>In article <v03130300b8e67e755432@[209.46.8.69]> you write:
>>>Apr 19 06:15:39 mercy syslogd: restart
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xb880
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xb8c0
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xb900
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xb940
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xb980
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xb9c0
>>>Apr 19 06:15:40 mercy /netbsd: cy0: Do manual int svc! icount 0xba00
>>
>>Besides the jump in times from the 17th to the 19th, notice the messages
>>about "Do manual int svc". These are kernel messages that are generated by
>>the Cyclades driver that I modified.  The number is a counter of "missed"
>>interrupts since boot.  I normally get several of these a day.  Note that
>>this counter is a compiled *static*, which gets re-initialized to zero on
>>boot.  Note that on reboot these appeared between the restart and the
>>kernel banner.
>>
>>I did not restart this machine gently.  I hit the HW reset button.  How did
>>these values survive to come out after the reboot???
>
>I'm not sure about macppc, but on a lot of architectures, the kernel message
>buffer is always written to the same physical address in memory, and the
>kernel startup code tries to preserve the message buffer if it finds a valid
>one when it starts.  This means that messages written by the kernel but not
>collected by syslogd will get preserved over a reboot and logged by syslogd
>when it restarts.
>
>In other words "It's a feature".

... and a pretty darn nice one, if you ask me.

This means that even if you hit the "red button", you can still see the last
few bits of info from the kernel, as long as you don't power down.
This is pretty useful for situations like mine, or with hard hangs,
where debugging and/or HW error messages might be lying in wait, but
you can't get to them.

Thanks,

-dgl-