On Thu, Jan 21, 2021 at 1:40 AM Jiri Slaby [off-list ref] wrote:

Ok, here belongs my earlier note: "if ld == NULL => crash here." That is
if hangup happens during the ldisc wait, the kernel will crash in
tty_ldisc_deref.

Right you are, good catch.

Is there a reason not to convert hung_up_tty_fops too and leave the
return hung_up_tty_write here intact? This would also solve the comments
above.

No, no reason. I started out just changing that one tty_write, then
noticed that I had to change the redirect case too, but never then got
to "yeah, I should have changed the hup case as well".

Greg, do you prefer a new series, or incremental patches?

               Linus