On Fri, 6 Sep 2024 16:49:32 +0200 Paolo Abeni wrote:
> I forgot to mention there is another, easier, alternative: keep the max 
> queue id check in the drivers. The driver will have to acquire and held 
> in the shaper callbacks the relevant driver-specific lock - 'crit_lock', 
> in the iavf case.
> 
> Would you be ok with such 2nd option?

I'd strongly prefer if you implemented what was suggested.

> Side note: I think the iavf should have to acquire such lock in the 
> callbacks no matter what or access/write to the rings info could be racy.

Reply via email to