[PATCH] eliminate spurious pointless WARN_ON()s

[PATCH] eliminate spurious pointless WARN_ON()s

[Jan Beulich]

Namely during early boot, the panic() or BUG() paths may end up in
smp_call_function_*() with just a single online CPU. In that situation
the warnings generated are not only meaningless, but also result in
relevant output being cluttered.

Therefore, defer the WARN_ON() checks until after the (unaffected from
the problem that is being attempted to be detected here) cases have
been handled.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Cc: Andi Kleen <andi@firstfloor.org>

---
 kernel/smp.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

--- linux-2.6.29-rc7/kernel/smp.c	2009-03-04 09:11:00.000000000 +0100
+++ 2.6.29-rc7-smp-call-defer-warn/kernel/smp.c	2009-02-13 12:17:02.000000000 +0100
@@ -222,9 +222,6 @@ int smp_call_function_single(int cpu, vo
 	int me = get_cpu();
 	int err = 0;
 
-	/* Can deadlock when called with interrupts disabled */
-	WARN_ON(irqs_disabled());
-
 	if (cpu == me) {
 		local_irq_save(flags);
 		func(info);
@@ -232,6 +229,9 @@ int smp_call_function_single(int cpu, vo
 	} else if ((unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) {
 		struct call_single_data *data;
 
+		/* Can deadlock when called with interrupts disabled */
+		WARN_ON(irqs_disabled());
+
 		if (!wait) {
 			/*
 			 * We are calling a function on a single CPU
@@ -325,9 +325,6 @@ void smp_call_function_many(const struct
 	unsigned long flags;
 	int cpu, next_cpu;
 
-	/* Can deadlock when called with interrupts disabled */
-	WARN_ON(irqs_disabled());
-
 	/* So, what's a CPU they want?  Ignoring this one. */
 	cpu = cpumask_first_and(mask, cpu_online_mask);
 	if (cpu == smp_processor_id())
@@ -347,6 +344,9 @@ void smp_call_function_many(const struct
 		return;
 	}
 
+	/* Can deadlock when called with interrupts disabled */
+	WARN_ON(irqs_disabled());
+
 	data = kmalloc(sizeof(*data) + cpumask_size(), GFP_ATOMIC);
 	if (unlikely(!data)) {
 		/* Slow path. */

Re: [PATCH] eliminate spurious pointless WARN_ON()s

[Ingo Molnar]

* Jan Beulich <jbeulich@novell.com> wrote:

> Namely during early boot, the panic() or BUG() paths may end 
> up in smp_call_function_*() with just a single online CPU. In 
> that situation the warnings generated are not only 
> meaningless, but also result in relevant output being 
> cluttered.
> 
> Therefore, defer the WARN_ON() checks until after the 
> (unaffected from the problem that is being attempted to be 
> detected here) cases have been handled.

Makes sense probably - but please send a patch against -tip as 
we've got quite many changes queued up in that area and your 
patch does not apply cleanly:

 patching file kernel/smp.c
 Hunk #1 FAILED at 222.
 Hunk #2 FAILED at 229.
 Hunk #3 FAILED at 325.
 Hunk #4 FAILED at 344.
 4 out of 4 hunks FAILED -- rejects in file kernel/smp.c

Thanks,

	Ingo