scsi: lpfc: Improve PCI EEH Error and Recovery Handling

Following EEH errors, the driver can crash or hang when deleting the
localport or when attempting to unload.

The EEH handlers in the driver did not notify the NVMe-FC transport before
tearing the driver down. This was delayed until the resume steps. This
worked for SCSI because lpfc_block_scsi() would notify the
scsi_fc_transport that the target was not available but it would not clean
up all the references to the ndlp.

The SLI3 prep for dev reset handler did the lpfc_offline_prep() and
lpfc_offline() calls to get the port stopped before restarting. The SLI4
version of the prep for dev reset just destroyed the queues and did not
stop NVMe from continuing.  Also because the port was not really stopped
the localport destroy would hang because the transport was still waiting
for I/O. Additionally, a devloss tmo can fire and post events to a stopped
worker thread creating another hang condition.

lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and
lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI
and NVMe transports are notified to block I/O to the driver.

Logic is added to devloss handler and worker thread to clean up ndlp
references and quiesce appropriately.

Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/lpfc/lpfc.h

@@ -896,6 +896,11 @@ enum lpfc_irq_chann_mode {
 	NHT_MODE,
 };
 
+enum lpfc_hba_bit_flags {
+	FABRIC_COMANDS_BLOCKED,
+	HBA_PCI_ERR,
+};
+
 struct lpfc_hba {
 	/* SCSI interface function jump table entries */
 	struct lpfc_io_buf * (*lpfc_get_scsi_buf)
@@ -1042,7 +1047,6 @@ struct lpfc_hba {
 					 * Firmware supports Forced Link Speed
 					 * capability
 					 */
-#define HBA_PCI_ERR		0x80000 /* The PCI slot is offline */
 #define HBA_FLOGI_ISSUED	0x100000 /* FLOGI was issued */
 #define HBA_SHORT_CMF		0x200000 /* shorter CMF timer routine */
 #define HBA_CGN_DAY_WRAP	0x400000 /* HBA Congestion info day wraps */
@@ -1349,7 +1353,6 @@ struct lpfc_hba {
 	atomic_t fabric_iocb_count;
 	struct timer_list fabric_block_timer;
 	unsigned long bit_flags;
-#define	FABRIC_COMANDS_BLOCKED	0
 	atomic_t num_rsrc_err;
 	atomic_t num_cmd_success;
 	unsigned long last_rsrc_error_time;

drivers/scsi/lpfc/lpfc_crtn.h

@@ -670,3 +670,6 @@ struct lpfc_vmid *lpfc_get_vmid_from_hashtable(struct lpfc_vport *vport,
 					      uint32_t hash, uint8_t *buf);
 void lpfc_vmid_vport_cleanup(struct lpfc_vport *vport);
 int lpfc_issue_els_qfpa(struct lpfc_vport *vport);
+
+void lpfc_sli_rpi_release(struct lpfc_vport *vport,
+			  struct lpfc_nodelist *ndlp);

drivers/scsi/lpfc/lpfc_hbadisc.c

@@ -109,8 +109,8 @@ lpfc_rport_invalid(struct fc_rport *rport)
 
 	ndlp = rdata->pnode;
 	if (!rdata->pnode) {
-		pr_err("**** %s: NULL ndlp on rport x%px SID x%x\n",
-		       __func__, rport, rport->scsi_target_id);
+		pr_info("**** %s: NULL ndlp on rport x%px SID x%x\n",
+			__func__, rport, rport->scsi_target_id);
 		return -EINVAL;
 	}
 
@@ -169,9 +169,10 @@ lpfc_dev_loss_tmo_callbk(struct fc_rport *rport)
 
 	lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
 			 "3181 dev_loss_callbk x%06x, rport x%px flg x%x "
-			 "load_flag x%x refcnt %d\n",
+			 "load_flag x%x refcnt %d state %d xpt x%x\n",
 			 ndlp->nlp_DID, ndlp->rport, ndlp->nlp_flag,
-			 vport->load_flag, kref_read(&ndlp->kref));
+			 vport->load_flag, kref_read(&ndlp->kref),
+			 ndlp->nlp_state, ndlp->fc4_xpt_flags);
 
 	/* Don't schedule a worker thread event if the vport is going down.
 	 * The teardown process cleans up the node via lpfc_drop_node.
@@ -181,6 +182,11 @@ lpfc_dev_loss_tmo_callbk(struct fc_rport *rport)
 		ndlp->rport = NULL;
 
 		ndlp->fc4_xpt_flags &= ~SCSI_XPT_REGD;
+		/* clear the NLP_XPT_REGD if the node is not registered
+		 * with nvme-fc
+		 */
+		if (ndlp->fc4_xpt_flags == NLP_XPT_REGD)
+			ndlp->fc4_xpt_flags &= ~NLP_XPT_REGD;
 
 		/* Remove the node reference from remote_port_add now.
 		 * The driver will not call remote_port_delete.
@@ -225,18 +231,36 @@ lpfc_dev_loss_tmo_callbk(struct fc_rport *rport)
 	ndlp->rport = NULL;
 	spin_unlock_irqrestore(&ndlp->lock, iflags);
 
-	/* We need to hold the node by incrementing the reference
-	 * count until this queued work is done
-	 */
-	evtp->evt_arg1 = lpfc_nlp_get(ndlp);
+	if (phba->worker_thread) {
+		/* We need to hold the node by incrementing the reference
+		 * count until this queued work is done
+		 */
+		evtp->evt_arg1 = lpfc_nlp_get(ndlp);
+
+		spin_lock_irqsave(&phba->hbalock, iflags);
+		if (evtp->evt_arg1) {
+			evtp->evt = LPFC_EVT_DEV_LOSS;
+			list_add_tail(&evtp->evt_listp, &phba->work_list);
+			lpfc_worker_wake_up(phba);
+		}
+		spin_unlock_irqrestore(&phba->hbalock, iflags);
+	} else {
+		lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
+				 "3188 worker thread is stopped %s x%06x, "
+				 " rport x%px flg x%x load_flag x%x refcnt "
+				 "%d\n", __func__, ndlp->nlp_DID,
+				 ndlp->rport, ndlp->nlp_flag,
+				 vport->load_flag, kref_read(&ndlp->kref));
+		if (!(ndlp->fc4_xpt_flags & NVME_XPT_REGD)) {
+			spin_lock_irqsave(&ndlp->lock, iflags);
+			/* Node is in dev loss.  No further transaction. */
+			ndlp->nlp_flag &= ~NLP_IN_DEV_LOSS;
+			spin_unlock_irqrestore(&ndlp->lock, iflags);
+			lpfc_disc_state_machine(vport, ndlp, NULL,
+						NLP_EVT_DEVICE_RM);
+		}
 
-	spin_lock_irqsave(&phba->hbalock, iflags);
-	if (evtp->evt_arg1) {
-		evtp->evt = LPFC_EVT_DEV_LOSS;
-		list_add_tail(&evtp->evt_listp, &phba->work_list);
-		lpfc_worker_wake_up(phba);
 	}
-	spin_unlock_irqrestore(&phba->hbalock, iflags);
 
 	return;
 }
@@ -503,11 +527,12 @@ lpfc_dev_loss_tmo_handler(struct lpfc_nodelist *ndlp)
 		lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
 				 "0203 Devloss timeout on "
 				 "WWPN %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x "
-				 "NPort x%06x Data: x%x x%x x%x\n",
+				 "NPort x%06x Data: x%x x%x x%x refcnt %d\n",
 				 *name, *(name+1), *(name+2), *(name+3),
 				 *(name+4), *(name+5), *(name+6), *(name+7),
 				 ndlp->nlp_DID, ndlp->nlp_flag,
-				 ndlp->nlp_state, ndlp->nlp_rpi);
+				 ndlp->nlp_state, ndlp->nlp_rpi,
+				 kref_read(&ndlp->kref));
 	} else {
 		lpfc_printf_vlog(vport, KERN_INFO, LOG_TRACE_EVENT,
 				 "0204 Devloss timeout on "
@@ -755,18 +780,22 @@ lpfc_work_list_done(struct lpfc_hba *phba)
 	int free_evt;
 	int fcf_inuse;
 	uint32_t nlp_did;
+	bool hba_pci_err;
 
 	spin_lock_irq(&phba->hbalock);
 	while (!list_empty(&phba->work_list)) {
 		list_remove_head((&phba->work_list), evtp, typeof(*evtp),
 				 evt_listp);
 		spin_unlock_irq(&phba->hbalock);
+		hba_pci_err = test_bit(HBA_PCI_ERR, &phba->bit_flags);
 		free_evt = 1;
 		switch (evtp->evt) {
 		case LPFC_EVT_ELS_RETRY:
 			ndlp = (struct lpfc_nodelist *) (evtp->evt_arg1);
-			lpfc_els_retry_delay_handler(ndlp);
-			free_evt = 0; /* evt is part of ndlp */
+			if (!hba_pci_err) {
+				lpfc_els_retry_delay_handler(ndlp);
+				free_evt = 0; /* evt is part of ndlp */
+			}
 			/* decrement the node reference count held
 			 * for this queued work
 			 */
@@ -788,8 +817,10 @@ lpfc_work_list_done(struct lpfc_hba *phba)
 			break;
 		case LPFC_EVT_RECOVER_PORT:
 			ndlp = (struct lpfc_nodelist *)(evtp->evt_arg1);
-			lpfc_sli_abts_recover_port(ndlp->vport, ndlp);
-			free_evt = 0;
+			if (!hba_pci_err) {
+				lpfc_sli_abts_recover_port(ndlp->vport, ndlp);
+				free_evt = 0;
+			}
 			/* decrement the node reference count held for
 			 * this queued work
 			 */
@@ -859,14 +890,18 @@ lpfc_work_done(struct lpfc_hba *phba)
 	struct lpfc_vport **vports;
 	struct lpfc_vport *vport;
 	int i;
+	bool hba_pci_err;
 
+	hba_pci_err = test_bit(HBA_PCI_ERR, &phba->bit_flags);
 	spin_lock_irq(&phba->hbalock);
 	ha_copy = phba->work_ha;
 	phba->work_ha = 0;
 	spin_unlock_irq(&phba->hbalock);
+	if (hba_pci_err)
+		ha_copy = 0;
 
 	/* First, try to post the next mailbox command to SLI4 device */
-	if (phba->pci_dev_grp == LPFC_PCI_DEV_OC)
+	if (phba->pci_dev_grp == LPFC_PCI_DEV_OC && !hba_pci_err)
 		lpfc_sli4_post_async_mbox(phba);
 
 	if (ha_copy & HA_ERATT) {
@@ -886,7 +921,7 @@ lpfc_work_done(struct lpfc_hba *phba)
 		lpfc_handle_latt(phba);
 
 	/* Handle VMID Events */
-	if (lpfc_is_vmid_enabled(phba)) {
+	if (lpfc_is_vmid_enabled(phba) && !hba_pci_err) {
 		if (phba->pport->work_port_events &
 		    WORKER_CHECK_VMID_ISSUE_QFPA) {
 			lpfc_check_vmid_qfpa_issue(phba);
@@ -936,6 +971,8 @@ lpfc_work_done(struct lpfc_hba *phba)
 			work_port_events = vport->work_port_events;
 			vport->work_port_events &= ~work_port_events;
 			spin_unlock_irq(&vport->work_port_lock);
+			if (hba_pci_err)
+				continue;
 			if (work_port_events & WORKER_DISC_TMO)
 				lpfc_disc_timeout_handler(vport);
 			if (work_port_events & WORKER_ELS_TMO)
@@ -1173,12 +1210,14 @@ lpfc_linkdown(struct lpfc_hba *phba)
 	struct lpfc_vport **vports;
 	LPFC_MBOXQ_t          *mb;
 	int i;
+	int offline;
 
 	if (phba->link_state == LPFC_LINK_DOWN)
 		return 0;
 
 	/* Block all SCSI stack I/Os */
 	lpfc_scsi_dev_block(phba);
+	offline = pci_channel_offline(phba->pcidev);
 
 	phba->defer_flogi_acc_flag = false;
 
@@ -1219,7 +1258,7 @@ lpfc_linkdown(struct lpfc_hba *phba)
 	lpfc_destroy_vport_work_array(phba, vports);
 
 	/* Clean up any SLI3 firmware default rpi's */
-	if (phba->sli_rev > LPFC_SLI_REV3)
+	if (phba->sli_rev > LPFC_SLI_REV3 || offline)
 		goto skip_unreg_did;
 
 	mb = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
@@ -4712,6 +4751,11 @@ lpfc_nlp_unreg_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
 	spin_lock_irqsave(&ndlp->lock, iflags);
 	if (!(ndlp->fc4_xpt_flags & NLP_XPT_REGD)) {
 		spin_unlock_irqrestore(&ndlp->lock, iflags);
+		lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
+				 "0999 %s Not regd: ndlp x%px rport x%px DID "
+				 "x%x FLG x%x XPT x%x\n",
+				  __func__, ndlp, ndlp->rport, ndlp->nlp_DID,
+				  ndlp->nlp_flag, ndlp->fc4_xpt_flags);
 		return;
 	}
 
@@ -4722,6 +4766,13 @@ lpfc_nlp_unreg_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
 	    ndlp->fc4_xpt_flags & SCSI_XPT_REGD) {
 		vport->phba->nport_event_cnt++;
 		lpfc_unregister_remote_port(ndlp);
+	} else if (!ndlp->rport) {
+		lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
+				 "1999 %s NDLP in devloss x%px DID x%x FLG x%x"
+				 " XPT x%x refcnt %d\n",
+				 __func__, ndlp, ndlp->nlp_DID, ndlp->nlp_flag,
+				 ndlp->fc4_xpt_flags,
+				 kref_read(&ndlp->kref));
 	}
 
 	if (ndlp->fc4_xpt_flags & NVME_XPT_REGD) {
@@ -6097,12 +6148,34 @@ lpfc_disc_flush_list(struct lpfc_vport *vport)
 	}
 }
 
+/*
+ * lpfc_notify_xport_npr - notifies xport of node disappearance
+ * @vport: Pointer to Virtual Port object.
+ *
+ * Transitions all ndlps to NPR state.  When lpfc_nlp_set_state
+ * calls lpfc_nlp_state_cleanup, the ndlp->rport is unregistered
+ * and transport notified that the node is gone.
+ * Return Code:
+ *	none
+ */
+static void
+lpfc_notify_xport_npr(struct lpfc_vport *vport)
+{
+	struct lpfc_nodelist *ndlp, *next_ndlp;
+
+	list_for_each_entry_safe(ndlp, next_ndlp, &vport->fc_nodes,
+				 nlp_listp) {
+		lpfc_nlp_set_state(vport, ndlp, NLP_STE_NPR_NODE);
+	}
+}
 void
 lpfc_cleanup_discovery_resources(struct lpfc_vport *vport)
 {
 	lpfc_els_flush_rscn(vport);
 	lpfc_els_flush_cmd(vport);
 	lpfc_disc_flush_list(vport);
+	if (pci_channel_offline(vport->phba->pcidev))
+		lpfc_notify_xport_npr(vport);
 }
 
 /*****************************************************************************/