xfs
[Top] [All Lists]

Re: [PATCH 13/16] xfs: implement demand load of utf8norm.ko

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 13/16] xfs: implement demand load of utf8norm.ko
From: Ben Myers <bpm@xxxxxxx>
Date: Thu, 9 Oct 2014 10:19:56 -0500
Cc: linux-fsdevel@xxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx, olaf@xxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20141004071651.GB5168@xxxxxxxxxxxxx>
References: <20141003214758.GY1865@xxxxxxx> <20141003220357.GM1865@xxxxxxx> <20141004071651.GB5168@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.20 (2009-06-14)
On Sat, Oct 04, 2014 at 12:16:51AM -0700, Christoph Hellwig wrote:
> > +int
> > +xfs_init_utf8_module(struct xfs_mount      *mp)
> > +{
> > +   request_module("utf8norm");
> > +
> > +   spin_lock(&utf8norm_lock);
> > +   if (utf8norm_initialized) {
> > +           spin_unlock(&utf8norm_lock);
> > +           return 0;
> > +   }
> > +
> > +   utf8version_is_supported_func = symbol_get(utf8version_is_supported);
> > +   if (!utf8version_is_supported_func)
> > +           goto error;
> > +
> > +   utf8nfkdi_func = symbol_get(utf8nfkdi);
> > +   if (!utf8nfkdi_func)
> > +           goto error;
> 
> Please export a structure with a function pointes so that we just need
> a single symbol_get call.  I'd have to look up how symbol_get works,
> but unless there's something that speaks against this it might be
> simpler to than just do a symbol_get per mounst structure that uses
> utf8 and can point to that structure so that we don't have to add
> additional reference counting infrastructure around it.

Sure, sounds good.  I believe I've seen the approach you are suggesting
used elsewhere and it works fine.

-Ben

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