[880] in Release_7.7_team
More on the KNFS problem
daemon@ATHENA.MIT.EDU (Greg Hudson)
Tue Mar 4 02:57:20 1997
Date: Tue, 4 Mar 1997 02:57:17 -0500
From: Greg Hudson <ghudson@MIT.EDU>
To: release-team@MIT.EDU
It turns out that Solaris KNFS servers do not behave substantially
differently from Decstation KNFS servers, I think, and I don't think
we need to put out the attach modification that I just submitted to
source-reviewers.
There is a bug in the current attach code, but it only affects
explicit mounts. So you can reproduce the bug by doing "attach -e
jason:/u1/bitbucket" just as well as by doing "attach -e
gypsy:/forfun". But if you do "attach bitbucket", a Hesiod lookup
gets done and, because there happens to be a socket with fd 7 after
the Hesiod lookup, the bug isn't triggered. (It seems that what
clntudp_create() passes back is not actually a socket, but a TLI
frob.)
So Decstation KNFS servers can be replaced with Solaris ones, and
nothing should go wrong that isn't already going wrong.
As for the correctness of my source change, I need to look at the
source for clntudp_create() on various operating systems and make sure
that it always returns a bound socket, and I need to test it under
Irix. But this is no longer an issue relevant to the release team, I
think.