Re: [PATCH] Re: sound and control-c

Wakko Warner (wakko@animx.eu.org)
Sun, 6 Jun 1999 22:05:33 -0400

FYI, this happens on the sparc arch as well, No, I can't test it. I no
longer have the speaker box, it never would work for me (the exp dbri driver
kept crashing the kernel. No, I don't have logs of that because I started
over).

I just thought you'd like to know that x86 wasn't the only arch having that
prob.

> [I CC this to Alan because it is, in a way, sound related, and I couldn't
> find anyone maintaining signal handling...]
>
> > I've noticed that the problem also occurs with the wavplay and
> > playmidi programs (and probably others). I just gave 'cat' as the
> > simplest example...
> >
> > > Maybe this is a non-issue, hence. When I get some time, I'll take a look
> > > again.
>
> Okay. I had some time (actually, that's a lie. I don't have any time. I just
> decided to postpone something else), so I decided to figure out what's wrong
> with ctrl-c and sound. Here are my results.
>
> Stopping pending sound with the same signal that caused the writing program
> to terminate - doesn't work. For example, when playing sound with
> 'cat soundfile.wav > /dev/dsp', and using SIGKILL to kill it, another
> SIGKILL wouldn't stop the pending sound. Some other signal would.
>
> This happens due to the following bit of code
> (linux/arch/i386/kernel/signal.c:do_signal() - and applies to other arches
> quite equally):
>
> default:
> lock_kernel();
> sigaddset(&current->signal, signr);
> current->flags |= PF_SIGNALED;
> do_exit(exit_code);
> /* NOTREACHED */
>
> This is executed when a signal is sent to a process, such that the signal is
> supposed to kill the process (like, a signal that isn't captured).
>
> What happens is that the pending signal is added back to the signal set
> (it was previously removed, when the processing started). However, the
> sigpending bit of the task isn't updated. This simply means that this signal
> can have no effect from this point on, unless some other signal is sent.
>
> The patch below applies on 2.2.9 (and probably on mostly anything else). One
> line. Works fine here (I've been running with it for the past few hours).
> If this seems to be the correct thing to do, it should probably be added for
> the other arches too.
>
>
> --- linux/arch/i386/kernel/signal.c.ORIG Tue Dec 1 21:28:24 1998
> +++ linux/arch/i386/kernel/signal.c Sat Jun 5 11:44:02 1999
> @@ -698,6 +698,7 @@
> default:
> lock_kernel();
> sigaddset(&current->signal, signr);
> + recalc_sigpending(current);
> current->flags |= PF_SIGNALED;
> do_exit(exit_code);
> /* NOTREACHED */
>
>
>
> Nimrod
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/