Re: [PATCH v3] fuse: In fuse_flush only wait if someone wants the return code

[Serge E. Hallyn]

On Tue, Jan 03, 2023 at 07:51:22AM -0700, Tycho Andersen wrote:
> On Mon, Dec 19, 2022 at 12:16:50PM -0700, Tycho Andersen wrote:
> > On Thu, Dec 08, 2022 at 10:49:30AM -0700, Tycho Andersen wrote:
> > > On Thu, Dec 08, 2022 at 03:26:19PM +0100, Miklos Szeredi wrote:
> > > > On Mon, 28 Nov 2022 at 16:01, Tycho Andersen <tycho@tycho.pizza> wrote:
> > > > >
> > > > > Hi Milkos,
> > > > >
> > > > > On Mon, Nov 14, 2022 at 09:02:09AM -0700, Tycho Andersen wrote:
> > > > > > v3: use schedule_work() to avoid other sleeps in inode_write_now() and
> > > > > >     fuse_sync_writes(). Fix a UAF of the stack-based inarg.
> > > > >
> > > > > Thoughts on this version?
> > > > 
> > > > Skipping attr invalidation on success is wrong.
> > > 
> > > Agreed, that looks like my mistake.
> > > 
> > > > How about the attached (untested) patch?
> > > 
> > > It passes my reproducer with no warnings or anything. Feel free to
> > > add:
> > > 
> > > Tested-by: Tycho Andersen <tycho@tycho.pizza>
> > > 
> > > if you want to commit it.
> > 
> > Ping, thoughts on landing this?
> 
> Happy new year all. Any update here?
> 
> Thanks,
> 
> Tycho

Thanks for pushing on this, Tycho.  I'd suggest sending a clean new version
incorporating Miklos' fix.

-serge