Subject: Re: kernel crash dumps
To: Jonathan Stone <jonathan@DSG.STANFORD.EDU>
From: Chris G Demetriou <Chris_G_Demetriou@BALVENIE.PDL.CS.CMU.EDU>
List: port-pmax
Date: 07/12/1995 04:25:14
> Is it reasonable to assume that blocks in aa swap partition can always
> be read (or written) with a 6-byte SCSI command, instead of a 10-byte
> group1 extended command?   I hope so...

uh, i doubt it.  doesn't that limit the blocks which can be written
to?

there's no limit on placement of a 'b' partition, and, unless i'm
mistaken, other partitions can be used for swap, as well.

i think that the machine-independent SCSI code _always_ uses
long commands, but am not 100% sure...

> Also, what's a good way to force a crashdump?   Is adding a -d option
> (meaning do RB_DUMP) to /sbin/reboot  a reasonable option?

i don't really think that's necessary or appropriate.  forcing a crash
dump is something that one would want to do very infrequently, and
(usually) only for debugging.  It's also not clear whether it's a good
idea to easily expose the ability to create crash dumps, though it
might be.

I think it's reasonable to require a person who wants to do it to
write a tiny little program that calls reboot() with the appoprate
flags...


chris