DR800 usbnet (USB networking to make development easier) for DR800

[GregorRichards]

Originally Posted by Iñigo

It works perfectly with my DR800S.

*whew*

After it didn't work on the DR1000, I was afraid I'd made something that for some reason only works on my device :P

[owl123]

Gregor: Despite the segmentation fault, the first time I ran the app there were some keys being generated in the program. Then the usual linux prompt/console appeared enabling me to put in some commands.

Now it's just the errors appearing and then the console. When I try executing 'route add default gw 192.168.200.1' there's a 'network unreachable error' being returned.

[GregorRichards]

The script always continues in spite of errors, the seg fault is still fatal in terms of actually getting the network to work ... that error message confirms that the interface failed to load. I'm going to hack up a custom script tonight for you to replace usbnet2 with, which will make some output that I may be able to do some debugging with (but don't get your hopes up)

And the fact that the keygen doesn't appear again is simply because the keys only need to be generated once.

[euzen]

thank you!
And I connect DR800SG via WindowsXP.
here the Drivers.

[Iñigo]

Originally Posted by GregorRichards

Fair enough. Basically what you'd have to do is configure your host to do IP forwarding (honestly I don't know how this is done on any modern system ... ), then on the DR:

route add default gw 192.168.200.1
echo 'nameserver <some nameserver>' > /etc/resolv.conf

You could put those two parts in the usbnet2 script to avoid rewriting them.

(Haven't tested this, YMMV)

Instructions for your PC (linux only), type on command line as root:

# echo 1 > /proc/sys/net/ipv4/ip_forward
# /sbin/iptables -t nat -F
# /sbin/iptables -t nat -A POSTROUTING -j SNAT -o eth0 --to your_pc_external_ip

if you use wifi, use wlan0 instead of eth0 (type ifconfig to check if the proper wifi device is wlan0 or other).
your_pc_external_ip should be the PC own IP, not the IP of usb0 (the IP shown for that device in ifconfig command output).

And then, on your DR, the commands GregorRichards wrote above:

# route add default gw 192.168.200.1
# echo 'nameserver <some nameserver>' > /etc/resolv.conf

On my DR800S with firmware 2.0beta I can't use erbrowse to navigate internet. The error message is "This device does not support opening internet links; cannot open http://www.google.cat/". But, from the system point of view I can't see any limitation, so once midori or other browser is ported, we could navigate internet freely (as in freedom ;-).

[GregorRichards]

Originally Posted by kapoira

Can you do the reverse way?

to put a aplication of X in the iliad, many thanks in advance.

Excluding the fact that this runs on the DR and not the iLiad, the answer can only be "yes and no". You could get X11 forwarding, but any host app wouldn't know how to do the necessary screen refreshes, so in spite of the fact that it would be drawing into the internal framebuffer, you would never see it drawn on the screen unless you put a LOT of work into it.

[sigius]

Originally Posted by GregorRichards

Another use I discovered for this is to run an X server on the host and connect to it from the iRex. This effectively isolates slowdown from the processing speed from display slowdown, so you can figure out what problems are where. To do this, on the host:

Xephyr :1 -ac -screen 1024x768 -host-cursor

Then on the guest:

DISPLAY=192.168.200.1:1 run-parts /etc/X11/Xsession.d/

It's kinda funny seeing this GUI on my desktop :P. Also, argh, getting screen updates done efficiently and at the right time is really tough >_>

Thanks, Gregor, great post!
It works for me. But I got some errors&warnings on DR800 side:

root@dr800:~# DISPLAY=192.168.200.1:1 run-parts /etc/X11/Xsession.d/
Error invoking bind: Address already in use
mbpixbuf: unable to use XShm. DISPLAY remote?
[settings daemon] Forking. run with -n to prevent fork
(sysd_E)system.c:216,sys_set_services() --- sysset0 not found or not unique
(sysd_E)system.c:228,sys_set_services() --- ionpower not found or not unique
(sysd_E)system.c:240,sys_set_services() --- serio not found or not unique
Error invoking bind: Address already in use
Error invoking bind: Address already in use
(popupmenu_W)main.c:171,main() gtk_rc_parse [/usr/share/popupmenu/popupmenu.rc]
(sysd_W)system.c:878,sys_get_battery() Failed to read battery information
(ctb:2452): Gtk-CRITICAL **: _gtk_tree_view_find_path: assertion `node != NULL' failed
(ctb_E)filetypes.c:469,get_icon_from_file() --- cannot load iconfile [/usr/share/ctb/icon-books-mini.png] error [Failed to open file '/usr/share/ctb/icon-books-mini.png': No such file or directory]

And following message every ~6 seconds:

(sysd_W)system.c:878,sys_get_battery() Failed to read battery information

It seems like some troubles in access to hardware info.
Have you any idea about it?

[sigius]

Ok, I undestood,because of
mount -o bind /tmp/usbnetempty /sys/devices

[scotsman]

Hi there,

I just wanted to ask if you maybe ported your g_ether module to the "new" 2.0 RC3 firmware. That would increase the chances of running it on the DR1000 as well.

Many thanks!

[Iñigo]

Originally Posted by scotsman

Hi there,

I just wanted to ask if you maybe ported your g_ether module to the "new" 2.0 RC3 firmware. That would increase the chances of running it on the DR1000 as well.

Many thanks!

It works with 2.0RC3 in the DR800.

[Mackx]

I think there is a difference between the DR800 and DR1000 kernel. If I remember well this was written here by Matthijs or Gertjan from iRex. This might be the reason that the current procedure works for DR800 and not for DR1000. Maybe some extra steps will help?

[Iñigo]

Originally Posted by Mackx

I think there is a difference between the DR800 and DR1000 kernel. If I remember well this was written here by Matthijs or Gertjan from iRex. This might be the reason that the current procedure works for DR800 and not for DR1000. Maybe some extra steps will help?

Yes, I know it. What I meant is that usbnet didn't work for DR1000 with previous firmware versions.
My limited English and scarce free time made me to explain badly.

[scotsman]

Thanks for following up on the DR1000 issue. I am running 2.0RC3, and I get the same segmentation fault error that was reported here.
Would it help if I get you the dmesg output before and after insmodding the module?

[scotsman]

OK, what I get is this:
execute: modprobe arcotg_usb
this gives me output
ARC USBOTG Device Controller driver (1 August 2005)
usbotg_active: usbotg_init()
and no further line in dmesg output

when I then insmod g_ether.ko,
I get a segmentation fault immediately
dmesg -c gives me

g_ether: no version for "struct_module" found: kernel tainted.
Unable to handle kernel NULL pointer dereference at virtual address 0000009c
pgd = c72cc000
[0000009c] *pgd=872b0031, *pte=00000000, *ppte=00000000
Internal error: Oops: 817 [#1] PREEMPT
Modules linked in: g_ether(F) arcotg_udc
CPU: 0 Tainted: GF (2.6.24-irex #2000)
PC is at eth_bind+0x290/0x6cc [g_ether]
LR is at alloc_netdev_mq+0xc4/0xdc
pc : [<bf007bdc>] lr : [<c01900bc>] psr: a0000013
sp : c72bde68 ip : c72bde30 fp : c72bdeb4
r10: c72a9540 r9 : c72a94c0 r8 : c7ffda00
r7 : bf00bfc0 r6 : 00000082 r5 : c7dfd800 r4 : 00000000
r3 : 00000000 r2 : 000000a0 r1 : c025f837 r0 : c7dfd800
Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment user
Control: 00e5387f Table: 872cc000 DAC: 00000015
Process insmod (pid: 2296, stack limit = 0xc72bc260)
Stack: (0xc72bde68 to 0xc72be000)
de60: c027bb9a bf002570 00000001 00000028 c72a9248 0000002a
de80: c72bde9c c72a9480 c007234c c72651e0 bf00bf2c bf0048c8 00000001 00000028
dea0: c72a9248 0000002a c72bded4 c72bdeb8 bf000c70 bf007958 00000001 c72651e0
dec0: 00000001 bf00c220 c72bdee4 c72bded8 bf00e014 bf000b90 c72bdfa4 c72bdee8
dee0: c00667a0 bf00e00c 00000000 00000000 00000000 00000000 00000000 00000000
df00: 00000000 00000000 0000000d 00000000 00000000 00000000 0000000a 00000000
df20: 00000000 00000000 0000000b c88f62c0 c72cb680 c88decb9 c8898000 c88df448
df40: c88df0b0 c008dcc8 00000141 00000141 c72bc000 bf00c268 00000027 c88df470
df60: bf00c22c c88df498 00000029 c01f6174 c0269080 00000000 c72bdfa4 00000000
df80: 00080000 40149008 00000080 c0022084 c72bc000 00000003 00000000 c72bdfa8
dfa0: c0021f00 c00651a0 00000000 00080000 40149008 0005ef31 00012008 40146480
dfc0: 00000000 00080000 40149008 00000080 00012008 00000000 00000003 beeaab04
dfe0: beeaad69 beeaaacc 00008e30 400e5db4 60000010 40149008 00000000 00000000
Backtrace:
[<bf00794c>] (eth_bind+0x0/0x6cc [g_ether]) from [<bf000c70>] (usb_gadget_register_driver+0xec/0x210 [arcotg_udc])
[<bf000b84>] (usb_gadget_register_driver+0x0/0x210 [arcotg_udc]) from [<bf00e014>] (init+0x14/0x1c [g_ether])
r6:bf00c220 r5:00000001 r4:c72651e0
[<bf00e000>] (init+0x0/0x1c [g_ether]) from [<c00667a0>] (sys_init_module+0x160c/0x16ac)
[<c0065194>] (sys_init_module+0x0/0x16ac) from [<c0021f00>] (ret_fast_syscall+0x0/0x2c)
Code: 0a0000d5 e5954194 e3a03000 e28420a0 (e584309c)
---[ end trace a725123fb95d8b82 ]---

when I then lsmod, I get
Module Size Used by
g_ether 25672 1
arcotg_udc 18636 1 g_ether

Note that g_ether is listed as used. I then cannot remove g_ether by rmmod g_ether, as I then get an error that the module is in use

[scotsman]

Have a look at all the modules provided with firmware 2.0RC3 (modprobe -l)

/lib/modules/2.6.24-irex/kernel/lib/crc-ccitt.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/ppp_deflate.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/slhc.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/ppp_generic.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/ppp_async.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/mii.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/usb/pegasus.ko
/lib/modules/2.6.24-irex/kernel/drivers/net/usb/usbnet.ko
/lib/modules/2.6.24-irex/kernel/drivers/usb/core/usbcore.ko
/lib/modules/2.6.24-irex/kernel/drivers/usb/host/ehci-hcd.ko
/lib/modules/2.6.24-irex/kernel/drivers/usb/gadget/g_file_storage.ko
/lib/modules/2.6.24-irex/kernel/drivers/usb/gadget/arcotg_udc.ko
/lib/modules/2.6.24-irex/kernel/drivers/usb/gadget/gadgetfs.ko
/lib/modules/2.6.24-irex/kernel/drivers/w1/slaves/w1_smem.ko
/lib/modules/2.6.24-irex/kernel/drivers/w1/wire.ko
/lib/modules/2.6.24-irex/kernel/drivers/w1/masters/mxc_w1.ko

Three things strike me as peculiar:
1) look at the comments in the usb networking section in /etc/network/interfaces:

# Ethernet/RNDIS gadget (g_ether)
# ... or on host side, usbnet and random hwaddr
auto usb0
iface usb0 inet static
address 192.168.0.202
netmask 255.255.255.0
network 192.168.0.0
gateway 192.168.0.200

It mentions g_ether, but the firmware does not include the module. Maybe they have it in some developer version / kit?

2) pegasus.ko is a module that drives an usb LAN adaptor, if I found that correctly. Makes little sense to me to include that module but not g_ether

3) usbnet.ko is the module that drives the host side of a usb networking connection. So I tried to load this module and network with my Nokia N810 which runs g_ether and USB networks fine with my laptop. The module loads without fault, but the Nokia would not detect a device at the other side, nor would the DR1000 detect the Nokia when I execute "ifup usb0". So it seems that the usbnet module does not work (which is no suprise, as it is supposedly not meant to drive USB OTG devices -- but why is it included here?).

OK, so far some more data. I hope you folk can make good use of it and figure something out. Do let me know if there is more I can do in terms of trying out modules, documenting output etc. And many thanks for your efforts!