[Cerowrt-devel] Any chance for the LUCI GUI for cake appearing?

Sebastian Moeller moeller0 at gmx.de
Fri Jun 26 16:47:49 EDT 2015


Hi Dave, hi List,

On Jun 26, 2015, at 21:48 , Dave Taht <dave.taht at gmail.com> wrote:

> It turns out that I am pulling from openwrt for luci-app-sqm, not
> ceropackages, and overriding it was a headache. Fixed now, see below
> 

	I guess Toke and I need to find a better way of updating the relevant repository… I had thought that ceropackages-3.0 was a relatively confined and hence “safe” testing group, so my idea was if it does worl for the ceopackages crowd, push it onto the openwrt community; but it seems that step is too large an impedance mismatch. So I will try to find a way with Toke to centralize development a bit. That scares me a bit as, knowing my level of detail, there will be lots of fix-up work; on the other hand, changes without people to test them is somewhat lame as well...

> router# cd /tmp; wget
> http://snapon.lab.bufferbloat.net/~cero3/luci-app-sqm_3-5_all.ipk ;
> opkg install ./luci-app-sqm_3-5_all.ipk

	That seems recent. I need to remember to bump so,e number on any relevant change. My target is to change on ever bug fixes and feature additions, but I am not in the best position to judge bug fixes truthfully...

> router # /etc/init.d/uhttpd restart
> 
> if it breaks, you get to keep both pieces, but it worked for me, and
> yea! cake in the gui! thx sebastian!

	Yippi ya ya yippy yippy yeah.

> 
> 
> I tried -
> 
> ./scripts feeds uninstall luci-app-sqm; ./scripts feeds install -p
> cero luci-app-sqm # seemed sane, did not work
> git rm in the feed itself did not work
> doing the uninstall, deleting the symlink, and the git rm, nuking the
> bin/ar71xx/packages/packages/luci-app-sqm*
> and doing a complete rebuild... eventually did work... but strikes me
> as a bit more brute force than I wanted.

	Fair enough, see the point above of Toke and me coming up with a better plan to get relevant changes into tester’s hands quicker...

> 
> I am not in a position (aformentioned build is my specialized for
> pico/nano build at the moment) to update the packages db,
> but I stuck up the correct new package... which you can install via
> 
> 
> Apologies for the confusion, I simply figured that the -p option would
> work, and never checked.
> 
> I also note that I think I am building things compile tuned for the
> archer mips34c, rather thanthe wndr3800´s 24c. However I kept the
> unaligned access hacks (not needed on the 34c derived archer, but
> needed on the 24c wndr3800)

	Looking at the readily available cpu information, MIPS  seems to be severely EOL’d. Mikael’s result with the armada 385 seem way more future proof (even without BQL)

> 
> Sigh.
> 
> maintaining 3 separate trees for the platforms we sort of support is a
> PITA.

	I am happy to buy a new router to be back on board with your development, but the mips 34c seems a bit underwhelming for the goal of NAT-shapimg roughly 300 Mbps bi-directionally. The edgerouterX’s numbers also do not look too hot; Mikeal’s numbers really made made me reconsider my plan to get another HP priloant microsrver G8 to do duty as PPPoE gateway router and traffic shaper. (The G8 retails fr 240 EUR with 2.3 GHz and 2GB ram, so 200 EUR for a wrt1200ac definitely looks like too much, even though the G* does not do wifi, but I still have the wbdr 3700v2 to serve wireless clients...)

> Then adding in something that can mesh by default on 2 others.
> and then trying to get anything to compile for another arch entirely
> and failing…
> 

	As much as I would like to participate in that fun, a 1 router household is just not interesting enough I guess )

Best Regards	

	Sebastian

> 
> On Fri, Jun 26, 2015 at 11:49 AM, Sebastian Moeller <moeller0 at gmx.de> wrote:
>> Hi Jim,
>> 
>> 
>> On Jun 26, 2015, at 19:21 , Jim Gettys <jg at freedesktop.org> wrote:
>> 
>>> Inquiring minds would like to know….
>> 
>>        I guess, that depends on your definition of “LUCI GUI for cake”; as far as I know you can select cake in the “Queueing Discipline” drop down in the “Queue Discipline” tab of luci-app-sqm. So if the machine has the cake module and a sufficiently recent (out-of-tree) tc binary installed it should just work. Using the “Advanced option strings…” it even should be possible to pass arbitrary strings to cake (without any error checking), so I assume we are all set up for cake testing (or should that be cake tasting?) from the LUCI GUI. But I have not managed to build cake locally, nor have I dared to risk my families internet connectivity by trying one of Dave’s newer "openwrt trunk with cake” builds. So these changes linger in the ceropackages-3.10 repository until we get enough tasting to convince Toke to pull them into the openwrt repositories (these changes affect both sqm-scripts and luci-app-sqm). Dave graciously allowed me access to one of his test machines, and the changes allow to set up cake from the GUI, but I only did limited functionality testing. (And we discovered that something was of in that cake worked but did not accept all the keywords it should have, but unless you are on a XDSL-Link you most likely do not care about specifying per-packet-overhead)
>>        If you have any specific thing you are missing, let me know and I will try to help.
>> 
>> Best Regards
>>        Sebastian
>> 
>> 
>>>                     - Jim
>>> 
>>> _______________________________________________
>>> Cerowrt-devel mailing list
>>> Cerowrt-devel at lists.bufferbloat.net
>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>> 
>> _______________________________________________
>> Cerowrt-devel mailing list
>> Cerowrt-devel at lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
> 
> 
> 
> -- 
> Dave Täht
> worldwide bufferbloat report:
> http://www.dslreports.com/speedtest/results/bufferbloat
> And:
> What will it take to vastly improve wifi for everyone?
> https://plus.google.com/u/0/explore/makewififast




More information about the Cerowrt-devel mailing list