Re: [PATCH RT 2/2 v4] preempt-rt/x86: Delay calling signals in int3

On Sun, 2012-02-05 at 20:31 +0100, Oleg Nesterov wrote:

Damn. Sorry for noise...

On 02/05, Oleg Nesterov wrote:

quoted

+int force_sig_info(int sig, struct siginfo *info, struct task_struct *t)
+{
+#ifdef CONFIG_PREEMPT_RT_FULL
+	if (in_atomic()) {
+		if (WARN_ON_ONCE(t != current))

This is certainly wrong in upstream kernel. It does use force_
this way although it shouldn't imho.

It's wrong in upstream even with the #ifdef define here?

But _probably_ this is fine for rt? We are going to take the mutex,
we shouldn't do this in atomic context. But, once again, I do not
really know what in_atomic() means in rt.

in_atomic() is the same in rt as in mainline. It should still work.

-- Steve