Re: [PATCH] coredump: Retry writes where appropriate

[Oleg Nesterov <oleg@redhat.com>]

On 06/01, Alan Cox wrote:
>
> > > If you are on the command line then SIGINT/SIGQUIT would be the obvious
> > > ones for this ?
> >
> > Not sure I understand. Do you mean we should treat the tty signals
> > specially ?
>
> Yes
>
> > Personally, I don't think we should. If we decide that only SIGKILL
> > interrupts the coredumping, then I think ^C should not interrupt.
>
> Wait until you have a remote session over ssh that core dumps a 2GB core.
> Then you'll understand why being able to ^C or ^\ it is useful.

Sure. But you have the same problem with

	$ perl -e '$SIG{INT} = $SIG{QUIT} = IGNORE; sleep'
	^C^C^C^\^\^\

over ssh.

And what if the coredumping task already has the pending SIGINT/SIGQUIT
or blocks/ignores them?

But don't get me wrong, I do agree this is useful, and this can be
implemented. But in this case, imho kill(SIGINT) should work as well,
not just ^C.


In short, I agree in advance with any authoritative decision. But
if we add more power to ^C (compared to kill), this should be a
separate patch imho.

Oleg.