Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Jon Pike]

[-- Attachment #1: Type: text/plain, Size: 2943 bytes --]

Hmmm...  There's a guy maintaining a tweaked and optmized for the C7 build
of LEDE.   He's got a long thread over on the LEDE Forums.

Might be a good way to get a wider group of testers running the patch,  and
thereby get the mainstream dev's more comfortable about rolling it in?
I'd mention such over there, if that sounds like an idea...

On Sep 6, 2017 9:00 AM, <make-wifi-fast-request@lists.bufferbloat.net>
wrote:

> Send Make-wifi-fast mailing list submissions to
>         make-wifi-fast@lists.bufferbloat.net
>
> To subscribe or unsubscribe via the World Wide Web, visit
>         https://lists.bufferbloat.net/listinfo/make-wifi-fast
> or, via email, send a message with subject or body 'help' to
>         make-wifi-fast-request@lists.bufferbloat.net
>
> You can reach the person managing the list at
>         make-wifi-fast-owner@lists.bufferbloat.net
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Make-wifi-fast digest..."
>
> Today's Topics:
>
>    1. Upstream Ath10k FQ_Codel Patch? (Noah Causin)
>    2. Re: Upstream Ath10k FQ_Codel Patch? (Toke Høiland-Jørgensen)
>
>
> ---------- Forwarded message ----------
> From: Noah Causin <n0manletter@gmail.com>
> To: make-wifi-fast@lists.bufferbloat.net
> Cc:
> Bcc:
> Date: Tue, 5 Sep 2017 15:30:04 -0400
> Subject: [Make-wifi-fast] Upstream Ath10k FQ_Codel Patch?
> Hi,
>
> I have been using the patch from Toke's website which enables mac80211
> soft queues for the ath10k on LEDE firmware for about 6 months now.
>
> http://kau.toke.dk/git/lede/
>
> It has been running really stable on both my Archer c7 v2 and v3.
>
> I was wondering if there were any plans to upstream it to the LEDE project.
>
> Noah
>
>
>
>
> ---------- Forwarded message ----------
> From: "Toke Høiland-Jørgensen" <toke@toke.dk>
> To: Noah Causin <n0manletter@gmail.com>,
> make-wifi-fast@lists.bufferbloat.net
> Cc:
> Bcc:
> Date: Wed, 06 Sep 2017 11:07:25 +0200
> Subject: Re: [Make-wifi-fast] Upstream Ath10k FQ_Codel Patch?
> Noah Causin <n0manletter@gmail.com> writes:
>
> > Hi,
> >
> > I have been using the patch from Toke's website which enables mac80211
> > soft queues for the ath10k on LEDE firmware for about 6 months now.
> >
> > http://kau.toke.dk/git/lede/
> >
> > It has been running really stable on both my Archer c7 v2 and v3.
> >
> > I was wondering if there were any plans to upstream it to the LEDE
> > project.
>
> No concrete plans, no. I was hoping that the underlying issue would be
> fixed upstream, but I'm not sure the original developer is working on
> this anymore.
>
> I can submit the patch to LEDE, but no idea if they will take it...
>
> -Toke
>
>
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast
>

[-- Attachment #2: Type: text/html, Size: 4472 bytes --]

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Toke Høiland-Jørgensen]

Jon Pike <jonpike54@gmail.com> writes:

> Hmmm...  There's a guy maintaining a tweaked and optmized for the C7 build
> of LEDE.   He's got a long thread over on the LEDE Forums.
>
> Might be a good way to get a wider group of testers running the patch,  and
> thereby get the mainstream dev's more comfortable about rolling it in?
> I'd mention such over there, if that sounds like an idea...

Sure. I submitted it to LEDE now, should be simple to pull in:
https://patchwork.ozlabs.org/patch/810573/

-Toke

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Adrian Popescu]

QCA9980 gets 2-6 Mbps with this patch. I was able to get up to 400-500
Mbps without the patch.

The QCA9980 firmware doesn't have push-pull support.

On Mon, Sep 11, 2017 at 10:33 AM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Jon Pike <jonpike54@gmail.com> writes:
>
>> Hmmm...  There's a guy maintaining a tweaked and optmized for the C7 build
>> of LEDE.   He's got a long thread over on the LEDE Forums.
>>
>> Might be a good way to get a wider group of testers running the patch,  and
>> thereby get the mainstream dev's more comfortable about rolling it in?
>> I'd mention such over there, if that sounds like an idea...
>
> Sure. I submitted it to LEDE now, should be simple to pull in:
> https://patchwork.ozlabs.org/patch/810573/
>
> -Toke
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Toke Høiland-Jørgensen]

Adrian Popescu <adriannnpopescu@gmail.com> writes:

> QCA9980 gets 2-6 Mbps with this patch. I was able to get up to 400-500
> Mbps without the patch.
>
> The QCA9980 firmware doesn't have push-pull support.

Ah, bugger. Guess we'll have to drop it, then. Which device is this in?

-Toke

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Adrian Popescu]

We need to put some pressure on QCA to add push-pull support to the
firmware. This is MU-MIMO capable hardware which should be able to
handle push-pull without breaking a sweat.

I have tested this on a brand new Archer C2600.

On Mon, Sep 11, 2017 at 3:24 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Adrian Popescu <adriannnpopescu@gmail.com> writes:
>
>> QCA9980 gets 2-6 Mbps with this patch. I was able to get up to 400-500
>> Mbps without the patch.
>>
>> The QCA9980 firmware doesn't have push-pull support.
>
> Ah, bugger. Guess we'll have to drop it, then. Which device is this in?
>
> -Toke

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Toke Høiland-Jørgensen]

Adrian Popescu <adriannnpopescu@gmail.com> writes:

> We need to put some pressure on QCA to add push-pull support to the
> firmware. This is MU-MIMO capable hardware which should be able to
> handle push-pull without breaking a sweat.
>
> I have tested this on a brand new Archer C2600.

There's new hardware shipping without this feature? WTF?

Putting pressure on QCA is all well and good, but unfortunately it's not
something we've had a lot of luck with so far :/

-Toke

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Dave Taht]

Can anyone identify what hardware does have push/pull support?

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Dave Taht]

Also, is the candelatech firmware any better here?

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Adrian Popescu]

I've tried a bunch of setups: standard firmware, CT firmware, CT
firmware with the patch and standard firmware with the patch.

Both CT and default firmware have low throughput with this patch.

I can't waste money on more ath10k hardware. QCA9980 was supposed to
be the shiny new thing with push-pull support.

On Mon, Sep 11, 2017 at 8:09 PM, Dave Taht <dave.taht@gmail.com> wrote:
> Also, is the candelatech firmware any better here?

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Kevin Darbyshire-Bryant]

On 11/09/17 19:33, Adrian Popescu wrote:
> I've tried a bunch of setups: standard firmware, CT firmware, CT
> firmware with the patch and standard firmware with the patch.
> 
> Both CT and default firmware have low throughput with this patch.

It's been committed in LEDE 
https://git.lede-project.org/?p=source.git;a=commit;h=76c3a033f63aba10047a141ccdae303bc9db571e

> 
> I can't waste money on more ath10k hardware. QCA9980 was supposed to
> be the shiny new thing with push-pull support.
> 
> On Mon, Sep 11, 2017 at 8:09 PM, Dave Taht <dave.taht@gmail.com> wrote:
>> Also, is the candelatech firmware any better here?
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast
> 

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Toke Høiland-Jørgensen]

Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk> writes:

> On 11/09/17 19:33, Adrian Popescu wrote:
>> I've tried a bunch of setups: standard firmware, CT firmware, CT
>> firmware with the patch and standard firmware with the patch.
>> 
>> Both CT and default firmware have low throughput with this patch.
>
> It's been committed in LEDE 
> https://git.lede-project.org/?p=source.git;a=commit;h=76c3a033f63aba10047a141ccdae303bc9db571e

Ah! So it should be in current nightlies. I'll flash an Archer C7 and
see if there are any issues. Anyone else care to test out the current
nightly? :)

-Toke

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Toke Høiland-Jørgensen]

[-- Attachment #1: Type: text/plain, Size: 923 bytes --]

Toke Høiland-Jørgensen <toke@toke.dk> writes:

> Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk> writes:
>
>> On 11/09/17 19:33, Adrian Popescu wrote:
>>> I've tried a bunch of setups: standard firmware, CT firmware, CT
>>> firmware with the patch and standard firmware with the patch.
>>> 
>>> Both CT and default firmware have low throughput with this patch.
>>
>> It's been committed in LEDE 
>> https://git.lede-project.org/?p=source.git;a=commit;h=76c3a033f63aba10047a141ccdae303bc9db571e
>
> Ah! So it should be in current nightlies. I'll flash an Archer C7 and
> see if there are any issues. Anyone else care to test out the current
> nightly? :)

So testing it, I see no difference in throughput on my Archer C7; my
802.11ac-capable phone gets ~200 Mbps and my 802.11n laptop around 80.
However, there's no improvement in latency under load either. Meh...

Plot attached.

-Toke


[-- Attachment #2: archer-c7-ath10k-aqm-test.png --]
[-- Type: image/png, Size: 40453 bytes --]

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Adrian Popescu]

[-- Attachment #1: Type: text/plain, Size: 1347 bytes --]

I was testing on QCA9*9*80, not QCA9*8*80. It probably works fine on C7. It
doesn't work on the C2600.

On Tue, Sep 12, 2017 at 12:51 PM, Toke Høiland-Jørgensen <toke@toke.dk>
wrote:

> Toke Høiland-Jørgensen <toke@toke.dk> writes:
>
> > Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk> writes:
> >
> >> On 11/09/17 19:33, Adrian Popescu wrote:
> >>> I've tried a bunch of setups: standard firmware, CT firmware, CT
> >>> firmware with the patch and standard firmware with the patch.
> >>>
> >>> Both CT and default firmware have low throughput with this patch.
> >>
> >> It's been committed in LEDE
> >> https://git.lede-project.org/?p=source.git;a=commit;h=
> 76c3a033f63aba10047a141ccdae303bc9db571e
> >
> > Ah! So it should be in current nightlies. I'll flash an Archer C7 and
> > see if there are any issues. Anyone else care to test out the current
> > nightly? :)
>
> So testing it, I see no difference in throughput on my Archer C7; my
> 802.11ac-capable phone gets ~200 Mbps and my 802.11n laptop around 80.
> However, there's no improvement in latency under load either. Meh...
>
> Plot attached.
>
> -Toke
>
>
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast
>

[-- Attachment #2: Type: text/html, Size: 2307 bytes --]

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Toke Høiland-Jørgensen]

On 12 September 2017 16:03:36 CEST, Adrian Popescu <adriannnpopescu@gmail.com> wrote:
>I was testing on QCA9*9*80, not QCA9*8*80. It probably works fine on
>C7. It
>doesn't work on the C2600.

Ah, right. Bugger. Could you confirm that the latest nightly is broken? Then I'll ask John to revert the patch...

-Toke

[Make-wifi-fast] ath10k latency under load?

[Noah Causin]

[-- Attachment #1: Type: text/plain, Size: 1363 bytes --]

Last year, mac80211 softqueues were re-enabled for ath10k devices in 
LEDE, which brought back fq_codel support, but did not improve latency 
under load.

Is this latency under load still an issue today?

Noah

On 9/12/2017 5:51 AM, Toke Høiland-Jørgensen wrote:
> Toke Høiland-Jørgensen <toke@toke.dk> writes:
>
>> Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk> writes:
>>
>>> On 11/09/17 19:33, Adrian Popescu wrote:
>>>> I've tried a bunch of setups: standard firmware, CT firmware, CT
>>>> firmware with the patch and standard firmware with the patch.
>>>>
>>>> Both CT and default firmware have low throughput with this patch.
>>> It's been committed in LEDE
>>> https://git.lede-project.org/?p=source.git;a=commit;h=76c3a033f63aba10047a141ccdae303bc9db571e
>> Ah! So it should be in current nightlies. I'll flash an Archer C7 and
>> see if there are any issues. Anyone else care to test out the current
>> nightly? :)
> So testing it, I see no difference in throughput on my Archer C7; my
> 802.11ac-capable phone gets ~200 Mbps and my 802.11n laptop around 80.
> However, there's no improvement in latency under load either. Meh...
>
> Plot attached.
>
> -Toke
>
>
>
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast


[-- Attachment #2: Type: text/html, Size: 2603 bytes --]

Re: [Make-wifi-fast] ath10k latency under load?

[Toke Høiland-Jørgensen]

Noah Causin <n0manletter@gmail.com> writes:

> Last year, mac80211 softqueues were re-enabled for ath10k devices in 
> LEDE, which brought back fq_codel support, but did not improve latency 
> under load.
>
> Is this latency under load still an issue today?

Dunno. I do not have any ath10k hardware in my testbed, so have not been
able to run any thorough benchmarks. I hope to remedy that, though, as I
also want to experiment with enabling airtime fairness in ath10k :)

-Toke

Re: [Make-wifi-fast] ath10k latency under load?

[Jim Gettys]

[-- Attachment #1: Type: text/plain, Size: 1041 bytes --]

I did a bit of test testing on my Ubiquity router some months ago, and it
looked like
latency was controlled in my (few) tests...

I should probably do more testing when air-time-fairness is enabled....
                                          - Jim


On Sun, Feb 11, 2018 at 7:54 AM, Toke Høiland-Jørgensen <toke@toke.dk>
wrote:

> Noah Causin <n0manletter@gmail.com> writes:
>
> > Last year, mac80211 softqueues were re-enabled for ath10k devices in
> > LEDE, which brought back fq_codel support, but did not improve latency
> > under load.
> >
> > Is this latency under load still an issue today?
>
> Dunno. I do not have any ath10k hardware in my testbed, so have not been
> able to run any thorough benchmarks. I hope to remedy that, though, as I
> also want to experiment with enabling airtime fairness in ath10k :)
>
> -Toke
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast
>

[-- Attachment #2: Type: text/html, Size: 2021 bytes --]

Re: [Make-wifi-fast] Make-wifi-fast Digest, Vol 29, Issue 1

[Kevin Darbyshire-Bryant]

On 11/09/17 04:11, Jon Pike wrote:
> Hmmm...  There's a guy maintaining a tweaked and optmized for the C7 
> build of LEDE.   He's got a long thread over on the LEDE Forums.
> 
> Might be a good way to get a wider group of testers running the patch, 
>   and thereby get the mainstream dev's more comfortable about rolling it 
> in?   I'd mention such over there, if that sounds like an idea...

Personally I wouldn't touch it with a barge pole.  Patch descriptions vs 
content mismatch - descriptions lacking/mostly missing.  It's never 
going to get committed in that state.

code provenance?  GPL?  Any attempt at all to get any of it upstream in 
kernel/LEDE?