[Top] [All Lists]

RE: SGI Version 1 XFS on Linux

To: Ben Myers <bpm@xxxxxxx>
Subject: RE: SGI Version 1 XFS on Linux
From: "Scott, Edmund J @ SSG - Link" <edmund.scott@xxxxxxxxxx>
Date: Wed, 6 May 2015 19:47:51 +0000
Accept-language: en-US
Cc: "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20150506171222.GA22744@xxxxxxx>
References: <2994A55DE0FBBE41B8FE47F9885D74221BC0569C@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20150506171222.GA22744@xxxxxxx>
Thread-index: AdCIDxpplU8CdHWQQXuPofGKd7w1IgAOp8YAAAXQpjA=
Thread-topic: SGI Version 1 XFS on Linux

We are in a thunderstorm right now so we had to shut down, be a while until we 
get things back up. We're in Utah with a rolling t-storm coming through. 

We did find documentation that XFS version 1 was pulled out of the linux kernel 
around or after kernel 2.4. and the messages kind of support that. I'm looking 
into digging up an old Red Hat OS to see if it is old enough to have the 
version 1 XFS in it.

While it was up, we've done the blkid command and it recognizes at least 1 
partition on the SGI drive to be xfs, others were efs. We also used the 
hardware browser and got the same readings.

I did try using cpio on the one xfs partition it does recognize and got off 
some info, mainly only an indy patch directory and files. When I did a dd, cpio 
or a strings command and just let it stream from the partition, I could see a 
lot more information and files that I recognized as being the normal info that 
would be on that partition.

It's almost like I need to have some sort of combination of cpio, dd or strings 
in a script to strip the data off of the old partition without getting the 
filesystem info and save it into a desired directory.

We've also tried to make a disk image. If I try to mount the image on the linux 
computer, it will essentially give me the same error about version 1 xfs. Not 
that this was unexpected, it should error out just like the drive did. I also 
happen to have an O2 up and running, so I copied the image over to the O2 
expecting to try and mount the image so it would/should automatically handle 
the xfs issue, but it looks like the O2 mount command does not have the loop or 
knows how to handle mounting an image like the Linux does. Right now I cannot 
restore the image, because it would wipe out my one good O2 drive, at least 
until I get some clones made up. You also might be thinking to just mount the 
Onyx drive to the O2, but the O2 does not have the interface to handle a wide 
differential high voltage SCSI interface.

If it's not one thing it's something else. I guess it would be boring if not 
for things like this.

I'll get the info you requested later when we come back up.

Thanks for the help,

-----Original Message-----
From: Ben Myers [mailto:bpm@xxxxxxx] 
Sent: Wednesday, May 06, 2015 11:12 AM
To: Scott, Edmund J @ SSG - Link
Cc: xfs@xxxxxxxxxxx
Subject: Re: SGI Version 1 XFS on Linux

Hi Edmund,

On Wed, May 06, 2015 at 03:12:59PM +0000, Scott, Edmund J @ SSG - Link
> We have several older SGI hard drives we are trying to recover data 
> from. However, since these drives came from an SGI Onyx running IRIX 
> 6.2, it uses the older version 1 XFS file system. I can see the SCSI 
> drive on my Linux box (CentOS 5.11) and I know data is on the drive, I 
> can list out some of the info using dd, cpio or strings, but cannot 
> copy the data off of it. I have tried using xfs_copy to a version 2 
> XFS partition, but it just converts the target partition to version 1, 
> which I cannot mount. I have also looked at xfs_dump/restore, but 
> those require mounted filesystems.
> Is there any script or program or option I can use to copy the data 
> off of the un-mounted SGI XFS partition. I'm guessing some combination 
> of the cpio and dd, but I have not figured out the right combination 
> yet.

Apparently CentOS 5.11 should be a 2.6.18 kernel.  It looks like it should 
still have support for version 1, so I'm surprised you're having trouble.  
Could you provide a little more information?  

1) dmesg output when you try to mount the filesystem
2) xfs_info output, e.g.
        xfs_info /dev/sda5
3) hexdump of the superblock, e.g. 
        dd if=/dev/sda5 count=1 | hexdump -C


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