Re: [PATCH V10 12/15] ptp: Added a brand new class driver for ptp clocks.

On Fri, Feb 11, 2011 at 09:15:24AM +0100, Richard Cochran wrote:

On Tue, Feb 01, 2011 at 06:00:31PM -0800, John Stultz wrote:

quoted

So what is serializing access to the timestamp_event_queue here? I don't
see any usage of tsevq_mux by the callers. Am I missing it? It looks
like its called from interrupt context, so do you really need a spinlock
and not a mutex here?

The external timestamp FIFO is written only from interrupt context.

Oops, I lied. After reworking the phyter driver, I find that I want to
enqueue time stamps from non-interrupt contexts, too.

So, you are right. This will need a spin lock...

Thanks,

Richard