[openib-general] APM: QP migration state change when failover triggered by hw

[Sean Hefty]

> I am testing APM with kernel module which directly interfaces with
>ib_verbs.ko and ib_cm.ko.
>Yes, I do receive IB_MIG_MIGRATED event, but the QP's mig_state is not
>actually changed to MIGRATED. So I had to do this from my module.

The ib_cm does not perform QP state transitions.  That is left to the ULPs.
(It's difficult to push this into the ib_cm because of the potential race
between the IB CM modifying the QP at the same time a ULP tries to destroy it.
The ULP may also want to post receives, or modify the QP size, etc. based on
connection information.)

>It could be a bug with ib_cm code, which may not be transitioning the QP
>state correctly. But the HW may be thinking that it has migrated. I am
>not sure how exactly ib_cm  should notice this event and should should
>transition the QP state. Any thoughts and suggestions are welcome. I can
>code it and test it.

There is a pending patch that was recently posted (dispatch communication
establish event) that can be extended to pass path migration events to the
ib_cm.  The purpose of passing path migration events to the ib_cm would be
limited to changing the path that future CM messages, and not related to QP
transitions.

- Sean