On Wed, Nov 13, 2024 at 03:54:21PM -0600, Terry Bowman wrote:
> Non-fatal CXL UCE errors will be treated as fatal.
Hm, I wonder why?
> --- a/drivers/pci/pcie/aer.c
> +++ b/drivers/pci/pcie/aer.c
> @@ -1048,7 +1048,10 @@ static void cxl_handle_error(struct pci_dev *dev, struct aer_err_info *info)
> pdrv->cxl_err_handler->cor_error_detected(dev);
>
> pcie_clear_device_status(dev);
> - }
> + } else if (info->severity == AER_NONFATAL)
> + cxl_do_recovery(dev);
> + else if (info->severity == AER_FATAL)
> + cxl_do_recovery(dev);
> }
Nit: Maybe use curly braces and collapse both if-block into one.
> + cxl_walk_bridge(bridge, cxl_report_error_detected, &status);
> + if (status)
> + panic("CXL cachemem error. Invoking panic");
Nit: This will be prefixed by "Kernel panic - not syncing: ",
so another "Invoking panic" message seems somewhat redundant.
Thanks,
Lukas
