Subject: Re: Is anyone seeing alignment faults on -current?
To: Frank van der Linden <email@example.com>
From: Mark H. Levine <firstname.lastname@example.org>
Date: 10/15/1997 00:22:39
There have been reports of my NFS changes causing alignment faults on
the Alpha and Sparc, probably because an off_t was fetched from
an address aligned to 4 bytes, not 8.
I checked in what I thought was a fix (making the changes I made in
dirent handling was off_t aligned), but I did get one report still
about alignment faults. However, the bug could not be reproduced
You might be interested in anecdotal evidence of an older 1.2G
Alpha system -- we noticed today that netboot failed with a
system whose name was "srdev3" but worked when we simply renamed
it "srd3". The error reported by mountd was "alignment error",
and the client machine simply reported NFS mount errors with
value -1 when using the longer name. I'd expect that the
alignment errors would be fixed up, and mountd kept running,
however we were amused that the NFS request was not being
fulfilled, and that of all things, the problem followed the
length of the hostname. Just in case you want to broaden your
search. I don't know that mountd has changed much of late.