[Intel-xe] [PATCH v5 4/4] drm/xe: Process fatal hardware errors.
Ghimiray, Himal Prasad
himal.prasad.ghimiray at intel.com
Tue Sep 26 10:24:46 UTC 2023
On 26-09-2023 09:51, Aravind Iddamsetty wrote:
> On 23/08/23 14:28, Himal Prasad Ghimiray wrote:
>> Fatal errors are reported as PCIe errors. When a PCIe error is asserted,
>> the OS will perform a device warm reset which causes the driver to reload.
>> The error registers are sticky and the values are maintained through a
>> warm reset. We read these registers during the boot flow of the driver and
>> increment the respective error counters.
>>
>> Bspec: 53076
> please mention 50875 as well.
sure.
>> Cc: Rodrigo Vivi <rodrigo.vivi at intel.com>
>> Cc: Aravind Iddamsetty <aravind.iddamsetty at intel.com>
>> Cc: Matthew Brost <matthew.brost at intel.com>
>> Cc: Matt Roper <matthew.d.roper at intel.com>
>> Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
>> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray at intel.com>
>> ---
>> drivers/gpu/drm/xe/regs/xe_regs.h | 3 +++
>> drivers/gpu/drm/xe/xe_hw_error.c | 37 ++++++++++++++++++++++++++++++-
>> drivers/gpu/drm/xe/xe_hw_error.h | 3 ++-
>> drivers/gpu/drm/xe/xe_irq.c | 2 +-
>> 4 files changed, 42 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/regs/xe_regs.h b/drivers/gpu/drm/xe/regs/xe_regs.h
>> index e223975a5acf..b8f2b1762d3f 100644
>> --- a/drivers/gpu/drm/xe/regs/xe_regs.h
>> +++ b/drivers/gpu/drm/xe/regs/xe_regs.h
>> @@ -90,5 +90,8 @@
>> #define GT_DW_IRQ(x) REG_BIT(x)
>> #define XE_ERROR_IRQ(x) REG_BIT(26 + (x))
>>
>> +#define DEV_PCIEERR_STATUS XE_REG(0x100180)
>> +#define DEV_PCIEERR_IS_FATAL(x) REG_BIT(x * 4 + 2)
> place it as per the address.
>> +
>> #define PVC_RP_STATE_CAP XE_REG(0x281014)
>> #endif
>> diff --git a/drivers/gpu/drm/xe/xe_hw_error.c b/drivers/gpu/drm/xe/xe_hw_error.c
>> index deb020a509d2..9595e3369656 100644
>> --- a/drivers/gpu/drm/xe/xe_hw_error.c
>> +++ b/drivers/gpu/drm/xe/xe_hw_error.c
>> @@ -183,7 +183,7 @@ static const struct err_msg_cntr_pair err_stat_gt_correctable_vectr_reg[] = {
>> [2 ... 3] = {"L3BANK", XE_GT_HW_ERR_L3BANK_CORR},
>> };
>>
>> -void xe_assign_hw_err_regs(struct xe_device *xe)
>> +static void xe_assign_hw_err_regs(struct xe_device *xe)
> change it in the patch that introduced it.
>> {
>> const struct err_msg_cntr_pair **dev_err_stat = xe->hw_err_regs.dev_err_stat;
>> const struct err_msg_cntr_pair **err_stat_gt = xe->hw_err_regs.err_stat_gt;
>> @@ -417,3 +417,38 @@ xe_hw_error_irq_handler(struct xe_tile *tile, const u32 master_ctl)
>> xe_hw_error_source_handler(tile, hw_err);
>> }
>> }
>> +
>> +/**
>> + * process_hw_errors - checks for the occurrence of HW errors
>> + *
>> + * This checks for the HW Errors including FATAL errors that might
>> + * have occurred in the previous boot of the driver which will
>> + * initiate PCIe FLR reset of the device and cause the
> this is not right, in fact it is better to rephrase as:
>
> "fatal will result in a card warm reset and driver will be reloaded."
ok
>
>> + * driver to reload.
>> + */
>> +void xe_process_hw_errors(struct xe_device *xe)
>> +{
>> + struct xe_tile *root_tile = xe_device_get_root_tile(xe);
>> + struct xe_gt *root_mmio = root_tile->primary_gt;
>> +
>> + u32 dev_pcieerr_status, master_ctl;
>> + struct xe_tile *tile;
>> + int i;
>> +
>> + xe_assign_hw_err_regs(xe);
> lets still have this in xe_irq_install, as it serves for a different purpose.
ok
>> +
>> + dev_pcieerr_status = xe_mmio_read32(root_mmio, DEV_PCIEERR_STATUS);
>> +
>> + for_each_tile(tile, xe, i) {
>> + struct xe_gt *mmio = tile->primary_gt;
>> +
>> + if (dev_pcieerr_status & DEV_PCIEERR_IS_FATAL(i))
>> + xe_hw_error_source_handler(tile, HARDWARE_ERROR_FATAL);
>> +
>> + master_ctl = xe_mmio_read32(mmio, GFX_MSTR_IRQ);
>> + xe_hw_error_irq_handler(tile, master_ctl);
>> + xe_mmio_write32(mmio, GFX_MSTR_IRQ, master_ctl);
>> + }
>> + if (dev_pcieerr_status)
>> + xe_mmio_write32(root_mmio, DEV_PCIEERR_STATUS, dev_pcieerr_status);
>> +}
>> diff --git a/drivers/gpu/drm/xe/xe_hw_error.h b/drivers/gpu/drm/xe/xe_hw_error.h
>> index 3fcbbcc338fe..2812407dd4bf 100644
>> --- a/drivers/gpu/drm/xe/xe_hw_error.h
>> +++ b/drivers/gpu/drm/xe/xe_hw_error.h
>> @@ -104,5 +104,6 @@ struct xe_device;
>> struct xe_tile;
>>
>> void xe_hw_error_irq_handler(struct xe_tile *tile, const u32 master_ctl);
>> -void xe_assign_hw_err_regs(struct xe_device *xe);
>> +void xe_process_hw_errors(struct xe_device *xe);
>> +
>> #endif
>> diff --git a/drivers/gpu/drm/xe/xe_irq.c b/drivers/gpu/drm/xe/xe_irq.c
>> index 48b933234342..be152ebd6ce9 100644
>> --- a/drivers/gpu/drm/xe/xe_irq.c
>> +++ b/drivers/gpu/drm/xe/xe_irq.c
>> @@ -573,7 +573,7 @@ int xe_irq_install(struct xe_device *xe)
>> return -EINVAL;
>> }
>>
>> - xe_assign_hw_err_regs(xe);
>> + xe_process_hw_errors(xe);
> this shall be called before xe_irq_reset.
>>
>> xe->irq.enabled = true;
>>
> Thanks,
> Aravind.
More information about the Intel-xe
mailing list