From: Matthew Wilcox <willy@infradead.org>
To: Gautham Ananthakrishna <gautham.ananthakrishna@oracle.com>
Cc: linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
linux-mm@kvack.org, viro@zeniv.linux.org.uk,
matthew.wilcox@oracle.com, khlebnikov@yandex-team.ru
Subject: Re: [PATCH RFC 0/6] fix the negative dentres bloating system memory usage
Date: Wed, 31 Mar 2021 15:23:55 +0100 [thread overview]
Message-ID: <20210331142355.GX351017@casper.infradead.org> (raw)
In-Reply-To: <1611235185-1685-1-git-send-email-gautham.ananthakrishna@oracle.com>
Ping? These patches are looking pretty good in our internal testing.
On Thu, Jan 21, 2021 at 06:49:39PM +0530, Gautham Ananthakrishna wrote:
> For most filesystems result of every negative lookup is cached, content of
> directories is usually cached too. Production of negative dentries isn't
> limited with disk speed. It's really easy to generate millions of them if
> system has enough memory.
>
> Getting this memory back ins't that easy because slab frees pages only when
> all related objects are gone. While dcache shrinker works in LRU order.
>
> Typical scenario is an idle system where some process periodically creates
> temporary files and removes them. After some time, memory will be filled
> with negative dentries for these random file names.
>
> Simple lookup of random names also generates negative dentries very fast.
> Constant flow of such negative denries drains all other inactive caches.
> Too many negative dentries in the system can cause memory fragmentation
> and memory compaction.
>
> Negative dentries are linked into siblings list along with normal positive
> dentries. Some operations walks dcache tree but looks only for positive
> dentries: most important is fsnotify/inotify. Hordes of negative dentries
> slow down these operations significantly.
>
> Time of dentry lookup is usually unaffected because hash table grows along
> with size of memory. Unless somebody especially crafts hash collisions.
>
> This patch set solves all of these problems:
>
> Move negative denries to the end of sliblings list, thus walkers could
> skip them at first sight (patches 1-4).
>
> Keep in dcache at most three unreferenced negative denties in row in each
> hash bucket (patches 5-6).
>
> We tested this patch set recently and found it limiting negative dentry to a
> small part of total memory. The following is the test result we ran on two
> types of servers, one is 256G memory with 24 CPUS and another is 3T memory
> with 384 CPUS. The test case is using a lot of processes to generate negative
> dentry in parallel, the following is the test result after 72 hours, the
> negative dentry number is stable around that number even after running longer
> for much longer time. Without the patch set, in less than half an hour 197G was
> taken by negative dentry on 256G system, in 1 day 2.4T was taken on 3T system.
>
> system memory neg-dentry-number neg-dentry-mem-usage
> 256G 55259084 10.6G
> 3T 202306756 38.8G
>
> For perf test, we ran the following, and no regression found.
>
> 1. create 1M negative dentry and then touch them to convert them to positive
> dentry
>
> 2. create 10K/100K/1M files
>
> 3. remove 10K/100K/1M files
>
> 4. kernel compile
>
> To verify the fsnotify fix, we used inotifywait to watch file create/open in
> some directory where there is a lot of negative dentry, without the patch set,
> the system would run into soft lockup, with it, no soft lockup was found.
>
> We also tried to defeat the limitation by making different processes generate
> negative dentry with the same name, that will make one negative dentry being
> accessed couple times around same time, DCACHE_REFERENCED will be set on it
> and it can't be trimmed easily.
>
> There were a lot of customer cases on this issue. It makes no sense to leave
> so many negative dentry, it just causes memory fragmentation and compaction
> and does not help a lot.
>
> Konstantin Khlebnikov (6):
> dcache: sweep cached negative dentries to the end of list of siblings
> fsnotify: stop walking child dentries if remaining tail is negative
> dcache: add action D_WALK_SKIP_SIBLINGS to d_walk()
> dcache: stop walking siblings if remaining dentries all negative
> dcache: push releasing dentry lock into sweep_negative
> dcache: prevent flooding with negative dentries
>
> fs/dcache.c | 135 +++++++++++++++++++++++++++++++++++++++++++++++--
> fs/libfs.c | 3 ++
> fs/notify/fsnotify.c | 6 ++-
> include/linux/dcache.h | 6 +++
> 4 files changed, 145 insertions(+), 5 deletions(-)
>
> --
> 1.8.3.1
>
>
next prev parent reply other threads:[~2021-03-31 14:24 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-21 13:19 [PATCH RFC 0/6] fix the negative dentres bloating system memory usage Gautham Ananthakrishna
2021-01-21 13:19 ` [PATCH RFC 1/6] dcache: sweep cached negative dentries to the end of list of siblings Gautham Ananthakrishna
2021-04-14 3:00 ` Al Viro
2021-04-15 16:50 ` Al Viro
2021-04-14 3:41 ` Al Viro
2021-04-15 16:25 ` Al Viro
2021-01-21 13:19 ` [PATCH RFC 2/6] fsnotify: stop walking child dentries if remaining tail is negative Gautham Ananthakrishna
2021-01-21 13:19 ` [PATCH RFC 3/6] dcache: add action D_WALK_SKIP_SIBLINGS to d_walk() Gautham Ananthakrishna
2021-01-21 13:19 ` [PATCH RFC 4/6] dcache: stop walking siblings if remaining dentries all negative Gautham Ananthakrishna
2021-01-21 13:19 ` [PATCH RFC 5/6] dcache: push releasing dentry lock into sweep_negative Gautham Ananthakrishna
2021-01-21 13:19 ` [PATCH RFC 6/6] dcache: prevent flooding with negative dentries Gautham Ananthakrishna
2021-04-14 3:56 ` Al Viro
2021-03-31 14:23 ` Matthew Wilcox [this message]
2021-04-14 2:40 ` [PATCH RFC 0/6] fix the negative dentres bloating system memory usage Al Viro
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210331142355.GX351017@casper.infradead.org \
--to=willy@infradead.org \
--cc=gautham.ananthakrishna@oracle.com \
--cc=khlebnikov@yandex-team.ru \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=matthew.wilcox@oracle.com \
--cc=viro@zeniv.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).