[Cerowrt-devel] cerowrt-3.10.32-1 dev build NOT released due to 5ghz wifi issues

[Dave Taht <dave.taht@gmail.com>]

DON'T get it at:

http://snapon.lab.bufferbloat.net/~cero2/cerowrt/wndr/bad/3.10.32-1

Tested for all of an hour  (as an interior gateway, not external).

Summary: We still have issues with the 5.x ghz channels

On the plus side, a device (Android nexus 7 2.4ghz) works now when
before it didn't. On the minus side another device (nexus 4 5ghz)
doesn't. No RX is ever seen
on that channel...

To keep my facts straight...

+ sync with openwrt head (with all changes pushed to date)
+ diff added back into system
+ radsecproxy added as an optional package (allows for enterprise wifi
logins securely)
+ updated kernel to current stable
+ SQM does more of the right thing with "target" at low bandwidths, has
   a few other tweaks, IMHO is nearly ready to go upstream to openwrt
+ ton of ath9k related fixes

- HT20 is still the default for wireless 5ghz.
+/- package signing is being reworked
+/- source specific dns stuff in there but not integrated with netifd
- no bcp38 still (help?)

I'm hoping we'll soon be able to call the kernel bits of this thing
"stable". I was hoping we've nailed the last of the major kernel bugs
at this point.
the wifi fixes looked good in theory...

https://dev.openwrt.org/changeset/39688
https://dev.openwrt.org/ticket/14092

NOTE: I'm out of time to work on this for the week, probably.

I will be doing some benchmarking of 3.10.28-16 but have to get on a
plane for england tuesday, have a lecture at Queen Mary college in
London thursday, and ietf
the week following (and another lecture at Cambridge the week after),
still have to pack, write a bunch of things, etc. I hope we get the
mdns hybrid proxy
and hnetd issues sorted at ietf.

I'd also like to be able to test things like huawei 3g devices
(commonly available in the EU, but not the US), am not sure all the required
modules are built.



-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html