Libertas Driver Locking Kernel

[Dan Williams]

On Wed, 2008-05-21 at 15:58 -0500, Nick Moszer wrote:
> Dan Williams wrote:
> > On Wed, 2008-05-21 at 11:21 -0500, Nick Moszer wrote:
> >   
> >
> > Any chance you could try an updated driver?  What's in kernel git right
> > now in wireless-testing should be able to be used with 2.6.24 without
> > much trouble.  I noticed SendSinglePacket() in your logs, which isn't in
> > the upstream sources any more.
> >   
> Trying wireless-testing as well as the driver from 2.6.25-4 both 
> resulted in kernel panics from null pointers.
> They both popped up running iwconfig, one in lbs_scan_networks (2.6.25) 
> and the other in lbs_process_command_response(wireless-testing)
> I just copied the contents of drivers/net/wireless/libertas into the 
> 2.6.24 svn blackfin kernel,  recompiled and made a few minor fixes.
> It wasn't very obvious what was causing either. 

Any chance you could grab the panic trace, and then see if you can get a
line # out of the panic message as described here?

http://madwifi.org/wiki/DevDocs/KernelOops

(in the section 'gdb' near the bottom).  I'd love to figure out what the
source of the NULL is and then fix it.  Failing that, I'll try latest
wireless-testing with my SD8686 reference board and see what I come up
with.

Dan

> > Interesting bits:
> >
> > Jan  1 00:04:03 blackfin user.warn kernel:
> > Jan  1 00:04:10 blackfin user.info kernel: NETDEV WATCHDOG: eth1: 
> > transmit timed out
> > Jan  1 00:04:10 blackfin user.debug kernel: libertas enter (INT): 
> > libertas_tx_timeout():583
> > Jan  1 00:04:10 blackfin user.err kernel: libertas: tx watch dog timeout
> > Jan  1 00:04:10 blackfin user.debug kernel: libertas leave (INT): 
> > libertas_tx_timeout():604
> >
> > this could indicate that the card or SDHC crashed or isn't accepting any
> > further data or commands.  The libertas driver in the upstream kernel
> > tree has a reworking of the command processing and interrupt handling
> > that might help here.
> >
> > Dan
> >
> >   
> 
> I might just have to wait for these updates to work their way into the 
> blackfin sources, unless there are other ideas?
> 
> 
> 
>  
>