linux/mm
Johannes Weiner 4b85afbdac mm: zero-seek shrinkers
The page cache and most shrinkable slab caches hold data that has been
read from disk, but there are some caches that only cache CPU work, such
as the dentry and inode caches of procfs and sysfs, as well as the subset
of radix tree nodes that track non-resident page cache.

Currently, all these are shrunk at the same rate: using DEFAULT_SEEKS for
the shrinker's seeks setting tells the reclaim algorithm that for every
two page cache pages scanned it should scan one slab object.

This is a bogus setting.  A virtual inode that required no IO to create is
not twice as valuable as a page cache page; shadow cache entries with
eviction distances beyond the size of memory aren't either.

In most cases, the behavior in practice is still fine.  Such virtual
caches don't tend to grow and assert themselves aggressively, and usually
get picked up before they cause problems.  But there are scenarios where
that's not true.

Our database workloads suffer from two of those.  For one, their file
workingset is several times bigger than available memory, which has the
kernel aggressively create shadow page cache entries for the non-resident
parts of it.  The workingset code does tell the VM that most of these are
expendable, but the VM ends up balancing them 2:1 to cache pages as per
the seeks setting.  This is a huge waste of memory.

These workloads also deal with tens of thousands of open files and use
/proc for introspection, which ends up growing the proc_inode_cache to
absurdly large sizes - again at the cost of valuable cache space, which
isn't a reasonable trade-off, given that proc inodes can be re-created
without involving the disk.

This patch implements a "zero-seek" setting for shrinkers that results in
a target ratio of 0:1 between their objects and IO-backed caches.  This
allows such virtual caches to grow when memory is available (they do
cache/avoid CPU work after all), but effectively disables them as soon as
IO-backed objects are under pressure.

It then switches the shrinkers for procfs and sysfs metadata, as well as
excess page cache shadow nodes, to the new zero-seek setting.

Link: http://lkml.kernel.org/r/20181009184732.762-5-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reported-by: Domas Mituzas <dmituzas@fb.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Rik van Riel <riel@surriel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-10-26 16:26:33 -07:00
..
kasan kernel/memremap, kasan: make ZONE_DEVICE with work with KASAN 2018-08-17 16:20:30 -07:00
backing-dev.c blkcg: delay blkg destruction until after writeback has finished 2018-08-31 14:48:56 -06:00
balloon_compaction.c virtio_balloon: fix deadlock on OOM 2017-11-14 23:57:38 +02:00
bootmem.c docs/mm: bootmem: add overview documentation 2018-08-02 12:17:27 -06:00
cleancache.c mm: use octal not symbolic permissions 2018-06-15 07:55:25 +09:00
cma_debug.c mm/cma: remove unsupported gfp_mask parameter from cma_alloc() 2018-08-17 16:20:32 -07:00
cma.c mm/cma: remove unsupported gfp_mask parameter from cma_alloc() 2018-08-17 16:20:32 -07:00
cma.h License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
compaction.c psi: pressure stall information for CPU, memory, and IO 2018-10-26 16:26:32 -07:00
debug_page_ref.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
debug.c mm: get rid of vmacache_flush_all() entirely 2018-09-13 15:18:04 -10:00
dmapool.c mm: use octal not symbolic permissions 2018-06-15 07:55:25 +09:00
early_ioremap.c mm/early_ioremap: Fix boot hang with earlyprintk=efi,keep 2017-12-11 14:54:44 +01:00
fadvise.c vfs: implement readahead(2) using POSIX_FADV_WILLNEED 2018-08-30 20:01:32 +02:00
failslab.c mm: use octal not symbolic permissions 2018-06-15 07:55:25 +09:00
filemap.c psi: pressure stall information for CPU, memory, and IO 2018-10-26 16:26:32 -07:00
frame_vector.c mm/frame_vector.c: release a semaphore in 'get_vaddr_frames()' 2017-12-14 16:00:48 -08:00
frontswap.c mm: use octal not symbolic permissions 2018-06-15 07:55:25 +09:00
gup_benchmark.c mm/gup_benchmark: fix unsigned comparison to zero in __gup_benchmark_ioctl 2018-10-05 16:32:04 -07:00
gup.c mm: Change return type int to vm_fault_t for fault handlers 2018-08-23 18:48:44 -07:00
highmem.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
hmm.c libnvdimm-for-4.19_dax-memory-failure 2018-08-25 18:43:59 -07:00
huge_memory.c mm: workingset: tell cache transitions from workingset thrashing 2018-10-26 16:26:32 -07:00
hugetlb_cgroup.c mm: rename page_counter's count/limit into usage/max 2018-06-07 17:34:35 -07:00
hugetlb.c hugetlb: take PMD sharing into account when flushing tlb/caches 2018-10-05 16:32:04 -07:00
hwpoison-inject.c mm/memory_failure: Remove unused trapno from memory_failure 2018-01-23 12:17:42 -06:00
init-mm.c mm: Allocate the mm_cpumask (mm->cpu_bitmap[]) dynamically based on nr_cpu_ids 2018-07-17 09:35:30 +02:00
internal.h mm: Change return type int to vm_fault_t for fault handlers 2018-08-23 18:48:44 -07:00
interval_tree.c mm/interval_tree.c: use vma_pages() helper 2018-01-31 17:18:37 -08:00
Kconfig mm: disable deferred struct page for 32-bit arches 2018-09-20 22:01:11 +02:00
Kconfig.debug mm: clarify CONFIG_PAGE_POISONING and usage 2018-08-22 10:52:44 -07:00
khugepaged.c mm: Change return type int to vm_fault_t for fault handlers 2018-08-23 18:48:44 -07:00
kmemleak-test.c
kmemleak.c kmemleak: add module param to print warnings to dmesg 2018-10-26 16:25:19 -07:00
ksm.c include/linux/compiler*.h: make compiler-*.h mutually exclusive 2018-08-22 17:31:34 -07:00
list_lru.c mm/list_lru: introduce list_lru_shrink_walk_irq() 2018-08-17 16:20:32 -07:00
maccess.c x86/fault: BUG() when uaccess helpers fault on kernel addresses 2018-09-03 15:12:09 +02:00
madvise.c mm: madvise(MADV_DODUMP): allow hugetlbfs pages 2018-10-05 16:32:05 -07:00
Makefile arm64 updates for 4.20: 2018-10-22 17:30:06 +01:00
memblock.c mm/memblock.c: replace u64 with phys_addr_t where appropriate 2018-08-17 16:20:30 -07:00
memcontrol.c mm: drain memcg stocks on css offlining 2018-10-26 16:25:19 -07:00
memfd.c alloc_file(): switch to passing O_... flags instead of FMODE_... mode 2018-07-12 10:02:57 -04:00
memory_hotplug.c mm/hugetlb: filter out hugetlb pages if HUGEPAGE migration is not supported. 2018-09-04 16:45:02 -07:00
memory-failure.c libnvdimm-for-4.19_dax-memory-failure 2018-08-25 18:43:59 -07:00
memory.c mm: convert insert_pfn() to vm_fault_t 2018-10-26 16:25:20 -07:00
mempolicy.c userfaultfd: allow get_mempolicy(MPOL_F_NODE|MPOL_F_ADDR) to trigger userfaults 2018-10-26 16:25:20 -07:00
mempool.c mm/mempool.c: add missing parameter description 2018-08-22 10:52:44 -07:00
memtest.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
migrate.c mm: workingset: tell cache transitions from workingset thrashing 2018-10-26 16:26:32 -07:00
mincore.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
mlock.c dax: remove VM_MIXEDMAP for fsdax and device dax 2018-08-17 16:20:27 -07:00
mm_init.c mm: access zone->node via zone_to_nid() and zone_set_nid() 2018-08-22 10:52:45 -07:00
mmap.c mm/mmap.c: don't clobber partially overlapping VMA with MAP_FIXED_NOREPLACE 2018-10-13 09:31:02 +02:00
mmu_context.c
mmu_gather.c mm/memory: Move mmu_gather and TLB invalidation code into its own file 2018-09-07 15:19:25 +01:00
mmu_notifier.c Revert "mm, mmu_notifier: annotate mmu notifiers with blockable invalidate callbacks" 2018-10-26 16:25:19 -07:00
mmzone.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
mprotect.c x86/speculation/l1tf: Disallow non privileged high MMIO PROT_NONE mappings 2018-06-20 19:10:01 +02:00
mremap.c mremap: properly flush TLB before releasing the page 2018-10-18 11:30:52 +02:00
msync.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
nobootmem.c mm/memblock: add a name for memblock flags enumeration 2018-08-02 12:17:27 -06:00
nommu.c mm: provide a fallback for PAGE_KERNEL_EXEC for architectures 2018-08-17 16:20:29 -07:00
oom_kill.c Merge branch 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace 2018-10-24 11:22:39 +01:00
page_alloc.c psi: pressure stall information for CPU, memory, and IO 2018-10-26 16:26:32 -07:00
page_counter.c memcg: introduce memory.min 2018-06-07 17:34:36 -07:00
page_ext.c mm/page_ext.c: constify lookup_page_ext() argument 2018-08-17 16:20:28 -07:00
page_idle.c mm: use octal not symbolic permissions 2018-06-15 07:55:25 +09:00
page_io.c blkcg: associate a blkg for pages being evicted by swap 2018-09-21 20:29:09 -06:00
page_isolation.c mm, migrate: remove reason argument from new_page_t 2018-04-11 10:28:32 -07:00
page_owner.c mm: use octal not symbolic permissions 2018-06-15 07:55:25 +09:00
page_poison.c mm/page_poison.c: make early_page_poison_param() __init 2018-04-05 21:36:26 -07:00
page_vma_mapped.c mm, page_vma_mapped: Introduce pfn_in_hpage() 2018-01-22 12:15:57 -08:00
page-writeback.c notifier: Remove notifier header file wherever not used 2018-08-30 12:56:40 +02:00
pagewalk.c mm: kernel-doc: add missing parameter descriptions 2018-04-05 21:36:27 -07:00
percpu-internal.h License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
percpu-km.c percpu: allow select gfp to be passed to underlying allocators 2018-02-18 05:33:01 -08:00
percpu-stats.c treewide: Use array_size() in vmalloc() 2018-06-12 16:19:22 -07:00
percpu-vm.c percpu: allow select gfp to be passed to underlying allocators 2018-02-18 05:33:01 -08:00
percpu.c percpu: stop leaking bitmap metadata blocks 2018-10-07 14:50:12 -07:00
pgtable-generic.c x86/mm: Page size aware flush_tlb_mm_range() 2018-10-09 16:51:11 +02:00
process_vm_access.c mm: docs: add blank lines to silence sphinx "Unexpected indentation" errors 2018-02-06 18:32:48 -08:00
quicklist.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
readahead.c vfs: implement readahead(2) using POSIX_FADV_WILLNEED 2018-08-30 20:01:32 +02:00
rmap.c mm: migration: fix migration of huge PMD shared pages 2018-10-05 16:32:04 -07:00
rodata_test.c mm: fix RODATA_TEST failure "rodata_test: test data was not read only" 2017-10-03 17:54:24 -07:00
shmem.c mm: shmem.c: Correctly annotate new inodes for lockdep 2018-09-20 22:01:11 +02:00
slab_common.c mm, slab: shorten kmalloc cache names for large sizes 2018-10-26 16:26:32 -07:00
slab.c mm, slab: combine kmalloc_caches and kmalloc_dma_caches 2018-10-26 16:26:31 -07:00
slab.h mm: introduce CONFIG_MEMCG_KMEM as combination of CONFIG_MEMCG && !CONFIG_SLOB 2018-08-17 16:20:30 -07:00
slob.c slab: __GFP_ZERO is incompatible with a constructor 2018-06-07 17:34:34 -07:00
slub.c mm, slab: combine kmalloc_caches and kmalloc_dma_caches 2018-10-26 16:26:31 -07:00
sparse-vmemmap.c mm/sparse: delete old sparse_init and enable new one 2018-08-17 16:20:32 -07:00
sparse.c mm/sparse: delete old sparse_init and enable new one 2018-08-17 16:20:32 -07:00
swap_cgroup.c License cleanup: add SPDX GPL-2.0 license identifier to files with no license 2017-11-02 11:10:55 +01:00
swap_slots.c mm, swap, get_swap_pages: use entry_size instead of cluster in parameter 2018-08-22 10:52:44 -07:00
swap_state.c mm: workingset: tell cache transitions from workingset thrashing 2018-10-26 16:26:32 -07:00
swap.c mm: introduce MEMORY_DEVICE_FS_DAX and CONFIG_DEV_PAGEMAP_OPS 2018-05-22 06:59:39 -07:00
swapfile.c mm/swapfile.c: clear si->swap_map[] in swap_free_cluster() 2018-10-26 16:25:19 -07:00
truncate.c page cache: use xa_lock 2018-04-11 10:28:39 -07:00
usercopy.c usercopy: Allow boot cmdline disabling of hardening 2018-07-04 08:04:52 -07:00
userfaultfd.c userfaultfd: prevent non-cooperative events vs mcopy_atomic races 2018-06-07 17:34:38 -07:00
util.c mm: rename and change semantics of nr_indirectly_reclaimable_bytes 2018-10-26 16:26:32 -07:00
vmacache.c mm: get rid of vmacache_flush_all() entirely 2018-09-13 15:18:04 -10:00
vmalloc.c mm: provide a fallback for PAGE_KERNEL_EXEC for architectures 2018-08-17 16:20:29 -07:00
vmpressure.c mm/vmpressure.c: convert to use match_string() helper 2018-06-07 17:34:36 -07:00
vmscan.c mm: zero-seek shrinkers 2018-10-26 16:26:33 -07:00
vmstat.c mm: workingset: add vmstat counter for shadow nodes 2018-10-26 16:26:33 -07:00
workingset.c mm: zero-seek shrinkers 2018-10-26 16:26:33 -07:00
z3fold.c z3fold: fix reclaim lock-ups 2018-05-11 17:28:45 -07:00
zbud.c mm: docs: fix parameter names mismatch 2018-02-06 18:32:48 -08:00
zpool.c mm/zpool.c: zpool_evictable: fix mismatch in parameter name and kernel-doc 2018-02-21 15:35:43 -08:00
zsmalloc.c mm/zsmalloc.c: make several functions and a struct static 2018-08-17 16:20:30 -07:00
zswap.c zswap: re-check zswap_is_full() after do zswap_shrink() 2018-07-26 19:38:03 -07:00