Randy> I think I have heard this problem mentioned before on this
Randy> list, but I don't remember there being a resolution. The
Randy> problem is that I have more entries in /proc/mounts than what
Randy> "mount" claims.
---8<---
From: Linus Torvalds <torvalds@transmeta.com>
Subject: Re: Bug report: Problems with NFS umounts on busy servers
To: Martin Kasper Petersen <mkp@socsci.auc.dk>
Date: Wed, 15 Oct 1997 14:42:20 -0700 (PDT)
X-From-Line: torvalds@transmeta.com Wed Oct 15 23:42:25 1997
On 15 Oct 1997, Martin Kasper Petersen wrote:
> I've been experiencing NR_SUPER exhaustion on my mail hub due to some
> NFS mounts not being correctly umounted. The mount entries disappear
> ok in /etc/mtab, but remain in /proc/mounts.
Right ho.
This should _hopefully_ be fixed in 2.1.57, which serializes mounts and
unmounts to make sure there are no race conditions. I don't know if it
works 100%, but I feel that it should. Have you tried 57 already and found
it to do the same thing?
Linus
---8<---
I went back to using amd (shudder!) on both my mailhub and my
WWW-server. The rest of my machines run fine with autofs (i.e. they
don't mount/umount a gazillion filesystems/second).
The tests I did back then seemed to indicate that 2.1.x fixed the
problem, though.
My test-machine suffers from severe glibc-NIS-brokenness at the
moment, so I haven't played around with autofs under 2.1 for a while.
-- Martin Kasper Petersen BOFH, IC1&2, Aalborg University, DK mailto:mkp@SunSITE.auc.dk http://www.socsci.auc.dk/~mkp/