scsi: mpt3sas: Fix sync irqs

_base_process_reply_queue() called from _base_interrupt() may schedule a
new irq poll. Fix this by calling synchronize_irq() first.

Also ensure that enable_irq() is called only when necessary to avoid
"Unbalanced enable for IRQ..." errors.

Link: https://lore.kernel.org/r/20200910142126.8147-1-thenzl@redhat.com
Fixes: 320e77acb3 ("scsi: mpt3sas: Irq poll to avoid CPU hard lockups")
Acked-by: Sreekanth Reddy <sreekanth.reddy@broadcom.com>
Signed-off-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/mpt3sas/mpt3sas_base.c

@@ -1809,18 +1809,22 @@ mpt3sas_base_sync_reply_irqs(struct MPT3SAS_ADAPTER *ioc, u8 poll)
 		/* TMs are on msix_index == 0 */
 		if (reply_q->msix_index == 0)
 			continue;
+		synchronize_irq(pci_irq_vector(ioc->pdev, reply_q->msix_index));
 		if (reply_q->irq_poll_scheduled) {
 			/* Calling irq_poll_disable will wait for any pending
 			 * callbacks to have completed.
 			 */
 			irq_poll_disable(&reply_q->irqpoll);
 			irq_poll_enable(&reply_q->irqpoll);
+			/* check how the scheduled poll has ended,
+			 * clean up only if necessary
+			 */
+			if (reply_q->irq_poll_scheduled) {
 				reply_q->irq_poll_scheduled = false;
 				reply_q->irq_line_enable = true;
 				enable_irq(reply_q->os_irq);
-			continue;
 			}
-		synchronize_irq(pci_irq_vector(ioc->pdev, reply_q->msix_index));
+		}
 	}
 	if (poll)
 		_base_process_reply_queue(reply_q);