[Top] [All Lists]

Re: [REVIEW] cleanup - remove bhv_vname_t

To: "Christoph Hellwig" <hch@xxxxxxxxxxxxx>
Subject: Re: [REVIEW] cleanup - remove bhv_vname_t
From: "Barry Naujok" <bnaujok@xxxxxxx>
Date: Tue, 08 Apr 2008 17:48:58 +1000
Cc: "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>, xfs-dev <xfs-dev@xxxxxxx>
In-reply-to: <20080408074106.GA3207@xxxxxxxxxxxxx>
Organization: SGI
References: <op.t89zp3j63jf8g2@xxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20080408063822.GA21876@xxxxxxxxxxxxx> <op.t895ac193jf8g2@xxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20080408074106.GA3207@xxxxxxxxxxxxx>
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Opera Mail/9.24 (Win32)
On Tue, 08 Apr 2008 17:41:06 +1000, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:

On Tue, Apr 08, 2008 at 04:50:26PM +1000, Barry Naujok wrote:
+static inline struct xfs_name *
+       struct xfs_name *namep,
+       struct dentry   *dentry)
+       namep->name = dentry->d_name.name;
+       namep->len = dentry->d_name.len;
+       return namep;

As mentioned in my comment to the CI series:  shouldn't you just use
a struct qstr instead of adding a new struct xfs_name?

Keeping a clean line between fs/xfs and fs/xfs/linux-2.6.

But this means more stack useage and more copies in every namespace
related operation.  I don'y yhink it's a good tradeoff.

This will happen anyway for Unicode NLS support. Also, this patch
doesn't increase stack usage anywhere and some have reduced stack usage.

If you really care about a clear separatation add a #define or typedef
for xfs_name to struct qstr.

and qstr is platform dependant (yes, I know you said it's premature!)


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