LKML News v6.5-rc1

[GIT PULL] MM updates for 6.5-rc1 (Andrew Morton)

https://lkml.kernel.org/r/20230626085035.e66992e96b4c6d37dad54bd9@linux-foundation.org

Andrew Morton sent the memory management subsystem pull request for v6.5-rc1. Among the change, cachestata() syscall and Mel Gorman’s compaction optimization looks interesting to me.

[GIT PULL] slab updates for 6.5 (Vlastimil Babka)

https://lkml.kernel.org/r/1c39c9b0-ec37-f910-2b09-cedf7acf6e91@suse.cz

Vlastimil sent the slab pull request for v6.5 to Linus Torvalds. This contains the SLAB config deprecation renaming (s/CONFIG_SLAB/CONFIG_SLAB_DEPRECATED/).

[PATCH v1 0/5] mm/memory_hotplug: make offline_and_remove_memory() timeout instead of failing on fatal signals (David Hildenbrand)

https://lkml.kernel.org/r/20230627112220.229240-1-david@redhat.com

offline_and_remove_memory() checks fatal signals to allow timeout. It works for sysfs-based direct offlining, but no much sense for out-of-tree drivers. This patchset implements timeout feature directly for the function.

[PATCH] writeback: Account the number of pages written back (Matthew Wilcox)

https://lkml.kernel.org/r/20230628185548.981888-1-willy@infradead.org

Currently nr_to_write of struct writeback_contro is wrongly set because it’s not accounting the number of pages written back. This commit fixes it to account by decreasing the number of pages in the folio we just wrote.

[PATCH v6 0/6] Per-VMA lock support for swap and userfaults (Suren Baghdasaryan)

https://lkml.kernel.org/r/20230630020436.1066016-1-surenb@google.com

The sixth version of the patchset that adding support of per-vma lock for swap and userfaults.

[PATCH v2] Documentation/mm: Add information about kmap_local_folio() (Fabio M. De Francesco)

https://lkml.kernel.org/r/20230701142041.9992-1-fmdefrancesco@gmail.com

This patch updates the highmem documentation to add information about kmap_local_folio().

[PATCH 1/1] mm: disable CONFIG_PER_VMA_LOCK by default until its fixed (Suren Baghdasaryan)

https://lkml.kernel.org/r/20230703182150.2193578-1-surenb@google.com

A memory corruption that supposed to be occurred due to per-VMA lock enabling has found. To fix the regression first, this patch disables per-VMA lock by default, until the issue is fixed.

[PATCH 1/1] fork: lock VMAs of the parent process when forking (Suren Baghadasaryan)

https://lkml.kernel.org/r/20230704200656.2526715-1-surenb@google.com

This patch fixes the memory corruption problem due to the per-vma locking. It makes fork() slower, though. On stress testing, about 5% regression was found. A solution for avoiding the regression is turning CONFIG_PER_VMA_LOCK off.

[PATCH] memcg: drop kmem.limit_in_bytes (Michal Hocko)

https://lkml.kernel.org/r/20230704115240.14672-1-mhocko@kernel.org

kmem.limit memcg file has been deprecated since v5.16. Michal tries to entirely remove it, since existence of it only causes confusion and bugs. He further wants the removal to be applied to stable kernels.

[RFC PATCH 2/2] zram: charge the compressed RAM to the page’s memcgroup (Zhongkun He)

https://lkml.kernel.org/r/20230707044707.1169285-1-hezhongkun.hzk@bytedance.com

The compressed memory for zram is not charged to any memcg but kernel. This patch makes those to be charged to the memcg.

[PATCH] mm:vmscan: fix inaccurate reclaim during proactive reclaim (Efly Young)

https://lkml.kernel.org/r/20230707103226.38496-1-yangyifei03@kuaishou.com

memory.reclaim memcg file based reclaim is still inaccurate, in terms of the amount of memory it reclaims. This commit further fixes it.

Linux 6.5-rc1 (Linus Torvalds)

https://lkml.kernel.org/r/CAHk-=wj8sPDVoWgaceAs1AiwZrHV8mtC3vQNGbeV6-RypJi6aw@mail.gmail.com

After the two weeks of the merge window for Linux v6.5, the first release candidate has released. Nothing unusual stands out.

Below is the diffstat of the releases in the last two years.

Note that the y-axis is in logarithm. I draw it using https://github.com/sjp38/relstat and https://github.com/sjp38/lazybox using below command:

$ relstat.py --since 2021-07-10 | ~/lazybox/gnuplot/plot.py \
    --data_fmt table --type labeled-lines --xtics_rotate -90 \
    --font "Times New Roman, 5pt" --ylog --pointsize 0.3

And, below is the diffstat of the -rc1 releases in the last two years.