601fcb5f05a62c20857a39dac175285c3b664743 Mon Sep 17 00:00:00 2001 From: Wayne Lin Date: Fri, 13 Mar 2026 12:05:40 +0800 Subject: [PATCH 45/99] drm/amd/display: Remove unnecessary completion flag for secure display The completion flag is not used in secure display today. Remove unnecessary code. Reviewed-by: Tom Chung Signed-off-by: Wayne Lin Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.c | 7 ------- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h | 1 - 2 files changed, 8 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 b96fbc03c371..62573173e2ac 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 @@ -812,7 +812,6 @@ void amdgpu_dm_crtc_handle_crc_window_irq(struct drm_crtc *crtc) unsigned long flags1; bool forward_roi_change = false; bool notify_ta = false; - bool all_crc_ready = true; struct dc_stream_state *stream_state; int i; @@ -936,9 +935,6 @@ void amdgpu_dm_crtc_handle_crc_window_irq(struct drm_crtc *crtc) continue; } - if (!crtc_ctx->crc_info.crc[i].crc_ready) - all_crc_ready = false; - if (reset_crc_frame_count[i] || crtc_ctx->crc_info.crc[i].frame_count == UINT_MAX) /* Reset the reference frame count after user update the ROI * or it reaches the maximum value. @@ -948,9 +944,6 @@ void amdgpu_dm_crtc_handle_crc_window_irq(struct drm_crtc *crtc) crtc_ctx->crc_info.crc[i].frame_count += 1; } spin_unlock_irqrestore(&crtc_ctx->crc_info.lock, flags1); - - if (all_crc_ready) - complete_all(&crtc_ctx->crc_info.completion); } void amdgpu_dm_crtc_secure_display_create_contexts(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h index 95bdb8699d7f..8538513ea879 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crc.h @@ -70,7 +70,6 @@ struct crc_data { struct crc_info { struct crc_data crc[MAX_CRC_WINDOW_NUM]; - struct completion completion; spinlock_t lock; }; From 94cd4c4d71e67e84c713492c5eea6c2370601bf2 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Thu, 12 Mar 2026 19:33:33 -0400 Subject: [PATCH 46/99] drm/amd/display: add dccg FGCG mask init [why] missing DCCG_GLOBAL_FGCG_REP_DIS mask macro init Reviewed-by: Swapnil Patel Signed-off-by: Charlene Liu Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h index d9831b0f8235..2076565b1caa 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h @@ -122,6 +122,7 @@ DCCG_SF(DCCG_GATE_DISABLE_CNTL2, PHYBSYMCLK_ROOT_GATE_DISABLE, mask_sh),\ DCCG_SF(DCCG_GATE_DISABLE_CNTL2, PHYCSYMCLK_ROOT_GATE_DISABLE, mask_sh),\ DCCG_SF(DCCG_GATE_DISABLE_CNTL2, PHYDSYMCLK_ROOT_GATE_DISABLE, mask_sh),\ + DCCG_SF(DCCG_GLOBAL_FGCG_REP_CNTL, DCCG_GLOBAL_FGCG_REP_DIS, mask_sh),\ DCCG_SFII(OTG, PIXEL_RATE_CNTL, DP_DTO, ENABLE, 0, mask_sh),\ DCCG_SFII(OTG, PIXEL_RATE_CNTL, DP_DTO, ENABLE, 1, mask_sh),\ DCCG_SFII(OTG, PIXEL_RATE_CNTL, DP_DTO, ENABLE, 2, mask_sh),\ From 3967ab83a5075c05be3c614f1dc8460f66ee77ad Mon Sep 17 00:00:00 2001 From: Joshua Aberback Date: Thu, 12 Mar 2026 18:33:49 -0400 Subject: [PATCH 47/99] drm/amd/display: Restore full update for tiling change to linear [Why] There was previously a dc debug flag to indicate that tiling changes should only be a medium update instead of full. The function get_plane_info_type was refactored to not rely on dc state, but in the process the logic was unintentionally changed, which leads to screen corruption in some cases. [How] - add flag to tiling struct to avoid full update when necessary Reviewed-by: Nicholas Kazlauskas Reviewed-by: Aric Cyr Signed-off-by: Joshua Aberback Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 26 ++++---------------- drivers/gpu/drm/amd/display/dc/dc_hw_types.h | 4 +++ 2 files changed, 9 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 1c57bfd1fd2d..7dac3f35f0e8 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -2769,28 +2769,12 @@ static struct surface_update_descriptor get_plane_info_update_type(const struct if (memcmp(tiling, &u->surface->tiling_info, sizeof(*tiling)) != 0) { update_flags->bits.swizzle_change = 1; - elevate_update_type(&update_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); - switch (tiling->gfxversion) { - case DcGfxVersion9: - case DcGfxVersion10: - case DcGfxVersion11: - if (tiling->gfx9.swizzle != DC_SW_LINEAR) { - update_flags->bits.bandwidth_change = 1; - elevate_update_type(&update_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); - } - break; - case DcGfxAddr3: - if (tiling->gfx_addr3.swizzle != DC_ADDR3_SW_LINEAR) { - update_flags->bits.bandwidth_change = 1; - elevate_update_type(&update_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); - } - break; - case DcGfxVersion7: - case DcGfxVersion8: - case DcGfxVersionUnknown: - default: - break; + if (tiling->flags.avoid_full_update_on_tiling_change) { + elevate_update_type(&update_type, UPDATE_TYPE_MED, LOCK_DESCRIPTOR_STREAM); + } else { + update_flags->bits.bandwidth_change = 1; + elevate_update_type(&update_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); } } diff --git a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h index 9bf853edc46f..595e082db281 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h @@ -445,6 +445,10 @@ enum dc_gfxversion { enum swizzle_mode_addr3_values swizzle; } gfx_addr3;/*gfx with addr3 and above*/ }; + + struct { + bool avoid_full_update_on_tiling_change; + } flags; }; /* Rotation angle */ From 19b79e4f21825b5e61ac9ac6db76327677dd8b6e Mon Sep 17 00:00:00 2001 From: Relja Vojvodic Date: Wed, 11 Mar 2026 17:02:24 -0400 Subject: [PATCH 48/99] drm/amd/display: Rework YCbCr422 DSC policy - Reworked YCbCr4:2:2 Native/Simple policy decision making with DSC enabled based on DSC caps and stream signal type Reviewed-by: Wenjing Liu Signed-off-by: Relja Vojvodic Signed-off-by: Chuanyu Tseng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc.h | 2 +- drivers/gpu/drm/amd/display/dc/dc_dsc.h | 1 + .../amd/display/dc/dml2_0/dml2_translation_helper.c | 6 +++--- drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c | 13 ++++++------- .../gpu/drm/amd/display/dc/dsc/dcn20/dcn20_dsc.c | 2 +- .../gpu/drm/amd/display/dc/dsc/dcn35/dcn35_dsc.c | 2 +- .../gpu/drm/amd/display/dc/dsc/dcn401/dcn401_dsc.c | 2 +- .../gpu/drm/amd/display/dc/link/link_detection.c | 10 +++++----- drivers/gpu/drm/amd/display/dc/link/link_dpms.c | 3 ++- .../amd/display/dc/resource/dcn31/dcn31_resource.c | 2 ++ .../display/dc/resource/dcn315/dcn315_resource.c | 2 ++ 11 files changed, 25 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 1dd0a2cf786c..2bb4c7751c01 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -563,6 +563,7 @@ struct dc_config { bool frame_update_cmd_version2; struct spl_sharpness_range dcn_sharpness_range; struct spl_sharpness_range dcn_override_sharpness_range; + bool no_native422_support; }; enum visual_confirm { @@ -987,7 +988,6 @@ struct link_service; * causing an issue or not. */ struct dc_debug_options { - bool native422_support; bool disable_dsc; enum visual_confirm visual_confirm; int visual_confirm_rect_height; diff --git a/drivers/gpu/drm/amd/display/dc/dc_dsc.h b/drivers/gpu/drm/amd/display/dc/dc_dsc.h index 9d18f1c08079..101bce6b8de6 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dsc.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dsc.h @@ -52,6 +52,7 @@ struct dc_dsc_policy { uint32_t max_target_bpp; uint32_t min_target_bpp; bool enable_dsc_when_not_needed; + bool ycbcr422_simple; }; struct dc_dsc_config_options { diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_translation_helper.c index d834cb595afa..e25b88f2d6b9 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_translation_helper.c @@ -1174,12 +1174,12 @@ static unsigned int map_plane_to_dml_display_cfg(const struct dml2_context *dml2 const struct dc_state *context, const struct dml_display_cfg_st *dml_dispcfg, unsigned int stream_id, int plane_index) { unsigned int plane_id; - int i = 0; - int location = -1; + unsigned int i = 0; + unsigned int location = UINT_MAX; if (!get_plane_id(context->bw_ctx.dml2, context, plane, stream_id, plane_index, &plane_id)) { ASSERT(false); - return -1; + return UINT_MAX; } for (i = 0; i < __DML2_WRAPPER_MAX_STREAMS_PLANES__; i++) { diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c index 5b3584ad5b6b..8dfb6dd14eb2 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c @@ -680,9 +680,6 @@ static void get_dsc_enc_caps( } else { build_dsc_enc_caps(dsc, dsc_enc_caps); } - - if (dsc->ctx->dc->debug.native422_support) - dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 1; } /* Returns 'false' if no intersection was found for at least one capability. @@ -1100,13 +1097,14 @@ static bool setup_dsc_config( branch_max_throughput_mps = dsc_sink_caps->branch_overall_throughput_0_mps; break; case PIXEL_ENCODING_YCBCR422: - is_dsc_possible = (bool)dsc_common_caps.color_formats.bits.YCBCR_NATIVE_422; - sink_per_slice_throughput_mps = dsc_sink_caps->throughput_mode_1_mps; - branch_max_throughput_mps = dsc_sink_caps->branch_overall_throughput_1_mps; - if (!is_dsc_possible) { + if (policy.ycbcr422_simple) { is_dsc_possible = (bool)dsc_common_caps.color_formats.bits.YCBCR_SIMPLE_422; dsc_cfg->ycbcr422_simple = is_dsc_possible; sink_per_slice_throughput_mps = dsc_sink_caps->throughput_mode_0_mps; + } else { + is_dsc_possible = (bool)dsc_common_caps.color_formats.bits.YCBCR_NATIVE_422; + sink_per_slice_throughput_mps = dsc_sink_caps->throughput_mode_1_mps; + branch_max_throughput_mps = dsc_sink_caps->branch_overall_throughput_1_mps; } break; case PIXEL_ENCODING_YCBCR420: @@ -1406,6 +1404,7 @@ void dc_dsc_get_policy_for_timing(const struct dc_crtc_timing *timing, policy->min_target_bpp = 8; /* DP specs limits to 3 x bpc */ policy->max_target_bpp = 3 * bpc; + policy->ycbcr422_simple = true; break; case PIXEL_ENCODING_YCBCR420: /* DP specs limits to 6 */ diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dcn20/dcn20_dsc.c b/drivers/gpu/drm/amd/display/dc/dsc/dcn20/dcn20_dsc.c index 242f1e6f0d8f..6e1e759462bf 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/dcn20/dcn20_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/dcn20/dcn20_dsc.c @@ -100,7 +100,7 @@ void dsc2_get_enc_caps(struct dsc_enc_caps *dsc_enc_caps, int pixel_clock_100Hz) dsc_enc_caps->color_formats.bits.RGB = 1; dsc_enc_caps->color_formats.bits.YCBCR_444 = 1; dsc_enc_caps->color_formats.bits.YCBCR_SIMPLE_422 = 1; - dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 0; + dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 1; dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_420 = 1; dsc_enc_caps->color_depth.bits.COLOR_DEPTH_8_BPC = 1; diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dcn35/dcn35_dsc.c b/drivers/gpu/drm/amd/display/dc/dsc/dcn35/dcn35_dsc.c index e712985f7abd..17acb64a9d80 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/dcn35/dcn35_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/dcn35/dcn35_dsc.c @@ -128,7 +128,7 @@ void dsc35_get_single_enc_caps(struct dsc_enc_caps *dsc_enc_caps, unsigned int m dsc_enc_caps->color_formats.bits.RGB = 1; dsc_enc_caps->color_formats.bits.YCBCR_444 = 1; dsc_enc_caps->color_formats.bits.YCBCR_SIMPLE_422 = 1; - dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 0; + dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 1; dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_420 = 1; dsc_enc_caps->color_depth.bits.COLOR_DEPTH_8_BPC = 1; diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dcn401/dcn401_dsc.c b/drivers/gpu/drm/amd/display/dc/dsc/dcn401/dcn401_dsc.c index c1bdbb38c690..bbb8b5b18a4e 100644 --- a/drivers/gpu/drm/amd/display/dc/dsc/dcn401/dcn401_dsc.c +++ b/drivers/gpu/drm/amd/display/dc/dsc/dcn401/dcn401_dsc.c @@ -78,7 +78,7 @@ static void dsc401_get_single_enc_caps(struct dsc_enc_caps *dsc_enc_caps, unsign dsc_enc_caps->color_formats.bits.RGB = 1; dsc_enc_caps->color_formats.bits.YCBCR_444 = 1; dsc_enc_caps->color_formats.bits.YCBCR_SIMPLE_422 = 1; - dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 0; + dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_422 = 1; dsc_enc_caps->color_formats.bits.YCBCR_NATIVE_420 = 1; dsc_enc_caps->color_depth.bits.COLOR_DEPTH_8_BPC = 1; diff --git a/drivers/gpu/drm/amd/display/dc/link/link_detection.c b/drivers/gpu/drm/amd/display/dc/link/link_detection.c index f992c2d16748..7f1761080aba 100644 --- a/drivers/gpu/drm/amd/display/dc/link/link_detection.c +++ b/drivers/gpu/drm/amd/display/dc/link/link_detection.c @@ -781,7 +781,6 @@ static void restore_phy_clocks_for_destructive_link_verification(const struct dc } static void verify_link_capability_destructive(struct dc_link *link, - struct dc_sink *sink, enum dc_detect_reason reason) { bool should_prepare_phy_clocks = @@ -855,11 +854,11 @@ static bool should_verify_link_capability_destructively(struct dc_link *link, return destrictive; } -static void verify_link_capability(struct dc_link *link, struct dc_sink *sink, +static void verify_link_capability(struct dc_link *link, enum dc_detect_reason reason) { if (should_verify_link_capability_destructively(link, reason)) - verify_link_capability_destructive(link, sink, reason); + verify_link_capability_destructive(link, reason); else verify_link_capability_non_destructive(link); } @@ -1453,8 +1452,9 @@ bool link_detect(struct dc_link *link, enum dc_detect_reason reason) is_local_sink_detect_success = detect_link_and_local_sink(link, reason); - if (is_local_sink_detect_success && link->local_sink) - verify_link_capability(link, link->local_sink, reason); + if (is_local_sink_detect_success && link->local_sink) { + verify_link_capability(link, reason); + } DC_LOG_DC("%s: link_index=%d is_local_sink_detect_success=%d pre_link_type=%d link_type=%d\n", __func__, link->link_index, is_local_sink_detect_success, pre_link_type, link->type); diff --git a/drivers/gpu/drm/amd/display/dc/link/link_dpms.c b/drivers/gpu/drm/amd/display/dc/link/link_dpms.c index b4f46408a000..e12c25896364 100644 --- a/drivers/gpu/drm/amd/display/dc/link/link_dpms.c +++ b/drivers/gpu/drm/amd/display/dc/link/link_dpms.c @@ -181,7 +181,8 @@ void link_set_all_streams_dpms_off_for_link(struct dc_link *link) /* link can be also enabled by vbios. In this case it is not recorded * in pipe_ctx. Disable link phy here to make sure it is completely off */ - dp_disable_link_phy(link, &link_res, link->connector_signal); + if (dc_is_dp_signal(link->connector_signal)) + dp_disable_link_phy(link, &link_res, link->connector_signal); } void link_resume(struct dc_link *link) 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 046724c86c7a..6ce6b2b1f288 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 @@ -1963,6 +1963,8 @@ static bool dcn31_resource_construct( dc->config.use_pipe_ctx_sync_logic = true; dc->config.disable_hbr_audio_dp2 = true; + dc->config.no_native422_support = true; + /* read VBIOS LTTPR caps */ { if (ctx->dc_bios->funcs->get_lttpr_caps) { diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c index 4d5fcd7a0b00..0e0c52128c55 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c @@ -1925,6 +1925,8 @@ static bool dcn315_resource_construct( dc->caps.color.mpc.ogam_rom_caps.hlg = 0; dc->caps.color.mpc.ocsc = 1; + dc->config.no_native422_support = true; + /* read VBIOS LTTPR caps */ { if (ctx->dc_bios->funcs->get_lttpr_caps) { From 38f471d38be6629a38ed12377937927b8086894e Mon Sep 17 00:00:00 2001 From: Taimur Hassan Date: Fri, 13 Mar 2026 17:42:59 -0500 Subject: [PATCH 49/99] drm/amd/display: Promote DC to 3.2.375 This version brings along following fixes: - Rework YCbCr422 DSC policy - Restore full update for tiling change to linear - add dccg FGCG mask init - Remove unnecessary completion flag for secure display - Agument live + capture with CVT case. - remove dc_clock_limit for apu - Fix Signed/Unsigned Int Usage Compiler Warning - Hardcode dtbclk value in bw_params - Revert inbox0 lock for cursor due to deadlock - Add 3DLUT DMA broadcast support - Fix Silence warnings - export get_power_profile interface for later use - pg cntl update based on previous asic. - remove disable_sutter touch pstate debug code - Refactor DC update checks - Fix drm_edid leak in amdgpu_dm - Add Extra SMU Log for dtbclk - Clamp min DS DCFCLK value to DCN limit - Update dpia supported configuration - Multiple DCN42 updates Acked-by: ChiaHsuan Chung Signed-off-by: Taimur Hassan Signed-off-by: Chuanyu Tseng 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 2bb4c7751c01..80e217c5a23d 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -63,7 +63,7 @@ struct dcn_dsc_reg_state; struct dcn_optc_reg_state; struct dcn_dccg_reg_state; -#define DC_VER "3.2.374" +#define DC_VER "3.2.375" /** * MAX_SURFACES - representative of the upper bound of surfaces that can be piped to a single CRTC From e65e749941bb686bb7a6dce996c44707bb8474be Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Thu, 26 Feb 2026 11:41:48 +0800 Subject: [PATCH 50/99] drm/amdgpu: Add amdgpu_regs_pcie64 debugfs node Add amdgpu_regs_pcie64 debugfs node to read/write 64bit PCIE registers. Signed-off-by: Stanley.Yang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 112 ++++++++++++++++++++ 1 file changed, 112 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index aa9239b310a3..092fd3309099 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -617,6 +617,110 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user return r; } +/** + * amdgpu_debugfs_regs_pcie64_read - Read from a 64-bit PCIE register + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + */ +static ssize_t amdgpu_debugfs_regs_pcie64_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x7 || *pos & 0x7) + return -EINVAL; + + r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + r = amdgpu_virt_enable_access_debugfs(adev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + while (size) { + uint64_t value; + + value = RREG64_PCIE_EXT(*pos); + + r = put_user(value, (uint64_t *)buf); + if (r) + goto out; + + result += 8; + buf += 8; + *pos += 8; + size -= 8; + } + + r = result; +out: + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + amdgpu_virt_disable_access_debugfs(adev); + return r; +} + +/** + * amdgpu_debugfs_regs_pcie64_write - Write to a 64-bit PCIE register + * + * @f: open file handle + * @buf: User buffer to write data from + * @size: Number of bytes to write + * @pos: Offset to seek to + */ +static ssize_t amdgpu_debugfs_regs_pcie64_write(struct file *f, const char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x7 || *pos & 0x7) + return -EINVAL; + + r = pm_runtime_get_sync(adev_to_drm(adev)->dev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + r = amdgpu_virt_enable_access_debugfs(adev); + if (r < 0) { + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + return r; + } + + while (size) { + uint64_t value; + + r = get_user(value, (uint64_t *)buf); + if (r) + goto out; + + WREG64_PCIE_EXT(*pos, value); + + result += 8; + buf += 8; + *pos += 8; + size -= 8; + } + + r = result; +out: + pm_runtime_put_autosuspend(adev_to_drm(adev)->dev); + amdgpu_virt_disable_access_debugfs(adev); + return r; +} + /** * amdgpu_debugfs_regs_didt_read - Read from a DIDT register * @@ -1525,6 +1629,12 @@ static const struct file_operations amdgpu_debugfs_regs_pcie_fops = { .write = amdgpu_debugfs_regs_pcie_write, .llseek = default_llseek }; +static const struct file_operations amdgpu_debugfs_regs_pcie64_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_regs_pcie64_read, + .write = amdgpu_debugfs_regs_pcie64_write, + .llseek = default_llseek +}; static const struct file_operations amdgpu_debugfs_regs_smc_fops = { .owner = THIS_MODULE, .read = amdgpu_debugfs_regs_smc_read, @@ -1587,6 +1697,7 @@ static const struct file_operations *debugfs_regs[] = { &amdgpu_debugfs_gprwave_fops, &amdgpu_debugfs_regs_didt_fops, &amdgpu_debugfs_regs_pcie_fops, + &amdgpu_debugfs_regs_pcie64_fops, &amdgpu_debugfs_regs_smc_fops, &amdgpu_debugfs_gca_config_fops, &amdgpu_debugfs_sensors_fops, @@ -1604,6 +1715,7 @@ static const char * const debugfs_regs_names[] = { "amdgpu_gprwave", "amdgpu_regs_didt", "amdgpu_regs_pcie", + "amdgpu_regs_pcie64", "amdgpu_regs_smc", "amdgpu_gca_config", "amdgpu_sensors", From 17f11bbbc76c8e83c8474ea708316b1e3631d927 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 18 Mar 2026 13:48:30 +0800 Subject: [PATCH 51/99] drm/amd/pm: Skip redundant UCLK restore in smu_v13_0_6 Only reapply UCLK soft limits during PP_OD_RESTORE_DEFAULT when the current max differs from the DPM table max. This avoids redundant SMC updates and prevents -EINVAL on restore when no change is needed. Fixes: b7a900344546 ("drm/amd/pm: Allow setting max UCLK on SMU v13.0.6") 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 | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) 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 bc361fee5777..13a544e8cf85 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 @@ -2125,6 +2125,7 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, { struct smu_dpm_context *smu_dpm = &(smu->smu_dpm); struct smu_13_0_dpm_context *dpm_context = smu_dpm->dpm_context; + struct smu_dpm_table *uclk_table = &dpm_context->dpm_tables.uclk_table; struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; uint32_t min_clk; uint32_t max_clk; @@ -2224,14 +2225,16 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, if (ret) return ret; - min_clk = SMU_DPM_TABLE_MIN( - &dpm_context->dpm_tables.uclk_table); - max_clk = SMU_DPM_TABLE_MAX( - &dpm_context->dpm_tables.uclk_table); - ret = smu_v13_0_6_set_soft_freq_limited_range( - smu, SMU_UCLK, min_clk, max_clk, false); - if (ret) - return ret; + if (SMU_DPM_TABLE_MAX(uclk_table) != + pstate_table->uclk_pstate.curr.max) { + min_clk = SMU_DPM_TABLE_MIN(&dpm_context->dpm_tables.uclk_table); + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.uclk_table); + ret = smu_v13_0_6_set_soft_freq_limited_range(smu, + SMU_UCLK, min_clk, + max_clk, false); + if (ret) + return ret; + } smu_v13_0_reset_custom_level(smu); } break; From d82e0a72d9189e8acd353988e1a57f85ce479e37 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 18 Mar 2026 13:52:57 +0800 Subject: [PATCH 52/99] drm/amd/pm: Return -EOPNOTSUPP for unsupported OD_MCLK on smu_v13_0_6 When SET_UCLK_MAX capability is absent, return -EOPNOTSUPP from smu_v13_0_6_emit_clk_levels() for OD_MCLK instead of 0. This makes unsupported OD_MCLK reporting consistent with other clock types and allows callers to skip the entry cleanly. Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) 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 13a544e8cf85..1af1c4f6dc09 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 @@ -1394,7 +1394,7 @@ static int smu_v13_0_6_emit_clk_levels(struct smu_context *smu, break; case SMU_OD_MCLK: if (!smu_v13_0_6_cap_supported(smu, SMU_CAP(SET_UCLK_MAX))) - return 0; + return -EOPNOTSUPP; size += sysfs_emit_at(buf, size, "%s:\n", "OD_MCLK"); size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMhz\n", From 069a3b425317de22ac5d14163d4045ad6aba164a Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 13 Mar 2026 13:39:21 +0800 Subject: [PATCH 53/99] drm/amd/pm: Add OD_FCLK interface Add OD_FCLK interface to set customa fclk max v2: Merge patch1 & 3, check EOPNOTSUPP for all clks (Lijo) v3: Remove redundant check (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/include/kgd_pp_interface.h | 2 ++ drivers/gpu/drm/amd/pm/amdgpu_pm.c | 18 +++++++++++------- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 2 ++ drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h | 1 + 4 files changed, 16 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index a9b73f4fd466..33a1404bb666 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -119,6 +119,7 @@ enum pp_clock_type { PP_ISPXCLK, OD_SCLK, OD_MCLK, + OD_FCLK, OD_VDDC_CURVE, OD_RANGE, OD_VDDGFX_OFFSET, @@ -208,6 +209,7 @@ enum { enum PP_OD_DPM_TABLE_COMMAND { PP_OD_EDIT_SCLK_VDDC_TABLE, PP_OD_EDIT_MCLK_VDDC_TABLE, + PP_OD_EDIT_FCLK_TABLE, PP_OD_EDIT_CCLK_VDDC_TABLE, PP_OD_EDIT_VDDC_CURVE, PP_OD_RESTORE_DEFAULT_TABLE, diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 9e2f69663f96..dbf3ae2f5e13 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -680,6 +680,8 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * - minimum(not available for Vega20 and Navi1x) and maximum memory * clock labeled OD_MCLK * + * - minimum and maximum fabric clock labeled OD_FCLK (SMU13) + * * - three points labeled OD_VDDC_CURVE. * They can be used to calibrate the sclk voltage curve. This is * available for Vega20 and NV1X. @@ -715,10 +717,11 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * - First select manual using power_dpm_force_performance_level * * - For clock frequency setting, enter a new value by writing a - * string that contains "s/m index clock" to the file. The index + * string that contains "s/m/f index clock" to the file. The index * should be 0 if to set minimum clock. And 1 if to set maximum * clock. E.g., "s 0 500" will update minimum sclk to be 500 MHz. - * "m 1 800" will update maximum mclk to be 800Mhz. For core + * "m 1 800" will update maximum mclk to be 800Mhz. "f 1 1600" will + * update maximum fabric clock to be 1600Mhz. For core * clocks on VanGogh, the string contains "p core index clock". * E.g., "p 2 0 800" would set the minimum core clock on core * 2 to 800Mhz. @@ -768,6 +771,8 @@ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, type = PP_OD_EDIT_CCLK_VDDC_TABLE; else if (*buf == 'm') type = PP_OD_EDIT_MCLK_VDDC_TABLE; + else if (*buf == 'f') + type = PP_OD_EDIT_FCLK_TABLE; else if (*buf == 'r') type = PP_OD_RESTORE_DEFAULT_TABLE; else if (*buf == 'c') @@ -843,9 +848,10 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, struct amdgpu_device *adev = drm_to_adev(ddev); int size = 0; int ret; - enum pp_clock_type od_clocks[6] = { + enum pp_clock_type od_clocks[] = { OD_SCLK, OD_MCLK, + OD_FCLK, OD_VDDC_CURVE, OD_RANGE, OD_VDDGFX_OFFSET, @@ -857,10 +863,8 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, if (ret) return ret; - for (clk_index = 0 ; clk_index < 6 ; clk_index++) { - ret = amdgpu_dpm_emit_clock_levels(adev, od_clocks[clk_index], buf, &size); - if (ret) - break; + for (clk_index = 0 ; clk_index < ARRAY_SIZE(od_clocks) ; clk_index++) { + amdgpu_dpm_emit_clock_levels(adev, od_clocks[clk_index], buf, &size); } if (size == 0) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 26b43a6276d0..7a2c0a5fd140 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -3056,6 +3056,8 @@ static enum smu_clk_type smu_convert_to_smuclk(enum pp_clock_type type) clk_type = SMU_OD_SCLK; break; case OD_MCLK: clk_type = SMU_OD_MCLK; break; + case OD_FCLK: + clk_type = SMU_OD_FCLK; break; case OD_VDDC_CURVE: clk_type = SMU_OD_VDDC_CURVE; break; case OD_RANGE: diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h index 584c4cfd0c16..8cdbaf32492e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h @@ -324,6 +324,7 @@ enum smu_clk_type { SMU_OD_CCLK, SMU_OD_SCLK, SMU_OD_MCLK, + SMU_OD_FCLK, SMU_OD_VDDC_CURVE, SMU_OD_RANGE, SMU_OD_VDDGFX_OFFSET, From 5da89a8afca14a266e82d2c99e0bd5ae6db23611 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 13 Mar 2026 17:11:03 +0800 Subject: [PATCH 54/99] drm/amd/pm: Add custom fclk setting support Add custom fclk setting support for smu_v13_x_x v2: Move uclk fix to separate patch, return EOPNOTSUPP in case of dpm disabled (Lijo) v3: remove dpm check for filling fclk pstate table (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c | 2 + .../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 82 ++++++++++++++++++- 2 files changed, 83 insertions(+), 1 deletion(-) 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 a3b755c61b1f..be9a7a32de99 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 @@ -2466,4 +2466,6 @@ void smu_v13_0_reset_custom_level(struct smu_context *smu) pstate_table->uclk_pstate.custom.max = 0; pstate_table->gfxclk_pstate.custom.min = 0; pstate_table->gfxclk_pstate.custom.max = 0; + pstate_table->fclk_pstate.custom.min = 0; + pstate_table->fclk_pstate.custom.max = 0; } 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 1af1c4f6dc09..312ed1475e6b 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 @@ -1199,6 +1199,7 @@ static int smu_v13_0_6_populate_umd_state_clk(struct smu_context *smu) struct smu_dpm_table *gfx_table = &dpm_context->dpm_tables.gfx_table; struct smu_dpm_table *mem_table = &dpm_context->dpm_tables.uclk_table; struct smu_dpm_table *soc_table = &dpm_context->dpm_tables.soc_table; + struct smu_dpm_table *fclk_table = &dpm_context->dpm_tables.fclk_table; struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; pstate_table->gfxclk_pstate.min = SMU_DPM_TABLE_MIN(gfx_table); @@ -1216,6 +1217,12 @@ static int smu_v13_0_6_populate_umd_state_clk(struct smu_context *smu) pstate_table->socclk_pstate.curr.min = SMU_DPM_TABLE_MIN(soc_table); pstate_table->socclk_pstate.curr.max = SMU_DPM_TABLE_MAX(soc_table); + pstate_table->fclk_pstate.min = SMU_DPM_TABLE_MIN(fclk_table); + pstate_table->fclk_pstate.peak = SMU_DPM_TABLE_MAX(fclk_table); + pstate_table->fclk_pstate.curr.min = SMU_DPM_TABLE_MIN(fclk_table); + pstate_table->fclk_pstate.curr.max = SMU_DPM_TABLE_MAX(fclk_table); + pstate_table->fclk_pstate.standard = SMU_DPM_TABLE_MIN(fclk_table); + if (gfx_table->count > SMU_13_0_6_UMD_PSTATE_GFXCLK_LEVEL && mem_table->count > SMU_13_0_6_UMD_PSTATE_MCLK_LEVEL && soc_table->count > SMU_13_0_6_UMD_PSTATE_SOCCLK_LEVEL) { @@ -1401,7 +1408,15 @@ static int smu_v13_0_6_emit_clk_levels(struct smu_context *smu, pstate_table->uclk_pstate.curr.min, pstate_table->uclk_pstate.curr.max); break; + case SMU_OD_FCLK: + if (!smu_cmn_feature_is_enabled(smu, SMU_FEATURE_DPM_FCLK_BIT)) + return -EOPNOTSUPP; + size += sysfs_emit_at(buf, size, "%s:\n", "OD_FCLK"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMhz\n", + pstate_table->fclk_pstate.curr.min, + pstate_table->fclk_pstate.curr.max); + break; case SMU_SCLK: case SMU_GFXCLK: single_dpm_table = &(dpm_context->dpm_tables.gfx_table); @@ -2043,7 +2058,7 @@ static int smu_v13_0_6_set_soft_freq_limited_range(struct smu_context *smu, int ret = 0; if (clk_type != SMU_GFXCLK && clk_type != SMU_SCLK && - clk_type != SMU_UCLK) + clk_type != SMU_UCLK && clk_type != SMU_FCLK) return -EINVAL; if ((smu_dpm->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) && @@ -2084,6 +2099,15 @@ static int smu_v13_0_6_set_soft_freq_limited_range(struct smu_context *smu, pstate_table->uclk_pstate.curr.max = max; } + if (clk_type == SMU_FCLK) { + if (max == pstate_table->fclk_pstate.curr.max) + return 0; + + ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_FCLK, 0, max, false); + if (!ret) + pstate_table->fclk_pstate.curr.max = max; + } + return ret; } @@ -2126,6 +2150,7 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, struct smu_dpm_context *smu_dpm = &(smu->smu_dpm); struct smu_13_0_dpm_context *dpm_context = smu_dpm->dpm_context; struct smu_dpm_table *uclk_table = &dpm_context->dpm_tables.uclk_table; + struct smu_dpm_table *fclk_table = &dpm_context->dpm_tables.fclk_table; struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; uint32_t min_clk; uint32_t max_clk; @@ -2206,6 +2231,40 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, pstate_table->uclk_pstate.custom.max = input[1]; } break; + case PP_OD_EDIT_FCLK_TABLE: + if (size != 2) { + dev_err(smu->adev->dev, + "Input parameter number not correct\n"); + return -EINVAL; + } + + if (!smu_cmn_feature_is_enabled(smu, + SMU_FEATURE_DPM_FCLK_BIT)) { + dev_warn(smu->adev->dev, + "FCLK limits setting not supported!\n"); + return -EOPNOTSUPP; + } + + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.fclk_table); + if (input[0] == 0) { + dev_info(smu->adev->dev, + "Setting min FCLK level is not supported\n"); + return -EOPNOTSUPP; + } else if (input[0] == 1) { + if (input[1] > max_clk) { + dev_warn(smu->adev->dev, + "Maximum FCLK (%ld) MHz specified is greater than the maximum allowed (%d) MHz\n", + input[1], max_clk); + pstate_table->fclk_pstate.custom.max = + pstate_table->fclk_pstate.curr.max; + return -EINVAL; + } + + pstate_table->fclk_pstate.custom.max = input[1]; + } else { + return -EINVAL; + } + break; case PP_OD_RESTORE_DEFAULT_TABLE: if (size != 0) { @@ -2235,6 +2294,17 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, if (ret) return ret; } + + if (SMU_DPM_TABLE_MAX(fclk_table) != + pstate_table->fclk_pstate.curr.max) { + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.fclk_table); + min_clk = SMU_DPM_TABLE_MIN(&dpm_context->dpm_tables.fclk_table); + ret = smu_v13_0_6_set_soft_freq_limited_range(smu, + SMU_FCLK, min_clk, + max_clk, false); + if (ret) + return ret; + } smu_v13_0_reset_custom_level(smu); } break; @@ -2261,6 +2331,16 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, if (ret) return ret; + if (pstate_table->fclk_pstate.custom.max) { + min_clk = pstate_table->fclk_pstate.curr.min; + max_clk = pstate_table->fclk_pstate.custom.max; + ret = smu_v13_0_6_set_soft_freq_limited_range(smu, + SMU_FCLK, min_clk, + max_clk, false); + if (ret) + return ret; + } + if (!pstate_table->uclk_pstate.custom.max) return 0; From 8f402ddd4f2321485fff7384c003124681a08146 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Fri, 20 Mar 2026 17:29:01 +0530 Subject: [PATCH 55/99] drm/amdgpu/userq: cleanup amdgpu_userq_get/put where not needed MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit amdgpu_userq_put/get are not needed in case we already holding the userq_mutex and reference is valid already from queue create time or from signal ioctl. These additional get/put could be a potential reason for deadlock in case the ref count reaches zero and destroy is called which again try to take the userq_mutex. Due to the above change we avoid deadlock between suspend/restore calling destroy queues trying to take userq_mutex again. Cc: Prike Liang Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 21 +++------------------ 1 file changed, 3 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index d94f4966fea9..b75b44a3169b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -999,15 +999,11 @@ amdgpu_userq_restore_all(struct amdgpu_userq_mgr *uq_mgr) /* Resume all the queues for this process */ xa_for_each(&uq_mgr->userq_xa, queue_id, queue) { - queue = amdgpu_userq_get(uq_mgr, queue_id); - if (!queue) - continue; if (!amdgpu_userq_buffer_vas_mapped(queue)) { drm_file_err(uq_mgr->file, "trying restore queue without va mapping\n"); queue->state = AMDGPU_USERQ_STATE_INVALID_VA; - amdgpu_userq_put(queue); continue; } @@ -1015,7 +1011,6 @@ amdgpu_userq_restore_all(struct amdgpu_userq_mgr *uq_mgr) if (r) ret = r; - amdgpu_userq_put(queue); } if (ret) @@ -1252,13 +1247,9 @@ amdgpu_userq_evict_all(struct amdgpu_userq_mgr *uq_mgr) amdgpu_userq_detect_and_reset_queues(uq_mgr); /* Try to unmap all the queues in this process ctx */ xa_for_each(&uq_mgr->userq_xa, queue_id, queue) { - queue = amdgpu_userq_get(uq_mgr, queue_id); - if (!queue) - continue; r = amdgpu_userq_preempt_helper(queue); if (r) ret = r; - amdgpu_userq_put(queue); } if (ret) @@ -1291,24 +1282,18 @@ amdgpu_userq_wait_for_signal(struct amdgpu_userq_mgr *uq_mgr) int ret; xa_for_each(&uq_mgr->userq_xa, queue_id, queue) { - queue = amdgpu_userq_get(uq_mgr, queue_id); - if (!queue) - continue; - struct dma_fence *f = queue->last_fence; - if (!f || dma_fence_is_signaled(f)) { - amdgpu_userq_put(queue); + if (!f || dma_fence_is_signaled(f)) continue; - } + ret = dma_fence_wait_timeout(f, true, msecs_to_jiffies(100)); if (ret <= 0) { drm_file_err(uq_mgr->file, "Timed out waiting for fence=%llu:%llu\n", f->context, f->seqno); - amdgpu_userq_put(queue); + return -ETIMEDOUT; } - amdgpu_userq_put(queue); } return 0; From 2beedc3a92b726f05c81127b7578faa505e4ba1e Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Mon, 3 Nov 2025 13:39:38 +0800 Subject: [PATCH 56/99] drm/amd/pm: Add initial support for smu v15_0_8 smu v15_0_8 is the new generation of smu ip block v2: Squash in rebase changes (Alex) v3: Squash in fw version check changes (Alex) Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 1 + drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h | 15 +- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h | 6 + drivers/gpu/drm/amd/pm/swsmu/smu15/Makefile | 2 +- .../gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c | 81 ++- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 483 ++++++++++++++++++ .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h | 28 + 7 files changed, 564 insertions(+), 52 deletions(-) create mode 100644 drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c create mode 100644 drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 7a2c0a5fd140..013d34e2f842 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -47,6 +47,7 @@ #include "smu_v14_0_0_ppt.h" #include "smu_v14_0_2_ppt.h" #include "smu_v15_0_0_ppt.h" +#include "smu_v15_0_8_ppt.h" #include "amd_pcie.h" /* diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h index 8cdbaf32492e..c7bfb4fc0e9a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h @@ -292,7 +292,12 @@ __SMU_DUMMY_MAP(AllowZstates), \ __SMU_DUMMY_MAP(GetSmartShiftStatus), \ __SMU_DUMMY_MAP(EnableLSdma), \ - __SMU_DUMMY_MAP(DisableLSdma), + __SMU_DUMMY_MAP(DisableLSdma), \ + __SMU_DUMMY_MAP(InitializeGfx), \ + __SMU_DUMMY_MAP(SetSoftMaxFclk), \ + __SMU_DUMMY_MAP(SetSoftMaxGl2clk), \ + __SMU_DUMMY_MAP(SetSoftMinGl2clk), \ + __SMU_DUMMY_MAP(GetSystemMetricsVersion), #undef __SMU_DUMMY_MAP #define __SMU_DUMMY_MAP(type) SMU_MSG_##type @@ -473,6 +478,14 @@ enum smu_clk_type { __SMU_DUMMY_MAP(GFX_DIDT_XVMIN), \ __SMU_DUMMY_MAP(FAN_ABNORMAL), \ __SMU_DUMMY_MAP(PIT), \ + __SMU_DUMMY_MAP(DS_DMABECLK), \ + __SMU_DUMMY_MAP(DS_MPIFOECLK), \ + __SMU_DUMMY_MAP(DS_MPRASCLK), \ + __SMU_DUMMY_MAP(DS_MPNHTCLK), \ + __SMU_DUMMY_MAP(DS_FIOCLK), \ + __SMU_DUMMY_MAP(DS_DXIOCLK), \ + __SMU_DUMMY_MAP(DS_GL2CLK), \ + __SMU_DUMMY_MAP(DPM_GL2CLK), \ __SMU_DUMMY_MAP(HROM_EN), #undef __SMU_DUMMY_MAP diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h index 09743ccd0d15..95c77e926e1f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h @@ -41,6 +41,10 @@ #define smnMP1_FIRMWARE_FLAGS 0x3010024 #define smnMP1_PUB_CTRL 0x3010d10 +#define SMU15_DRIVER_IF_VERSION_SMU_V15_0_8 0x007D0000 + +#define FEATURE_MASK(feature) (1ULL << feature) + #define MAX_DPM_LEVELS 16 #define MAX_PCIE_CONF 3 @@ -104,6 +108,7 @@ struct smu_15_0_dpm_context { struct smu_15_0_dpm_tables dpm_tables; uint32_t workload_policy_mask; uint32_t dcef_min_ds_clk; + uint64_t caps; }; enum smu_15_0_power_state { @@ -118,6 +123,7 @@ struct smu_15_0_power_context { uint32_t power_source; uint8_t in_power_limit_boost_mode; enum smu_15_0_power_state power_state; + atomic_t throttle_status; }; #if defined(SWSMU_CODE_LAYER_L2) || defined(SWSMU_CODE_LAYER_L3) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/Makefile b/drivers/gpu/drm/amd/pm/swsmu/smu15/Makefile index 7f59a0aabdeb..fa083ad46c0f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/Makefile +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/Makefile @@ -23,7 +23,7 @@ # Makefile for the 'smu manager' sub-component of powerplay. # It provides the smu management services for the driver. -SMU15_MGR = smu_v15_0.o smu_v15_0_0_ppt.o +SMU15_MGR = smu_v15_0.o smu_v15_0_0_ppt.o smu_v15_0_8_ppt.o AMD_SWSMU_SMU15MGR = $(addprefix $(AMD_SWSMU_PATH)/smu15/,$(SMU15_MGR)) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c index 09bb3d91a001..21df9705fda9 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c @@ -589,71 +589,52 @@ int smu_v15_0_notify_memory_pool_location(struct smu_context *smu) { struct smu_table_context *smu_table = &smu->smu_table; struct smu_table *memory_pool = &smu_table->memory_pool; - int ret = 0; - uint64_t address; - uint32_t address_low, address_high; + struct smu_msg_args args = { + .msg = SMU_MSG_DramLogSetDramAddr, + .num_args = 3, + .num_out_args = 0, + }; if (memory_pool->size == 0 || memory_pool->cpu_addr == NULL) - return ret; + return 0; - address = memory_pool->mc_address; - address_high = (uint32_t)upper_32_bits(address); - address_low = (uint32_t)lower_32_bits(address); + /* SMU_MSG_DramLogSetDramAddr: ARG0=low, ARG1=high, ARG2=size */ + args.args[0] = lower_32_bits(memory_pool->mc_address); + args.args[1] = upper_32_bits(memory_pool->mc_address); + args.args[2] = (u32)memory_pool->size; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_DramLogSetDramAddrHigh, - address_high, NULL); - if (ret) - return ret; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_DramLogSetDramAddrLow, - address_low, NULL); - if (ret) - return ret; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_DramLogSetDramSize, - (uint32_t)memory_pool->size, NULL); - if (ret) - return ret; - - return ret; + return smu->msg_ctl.ops->send_msg(&smu->msg_ctl, &args); } int smu_v15_0_set_driver_table_location(struct smu_context *smu) { struct smu_table *driver_table = &smu->smu_table.driver_table; - int ret = 0; + struct smu_msg_args args = { + .msg = SMU_MSG_SetDriverDramAddrHigh, + .num_args = 2, + .num_out_args = 0, + }; - if (driver_table->mc_address) { - ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetDriverDramAddrHigh, - upper_32_bits(driver_table->mc_address), - NULL); - if (!ret) - ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetDriverDramAddrLow, - lower_32_bits(driver_table->mc_address), - NULL); - } + args.args[0] = lower_32_bits(driver_table->mc_address); + args.args[1] = upper_32_bits(driver_table->mc_address); - return ret; + return smu->msg_ctl.ops->send_msg(&smu->msg_ctl, &args); } int smu_v15_0_set_tool_table_location(struct smu_context *smu) { - int ret = 0; struct smu_table *tool_table = &smu->smu_table.tables[SMU_TABLE_PMSTATUSLOG]; + struct smu_msg_args args = { + .msg = SMU_MSG_SetToolsDramAddrHigh, + .num_args = 2, + .num_out_args = 0, + }; - if (tool_table->mc_address) { - ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetToolsDramAddrHigh, - upper_32_bits(tool_table->mc_address), - NULL); - if (!ret) - ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetToolsDramAddrLow, - lower_32_bits(tool_table->mc_address), - NULL); - } + /* SMU_MSG_SetToolsDramAddr: ARG0=low, ARG1=high */ + args.args[0] = lower_32_bits(tool_table->mc_address); + args.args[1] = upper_32_bits(tool_table->mc_address); - return ret; + return smu->msg_ctl.ops->send_msg(&smu->msg_ctl, &args); } int smu_v15_0_set_allowed_mask(struct smu_context *smu) @@ -700,8 +681,7 @@ int smu_v15_0_gfx_off_control(struct smu_context *smu, bool enable) return ret; } -int smu_v15_0_system_features_control(struct smu_context *smu, - bool en) +int smu_v15_0_system_features_control(struct smu_context *smu, bool en) { return smu_cmn_send_smc_msg(smu, (en ? SMU_MSG_EnableAllSmuFeatures : SMU_MSG_DisableAllSmuFeatures), NULL); @@ -905,7 +885,8 @@ static int smu_v15_0_wait_for_reset_complete(struct smu_context *smu, return ret; } -int smu_v15_0_wait_for_event(struct smu_context *smu, enum smu_event_type event, +int smu_v15_0_wait_for_event(struct smu_context *smu, + enum smu_event_type event, uint64_t event_arg) { int ret = -EINVAL; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c new file mode 100644 index 000000000000..c4da0cdc1222 --- /dev/null +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -0,0 +1,483 @@ +/* + * 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 SWSMU_CODE_LAYER_L2 + +#include +#include "amdgpu.h" +#include "amdgpu_smu.h" +#include "smu_v15_0_8_pmfw.h" +#include "smu15_driver_if_v15_0_8.h" +#include "smu_v15_0_8_ppsmc.h" +#include "smu_v15_0_8_ppt.h" +#include +#include "smu_cmn.h" +#include "mp/mp_15_0_8_offset.h" +#include "mp/mp_15_0_8_sh_mask.h" +#include "smu_v15_0.h" + +#undef MP1_Public + +/* address block */ +#define MP1_Public 0x03b00000 +#define smnMP1_FIRMWARE_FLAGS_15_0_8 0x3010024 +/* + * DO NOT use these for err/warn/info/debug messages. + * Use dev_err, dev_warn, dev_info and dev_dbg instead. + * They are more MGPU friendly. + */ +#undef pr_err +#undef pr_warn +#undef pr_info +#undef pr_debug + +#define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) + +#define SMU_15_0_8_FEA_MAP(smu_feature, smu_15_0_8_feature) \ + [smu_feature] = { 1, (smu_15_0_8_feature) } + +#define FEATURE_MASK(feature) (1ULL << feature) + +static const struct smu_feature_bits smu_v15_0_8_dpm_features = { + .bits = { SMU_FEATURE_BIT_INIT(FEATURE_ID_DATA_CALCULATION), + SMU_FEATURE_BIT_INIT(FEATURE_ID_DPM_GFXCLK), + SMU_FEATURE_BIT_INIT(FEATURE_ID_DPM_UCLK), + SMU_FEATURE_BIT_INIT(FEATURE_ID_DPM_FCLK), + SMU_FEATURE_BIT_INIT(FEATURE_ID_DPM_GL2CLK) } +}; + +static const struct cmn2asic_msg_mapping smu_v15_0_8_message_map[SMU_MSG_MAX_COUNT] = { + MSG_MAP(TestMessage, PPSMC_MSG_TestMessage, 0), + MSG_MAP(GetSmuVersion, PPSMC_MSG_GetSmuVersion, 1), + MSG_MAP(GfxDeviceDriverReset, PPSMC_MSG_GfxDriverReset, SMU_MSG_RAS_PRI | SMU_MSG_NO_PRECHECK), + MSG_MAP(GetDriverIfVersion, PPSMC_MSG_GetDriverIfVersion, 1), + MSG_MAP(EnableAllSmuFeatures, PPSMC_MSG_EnableAllSmuFeatures, 0), + MSG_MAP(GetMetricsVersion, PPSMC_MSG_GetMetricsVersion, 1), + MSG_MAP(GetMetricsTable, PPSMC_MSG_GetMetricsTable, 1), + MSG_MAP(GetEnabledSmuFeatures, PPSMC_MSG_GetEnabledSmuFeatures, 1), + MSG_MAP(SetDriverDramAddrHigh, PPSMC_MSG_SetDriverDramAddr, 1), + MSG_MAP(SetToolsDramAddrHigh, PPSMC_MSG_SetToolsDramAddr, 0), + MSG_MAP(SetSoftMaxByFreq, PPSMC_MSG_SetSoftMaxByFreq, 1), + MSG_MAP(SetPptLimit, PPSMC_MSG_SetPptLimit, 0), + MSG_MAP(GetPptLimit, PPSMC_MSG_GetPptLimit, 1), + MSG_MAP(DramLogSetDramAddr, PPSMC_MSG_DramLogSetDramAddr, 0), + MSG_MAP(HeavySBR, PPSMC_MSG_HeavySBR, 0), + MSG_MAP(DFCstateControl, PPSMC_MSG_DFCstateControl, 0), + MSG_MAP(GfxDriverResetRecovery, PPSMC_MSG_GfxDriverResetRecovery, 0), + MSG_MAP(SetSoftMinGfxclk, PPSMC_MSG_SetSoftMinGfxClk, 1), + MSG_MAP(SetSoftMaxGfxClk, PPSMC_MSG_SetSoftMaxGfxClk, 1), + MSG_MAP(PrepareMp1ForUnload, PPSMC_MSG_PrepareForDriverUnload, 0), + MSG_MAP(QueryValidMcaCount, PPSMC_MSG_QueryValidMcaCount, SMU_MSG_RAS_PRI), + MSG_MAP(McaBankDumpDW, PPSMC_MSG_McaBankDumpDW, SMU_MSG_RAS_PRI), + MSG_MAP(ClearMcaOnRead, PPSMC_MSG_ClearMcaOnRead, 0), + MSG_MAP(QueryValidMcaCeCount, PPSMC_MSG_QueryValidMcaCeCount, SMU_MSG_RAS_PRI), + MSG_MAP(McaBankCeDumpDW, PPSMC_MSG_McaBankCeDumpDW, SMU_MSG_RAS_PRI), + MSG_MAP(SelectPLPDMode, PPSMC_MSG_SelectPLPDMode, 0), + MSG_MAP(SetThrottlingPolicy, PPSMC_MSG_SetThrottlingPolicy, 0), + MSG_MAP(ResetSDMA, PPSMC_MSG_ResetSDMA, 0), + MSG_MAP(GetRASTableVersion, PPSMC_MSG_GetRasTableVersion, 0), + MSG_MAP(SetTimestamp, PPSMC_MSG_SetTimestamp, 0), + MSG_MAP(GetTimestamp, PPSMC_MSG_GetTimestamp, 0), + MSG_MAP(GetBadPageIpid, PPSMC_MSG_GetBadPageIpIdLoHi, 0), + MSG_MAP(EraseRasTable, PPSMC_MSG_EraseRasTable, 0), + MSG_MAP(GetStaticMetricsTable, PPSMC_MSG_GetStaticMetricsTable, 1), + MSG_MAP(GetSystemMetricsTable, PPSMC_MSG_GetSystemMetricsTable, 1), + MSG_MAP(GetSystemMetricsVersion, PPSMC_MSG_GetSystemMetricsVersion, 0), + MSG_MAP(ResetVCN, PPSMC_MSG_ResetVCN, 0), + MSG_MAP(SetFastPptLimit, PPSMC_MSG_SetFastPptLimit, 0), + MSG_MAP(GetFastPptLimit, PPSMC_MSG_GetFastPptLimit, 0), + MSG_MAP(SetSoftMinGl2clk, PPSMC_MSG_SetSoftMinGl2clk, 0), + MSG_MAP(SetSoftMaxGl2clk, PPSMC_MSG_SetSoftMaxGl2clk, 0), + MSG_MAP(SetSoftMinFclk, PPSMC_MSG_SetSoftMinFclk, 0), + MSG_MAP(SetSoftMaxFclk, PPSMC_MSG_SetSoftMaxFclk, 0), +}; + +/* TODO: Update the clk map once enum PPCLK is updated in smu15_driver_if_v15_0_8.h */ +static struct cmn2asic_mapping smu_v15_0_8_clk_map[SMU_CLK_COUNT] = { + CLK_MAP(UCLK, PPCLK_UCLK), +}; + +static const struct cmn2asic_mapping smu_v15_0_8_feature_mask_map[SMU_FEATURE_COUNT] = { + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DATA_CALCULATIONS_BIT, FEATURE_ID_DATA_CALCULATION), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DPM_GFXCLK_BIT, FEATURE_ID_DPM_GFXCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DPM_UCLK_BIT, FEATURE_ID_DPM_UCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DPM_FCLK_BIT, FEATURE_ID_DPM_FCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DPM_GL2CLK_BIT, FEATURE_ID_DPM_GL2CLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_GFXCLK_BIT, FEATURE_ID_DS_GFXCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_SOCCLK_BIT, FEATURE_ID_DS_SOCCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_LCLK_BIT, FEATURE_ID_DS_LCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_FCLK_BIT, FEATURE_ID_DS_FCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_DMABECLK_BIT, FEATURE_ID_DS_DMABECLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_MPIFOECLK_BIT, FEATURE_ID_DS_MPIFOECLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_MPRASCLK_BIT, FEATURE_ID_DS_MPRASCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_MPNHTCLK_BIT, FEATURE_ID_DS_MPNHTCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_FIOCLK_BIT, FEATURE_ID_DS_FIOCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_DXIOCLK_BIT, FEATURE_ID_DS_DXIOCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_GL2CLK_BIT, FEATURE_ID_DS_GL2CLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_PPT_BIT, FEATURE_ID_PPT), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_TDC_BIT, FEATURE_ID_TDC), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_MP1_CG_BIT, FEATURE_ID_SMU_CG), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_FW_CTF_BIT, FEATURE_ID_FW_CTF), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_THERMAL_BIT, FEATURE_ID_THERMAL), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_SOC_PCC_BIT, FEATURE_ID_SOC_PCC), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_XGMI_PER_LINK_PWR_DWN_BIT, FEATURE_ID_XGMI_PER_LINK_PWR_DOWN), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_VCN_BIT, FEATURE_ID_DS_VCN), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_MP1CLK_BIT, FEATURE_ID_DS_MP1CLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_MPIOCLK_BIT, FEATURE_ID_DS_MPIOCLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_DS_MP0CLK_BIT, FEATURE_ID_DS_MP0CLK), + SMU_15_0_8_FEA_MAP(SMU_FEATURE_PIT_BIT, FEATURE_ID_PIT), +}; + +#define TABLE_PMSTATUSLOG 0 +#define TABLE_SMU_METRICS 1 +#define TABLE_I2C_COMMANDS 2 +#define TABLE_COUNT 3 + +static const struct cmn2asic_mapping smu_v15_0_8_table_map[SMU_TABLE_COUNT] = { + TAB_MAP(PMSTATUSLOG), + TAB_MAP(SMU_METRICS), + TAB_MAP(I2C_COMMANDS), +}; + +static int smu_v15_0_8_tables_init(struct smu_context *smu) +{ + return 0; +} + +static int smu_v15_0_8_allocate_dpm_context(struct smu_context *smu) +{ + struct smu_dpm_context *smu_dpm = &smu->smu_dpm; + + smu_dpm->dpm_context = + kzalloc(sizeof(struct smu_15_0_dpm_context), GFP_KERNEL); + if (!smu_dpm->dpm_context) + return -ENOMEM; + smu_dpm->dpm_context_size = sizeof(struct smu_15_0_dpm_context); + + smu_dpm->dpm_policies = + kzalloc(sizeof(struct smu_dpm_policy_ctxt), GFP_KERNEL); + if (!smu_dpm->dpm_policies) { + kfree(smu_dpm->dpm_context); + return -ENOMEM; + } + + return 0; +} + +static int smu_v15_0_8_init_smc_tables(struct smu_context *smu) +{ + int ret = 0; + + ret = smu_v15_0_8_tables_init(smu); + if (ret) + return ret; + + ret = smu_v15_0_8_allocate_dpm_context(smu); + + return ret; +} + +static int smu_v15_0_8_init_allowed_features(struct smu_context *smu) +{ + /* pptable will handle the features to enable */ + smu_feature_list_set_all(smu, SMU_FEATURE_LIST_ALLOWED); + + return 0; +} + +static int smu_v15_0_8_set_default_dpm_table(struct smu_context *smu) +{ + return 0; +} + +static int smu_v15_0_8_setup_pptable(struct smu_context *smu) +{ + struct smu_table_context *table_context = &smu->smu_table; + + /* TODO: PPTable is not available. + * 1) Find an alternate way to get 'PPTable values' here. + * 2) Check if there is SW CTF + */ + table_context->thermal_controller_type = 0; + + return 0; +} + +static int smu_v15_0_8_check_fw_status(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + uint32_t mp1_fw_flags; + + mp1_fw_flags = RREG32_PCIE(MP1_Public | + (smnMP1_FIRMWARE_FLAGS_15_0_8 & 0xffffffff)); + + if ((mp1_fw_flags & MP1_CRU1_MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) >> + MP1_CRU1_MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED__SHIFT) + return 0; + + return -EIO; +} + +static int smu_v15_0_8_irq_process(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + struct smu_context *smu = adev->powerplay.pp_handle; + struct smu_power_context *smu_power = &smu->smu_power; + struct smu_15_0_power_context *power_context = smu_power->power_context; + uint32_t client_id = entry->client_id; + uint32_t ctxid = entry->src_data[0]; + uint32_t src_id = entry->src_id; + uint32_t data; + + if (client_id == SOC_V1_0_IH_CLIENTID_MP1) { + if (src_id == IH_INTERRUPT_ID_TO_DRIVER) { + /* ACK SMUToHost interrupt */ + data = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL); + data = REG_SET_FIELD(data, MP1_SMN_IH_SW_INT_CTRL, INT_ACK, 1); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL, data); + /* + * ctxid is used to distinguish different events for SMCToHost + * interrupt. + */ + switch (ctxid) { + case IH_INTERRUPT_CONTEXT_ID_THERMAL_THROTTLING: + /* + * Increment the throttle interrupt counter + */ + atomic64_inc(&smu->throttle_int_counter); + + if (!atomic_read(&adev->throttling_logging_enabled)) + return 0; + + /* This uses the new method which fixes the + * incorrect throttling status reporting + * through metrics table. For older FWs, + * it will be ignored. + */ + if (__ratelimit(&adev->throttling_logging_rs)) { + atomic_set( + &power_context->throttle_status, + entry->src_data[1]); + schedule_work(&smu->throttling_logging_work); + } + break; + default: + dev_dbg(adev->dev, "Unhandled context id %d from client:%d!\n", + ctxid, client_id); + break; + } + } + } + + return 0; +} + +static int smu_v15_0_8_set_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + uint32_t val = 0; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + /* For MP1 SW irqs */ + val = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT_CTRL, INT_MASK, 1); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL, val); + + break; + case AMDGPU_IRQ_STATE_ENABLE: + /* For MP1 SW irqs */ + val = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT, ID, 0xFE); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT, VALID, 0); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT, val); + + val = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT_CTRL, INT_MASK, 0); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL, val); + + break; + default: + break; + } + + return 0; +} + +static const struct amdgpu_irq_src_funcs smu_v15_0_8_irq_funcs = { + .set = smu_v15_0_8_set_irq_state, + .process = smu_v15_0_8_irq_process, +}; + +static int smu_v15_0_8_register_irq_handler(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + struct amdgpu_irq_src *irq_src = &smu->irq_source; + int ret = 0; + + if (amdgpu_sriov_vf(adev)) + return 0; + + irq_src->num_types = 1; + irq_src->funcs = &smu_v15_0_8_irq_funcs; + + ret = amdgpu_irq_add_id(adev, SOC_V1_0_IH_CLIENTID_MP1, + IH_INTERRUPT_ID_TO_DRIVER, + irq_src); + if (ret) + return ret; + + return ret; +} + +static int smu_v15_0_8_notify_unload(struct smu_context *smu) +{ + if (amdgpu_in_reset(smu->adev)) + return 0; + + dev_dbg(smu->adev->dev, "Notify PMFW about driver unload"); + /* Ignore return, just intimate FW that driver is not going to be there */ + smu_cmn_send_smc_msg(smu, SMU_MSG_PrepareMp1ForUnload, NULL); + + return 0; +} + + +static int smu_v15_0_8_system_features_control(struct smu_context *smu, + bool enable) +{ + struct amdgpu_device *adev = smu->adev; + int ret = 0; + + if (amdgpu_sriov_vf(adev)) + return 0; + + if (enable) + ret = smu_v15_0_system_features_control(smu, enable); + else + smu_v15_0_8_notify_unload(smu); + + return ret; +} + +/** + * smu_v15_0_8_get_enabled_mask - Get enabled SMU features (128-bit) + * @smu: SMU context + * @feature_mask: feature mask structure + * + * SMU 15 returns all 128 feature bits in a single message via out_args[0..3]. + * For backward compatibility, this function returns only the first 64 bits. + * + * Return: 0 on success, negative errno on failure + */ +static int smu_v15_0_8_get_enabled_mask(struct smu_context *smu, + struct smu_feature_bits *feature_mask) +{ + struct smu_msg_args args = { + .msg = SMU_MSG_GetEnabledSmuFeatures, + .num_args = 0, + .num_out_args = 2, + }; + int ret; + + if (!feature_mask) + return -EINVAL; + + ret = smu->msg_ctl.ops->send_msg(&smu->msg_ctl, &args); + + if (ret) + return ret; + + smu_feature_bits_from_arr32(feature_mask, args.out_args, + SMU_FEATURE_NUM_DEFAULT); + + return 0; +} + +static bool smu_v15_0_8_is_dpm_running(struct smu_context *smu) +{ + int ret = 0; + struct smu_feature_bits feature_enabled; + + ret = smu_v15_0_8_get_enabled_mask(smu, &feature_enabled); + if (ret) + return false; + + return smu_feature_bits_test_mask(&feature_enabled, + smu_v15_0_8_dpm_features.bits); +} + +static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { + .init_allowed_features = smu_v15_0_8_init_allowed_features, + .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, + .is_dpm_running = smu_v15_0_8_is_dpm_running, + .init_smc_tables = smu_v15_0_8_init_smc_tables, + .fini_smc_tables = smu_v15_0_fini_smc_tables, + .init_power = smu_v15_0_init_power, + .fini_power = smu_v15_0_fini_power, + .check_fw_status = smu_v15_0_8_check_fw_status, + .check_fw_version = smu_cmn_check_fw_version, + .set_driver_table_location = smu_v15_0_set_driver_table_location, + .set_tool_table_location = smu_v15_0_set_tool_table_location, + .notify_memory_pool_location = smu_v15_0_notify_memory_pool_location, + .system_features_control = smu_v15_0_8_system_features_control, + .get_enabled_mask = smu_v15_0_8_get_enabled_mask, + .feature_is_enabled = smu_cmn_feature_is_enabled, + .register_irq_handler = smu_v15_0_8_register_irq_handler, + .setup_pptable = smu_v15_0_8_setup_pptable, + .get_pp_feature_mask = smu_cmn_get_pp_feature_mask, + .wait_for_event = smu_v15_0_wait_for_event, +}; + +static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, + const struct cmn2asic_msg_mapping *message_map) +{ + struct amdgpu_device *adev = smu->adev; + struct smu_msg_ctl *ctl = &smu->msg_ctl; + + ctl->smu = smu; + mutex_init(&ctl->lock); + ctl->config.msg_reg = SOC15_REG_OFFSET(MP1, 0, regMP1_SMN_C2PMSG_40); + ctl->config.resp_reg = SOC15_REG_OFFSET(MP1, 0, regMP1_SMN_C2PMSG_41); + ctl->config.arg_regs[0] = SOC15_REG_OFFSET(MP1, 0, regMP1_SMN_C2PMSG_42); + ctl->config.arg_regs[1] = SOC15_REG_OFFSET(MP1, 0, regMP1_SMN_C2PMSG_43); + ctl->config.arg_regs[2] = SOC15_REG_OFFSET(MP1, 0, regMP1_SMN_C2PMSG_44); + ctl->config.arg_regs[3] = SOC15_REG_OFFSET(MP1, 0, regMP1_SMN_C2PMSG_45); + ctl->config.num_arg_regs = 4; + ctl->ops = &smu_msg_v1_ops; + ctl->default_timeout = adev->usec_timeout * 20; + ctl->message_map = message_map; +} + +void smu_v15_0_8_set_ppt_funcs(struct smu_context *smu) +{ + smu->ppt_funcs = &smu_v15_0_8_ppt_funcs; + smu->clock_map = smu_v15_0_8_clk_map; + smu->feature_map = smu_v15_0_8_feature_mask_map; + smu->table_map = smu_v15_0_8_table_map; + smu_v15_0_8_init_msg_ctl(smu, smu_v15_0_8_message_map); + smu->smc_driver_if_version = SMU15_DRIVER_IF_VERSION_SMU_V15_0_8; +} diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h new file mode 100644 index 000000000000..40c410928c96 --- /dev/null +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h @@ -0,0 +1,28 @@ +/* + * 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. + * + */ +#ifndef __SMU_15_0_8_PPT_H__ +#define __SMU_15_0_8_PPT_H__ + +extern void smu_v15_0_8_set_ppt_funcs(struct smu_context *smu); + +#endif From fde815f76fa060ce5540776f2365f1bdf5cda845 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Tue, 25 Nov 2025 01:19:13 +0800 Subject: [PATCH 57/99] drm/amd/pm: Add mode2 support for smu_v15_0_8 Add initial mode2 support for smu_v15_0_8 v2: Move out non smu code, remove pci save/restore logic (Lijo) v3: squash in updated msg (Alex) Signed-off-by: Asad Kamal Reviewed-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 44 +++++++++++++++++++ 1 file changed, 44 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index c4da0cdc1222..b3dc04ea5e71 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -430,6 +430,49 @@ static bool smu_v15_0_8_is_dpm_running(struct smu_context *smu) smu_v15_0_8_dpm_features.bits); } +static int smu_v15_0_8_mode2_reset(struct smu_context *smu) +{ + struct smu_msg_ctl *ctl = &smu->msg_ctl; + struct amdgpu_device *adev = smu->adev; + int timeout = 10; + int ret = 0; + + mutex_lock(&ctl->lock); + + ret = smu_msg_send_async_locked(ctl, SMU_MSG_GfxDeviceDriverReset, + SMU_RESET_MODE_2); + + if (ret) + goto out; + + /* Reset takes a bit longer, wait for 200ms. */ + msleep(200); + + dev_dbg(adev->dev, "wait for reset ack\n"); + do { + ret = smu_msg_wait_response(ctl, 0); + /* Wait a bit more time for getting ACK */ + if (ret == -ETIME) { + --timeout; + usleep_range(500, 1000); + continue; + } + + if (ret) + goto out; + + } while (ret == -ETIME && timeout); + +out: + mutex_unlock(&ctl->lock); + + if (ret) + dev_err(adev->dev, "failed to send mode2 reset, error code %d", + ret); + + return ret; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -450,6 +493,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .setup_pptable = smu_v15_0_8_setup_pptable, .get_pp_feature_mask = smu_cmn_get_pp_feature_mask, .wait_for_event = smu_v15_0_wait_for_event, + .mode2_reset = smu_v15_0_8_mode2_reset, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, From 611e5af419c2d220c6fc7b45a013382fd1eef49b Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Wed, 26 Nov 2025 16:02:41 +0800 Subject: [PATCH 58/99] drm/amd/pm: Setup driver pptable for smu 15.0.8 Setup driver pptable and initialize data from static metrics table for smu_v15_0_8 v2: Remove unrelated changes and update description (Lijo) v3: Use ARRAY_SIZE (Lijo) v4: Move structure to header file v5: squash in static metrics support (Asad) Signed-off-by: Yang Wang Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h | 1 + .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 232 +++++++++++++++++- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h | 30 +++ 4 files changed, 258 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 892c90b8d063..49e7881750fa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -690,6 +690,7 @@ enum amdgpu_uid_type { AMDGPU_UID_TYPE_XCD, AMDGPU_UID_TYPE_AID, AMDGPU_UID_TYPE_SOC, + AMDGPU_UID_TYPE_MID, AMDGPU_UID_TYPE_MAX }; diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h index 95c77e926e1f..06842d38c92b 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h @@ -109,6 +109,7 @@ struct smu_15_0_dpm_context { uint32_t workload_policy_mask; uint32_t dcef_min_ds_clk; uint64_t caps; + uint32_t board_volt; }; enum smu_15_0_power_state { diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index b3dc04ea5e71..82b09fe7fccd 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -35,6 +35,7 @@ #include "mp/mp_15_0_8_offset.h" #include "mp/mp_15_0_8_sh_mask.h" #include "smu_v15_0.h" +#include "amdgpu_fru_eeprom.h" #undef MP1_Public @@ -51,6 +52,10 @@ #undef pr_info #undef pr_debug +#define SMUQ10_TO_UINT(x) ((x) >> 10) +#define SMUQ10_FRAC(x) ((x) & 0x3ff) +#define SMUQ10_ROUND(x) ((SMUQ10_TO_UINT(x)) + ((SMUQ10_FRAC(x)) >= 0x200)) + #define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) #define SMU_15_0_8_FEA_MAP(smu_feature, smu_15_0_8_feature) \ @@ -161,6 +166,35 @@ static const struct cmn2asic_mapping smu_v15_0_8_table_map[SMU_TABLE_COUNT] = { static int smu_v15_0_8_tables_init(struct smu_context *smu) { + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_table *tables = smu_table->tables; + int gpu_metrcs_size = sizeof(MetricsTable_t); + void *driver_pptable __free(kfree) = NULL; + void *metrics_table __free(kfree) = NULL; + + SMU_TABLE_INIT(tables, SMU_TABLE_PMSTATUSLOG, SMU15_TOOL_SIZE, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); + + SMU_TABLE_INIT(tables, SMU_TABLE_SMU_METRICS, + gpu_metrcs_size, + PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); + + metrics_table = kzalloc(gpu_metrcs_size, GFP_KERNEL); + if (!metrics_table) + return -ENOMEM; + + smu_table->metrics_time = 0; + + driver_pptable = kzalloc(sizeof(PPTable_t), GFP_KERNEL); + if (!driver_pptable) + return -ENOMEM; + + smu_table->metrics_table = no_free_ptr(metrics_table); + smu_table->driver_pptable = no_free_ptr(driver_pptable); + + mutex_init(&smu_table->metrics_lock); + return 0; } @@ -197,6 +231,30 @@ static int smu_v15_0_8_init_smc_tables(struct smu_context *smu) return ret; } +static int smu_v15_0_8_tables_fini(struct smu_context *smu) +{ + struct smu_table_context *smu_table = &smu->smu_table; + + mutex_destroy(&smu_table->metrics_lock); + + return 0; +} + +static int smu_v15_0_8_fini_smc_tables(struct smu_context *smu) +{ + int ret; + + ret = smu_v15_0_8_tables_fini(smu); + if (ret) + return ret; + + ret = smu_v15_0_fini_smc_tables(smu); + if (ret) + return ret; + + return ret; +} + static int smu_v15_0_8_init_allowed_features(struct smu_context *smu) { /* pptable will handle the features to enable */ @@ -205,11 +263,6 @@ static int smu_v15_0_8_init_allowed_features(struct smu_context *smu) return 0; } -static int smu_v15_0_8_set_default_dpm_table(struct smu_context *smu) -{ - return 0; -} - static int smu_v15_0_8_setup_pptable(struct smu_context *smu) { struct smu_table_context *table_context = &smu->smu_table; @@ -238,6 +291,173 @@ static int smu_v15_0_8_check_fw_status(struct smu_context *smu) return -EIO; } +static int smu_v15_0_8_get_static_metrics_table(struct smu_context *smu) +{ + struct smu_table_context *smu_table = &smu->smu_table; + uint32_t table_size = smu_table->tables[SMU_TABLE_SMU_METRICS].size; + struct smu_table *table = &smu_table->driver_table; + int ret; + + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetStaticMetricsTable, NULL); + if (ret) { + dev_err(smu->adev->dev, + "Failed to export static metrics table!\n"); + return ret; + } + + amdgpu_hdp_invalidate(smu->adev, NULL); + memcpy(smu_table->metrics_table, table->cpu_addr, table_size); + + return 0; +} + +static int smu_v15_0_8_fru_get_product_info(struct smu_context *smu, + StaticMetricsTable_t *static_metrics) +{ + struct amdgpu_fru_info *fru_info; + struct amdgpu_device *adev = smu->adev; + + if (!adev->fru_info) { + adev->fru_info = kzalloc(sizeof(*adev->fru_info), GFP_KERNEL); + if (!adev->fru_info) + return -ENOMEM; + } + + fru_info = adev->fru_info; + strscpy(fru_info->product_number, static_metrics->ProductInfo.ModelNumber, + sizeof(fru_info->product_number)); + strscpy(fru_info->product_name, static_metrics->ProductInfo.Name, + sizeof(fru_info->product_name)); + strscpy(fru_info->serial, static_metrics->ProductInfo.Serial, + sizeof(fru_info->serial)); + strscpy(fru_info->manufacturer_name, static_metrics->ProductInfo.ManufacturerName, + sizeof(fru_info->manufacturer_name)); + strscpy(fru_info->fru_id, static_metrics->ProductInfo.FruId, + sizeof(fru_info->fru_id)); + + return 0; +} + +static void smu_v15_0_8_init_xgmi_data(struct smu_context *smu, + StaticMetricsTable_t *static_metrics) +{ + uint16_t max_speed; + uint8_t max_width; + + max_width = (uint8_t)static_metrics->MaxXgmiWidth; + max_speed = (uint16_t)static_metrics->MaxXgmiBitrate; + amgpu_xgmi_set_max_speed_width(smu->adev, max_speed, max_width); +} + +static int smu_v15_0_8_set_driver_pptable(struct smu_context *smu) +{ + struct smu_15_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; + struct smu_table_context *smu_table = &smu->smu_table; + StaticMetricsTable_t *static_metrics = (StaticMetricsTable_t *)smu_table->metrics_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + int ret, i, n; + uint32_t table_version; + + if (!pptable->init) { + ret = smu_v15_0_8_get_static_metrics_table(smu); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetMetricsVersion, + &table_version); + if (ret) + return ret; + smu_table->tables[SMU_TABLE_SMU_METRICS].version = + table_version; + + pptable->MaxSocketPowerLimit = + SMUQ10_ROUND(static_metrics->MaxSocketPowerLimit); + pptable->MaxGfxclkFrequency = + SMUQ10_ROUND(static_metrics->MaxGfxclkFrequency); + pptable->MinGfxclkFrequency = + SMUQ10_ROUND(static_metrics->MinGfxclkFrequency); + pptable->MaxFclkFrequency = + SMUQ10_ROUND(static_metrics->MaxFclkFrequency); + pptable->MinFclkFrequency = + SMUQ10_ROUND(static_metrics->MinFclkFrequency); + pptable->MaxGl2clkFrequency = + SMUQ10_ROUND(static_metrics->MaxGl2clkFrequency); + pptable->MinGl2clkFrequency = + SMUQ10_ROUND(static_metrics->MinGl2clkFrequency); + + for (i = 0; i < ARRAY_SIZE(static_metrics->UclkFrequencyTable); ++i) + pptable->UclkFrequencyTable[i] = + SMUQ10_ROUND(static_metrics->UclkFrequencyTable[i]); + + pptable->SocclkFrequency = SMUQ10_ROUND(static_metrics->SocclkFrequency); + pptable->LclkFrequency = SMUQ10_ROUND(static_metrics->LclkFrequency); + pptable->VclkFrequency = SMUQ10_ROUND(static_metrics->VclkFrequency); + pptable->DclkFrequency = SMUQ10_ROUND(static_metrics->DclkFrequency); + + pptable->CTFLimitMID = SMUQ10_ROUND(static_metrics->CTFLimit_MID); + pptable->CTFLimitAID = SMUQ10_ROUND(static_metrics->CTFLimit_AID); + pptable->CTFLimitXCD = SMUQ10_ROUND(static_metrics->CTFLimit_XCD); + pptable->CTFLimitHBM = SMUQ10_ROUND(static_metrics->CTFLimit_HBM); + pptable->ThermalLimitMID = SMUQ10_ROUND(static_metrics->ThermalLimit_MID); + pptable->ThermalLimitAID = SMUQ10_ROUND(static_metrics->ThermalLimit_AID); + pptable->ThermalLimitXCD = SMUQ10_ROUND(static_metrics->ThermalLimit_XCD); + pptable->ThermalLimitHBM = SMUQ10_ROUND(static_metrics->ThermalLimit_HBM); + + /* use MID0 serial number by default */ + pptable->PublicSerialNumberMID = + static_metrics->PublicSerialNumber_MID[0]; + + amdgpu_device_set_uid(smu->adev->uid_info, AMDGPU_UID_TYPE_SOC, + 0, pptable->PublicSerialNumberMID); + pptable->PublicSerialNumberAID = + static_metrics->PublicSerialNumber_AID[0]; + pptable->PublicSerialNumberXCD = + static_metrics->PublicSerialNumber_XCD[0]; + n = ARRAY_SIZE(static_metrics->PublicSerialNumber_MID); + for (i = 0; i < n; i++) { + amdgpu_device_set_uid(smu->adev->uid_info, AMDGPU_UID_TYPE_MID, i, + static_metrics->PublicSerialNumber_MID[i]); + } + n = ARRAY_SIZE(static_metrics->PublicSerialNumber_AID); + for (i = 0; i < n; i++) { + amdgpu_device_set_uid(smu->adev->uid_info, AMDGPU_UID_TYPE_AID, i, + static_metrics->PublicSerialNumber_AID[i]); + } + n = ARRAY_SIZE(static_metrics->PublicSerialNumber_XCD); + for (i = 0; i < n; i++) { + amdgpu_device_set_uid(smu->adev->uid_info, AMDGPU_UID_TYPE_XCD, i, + static_metrics->PublicSerialNumber_XCD[i]); + } + + ret = smu_v15_0_8_fru_get_product_info(smu, static_metrics); + if (ret) + return ret; + pptable->PPT1Max = static_metrics->PPT1Max; + pptable->PPT1Min = static_metrics->PPT1Min; + pptable->PPT1Default = static_metrics->PPT1Default; + + if (static_metrics->pldmVersion[0] != 0xFFFFFFFF) + smu->adev->firmware.pldm_version = + static_metrics->pldmVersion[0]; + dpm_context->board_volt = static_metrics->InputTelemetryVoltageInmV; + smu_v15_0_8_init_xgmi_data(smu, static_metrics); + pptable->init = true; + } + + return 0; +} + +static int smu_v15_0_8_set_default_dpm_table(struct smu_context *smu) +{ + int ret; + + ret = smu_v15_0_8_set_driver_pptable(smu); + if (ret) + return ret; + + return 0; +} + static int smu_v15_0_8_irq_process(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -478,7 +698,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, .is_dpm_running = smu_v15_0_8_is_dpm_running, .init_smc_tables = smu_v15_0_8_init_smc_tables, - .fini_smc_tables = smu_v15_0_fini_smc_tables, + .fini_smc_tables = smu_v15_0_8_fini_smc_tables, .init_power = smu_v15_0_init_power, .fini_power = smu_v15_0_fini_power, .check_fw_status = smu_v15_0_8_check_fw_status, diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h index 40c410928c96..6c85f23d3111 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h @@ -25,4 +25,34 @@ extern void smu_v15_0_8_set_ppt_funcs(struct smu_context *smu); +typedef struct { + uint32_t MaxSocketPowerLimit; + uint32_t MaxGfxclkFrequency; + uint32_t MinGfxclkFrequency; + uint32_t MaxFclkFrequency; + uint32_t MinFclkFrequency; + uint32_t MaxGl2clkFrequency; + uint32_t MinGl2clkFrequency; + uint32_t UclkFrequencyTable[4]; + uint32_t SocclkFrequency; + uint32_t LclkFrequency; + uint32_t VclkFrequency; + uint32_t DclkFrequency; + uint32_t CTFLimitMID; + uint32_t CTFLimitAID; + uint32_t CTFLimitXCD; + uint32_t CTFLimitHBM; + uint32_t ThermalLimitMID; + uint32_t ThermalLimitAID; + uint32_t ThermalLimitXCD; + uint32_t ThermalLimitHBM; + uint64_t PublicSerialNumberMID; + uint64_t PublicSerialNumberAID; + uint64_t PublicSerialNumberXCD; + uint32_t PPT1Max; + uint32_t PPT1Min; + uint32_t PPT1Default; + bool init; +} PPTable_t; + #endif From f7bee962d92ad628f0ba8d14db6069547b84bd40 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Tue, 23 Dec 2025 23:35:32 +0800 Subject: [PATCH 59/99] drm/amd/pm: Update dpm table structs for smu_v15_0 Update dpm table structs to use common definitions for smu_15_0 Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h | 47 +++-------- .../gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c | 79 +++++++++---------- 2 files changed, 49 insertions(+), 77 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h index 06842d38c92b..3c8c086f0f9d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h @@ -45,7 +45,6 @@ #define FEATURE_MASK(feature) (1ULL << feature) -#define MAX_DPM_LEVELS 16 #define MAX_PCIE_CONF 3 #define SMU15_TOOL_SIZE 0x19000 @@ -69,39 +68,19 @@ struct smu_15_0_max_sustainable_clocks { uint32_t soc_clock; }; -struct smu_15_0_dpm_clk_level { - bool enabled; - uint32_t value; -}; - -struct smu_15_0_dpm_table { - uint32_t min; /* MHz */ - uint32_t max; /* MHz */ - uint32_t count; - bool is_fine_grained; - struct smu_15_0_dpm_clk_level dpm_levels[MAX_DPM_LEVELS]; -}; - -struct smu_15_0_pcie_table { - uint8_t pcie_gen[MAX_PCIE_CONF]; - uint8_t pcie_lane[MAX_PCIE_CONF]; - uint16_t clk_freq[MAX_PCIE_CONF]; - uint32_t num_of_link_levels; -}; - struct smu_15_0_dpm_tables { - struct smu_15_0_dpm_table soc_table; - struct smu_15_0_dpm_table gfx_table; - struct smu_15_0_dpm_table uclk_table; - struct smu_15_0_dpm_table eclk_table; - struct smu_15_0_dpm_table vclk_table; - struct smu_15_0_dpm_table dclk_table; - struct smu_15_0_dpm_table dcef_table; - struct smu_15_0_dpm_table pixel_table; - struct smu_15_0_dpm_table display_table; - struct smu_15_0_dpm_table phy_table; - struct smu_15_0_dpm_table fclk_table; - struct smu_15_0_pcie_table pcie_table; + struct smu_dpm_table soc_table; + struct smu_dpm_table gfx_table; + struct smu_dpm_table uclk_table; + struct smu_dpm_table eclk_table; + struct smu_dpm_table vclk_table; + struct smu_dpm_table dclk_table; + struct smu_dpm_table dcef_table; + struct smu_dpm_table pixel_table; + struct smu_dpm_table display_table; + struct smu_dpm_table phy_table; + struct smu_dpm_table fclk_table; + struct smu_pcie_table pcie_table; }; struct smu_15_0_dpm_context { @@ -204,7 +183,7 @@ int smu_v15_0_set_power_source(struct smu_context *smu, int smu_v15_0_set_single_dpm_table(struct smu_context *smu, enum smu_clk_type clk_type, - struct smu_15_0_dpm_table *single_dpm_table); + struct smu_dpm_table *single_dpm_table); int smu_v15_0_gfx_ulv_control(struct smu_context *smu, bool enablement); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c index 21df9705fda9..eb3fa767c848 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c @@ -1058,18 +1058,12 @@ int smu_v15_0_set_performance_level(struct smu_context *smu, { struct smu_15_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; - struct smu_15_0_dpm_table *gfx_table = - &dpm_context->dpm_tables.gfx_table; - struct smu_15_0_dpm_table *mem_table = - &dpm_context->dpm_tables.uclk_table; - struct smu_15_0_dpm_table *soc_table = - &dpm_context->dpm_tables.soc_table; - struct smu_15_0_dpm_table *vclk_table = - &dpm_context->dpm_tables.vclk_table; - struct smu_15_0_dpm_table *dclk_table = - &dpm_context->dpm_tables.dclk_table; - struct smu_15_0_dpm_table *fclk_table = - &dpm_context->dpm_tables.fclk_table; + struct smu_dpm_table *gfx_table = &dpm_context->dpm_tables.gfx_table; + struct smu_dpm_table *mem_table = &dpm_context->dpm_tables.uclk_table; + struct smu_dpm_table *soc_table = &dpm_context->dpm_tables.soc_table; + struct smu_dpm_table *vclk_table = &dpm_context->dpm_tables.vclk_table; + struct smu_dpm_table *dclk_table = &dpm_context->dpm_tables.dclk_table; + struct smu_dpm_table *fclk_table = &dpm_context->dpm_tables.fclk_table; struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; struct amdgpu_device *adev = smu->adev; @@ -1084,34 +1078,34 @@ int smu_v15_0_set_performance_level(struct smu_context *smu, switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: - sclk_min = sclk_max = gfx_table->max; - mclk_min = mclk_max = mem_table->max; - socclk_min = socclk_max = soc_table->max; - vclk_min = vclk_max = vclk_table->max; - dclk_min = dclk_max = dclk_table->max; - fclk_min = fclk_max = fclk_table->max; + sclk_min = sclk_max = SMU_DPM_TABLE_MAX(gfx_table); + mclk_min = mclk_max = SMU_DPM_TABLE_MAX(mem_table); + socclk_min = socclk_max = SMU_DPM_TABLE_MAX(soc_table); + vclk_min = vclk_max = SMU_DPM_TABLE_MAX(vclk_table); + dclk_min = dclk_max = SMU_DPM_TABLE_MAX(dclk_table); + fclk_min = fclk_max = SMU_DPM_TABLE_MAX(fclk_table); break; case AMD_DPM_FORCED_LEVEL_LOW: - sclk_min = sclk_max = gfx_table->min; - mclk_min = mclk_max = mem_table->min; - socclk_min = socclk_max = soc_table->min; - vclk_min = vclk_max = vclk_table->min; - dclk_min = dclk_max = dclk_table->min; - fclk_min = fclk_max = fclk_table->min; + sclk_min = sclk_max = SMU_DPM_TABLE_MIN(gfx_table); + mclk_min = mclk_max = SMU_DPM_TABLE_MIN(mem_table); + socclk_min = socclk_max = SMU_DPM_TABLE_MIN(soc_table); + vclk_min = vclk_max = SMU_DPM_TABLE_MIN(vclk_table); + dclk_min = dclk_max = SMU_DPM_TABLE_MIN(dclk_table); + fclk_min = fclk_max = SMU_DPM_TABLE_MIN(fclk_table); break; case AMD_DPM_FORCED_LEVEL_AUTO: - sclk_min = gfx_table->min; - sclk_max = gfx_table->max; - mclk_min = mem_table->min; - mclk_max = mem_table->max; - socclk_min = soc_table->min; - socclk_max = soc_table->max; - vclk_min = vclk_table->min; - vclk_max = vclk_table->max; - dclk_min = dclk_table->min; - dclk_max = dclk_table->max; - fclk_min = fclk_table->min; - fclk_max = fclk_table->max; + sclk_min = SMU_DPM_TABLE_MIN(gfx_table); + sclk_max = SMU_DPM_TABLE_MAX(gfx_table); + mclk_min = SMU_DPM_TABLE_MIN(mem_table); + mclk_max = SMU_DPM_TABLE_MAX(mem_table); + socclk_min = SMU_DPM_TABLE_MIN(soc_table); + socclk_max = SMU_DPM_TABLE_MAX(soc_table); + vclk_min = SMU_DPM_TABLE_MIN(vclk_table); + vclk_max = SMU_DPM_TABLE_MAX(vclk_table); + dclk_min = SMU_DPM_TABLE_MIN(dclk_table); + dclk_max = SMU_DPM_TABLE_MAX(dclk_table); + fclk_min = SMU_DPM_TABLE_MIN(fclk_table); + fclk_max = SMU_DPM_TABLE_MAX(fclk_table); auto_level = true; break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: @@ -1333,10 +1327,11 @@ static int smu_v15_0_get_fine_grained_status(struct smu_context *smu, int smu_v15_0_set_single_dpm_table(struct smu_context *smu, enum smu_clk_type clk_type, - struct smu_15_0_dpm_table *single_dpm_table) + struct smu_dpm_table *single_dpm_table) { int ret = 0; uint32_t clk; + bool is_fine_grained; int i; ret = smu_v15_0_get_dpm_level_count(smu, @@ -1349,12 +1344,15 @@ int smu_v15_0_set_single_dpm_table(struct smu_context *smu, ret = smu_v15_0_get_fine_grained_status(smu, clk_type, - &single_dpm_table->is_fine_grained); + &is_fine_grained); if (ret) { dev_err(smu->adev->dev, "[%s] failed to get fine grained status!\n", __func__); return ret; } + if (is_fine_grained) + single_dpm_table->flags |= SMU_DPM_TABLE_FINE_GRAINED; + for (i = 0; i < single_dpm_table->count; i++) { ret = smu_v15_0_get_dpm_freq_by_index(smu, clk_type, @@ -1367,11 +1365,6 @@ int smu_v15_0_set_single_dpm_table(struct smu_context *smu, single_dpm_table->dpm_levels[i].value = clk; single_dpm_table->dpm_levels[i].enabled = true; - - if (i == 0) - single_dpm_table->min = clk; - else if (i == single_dpm_table->count - 1) - single_dpm_table->max = clk; } return 0; From 34b19dab0c075e1a715537590af5f4cee5610633 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 24 Dec 2025 00:33:49 +0800 Subject: [PATCH 60/99] drm/amd/pm: Add default dpm table support for smu 15.0.8 Add default dpm table support for smu 15.0.8 v2: Remove lclk, move pptable check up, add missing clk (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h | 1 + drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h | 1 + .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 211 ++++++++++++++++++ 3 files changed, 213 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h index c7bfb4fc0e9a..fc31cc5e97c8 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h @@ -340,6 +340,7 @@ enum smu_clk_type { SMU_OD_FAN_MINIMUM_PWM, SMU_OD_FAN_ZERO_RPM_ENABLE, SMU_OD_FAN_ZERO_RPM_STOP_TEMP, + SMU_GL2CLK, SMU_CLK_COUNT, }; diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h index 3c8c086f0f9d..e6fd8be2cc4a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v15_0.h @@ -81,6 +81,7 @@ struct smu_15_0_dpm_tables { struct smu_dpm_table phy_table; struct smu_dpm_table fclk_table; struct smu_pcie_table pcie_table; + struct smu_dpm_table gl2_table; }; struct smu_15_0_dpm_context { diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 82b09fe7fccd..149421b1c6cb 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -263,6 +263,212 @@ static int smu_v15_0_8_init_allowed_features(struct smu_context *smu) return 0; } +static int smu_v15_0_8_get_dpm_ultimate_freq(struct smu_context *smu, + enum smu_clk_type clk_type, + uint32_t *min, uint32_t *max) +{ + struct smu_15_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; + struct smu_table_context *smu_table = &smu->smu_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + struct smu_dpm_table *dpm_table; + uint32_t min_clk = 0, max_clk = 0; + + if (!pptable->init) + return -EINVAL; + + /* Try cached DPM tables first */ + if (dpm_context) { + switch (clk_type) { + case SMU_MCLK: + case SMU_UCLK: + dpm_table = &dpm_context->dpm_tables.uclk_table; + break; + case SMU_GFXCLK: + case SMU_SCLK: + dpm_table = &dpm_context->dpm_tables.gfx_table; + break; + case SMU_SOCCLK: + dpm_table = &dpm_context->dpm_tables.soc_table; + break; + case SMU_FCLK: + dpm_table = &dpm_context->dpm_tables.fclk_table; + break; + case SMU_GL2CLK: + dpm_table = &dpm_context->dpm_tables.gl2_table; + break; + case SMU_VCLK: + dpm_table = &dpm_context->dpm_tables.vclk_table; + break; + case SMU_DCLK: + dpm_table = &dpm_context->dpm_tables.dclk_table; + break; + default: + dpm_table = NULL; + break; + } + + if (dpm_table && dpm_table->count > 0) { + min_clk = SMU_DPM_TABLE_MIN(dpm_table); + max_clk = SMU_DPM_TABLE_MAX(dpm_table); + + if (min_clk && max_clk) { + if (min) + *min = min_clk; + if (max) + *max = max_clk; + return 0; + } + } + } + + /* Fall back to pptable */ + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + min_clk = pptable->MinGfxclkFrequency; + max_clk = pptable->MaxGfxclkFrequency; + break; + case SMU_FCLK: + min_clk = pptable->MinFclkFrequency; + max_clk = pptable->MaxFclkFrequency; + break; + case SMU_GL2CLK: + min_clk = pptable->MinGl2clkFrequency; + max_clk = pptable->MaxGl2clkFrequency; + break; + case SMU_MCLK: + case SMU_UCLK: + min_clk = pptable->UclkFrequencyTable[0]; + max_clk = pptable->UclkFrequencyTable[ARRAY_SIZE(pptable->UclkFrequencyTable) - 1]; + break; + case SMU_SOCCLK: + min_clk = pptable->SocclkFrequency; + max_clk = pptable->SocclkFrequency; + break; + case SMU_VCLK: + min_clk = pptable->VclkFrequency; + max_clk = pptable->VclkFrequency; + break; + case SMU_DCLK: + min_clk = pptable->DclkFrequency; + max_clk = pptable->DclkFrequency; + break; + default: + return -EINVAL; + } + + if (min) + *min = min_clk; + if (max) + *max = max_clk; + + return 0; +} + +static int smu_v15_0_8_set_dpm_table(struct smu_context *smu) +{ + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_15_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; + struct smu_dpm_table *dpm_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + int i, ret; + uint32_t gfxclkmin, gfxclkmax; + + /* gfxclk dpm table setup - fine-grained */ + dpm_table = &dpm_context->dpm_tables.gfx_table; + dpm_table->clk_type = SMU_GFXCLK; + dpm_table->flags = SMU_DPM_TABLE_FINE_GRAINED; + if (smu_cmn_feature_is_enabled(smu, SMU_FEATURE_DPM_GFXCLK_BIT)) { + ret = smu_v15_0_8_get_dpm_ultimate_freq(smu, SMU_GFXCLK, + &gfxclkmin, &gfxclkmax); + if (ret) + return ret; + + dpm_table->count = 2; + dpm_table->dpm_levels[0].value = gfxclkmin; + dpm_table->dpm_levels[0].enabled = true; + dpm_table->dpm_levels[1].value = gfxclkmax; + dpm_table->dpm_levels[1].enabled = true; + } else { + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->MinGfxclkFrequency; + dpm_table->dpm_levels[0].enabled = true; + } + + /* fclk dpm table setup - fine-grained */ + dpm_table = &dpm_context->dpm_tables.fclk_table; + dpm_table->clk_type = SMU_FCLK; + dpm_table->flags = SMU_DPM_TABLE_FINE_GRAINED; + if (smu_cmn_feature_is_enabled(smu, SMU_FEATURE_DPM_FCLK_BIT)) { + dpm_table->count = 2; + dpm_table->dpm_levels[0].value = pptable->MinFclkFrequency; + dpm_table->dpm_levels[0].enabled = true; + dpm_table->dpm_levels[1].value = pptable->MaxFclkFrequency; + dpm_table->dpm_levels[1].enabled = true; + } else { + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->MinFclkFrequency; + dpm_table->dpm_levels[0].enabled = true; + } + + /* gl2clk dpm table setup - fine-grained */ + dpm_table = &dpm_context->dpm_tables.gl2_table; + dpm_table->flags = SMU_DPM_TABLE_FINE_GRAINED; + if (smu_cmn_feature_is_enabled(smu, SMU_FEATURE_DPM_GL2CLK_BIT)) { + dpm_table->count = 2; + dpm_table->dpm_levels[0].value = pptable->MinGl2clkFrequency; + dpm_table->dpm_levels[0].enabled = true; + dpm_table->dpm_levels[1].value = pptable->MaxGl2clkFrequency; + dpm_table->dpm_levels[1].enabled = true; + } else { + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->MinGl2clkFrequency; + dpm_table->dpm_levels[0].enabled = true; + } + + /* uclk dpm table setup - discrete levels */ + dpm_table = &dpm_context->dpm_tables.uclk_table; + dpm_table->clk_type = SMU_UCLK; + dpm_table->flags = 0; + if (smu_cmn_feature_is_enabled(smu, SMU_FEATURE_DPM_UCLK_BIT)) { + dpm_table->count = ARRAY_SIZE(pptable->UclkFrequencyTable); + for (i = 0; i < dpm_table->count; ++i) { + dpm_table->dpm_levels[i].value = pptable->UclkFrequencyTable[i]; + dpm_table->dpm_levels[i].enabled = true; + } + } else { + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->UclkFrequencyTable[0]; + dpm_table->dpm_levels[0].enabled = true; + } + + /* socclk dpm table setup - single boot-time value */ + dpm_table = &dpm_context->dpm_tables.soc_table; + dpm_table->clk_type = SMU_SOCCLK; + dpm_table->flags = 0; + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->SocclkFrequency; + dpm_table->dpm_levels[0].enabled = true; + + /* vclk dpm table setup - single boot-time value */ + dpm_table = &dpm_context->dpm_tables.vclk_table; + dpm_table->clk_type = SMU_VCLK; + dpm_table->flags = 0; + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->VclkFrequency; + dpm_table->dpm_levels[0].enabled = true; + + /* dclk dpm table setup - single boot-time value */ + dpm_table = &dpm_context->dpm_tables.dclk_table; + dpm_table->clk_type = SMU_DCLK; + dpm_table->flags = 0; + dpm_table->count = 1; + dpm_table->dpm_levels[0].value = pptable->DclkFrequency; + dpm_table->dpm_levels[0].enabled = true; + + return 0; +} + static int smu_v15_0_8_setup_pptable(struct smu_context *smu) { struct smu_table_context *table_context = &smu->smu_table; @@ -455,6 +661,10 @@ static int smu_v15_0_8_set_default_dpm_table(struct smu_context *smu) if (ret) return ret; + ret = smu_v15_0_8_set_dpm_table(smu); + if (ret) + return ret; + return 0; } @@ -714,6 +924,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .get_pp_feature_mask = smu_cmn_get_pp_feature_mask, .wait_for_event = smu_v15_0_wait_for_event, .mode2_reset = smu_v15_0_8_mode2_reset, + .get_dpm_ultimate_freq = smu_v15_0_8_get_dpm_ultimate_freq, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, From 6bcea37ff26bcbfd04d39b00855e72f890412c5b Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 26 Nov 2025 18:41:34 +0800 Subject: [PATCH 61/99] drm/amd/pm: Add get_pm_metrics support for smu 15.0.8 export .get_pm_metrics interface for smu 15.0.8. v2: Make tmo as unsigned (Lijo) Signed-off-by: Yang Wang Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 65 +++++++++++++++++++ 1 file changed, 65 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 149421b1c6cb..60a43ce5648a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -263,6 +263,37 @@ static int smu_v15_0_8_init_allowed_features(struct smu_context *smu) return 0; } +static int smu_v15_0_8_get_metrics_table_internal(struct smu_context *smu, uint32_t tmo, void *data) +{ + struct smu_table_context *smu_table = &smu->smu_table; + uint32_t table_size = smu_table->tables[SMU_TABLE_SMU_METRICS].size; + struct smu_table *table = &smu_table->driver_table; + struct amdgpu_device *adev = smu->adev; + int ret; + + mutex_lock(&smu_table->metrics_lock); + + if (!tmo || !smu_table->metrics_time || + time_after(jiffies, smu_table->metrics_time + msecs_to_jiffies(tmo))) { + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetMetricsTable, NULL); + if (ret) { + dev_info(adev->dev, + "Failed to export SMU metrics table!\n"); + return ret; + } + + amdgpu_device_invalidate_hdp(smu->adev, NULL); + memcpy(smu_table->metrics_table, table->cpu_addr, table_size); + + smu_table->metrics_time = jiffies; + } + + if (data) + memcpy(data, smu_table->metrics_table, table_size); + mutex_unlock(&smu_table->metrics_lock); + return ret; +} + static int smu_v15_0_8_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max) @@ -860,6 +891,39 @@ static bool smu_v15_0_8_is_dpm_running(struct smu_context *smu) smu_v15_0_8_dpm_features.bits); } +static ssize_t smu_v15_0_8_get_pm_metrics(struct smu_context *smu, + void *metrics, size_t max_size) +{ + struct smu_table_context *smu_table = &smu->smu_table; + struct amdgpu_pm_metrics *pm_metrics = (struct amdgpu_pm_metrics *)metrics; + uint32_t table_version = smu_table->tables[SMU_TABLE_SMU_METRICS].version; + uint32_t table_size = smu_table->tables[SMU_TABLE_SMU_METRICS].size; + uint32_t pmfw_version; + int ret; + + if (!pm_metrics || !max_size) + return -EINVAL; + + if (max_size < (table_size + sizeof(pm_metrics->common_header))) + return -EOVERFLOW; + + /* Don't use cached metrics data */ + ret = smu_v15_0_8_get_metrics_table_internal(smu, 0, pm_metrics->data); + if (ret) + return ret; + + smu_cmn_get_smc_version(smu, NULL, &pmfw_version); + memset(&pm_metrics->common_header, 0, sizeof(pm_metrics->common_header)); + pm_metrics->common_header.mp1_ip_discovery_version = + amdgpu_ip_version(smu->adev, MP1_HWIP, 0); + pm_metrics->common_header.pmfw_version = pmfw_version; + pm_metrics->common_header.pmmetrics_version = table_version; + pm_metrics->common_header.structure_size = + sizeof(pm_metrics->common_header) + table_size; + + return pm_metrics->common_header.structure_size; +} + static int smu_v15_0_8_mode2_reset(struct smu_context *smu) { struct smu_msg_ctl *ctl = &smu->msg_ctl; @@ -923,6 +987,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .setup_pptable = smu_v15_0_8_setup_pptable, .get_pp_feature_mask = smu_cmn_get_pp_feature_mask, .wait_for_event = smu_v15_0_wait_for_event, + .get_pm_metrics = smu_v15_0_8_get_pm_metrics, .mode2_reset = smu_v15_0_8_mode2_reset, .get_dpm_ultimate_freq = smu_v15_0_8_get_dpm_ultimate_freq, }; From 0a66ca3b351f107fdc485cb5c3b4b76ab0c1e4e5 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Thu, 23 Oct 2025 11:04:40 +0800 Subject: [PATCH 62/99] drm/amd/pm: add get_gpu_metrics support for 15.0.8 export .get_gpu_metrics interface for 15.0.8 v2: Remove members already exposed by other interfaces, use mask, logical conversion (Lijo) v3: Use correct logic for hbm stacks loop (Lijo) Remove buffer allocation v4: Make out of bound check outside loop (Lijo) v5: fix locking in error case (Alex) Signed-off-by: Yang Wang Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../gpu/drm/amd/include/kgd_pp_interface.h | 1 + .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 180 +++++++++++++++++- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h | 126 ++++++++++++ 3 files changed, 303 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 33a1404bb666..89b440edd15b 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -587,6 +587,7 @@ enum amdgpu_metrics_attr_id { AMDGPU_METRICS_ATTR_ID_GFX_LOW_UTILIZATION_ACC, AMDGPU_METRICS_ATTR_ID_GFX_BELOW_HOST_LIMIT_TOTAL_ACC, AMDGPU_METRICS_ATTR_ID_TEMPERATURE_HBM, + AMDGPU_METRICS_ATTR_ID_TEMPERATURE_MID, AMDGPU_METRICS_ATTR_ID_TEMPERATURE_AID, AMDGPU_METRICS_ATTR_ID_TEMPERATURE_XCD, AMDGPU_METRICS_ATTR_ID_MAX, diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 60a43ce5648a..b749f983b3a0 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -56,6 +56,13 @@ #define SMUQ10_FRAC(x) ((x) & 0x3ff) #define SMUQ10_ROUND(x) ((SMUQ10_TO_UINT(x)) + ((SMUQ10_FRAC(x)) >= 0x200)) +#define hbm_stack_mask_valid(umc_mask) \ + (((umc_mask) & 0xF) == 0xF) + +#define for_each_hbm_stack(stack_idx, umc_mask) \ + for ((stack_idx) = 0; (umc_mask); \ + (umc_mask) >>= 4, (stack_idx)++) \ + #define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) #define SMU_15_0_8_FEA_MAP(smu_feature, smu_15_0_8_feature) \ @@ -167,8 +174,9 @@ static const struct cmn2asic_mapping smu_v15_0_8_table_map[SMU_TABLE_COUNT] = { static int smu_v15_0_8_tables_init(struct smu_context *smu) { struct smu_table_context *smu_table = &smu->smu_table; + int ret, gpu_metrcs_size = sizeof(MetricsTable_t); struct smu_table *tables = smu_table->tables; - int gpu_metrcs_size = sizeof(MetricsTable_t); + struct smu_v15_0_8_gpu_metrics *gpu_metrics; void *driver_pptable __free(kfree) = NULL; void *metrics_table __free(kfree) = NULL; @@ -190,6 +198,15 @@ static int smu_v15_0_8_tables_init(struct smu_context *smu) if (!driver_pptable) return -ENOMEM; + ret = smu_driver_table_init(smu, SMU_DRIVER_TABLE_GPU_METRICS, + sizeof(struct smu_v15_0_8_gpu_metrics), + SMU_GPU_METRICS_CACHE_INTERVAL); + if (ret) + return ret; + + gpu_metrics = (struct smu_v15_0_8_gpu_metrics *)smu_driver_table_ptr(smu, + SMU_DRIVER_TABLE_GPU_METRICS); + smu_v15_0_8_gpu_metrics_init(gpu_metrics, 1, 9); smu_table->metrics_table = no_free_ptr(metrics_table); smu_table->driver_pptable = no_free_ptr(driver_pptable); @@ -269,16 +286,16 @@ static int smu_v15_0_8_get_metrics_table_internal(struct smu_context *smu, uint3 uint32_t table_size = smu_table->tables[SMU_TABLE_SMU_METRICS].size; struct smu_table *table = &smu_table->driver_table; struct amdgpu_device *adev = smu->adev; - int ret; mutex_lock(&smu_table->metrics_lock); if (!tmo || !smu_table->metrics_time || time_after(jiffies, smu_table->metrics_time + msecs_to_jiffies(tmo))) { - ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetMetricsTable, NULL); + int ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetMetricsTable, NULL); if (ret) { dev_info(adev->dev, "Failed to export SMU metrics table!\n"); + mutex_unlock(&smu_table->metrics_lock); return ret; } @@ -291,7 +308,7 @@ static int smu_v15_0_8_get_metrics_table_internal(struct smu_context *smu, uint3 if (data) memcpy(data, smu_table->metrics_table, table_size); mutex_unlock(&smu_table->metrics_lock); - return ret; + return 0; } static int smu_v15_0_8_get_dpm_ultimate_freq(struct smu_context *smu, @@ -967,6 +984,160 @@ static int smu_v15_0_8_mode2_reset(struct smu_context *smu) return ret; } +static ssize_t smu_v15_0_8_get_gpu_metrics(struct smu_context *smu, void **table) +{ + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_v15_0_8_gpu_metrics *gpu_metrics; + struct amdgpu_device *adev = smu->adev; + int ret = 0, xcc_id, inst, i, j, idx; + uint32_t aid_mask = adev->aid_mask; + uint32_t mid_mask = adev->aid_mask; + MetricsTable_t *metrics; + + metrics = kzalloc(sizeof(MetricsTable_t), GFP_KERNEL); + + ret = smu_v15_0_8_get_metrics_table_internal(smu, 1, NULL); + if (ret) + return ret; + + metrics = (MetricsTable_t *)smu_table->metrics_table; + gpu_metrics = (struct smu_v15_0_8_gpu_metrics *)smu_driver_table_ptr(smu, + SMU_DRIVER_TABLE_GPU_METRICS); + + gpu_metrics->system_clock_counter = ktime_get_boottime_ns(); + + gpu_metrics->temperature_hotspot = SMUQ10_ROUND(metrics->MaxSocketTemperature); + + /* Per-HBM stack temperatures */ + if (adev->umc.active_mask) { + u64 mask = adev->umc.active_mask; + int out_idx = 0; + int stack_idx; + + if (unlikely(hweight64(mask)/4 > SMU_15_0_8_MAX_HBM_STACKS)) + dev_warn(adev->dev, "Invalid umc mask %lld\n", mask); + else { + for_each_hbm_stack(stack_idx, mask) { + if (!hbm_stack_mask_valid(mask)) + continue; + gpu_metrics->temperature_hbm[out_idx++] = + SMUQ10_ROUND(metrics->HbmTemperature[stack_idx]); + } + } + } + + /* Reports max temperature of all voltage rails */ + gpu_metrics->temperature_vrsoc = SMUQ10_ROUND(metrics->MaxVrTemperature); + /* MID, AID, XCD temperatures */ + idx = 0; + for_each_inst(i, mid_mask) { + gpu_metrics->temperature_mid[idx] = SMUQ10_ROUND(metrics->MidTemperature[i]); + idx++; + } + + idx = 0; + for_each_inst(i, aid_mask) { + gpu_metrics->temperature_aid[idx] = SMUQ10_ROUND(metrics->AidTemperature[i]); + idx++; + } + + for (i = 0; i < NUM_XCC(adev->gfx.xcc_mask); ++i) { + xcc_id = GET_INST(GC, i); + if (xcc_id >= 0) + gpu_metrics->temperature_xcd[i] = SMUQ10_ROUND(metrics->XcdTemperature[xcc_id]); + } + /* Power */ + gpu_metrics->curr_socket_power = SMUQ10_ROUND(metrics->SocketPower); + + gpu_metrics->average_gfx_activity = SMUQ10_ROUND(metrics->SocketGfxBusy); + gpu_metrics->average_umc_activity = SMUQ10_ROUND(metrics->DramBandwidthUtilization); + gpu_metrics->mem_max_bandwidth = SMUQ10_ROUND(metrics->MaxDramBandwidth); + + /* Energy counter reported in 15.259uJ (2^-16) units */ + gpu_metrics->energy_accumulator = metrics->SocketEnergyAcc; + + for (i = 0; i < NUM_XCC(adev->gfx.xcc_mask); ++i) { + xcc_id = GET_INST(GC, i); + if (xcc_id >= 0) { + gpu_metrics->current_gfxclk[i] = + SMUQ10_ROUND(metrics->GfxclkFrequency[xcc_id]); + } + } + + /* Per-MID clocks */ + idx = 0; + for_each_inst(i, mid_mask) { + gpu_metrics->current_socclk[idx] = SMUQ10_ROUND(metrics->SocclkFrequency[i]); + idx++; + } + + /* Per-VCN clocks */ + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + inst = GET_INST(VCN, i); + if (inst >= 0) { + gpu_metrics->current_vclk0[i] = SMUQ10_ROUND(metrics->VclkFrequency[inst]); + gpu_metrics->current_dclk0[i] = SMUQ10_ROUND(metrics->DclkFrequency[inst]); + } + } + + /* Per-AID clocks */ + idx = 0; + for_each_inst(i, aid_mask) { + gpu_metrics->current_uclk[idx] = SMUQ10_ROUND(metrics->UclkFrequency[i]); + idx++; + } + + /* Total accumulated cycle counter */ + gpu_metrics->accumulation_counter = metrics->AccumulationCounter; + + /* Accumulated throttler residencies */ + gpu_metrics->prochot_residency_acc = metrics->ProchotResidencyAcc; + gpu_metrics->ppt_residency_acc = metrics->PptResidencyAcc; + gpu_metrics->socket_thm_residency_acc = metrics->SocketThmResidencyAcc; + gpu_metrics->vr_thm_residency_acc = metrics->VrThmResidencyAcc; + gpu_metrics->hbm_thm_residency_acc = metrics->HbmThmResidencyAcc; + + gpu_metrics->gfx_activity_acc = SMUQ10_ROUND(metrics->SocketGfxBusyAcc); + gpu_metrics->mem_activity_acc = SMUQ10_ROUND(metrics->DramBandwidthUtilizationAcc); + + for (i = 0; i < NUM_XGMI_LINKS; i++) { + j = amdgpu_xgmi_get_ext_link(adev, i); + if (j < 0 || j >= NUM_XGMI_LINKS) + continue; + ret = amdgpu_get_xgmi_link_status(adev, i); + if (ret >= 0) + gpu_metrics->xgmi_link_status[j] = ret; + } + + gpu_metrics->xgmi_read_data_acc = SMUQ10_ROUND(metrics->XgmiReadBandwidthAcc); + gpu_metrics->xgmi_write_data_acc = SMUQ10_ROUND(metrics->XgmiWriteBandwidthAcc); + + for (i = 0; i < NUM_XCC(adev->gfx.xcc_mask); ++i) { + inst = GET_INST(GC, i); + gpu_metrics->gfx_busy_inst[i] = SMUQ10_ROUND(metrics->GfxBusy[inst]); + gpu_metrics->gfx_busy_acc[i] = SMUQ10_ROUND(metrics->GfxBusyAcc[inst]); + gpu_metrics->gfx_below_host_limit_ppt_acc[i] = + SMUQ10_ROUND(metrics->GfxclkBelowHostLimitPptAcc[inst]); + gpu_metrics->gfx_below_host_limit_thm_acc[i] = + SMUQ10_ROUND(metrics->GfxclkBelowHostLimitThmAcc[inst]); + gpu_metrics->gfx_low_utilization_acc[i] = + SMUQ10_ROUND(metrics->GfxclkLowUtilizationAcc[inst]); + gpu_metrics->gfx_below_host_limit_total_acc[i] = + SMUQ10_ROUND(metrics->GfxclkBelowHostLimitTotalAcc[inst]); + } + + gpu_metrics->xgmi_link_width = metrics->XgmiWidth; + gpu_metrics->xgmi_link_speed = metrics->XgmiBitrate; + + gpu_metrics->firmware_timestamp = metrics->Timestamp; + + *table = gpu_metrics; + + smu_driver_table_update_cache_time(smu, SMU_DRIVER_TABLE_GPU_METRICS); + + return sizeof(*gpu_metrics); +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -990,6 +1161,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .get_pm_metrics = smu_v15_0_8_get_pm_metrics, .mode2_reset = smu_v15_0_8_mode2_reset, .get_dpm_ultimate_freq = smu_v15_0_8_get_dpm_ultimate_freq, + .get_gpu_metrics = smu_v15_0_8_get_gpu_metrics, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h index 6c85f23d3111..8fc16796788b 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h @@ -23,6 +23,15 @@ #ifndef __SMU_15_0_8_PPT_H__ #define __SMU_15_0_8_PPT_H__ +#define SMU_15_0_8_NUM_XGMI_LINKS 8 +#define SMU_15_0_8_MAX_GFX_CLKS 8 +#define SMU_15_0_8_MAX_CLKS 4 +#define SMU_15_0_8_MAX_XCC 8 +#define SMU_15_0_8_MAX_VCN 4 +#define SMU_15_0_8_MAX_JPEG 40 +#define SMU_15_0_8_MAX_AID 2 +#define SMU_15_0_8_MAX_MID 2 +#define SMU_15_0_8_MAX_HBM_STACKS 12 extern void smu_v15_0_8_set_ppt_funcs(struct smu_context *smu); typedef struct { @@ -55,4 +64,121 @@ typedef struct { bool init; } PPTable_t; +#if defined(SWSMU_CODE_LAYER_L2) +#include "smu_cmn.h" + +/* SMUv 15.0.8 GPU metrics*/ +#define SMU_15_0_8_METRICS_FIELDS(SMU_SCALAR, SMU_ARRAY) \ + SMU_SCALAR(SMU_MATTR(TEMPERATURE_HOTSPOT), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_hotspot); \ + SMU_SCALAR(SMU_MATTR(TEMPERATURE_MEM), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_mem); \ + SMU_SCALAR(SMU_MATTR(TEMPERATURE_VRSOC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_vrsoc); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_HBM), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_hbm, \ + SMU_15_0_8_MAX_HBM_STACKS); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_MID), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_mid, SMU_15_0_8_MAX_MID); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_AID), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_aid, SMU_15_0_8_MAX_AID); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_XCD), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_xcd, SMU_15_0_8_MAX_XCC); \ + SMU_SCALAR(SMU_MATTR(CURR_SOCKET_POWER), SMU_MUNIT(POWER_1), \ + SMU_MTYPE(U16), curr_socket_power); \ + SMU_SCALAR(SMU_MATTR(AVERAGE_GFX_ACTIVITY), SMU_MUNIT(PERCENT), \ + SMU_MTYPE(U16), average_gfx_activity); \ + SMU_SCALAR(SMU_MATTR(AVERAGE_UMC_ACTIVITY), SMU_MUNIT(PERCENT), \ + SMU_MTYPE(U16), average_umc_activity); \ + SMU_SCALAR(SMU_MATTR(MEM_MAX_BANDWIDTH), SMU_MUNIT(BW_1), \ + SMU_MTYPE(U64), mem_max_bandwidth); \ + SMU_SCALAR(SMU_MATTR(ENERGY_ACCUMULATOR), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), energy_accumulator); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_CLOCK_COUNTER), SMU_MUNIT(TIME_1), \ + SMU_MTYPE(U64), system_clock_counter); \ + SMU_SCALAR(SMU_MATTR(ACCUMULATION_COUNTER), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), accumulation_counter); \ + SMU_SCALAR(SMU_MATTR(PROCHOT_RESIDENCY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), prochot_residency_acc); \ + SMU_SCALAR(SMU_MATTR(PPT_RESIDENCY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), ppt_residency_acc); \ + SMU_SCALAR(SMU_MATTR(SOCKET_THM_RESIDENCY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), socket_thm_residency_acc); \ + SMU_SCALAR(SMU_MATTR(VR_THM_RESIDENCY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), vr_thm_residency_acc); \ + SMU_SCALAR(SMU_MATTR(HBM_THM_RESIDENCY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), hbm_thm_residency_acc); \ + SMU_SCALAR(SMU_MATTR(GFXCLK_LOCK_STATUS), SMU_MUNIT(NONE), \ + SMU_MTYPE(U32), gfxclk_lock_status); \ + SMU_SCALAR(SMU_MATTR(PCIE_LINK_WIDTH), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), pcie_link_width); \ + SMU_SCALAR(SMU_MATTR(PCIE_LINK_SPEED), SMU_MUNIT(SPEED_2), \ + SMU_MTYPE(U16), pcie_link_speed); \ + SMU_SCALAR(SMU_MATTR(XGMI_LINK_WIDTH), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), xgmi_link_width); \ + SMU_SCALAR(SMU_MATTR(XGMI_LINK_SPEED), SMU_MUNIT(SPEED_1), \ + SMU_MTYPE(U16), xgmi_link_speed); \ + SMU_SCALAR(SMU_MATTR(GFX_ACTIVITY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), gfx_activity_acc); \ + SMU_SCALAR(SMU_MATTR(MEM_ACTIVITY_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), mem_activity_acc); \ + SMU_ARRAY(SMU_MATTR(PCIE_BANDWIDTH_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), pcie_bandwidth_acc, SMU_15_0_8_MAX_MID); \ + SMU_ARRAY(SMU_MATTR(PCIE_BANDWIDTH_INST), SMU_MUNIT(BW_1), \ + SMU_MTYPE(U32), pcie_bandwidth_inst, SMU_15_0_8_MAX_MID); \ + SMU_SCALAR(SMU_MATTR(PCIE_L0_TO_RECOV_COUNT_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), pcie_l0_to_recov_count_acc); \ + SMU_SCALAR(SMU_MATTR(PCIE_REPLAY_COUNT_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), pcie_replay_count_acc); \ + SMU_SCALAR(SMU_MATTR(PCIE_REPLAY_ROVER_COUNT_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), pcie_replay_rover_count_acc); \ + SMU_SCALAR(SMU_MATTR(PCIE_NAK_SENT_COUNT_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), pcie_nak_sent_count_acc); \ + SMU_SCALAR(SMU_MATTR(PCIE_NAK_RCVD_COUNT_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), pcie_nak_rcvd_count_acc); \ + SMU_ARRAY(SMU_MATTR(XGMI_LINK_STATUS), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), xgmi_link_status, \ + SMU_15_0_8_NUM_XGMI_LINKS); \ + SMU_SCALAR(SMU_MATTR(XGMI_READ_DATA_ACC), SMU_MUNIT(DATA_1), \ + SMU_MTYPE(U64), xgmi_read_data_acc); \ + SMU_SCALAR(SMU_MATTR(XGMI_WRITE_DATA_ACC), SMU_MUNIT(DATA_1), \ + SMU_MTYPE(U64), xgmi_write_data_acc); \ + SMU_SCALAR(SMU_MATTR(FIRMWARE_TIMESTAMP), SMU_MUNIT(TIME_2), \ + SMU_MTYPE(U64), firmware_timestamp); \ + SMU_ARRAY(SMU_MATTR(CURRENT_GFXCLK), SMU_MUNIT(CLOCK_1), \ + SMU_MTYPE(U16), current_gfxclk, SMU_15_0_8_MAX_GFX_CLKS); \ + SMU_ARRAY(SMU_MATTR(CURRENT_SOCCLK), SMU_MUNIT(CLOCK_1), \ + SMU_MTYPE(U16), current_socclk, SMU_15_0_8_MAX_MID); \ + SMU_ARRAY(SMU_MATTR(CURRENT_VCLK0), SMU_MUNIT(CLOCK_1), \ + SMU_MTYPE(U16), current_vclk0, SMU_15_0_8_MAX_VCN); \ + SMU_ARRAY(SMU_MATTR(CURRENT_DCLK0), SMU_MUNIT(CLOCK_1), \ + SMU_MTYPE(U16), current_dclk0, SMU_15_0_8_MAX_VCN); \ + SMU_ARRAY(SMU_MATTR(CURRENT_UCLK), SMU_MUNIT(CLOCK_1), \ + SMU_MTYPE(U16), current_uclk, SMU_15_0_8_MAX_AID); \ + SMU_SCALAR(SMU_MATTR(PCIE_LC_PERF_OTHER_END_RECOVERY), \ + SMU_MUNIT(NONE), SMU_MTYPE(U64), \ + pcie_lc_perf_other_end_recovery); \ + SMU_ARRAY(SMU_MATTR(GFX_BUSY_INST), SMU_MUNIT(PERCENT), \ + SMU_MTYPE(U32), gfx_busy_inst, SMU_15_0_8_MAX_XCC); \ + SMU_ARRAY(SMU_MATTR(JPEG_BUSY), SMU_MUNIT(PERCENT), SMU_MTYPE(U16), \ + jpeg_busy, SMU_15_0_8_MAX_JPEG); \ + SMU_ARRAY(SMU_MATTR(VCN_BUSY), SMU_MUNIT(PERCENT), SMU_MTYPE(U16), \ + vcn_busy, SMU_15_0_8_MAX_VCN); \ + SMU_ARRAY(SMU_MATTR(GFX_BUSY_ACC), SMU_MUNIT(NONE), SMU_MTYPE(U64), \ + gfx_busy_acc, SMU_15_0_8_MAX_XCC); \ + SMU_ARRAY(SMU_MATTR(GFX_BELOW_HOST_LIMIT_PPT_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), gfx_below_host_limit_ppt_acc, \ + SMU_15_0_8_MAX_XCC); \ + SMU_ARRAY(SMU_MATTR(GFX_BELOW_HOST_LIMIT_THM_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), gfx_below_host_limit_thm_acc, \ + SMU_15_0_8_MAX_XCC); \ + SMU_ARRAY(SMU_MATTR(GFX_LOW_UTILIZATION_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), gfx_low_utilization_acc, \ + SMU_15_0_8_MAX_XCC); \ + SMU_ARRAY(SMU_MATTR(GFX_BELOW_HOST_LIMIT_TOTAL_ACC), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), gfx_below_host_limit_total_acc, \ + SMU_15_0_8_MAX_XCC); + +DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_gpu_metrics, SMU_15_0_8_METRICS_FIELDS); +#endif #endif From 39e0a73bdeb0890718260d0bde256ef50288bb80 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Wed, 22 Oct 2025 21:08:13 +0800 Subject: [PATCH 63/99] drm/amd/pm: add get_unique_id support for smu 15.0.8 export .get_unique_id interface for smu 15.0.8 Signed-off-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index b749f983b3a0..f3aa25c59f9c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -1138,6 +1138,15 @@ static ssize_t smu_v15_0_8_get_gpu_metrics(struct smu_context *smu, void **table return sizeof(*gpu_metrics); } +static void smu_v15_0_8_get_unique_id(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + struct smu_table_context *smu_table = &smu->smu_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + + adev->unique_id = pptable->PublicSerialNumberMID; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -1162,6 +1171,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .mode2_reset = smu_v15_0_8_mode2_reset, .get_dpm_ultimate_freq = smu_v15_0_8_get_dpm_ultimate_freq, .get_gpu_metrics = smu_v15_0_8_get_gpu_metrics, + .get_unique_id = smu_v15_0_8_get_unique_id, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, From e20e47bcb3f1e90e2fd8bd8b543224d1bdd3ba3f Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Wed, 22 Oct 2025 21:32:55 +0800 Subject: [PATCH 64/99] drm/amd/pm: add set{get}_power_limit support for smu 15.0.8 export .set_power_limit & .get_power_limit interface for smu 15.0.8 Signed-off-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 1 + .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 34 +++++++++++++++++++ 2 files changed, 35 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 013d34e2f842..eee2fe0743b2 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -2966,6 +2966,7 @@ int smu_get_power_limit(void *handle, case IP_VERSION(11, 0, 11): case IP_VERSION(11, 0, 12): case IP_VERSION(11, 0, 13): + case IP_VERSION(15, 0, 8): ret = smu_get_asic_power_limits(smu, &smu->current_power_limit, NULL, NULL, NULL); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index f3aa25c59f9c..0c06b6500121 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -1147,6 +1147,38 @@ static void smu_v15_0_8_get_unique_id(struct smu_context *smu) adev->unique_id = pptable->PublicSerialNumberMID; } +static int smu_v15_0_8_get_power_limit(struct smu_context *smu, + uint32_t *current_power_limit, + uint32_t *default_power_limit, + uint32_t *max_power_limit, + uint32_t *min_power_limit) +{ + struct smu_table_context *smu_table = &smu->smu_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + uint32_t power_limit = 0; + int ret; + + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetPptLimit, &power_limit); + if (ret) { + dev_err(smu->adev->dev, "Couldn't get PPT limit"); + return -EINVAL; + } + + if (current_power_limit) + *current_power_limit = power_limit; + + if (default_power_limit) + *max_power_limit = pptable->MaxSocketPowerLimit; + + if (max_power_limit) + *max_power_limit = pptable->MaxSocketPowerLimit; + + if (min_power_limit) + *min_power_limit = 0; + + return 0; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -1172,6 +1204,8 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .get_dpm_ultimate_freq = smu_v15_0_8_get_dpm_ultimate_freq, .get_gpu_metrics = smu_v15_0_8_get_gpu_metrics, .get_unique_id = smu_v15_0_8_get_unique_id, + .get_power_limit = smu_v15_0_8_get_power_limit, + .set_power_limit = smu_v15_0_set_power_limit, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, From bf39c461ada3f81e2cae487a402c03ce282a545c Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 24 Dec 2025 03:54:11 +0800 Subject: [PATCH 65/99] drm/amd/pm: Add emit clock support Add emit clock support and fetching other metrics data like temperature, clock for smu_v15_0_8 v2: Use umc count for hbm stack temperature (Lijo) v3: Use correct logic for hbm stacks (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 198 ++++++++++++++++++ 1 file changed, 198 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 0c06b6500121..2a74acb2790f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -311,6 +311,203 @@ static int smu_v15_0_8_get_metrics_table_internal(struct smu_context *smu, uint3 return 0; } +static int smu_v15_0_8_get_smu_metrics_data(struct smu_context *smu, + MetricsMember_t member, uint32_t *value) +{ + struct smu_table_context *smu_table = &smu->smu_table; + MetricsTable_t *metrics = (MetricsTable_t *)smu_table->metrics_table; + struct amdgpu_device *adev = smu->adev; + int ret, xcc_id; + + ret = smu_v15_0_8_get_metrics_table_internal(smu, 10, NULL); + if (ret) + return ret; + + switch (member) { + case METRICS_CURR_GFXCLK: + case METRICS_AVERAGE_GFXCLK: + xcc_id = GET_INST(GC, 0); + *value = SMUQ10_ROUND(metrics->GfxclkFrequency[xcc_id]); + break; + case METRICS_CURR_SOCCLK: + case METRICS_AVERAGE_SOCCLK: + *value = SMUQ10_ROUND(metrics->SocclkFrequency[0]); + break; + case METRICS_CURR_UCLK: + case METRICS_AVERAGE_UCLK: + *value = SMUQ10_ROUND(metrics->UclkFrequency[0]); + break; + case METRICS_CURR_VCLK: + *value = SMUQ10_ROUND(metrics->VclkFrequency[0]); + break; + case METRICS_CURR_DCLK: + *value = SMUQ10_ROUND(metrics->DclkFrequency[0]); + break; + case METRICS_CURR_FCLK: + *value = SMUQ10_ROUND(metrics->FclkFrequency[0]); + break; + case METRICS_AVERAGE_GFXACTIVITY: + *value = SMUQ10_ROUND(metrics->SocketGfxBusy); + break; + case METRICS_AVERAGE_MEMACTIVITY: + *value = SMUQ10_ROUND(metrics->DramBandwidthUtilization); + break; + case METRICS_CURR_SOCKETPOWER: + *value = SMUQ10_ROUND(metrics->SocketPower) << 8; + break; + case METRICS_TEMPERATURE_HOTSPOT: + *value = SMUQ10_ROUND(metrics->MaxSocketTemperature) * + SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + break; + case METRICS_TEMPERATURE_MEM: + { + struct amdgpu_device *adev = smu->adev; + u32 max_hbm_temp = 0; + + /* Find max temperature across all HBM stacks */ + if (adev->umc.active_mask) { + u64 mask = adev->umc.active_mask; + int stack_idx; + + for_each_hbm_stack(stack_idx, mask) { + u32 temp; + + if (!hbm_stack_mask_valid(mask)) + continue; + + temp = SMUQ10_ROUND(metrics->HbmTemperature[stack_idx]); + if (temp > max_hbm_temp) + max_hbm_temp = temp; + } + } + *value = max_hbm_temp * SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + break; + } + /* This is the max of all VRs and not just SOC VR. + */ + case METRICS_TEMPERATURE_VRSOC: + *value = SMUQ10_ROUND(metrics->MaxVrTemperature) * + SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + break; + default: + *value = UINT_MAX; + break; + } + + return 0; +} + +static int smu_v15_0_8_get_current_clk_freq_by_table(struct smu_context *smu, + enum smu_clk_type clk_type, + uint32_t *value) +{ + MetricsMember_t member_type; + + if (!value) + return -EINVAL; + + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + member_type = METRICS_CURR_GFXCLK; + break; + case SMU_UCLK: + case SMU_MCLK: + member_type = METRICS_CURR_UCLK; + break; + case SMU_SOCCLK: + member_type = METRICS_CURR_SOCCLK; + break; + case SMU_VCLK: + member_type = METRICS_CURR_VCLK; + break; + case SMU_DCLK: + member_type = METRICS_CURR_DCLK; + break; + case SMU_FCLK: + member_type = METRICS_CURR_FCLK; + break; + default: + return -EINVAL; + } + + return smu_v15_0_8_get_smu_metrics_data(smu, member_type, value); +} + +static int smu_v15_0_8_emit_clk_levels(struct smu_context *smu, + enum smu_clk_type type, char *buf, + int *offset) +{ + struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; + struct smu_15_0_dpm_context *dpm_context; + struct smu_dpm_table *single_dpm_table = NULL; + struct smu_dpm_context *smu_dpm = &smu->smu_dpm; + int ret, now, size = *offset; + + if (amdgpu_ras_intr_triggered()) { + sysfs_emit_at(buf, size, "unavailable\n"); + return -EBUSY; + } + + dpm_context = smu_dpm->dpm_context; + + switch (type) { + case SMU_OD_SCLK: + size += sysfs_emit_at(buf, size, "%s:\n", "OD_SCLK"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMhz\n", + pstate_table->gfxclk_pstate.curr.min, + pstate_table->gfxclk_pstate.curr.max); + break; + case SMU_OD_MCLK: + size += sysfs_emit_at(buf, size, "%s:\n", "OD_MCLK"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMhz\n", + pstate_table->uclk_pstate.curr.min, + pstate_table->uclk_pstate.curr.max); + break; + case SMU_SCLK: + case SMU_GFXCLK: + single_dpm_table = &dpm_context->dpm_tables.gfx_table; + break; + case SMU_MCLK: + case SMU_UCLK: + single_dpm_table = &dpm_context->dpm_tables.uclk_table; + break; + case SMU_SOCCLK: + single_dpm_table = &dpm_context->dpm_tables.soc_table; + break; + case SMU_FCLK: + single_dpm_table = &dpm_context->dpm_tables.fclk_table; + break; + case SMU_VCLK: + single_dpm_table = &dpm_context->dpm_tables.vclk_table; + break; + case SMU_DCLK: + single_dpm_table = &dpm_context->dpm_tables.dclk_table; + break; + default: + break; + } + + if (single_dpm_table) { + ret = smu_v15_0_8_get_current_clk_freq_by_table(smu, type, &now); + if (ret) { + dev_err(smu->adev->dev, + "Attempt to get current clk Failed!"); + return ret; + } + ret = smu_cmn_print_dpm_clk_levels(smu, single_dpm_table, now, + buf, offset); + if (ret < 0) + return ret; + + return 0; + } + + *offset = size; + + return 0; +} + static int smu_v15_0_8_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max) @@ -1206,6 +1403,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .get_unique_id = smu_v15_0_8_get_unique_id, .get_power_limit = smu_v15_0_8_get_power_limit, .set_power_limit = smu_v15_0_set_power_limit, + .emit_clk_levels = smu_v15_0_8_emit_clk_levels, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, From b4a8a2aba4e3a02ebbf8d2708fcafd491ebd4644 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Sun, 15 Mar 2026 20:11:54 +0530 Subject: [PATCH 66/99] drm/amd/display: Add clk_mgr NULL checks in dcn32_initialize_min_clocks() dcn32_init_hw() checks dc->clk_mgr before calling init_clocks(), so the clock manager is not treated as unconditionally present on this path. However, dcn32_initialize_min_clocks() later dereferences dc->clk_mgr, bw_params, and clk_mgr callbacks without validating them. Add the required guards in dcn32_initialize_min_clocks() before accessing clk_mgr-dependent state, and check callback presence before calling get_dispclk_from_dentist() and update_clocks(). Also guard the later update_bw_bounding_box() call in the FAMS2-disabled path since it also dereferences dc->clk_mgr->bw_params. This keeps clk_mgr handling consistent in the DCN32 HW init flow and avoids possible NULL pointer dereferences reported by Smatch. Fixes the below: drivers/gpu/drm/amd/amdgpu/../display/dc/hwss/dcn32/dcn32_hwseq.c:1012 dcn32_init_hw() error: we previously assumed 'dc->clk_mgr' could be null (see line 978) Cc: Roman Li Cc: Alex Hung Cc: Jerry Zuo Cc: Sun peng Li Cc: Tom Chung Cc: Dan Carpenter Cc: Aurabindo Pillai Signed-off-by: Srinivasan Shanmugam Reviewed-by: Alex Hung Signed-off-by: Alex Deucher --- .../drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c index a0aaa727e9fa..e5d93dd348dd 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c @@ -757,6 +757,9 @@ static void dcn32_initialize_min_clocks(struct dc *dc) { struct dc_clocks *clocks = &dc->current_state->bw_ctx.bw.dcn.clk; + if (!dc->clk_mgr || !dc->clk_mgr->bw_params || !dc->clk_mgr->funcs) + return; + clocks->dcfclk_deep_sleep_khz = DCN3_2_DCFCLK_DS_INIT_KHZ; clocks->dcfclk_khz = dc->clk_mgr->bw_params->clk_table.entries[0].dcfclk_mhz * 1000; clocks->socclk_khz = dc->clk_mgr->bw_params->clk_table.entries[0].socclk_mhz * 1000; @@ -765,9 +768,10 @@ static void dcn32_initialize_min_clocks(struct dc *dc) clocks->ref_dtbclk_khz = dc->clk_mgr->bw_params->clk_table.entries[0].dtbclk_mhz * 1000; clocks->fclk_p_state_change_support = true; clocks->p_state_change_support = true; + if (dc->debug.disable_boot_optimizations) { clocks->dispclk_khz = dc->clk_mgr->bw_params->clk_table.entries[0].dispclk_mhz * 1000; - } else { + } else if (dc->clk_mgr->funcs->get_dispclk_from_dentist) { /* Even though DPG_EN = 1 for the connected display, it still requires the * correct timing so we cannot set DISPCLK to min freq or it could cause * audio corruption. Read current DISPCLK from DENTIST and request the same @@ -776,10 +780,10 @@ static void dcn32_initialize_min_clocks(struct dc *dc) clocks->dispclk_khz = dc->clk_mgr->funcs->get_dispclk_from_dentist(dc->clk_mgr); } - dc->clk_mgr->funcs->update_clocks( - dc->clk_mgr, - dc->current_state, - true); + if (dc->clk_mgr->funcs->update_clocks) + dc->clk_mgr->funcs->update_clocks(dc->clk_mgr, + dc->current_state, + true); } void dcn32_init_hw(struct dc *dc) @@ -1007,7 +1011,8 @@ void dcn32_init_hw(struct dc *dc) DMUB_FW_VERSION(7, 0, 35)) { /* FAMS2 is disabled */ dc->debug.fams2_config.bits.enable = false; - if (dc->debug.using_dml2 && dc->res_pool->funcs->update_bw_bounding_box) { + if (dc->debug.using_dml2 && dc->res_pool->funcs->update_bw_bounding_box && + dc->clk_mgr && dc->clk_mgr->bw_params) { /* update bounding box if FAMS2 disabled */ dc->res_pool->funcs->update_bw_bounding_box(dc, dc->clk_mgr->bw_params); } From 3f5c75634fb5dfa4896e74f39e8b04e879fb4de9 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Tue, 17 Mar 2026 08:38:38 +0530 Subject: [PATCH 67/99] drm/amd/display: Fix NULL pointer assumptions in dcn42_init_hw() dcn42_init_hw() calls update_bw_bounding_box() when FAMS2 is disabled or when the dchub reference clock changes. However the existing condition mixes the callback pointer check with only one side of the || expression: ((!fams2_enable && update_bw_bounding_box) || freq_changed) This allows the block to be entered through the freq_changed path even when update_bw_bounding_box() is NULL. The function is then called unconditionally inside the block, which can lead to a NULL pointer dereference. Additionally, the code dereferences dc->clk_mgr->bw_params without verifying that dc->clk_mgr and bw_params are valid. Restructure the condition so that the update trigger remains the same (FAMS2 disabled or dchub ref clock changed), but guard the call with explicit checks for: - update_bw_bounding_box callback - dc->clk_mgr - dc->clk_mgr->bw_params Also introduce a helper boolean (dchub_ref_freq_changed) to improve readability of the clock-change condition. This fixes Smatch warnings about inconsistent NULL assumptions in dcn42_init_hw(). drivers/gpu/drm/amd/amdgpu/../display/dc/hwss/dcn42/dcn42_hwseq.c:264 dcn42_init_hw() error: we previously assumed 'dc->clk_mgr' could be null (see line 253) drivers/gpu/drm/amd/amdgpu/../display/dc/hwss/dcn42/dcn42_hwseq.c:278 dcn42_init_hw() error: we previously assumed 'dc->res_pool->funcs->update_bw_bounding_box' could be null (see line 274) Cc: Roman Li Cc: Alex Hung Cc: Jerry Zuo Cc: Sun peng Li Cc: Tom Chung Cc: Dan Carpenter Cc: Aurabindo Pillai Signed-off-by: Srinivasan Shanmugam Reviewed-by: Alex Hung Signed-off-by: Alex Deucher --- .../amd/display/dc/hwss/dcn42/dcn42_hwseq.c | 24 +++++++++++++------ 1 file changed, 17 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c index 7bf94eae30a6..7f9c121c00e6 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c @@ -69,6 +69,7 @@ void dcn42_init_hw(struct dc *dc) int edp_num; uint32_t backlight = MAX_BACKLIGHT_LEVEL; uint32_t user_level = MAX_BACKLIGHT_LEVEL; + bool dchub_ref_freq_changed; int current_dchub_ref_freq = 0; if (dc->clk_mgr && dc->clk_mgr->funcs && dc->clk_mgr->funcs->init_clocks) { @@ -261,8 +262,12 @@ void dcn42_init_hw(struct dc *dc) if (dc->res_pool->hubbub->funcs->init_crb) dc->res_pool->hubbub->funcs->init_crb(dc->res_pool->hubbub); - if (dc->res_pool->hubbub->funcs->set_request_limit && dc->config.sdpif_request_limit_words_per_umc > 0) - dc->res_pool->hubbub->funcs->set_request_limit(dc->res_pool->hubbub, dc->clk_mgr->bw_params->num_channels, dc->config.sdpif_request_limit_words_per_umc); + if (dc->res_pool->hubbub->funcs->set_request_limit && + dc->clk_mgr && dc->clk_mgr->bw_params && + dc->config.sdpif_request_limit_words_per_umc > 0) + dc->res_pool->hubbub->funcs->set_request_limit(dc->res_pool->hubbub, + dc->clk_mgr->bw_params->num_channels, + dc->config.sdpif_request_limit_words_per_umc); // Get DMCUB capabilities if (dc->ctx->dmub_srv) { @@ -270,13 +275,18 @@ void dcn42_init_hw(struct dc *dc) dc->caps.dmub_caps.psr = dc->ctx->dmub_srv->dmub->feature_caps.psr; dc->caps.dmub_caps.mclk_sw = dc->ctx->dmub_srv->dmub->feature_caps.fw_assisted_mclk_switch_ver > 0; dc->caps.dmub_caps.fams_ver = dc->ctx->dmub_srv->dmub->feature_caps.fw_assisted_mclk_switch_ver; + + /* sw and fw FAMS versions must match for support */ dc->debug.fams2_config.bits.enable &= - dc->caps.dmub_caps.fams_ver == dc->debug.fams_version.ver; // sw & fw fams versions must match for support - if ((!dc->debug.fams2_config.bits.enable && dc->res_pool->funcs->update_bw_bounding_box) - || res_pool->ref_clocks.dchub_ref_clock_inKhz / 1000 != current_dchub_ref_freq) { + dc->caps.dmub_caps.fams_ver == dc->debug.fams_version.ver; + dchub_ref_freq_changed = + res_pool->ref_clocks.dchub_ref_clock_inKhz / 1000 != current_dchub_ref_freq; + + if ((!dc->debug.fams2_config.bits.enable || dchub_ref_freq_changed) && + dc->res_pool->funcs->update_bw_bounding_box && + dc->clk_mgr && dc->clk_mgr->bw_params) { /* update bounding box if FAMS2 disabled, or if dchub clk has changed */ - if (dc->clk_mgr) - dc->res_pool->funcs->update_bw_bounding_box(dc, dc->clk_mgr->bw_params); + dc->res_pool->funcs->update_bw_bounding_box(dc, dc->clk_mgr->bw_params); } } if (dc->res_pool->pg_cntl) { From c7de5a863c63de621843ed964a10662408e974f2 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Thu, 25 Dec 2025 23:44:20 +0800 Subject: [PATCH 68/99] drm/amd/pm: add populate_umd_state_clk support add populate_umd_state_clk support for smu 15.0.8 v2: remove gl2clk/socclk/fclk, restrict to only current min/max (Lijo) Signed-off-by: Asad Kamal Signed-off-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 2a74acb2790f..c13d9e506082 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -1376,6 +1376,21 @@ static int smu_v15_0_8_get_power_limit(struct smu_context *smu, return 0; } +static int smu_v15_0_8_populate_umd_state_clk(struct smu_context *smu) +{ + struct smu_15_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; + struct smu_dpm_table *gfx_table = &dpm_context->dpm_tables.gfx_table; + struct smu_dpm_table *mem_table = &dpm_context->dpm_tables.uclk_table; + struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; + + pstate_table->gfxclk_pstate.curr.min = SMU_DPM_TABLE_MIN(gfx_table); + pstate_table->gfxclk_pstate.curr.max = SMU_DPM_TABLE_MAX(gfx_table); + + pstate_table->uclk_pstate.curr.min = SMU_DPM_TABLE_MIN(mem_table); + pstate_table->uclk_pstate.curr.max = SMU_DPM_TABLE_MAX(mem_table); + return 0; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -1404,7 +1419,8 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .get_power_limit = smu_v15_0_8_get_power_limit, .set_power_limit = smu_v15_0_set_power_limit, .emit_clk_levels = smu_v15_0_8_emit_clk_levels, -}; + .populate_umd_state_clk = smu_v15_0_8_populate_umd_state_clk, + }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, const struct cmn2asic_msg_mapping *message_map) From 422b399b09c7b4fd60b01a5635fbb73e8d9a1154 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 26 Dec 2025 00:13:13 +0800 Subject: [PATCH 69/99] drm/amd/pm: Add od_edit_dpm_table support Add od_edit_dpm_table support for smu_v15_0_8 v2: Skip Gl2clk/Fclk (Lijo) v3: sqaush in set_performance_support (Asad) Signed-off-by: Asad Kamal Signed-off-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 296 +++++++++++++++++- drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 10 + drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h | 1 + 3 files changed, 306 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index c13d9e506082..6283e6b404f0 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -1391,6 +1391,298 @@ static int smu_v15_0_8_populate_umd_state_clk(struct smu_context *smu) return 0; } +static int smu_v15_0_8_set_gfx_soft_freq_limited_range(struct smu_context *smu, + uint32_t min, + uint32_t max) +{ + int ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMaxGfxClk, + max & 0xffff, NULL); + if (ret) + return ret; + + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMinGfxclk, + min & 0xffff, NULL); + + return ret; +} + +static int smu_v15_0_8_set_performance_level(struct smu_context *smu, + enum amd_dpm_forced_level level) +{ + struct smu_dpm_context *smu_dpm = &smu->smu_dpm; + struct smu_15_0_dpm_context *dpm_context = smu_dpm->dpm_context; + struct smu_dpm_table *gfx_table = &dpm_context->dpm_tables.gfx_table; + struct smu_dpm_table *uclk_table = &dpm_context->dpm_tables.uclk_table; + struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; + int ret; + + switch (level) { + case AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM: + /* Determinism not supported on SMU v15.0.8 */ + ret = -EOPNOTSUPP; + break; + + case AMD_DPM_FORCED_LEVEL_AUTO: + /* Restore GFXCLK to default range */ + if ((SMU_DPM_TABLE_MIN(gfx_table) != + pstate_table->gfxclk_pstate.curr.min) || + (SMU_DPM_TABLE_MAX(gfx_table) != + pstate_table->gfxclk_pstate.curr.max)) { + ret = smu_v15_0_8_set_gfx_soft_freq_limited_range( + smu, SMU_DPM_TABLE_MIN(gfx_table), + SMU_DPM_TABLE_MAX(gfx_table)); + if (ret) + goto out; + + pstate_table->gfxclk_pstate.curr.min = + SMU_DPM_TABLE_MIN(gfx_table); + pstate_table->gfxclk_pstate.curr.max = + SMU_DPM_TABLE_MAX(gfx_table); + } + + /* Restore UCLK to default max */ + if (SMU_DPM_TABLE_MAX(uclk_table) != + pstate_table->uclk_pstate.curr.max) { + /* Min UCLK is not expected to be changed */ + ret = smu_v15_0_set_soft_freq_limited_range(smu, + SMU_UCLK, 0, + SMU_DPM_TABLE_MAX(uclk_table), + false); + if (ret) + goto out; + + pstate_table->uclk_pstate.curr.max = + SMU_DPM_TABLE_MAX(uclk_table); + } + + if (ret) + goto out; + + smu_cmn_reset_custom_level(smu); + + break; + case AMD_DPM_FORCED_LEVEL_MANUAL: + ret = 0; + break; + default: + ret = -EOPNOTSUPP; + break; + } + +out: + return ret; +} + +static int smu_v15_0_8_set_soft_freq_limited_range(struct smu_context *smu, + enum smu_clk_type clk_type, + uint32_t min, uint32_t max, + bool automatic) +{ + struct smu_dpm_context *smu_dpm = &smu->smu_dpm; + struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; + int ret = 0; + + if (clk_type != SMU_GFXCLK && clk_type != SMU_SCLK && + clk_type != SMU_UCLK) + return -EINVAL; + + if (smu_dpm->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) + return -EINVAL; + + if (smu_dpm->dpm_level == AMD_DPM_FORCED_LEVEL_MANUAL) { + if (min >= max) { + dev_err(smu->adev->dev, + "Minimum clk should be less than the maximum allowed clock\n"); + return -EINVAL; + } + + if (clk_type == SMU_GFXCLK || clk_type == SMU_SCLK) { + if ((min == pstate_table->gfxclk_pstate.curr.min) && + (max == pstate_table->gfxclk_pstate.curr.max)) + return 0; + + ret = smu_v15_0_8_set_gfx_soft_freq_limited_range(smu, + min, max); + if (!ret) { + pstate_table->gfxclk_pstate.curr.min = min; + pstate_table->gfxclk_pstate.curr.max = max; + } + } + + if (clk_type == SMU_UCLK) { + if (max == pstate_table->uclk_pstate.curr.max) + return 0; + + ret = smu_v15_0_set_soft_freq_limited_range(smu, + SMU_UCLK, + 0, max, + false); + if (!ret) + pstate_table->uclk_pstate.curr.max = max; + } + + return ret; + } + + return 0; +} + +static int smu_v15_0_8_od_edit_dpm_table(struct smu_context *smu, + enum PP_OD_DPM_TABLE_COMMAND type, + long input[], uint32_t size) +{ + struct smu_dpm_context *smu_dpm = &smu->smu_dpm; + struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; + struct smu_15_0_dpm_context *dpm_context = smu_dpm->dpm_context; + uint32_t min_clk, max_clk; + int ret; + + /* Only allowed in manual mode */ + if (smu_dpm->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) + return -EINVAL; + + switch (type) { + case PP_OD_EDIT_SCLK_VDDC_TABLE: + if (size != 2) { + dev_err(smu->adev->dev, + "Input parameter number not correct\n"); + return -EINVAL; + } + min_clk = SMU_DPM_TABLE_MIN(&dpm_context->dpm_tables.gfx_table); + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.gfx_table); + if (input[0] == 0) { + if (input[1] < min_clk) { + dev_warn(smu->adev->dev, + "Minimum GFX clk (%ld) MHz specified is less than the minimum allowed (%d) MHz\n", + input[1], min_clk); + pstate_table->gfxclk_pstate.custom.min = + pstate_table->gfxclk_pstate.curr.min; + return -EINVAL; + } + + pstate_table->gfxclk_pstate.custom.min = input[1]; + } else if (input[0] == 1) { + if (input[1] > max_clk) { + dev_warn(smu->adev->dev, + "Maximum GFX clk (%ld) MHz specified is greater than the maximum allowed (%d) MHz\n", + input[1], max_clk); + pstate_table->gfxclk_pstate.custom.max = + pstate_table->gfxclk_pstate.curr.max; + return -EINVAL; + } + + pstate_table->gfxclk_pstate.custom.max = input[1]; + } else { + return -EINVAL; + } + break; + case PP_OD_EDIT_MCLK_VDDC_TABLE: + if (size != 2) { + dev_err(smu->adev->dev, + "Input parameter number not correct\n"); + return -EINVAL; + } + + if (!smu_cmn_feature_is_enabled(smu, SMU_FEATURE_DPM_UCLK_BIT)) { + dev_warn(smu->adev->dev, + "UCLK_LIMITS setting not supported!\n"); + return -EOPNOTSUPP; + } + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.uclk_table); + if (input[0] == 0) { + dev_info(smu->adev->dev, + "Setting min UCLK level is not supported"); + return -EINVAL; + } else if (input[0] == 1) { + if (input[1] > max_clk) { + dev_warn(smu->adev->dev, + "Maximum UCLK (%ld) MHz specified is greater than the maximum allowed (%d) MHz\n", + input[1], max_clk); + pstate_table->uclk_pstate.custom.max = + pstate_table->uclk_pstate.curr.max; + + return -EINVAL; + } + + pstate_table->uclk_pstate.custom.max = input[1]; + } + break; + case PP_OD_RESTORE_DEFAULT_TABLE: + if (size != 0) { + dev_err(smu->adev->dev, + "Input parameter number not correct\n"); + return -EINVAL; + } + + /* Use the default frequencies for manual mode */ + min_clk = SMU_DPM_TABLE_MIN(&dpm_context->dpm_tables.gfx_table); + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.gfx_table); + + ret = smu_v15_0_8_set_soft_freq_limited_range(smu, + SMU_GFXCLK, + min_clk, max_clk, + false); + if (ret) + return ret; + + min_clk = SMU_DPM_TABLE_MIN(&dpm_context->dpm_tables.uclk_table); + max_clk = SMU_DPM_TABLE_MAX(&dpm_context->dpm_tables.uclk_table); + ret = smu_v15_0_8_set_soft_freq_limited_range(smu, + SMU_UCLK, + min_clk, max_clk, + false); + if (ret) + return ret; + + smu_cmn_reset_custom_level(smu); + break; + case PP_OD_COMMIT_DPM_TABLE: + if (size != 0) { + dev_err(smu->adev->dev, + "Input parameter number not correct\n"); + return -EINVAL; + } + + if (!pstate_table->gfxclk_pstate.custom.min) + pstate_table->gfxclk_pstate.custom.min = + pstate_table->gfxclk_pstate.curr.min; + + if (!pstate_table->gfxclk_pstate.custom.max) + pstate_table->gfxclk_pstate.custom.max = + pstate_table->gfxclk_pstate.curr.max; + + min_clk = pstate_table->gfxclk_pstate.custom.min; + max_clk = pstate_table->gfxclk_pstate.custom.max; + + ret = smu_v15_0_8_set_soft_freq_limited_range(smu, + SMU_GFXCLK, + min_clk, max_clk, + false); + if (ret) + return ret; + + /* Commit UCLK custom range (only max supported) */ + if (pstate_table->uclk_pstate.custom.max) { + min_clk = pstate_table->uclk_pstate.curr.min; + max_clk = pstate_table->uclk_pstate.custom.max; + ret = smu_v15_0_8_set_soft_freq_limited_range(smu, + SMU_UCLK, + min_clk, max_clk, + false); + if (ret) + return ret; + } + + break; + default: + return -ENOSYS; + } + + return 0; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -1420,7 +1712,9 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .set_power_limit = smu_v15_0_set_power_limit, .emit_clk_levels = smu_v15_0_8_emit_clk_levels, .populate_umd_state_clk = smu_v15_0_8_populate_umd_state_clk, - }; + .set_performance_level = smu_v15_0_8_set_performance_level, + .od_edit_dpm_table = smu_v15_0_8_od_edit_dpm_table, +}; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, const struct cmn2asic_msg_mapping *message_map) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index 5ec2b4e6f3f3..7bd8c435466a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -1301,6 +1301,16 @@ void smu_cmn_get_backend_workload_mask(struct smu_context *smu, } } +void smu_cmn_reset_custom_level(struct smu_context *smu) +{ + struct smu_umd_pstate_table *pstate_table = &smu->pstate_table; + + pstate_table->gfxclk_pstate.custom.min = 0; + pstate_table->gfxclk_pstate.custom.max = 0; + pstate_table->uclk_pstate.custom.min = 0; + pstate_table->uclk_pstate.custom.max = 0; +} + static inline bool smu_cmn_freqs_match(uint32_t freq1, uint32_t freq2) { /* Frequencies within 25 MHz are considered equal */ diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h index e4d282d8bcae..b76e86df5da7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h @@ -204,6 +204,7 @@ int smu_cmn_print_pcie_levels(struct smu_context *smu, struct smu_pcie_table *pcie_table, uint32_t cur_gen, uint32_t cur_lane, char *buf, int *offset); +void smu_cmn_reset_custom_level(struct smu_context *smu); int smu_cmn_dpm_pcie_gen_idx(int gen); int smu_cmn_dpm_pcie_width_idx(int width); From bc296e6c9564b952d9f2ecae5de6c8c7d38c3a25 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 26 Dec 2025 00:30:41 +0800 Subject: [PATCH 70/99] drm/amd/pm: Add get_thermal_temperature_range support Add get_thermal_temperature_range support smu_v15_0_8 v2: Remove sriov check (Lijo) v3: Restrict to 1VF mode(Lijo) Signed-off-by: Asad Kamal Signed-off-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 33 +++++++++++++++++++ 1 file changed, 33 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 6283e6b404f0..904779f67b06 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -1683,6 +1683,38 @@ static int smu_v15_0_8_od_edit_dpm_table(struct smu_context *smu, return 0; } +static int smu_v15_0_8_get_thermal_temperature_range(struct smu_context *smu, + struct smu_temperature_range *range) +{ + struct smu_table_context *smu_table = &smu->smu_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + uint32_t max_ctf, max_thm; + + if (amdgpu_sriov_multi_vf_mode(smu->adev)) + return 0; + + if (!range) + return -EINVAL; + + /* CTF (Critical Temperature Fault) limits */ + max_ctf = max3(pptable->CTFLimitMID, pptable->CTFLimitXCD, + pptable->CTFLimitAID); + range->hotspot_emergency_max = max_ctf * SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + + range->mem_emergency_max = pptable->CTFLimitHBM * + SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + + /* Thermal throttling limits */ + max_thm = max3(pptable->ThermalLimitMID, pptable->ThermalLimitXCD, + pptable->ThermalLimitAID); + range->hotspot_crit_max = max_thm * SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + + range->mem_crit_max = pptable->ThermalLimitHBM * + SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; + + return 0; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -1714,6 +1746,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .populate_umd_state_clk = smu_v15_0_8_populate_umd_state_clk, .set_performance_level = smu_v15_0_8_set_performance_level, .od_edit_dpm_table = smu_v15_0_8_od_edit_dpm_table, + .get_thermal_temperature_range = smu_v15_0_8_get_thermal_temperature_range, }; static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, From a6c2ecd95ef0e6e06db4a76fbeafdea1f742e280 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 26 Dec 2025 13:00:48 +0800 Subject: [PATCH 71/99] drm/amd/pm: Add ppt1 support Add ppt1 support for smu_v15_0_8 Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 75 ++++++++++++++++++- 1 file changed, 74 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 904779f67b06..abfaf8b06e50 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -1715,6 +1715,78 @@ static int smu_v15_0_8_get_thermal_temperature_range(struct smu_context *smu, return 0; } +static int smu_v15_0_8_set_power_limit(struct smu_context *smu, + enum smu_ppt_limit_type limit_type, + uint32_t limit) +{ + struct smu_table_context *smu_table = &smu->smu_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + int ret; + + if (limit_type == SMU_FAST_PPT_LIMIT) { + if (!pptable->PPT1Max) + return -EOPNOTSUPP; + + if (limit > pptable->PPT1Max || limit < pptable->PPT1Min) { + dev_err(smu->adev->dev, + "New PPT1 limit (%d) should be between min %d and max %d\n", + limit, pptable->PPT1Min, pptable->PPT1Max); + return -EINVAL; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetFastPptLimit, + limit, NULL); + if (ret) + dev_err(smu->adev->dev, "Set fast PPT limit failed!\n"); + + return ret; + } + + return smu_v15_0_set_power_limit(smu, limit_type, limit); +} + +static int smu_v15_0_8_get_ppt_limit(struct smu_context *smu, + uint32_t *ppt_limit, + enum smu_ppt_limit_type type, + enum smu_ppt_limit_level level) +{ + struct smu_table_context *smu_table = &smu->smu_table; + PPTable_t *pptable = (PPTable_t *)smu_table->driver_pptable; + int ret = 0; + + if (!ppt_limit) + return -EINVAL; + + if (type == SMU_FAST_PPT_LIMIT) { + if (!pptable->PPT1Max) + return -EOPNOTSUPP; + + switch (level) { + case SMU_PPT_LIMIT_MAX: + *ppt_limit = pptable->PPT1Max; + break; + case SMU_PPT_LIMIT_CURRENT: + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetFastPptLimit, + ppt_limit); + if (ret) + dev_err(smu->adev->dev, + "Get fast PPT limit failed!\n"); + break; + case SMU_PPT_LIMIT_DEFAULT: + *ppt_limit = pptable->PPT1Default; + break; + case SMU_PPT_LIMIT_MIN: + *ppt_limit = pptable->PPT1Min; + break; + default: + return -EOPNOTSUPP; + } + return ret; + } + + return -EOPNOTSUPP; +} + static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .init_allowed_features = smu_v15_0_8_init_allowed_features, .set_default_dpm_table = smu_v15_0_8_set_default_dpm_table, @@ -1741,7 +1813,8 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .get_gpu_metrics = smu_v15_0_8_get_gpu_metrics, .get_unique_id = smu_v15_0_8_get_unique_id, .get_power_limit = smu_v15_0_8_get_power_limit, - .set_power_limit = smu_v15_0_set_power_limit, + .set_power_limit = smu_v15_0_8_set_power_limit, + .get_ppt_limit = smu_v15_0_8_get_ppt_limit, .emit_clk_levels = smu_v15_0_8_emit_clk_levels, .populate_umd_state_clk = smu_v15_0_8_populate_umd_state_clk, .set_performance_level = smu_v15_0_8_set_performance_level, From 1415503db03b0e44f2d70d2c2c87c83c6fed9f46 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 26 Dec 2025 13:12:56 +0800 Subject: [PATCH 72/99] drm/amd/pm: Add read sensor support Add read sensor support for smu_v15_0_8 v2: Remove gfx voltage support (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 112 ++++++++++++++++++ 1 file changed, 112 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index abfaf8b06e50..8a49f1a190af 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -434,6 +434,117 @@ static int smu_v15_0_8_get_current_clk_freq_by_table(struct smu_context *smu, return smu_v15_0_8_get_smu_metrics_data(smu, member_type, value); } +static int smu_v15_0_8_get_current_activity_percent(struct smu_context *smu, + enum amd_pp_sensors sensor, + uint32_t *value) +{ + int ret = 0; + + if (!value) + return -EINVAL; + + switch (sensor) { + case AMDGPU_PP_SENSOR_GPU_LOAD: + ret = smu_v15_0_8_get_smu_metrics_data(smu, + METRICS_AVERAGE_GFXACTIVITY, value); + break; + case AMDGPU_PP_SENSOR_MEM_LOAD: + ret = smu_v15_0_8_get_smu_metrics_data(smu, + METRICS_AVERAGE_MEMACTIVITY, value); + break; + default: + dev_err(smu->adev->dev, + "Invalid sensor for retrieving clock activity\n"); + return -EINVAL; + } + + return ret; +} + +static int smu_v15_0_8_thermal_get_temperature(struct smu_context *smu, + enum amd_pp_sensors sensor, + uint32_t *value) +{ + int ret = 0; + + if (!value) + return -EINVAL; + + switch (sensor) { + case AMDGPU_PP_SENSOR_HOTSPOT_TEMP: + ret = smu_v15_0_8_get_smu_metrics_data(smu, + METRICS_TEMPERATURE_HOTSPOT, value); + break; + case AMDGPU_PP_SENSOR_MEM_TEMP: + ret = smu_v15_0_8_get_smu_metrics_data(smu, + METRICS_TEMPERATURE_MEM, value); + break; + default: + dev_err(smu->adev->dev, "Invalid sensor for retrieving temp\n"); + return -EINVAL; + } + + return ret; +} + +static int smu_v15_0_8_read_sensor(struct smu_context *smu, + enum amd_pp_sensors sensor, void *data, + uint32_t *size) +{ + struct smu_15_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; + int ret = 0; + + if (amdgpu_ras_intr_triggered()) + return 0; + + if (!data || !size) + return -EINVAL; + + switch (sensor) { + case AMDGPU_PP_SENSOR_MEM_LOAD: + case AMDGPU_PP_SENSOR_GPU_LOAD: + ret = smu_v15_0_8_get_current_activity_percent(smu, sensor, + (uint32_t *)data); + *size = 4; + break; + case AMDGPU_PP_SENSOR_GPU_INPUT_POWER: + ret = smu_v15_0_8_get_smu_metrics_data(smu, + METRICS_CURR_SOCKETPOWER, + (uint32_t *)data); + *size = 4; + break; + case AMDGPU_PP_SENSOR_HOTSPOT_TEMP: + case AMDGPU_PP_SENSOR_MEM_TEMP: + ret = smu_v15_0_8_thermal_get_temperature(smu, sensor, + (uint32_t *)data); + *size = 4; + break; + case AMDGPU_PP_SENSOR_GFX_MCLK: + ret = smu_v15_0_8_get_current_clk_freq_by_table(smu, + SMU_UCLK, (uint32_t *)data); + /* the output clock frequency in 10K unit */ + *(uint32_t *)data *= 100; + *size = 4; + break; + case AMDGPU_PP_SENSOR_GFX_SCLK: + ret = smu_v15_0_8_get_current_clk_freq_by_table(smu, + SMU_GFXCLK, (uint32_t *)data); + *(uint32_t *)data *= 100; + *size = 4; + break; + case AMDGPU_PP_SENSOR_VDDBOARD: + *(uint32_t *)data = dpm_context->board_volt; + *size = 4; + break; + case AMDGPU_PP_SENSOR_GPU_AVG_POWER: + default: + ret = -EOPNOTSUPP; + break; + } + + return ret; +} + static int smu_v15_0_8_emit_clk_levels(struct smu_context *smu, enum smu_clk_type type, char *buf, int *offset) @@ -1816,6 +1927,7 @@ static const struct pptable_funcs smu_v15_0_8_ppt_funcs = { .set_power_limit = smu_v15_0_8_set_power_limit, .get_ppt_limit = smu_v15_0_8_get_ppt_limit, .emit_clk_levels = smu_v15_0_8_emit_clk_levels, + .read_sensor = smu_v15_0_8_read_sensor, .populate_umd_state_clk = smu_v15_0_8_populate_umd_state_clk, .set_performance_level = smu_v15_0_8_set_performance_level, .od_edit_dpm_table = smu_v15_0_8_od_edit_dpm_table, From e3b96f5b20ab5178ded733351048ce64bb989a9e Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 30 Jan 2026 21:10:47 +0800 Subject: [PATCH 73/99] drm/amd/pm: Add gpuboard temperature metrics support Add gpuboard temperature metrics support via system metrics table for smu_v15_0_8 v3: Use per sensor attr id (Lijo) v4: Use s16 for temp, remove cast, use separate function to fill gpuboard temperature metrics data (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../gpu/drm/amd/include/kgd_pp_interface.h | 35 ++++ .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 172 ++++++++++++++++++ .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h | 71 ++++++++ 3 files changed, 278 insertions(+) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 89b440edd15b..ff91f16af05c 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -590,6 +590,35 @@ enum amdgpu_metrics_attr_id { AMDGPU_METRICS_ATTR_ID_TEMPERATURE_MID, AMDGPU_METRICS_ATTR_ID_TEMPERATURE_AID, AMDGPU_METRICS_ATTR_ID_TEMPERATURE_XCD, + AMDGPU_METRICS_ATTR_ID_LABEL_VERSION, + AMDGPU_METRICS_ATTR_ID_NODE_ID, + AMDGPU_METRICS_ATTR_ID_NODE_TEMP_RETIMER, + AMDGPU_METRICS_ATTR_ID_NODE_TEMP_IBC, + AMDGPU_METRICS_ATTR_ID_NODE_TEMP_IBC_2, + AMDGPU_METRICS_ATTR_ID_NODE_TEMP_VDD18_VR, + AMDGPU_METRICS_ATTR_ID_NODE_TEMP_04_HBM_B_VR, + AMDGPU_METRICS_ATTR_ID_NODE_TEMP_04_HBM_D_VR, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_SOCIO_A, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_SOCIO_C, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_X0, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_X1, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_HBM_B, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_HBM_D, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_04_HBM_B, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_04_HBM_D, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_HBM_B, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_HBM_D, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_075_HBM_B, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_075_HBM_D, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_11_GTA_A, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_11_GTA_C, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDAN_075_GTA_A, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDAN_075_GTA_C, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDCR_075_UCIE, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_065_UCIEAA, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_065_UCIEAM_A, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_065_UCIEAM_C, + AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDAN_075, AMDGPU_METRICS_ATTR_ID_MAX, }; @@ -1842,4 +1871,10 @@ enum amdgpu_xgmi_link_status { AMDGPU_XGMI_LINK_NA = 2, }; +struct amdgpu_gpuboard_temp_metrics_v1_1 { + struct metrics_table_header common_header; + int attr_count; + struct gpu_metrics_attr metrics_attrs[]; +}; + #endif diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 8a49f1a190af..4c1a8f37d12c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -171,8 +171,14 @@ static const struct cmn2asic_mapping smu_v15_0_8_table_map[SMU_TABLE_COUNT] = { TAB_MAP(I2C_COMMANDS), }; +static size_t smu_v15_0_8_get_system_metrics_size(void) +{ + return sizeof(SystemMetricsTable_t); +} + static int smu_v15_0_8_tables_init(struct smu_context *smu) { + struct smu_v15_0_8_gpuboard_temp_metrics *gpuboard_temp_metrics; struct smu_table_context *smu_table = &smu->smu_table; int ret, gpu_metrcs_size = sizeof(MetricsTable_t); struct smu_table *tables = smu_table->tables; @@ -187,6 +193,9 @@ static int smu_v15_0_8_tables_init(struct smu_context *smu) gpu_metrcs_size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); + SMU_TABLE_INIT(tables, SMU_TABLE_PMFW_SYSTEM_METRICS, + smu_v15_0_8_get_system_metrics_size(), PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); metrics_table = kzalloc(gpu_metrcs_size, GFP_KERNEL); if (!metrics_table) @@ -207,6 +216,25 @@ static int smu_v15_0_8_tables_init(struct smu_context *smu) gpu_metrics = (struct smu_v15_0_8_gpu_metrics *)smu_driver_table_ptr(smu, SMU_DRIVER_TABLE_GPU_METRICS); smu_v15_0_8_gpu_metrics_init(gpu_metrics, 1, 9); + + ret = smu_table_cache_init(smu, SMU_TABLE_PMFW_SYSTEM_METRICS, + smu_v15_0_8_get_system_metrics_size(), 5); + + if (ret) + return ret; + + /* Initialize GPU board temperature metrics */ + ret = smu_driver_table_init(smu, SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS, + sizeof(*gpuboard_temp_metrics), 50); + if (ret) { + smu_table_cache_fini(smu, SMU_TABLE_PMFW_SYSTEM_METRICS); + return ret; + } + gpuboard_temp_metrics = (struct smu_v15_0_8_gpuboard_temp_metrics *) + smu_driver_table_ptr(smu, + SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS); + smu_v15_0_8_gpuboard_temp_metrics_init(gpuboard_temp_metrics, 1, 1); + smu_table->metrics_table = no_free_ptr(metrics_table); smu_table->driver_pptable = no_free_ptr(driver_pptable); @@ -252,6 +280,8 @@ static int smu_v15_0_8_tables_fini(struct smu_context *smu) { struct smu_table_context *smu_table = &smu->smu_table; + smu_driver_table_fini(smu, SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS); + smu_table_cache_fini(smu, SMU_TABLE_PMFW_SYSTEM_METRICS); mutex_destroy(&smu_table->metrics_lock); return 0; @@ -487,6 +517,33 @@ static int smu_v15_0_8_thermal_get_temperature(struct smu_context *smu, return ret; } +static int smu_v15_0_8_get_system_metrics_table(struct smu_context *smu) +{ + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_table *table = &smu_table->driver_table; + struct smu_table *tables = smu_table->tables; + struct smu_table *sys_table; + int ret; + + sys_table = &tables[SMU_TABLE_PMFW_SYSTEM_METRICS]; + if (smu_table_cache_is_valid(sys_table)) + return 0; + + ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetSystemMetricsTable, NULL); + if (ret) { + dev_info(smu->adev->dev, + "Failed to export system metrics table!\n"); + return ret; + } + + amdgpu_hdp_invalidate(smu->adev, NULL); + smu_table_cache_update_time(sys_table, jiffies); + memcpy(sys_table->cache.buffer, table->cpu_addr, + sizeof(SystemMetricsTable_t)); + + return 0; +} + static int smu_v15_0_8_read_sensor(struct smu_context *smu, enum amd_pp_sensors sensor, void *data, uint32_t *size) @@ -1292,6 +1349,115 @@ static int smu_v15_0_8_mode2_reset(struct smu_context *smu) return ret; } +static bool smu_v15_0_8_is_temp_metrics_supported(struct smu_context *smu, + enum smu_temp_metric_type type) +{ + switch (type) { + case SMU_TEMP_METRIC_GPUBOARD: + return true; + default: + return false; + } +} + +static void smu_v15_0_8_fill_gpuboard_temp_metrics( + struct smu_v15_0_8_gpuboard_temp_metrics *gpuboard_temp_metrics, + const SystemMetricsTable_t *metrics) +{ + gpuboard_temp_metrics->accumulation_counter = metrics->AccumulationCounter; + gpuboard_temp_metrics->label_version = metrics->LabelVersion; + gpuboard_temp_metrics->node_id = metrics->NodeIdentifier; + + gpuboard_temp_metrics->node_temp_retimer = + metrics->NodeTemperatures[NODE_TEMP_RETIMER]; + gpuboard_temp_metrics->node_temp_ibc = + metrics->NodeTemperatures[NODE_TEMP_IBC_TEMP]; + gpuboard_temp_metrics->node_temp_ibc_2 = + metrics->NodeTemperatures[NODE_TEMP_IBC_2_TEMP]; + gpuboard_temp_metrics->node_temp_vdd18_vr = + metrics->NodeTemperatures[NODE_TEMP_VDD18_VR_TEMP]; + gpuboard_temp_metrics->node_temp_04_hbm_b_vr = + metrics->NodeTemperatures[NODE_TEMP_04_HBM_B_VR_TEMP]; + gpuboard_temp_metrics->node_temp_04_hbm_d_vr = + metrics->NodeTemperatures[NODE_TEMP_04_HBM_D_VR_TEMP]; + + gpuboard_temp_metrics->vr_temp_vddcr_socio_a = + metrics->VrTemperatures[SVI_PLANE_VDDCR_SOCIO_A_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_socio_c = + metrics->VrTemperatures[SVI_PLANE_VDDCR_SOCIO_C_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_x0 = + metrics->VrTemperatures[SVI_PLANE_VDDCR_X0_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_x1 = + metrics->VrTemperatures[SVI_PLANE_VDDCR_X1_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_hbm_b = + metrics->VrTemperatures[SVI_PLANE_VDDIO_HBM_B_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_hbm_d = + metrics->VrTemperatures[SVI_PLANE_VDDIO_HBM_D_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_04_hbm_b = + metrics->VrTemperatures[SVI_PLANE_VDDIO_04_HBM_B_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_04_hbm_d = + metrics->VrTemperatures[SVI_PLANE_VDDIO_04_HBM_D_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_hbm_b = + metrics->VrTemperatures[SVI_PLANE_VDDCR_HBM_B_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_hbm_d = + metrics->VrTemperatures[SVI_PLANE_VDDCR_HBM_D_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_075_hbm_b = + metrics->VrTemperatures[SVI_PLANE_VDDCR_075_HBM_B_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_075_hbm_d = + metrics->VrTemperatures[SVI_PLANE_VDDCR_075_HBM_D_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_11_gta_a = + metrics->VrTemperatures[SVI_PLANE_VDDIO_11_GTA_A_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_11_gta_c = + metrics->VrTemperatures[SVI_PLANE_VDDIO_11_GTA_C_TEMP]; + gpuboard_temp_metrics->vr_temp_vddan_075_gta_a = + metrics->VrTemperatures[SVI_PLANE_VDDAN_075_GTA_A_TEMP]; + gpuboard_temp_metrics->vr_temp_vddan_075_gta_c = + metrics->VrTemperatures[SVI_PLANE_VDDAN_075_GTA_C_TEMP]; + gpuboard_temp_metrics->vr_temp_vddcr_075_ucie = + metrics->VrTemperatures[SVI_PLANE_VDDCR_075_UCIE_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_065_ucieaa = + metrics->VrTemperatures[SVI_PLANE_VDDIO_065_UCIEAA_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_065_ucieam_a = + metrics->VrTemperatures[SVI_PLANE_VDDIO_065_UCIEAM_A_TEMP]; + gpuboard_temp_metrics->vr_temp_vddio_065_ucieam_c = + metrics->VrTemperatures[SVI_PLANE_VDDIO_065_UCIEAM_C_TEMP]; + gpuboard_temp_metrics->vr_temp_vddan_075 = + metrics->VrTemperatures[SVI_PLANE_VDDAN_075_TEMP]; +} + +static ssize_t smu_v15_0_8_get_temp_metrics(struct smu_context *smu, + enum smu_temp_metric_type type, + void *table) +{ + struct smu_v15_0_8_gpuboard_temp_metrics *gpuboard_temp_metrics; + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_table *tables = smu_table->tables; + enum smu_driver_table_id table_id; + SystemMetricsTable_t *metrics; + struct smu_table *sys_table; + ssize_t size; + int ret; + + table_id = SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS; + gpuboard_temp_metrics = + (struct smu_v15_0_8_gpuboard_temp_metrics *) + smu_driver_table_ptr(smu, table_id); + size = sizeof(*gpuboard_temp_metrics); + + ret = smu_v15_0_8_get_system_metrics_table(smu); + if (ret) + return ret; + + sys_table = &tables[SMU_TABLE_PMFW_SYSTEM_METRICS]; + metrics = (SystemMetricsTable_t *)sys_table->cache.buffer; + smu_driver_table_update_cache_time(smu, table_id); + + smu_v15_0_8_fill_gpuboard_temp_metrics(gpuboard_temp_metrics, + metrics); + memcpy(table, gpuboard_temp_metrics, size); + return size; +} + static ssize_t smu_v15_0_8_get_gpu_metrics(struct smu_context *smu, void **table) { struct smu_table_context *smu_table = &smu->smu_table; @@ -1954,6 +2120,11 @@ static void smu_v15_0_8_init_msg_ctl(struct smu_context *smu, ctl->message_map = message_map; } +static const struct smu_temp_funcs smu_v15_0_8_temp_funcs = { + .temp_metrics_is_supported = smu_v15_0_8_is_temp_metrics_supported, + .get_temp_metrics = smu_v15_0_8_get_temp_metrics, +}; + void smu_v15_0_8_set_ppt_funcs(struct smu_context *smu) { smu->ppt_funcs = &smu_v15_0_8_ppt_funcs; @@ -1961,5 +2132,6 @@ void smu_v15_0_8_set_ppt_funcs(struct smu_context *smu) smu->feature_map = smu_v15_0_8_feature_mask_map; smu->table_map = smu_v15_0_8_table_map; smu_v15_0_8_init_msg_ctl(smu, smu_v15_0_8_message_map); + smu->smu_temp.temp_funcs = &smu_v15_0_8_temp_funcs; smu->smc_driver_if_version = SMU15_DRIVER_IF_VERSION_SMU_V15_0_8; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h index 8fc16796788b..0856d11d8e55 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h @@ -180,5 +180,76 @@ typedef struct { SMU_15_0_8_MAX_XCC); DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_gpu_metrics, SMU_15_0_8_METRICS_FIELDS); + +/* Maximum temperature sensor counts for system metrics */ +#define SMU_15_0_8_MAX_NODE_TEMP_ENTRIES 12 +#define SMU_15_0_8_MAX_VR_TEMP_ENTRIES 22 + +/* SMUv 15.0.8 GPU board temperature metrics */ +#define SMU_15_0_8_GPUBOARD_TEMP_METRICS_FIELDS(SMU_SCALAR, SMU_ARRAY) \ + SMU_SCALAR(SMU_MATTR(ACCUMULATION_COUNTER), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), accumulation_counter); \ + SMU_SCALAR(SMU_MATTR(LABEL_VERSION), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), label_version); \ + SMU_SCALAR(SMU_MATTR(NODE_ID), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), node_id); \ + SMU_SCALAR(SMU_MATTR(NODE_TEMP_RETIMER), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), node_temp_retimer); \ + SMU_SCALAR(SMU_MATTR(NODE_TEMP_IBC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), node_temp_ibc); \ + SMU_SCALAR(SMU_MATTR(NODE_TEMP_IBC_2), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), node_temp_ibc_2); \ + SMU_SCALAR(SMU_MATTR(NODE_TEMP_VDD18_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), node_temp_vdd18_vr); \ + SMU_SCALAR(SMU_MATTR(NODE_TEMP_04_HBM_B_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), node_temp_04_hbm_b_vr); \ + SMU_SCALAR(SMU_MATTR(NODE_TEMP_04_HBM_D_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), node_temp_04_hbm_d_vr); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_SOCIO_A), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_socio_a); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_SOCIO_C), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_socio_c); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_X0), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_x0); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_X1), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_x1); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_HBM_B), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_hbm_b); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_HBM_D), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_hbm_d); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_04_HBM_B), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_04_hbm_b); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_04_HBM_D), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_04_hbm_d); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_HBM_B), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_hbm_b); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_HBM_D), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_hbm_d); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_075_HBM_B), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_075_hbm_b); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_075_HBM_D), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_075_hbm_d); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_11_GTA_A), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_11_gta_a); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_11_GTA_C), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_11_gta_c); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDAN_075_GTA_A), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddan_075_gta_a); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDAN_075_GTA_C), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddan_075_gta_c); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDCR_075_UCIE), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddcr_075_ucie); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_065_UCIEAA), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_065_ucieaa); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_065_UCIEAM_A), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_065_ucieam_a); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDIO_065_UCIEAM_C), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddio_065_ucieam_c); \ + SMU_SCALAR(SMU_MATTR(VR_TEMP_VDDAN_075), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), vr_temp_vddan_075); + +DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_gpuboard_temp_metrics, + SMU_15_0_8_GPUBOARD_TEMP_METRICS_FIELDS); + #endif #endif From 8847d59969eb38519b4790404248fa6828956daf Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 30 Jan 2026 21:18:29 +0800 Subject: [PATCH 74/99] drm/amd/pm: Add baseboard temperature metrics support Add baseboard temperature metrics support via system metrics table for smu_v15_0_8 v4: Add separate function to fill baseboard temperature, use 16, remove casting v5: Optimize to use single switch case (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../gpu/drm/amd/include/kgd_pp_interface.h | 29 +++++ .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 110 +++++++++++++++--- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h | 58 +++++++++ 3 files changed, 184 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index ff91f16af05c..1bbf531de5ed 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -619,6 +619,29 @@ enum amdgpu_metrics_attr_id { AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_065_UCIEAM_A, AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDIO_065_UCIEAM_C, AMDGPU_METRICS_ATTR_ID_VR_TEMP_VDDAN_075, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_FPGA, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_FRONT, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_BACK, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_OAM7, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_IBC, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_UFPGA, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_OAM1, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_OAM_0_1_HSC, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_OAM_2_3_HSC, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_OAM_4_5_HSC, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_OAM_6_7_HSC, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_FPGA_0V72_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_UBB_FPGA_3V3_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_RETIMER_0_1_2_3_1V2_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_RETIMER_4_5_6_7_1V2_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_RETIMER_0_1_0V9_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_RETIMER_4_5_0V9_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_RETIMER_2_3_0V9_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_RETIMER_6_7_0V9_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_OAM_0_1_2_3_3V3_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_OAM_4_5_6_7_3V3_VR, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_IBC_HSC, + AMDGPU_METRICS_ATTR_ID_SYSTEM_TEMP_IBC, AMDGPU_METRICS_ATTR_ID_MAX, }; @@ -1877,4 +1900,10 @@ struct amdgpu_gpuboard_temp_metrics_v1_1 { struct gpu_metrics_attr metrics_attrs[]; }; +struct amdgpu_baseboard_temp_metrics_v1_1 { + struct metrics_table_header common_header; + int attr_count; + struct gpu_metrics_attr metrics_attrs[]; +}; + #endif diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 4c1a8f37d12c..c850578f735c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -178,6 +178,7 @@ static size_t smu_v15_0_8_get_system_metrics_size(void) static int smu_v15_0_8_tables_init(struct smu_context *smu) { + struct smu_v15_0_8_baseboard_temp_metrics *baseboard_temp_metrics; struct smu_v15_0_8_gpuboard_temp_metrics *gpuboard_temp_metrics; struct smu_table_context *smu_table = &smu->smu_table; int ret, gpu_metrcs_size = sizeof(MetricsTable_t); @@ -223,11 +224,23 @@ static int smu_v15_0_8_tables_init(struct smu_context *smu) if (ret) return ret; + /* Initialize base board temperature metrics */ + ret = smu_driver_table_init(smu, + SMU_DRIVER_TABLE_BASEBOARD_TEMP_METRICS, + sizeof(*baseboard_temp_metrics), 50); + if (ret) + return ret; + baseboard_temp_metrics = (struct smu_v15_0_8_baseboard_temp_metrics *) + smu_driver_table_ptr(smu, + SMU_DRIVER_TABLE_BASEBOARD_TEMP_METRICS); + smu_v15_0_8_baseboard_temp_metrics_init(baseboard_temp_metrics, 1, 1); /* Initialize GPU board temperature metrics */ ret = smu_driver_table_init(smu, SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS, sizeof(*gpuboard_temp_metrics), 50); if (ret) { smu_table_cache_fini(smu, SMU_TABLE_PMFW_SYSTEM_METRICS); + smu_driver_table_fini(smu, + SMU_DRIVER_TABLE_BASEBOARD_TEMP_METRICS); return ret; } gpuboard_temp_metrics = (struct smu_v15_0_8_gpuboard_temp_metrics *) @@ -280,6 +293,7 @@ static int smu_v15_0_8_tables_fini(struct smu_context *smu) { struct smu_table_context *smu_table = &smu->smu_table; + smu_driver_table_fini(smu, SMU_DRIVER_TABLE_BASEBOARD_TEMP_METRICS); smu_driver_table_fini(smu, SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS); smu_table_cache_fini(smu, SMU_TABLE_PMFW_SYSTEM_METRICS); mutex_destroy(&smu_table->metrics_lock); @@ -1353,6 +1367,10 @@ static bool smu_v15_0_8_is_temp_metrics_supported(struct smu_context *smu, enum smu_temp_metric_type type) { switch (type) { + case SMU_TEMP_METRIC_BASEBOARD: + if (smu->adev->gmc.xgmi.physical_node_id == 0) + return true; + return false; case SMU_TEMP_METRIC_GPUBOARD: return true; default: @@ -1360,6 +1378,62 @@ static bool smu_v15_0_8_is_temp_metrics_supported(struct smu_context *smu, } } +static void smu_v15_0_8_fill_baseboard_temp_metrics( + struct smu_v15_0_8_baseboard_temp_metrics *baseboard_temp_metrics, + const SystemMetricsTable_t *metrics) +{ + baseboard_temp_metrics->accumulation_counter = metrics->AccumulationCounter; + baseboard_temp_metrics->label_version = metrics->LabelVersion; + baseboard_temp_metrics->node_id = metrics->NodeIdentifier; + + baseboard_temp_metrics->system_temp_ubb_fpga = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_FPGA]; + baseboard_temp_metrics->system_temp_ubb_front = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_FRONT]; + baseboard_temp_metrics->system_temp_ubb_back = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_BACK]; + baseboard_temp_metrics->system_temp_ubb_oam7 = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_OAM7]; + baseboard_temp_metrics->system_temp_ubb_ibc = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_IBC]; + baseboard_temp_metrics->system_temp_ubb_ufpga = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_UFPGA]; + baseboard_temp_metrics->system_temp_ubb_oam1 = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_OAM1]; + baseboard_temp_metrics->system_temp_oam_0_1_hsc = + metrics->SystemTemperatures[SYSTEM_TEMP_OAM_0_1_HSC]; + baseboard_temp_metrics->system_temp_oam_2_3_hsc = + metrics->SystemTemperatures[SYSTEM_TEMP_OAM_2_3_HSC]; + baseboard_temp_metrics->system_temp_oam_4_5_hsc = + metrics->SystemTemperatures[SYSTEM_TEMP_OAM_4_5_HSC]; + baseboard_temp_metrics->system_temp_oam_6_7_hsc = + metrics->SystemTemperatures[SYSTEM_TEMP_OAM_6_7_HSC]; + baseboard_temp_metrics->system_temp_ubb_fpga_0v72_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_FPGA_0V72_VR]; + baseboard_temp_metrics->system_temp_ubb_fpga_3v3_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_UBB_FPGA_3V3_VR]; + baseboard_temp_metrics->system_temp_retimer_0_1_2_3_1v2_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_RETIMER_0_1_2_3_1V2_VR]; + baseboard_temp_metrics->system_temp_retimer_4_5_6_7_1v2_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_RETIMER_4_5_6_7_1V2_VR]; + baseboard_temp_metrics->system_temp_retimer_0_1_0v9_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_RETIMER_0_1_0V9_VR]; + baseboard_temp_metrics->system_temp_retimer_4_5_0v9_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_RETIMER_4_5_0V9_VR]; + baseboard_temp_metrics->system_temp_retimer_2_3_0v9_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_RETIMER_2_3_0V9_VR]; + baseboard_temp_metrics->system_temp_retimer_6_7_0v9_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_RETIMER_6_7_0V9_VR]; + baseboard_temp_metrics->system_temp_oam_0_1_2_3_3v3_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_OAM_0_1_2_3_3V3_VR]; + baseboard_temp_metrics->system_temp_oam_4_5_6_7_3v3_vr = + metrics->SystemTemperatures[SYSTEM_TEMP_OAM_4_5_6_7_3V3_VR]; + baseboard_temp_metrics->system_temp_ibc_hsc = + metrics->SystemTemperatures[SYSTEM_TEMP_IBC_HSC]; + baseboard_temp_metrics->system_temp_ibc = + metrics->SystemTemperatures[SYSTEM_TEMP_IBC]; +} + static void smu_v15_0_8_fill_gpuboard_temp_metrics( struct smu_v15_0_8_gpuboard_temp_metrics *gpuboard_temp_metrics, const SystemMetricsTable_t *metrics) @@ -1429,33 +1503,43 @@ static ssize_t smu_v15_0_8_get_temp_metrics(struct smu_context *smu, enum smu_temp_metric_type type, void *table) { + struct smu_v15_0_8_baseboard_temp_metrics *baseboard_temp_metrics; struct smu_v15_0_8_gpuboard_temp_metrics *gpuboard_temp_metrics; struct smu_table_context *smu_table = &smu->smu_table; struct smu_table *tables = smu_table->tables; - enum smu_driver_table_id table_id; SystemMetricsTable_t *metrics; struct smu_table *sys_table; - ssize_t size; int ret; - table_id = SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS; - gpuboard_temp_metrics = - (struct smu_v15_0_8_gpuboard_temp_metrics *) - smu_driver_table_ptr(smu, table_id); - size = sizeof(*gpuboard_temp_metrics); - ret = smu_v15_0_8_get_system_metrics_table(smu); if (ret) return ret; sys_table = &tables[SMU_TABLE_PMFW_SYSTEM_METRICS]; metrics = (SystemMetricsTable_t *)sys_table->cache.buffer; - smu_driver_table_update_cache_time(smu, table_id); - smu_v15_0_8_fill_gpuboard_temp_metrics(gpuboard_temp_metrics, - metrics); - memcpy(table, gpuboard_temp_metrics, size); - return size; + switch (type) { + case SMU_TEMP_METRIC_GPUBOARD: + gpuboard_temp_metrics = + (struct smu_v15_0_8_gpuboard_temp_metrics *) + smu_driver_table_ptr(smu, SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS); + smu_driver_table_update_cache_time(smu, SMU_DRIVER_TABLE_GPUBOARD_TEMP_METRICS); + smu_v15_0_8_fill_gpuboard_temp_metrics(gpuboard_temp_metrics, + metrics); + memcpy(table, gpuboard_temp_metrics, sizeof(*gpuboard_temp_metrics)); + return sizeof(*gpuboard_temp_metrics); + case SMU_TEMP_METRIC_BASEBOARD: + baseboard_temp_metrics = + (struct smu_v15_0_8_baseboard_temp_metrics *) + smu_driver_table_ptr(smu, SMU_DRIVER_TABLE_BASEBOARD_TEMP_METRICS); + smu_driver_table_update_cache_time(smu, SMU_DRIVER_TABLE_BASEBOARD_TEMP_METRICS); + smu_v15_0_8_fill_baseboard_temp_metrics(baseboard_temp_metrics, + metrics); + memcpy(table, baseboard_temp_metrics, sizeof(*baseboard_temp_metrics)); + return sizeof(*baseboard_temp_metrics); + default: + return -EINVAL; + } } static ssize_t smu_v15_0_8_get_gpu_metrics(struct smu_context *smu, void **table) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h index 0856d11d8e55..398ce4482174 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.h @@ -182,6 +182,7 @@ typedef struct { DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_gpu_metrics, SMU_15_0_8_METRICS_FIELDS); /* Maximum temperature sensor counts for system metrics */ +#define SMU_15_0_8_MAX_SYSTEM_TEMP_ENTRIES 32 #define SMU_15_0_8_MAX_NODE_TEMP_ENTRIES 12 #define SMU_15_0_8_MAX_VR_TEMP_ENTRIES 22 @@ -251,5 +252,62 @@ DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_gpu_metrics, SMU_15_0_8_METRICS_FIELDS); DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_gpuboard_temp_metrics, SMU_15_0_8_GPUBOARD_TEMP_METRICS_FIELDS); +/* SMUv 15.0.8 Baseboard temperature metrics - ID-based approach */ +#define SMU_15_0_8_BASEBOARD_TEMP_METRICS_FIELDS(SMU_SCALAR, SMU_ARRAY) \ + SMU_SCALAR(SMU_MATTR(ACCUMULATION_COUNTER), SMU_MUNIT(NONE), \ + SMU_MTYPE(U64), accumulation_counter); \ + SMU_SCALAR(SMU_MATTR(LABEL_VERSION), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), label_version); \ + SMU_SCALAR(SMU_MATTR(NODE_ID), SMU_MUNIT(NONE), \ + SMU_MTYPE(U16), node_id); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_FPGA), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_fpga); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_FRONT), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_front); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_BACK), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_back); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_OAM7), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_oam7); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_IBC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_ibc); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_UFPGA), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_ufpga); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_OAM1), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_oam1); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_OAM_0_1_HSC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_oam_0_1_hsc); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_OAM_2_3_HSC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_oam_2_3_hsc); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_OAM_4_5_HSC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_oam_4_5_hsc); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_OAM_6_7_HSC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_oam_6_7_hsc); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_FPGA_0V72_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_fpga_0v72_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_UBB_FPGA_3V3_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ubb_fpga_3v3_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_RETIMER_0_1_2_3_1V2_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_retimer_0_1_2_3_1v2_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_RETIMER_4_5_6_7_1V2_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_retimer_4_5_6_7_1v2_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_RETIMER_0_1_0V9_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_retimer_0_1_0v9_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_RETIMER_4_5_0V9_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_retimer_4_5_0v9_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_RETIMER_2_3_0V9_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_retimer_2_3_0v9_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_RETIMER_6_7_0V9_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_retimer_6_7_0v9_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_OAM_0_1_2_3_3V3_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_oam_0_1_2_3_3v3_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_OAM_4_5_6_7_3V3_VR), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_oam_4_5_6_7_3v3_vr); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_IBC_HSC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ibc_hsc); \ + SMU_SCALAR(SMU_MATTR(SYSTEM_TEMP_IBC), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(S16), system_temp_ibc); + +DECLARE_SMU_METRICS_CLASS(smu_v15_0_8_baseboard_temp_metrics, + SMU_15_0_8_BASEBOARD_TEMP_METRICS_FIELDS); #endif #endif From 6a609b800c9ed8d4bdb44362c8d04afa41d44d21 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Fri, 30 Jan 2026 21:20:58 +0800 Subject: [PATCH 75/99] drm/amd/pm: Add NPM support for smu_v15_0_8 Add node power management support for smu_v15_0_8 Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 49 +++++++++++++++++++ 1 file changed, 49 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index c850578f735c..13c2b0cb7ad1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -558,6 +558,46 @@ static int smu_v15_0_8_get_system_metrics_table(struct smu_context *smu) return 0; } +static int smu_v15_0_8_get_npm_data(struct smu_context *smu, + enum amd_pp_sensors sensor, + uint32_t *value) +{ + struct smu_table_context *smu_table = &smu->smu_table; + struct smu_table *tables = smu_table->tables; + SystemMetricsTable_t *metrics; + struct smu_table *sys_table; + int ret; + + if (sensor == AMDGPU_PP_SENSOR_MAXNODEPOWERLIMIT) { + /*TBD as of now put 0 */ + *value = 0; + return 0; + } + + ret = smu_v15_0_8_get_system_metrics_table(smu); + if (ret) + return ret; + + sys_table = &tables[SMU_TABLE_PMFW_SYSTEM_METRICS]; + metrics = (SystemMetricsTable_t *)sys_table->cache.buffer; + + switch (sensor) { + case AMDGPU_PP_SENSOR_NODEPOWERLIMIT: + *value = SMUQ10_ROUND(metrics->NodePowerLimit); + break; + case AMDGPU_PP_SENSOR_NODEPOWER: + *value = SMUQ10_ROUND(metrics->NodePower); + break; + case AMDGPU_PP_SENSOR_GPPTRESIDENCY: + *value = SMUQ10_ROUND(metrics->GlobalPPTResidencyAcc); + break; + default: + return -EINVAL; + } + + return 0; +} + static int smu_v15_0_8_read_sensor(struct smu_context *smu, enum amd_pp_sensors sensor, void *data, uint32_t *size) @@ -607,6 +647,15 @@ static int smu_v15_0_8_read_sensor(struct smu_context *smu, *(uint32_t *)data = dpm_context->board_volt; *size = 4; break; + case AMDGPU_PP_SENSOR_NODEPOWERLIMIT: + case AMDGPU_PP_SENSOR_NODEPOWER: + case AMDGPU_PP_SENSOR_GPPTRESIDENCY: + case AMDGPU_PP_SENSOR_MAXNODEPOWERLIMIT: + ret = smu_v15_0_8_get_npm_data(smu, sensor, (uint32_t *)data); + if (ret) + return ret; + *size = 4; + break; case AMDGPU_PP_SENSOR_GPU_AVG_POWER: default: ret = -EOPNOTSUPP; From df4929d76aa29f32c255f60bab4ebe862dc7c9a3 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sun, 2 Nov 2025 19:55:53 +0800 Subject: [PATCH 76/99] drm/amdgpu: Add smu v15_0_8 ip block Add smu v15_0_8 ip block Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 1 + drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 3 +++ 2 files changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 97fd66c578a9..c53cfd5894cd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -2341,6 +2341,7 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &smu_v14_0_ip_block); break; case IP_VERSION(15, 0, 0): + case IP_VERSION(15, 0, 8): amdgpu_device_ip_block_add(adev, &smu_v15_0_ip_block); break; default: diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index eee2fe0743b2..5c43e2ad16e0 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -803,6 +803,9 @@ static int smu_set_funcs(struct amdgpu_device *adev) case IP_VERSION(15, 0, 0): smu_v15_0_0_set_ppt_funcs(smu); break; + case IP_VERSION(15, 0, 8): + smu_v15_0_8_set_ppt_funcs(smu); + break; default: return -EINVAL; } From 7aaa09ab30c788038e3996c17c3fb5a7be53e6cd Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 7 Jan 2026 11:50:21 +0800 Subject: [PATCH 77/99] drm/amd/pm: Enable user specified gfx clock ranges Enable user specified gfx clock ranges for smu_15_0_8 Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 5c43e2ad16e0..8faf7de7aaa9 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -805,6 +805,7 @@ static int smu_set_funcs(struct amdgpu_device *adev) break; case IP_VERSION(15, 0, 8): smu_v15_0_8_set_ppt_funcs(smu); + smu->od_enabled = true; break; default: return -EINVAL; From d31ed58e2988d4a733242d6900920c9af39db8bb Mon Sep 17 00:00:00 2001 From: Hou Wenlong Date: Mon, 16 Mar 2026 11:46:29 +0800 Subject: [PATCH 78/99] drm/amd/display: Rename enum 'pixel_format' to 'dc_pixel_format' Rename the enum 'pixel_format' to 'dc_pixel_format' to avoid potential name conflicts with the pixel_format struct defined in include/video/pixel_format.h. Signed-off-by: Hou Wenlong Reviewed-by: Alex Hung Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc_resource.c | 4 ++-- drivers/gpu/drm/amd/display/dc/dc_hw_types.h | 2 +- drivers/gpu/drm/amd/display/dc/dc_spl_translate.c | 3 ++- drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp_dscl.c | 4 ++-- drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c | 4 ++-- drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h | 2 +- drivers/gpu/drm/amd/display/dc/inc/hw/transform.h | 2 +- 7 files changed, 11 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index 8271b12c1a66..727bcf08a84f 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -749,10 +749,10 @@ struct clock_source *resource_find_used_clk_src_for_sharing( return NULL; } -static enum pixel_format convert_pixel_format_to_dalsurface( +static enum dc_pixel_format convert_pixel_format_to_dalsurface( enum surface_pixel_format surface_pixel_format) { - enum pixel_format dal_pixel_format = PIXEL_FORMAT_UNKNOWN; + enum dc_pixel_format dal_pixel_format = PIXEL_FORMAT_UNKNOWN; switch (surface_pixel_format) { case SURFACE_PIXEL_FORMAT_GRPH_PALETA_256_COLORS: diff --git a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h index 595e082db281..c2ca08d26e37 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h @@ -218,7 +218,7 @@ enum surface_pixel_format { /* Pixel format */ -enum pixel_format { +enum dc_pixel_format { /*graph*/ PIXEL_FORMAT_UNINITIALIZED, PIXEL_FORMAT_INDEX8, diff --git a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c index ba7bf23f2b2f..52ed8deebf63 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c +++ b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c @@ -63,7 +63,8 @@ static void populate_inits_from_splinits(struct scl_inits *inits, inits->h_c = dc_fixpt_from_int_dy(spl_inits->h_filter_init_int_c, spl_inits->h_filter_init_frac_c >> 5, 0, 19); inits->v_c = dc_fixpt_from_int_dy(spl_inits->v_filter_init_int_c, spl_inits->v_filter_init_frac_c >> 5, 0, 19); } -static void populate_splformat_from_format(enum spl_pixel_format *spl_pixel_format, const enum pixel_format pixel_format) +static void populate_splformat_from_format(enum spl_pixel_format *spl_pixel_format, + const enum dc_pixel_format pixel_format) { if (pixel_format < PIXEL_FORMAT_INVALID) *spl_pixel_format = (enum spl_pixel_format)pixel_format; diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp_dscl.c index 808bca9fb804..0d2c9fcd3362 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp_dscl.c @@ -102,7 +102,7 @@ static int dpp1_dscl_get_pixel_depth_val(enum lb_pixel_depth depth) } } -static bool dpp1_dscl_is_video_format(enum pixel_format format) +static bool dpp1_dscl_is_video_format(enum dc_pixel_format format) { if (format >= PIXEL_FORMAT_VIDEO_BEGIN && format <= PIXEL_FORMAT_VIDEO_END) @@ -111,7 +111,7 @@ static bool dpp1_dscl_is_video_format(enum pixel_format format) return false; } -static bool dpp1_dscl_is_420_format(enum pixel_format format) +static bool dpp1_dscl_is_420_format(enum dc_pixel_format format) { if (format == PIXEL_FORMAT_420BPP8 || format == PIXEL_FORMAT_420BPP10) diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c index 8b6155f9122f..8b7e55e337d3 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c @@ -94,7 +94,7 @@ static int dpp401_dscl_get_pixel_depth_val(enum lb_pixel_depth depth) } } -static bool dpp401_dscl_is_video_format(enum pixel_format format) +static bool dpp401_dscl_is_video_format(enum dc_pixel_format format) { if (format >= PIXEL_FORMAT_VIDEO_BEGIN && format <= PIXEL_FORMAT_VIDEO_END) @@ -103,7 +103,7 @@ static bool dpp401_dscl_is_video_format(enum pixel_format format) return false; } -static bool dpp401_dscl_is_420_format(enum pixel_format format) +static bool dpp401_dscl_is_420_format(enum dc_pixel_format format) { if (format == PIXEL_FORMAT_420BPP8 || format == PIXEL_FORMAT_420BPP10) diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h index b152f6879495..51581c10fd6b 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h @@ -256,7 +256,7 @@ struct default_adjustment { enum dc_color_space out_color_space; enum dc_color_space in_color_space; enum dc_color_depth color_depth; - enum pixel_format surface_pixel_format; + enum dc_pixel_format surface_pixel_format; enum graphics_csc_adjust_type csc_adjust_type; bool force_hw_default; }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h b/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h index 5a1d9b708a9d..30990355985d 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h @@ -160,7 +160,7 @@ struct scaler_data { struct scaling_ratios ratios; struct scl_inits inits; struct sharpness_adj sharpness; - enum pixel_format format; + enum dc_pixel_format format; struct line_buffer_params lb_params; // Below struct holds the scaler values to program hw registers struct dscl_prog_data dscl_prog_data; From 896ffa9847c3d00895fa7f6da5b7e2c141189c9b Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Mon, 23 Feb 2026 16:28:22 +0530 Subject: [PATCH 79/99] drm/amdgpu: Use stack variable to fetch nps info Instead of a dynamic allocation, use stack variable and let the caller pass the maximum ranges that can be held in the buffer. Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 21 ++++++++++--------- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 10 ++++----- 3 files changed, 16 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index c53cfd5894cd..a20b73eab015 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -1984,11 +1984,10 @@ static int amdgpu_discovery_refresh_nps_info(struct amdgpu_device *adev, int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev, uint32_t *nps_type, - struct amdgpu_gmc_memrange **ranges, + struct amdgpu_gmc_memrange *ranges, int *range_cnt, bool refresh) { uint8_t *discovery_bin = adev->discovery.bin; - struct amdgpu_gmc_memrange *mem_ranges; struct table_info *info; union nps_info *nps_info; union nps_info nps_data; @@ -2026,20 +2025,22 @@ int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev, switch (le16_to_cpu(nps_info->v1.header.version_major)) { case 1: - mem_ranges = kvzalloc_objs(*mem_ranges, nps_info->v1.count); - if (!mem_ranges) - return -ENOMEM; *nps_type = nps_info->v1.nps_type; + if (*range_cnt < nps_info->v1.count) { + dev_dbg(adev->dev, + "not enough space for nps ranges: %d < %d\n", + *range_cnt, nps_info->v1.count); + return -ENOSPC; + } *range_cnt = nps_info->v1.count; for (i = 0; i < *range_cnt; i++) { - mem_ranges[i].base_address = + ranges[i].base_address = nps_info->v1.instance_info[i].base_address; - mem_ranges[i].limit_address = + ranges[i].limit_address = nps_info->v1.instance_info[i].limit_address; - mem_ranges[i].nid_mask = -1; - mem_ranges[i].flags = 0; + ranges[i].nid_mask = -1; + ranges[i].flags = 0; } - *ranges = mem_ranges; break; default: dev_err(adev->dev, "Unhandled NPS info table %d.%d\n", diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h index a7aeb47887a3..0ff1a7923eed 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h @@ -46,7 +46,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev); int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev, uint32_t *nps_type, - struct amdgpu_gmc_memrange **ranges, + struct amdgpu_gmc_memrange *ranges, int *range_cnt, bool refresh); void amdgpu_discovery_dump(struct amdgpu_device *adev, struct drm_printer *p); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index a0940db1cd36..860a4405f7dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -1374,18 +1374,18 @@ int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev, struct amdgpu_mem_partition_info *mem_ranges, uint8_t *exp_ranges) { - struct amdgpu_gmc_memrange *ranges; + struct amdgpu_gmc_memrange ranges[AMDGPU_MAX_MEM_RANGES]; int range_cnt, ret, i, j; uint32_t nps_type; bool refresh; if (!mem_ranges || !exp_ranges) return -EINVAL; - + range_cnt = AMDGPU_MAX_MEM_RANGES; refresh = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) && (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS); - ret = amdgpu_discovery_get_nps_info(adev, &nps_type, &ranges, - &range_cnt, refresh); + ret = amdgpu_discovery_get_nps_info(adev, &nps_type, ranges, &range_cnt, + refresh); if (ret) return ret; @@ -1446,8 +1446,6 @@ int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev, if (!*exp_ranges) *exp_ranges = range_cnt; err: - kvfree(ranges); - return ret; } From 470891606c5a97b1d0d937e0aa67a3bed9fcb056 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Thu, 19 Mar 2026 21:17:38 -0400 Subject: [PATCH 80/99] drm/amd/pm: disable OD_FAN_CURVE if temp or pwm range invalid for smu v13 Forcibly disable the OD_FAN_CURVE feature when temperature or PWM range is invalid, otherwise PMFW will reject this configuration on smu v13.0.x example: $ sudo cat /sys/bus/pci/devices//gpu_od/fan_ctrl/fan_curve OD_FAN_CURVE: 0: 0C 0% 1: 0C 0% 2: 0C 0% 3: 0C 0% 4: 0C 0% OD_RANGE: FAN_CURVE(hotspot temp): 0C 0C FAN_CURVE(fan speed): 0% 0% $ echo "0 50 40" | sudo tee fan_curve kernel log: [ 756.442527] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]! [ 777.345800] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]! Closes: https://github.com/ROCm/amdgpu/issues/208 Signed-off-by: Yang Wang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 33 ++++++++++++++++++- .../drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 33 ++++++++++++++++++- 2 files changed, 64 insertions(+), 2 deletions(-) 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 59720372fc17..b414a74d29fd 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 @@ -59,6 +59,10 @@ #define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) +static void smu_v13_0_0_get_od_setting_limits(struct smu_context *smu, + int od_feature_bit, + int32_t *min, int32_t *max); + static const struct smu_feature_bits smu_v13_0_0_dpm_features = { .bits = { SMU_FEATURE_BIT_INIT(FEATURE_DPM_GFXCLK_BIT), @@ -1043,8 +1047,35 @@ static bool smu_v13_0_0_is_od_feature_supported(struct smu_context *smu, PPTable_t *pptable = smu->smu_table.driver_pptable; const OverDriveLimits_t * const overdrive_upperlimits = &pptable->SkuTable.OverDriveLimitsBasicMax; + int32_t min_value, max_value; + bool feature_enabled; - return overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit); + switch (od_feature_bit) { + case PP_OD_FEATURE_FAN_CURVE_BIT: + feature_enabled = !!(overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit)); + if (feature_enabled) { + smu_v13_0_0_get_od_setting_limits(smu, PP_OD_FEATURE_FAN_CURVE_TEMP, + &min_value, &max_value); + if (!min_value && !max_value) { + feature_enabled = false; + goto out; + } + + smu_v13_0_0_get_od_setting_limits(smu, PP_OD_FEATURE_FAN_CURVE_PWM, + &min_value, &max_value); + if (!min_value && !max_value) { + feature_enabled = false; + goto out; + } + } + break; + default: + feature_enabled = !!(overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit)); + break; + } + +out: + return feature_enabled; } static void smu_v13_0_0_get_od_setting_limits(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 88e0d99b8ad2..fd0b6215364f 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 @@ -59,6 +59,10 @@ #define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) +static void smu_v13_0_7_get_od_setting_limits(struct smu_context *smu, + int od_feature_bit, + int32_t *min, int32_t *max); + static const struct smu_feature_bits smu_v13_0_7_dpm_features = { .bits = { SMU_FEATURE_BIT_INIT(FEATURE_DPM_GFXCLK_BIT), @@ -1053,8 +1057,35 @@ static bool smu_v13_0_7_is_od_feature_supported(struct smu_context *smu, PPTable_t *pptable = smu->smu_table.driver_pptable; const OverDriveLimits_t * const overdrive_upperlimits = &pptable->SkuTable.OverDriveLimitsBasicMax; + int32_t min_value, max_value; + bool feature_enabled; - return overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit); + switch (od_feature_bit) { + case PP_OD_FEATURE_FAN_CURVE_BIT: + feature_enabled = !!(overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit)); + if (feature_enabled) { + smu_v13_0_7_get_od_setting_limits(smu, PP_OD_FEATURE_FAN_CURVE_TEMP, + &min_value, &max_value); + if (!min_value && !max_value) { + feature_enabled = false; + goto out; + } + + smu_v13_0_7_get_od_setting_limits(smu, PP_OD_FEATURE_FAN_CURVE_PWM, + &min_value, &max_value); + if (!min_value && !max_value) { + feature_enabled = false; + goto out; + } + } + break; + default: + feature_enabled = !!(overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit)); + break; + } + +out: + return feature_enabled; } static void smu_v13_0_7_get_od_setting_limits(struct smu_context *smu, From 8b9e5259adc385b61a6590a13b82ae0ac2bd3482 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Mon, 23 Mar 2026 13:41:18 +0530 Subject: [PATCH 81/99] drm/amdgpu: Fix fence put before wait in amdgpu_amdkfd_submit_ib MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit amdgpu_amdkfd_submit_ib() submits a GPU job and gets a fence from amdgpu_ib_schedule(). This fence is used to wait for job completion. Currently, the code drops the fence reference using dma_fence_put() before calling dma_fence_wait(). If dma_fence_put() releases the last reference, the fence may be freed before dma_fence_wait() is called. This can lead to a use-after-free. Fix this by waiting on the fence first and releasing the reference only after dma_fence_wait() completes. Fixes the below: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c:697 amdgpu_amdkfd_submit_ib() warn: passing freed memory 'f' (line 696) Fixes: 9ae55f030dc5 ("drm/amdgpu: Follow up change to previous drm scheduler change.") Cc: Felix Kuehling Cc: Dan Carpenter Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 40c22438b1d2..4f27c75abedb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -692,9 +692,9 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev, goto err_ib_sched; } - /* Drop the initial kref_init count (see drm_sched_main as example) */ - dma_fence_put(f); ret = dma_fence_wait(f, false); + /* Drop the returned fence reference after the wait completes */ + dma_fence_put(f); err_ib_sched: amdgpu_job_free(job); From 3f797396d7f4eb9bb6eded184bbc6f033628a6f6 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 20 Mar 2026 12:33:48 -0400 Subject: [PATCH 82/99] drm/amd/display: check if ext_caps is valid in BL setup LVDS connectors don't have extended backlight caps so check if the pointer is valid before accessing it. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5012 Fixes: 1454642960b0 ("drm/amd: Re-introduce property to control adaptive backlight modulation") Cc: Mario Limonciello Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) 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 fa88eb548404..9e967007d677 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -5418,7 +5418,7 @@ static void setup_backlight_device(struct amdgpu_display_manager *dm, caps = &dm->backlight_caps[aconnector->bl_idx]; /* Only offer ABM property when non-OLED and user didn't turn off by module parameter */ - if (!caps->ext_caps->bits.oled && amdgpu_dm_abm_level < 0) + if (caps->ext_caps && !caps->ext_caps->bits.oled && amdgpu_dm_abm_level < 0) drm_object_attach_property(&aconnector->base.base, dm->adev->mode_info.abm_level_property, ABM_SYSFS_CONTROL); From 19d4149b22f57094bfc4b86b742381b3ca394ead Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Mon, 23 Mar 2026 14:28:57 +0530 Subject: [PATCH 83/99] drm/amdkfd: Fix NULL pointer check order in kfd_ioctl_create_process MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In kfd_ioctl_create_process(), the pointer 'p' is used before checking if it is NULL. The code accesses p->context_id before validating 'p'. This can lead to a possible NULL pointer dereference. Move the NULL check before using 'p' so that the pointer is validated before access. Fixes the below: drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_chardev.c:3177 kfd_ioctl_create_process() warn: variable dereferenced before check 'p' (see line 3174) Fixes: cc6b66d661fd ("amdkfd: introduce new ioctl AMDKFD_IOC_CREATE_PROCESS") Cc: Zhu Lingshan Cc: Felix Kuehling Cc: Christian König Cc: Alex Deucher Cc: Dan Carpenter Signed-off-by: Srinivasan Shanmugam Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 09dabb3b3297..462a32abf720 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -3170,11 +3170,11 @@ static int kfd_ioctl_create_process(struct file *filep, struct kfd_process *p, v struct kfd_process *process; int ret; - /* Each FD owns only one kfd_process */ - if (p->context_id != KFD_CONTEXT_ID_PRIMARY) + if (!filep->private_data || !p) return -EINVAL; - if (!filep->private_data || !p) + /* Each FD owns only one kfd_process */ + if (p->context_id != KFD_CONTEXT_ID_PRIMARY) return -EINVAL; mutex_lock(&kfd_processes_mutex); From ab4905d466b60f170d85e19ca2a5d2b159aeb780 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Thu, 19 Mar 2026 03:36:50 -0400 Subject: [PATCH 84/99] drm/amd/pm: disable OD_FAN_CURVE if temp or pwm range invalid for smu v14 Forcibly disable the OD_FAN_CURVE feature when temperature or PWM range is invalid, otherwise PMFW will reject this configuration on smu v14.0.2/14.0.3. example: $ sudo cat /sys/bus/pci/devices//gpu_od/fan_ctrl/fan_curve OD_FAN_CURVE: 0: 0C 0% 1: 0C 0% 2: 0C 0% 3: 0C 0% 4: 0C 0% OD_RANGE: FAN_CURVE(hotspot temp): 0C 0C FAN_CURVE(fan speed): 0% 0% $ echo "0 50 40" | sudo tee fan_curve kernel log: [ 969.761627] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]! [ 1010.897800] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]! Signed-off-by: Yang Wang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 33 ++++++++++++++++++- 1 file changed, 32 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c index 73762d9b5969..c3ebfac062a7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c @@ -56,6 +56,10 @@ #define to_amdgpu_device(x) (container_of(x, struct amdgpu_device, pm.smu_i2c)) +static void smu_v14_0_2_get_od_setting_limits(struct smu_context *smu, + int od_feature_bit, + int32_t *min, int32_t *max); + static const struct smu_feature_bits smu_v14_0_2_dpm_features = { .bits = { SMU_FEATURE_BIT_INIT(FEATURE_DPM_GFXCLK_BIT), SMU_FEATURE_BIT_INIT(FEATURE_DPM_UCLK_BIT), @@ -922,8 +926,35 @@ static bool smu_v14_0_2_is_od_feature_supported(struct smu_context *smu, PPTable_t *pptable = smu->smu_table.driver_pptable; const OverDriveLimits_t * const overdrive_upperlimits = &pptable->SkuTable.OverDriveLimitsBasicMax; + int32_t min_value, max_value; + bool feature_enabled; - return overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit); + switch (od_feature_bit) { + case PP_OD_FEATURE_FAN_CURVE_BIT: + feature_enabled = !!(overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit)); + if (feature_enabled) { + smu_v14_0_2_get_od_setting_limits(smu, PP_OD_FEATURE_FAN_CURVE_TEMP, + &min_value, &max_value); + if (!min_value && !max_value) { + feature_enabled = false; + goto out; + } + + smu_v14_0_2_get_od_setting_limits(smu, PP_OD_FEATURE_FAN_CURVE_PWM, + &min_value, &max_value); + if (!min_value && !max_value) { + feature_enabled = false; + goto out; + } + } + break; + default: + feature_enabled = !!(overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit)); + break; + } + +out: + return feature_enabled; } static void smu_v14_0_2_get_od_setting_limits(struct smu_context *smu, From bdb2b9e1e08f34a4897c9f2558b4c9941638a6e5 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Sun, 22 Mar 2026 21:48:39 -0400 Subject: [PATCH 85/99] drm/amd/pm: add dedicated dram addr msg for smu v15 Add dedicated SMU Dram MSG mapping to avoid conflicts in SMU IP v15 common code for upcoming ASICs. add new smu msg: - SMU_MSG_SetDriverDramAddr - SMU_MSG_SetToolsDramAddr Signed-off-by: Yang Wang Reviewed-by: Lijo Lazar Reviewed-by: Asad Kamal Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h | 2 ++ drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c | 4 ++-- drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c | 4 ++-- 3 files changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h index fc31cc5e97c8..636ff90923d9 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h @@ -42,8 +42,10 @@ __SMU_DUMMY_MAP(SetPptLimit), \ __SMU_DUMMY_MAP(SetDriverDramAddrHigh), \ __SMU_DUMMY_MAP(SetDriverDramAddrLow), \ + __SMU_DUMMY_MAP(SetDriverDramAddr), \ __SMU_DUMMY_MAP(SetToolsDramAddrHigh), \ __SMU_DUMMY_MAP(SetToolsDramAddrLow), \ + __SMU_DUMMY_MAP(SetToolsDramAddr), \ __SMU_DUMMY_MAP(TransferTableSmu2Dram), \ __SMU_DUMMY_MAP(TransferTableDram2Smu), \ __SMU_DUMMY_MAP(UseDefaultPPTable), \ diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c index eb3fa767c848..c3cb36813806 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0.c @@ -610,7 +610,7 @@ int smu_v15_0_set_driver_table_location(struct smu_context *smu) { struct smu_table *driver_table = &smu->smu_table.driver_table; struct smu_msg_args args = { - .msg = SMU_MSG_SetDriverDramAddrHigh, + .msg = SMU_MSG_SetDriverDramAddr, .num_args = 2, .num_out_args = 0, }; @@ -625,7 +625,7 @@ int smu_v15_0_set_tool_table_location(struct smu_context *smu) { struct smu_table *tool_table = &smu->smu_table.tables[SMU_TABLE_PMSTATUSLOG]; struct smu_msg_args args = { - .msg = SMU_MSG_SetToolsDramAddrHigh, + .msg = SMU_MSG_SetToolsDramAddr, .num_args = 2, .num_out_args = 0, }; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c index 13c2b0cb7ad1..cc2babc6a341 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu15/smu_v15_0_8_ppt.c @@ -87,8 +87,8 @@ static const struct cmn2asic_msg_mapping smu_v15_0_8_message_map[SMU_MSG_MAX_COU MSG_MAP(GetMetricsVersion, PPSMC_MSG_GetMetricsVersion, 1), MSG_MAP(GetMetricsTable, PPSMC_MSG_GetMetricsTable, 1), MSG_MAP(GetEnabledSmuFeatures, PPSMC_MSG_GetEnabledSmuFeatures, 1), - MSG_MAP(SetDriverDramAddrHigh, PPSMC_MSG_SetDriverDramAddr, 1), - MSG_MAP(SetToolsDramAddrHigh, PPSMC_MSG_SetToolsDramAddr, 0), + MSG_MAP(SetDriverDramAddr, PPSMC_MSG_SetDriverDramAddr, 1), + MSG_MAP(SetToolsDramAddr, PPSMC_MSG_SetToolsDramAddr, 0), MSG_MAP(SetSoftMaxByFreq, PPSMC_MSG_SetSoftMaxByFreq, 1), MSG_MAP(SetPptLimit, PPSMC_MSG_SetPptLimit, 0), MSG_MAP(GetPptLimit, PPSMC_MSG_GetPptLimit, 1), From 5c82adf957dcb187add5ee3209439c67b63df538 Mon Sep 17 00:00:00 2001 From: "Jesse.Zhang" Date: Mon, 23 Mar 2026 13:31:54 +0800 Subject: [PATCH 86/99] drm/amdgpu: use DISCOVERY_TMR_SIZE in ACPI TMR fallback amdgpu_acpi_get_tmr_info() returns the full TMR region size, not the IP discovery table size. Using tmr_size as discovery.size can lead to oversized allocations and probe failure. In the ACPI fallback path, keep discovery.size as DISCOVERY_TMR_SIZE and only use ACPI data for offset calculation. Fixes: 01bdc7e219c4 ("drm/amdgpu: New interface to get IP discovery binary v3") Reviewed-by: Lijo Lazar Suggested-by: Lijo Lazar Signed-off-by: Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index a20b73eab015..8ec5465c3349 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -324,7 +324,7 @@ static int amdgpu_discovery_get_tmr_info(struct amdgpu_device *adev, ret = amdgpu_acpi_get_tmr_info(adev, &tmr_offset, &tmr_size); if (ret) return ret; - adev->discovery.size = (u32)tmr_size; + adev->discovery.size = DISCOVERY_TMR_SIZE; adev->discovery.offset = tmr_offset + tmr_size - DISCOVERY_TMR_OFFSET; } } From 6728daa2592788ab040b6e0389f86b197d4d8f6c Mon Sep 17 00:00:00 2001 From: "Jesse.Zhang" Date: Thu, 19 Mar 2026 15:54:38 +0800 Subject: [PATCH 87/99] drm/amd/pm: Enable VCN reset for pgm=4 with appropriate FW version Extend the VCN reset capability to include pgm=4 variants when the firmware version meets the required threshold (>= 0x04557100). This follows the existing pattern for pgm=0 and pgm=7, ensuring that VCN reset is enabled only on configurations where it is supported by the firmware. Reviewed-by: Lijo Lazar Acked-by: Alex Deucher Signed-off-by: Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 1 + 1 file changed, 1 insertion(+) 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 312ed1475e6b..475541189782 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 @@ -461,6 +461,7 @@ static void smu_v13_0_6_init_caps(struct smu_context *smu) smu_v13_0_6_cap_set(smu, SMU_CAP(SDMA_RESET)); if ((pgm == 0 && fw_ver >= 0x00558200) || + (pgm == 4 && fw_ver >= 0x04557100) || (pgm == 7 && fw_ver >= 0x07551400)) smu_v13_0_6_cap_set(smu, SMU_CAP(VCN_RESET)); } From 16df395a55c635649dd3cf41d61bd685d3fd7a98 Mon Sep 17 00:00:00 2001 From: Jesse Zhang Date: Fri, 20 Mar 2026 16:16:16 +0800 Subject: [PATCH 88/99] drm/amdgpu/vcn4.0.3: gate per-queue reset by PSP SOS program version Add a PSP SOS firmware compatibility check before enabling VCN per-queue reset on vcn_v4_0_3. Per review, program check is sufficient: when PSP SOS program is 0x01, require fw version >= 0x0036015f; otherwise allow per-queue reset. Reviewed-by: Lijo Lazar Suggested-by: Lijo Lazar Signed-off-by: Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c index e78526a4e521..ff3013b97abd 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c @@ -134,6 +134,21 @@ static int vcn_v4_0_3_early_init(struct amdgpu_ip_block *ip_block) return 0; } +static bool vcn_v4_0_3_is_psp_fw_reset_supported(struct amdgpu_device *adev) +{ + uint32_t fw_ver = adev->psp.sos.fw_version; + uint32_t pgm = (fw_ver >> 8) & 0xFF; + + /* + * FWDEV-159155: PSP SOS FW must be >= 0x0036015f for program 0x01 + * before enabling VCN per-queue reset. + */ + if (pgm == 1) + return fw_ver >= 0x0036015f; + + return true; +} + static int vcn_v4_0_3_late_init(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; @@ -141,7 +156,9 @@ static int vcn_v4_0_3_late_init(struct amdgpu_ip_block *ip_block) adev->vcn.supported_reset = amdgpu_get_soft_full_reset_mask(&adev->vcn.inst[0].ring_enc[0]); - if (amdgpu_dpm_reset_vcn_is_supported(adev) && !amdgpu_sriov_vf(adev)) + if (amdgpu_dpm_reset_vcn_is_supported(adev) && + vcn_v4_0_3_is_psp_fw_reset_supported(adev) && + !amdgpu_sriov_vf(adev)) adev->vcn.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE; return 0; From a0f0b6d31a53a7607ed44f7623faafc628333258 Mon Sep 17 00:00:00 2001 From: Prike Liang Date: Thu, 19 Mar 2026 14:47:11 +0800 Subject: [PATCH 89/99] drm/amdgpu: fix syncobj leak for amdgpu_gem_va_ioctl() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It requires freeing the syncobj and chain alloction resource. Signed-off-by: Prike Liang Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 4ebb99e22ff2..5376035d32fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -107,6 +107,7 @@ amdgpu_gem_update_timeline_node(struct drm_file *filp, *chain = dma_fence_chain_alloc(); if (!*chain) { drm_syncobj_put(*syncobj); + *syncobj = NULL; return -ENOMEM; } @@ -980,6 +981,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, timeline_chain, fence, args->vm_timeline_point); + timeline_chain = NULL; } } dma_fence_put(fence); @@ -987,6 +989,9 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, } error: + dma_fence_chain_free(timeline_chain); + if (timeline_syncobj) + drm_syncobj_put(timeline_syncobj); drm_exec_fini(&exec); error_put_gobj: drm_gem_object_put(gobj); From 473527e70e73b6f29b54242b8de0dcb459829856 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Tue, 24 Mar 2026 13:18:28 +0530 Subject: [PATCH 90/99] drm/amdgpu/userq: dont use goto to jump when at end of function MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In function amdgpu_userq_restore_worker we dont need to use goto as we already in the end of function and it will exit naturally. Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index b75b44a3169b..f3fdfe535d02 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -1227,10 +1227,8 @@ static void amdgpu_userq_restore_worker(struct work_struct *work) } ret = amdgpu_userq_restore_all(uq_mgr); - if (ret) { + if (ret) drm_file_err(uq_mgr->file, "Failed to restore all queues\n"); - goto unlock; - } unlock: mutex_unlock(&uq_mgr->userq_mutex); From 392848cdd0aa16e262387c46102eeec821ae866b Mon Sep 17 00:00:00 2001 From: Adriano Vero Date: Tue, 17 Mar 2026 21:36:13 +0100 Subject: [PATCH 91/99] drm/amd/display: clean up typecasts and constants in dcn4_calcs Signed-off-by: Adriano Vero Signed-off-by: Alex Deucher --- .../src/dml2_core/dml2_core_dcn4_calcs.c | 32 +++++++++---------- 1 file changed, 16 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4_calcs.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4_calcs.c index b30d16474ceb..f6402e199354 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4_calcs.c @@ -202,7 +202,7 @@ static unsigned int dml_round_to_multiple(unsigned int num, unsigned int multipl return (num - remainder); } -static unsigned int dml_get_num_active_pipes(int unsigned num_planes, const struct core_display_cfg_support_info *cfg_support_info) +static unsigned int dml_get_num_active_pipes(unsigned int num_planes, const struct core_display_cfg_support_info *cfg_support_info) { unsigned int num_active_pipes = 0; @@ -546,9 +546,9 @@ static bool dml_is_vertical_rotation(enum dml2_rotation_angle Scan) return is_vert; } -static int unsigned dml_get_gfx_version(enum dml2_swizzle_mode sw_mode) +static unsigned int dml_get_gfx_version(enum dml2_swizzle_mode sw_mode) { - int unsigned version = 0; + unsigned int version = 0; if (sw_mode == dml2_sw_linear || sw_mode == dml2_sw_256b_2d || @@ -1761,7 +1761,7 @@ static unsigned int CalculateVMAndRowBytes(struct dml2_core_shared_calculate_vm_ *p->PixelPTEBytesPerRow = (unsigned int)((double)*p->dpte_row_width_ub / (double)*p->PixelPTEReqWidth * *p->PTERequestSize); // VBA_DELTA, VBA doesn't have programming value for pte row height linear. - *p->dpte_row_height_linear = (unsigned int)1 << (unsigned int)math_floor2(math_log((float)(p->PTEBufferSizeInRequests * PixelPTEReqWidth_linear / p->Pitch), 2.0), 1); + *p->dpte_row_height_linear = 1U << (unsigned int)math_floor2(math_log((float)(p->PTEBufferSizeInRequests * PixelPTEReqWidth_linear / p->Pitch), 2.0), 1); if (*p->dpte_row_height_linear > 128) *p->dpte_row_height_linear = 128; @@ -3377,7 +3377,7 @@ static void calculate_cursor_req_attributes( DML_LOG_VERBOSE("DML::%s: cursor_bytes_per_line = %d\n", __func__, *cursor_bytes_per_line); DML_LOG_VERBOSE("DML::%s: cursor_bytes_per_chunk = %d\n", __func__, *cursor_bytes_per_chunk); DML_LOG_VERBOSE("DML::%s: cursor_bytes = %d\n", __func__, *cursor_bytes); - DML_LOG_VERBOSE("DML::%s: cursor_pitch = %d\n", __func__, cursor_bpp == 2 ? 256 : (unsigned int)1 << (unsigned int)math_ceil2(math_log((float)cursor_width, 2), 1)); + DML_LOG_VERBOSE("DML::%s: cursor_pitch = %d\n", __func__, cursor_bpp == 2 ? 256 : 1U << (unsigned int)math_ceil2(math_log((float)cursor_width, 2), 1)); #endif } @@ -12205,15 +12205,15 @@ static void rq_dlg_get_wm_regs(const struct dml2_display_cfg *display_cfg, const { double refclk_freq_in_mhz = (display_cfg->overrides.hw.dlg_ref_clk_mhz > 0) ? (double)display_cfg->overrides.hw.dlg_ref_clk_mhz : mode_lib->soc.dchub_refclk_mhz; - wm_regs->fclk_pstate = (int unsigned)(mode_lib->mp.Watermark.FCLKChangeWatermark * refclk_freq_in_mhz); - wm_regs->sr_enter = (int unsigned)(mode_lib->mp.Watermark.StutterEnterPlusExitWatermark * refclk_freq_in_mhz); - wm_regs->sr_exit = (int unsigned)(mode_lib->mp.Watermark.StutterExitWatermark * refclk_freq_in_mhz); - wm_regs->sr_enter_z8 = (int unsigned)(mode_lib->mp.Watermark.Z8StutterEnterPlusExitWatermark * refclk_freq_in_mhz); - wm_regs->sr_exit_z8 = (int unsigned)(mode_lib->mp.Watermark.Z8StutterExitWatermark * refclk_freq_in_mhz); - wm_regs->temp_read_or_ppt = (int unsigned)(mode_lib->mp.Watermark.temp_read_or_ppt_watermark_us * refclk_freq_in_mhz); - wm_regs->uclk_pstate = (int unsigned)(mode_lib->mp.Watermark.DRAMClockChangeWatermark * refclk_freq_in_mhz); - wm_regs->urgent = (int unsigned)(mode_lib->mp.Watermark.UrgentWatermark * refclk_freq_in_mhz); - wm_regs->usr = (int unsigned)(mode_lib->mp.Watermark.USRRetrainingWatermark * refclk_freq_in_mhz); + wm_regs->fclk_pstate = (unsigned int)(mode_lib->mp.Watermark.FCLKChangeWatermark * refclk_freq_in_mhz); + wm_regs->sr_enter = (unsigned int)(mode_lib->mp.Watermark.StutterEnterPlusExitWatermark * refclk_freq_in_mhz); + wm_regs->sr_exit = (unsigned int)(mode_lib->mp.Watermark.StutterExitWatermark * refclk_freq_in_mhz); + wm_regs->sr_enter_z8 = (unsigned int)(mode_lib->mp.Watermark.Z8StutterEnterPlusExitWatermark * refclk_freq_in_mhz); + wm_regs->sr_exit_z8 = (unsigned int)(mode_lib->mp.Watermark.Z8StutterExitWatermark * refclk_freq_in_mhz); + wm_regs->temp_read_or_ppt = (unsigned int)(mode_lib->mp.Watermark.temp_read_or_ppt_watermark_us * refclk_freq_in_mhz); + wm_regs->uclk_pstate = (unsigned int)(mode_lib->mp.Watermark.DRAMClockChangeWatermark * refclk_freq_in_mhz); + wm_regs->urgent = (unsigned int)(mode_lib->mp.Watermark.UrgentWatermark * refclk_freq_in_mhz); + wm_regs->usr = (unsigned int)(mode_lib->mp.Watermark.USRRetrainingWatermark * refclk_freq_in_mhz); wm_regs->refcyc_per_trip_to_mem = (unsigned int)(mode_lib->mp.UrgentLatency * refclk_freq_in_mhz); wm_regs->refcyc_per_meta_trip_to_mem = (unsigned int)(mode_lib->mp.MetaTripToMemory * refclk_freq_in_mhz); wm_regs->frac_urg_bw_flip = (unsigned int)(mode_lib->mp.FractionOfUrgentBandwidthImmediateFlip * 1000); @@ -12692,7 +12692,7 @@ static void rq_dlg_get_dlg_reg( disp_dlg_regs->refcyc_per_vm_req_flip = (unsigned int)(math_pow(2, 23) - 1); - DML_ASSERT(disp_dlg_regs->dst_y_after_scaler < (unsigned int)8); + DML_ASSERT(disp_dlg_regs->dst_y_after_scaler < 8U); DML_ASSERT(disp_dlg_regs->refcyc_x_after_scaler < (unsigned int)math_pow(2, 13)); if (disp_dlg_regs->dst_y_per_pte_row_nom_l >= (unsigned int)math_pow(2, 17)) { @@ -13248,7 +13248,7 @@ void dml2_core_calcs_get_informative(const struct dml2_core_internal_display_mod out->informative.misc.cstate_max_cap_mode = dml_get_cstate_max_cap_mode(mode_lib); - out->min_clocks.dcn4x.dpprefclk_khz = (int unsigned)dml_get_global_dppclk_khz(mode_lib); + out->min_clocks.dcn4x.dpprefclk_khz = (unsigned int)dml_get_global_dppclk_khz(mode_lib); out->informative.qos.max_active_fclk_change_latency_supported = dml_get_fclk_change_latency(mode_lib); From 94c7e9669be1a29b8e44fa6ca3d2d8943b60a91e Mon Sep 17 00:00:00 2001 From: Kexin Sun Date: Sat, 21 Mar 2026 18:57:17 +0800 Subject: [PATCH 92/99] drm/amd/display: update outdated comments for renamed vblank_control_worker() The function vblank_control_worker() was renamed to amdgpu_dm_crtc_vblank_control_worker() by commit 6ce4f9ee25ff ("drm/amd/display: Add prefix to amdgpu crtc functions"). Update the two stale references in amdgpu_dm.c. Assisted-by: unnamed:deepseek-v3.2 coccinelle Signed-off-by: Kexin Sun Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 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 9e967007d677..1e304c9d2030 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -9899,7 +9899,8 @@ static void amdgpu_dm_enable_self_refresh(struct amdgpu_crtc *acrtc_attach, * a vblank event disable request to enable PSR/RP. PSR SU/RP * can be enabled immediately once OS demonstrates an * adequate number of fast atomic commits to notify KMD - * of update events. See `vblank_control_worker()`. + * of update events. + * See `amdgpu_dm_crtc_vblank_control_worker()`. */ if (!vrr_active && acrtc_attach->dm_irq_params.allow_sr_entry && @@ -10067,8 +10068,9 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, /* * If the dirty regions changed, PSR-SU need to be disabled temporarily * and enabled it again after dirty regions are stable to avoid video glitch. - * PSR-SU will be enabled in vblank_control_worker() if user pause the video - * during the PSR-SU was disabled. + * PSR-SU will be enabled in + * amdgpu_dm_crtc_vblank_control_worker() if user + * pause the video during the PSR-SU was disabled. */ if (acrtc_state->stream->link->psr_settings.psr_version >= DC_PSR_VERSION_SU_1 && acrtc_attach->dm_irq_params.allow_sr_entry && From 074fe395fb13247b057f60004c7ebcca9f38ef46 Mon Sep 17 00:00:00 2001 From: Donet Tom Date: Mon, 23 Mar 2026 09:58:36 +0530 Subject: [PATCH 93/99] drm/amdgpu: Handle GPU page faults correctly on non-4K page systems MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit During a GPU page fault, the driver restores the SVM range and then maps it into the GPU page tables. The current implementation passes a GPU-page-size (4K-based) PFN to svm_range_restore_pages() to restore the range. SVM ranges are tracked using system-page-size PFNs. On systems where the system page size is larger than 4K, using GPU-page-size PFNs to restore the range causes two problems: Range lookup fails: Because the restore function receives PFNs in GPU (4K) units, the SVM range lookup does not find the existing range. This will result in a duplicate SVM range being created. VMA lookup failure: The restore function also tries to locate the VMA for the faulting address. It converts the GPU-page-size PFN into an address using the system page size, which results in an incorrect address on non-4K page-size systems. As a result, the VMA lookup fails with the message: "address 0xxxx VMA is removed". This patch passes the system-page-size PFN to svm_range_restore_pages() so that the SVM range is restored correctly on non-4K page systems. Acked-by: Christian König Signed-off-by: Donet Tom Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index ce5794d1f518..73abac6be5b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2992,14 +2992,14 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid, if (!root) return false; - addr /= AMDGPU_GPU_PAGE_SIZE; - if (is_compute_context && !svm_range_restore_pages(adev, pasid, vmid, - node_id, addr, ts, write_fault)) { + node_id, addr >> PAGE_SHIFT, ts, write_fault)) { amdgpu_bo_unref(&root); return true; } + addr /= AMDGPU_GPU_PAGE_SIZE; + r = amdgpu_bo_reserve(root, true); if (r) goto error_unref; From 6d1bb522924a31cf492a0c44f41a8178eba40c90 Mon Sep 17 00:00:00 2001 From: Chen Ni Date: Fri, 20 Mar 2026 16:32:47 +0800 Subject: [PATCH 94/99] drm/amdgpu/userq: convert comma to semicolon Using a ',' in place of a ';' can have unintended side effects. Although that is not the case here, it seems best to use ';' unless ',' is intended. Found by inspection. No functional change intended. Compile tested only. Signed-off-by: Chen Ni Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c index f77fc210cb93..fe6d83e859a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c @@ -967,13 +967,13 @@ int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data, } num_read_bo_handles = wait_info->num_bo_read_handles; - ptr = u64_to_user_ptr(wait_info->bo_read_handles), + ptr = u64_to_user_ptr(wait_info->bo_read_handles); r = drm_gem_objects_lookup(filp, ptr, num_read_bo_handles, &gobj_read); if (r) goto free_timeline_points; num_write_bo_handles = wait_info->num_bo_write_handles; - ptr = u64_to_user_ptr(wait_info->bo_write_handles), + ptr = u64_to_user_ptr(wait_info->bo_write_handles); r = drm_gem_objects_lookup(filp, ptr, num_write_bo_handles, &gobj_write); if (r) From ab2fcdd7021908d5b1537f97be5a1994339fd6da Mon Sep 17 00:00:00 2001 From: Kexin Sun Date: Sat, 21 Mar 2026 18:57:28 +0800 Subject: [PATCH 95/99] drm/amdgpu: update outdated comment for renamed amdgpu_fence_driver_init() The function amdgpu_fence_driver_init() was renamed to amdgpu_fence_driver_sw_init() by commit 067f44c8b459 ("drm/amdgpu: avoid over-handle of fence driver fini in s3 test (v2)"). Update the stale reference in the amdgpu_fence_driver_init_ring() kdoc. Assisted-by: unnamed:deepseek-v3.2 coccinelle Signed-off-by: Kexin Sun Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index d209591e3710..8048a4c04b47 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -438,7 +438,7 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, * @ring: ring to init the fence driver on * * Init the fence driver for the requested ring (all asics). - * Helper function for amdgpu_fence_driver_init(). + * Helper function for amdgpu_fence_driver_sw_init(). */ int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring) { From 3b5620f7ee688177fcf65cf61588c5435bce1872 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 26 Feb 2026 17:12:08 -0500 Subject: [PATCH 96/99] drm/amd/display: Fix DCE LVDS handling LVDS does not use an HPD pin so it may be invalid. Handle this case correctly in link encoder creation. Fixes: 7c8fb3b8e9ba ("drm/amd/display: Add hpd_source index check for DCE60/80/100/110/112/120 link encoders") Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5012 Cc: Srinivasan Shanmugam Cc: Roman Li Reviewed-by: Roman Li Reviewed-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher --- .../display/dc/resource/dce100/dce100_resource.c | 6 ++---- .../display/dc/resource/dce110/dce110_resource.c | 5 +++-- .../display/dc/resource/dce112/dce112_resource.c | 5 +++-- .../display/dc/resource/dce120/dce120_resource.c | 5 +++-- .../amd/display/dc/resource/dce60/dce60_resource.c | 14 ++++++-------- .../amd/display/dc/resource/dce80/dce80_resource.c | 6 ++---- 6 files changed, 19 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/resource/dce100/dce100_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dce100/dce100_resource.c index 92c123aca0c9..fdcf8db6be50 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dce100/dce100_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dce100/dce100_resource.c @@ -650,9 +650,6 @@ static struct link_encoder *dce100_link_encoder_create( return &enc110->base; } - if (enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs)) - return NULL; - link_regs_id = map_transmitter_id_to_phy_instance(enc_init_data->transmitter); @@ -661,7 +658,8 @@ static struct link_encoder *dce100_link_encoder_create( &link_enc_feature, &link_enc_regs[link_regs_id], &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ? + NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]); return &enc110->base; } diff --git a/drivers/gpu/drm/amd/display/dc/resource/dce110/dce110_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dce110/dce110_resource.c index 95852d277c22..ab71f645c90e 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dce110/dce110_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dce110/dce110_resource.c @@ -671,7 +671,7 @@ static struct link_encoder *dce110_link_encoder_create( kzalloc_obj(struct dce110_link_encoder); int link_regs_id; - if (!enc110 || enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs)) + if (!enc110) return NULL; link_regs_id = @@ -682,7 +682,8 @@ static struct link_encoder *dce110_link_encoder_create( &link_enc_feature, &link_enc_regs[link_regs_id], &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ? + NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]); return &enc110->base; } diff --git a/drivers/gpu/drm/amd/display/dc/resource/dce112/dce112_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dce112/dce112_resource.c index 58c6a00397cf..b7051bfd4326 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dce112/dce112_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dce112/dce112_resource.c @@ -632,7 +632,7 @@ static struct link_encoder *dce112_link_encoder_create( kzalloc_obj(struct dce110_link_encoder); int link_regs_id; - if (!enc110 || enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs)) + if (!enc110) return NULL; link_regs_id = @@ -643,7 +643,8 @@ static struct link_encoder *dce112_link_encoder_create( &link_enc_feature, &link_enc_regs[link_regs_id], &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ? + NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]); return &enc110->base; } diff --git a/drivers/gpu/drm/amd/display/dc/resource/dce120/dce120_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dce120/dce120_resource.c index 71d76b021375..7ee70f7b3aa7 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dce120/dce120_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dce120/dce120_resource.c @@ -716,7 +716,7 @@ static struct link_encoder *dce120_link_encoder_create( kzalloc_obj(struct dce110_link_encoder); int link_regs_id; - if (!enc110 || enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs)) + if (!enc110) return NULL; link_regs_id = @@ -727,7 +727,8 @@ static struct link_encoder *dce120_link_encoder_create( &link_enc_feature, &link_enc_regs[link_regs_id], &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ? + NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]); return &enc110->base; } diff --git a/drivers/gpu/drm/amd/display/dc/resource/dce60/dce60_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dce60/dce60_resource.c index a57d68427812..6a25dcfcdf17 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dce60/dce60_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dce60/dce60_resource.c @@ -746,18 +746,16 @@ static struct link_encoder *dce60_link_encoder_create( return &enc110->base; } - if (enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs)) - return NULL; - link_regs_id = map_transmitter_id_to_phy_instance(enc_init_data->transmitter); dce60_link_encoder_construct(enc110, - enc_init_data, - &link_enc_feature, - &link_enc_regs[link_regs_id], - &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + enc_init_data, + &link_enc_feature, + &link_enc_regs[link_regs_id], + &link_enc_aux_regs[enc_init_data->channel - 1], + enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ? + NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]); return &enc110->base; } diff --git a/drivers/gpu/drm/amd/display/dc/resource/dce80/dce80_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dce80/dce80_resource.c index d66d8ac6d897..89927727a0d9 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dce80/dce80_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dce80/dce80_resource.c @@ -752,9 +752,6 @@ static struct link_encoder *dce80_link_encoder_create( return &enc110->base; } - if (enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs)) - return NULL; - link_regs_id = map_transmitter_id_to_phy_instance(enc_init_data->transmitter); @@ -763,7 +760,8 @@ static struct link_encoder *dce80_link_encoder_create( &link_enc_feature, &link_enc_regs[link_regs_id], &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + enc_init_data->hpd_source >= ARRAY_SIZE(link_enc_hpd_regs) ? + NULL : &link_enc_hpd_regs[enc_init_data->hpd_source]); return &enc110->base; } From 46f4458c236956ad4d185cd46d70d60dfd996f3d Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Mon, 23 Mar 2026 22:43:48 +0000 Subject: [PATCH 97/99] drm/amdgpu/mes12_1: emove extra ; from declaration statement MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is a declaration statement that has a ;; at the end, remove the extraneous ; Reviewed-by: Christian König Signed-off-by: Colin Ian King Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/mes_v12_1.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c index 5dcc2c32644a..0e9089544769 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c @@ -2227,7 +2227,7 @@ static int mes_v12_1_self_test(struct amdgpu_device *adev, int xcc_id) struct amdgpu_bo *meta_bo = NULL, *ctx_bo = NULL; void *meta_ptr = NULL, *ctx_ptr = NULL; u64 meta_gpu_addr, ctx_gpu_addr; - int size, i, r, pasid;; + int size, i, r, pasid; pasid = amdgpu_pasid_alloc(16); if (pasid < 0) From a3057aa926ad024e79c6652d891f311d6d0faa5c Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Tue, 24 Mar 2026 20:46:34 +0530 Subject: [PATCH 98/99] drm/amdgpu/userq: schedule_delayed_work should be after fence signalled MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Reorganise the amdgpu_eviction_fence_suspend_worker code so schedule_delayed_work is the last thing we do after amdgpu_userq_evict is complete and the eviction fence is signalled. Suggested-by: Christian König Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c | 6 +++++- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 4 +--- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h | 3 +-- 3 files changed, 7 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c index 9ca21a32bc98..4c5e38dea4c2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c @@ -76,7 +76,7 @@ amdgpu_eviction_fence_suspend_worker(struct work_struct *work) cookie = dma_fence_begin_signalling(); ev_fence = amdgpu_evf_mgr_get_fence(evf_mgr); - amdgpu_userq_evict(uq_mgr, !evf_mgr->shutdown); + amdgpu_userq_evict(uq_mgr); /* * Signaling the eviction fence must be done while holding the @@ -86,6 +86,10 @@ amdgpu_eviction_fence_suspend_worker(struct work_struct *work) dma_fence_signal(ev_fence); dma_fence_end_signalling(cookie); dma_fence_put(ev_fence); + + if (!evf_mgr->shutdown) + schedule_delayed_work(&uq_mgr->resume_work, 0); + mutex_unlock(&uq_mgr->userq_mutex); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index f3fdfe535d02..7f64b783954a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -1298,7 +1298,7 @@ amdgpu_userq_wait_for_signal(struct amdgpu_userq_mgr *uq_mgr) } void -amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr, bool schedule_resume) +amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr) { struct amdgpu_device *adev = uq_mgr->adev; int ret; @@ -1312,8 +1312,6 @@ amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr, bool schedule_resume) if (ret) dev_err(adev->dev, "Failed to evict userqueue\n"); - if (schedule_resume) - schedule_delayed_work(&uq_mgr->resume_work, 0); } int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *file_priv, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h index f0abc16d02cc..a4d44abf24fa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h @@ -133,8 +133,7 @@ int amdgpu_userq_create_object(struct amdgpu_userq_mgr *uq_mgr, void amdgpu_userq_destroy_object(struct amdgpu_userq_mgr *uq_mgr, struct amdgpu_userq_obj *userq_obj); -void amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr, - bool schedule_resume); +void amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr); void amdgpu_userq_ensure_ev_fence(struct amdgpu_userq_mgr *userq_mgr, struct amdgpu_eviction_fence_mgr *evf_mgr); From 68178644c35fca972ed970dc84933281b4913bff Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 20 Mar 2026 12:56:20 -0400 Subject: [PATCH 99/99] drm/amd/display: add a no_hpd link_encoder_funcs variant For link encoders without HPD (analog or LVDS), add a link_encoder_funcs structure with no hpd enable callbacks. The enable and disable hpd callbacks are currently not used outside of a special case in debugfs which checks if the hpd is valid before using it, but this will protect us if they ever are. Reviewed-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher --- .../drm/amd/display/dc/dce/dce_link_encoder.c | 64 ++++++++++++++++++- 1 file changed, 62 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c index 2ba3d3a3aac5..a368802ba51d 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c @@ -122,6 +122,33 @@ static const struct link_encoder_funcs dce110_lnk_enc_funcs = { .program_hpd_filter = dce110_program_hpd_filter, }; +static const struct link_encoder_funcs dce110_lnk_enc_funcs_no_hpd = { + .validate_output_with_stream = + dce110_link_encoder_validate_output_with_stream, + .hw_init = dce110_link_encoder_hw_init, + .setup = dce110_link_encoder_setup, + .enable_tmds_output = dce110_link_encoder_enable_tmds_output, + .enable_dp_output = dce110_link_encoder_enable_dp_output, + .enable_dp_mst_output = dce110_link_encoder_enable_dp_mst_output, + .enable_lvds_output = dce110_link_encoder_enable_lvds_output, + .enable_analog_output = dce110_link_encoder_enable_analog_output, + .disable_output = dce110_link_encoder_disable_output, + .dp_set_lane_settings = dce110_link_encoder_dp_set_lane_settings, + .dp_set_phy_pattern = dce110_link_encoder_dp_set_phy_pattern, + .update_mst_stream_allocation_table = + dce110_link_encoder_update_mst_stream_allocation_table, + .psr_program_dp_dphy_fast_training = + dce110_psr_program_dp_dphy_fast_training, + .psr_program_secondary_packet = dce110_psr_program_secondary_packet, + .connect_dig_be_to_fe = dce110_link_encoder_connect_dig_be_to_fe, + .is_dig_enabled = dce110_is_dig_enabled, + .destroy = dce110_link_encoder_destroy, + .get_max_link_cap = dce110_link_encoder_get_max_link_cap, + .get_dig_frontend = dce110_get_dig_frontend, + .get_hpd_state = dce110_get_hpd_state, + .program_hpd_filter = dce110_program_hpd_filter, +}; + static enum bp_result link_transmitter_control( struct dce110_link_encoder *enc110, struct bp_transmitter_control *cntl) @@ -865,7 +892,10 @@ void dce110_link_encoder_construct( const struct dc_vbios_funcs *bp_funcs = init_data->ctx->dc_bios->funcs; enum bp_result result = BP_RESULT_OK; - enc110->base.funcs = &dce110_lnk_enc_funcs; + if (hpd_regs) + enc110->base.funcs = &dce110_lnk_enc_funcs; + else + enc110->base.funcs = &dce110_lnk_enc_funcs_no_hpd; enc110->base.ctx = init_data->ctx; enc110->base.id = init_data->encoder; enc110->base.analog_id = init_data->analog_encoder; @@ -1855,6 +1885,33 @@ static const struct link_encoder_funcs dce60_lnk_enc_funcs = { .program_hpd_filter = dce110_program_hpd_filter, }; +static const struct link_encoder_funcs dce60_lnk_enc_funcs_no_hpd = { + .validate_output_with_stream = + dce110_link_encoder_validate_output_with_stream, + .hw_init = dce110_link_encoder_hw_init, + .setup = dce110_link_encoder_setup, + .enable_tmds_output = dce110_link_encoder_enable_tmds_output, + .enable_dp_output = dce60_link_encoder_enable_dp_output, + .enable_dp_mst_output = dce60_link_encoder_enable_dp_mst_output, + .enable_lvds_output = dce110_link_encoder_enable_lvds_output, + .enable_analog_output = dce110_link_encoder_enable_analog_output, + .disable_output = dce110_link_encoder_disable_output, + .dp_set_lane_settings = dce110_link_encoder_dp_set_lane_settings, + .dp_set_phy_pattern = dce60_link_encoder_dp_set_phy_pattern, + .update_mst_stream_allocation_table = + dce110_link_encoder_update_mst_stream_allocation_table, + .psr_program_dp_dphy_fast_training = + dce110_psr_program_dp_dphy_fast_training, + .psr_program_secondary_packet = dce110_psr_program_secondary_packet, + .connect_dig_be_to_fe = dce110_link_encoder_connect_dig_be_to_fe, + .is_dig_enabled = dce110_is_dig_enabled, + .destroy = dce110_link_encoder_destroy, + .get_max_link_cap = dce110_link_encoder_get_max_link_cap, + .get_dig_frontend = dce110_get_dig_frontend, + .get_hpd_state = dce110_get_hpd_state, + .program_hpd_filter = dce110_program_hpd_filter, +}; + void dce60_link_encoder_construct( struct dce110_link_encoder *enc110, const struct encoder_init_data *init_data, @@ -1867,7 +1924,10 @@ void dce60_link_encoder_construct( const struct dc_vbios_funcs *bp_funcs = init_data->ctx->dc_bios->funcs; enum bp_result result = BP_RESULT_OK; - enc110->base.funcs = &dce60_lnk_enc_funcs; + if (hpd_regs) + enc110->base.funcs = &dce60_lnk_enc_funcs; + else + enc110->base.funcs = &dce60_lnk_enc_funcs_no_hpd; enc110->base.ctx = init_data->ctx; enc110->base.id = init_data->encoder; enc110->base.analog_id = init_data->analog_encoder;