NFS write from emacs (on an SGI) to lispm hangs-up forever

[rwk@Crl.dec.com (Robert W. Kerns)]

    Date: Thu, 19 Sep 91 08:05 EDT
    From: rwk (Robert W. Kerns)

	Date: Wed, 18 Sep 91 14:57:51 EDT
	From: aboulang@bbn.com
	I did this in my original bug report which I failed to keep around. It
	(the lisp side) is hanging up in the state "Truncate File" as viewed
	from the peek display. One can reset the process (lisp side) and
	things finish ok. Note that normal file operations work ok (they seem
	to be slower than from a Sun box however) -- it is only writing from
	EMACS that hangs up.  I saw a message describing a problem like this
	come over SLUG a couple of months ago. (Where are the archives kept by
	the way?)

    We also see this here on an annoyingly frequent basis.
    The NFS server process is *LOOPING* truncating the file.
    It's not just waiting for something, and it frequently
    spends time in Disk Wait as well.

    I think I might have seen it a time or two with Lucid, but
    I'm not sure.  Certainly it has a less than .005 probability
    with a write from Lucid.  I'd estimate around a .3 or .4
    probability from a Decstation 5000 to a 3600.
That last was supposed to be ".3 or .4 probability from from emacs.
All this on a Decstation 5000 to a 3600."

    Sometimes killing the process leaves things happy on both
    ends, and sometimes it leaves the file, well, truncated to 0.
    Lately, it seems it's always left the file truncated.  This
    is Not OK.

    Needless to say, this is *extremely* irritating with autosave
    turned on!

    I've been holding off on sending a bug report until we get 8.1
    up here, but I can confirm it happens, and not just on Suns.