November 15, 2019, 08:37:33 pm

News:

Have you visited the Allwinner Chipset wiki? - http://linux-sunxi.org/


Buster, Xenial and Bionic - updated 6.7.2019

Started by slovenia, December 24, 2013, 03:23:11 pm

Previous topic - Next topic

arrhenius

Slovenia, tested your latest script from github and works perfect  :D

thanks.

fairlane

Hello,

I use your image and this is work fine.
I would move the rootfs to my SATA drive, but this image has one fs (boot + root in one )
Do you have an advice how to make that?

thank you,
fairlane

slovenia

Quote from: fairlane on January 11, 2014, 03:18:25 pm
Hello,

I use your image and this is work fine.
I would move the rootfs to my SATA drive, but this image has one fs (boot + root in one )
Do you have an advice how to make that?

thank you,
fairlane


Anyway you can not boot directly from SATA ... so copy rootfs to SATA drive and change(for example first sata partition) root=/dev/sda1 in /boot/uEnv.txt on your SDcard

How to copy:

Create install-exclude with this content:
/dev/*
/proc/*
/sys/*
/media/*
/mnt/*
/run/*
/tmp/*

rsync -aH --exclude-from=install-exclude  /  /mnt

where /mnt is already mounted SATA drives partion.

reboot

Debian and Ubuntu images with kernel 3.4.110, 4.3.3, 4.4
http://www.armbian.com

Jono

v1.4 eth0 network is not work
even if manually set ip address, it is not in the network

slovenia

I test each build - it should work. The only thing which is changed from previous releases and can be related is removing/commenting sunxi_gmac in /etc/modules .... because Ethernet driver is build in kernel now.

Try enabling it (remove #, save & reboot) or let's debug a little - What you get if you try this:

dmesg | grep gmac
ifconfig
cat /etc/network/interfaces
Debian and Ubuntu images with kernel 3.4.110, 4.3.3, 4.4
http://www.armbian.com

maxx

Hi slovenia,

thanks a lot for your build. Using CT VGA v1.4, that works fine for me.
Testing your NAND install script, really smooth and cool. Thanks.

No problems here, to get eth0 working so far.


# dmesg | grep gmac
[    1.917432] [gmac]: sun6i_gmac platform driver registration completed
[   13.872947] sunxi_gmac: probed
[   13.886049] eth0: PHY ID 001cc915 at 0 IRQ 0 (sunxi_gmac-0:00) active
[   13.901863] eth0: PHY ID 001cc915 at 1 IRQ 0 (sunxi_gmac-0:01)
[   15.912031] PHY: sunxi_gmac-0:00 - Link is Up - 1000/Full



# ifconfig eth0
eth0      Link encap:Ethernet  HWaddr ae:50:30:27:5a:cf 
          inet addr:172.16.10.201  Bcast:172.16.10.255  Mask:255.255.255.0
          inet6 addr: 2001:470:1453:0:ac50:30ff:fe27:5acf/64 Scope:Global
          inet6 addr: fe80::ac50:30ff:fe27:5acf/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:3838  Metric:1
          RX packets:2273839 errors:0 dropped:0 overruns:0 frame:0
          TX packets:6931039 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:150098081 (143.1 MiB)  TX bytes:1849894882 (1.7 GiB)
          Interrupt:117 Base address:0xc000



# cat /etc/network/interfaces
# interfaces(5) file used by ifup(8) and ifdown(8)
auto lo
iface lo inet loopback
auto eth0
allow-hotplug eth0
iface eth0 inet dhcp
        hwaddress ether AE:50:30:27:5A:CF # change this
        pre-up /sbin/ifconfig eth0 mtu 3838 # setting MTU for DHCP, static just: mtu 3838
#auto wlan0
#allow-hotplug wlan0
#iface wlan0 inet dhcp
#    wpa-ssid SSID
#    wpa-psk xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
# to generate proper encrypted key: wpa_passphrase yourSSID yourpassword



But I notice, a lacking/poor network performance, tested with "iperf".


# iperf -c 172.16.10.104 -t 60
------------------------------------------------------------
Client connecting to 172.16.10.104, TCP port 5001
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  3] local 172.16.10.201 port 47314 connected with 172.16.10.104 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-60.0 sec  3.11 GBytes   446 Mbits/sec

# uname -a
Linux cubietruck 3.4.75-sun7i+ #1 SMP PREEMPT Sun Jan 12 14:14:31 CET 2014 armv7l GNU/Linux


Double-check with a other ARM based machine:

> iperf -c 172.16.10.104 -t 30     
------------------------------------------------------------
Client connecting to 172.16.10.104, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[  3] local 172.16.10.12 port 46826 connected with 172.16.10.104 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-30.0 sec  3.23 GBytes    925 Mbits/sec

> uname -a
Linux DS112 2.6.32.12 #3810 Mon Dec 2 17:14:23 CST 2013 armv5tel GNU/Linux synology_88f6282_112+


So, I fiddle around, whats going on.
"htop" show me, while "iperf" is running, only one cpu is working now.


cpu1  :  0.0% sys:100.0% low:  0.0%                              Tasks: 32, 11 thr; 2 running
cpu2  :  1.3% sys:  2.0% low:  0.0%                                Load average: 0.60 0.19 0.16
cpu  Avg:  0.7% sys: 51.5% low:  0.0%                            Hostname: cubietruck
  1  [||||||||||||||||100.0%]   2  [||                3.3%]     


Hmmm ...  :-[

Could you please have a look on it or check yourself?

Thanks in anticipation.

cu
maxx

Jono

January 13, 2014, 01:18:37 pm #66 Last Edit: January 13, 2014, 01:30:21 pm by Jono
Quote from: slovenia on January 12, 2014, 01:34:03 pm
I test each build - it should work. The only thing which is changed from previous releases and can be related is removing/commenting sunxi_gmac in /etc/modules .... because Ethernet driver is build in kernel now.

Try enabling it (remove #, save & reboot) or let's debug a little - What you get if you try this:

dmesg | grep gmac
ifconfig
cat /etc/network/interfaces

I put your previous assembly and the network is not raised machine.
Configuration options were and to give dynamic ip or static ip (Cube was not in the network)
I do not have usb keyboard, all settings spend only through ssh.
Problem was solved by booting from sd card (cubuntu):
-connected partition roofts on Nand
mount /dev/nandb /mnt/rootfs
chroot /mnt/rootfs

-edited interfaces nano /etc/network/interfaces
# interfaces(5) file used by ifup(8) and ifdown(8)
auto lo
iface lo inet loopback
auto eth0
allow-hotplug eth0
iface eth0 inet dhcp
hwaddress ether 8E:E7:E9:C1:83:93

-restarted the network interfaces by command /etc/init.d/networking restart
-umont rootfs
-reboot to flash
Cube is now visible on the net!

arrhenius

Quote from: maxx on January 13, 2014, 11:40:44 am

Could you please have a look on it or check yourself?

Thanks in anticipation.

cu
maxx



hi:

i tested myself , i get same results (474 Mbits), but i though it was not a bad result for a cubieboard. i tried then with 4 threads (-P 4 option ) and got 573 Mbits. the second cpu seems to work more, but i cant get both 100% working.

What arm machine do you use to get 900Mbit?


root@cubie:~# iperf -c 192.168.1.202
------------------------------------------------------------
Client connecting to 192.168.1.202, TCP port 5001
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  3] local 192.168.1.208 port 49794 connected with 192.168.1.202 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   566 MBytes   474 Mbits/sec
root@cubie:~# iperf -c 192.168.1.202 -P 4
------------------------------------------------------------
Client connecting to 192.168.1.202, TCP port 5001
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  5] local 192.168.1.208 port 49798 connected with 192.168.1.202 port 5001
[  4] local 192.168.1.208 port 49795 connected with 192.168.1.202 port 5001
[  6] local 192.168.1.208 port 49799 connected with 192.168.1.202 port 5001
[  3] local 192.168.1.208 port 49796 connected with 192.168.1.202 port 5001
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.0 sec   155 MBytes   129 Mbits/sec
[  6]  0.0-10.0 sec   176 MBytes   147 Mbits/sec
[  3]  0.0-10.0 sec   180 MBytes   151 Mbits/sec
[  5]  0.0-10.0 sec   175 MBytes   146 Mbits/sec
[SUM]  0.0-10.0 sec   685 MBytes   573 Mbits/sec
root@cubie:~#

ellnic

Nice work Slovenia. :D This might be just what I am after. Looks your build has decent speeds on the Ethernet. I also especially like the fact that you have provided the option for SD or NAND.. excellent :) I can't wait to test this out fully. Thx

maxx

January 13, 2014, 02:51:21 pm #69 Last Edit: January 13, 2014, 03:02:41 pm by maxx
Hi arrhenius,

thanks for checking and response.

Quote from: arrhenius on January 13, 2014, 01:42:45 pm
What arm machine do you use to get 900Mbit?


Quote from: maxx on January 13, 2014, 11:40:44 am
...
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-30.0 sec  3.23 GBytes    925 Mbits/sec

> uname -a
Linux DS112 2.6.32.12 #3810 Mon Dec 2 17:14:23 CST 2013 armv5tel GNU/Linux synology_88f6282_112+


It's as  Synology DS112+ NAS  (http://www.synology.com/products/spec/DS112+) with Marvell Kirkwood (ARM5) 2.0 GHz / 512 MB DDR3, 1 year old.  MTU 1500, no jumbo packets enabled.
Network with HP ProCurve 1800-24G GigaBit Switch. Remote station is a iMac / OS X 10.7.5 .

Thanks for the hint, never use option (-P 4) before. In my use case, we mostly have only one thread.
And yes, it get better now. Same here, with -P 4, maximum has been reached.


[
cpu 1  :  0.0% sys:100.0% low:  0.0%                              Tasks: 29, 14 thr; 4 running
cpu 2  :  1.3% sys: 55.8% low:  0.0%                              Load average: 0.66 0.16 0.11
cpu  Avg:  0.7% sys: 77.9% low:  0.0%                             Hostname: cubietruck
  1  [||||||||||||||||100.0%]   2  [||||||||||||||   57.1%]     



~$ iperf -c 172.16.10.104 -P 4 -t 60
------------------------------------------------------------
Client connecting to 172.16.10.104, TCP port 5001
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  4] local 172.16.10.201 port 47327 connected with 172.16.10.104 port 5001
[  3] local 172.16.10.201 port 47326 connected with 172.16.10.104 port 5001
[  6] local 172.16.10.201 port 47329 connected with 172.16.10.104 port 5001
[  5] local 172.16.10.201 port 47328 connected with 172.16.10.104 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-60.0 sec  1.09 GBytes   156 Mbits/sec
[  4]  0.0-60.0 sec  1.14 GBytes   164 Mbits/sec
[  6]  0.0-60.0 sec  1.21 GBytes   174 Mbits/sec
[  5]  0.0-60.0 sec  1.06 GBytes   152 Mbits/sec
[SUM]  0.0-60.0 sec  4.51 GBytes   645 Mbits/sec

~$ uname -a
Linux cubietruck 3.4.75-sun7i+ #1 SMP PREEMPT Sun Jan 12 14:14:31 CET 2014 armv7l GNU/Linux


Maybe somewhere is a possibility for network performance improvement.
Network sunxi_gmac modul, Kernel, ... ? I hope so.

=====
# dmesg | grep gmac
[    1.917432] [gmac]: sun6i_gmac platform driver registration completed
[   13.872947] sunxi_gmac: probed
[   13.886049] eth0: PHY ID 001cc915 at 0 IRQ 0 (sunxi_gmac-0:00) active
[   13.901863] eth0: PHY ID 001cc915 at 1 IRQ 0 (sunxi_gmac-0:01)
[   15.912031] PHY: sunxi_gmac-0:00 - Link is Up - 1000/Full
=====


cu
maxx


arrhenius

Quote from: maxx on January 13, 2014, 02:51:21 pm
Hi arrhenius,

thanks for checking and response.

Quote from: arrhenius on January 13, 2014, 01:42:45 pm
What arm machine do you use to get 900Mbit?


Quote from: maxx on January 13, 2014, 11:40:44 am
...
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-30.0 sec  3.23 GBytes    925 Mbits/sec

> uname -a
Linux DS112 2.6.32.12 #3810 Mon Dec 2 17:14:23 CST 2013 armv5tel GNU/Linux synology_88f6282_112+


It's as  Synology DS112+ NAS  (http://www.synology.com/products/spec/DS112+) with Marvell Kirkwood (ARM5) 2.0 GHz / 512 MB DDR3, 1 year old.  MTU 1500, no jumbo packets enabled.
Network with HP ProCurve 1800-24G GigaBit Switch. Remote station is a iMac / OS X 10.7.5 .

Thanks for the hint, never use option (-P 4) before. In my use case, we mostly have only one thread.
And yes, it get better now. Same here, with -P 4, maximum has been reached.


[
cpu 1  :  0.0% sys:100.0% low:  0.0%                              Tasks: 29, 14 thr; 4 running
cpu 2  :  1.3% sys: 55.8% low:  0.0%                              Load average: 0.66 0.16 0.11
cpu  Avg:  0.7% sys: 77.9% low:  0.0%                             Hostname: cubietruck
  1  [||||||||||||||||100.0%]   2  [||||||||||||||   57.1%]     



~$ iperf -c 172.16.10.104 -P 4 -t 60
------------------------------------------------------------
Client connecting to 172.16.10.104, TCP port 5001
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  4] local 172.16.10.201 port 47327 connected with 172.16.10.104 port 5001
[  3] local 172.16.10.201 port 47326 connected with 172.16.10.104 port 5001
[  6] local 172.16.10.201 port 47329 connected with 172.16.10.104 port 5001
[  5] local 172.16.10.201 port 47328 connected with 172.16.10.104 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-60.0 sec  1.09 GBytes   156 Mbits/sec
[  4]  0.0-60.0 sec  1.14 GBytes   164 Mbits/sec
[  6]  0.0-60.0 sec  1.21 GBytes   174 Mbits/sec
[  5]  0.0-60.0 sec  1.06 GBytes   152 Mbits/sec
[SUM]  0.0-60.0 sec  4.51 GBytes   645 Mbits/sec

~$ uname -a
Linux cubietruck 3.4.75-sun7i+ #1 SMP PREEMPT Sun Jan 12 14:14:31 CET 2014 armv7l GNU/Linux


Maybe somewhere is a possibility for network performance improvement.
Network sunxi_gmac modul, Kernel, ... ? I hope so.

=====
# dmesg | grep gmac
[    1.917432] [gmac]: sun6i_gmac platform driver registration completed
[   13.872947] sunxi_gmac: probed
[   13.886049] eth0: PHY ID 001cc915 at 0 IRQ 0 (sunxi_gmac-0:00) active
[   13.901863] eth0: PHY ID 001cc915 at 1 IRQ 0 (sunxi_gmac-0:01)
[   15.912031] PHY: sunxi_gmac-0:00 - Link is Up - 1000/Full
=====


cu
maxx


it looks like there are margin for improvement , but no idea how, i hope Patwood and Slovenia can figure it out :D


PD: Slovenia, in the forums you stated this is version 1.4 but in github it says VERSION="CTDebian 1.3" i think you forgot updating build.sh.


thanks everyone for sharing your work.



slovenia

Thanks for the feedback. In this release I switched to Patwood's kernel and haven't changed much in config. Just a CPU governor. I also saw some other measurements and people are getting the same speed with different/older kernel.

QuoteMaybe somewhere is a possibility for network performance improvement.
Network sunxi_gmac modul, Kernel, ... ? I hope so.


Probably. I hope the same. Currently I have no idea if this is the HW limit but I will pay attention to this matter. There is also one very lame problem - I don't have a gigabit LAN and I think Patwood doesn't have a CT :)

QuoteSlovenia, in the forums you stated this is version 1.4 but in github it says VERSION="CTDebian 1.3" i think you forgot updating build.sh.


Changed  :P

Debian and Ubuntu images with kernel 3.4.110, 4.3.3, 4.4
http://www.armbian.com

slovenia

Quote from: Jono on January 13, 2014, 01:18:37 pm
# interfaces(5) file used by ifup(8) and ifdown(8)
auto lo
iface lo inet loopback
auto eth0
allow-hotplug eth0
iface eth0 inet dhcp
hwaddress ether 8E:E7:E9:C1:83:93


MTU settings caused troubles? Strange ... I will comment it out in the future.

Thanks.
Debian and Ubuntu images with kernel 3.4.110, 4.3.3, 4.4
http://www.armbian.com

maxx

January 13, 2014, 05:01:00 pm #73 Last Edit: January 13, 2014, 06:24:01 pm by maxx
Hi slovenia,

Quote from: slovenia on January 13, 2014, 03:38:29 pm
Probably. I hope the same. Currently I have no idea if this is the HW limit but I will pay attention to this matter.


thanks for response. Good to hear, you will have a eye on this.

I wonder about the second load at " at 1 IRQ 0 (sunxi_gmac-0:01) " in dmesg, but it's will be not active.
Currently this cause the issue ... but how knows.  :-\

Quote
There is also one very lame problem - I don't have a gigabit LAN and I think Patwood doesn't have a CT :)


Should I send you such a 5-Port Gigabit Ethernet Switch

Cubieboard.org should send a CT to Patwood  for his awesome work!   :)

I like the way you and Patwood keep the ball rolling. It's definitely not your fault. So both keep on your  excellent work!

cu
maxx


patwood

I just received my ct a little over a week ago  ;D , and I have a GB switch.  Gmac improvements are a long ways off, though -- gotta figure out why BT need so much hand holding to start up.