freebsd-arm Digest, Vol 740, Issue 7

[Oleksandr Tymoshenko]

Mark Millard (marklmi at yahoo.com) wrote:
> 
> 
> On 2020-Jul-6, at 13:47, Oleksandr Tymoshenko <gonzo at bluezbox.com> wrote:
> 
> > Furkan Salman (furkan at fkardame.com) wrote:
> >> Hello Peter,
> >> 
> >> I have rockpiE which is somewhat similar to Rock64, If s133pwa1k9r@ or gonzo@ can confirm if rockpie can be used to test RK3328 Lan issue then I am happy to help with testing.
> > 
> > 
> > Hi Furkan,
> > 
> > Yes, RockPi E seems to be a good test target. If you could check the
> > GigE interface before and after the patch. Whether it works/doesn't work
> > and if it works in both cases - try testing performance with iperf3,
> > just to see if performance was affected in any way.
> 
> For folks not familiar with the general type of activity
> or specifically with iperf3 (or other specifics), more
> detailed information to "collect and report . . ., collecting
> the information via the commands .  . ." could help: more
> step-by-step.

Good  idea. To test inteface performance you need two instances of the
iperf3 program: server and client. Server can be FreeBSD, Linux, or
macOS machine, shouldn't matter. Should be connected to the wired
network though, as close to the board as possible, to make sure we're
testing the board's interface and not the narrowest link between client
and server.

On the server machine install iperf3 package and run iperf -s
On the board run iperf3 -c <serverip>. This is normal mode:
client sends as much data as possible and server receives,
so the results are TX speed. Output should look like this:

Connecting to host 192.168.10.1, port 5201
[  5] local 192.168.10.117 port 56650 connected to 192.168.10.1 port 5201
[ ID] Interval           Transfer     Bitrate         Retr  Cwnd
[  5]   0.00-1.00   sec  55.3 MBytes   464 Mbits/sec    0    403 KBytes
[  5]   1.00-2.00   sec  54.6 MBytes   458 Mbits/sec    0    567 KBytes
[  5]   2.00-3.00   sec  54.2 MBytes   455 Mbits/sec    0    692 KBytes
[  5]   3.00-4.00   sec  54.0 MBytes   453 Mbits/sec    0    798 KBytes
[  5]   4.00-5.00   sec  54.1 MBytes   454 Mbits/sec    0    892 KBytes
[  5]   5.00-6.00   sec  54.5 MBytes   457 Mbits/sec    0    927 KBytes
[  5]   6.00-7.00   sec  54.5 MBytes   457 Mbits/sec    0    927 KBytes
[  5]   7.00-8.00   sec  54.7 MBytes   459 Mbits/sec    0    929 KBytes
[  5]   8.00-9.00   sec  54.7 MBytes   459 Mbits/sec    0    929 KBytes
[  5]   9.00-10.00  sec  54.3 MBytes   455 Mbits/sec    0    929 KBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-10.00  sec   545 MBytes   457 Mbits/sec    0             sender
[  5]   0.00-11.15  sec   545 MBytes   410 Mbits/sec                  receiver

To test RX speed run "iperf3 -R -c <serverip>". -R is for reverse mode:
server sends and client receives. The output format is simmilar:

root@:/ # iperf3 -R -c 192.168.10.1
Connecting to host 192.168.10.1, port 5201
Reverse mode, remote host 192.168.10.1 is sending
[  5] local 192.168.10.117 port 36908 connected to 192.168.10.1 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-1.00   sec  18.7 MBytes   156 Mbits/sec
[  5]   1.00-2.01   sec  8.31 MBytes  69.0 Mbits/sec
[  5]   2.01-3.01   sec  18.3 MBytes   154 Mbits/sec
[  5]   3.01-4.00   sec  20.3 MBytes   171 Mbits/sec
[  5]   4.00-5.01   sec  9.92 MBytes  82.6 Mbits/sec
[  5]   5.01-6.00   sec  20.3 MBytes   171 Mbits/sec
[  5]   6.00-7.01   sec  11.4 MBytes  95.5 Mbits/sec
[  5]   7.01-8.00   sec  22.3 MBytes   188 Mbits/sec
[  5]   8.00-9.00   sec  20.8 MBytes   174 Mbits/sec
[  5]   9.00-10.00  sec  20.2 MBytes   170 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-11.00  sec   171 MBytes   130 Mbits/sec  1789             sender
[  5]   0.00-10.00  sec   170 MBytes   143 Mbits/sec                  receiver

As you can see, download on my Firefly-RK3399 is much
slower and high number of Retr indicates either packet
corruption or packets dropped.
 
> Also: Do you care between debug kernels vs. non-debug
> kernels? Debug ones of the appropriate vintage for head
> are available via artifacts.ci.freebsd.org but there
> might be performance consequences to using such.

TX speed difference between GENERIC-NODEBUG and GENERIC
on the Firefly is ~40%, so I'd prefer results for NODEBUG,
but even with DEBUG kernel it'd be nice to know if there is
any effect of the commit.

-- 
gonzo