Re: [PATCH] vfs: transitive upgrade restrictions for fds

[Greg KH]

On Thu, Mar 26, 2026 at 12:09:08PM +0100, Jori Koolstra wrote:
> 
> > Op 24-03-2026 15:37 CET schreef Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx>:
> > 
> > Any chance to get a test for this as well to know if this keeps working
> > (or works at all)?
> > 
> 
> I do have some tests, can shape them up a bit so that they may go in selftest
> (if this fd stuff goes through). How should I do this? Send all as a v2, or
> inline it as a response here?

v2 when you send it.

> > >  /*
> > >   * Helper to directly jump to a known parsed path from ->get_link,
> > >   * caller must have taken a reference to path beforehand.
> > >   */
> > > -int nd_jump_link(const struct path *path)
> > > +int nd_jump_link_how(const struct path *path, const struct jump_how how)
> > 
> > Shouldn't that be "const struct jump_how *how"?
> > 
> > Or do you really want to pass this structure on the stack?
> > 
> 
> I save a pointer indirection, and jump_how is right now just an int. That was
> at least my reasoning here.

It looks "odd" as it's a struct, and we don't know the size, and it's
the only place in the patch that does it this way, so it "stood out".

thanks,

greg k-h