>We could make one of the methods be "startup", of course, and move the
>actual device initialization there (and leave just the "find driver" in
>the initcall logic), but that would not get rid of the initcalls, it would
>just split them into two parts.
>
doing this has been mentioned independently a couple times, in fact...
And this may be a tangent, or maybe not: like I was trying to explain
to Gerard about regarding SCSI devices, it is often valuable to separate
the two steps. Gerard complained about the new PCI API not being able
to register devices (_register_, not probe) in the order he wished. And
my response was... sure you can. Just break it up into two steps, find
devices, and register devices. The ordering of the register calls in
the second step are entirely up the driver and/or subsystem. Similarly,
the IDE subsystem could handle the mapping of the BIOS-ordered
/dev/hd[a-d] completely independently of probing and registering hosts
and devices. [which, in turn, is useful in moving to a more dynamic
/dev among other things] Right now the IDE probe order is a bit
delicate, and full of hueristics that could be cleaned up with such a
separation.
Jeff, in a rambling mood
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/