Re: Patch for Fio 1.21 plus changes up to de79c9158eae1459970e0df6efc6a2b09d920f15

On Wed, Sep 10 2008, Shaozhi Ye wrote:
> Jens,
>
> Thanks for catching this. I thought I had some problem with it but can
> not reproduce the bug again now. So let's keep it as it is.

Actually, I seem to recall it very recently caught a bug in ext4, so I
guess it's not totally unlikely that it could expose bugs elsewhere. But
that's just more a reason to keep it :-)

>
> Best,
> Shawn
>
> On Tue, Sep 9, 2008 at 11:59 PM, Jens Axboe <jens.axboe_at_oracle.com> wrote:
> >
> > The only one I have an issue with is this:
> >
> > @@ -193,6 +193,9 @@ static int __file_invalidate_cache(struc
> > {
> > int ret = 0;
> >
> > + if (td->o.odirect)
> > + return 0;
> > +
> >
> > since it's actually intentionally killing the cache of an O_DIRECT file.
> > If we have to lay out the file, it just seems prudent to kill off the
> > page cache of that file before we initiate direct IO on it. So can you
> > elaborate on why you made that change?
> >
> > I have committed the rest, thanks!
> >
> > --
> > Jens Axboe
> >
> >
>
>
>
> --
> Shaozhi Ye
> Platforms Software Engineer Intern
> Mountain View 1600 2F8D
> +1 (650) 214-0873
>

-- 
Jens Axboe