[pkg-fso-maint] Kernel and wifi

[Michael 'Mickey' Lauer]

NOTE: Adding Werner in the loop (Werner, here's some more lists for you
to subscribe: smartphones-{kernel|userland}, and if you want,
smartphones-standards at linuxtogo.org :)

Am Dienstag, den 03.03.2009, 09:46 +0100 schrieb Christian Adams:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> moinmoin,
> Am 02.03.2009 um 20:32 schrieb gregor herrmann:
> 
> [..]
> 
> > 2) boot with wicd disabled in runlevel 2
> > As we already know :) there's no oops in this case. Status after
> > everything has come up and before I did anything (except entering the
> > PIN):
> > * ifconfig and iwconfig don't know about eth0, o-p-p has the "wifi
> >   disabled" icon.
> > * No eth0 in /proc/net/dev.
> > * ar6000 loaded.
> >
> > Now let's try to produce a kernel oops:
> > * Turn on wifi (o-p-p): eth0 shows up in /proc/net/dev and in
> >   i(w|f)config output. No oops.
> > * /etc/init.d/wicd start: no changes, no oops.
> > * Start wicd-client: no changes, no oops.
> > * Open the wicd-client GUI; it scans for wireless networks; still no
> >   changes/problems.
> > * Connect to a (WPA encrypted) network via the GUI: wpa_supplicant
> >   and dhclient get called, both succeed, no problems. dmesg output:
> >
> > [  584.355000] ar6000_ioctl_giwscan(): data length 0
> > [  587.245000] AR6000 disconnected
> > [  587.630000] channel hint set to 2462
> > [  590.105000] eth0: no IPv6 routers present
> > [  592.135000] channel hint set to 2462
> > [  592.150000] AR6000 disconnected
> > [  592.225000] AR6000 connected event on freq 2462 with bssid  
> > 00:18:84:29:9a:57  listenInterval=100, beaconInterval = 100,  
> > beaconIeLen = 26 assocReqLen=60 assocRespLen =59
> > [  592.230000] Network: Infrastructure
> >
> > * Turn off wifi (o-p-p): Here we are: the kernel oopses; last lines
> >   of dmesg output attached.
> 
> i think i know now why i don't get the oops here:
> in o-p-p you have options for pre/post enable/disable - and for wifi
> i use the following:
> * post-enable: sleep 2 && iwlist scan && ifup eth0
> * pre-disable: ifdown eth0
> 
> @mickey:
> whouldn't it be an idea that frameworkd ensures the interface is down
> (via 'ifdown eth0') before 'disabling' it?

Sure we can; I did not do this yet since I had the assumption that
unbinding is doing that already in a proper way. Werner?

-- 
:M: