[1066] in SIPB_Linux_Development
Re: Hand-modifications to make Linux-Athena work?
daemon@ATHENA.MIT.EDU (Derek Atkins)
Mon Sep 18 22:10:07 1995
To: jered@MIT.EDU
Cc: "Theodore Ts'o" <tytso@MIT.EDU>, ddaniel@vigor.JPL.NASA.GOV,
Derek Atkins <warlord@MIT.EDU>, linux-dev@MIT.EDU, linux-help@MIT.EDU
In-Reply-To: Your message of "Mon, 18 Sep 1995 20:36:37 EDT."
<9509190036.AA21107@chicken-soup.MIT.EDU>
Date: Mon, 18 Sep 1995 22:09:35 EDT
From: Derek Atkins <warlord@MIT.EDU>
It is not the location of the AFS cache that matters. Rather, it is
the location of the afsd client program.
What is going on is that the afsd doesn't die properly, so the
partition still has an open file and therefore cannot unmount
properly. Because it cannot unmount, it tries to fsck.
I only say root is special because it doesn't require all the programs
to stop for it to unmount properly, whereas other partitions do. So
if afsd is on the root parition, it can still unmount properly. I
think the misnomer about putting the cache on the root is that usually
both afsd and the cache are in the same partition.
-derek
PS: I think another problem is that AFS may keep some of the
cachefiles open, in which case the cache partition wouldn't be
able to be unmounted, either. I'm working on getting afsd to
die properly -- I'm mostly there.
