Hacker News Reader

Zero details, nice.

> The reason turned out to be that SVM is discarding the guest PAT setting and uses the default (PA0=PA4=WB, PA1=PA5=WT, PA2=PA6=UC-, PA3=UC). The guest might be using a different setting, and especially might want write combining but isn't getting it (instead getting slow UC or UC- accesses). [0]

That's all it is. Not a big complicated bug, the patch is seven simple lines.

[0] https://patchwork.kernel.org/patch/10027523/

That patch might be small but I imagine the path to the patch had more details than mentioned in the article. It would be nice to hear that side of it to.

For someone benefiting from the patch it's a great story, from a developer viewpoint there aren't enough details for it to be satisfying.

Not much of a story [0] either. A quick benchmark or three, an assumption that something was missing... And it was, it was just stubbed.

This isn't a heisenbug, it was just never high enough priority for someone to notice.

[0] https://lists.linuxfoundation.org/pipermail/iommu/2017-Octob...

The links cover the details. There isn't much more to it.

https://lists.linuxfoundation.org/pipermail/iommu/2017-Octob...

https://patchwork.kernel.org/patch/10027523/

Reply on news.ycombinator.com