A quick report of the success of the latest kernel source changes
with regard to my machine locking up every 4-6 hours. It has been
up for 24hours with no problem.  The only nit that still seems to
still be happening are the silo overflows.  I have 18 log entries
in the past 24 hours. What seems to have helped that some is the
addition of the second line in this snipit from isa_machdep.c.

    /*
     * There are network and disk drivers that use free() at interrupt
     * time, so imp > (net | bio).
     */
    imask[IPL_IMP] |= imask[IPL_NET] | imask[IPL_BIO];
    imask[IPL_TTY] |= imask[IPL_NET];


The following is a snipit from 'iostat' on a 2 second interval while
doing an NFS read over the 28.8 connection. Baudrate is locked at
115200.

      tty
 tin tout 

10230   41
10971   41
10790   41
11231   41
10703   41
10876   41
11342  120
11188   41


BTW - Didn't someone post some patches awhile back to make 'iostat' 
work for SCSI disks?