[Cerowrt-devel] cerowrt 3.8.8-4 released

[Cerowrt-devel] cerowrt 3.8.8-4 released

[Dave Taht]

+ Refresh to openwrt barrier breaker head

   this now contains nearly all the patches formerly separately in cerowrt!

   ++ fq_codel is on by default on ALL interfaces with default quantum of 300
        (yes, openwrt has obsoleted pfifo_fast!)
   ++ unaligned access patches, etc, etc
   + dhcp-pd SERVER support
 the usual multitude of other openwrt fixes... all tested extensively
at the battlemesh conference.

+ Update to dnsmasq 2.67test2

Toke got really busy in building his own version of cero and adding

+ AQM scripts and gui
+ tahoe-lafs added (untested)
+ uftp4 updated

- no upnp/ssdp fix because I'm clueless

Yep, "AQM" gui now. /deep hat tip to toke for writing that. No need to
fiddle with any scripts now....

I guess of largest import here, underlying the gui, is that I had a
chance to deeply look at simple_qos.sh. There are several bugs in it,
that appear actually in the underlying tc subsystem.On ingress, the
priority queue was not being used properly, and a few other things
were odd in ipv6.

Out of frustration with that, and based on the data we'd got back from
cablelabs, I wrote the simplest possible rate limited fq_codel
implementation, using a reduced number of flows and a single tier of
htb only - and that works surprisingly well.

It's called "Simplest" in the new AQM screen. Give it a shot.

Secondly, the exercise of writing that and comparing it to the simple_qos script

.... seems to have exposed a bug in htb at low bandwidths.

http://snapon.lab.bufferbloat.net/~d/dsl384k-htb-bug.svg

Here I was trying at 384k up and 8Mbit down, and the middle portion of
the upload graph there is kind of... impossible. At higher rates I
haven't seen this happen.

-- 
Dave Täht

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

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Dave Taht]

On Wed, Apr 24, 2013 at 1:55 AM, Dave Taht <dave.taht@gmail.com> wrote:
> + Refresh to openwrt barrier breaker head
>
>    this now contains nearly all the patches formerly separately in cerowrt!
>
>    ++ fq_codel is on by default on ALL interfaces with default quantum of 300
>         (yes, openwrt has obsoleted pfifo_fast!)
>    ++ unaligned access patches, etc, etc
>    + dhcp-pd SERVER support
>  the usual multitude of other openwrt fixes... all tested extensively
> at the battlemesh conference.
>
> + Update to dnsmasq 2.67test2
>
> Toke got really busy in building his own version of cero and adding
>
> + AQM scripts and gui
> + tahoe-lafs added (untested)
> + uftp4 updated
>
> - no upnp/ssdp fix because I'm clueless
>
> Yep, "AQM" gui now. /deep hat tip to toke for writing that. No need to
> fiddle with any scripts now....
>
> I guess of largest import here, underlying the gui, is that I had a
> chance to deeply look at simple_qos.sh. There are several bugs in it,
> that appear actually in the underlying tc subsystem.On ingress, the
> priority queue was not being used properly, and a few other things
> were odd in ipv6.
>
> Out of frustration with that, and based on the data we'd got back from
> cablelabs, I wrote the simplest possible rate limited fq_codel
> implementation, using a reduced number of flows and a single tier of
> htb only - and that works surprisingly well.
>
> It's called "Simplest" in the new AQM screen. Give it a shot.
>
> Secondly, the exercise of writing that and comparing it to the simple_qos script
>
> .... seems to have exposed a bug in htb at low bandwidths.
>
> http://snapon.lab.bufferbloat.net/~d/dsl384k-htb-bug.svg
>
> Here I was trying at 384k up and 8Mbit down, and the middle portion of
> the upload graph there is kind of... impossible. At higher rates I
> haven't seen this happen.


And of course, you can get it from:

http://snapon.lab.bufferbloat.net/~cero2/cerowrt/wndr/3.8.8-4/

-- 
Dave Täht

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

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Maciej Soltysiak]

[-- Attachment #1: Type: text/plain, Size: 405 bytes --]

On Wed, Apr 24, 2013 at 10:55 AM, Dave Taht <dave.taht@gmail.com> wrote:

>    ++ fq_codel is on by default on ALL interfaces with default quantum of
> 300
>
Including loopback?

>          (yes, openwrt has obsoleted pfifo_fast!)
>
Yeah, if this gets picked up by regular distros, looks like some portions
of LARTC (lartc.org) might be needing an update soon :-)

Off to download latest cero :-)

Maciej

[-- Attachment #2: Type: text/html, Size: 1068 bytes --]

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Dave Taht]

[-- Attachment #1: Type: text/plain, Size: 1072 bytes --]

I should point out cero still defaults to nfq codel via the debloat script.
The setting  of the qdisc part now has to get separated from the ethtool
and bql settings part.  Or there needs to be more surgery in OpenWrt to do
the debloating of those two subsystems in kernel space also.

I still lack a good picture for any real differences between nfq and fq
codel and having lost faith in htb temporarily... Just don't know. But I
would like it if people tried the new OpenWrt defaults rather than ceros
which involves commenting out some stuff in the debloat script.
On Apr 24, 2013 10:07 AM, "Maciej Soltysiak" <maciej@soltysiak.com> wrote:

> On Wed, Apr 24, 2013 at 10:55 AM, Dave Taht <dave.taht@gmail.com> wrote:
>
>>    ++ fq_codel is on by default on ALL interfaces with default quantum of
>> 300
>>
> Including loopback?
>
>>          (yes, openwrt has obsoleted pfifo_fast!)
>>
> Yeah, if this gets picked up by regular distros, looks like some portions
> of LARTC (lartc.org) might be needing an update soon :-)
>
> Off to download latest cero :-)
>
> Maciej
>

[-- Attachment #2: Type: text/html, Size: 1981 bytes --]

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Rich Brown]

Folks,

Good news from the (limited) testing that I do. I installed 3.8.8-4 on my WNDR3800 without incident. I also ran the two configuration scripts in the wiki to set the configuration to set the timezone, put a password on my wireless and set up the IPv6 tunnel.

Thanks to Dave and Toke for adding the Network/AQM tab to the GUI. As Dave says, "No more fiddling with scripts". Any words of wisdom for choosing one of  "simple.qos" and "simplest.qos" over the other?

Finally, the About… and other pages at http://gw.home.lan/cerowrt/ have been updated to talk about this release. 

Nice work, Dave, Toke, and everyone else who contributed. 

Rich

Snag the binary at: http://snapon.lab.bufferbloat.net/~cero2/cerowrt/wndr/3.8.8-4/
See the automated scripts at: 
	http://www.bufferbloat.net/projects/cerowrt/wiki/Automated_Configuration_of_CeroWrt
	http://www.bufferbloat.net/projects/cerowrt/wiki/IPv6_Tunnel

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Dave Taht]

simplest most closely models the ns2 testing cablelabs did. By being
simpler, it has less bugs.

So if you could try that under a variety of workloads, particularly
with bittorrent I think it would be interesting... I still think the
more complex script is closer to the right thing in that case, but
lack data.

On Wed, Apr 24, 2013 at 5:06 PM, Rich Brown <richb.hanover@gmail.com> wrote:
> Folks,
>
> Good news from the (limited) testing that I do. I installed 3.8.8-4 on my WNDR3800 without incident. I also ran the two configuration scripts in the wiki to set the configuration to set the timezone, put a password on my wireless and set up the IPv6 tunnel.
>
> Thanks to Dave and Toke for adding the Network/AQM tab to the GUI. As Dave says, "No more fiddling with scripts". Any words of wisdom for choosing one of  "simple.qos" and "simplest.qos" over the other?
>
> Finally, the About… and other pages at http://gw.home.lan/cerowrt/ have been updated to talk about this release.

You know I totally forgot about that. Didn't you have a hand in that? :)

I'd meant to update the credits file on this go-round. Next time.
> Nice work, Dave, Toke, and everyone else who contributed.
>
> Rich
>
> Snag the binary at: http://snapon.lab.bufferbloat.net/~cero2/cerowrt/wndr/3.8.8-4/
> See the automated scripts at:
>         http://www.bufferbloat.net/projects/cerowrt/wiki/Automated_Configuration_of_CeroWrt
>         http://www.bufferbloat.net/projects/cerowrt/wiki/IPv6_Tunnel



-- 
Dave Täht

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

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Toke Høiland-Jørgensen]

[-- Attachment #1: Type: text/plain, Size: 737 bytes --]

Rich Brown <richb.hanover@gmail.com> writes:

> Thanks to Dave and Toke for adding the Network/AQM tab to the GUI. As
> Dave says, "No more fiddling with scripts". Any words of wisdom for
> choosing one of "simple.qos" and "simplest.qos" over the other?

Basically, "simplest.qos" is "simple.qos" without explicit diffserv
handling, leaving prioritisation in the hands of fq_codel's thin/thick
flow logic. Dave found some bugs in the diffserv matching code, which
led to simplest.qos as an alternative approach which seems to perform
well. So I would suggest trying out simplest.qos and seeing if that
works well for your workload. If it doesn't, see if simple.qos is
better. Bonus points for posting details on the difference :)

-Toke

[-- Attachment #2: Type: application/pgp-signature, Size: 489 bytes --]

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Maciej Soltysiak]

[-- Attachment #1: Type: text/plain, Size: 644 bytes --]

On Thu, Apr 25, 2013 at 7:56 AM, Dave Taht <dave.taht@gmail.com> wrote:

> simplest most closely models the ns2 testing cablelabs did. By being
> simpler, it has less bugs.
>
> So if you could try that under a variety of workloads, particularly
> with bittorrent I think it would be interesting... I still think the
> more complex script is closer to the right thing in that case, but
> lack data.

One thing I've noticed recently is that Bitcoin clients generate a bit
disruptive traffic.
Especially when synchronizing the blockchain.

I've yet to tcpdump/netstat this and see if classification could help as
with bittorrent.

Regards,
Maciej

[-- Attachment #2: Type: text/html, Size: 1005 bytes --]

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Maciej Soltysiak]

[-- Attachment #1: Type: text/plain, Size: 2831 bytes --]

Hi Dave,

I installed 3.8.8-4 more than a month ago but it took me this long to give
up  trying to get DLNA working.
For some reason those DLNA discovery (SSDP) multicasts seem not forward
between se00 and sw00 or sw10 so the devices cannot find themselves. This
is supposed to be done by minissdpd.

I know previously we've had it broken but you fixed it by adding some
missing config in the kernel for IGMP I believe.
But I'm unable to find the cause this time.

Would you suggest I have a go with a more recent cero release or does
anything come to mind when I describe the issue?

Best regards,
Maciej Soltysiak

On Wed, Apr 24, 2013 at 10:55 AM, Dave Taht <dave.taht@gmail.com> wrote:

> + Refresh to openwrt barrier breaker head
>
>    this now contains nearly all the patches formerly separately in cerowrt!
>
>    ++ fq_codel is on by default on ALL interfaces with default quantum of
> 300
>         (yes, openwrt has obsoleted pfifo_fast!)
>    ++ unaligned access patches, etc, etc
>    + dhcp-pd SERVER support
>  the usual multitude of other openwrt fixes... all tested extensively
> at the battlemesh conference.
>
> + Update to dnsmasq 2.67test2
>
> Toke got really busy in building his own version of cero and adding
>
> + AQM scripts and gui
> + tahoe-lafs added (untested)
> + uftp4 updated
>
> - no upnp/ssdp fix because I'm clueless
>
> Yep, "AQM" gui now. /deep hat tip to toke for writing that. No need to
> fiddle with any scripts now....
>
> I guess of largest import here, underlying the gui, is that I had a
> chance to deeply look at simple_qos.sh. There are several bugs in it,
> that appear actually in the underlying tc subsystem.On ingress, the
> priority queue was not being used properly, and a few other things
> were odd in ipv6.
>
> Out of frustration with that, and based on the data we'd got back from
> cablelabs, I wrote the simplest possible rate limited fq_codel
> implementation, using a reduced number of flows and a single tier of
> htb only - and that works surprisingly well.
>
> It's called "Simplest" in the new AQM screen. Give it a shot.
>
> Secondly, the exercise of writing that and comparing it to the simple_qos
> script
>
> .... seems to have exposed a bug in htb at low bandwidths.
>
> http://snapon.lab.bufferbloat.net/~d/dsl384k-htb-bug.svg
>
> Here I was trying at 384k up and 8Mbit down, and the middle portion of
> the upload graph there is kind of... impossible. At higher rates I
> haven't seen this happen.
>
> --
> Dave Täht
>
> Fixing bufferbloat with cerowrt:
> http://www.teklibre.com/cerowrt/subscribe.html
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>

[-- Attachment #2: Type: text/html, Size: 3731 bytes --]

Re: [Cerowrt-devel] cerowrt 3.8.8-4 released

[Dave Taht]

[-- Attachment #1: Type: text/plain, Size: 3903 bytes --]

I note that 3.8.X has been kind of a disappointing kernel series
(notably the ADSL bug) and I'm not intending to ship a cero based on it.
Please feel free to keep hacking cero forward or reverting to Modena.

Also, after my very long trip around the eu...

http://www.youtube.com/watch?v=quAaZKBHvs8

I'm taking a *vacation*. This will be my last read email for at least a
week.

In addition to the ongoing evals of other ath9k based boards...

I have a ton of boards like these on order or backorder:

http://linuxgizmos.com/wandboard-gains-quad-core-cpu-and-enhanced-gpu/

http://linuxgizmos.com/beaglebone-black-climbs-the-device-tree-with-linux-3-8/

On Sun, Jun 2, 2013 at 11:00 AM, Maciej Soltysiak <maciej@soltysiak.com>wrote:

> Hi Dave,
>
> I installed 3.8.8-4 more than a month ago but it took me this long to give
> up  trying to get DLNA working.
> For some reason those DLNA discovery (SSDP) multicasts seem not forward
> between se00 and sw00 or sw10 so the devices cannot find themselves. This
> is supposed to be done by minissdpd.
>

Have you tried disabling the firewall rules entirely?


>
> I know previously we've had it broken but you fixed it by adding some
> missing config in the kernel for IGMP I believe.
> But I'm unable to find the cause this time.
>
>
I am puzzled. IGMP was working in 3.7


> Would you suggest I have a go with a more recent cero release or does
> anything come to mind when I describe the issue?
>
> Best regards,
> Maciej Soltysiak
>
> On Wed, Apr 24, 2013 at 10:55 AM, Dave Taht <dave.taht@gmail.com> wrote:
>
>> + Refresh to openwrt barrier breaker head
>>
>>    this now contains nearly all the patches formerly separately in
>> cerowrt!
>>
>>    ++ fq_codel is on by default on ALL interfaces with default quantum of
>> 300
>>         (yes, openwrt has obsoleted pfifo_fast!)
>>    ++ unaligned access patches, etc, etc
>>    + dhcp-pd SERVER support
>>  the usual multitude of other openwrt fixes... all tested extensively
>> at the battlemesh conference.
>>
>> + Update to dnsmasq 2.67test2
>>
>> Toke got really busy in building his own version of cero and adding
>>
>> + AQM scripts and gui
>> + tahoe-lafs added (untested)
>> + uftp4 updated
>>
>> - no upnp/ssdp fix because I'm clueless
>>
>> Yep, "AQM" gui now. /deep hat tip to toke for writing that. No need to
>> fiddle with any scripts now....
>>
>> I guess of largest import here, underlying the gui, is that I had a
>> chance to deeply look at simple_qos.sh. There are several bugs in it,
>> that appear actually in the underlying tc subsystem.On ingress, the
>> priority queue was not being used properly, and a few other things
>> were odd in ipv6.
>>
>> Out of frustration with that, and based on the data we'd got back from
>> cablelabs, I wrote the simplest possible rate limited fq_codel
>> implementation, using a reduced number of flows and a single tier of
>> htb only - and that works surprisingly well.
>>
>> It's called "Simplest" in the new AQM screen. Give it a shot.
>>
>> Secondly, the exercise of writing that and comparing it to the simple_qos
>> script
>>
>> .... seems to have exposed a bug in htb at low bandwidths.
>>
>> http://snapon.lab.bufferbloat.net/~d/dsl384k-htb-bug.svg
>>
>> Here I was trying at 384k up and 8Mbit down, and the middle portion of
>> the upload graph there is kind of... impossible. At higher rates I
>> haven't seen this happen.
>>
>> --
>> Dave Täht
>>
>> Fixing bufferbloat with cerowrt:
>> http://www.teklibre.com/cerowrt/subscribe.html
>> _______________________________________________
>> Cerowrt-devel mailing list
>> Cerowrt-devel@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>>
>
>


-- 
Dave Täht

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

[-- Attachment #2: Type: text/html, Size: 5687 bytes --]