regression: tethering fails in 3.5 with iwlwifi

Hi Johannes,

We are using iwlwifi driver in Tizen IVI. We were using kernel v3.4 and
wifi tethering worked fine. But after upgrading to v3.5 it stopped
working.

I am trying to bisect this already many hours - very difficult, because
tethering did not work in kernels around v3.4-rcX for some different
reasons, so just 'git bisect' does not work.

Anyway, what I found out is:

bad:  cb62ab7 Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
good: 31ed8e6 Merge branch 'dentry-cleanups' (dcache access cleanups and optimizations)

I spent a lot of time trying to bisect - tough, no luck. Then I the
following trick.

I reset to the last good commit: 31ed8e6

Now I know that Dave's tree broke tethering (merge cb62ab7 by Linus).

I start _partioally_ merge Dave's tree into 31ed8e6.

I found out this way the following:

bad:   816a785 David S. Miller Merge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next
good:  011e3c6 David S. Miller Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Note, the second (good) line means that if you merge 011e3c6 to 31ed8e6
like this:

git reset --hard 31ed8e6
git merge 011e3c6

tethering works fine. The merge has no conflicts.

But if you merge 816a785 into 31ed8e6 like this:

git reset --hard 31ed8e6
git merge 011e3c6

then tethering does not work. There are non-trivial merge conflicts
though. Note also, that here I had to apply

bb3e10f mac80211: check IEEE80211_HW_QUEUE_CONTROL in ieee80211_check_queues()

because without this patch there was an WARN() and nothing worked at
all.

This all means that one of the following broke tehering:

1. David's merge of the wireless tree which contains a lot of your
   iwlwifi changes.
2. The conflicts resolution.

How I resolved conflicts? I actually did the same David did it his
merge:

0d6c4a2 Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

This is maximum I can do without considerable amount of effort. Any idea
what would the bug be?

Could you also take a look at the conflict resolutions for iwlwifi in
0d6c4a2?

Thanks a lot!

-- 
Best Regards,
Artem Bityutskiy