Acme Panic

Acme Panic

[dhog]

>have you tried ramfs?

No, because I don't want to use ramfs all the time, and the bug
doesn't appear to be reproducible.  It also occurs very infrequently.

It actually happened again tonight!  Forsyth pointed out to me that the
`phase error -- directory entry not allocated' means that acme tried to
do I/O on a file descriptor for which the file has been removed.  I checked,
and the temporary file had indeed been removed.  Now acme opens its
temp file with mode ORCLOSE, so the file will be removed if it is closed.
But then, if acme does close it (by accident) then it can't very well get that
error from still having it open.  So perhaps there is a bug in the handling
of ORCLOSE in both fileservers?  As an aside, I've noticed that old temp files
are hanging around, but I've always assumed that they were left overs from
file server panics/halts, or possibly severed connections (though I'd hope
in that case that the file server deallocates those fids and removes the
ORCLOSE ones, just as if they'd ben clunked).

Acme Panic

[Boyd]

have you tried ramfs?

Acme Panic

[dhog]

I have had the following error a couple of times with acme:

	acme: write error to temp file: phase error -- qid does not match
	acme 40: suicide: sys: trap: fault write addr=0x0 pc=0x28f08

This has happened both using the real file server, and kfs, as /tmp (and
everything else for that matter).  Has anyone else seen this?