From: "Theodore Ts'o" <tytso@mit.edu>
To: Dave Chinner <david@fromorbit.com>
Cc: fstests@vger.kernel.org, hughd@google.com
Subject: Re: [PATCH 10/12] generic: disable generic/027 for tmpfs
Date: Wed, 10 Feb 2016 10:48:10 -0500 [thread overview]
Message-ID: <20160210154810.GC26922@thunk.org> (raw)
In-Reply-To: <20160210055838.GT19486@dastard>
On Wed, Feb 10, 2016 at 04:58:39PM +1100, Dave Chinner wrote:
> > With tmpfs, we can limit the size of the file system, which limits the
> > block allocation, but there is no limit to the number of inodes that
> > can be created until kmalloc() fails --- or the OOM killer kills the
> > test. So this causes this test to run for a long, long time, and in
> > some cases the test or the test runner will get OOM killed instead.
> > We have other ENOSPC tests, so given that tmpfs is just so different
> > from all other file systems, it's simpler just to disable this test
> > for tmpfs than to try to make it work.
>
> This sounds like a bug in tmpfs and a potential user level DOS
> vector, too. Hence I dont think not running the test is the right
> thing to do here - tmpfs should be handling this gracefully by
> applying sane resource limits.
Well, it's not really that interesting of a DOS vector since if the
goal is to use up all available memory, there are other ways to do it
that are much more efficient. If you are using a memory cgroup, the
kmalloc does eventually fail and so the O_CREAT open(2) call will
return ENOMEM --- but it can take 20+ hours with a 8G memory
container. Without a memory cgroup in general the test runner gets
OOM killed first, but a much better DOS vector is to open one too many
tabs in Chrome, at which point your machine thrashes to death and the
X server goes unresponsive (and this is why many people have started
running Chrome inside a memory container).
OTOH, the fact that tmpfs doesn't have a inode limits is a bit weird.
What about having tmpfs enforce a per-mount "maxinodes" restriction
which by default is "maxsize / 1k", and which can be overriden using a
maxinodes mount option? Does that sound sane to you?
Or we could charge a minimum 512 bytes per inode against the size,
since between the kernel data structures and the file name, it's not
like a zero-length tmpfs file is free.
- Ted
next prev parent reply other threads:[~2016-02-10 15:48 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-02-10 1:49 [PATCH 00/12] xfstests: fix up various tmpfs failures Theodore Ts'o
2016-02-10 1:49 ` [PATCH 01/12] check: avoid error messages of tests/$FS does not exist Theodore Ts'o
2016-02-10 5:45 ` Dave Chinner
2016-02-10 15:34 ` Theodore Ts'o
2016-02-10 16:28 ` Christoph Hellwig
2016-02-10 23:17 ` Theodore Ts'o
2016-02-10 22:19 ` Dave Chinner
2016-02-10 23:32 ` Theodore Ts'o
2016-02-10 1:49 ` [PATCH 02/12] common: _scratch_mkfs_sized() for tmpfs Theodore Ts'o
2016-02-10 6:00 ` Dave Chinner
2016-02-10 15:58 ` Theodore Ts'o
2016-02-10 22:37 ` Dave Chinner
2016-02-10 1:49 ` [PATCH 03/12] generic: use mount point instead of device name Theodore Ts'o
2016-02-10 1:49 ` [PATCH 04/12] generic: add _require_odirect to three more tests Theodore Ts'o
2016-02-10 9:15 ` Eryu Guan
2016-02-10 16:11 ` Theodore Ts'o
2016-02-10 22:51 ` Dave Chinner
2016-02-10 23:21 ` Theodore Ts'o
2016-02-10 1:49 ` [PATCH 05/12] xfstests: do not unmount tmpfs during remount Theodore Ts'o
2016-02-10 6:07 ` Dave Chinner
2016-02-10 16:07 ` Theodore Ts'o
2016-02-10 18:04 ` Theodore Ts'o
2016-02-10 23:07 ` Dave Chinner
2016-02-10 23:28 ` Theodore Ts'o
2016-02-11 3:07 ` Dave Chinner
2016-02-11 15:25 ` Theodore Ts'o
2016-02-11 17:36 ` Darrick J. Wong
2016-02-10 1:49 ` [PATCH 06/12] generic: do not unmount before calling _check_scratch_fs() Theodore Ts'o
2016-02-10 1:49 ` [PATCH 07/12] generic: require fiemap for generic/009 Theodore Ts'o
2016-02-10 1:49 ` [PATCH 08/12] xfstests: fix generic/312 on tmpfs, ignore /proc/partitions Theodore Ts'o
2016-02-10 5:54 ` Dave Chinner
2016-02-10 23:39 ` Theodore Ts'o
2016-02-11 2:53 ` Dave Chinner
2016-02-10 1:49 ` [PATCH 09/12] xfstests: generic/079 requires chattr, not xattrs Theodore Ts'o
2016-02-10 9:09 ` Eryu Guan
2016-02-10 16:09 ` Theodore Ts'o
2016-02-10 1:49 ` [PATCH 10/12] generic: disable generic/027 for tmpfs Theodore Ts'o
2016-02-10 5:58 ` Dave Chinner
2016-02-10 15:48 ` Theodore Ts'o [this message]
2016-02-10 23:13 ` Dave Chinner
2016-02-10 1:50 ` [PATCH 11/12] xfstests: add executable permission to tests Theodore Ts'o
2016-02-10 9:07 ` Eryu Guan
2016-02-10 1:50 ` [PATCH 12/12] xfstests: increase tmpfs memory size Theodore Ts'o
2016-02-10 2:10 ` [PATCH 00/12] xfstests: fix up various tmpfs failures Theodore Ts'o
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=20160210154810.GC26922@thunk.org \
--to=tytso@mit.edu \
--cc=david@fromorbit.com \
--cc=fstests@vger.kernel.org \
--cc=hughd@google.com \
/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