Re: vfs: Add MS_FLUSHONFSYNC mount flag

From: Eric Sandeen <sandeen@redhat.com>
To: "Fernando Luis Vázquez Cao" <fernando@oss.ntt.co.jp>
Cc: Jan Kara <jack@suse.cz>, Theodore Tso <tytso@MIT.EDU>,
	Alan Cox <alan@lxorguk.ukuu.org.uk>, Pavel Machek <pavel@suse.cz>,
	kernel list <linux-kernel@vger.kernel.org>,
	Jens Axboe <jens.axboe@oracle.com>,
	fernando@kic.ac.jp, Ric Wheeler <rwheeler@redhat.com>
Subject: Re: vfs: Add MS_FLUSHONFSYNC mount flag
Date: Fri, 13 Feb 2009 00:07:26 -0600	[thread overview]
Message-ID: <49950E1E.8040307@redhat.com> (raw)
In-Reply-To: <1234489622.3795.49.camel@sebastian.kern.oss.ntt.co.jp>

Fernando Luis Vázquez Cao wrote:
> On Thu, 2009-02-12 at 15:30 -0600, Eric Sandeen wrote:
>> Jan Kara wrote:
>>> On Thu 12-02-09 11:13:37, Eric Sandeen wrote:
>> ...
>>
>>>> Also that way if you have 8 partitions on a battery-backed blockdev, you
>>>> can tune it once, instead of needing to mount all 8 filesystems with the
>>>> new option.
>>>   Yes, but OTOH we should give sysadmin a possibility to enable / disable
>>> it on just some partitions. I don't see a reasonable use for that but people
>>> tend to do strange things ;) and here isn't probably a strong reason to not
>>> allow them.
>>>
>>> 								Honza
>> But nobody has asked for that, have they?  So why offer it up a this point?
>>
>> They could use LD_PRELOAD to make fsync a no-op if they really don't
>> care for it, I guess... though that's not easily per-fs either.
>>
>> But do we really want to go out of our way to enable people to
>> short-circuit data integrity paths and then file bugs when their files
>> go missing? :)
> 
> Well, it is just a matter of using safe defaults. IMHO, a scenario where
> the administrator wants to optimize writes to a certain partition and
> _explicitly_ clears MS_FLUSHONFSYNC on that superblock is not completely
> unreasonable.

One case is "this device is safe with a write cache and flush is not
necessary for data consistency" - that's the per-bdev setting.

The other case is "I don't really care and I just want to go faster" -
that's the per-mount setting.

I'd be much more likely to support the first case, as it's needed for
maximum performance w/o sacrificing correctness, when properly used.

The latter case is really only for cutting corners and giving people
more rope than they need to hang themselves.

>> (I guess the blockdev tunable is similarly dangerous, but it more
>> clearly meets the explicit need (writecache-safe devices))
> 
> If distributions use sane defaults and we document the mount option or
> bdev tunable properly I guess it might make sense to allow system
> administrators to shoot themselves in the foot.
> 
> (By the way, in this patch-set a patch for mount(8) is included.)

... one of the advantages of making it a bdev tunable is that you don't
have to mess with mount(8) :)

-Eric