Re: [PATCH 1/2] VFS: change kern_path_locked() and user_path_locked_at() to never return negative dentry

[NeilBrown]

On Fri, 07 Feb 2025, Kent Overstreet wrote:
> On Fri, Feb 07, 2025 at 05:34:23PM +1100, NeilBrown wrote:
> > On Fri, 07 Feb 2025, Kent Overstreet wrote:
> > > On Fri, Feb 07, 2025 at 03:53:52PM +1100, NeilBrown wrote:
> > > > Do you think there could be a problem with changing the error returned
> > > > in this circumstance? i.e. if you try to destroy a subvolume with a
> > > > non-existant name on a different filesystem could getting -ENOENT
> > > > instead of -EXDEV be noticed?
> > > 
> > > -EXDEV is the standard error code for "we're crossing a filesystem
> > > boundary and we can't or aren't supposed to be", so no, let's not change
> > > that.
> > > 
> > 
> > OK.  As bcachefs is the only user of user_path_locked_at() it shouldn't
> > be too hard.
> 
> Hang on, why does that require keeping user_path_locked_at()? Just
> compare i_sb...
> 

I changed user_path_locked_at() to not return a dentry at all when the
full path couldn't be found.  If there is no dentry, then there is no
->d_sb.
(if there was an ->i_sb, there would be an inode and this all wouldn't
be an issue).

To recap: the difference happens if the path DOESN'T exist but the
parent DOES exist on a DIFFERENT filesystem.  It is very much a corner
case and the error code shouldn't matter.  But I had to ask...

NeilBrown