On Mon, Apr 6, 2020 at 10:40 AM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
> >
> > But look at anybody who ever worked more with NFS mounts, and they got
> > used to having the 'intr' mount flag set and incomplete reads and
> > -EAGAIN as a result.
>
> That's why you had me implement TASK_KILLABLE ;-)
Oh, absolutely. We can *NOT* do this in general. Applications _will_
break if you end up just randomly breaking POSIX behavior.
But network filesystems are almost never fully POSIX anyway. And yes,
they do break some apps. 'intr' may not be a thing any more, but
other differences wrt various atomicity guarantees (or file locking)
etc still exist.
So the whole "network filesystems do odd things in corner cases" isn't
exactly unusual.
I think the O_NONBLOCK difference is one of the more benign ones.
I just think it should be documented more.
Linus
