[SCSI] be2iscsi: Fix AER handling in driver

Signed-off-by: Minh Tran <minhduc.tran@emulex.com>
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

4 files changed, 270 insertions, 29 deletions

diff --git a/drivers/scsi/be2iscsi/be_cmds.c b/drivers/scsi/be2iscsi/be_cmds.c
index fce298ba4b41..3338391b64de 100644
--- a/drivers/scsi/be2iscsi/be_cmds.c
+++ b/drivers/scsi/be2iscsi/be_cmds.c
@@ -377,7 +377,7 @@ void beiscsi_async_link_state_process(struct beiscsi_hba *phba,
 	} else if ((evt->port_link_status & ASYNC_EVENT_LINK_UP) ||
 		    ((evt->port_link_status & ASYNC_EVENT_LOGICAL) &&
 		     (evt->port_fault == BEISCSI_PHY_LINK_FAULT_NONE))) {
-		phba->state = BE_ADAPTER_UP;
+		phba->state = BE_ADAPTER_LINK_UP;
 
 		beiscsi_log(phba, KERN_ERR,
 			    BEISCSI_LOG_CONFIG | BEISCSI_LOG_INIT,
diff --git a/drivers/scsi/be2iscsi/be_iscsi.c b/drivers/scsi/be2iscsi/be_iscsi.c
index e82ab8124958..ffadbee0b4d9 100644
--- a/drivers/scsi/be2iscsi/be_iscsi.c
+++ b/drivers/scsi/be2iscsi/be_iscsi.c
@@ -58,10 +58,15 @@ struct iscsi_cls_session *beiscsi_session_create(struct iscsi_endpoint *ep,
 	}
 	beiscsi_ep = ep->dd_data;
 	phba = beiscsi_ep->phba;
-	shost = phba->shost;
 
-	beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
-		    "BS_%d : In beiscsi_session_create\n");
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : PCI_ERROR Recovery\n");
+		return NULL;
+	} else {
+		beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In beiscsi_session_create\n");
+	}
 
 	if (cmds_max > beiscsi_ep->phba->params.wrbs_per_cxn) {
 		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
@@ -74,6 +79,7 @@ struct iscsi_cls_session *beiscsi_session_create(struct iscsi_endpoint *ep,
 		cmds_max = beiscsi_ep->phba->params.wrbs_per_cxn;
 	}
 
+	shost = phba->shost;
 	cls_session = iscsi_session_setup(&beiscsi_iscsi_transport,
 					  shost, cmds_max,
 					  sizeof(*beiscsi_sess),
@@ -477,6 +483,12 @@ int be2iscsi_iface_set_param(struct Scsi_Host *shost,
 	uint32_t rm_len = dt_len;
 	int ret = 0 ;
 
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In PCI_ERROR Recovery\n");
+		return -EBUSY;
+	}
+
 	nla_for_each_attr(attrib, data, dt_len, rm_len) {
 		iface_param = nla_data(attrib);
 
@@ -588,6 +600,12 @@ int be2iscsi_iface_get_param(struct iscsi_iface *iface,
 	struct be_cmd_get_def_gateway_resp gateway;
 	int len = -ENOSYS;
 
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In PCI_ERROR Recovery\n");
+		return -EBUSY;
+	}
+
 	switch (param) {
 	case ISCSI_NET_PARAM_IPV4_ADDR:
 	case ISCSI_NET_PARAM_IPV4_SUBNET:
@@ -737,7 +755,7 @@ static void beiscsi_get_port_state(struct Scsi_Host *shost)
 	struct beiscsi_hba *phba = iscsi_host_priv(shost);
 	struct iscsi_cls_host *ihost = shost->shost_data;
 
-	ihost->port_state = (phba->state == BE_ADAPTER_UP) ?
+	ihost->port_state = (phba->state == BE_ADAPTER_LINK_UP) ?
 		ISCSI_PORT_STATE_UP : ISCSI_PORT_STATE_DOWN;
 }
 
@@ -805,9 +823,16 @@ int beiscsi_get_host_param(struct Scsi_Host *shost,
 	struct beiscsi_hba *phba = iscsi_host_priv(shost);
 	int status = 0;
 
-	beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
-		    "BS_%d : In beiscsi_get_host_param,"
-		    " param= %d\n", param);
+
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In PCI_ERROR Recovery\n");
+		return -EBUSY;
+	} else {
+		beiscsi_log(phba, KERN_INFO, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In beiscsi_get_host_param,"
+			    " param = %d\n", param);
+	}
 
 	switch (param) {
 	case ISCSI_HOST_PARAM_HWADDRESS:
@@ -950,10 +975,19 @@ int beiscsi_conn_start(struct iscsi_cls_conn *cls_conn)
 	struct beiscsi_conn *beiscsi_conn = conn->dd_data;
 	struct beiscsi_endpoint *beiscsi_ep;
 	struct beiscsi_offload_params params;
+	struct beiscsi_hba *phba;
 
-	beiscsi_log(beiscsi_conn->phba, KERN_INFO,
-		    BEISCSI_LOG_CONFIG,
-		    "BS_%d : In beiscsi_conn_start\n");
+	phba = ((struct beiscsi_conn *)conn->dd_data)->phba;
+
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In PCI_ERROR Recovery\n");
+		return -EBUSY;
+	} else {
+		beiscsi_log(beiscsi_conn->phba, KERN_INFO,
+			    BEISCSI_LOG_CONFIG,
+			    "BS_%d : In beiscsi_conn_start\n");
+	}
 
 	memset(&params, 0, sizeof(struct beiscsi_offload_params));
 	beiscsi_ep = beiscsi_conn->ep;
@@ -1178,7 +1212,12 @@ beiscsi_ep_connect(struct Scsi_Host *shost, struct sockaddr *dst_addr,
 		return ERR_PTR(ret);
 	}
 
-	if (phba->state != BE_ADAPTER_UP) {
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		ret = -EBUSY;
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : In PCI_ERROR Recovery\n");
+		return ERR_PTR(ret);
+	} else if (phba->state & BE_ADAPTER_LINK_DOWN) {
 		ret = -EBUSY;
 		beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_CONFIG,
 			    "BS_%d : The Adapter Port state is Down!!!\n");
@@ -1303,6 +1342,12 @@ void beiscsi_ep_disconnect(struct iscsi_endpoint *ep)
 		tcp_upload_flag = CONNECTION_UPLOAD_ABORT;
 	}
 
+	if (phba->state & BE_ADAPTER_PCI_ERR) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_CONFIG,
+			    "BS_%d : PCI_ERROR Recovery\n");
+		goto free_ep;
+	}
+
 	tag = mgmt_invalidate_connection(phba, beiscsi_ep,
 					  beiscsi_ep->ep_cid,
 					  mgmt_invalidate_flag,
@@ -1315,6 +1360,7 @@ void beiscsi_ep_disconnect(struct iscsi_endpoint *ep)
 
 	beiscsi_mccq_compl(phba, tag, NULL, NULL);
 	beiscsi_close_conn(beiscsi_ep, tcp_upload_flag);
+free_ep:
 	beiscsi_free_ep(beiscsi_ep);
 	beiscsi_unbind_conn_to_cid(phba, beiscsi_ep->ep_cid);
 	iscsi_destroy_endpoint(beiscsi_ep->openiscsi_ep);
diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index 8f300534fc32..d84ecc5317ff 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -5140,10 +5140,12 @@ void beiscsi_hba_attrs_init(struct beiscsi_hba *phba)
 /*
  * beiscsi_quiesce()- Cleanup Driver resources
  * @phba: Instance Priv structure
+ * @unload_state:i Clean or EEH unload state
  *
  * Free the OS and HW resources held by the driver
  **/
-static void beiscsi_quiesce(struct beiscsi_hba *phba)
+static void beiscsi_quiesce(struct beiscsi_hba *phba,
+		uint32_t unload_state)
 {
 	struct hwi_controller *phwi_ctrlr;
 	struct hwi_context_memory *phwi_context;
@@ -5156,28 +5158,37 @@ static void beiscsi_quiesce(struct beiscsi_hba *phba)
 	if (phba->msix_enabled) {
 		for (i = 0; i <= phba->num_cpus; i++) {
 			msix_vec = phba->msix_entries[i].vector;
+			synchronize_irq(msix_vec);
 			free_irq(msix_vec, &phwi_context->be_eq[i]);
 			kfree(phba->msi_name[i]);
 		}
 	} else
-		if (phba->pcidev->irq)
+		if (phba->pcidev->irq) {
+			synchronize_irq(phba->pcidev->irq);
 			free_irq(phba->pcidev->irq, phba);
+		}
 	pci_disable_msix(phba->pcidev);
-	destroy_workqueue(phba->wq);
+
 	if (blk_iopoll_enabled)
 		for (i = 0; i < phba->num_cpus; i++) {
 			pbe_eq = &phwi_context->be_eq[i];
 			blk_iopoll_disable(&pbe_eq->iopoll);
 		}
 
-	beiscsi_clean_port(phba);
-	beiscsi_free_mem(phba);
+	if (unload_state == BEISCSI_CLEAN_UNLOAD) {
+		destroy_workqueue(phba->wq);
+		beiscsi_clean_port(phba);
+		beiscsi_free_mem(phba);
 
-	beiscsi_unmap_pci_function(phba);
-	pci_free_consistent(phba->pcidev,
-			    phba->ctrl.mbox_mem_alloced.size,
-			    phba->ctrl.mbox_mem_alloced.va,
-			    phba->ctrl.mbox_mem_alloced.dma);
+		beiscsi_unmap_pci_function(phba);
+		pci_free_consistent(phba->pcidev,
+				    phba->ctrl.mbox_mem_alloced.size,
+				    phba->ctrl.mbox_mem_alloced.va,
+				    phba->ctrl.mbox_mem_alloced.dma);
+	} else {
+		hwi_purge_eq(phba);
+		hwi_cleanup(phba);
+	}
 
 	cancel_delayed_work_sync(&phba->beiscsi_hw_check_task);
 }
@@ -5194,11 +5205,13 @@ static void beiscsi_remove(struct pci_dev *pcidev)
 	}
 
 	beiscsi_destroy_def_ifaces(phba);
-	beiscsi_quiesce(phba);
+	beiscsi_quiesce(phba, BEISCSI_CLEAN_UNLOAD);
 	iscsi_boot_destroy_kset(phba->boot_kset);
 	iscsi_host_remove(phba->shost);
 	pci_dev_put(phba->pcidev);
 	iscsi_host_free(phba->shost);
+	pci_disable_pcie_error_reporting(pcidev);
+	pci_set_drvdata(pcidev, NULL);
 	pci_disable_device(pcidev);
 }
 
@@ -5213,7 +5226,7 @@ static void beiscsi_shutdown(struct pci_dev *pcidev)
 		return;
 	}
 
-	beiscsi_quiesce(phba);
+	beiscsi_quiesce(phba, BEISCSI_CLEAN_UNLOAD);
 	pci_disable_device(pcidev);
 }
 
@@ -5251,6 +5264,167 @@ beiscsi_hw_health_check(struct work_struct *work)
 			      msecs_to_jiffies(1000));
 }
 
+
+static pci_ers_result_t beiscsi_eeh_err_detected(struct pci_dev *pdev,
+		pci_channel_state_t state)
+{
+	struct beiscsi_hba *phba = NULL;
+
+	phba = (struct beiscsi_hba *)pci_get_drvdata(pdev);
+	phba->state |= BE_ADAPTER_PCI_ERR;
+
+	beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+		    "BM_%d : EEH error detected\n");
+
+	beiscsi_quiesce(phba, BEISCSI_EEH_UNLOAD);
+
+	if (state == pci_channel_io_perm_failure) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+			    "BM_%d : EEH : State PERM Failure");
+		return PCI_ERS_RESULT_DISCONNECT;
+	}
+
+	pci_disable_device(pdev);
+
+	/* The error could cause the FW to trigger a flash debug dump.
+	 * Resetting the card while flash dump is in progress
+	 * can cause it not to recover; wait for it to finish.
+	 * Wait only for first function as it is needed only once per
+	 * adapter.
+	 **/
+	if (pdev->devfn == 0)
+		ssleep(30);
+
+	return PCI_ERS_RESULT_NEED_RESET;
+}
+
+static pci_ers_result_t beiscsi_eeh_reset(struct pci_dev *pdev)
+{
+	struct beiscsi_hba *phba = NULL;
+	int status = 0;
+
+	phba = (struct beiscsi_hba *)pci_get_drvdata(pdev);
+
+	beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+		    "BM_%d : EEH Reset\n");
+
+	status = pci_enable_device(pdev);
+	if (status)
+		return PCI_ERS_RESULT_DISCONNECT;
+
+	pci_set_master(pdev);
+	pci_set_power_state(pdev, PCI_D0);
+	pci_restore_state(pdev);
+
+	/* Wait for the CHIP Reset to complete */
+	status = be_chk_reset_complete(phba);
+	if (!status) {
+		beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
+			    "BM_%d : EEH Reset Completed\n");
+	} else {
+		beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
+			    "BM_%d : EEH Reset Completion Failure\n");
+		return PCI_ERS_RESULT_DISCONNECT;
+	}
+
+	pci_cleanup_aer_uncorrect_error_status(pdev);
+	return PCI_ERS_RESULT_RECOVERED;
+}
+
+static void beiscsi_eeh_resume(struct pci_dev *pdev)
+{
+	int ret = 0, i;
+	struct be_eq_obj *pbe_eq;
+	struct beiscsi_hba *phba = NULL;
+	struct hwi_controller *phwi_ctrlr;
+	struct hwi_context_memory *phwi_context;
+
+	phba = (struct beiscsi_hba *)pci_get_drvdata(pdev);
+	pci_save_state(pdev);
+
+	if (enable_msix)
+		find_num_cpus(phba);
+	else
+		phba->num_cpus = 1;
+
+	if (enable_msix) {
+		beiscsi_msix_enable(phba);
+		if (!phba->msix_enabled)
+			phba->num_cpus = 1;
+	}
+
+	ret = beiscsi_cmd_reset_function(phba);
+	if (ret) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+			    "BM_%d : Reset Failed\n");
+		goto ret_err;
+	}
+
+	ret = be_chk_reset_complete(phba);
+	if (ret) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+			    "BM_%d : Failed to get out of reset.\n");
+		goto ret_err;
+	}
+
+	beiscsi_get_params(phba);
+	phba->shost->max_id = phba->params.cxns_per_ctrl;
+	phba->shost->can_queue = phba->params.ios_per_ctrl;
+	ret = hwi_init_controller(phba);
+
+	for (i = 0; i < MAX_MCC_CMD; i++) {
+		init_waitqueue_head(&phba->ctrl.mcc_wait[i + 1]);
+		phba->ctrl.mcc_tag[i] = i + 1;
+		phba->ctrl.mcc_numtag[i + 1] = 0;
+		phba->ctrl.mcc_tag_available++;
+	}
+
+	phwi_ctrlr = phba->phwi_ctrlr;
+	phwi_context = phwi_ctrlr->phwi_ctxt;
+
+	if (blk_iopoll_enabled) {
+		for (i = 0; i < phba->num_cpus; i++) {
+			pbe_eq = &phwi_context->be_eq[i];
+			blk_iopoll_init(&pbe_eq->iopoll, be_iopoll_budget,
+					be_iopoll);
+			blk_iopoll_enable(&pbe_eq->iopoll);
+		}
+
+		i = (phba->msix_enabled) ? i : 0;
+		/* Work item for MCC handling */
+		pbe_eq = &phwi_context->be_eq[i];
+		INIT_WORK(&pbe_eq->work_cqs, beiscsi_process_all_cqs);
+	} else {
+		if (phba->msix_enabled) {
+			for (i = 0; i <= phba->num_cpus; i++) {
+				pbe_eq = &phwi_context->be_eq[i];
+				INIT_WORK(&pbe_eq->work_cqs,
+					  beiscsi_process_all_cqs);
+			}
+		} else {
+			pbe_eq = &phwi_context->be_eq[0];
+			INIT_WORK(&pbe_eq->work_cqs,
+				  beiscsi_process_all_cqs);
+		}
+	}
+
+	ret = beiscsi_init_irqs(phba);
+	if (ret < 0) {
+		beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+			    "BM_%d : beiscsi_eeh_resume - "
+			    "Failed to beiscsi_init_irqs\n");
+		goto ret_err;
+	}
+
+	hwi_enable_intr(phba);
+	phba->state &= ~BE_ADAPTER_PCI_ERR;
+
+	return;
+ret_err:
+	beiscsi_log(phba, KERN_ERR, BEISCSI_LOG_INIT,
+		    "BM_%d : AER EEH Resume Failed\n");
+}
+
 static int beiscsi_dev_probe(struct pci_dev *pcidev,
 			     const struct pci_device_id *id)
 {
@@ -5258,7 +5432,7 @@ static int beiscsi_dev_probe(struct pci_dev *pcidev,
 	struct hwi_controller *phwi_ctrlr;
 	struct hwi_context_memory *phwi_context;
 	struct be_eq_obj *pbe_eq;
-	int ret, i;
+	int ret = 0, i;
 
 	ret = beiscsi_enable_pci(pcidev);
 	if (ret < 0) {
@@ -5274,6 +5448,15 @@ static int beiscsi_dev_probe(struct pci_dev *pcidev,
 		goto disable_pci;
 	}
 
+	/* Enable EEH reporting */
+	ret = pci_enable_pcie_error_reporting(pcidev);
+	if (ret)
+		beiscsi_log(phba, KERN_WARNING, BEISCSI_LOG_INIT,
+			    "BM_%d : PCIe Error Reporting "
+			    "Enabling Failed\n");
+
+	pci_save_state(pcidev);
+
 	/* Initialize Driver configuration Paramters */
 	beiscsi_hba_attrs_init(phba);
 
@@ -5359,7 +5542,7 @@ static int beiscsi_dev_probe(struct pci_dev *pcidev,
 		goto free_port;
 	}
 
-	for (i = 0; i < MAX_MCC_CMD ; i++) {
+	for (i = 0; i < MAX_MCC_CMD; i++) {
 		init_waitqueue_head(&phba->ctrl.mcc_wait[i + 1]);
 		phba->ctrl.mcc_tag[i] = i + 1;
 		phba->ctrl.mcc_numtag[i + 1] = 0;
@@ -5463,6 +5646,12 @@ disable_pci:
 	return ret;
 }
 
+static struct pci_error_handlers beiscsi_eeh_handlers = {
+	.error_detected = beiscsi_eeh_err_detected,
+	.slot_reset = beiscsi_eeh_reset,
+	.resume = beiscsi_eeh_resume,
+};
+
 struct iscsi_transport beiscsi_iscsi_transport = {
 	.owner = THIS_MODULE,
 	.name = DRV_NAME,
@@ -5501,7 +5690,8 @@ static struct pci_driver beiscsi_pci_driver = {
 	.probe = beiscsi_dev_probe,
 	.remove = beiscsi_remove,
 	.shutdown = beiscsi_shutdown,
-	.id_table = beiscsi_pci_id_table
+	.id_table = beiscsi_pci_id_table,
+	.err_handler = &beiscsi_eeh_handlers
 };
 
 
diff --git a/drivers/scsi/be2iscsi/be_main.h b/drivers/scsi/be2iscsi/be_main.h
index a8ae6b82c3e1..aace072e033e 100644
--- a/drivers/scsi/be2iscsi/be_main.h
+++ b/drivers/scsi/be2iscsi/be_main.h
@@ -26,6 +26,7 @@
 #include <linux/in.h>
 #include <linux/ctype.h>
 #include <linux/module.h>
+#include <linux/aer.h>
 #include <scsi/scsi.h>
 #include <scsi/scsi_cmnd.h>
 #include <scsi/scsi_device.h>
@@ -96,8 +97,12 @@
 
 #define INVALID_SESS_HANDLE	0xFFFFFFFF
 
-#define BE_ADAPTER_UP		0x00000000
-#define BE_ADAPTER_LINK_DOWN	0x00000001
+#define BE_ADAPTER_LINK_UP	0x001
+#define BE_ADAPTER_LINK_DOWN	0x002
+#define BE_ADAPTER_PCI_ERR	0x004
+
+#define BEISCSI_CLEAN_UNLOAD	0x01
+#define BEISCSI_EEH_UNLOAD	0x02
 /**
  * hardware needs the async PDU buffers to be posted in multiples of 8
  * So have atleast 8 of them by default