[Cerowrt-devel] linksys 1200ac lockout problem still exists

[Cerowrt-devel] linksys 1200ac lockout problem still exists

[Dave Taht]

so, in a flood of optimism, I reflashed my linksys 1200ac with lede head.

If you hit it with more than one download flow in netperf(flent), all
successor flows still starve.

http://www.taht.net/~d/linksys1200aclockout.png

I was under the impression a fix (also adding BQL) had arrived for
this problem in mainline.

filed a bug on it just now, I should have done so long ago.

https://bugs.lede-project.org/index.php?do=details&task_id=294

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

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Dave Taht]

Has the omnia got this problem?

On Sun, Nov 20, 2016 at 11:29 AM, Dave Taht <dave.taht@gmail.com> wrote:
> so, in a flood of optimism, I reflashed my linksys 1200ac with lede head.
>
> If you hit it with more than one download flow in netperf(flent), all
> successor flows still starve.
>
> http://www.taht.net/~d/linksys1200aclockout.png
>
> I was under the impression a fix (also adding BQL) had arrived for
> this problem in mainline.
>
> filed a bug on it just now, I should have done so long ago.
>
> https://bugs.lede-project.org/index.php?do=details&task_id=294
>
> --
> 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

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Toke Høiland-Jørgensen]

Dave Taht <dave.taht@gmail.com> writes:

> Has the omnia got this problem?

Can check. Just finished fixing (well, working around) an annoying issue
with the switch in the omnia:
https://forum.turris.cz/t/swich-ports-offline-if-plugged-during-boot/1611


Steps to reproduce the lockout problem?

-Toke

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Sebastian Moeller]

Hi Dave,

> On Nov 20, 2016, at 20:30, Dave Taht <dave.taht@gmail.com> wrote:
> 
> Has the omnia got this problem?

Not sure, following your recipe from the LEDE bug, I attempted:
bash-3.2$ ./run-flent -H netperf-eu.bufferbloat.net –test-parameter=download_streams=12 tcp_ndown
Fatal error: Hostname lookup failed for host –test-parameter=download_streams=12: [Errno 8] nodename nor servname provided, or not known
bash-3.2$ 

But as you see that does not work. At least not against netperf-eu.bufferbloat.net even though:
bash-3.2$ ping -c 10 netperf-eu.bufferbloat.net
PING kau.toke.dk (130.243.26.64): 56 data bytes
64 bytes from 130.243.26.64: icmp_seq=0 ttl=46 time=46.657 ms
64 bytes from 130.243.26.64: icmp_seq=1 ttl=46 time=47.011 ms
64 bytes from 130.243.26.64: icmp_seq=2 ttl=46 time=46.643 ms
64 bytes from 130.243.26.64: icmp_seq=3 ttl=46 time=46.932 ms
64 bytes from 130.243.26.64: icmp_seq=4 ttl=46 time=47.062 ms
64 bytes from 130.243.26.64: icmp_seq=5 ttl=46 time=46.635 ms
64 bytes from 130.243.26.64: icmp_seq=6 ttl=46 time=46.479 ms
64 bytes from 130.243.26.64: icmp_seq=7 ttl=46 time=46.545 ms
64 bytes from 130.243.26.64: icmp_seq=8 ttl=46 time=46.585 ms
64 bytes from 130.243.26.64: icmp_seq=9 ttl=46 time=46.587 ms

--- kau.toke.dk ping statistics ---
10 packets transmitted, 10 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 46.479/46.714/47.062/0.197 ms

So I must be doing something wrong, but what? I have never tried the tcp_ndown test, so aybe just pilot’s error…

Best Regards
	Sebastian


> 
> On Sun, Nov 20, 2016 at 11:29 AM, Dave Taht <dave.taht@gmail.com> wrote:
>> so, in a flood of optimism, I reflashed my linksys 1200ac with lede head.
>> 
>> If you hit it with more than one download flow in netperf(flent), all
>> successor flows still starve.
>> 
>> http://www.taht.net/~d/linksys1200aclockout.png
>> 
>> I was under the impression a fix (also adding BQL) had arrived for
>> this problem in mainline.
>> 
>> filed a bug on it just now, I should have done so long ago.
>> 
>> https://bugs.lede-project.org/index.php?do=details&task_id=294
>> 
>> --
>> 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
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Sebastian Moeller]

[-- Attachment #1: Type: text/plain, Size: 2677 bytes --]

Hi Dave,

when I use tcp_12down instead of tcp_ndown I end up with something looking sane (see attached image?)

Also:
Summary of tcp_12down test run at 2016-11-20 20:16:54.350770:

                             avg /       median          # data pts
 Ping (ms) ICMP   :        54.08 /        53.75 ms              351
 TCP download avg :         3.52 /         3.57 Mbits/s         301
 TCP download sum :        42.24 /        42.79 Mbits/s         301
 TCP download::1  :         3.50 /         3.63 Mbits/s         301
 TCP download::10 :         3.55 /         3.61 Mbits/s         301
 TCP download::11 :         3.57 /         3.62 Mbits/s         301
 TCP download::12 :         3.54 /         3.66 Mbits/s         300
 TCP download::2  :         3.51 /         3.61 Mbits/s         300
 TCP download::3  :         3.55 /         3.61 Mbits/s         301
 TCP download::4  :         3.59 /         3.67 Mbits/s         300
 TCP download::5  :         3.46 /         3.55 Mbits/s         300
 TCP download::6  :         3.49 /         3.60 Mbits/s         301
 TCP download::7  :         3.50 /         3.59 Mbits/s         300
 TCP download::8  :         3.48 /         3.57 Mbits/s         301
 TCP download::9  :         3.50 /         3.58 Mbits/s         300


looks fine. But maybe tcp_12down is not exactly doing the same as tcp_ndown… I tested from a laptop attached to one of the omnia’s lan ports (omnia’s wan is connected to the lan port of a LEDE r2222 wndr3700v2 with double NAT).

Best Regards
	Sebastian

> On Nov 20, 2016, at 20:30, Dave Taht <dave.taht@gmail.com> wrote:
> 
> Has the omnia got this problem?
> 
> On Sun, Nov 20, 2016 at 11:29 AM, Dave Taht <dave.taht@gmail.com> wrote:
>> so, in a flood of optimism, I reflashed my linksys 1200ac with lede head.
>> 
>> If you hit it with more than one download flow in netperf(flent), all
>> successor flows still starve.
>> 
>> http://www.taht.net/~d/linksys1200aclockout.png
>> 
>> I was under the impression a fix (also adding BQL) had arrived for
>> this problem in mainline.
>> 
>> filed a bug on it just now, I should have done so long ago.
>> 
>> https://bugs.lede-project.org/index.php?do=details&task_id=294
>> 
>> --
>> 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
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel


[-- Attachment #2.1: Type: text/html, Size: 5304 bytes --]

[-- Attachment #2.2: tcp_12down_20161120.png --]
[-- Type: image/png, Size: 40178 bytes --]

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Toke Høiland-Jørgensen]

Sebastian Moeller <moeller0@gmx.de> writes:

> Hi Dave,
>
>> On Nov 20, 2016, at 20:30, Dave Taht <dave.taht@gmail.com> wrote:
>> 
>> Has the omnia got this problem?
>
> Not sure, following your recipe from the LEDE bug, I attempted:
> bash-3.2$ ./run-flent -H netperf-eu.bufferbloat.net –test-parameter=download_streams=12 tcp_ndown
> Fatal error: Hostname lookup failed for host –test-parameter=download_streams=12: [Errno 8] nodename nor servname provided, or not known
> bash-3.2$

Think that's a copy-paste error. The bug tracker has turned -- into a
long dash, which Flent doesn't understand and thus interprets the
parameter as a hostname (which it understandably fails to lookup).

-Toke

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Dave Taht]

this a to-the-router test, not a through the router test.

On Sun, Nov 20, 2016 at 12:32 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Sebastian Moeller <moeller0@gmx.de> writes:
>
>> Hi Dave,
>>
>>> On Nov 20, 2016, at 20:30, Dave Taht <dave.taht@gmail.com> wrote:
>>>
>>> Has the omnia got this problem?
>>
>> Not sure, following your recipe from the LEDE bug, I attempted:
>> bash-3.2$ ./run-flent -H netperf-eu.bufferbloat.net –test-parameter=download_streams=12 tcp_ndown
>> Fatal error: Hostname lookup failed for host –test-parameter=download_streams=12: [Errno 8] nodename nor servname provided, or not known
>> bash-3.2$
>
> Think that's a copy-paste error. The bug tracker has turned -- into a
> long dash, which Flent doesn't understand and thus interprets the
> parameter as a hostname (which it understandably fails to lookup).
>
> -Toke



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

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Sebastian Moeller]

[-- Attachment #1: Type: text/plain, Size: 2618 bytes --]

Hi Dave, hi Toke,

thanks for talking me through this:
bash-3.2$ ./run-flent -H 192.168.42.1 --test-parameter=download_streams=12 tcp_12down
Data file written to ./tcp_12down-2016-11-20T213901.312257.flent.gz.
Summary of tcp_12down test run at 2016-11-20 20:39:01.312257:

                             avg /       median          # data pts
 Ping (ms) ICMP   :       365.65 /       116.48 ms              302
 TCP download avg :        80.62 /        75.99 Mbits/s         301
 TCP download sum :       967.41 /       755.47 Mbits/s         301
 TCP download::1  :         2.91 /         1.03 Mbits/s         255
 TCP download::10 :         1.01 /         0.54 Mbits/s         276
 TCP download::11 :        23.44 /         0.01 Mbits/s           1
 TCP download::12 :       207.33 /       170.06 Mbits/s         299
 TCP download::2  :       146.25 /          N/A Mbits/s           0
 TCP download::3  :       137.22 /       147.26 Mbits/s         301
 TCP download::4  :         1.44 /         1.08 Mbits/s         259
 TCP download::5  :       176.87 /       164.81 Mbits/s         279
 TCP download::6  :         0.63 /         0.32 Mbits/s         271
 TCP download::7  :         0.73 /         0.50 Mbits/s         262
 TCP download::8  :       133.51 /       150.26 Mbits/s         296
 TCP download::9  :       136.07 /       150.13 Mbits/s         300
bash-3.2$





So this certainly looks like the omnia has similar issues…

Best Regards
	Sebastian



> On Nov 20, 2016, at 21:36, Dave Taht <dave.taht@gmail.com> wrote:
> 
> this a to-the-router test, not a through the router test.
> 
> On Sun, Nov 20, 2016 at 12:32 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>> Sebastian Moeller <moeller0@gmx.de> writes:
>> 
>>> Hi Dave,
>>> 
>>>> On Nov 20, 2016, at 20:30, Dave Taht <dave.taht@gmail.com> wrote:
>>>> 
>>>> Has the omnia got this problem?
>>> 
>>> Not sure, following your recipe from the LEDE bug, I attempted:
>>> bash-3.2$ ./run-flent -H netperf-eu.bufferbloat.net –test-parameter=download_streams=12 tcp_ndown
>>> Fatal error: Hostname lookup failed for host –test-parameter=download_streams=12: [Errno 8] nodename nor servname provided, or not known
>>> bash-3.2$
>> 
>> Think that's a copy-paste error. The bug tracker has turned -- into a
>> long dash, which Flent doesn't understand and thus interprets the
>> parameter as a hostname (which it understandably fails to lookup).
>> 
>> -Toke
> 
> 
> 
> -- 
> Dave Täht
> Let's go make home routers and wifi faster! With better software!
> http://blog.cerowrt.org


[-- Attachment #2.1: Type: text/html, Size: 5444 bytes --]

[-- Attachment #2.2: tcp_12down_20161120_netserver_on_omnia.png --]
[-- Type: image/png, Size: 65885 bytes --]

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Toke Høiland-Jørgensen]

[-- Attachment #1: Type: text/plain, Size: 196 bytes --]

Dave Taht <dave.taht@gmail.com> writes:

> this a to-the-router test, not a through the router test.

Yup, happens on the Omnia as well. See attached. Fairly old kernel on
that, though...

-Toke


[-- Attachment #2: tcp_ndown-2016-11-20T214220.946956.Test_12_streams_to_Omnia.flent.gz --]
[-- Type: application/octet-stream, Size: 27295 bytes --]

[-- Attachment #3: omnia_lockout.png --]
[-- Type: image/png, Size: 53105 bytes --]

Re: [Cerowrt-devel] linksys 1200ac lockout problem still exists

[Toke Høiland-Jørgensen]

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

> Dave Taht <dave.taht@gmail.com> writes:
>
>> this a to-the-router test, not a through the router test.
>
> Yup, happens on the Omnia as well. See attached. Fairly old kernel on
> that, though...

Not just new connections that die, BTW. My SSH session froze up as well.
Oh, and my test was running through the WAN port...

-Toke