On 04/20/2010 02:28 AM, Gerd Hoffmann wrote:
Hi,
Not much traffic on this thread ;-)
Indeed ;)
I can see the usefulness of an init_late() to generalize post device
setup issues.
I assume then that you didn't have any other issues with my patch,
other than general code structure concerns?
Yes, that is the major one. I think it is much saner to just have a
init_late() and collect everything there instead of creating a new
hook each time you figure you need one.
I think this also allows to make the ide changes less intrusive as all
the cmos setup logic stays local to pc.c. pc.c can simply keep a
pointer to the DeviceState structs of the ide interface(s) created in
pc_init1(). pc_init_late() then can check which drives are plugged in
and update cmos accordingly.
I'd personally prefer to see some sort of registration mechanism.
Something notifier based.
Regards,
Anthony Liguori
cheers,
Gerd
