[Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Dave Taht]

looks like it can be easily flashed once again as 49217...

https://wiki.openwrt.org/toh/tp-link/archer-c5-c7-wdr7500

but Arokh's comprehensive build (which includes sch_cake):
https://forum.openwrt.org/viewtopic.php?id=50914

is a bit out of date...

so I guess I should try lede for the first time via
https://downloads.lede-project.org/snapshots/targets/ar71xx/generic/

and try cross compiling sch_cake for it.

... and if I feel really, really, really, really ambitious, try
backporting one of michals lovely patchsets to 4.4.

http://blog.cerowrt.org/post/fq_codel_on_ath10k/

https://github.com/kazikcz/linux/tree/fqmac-v4+dqlrfc+cpuregrfix

But I don't feel that ambitious. It *will* be good to blow the dust
off of the tp-link that made me go to Washington...

here goes...


-- 
Dave Täht
Let's go make home routers and wifi faster! With better software!
http://blog.cerowrt.org

Re: [Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Kevin Darbyshire-Bryant]

[-- Attachment #1: Type: text/plain, Size: 1132 bytes --]

LEDE on the C7 isn't a problem.  Neither is sch_cake.  Unfortunately
AR71xx is on a 4.1 kernel.

In the very brief flirtation I had with 4.4 on the C7, it appeared to
work except for USB...at that point I backed out :-)

Kevin

On 11/05/16 22:53, Dave Taht wrote:
> looks like it can be easily flashed once again as 49217...
>
> https://wiki.openwrt.org/toh/tp-link/archer-c5-c7-wdr7500
>
> but Arokh's comprehensive build (which includes sch_cake):
> https://forum.openwrt.org/viewtopic.php?id=50914
>
> is a bit out of date...
>
> so I guess I should try lede for the first time via
> https://downloads.lede-project.org/snapshots/targets/ar71xx/generic/
>
> and try cross compiling sch_cake for it.
>
> ... and if I feel really, really, really, really ambitious, try
> backporting one of michals lovely patchsets to 4.4.
>
> http://blog.cerowrt.org/post/fq_codel_on_ath10k/
>
> https://github.com/kazikcz/linux/tree/fqmac-v4+dqlrfc+cpuregrfix
>
> But I don't feel that ambitious. It *will* be good to blow the dust
> off of the tp-link that made me go to Washington...
>
> here goes...
>
>



[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4816 bytes --]

Re: [Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Kevin Darbyshire-Bryant]

[-- Attachment #1: Type: text/plain, Size: 1689 bytes --]

Forgot to mention - in slightly better news, the recent fq_codel
flows_cnt = 128 recommendation from Eric has made it up into John
Crispin's lede staging tree, so I hope that'll appear in master soon.

The fq_codel 'batch drop' has sort of made it back to lede 4.4...well
there's 2 patches, one is a backport which also includes the api update
to support "update backlog for hierarchical qdiscs" as well as the batch
drop (not done by me),  there's a version which I did which just
backports the batch drop.

The batch drop alone to 4.1 is beyond me. 

I'm slowly coming around to the opinion that the 'batch drop alone'
patch is what should temporarily go into lede/openwrt until an official
backport is done.  I'm not totally convinced about backporting the API
change. 


On 11/05/16 22:53, Dave Taht wrote:
> looks like it can be easily flashed once again as 49217...
>
> https://wiki.openwrt.org/toh/tp-link/archer-c5-c7-wdr7500
>
> but Arokh's comprehensive build (which includes sch_cake):
> https://forum.openwrt.org/viewtopic.php?id=50914
>
> is a bit out of date...
>
> so I guess I should try lede for the first time via
> https://downloads.lede-project.org/snapshots/targets/ar71xx/generic/
>
> and try cross compiling sch_cake for it.
>
> ... and if I feel really, really, really, really ambitious, try
> backporting one of michals lovely patchsets to 4.4.
>
> http://blog.cerowrt.org/post/fq_codel_on_ath10k/
>
> https://github.com/kazikcz/linux/tree/fqmac-v4+dqlrfc+cpuregrfix
>
> But I don't feel that ambitious. It *will* be good to blow the dust
> off of the tp-link that made me go to Washington...
>
> here goes...
>
>



[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4816 bytes --]

Re: [Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Dave Taht]

On Wed, May 11, 2016 at 3:15 PM, Kevin Darbyshire-Bryant
<kevin@darbyshire-bryant.me.uk> wrote:
> Forgot to mention - in slightly better news, the recent fq_codel
> flows_cnt = 128 recommendation from Eric has made it up into John

I still kind of disagree with the reasoning on cutting the default to
128 queues, btw, but I felt it was worth testing.

IMHO in most circumstances having more queues will make you more
resilient to simple dumb floods as well as more complex ones, and give
better latency overall for all flows... But, lacking data on lower end
platforms, and being only willing to argue with y'all - (especially
eric), with it...

... is sort of why you find me (successfully) installing lede today on
the archer c7 v2.

I certainly agree strongly with the bulk drop portion of the patch. I
never liked scanning all the queues for every drop... I just couldn't
force myself to come up with something as drastic as eric did.

I sure wish cake, tc-adv, and the current sqm were up in lede...



> Crispin's lede staging tree, so I hope that'll appear in master soon.
>
> The fq_codel 'batch drop' has sort of made it back to lede 4.4...well
> there's 2 patches, one is a backport which also includes the api update
> to support "update backlog for hierarchical qdiscs" as well as the batch
> drop (not done by me),  there's a version which I did which just
> backports the batch drop.
>
> The batch drop alone to 4.1 is beyond me.
>
> I'm slowly coming around to the opinion that the 'batch drop alone'
> patch is what should temporarily go into lede/openwrt until an official
> backport is done.  I'm not totally convinced about backporting the API
> change.
>
>
> On 11/05/16 22:53, Dave Taht wrote:
>> looks like it can be easily flashed once again as 49217...
>>
>> https://wiki.openwrt.org/toh/tp-link/archer-c5-c7-wdr7500
>>
>> but Arokh's comprehensive build (which includes sch_cake):
>> https://forum.openwrt.org/viewtopic.php?id=50914
>>
>> is a bit out of date...
>>
>> so I guess I should try lede for the first time via
>> https://downloads.lede-project.org/snapshots/targets/ar71xx/generic/
>>
>> and try cross compiling sch_cake for it.
>>
>> ... and if I feel really, really, really, really ambitious, try
>> backporting one of michals lovely patchsets to 4.4.
>>
>> http://blog.cerowrt.org/post/fq_codel_on_ath10k/
>>
>> https://github.com/kazikcz/linux/tree/fqmac-v4+dqlrfc+cpuregrfix
>>
>> But I don't feel that ambitious. It *will* be good to blow the dust
>> off of the tp-link that made me go to Washington...
>>
>> here goes...
>>
>>
>
>
>
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake
>



-- 
Dave Täht
Let's go make home routers and wifi faster! With better software!
http://blog.cerowrt.org

Re: [Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Kevin Darbyshire-Bryant]

[-- Attachment #1: Type: text/plain, Size: 1262 bytes --]



On 11/05/16 23:39, Dave Taht wrote:
> On Wed, May 11, 2016 at 3:15 PM, Kevin Darbyshire-Bryant
> <kevin@darbyshire-bryant.me.uk> wrote:
>> Forgot to mention - in slightly better news, the recent fq_codel
>> flows_cnt = 128 recommendation from Eric has made it up into John
> I still kind of disagree with the reasoning on cutting the default to
> 128 queues, btw, but I felt it was worth testing.
>
> IMHO in most circumstances having more queues will make you more
> resilient to simple dumb floods as well as more complex ones, and give
> better latency overall for all flows... But, lacking data on lower end
> platforms, and being only willing to argue with y'all - (especially
> eric), with it...
>
> ... is sort of why you find me (successfully) installing lede today on
> the archer c7 v2.
>
> I certainly agree strongly with the bulk drop portion of the patch. I
> never liked scanning all the queues for every drop... I just couldn't
> force myself to come up with something as drastic as eric did.
>
> I sure wish cake, tc-adv, and the current sqm were up in lede...
The bulk drop has just gone in as a patch to 4.4 on LEDE.
https://git.lede-project.org/?p=source.git;a=commit;h=fad8bdfa40df8636a52d770bbab010086c1976ec



[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4816 bytes --]

Re: [Cake] tp-link archer c7 & cake & mac80211 fq_codel

[Kevin Darbyshire-Bryant]

[-- Attachment #1: Type: text/plain, Size: 1789 bytes --]

Oh and forgot, will mean you'll need this patch https://github.com/kdarbyshirebryant/ceropackages-3.10/commit/f26596763b3aecf2f6ec6a2401efe4d6966ecf27 if you want to build cake on lede's 4.4 bacported fq Codel fixes



On 12 May 2016, at 21:34, Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk<mailto:kevin@darbyshire-bryant.me.uk>> wrote:



On 11/05/16 23:39, Dave Taht wrote:
On Wed, May 11, 2016 at 3:15 PM, Kevin Darbyshire-Bryant
<kevin@darbyshire-bryant.me.uk<mailto:kevin@darbyshire-bryant.me.uk>> wrote:
Forgot to mention - in slightly better news, the recent fq_codel
flows_cnt = 128 recommendation from Eric has made it up into John
I still kind of disagree with the reasoning on cutting the default to
128 queues, btw, but I felt it was worth testing.

IMHO in most circumstances having more queues will make you more
resilient to simple dumb floods as well as more complex ones, and give
better latency overall for all flows... But, lacking data on lower end
platforms, and being only willing to argue with y'all - (especially
eric), with it...

... is sort of why you find me (successfully) installing lede today on
the archer c7 v2.

I certainly agree strongly with the bulk drop portion of the patch. I
never liked scanning all the queues for every drop... I just couldn't
force myself to come up with something as drastic as eric did.

I sure wish cake, tc-adv, and the current sqm were up in lede...
The bulk drop has just gone in as a patch to 4.4 on LEDE.
https://git.lede-project.org/?p=source.git;a=commit;h=fad8bdfa40df8636a52d770bbab010086c1976ec


_______________________________________________
Cake mailing list
Cake@lists.bufferbloat.net<mailto:Cake@lists.bufferbloat.net>
https://lists.bufferbloat.net/listinfo/cake

[-- Attachment #2: Type: text/html, Size: 3941 bytes --]