Re: [PATCH RFC 1/3] fs: introduce helper d_path_fast()

[Al Viro]

On Sat, May 08, 2021 at 03:47:38PM -0700, Linus Torvalds wrote:
> On Sat, May 8, 2021 at 3:42 PM Linus Torvalds
> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > But your READ_ONCE() is definitely the right thing to do (whether we
> > do your re-org or not, and whether we do this "prepend_buffer" thing
> > or not).
> 
> Oh, and looking at it some more, I think it would probably be a good
> thing to make __dentry_path() take a
> 
>        struct prepend_buffer *orig
> 
> argument, the same way prepend_path() does. Also, like prepend_path(),
> the terminating NUL should probably be done by the caller.
> 
> Doing those two changes would simplify the hackery we now have in
> "dentry_path()" due to the "//deleted" games. The whole "restore '/'
> that was overwritten by the NUL added by __dentry_path() is
> unbelievably ugly.

	Agreed.  Re READ_ONCE() - we are wrapped into
read_seqbegin_or_lock(&rename_lock, &seq) there, so it's more about
being explicit than about correctness considerations.