[lm-sensors] Kernel hangs with i2c-i801 driver?

[khali@linux-fr.org (Jean Delvare)]

Hi Daniel,

> Quoting Gigabyte support:
> 
> "Hi Daniel,
> 
> We are sorry for the late reply. Here are some of the answers from our R&D
> department:
> 
> The interrupt pin of the W83792D is connected to one of the GPIO pin of south
> bridge. And that GPIO can trigger SMI.
> 
> If the sensor limit interrupt enable bit is enabled and the sensor reading is
> beyond its limit, the chip will assert the interrupt pin.
> 
> The BIOS logs the events and does fan controls based on the interrupts of
> w83792. But BIOS always enables interrupt of W83792 even when miniBMC is
> disabled because BIOS needs it for fan controls.
> 
> Attached a simple diagram to show you how the W83792D chip is connected 
> to south
> bridge.
> 
> Best regards,
> 
> GBT Tech Support"

Wow, this is way better than they used me to. I'll try to remember
that. This doesn't help me as an individual, but at least they seem to
provide some useful technical detail.

> The picture attached is a very simple drawing showing the south bridge in one
> box and the w83792d in another box. There is one arrow from the south 
> bridge to
> the w83792d labelled "SMI bus" and the there is one arrow from the w83792d to
> the south bridge labelled interrupt.
> 
> Does this provide any more insight? Could this be more of an issue with the
> interrupt coming into the south bridge causing a hang since no code is
> servicing that interrupt? I don't know the kernel that well to guess any
> further though...

I don't know much either, but the schematics would need to mention pin
numbers or names so that we know *which* GPIOs of the i801 are
involved. Is that the case? Without that information, it probably
doesn't help us much.

-- 
Jean Delvare