xfs
[Top] [All Lists]

Re: Question about xfs

To: Sean Dougherty <sean@xxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Question about xfs
From: Steve Lord <lord@xxxxxxx>
Date: Mon, 12 Feb 2001 10:52:04 -0600
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: Message from Sean Dougherty <sean@cortex.ama.ttuhsc.edu> of "Mon, 12 Feb 2001 10:48:00 CST." <Pine.SUN.3.91.1010212103648.1634A-100000@cortex>
Sender: owner-linux-xfs@xxxxxxxxxxx
> 
> As I have been testing, I have crashed a few times :-).
> 
> When I do an xfs_repair of the crashed drive, that seem to run very 
> well.  Then I remount the drive and then delete everything on that drive 
> (getting set up for the next test) I have noticed that when I do a "df" 
> the drive shows oh between 30-40K still used.  Whereas when I created the 
> file system (mkfs -t xfs....) a df shows about 1200 bytes used.  

The quick answer is that XFS allocates inodes dynamically, so as you create
files, blocks of disk space are taken out of free space and made into
inode blocks. When you delete files this space does not go back to being
free again, but remains as free inode blocks. The free space calculation
is based on the amount of space which is not used for anything, so free
inode blocks do not count towards it. Hence the difference in the df
report. If you were using something like ext2 then inodes are allocated
at mkfs time and the space is gone from 'free space' immediately.

Steve

> 
> I am sure this has to do with the journaling and want to learn more, can 
> someone point me to reference where I can read up on this.


Apart from the documents hanging of the oss web site there is nothing
beyond the code and what is in peoples heads.

http://oss.sgi.com/projects/xfs/design_docs/

Steve

> 
> Thanks.
> 
> Sean Dougherty
> Manager of way too much
> TTUHSC At Amarillo
> 



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