what does ``pme'' mean in 4/280 crashes?

[George Nassiopoulos]

For some months now, our SUN 4/280 has been crashing
with messages similar to this one:

    vmunix: nfsd: Data fault
    vmunix: kernel write fault at addr=0xffffeb7c, pme=0xe30015ba
    vmunix: Bus Error Reg 20<TIMEOUT>

Other examples of the second line of the above have included:

    kernel write fault at addr=0xffffeddc, pme=0xe3001688
    kernel read fault at addr=0xff121f38, pme=0xe30015d5
    kernel read fault at addr=0xff651e2c, pme=0xe30017f7
    kernel read fault at addr=0xff664d8a, pme=0xe300167e
    kernel write fault at addr=0xff0aa19a, pme=0xe3001788
    kernel read fault at addr=0xff0aa198, pme=0xe3001788
    kernel read fault at addr=0xffffe858, pme=0xe3001551

There have been more. The fact the the values for ``pme'' are so
similar is suspicious.  I've suspected one of our memory boards,
but can't translate the ``pme'' value into anything reasonable.

Can anyone tell me what ``addr'' and ``pme'' refer to?  Is this
a disk or memory error that is causing the crashes?  I will gladly
RTFM if someone can tell me which one, where.

Thanks in advance,

George Nassiopoulos
nassio at cfa.harvard.edu