[SCSI] lpfc 8.3.38: Fixed NMI watch dog panic's when resetting the hba.

Fixed NMI watch dog panic's when resetting the hba.

Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

8 files changed, 70 insertions, 43 deletions

diff --git a/drivers/scsi/lpfc/lpfc_attr.c b/drivers/scsi/lpfc/lpfc_attr.c
index a364cae9e98..97921dcd862 100644
--- a/drivers/scsi/lpfc/lpfc_attr.c
+++ b/drivers/scsi/lpfc/lpfc_attr.c
@@ -692,7 +692,7 @@ lpfc_do_offline(struct lpfc_hba *phba, uint32_t type)
 	 */
 	for (i = 0; i < psli->num_rings; i++) {
 		pring = &psli->ring[i];
-		while (pring->txcmplq_cnt) {
+		while (!list_empty(&pring->txcmplq)) {
 			msleep(10);
 			if (cnt++ > 500) {  /* 5 secs */
 				lpfc_printf_log(phba,
diff --git a/drivers/scsi/lpfc/lpfc_bsg.c b/drivers/scsi/lpfc/lpfc_bsg.c
index 32d5683e618..18d3b69963a 100644
--- a/drivers/scsi/lpfc/lpfc_bsg.c
+++ b/drivers/scsi/lpfc/lpfc_bsg.c
@@ -1559,7 +1559,7 @@ lpfc_bsg_diag_mode_enter(struct lpfc_hba *phba)
 		scsi_block_requests(shost);
 	}
 
-	while (pring->txcmplq_cnt) {
+	while (!list_empty(&pring->txcmplq)) {
 		if (i++ > 500)  /* wait up to 5 seconds */
 			break;
 		msleep(10);
diff --git a/drivers/scsi/lpfc/lpfc_els.c b/drivers/scsi/lpfc/lpfc_els.c
index 08d156a9094..b7a61ff43ca 100644
--- a/drivers/scsi/lpfc/lpfc_els.c
+++ b/drivers/scsi/lpfc/lpfc_els.c
@@ -6225,7 +6225,7 @@ lpfc_els_timeout_handler(struct lpfc_vport *vport)
 		spin_unlock_irq(&phba->hbalock);
 	}
 
-	if (phba->sli.ring[LPFC_ELS_RING].txcmplq_cnt)
+	if (!list_empty(&phba->sli.ring[LPFC_ELS_RING].txcmplq))
 		mod_timer(&vport->els_tmofunc, jiffies + HZ * timeout);
 }
 
@@ -6279,7 +6279,6 @@ lpfc_els_flush_cmd(struct lpfc_vport *vport)
 			continue;
 
 		list_move_tail(&piocb->list, &completions);
-		pring->txq_cnt--;
 	}
 
 	list_for_each_entry_safe(piocb, tmp_iocb, &pring->txcmplq, list) {
@@ -6339,7 +6338,6 @@ lpfc_els_flush_all_cmd(struct lpfc_hba  *phba)
 		    cmd->ulpCommand == CMD_ABORT_XRI_CN)
 			continue;
 		list_move_tail(&piocb->list, &completions);
-		pring->txq_cnt--;
 	}
 	list_for_each_entry_safe(piocb, tmp_iocb, &pring->txcmplq, list) {
 		if (piocb->iocb_flag & LPFC_IO_LIBDFC)
@@ -8065,7 +8063,7 @@ lpfc_sli4_els_xri_aborted(struct lpfc_hba *phba,
 				rxid, 1);
 
 			/* Check if TXQ queue needs to be serviced */
-			if (pring->txq_cnt)
+			if (!(list_empty(&pring->txq)))
 				lpfc_worker_wake_up(phba);
 			return;
 		}
diff --git a/drivers/scsi/lpfc/lpfc_hbadisc.c b/drivers/scsi/lpfc/lpfc_hbadisc.c
index d7096ad94d3..915e1aa6c68 100644
--- a/drivers/scsi/lpfc/lpfc_hbadisc.c
+++ b/drivers/scsi/lpfc/lpfc_hbadisc.c
@@ -696,7 +696,8 @@ lpfc_work_done(struct lpfc_hba *phba)
 							(status &
 							 HA_RXMASK));
 		}
-		if ((phba->sli_rev == LPFC_SLI_REV4) && pring->txq_cnt)
+		if ((phba->sli_rev == LPFC_SLI_REV4) &
+				 (!list_empty(&pring->txq)))
 			lpfc_drain_txq(phba);
 		/*
 		 * Turn on Ring interrupts
@@ -4354,7 +4355,6 @@ lpfc_no_rpi(struct lpfc_hba *phba, struct lpfc_nodelist *ndlp)
 					   with an error */
 					list_move_tail(&iocb->list,
 						       &completions);
-					pring->txq_cnt--;
 				}
 			}
 			spin_unlock_irq(&phba->hbalock);
@@ -5055,7 +5055,6 @@ lpfc_free_tx(struct lpfc_hba *phba, struct lpfc_nodelist *ndlp)
 		    (icmd->ulpCommand == CMD_XMIT_ELS_RSP64_CX)) {
 
 			list_move_tail(&iocb->list, &completions);
-			pring->txq_cnt--;
 		}
 	}
 
diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c
index 314b4f61b9e..b28f4e7c3bc 100644
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/drivers/scsi/lpfc/lpfc_init.c
@@ -839,7 +839,6 @@ lpfc_hba_down_post_s3(struct lpfc_hba *phba)
 		 * way, nothing should be on txcmplq as it will NEVER complete.
 		 */
 		list_splice_init(&pring->txcmplq, &completions);
-		pring->txcmplq_cnt = 0;
 		spin_unlock_irq(&phba->hbalock);
 
 		/* Cancel all the IOCBs from the completions list */
diff --git a/drivers/scsi/lpfc/lpfc_nportdisc.c b/drivers/scsi/lpfc/lpfc_nportdisc.c
index 46128c67920..82f4d354228 100644
--- a/drivers/scsi/lpfc/lpfc_nportdisc.c
+++ b/drivers/scsi/lpfc/lpfc_nportdisc.c
@@ -226,7 +226,6 @@ lpfc_els_abort(struct lpfc_hba *phba, struct lpfc_nodelist *ndlp)
 		if (lpfc_check_sli_ndlp(phba, pring, iocb, ndlp)) {
 			/* It matches, so deque and call compl with anp error */
 			list_move_tail(&iocb->list, &completions);
-			pring->txq_cnt--;
 		}
 	}
 
diff --git a/drivers/scsi/lpfc/lpfc_scsi.c b/drivers/scsi/lpfc/lpfc_scsi.c
index 98af07c6e30..e75210d6797 100644
--- a/drivers/scsi/lpfc/lpfc_scsi.c
+++ b/drivers/scsi/lpfc/lpfc_scsi.c
@@ -732,7 +732,7 @@ lpfc_sli4_fcp_xri_aborted(struct lpfc_hba *phba,
 		psb = container_of(iocbq, struct lpfc_scsi_buf, cur_iocbq);
 		psb->exch_busy = 0;
 		spin_unlock_irqrestore(&phba->hbalock, iflag);
-		if (pring->txq_cnt)
+		if (!list_empty(&pring->txq))
 			lpfc_worker_wake_up(phba);
 		return;
 
@@ -4246,7 +4246,7 @@ static __inline__ void lpfc_poll_rearm_timer(struct lpfc_hba * phba)
 	unsigned long  poll_tmo_expires =
 		(jiffies + msecs_to_jiffies(phba->cfg_poll_tmo));
 
-	if (phba->sli.ring[LPFC_FCP_RING].txcmplq_cnt)
+	if (!list_empty(&phba->sli.ring[LPFC_FCP_RING].txcmplq))
 		mod_timer(&phba->fcp_poll_timer,
 			  poll_tmo_expires);
 }
diff --git a/drivers/scsi/lpfc/lpfc_sli.c b/drivers/scsi/lpfc/lpfc_sli.c
index 74b67d98e95..a0ef439058f 100644
--- a/drivers/scsi/lpfc/lpfc_sli.c
+++ b/drivers/scsi/lpfc/lpfc_sli.c
@@ -1008,6 +1008,18 @@ __lpfc_sli_release_iocbq_s4(struct lpfc_hba *phba, struct lpfc_iocbq *iocbq)
 	else
 		sglq = __lpfc_clear_active_sglq(phba, iocbq->sli4_lxritag);
 
+	/*
+	** This should have been removed from the txcmplq before calling
+	** iocbq_release. The normal completion
+	** path should have already done the list_del_init.
+	*/
+	if (unlikely(!list_empty(&iocbq->list))) {
+		if (iocbq->iocb_flag & LPFC_IO_ON_TXCMPLQ)
+			iocbq->iocb_flag &= ~LPFC_IO_ON_TXCMPLQ;
+		list_del_init(&iocbq->list);
+	}
+
+
 	if (sglq)  {
 		if ((iocbq->iocb_flag & LPFC_EXCHANGE_BUSY) &&
 			(sglq->state != SGL_XRI_ABORTED)) {
@@ -1024,7 +1036,7 @@ __lpfc_sli_release_iocbq_s4(struct lpfc_hba *phba, struct lpfc_iocbq *iocbq)
 				&phba->sli4_hba.lpfc_sgl_list);
 
 			/* Check if TXQ queue needs to be serviced */
-			if (pring->txq_cnt)
+			if (!list_empty(&pring->txq))
 				lpfc_worker_wake_up(phba);
 		}
 	}
@@ -1056,6 +1068,14 @@ __lpfc_sli_release_iocbq_s3(struct lpfc_hba *phba, struct lpfc_iocbq *iocbq)
 	size_t start_clean = offsetof(struct lpfc_iocbq, iocb);
 
 	/*
+	** This should have been removed from the txcmplq before calling
+	** iocbq_release. The normal completion
+	** path should have already done the list_del_init.
+	*/
+	if (unlikely(!list_empty(&iocbq->list)))
+		list_del_init(&iocbq->list);
+
+	/*
 	 * Clean all volatile data fields, preserve iotag and node struct.
 	 */
 	memset((char*)iocbq + start_clean, 0, sizeof(*iocbq) - start_clean);
@@ -1121,7 +1141,6 @@ lpfc_sli_cancel_iocbs(struct lpfc_hba *phba, struct list_head *iocblist,
 
 	while (!list_empty(iocblist)) {
 		list_remove_head(iocblist, piocb, struct lpfc_iocbq, list);
-
 		if (!piocb->iocb_cmpl)
 			lpfc_sli_release_iocbq(phba, piocb);
 		else {
@@ -1309,9 +1328,6 @@ lpfc_sli_ringtxcmpl_put(struct lpfc_hba *phba, struct lpfc_sli_ring *pring,
 {
 	list_add_tail(&piocb->list, &pring->txcmplq);
 	piocb->iocb_flag |= LPFC_IO_ON_TXCMPLQ;
-	pring->txcmplq_cnt++;
-	if (pring->txcmplq_cnt > pring->txcmplq_max)
-		pring->txcmplq_max = pring->txcmplq_cnt;
 
 	if ((unlikely(pring->ringno == LPFC_ELS_RING)) &&
 	   (piocb->iocb.ulpCommand != CMD_ABORT_XRI_CN) &&
@@ -1343,8 +1359,6 @@ lpfc_sli_ringtx_get(struct lpfc_hba *phba, struct lpfc_sli_ring *pring)
 	struct lpfc_iocbq *cmd_iocb;
 
 	list_remove_head((&pring->txq), cmd_iocb, struct lpfc_iocbq, list);
-	if (cmd_iocb != NULL)
-		pring->txq_cnt--;
 	return cmd_iocb;
 }
 
@@ -1613,8 +1627,9 @@ lpfc_sli_resume_iocb(struct lpfc_hba *phba, struct lpfc_sli_ring *pring)
 	 *  (c) link attention events can be processed (fcp ring only)
 	 *  (d) IOCB processing is not blocked by the outstanding mbox command.
 	 */
-	if (pring->txq_cnt &&
-	    lpfc_is_link_up(phba) &&
+
+	if (lpfc_is_link_up(phba) &&
+	    (!list_empty(&pring->txq)) &&
 	    (pring->ringno != phba->sli.fcp_ring ||
 	     phba->sli.sli_flag & LPFC_PROCESS_LA)) {
 
@@ -2611,7 +2626,6 @@ lpfc_sli_iocbq_lookup(struct lpfc_hba *phba,
 		cmd_iocb = phba->sli.iocbq_lookup[iotag];
 		list_del_init(&cmd_iocb->list);
 		if (cmd_iocb->iocb_flag & LPFC_IO_ON_TXCMPLQ) {
-			pring->txcmplq_cnt--;
 			cmd_iocb->iocb_flag &= ~LPFC_IO_ON_TXCMPLQ;
 		}
 		return cmd_iocb;
@@ -2649,7 +2663,6 @@ lpfc_sli_iocbq_lookup_by_tag(struct lpfc_hba *phba,
 			/* remove from txcmpl queue list */
 			list_del_init(&cmd_iocb->list);
 			cmd_iocb->iocb_flag &= ~LPFC_IO_ON_TXCMPLQ;
-			pring->txcmplq_cnt--;
 			return cmd_iocb;
 		}
 	}
@@ -3498,7 +3511,6 @@ lpfc_sli_abort_iocb_ring(struct lpfc_hba *phba, struct lpfc_sli_ring *pring)
 	 */
 	spin_lock_irq(&phba->hbalock);
 	list_splice_init(&pring->txq, &completions);
-	pring->txq_cnt = 0;
 
 	/* Next issue ABTS for everything on the txcmplq */
 	list_for_each_entry_safe(iocb, next_iocb, &pring->txcmplq, list)
@@ -3535,11 +3547,9 @@ lpfc_sli_flush_fcp_rings(struct lpfc_hba *phba)
 	spin_lock_irq(&phba->hbalock);
 	/* Retrieve everything on txq */
 	list_splice_init(&pring->txq, &txq);
-	pring->txq_cnt = 0;
 
 	/* Retrieve everything on the txcmplq */
 	list_splice_init(&pring->txcmplq, &txcmplq);
-	pring->txcmplq_cnt = 0;
 
 	/* Indicate the I/O queues are flushed */
 	phba->hba_flag |= HBA_FCP_IOQ_FLUSH;
@@ -7614,7 +7624,6 @@ __lpfc_sli_ringtx_put(struct lpfc_hba *phba, struct lpfc_sli_ring *pring,
 {
 	/* Insert the caller's iocb in the txq tail for later processing. */
 	list_add_tail(&piocb->list, &pring->txq);
-	pring->txq_cnt++;
 }
 
 /**
@@ -8386,7 +8395,7 @@ __lpfc_sli_issue_iocb_s4(struct lpfc_hba *phba, uint32_t ring_number,
 		    piocb->iocb.ulpCommand == CMD_CLOSE_XRI_CN)
 			sglq = NULL;
 		else {
-			if (pring->txq_cnt) {
+			if (!list_empty(&pring->txq)) {
 				if (!(flag & SLI_IOCB_RET_IOCB)) {
 					__lpfc_sli_ringtx_put(phba,
 						pring, piocb);
@@ -9054,7 +9063,6 @@ lpfc_sli_host_down(struct lpfc_vport *vport)
 			if (iocb->vport != vport)
 				continue;
 			list_move_tail(&iocb->list, &completions);
-			pring->txq_cnt--;
 		}
 
 		/* Next issue ABTS for everything on the txcmplq */
@@ -9123,8 +9131,6 @@ lpfc_sli_hba_down(struct lpfc_hba *phba)
 		 * given to the FW yet.
 		 */
 		list_splice_init(&pring->txq, &completions);
-		pring->txq_cnt = 0;
-
 	}
 	spin_unlock_irqrestore(&phba->hbalock, flags);
 
@@ -9965,6 +9971,9 @@ lpfc_sli_issue_iocb_wait(struct lpfc_hba *phba,
 	long timeleft, timeout_req = 0;
 	int retval = IOCB_SUCCESS;
 	uint32_t creg_val;
+	struct lpfc_iocbq *iocb;
+	int txq_cnt = 0;
+	int txcmplq_cnt = 0;
 	struct lpfc_sli_ring *pring = &phba->sli.ring[LPFC_ELS_RING];
 	/*
 	 * If the caller has provided a response iocbq buffer, then context2
@@ -10012,9 +10021,17 @@ lpfc_sli_issue_iocb_wait(struct lpfc_hba *phba,
 			retval = IOCB_TIMEDOUT;
 		}
 	} else if (retval == IOCB_BUSY) {
-		lpfc_printf_log(phba, KERN_INFO, LOG_SLI,
-			"2818 Max IOCBs %d txq cnt %d txcmplq cnt %d\n",
-			phba->iocb_cnt, pring->txq_cnt, pring->txcmplq_cnt);
+		if (phba->cfg_log_verbose & LOG_SLI) {
+			list_for_each_entry(iocb, &pring->txq, list) {
+				txq_cnt++;
+			}
+			list_for_each_entry(iocb, &pring->txcmplq, list) {
+				txcmplq_cnt++;
+			}
+			lpfc_printf_log(phba, KERN_INFO, LOG_SLI,
+				"2818 Max IOCBs %d txq cnt %d txcmplq cnt %d\n",
+				phba->iocb_cnt, txq_cnt, txcmplq_cnt);
+		}
 		return retval;
 	} else {
 		lpfc_printf_log(phba, KERN_INFO, LOG_SLI,
@@ -11297,16 +11314,25 @@ lpfc_sli4_sp_handle_els_wcqe(struct lpfc_hba *phba, struct lpfc_queue *cq,
 	struct lpfc_iocbq *irspiocbq;
 	unsigned long iflags;
 	struct lpfc_sli_ring *pring = cq->pring;
+	int txq_cnt = 0;
+	int txcmplq_cnt = 0;
+	int fcp_txcmplq_cnt = 0;
 
 	/* Get an irspiocbq for later ELS response processing use */
 	irspiocbq = lpfc_sli_get_iocbq(phba);
 	if (!irspiocbq) {
+		if (!list_empty(&pring->txq))
+			txq_cnt++;
+		if (!list_empty(&pring->txcmplq))
+			txcmplq_cnt++;
+		if (!list_empty(&phba->sli.ring[LPFC_FCP_RING].txcmplq))
+			fcp_txcmplq_cnt++;
 		lpfc_printf_log(phba, KERN_ERR, LOG_SLI,
 			"0387 NO IOCBQ data: txq_cnt=%d iocb_cnt=%d "
 			"fcp_txcmplq_cnt=%d, els_txcmplq_cnt=%d\n",
-			pring->txq_cnt, phba->iocb_cnt,
-			phba->sli.ring[LPFC_FCP_RING].txcmplq_cnt,
-			phba->sli.ring[LPFC_ELS_RING].txcmplq_cnt);
+			txq_cnt, phba->iocb_cnt,
+			fcp_txcmplq_cnt,
+			txcmplq_cnt);
 		return false;
 	}
 
@@ -16244,14 +16270,19 @@ lpfc_drain_txq(struct lpfc_hba *phba)
 	char *fail_msg = NULL;
 	struct lpfc_sglq *sglq;
 	union lpfc_wqe wqe;
+	int txq_cnt = 0;
 
 	spin_lock_irqsave(&phba->hbalock, iflags);
-	if (pring->txq_cnt > pring->txq_max)
-		pring->txq_max = pring->txq_cnt;
+	list_for_each_entry(piocbq, &pring->txq, list) {
+		txq_cnt++;
+	}
+
+	if (txq_cnt > pring->txq_max)
+		pring->txq_max = txq_cnt;
 
 	spin_unlock_irqrestore(&phba->hbalock, iflags);
 
-	while (pring->txq_cnt) {
+	while (!list_empty(&pring->txq)) {
 		spin_lock_irqsave(&phba->hbalock, iflags);
 
 		piocbq = lpfc_sli_ringtx_get(phba, pring);
@@ -16259,7 +16290,7 @@ lpfc_drain_txq(struct lpfc_hba *phba)
 			spin_unlock_irqrestore(&phba->hbalock, iflags);
 			lpfc_printf_log(phba, KERN_ERR, LOG_SLI,
 				"2823 txq empty and txq_cnt is %d\n ",
-				pring->txq_cnt);
+				txq_cnt);
 			break;
 		}
 		sglq = __lpfc_sli_get_sglq(phba, piocbq);
@@ -16268,6 +16299,7 @@ lpfc_drain_txq(struct lpfc_hba *phba)
 			spin_unlock_irqrestore(&phba->hbalock, iflags);
 			break;
 		}
+		txq_cnt--;
 
 		/* The xri and iocb resources secured,
 		 * attempt to issue request
@@ -16299,5 +16331,5 @@ lpfc_drain_txq(struct lpfc_hba *phba)
 	lpfc_sli_cancel_iocbs(phba, &completions, IOSTAT_LOCAL_REJECT,
 				IOERR_SLI_ABORTED);
 
-	return pring->txq_cnt;
+	return txq_cnt;
 }