[pnfs] [PATCH] pnfsd: Notify device ID changes
Marc Eshel
eshel at almaden.ibm.com
Wed May 21 19:19:48 EDT 2008
Benny Halevy <bhalevy.lists at gmail.com> wrote on 05/20/2008 02:09:56 AM:
> >
> > static struct svc_export *svc_export_update(struct svc_export *new,
> > @@ -434,6 +444,8 @@ static int check_export(struct inode *inode,
> int flags, unsigned char *uuid)
> > inode->i_sb->s_export_op->cb_change_state = cb_change_state;
> > if (!inode->i_sb->s_export_op->cb_layout_recall)
> > inode->i_sb->s_export_op->cb_layout_recall = cb_layout_recall;
> > + if (!inode->i_sb->s_export_op->cb_device_notify)
>
> Unrelated to this patch but why do we check this?
> Are there any other implementations?
>
> If not we should just EXPORT_SYMBOL the pnfsd callback entry points
> and forget about the function pointers in struct export_operations
> (they are pointing in the wrong direction anyway)
>
> Also, why the extra indirection?
> cb_device_notify just calls nfsd_device_notify_cb.
> The file system can^Wshould do that too.
>
The fs module can not depend on symbols that nfsd exports, other wise it
can not be loaded until the nfsd module is loaded.
Marc.
More information about the pNFS
mailing list