rambo | # 51 | 2010-01-03, 20:57 | Report

I'm working (when I happen to have time) on something slightly more comprehensive: the mobile hotspot.

The hotspot_backend.py script will do the setup/teardown cleanly eventually (now it's a bit work in progress still and there is absolutely no UI [or any other way than editing the script] to select GPRS connection name etc)

As long as the modules are not loaded into the kernel they just use space (and not much of it either if you have only modules that you intend to use from time to time installed...)

iptables is small (400k deb) but I don't think there isn't any reason why it couldn't be optified.

In theory even the modules could be installed outside the rootfs but that's asking for trouble.

jebba | # 52 | 2010-01-03, 22:36 | Report

Yes, mobilehotspot is the way to go for the future. This was just quick and dirty.

I think you need to delete the "1" from this line in hotspot_backend.py.

Ya, could be optified. Most of that 400k is probably docs that don't get installed.

That is *definitely* asking for trouble. Won't be available at boot, for one....

P.S. Edit: ah, i think i see why you have that "1" there, so the wifi connection always gets .1 in the network?

---

Last edited by jebba; 2010-01-03 at 22:38. Reason: ah, i think i get it.

rambo | # 53 | 2010-01-04, 06:24 | Report

Yep, the self.network is missing the last octet of the IPv4 address (we need the network part for the interface and for dnsmasq DHCP setup)

edit self.connection -> self.network

R-R | # 54 | 2010-01-04, 06:40 | Report

Not too sure how ad-hoc differs from infrastructure mode... Is it only about power management or does it restrict it to only 1 user or something similar?

Anyway, this is going to be a killer feature, thanks to both of you :-)

rambo | # 55 | 2010-01-04, 22:13 | Report

Wikipedia:Wireless_LAN#Peer-to-peer

Number of users is not artificially limited but physical network constraints apply. The N900 wireless module refuses to go into infrastructure mode (as-in act as basestation, being a client to a basestation is different mode) anyway so it's a moot point.

The backend script should now do somewhat sane setup/teardown (it needs to run as root/sudo, eventually the packages will setup sudoers right), however I trigger some watchdog on my device when I connect to the hotspot and the device boots soon after... laptop got IP from dnsmasq though.

bmouring | # 56 | 2010-01-04, 23:52 | Report

Was this through use of iwconfig (supposedly not supported at all anymore, any device will return "not supported" if attempting this) or through iw (using the forced __ap type, useful for testing only as it doesn't expose what's needed) or through hostapd (the supported way of entering AP mode)? Link for reference

rambo | # 57 | 2010-01-05, 08:08 | Report

Ah, thanks for the info. I didn't go digging around then it failed with iwconfig and just supposed it cannot work (which proves again that supposion is the mother of all f**kups).

Anyways for this use case I see no problem using ad-hoc mode, though mesh-routing would be mighty cool to do at some point...

edit: typo fix

aboaboit | # 58 | 2010-01-05, 10:00 | Report

I filed bug 6097 for this a while ago.

bocaJ | # 59 | 2010-01-06, 07:40 | Report

Looks like it was changed to WONTFIX. Are there any alternatives?

rambo | # 60 | 2010-01-06, 11:15 | Report

I suggest reading the highly informative post by bmouring above...

edit: also INVALID, not WONTFIX, major difference...