Re: Short update and pause in 2.6.27 merge window

From: J. Bruce Fields
Date: Fri Jul 18 2008 - 14:58:50 EST

On Thu, Jul 17, 2008 at 09:47:45AM -0700, Linus Torvalds wrote:
> This is just a quick note to let people know that I'll be off for an
> extended weekend starting later today, so the next few days will be very
> quiet from a merge standpoint.
> Feel free to send the merge requests, just don't get worried when I don't
> act on them ;)
> (There's also a few merge requests that I didn't act on yet just because I
> wanted to take a closer look - but if you're worried about it there is
> nothing wrong with re-sending it. I _do_ just drop requests in the
> bottomless pit that is my mailbox at times)
> We seem to be roughly half-way through the merge window, judging purely by
> number of commits, although part of my calculations there is that I'm
> actually hoping/expecting that because a number of people are on vacation
> etc, this merge window might be smaller than some of the other recent ones
> (2.6.25-rc1 in particular was huge).
> In the last couple of days I _have_ merged 50+ trees, and while there's
> been some 'heated discussion' about some of them (you know who you are ;),
> I'm hoping that we're actually in reasonably good shape even though it's
> in the middle of the merge window, and that people will test out the
> snapshot kernels even though I'm not ready to do a -rc1 release.
> So go out and test.
> The fact that much of it has been in linux-next may or may not have
> helped, but it definitely meant that people knew of _some_ problems early.
> Let's see how people feel about that after the whole release is done, but
> it certainly doesn't seem to have hurt.
> My personal favorite merge is the BKL pushdown, which while not likely to
> matter hugely in practice for lots of people is nice to finally see. We
> kind of dropped the ball on the BKL once it got to be small enough that it
> wasn't very high on peoples radar any more. Now the most noticeable user
> of the BKL in "core" kernel code seems to be fs/locks.c.
> Hint, hint. I think somebody had some patches for that one too. Matthew?

I recall there being worries about Matthew's patch, but I don't think
anyone thought them through carefully. For example, the lockd thread
currently spends its entire life under the BKL, so we wondered if there
could be undocumented dependencies on lock.c's use of the BKL. But that
may not really be a problem. I'll set aside some time to look at this
if noone else beats me to it....


> But we've had lots of other things (in fact, the BKL changes ended up
> being much smaller than I expected), and the dirstat so far is
> 4.7% arch/arm/
> 3.3% arch/mips/
> 3.5% arch/ppc/platforms/
> 14.9% arch/ppc/
> 3.3% arch/x86/
> 29.3% arch/
> 8.3% drivers/char/drm/
> 9.3% drivers/char/
> 7.3% drivers/gpu/drm/
> 7.3% drivers/gpu/
> 3.3% drivers/s390/
> 10.4% drivers/usb/misc/
> 6.0% drivers/usb/serial/
> 17.2% drivers/usb/
> 45.0% drivers/
> 7.1% firmware/
> 4.3% fs/ubifs/
> 5.0% fs/
> 6.7% include/
> 3.3% sound/
> where some of those percentages are a bit misleading (the "gpu" part, for
> example, goes away if you enable renames, since that was all really just
> moving files around from drivers/char/ - but at the same time it's one of
> the things people may notice more, so I chose the non-rename version of
> dirstat on purpose).
> No network or driver core/USB merges yet (the USB stuff in the dirstat is
> from the ARM, BKL and firmware merges) so there are complete big
> subsystems lacking still, but it would be good to have people test the
> things we _have_ merged.
> Linus
> PS. And to get wider distribution for this message: Digg users - you're
> all a bunch of Wanking Walruses. And you can quote me on that.
> --
> 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
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