OSDN Git Service

scsi: qedf: Keep track of num of pending flogi
authorSaurav Kashyap <skashyap@marvell.com>
Thu, 16 Apr 2020 08:43:06 +0000 (01:43 -0700)
committerMartin K. Petersen <martin.petersen@oracle.com>
Fri, 17 Apr 2020 21:55:23 +0000 (17:55 -0400)
If a port is brought down for an extended period of time, the fipvlan
counter gets exhausted and the driver will fall back to default VLAN 1002
and call fcoe_ctlr_link_up to log in. However, the switch will discard the
FLOGI attempt because the VLAN is now different.

Keep track of the number of FLOGI attempts and if a threshold of
QEDF_FLOGI_RETRY_CNT is exceeded, perform a context soft reset.

Link: https://lore.kernel.org/r/20200416084314.18851-2-skashyap@marvell.com
Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/qedf/qedf.h
drivers/scsi/qedf/qedf_main.c

index f3f399f..042ebf6 100644 (file)
@@ -388,6 +388,7 @@ struct qedf_ctx {
        mempool_t *io_mempool;
        struct workqueue_struct *dpc_wq;
        struct delayed_work grcdump_work;
+       struct delayed_work stag_work;
 
        u32 slow_sge_ios;
        u32 fast_sge_ios;
@@ -403,6 +404,7 @@ struct qedf_ctx {
 
        u32 flogi_cnt;
        u32 flogi_failed;
+       u32 flogi_pending;
 
        /* Used for fc statistics */
        struct mutex stats_mutex;
index 5b19f51..89d5c12 100644 (file)
@@ -282,6 +282,7 @@ static void qedf_flogi_resp(struct fc_seq *seq, struct fc_frame *fp,
        else if (fc_frame_payload_op(fp) == ELS_LS_ACC) {
                /* Set the source MAC we will use for FCoE traffic */
                qedf_set_data_src_addr(qedf, fp);
+               qedf->flogi_pending = 0;
        }
 
        /* Complete flogi_compl so we can proceed to sending ADISCs */
@@ -307,6 +308,11 @@ static struct fc_seq *qedf_elsct_send(struct fc_lport *lport, u32 did,
         */
        if (resp == fc_lport_flogi_resp) {
                qedf->flogi_cnt++;
+               if (qedf->flogi_pending >= QEDF_FLOGI_RETRY_CNT) {
+                       schedule_delayed_work(&qedf->stag_work, 2);
+                       return NULL;
+               }
+               qedf->flogi_pending++;
                return fc_elsct_send(lport, did, fp, op, qedf_flogi_resp,
                    arg, timeout);
        }
@@ -850,6 +856,7 @@ void qedf_ctx_soft_reset(struct fc_lport *lport)
 
        qedf = lport_priv(lport);
 
+       qedf->flogi_pending = 0;
        /* For host reset, essentially do a soft link up/down */
        atomic_set(&qedf->link_state, QEDF_LINK_DOWN);
        QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC,
@@ -3197,6 +3204,7 @@ static int __qedf_probe(struct pci_dev *pdev, int mode)
                init_completion(&qedf->fipvlan_compl);
                mutex_init(&qedf->stats_mutex);
                mutex_init(&qedf->flush_mutex);
+               qedf->flogi_pending = 0;
 
                QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_INFO,
                   "QLogic FastLinQ FCoE Module qedf %s, "
@@ -3227,6 +3235,7 @@ static int __qedf_probe(struct pci_dev *pdev, int mode)
        INIT_DELAYED_WORK(&qedf->link_update, qedf_handle_link_update);
        INIT_DELAYED_WORK(&qedf->link_recovery, qedf_link_recovery);
        INIT_DELAYED_WORK(&qedf->grcdump_work, qedf_wq_grcdump);
+       INIT_DELAYED_WORK(&qedf->stag_work, qedf_stag_change_work);
        qedf->fipvlan_retries = qedf_fipvlan_retries;
        /* Set a default prio in case DCBX doesn't converge */
        if (qedf_default_prio > -1) {
@@ -3762,6 +3771,20 @@ void qedf_get_protocol_tlv_data(void *dev, void *data)
        fcoe->scsi_tsk_full = qedf->task_set_fulls;
 }
 
+/* Deferred work function to perform soft context reset on STAG change */
+void qedf_stag_change_work(struct work_struct *work)
+{
+       struct qedf_ctx *qedf =
+           container_of(work, struct qedf_ctx, stag_work.work);
+
+       if (!qedf) {
+               QEDF_ERR(&qedf->dbg_ctx, "qedf is NULL");
+               return;
+       }
+       QEDF_ERR(&qedf->dbg_ctx, "Performing software context reset.\n");
+       qedf_ctx_soft_reset(qedf->lport);
+}
+
 static void qedf_shutdown(struct pci_dev *pdev)
 {
        __qedf_remove(pdev, QEDF_MODE_NORMAL);