Re: checksignals hackery?

Linus Torvalds (torvalds@transmeta.com)
Sat, 5 Sep 1998 16:53:07 -0700 (PDT)

On Sat, 5 Sep 1998, Richard Henderson wrote:
>
> void checksignals(void)
> {
> sigset_t *blocked = &current->blocked;
> unsigned long mask = blocked->sig[0] | sigmask(SIGKILL) | sigmask(SIGINT) | sigmask(SIGQUIT);
> mask &= blocked->sig[1];
> if (~mask) {
> printk("Bad signal mask\n");
> *(int *) 0 = 0;
> }
> }
>
> It seems to be confused wrt sig[1]. Should that have been
> an IOR instead of an AND?

No.

The thing makes sure that the blocked set contains _everything_ except for
SIGKILL/SIGINT/SIGQUIT, because the RPC code really wouldn't work if there
are other signals that could come through.

So the "and" is there to make sure that all bits are set.

> And why, if it is important at all, it was implemented in
> arch/i386/kernel/signal.c rather than kernel/signal.c? Is
> this just some big of debug hackery that was accidentally
> left in?

It wasn't exactly accidental, and it already found one case where we
allowed signals we really didn't want to allow. But yes, it's debugging
code, and I expect to remove it very shortly when I feel comfortable that
all NFS paths do indeed block signals properly..

Linus

-
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/faq.html