Re: [PATCH] NMI request/release, version 3

[Dipankar Sarma]

On Wed, Oct 23, 2002 at 01:03:11PM -0500, Corey Minyard wrote:
> >
> Yes, it's not correct, I will fix it.  I didn't realize there was an 
> rcu-safe list.  That should make things simpler.

Yeah, RCU documentation needs some serious updating :(

One other thing, it might be better to do all handler checks
in request_nmi and release_nmi in the spinlock serialized
critical section to minimize memory barrier issues unlike in the code
snippet I mailed - things like the list_empty() check etc.

That way you need to look at memory barrier issues with only
one piece of lockfree code - call_nmi_handlers.

Thanks
Dipankar