drm/amd/pm: use debug port for mode1 reset request on smu 13&14
Subsystems (3)
Kenneth Feng
maintainerPATCH 1/2 drm/amd/pm: use debug port for mode1 reset request on smu 13&14
use debug port for mode1 reset request so fw can handle mode1 reset even when it is stuck.
--- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 14 ++----------- .../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 6 ++++++ .../drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 21 ++++++++++++++++++- .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 6 +++++- 4 files changed, 33 insertions(+), 14 deletions(-)
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +2 -12
@@ -2617,21 +2617,11 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu,static bool smu_v13_0_0_is_mode1_reset_supported (struct smu_context *smu) {struct amdgpu_device *adev = smu->adev; -u32 smu_version; -int ret;/* SRIOV does not support SMU mode1 reset */ if (amdgpu_sriov_vf (adev))return false; -/* PMFW support is available since 78.41 */ - ret =smu_cmn_get_smc_version (smu,NULL , &smu_version); -if (ret) -return false; - -if (smu_version <0x004e2900 ) -return false; -return true; }
@@ -2830,8 +2820,8 @@ static int smu_v13_0_0_mode1_reset(struct smu_context *smu)/* SMU 13_0_0 PMFW supports RAS fatal error reset from 78.77 */ smu_v13_0_0_set_mode1_reset_param (smu,0x004e4d00 , ¶m); - ret =smu_cmn_send_smc_msg_with_param (smu, - SMU_MSG_Mode1Reset, param,NULL ); + ret =smu_cmn_send_debug_smc_msg_with_param (smu, + DEBUGSMC_MSG_Mode1Reset, param);break ;case IP_VERSION (13 ,0 ,10 ):
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +6 -0
@@ -3055,6 +3055,12 @@ static int smu_v13_0_6_link_reset(struct smu_context *smu)static bool smu_v13_0_6_is_mode1_reset_supported (struct smu_context *smu) { +struct amdgpu_device *adev = smu->adev; + +/* SRIOV does not support SMU mode1 reset */ +if (amdgpu_sriov_vf (adev)) +return false; +return true; }
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c +20 -1
@@ -72,6 +72,8 @@#define MP0_MP1_DATA_REGION_SIZE_COMBOPPTABLE 0x4000 +#define DEBUGSMC_MSG_Mode1Reset 2 +#define PP_OD_FEATURE_GFXCLK_FMIN 0#define PP_OD_FEATURE_GFXCLK_FMAX 1#define PP_OD_FEATURE_UCLK_FMIN 2
@@ -2735,6 +2737,23 @@ static int smu_v13_0_7_update_pcie_parameters(struct smu_context *smu,return ret; } +static int smu_v13_0_7_mode1_reset (struct smu_context *smu) +{ +int ret; + + ret =smu_cmn_send_debug_smc_msg (smu, DEBUGSMC_MSG_Mode1Reset); + +if (!ret) { +/* disable mmio access while doing mode 1 reset*/ + smu->adev->no_hw_access = true; +/* ensure no_hw_access is globally visible before any MMIO */ +smp_mb (); +msleep (SMU13_MODE1_RESET_WAIT_TIME_IN_MS ); + } + +return ret; +} +static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .get_allowed_feature_mask = smu_v13_0_7_get_allowed_feature_mask, .set_default_dpm_table = smu_v13_0_7_set_default_dpm_table,
@@ -2796,7 +2815,7 @@ static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .baco_enter = smu_v13_0_baco_enter, .baco_exit = smu_v13_0_baco_exit, .mode1_reset_is_support = smu_v13_0_7_is_mode1_reset_supported, - .mode1_reset = smu_v13_0_mode1_reset, + .mode1_reset = smu_v13_0_7_mode1_reset, .set_mp1_state = smu_v13_0_7_set_mp1_state, .set_df_cstate = smu_v13_0_7_set_df_cstate, .gpo_control = smu_v13_0_gpo_control,
drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +5 -1
@@ -1899,7 +1899,11 @@ static int smu_v14_0_2_baco_exit(struct smu_context *smu)static bool smu_v14_0_2_is_mode1_reset_supported (struct smu_context *smu) { -// TODO +struct amdgpu_device *adev = smu->adev; + +/* SRIOV does not support SMU mode1 reset */ +if (amdgpu_sriov_vf (adev)) +return false;return true; }
Signature · 1 line
2.34.1
Kenneth Feng
maintainerPATCH 2/2 drm/amd/pm: send unload command to smu during modprobe -r amdgpu
Send unload command to smu during modprobe -r amdgpu for smu 13/14.
1. This can fix the high voltage/temperatue issue after driver is unloaded.
2. Reloading driver could fail but with the debug port based mode1 reset during driver is reloaded, it is good and safe.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 7 +++---- drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h | 1 - drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 8 +------- 4 files changed, 4 insertions(+), 15 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -3
@@ -4657,9 +4657,6 @@ int amdgpu_device_init(struct amdgpu_device *adev,dev_info (adev->dev,"Pending hive reset.\n" );amdgpu_set_init_level (adev,AMDGPU_INIT_LEVEL_MINIMAL_XGMI ); - }else if (amdgpu_ip_version (adev,MP1_HWIP ,0 ) ==IP_VERSION (13 ,0 ,10 ) && - !amdgpu_device_has_display_hardware (adev)) { - r =psp_gpu_reset (adev); }else { tmp = amdgpu_reset_method;/* It should do a default reset when loading or reloading the driver,
drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +3 -4
@@ -820,7 +820,7 @@ static int smu_early_init(struct amdgpu_ip_block *ip_block) smu->adev = adev; smu->pm_enabled = !!amdgpu_dpm; smu->is_apu = false; - smu->smu_baco.state =SMU_BACO_STATE_NONE ; + smu->smu_baco.state =SMU_BACO_STATE_EXIT ; smu->smu_baco.platform_support = false; smu->smu_baco.maco_support = false; smu->user_dpm_profile.fan_mode =-1 ;
@@ -2134,9 +2134,8 @@ static int smu_reset_mp1_state(struct smu_context *smu)int ret =0 ;if ((!adev->in_runpm) && (!adev->in_suspend) && - (!amdgpu_in_reset (adev)) &&amdgpu_ip_version (adev,MP1_HWIP ,0 ) == -IP_VERSION (13 ,0 ,10 ) && - !amdgpu_device_has_display_hardware (adev)) + (!amdgpu_in_reset (adev)) && !smu->is_apu && +amdgpu_ip_version (adev,MP1_HWIP ,0 ) >=IP_VERSION (13 ,0 ,0 )) ret =smu_set_mp1_state (smu,PP_MP1_STATE_UNLOAD );return ret;
drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h +0 -1
@@ -508,7 +508,6 @@ enum smu_reset_mode {enum smu_baco_state {SMU_BACO_STATE_ENTER =0 ,SMU_BACO_STATE_EXIT , -SMU_BACO_STATE_NONE , };struct smu_baco_context {
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +1 -7
@@ -2770,13 +2770,7 @@ static int smu_v13_0_0_set_mp1_state(struct smu_context *smu,switch (mp1_state) {case PP_MP1_STATE_UNLOAD : - ret =smu_cmn_send_smc_msg_with_param (smu, - SMU_MSG_PrepareMp1ForUnload, -0x55 ,NULL ); - -if (!ret && smu->smu_baco.state ==SMU_BACO_STATE_EXIT ) - ret =smu_v13_0_disable_pmfw_state (smu); - + ret =smu_cmn_set_mp1_state (smu, mp1_state);break ;default :/* Ignore others */
Signature · 1 line
2.34.1
[AMD Official Use Only - AMD Internal Distribution Only]
Please split the patch into two parts, one for vf mode reset support and the other for modifying msg to debug port.
With that fixed, the series is
Best Regards,
Kevin
-----Original Message-----
From: Kenneth Feng <[email protected]>
Sent: Friday, January 23, 2026 13:57
To: [email protected]
Cc: Wang, Yang(Kevin) <[email protected]>; Deucher, Alexander <[email protected]>; Feng, Kenneth <[email protected]>
Subject: [PATCH 1/2] drm/amd/pm: use debug port for mode1 reset request on smu 13&14
use debug port for mode1 reset request so fw can handle mode1 reset even when it is stuck.
--- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 14 ++----------- .../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 6 ++++++ .../drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 21 ++++++++++++++++++- .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 6 +++++- 4 files changed, 33 insertions(+), 14 deletions(-)
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +2 -12
@@ -2617,21 +2617,11 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, static bool smu_v13_0_0_is_mode1_reset_supported(struct smu_context *smu) {struct amdgpu_device *adev = smu->adev; -u32 smu_version; -int ret;/* SRIOV does not support SMU mode1 reset */ if (amdgpu_sriov_vf (adev))return false; -/* PMFW support is available since 78.41 */ - ret =smu_cmn_get_smc_version (smu,NULL , &smu_version); -if (ret) -return false; - -if (smu_version <0x004e2900 ) -return false; -return true; }
@@ -2830,8 +2820,8 @@ static int smu_v13_0_0_mode1_reset(struct smu_context *smu)/* SMU 13_0_0 PMFW supports RAS fatal error reset from 78.77 */ smu_v13_0_0_set_mode1_reset_param (smu,0x004e4d00 , ¶m); - ret =smu_cmn_send_smc_msg_with_param (smu, - SMU_MSG_Mode1Reset, param,NULL ); + ret =smu_cmn_send_debug_smc_msg_with_param (smu, + DEBUGSMC_MSG_Mode1Reset, param);break ;case IP_VERSION (13 ,0 ,10 ):
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +6 -0
@@ -3055,6 +3055,12 @@ static int smu_v13_0_6_link_reset(struct smu_context *smu)static bool smu_v13_0_6_is_mode1_reset_supported (struct smu_context *smu) { +struct amdgpu_device *adev = smu->adev; + +/* SRIOV does not support SMU mode1 reset */ +if (amdgpu_sriov_vf (adev)) +return false; +return true; }
drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c +19 -1
@@ -72,6 +72,8 @@#define MP0_MP1_DATA_REGION_SIZE_COMBOPPTABLE 0x4000 +#define DEBUGSMC_MSG_Mode1Reset 2 +#define PP_OD_FEATURE_GFXCLK_FMIN 0#define PP_OD_FEATURE_GFXCLK_FMAX 1#define PP_OD_FEATURE_UCLK_FMIN 2
@@ -2735,6 +2737,23 @@ static int smu_v13_0_7_update_pcie_parameters(struct smu_context *smu,return ret; } +static int smu_v13_0_7_mode1_reset (struct smu_context *smu) { +int ret; + + ret =smu_cmn_send_debug_smc_msg (smu, DEBUGSMC_MSG_Mode1Reset); + +if (!ret) { +/* disable mmio access while doing mode 1 reset*/ + smu->adev->no_hw_access = true; +/* ensure no_hw_access is globally visible before any MMIO */ +smp_mb (); +msleep (SMU13_MODE1_RESET_WAIT_TIME_IN_MS ); + } + +return ret; +} +static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .get_allowed_feature_mask = smu_v13_0_7_get_allowed_feature_mask, .set_default_dpm_table = smu_v13_0_7_set_default_dpm_table,
@@ -2796,7 +2815,7 @@ static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .baco_enter = smu_v13_0_baco_enter, .baco_exit = smu_v13_0_baco_exit, .mode1_reset_is_support = smu_v13_0_7_is_mode1_reset_supported, - .mode1_reset = smu_v13_0_mode1_reset, + .mode1_reset = smu_v13_0_7_mode1_reset, .set_mp1_state = smu_v13_0_7_set_mp1_state, .set_df_cstate = smu_v13_0_7_set_df_cstate, .gpo_control = smu_v13_0_gpo_control,
drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +5 -2
@@ -1899,7 +1899,11 @@ static int smu_v14_0_2_baco_exit(struct smu_context *smu)static bool smu_v14_0_2_is_mode1_reset_supported (struct smu_context *smu) { -// TODO +struct amdgpu_device *adev = smu->adev; + +/* SRIOV does not support SMU mode1 reset */ +if (amdgpu_sriov_vf (adev)) +return false;return true; } --
2.34.1