Skip to contentSkip to conversation

kgdb 2.6.27-rc8 fixes

GIT PULL New Jason Wessel 1 participant 3 messages Oct 6, 2008 x86

lore
Share & tools
lore
#0 lore raw

Linus, please pull the kgdb git tree fixes for 2.6.27-rc8

git://git.kernel.org/pub/scm/linux/kernel/git/jwessel/linux-2.6-kgdb.git for_linus

Summary:

Testing the kgdb with kvm has found 2 problems:
- NMI code has a race where it can be invoked twice
- The softlockup watchdog must be touched when running with CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE=1, else the kernel executes a reboot if debugged for too long.

Short log follows:

---

Jan Kiszka (1):
kgdb, x86: Avoid invoking kgdb_nmicallback twice per NMI

Jason Wessel (1):
kgdb: call touch_softlockup_watchdog on resume

 arch/x86/kernel/kgdb.c |    7 +------
 kernel/kgdb.c          |    3 +++
 2 files changed, 4 insertions(+), 6 deletions(-)
#1 lore raw

PATCH 1/2 kgdb, x86: Avoid invoking kgdb_nmicallback twice per NMI

From: Jan Kiszka <[email protected]>

Stress-testing KVM's latest NMI support with kgdbts inside an SMP guest, I came across spurious unhandled NMIs while running the singlestep test. Looking closer at the code path each NMI takes when KGDB is enabled, I noticed that kgdb_nmicallback is called twice per event: One time via DIE_NMI_IPI notification, the second time on DIE_NMI. Removing the first invocation cures the unhandled NMIs here.

---
 arch/x86/kernel/kgdb.c |    7 +------
 1 files changed, 1 insertions(+), 6 deletions(-)
arch/x86/kernel/kgdb.c +1 -6
index 8282a21..10435a1 100644
@@ -455,12 +455,7 @@ static int __kgdb_notify(struct die_args *args, unsigned long cmd)
 		return NOTIFY_DONE;
 
 	case DIE_NMI_IPI:
-		if (atomic_read(&kgdb_active) != -1) {
-			/* KGDB CPU roundup */
-			kgdb_nmicallback(raw_smp_processor_id(), regs);
-			was_in_debug_nmi[raw_smp_processor_id()] = 1;
-			touch_nmi_watchdog();
-		}
+		/* Just ignore, we will handle the roundup on DIE_NMI. */
 		return NOTIFY_DONE;
 
 	case DIE_NMIUNKNOWN:
Signature · 1 line
1.6.0.2
#2 lore raw

PATCH 2/2 kgdb: call touch_softlockup_watchdog on resume

The softlockup watchdog needs to be touched when resuming the from the kgdb stopped state to avoid the printk that a CPU is stuck if the debugger was active for longer than the softlockup threshold.

---
 kernel/kgdb.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
kernel/kgdb.c +3 -0
index 25d955d..e4dcfb2 100644
@@ -590,6 +590,7 @@ static void kgdb_wait(struct pt_regs *regs)
 
 	/* Signal the primary CPU that we are done: */
 	atomic_set(&cpu_in_kgdb[cpu], 0);
+	touch_softlockup_watchdog();
 	clocksource_touch_watchdog();
 	local_irq_restore(flags);
 }
@@ -1432,6 +1433,7 @@ acquirelock:
 	    atomic_read(&kgdb_cpu_doing_single_step) != cpu) {
 
 		atomic_set(&kgdb_active, -1);
+		touch_softlockup_watchdog();
 		clocksource_touch_watchdog();
 		local_irq_restore(flags);
 
@@ -1524,6 +1526,7 @@ acquirelock:
 kgdb_restore:
 	/* Free kgdb_active */
 	atomic_set(&kgdb_active, -1);
+	touch_softlockup_watchdog();
 	clocksource_touch_watchdog();
 	local_irq_restore(flags);
 
Signature · 1 line
1.6.0.2