On Mon, 17 Dec 2001, Jan Harkes wrote:
> On Fri, Dec 14, 2001 at 01:51:36PM +0100, Trond Myklebust wrote:
> > Just one comment: Isn't it easier to do this in generic_file_llseek()
> > itself using inode->i_op->revalidate()? That would make it work for
> > coda and smbfs too...
>
> Actually, as far as Coda is concerned this only adds overhead. Coda uses
> AFS2 session semantics instead of UNIX semantics, so updates are only
> propagated when a file is closed.
>
> Adding this to the generic_file_llseek will force an useless but
> expensive upcall (and RPC call to the server) to every seek to check for
> an updated i_size while we already know that the i_size of the file
> won't have to change until it is closed and reopened.
>
> I guess we're just (mis-)using the revalidate call as a replacement of a
> missing call to i_ops->getattr from sys_stat. So perhaps adding the
> revalidate to the generic_llseek is fine, but I'll just have to get that
> missing getattr call into the tree.
As far as I'm concerned it's not fine.
->getattr() is needed and will be added (patch exists), but the thing
about ->revalidate()... It's a bloody mess that will need serious
cleanups. And I'd rather have fewer code paths involved into that
cleanup.
So please, do it in nfs_lseek() explicitly. If you want to use
generic_file_lseek() - wonderful, call it from nfs_lseek() after you've
done revalidation.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Dec 23 2001 - 21:00:13 EST