I think that the namei changes that handle the emulated root could
be at fault here, or NFS. Have you tried using a loopback mount?
No, but I will try that now then.
Unfortunately, we did try to go for loopback mounts under netbsd-4,
(worked well under netbsd-3-1) but it proved to have some sort of
small memory leak, so the system crashed after many days of uptime
and
huge numbers of Gigbytes of data reads for the backups.
he@(netbsd.org) did tracing on that problem, but I don't know if he
managed to gather enough data for pinpointing the actual code that
caused these leaks.
Yes, I did some debugging and tracing in an attempt to narrow down
what was causing the kernel memory leakage. Unfortunately, I never
did manage to narrow down the root cause, and we moved to doing NFS
mounts to work around the problem, so the problem which is triggered
by use of null mounts and our particular useage pattern is most
probably still with us to this day. :(
I now see that the system in question became unable to fulfill it's
duties during the night due to
WARNING: mclpool limit reached; increase NMBCLUSTERS
and that someone else rebooted it before I got a chance to have a
closer look. :( BTW, as far as I can see, it runs with the default
setting for NMBCLUSTERS, which is most probably too small. As a
matter of fact, for reasonably-modern i386 systems I consider the
default NMBCLUSTERS setting to be too low in general...