Re: [PATCH] NFSD: Fix nfs4_verifier memory alignment

["J. Bruce Fields" <bfields@fieldses.org>]

On Mon, Mar 12, 2012 at 11:24:18AM -0400, Chuck Lever wrote:
> 
> On Mar 12, 2012, at 11:04 AM, J. Bruce Fields wrote:
> 
> > On Mon, Mar 12, 2012 at 10:57:32AM -0400, Chuck Lever wrote:
> >> 
> >> On Mar 12, 2012, at 10:20 AM, J. Bruce Fields wrote:
> >> 
> >>> On Fri, Mar 02, 2012 at 05:13:50PM -0500, Chuck Lever wrote:
> >>>> Clean up due to code review.
> >>>> 
> >>>> The nfs4_verifier's data field is not guaranteed to be u32-aligned.
> >>>> Casting an array of chars to a u32 * is considered generally
> >>>> hazardous.
> >>>> 
> >>>> We can fix most of this by using a __be32 array to generate the
> >>>> verifier's contents and then byte-copying it into the verifier field.
> >>>> 
> >>>> However, there is one spot where there is a backwards compatibility
> >>>> constraint: the do_nfsd_create() call expects a verifier which is
> >>>> 32-bit aligned.  Fix this spot by forcing the alignment of the create
> >>>> verifier in the nfsd4_open args structure.
> >>>> 
> >>>> Also, sizeof(nfs4_verifer) is the size of the in-core verifier data
> >>>> structure, but NFS4_VERIFIER_SIZE is the number of octets in an XDR'd
> >>>> verifier.  The two are not interchangeable, even if they happen to
> >>>> have the same value.
> >>>> 
> >>>> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
> >>>> ---
> >>>> 
> >>>> Hi Bruce-
> >>>> 
> >>>> Compile-tested only.  Does this look reasonable?
> >>> 
> >>> Looks fine, but the setclientid verifier stuff belongs in a separate
> >>> patch.
> >> 
> >> Thanks for the review.  I'm not clear on exactly which hunks you would like split.
> > 
> > This:
> > 
> >>>> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> >>>> index c5cddd6..9f0e139 100644
> >>>> --- a/fs/nfsd/nfs4state.c
> >>>> +++ b/fs/nfsd/nfs4state.c
> >>>> @@ -1138,12 +1138,12 @@ static void gen_clid(struct nfs4_client *clp)
> >>>> 
> >>>> static void gen_confirm(struct nfs4_client *clp)
> >>>> {
> >>>> +	__be32 verf[2];
> >>>> 	static u32 i;
> >>>> -	u32 *p;
> >>>> 
> >>>> -	p = (u32 *)clp->cl_confirm.data;
> >>>> -	*p++ = get_seconds();
> >>>> -	*p++ = i++;
> >>>> +	verf[0] = (__be32)get_seconds();
> >>>> +	verf[1] = (__be32)i++;
> >>>> +	memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
> > 
> > This cl_confirm verifier really has nothing to do with the write
> > verifier (though maybe it has a similar problem).
> 
> The patch fixes nfs4_verifiers, of which cl_confirm is one.  We can't guarantee access to an nfs4_verifier field, which is an array of char, using (u32 *).  It just happens to work now on architectures we test regularly.
> 
> This seems perfectly relevant to the patch description to me.  Do you still want this hunk split into a separate patch?

Nah, I guess I can live with it as is.

Applying (pending some testing), thanks.--b.