> What makes you think there is a relation between SMM and ACPI? There should be. See ACPI specs and Global lock chapter. It should be a hw lock between ACPI and SMM code touching the hw. Maybe we are lucky and something like this is really implemented. But since the 1) is true I would suspect that we just see some other transactions. >> 3) Make a driver more bullet proof. Maybe it would be possible somehow to force >> the driver to have a longer timeouts, more retries etc. > > This is pointless. If another entity is accessing the chip without > proper locking, no amount of timeouts or retries will help. In this > scenario, reported errors are in fact the best thing than can happen. > The worst case is silent misbehavior. It is not pointless if this entity is just observed. I mean it is not using the i2c hw like our driver but it is using the bus itself. Thanks, Rudolf
