1
2Is your ix86 system locking up unpredictably? No keyboard activity, just
3a frustrating complete hard lockup? Do you want to help us debugging
4such lockups? If all yes then this document is definitely for you.
5
6On Intel and similar ix86 type hardware there is a feature that enables
7us to generate 'watchdog NMI interrupts'.  (NMI: Non Maskable Interrupt
8which get executed even if the system is otherwise locked up hard).
9This can be used to debug hard kernel lockups.  By executing periodic
10NMI interrupts, the kernel can monitor whether any CPU has locked up,
11and print out debugging messages if so.  You must enable the NMI
12watchdog at boot time with the 'nmi_watchdog=n' boot parameter.  Eg.
13the relevant lilo.conf entry:
14
15        append="nmi_watchdog=1"
16
17For SMP machines and UP machines with an IO-APIC use nmi_watchdog=1.
18For UP machines without an IO-APIC use nmi_watchdog=2, this only works
19for some processor types.  If in doubt, boot with nmi_watchdog=1 and
20check the NMI count in /proc/interrupts; if the count is zero then
21reboot with nmi_watchdog=2 and check the NMI count.  If it is still
22zero then log a problem, you probably have a processor that needs to be
23added to the nmi code.
24
25A 'lockup' is the following scenario: if any CPU in the system does not
26execute the period local timer interrupt for more than 5 seconds, then
27the NMI handler generates an oops and kills the process. This
28'controlled crash' (and the resulting kernel messages) can be used to
29debug the lockup. Thus whenever the lockup happens, wait 5 seconds and
30the oops will show up automatically. If the kernel produces no messages
31then the system has crashed so hard (eg. hardware-wise) that either it
32cannot even accept NMI interrupts, or the crash has made the kernel
33unable to print messages.
34
35NOTE: starting with 2.4.2-ac18 the NMI-oopser is disabled by default,
36you have to enable it with a boot time parameter.  Prior to 2.4.2-ac18
37the NMI-oopser is enabled unconditionally on x86 SMP boxes.
38
39[ feel free to send bug reports, suggestions and patches to
40  Ingo Molnar <mingo@redhat.com> or the Linux SMP mailing
41  list at <linux-smp@vger.kernel.org> ]
42
43