Sure

It was a belt-and-braces approach. I had without initially and it seemed to have no effect, and I noticed ixgbe quiesce does something similar, so I thought it would be a good idea.

I wanted to do this but the core function to do this acquires locks which is a no-no for the quiesce. I have seen no ill effects from it, but instead I could do (and have tested) this:
/*
 * Taken from the common code to shutdown an adminq ring
 */
static void
i40e_stop_adminq_ring(i40e_hw_t *hw, struct i40e_adminq_ring *q)
{
        I40E_WRITE_REG(hw, q->head, 0);
        I40E_WRITE_REG(hw, q->tail, 0);
        I40E_WRITE_REG(hw, q->len, 0);
        I40E_WRITE_REG(hw, q->bal, 0);
        I40E_WRITE_REG(hw, q->bah, 0);
}



called in i40e_quiesce:
        i40e_shutdown_rx_rings(i40e);
        i40e_shutdown_tx_rings(i40e);

        /*
         * Stop the adminq send and receive 

         */
        i40e_stop_adminq_ring(hw, &hw->aq.asq);
        i40e_stop_adminq_ring(hw, &hw->aq.arq);

        i40e_intr_chip_fini(i40e);

        (void) i40e_shutdown_lan_hmc(hw);



The call to i40e_intr_chip_fini() will ensure the interrupts for adminq are disabled.