On Wed, Feb 6, 2019 at 4:28 PM Eric Dumazet <eric.duma...@gmail.com> wrote: > > > > On 02/06/2019 04:04 PM, Cong Wang wrote: > > > synchronize_irq() is called before mlx5_cq_put(), so I don't > > see why readers could get 0 refcnt. > > Then the more reasons to get rid of the refcount increment/decrement > completely ... > > Technically, even the rcu_read_lock() and rcu_read_unlock() are not needed, > since synchronize_irq() is enough. >
I already suggested this, quoting myself from my first reply to this patch V0: "another way to do it is not to do any refcounting in the irq handler and fence cq removal via synchronize_irq(eq->irqn) on mlx5_eq_del_cq." I already have a patch I was just waiting for Cong to push V2.
