Skip to contentSkip to conversation

Enable devlink support

PATCH Applied Javed Hasan 2 participants 5 messages Mar 31, 2021 – Apr 16, 2021

Subsystems (2)
lore
Share & tools
lore

Javed Hasan

author maintainer
#0 lore raw

cover Enable devlink support

Devlink instance lifecycle was linked to qed_dev object, that caused devlink to be recreated on each recovery. Use devlink_health_report to push error indications.

Kindly apply this series to scsi-queue at your earliest convenience.

Javed Hasan (2):
qedf: Enable devlink support
qedf: use devlink to report errors and recovery

 drivers/scsi/qedf/qedf.h      |  1 +
 drivers/scsi/qedf/qedf_main.c | 18 ++++++++++++++++--
 2 files changed, 17 insertions(+), 2 deletions(-)
Signature · 1 line
2.18.2
author maintainer
#1 lore raw

PATCH 1/2 qedf: Enable devlink support

Devlink instance lifecycle was linked to qed_dev object, that caused devlink to be recreated on each recovery.

Changing it by making higher level driver (qede) responsible for its life. This way devlink now survives recoveries.

qede now stores devlink structure pointer as a part of its device object, devlink private data contains a linkage structure, qed_devlink.

---
 drivers/scsi/qedf/qedf.h      |  1 +
 drivers/scsi/qedf/qedf_main.c | 13 +++++++++++++
 2 files changed, 14 insertions(+)
drivers/scsi/qedf/qedf.h +1 -0
index 88a592d09433..36dde38608dd 100644
@@ -335,6 +335,7 @@ struct qedf_ctx {
 	unsigned int curr_conn_id;
 	struct workqueue_struct *ll2_recv_wq;
 	struct workqueue_struct *link_update_wq;
+	struct devlink *devlink;
 	struct delayed_work link_update;
 	struct delayed_work link_recovery;
 	struct completion flogi_compl;
drivers/scsi/qedf/qedf_main.c +13 -0
index 46d185cb9ea8..6b6c3fd97798 100644
@@ -3408,6 +3408,14 @@ static int __qedf_probe(struct pci_dev *pdev, int mode)
 		goto err2;
 	}
 
+	if (mode != QEDF_MODE_RECOVERY) {
+		qedf->devlink = qed_ops->common->devlink_register(qedf->cdev);
+		if (IS_ERR(qedf->devlink)) {
+			QEDF_ERR(&qedf->dbg_ctx, "Cannot register devlink\n");
+			qedf->devlink = NULL;
+		}
+	}
+
 	/* Record BDQ producer doorbell addresses */
 	qedf->bdq_primary_prod = qedf->dev_info.primary_dbq_rq_addr;
 	qedf->bdq_secondary_prod = qedf->dev_info.secondary_bdq_rq_addr;
@@ -3789,6 +3797,11 @@ static void __qedf_remove(struct pci_dev *pdev, int mode)
 		QEDF_ERR(&(qedf->dbg_ctx),
 			"Failed to send drv state to MFW.\n");
 
+	if (mode != QEDF_MODE_RECOVERY && qedf->devlink) {
+		qed_ops->common->devlink_unregister(qedf->devlink);
+		qedf->devlink = NULL;
+	}
+
 	qed_ops->common->slowpath_stop(qedf->cdev);
 	qed_ops->common->remove(qedf->cdev);
 
Signature · 1 line
2.18.2
author maintainer
#2 lore raw

PATCH 2/2 qedf: use devlink to report errors and recovery

In reply to Javed Hasan

Use devlink_health_report to push error indications.
We implement this in qede via callback function to make it possible to reuse the same for other drivers sitting on top of qed in future. Also, remove forcible recovery trigger and put it as a normal devlink callback in qed module.

This allows user to enable/disable it via

devlink health set pci/xxxx:xx:xx.x reporter fw_fatal auto_recover false

---
 drivers/scsi/qedf/qedf_main.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/scsi/qedf/qedf_main.c +3 -2
index 6b6c3fd97798..ad125b0ff1de 100644
@@ -3859,8 +3859,9 @@ void qedf_schedule_hw_err_handler(void *dev, enum qed_hw_err_type err_type)
 		/* 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);
+		if (qedf_enable_recovery && qedf->devlink)
+			qed_ops->common->report_fatal_error(qedf->devlink,
+				err_type);
 
 		break;
 	default:
Signature · 1 line
2.18.2
#3 lore raw
In reply to Javed Hasan

Javed,

Quoted 3 lines

Devlink instance lifecycle was linked to qed_dev object, that caused devlink to be recreated on each recovery. Use devlink_health_report to push error indications.

Applied to 5.13/scsi-staging, thanks!

Signature · 1 line
Martin K. Petersen	Oracle Linux Engineering
#4 lore raw
In reply to Javed Hasan
Quoted 11 lines · Javed Hasan

Devlink instance lifecycle was linked to qed_dev object, that caused devlink to be recreated on each recovery. Use devlink_health_report to push error indications.

Kindly apply this series to scsi-queue at your earliest convenience.

Javed Hasan (2):
qedf: Enable devlink support
qedf: use devlink to report errors and recovery

[...]

Applied to 5.13/scsi-queue, thanks!

[1/2] qedf: Enable devlink support

      https://git.kernel.org/mkp/scsi/c/4aab946f789e
[2/2] qedf: use devlink to report errors and recovery
      https://git.kernel.org/mkp/scsi/c/9d6f87c25ad5
Signature · 1 line
Martin K. Petersen	Oracle Linux Engineering