On Mon, 2007-11-26 at 17:30 +1100, Jeremy Visser wrote:
> On Sun, 2007-11-25 at 23:18 -0500, Pavel Roskin wrote:
> > Thanks! It looks good. I'm planning to collect your patches, test them
> > and send them for inclusion into the kernel.
>
> Is orinoco_usb anywhere near being stable enough for kernel inclusion? I
> mean, obviously, it's not that stable now, but if we organised a weekend
> where we got together on IRC, would we be able to improve the state of
> things?
One problem is that the driver is tied to the common code (orinoco.c),
which enables code sharing, but requires some ugly hacks that make it
impossible to be included into the kernel.
The only way to get rid of those hacks while keeping code sharing is to
change the way the non-USB driver operates. More stuff should be done
in process context, as opposed to irq and softirq context.
Another approach would be to abandon code sharing and share only some
headers. That means another driver to maintain and more work
synchronizing changes. Although it would allow a much more aggressive
cleanup.
An additional problem is that only mac80211 based drivers are accepted
to the kernel. This means that the driver should be rewritten to use
mac80211 layer instead of the netdev layer. That's a lot of work.
Besides, the card would have to be dumbed down, automatic scanning and
joining will need to be disabled.
> Well, that is, if there is still interest in the driver. No new devices
> use orinoco_usb, do they? How many people still use this driver? Is it
> worth maintaining anymore, or should we put it down now so as to not let
> it die a slow and painful death?
As I said, a decision about code sharing will need to be made. If there
is somebody willing to maintain Orinoco USB as a separate driver, I can
offer help with initial cleanup and testing.
If somebody wants to help with the conversion of the whole Orinoco suite
to the process context, that would be welcome too.
--
Regards,
Pavel Roskin
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
|
