LKML News v5.10-rc4

[PATCH -V3 0/2] autonuma: Migrate on fault among multiple bound nodes (Huang Ying)

https://lkml.kernel.org/r/20201110055951.85085-1-ying.huang@intel.com

This patchset makes AutoNUMA’s cross-socket memory access optimization possible even if the application’s memory is bound to multiple NUMA nodes.

[PATCH] kfence: Avoid stalling work queue task without allocations (Marco Elver)

https://lkml.kernel.org/r/20201110135320.3309507-1-elver@google.com

This commit makes the toggle allocation gate of kfence to have 1 second timeout, to be prepared for allocation stallings.

[RFC PATCH 0/6] Split huge pages to any lower order pages. (Zi Yan)

https://lkml.kernel.org/r/20201111204008.21332-1-zi.yan@sent.com

Now pagecache can handle any size of THP, however the split_huge_page() splits a THP to only order-0 pages. For the reason, this patchset supports the function to be able to split a THP into any size smaller pages.

[PATCH v4 00/16] Overhaul multi-page lookups for THP (Matthew Wilcox)

https://lkml.kernel.org/r/20201112212641.27837-1-willy@infradead.org

This is a preparation of next changes Matthew will post. It first makes several page cache iteration APIs to only return head pages.

Are THPs the right model for the pagecache? (Matthew Wilcox)

https://lkml.kernel.org/r/20201113044652.GD17076@casper.infradead.org

THP is for architectures supporting large TLB entries. However, the point of enabling THP for page cache is only reducing the software works, so the definition doesn’t need to be true in there. Matthew suggests to use another model for the case and lists concerns about it.

Linux 5.10-rc4 (Linus Torvalds)

https://lkml.kernel.org/r/CAHk-=wjFfAktnadOPb_iV5nKh=V5Am1sG-gciYveswRtuEkrLQ@mail.gmail.com

It’s already rc4. The next long term support kernel, v5.10, will be released about 4-5 weeks later. Now Linus want the things to start calming down, but it wasn’t so much. That said, he says it’s not really worrisome but looks good.