be2net: Prevent CQ full condition for Lancer

Indicate to HW that the CQ is cleaned up before posting new RX buffers.
This prevents the HW to go into CQ full error condition.

Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
Padmanabh Ratnakar 2011-11-03 01:49:55 +00:00 коммит произвёл David S. Miller
Родитель 1610c79f1e
Коммит 9372cacb30
1 изменённых файлов: 4 добавлений и 4 удалений

Просмотреть файл

@ -1905,6 +1905,8 @@ loop_continue:
be_rx_stats_update(rxo, rxcp);
}
be_cq_notify(adapter, rx_cq->id, false, work_done);
/* Refill the queue */
if (work_done && atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM)
be_post_rx_frags(rxo, GFP_ATOMIC);
@ -1912,10 +1914,8 @@ loop_continue:
/* All consumed */
if (work_done < budget) {
napi_complete(napi);
be_cq_notify(adapter, rx_cq->id, true, work_done);
} else {
/* More to be consumed; continue with interrupts disabled */
be_cq_notify(adapter, rx_cq->id, false, work_done);
/* Arm CQ */
be_cq_notify(adapter, rx_cq->id, true, 0);
}
return work_done;
}