On 2024-06-12 12:49:21 [-0700], Vinicius Costa Gomes wrote: > > diff --git a/drivers/net/ethernet/intel/igc/igc_main.c > > b/drivers/net/ethernet/intel/igc/igc_main.c > > index 305e05294a26..e666739dfac7 100644 > > --- a/drivers/net/ethernet/intel/igc/igc_main.c > > +++ b/drivers/net/ethernet/intel/igc/igc_main.c > > @@ -5811,11 +5815,23 @@ static void igc_watchdog_task(struct work_struct > > *work) > > if (adapter->flags & IGC_FLAG_HAS_MSIX) { > > u32 eics = 0; > > > > - for (i = 0; i < adapter->num_q_vectors; i++) > > - eics |= adapter->q_vector[i]->eims_value; > > - wr32(IGC_EICS, eics); > > + for (i = 0; i < adapter->num_q_vectors; i++) { > > + struct igc_ring *rx_ring = adapter->rx_ring[i]; > > + > > + if (test_bit(IGC_RING_FLAG_RX_ALLOC_FAILED, > > &rx_ring->flags)) { > > Minor and optional: I guess you can replace test_bit() -> clear_bit() > with __test_and_clear_bit() here and below.
That are two steps, first test+clear is merged into one and then __ is added. The former is doable but it will always lead to a write operation while in the common case the flag isn't set so it will be skipped. Adding the __ leads to an unlocked operation and I don't see how this is synchronized against the other writes. In fact, nobody else is doing it. Sebastian