On Wed, Aug 01, 2012 at 08:40:52PM +0200, Hartmut Knaack wrote:
> Hi,
> my impression is, that a kernel version makes it into trunk if it is either a 
> long term kernel, or it brings essential new functions. For 3.3 this was most 
> certainly the introduction of BQL code. Keeping in mind that our main targets 
> are network routers, the bufferbloat issue probably concerns most of the 
> maintainers.

I have been away from my mail for a while.

Van Jacobson gave a great talk about bufferbloat, BQL, codel, and fq_codel
at last week's ietf meeting. Well worth watching. At the end he outlines
the deployment problems in particular.

http://recordings.conf.meetecho.com/Recordings/watch.jsp?recording=IETF84_TSVAREA&chapter=part_3

Far more interesting than this email! 

While we have made great progress towards addressing bufferbloat in openwrt, it
is only barely addressed in the openwrt QoS system, (it would be nice
if fq_codel ran on all interfaces with BQL support), BQL support only exists
for a few drivers in the openwrt tree...

And fixing wifi with similar techniques is going to take months and
months longer to do, even if we could get the chipset makers focused on
the problem.

http://www.bufferbloat.net/projects/cerowrt/wiki/Fq_Codel_on_Wireless

So the upcoming 3.3 release of openwrt is a start, only, towards fixing
bufferbloat. I'm very happy we made so much progress in a year, with
perhaps another year of effort and we'll have it thoroughly licked in linux
and in openwrt. That leads towards difficult decision-making for declaring
a "stable" release this year for openwrt.

The current trunk of openwrt is unquestionably better than last years
release, and deserves wider use.

As for the long term support issue for 3.3, it does concern me very
much that 3.3 is already EOL'd.

I would like it if openwrt planned on a stable release on the next go-round
built around a kernel that has long term support (perhaps 3.8?).

I do note that openwrt's "3.3" is not actually 3.3 to a large extent,
it has wireless-next, and the fq_codel implementation
from linux 3.5, and tons and tons of patches for driver and chipset
support that haven't made it up to mainline.

There are 160 generic linux patches, and for the ar71xx architecture
*alone* there are 136 patches, and 95 new driver related files.

It would be good if efforts were expended to merge up the
enormous patch burden into the kernel mainline, so as to make it
easier to move openwrt forward in conjuction with it.

The patch burden is the enormous problem impeding progress forward, and
makes bi-directional patching from the Linux kernel head much harder as
time goes by...

The patch burden and driver collisions in the arm world got
so bad 2 years ago that linaro was formed to fix it. That effort
has largely been successful and it would help a lot to do something similar
here.

It would be my hope that those that build products around openwrt would
recognise this infrastructural problem and step up.

(side note, the bufferbloat.net effort has loaned 4 boxes to the buildbot
 system and it would be very helpful in terms of cycle time if the buildbot
 cluster could be even more dramatically expanded)

IMHO, I think it's worth creating a stable release now, but also working
hard towards being able to create another stable release on a new kernel
base *this year* - which is going to be hard due to these other problems.


> 
> Emmanuel Deloget schrieb:
> > Hello, 
> >
> > I understand that a lot of effort has been pushed in making 
> > Linux 3.3 the trunk kernel, and I understand that I probably 
> > missed long (IRC?) discussions on this very subject, but since 
> > 3.3.8 is going to be the last supported kernel in the 3.3.x 
> > branch it might be a good idea to move on to another, more 
> > recent kernel version - and to do it slightly better. Not 
> > that anything is really bad, but there were obviously better 
> > choices that 3.3 at the time it came up. 
> >
> 
> _______________________________________________
> openwrt-devel mailing list
> [email protected]
> https://lists.openwrt.org/mailman/listinfo/openwrt-devel
_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/mailman/listinfo/openwrt-devel

Reply via email to