Skip to contentSkip to conversation

prevent nested panic from soft lockup detection

PATCH Andi Kleen 2 participants 2 messages Jan 31, 2006 2 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Andi Kleen

author
#2 lore raw
Quoted 25 lines · Dave Jones
Quoted 23 lines · Jan Beulich

From: Jan Beulich <[email protected]>

Suppress triggering a nested panic due to soft lockup detection.

diff -Npru /home/jbeulich/tmp/linux-2.6.16-rc1/kernel/panic.c 2.6.16-rc1-panic-softlockup/kernel/panic.c

kernel/panic.c renamed from home/jbeulich/tmp/linux-2.6.16-rc1/kernel/panic.c +2 -0
@@ -107,6 +107,7 @@ NORET_TYPE void panic(const char * fmt, 
 		printk(KERN_EMERG "Rebooting in %d seconds..",panic_timeout);
 		for (i = 0; i < panic_timeout*1000; ) {
 			touch_nmi_watchdog();
+			touch_softlockup_watchdog();
 			i += panic_blink(i);
 			mdelay(1);
 			i++;
@@ -130,6 +131,7 @@ NORET_TYPE void panic(const char * fmt, 
 #endif
 	local_irq_enable();
 	for (i = 0;;) {
+		touch_softlockup_watchdog();
 		i += panic_blink(i);
Quoted 8 lines

kernel/panic.c NORET_TYPE void panic(const char * fmt, · line 137

Quoted 2 lines
kernel/panic.c renamed from home/jbeulich/tmp/linux-2.6.16-rc1/kernel/panic.c +0 -0
 		mdelay(1);
 		i++;

I've been wondering for a while why we don't just make touch_nmi_watchdog do an implicit call to touch_softlockup_watchdog. I can't think of a situation where we'd want to do one but not the other, and adding patches like this seems to be an uphill battle (I know at least two other places that need this off the top of my head).

Very good idea.

Someone did it already in the SUSE kernel and it helped considerably there.

-Andi

#3 lore raw
Quoted 10 lines
Quoted 5 lines

I've been wondering for a while why we don't just make touch_nmi_watchdog do an implicit call to touch_softlockup_watchdog. I can't think of a situation where we'd want to do one but not the other, and adding patches like this seems to be an uphill battle (I know at least two other places that need this off the top of my head).

Very good idea.

Someone did it already in the SUSE kernel and it helped considerably there.

Actually, plain 2.6.15 already has this (for i386 and x86-64 at least). Hence the first of the two hunks the patch consists of is superfluous. The second hunk, however, is still necessary (as there's no pre-existing touch_nmi_watchdog() call there, and there also shouldn't be one as interrupts get re-enabled before getting there).

Jan