Re: [KVM-RFC PATCH 1/2] eventfd: add an explicit srcu based notifier interface

[Gregory Haskins <ghaskins@novell.com>]

Davide Libenzi wrote:
> On Thu, 18 Jun 2009, Gregory Haskins wrote:
>
>>
>> But the fact is: I do not see any way to actually use your referenceless
>> POLLHUP release code in a race free way without doing something like I
>> propose in 3/4, 4/4.   Lets keep the discussion focused on that for now,
>> if we could.
>
> OK, since I got literally swamped by the amount of talks and patches over 
> this theoretically simple topic, would you mind  1) posting the global 
> patch over eventfd

Done.  Should show up as replies to this email.  These apply to v2.6.30 and
are devoid of any KVM-isms. :)

(Note-1: I've built and run these patches against additional kvm/irqfd
patches and verified they all work properly)

(Note-2: I included your POLLHUP patch as 1/3 since it currently only exists
in kvm.git, and is a pre-req for the other two)

> 2) describe exactly what races are you talking about


Covered in the patch headers (3/3 probably has the most detail)

 
> 3) explain why this should be any concern of eventfd at all?
>

To support any  in-kernel eventfd client that wants to get both "signal"
and "release" notifications (based on using your POLLHUP patch) in a
race-free way.  Without 2/3, 3/3, I cannot see how this can be done.  Yet
your 1/3 patch is an important enhancement (to at least one client).  I
suspect any other in-kernel users will find this a good feature as well.

Thanks Davide,
-Greg