On Tue, 2010-10-12 at 16:56 -0400, John W. Linville wrote:
On Tue, Oct 12, 2010 at 09:11:50PM +0100, Matthew Garrett wrote:
> On Tue, Oct 12, 2010 at 03:37:28PM -0400, Jon Masters wrote:
> > Sure, it'll hopefully land in staging and be in the rawhide kernel in
> > due course, but given that I'm lazy and bought the netbook in question
> > specifically to run rawhide(!), would there be huge objections to a
> > patch sooner that pulled in this driver until it's there officially?
> > Sure, it's experimental, but it's self-contained and this is rawhide.
> > Let me know if this would be just a silly/bad idea.
> Pretty much. We're reluctant to enable anything in staging until it's
> clear that it's going to form the basis of ongoing development (rather
> than being neglected and b43 gaining support), and we'd want an
> indication that John Linville's willing to deal with the bugs. Breaking
> suspend/resume is an absolute dealbreaker at present.
I would prefer not to 'jump the gun', for the reasons Matthew stated.
Agreed. I would just like to clarify for Matthew (since you may not have
this hardware amongst your collection of fun gadgets) that this doesn't
actually break suspend per se. Everything else works, but the wireless
interface doesn't come up cleanly. There's no crash though, so I'd argue
you go from having no WiFi to WiFi only if you don't suspend. Not really
a regression, just not an entirely polished new feature.
Anyway. Good to know what the "staging policy" is.
That said, I would strongly encourage you (i.e. jcm) and anyone else
to continue contributing to brcm80211 in order to help it transition
to drivers/net/wireless ASAP. If you would prefer to use brcm80211
as a hardware guide for improving b43, that's find by me too. :-)
I intend to. At this moment I consider that driver a nice gesture (which
is why I want to help), but it is very unpolished, has a lot of obvious
bugs (from visual inspection) and is clearly based on some generic stuff
from Broadcom they are using in other drivers (messy). All that said, I
think it is a great thing to see, I am able to use the driver already,
and if I can fix suspend/resume too then it'll be good enough for me to
use. At that point, I think it'll be great to see wider testing.
I don't know any of the b43 folks so I'm not sure what their views would
be about pulling in such features. I do think Broadcom should be
encouraged as much as possible, especially if they're going to keep
these three or four people tasked with Open Linux drivers :)
P.S. One thing I do enjoy about your stack is that I didn't really know
it that well before last night but was able to go over the core code
fairly quickly and figure out enough to see how a driver is supposed to
be written. This will be a nice little learning exercise.