linux-kernel - Re: [RFC] [PATCH] Avoid livelock for fsync

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Wed, 4 Nov 2009 19:32:11 +0800
From:	Wu Fengguang <fengguang.wu@...el.com>
To:	Jan Kara <jack@...e.cz>
Cc:	"npiggin@...e.de" <npiggin@...e.de>,
	Andrew Morton <akpm@...ux-foundation.org>,
	LKML <linux-kernel@...r.kernel.org>,
	"linux-mm@...ck.org" <linux-mm@...ck.org>,
	"hch@...radead.org" <hch@...radead.org>,
	"chris.mason@...cle.com" <chris.mason@...cle.com>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>
Subject: Re: [RFC] [PATCH] Avoid livelock for fsync

Jan,

On Tue, Nov 03, 2009 at 10:56:54PM +0800, Jan Kara wrote:
[snip]
> > > inodes really does not make any sence - why should be a page in a file
> > > penalized and written later only because there are lots of other dirty
> > > pages in the file? It is enough to make sure that we don't write one file
> > > indefinitely when there are new dirty pages continuously created - and my
> > > patch achieves that.
> > 
> > This is a big policy change. Imagine dirty files A=4GB, B=C=D=1MB.
> > With current policy, it could be
> > 
> >         sync 4MB of A
> >         sync B
> >         sync C
> >         sync D
> >         sync 4MB of A
> >         sync 4MB of A
> >         ...
> > 
> > And you want to change to
> > 
> >         sync A (all 4GB)
> >         sync B
> >         sync C
> >         sync D
> > 
> > This means the writeback of B,C,D won't be able to start at 30s, but
> > delayed to 80s because of A. This is not entirely fair. IMHO writeback
> > of big files shall not delay small files too much. 
>   Yes, I'm aware of this change. It's just that I'm not sure we really
> care. There are few reasons to this: What advantage does it bring that we
> are "fair among files"? User can only tell the difference if after a crash,

I'm not all that sure, too. The perception is, big files normally
contain less valuable information per-page than small files ;)

If crashed, it's much better to lose one single big file, than to lose
all the (big and small) files.

Maybe nobody really care that - sync() has always been working file
after file (ignoring nr_to_write) and no one complained.

> files he wrote long time ago are still not on disk. But we shouldn't
> accumulate too many dirty data (like minutes of writeback) in caches
> anyway... So the difference should not be too big. Also how is the case
> "one big and a few small files" different from the case "many small files"
> where to be fair among files does not bring anything? 
>   It's just that see some substantial code complexity and also performance
> impact (because of smaller chunks of sequential IO) in trying to be fair
> among files and I don't really see adequate advantages of that approach.
> That's why I'm suggesting we should revisit the decision and possibly go in
> a different direction.

Anyway, if this is not a big concern, nr_to_write could be removed.

Note that requeue_io() (or requeue_io_wait) still cannot be removed
because sometimes we have (temporary) problems on writeback an inode.

Thanks,
Fengguang
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/