[SCSI] lpfc 8.1.9 : Misc Bug Fixes

Misc Bug Fixes:
- Cap MBX_DOWN_LINK command timeout to 60 seconds
- Fix double free of ndlp object
- Don't free mbox structures on error. The completion handlers expect to do so.
- Clear host attention work items when going offline
- Fixed discovery issues in multi-initiator environments.

Signed-off-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

5 files changed, 38 insertions, 5 deletions

diff --git a/drivers/scsi/lpfc/lpfc_attr.c b/drivers/scsi/lpfc/lpfc_attr.c
index 76f8bd5..d384c16 100644
--- a/drivers/scsi/lpfc/lpfc_attr.c
+++ b/drivers/scsi/lpfc/lpfc_attr.c
@@ -222,7 +222,7 @@ lpfc_issue_lip(struct Scsi_Host *host)
 	pmboxq->mb.mbxCommand = MBX_DOWN_LINK;
 	pmboxq->mb.mbxOwner = OWN_HOST;
 
-	mbxstatus = lpfc_sli_issue_mbox_wait(phba, pmboxq, phba->fc_ratov * 2);
+	mbxstatus = lpfc_sli_issue_mbox_wait(phba, pmboxq, LPFC_MBOX_TMO * 2);
 
 	if ((mbxstatus == MBX_SUCCESS) && (pmboxq->mb.mbxStatus == 0)) {
 		memset((void *)pmboxq, 0, sizeof (LPFC_MBOXQ_t));
diff --git a/drivers/scsi/lpfc/lpfc_els.c b/drivers/scsi/lpfc/lpfc_els.c
index 60f5cca..3567de61 100644
--- a/drivers/scsi/lpfc/lpfc_els.c
+++ b/drivers/scsi/lpfc/lpfc_els.c
@@ -1848,9 +1848,12 @@ static void
 lpfc_cmpl_els_acc(struct lpfc_hba * phba, struct lpfc_iocbq * cmdiocb,
 		  struct lpfc_iocbq * rspiocb)
 {
+	IOCB_t *irsp;
 	struct lpfc_nodelist *ndlp;
 	LPFC_MBOXQ_t *mbox = NULL;
 
+	irsp = &rspiocb->iocb;
+
 	ndlp = (struct lpfc_nodelist *) cmdiocb->context1;
 	if (cmdiocb->context_un.mbox)
 		mbox = cmdiocb->context_un.mbox;
@@ -1893,9 +1896,15 @@ lpfc_cmpl_els_acc(struct lpfc_hba * phba, struct lpfc_iocbq * cmdiocb,
 			mempool_free( mbox, phba->mbox_mem_pool);
 		} else {
 			mempool_free( mbox, phba->mbox_mem_pool);
-			if (ndlp->nlp_flag & NLP_ACC_REGLOGIN) {
-				lpfc_nlp_list(phba, ndlp, NLP_NO_LIST);
-				ndlp = NULL;
+			/* Do not call NO_LIST for lpfc_els_abort'ed ELS cmds */
+			if (!((irsp->ulpStatus == IOSTAT_LOCAL_REJECT) &&
+			      ((irsp->un.ulpWord[4] == IOERR_SLI_ABORTED) ||
+			       (irsp->un.ulpWord[4] == IOERR_LINK_DOWN) ||
+			       (irsp->un.ulpWord[4] == IOERR_SLI_DOWN)))) {
+				if (ndlp->nlp_flag & NLP_ACC_REGLOGIN) {
+					lpfc_nlp_list(phba, ndlp, NLP_NO_LIST);
+					ndlp = NULL;
+				}
 			}
 		}
 	}
diff --git a/drivers/scsi/lpfc/lpfc_hbadisc.c b/drivers/scsi/lpfc/lpfc_hbadisc.c
index 1c3f268..b2f1552 100644
--- a/drivers/scsi/lpfc/lpfc_hbadisc.c
+++ b/drivers/scsi/lpfc/lpfc_hbadisc.c
@@ -1557,6 +1557,8 @@ lpfc_freenode(struct lpfc_hba * phba, struct lpfc_nodelist * ndlp)
 			mb->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
 		}
 	}
+
+	spin_lock_irq(phba->host->host_lock);
 	list_for_each_entry_safe(mb, nextmb, &phba->sli.mboxq, list) {
 		if ((mb->mb.mbxCommand == MBX_REG_LOGIN64) &&
 		   (ndlp == (struct lpfc_nodelist *) mb->context2)) {
@@ -1569,6 +1571,7 @@ lpfc_freenode(struct lpfc_hba * phba, struct lpfc_nodelist * ndlp)
 			mempool_free(mb, phba->mbox_mem_pool);
 		}
 	}
+	spin_unlock_irq(phba->host->host_lock);
 
 	lpfc_els_abort(phba,ndlp,0);
 	spin_lock_irq(phba->host->host_lock);
diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c
index 16dc8c8..f6948ff 100644
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/drivers/scsi/lpfc/lpfc_init.c
@@ -1379,6 +1379,7 @@ lpfc_offline(struct lpfc_hba * phba)
 	/* stop all timers associated with this hba */
 	lpfc_stop_timer(phba);
 	phba->work_hba_events = 0;
+	phba->work_ha = 0;
 
 	lpfc_printf_log(phba,
 		       KERN_WARNING,
diff --git a/drivers/scsi/lpfc/lpfc_nportdisc.c b/drivers/scsi/lpfc/lpfc_nportdisc.c
index b38021a..20449a8 100644
--- a/drivers/scsi/lpfc/lpfc_nportdisc.c
+++ b/drivers/scsi/lpfc/lpfc_nportdisc.c
@@ -393,6 +393,20 @@ lpfc_rcv_plogi(struct lpfc_hba * phba,
 	mbox->context2  = ndlp;
 	ndlp->nlp_flag |= (NLP_ACC_REGLOGIN | NLP_RCV_PLOGI);
 
+	/*
+	 * If there is an outstanding PLOGI issued, abort it before
+	 * sending ACC rsp for received PLOGI. If pending plogi
+	 * is not canceled here, the plogi will be rejected by
+	 * remote port and will be retried. On a configuration with
+	 * single discovery thread, this will cause a huge delay in
+	 * discovery. Also this will cause multiple state machines
+	 * running in parallel for this node.
+	 */
+	if (ndlp->nlp_state == NLP_STE_PLOGI_ISSUE) {
+		/* software abort outstanding PLOGI */
+		lpfc_els_abort(phba, ndlp, 1);
+	}
+
 	lpfc_els_rsp_acc(phba, ELS_CMD_PLOGI, cmdiocb, ndlp, mbox, 0);
 	return 1;
 
@@ -1601,7 +1615,13 @@ lpfc_rcv_padisc_npr_node(struct lpfc_hba * phba,
 
 	lpfc_rcv_padisc(phba, ndlp, cmdiocb);
 
-	if (!(ndlp->nlp_flag & NLP_DELAY_TMO)) {
+	/*
+	 * Do not start discovery if discovery is about to start
+	 * or discovery in progress for this node. Starting discovery
+	 * here will affect the counting of discovery threads.
+	 */
+	if ((!(ndlp->nlp_flag & NLP_DELAY_TMO)) &&
+		(ndlp->nlp_flag & NLP_NPR_2B_DISC)){
 		if (ndlp->nlp_flag & NLP_NPR_ADISC) {
 			ndlp->nlp_prev_state = NLP_STE_NPR_NODE;
 			ndlp->nlp_state = NLP_STE_ADISC_ISSUE;