Re: [Intel-wired-lan] [PATCH] igc: Mask replay rollover/timeout errors in I225_LMVP

From: Bjorn Helgaas
Date: Tue Jan 03 2023 - 14:15:15 EST


On Tue, Jan 03, 2023 at 06:58:36AM +0000, Khandelwal, Rajat wrote:
> ...
> The reason I masked inherently is I witnessed a function
> netxen_mask_aer_correctable() inside
> net/ethernet/qlogic/netxen/netxen_nic_main.c, which masks the
> correctable errors in the corresponding PCIe device.

In my opinion, netxen_mask_aer_correctable() should not exist. The
PCI core should own the PCI_ERR_COR_MASK register.

netxen_mask_aer_correctable() was added by dce87b960cf4 ("netxen: mask
correctable error") with the note that it is a "HW workaround." Maybe
it covers up some hardware defect in the device, although it doesn't
include any evidence of this.

But if we do actually need it, I would rather have the driver set a
quirk bit that the PCI core can use to mask correctable errors so the
AER configuration is all in one place.

Bjorn