Re: [PATCH 1/2] pipe: change pipe_write() to never add a zero-sized buffer

[Linus Torvalds]

On Sun, 9 Feb 2025 at 07:08, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
>
> This is no longer necessary after c73be61cede5 ("pipe: Add general notification
> queue support"), pipe_write() checks pipe_has_watch_queue() and returns -EXDEV
> at the start. And can't help in any case, pipe_write() no longer takes this
> rd_wait.lock spinlock.

Ack. This code all goes back to the two-level locking thing with
notifications using just the spinlock side.

The locking was removed from this code in commit dfaabf916b1c
("fs/pipe: remove unnecessary spinlock from pipe_write()"), but the
"pre-allocate" logic remained.

This patch seems to be the right thing to do and removes the vestiges
of the old model.

But I don't think you need that pipe_buf_assert_len() thing.  And if
you do, please don't make it a pointless inline helper that only hides
what it does.

             Linus