xfs
[Top] [All Lists]

Re: some XFS questions

To: Kwon SoonSon <ksoonson@xxxxxxxxxxx>
Subject: Re: some XFS questions
From: Eric Sandeen <sandeen@xxxxxxx>
Date: 27 Jan 2003 16:22:07 -0600
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <20030127083818.84351.qmail@web7306.mail.yahoo.co.kr>
References: <20030127083818.84351.qmail@web7306.mail.yahoo.co.kr>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Mon, 2003-01-27 at 02:38, Kwon SoonSon wrote:

> 2. in the superblock(xfs_sb), I see "realtime block"
> and
> "realtime extent". Can anyone please let me know
> what the realtime block/extent mean?

xfs has a "realtime" feature (somewhat misnamed) where data goes to a
different partition, and uses a more deterministic allocator - generally
allocates in larger chunks.  you do an fcntl on the file to designate it
as realtime, then do direct I/O to it.  This is still buggy on Linux,
and is not supported.  See the mkfs.xfs man page for how to make an fs
w/ a realtime subvolume.

 
> 3. in xfs_dinode_core, there is di_projid.
> What is this for?

Irix has the concepts of user, group, and project, I believe.  Probably
not used on Linux, but retained for on-disk compat.

> 4. according to the 1993 design document, the inode
> number of xfs_agi used to be devided into 3 parts
> but now it is 2 parts. I think this should have some
> reasoning but don't know the details. Why?

That one is probably lost in the mists of time.  Perhaps Steve knows.
Actually, I thought that the inode number -was- still 3 parts, for ag #,
ag block #, and inode # in that block.


-Eric

-- 
Eric Sandeen      XFS for Linux     http://oss.sgi.com/projects/xfs
sandeen@xxxxxxx   SGI, Inc.         651-683-3102


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