[Top] [All Lists]

Re: Errr... what's all this about then?

To: linux-xfs@xxxxxxxxxxx
Subject: Re: Errr... what's all this about then?
From: Bruce Tenison <btenison@xxxxxxxxxxxxx>
Date: Sun, 27 May 2001 21:14:05 -0500
References: <LGECJLHBGAPMGAJFMAFIEEFJCAAA.juha@saarinen.org>
Sender: owner-linux-xfs@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux 2.4.4-xfs i686; en-US; rv:0.9) Gecko/20010507
I've done a bit more work on this today, and now have a drive that doesn't seem to generate the errors I was having.
The solution for me seemed to be enabling devfs support. I already had it compiled into the kernel, but not mounted.
To begin with I noticed that once I used the drive, and rebooted to a non-xfs partition (ext2) as root (with an xfs enabled
kernel) and issued an xfs_check on the xfs partition (that was previously root) it returned all kinds of errors.

I then, instinctively (I guess), issued an xfs_repair on the drive, and wound up with some entries in lost+found, mostly from
either log files or pid files in /var/run (and sometimes the directory itself)...

Well, I noticed that a lot of times I was getting an xfs recovery after just issuing a reboot command. So I started looking at
/etc/rc.d/init.d/halt to see how the filesystems were being umounted. It looks at /proc/mounts, so I looked there, and what did
I see?
/dev/ide/host0/bus0/target1/lun1/part3 / xfs rw 0 0

Hmmm. Since I have devfs NOT mounted, just compiled in.... Well I rebooted, to the non-xfs, mounted the xfs partition, let it
recover, umounted it, then ran xfs_check. No errors!!!! Hmm, maybe I was overzealous on the xfs_repair earlier. Shouldn't
xfs_repair check for this case?

Rebooted to the xfs partition with devfs=mount as an option. Everything seems to work after that....

Also, I noticed that, near the bottom of /etc/rc.d/init.d/halt there are the following lines (note the ext2):
# Remount read only anything that's left mounted.
#echo $"Remounting remaining filesystems (if any) readonly"
mount | awk '/ext2/ { print $3 }' | while read line; do
mount -n -o ro,remount $line

So I guess this is looking for an ext2 partition as the root and maybe the only partition left to umount. Since it's root, I guess
we can only remount it ro. So I changed it to xfs, and placed a couple of mount commands to show me whether or not it worked.
It doesn't seem to remount the root xfs partition read-only (at least mount reports (rw) both before and after the remount).

But, at least I'm not seeing the errors anymore (I hope)

Hope this helps.


Juha Saarinen wrote:

:: I've also got a Promise PDC 20262 w/ an IBM Deskstar running in UDMA
:: mode 4 w/ no problems whatsoever. Additionally a Maxtor Diamond Max
:: connected to the stock BX MB w/ no problems. Early on (5-6 months ago) I

Yep... you're using a BX440 motherboard. Bruce and I have VIA chip set ones.
I really, really, really regret buying that Tyan motherboard. :-(

-- Juha

<Prev in Thread] Current Thread [Next in Thread>