news

LKML News v5.17-rc6

[PATCH -V13 0/3] NUMA balancing: optimize memory placement for memory tiering system (Huang Ying) https://lkml.kernel.org/r/20220221084529.1052339-1-ying.huang@intel.com The AutoNUMA patchset for tiered memory system has revised again. [LSF/MM/BPF TOPIC] page table reclaim (David Hildenbrand) https://lkml.kernel.org/r/7b908208-02f8-6fde-4dfc-13d5e00310a6@redhat.com Some workloads allocates a lot of page tables. As page tables are unmovable and unreclaimable, this can contribute to memory pressure. David is proposing making it reclaimable and suggesting discussion in LSF/MM/BPF summit for the idea. [RFC PATCH 00/47] Address Space Isolation for KVM (Junaid Shahid) https://lkml.

LKML News v5.17-rc5

[PATCH v3 0/3] Add hugetlb MADV_DONTNEED support (Mike Kravetz) https://lkml.kernel.org/r/20220215002348.128823-1-mike.kravetz@oracle.com MADV_DONTNEED can be used for hugetlb but currently prohibits it with a simple check. This commit removes the check to enable MADV_DONENEED for hugetlb. mmotm 2022-02-14-17-46 uploaded (Andrew Morton) https://lkml.kernel.org/r/20220215014633.9CD9CC340E9@smtp.kernel.org mmotm has updated. It’s rebased on 5.17-rc4, added 20 new patches, modified 295 patches, and dropped 10 patches. In total, 316 patches in the queue. [PATCH v6 00/71] Introducing the Maple Tree (Liam Howlett) https://lkml.

LKML News v5.17-rc4

[PATCH v2 1/2] arm64: avoid flushing icache multiple times on contiguous HugeTLB (Muchun Song) https://lkml.kernel.org/r/20220208054632.66534-1-songmuchun@bytedance.com When a contiguous HugeTLB page is mapped, set_pte_at() is called CONT_PTES / CONT_PMDS times. Because the first flushing covers subsequent flushes, this is unnecessary overhead, so this patchset avoids it. [PATCH v7 00/12] Multigenerational LRU Framework (Yu Zhao) https://lkml.kernel.org/r/20220208081902.3550911-1-yuzhao@google.com Seventh version of the MGLRU patchset, which aims to reduce performance overhead and improve eviction target pages selection of reclamation logic is posted.

LKML News v5.17-rc3

[PATCH] selftests: kselftest framework: Provide “finished” helper (Kees Cook) https://lkml.kernel.org/r/20220201013717.2464392-1-keescook@chromium.org Users who want to use ksft_exit() need to figure out the internals of kselftest. This commit adds a helper function for that purpose, namely, ksft_finished(). [LSF/MM/BPF TOPIC] Handling of VMAs (Liam Howlett) https://lkml.kernel.org/r/20220201150521.owa5axfsmpdvc7uo@revolver Liam is pointing out the VMAs are quite complicately managed with their cache, tree, linked list, and even reverse mapping. As it now shows its limit with the lock contention and complexity of the handling code, he wants to propose a simpler new way for their management.

LKML News v5.17-rc2

[PATCH 0/7] Allocate module text and data separately (Christophe Leroy) https://lkml.kernel.org/r/cover.1643015752.git.christophe.leroy@csgroup.eu Powerpc book3s/32 can set execute permission in 256 MiB segments granularity instead of page granularity. This patchset therefore allow module data be allocated in vmalloc rea instead of module area. [PATCH 0/4] mm/memcg: Address PREEMPT_RT problems instead of disabling it. (Sebastian Adrzej Siewior) https://lkml.kernel.org/r/20220125164337.2071854-1-bigeasy@linutronix.de This patchset enables MEMCG for PREEMPT_RT instead of disabling it. [LSF/MM/BPF TOPIC] Folios (Matthew Wilcox) https://lkml.

LKML News v5.17-rc1

[PATCH 0/3] usable memory range fixes (arm64/fdt/efi) (Frank van der Linden) https://lkml.kernel.org/r/20220110210809.3528-1-fllinden@amazon.com This patchset fixes the usable memory range behavior when both DT and EFI are given. [PATCHv2 0/7] Implement support for unaccepted memory (Kirill A. Shutemov) https://lkml.kernel.org/r/20220111113314.27173-1-kirill.shutemov@linux.intel.com UEFI spec 2.9 introduces a concept of memory acceptance. It requires virtual machine platforms such as Intel TDX or AMD SEV-SNP to accept memory before use it. This patchset makes Linux kernel to support it, and therefore be able to reduce VM booting time by optimizing the acceptance.

LKML News v5.16

[PATCH v4 00/32] Separate struct slab from struct page (Vlastimil Babka) https://lkml.kernel.org/r/20220104001046.12263-1-vbabka@suse.cz The work for seperating struct slab from struct page for making the struct page simpler is updated to fourth version. As the patchset was in -next tree for a while, Vlastimil would sent a pull request of this to Linus. [PATCH 0000/2297] [ANNOUNCE, RFC] “Fast Kernel Headers” Tree -v1: Eliminate the Linux kernel’s “Dependency Hell” (Ingo Molnar) https://lkml.kernel.org/r/YdIfz+LMewetSaEB@gmail.com

LKML News v5.16-rc8

[RFC v2] mm: introduce page pin owner (Minchan Kim) https://lkml.kernel.org/r/20211228175904.3739751-1-minchan@kernel.org Second version of the RFC for page pinning owner feature, which can be useful for tracking down who pins which page and therefore why contiguous pages allocation or migration fails. mmotm 2021-12-29-20-07 uploaded (Andrew Morton) https://lkml.kernel.org/r/20211230040740.SbquJAFf5%akpm@linux-foundation.org MM of the moment has updated. It’s rebased on v5.16-rc7, adds 32 new patches to the queue, drops 11 patches from the queue, and modifies 100 existing patches.

LKML News v5.16-rc7

[PATCH v5 00/16] Optimize list lru memory consumption (Muchun Song) https://lkml.kernel.org/r/20211220085649.8196-1-songmuchun@bytedance.com Fifth version of the patchset for reducing lru memory consumption. This version fixes a sleep-in-atomic-context bug which is reported by the kernel test robot, applies some requests from Johannes, and collect Acked-by from Johannes. [PATCH 0/2] Add a new scheme to support demotion on tiered memory system (Baolin Wang) https://lkml.kernel.org/r/cover.1640077468.git.baolin.wang@linux.alibaba.com This patchset makes DAMOS supports demotion of memory regions.

LKML News v5.16-rc6

[PATCH v4 00/17] Optimize list lru memory consumption (Muchun Song) https://lkml.kernel.org/r/20211213165342.74704-1-songmuchun@bytedance.com Muchun found 6GB of memory was consumed by kmalloc-32 slab cache on a machine having 952 super blocks. That was due to list_lru_one allocation from memcg_nr_cache_ids. This patchset aims to fix the issue. According to a simple test, this reduces memory consumption under a situation from 24GiB to 4GiB. mmotm 2021-12-13-21-28 uploaded (Andrew Morton) https://lkml.kernel.org/r/20211214052904.VfRYfitp0%akpm@linux-foundation.org Andrew Morton uploaded new mmotm.