Re: watchdog?

[Masao Uebayashi <uebayasi%gmail.com@localhost>]

On Mon, Jun 17, 2013 at 10:37 AM, Jonathan A. Kollasch
<jakllsch%kollasch.net@localhost> wrote:
> On Mon, Jun 17, 2013 at 10:28:43AM +0900, Masao Uebayashi wrote:
>> What do people do to detect hard hang-up and forcibly reboot the system?
>>
>> My understanding is that x86 has no standard way to do this, right?
>>
>> I'm considering to implement something similar to Linux's "NMI
>> watchdog", but before doing that, I want to make sure I'm in the right
>> direction ...
>
> There is an ACPI spec (or two) for watchdog hardware, but AFAIK they are
> not widely implemented.

IIUC, ACPI only provides configuration information of underlying H/W,
correct?  If H/W has no watchdog device, there's no chance for ACPI to
show watchdog info.

I also saw docs mentioning that some Intel chipsets have WDT.  But
info is very limited.  Not sure if it's real.

> LAPIC-counter-triggered NMI is going to be a lot easier to come by.
> Also, NMI can get you into ddb, if you want.

Do you mean "LAPIC-counter" == Performance Monitor overflow interrupt
hooked to NMI via LAPIC?  That's what Linux does.