[SCSI] lpfc 8.3.15: BSG, Discovery, and Misc fixes

- BSG interface related:
  - Fix node reference count if node is active
  - Warn if we're overwriting an active CT context

- Discovery related:
  - Clear "Ignore Reg Login" flag when purging mailbox queue
  - Pay attention to return code for fc_block_scsi_eh()
  - Stall device loss code if we're almost done when it fires
    (we're logged in, but PRLI is outstanding)

- Bugs
  - Correct DIF code for endianness issues
  - Correct where we had missed points to check txq on i/o
    completion/cleanup

Signed-off-by: Alex Iannicelli <alex.iannicelli@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
diff --git a/drivers/scsi/lpfc/lpfc_els.c b/drivers/scsi/lpfc/lpfc_els.c
index 017c933..f801562 100644
--- a/drivers/scsi/lpfc/lpfc_els.c
+++ b/drivers/scsi/lpfc/lpfc_els.c
@@ -6962,6 +6962,7 @@
 	uint16_t xri = bf_get(lpfc_wcqe_xa_xri, axri);
 	struct lpfc_sglq *sglq_entry = NULL, *sglq_next = NULL;
 	unsigned long iflag = 0;
+	struct lpfc_sli_ring *pring = &phba->sli.ring[LPFC_ELS_RING];
 
 	spin_lock_irqsave(&phba->hbalock, iflag);
 	spin_lock(&phba->sli4_hba.abts_sgl_list_lock);
@@ -6974,6 +6975,10 @@
 			sglq_entry->state = SGL_FREED;
 			spin_unlock(&phba->sli4_hba.abts_sgl_list_lock);
 			spin_unlock_irqrestore(&phba->hbalock, iflag);
+
+			/* Check if TXQ queue needs to be serviced */
+			if (pring->txq_cnt)
+				lpfc_worker_wake_up(phba);
 			return;
 		}
 	}