Re: [PATCH 08/10] NFSD: allow parallel creates from nfsd

[NeilBrown]

On Sat, 27 Aug 2022, Al Viro wrote:
> On Fri, Aug 26, 2022 at 12:10:43PM +1000, NeilBrown wrote:
> 
> >  	if (is_create_with_attrs(open))
> >  		nfsd4_acl_to_attr(NF4REG, open->op_acl, &attrs);
> > +	inode = d_inode(path.dentry);
> >  
> > -	inode_lock_nested(inode, I_MUTEX_PARENT);
> > +	child = filename_create_one_len(open->op_fname,
> > +					open->op_fnamelen,
> > +					&path, 0, &wq);
> >  
> > -	child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
> > -	if (IS_ERR(child)) {
> > -		status = nfserrno(PTR_ERR(child));
> > -		goto out;
> > -	}
> > +	if (IS_ERR(child))
> > +		return nfserrno(PTR_ERR(child));
> 
> Leaks acls, by the look of it?

Yep - I think that fell through a crack when I reordered patches to get
get some clean-ups into nfsd before this repost.

> 
> > +	if (!IS_PAR_UPDATE(fhp->fh_dentry->d_inode) &&
> > +	    inode_trylock_shared(fhp->fh_dentry->d_inode)) {
> > +		/* only have a shared lock */
> > +		inode_unlock_shared(fhp->fh_dentry->d_inode);
> > +		fhp->fh_no_atomic_attr = true;
> > +		fhp->fh_no_wcc = true;
> 
> Er...  Shouldn't that be IS_PAR_UPDATE() && ... ?
> 
Definitely.  Thanks!

NeilBrown