Re: [PATCH 2/7] cxl/mbox: Add background cmd handling machinery

[Dave Jiang <dave.jiang@intel.com>]

On 4/28/23 10:18 AM, Davidlohr Bueso wrote:
> On Fri, 28 Apr 2023, Dave Jiang wrote:
> 
>>> +static DECLARE_WAIT_QUEUE_HEAD(mbox_wait);
>>
>> I apologize if I've asked this before and you've already answered.
>> What is the reason the mbox_wait a module global wq instead of a per
>> device wq? Just thinking when you tear down a device, you may want to
>> wake all pending for that device to clean up.
> 
> Yes, I agree that doing the wait per-device is better, and not only
> for tear down reasons. By doing it globally, the queue reflects waits
> from different devices, but the driver really has no control about
> the order of the wait of each device is, so upon a blind wake_up(),
> it could perfectly well be that the first node in the waitq is not
> the node for that device.
> 
> This all goes away with per-device, with the note that because of the
> mbox_mutex there never will only ever be a single waiter, so no concept
> of a queue really.

If there's only a single waiter then I think a 'completion' can be used 
instead right?

> 
> Thanks,
> Davidlohr