[openib-general] RDMA Generic Connection Management

[Caitlin Bestler]

> -----Original Message-----
> From: openib-general-bounces at openib.org 
> [mailto:openib-general-bounces at openib.org] On Behalf Of Roland Dreier
> Sent: Wednesday, August 31, 2005 10:53 AM
> To: Sean Hefty
> Cc: openib-general at openib.org
> Subject: Re: [openib-general] RDMA Generic Connection Management
> 
>     Sean> For this to work, we should ensure that a client is never
>     Sean> given a reference to a device that they may have received
>     Sean> removal notification for.
> 
> This is the hard part: one CPU could start calling into a 
> consumer with a valid device, but get delayed by an interrupt 
> or something.  In the meantime, another CPU could remove that 
> device from the consumer, and then when the first 
> notification finally arrives, it's no longer valid.
> 

A further complication is ensuring that the problem is not
solved twice. The device layer may already have logic to
ensure that concurrent callbacks from the driver layer 
do not conflict (including the simplest of not allowing
concurrent callbacks).