diff options
Diffstat (limited to '')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c | 35 |
1 files changed, 27 insertions, 8 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index 0db458f9fafc..65eb378fa035 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -52,6 +52,9 @@ #define BIF_MMSCH1_DOORBELL_RANGE__OFFSET_MASK 0x00000FFCL #define BIF_MMSCH1_DOORBELL_RANGE__SIZE_MASK 0x001F0000L +static void nbio_v7_4_query_ras_error_count(struct amdgpu_device *adev, + void *ras_error_status); + static void nbio_v7_4_remap_hdp_registers(struct amdgpu_device *adev) { WREG32_SOC15(NBIO, 0, mmREMAP_HDP_MEM_FLUSH_CNTL, @@ -314,6 +317,7 @@ static void nbio_v7_4_init_registers(struct amdgpu_device *adev) static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device *adev) { uint32_t bif_doorbell_intr_cntl; + struct ras_manager *obj = amdgpu_ras_find_obj(adev, adev->nbio.ras_if); bif_doorbell_intr_cntl = RREG32_SOC15(NBIO, 0, mmBIF_DOORBELL_INT_CNTL); if (REG_GET_FIELD(bif_doorbell_intr_cntl, @@ -324,7 +328,18 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device RAS_CNTLR_INTERRUPT_CLEAR, 1); WREG32_SOC15(NBIO, 0, mmBIF_DOORBELL_INT_CNTL, bif_doorbell_intr_cntl); - amdgpu_ras_global_ras_isr(adev); + /* + * clear error status after ras_controller_intr according to + * hw team and count ue number for query + */ + nbio_v7_4_query_ras_error_count(adev, &obj->err_data); + + DRM_WARN("RAS controller interrupt triggered by NBIF error\n"); + + /* ras_controller_int is dedicated for nbif ras error, + * not the global interrupt for sync flood + */ + amdgpu_ras_reset_gpu(adev); } } @@ -441,10 +456,8 @@ static int nbio_v7_4_init_ras_controller_interrupt (struct amdgpu_device *adev) r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_BIF, NBIF_7_4__SRCID__RAS_CONTROLLER_INTERRUPT, &adev->nbio.ras_controller_irq); - if (r) - return r; - return 0; + return r; } static int nbio_v7_4_init_ras_err_event_athub_interrupt (struct amdgpu_device *adev) @@ -461,16 +474,16 @@ static int nbio_v7_4_init_ras_err_event_athub_interrupt (struct amdgpu_device *a r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_BIF, NBIF_7_4__SRCID__ERREVENT_ATHUB_INTERRUPT, &adev->nbio.ras_err_event_athub_irq); - if (r) - return r; - return 0; + return r; } +#define smnPARITY_ERROR_STATUS_UNCORR_GRP2 0x13a20030 + static void nbio_v7_4_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { - uint32_t global_sts, central_sts, int_eoi; + uint32_t global_sts, central_sts, int_eoi, parity_sts; uint32_t corr, fatal, non_fatal; struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -479,6 +492,7 @@ static void nbio_v7_4_query_ras_error_count(struct amdgpu_device *adev, fatal = REG_GET_FIELD(global_sts, RAS_GLOBAL_STATUS_LO, ParityErrFatal); non_fatal = REG_GET_FIELD(global_sts, RAS_GLOBAL_STATUS_LO, ParityErrNonFatal); + parity_sts = RREG32_PCIE(smnPARITY_ERROR_STATUS_UNCORR_GRP2); if (corr) err_data->ce_count++; @@ -490,6 +504,11 @@ static void nbio_v7_4_query_ras_error_count(struct amdgpu_device *adev, /* clear error status register */ WREG32_PCIE(smnRAS_GLOBAL_STATUS_LO, global_sts); + if (fatal) + /* clear parity fatal error indication field */ + WREG32_PCIE(smnPARITY_ERROR_STATUS_UNCORR_GRP2, + parity_sts); + if (REG_GET_FIELD(central_sts, BIFL_RAS_CENTRAL_STATUS, BIFL_RasContller_Intr_Recv)) { /* clear interrupt status register */ |