| Message ID | 20200403120957.2431-2-skashyap@marvell.com |
|---|---|
| State | Changes Requested |
| Delegated to: | David Miller |
| Headers | show |
| Series | qed/qedf: Firmware recovery, bw update and misc fixes. | expand |
Saurav, Please, no "." at the end of Subject: lines. > - Problem: Port not coming up after bringing down the port > for longer duration. > - Bring down the port from the switch > - wait for fipvlan to exhaust, driver will use > default vlan (1002) and call fcoe_ctlr_link_up > - libfc/fcoe will start sending FLOGI > - bring back the port and switch discard FLOGI > because vlan is different. > - keep track of pending flogi and if it increases > certain number then do ctx reset and it will do > fipvlan again. That doesn't look like a proper commit message. How about something like: 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.
Hi Martin, > -----Original Message----- > From: Martin K. Petersen <martin.petersen@oracle.com> > Sent: Tuesday, April 14, 2020 6:50 AM > To: Saurav Kashyap <skashyap@marvell.com> > Cc: martin.petersen@oracle.com; GR-QLogic-Storage-Upstream <GR-QLogic- > Storage-Upstream@marvell.com>; linux-scsi@vger.kernel.org; Javed Hasan > <jhasan@marvell.com>; netdev@vger.kernel.org > Subject: [EXT] Re: [PATCH v3 1/7] qedf: Keep track of num of pending flogi. > > External Email > > ---------------------------------------------------------------------- > > Saurav, > > Please, no "." at the end of Subject: lines. > > > - Problem: Port not coming up after bringing down the port > > for longer duration. > > - Bring down the port from the switch > > - wait for fipvlan to exhaust, driver will use > > default vlan (1002) and call fcoe_ctlr_link_up > > - libfc/fcoe will start sending FLOGI > > - bring back the port and switch discard FLOGI > > because vlan is different. > > - keep track of pending flogi and if it increases > > certain number then do ctx reset and it will do > > fipvlan again. > > That doesn't look like a proper commit message. > > How about something like: > > 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. <SK> Looks good. Thanks, ~Saurav > > -- > Martin K. Petersen Oracle Linux Engineering
diff --git a/drivers/scsi/qedf/qedf.h b/drivers/scsi/qedf/qedf.h index f3f399f..042ebf6 100644 --- a/drivers/scsi/qedf/qedf.h +++ b/drivers/scsi/qedf/qedf.h @@ -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; diff --git a/drivers/scsi/qedf/qedf_main.c b/drivers/scsi/qedf/qedf_main.c index 604856e..ee468102 100644 --- a/drivers/scsi/qedf/qedf_main.c +++ b/drivers/scsi/qedf/qedf_main.c @@ -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, @@ -3205,6 +3212,7 @@ 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, " @@ -3235,6 +3243,7 @@ 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) { @@ -3770,6 +3779,20 @@ 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);
- Problem: Port not coming up after bringing down the port for longer duration. - Bring down the port from the switch - wait for fipvlan to exhaust, driver will use default vlan (1002) and call fcoe_ctlr_link_up - libfc/fcoe will start sending FLOGI - bring back the port and switch discard FLOGI because vlan is different. - keep track of pending flogi and if it increases certain number then do ctx reset and it will do fipvlan again. Signed-off-by: Saurav Kashyap <skashyap@marvell.com> --- drivers/scsi/qedf/qedf.h | 2 ++ drivers/scsi/qedf/qedf_main.c | 23 +++++++++++++++++++++++ 2 files changed, 25 insertions(+)