[ofw] RE: ibal pnp question

[Fab Tillier]

>> I'm debugging WinMad and hitting into an issue where the winmad driver
>> has received pnp notification that a CA has been added and is in the D0
>> power state, but the CA is not yet known to ibal.  Does anyone know
>> what Windows PnP corresponds to when IBAL is aware of a device
> existing?
>
> The best that I can tell is that there is no correlation between a
> Windows AddDevice PnP event and an IBAL AddDevice PnP event.  Does
> anyone know the code well enough to confirm this?

They aren't correlated.  However, the HCA driver registers itself with IBAL from its start device handler I believe, and that will queue a CA_ADD IBAL PnP event for processing.  So the Windows PnP event triggers the IBAL PnP event, but there's no synchronization between the two - the IBAL PnP event could lag the Windows one by an indeterminate amount.

> Can an IBAL
> AddDevice PnP event occur after the device has been removed?  (It
> looks like it to me, but I may not be seeing where that
> synchronization is.)

No.  When pnp_ca_event runs for CA_ADD, a reference is taken on the ci_ca object.  That reference is released after the async callback item for CA_REMOVE has run.  This effectively prevents the ci_ca from being destroyed until the IBAL PnP event queue is drained (at least through the CA_REMOVE event).

-Fab