Re: [PATCH RFC 18/20] fs: add f_pipe

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

On Fri, Aug 30, 2024 at 03:04:59PM +0200, Christian Brauner wrote:
> Only regular files with FMODE_ATOMIC_POS and directories need
> f_pos_lock. Place a new f_pipe member in a union with f_pos_lock
> that they can use and make them stop abusing f_version in follow-up
> patches.

Not sure I like that - having lseek(2) use a separate primitive
instead of fdget_pos(), grabbing ->f_pos_lock for _everything_ that
has FMODE_LSEEK, directory or no directory, would simplify quite
a few things.  OTOH, that will affect only the explanation of validity -
pipes do *not* have FMODE_LSEEK, so it becomes "fdget_pos() and
fdget_seek() are the only things that might want ->f_pos_lock, and
neither touch it for pipes - fdget_pos() because FMODE_ATOMIC_POS
is not there and fdget_seek() because FMODE_LSEEK isn't".

Oh, well...