TCP TSQ regression affecting ath10k throughput

[Kalle Valo]

Hi,

Sujith reported a regression in TCP which caused a major decrease in
ath9k and ath10k throughput:

http://thread.gmane.org/gmane.linux.network/290269

The commit which broke this is c9eeec26e32e ("tcp: TSQ can use a dynamic
limit") and this patch seems to fix it:

http://patchwork.ozlabs.org/patch/290672/

I have cherry picked the fix to ath.git master branch to make it easier
for ath10k developers.

-- 
Kalle Valo