Intel and using pins for multiple purposes

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
I was wondering why I couldn't use the watchdog timer function on one of  
my computers - the driver says that it's been disabled in hardware.

A look at the chipset page reveals that when the system is powered up, a  
pin is sampled, and if it is high, the watchdog timer function is indeed  
disabled.

The pin has a pulldown resistor, which the spec warns can be as high as 50K.

Only thing is, the rest of the time, the pin is the output intended to  
drive (indirectly) the computer speaker.

So if the speaker driver circuitry is implemented in a way that could  
behave as a modest pullup, it could unintentionally disable the watchdog  
timer function. I rather suspect that this is what is happening, since I  
can't see any reason for a MB manufacturer to deliberately disable such  
an obscure function.

Which leaves two questions.

1) Was the speaker output pin really the best choice for this?

2) Where is the need to disable the watchdog timer function at all? If a  
system doesn't program it to operate, then it won't. There seems no  
earthly need to make it impossible for software to use it. It's not even  
the kind of thing that would make for product differentiation (charge  
more for a board without the functionality disabled).

Interested readers can find the spec here:

<http://www.intel.com/content/dam/doc/datasheet/6-chipset-c200-chipset-datasheet.pdf

Search for "No Reboot (NR)" without the quotes.

Grrrrr!

Sylvia.


Re: Intel and using pins for multiple purposes

Quoted text here. Click to load it

What do you need the watchdog for?

Quoted text here. Click to load it



Re: Intel and using pins for multiple purposes
On 22/09/2013 7:57 PM, yaputya wrote:
Quoted text here. Click to load it

Restart the system if it hangs.

Sylvia.


Re: Intel and using pins for multiple purposes

Quoted text here. Click to load it

Yeah but why do you need it running unattended?
If it hangs you must have a software/hardware problem.
You shouldn't be using a Windoze PC for anything critical anyway.





Re: Intel and using pins for multiple purposes
Quoted text here. Click to load it

It could also be something as simple as a flipped bit pursuant to a  
voltage/magnetic surge. There's a reason why 100% of dedicated servers  
actually have a functional watchdog implementation.

Quoted text here. Click to load it

I didn't see anywhere in her post that she'd written she was using Windows.

Just for the record though, I have a server sitting here which has been  
running Windows (2K Server) continuously since late 2002 and hasn't once  
crashed due to software failure (I'll admit that it's had two PSUs and one  
HDD replaced in that time - but neither of these have anything to do with  
choice of operating system). I also have a Debian machine which falls over  
of its own accord at least once a week due to memory leaks in a driver.

--  
Bob Milutinovic
Cognicom


Re: Intel and using pins for multiple purposes
On 23/09/2013 1:08 AM, yaputya wrote:
Quoted text here. Click to load it

Who said it was Windows?

I say I want to use the watchdog timer. I can't for the life of me see  
how you can conclude that I shouldn't want to.

Sylvia.


Re: Intel and using pins for multiple purposes

Quoted text here. Click to load it

I assumed so, around 90% of PCs use it.
Quoted text here. Click to load it

You normally only use a watchdog to reboot when a problem occurs.
If you are having PC problems this ng might be able to help you fix them directly.
So, why do you need a watchdog in the first place?


Quoted text here. Click to load it



Re: Intel and using pins for multiple purposes
On 23/09/2013 4:04 PM, yaputya wrote:
Quoted text here. Click to load it

So you assumed that I was doing something critical, but that I was using  
Windows for it, despite, in your view, the inappropriateness of that.  
Seems a bit arrogant to me.

Quoted text here. Click to load it

I can't fix a problem until it exists, and if one arises, then without  
the watchdog timer, the system will hang until such time as I notice,  
which I don't want.

Pretty much the standard use of watchdog timer, indeed.

Sylvia.


Re: Intel and using pins for multiple purposes

Quoted text here. Click to load it

Not really, and no offence intended.
I simply meant *if* it was critical, you shouldn't depend on Windows.
Since you haven't given a reason why you want to go to the trouble of using
a watchdog, I still don't know if your application is critical or not.
If you are using Windows, just say so.

Quoted text here. Click to load it

A watchdog reset won't fix hardware software problems.
And if you aren't around to notice a crash, you won't be around to see the watchdog stuck
in an endless reboot sequence either, if the fault is still present. Unless you write recovery
code your PC will not resume where it left off either.

Quoted text here. Click to load it

You still haven't explained what you are doing with the PC that justifies
a watchdog.  You may be better off with a more stable O/S instead.

Quoted text here. Click to load it



Re: Intel and using pins for multiple purposes
On 23/09/2013 5:14 PM, yaputya wrote:
Quoted text here. Click to load it

It already resumes its critical task on a reboot.

Quoted text here. Click to load it

Remind me why it needs to be justified.

Sylvia.


Re: Intel and using pins for multiple purposes
Quoted text here. Click to load it

I can only imagine that Intel provided a method for disabling the hardware  
watchdog in order to avoid an infinite boot loop in the event that the timer  
was set to too short a duration.

I haven't looked into the architecture, but the speaker output would likely  
be the least critical pin (system function wise), and would tie in well with  
the "system diagnostics" side of things - the motherboard manufacturer  
definitely seems to have botched it.

If you have a decent (i.e., very high impedance) multimeter, check the  
voltage on that pin at boot time. A decent DSO would be even better, but I  
suspect it's a simple resistive pull up to +Vcc rather than a pulse in  
response to something else.

If it is indeed pulling it up, simply cut the speaker circuit off - I'd  
suspect in your scenario watchdog time functionality would be more  
beneficial than the provision of a power-on beep. If you still want  
provision for power-on diagnostics, put a switch in there, to switch the  
speaker circuit back in if the system fails.

--  
Bob Milutinovic
Cognicom


Re: Intel and using pins for multiple purposes
On 23/09/2013 12:12 AM, Bob Milutinovic wrote:
Quoted text here. Click to load it

Though the only thing that could stuff things up that badly would be the  
BIOS, and if a BIOS developer gets things that wrong on a system with a  
BIOS rom that's soldered into place, and thereby not reflashable, they  
strongly deserve to have to buy a new motherboard.
Quoted text here. Click to load it

After posting, I did reflect on it some more, and it certainly seems  
that the most obvious speaker driver implementation (and we're not  
talking Hi-Fi here), is not likely to function as a pull-up. I've  
entered a support ticket with the board manufacturer - but what are the  
odds that they'll just tell me to reflash to the most recent BIOS? I  
already did anyway, before looking at the hardware details. It didn't  
help, of course.

Quoted text here. Click to load it

Unfortunately, I can't even reach the pin - it's a surface mounted  
device soldered to the board with an array of pins. I've tried  
backtracking from the loudspeaker driver (using a magnifying glass), but  
I quickly get lost, and I suspect this is a multi-layer board. Buying an  
X-ray machine to diagnose this seems a bit over-the-top.

Quoted text here. Click to load it

If necessary, I'll just buy a third-party watchdog card.

Sylvia.

Re: Intel and using pins for multiple purposes
On 21/09/2013 7:19 PM, Sylvia Else wrote:
Quoted text here. Click to load it

I submitted a support ticket to the board manufacturer, whose initial  
response was that they don't support Linux, and that I should install a  
Windows operating system and try it there.

I said I'be be willing to try, but that as far I can see there's no  
official driver for the hardware.

Their response is now that I contact Microsort for further information.

Perhaps they mean that if I get enough information from Microsoft to  
write a driver for the hardare, do so, and then it doesn't work, they'll  
take the matter seriously.

The real situation seems to be that they don't support any chipset  
feature that Windows doesn't use, and it's pot-luck whether it'll work.

The board, BTW, is a GA-H61M-S2PV (rev 2.0). I've also tried a revision  
2.2, with the same result.

Sylvia.

Site Timeline