this has been reported by several people, but so far it's been
a very infrequent problem.  if you can reproduce it easily,
we should be able to figure it out without much fuss.

I don't think dumps to nfs swap are supported at all yet,
but we may not need a dump since we can use tcpdump to see
what's blocks in the file are being read, and that might be
enough to figure out what the problem is.

let me know if you can reproduce it and I'll work with you
to figure it out.

-Chuck


Miles Nordin writes:
> I got this running a 19990527 kernel with a 19990423 userland, during
> multiuser boot at the point where filesystems are NFS-mounted I assume.
> 
> panic: getblk: block size invariant failed
> Stopped in mount at _Debugger+0x6: unlk a6
> db> trace
> _Debugger(2004,0,0,e772a7c,f1d9d20) + 6
> _panic(e02fb1d,2000,0,f1b8500,0) + 54
> _getblk(f1ae924,0,2000,0,0) + c0
> _nfs_getcacheblk(f1ae924,0,2000,f1b8500) + 84
> _nfs_bioread(f1ae924,f1d9ef0,0,e16db80,0) + 62a
> _nfs_read(f1d9ea0) + 24
> _vn_read(f1d4058,f1d4070,f1d9ef0,e16db80,1) + ba
> _dofileread(f1b8500,3,f1d4058,3c000,200) + 8c
> _sys_read(f1b8500,f1d9f84,f1d9f7c) + 6e
> _syscall(3) + 114
> _trap0() + e
> db> 
> 
> The machine is NFS-rooted.  It runs well with a 19990423 kernel.
> Debugging suggestions are welcome--i'm building a kernel with -g symbols
> now.  Anyone know how to get crash dumps on an NFS-swaped machine?
> 
> -- 
> Miles Nordin / 1-888-857-2723
> 555 Bryant Street PMB 182 / Palo Alto, CA 94301-1700