[SCSI] lpfc 8.3.22: Add support for PCI Adapter Failure

Periodically poll adapter registers to detect pci adapter failure
(reads return -1). On failure, take port offline, set error indicators
and wake up worker threads. Threads will take adapter offline.
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>

drivers/scsi/lpfc/lpfc.h

@@ -897,7 +897,18 @@ lpfc_worker_wake_up(struct lpfc_hba *phba)
 	return;
 }
 
-static inline void
+static inline int
+lpfc_readl(void __iomem *addr, uint32_t *data)
+{
+	uint32_t temp;
+	temp = readl(addr);
+	if (temp == 0xffffffff)
+		return -EIO;
+	*data = temp;
+	return 0;
+}
+
+static inline int
 lpfc_sli_read_hs(struct lpfc_hba *phba)
 {
 	/*
@@ -906,15 +917,17 @@ lpfc_sli_read_hs(struct lpfc_hba *phba)
 	 */
 	phba->sli.slistat.err_attn_event++;
 
-	/* Save status info */
-	phba->work_hs = readl(phba->HSregaddr);
-	phba->work_status[0] = readl(phba->MBslimaddr + 0xa8);
-	phba->work_status[1] = readl(phba->MBslimaddr + 0xac);
+	/* Save status info and check for unplug error */
+	if (lpfc_readl(phba->HSregaddr, &phba->work_hs) ||
+		lpfc_readl(phba->MBslimaddr + 0xa8, &phba->work_status[0]) ||
+		lpfc_readl(phba->MBslimaddr + 0xac, &phba->work_status[1])) {
+		return -EIO;
+	}
 
 	/* Clear chip Host Attention error bit */
 	writel(HA_ERATT, phba->HAregaddr);
 	readl(phba->HAregaddr); /* flush */
 	phba->pport->stopped = 1;
 
-	return;
+	return 0;
 }

drivers/scsi/lpfc/lpfc_attr.c

@@ -1224,7 +1224,10 @@ lpfc_poll_store(struct device *dev, struct device_attribute *attr,
 	if (val & ENABLE_FCP_RING_POLLING) {
 		if ((val & DISABLE_FCP_RING_INT) &&
 		    !(old_val & DISABLE_FCP_RING_INT)) {
-			creg_val = readl(phba->HCregaddr);
+			if (lpfc_readl(phba->HCregaddr, &creg_val)) {
+				spin_unlock_irq(&phba->hbalock);
+				return -EINVAL;
+			}
 			creg_val &= ~(HC_R0INT_ENA << LPFC_FCP_RING);
 			writel(creg_val, phba->HCregaddr);
 			readl(phba->HCregaddr); /* flush */
@@ -1242,7 +1245,10 @@ lpfc_poll_store(struct device *dev, struct device_attribute *attr,
 		spin_unlock_irq(&phba->hbalock);
 		del_timer(&phba->fcp_poll_timer);
 		spin_lock_irq(&phba->hbalock);
-		creg_val = readl(phba->HCregaddr);
+		if (lpfc_readl(phba->HCregaddr, &creg_val)) {
+			spin_unlock_irq(&phba->hbalock);
+			return -EINVAL;
+		}
 		creg_val |= (HC_R0INT_ENA << LPFC_FCP_RING);
 		writel(creg_val, phba->HCregaddr);
 		readl(phba->HCregaddr); /* flush */

drivers/scsi/lpfc/lpfc_bsg.c

@@ -348,7 +348,10 @@ lpfc_bsg_send_mgmt_cmd(struct fc_bsg_job *job)
 	dd_data->context_un.iocb.bmp = bmp;
 
 	if (phba->cfg_poll & DISABLE_FCP_RING_INT) {
-		creg_val = readl(phba->HCregaddr);
+		if (lpfc_readl(phba->HCregaddr, &creg_val)) {
+			rc = -EIO ;
+			goto free_cmdiocbq;
+		}
 		creg_val |= (HC_R0INT_ENA << LPFC_FCP_RING);
 		writel(creg_val, phba->HCregaddr);
 		readl(phba->HCregaddr); /* flush */
@@ -599,7 +602,10 @@ lpfc_bsg_rport_els(struct fc_bsg_job *job)
 	dd_data->context_un.iocb.ndlp = ndlp;
 
 	if (phba->cfg_poll & DISABLE_FCP_RING_INT) {
-		creg_val = readl(phba->HCregaddr);
+		if (lpfc_readl(phba->HCregaddr, &creg_val)) {
+			rc = -EIO;
+			goto linkdown_err;
+		}
 		creg_val |= (HC_R0INT_ENA << LPFC_FCP_RING);
 		writel(creg_val, phba->HCregaddr);
 		readl(phba->HCregaddr); /* flush */
@@ -613,6 +619,7 @@ lpfc_bsg_rport_els(struct fc_bsg_job *job)
 	else
 		rc = -EIO;
 
+linkdown_err:
 	pci_unmap_sg(phba->pcidev, job->request_payload.sg_list,
 		     job->request_payload.sg_cnt, DMA_TO_DEVICE);
 	pci_unmap_sg(phba->pcidev, job->reply_payload.sg_list,
@@ -1357,7 +1364,10 @@ lpfc_issue_ct_rsp(struct lpfc_hba *phba, struct fc_bsg_job *job, uint32_t tag,
 	dd_data->context_un.iocb.ndlp = ndlp;
 
 	if (phba->cfg_poll & DISABLE_FCP_RING_INT) {
-		creg_val = readl(phba->HCregaddr);
+		if (lpfc_readl(phba->HCregaddr, &creg_val)) {
+			rc = -IOCB_ERROR;
+			goto issue_ct_rsp_exit;
+		}
 		creg_val |= (HC_R0INT_ENA << LPFC_FCP_RING);
 		writel(creg_val, phba->HCregaddr);
 		readl(phba->HCregaddr); /* flush */

drivers/scsi/lpfc/lpfc_els.c

@@ -89,7 +89,8 @@ lpfc_els_chk_latt(struct lpfc_vport *vport)
 		return 0;
 
 	/* Read the HBA Host Attention Register */
-	ha_copy = readl(phba->HAregaddr);
+	if (lpfc_readl(phba->HAregaddr, &ha_copy))
+		return 1;
 
 	if (!(ha_copy & HA_LATT))
 		return 0;

drivers/scsi/lpfc/lpfc_hw.h

@@ -1344,7 +1344,7 @@ typedef struct {		/* FireFly BIU registers */
 #define HS_FFER1       0x80000000	/* Bit 31 */
 #define HS_CRIT_TEMP   0x00000100	/* Bit 8  */
 #define HS_FFERM       0xFF000100	/* Mask for error bits 31:24 and 8 */
-
+#define UNPLUG_ERR     0x00000001	/* Indicate pci hot unplug */
 /* Host Control Register */
 
 #define HC_REG_OFFSET  12	/* Byte offset from register base address */

drivers/scsi/lpfc/lpfc_init.c

@@ -507,7 +507,10 @@ lpfc_config_port_post(struct lpfc_hba *phba)
 	phba->hba_flag &= ~HBA_ERATT_HANDLED;
 
 	/* Enable appropriate host interrupts */
-	status = readl(phba->HCregaddr);
+	if (lpfc_readl(phba->HCregaddr, &status)) {
+		spin_unlock_irq(&phba->hbalock);
+		return -EIO;
+	}
 	status |= HC_MBINT_ENA | HC_ERINT_ENA | HC_LAINT_ENA;
 	if (psli->num_rings > 0)
 		status |= HC_R0INT_ENA;
@@ -1222,7 +1225,10 @@ lpfc_handle_deferred_eratt(struct lpfc_hba *phba)
 	/* Wait for the ER1 bit to clear.*/
 	while (phba->work_hs & HS_FFER1) {
 		msleep(100);
-		phba->work_hs = readl(phba->HSregaddr);
+		if (lpfc_readl(phba->HSregaddr, &phba->work_hs)) {
+			phba->work_hs = UNPLUG_ERR ;
+			break;
+		}
 		/* If driver is unloading let the worker thread continue */
 		if (phba->pport->load_flag & FC_UNLOADING) {
 			phba->work_hs = 0;
@@ -5386,13 +5392,16 @@ lpfc_sli4_post_status_check(struct lpfc_hba *phba)
 	int i, port_error = 0;
 	uint32_t if_type;
 
+	memset(&portsmphr_reg, 0, sizeof(portsmphr_reg));
+	memset(&reg_data, 0, sizeof(reg_data));
 	if (!phba->sli4_hba.PSMPHRregaddr)
 		return -ENODEV;
 
 	/* Wait up to 30 seconds for the SLI Port POST done and ready */
 	for (i = 0; i < 3000; i++) {
-		portsmphr_reg.word0 = readl(phba->sli4_hba.PSMPHRregaddr);
-		if (bf_get(lpfc_port_smphr_perr, &portsmphr_reg)) {
+		if (lpfc_readl(phba->sli4_hba.PSMPHRregaddr,
+			&portsmphr_reg.word0) ||
+			(bf_get(lpfc_port_smphr_perr, &portsmphr_reg))) {
 			/* Port has a fatal POST error, break out */
 			port_error = -ENODEV;
 			break;
@@ -5473,9 +5482,9 @@ lpfc_sli4_post_status_check(struct lpfc_hba *phba)
 			break;
 		case LPFC_SLI_INTF_IF_TYPE_2:
 			/* Final checks.  The port status should be clean. */
-			reg_data.word0 =
-				readl(phba->sli4_hba.u.if_type2.STATUSregaddr);
-			if (bf_get(lpfc_sliport_status_err, &reg_data)) {
+			if (lpfc_readl(phba->sli4_hba.u.if_type2.STATUSregaddr,
+				&reg_data.word0) ||
+				bf_get(lpfc_sliport_status_err, &reg_data)) {
 				phba->work_status[0] =
 					readl(phba->sli4_hba.u.if_type2.
 					      ERR1regaddr);
@@ -6761,9 +6770,11 @@ lpfc_pci_function_reset(struct lpfc_hba *phba)
 			 * the loop again.
 			 */
 			for (rdy_chk = 0; rdy_chk < 1000; rdy_chk++) {
-				reg_data.word0 =
-					readl(phba->sli4_hba.u.if_type2.
-					      STATUSregaddr);
+				if (lpfc_readl(phba->sli4_hba.u.if_type2.
+					      STATUSregaddr, &reg_data.word0)) {
+					rc = -ENODEV;
+					break;
+				}
 				if (bf_get(lpfc_sliport_status_rdy, &reg_data))
 					break;
 				if (bf_get(lpfc_sliport_status_rn, &reg_data)) {
@@ -6784,8 +6795,11 @@ lpfc_pci_function_reset(struct lpfc_hba *phba)
 			}
 
 			/* Detect any port errors. */
-			reg_data.word0 = readl(phba->sli4_hba.u.if_type2.
-					       STATUSregaddr);
+			if (lpfc_readl(phba->sli4_hba.u.if_type2.STATUSregaddr,
+				 &reg_data.word0)) {
+				rc = -ENODEV;
+				break;
+			}
 			if ((bf_get(lpfc_sliport_status_err, &reg_data)) ||
 			    (rdy_chk >= 1000)) {
 				phba->work_status[0] = readl(