linuxppc-dev.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
* [RFC PATCH] powerpc: improve accounting of non maskable interrupts
@ 2017-07-29 13:22 Nicholas Piggin
  0 siblings, 0 replies; 3+ messages in thread
From: Nicholas Piggin @ 2017-07-29 13:22 UTC (permalink / raw)
  To: linuxppc-dev; +Cc: Nicholas Piggin

This fixes a case of double counting MCEs on PowerNV.

Adds a counter for the system reset interrupt, which will
see more use as a debugging NMI.

Adds a soft-NMI counter for the 64s watchdog. Although this could cause
confusion because it only fires when interrupts are soft-disabled, so it
won't increment much even when the watchdog is running.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
I can split these out or drop any objectionable bits. At least the
MCE we should fix, not sure if the other bits are wanted.

Thanks,
Nick

 arch/powerpc/include/asm/hardirq.h |  4 ++++
 arch/powerpc/kernel/irq.c          | 16 ++++++++++++++++
 arch/powerpc/kernel/traps.c        |  9 +++++++++
 arch/powerpc/kernel/watchdog.c     |  3 +++
 4 files changed, 32 insertions(+)

diff --git a/arch/powerpc/include/asm/hardirq.h b/arch/powerpc/include/asm/hardirq.h
index 8add8b861e8d..a3c83ec416c6 100644
--- a/arch/powerpc/include/asm/hardirq.h
+++ b/arch/powerpc/include/asm/hardirq.h
@@ -12,6 +12,10 @@ typedef struct {
 	unsigned int mce_exceptions;
 	unsigned int spurious_irqs;
 	unsigned int hmi_exceptions;
+	unsigned int sreset_irqs;
+#if defined(CONFIG_HARDLOCKUP_DETECTOR) && defined(CONFIG_HAVE_HARDLOCKUP_DETECTOR_ARCH)
+	unsigned int soft_nmi_irqs;
+#endif
 #ifdef CONFIG_PPC_DOORBELL
 	unsigned int doorbell_irqs;
 #endif
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 0bcec745a672..36250df64615 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -470,6 +470,18 @@ int arch_show_interrupts(struct seq_file *p, int prec)
 		seq_printf(p, "  Hypervisor Maintenance Interrupts\n");
 	}
 
+	seq_printf(p, "%*s: ", prec, "NMI");
+	for_each_online_cpu(j)
+		seq_printf(p, "%10u ", per_cpu(irq_stat, j).sreset_irqs);
+	seq_printf(p, "  System Reset interrupts\n");
+
+#if defined(CONFIG_HARDLOCKUP_DETECTOR) && defined(CONFIG_HAVE_HARDLOCKUP_DETECTOR_ARCH)
+	seq_printf(p, "%*s: ", prec, "WDG");
+	for_each_online_cpu(j)
+		seq_printf(p, "%10u ", per_cpu(irq_stat, j).soft_nmi_irqs);
+	seq_printf(p, "  Watchdog soft-NMI interrupts\n");
+#endif
+
 #ifdef CONFIG_PPC_DOORBELL
 	if (cpu_has_feature(CPU_FTR_DBELL)) {
 		seq_printf(p, "%*s: ", prec, "DBL");
@@ -494,6 +506,10 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
 	sum += per_cpu(irq_stat, cpu).spurious_irqs;
 	sum += per_cpu(irq_stat, cpu).timer_irqs_others;
 	sum += per_cpu(irq_stat, cpu).hmi_exceptions;
+	sum += per_cpu(irq_stat, cpu).sreset_irqs;
+#ifdef CONFIG_HARDLOCKUP_DETECTOR
+	sum += per_cpu(irq_stat, cpu).soft_nmi_irqs;
+#endif
 #ifdef CONFIG_PPC_DOORBELL
 	sum += per_cpu(irq_stat, cpu).doorbell_irqs;
 #endif
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index bfcfd9ef09f2..6a892ca7bf18 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -288,6 +288,8 @@ void system_reset_exception(struct pt_regs *regs)
 	if (!nested)
 		nmi_enter();
 
+	__this_cpu_inc(irq_stat.sreset_irqs);
+
 	/* See if any machine dependent calls */
 	if (ppc_md.system_reset_exception) {
 		if (ppc_md.system_reset_exception(regs))
@@ -755,7 +757,14 @@ void machine_check_exception(struct pt_regs *regs)
 	enum ctx_state prev_state = exception_enter();
 	int recover = 0;
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	/* 64s accounts the mce in machine_check_early when in HVMODE */
+	if (!cpu_has_feature(CPU_FTR_HVMODE))
+		__this_cpu_inc(irq_stat.mce_exceptions);
+#else
 	__this_cpu_inc(irq_stat.mce_exceptions);
+#endif
+
 
 	add_taint(TAINT_MACHINE_CHECK, LOCKDEP_NOW_UNRELIABLE);
 
diff --git a/arch/powerpc/kernel/watchdog.c b/arch/powerpc/kernel/watchdog.c
index b67f8b03a32d..4b9a567c9975 100644
--- a/arch/powerpc/kernel/watchdog.c
+++ b/arch/powerpc/kernel/watchdog.c
@@ -204,6 +204,9 @@ void soft_nmi_interrupt(struct pt_regs *regs)
 		return;
 
 	nmi_enter();
+
+	__this_cpu_inc(irq_stat.soft_nmi_irqs);
+
 	tb = get_tb();
 	if (tb - per_cpu(wd_timer_tb, cpu) >= wd_panic_timeout_tb) {
 		per_cpu(wd_timer_tb, cpu) = tb;
-- 
2.11.0

^ permalink raw reply related	[flat|nested] 3+ messages in thread
* Re: [RFC PATCH] powerpc: improve accounting of non maskable interrupts
  2017-07-29 13:22 Nicholas Piggin
@ 2017-07-31 10:17 ` Michael Ellerman
  0 siblings, 0 replies; 3+ messages in thread
From: Michael Ellerman @ 2017-07-31 10:17 UTC (permalink / raw)
  To: Nicholas Piggin, linuxppc-dev; +Cc: Nicholas Piggin

Nicholas Piggin <npiggin@gmail.com> writes:

> This fixes a case of double counting MCEs on PowerNV.
>
> Adds a counter for the system reset interrupt, which will
> see more use as a debugging NMI.
>
> Adds a soft-NMI counter for the 64s watchdog. Although this could cause
> confusion because it only fires when interrupts are soft-disabled, so it
> won't increment much even when the watchdog is running.
>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
> I can split these out or drop any objectionable bits. At least the
> MCE we should fix, not sure if the other bits are wanted.

Yeah if you can split it please that would be good.

I'm not sure how useful it is to count the SOFT NMIs. I guess it's not
much overhead so we may as well, if nothing else it might be handy for
us debugging.

If you can make the ifdef look less horrendous that would be great :D -
perhaps a kconfig symbol that keys off both.

cheers

^ permalink raw reply	[flat|nested] 3+ messages in thread
* [RFC PATCH] powerpc: improve accounting of non maskable interrupts
@ 2017-07-29 13:22 Nicholas Piggin
  2017-07-31 10:17 ` Michael Ellerman
  0 siblings, 1 reply; 3+ messages in thread
From: Nicholas Piggin @ 2017-07-29 13:22 UTC (permalink / raw)
  To: linuxppc-dev; +Cc: Nicholas Piggin

This fixes a case of double counting MCEs on PowerNV.

Adds a counter for the system reset interrupt, which will
see more use as a debugging NMI.

Adds a soft-NMI counter for the 64s watchdog. Although this could cause
confusion because it only fires when interrupts are soft-disabled, so it
won't increment much even when the watchdog is running.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
I can split these out or drop any objectionable bits. At least the
MCE we should fix, not sure if the other bits are wanted.

Thanks,
Nick

 arch/powerpc/include/asm/hardirq.h |  4 ++++
 arch/powerpc/kernel/irq.c          | 16 ++++++++++++++++
 arch/powerpc/kernel/traps.c        |  9 +++++++++
 arch/powerpc/kernel/watchdog.c     |  3 +++
 4 files changed, 32 insertions(+)

diff --git a/arch/powerpc/include/asm/hardirq.h b/arch/powerpc/include/asm/hardirq.h
index 8add8b861e8d..a3c83ec416c6 100644
--- a/arch/powerpc/include/asm/hardirq.h
+++ b/arch/powerpc/include/asm/hardirq.h
@@ -12,6 +12,10 @@ typedef struct {
 	unsigned int mce_exceptions;
 	unsigned int spurious_irqs;
 	unsigned int hmi_exceptions;
+	unsigned int sreset_irqs;
+#if defined(CONFIG_HARDLOCKUP_DETECTOR) && defined(CONFIG_HAVE_HARDLOCKUP_DETECTOR_ARCH)
+	unsigned int soft_nmi_irqs;
+#endif
 #ifdef CONFIG_PPC_DOORBELL
 	unsigned int doorbell_irqs;
 #endif
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 0bcec745a672..36250df64615 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -470,6 +470,18 @@ int arch_show_interrupts(struct seq_file *p, int prec)
 		seq_printf(p, "  Hypervisor Maintenance Interrupts\n");
 	}
 
+	seq_printf(p, "%*s: ", prec, "NMI");
+	for_each_online_cpu(j)
+		seq_printf(p, "%10u ", per_cpu(irq_stat, j).sreset_irqs);
+	seq_printf(p, "  System Reset interrupts\n");
+
+#if defined(CONFIG_HARDLOCKUP_DETECTOR) && defined(CONFIG_HAVE_HARDLOCKUP_DETECTOR_ARCH)
+	seq_printf(p, "%*s: ", prec, "WDG");
+	for_each_online_cpu(j)
+		seq_printf(p, "%10u ", per_cpu(irq_stat, j).soft_nmi_irqs);
+	seq_printf(p, "  Watchdog soft-NMI interrupts\n");
+#endif
+
 #ifdef CONFIG_PPC_DOORBELL
 	if (cpu_has_feature(CPU_FTR_DBELL)) {
 		seq_printf(p, "%*s: ", prec, "DBL");
@@ -494,6 +506,10 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
 	sum += per_cpu(irq_stat, cpu).spurious_irqs;
 	sum += per_cpu(irq_stat, cpu).timer_irqs_others;
 	sum += per_cpu(irq_stat, cpu).hmi_exceptions;
+	sum += per_cpu(irq_stat, cpu).sreset_irqs;
+#ifdef CONFIG_HARDLOCKUP_DETECTOR
+	sum += per_cpu(irq_stat, cpu).soft_nmi_irqs;
+#endif
 #ifdef CONFIG_PPC_DOORBELL
 	sum += per_cpu(irq_stat, cpu).doorbell_irqs;
 #endif
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index bfcfd9ef09f2..6a892ca7bf18 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -288,6 +288,8 @@ void system_reset_exception(struct pt_regs *regs)
 	if (!nested)
 		nmi_enter();
 
+	__this_cpu_inc(irq_stat.sreset_irqs);
+
 	/* See if any machine dependent calls */
 	if (ppc_md.system_reset_exception) {
 		if (ppc_md.system_reset_exception(regs))
@@ -755,7 +757,14 @@ void machine_check_exception(struct pt_regs *regs)
 	enum ctx_state prev_state = exception_enter();
 	int recover = 0;
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	/* 64s accounts the mce in machine_check_early when in HVMODE */
+	if (!cpu_has_feature(CPU_FTR_HVMODE))
+		__this_cpu_inc(irq_stat.mce_exceptions);
+#else
 	__this_cpu_inc(irq_stat.mce_exceptions);
+#endif
+
 
 	add_taint(TAINT_MACHINE_CHECK, LOCKDEP_NOW_UNRELIABLE);
 
diff --git a/arch/powerpc/kernel/watchdog.c b/arch/powerpc/kernel/watchdog.c
index b67f8b03a32d..4b9a567c9975 100644
--- a/arch/powerpc/kernel/watchdog.c
+++ b/arch/powerpc/kernel/watchdog.c
@@ -204,6 +204,9 @@ void soft_nmi_interrupt(struct pt_regs *regs)
 		return;
 
 	nmi_enter();
+
+	__this_cpu_inc(irq_stat.soft_nmi_irqs);
+
 	tb = get_tb();
 	if (tb - per_cpu(wd_timer_tb, cpu) >= wd_panic_timeout_tb) {
 		per_cpu(wd_timer_tb, cpu) = tb;
-- 
2.11.0

^ permalink raw reply related	[flat|nested] 3+ messages in thread
end of thread, other threads:[~2017-07-31 10:17 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-07-29 13:22 [RFC PATCH] powerpc: improve accounting of non maskable interrupts Nicholas Piggin
2017-07-29 13:22 Nicholas Piggin
2017-07-31 10:17 ` Michael Ellerman

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).