Did you get `NBUF == BUFPAGES'. I remember a similar panic on the sparc port
some months ago. I don't think the buffer cache is already fixed for the
`NBUF != BUFPAGES' case.

> gunnar@bitcon.no (Gunnar Helliesen) writes:
> >Anyone know what causes this:
> 
> >panic: cluster_rbuild: too much memory
> >Stopped at     _Debugger+0x4:  leave
> >db> 
> 
> After looking at the code a bit I'd say this is a bug in the vfs_cluster
> code.  At least the condition for the panic doesn't look right.
> 
> >System info:
> 
> >NetBSD 1.2E (ATLAS) #0: Sun Jun 15 11:02:36 MET DST 1997
> >    root@atlas.bitcon.no:/usr/src/sys/arch/i386/compile/ATLAS
> >cpu0: family 6 model 1 step 7
> >cpu0: Intel Pentium Pro (686-class)
> >real mem  = 133824512
> >avail mem = 95916032
> >using 1388 buffers containing 33980416 bytes of memory
> 
> That is more memory and less buffers than the kernel would have allocated
> by default.  Have you used options BUFPAGES and NBUF in the config file?
> If so, what are the values.
> 
> You should probably use at least 4200 buffers with that large a buffer
> cache.  Perhaps this helps.
> 
> It's probably worth submitting a PR.  Please include the values for
> NBUF and BUFPAGES, if you have used them in the kernel config file.
> A stack trace might be a good idea too.
> 
> -- 
> Christoph Badura
> 
> Now available in print: Lion's Commentary on UNIX 6th Edition, with Source Code
> 			http://www.peer-to-peer.com/
> 


-- 
Juergen Hannken-Illjes - hannken@eis.cs.tu-bs.de - TU Braunschweig (Germany)