On Mon, Feb 8, 2021 at 11:19 PM Pkshih <pkshih@xxxxxxxxxxx> wrote: > Then, I have another simple thinking -- enable NAPI only if interrupt is > enabled. Other operations with RX ring are working only if interrupt is > disabled. So, we don't need a lock to protect RX ring at all. That makes more sense; thanks for the update. > The irq_lock is still used to protect TX ring/queue, and now it also used > to protect switching IMR. Some comments are added to describe about this. > > Above is implemented in v5. I've taken a brief look, and that looks better. I'll likely provide my Reviewed-by there. Thanks, Brian
