Re: [2.6.29-git12] BUG: scheduling while atomic: swapper/0/0x10000100

[Thomas Gleixner <tglx@linutronix.de>]

On Fri, 10 Apr 2009, Maciej Rutecki wrote:

> 2009/4/10 Andrew Morton <akpm@linux-foundation.org>
> 
> > Did this get fixed?
> 
> On 2.6.30-rc1-git3 still the same:
> [  162.541793] BUG: scheduling while atomic: swapper/0/0x10000100

Yikes.	

0x10000100
       ^----- softirq count leaked.

We have a check for that in do_softirq, but none in do_IRQ. Can you
please apply the patch below and test whether it triggers ?

Thanks,

	tglx
---
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index c3fe010..d164396 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -206,6 +206,7 @@ unsigned int __irq_entry do_IRQ(struct pt_regs *regs)
 	/* high bit used in ret_from_ code  */
 	unsigned vector = ~regs->orig_ax;
 	unsigned irq;
+	int prev_count = preempt_count();
 
 	exit_idle();
 	irq_enter();
@@ -225,6 +226,13 @@ unsigned int __irq_entry do_IRQ(struct pt_regs *regs)
 
 	irq_exit();
 
+	if (unlikely(prev_count != preempt_count())) {
+		printk(KERN_ERR "huh, entered interrupt %d"
+		       "with preempt_count %08x, exited with %08x?\n", irq,
+		       prev_count, preempt_count());
+		preempt_count() = prev_count;
+	}
+
 	set_irq_regs(old_regs);
 	return 1;
 }