Skip to content

Rate limit reporting of Correctable Errors

Jon Pan-Doh linux-pci 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest2 of 4 patches 2 repliesIn review

2 of 4 patches are in this archive. The cover letter announces 4.

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/4PCI/AER: Use the same log level for all messages+0 -00 replies
2/4PCI/AER: Add Correctable Errors rate limiting+0 -00 replies

Reviews trailers per patch

1 of 2 patches has a Reviewed-by or Acked-by.

PatchJon Pan-Doh
1/4R-b
2/4

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v1 thread

Cover letter [PATCH RESEND 0/4] Rate limit reporting of Correctable Errors
Quoted 10 lines · On Wed, 8 Jan 2025 13:55:30 +0000 Karolina Stolarek…

TL;DR
====

We are getting multiple reports about excessive logging of Correctable Errors with no clear common root cause. As these errors are already corrected by hardware, it makes sense to limit them. Introduce a ratelimit state definition to pci_dev to control the number of messages reported by a Root Port within a specified time interval. The series adds other improvements in the area, as outlined in the Proposal section.

Hi Karolina,

This is a common impediment for many folks that want to enable AER. The excessive logging stalls execution, making machines unusable. I've been working on a similar solution[1] to yours (i.e. ratelimiting) with a few differences:

- ratelimit uncorrectable errors
- ratelimit IRQs
- configure ratelimits from userspace (sysfs knobs)

Hoping we can collaborate on a solution (i.e. take best parts of both patch series).

Thanks,
Jon

[1] https://lore.kernel.org/linux-pci/[email protected]/