[ath9k-devel] ath9k causes lockups since kernel 2.6.35

[Tony Houghton <h@realh.co.uk>]

Good news below, read on :-).

On Fri, 10 Jun 2011 21:18:37 +0800
Adrian Chadd <adrian@freebsd.org> wrote:

> It's very likely in conjunction with your chipset and kernel build
> options.
> 
> I have a nice selection of AR9285's to choose from too. They all work
> fine in whatever PCIe hardware I have. It's going to be something
> strange with the combination of linux kernel version/build and the
> chipset/bios features.

And/or possibly even something specific to the UK or Europe? I think I
heard there are some subtle differences in the wifi standard depending
on region.

I have tried quite a few different kernel versions, including Ubuntu
(2.6.35) as well as Debian 2.6.37 through -39 and the crash has been
very consistent on everything since 2.6.35 (or 2.6.33 with the "bad"
commit).

> Let's just try the patch I emailed you to apply to wireless-testing.
> Hopefully that restores functionality enough and we can punt it to the
> Atheros guys to do some further investigation.

That change seems to have fixed it! I've managed at least 3
suspend/resumes and an rmmod/modprobe without crashing. Perhaps a little
early to say it's definitely fixed, but I don't think I've ever managed
more than 2 attempts in a row without a "bad" version crashing.

PS I used compat-wireless rather than testing. I tried to build it for
2.6.39-2-686-pae (the name for the Debian stock kernel) while running on
2.6.32, but I couldn't work out how. The kernel's top-level Makefile
offers no clues. From compat-wireless' Makefile I deduced I could set
KLIB=/lib/modules/2.6.39-2-686-pae. It built OK but installed the
modules under
/lib/modules/2.6.39-2-686-pae/lib/modules/2.6.39-2-686-pae/updates
instead of /lib/modules/2.6.39-2-686-pae/updates/ and depmod didn't seem
to recognise that. What variable should I have set?