Re: Pending splice(file -> FIFO) excludes all other FIFO operations forever (was: ... always blocks read(FIFO), regardless of O_NONBLOCK on read side?)

[Ahelenia Ziemiańska <nabijaczleweli@xxxxxxxxxxxxxxxxxx>]

On Mon, Jun 26, 2023 at 05:56:28PM +0200, Christian Brauner wrote:
> I mean, one workaround would probably be poll() even with O_NONBLOCK but
> I get why that's annoying and half of a solution.
poll() doesn't really change anything since it turns into a hotloop of
.revents=POLLHUP with a disconnected writer, cf.
  https://lore.kernel.org/linux-fsdevel/jbyihkyk5dtaohdwjyivambb2gffyjs3dodpofafnkkunxq7bu@jngkdxx65pux/t/#m747e2bbd0c5cffb6baaf1955f6f8b0d97e216839
The only apparently-supported way to poll pipes under linux is to
sleep()/read(O_NONBLOCK) like in the bad old days.

And even if that was a working work-around, the fundamental problem of
./spl>fifo excluding all other access to fifo is quite unfortunate too.

> tl;dr it by splicing from a regular file to a pipe where the regular
> file in splice isn't O_NONBLOCK
(Most noticeable when the "regular file" is a tty and thus the I/O never
 completes by design.)

Attachment: signature.asc
Description: PGP signature