* Wait some time to allow driver to initialize before
* starting configuring the driver. This seems to be
* needed at least some drivers that load firmware etc.
* when the interface is set up.
which drivers are those, and why can't we fix the drivers instead? This
seems like a bug in the drivers that they return from their open
callback prematurely.
johannes
On Tue, 2008-10-07 at 20:16 +0200, Johannes Berg wrote:
> * Wait some time to allow driver to initialize before
> * starting configuring the driver. This seems to be
> * needed at least some drivers that load firmware etc.
> * when the interface is set up.
>
> which drivers are those, and why can't we fix the drivers instead? This
> seems like a bug in the drivers that they return from their open
> callback prematurely.
In any case, this delay should certainly be removed from the nl80211
driver; when porting drivers to the new APIs we can make sure this is
done correctly.
johannes
On Tue, Oct 07, 2008 at 08:16:01PM +0200, Johannes Berg wrote:
> * Wait some time to allow driver to initialize before
> * starting configuring the driver. This seems to be
> * needed at least some drivers that load firmware etc.
> * when the interface is set up.
>
> which drivers are those, and why can't we fix the drivers instead? This
> seems like a bug in the drivers that they return from their open
> callback prematurely.
I've observed issues with a driver that (at least at the time) was not
fully open source. I didn't feel like fixing the driver at the time.
Anyway, I'm perfectly fine with removing this from nl80211. As far as
driver_wext.c is concerned, I would like to first verify what the
original issue is and whether the driver could be fixed easily.
--
Jouni Malinen PGP id EFC895FA
On Tue, Oct 07, 2008 at 08:18:18PM +0200, Johannes Berg wrote:
> On Tue, 2008-10-07 at 20:16 +0200, Johannes Berg wrote:
> > * Wait some time to allow driver to initialize before
> > * starting configuring the driver. This seems to be
> > * needed at least some drivers that load firmware etc.
> > * when the interface is set up.
> >
> > which drivers are those, and why can't we fix the drivers instead? This
> > seems like a bug in the drivers that they return from their open
> > callback prematurely.
>
> In any case, this delay should certainly be removed from the nl80211
> driver; when porting drivers to the new APIs we can make sure this is
> done correctly.
ACK
--
John W. Linville Linux should be at the core
[email protected] of your literate lifestyle.