On Mon, 2004-07-26 at 20:07 +0200, Hans de Goede wrote:
Because of these dependencies parralellisation really isn't any help. Getting a signal that the kernel has completed its initial scan of all the usb busses however would be a bug help, because I believe the current sleep 5 is much longer then is needed on most systems.
Realistically, the sleep 5 isn't sufficient in other cases. We should probably instead do something and verify that /proc/bus/usb/devices is stable (ie, that two checks one second apart haven't changed) similar to what anaconda does in anaconda/loader2/usb.c:sleepUntilUsbIsStable()
This will lead to a shorter sleep on "normal" systems and more correct behavior on systems with a longer USB device tree to enumerate.
Jeremy
Since anaconda appearantly hits the smae problem, why not fix tbis where it should be fixed, in the kernel. The kernel is the only "process" which knows when the initial enumeration is done!
Regards,
Hans
-- EuropeSwPatentFree http://EuropeSwPatentFree.hispalinux.es
