scsi: qedf: Add schedule recovery handler

Implement recovery handler to be used by QED to signal the need for
recovery to come out of an error condition like ramrod struck and firmware
context reset.

Link: https://lore.kernel.org/r/20200416084314.18851-8-skashyap@marvell.com
Signed-off-by: Chad Dupuis <cdupuis@marvell.com>
Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

diff --git a/drivers/scsi/qedf/qedf.h b/drivers/scsi/qedf/qedf.h
index aaa2ac95ea23ab07b2c23937443ea930b16f11c9..f8a98e5e3cb508f3b27ff16416f6d0aa4b8dc4ab 100644 (file)
--- a/drivers/scsi/qedf/qedf.h
+++ b/drivers/scsi/qedf/qedf.h
@@ -387,6 +387,7 @@ struct qedf_ctx {
 #define QEDF_IO_WORK_MIN               64
        mempool_t *io_mempool;
        struct workqueue_struct *dpc_wq;
+       struct delayed_work recovery_work;
        struct delayed_work grcdump_work;
        struct delayed_work stag_work;
 

diff --git a/drivers/scsi/qedf/qedf_main.c b/drivers/scsi/qedf/qedf_main.c
index 50557c4dbf217d35887d2dd74c5c86c307795c5e..3823cdfec46a7926e9168aa62520fa7e2ed36865 100644 (file)
--- a/drivers/scsi/qedf/qedf_main.c
+++ b/drivers/scsi/qedf/qedf_main.c
@@ -28,6 +28,8 @@ const struct qed_fcoe_ops *qed_ops;
 static int qedf_probe(struct pci_dev *pdev, const struct pci_device_id *id);
 static void qedf_remove(struct pci_dev *pdev);
 static void qedf_shutdown(struct pci_dev *pdev);
+static void qedf_schedule_recovery_handler(void *dev);
+static void qedf_recovery_handler(struct work_struct *work);
 
 /*
  * Driver module parameters.
@@ -662,6 +664,7 @@ static struct qed_fcoe_cb_ops qedf_cb_ops = {
        {
                .link_update = qedf_link_update,
                .bw_update = qedf_bw_update,
+               .schedule_recovery_handler = qedf_schedule_recovery_handler,
                .dcbx_aen = qedf_dcbx_handler,
                .get_generic_tlv_data = qedf_get_generic_tlv_data,
                .get_protocol_tlv_data = qedf_get_protocol_tlv_data,
@@ -3502,6 +3505,7 @@ static int __qedf_probe(struct pci_dev *pdev, int mode)
                    qedf->lport->host->host_no);
                qedf->dpc_wq = create_workqueue(host_buf);
        }
+       INIT_DELAYED_WORK(&qedf->recovery_work, qedf_recovery_handler);
 
        /*
         * GRC dump and sysfs parameters are not reaped during the recovery
@@ -3817,6 +3821,45 @@ static void qedf_shutdown(struct pci_dev *pdev)
        __qedf_remove(pdev, QEDF_MODE_NORMAL);
 }
 
+/*
+ * Recovery handler code
+ */
+static void qedf_schedule_recovery_handler(void *dev)
+{
+       struct qedf_ctx *qedf = dev;
+
+       QEDF_ERR(&qedf->dbg_ctx, "Recovery handler scheduled.\n");
+       schedule_delayed_work(&qedf->recovery_work, 0);
+}
+
+static void qedf_recovery_handler(struct work_struct *work)
+{
+       struct qedf_ctx *qedf =
+           container_of(work, struct qedf_ctx, recovery_work.work);
+
+       if (test_and_set_bit(QEDF_IN_RECOVERY, &qedf->flags))
+               return;
+
+       /*
+        * Call common_ops->recovery_prolog to allow the MFW to quiesce
+        * any PCI transactions.
+        */
+       qed_ops->common->recovery_prolog(qedf->cdev);
+
+       QEDF_ERR(&qedf->dbg_ctx, "Recovery work start.\n");
+       __qedf_remove(qedf->pdev, QEDF_MODE_RECOVERY);
+       /*
+        * Reset link and dcbx to down state since we will not get a link down
+        * event from the MFW but calling __qedf_remove will essentially be a
+        * link down event.
+        */
+       atomic_set(&qedf->link_state, QEDF_LINK_DOWN);
+       atomic_set(&qedf->dcbx, QEDF_DCBX_PENDING);
+       __qedf_probe(qedf->pdev, QEDF_MODE_RECOVERY);
+       clear_bit(QEDF_IN_RECOVERY, &qedf->flags);
+       QEDF_ERR(&qedf->dbg_ctx, "Recovery work complete.\n");
+}
+
 /* Generic TLV data callback */
 void qedf_get_generic_tlv_data(void *dev, struct qed_generic_tlvs *data)
 {