Re: the " 'official' point of view" expressed by regarding reiser4 inclusion

From: Olivier Galibert
Date: Mon Jul 24 2006 - 11:37:24 EST

On Mon, Jul 24, 2006 at 09:39:39AM -0400, Theodore Tso wrote:
> On Mon, Jul 24, 2006 at 01:35:34PM +0200, Olivier Galibert wrote:
> >Ext patches don't get reviewed much
> > outside of the developpers, and they go in pretty much without
> > discussion in any case, except when Linus blows a fuse.
> Um, you're kidding, right? We certainly don't make the assumption
> that we can violate CodingStyle willy nilly and stuff in yacc grammers
> into ext3 and assume that no one will push back.

I'm not kidding. I do recognise that the ext* maintainers do a very
good and clean job.

> In fact we did a lot of work to make sure the patches were clean and
> mostly ready to be accepted to mainline even before we made the first
> proposal to push extents to LKML.

I'm no talking about extends only. Ext3 now is very, very different
than the ext2+journal it was at the start, with backwards-incompatible
format changes added all the time[1]. These changes went in

> > I think there is something of a problem currently, tough. It is
> > getting too hard to get code in if you're not a maintainer for an
> > existing subsystem (reiser4, suspend2...), and too easy when you're a
> > maintainer (ext4, uswsusp...).
> It's not fair to assume that the only reason why non-maintainers have
> a harder time getting changes is because their changes are getting
> more intensive review.

"only", no, definitively not. The impact is non-negligible though.

> (Although it is the case that we probably do
> need to get better at reviewing changes that go in via git trees.)

Ohh yes, a lot better. Just look at ALSA, most of SNDRV_HWDEP* should
never have gotten in in the first place, especially some recent ones
are security holes the size of Cleveland. I need to continue
documenting the Alsa kernel interface, but I need a new bucket, the
first one is overflowing.

> A much more important effect is that non-maintainers aren't familiar
> with coding and patch submission guidelines. For example, in
> suspend2, Nigel first tried with patches that were too monolithic,
> and then his next series was too broken down such that it was too
> hard to review (and "git bisect" wouldn't work).

All his submissions since 2004 or so? It's a little easy to limit
oneself to the last two ones.

> And of course, there are people who assume that the rules shouldn't
> apply to their filesystem...

It may be a little hard to remove XFS at that point though... And,
while it's not a filesystem, I'd love to be pointed to the technical
discussion deciding whether uswsusp is a good idea.


[1] All optional, I know.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at