Re: [PATCH 0/3] nfsd: plug some filecache refcount leaks

[Jeff Layton]

On Wed, 2024-07-10 at 10:35 -0400, Chuck Lever wrote:
> On Wed, Jul 10, 2024 at 09:05:30AM -0400, Jeff Layton wrote:
> > Youzhong Yang sent an email to the list (along with some draft
> > patches)
> > that indicated some nfsd_file refcount leaks. I went crawling over
> > the
> > filecache code (again) and found a couple of places where we didn't
> > put
> > references when we should. I'm not sure if it'll fix the problem
> > they
> > reported, but they are bugs.
> > 
> > Plus, let's start counting nfsd_file allocations. The last patch
> > adds
> > support for this.
> > 
> > Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > ---
> > Jeff Layton (3):
> >       nfsd: fix refcount leak when failing to hash nfsd_file
> >       nfsd: fix refcount leak when file is unhashed after being
> > found
> >       nfsd: count nfsd_file allocations
> > 
> >  fs/nfsd/filecache.c | 14 +++++++++++---
> >  1 file changed, 11 insertions(+), 3 deletions(-)
> > ---
> > base-commit: 24decb225ed20a5ba46a79f4609e109cb0e4a359
> > change-id: 20240710-nfsd-next-01e2afdebb31
> > 
> > Best regards,
> > -- 
> > Jeff Layton <jlayton@xxxxxxxxxx>
> > 
> 
> I've browsed these, they seem straightforward.
> 
> Since this is the week before a merge window, I would like to save
> these and Youzhong's patch for v6.12. The Fixes: tags will get them
> pulled back into the stable kernels once they are merged.
> 
> 'Salright?
> 

Agreed. Leaks suck, but they (usually) aren't fatal, and interactions
in the filecache can be...subtle. I wouldn't mind a bit more testing
before we send this on.

Thanks!
-- 
Jeff Layton <jlayton@xxxxxxxxxx>