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

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

stats collection doesn't make a lot of sense.

root@archer-2:/sys/kernel/debug/ieee80211/phy1/ath9k# cat xmit
                            BE         BK        VI        VO

MPDUs Queued:                0          0         0       811
MPDUs Completed:            78          1         1       828
MPDUs XRetried:              0          0         0         1
Aggregates:              46528         99     13023         0 ?
AMPDUs Queued HW:            0          0         0         0
AMPDUs Completed:       986000       4544    244768         0 ??
AMPDUs Retried:          18377         23      4073         0

On Thu, Nov 17, 2016 at 11:08 AM, Dave Taht <dave.taht@gmail.com> wrote:
> I tested this against the archer c7v2 just now.
>
> It doesn't crash after an hour of extensive testing.
>
> For testers, I note the airtime fairness portion of these patches is ath9k only.
>
> The ath10k firmware build on toke's site is out of date, but the qca
> firmware can be installed for the ath10k regardless after modifying
> /etc/opkg/distfeeds.conf to point to his site.
>
> Unencrypted, using an osx box as the client, using flent with
> tcp_ndown and --test-parameter=download_streams=1 8, and 24
>
> The ath10k works - with download rates in vht80 ranging from 210 to
> 140, eating 97% of cpu while forwarding, with less than < 30ms latency
> at the 90th percentile.
> The ath9k works in HT20 mode, eating 40% of cpu while forwarding, with
> less than <30ms also, speed in the 50mbit range.
>
> I am about 10 ft from the AP in the SF apt, in a fairly noisy space
> (can hear over 50 other APs). The only sane way to get a before/after
> comparison is for me to reflash...
>
> but: I will try a bridged connection next, then try crypted, then go
> back to looking at the TSQ issue on x86.
>
>
> On Thu, Nov 17, 2016 at 8:30 AM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>> 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/
>>
>> There's an assortment of packages included in the build. I *think* it's
>> possible to install more packages from the LEDE nightly builds by simply
>> doing an 'opkg update'.
>>
>> I was going to do a build for the Turris Omnia as well, but turns out
>> they haven't upstreamed their code. On the other hand, they seem to be
>> tracking both LEDE and OpenWrt fairly closely, so the intermediate
>> queueing patches for ath9k are included in their builds.
>>
>> I make no promises of regular updated builds, but feel free to take this
>> one for a spin :)
>>
>> -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