2003-07-14 12:50:13

by Alan

[permalink] [raw]
Subject: -- END OF BLOCK --


That ends the push of -ac contributions that want adding now. There is a
chunk of quota stuff that vanished with hch's merge that looks important
as it avoids deadlocks in ext3. Perhaps Christoph can take a glance at
the ext3/super differences and the related code ?

I've resynched -ac to the quota code in pre5 and added the automatic
quota loader on top again.


2003-07-14 12:47:15

by Alan

[permalink] [raw]
Subject: Re: -- END OF BLOCK --

On Llu, 2003-07-14 at 13:50, Marcelo Tosatti wrote:
> > I've resynched -ac to the quota code in pre5 and added the automatic
> > quota loader on top again.
>
> And the deadlock avoidance patches too right?
>
> Would you mind sending me the automatic quota loader diff and the deadlock
> avoidance diff ?

The merge is non trivial. I'll let Christoph sort the mess out since I don't
have time to waste on it, and the old -ac quota code works perfectly well for
me.

2003-07-14 12:41:29

by Marcelo Tosatti

[permalink] [raw]
Subject: Re: -- END OF BLOCK --



On Mon, 14 Jul 2003, Alan Cox wrote:

>
> That ends the push of -ac contributions that want adding now.

I've applied some of the patches already, will keep looking at them now...

Thanks.

> There is a chunk of quota stuff that vanished with hch's merge that
> looks important as it avoids deadlocks in ext3. Perhaps Christoph can
> take a glance at the ext3/super differences and the related code ?

Christoph?

> I've resynched -ac to the quota code in pre5 and added the automatic
> quota loader on top again.

And the deadlock avoidance patches too right?

Would you mind sending me the automatic quota loader diff and the deadlock
avoidance diff ?


2003-07-14 12:54:14

by Marcelo Tosatti

[permalink] [raw]
Subject: Re: -- END OF BLOCK --



On Mon, 14 Jul 2003, Alan Cox wrote:

> On Llu, 2003-07-14 at 13:50, Marcelo Tosatti wrote:
> > > I've resynched -ac to the quota code in pre5 and added the automatic
> > > quota loader on top again.
> >
> > And the deadlock avoidance patches too right?
> >
> > Would you mind sending me the automatic quota loader diff and the deadlock
> > avoidance diff ?
>
> The merge is non trivial. I'll let Christoph sort the mess out since I don't
> have time to waste on it, and the old -ac quota code works perfectly well for
> me.

Okay.

The quota code you have in -ac is new Jan Kara's stuff (which supports
both formats, etc) plus the ext3 deadlock avoidance and the compat stuff ?

Christoph, for what reason have you removed the ext3 deadlock avoidance
patches? And also, what else have you changed wrt originals Jan code ?

Lets have -pre6 soon to fix up this mess (which is causing DEADLOCKS), ok?

2003-07-14 17:32:28

by Jan Kara

[permalink] [raw]
Subject: Re: -- END OF BLOCK --

> On Mon, 14 Jul 2003, Alan Cox wrote:
>
> > On Llu, 2003-07-14 at 13:50, Marcelo Tosatti wrote:
> > > > I've resynched -ac to the quota code in pre5 and added the automatic
> > > > quota loader on top again.
> > >
> > > And the deadlock avoidance patches too right?
> > >
> > > Would you mind sending me the automatic quota loader diff and the deadlock
> > > avoidance diff ?
> >
> > The merge is non trivial. I'll let Christoph sort the mess out since I don't
> > have time to waste on it, and the old -ac quota code works perfectly well for
> > me.
>
> Okay.
>
> The quota code you have in -ac is new Jan Kara's stuff (which supports
> both formats, etc) plus the ext3 deadlock avoidance and the compat stuff ?
The deadlock avoidance, quota module autoload etc. are patches which
are not existing for a long time (two weeks or so) and they were merged
just into the last 2.5 kernel so it's not Christophs fault he didn't
send them to you. Actually I just wanted to send them to you myself today...
Anyway now it seems everything is sorted out.

Honza

--
Jan Kara <[email protected]>
SuSE CR Labs

2003-07-14 18:30:30

by Alan

[permalink] [raw]
Subject: Re: -- END OF BLOCK --

On Llu, 2003-07-14 at 18:41, Jan Kara wrote:
> just into the last 2.5 kernel so it's not Christophs fault he didn't
> send them to you. Actually I just wanted to send them to you myself today...
> Anyway now it seems everything is sorted out.

Yes I think so. Its also somewhat cleaner than the older -ac code which is
nice.