[OpenWrt-Devel] [PATCH][ar71xx] Mikrotik Routerboard RB2011 switch fix

[George Chriss]

On Thu, April 23, 2015 06:16:08 CEST, Toerless Eckert wrote:
> This mail thread seem to have gone dark since december with seemingly
> no conclusion.
>
> I have tried to collect the experiences reported on the wiki page:
>
> wiki.openwrt.org/toh/mikrotik/rb2011uias
>
> As you can see from the table, for me it only works with 0x6f.
> Chris, for whom both 0x3e and 0x6f work has a rev 2 AR9344. I have
> a rev 3. AR8327 is rev. 4 on both our routerboards. No info what rev Matt
> has, he is the other one reporting that only 0x6f works.
>
> I am also observing some amount of increases in switch 0 port 0
> RxBadByte (aka: from CPU to AR8327). It seems to happen for
> all type of traffic sent from CPU to switch, eg: whether i inject it from
> WiFi or from a 100 Mbps port, and whether i send it out
> on an untagged port (lan) or tagged one (from CPU, Wan).

I'm under the impression that some of the "OK/Not OK" reports are based on
ping packet loss which provides limited insight into actual throughput.  My
results with a RB2011UiAS-IN (no WiFi, empty SFP cage, AR9344 Rev. 2, CPU @
600MHz powered from the DC-in jack) on patched 15.05:

0x06000000 (Unpatched)
No TCP/IP 2-way connection [Not OK]

======

0x3e000000
(MikroTik Eth1 -> laptop, crossover cable)
/bin/bash -c "dd if=/dev/zero bs=1M count=1000 | netcat 192.168.1.101 7777"
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 1589.44 s, 660 kB/s  [Not OK]

(laptop -> MikroTik Eth1, crossover cable)
dd if=/dev/zero bs=1M count=1000 | nc 192.168.1.121 7777
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 49.6707 s, 21.1 MB/s  [OK]

======

0x6f000000
(MikroTik Eth1 -> laptop, crossover cable)
/bin/bash -c "dd if=/dev/zero bs=1M count=1000 | netcat 192.168.
1.101 7777"
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 176.224 s, 6.0 MB/s  [OK?]

(laptop -> MikroTik Eth1, crossover cable)
dd if=/dev/zero bs=1M count=1000 | nc 192.168.1.121 7777
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 50.7454 s, 20.7 MB/s  [OK]


Any word on the bootloader patch?  Happy to test more values as needed.

Sincerely,
George


> I have not really identified any noticeable performance impact from
> this effect through.
>
> Would be nice to hear some logic why 0x6f is the right value.
>
>> From the discussion it looks a bit like trial and error.
>>
>
> Cheers
> Toerless
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/openwrt-devel/attachments/20151015/c252a5c3/attachment.htm>
-------------- next part --------------
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel