Note that the processors on a JS20 are noticably slower to handshake
than those on a JS21. I had to rid of the hard-coded 1024 timebase
ticks and replace it with calculating five seconds from the timebase
frequency, as the timeout logic was firing on my JS20 blade.
What should we do about CPUs that successfully start but take too
long to ack, not sure where they'll end up.
This is prolly overkill but perhaps we shoud:
If it takes more than the proposed 5 secs, why bother -- your
system is dead anyway ;-)
Xen-ppc-devel mailing list