xfs
[Top] [All Lists]

Re: [RFC, PATCH 0/2] fiemap: filesystem free space mapping

To: Chris Mason <chris.mason@xxxxxxxxxxxx>, Dave Chinner <david@xxxxxxxxxxxxx>, Christoph Hellwig <hch@xxxxxxxxxxxxx>, "linux-fsdevel@xxxxxxxxxxxxxxx" <linux-fsdevel@xxxxxxxxxxxxxxx>, "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>
Subject: Re: [RFC, PATCH 0/2] fiemap: filesystem free space mapping
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Wed, 24 Oct 2012 11:09:51 -0400
In-reply-to: <20121024114711.GB11262@shiny>
References: <1350537079-16246-1-git-send-email-david@xxxxxxxxxxxxx> <20121023123044.GG7341@xxxxxxxxxxxxx> <20121023215313.GQ4291@dastard> <20121024114711.GB11262@shiny>
User-agent: Mutt/1.5.21 (2010-09-15)
On Wed, Oct 24, 2012 at 07:47:17AM -0400, Chris Mason wrote:
> I'm all for it in the main fiemap call, it makes much more sense for the
> users I think.

How so?  Current fiemap is a per-inode information, Daves new call is
per-fs.  Making one a flag of another is a gross user interface.  In
addition we're bound to get issue where filesystems fail to wire up
fiemap to the tons of different iops just for this operation, or
accidentally wire up "real" fiemap to things like special files or
pipes.

Btw, I'd like t orestate that I really love to see this functionality in
the VFS, just not multiplexed over FIEMAP.

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