From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-iy0-f171.google.com (mail-iy0-f171.google.com [209.85.210.171]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (Client CN "smtp.gmail.com", Issuer "Google Internet Authority" (verified OK)) by huchra.bufferbloat.net (Postfix) with ESMTPS id BDBCE20067D for ; Mon, 19 Dec 2011 01:50:12 -0800 (PST) Received: by iagw33 with SMTP id w33so10231307iag.16 for ; Mon, 19 Dec 2011 01:50:11 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=gha4+F71MrfvOkgU/p5hPeW29qgz0m7ldRmld1hrJIE=; b=Ub6vYzF2C/Xx1tHUCEEEkDPkRJOTzTN3oDo0ah/ueWtteoCVBbIzd2yAl8k9KFB6K+ bWAidir5aY7gD8lANN62qwcnMxz/YGkIFAj4avk6jP7Vm0x9K8JARhWVE3cYRYlPVrRp OMabvrQxhVhtjo3zpr9Y8/IKqx/19bC49ahlE= MIME-Version: 1.0 Received: by 10.50.85.136 with SMTP id h8mr26596622igz.56.1324288210420; Mon, 19 Dec 2011 01:50:10 -0800 (PST) Received: by 10.231.204.83 with HTTP; Mon, 19 Dec 2011 01:50:10 -0800 (PST) Date: Mon, 19 Dec 2011 10:50:10 +0100 Message-ID: From: Dave Taht To: cerowrt-devel@lists.bufferbloat.net Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Subject: [Cerowrt-devel] cerowrt-bql-3 available X-BeenThere: cerowrt-devel@lists.bufferbloat.net X-Mailman-Version: 2.1.13 Precedence: list List-Id: Development issues regarding the cerowrt test router project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 19 Dec 2011 09:50:13 -0000 is up at: http://huchra.bufferbloat.net/~cero1/bql-smoketests/bql-3/ Although this is named bql-3, I have yet to successfully backport byte queue limits to 3.1.5 - (or if have, I managed to break something somewhere else) so that is not in this. I may well defer the bql work to Q2 after it stabilizes in mainline. I definitely need to come up with better names than 'rc' or 'smoketest'. Current Features kernel 3.1.5 - notably this has Eric's fix to RED (Linux RED implementation had been broken for 3 years) - ALL known tcp congestion control algorithms - RED, QFQ, CBQ, SFB, DRR schedulers are available - iproute2 and iptables updated for 3.1 - ESFQ is removed - radios work correctly now on boot - clock interrupt (temporarily?) increased to 256 from 100. - tcp timestamping is now enabled - isc ntp and named (temporarily?) replaced with dnsmasq and busybox ntp - samba361 is in there but totally underconfigured - wins is supplied via dhcp. - dscp improvements for wireless temporarily removed - iperf-mt and netperf is installed by default - ahcp still fails to start - wndr3700v2 and wndr3800 fully supported - wndrmacv2 is also built but untested Getting ntp and the radios working on boot were the two big bugs left in rc7-smoketest10. However I did not intend to 'solve' those problems in this way and losing esfq and the dscp hacks is a regression on the qos system, and I already miss having named... I have spent a bunch of time working on cleaning up the patch set, and will be submitting them upstream over the next few weeks, then rebasing and making sources available again. the rc7 sources will remain available, but they no longer apply on top of openwrt (which is why I'm rebasing) I also intend to move netperf forward to svn head so as to be able to better test alternate congestion control algorithms. This is why iperf is also now in by default, as I find using them both is workwhile... I also plan a custom version of rsync, with diffserv and congestion algorithm support... As for what else, going forward...? Been posting my thoughts here in the hope of flushing out more feedback as to what was useful and what would be good, going forward... so speak up, please!!! As for me, I've been working primarily on coming up with a long term plan for the bufferbloat projects (note the s!), and on bugs #306, and #305. I note either there's a bug in the driver, QFQ, pfifo_drop_head, or new scripts that can cause the interface to lock up after an hour or so of saturating load. I can't (thus far) convince the interface to lock up otherwise... so it seems stable, at present, without bql or those new scripts, and that's what's up on the website now. I'd like it very much if other folk experimented with shapers and schedulers as well. Getting a DRR script going, fiddling with wondershaper, rolling your own... - I hope we can have a bakeoff this coming year, of shapers and AQMs. If you get something reasonable please submit to: http://www.bufferbloat.net/issues/301 (PLEASE NOTE: you can prototype this sort of work just fine on an x86 box) My own principal motivation for the 'bql' release at the moment is to thoroughly exercise the packet schedulers and AQMs - not to get the AQM/scheduling thing 'right' - but to expose any more bugs that may lurk within this neglected subsystems. I've been tossing scripts as they stabilize into my deBloat repo on github, however I've decided to rewrite everything into lua... which is going to take a while. Anyway - Install this 'smoketest' at your own risk, ave fun inside of the above limitations... and please please please don't experiment on wives or small children. I would like it very much if the samba/wins issues could be put to bed, and the qos system brought forward to work correctly with the new improved sfq system. (there was a patch for that I seem to have misplaced) There are tons of bugs left to look at and prioritize on the bug list, too. I look forward to hearing about any new ones. happy hacking, and have a Merry Christmas. --=20 Dave T=E4ht SKYPE: davetaht US Tel: 1-239-829-5608 FR Tel: 0638645374 http://www.bufferbloat.net