Re: journaling integration in current?

[Michael Lorenz <macallan%netbsd.org@localhost>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Apr 9, 2008, at 09:27, Joerg Sonnenberger wrote:
> On Wed, Apr 09, 2008 at 01:45:14PM +1000, Simon Burge wrote:
> >  - kern/38057.  This has two effects: your root FS can't be journaled
> >    and you need a ffs /dev (ie, you can't use an mfs or tmpfs /dev).
>
> Not a show-stopper for commit IMO.

Yeah, at least my / is usually small enough to fsck in a matter of  
seconds. The bigger partitions that actually contain lots of data are  
what I want journalling for.

> >  - Joerg has seen some panics when bulk-building.  Both have the same
> >    trace:
>
> It wouldn't be the first layerfs issue, so I wouldn't call this a show
> stopper either. I would expect that both are a side effect of some
> concurrency issue. The bulk build is heavily utilising the unlink path
> as well.

I see deadlocks and locking panics from vfs in parallel builds  
without this.

> >  - Joerg and I have both seen processes block on "tstile" and the  
> > system
> >    then slowly grinds to a halt.  At least for me this is with a SMP
> >    machine, I'm not sure what Joerg has.
>
> Me too. I have also seen one instance where the code seems to have  
> been
> spinning inside the kernel.

May or may not be journalling related.

> Nevertheless I do not believe this problems to be show-stoppers for
> merging the code into HEAD.

Agreed. That would certainly get more people to play with it.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBR/zSBMpnzkX8Yg2nAQJjjgf/fCJiadVS/kinMb9PTLv8SXqXLUatH12H
REwjB6Swev37U4u63nT/iqEZjsXVsQZ4kbtv4jU0abImEf7GS1ouNjfRMOYmghtg
8nqWMMdG4QLIiVdF+FcTHqHyWkxNhUhfmdO1a+z/r/ztIEB9UZbA1n9XJ9uKryp4
8a7wPmCEuYxrJkMdpc9kB4gCMrnc4LUGHywuMLJ8qhyDuef4pSqZMu7VBJqaPkaK
kfSlEcJF47s1j+u8662py/AANVS7fmc4CHHR/IxDZ8uiCoDPny/IV3xy3lw3g6EU
gvOFXhh0zP2CPl9xZFNR7A292+vPQ7ayJ39ySIn9pBW4SP8sbhj5Tw==
=zLDK
-----END PGP SIGNATURE-----