someone meantioned here rather recently that the audio interrupts are
serviced at a higher priority than serial interrupts.  i just noticed
that in the oct 2 -> oct 19 patch i made last year (the first time
that *i* noticed extra lossage), i see that the sparc audio driver was
changed...

other changes of note:
	- bitmask_sprintf() went in.
	- ntp_xxx() system calls went in.
	- sparc went to new kernel crash dump format.
	- (sparc) MAXTSIZ, DFLDSIZ and MAXDSIZ were increased.
	- midway (atm) driver went in.
	- sockets are now kept on a doubly linked list (prior was single)

thats about it...


note that the audio changes are the port to the sun4m.  from what i can
see in my diff, with only SUN4C defined, the code should come out
*exactly* the same...so this may not be the problem.  sigh.


.mrg.