Re: [Cake] Testing wanted: Statistics API rework

[Toke Høiland-Jørgensen]

Pete Heist  writes:

>> On Apr 25, 2018, at 11:15 AM, Toke Høiland-Jørgensen  wrote:
>> 
>> Pete Heist  writes:
>> 
>>> I requested a re-build over in the Ubiquiti EdgeOS forums. Between
>>> ER-X and ER-L I think that would cover 32-bit mips (little endian) and
>>> 64-bit mips (big endian). Anything specific to test besides “it
>>> works"? :)
>> 
>> We need to verify that it will still print statistics and that the
>> values do not look weird (especially the per-tin 'bytes' counter since
>> that is a 64bit value).
>
>
> In case you didn’t see it before, Lochnair has automatic builds set up
> (cool) for EdgeOS that are failing as of yesterday:
>
> https://build.lochnair.net/job/ubiquiti/job/bufferbloat/job/cake_cobalt/
>
> It may very well be due to the 3.10(!) kernel it ships with, but in
> case it helps somehow, go to the latest build and “Console Output”. In
> case you commit a change, the git repo is polled once an hour,
> apparently. Otherwise, I don’t have an easy way to test mips…

Yup, definitely an issue with kernel versions. Guess we'll have to go
back and add in compatibility stuff later.

Kevin was nice enough to test on an Archer C7, so I think we are good
for now :)

-Toke

___
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] Testing wanted: Statistics API rework

[Pete Heist]

> On Apr 25, 2018, at 11:15 AM, Toke Høiland-Jørgensen  wrote:
> 
> Pete Heist  writes:
> 
>> I requested a re-build over in the Ubiquiti EdgeOS forums. Between
>> ER-X and ER-L I think that would cover 32-bit mips (little endian) and
>> 64-bit mips (big endian). Anything specific to test besides “it
>> works"? :)
> 
> We need to verify that it will still print statistics and that the
> values do not look weird (especially the per-tin 'bytes' counter since
> that is a 64bit value).


In case you didn’t see it before, Lochnair has automatic builds set up (cool) 
for EdgeOS that are failing as of yesterday:

https://build.lochnair.net/job/ubiquiti/job/bufferbloat/job/cake_cobalt/

It may very well be due to the 3.10(!) kernel it ships with, but in case it 
helps somehow, go to the latest build and “Console Output”. In case you commit 
a change, the git repo is polled once an hour, apparently. Otherwise, I don’t 
have an easy way to test mips…

___
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] Testing wanted: Statistics API rework

[Toke Høiland-Jørgensen]

Pete Heist  writes:

>> On Apr 24, 2018, at 11:30 PM, Toke Høiland-Jørgensen  wrote:
>> 
>> If someone has time to test it (especially on mips!), that would be
>> great; I think we still have a few days left of net-next being open, so
>> if it works for you, I'll resubmit tomorrow...
>
> I requested a re-build over in the Ubiquiti EdgeOS forums. Between
> ER-X and ER-L I think that would cover 32-bit mips (little endian) and
> 64-bit mips (big endian). Anything specific to test besides “it
> works"? :)

We need to verify that it will still print statistics and that the
values do not look weird (especially the per-tin 'bytes' counter since
that is a 64bit value).

-Toke
___
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] Testing wanted: Statistics API rework

[Marcelo Ricardo Leitner]

On Tue, Apr 24, 2018 at 11:30:35PM +0200, Toke Høiland-Jørgensen wrote:
> Hey everyone
>
> As you have probably seen, the posting of CAKE to netdev resulted in
> feedback saying that versioned structs are a no-go. Well, I've just
> pushed a change to both the cake and tc repos that changes the
> statistics reporting to used nested netlink attributes.
>
> If someone has time to test it (especially on mips!), that would be
> great; I think we still have a few days left of net-next being open, so

We should still have ~4 weeks of net-next open. It will close when
Linus releases the v4.17, which is when v4.18 merge window starts (and
thus why net-next gets closed).  Considering 1 -rc per week and at
least 4 more -rc's, ~4 weeks.

To be more precise, if that is possible, we can use the crystal ball
:-)
http://phb-crystal-ball.org/
Jun 10th.

> if it works for you, I'll resubmit tomorrow...
>
> -Toke
> ___
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake
___
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake