scsi: qedf: Add schedule_hw_err_handler callback for fan failure

On fan failure, disable the PCI function and initiate recovery for ramrod
failure.

Link: https://lore.kernel.org/r/20200907121443.5150-7-jhasan@marvell.comSigned-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/qedf/qedf.h

@@ -389,6 +389,7 @@ struct qedf_ctx {
 	mempool_t *io_mempool;
 	struct workqueue_struct *dpc_wq;
 	struct delayed_work recovery_work;
+	struct delayed_work board_disable_work;
 	struct delayed_work grcdump_work;
 	struct delayed_work stag_work;
 
@@ -541,6 +542,9 @@ extern void qedf_get_generic_tlv_data(void *dev, struct qed_generic_tlvs *data);
 extern void qedf_wq_grcdump(struct work_struct *work);
 void qedf_stag_change_work(struct work_struct *work);
 void qedf_ctx_soft_reset(struct fc_lport *lport);
+extern void qedf_board_disable_work(struct work_struct *work);
+extern void qedf_schedule_hw_err_handler(void *dev,
+		enum qed_hw_err_type err_type);
 
 #define FCOE_WORD_TO_BYTE  4
 #define QEDF_MAX_TASK_NUM	0xFFFF

drivers/scsi/qedf/qedf_main.c

@@ -105,6 +105,12 @@ module_param_named(dp_level, qedf_dp_level, uint, S_IRUGO);
 MODULE_PARM_DESC(dp_level, " printk verbosity control passed to qed module  "
 	"during probe (0-3: 0 more verbose).");
 
+static bool qedf_enable_recovery = true;
+module_param_named(enable_recovery, qedf_enable_recovery,
+		bool, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(enable_recovery, "Enable/disable recovery on driver/firmware "
+		"interface level errors 0 = Disabled, 1 = Enabled (Default: 1).");
+
 struct workqueue_struct *qedf_io_wq;
 
 static struct fcoe_percpu_s qedf_global;
@@ -690,6 +696,7 @@ static struct qed_fcoe_cb_ops qedf_cb_ops = {
 		.dcbx_aen = qedf_dcbx_handler,
 		.get_generic_tlv_data = qedf_get_generic_tlv_data,
 		.get_protocol_tlv_data = qedf_get_protocol_tlv_data,
+		.schedule_hw_err_handler = qedf_schedule_hw_err_handler,
 	}
 };
 
@@ -3799,6 +3806,44 @@ void qedf_wq_grcdump(struct work_struct *work)
 	qedf_capture_grc_dump(qedf);
 }
 
+void qedf_schedule_hw_err_handler(void *dev, enum qed_hw_err_type err_type)
+{
+	struct qedf_ctx *qedf = dev;
+
+	QEDF_ERR(&(qedf->dbg_ctx),
+			"Hardware error handler scheduled, event=%d.\n",
+			err_type);
+
+	if (test_bit(QEDF_IN_RECOVERY, &qedf->flags)) {
+		QEDF_ERR(&(qedf->dbg_ctx),
+				"Already in recovery, not scheduling board disable work.\n");
+		return;
+	}
+
+	switch (err_type) {
+	case QED_HW_ERR_FAN_FAIL:
+		schedule_delayed_work(&qedf->board_disable_work, 0);
+		break;
+	case QED_HW_ERR_MFW_RESP_FAIL:
+	case QED_HW_ERR_HW_ATTN:
+	case QED_HW_ERR_DMAE_FAIL:
+	case QED_HW_ERR_FW_ASSERT:
+		/* Prevent HW attentions from being reasserted */
+		qed_ops->common->attn_clr_enable(qedf->cdev, true);
+		break;
+	case QED_HW_ERR_RAMROD_FAIL:
+		/* Prevent HW attentions from being reasserted */
+		qed_ops->common->attn_clr_enable(qedf->cdev, true);
+
+		if (qedf_enable_recovery)
+			qed_ops->common->recovery_process(qedf->cdev);
+
+		break;
+	default:
+		break;
+	}
+}
+
 /*
  * Protocol TLV handler
  */