On Thursday 26 November 2009 16:14:15 JiSheng Zhang wrote:
> 2009/11/25 Laurence Withers <l@lwithers.me.uk>:
> > On Wed, Nov 25, 2009 at 03:27:26PM +1300, Charles Manning wrote:
> >> What I have now only fails if a mmap write is followed by a truncate
> >> down then truncate up.
> >
> > I'm not sure if this is relevant, or sheds any light on the situation,
> > but have you read this article from LWN?
> >
> >  File holes, races, and mmap()
> >  http://lwn.net/Articles/357767/
>
> I read the article a few weeks ago. If I understand correctly, the bug
> described by this article is triggered when quota exhausts or the
> filesystem has no more space or two process and one of it be killed.
> However, this is not the case in fsx

The article is all about new kernel mods and race conditions. The problem of 
yaffs not running fsx predates these changes.

Unfortunately the kernel VFS interface is one of the most changing interfaces 
in the whole kernel and it is really challenging to keep up. Each file system 
handles the interface in slightly different ways. What would be nice is 
simplified fs interfacing layer.

-- Charles