> Additionally, this patch also contains changes for prep and sparc64 to use MI
> mc146818(4) todclock driver.

 Could this change have caused an error like this?  This morning my Ultra10
gave a bogus NTP time warning shortly after I booted my new sparc64-current
kernel:

Nov  3 07:23:52 struck /netbsd: root file system type: ffs
Nov  3 07:23:52 struck savecore: no core dump
Nov  3 07:23:53 struck ntpd[402]: ntpd 4.0.99i Tue Oct 28 21:27:54 EST 2003 (1)
 [...]
Nov  3 07:27:24 struck ntpd[402]: time reset 1.405487 s
Nov  3 07:27:24 struck ntpd[402]: kernel pll status change 41
Nov  3 07:38:09 struck ntpd[402]: time error 2069188655 over 1000 seconds; set clock manually

 I saw the error on the console at 8:14 and a manual check revealed the time
was off by only a few seconds.  I restarted ntpd, without changing the time,
and it worked without error.

Nov  3 08:15:19 struck ntpd[9890]: ntpd 4.0.99i Tue Oct 28 21:27:54 EST 2003 (1)
Nov  3 08:15:19 struck ntpd[9890]: precision = 7 usec
Nov  3 08:15:19 struck ntpd[9890]: using kernel phase-lock loop 0001
Nov  3 08:15:19 struck ntpd[9890]: frequency initialized 59.282 from /var/db/ntp.drift
Nov  3 08:15:19 struck ntpd[9890]: using kernel phase-lock loop 0041