[ofa-general] Re: [PATCH 2.6.24] rdma/cm: fix deadlock destroyinglisten requests

[Sean Hefty]

>I don't understand the reason for callback initiated destruct in the
>first place, but thats too off topic ...

This was a feature request by users.  It ends up being convenient from the
user's perspective, since it avoids needing to queue the id's to another thread
for destruction.  This is particularly important for connection requests, where
the id is a new id.  The user may not be able to allocate space for their
context.  This is how it is used by the ucma and sdp (out of tree module).

>With this new information, I will revisit the thread posted at
>http://lists.openfabrics.org/pipermail/general/2007-September/040614.html
>to see if really the problem being talked about there is non existant.

The expected behavior of the IB and iWarp CMs is that a destroy call for a given
id will block while there's a callback in progress which references it.  The
rdma_cm will only work if the underlying CMs adhere to this.  I would personally
treat issues in iWarp CM synchronization as separate. 

- Sean