Re: [PATCH 22/23] fs: default to generic_file_splice_read for files having ->read_iter

[Al Viro]

On Thu, Jul 30, 2020 at 05:17:01PM +0100, Al Viro wrote:
> On Thu, Jul 30, 2020 at 05:20:46PM +0200, Christoph Hellwig wrote:
> 
> > Fortunately I think the fix is pretty easy - remove the special pipe
> > zero copy optimization from copy_page_to_iter, and just have the
> > callers actually want it because they have pagecache or similar
> > refcountable pages use it explicitly for the ITER_PIPE case.  That gives
> > us a safe default with an opt-in into the optimized variant.  I'm
> > currently auditing all the users of for how it is used and that looks
> > pretty promising.
> 
> Huh?  What does that have to do with anything?

FWIW, none of the dubious (and outright broken) cases I've found go anywhere
near that.  And it definitely won't help tun/tap...