Quantcast

Regression with ethernet receive speed

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Regression with ethernet receive speed

Scott Ellis
I noticed a regression in the Overo ethernet speeds.

3.2 kernel

RX 53.1 Mb/s
TX 93.2 Mb/s

2.6.34 kernel

RX 92.7 Mb/s
TX 94.1 Mb/s

Is this a known issue?

Same MLO and U-boot and the same hardware. I tried two sets of COM/
expansion board just to be sure.

I also tried with the new SPL U-Boot from a Yocto build with the same
results.

I used iperf for the testing.

If it's not a known problem, any ideas on where to start looking?

I already dropped some printks into board-overo.c after the smsc911x
initialization to verify the GPMC[5] registers still have the same values
that were set in u-boot and they do.

Scott
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Regression with ethernet receive speed

nikkadim
Did you use in both tests the same SD card for writing the transferred file?
The best way to test is use the /media/ram partition.

--
Rogoshchenkov Nikolay


On Thu, Jul 5, 2012 at 11:05 AM, jumpnowdev <[hidden email]> wrote:
I noticed a regression in the Overo ethernet speeds.

3.2 kernel

RX 53.1 Mb/s
TX 93.2 Mb/s

2.6.34 kernel

RX 92.7 Mb/s
TX 94.1 Mb/s

Is this a known issue?

Same MLO and U-boot and the same hardware. I tried two sets of COM/
expansion board just to be sure.

I also tried with the new SPL U-Boot from a Yocto build with the same
results.

I used iperf for the testing.

If it's not a known problem, any ideas on where to start looking?

I already dropped some printks into board-overo.c after the smsc911x
initialization to verify the GPMC[5] registers still have the same values
that were set in u-boot and they do.

Scott

--
View this message in context: http://gumstix.8.n6.nabble.com/Regression-with-ethernet-receive-speed-tp4964825.html
Sent from the Gumstix mailing list archive at Nabble.com.

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
gumstix-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/gumstix-users


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
gumstix-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/gumstix-users
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Regression with ethernet receive speed

Scott Ellis
There was no writing involved. iperf discards the packets.

But anyway, there were a number of SD cards involved.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Regression with ethernet receive speed

bhamadicharef
Scott

I just install iperf ... what settings you use to test ?

Brahim
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Regression with ethernet receive speed

Scott Ellis
Gumstix: 192.168.10.112
Workstation: 192.168.10.3

Testing RX speed

On Gumstix
root@overo:~# iperf -s -p5000

On Workstation
scott@hex:~$ iperf -c192.168.10.112 -p5000 -t60

Result
root@overo:~# iperf -s -p5000
------------------------------------------------------------
Server listening on TCP port 5000
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  4] local 192.168.10.112 port 5000 connected with 192.168.10.3 port 35403
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-60.0 sec    378 MBytes  52.9 Mbits/sec


Test TX speed

On Workstation
scott@hex:~$ iperf -s -p5000

On Gumstix
root@overo:~# iperf -c192.168.10.3 -p5000 -t60

Result
root@overo:~# iperf -c192.168.10.3 -p5000 -t60
------------------------------------------------------------
Client connecting to 192.168.10.3, TCP port 5000
TCP window size: 20.7 KByte (default)
------------------------------------------------------------
[  3] local 192.168.10.112 port 35872 connected with 192.168.10.3 port 5000
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-60.0 sec    666 MBytes  93.1 Mbits/sec


On older kernels you'll get results greater then 92Mbits/sec in both directions.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Regression with ethernet receive speed

bhamadicharef
My tests using the same iperf settings I get 79.5 Mbits/sec

VMware 12.04: 192.168.61.131
Gumstix: 192.168.1.134

root@omap3-multi:~# uname -a
Linux omap3-multi 3.0.0 #1 Tue Apr 24 16:17:43 SGT 2012 armv7l GNU/Linux

Brahim
Loading...