Re: [RFC PATCH v1 00/13] lru_lock scalability

From: Daniel Jordan
Date: Thu Feb 01 2018 - 23:18:31 EST



On 02/01/2018 10:54 AM, Steven Whitehouse wrote:
Hi,


On 31/01/18 23:04, daniel.m.jordan@xxxxxxxxxx wrote:
lru_lock, a per-node* spinlock that protects an LRU list, is one of the
hottest locks in the kernel. On some workloads on large machines, it
shows up at the top of lock_stat.

One way to improve lru_lock scalability is to introduce an array of locks,
with each lock protecting certain batches of LRU pages.

ÂÂÂÂÂÂÂÂ *ooooooooooo**ooooooooooo**ooooooooooo**oooo ...
ÂÂÂÂÂÂÂÂ |ÂÂÂÂÂÂÂÂÂÂ ||ÂÂÂÂÂÂÂÂÂÂ ||ÂÂÂÂÂÂÂÂÂÂ ||
ÂÂÂÂÂÂÂÂÂ \ batch 1 /Â \ batch 2 /Â \ batch 3 /

In this ASCII depiction of an LRU, a page is represented with either '*'
or 'o'. An asterisk indicates a sentinel page, which is a page at the
edge of a batch. An 'o' indicates a non-sentinel page.

To remove a non-sentinel LRU page, only one lock from the array is
required. This allows multiple threads to remove pages from different
batches simultaneously. A sentinel page requires lru_lock in addition to
a lock from the array.

Full performance numbers appear in the last patch in this series, but this
prototype allows a microbenchmark to do up to 28% more page faults per
second with 16 or more concurrent processes.

This work was developed in collaboration with Steve Sistare.

Note: This is an early prototype. I'm submitting it now to support my
request to attend LSF/MM, as well as get early feedback on the idea. Any
comments appreciated.


* lru_lock is actually per-memcg, but without memcg's in the picture it
ÂÂ becomes per-node.
GFS2 has an lru list for glocks, which can be contended under certain workloads. Work is still ongoing to figure out exactly why, but this looks like it might be a good approach to that issue too. The main purpose of GFS2's lru list is to allow shrinking of the glocks under memory pressure via the gfs2_scan_glock_lru() function, and it looks like this type of approach could be used there to improve the scalability,

Glad to hear that this could help in gfs2 as well.

Hopefully struct gfs2_glock is less space constrained than struct page for storing the few bits of metadata that this approach requires.

Daniel


Steve.


Aaron Lu (1):
ÂÂ mm: add a percpu_pagelist_batch sysctl interface

Daniel Jordan (12):
ÂÂ mm: allow compaction to be disabled
ÂÂ mm: add lock array to pgdat and batch fields to struct page
ÂÂ mm: introduce struct lru_list_head in lruvec to hold per-LRU batch
ÂÂÂÂ info
ÂÂ mm: add batching logic to add/delete/move API's
ÂÂ mm: add lru_[un]lock_all APIs
ÂÂ mm: convert to-be-refactored lru_lock callsites to lock-all API
ÂÂ mm: temporarily convert lru_lock callsites to lock-all API
ÂÂ mm: introduce add-only version of pagevec_lru_move_fn
ÂÂ mm: add LRU batch lock API's
ÂÂ mm: use lru_batch locking in release_pages
ÂÂ mm: split up release_pages into non-sentinel and sentinel passes
ÂÂ mm: splice local lists onto the front of the LRU

 include/linux/mm_inline.h | 209 +++++++++++++++++++++++++++++++++++++++++++++-
 include/linux/mm_types.h | 5 ++
 include/linux/mmzone.h | 25 +++++-
 kernel/sysctl.c | 9 ++
 mm/Kconfig | 1 -
 mm/huge_memory.c | 6 +-
 mm/memcontrol.c | 5 +-
 mm/mlock.c | 11 +--
 mm/mmzone.c | 7 +-
 mm/page_alloc.c | 43 +++++++++-
 mm/page_idle.c | 4 +-
 mm/swap.c | 208 ++++++++++++++++++++++++++++++++++++---------
 mm/vmscan.c | 49 +++++------
 13 files changed, 500 insertions(+), 82 deletions(-)