Re: [2.6.26 PATCH, RESEND]: fs_stack/eCryptfs: fsstack_copy_* updates

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 30 Apr 2008 17:09:15 -0400
Erez Zadok <ezk@cs.sunysb.edu> wrote:

> In message <20080430101704.9cbd6384.akpm@linux-foundation.org>, Andrew Morton writes:
> > On Mon, 21 Apr 2008 02:50:42 -0400
> > Erez Zadok <ezk@cs.sunysb.edu> wrote:
> [...]
> > Can we avoid having to think?
> > 
> > void fsstack_copy_inode_size(struct inode *dst, const struct inode *src)
> > {
> > 	blkcnt_t i_blocks;
> > 	loff_t i_size;
> > 
> > 	i_size = i_size_read(src);
> > 	spin_lock_32bit(&src->i_lock);
> > 	i_blocks = src->i_blocks;
> > 	spin_unlock_32bit(&src->i_lock);
> > 
> > 	i_size_write(dst, i_size);
> > 	spin_lock_32bit(&dst->i_lock)
> > 	dst->i_blocks = i_blocks;
> > 	spin_unlock_32bit(&dst->i_lock)
> > }
> 
> Thanks.  I can't say that I'm an expert in these SMP issues.  But I'll run
> your rewritten function through my 32 and 64-bit SMP and non-SMP systems,
> and see how it behaves.
> 

The obvious risk here is that there's no synchronisation between the
copying of i_size and i_blocks.  If that's a problem, I _expect_ that
i_mutex wold give pretty good coverage (but insufficient for
mmap-write-over-a-hole, I guess).

And someone needs to write spin_lock_32bit() ;)