[Top] [All Lists]

Re: [PATCH] xfstests: test for file clone functionality of btrfs ("refli

To: Koen De Wit <koen.de.wit@xxxxxxxxxx>
Subject: Re: [PATCH] xfstests: test for file clone functionality of btrfs ("reflinks")
From: David Sterba <dsterba@xxxxxxx>
Date: Thu, 10 Jan 2013 14:15:25 +0100
Cc: xfs@xxxxxxxxxxx
In-reply-to: <50C762DA.2090907@xxxxxxxxxx>
References: <50C762DA.2090907@xxxxxxxxxx>
Reply-to: dsterba@xxxxxxx
User-agent: Mutt/1.5.21 (2012-12-30)

On Tue, Dec 11, 2012 at 05:44:10PM +0100, Koen De Wit wrote:
> Tests included:
>   - Creating a reflink and overwriting the original contents
>   - Reflinking a directory tree
>   - Moving/deleting reflinks
>   - Diskspace consumption checks
>   - Cross-filesystem copies (should fail)
>   - Cross-subvolume copies

David did a review and I agree with all of his comments. I have a few
suggestions to extend test scope:

* check if reflink expectedly fails when it's done on the same
  filesystem but mounted to 2 different directories (for example bind
  mount the TEST partition to SCRATCH dir and do the reflinks)

* most of the files are very small and will get inlined, this
  covers only a small range of problems that could arise from reflinks.
  I'd like to see this included in the initial test version so the test
  covers majority of the cases and we can add specific corner cases

* larger files allow to test reflinks on files that comprise of
  interesting mix of holes and data (eg. what happens if I reflink half
  of hole and data extent at once?), reflink beginning or end of a file
  to another file


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