[Top] [All Lists]

Re: xfs module insert fails with "xfs: Unknown symbol exportfs_decode_fh

To: Ben Myers <bpm@xxxxxxx>
Subject: Re: xfs module insert fails with "xfs: Unknown symbol exportfs_decode_fh (err 0)"
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Sat, 27 Apr 2013 07:46:26 +1000
Cc: Chandra Seetharaman <sekharan@xxxxxxxxxx>, XFS mailing list <xfs@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20130426181624.GG29359@xxxxxxx>
References: <1366997050.4098.14.camel@xxxxxxxxxxxxxxxxxx> <20130426181624.GG29359@xxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Fri, Apr 26, 2013 at 01:16:24PM -0500, Ben Myers wrote:
> Hey Chandra,
> On Fri, Apr 26, 2013 at 12:24:09PM -0500, Chandra Seetharaman wrote:
> > When I tried to insert xfs module in my test machine it failed with 
> > 
> >     xfs: Unknown symbol exportfs_decode_fh (err 0).
> > 
> > Found out that I did not have CONFIG_EXPORTFS enabled in my kernel
> > config. 
> Most people are running with NFS enabled in their kernel config, I think.
> > All .*by_handle code in XFS is open coded. Wondering if they should be
> > under #ifdef CONFIG_EXPORTFS ?
> >
> > I can create a patch, but wanted to find out if it is indeed needed :)
> foo_by_handle
>   xfs_handlereq_to_dentry
>     xfs_handle_to_dentry
>       exportfs_decode_fh
> xfsdump does open by handle, so making that conditional on CONFIG_EXPORTFS 
> will
> break things.  I think would be better to just make CONFIG_XFS depend upon
> CONFIG_EXPORTFS in fs/xfs/Kconfig.  Looks like we'll have the same problem 
> with
> libcrc32c.

AFAIK, "select" isthe right keyword for these dependencies. They are
like saying "depends on", but instead of having the XFS config options
only appear in the menus if CONFIG_EXPORTFS is selected, it will

So I don't think there are any problems here...


Dave Chinner

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