[openib-general] Re: RFC: ib_cache_event problems

[Roland Dreier]

    Michael> Hmm. Thinking about it some more - how about generating
    Michael> the events from a mad thread in core rather than from
    Michael> povider?  Then this would be thread context so cache
    Michael> could simply perform updates inline in event handler.

You have the same problem with allocating storage in atomic context
for an arbitrary-length queue of events.

 - R.