Re: 2.6.28-rc9: oprofile regression

[Thomas Gleixner <tglx@linutronix.de>]

Tim,

On Wed, 14 Jan 2009, Tim Blechmann wrote:
> this code (line 81/82), changes counter_width from 32 to 40.
> 
> if (counter_width < eax.split.bit_width)
>         counter_width = eax.split.bit_width;
> 
> however when removing these lines, and thus keeping the value 32 for
> counter_width, doesn't change the behavior, only one NMI per cpu.

It would only help, when the reported bit_width would be bogus. We
know that you get at least one NMI, so lets look at the results we get
there.

Thanks,

	tglx
---
 arch/x86/oprofile/op_model_ppro.c |    4 ++++
 1 file changed, 4 insertions(+)

Index: linux-2.6/arch/x86/oprofile/op_model_ppro.c
===================================================================
--- linux-2.6.orig/arch/x86/oprofile/op_model_ppro.c
+++ linux-2.6/arch/x86/oprofile/op_model_ppro.c
@@ -126,11 +126,15 @@ static int ppro_check_ctrs(struct pt_reg
 	u64 val;
 	int i;
 
+	printk(KERN_INFO "ppro_check_ctrs\n");
+
 	for (i = 0 ; i < num_counters; ++i) {
 		if (!reset_value[i])
 			continue;
 		rdmsrl(msrs->counters[i].addr, val);
+		printk(KERN_INFO "cntr %d: 0x%016lx\n", i, (unsigned long) val);
 		if (CTR_OVERFLOWED(val)) {
+			printk(KERN_INFO "cntr %d overflowed\n", i);
 			oprofile_add_sample(regs, i);
 			wrmsrl(msrs->counters[i].addr, -reset_value[i]);
 		}