Re: heavy handed exit() in latest BK

[Ingo Molnar <mingo@elte.hu>]

On Sun, 9 Feb 2003, Roland McGrath wrote:

> >  - a read_lock(&tasklist_lock) is missing around the group_send_sig_info()
> >    in send_sig_info().
> 
> Indeed.  I still intend to clean up those entry points and haven't
> gotten to it, so I hadn't bothered with this yet either (though I think
> I sent it to you for the backport).  It certainly does bite in practice,
> e.g. SIGPIPE.

it does bite - here's the correctness fix meanwhile, until the interface 
is cleaned up.

	Ingo

--- linux/kernel/signal.c.orig	
+++ linux/kernel/signal.c	
@@ -1083,17 +1083,19 @@
 int
 send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
 {
+	int ret;
+
 	/* XXX should nix these interfaces and update the kernel */
-	if (T(sig, SIG_KERNEL_BROADCAST_MASK))
-		/* XXX do callers really always hold the tasklist_lock?? */
-		return group_send_sig_info(sig, info, p);
-	else {
-		int error;
+	if (T(sig, SIG_KERNEL_BROADCAST_MASK)) {
+		read_lock(&tasklist_lock);
+		ret = group_send_sig_info(sig, info, p);
+		read_unlock(&tasklist_lock);
+	} else {
 		spin_lock_irq(&p->sighand->siglock);
-		error = specific_send_sig_info(sig, info, p);
+		ret = specific_send_sig_info(sig, info, p);
 		spin_unlock_irq(&p->sighand->siglock);
-		return error;
 	}
+	return ret;
 }
 
 int