Re: offset broken for randwrite?

On Tue, Feb 19 2008, Jens Axboe wrote:
> On Tue, Feb 19 2008, Ryan Thomas wrote:
> >
> > Your patch does fix the segfaults. But now I'm really confused about the I/O pattern realized when using randwrite and a non-zero offset. The I/O trace shows multiple opens/closes anad the same block being written over multiple times. Here's what I got:
> >
> > fio version 2 iolog
> > /dev/sdf add
> > /dev/sdf open
> > /dev/sdf write 68719411200 65536
> > /dev/sdf write 131072 65536
> > /dev/sdf write 196608 0
> > /dev/sdf close
> > /dev/sdf open
> > /dev/sdf write 68719411200 65536
> > /dev/sdf write 131072 65536
> > /dev/sdf write 196608 0
> > /dev/sdf close
> > /dev/sdf open
> > /dev/sdf write 68719411200 65536
> > /dev/sdf write 2164391936 65536
> > /dev/sdf write 196608 0
> > /dev/sdf close
> > /dev/sdf open
> > /dev/sdf write 68719411200 65536
> > /dev/sdf write 62356848640 65536
> > /dev/sdf write 131072 65536
> > /dev/sdf write 38256377856 65536
> > /dev/sdf write 57294323712 65536
> > /dev/sdf write 196608 0
> > /dev/sdf close
> > /dev/sdf open
> > /dev/sdf write 68719411200 65536
> > /dev/sdf write 131072 65536
> > /dev/sdf write 196608 0
> > /dev/sdf close
> > /dev/sdf open
>
> Indeed, there's something screwy with the recent builds. Using
> norandommap makes things even worse in fact. I'll audit the whole path,
> it must be the offset. Strange.

OK, can you either repull the git repo or download and try again?

http://brick.kernel.dk/snaps/fio-git-latest.tar.gz

-- 
Jens Axboe