maemo.org - Talk

maemo.org - Talk (https://talk.maemo.org/index.php)
-   Applications (https://talk.maemo.org/forumdisplay.php?f=41)
-   -   N900 as a Wifi Hotspot, part 2 (https://talk.maemo.org/showthread.php?t=48075)

5416u0d 2010-06-02 22:59

Re: N900 as a Wifi Hotspot, part 2
 
thanks. what repo is it in? i have extras, extras-devel, and testing but the package manager and apt are both not able to find iptables or wireless tools. maybe they arent available for pr1.2 yet?

5416u0d 2010-06-03 00:56

Re: N900 as a Wifi Hotspot, part 2
 
got it! forgot about apt-get update.

Kasperl 2010-06-03 07:50

Re: N900 as a Wifi Hotspot, part 2
 
I'm currently running whatever kernel appears in the repos as the power management kernel, since installing the other kernel didnt work due to a unresolved dependancy (fiasco-flasher). starting the hotspot works noth from command.ine and gui, connecting from two debian lenny macgines as well, but it wont dhcp and filling weverything in via ifconfig on the laptops doesnt work either.

debugging steps?

(sorry for spelling and such, posting from my n900)

rambo 2010-06-04 10:12

Re: N900 as a Wifi Hotspot, part 2
 
Quote:

Originally Posted by Kasperl (Post 697208)
I'm currently running whatever kernel appears in the repos as the power management kernel, since installing the other kernel didnt work due to a unresolved dependancy (fiasco-flasher). starting the hotspot works noth from command.ine and gui, connecting from two debian lenny macgines as well, but it wont dhcp and filling weverything in via ifconfig on the laptops doesnt work either.

Works noth ? Is this "works not" or what ? post output for CLI backend start. post output of "uname -a"

kodomo 2010-06-04 16:04

Re: N900 as a Wifi Hotspot, part 2
 
Quote:

Originally Posted by meepha12 (Post 687136)
/home/user # /usr/sbin/mobilehotspot_backend start ; echo$?DEBUG: executing uname -r
DEBUG: executing uname -r
DEBUG: executing lsmod
DEBUG: executing insmod /lib/modules/2.6.28.10power37/nf_conntrack.ko
DEBUG: executing lsmod
DEBUG: executing insmod /lib/modules/2.6.28.10power37/nf_defrag_ipv4.ko
DEBUG: executing lsmod
DEBUG: executing insmod /lib/modules/2.6.28.10power37/nf_conntrack_ipv4.ko
DEBUG: executing lsmod
DEBUG: executing lsmod
DEBUG: executing lsmod
DEBUG: executing insmod /lib/modules/2.6.28.10power37/nf_nat.ko
DEBUG: executing lsmod
DEBUG: executing insmod /lib/modules/2.6.28.10power37/iptable_nat.ko
DEBUG: executing lsmod
DEBUG: executing insmod /lib/modules/2.6.28.10power37/ipt_MASQUERADE.ko
ERROR:dbus.proxies:Introspect error on :1.86:/com/nokia/icd2: dbus.exceptions.DBusException: org.freedesktop.DBus.Error.NotSupported: Unsupported interface or method
icd2_state_listener got args:

(dbus.String(u''), dbus.UInt32(0L), dbus.String(u''), dbus.String(u'WLAN_INFRA'), dbus.UInt32(83886080L), dbus.ByteArray('4c0d54c3-f934-40ef-9477-7b41415a8804\x00'), dbus.String(u''), dbus.UInt32(2L))
state['icd']:

{'old_connection_type': 'WLAN_INFRA', 'old_connection_name': '4c0d54c3-f934-40ef-9477-7b41415a8804'}
DEBUG: executing ifconfig wlan0 down
DEBUG: executing iwconfig wlan0 mode ad-hoc
DEBUG: executing ifconfig wlan0 up
DEBUG: executing iwconfig wlan0 essid "MobileHotSpot"
DEBUG: executing ifconfig wlan0 10.23.136.1 netmask 255.255.255.0 up
DEBUG: executing /usr/sbin/dnsmasq -i wlan0 -a 10.23.136.1 -I lo -z -x /var/run/dnsmasq.wlan0.pid --dhcp-range=10.23.136.10,10.23.136.100,6h --dhcp-option=3,10.23.136.1 --dhcp-option=6,10.23.136.1
DEBUG: executing iptables --flush
DEBUG: executing iptables --flush -t nat
DEBUG: executing iptables -P FORWARD ACCEPT
DEBUG: executing iptables -P INPUT ACCEPT
DEBUG: executing iptables -P OUTPUT ACCEPT
DEBUG: executing iptables --table nat --append POSTROUTING --out-interface gprs0 -j MASQUERADE
/bin/sh: echo0DEBUG:: not found

Hi Folks!
I've been wondering about one thing: Was there a conscious decision not to avoid IP packet fragmentation?
IP packet fragmentation involves quite some overhead and to my experience there seem some bugs and providers blocking fragmented packets out in the wild creating spurious connectivity problems.

If not, I'd like to propose following the common practice of changing the TCPMSS, avoiding unnecessary fragmentation at least for TCP packets:
iptables -A FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu
(doesn't work right now on the N900, as the corresponding module seems to be missing in the power user kernel - maybe someone could add it? :P )

(For those who don't know what I'm talking about: look at the MTU of your wlan0 and gprs0 interface - the latter will probably be smaller - and consider this explanation of ip fragmentation in the context of tunnels :) :
http://www.cisco.com/en/US/tech/tk82...800d6979.shtml )

raisemyspirit 2010-06-04 16:55

Re: N900 as a Wifi Hotspot, part 2
 
@Kodomo:

Agreed 100%

But is it missing in the kernel ... ?

CONFIG_IP_NF_MATCH_TCPMSS=y
CONFIG_IP_NF_TARGET_TCPMSS=y

Or in the iptables userspace modules ... ?

libipt_TCPMSS.so

Please check that briefly.

I am still running PR1.1.1 and using the device productively, so when I can afford some time to play I will update and go to an enhanced kernel myself.

In dieser Stelle ein dickes Danke! an Tom Tanner für seine Arbeit am Kernel. (A big thank you to Tom Tanner for his work).

Quote:

Originally Posted by kodomo (Post 699744)
Hi Folks!
I've been wondering about one thing: Was there a conscious decision not to avoid IP packet fragmentation?
IP packet fragmentation involves quite some overhead and to my experience there seem some bugs and providers blocking fragmented packets out in the wild creating spurious connectivity problems.


rambo 2010-06-04 18:37

Re: N900 as a Wifi Hotspot, part 2
 
Quote:

Originally Posted by raisemyspirit (Post 699802)
@Kodomo:

Agreed 100%


No decision, initially focus was just getting it working somehow and then many features were left disabled, ask titan and the needed moduless will be in the power kernel and once that's done I can add the options to the hotspot backend script.

kodomo 2010-06-05 11:26

Re: N900 as a Wifi Hotspot, part 2
 
Quote:

Originally Posted by raisemyspirit (Post 699802)
@Kodomo:

Agreed 100%

But is it missing in the kernel ... ?

CONFIG_IP_NF_MATCH_TCPMSS=y
CONFIG_IP_NF_TARGET_TCPMSS=y

Or in the iptables userspace modules ... ?

libipt_TCPMSS.so

Please check that briefly.

I am still running PR1.1.1 and using the device productively, so when I can afford some time to play I will update and go to an enhanced kernel myself.

In dieser Stelle ein dickes Danke! an Tom Tanner für seine Arbeit am Kernel. (A big thank you to Tom Tanner for his work).

I'm using 2.6.28.10power37 , and the iptables command fails. There's neither libipt_TCPMSS.so , nor xt_TCPMSS.so (think, I remember that there was a name change in recent kernels)

JayBEE 2010-06-05 18:24

Re: N900 as a Wifi Hotspot, part 2
 
Rambo,

Thanks for getting this going! I'm up and running with it and life is good! Also thanks to Flynx for his modified backend.py!

Now if we can get Flash 10.1, life with my N900 would be great!

Thanks again!

Kasperl 2010-06-07 07:18

Re: N900 as a Wifi Hotspot, part 2
 
2 Attachment(s)
Quote:

Originally Posted by rambo (Post 699293)
Works noth ? Is this "works not" or what ? post output for CLI backend start. post output of "uname -a"

noth=both, I nearly hit the right keys.

Output attached.


All times are GMT. The time now is 16:17.

vBulletin® Version 3.8.8