Re: memory leak in nfs_get_client

[Benjamin Coddington]

On 2 Jul 2019, at 12:11, Eric Biggers wrote:

> On Tue, Jul 02, 2019 at 07:23:32AM -0400, Benjamin Coddington wrote:
> > On 2 Jul 2019, at 2:31, Eric Biggers wrote:
> >
> > > On Tue, Jun 11, 2019 at 12:23:12PM -0400, Benjamin Coddington wrote:
> > > > Ugh.. Now that you can cancel the wait, you have to also handle if
> > > > "new" was
> > > > allocated.  I think this needs:
> > > >
> > > > diff --git a/fs/nfs/client.c b/fs/nfs/client.c
> > > > index d7e4f0848e28..4d90f5bf0b0a 100644
> > > > --- a/fs/nfs/client.c
> > > > +++ b/fs/nfs/client.c
> > > > @@ -406,10 +406,10 @@ struct nfs_client *nfs_get_client(const 
> > > > struct
> > > > nfs_client_initdata *cl_init)
> > > >                 clp = nfs_match_client(cl_init);
> > > >                 if (clp) {
> > > >                         spin_unlock(&nn->nfs_client_lock);
> > > > -                       if (IS_ERR(clp))
> > > > -                               return clp;
> > > >                         if (new)
> > > >                                 new->rpc_ops->free_client(new);
> > > > +                       if (IS_ERR(clp))
> > > > +                               return clp;
> > > >                         return nfs_found_client(cl_init, clp);
> > > >                 }
> > > >                 if (new) {
> > > >
> > > > I'll patch/test and send it along.
> > > >
> > > > Ben
> > >
> > > Hi Ben, what happened to this patch?
> >
> > I sent it along:
> >
> > https://lore.kernel.org/linux-nfs/65b675cec79d140df64bc30def88b1def32bf87e.1560272160.git.bcodding@xxxxxxxxxx/
> >
> > I don't think it will go in 5.2.. it's not a huge problem.
> >
> > Ben
>
> Okay, great.  I didn't see it in linux-next and there was no further 
> reply to
> this thread, which usually (having seen it happen on lots of syzbot 
> bugs) means
> the person forgot about it.
>
> Tip: you can use the '--in-reply-to=<MESSAGE_ID>' option to 'git 
> send-email' or
> 'git format-patch' to send the patch in response to the original 
> thread, which
> makes it very easy to see that a patch was actually sent out.

Yep, that's nice for those following along, but sometimes I think that 
makes
it harder for the maintainers to scrape the patches off the list.  I've 
had
patches get dropped that were buried at the end of a long discussion, so
I've made a habit of always making patches the top.

Ben