Linux 6.11 "MM" Patches Include Many Improvements, A 10x Speedup For One Optimization
There's a lot as usual in the MM code for this next Linux kernel cycle with many THP enhancements, various memory management optimizations, and other low-level kernel code improvements. Some of the MM enhancements in Linux 6.11 include:
- Fixes to writeback throttling code.
- Continued folio conversions.
- Cleanups to the core TLB handling code along with some small optimizations.
- Multi-size THP (mTHP) support for anonymous SHMEM that can yield a "significant improvement" around page fault latency.
- Reclaiming lazy-free THP without splitting to lower the latency of reclaim of PMD-mapped transparent hugepages (THPs) under common circumstances. In a micro-benchmark this has shown a 10x speed-up. The patch series explains of the work:
"This series adds support for reclaiming PMD-mapped THP marked as lazyfree without needing to first split the large folio via split_huge_pmd_address().
When the user no longer requires the pages, they would use madvise(MADV_FREE) to mark the pages as lazy free. Subsequently, they typically would not re-write to that memory again.
During memory reclaim, if we detect that the large folio and its PMD are both still marked as clean and there are no unexpected references(such as GUP), so we can just discard the memory lazily, improving the efficiency of memory reclamation in this case."
- DAMON functionality to help determine the efficiency of placement of CXL/PCIe-attached DRAM.
- Various updates to the NILFS2 file-system driver.
- Kernel Memory Sanitizer (KMSAN) support on s390.
More details on all the MM feature patches for Linux 6.11 via this pull request.
Andrew Morton also submitted the non-MM pull request at the same time with some clean-ups and other code improvements as noted in that pull.
I'll begin my Linux 6.11 kernel benchmarking and testing on different systems following the merge window wrapping up next weekend.
1 Comment
