On 17 Jul 2007, at 21:47, Tyler Trafford wrote: > On Tue, Jul 17, 2007 at 04:25:13PM -0400, Sander Sweers wrote: >> On 7/17/07, Nicholas Fearnley <[EMAIL PROTECTED]> wrote: >>> On 17 Jul 2007, at 18:11, Tyler Trafford wrote: >>>> http://ivtvdriver.org/pipermail/ivtv-devel/2007-February/ >>>> 004501.html >>>> >>>> I think it might be a good idea for drivers that require >>>> firmware to >>>> delay initialization until triggered by the user. >>> >>> I agree, the budget dvb cards do just that on firing up the driver. >>> It's disappointing that something in the kernel *has* to be compiled >>> as a module, I totally understand the technical constraint in this >>> case though. >> >> And this is why uevents/udev have been created and /sbin/hotplug has >> been deprecated. >> >> Device discovered by kernel -> uevent -> userspace (udev) listens for >> these on netlink -> module loaded (if not build-in) -> device created >> -> firmware loaded. >> >> AFAIK udev does not care whether the device is loaded as a module or >> build in it just handles the event and loads the needed firmware. >> >> So get rid of /sbin/hotplug ;-) > > The problem though, is when ivtv requests the firmware before udev (or > the filesystem) are there. > > I think this problem exists with the ipw* wireless drivers, although I > haven't looked at them in a long time. > -- > Tyler Trafford >
True, that is the problem. I acknowledge for a desktop system (and some more dynamic servers) that udev is cooler than hotplug. But my system is very static. It never changes. It just wants firmware loaded on boot and then left alone, for ever, there's really no benefit in swapping for udev I think. I'm going to try and modify ivtv to load up the driver on request, so I can build into kernel. If I get it working I'll feedback. Need to try Hans hotplug script to at least get the module loaded up before I proceed. _______________________________________________ ivtv-devel mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-devel
