[Top] [All Lists]

Re: [PATCH 03/14] repair: kill B_IS_META flag

To: Eric Sandeen <sandeen@xxxxxxxxxxx>
Subject: Re: [PATCH 03/14] repair: kill B_IS_META flag
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Tue, 13 Oct 2009 18:16:46 -0400
Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <4AD38744.3010702@xxxxxxxxxxx>
References: <20090902175531.469184575@xxxxxxxxxxxxxxxxxxxxxx> <20090902175840.224768080@xxxxxxxxxxxxxxxxxxxxxx> <4AD38744.3010702@xxxxxxxxxxx>
User-agent: Mutt/1.5.19 (2009-01-05)
On Mon, Oct 12, 2009 at 02:45:08PM -0500, Eric Sandeen wrote:
> Christoph Hellwig wrote:
>> B_IS_META is the inverse flag of B_IS_INODE which is not really obvious
>> from it's use.  So just use !B_IS_INODE to make it more clear.
> Logic-wise it's fine, but is this change really helpful?   The comment says:
> /*
> * Test if bit 0 or 2 is set in the "priority tag" of the buffer to see if
> * the buffer is for an inode or other metadata.
> */
> so basically it distinguishes inodes from other metadata right.

Yes, with the key on other.  In my books inodes are meta-data.

> B_IS_INODE is clear; B_IS_META is pretty clear, "!B_IS_INODE" seems muddy; so
> very many things are "not inodes" :)

In a buffercache (and in fact a whole application) that only deals with
metadata at all !B_IS_INODE meaning other metadata seems a lot more
clear to me than B_IS_META.

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