Thread (24 messages) 24 messages, 5 authors, 2025-05-09
STALE416d
Revisions (3)
  1. v5 current
  2. v6 [diff vs current]
  3. v7 [diff vs current]

[PATCH v5 9/9] x86/nmi: Include NMI-source information in tracepoint and debug prints

From: Sohil Mehta <hidden>
Date: 2025-05-07 01:21:30
Also in: kvm, linux-edac, linux-perf-users, linux-pm, lkml
Subsystem: the rest, tracing, x86 architecture (32-bit and 64-bit) · Maintainers: Linus Torvalds, Steven Rostedt, Masami Hiramatsu, Thomas Gleixner, Ingo Molnar, Borislav Petkov, Dave Hansen

The NMI-source bitmap is the most critical information provided by the
NMI-source reporting feature. It can help identify issues when multiple
NMIs occur simultaneously or if certain NMI handlers consistently
misbehave. It is also very useful in debugging unknown NMIs since it can
pinpoint the exact source that caused the NMI.

Add the source bitmap to the nmi_handler() tracepoint. Also, print the
bitmap along with the "Unknown NMI" kernel log message.

Signed-off-by: Sohil Mehta <redacted>
---
v5: New patch
---
 arch/x86/kernel/nmi.c      |  5 ++++-
 include/trace/events/nmi.h | 13 ++++++++-----
 2 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index 183e3e717326..b9ece0b63ca7 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -202,7 +202,7 @@ static int nmi_handle(unsigned int type, struct pt_regs *regs)
 		thishandled = a->handler(type, regs);
 		handled += thishandled;
 		delta = sched_clock() - delta;
-		trace_nmi_handler(a->handler, (int)delta, thishandled);
+		trace_nmi_handler(a->handler, (int)delta, thishandled, source_bitmap);
 
 		nmi_check_duration(a, delta);
 	}
@@ -387,6 +387,9 @@ unknown_nmi_error(unsigned char reason, struct pt_regs *regs)
 	pr_emerg_ratelimited("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
 			     reason, smp_processor_id());
 
+	if (cpu_feature_enabled(X86_FEATURE_NMI_SOURCE))
+		pr_emerg_ratelimited("NMI-source bitmap is 0x%lx\n", fred_event_data(regs));
+
 	if (unknown_nmi_panic || panic_on_unrecovered_nmi)
 		nmi_panic(regs, "NMI: Not continuing");
 
diff --git a/include/trace/events/nmi.h b/include/trace/events/nmi.h
index 18e0411398ba..6e4a1ff70a44 100644
--- a/include/trace/events/nmi.h
+++ b/include/trace/events/nmi.h
@@ -10,29 +10,32 @@
 
 TRACE_EVENT(nmi_handler,
 
-	TP_PROTO(void *handler, s64 delta_ns, int handled),
+	TP_PROTO(void *handler, s64 delta_ns, int handled, unsigned long source_bitmap),
 
-	TP_ARGS(handler, delta_ns, handled),
+	TP_ARGS(handler, delta_ns, handled, source_bitmap),
 
 	TP_STRUCT__entry(
 		__field(	void *,		handler	)
 		__field(	s64,		delta_ns)
 		__field(	int,		handled	)
+		__field(	unsigned long,	source_bitmap)
 	),
 
 	TP_fast_assign(
 		__entry->handler = handler;
 		__entry->delta_ns = delta_ns;
 		__entry->handled = handled;
+		__entry->source_bitmap = source_bitmap;
 	),
 
-	TP_printk("%ps() delta_ns: %lld handled: %d",
+	TP_printk("%ps() delta_ns: %lld handled: %d source_bitmap: 0x%lx",
 		__entry->handler,
 		__entry->delta_ns,
-		__entry->handled)
+		__entry->handled,
+		__entry->source_bitmap)
 );
 
 #endif /* _TRACE_NMI_H */
 
-/* This part ust be outside protection */
+/* This part must be outside protection */
 #include <trace/define_trace.h>
-- 
2.43.0
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help