rmarcus | # 121 | 2010-01-29, 22:52 | Report

Keep up with the good job Rambo! We need you...your country needs you,,,

rambo | # 122 | 2010-01-30, 12:00 | Report

Crap, it seems the patched iphb module does not solve the reboot-on-traffick issue fully (and in fact it seems stopping iphbd does not either).

I can (with iphbd enabled and hotspot backend started) browse and download on the device as much as I can, but when connect the laptop and start causing traffic from there the load first jumps (kinda obvious, the kernel has plenty of work to do with routing and NAT there) and then after few seconds the device reboots. This seems to happen with iphbd stopped as well.

I got the flu (Talk about timing...) so I can't wrap my head around kernel core dumps now, if someone else could try with the kernel and modules from the mobilehotspot svn I would be grateful. To get the exact same patches checkout the patches dir from svn, "apt-get source kernel" (use clean dir) and from this directory call the hotspotize.sh in the patches checkout.

Meanwhile I'll see if I can do something about the GUI.

eber42 | # 123 | 2010-01-30, 14:50 | Report

I confirm i've got no more crashes with iphb kernel patch.

I've tried again with my kernel (with bluetooth, wlan & phone), and yours (only routing from wlan to wlan - I downloaded your latest binary modules deb and zImage)

I've used my usual test case: reloading all my firefox tabs (session manager tells me i've got 47 tabs) multiple times, and i've also tried to download a kernel source archive at the same time.

In both case, iphbd was always up and CPU usage was around 10% (just a quick look at top)

Just an issue with your kernel: I've got a lot of these:

Maybe there is another cause to your crashes ?

More testers would be nice (at least to make sure the reboot problem is solved or not)

aboaboit | # 124 | 2010-01-31, 11:01 | Report

Could you file a bug report for this issue and attach your patch?

hcm | # 125 | 2010-01-31, 15:25 | Report

Hi guys,

great work! Many thanks for your effort!

I just tried it out and would like to share my experience:

• The installation worked without any problems
• When starting the backend, I get DBUS-errors for icd and icd2:
  
  Code:

  ERROR:dbus.proxies:Introspect error on :1.79:/com/nokia/icd: dbus.exceptions.DBusException: org.freedesktop.DBus.Error.NotSupported: Unsupported interface or method

  (and the same for icd2). But the WLAN is available nevertheless…
• My N900 rebooted once, some 10 seconds after I started the backend the first time. I wasn't paying attention at that moment, so I'm not sure why. After that, no more reboots when starting the backend.
• When connecting to the network with another device, I experienced the problem that *sometimes* the DNS did not work… Sometimes I had to give a manual DNS (had to try several servers), sometimes it worked with autoassigning…

And then, some questions:

• Is it possible to connect multiple devices at the same time? The network shows up as ad-hoc, but apparently I was able to connect with two devices at the same time (but with DNS-problems, so I'm not 100% sure about it…)
• What changes are made in the custom kernel, apart from NAT-stuff and framebuffer? Is it likely that some functionality of the original kernel is missing or that power-consumption is higher (when not using the hotspot)?

keep up the work!
hcm

---

Last edited by hcm; 2010-01-31 at 15:30. Reason: forgot the reboot thing

rambo | # 126 | 2010-01-31, 16:45 | Report

Many thanks for testing

IcD[1,2] does not provide DBUS introspection, this sucks a lot in more ways than just getting this cosmetic problem, it's safe to ignore this error.

Sometimes I have problem with not getting IP from the DHCP server in dnsmasq, though then my laptop refuses to work with the n900 as router even if I assign ip etc manually.

WiFi Ad-Hoc mode itself has no artificial limit on how many devices may talk to each other, dnsmasq DHCP range in this case is 10-100 (which is way overkill but costs nothing).

Now there will be some practical issues when number of devices using the N900 as router goes above X (such as bandwidth, processor and other resource constraints), determining X will take some experimentation.

The configuration patch (apply this to the .config generated via "make rx51_defconfig") is in svn, but basically I have added:

• Framebuffer
• NAT related modules
• QoS modules

If all modules are unloaded correctly (big if...) on hotspot stop then in theory it should not consume (noticeably) more power, if the connection tracking stuff is loaded in kernel while the hotspot is no longer active I would guess network activity would take slightly more processing power than it normally would.

The QoS stuff might be removed, it adds size to the base kernel too and I'm not sure if we're going to use it for anything (if there was a good way to know the actual speed of the GPRS connection we could automagically run wondershaper to get nicer network performance)

In other news I have made some improvements to the backend and frontend, see updated instructions on project home page. NOTE: for debugging any backend issues it's best to run the hotspot_backend.py manually.

hcm | # 127 | 2010-01-31, 19:10 | Report

ok with the new version, the first icd-error has vanished.

but right after the icd2-error, the script hangs sometimes (I canceled after like 4 minutes…)
Apparently, it only hangs if I'm not connected to either GPRS or WLAN (but not always, when I'm not connected). If it hangs, I can go to the connection-selection menu (no need to select a connection, just go there) and the script continues.

The next thing I noticed: When I didn't have a connection before, the backend closes the connection when I stop it. But when I wanted to connect to GPRS manually after that, the GPRS-Connection was not there, just the WLANs!
This was reproducible quite a few times (5 or so), but the 6th time, everything was fine again ^^

regards
hcm

rambo | # 128 | 2010-01-31, 20:48 | Report

IcD1 interface is no longer instantiated (it wasn't really used for anything usefull earlier)

Stupid me, hadn't though about what happens when IcD2 state_req returns 0 (though as I understand the API docs it should send a signal even if there are no connection active), also added a SIGALARM to make sure that we get out of the DBUS mainloop.

The reason opening the connection menu solves this is the connections scan triggers state signals when the scan is done. And thus the state signal reaches listener in hotspot_backend.py which can then continue and exit the mainloop where it was waiting for a signal to arrive

After getting out of the loop by opening the connection menu ? Not that it should matter.

hcm | # 129 | 2010-01-31, 22:03 | Report

great! it doesn't hang anymore. and my GPRS connection is also still present. I'm not sure at which condition it disappeared before...

perhaps I can do some more testing tomorrow. I would like to try a kernel with minimal changes (i.e. only absolutely needed changes).

And should it be possible to activate WEP just by configuring a key in the backend script? I would like to try that also, because at least it would require a little more knowledge/software/criminal intend to break into my hotspot (my 3G volume is limited )

rambo | # 130 | 2010-01-31, 22:08 | Report

Nope, you need to add a check if the key is defined and add correct iwconfig command to start_configure_wlan() in hotspot_backend.py

It's 4 lines of code but I never got around adding them (and since I need to go to sleep now I won't be doing it now, don't want to blindly commit untested code)