Re: [Make-wifi-fast] Updated LEDE build with airtime patches

[Dave Taht <dave.taht@gmail.com>]

on the rrul test, the BK queue got no connection, and I lost the
connection several times during it. While I was testing
the ath9k, a new message showed up in the ath10k log.

It would not surprise me if a shorter codel target made the hardware
queues work better.

Otherwise the short test runs I just did (tcp_nup, tcp_ndown) were fine.

 0x00001000
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.777307] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.784831] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.792409] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.799962] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.807518] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.815016] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.822565] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.830113] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.837657] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon
Thu Nov 24 17:50:58 2016 kern.warn kernel: [ 2760.845181] ath10k_pci
0000:01:00.0: SWBA overrun on vdev 0, skipped old beacon

On Thu, Nov 24, 2016 at 9:35 AM, Dave Taht <dave.taht@gmail.com> wrote:
> It's more the quantum than the target that is an issue on *wrt, IMHO.
>
> (I got no problem with testing each of these in isolation)
>
> I am starting a test run now. So far no crashes on the ath9k side on the archer.
>
> Two things I am not testing on this platform are HT40 and crypto. (I
> was using the x86 versions for that). Nor was I monitoring cpu usage
> on the previous archer test runs (sigh), and for some value of
> consistency, am not doing it on these either. I'd *really* like more
> ht40.....
>
> I will finish bootstrapping up my "c.h.i.p." nodes to do better
> airtime fairness testing soon via the rtt_var tests.
>
> Ideally I'd like more to join in on this hopefully last push to get
> the core patches upstream.
>
> Note that in the US...
>
> it's the thanksgiving holiday which can be interpreted as either
> people spending more time with family - or being free enough from work
> to dig in and test the heck out of this stuff (or in my case, fire off
> a big automated test, while spending time with family). :)
>
>
> On Thu, Nov 24, 2016 at 9:14 AM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>> Yup. Planning to look into the codel tuning issue next. Until then I figured I'd keep the patch set small...
>>
>> On 24 November 2016 18:01:32 CET, Dave Taht <dave.taht@gmail.com> wrote:
>>>It looks like you dropped the quantum 1514/codel 5ms from the patchset?
>>>
>>>On Thu, Nov 24, 2016 at 7:01 AM, Toke Høiland-Jørgensen <toke@toke.dk>
>>>wrote:
>>>> Toke Høiland-Jørgensen <toke@toke.dk> writes:
>>>>
>>>>> I created an updated build of current LEDE head with added airtime
>>>>> fairness patches, and the patch to re-enable the intermediate queues
>>>for
>>>>> ath10k. I.e., the contents of my repo here:
>>>>>
>>>>> https://kau.toke.dk/git/lede/
>>>>>
>>>>> Builds for some common ar71xx devices are here:
>>>>>
>>>>> https://kau.toke.dk/lede/airtime-fairness-builds/ar71xx/generic/
>>>>
>>>>
>>>> These builds have now been updated to a new version of the airtime
>>>> fairness patch (that was also posted to the list a few hours ago).
>>>> Please take them for a spin. Once I have confirmation that they don't
>>>> blow up catastrophically, I'll submit the patch for inclusion in LEDE
>>>:)
>>>>
>>>> -Toke
>>>> _______________________________________________
>>>> Make-wifi-fast mailing list
>>>> Make-wifi-fast@lists.bufferbloat.net
>>>> https://lists.bufferbloat.net/listinfo/make-wifi-fast
>>
>
>
>
> --
> Dave Täht
> Let's go make home routers and wifi faster! With better software!
> http://blog.cerowrt.org



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