Re: Problem with f_rndis.c

[David Brownell <david-b@pacbell.net>]

On Friday 03 April 2009, Eugene T. Bordenkircher wrote:
> The following is an artifact in f_rndis.c that is present because of the USB
> gadget refactoring
> 
> 679         rndis_set_param_medium(rndis->config, NDIS_MEDIUM_802_3, 0);
> 680         rndis_set_host_mac(rndis->config, rndis->ethaddr);
> 681
> 682 #if 0
> 683 // FIXME
> 684         if (rndis_set_param_vendor(rndis->config, vendorID,
> 685                                 manufacturer))
> 686                 goto fail0;
> 687 #endif
> 688
> 689         /* NOTE:  all that is done without knowing or caring about
> 690          * the network link ... which is unavailable to this code
> 691          * until we're activated via set_alt().
> 
> The bad news is that I'm hitting an oops because of it.
> 
> Newer versions of the windows rndis driver never seem to call
> OID_GEN_VENDOR_DESCRIPTION so this isn't a problem.  However, older versions of
> XP SP2 and earlier machines still call this, and since the vendor string is
> never set in rndis_per_dev_params(rndis.c) the kernel ends up SEGV'ing by
> accessing the null pointer for the vendor string.
> 
> I played around with this for about 3-4 hours yesterday trying to get the
> vendorID and manufacturer in some reasonable manner, but could not find a way
> to backtrack to them.
> 
> Any suggestions to help me fix it? or even better solutions? :) I have plenty
> of boxes that are oops'ing so I can test solutions as needed.

Maybe just add more params to rndis_bind_config() and
call that code from there.