Re: [OP] v2.6.22.22-op1

["Oliver Pinter" <oliver.pntr@gmail.com>]

sure, bad branch see I

On 5/1/08, Oliver Pinter <oliver.pntr@gmail.com> wrote:
> On 5/1/08, Jeff Moyer <jmoyer@redhat.com> wrote:
> > <oliver@zipernowsky.hu> writes:
> >
> > > I announced the v2.6.22.22-op1 "stable" kernel.
> >
> > >  fs/aio.c                    |   23 ++++++++++++-----------
> >
> > > Jeff Moyer (1):
> > >       aio: account I/O wait time properly
> >
> > This version of the patch introduced a regression for UML.  It was
> > seeing 100% iowait time when it wasn't waiting on any I/O.  There was a
> > follow-up patch which did the io_schedule conditionally, see commit id:
> >
> >   e00ba3dae077f54cfd2af42e939a618caa7a3bca
> >
> > Cheers,
> >
> > Jeff
> >
> > > diff --git a/fs/aio.c b/fs/aio.c
> > > index dbe699e..b3419c5 100644
> > > --- a/fs/aio.c
> > > +++ b/fs/aio.c
> > > @@ -1170,7 +1171,7 @@ retry:
> > >  			ret = 0;
> > >  			if (to.timed_out)	/* Only check after read evt */
> > >  				break;
> > > -			schedule();
> > > +			io_schedule();
> > >  			if (signal_pending(tsk)) {
> > >  				ret = -EINTR;
> > >  				break;
> >
>
>
> hmm, this is the relevant code in 2.6.22:
> ----8<-------------->8-----
>                         if (min_nr <= i)
>                                 break;
>                         ret = 0;
>                         if (to.timed_out)       /* Only check after read evt
> */
>                                 break;
>                         schedule(); <-----------
>                         if (signal_pending(tsk)) {
>                                 ret = -EINTR;
>                                 break;
>                         }
>                         /*ret = aio_read_evt(ctx, &ent);*/
> ----8<-------------->8-----
>
> --
> Thanks,
> Oliver
>


-- 
Thanks,
Oliver