Re: [viro-vfs:work.fdtable 13/13] kernel/fork.c:3242 unshare_fd() warn: passing a valid pointer to 'PTR_ERR'

[Al Viro <viro@zeniv.linux.org.uk>]

On Wed, Aug 21, 2024 at 12:38:48AM -0600, Shuah Khan wrote:
> On 8/13/24 19:03, Al Viro wrote:
> > On Tue, Aug 13, 2024 at 07:16:00PM +0100, Al Viro wrote:
> > > On Tue, Aug 13, 2024 at 11:00:04AM +0300, Dan Carpenter wrote:
> > > > 3f4b0acefd818e Al Viro           2024-08-06  3240  		if (IS_ERR(*new_fdp)) {
> > > > 3f4b0acefd818e Al Viro           2024-08-06  3241  			*new_fdp = NULL;
> > > > 3f4b0acefd818e Al Viro           2024-08-06 @3242  			return PTR_ERR(new_fdp);
> > > >                                                                                 ^^^^^^^^^^^^^^^^
> > > > 	err = PTR_ERR(*new_fdp);
> > > > 	*new_fdp = NULL;
> > > > 	return err;
> > > 
> > > Argh...  Obvious braino, but what it shows is that failures of that
> > > thing are not covered by anything in e.g. LTP.  Or in-kernel
> > > self-tests, for that matter...
> > 
> > FWIW, this does exercise that codepath, but I would really like to
> > have kselftest folks to comment on the damn thing - I'm pretty sure
> > that it's _not_ a good style for those.
> 
> Looks good to me. Would you be able to send a patch for this new test?

Umm...  Send as in "mail somewhere specific", or as "push into vfs.git",
or...?