Skip to contentSkip to conversation

nvme: send uevent once a multipath namespace is operational again

PATCH In review Hannes Reinecke 4 participants 5 messages Mar 3, 2021 – Mar 11, 2021 NVMe

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 nvme: send uevent once a multipath namespace is operational again

In an all paths down scenario I/O will be requeued or aborted, and no further I/O will be ongoing on this namespace.
This leaves the upper layers unable to determine if the namespace becomes operational again eg. after a successful controller reset. With this patch a 'change' uevent will be sent per multipathed namespace once the underlying controller moved to LIVE and started I/O processing by calling nvme_kick_requeue_lists().

---
 drivers/nvme/host/multipath.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
drivers/nvme/host/multipath.c +5 -2
index 0696319adaf6..1a9e14ebe11b 100644
@@ -97,8 +97,11 @@ void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl)
 
 	down_read(&ctrl->namespaces_rwsem);
 	list_for_each_entry(ns, &ctrl->namespaces, list) {
-		if (ns->head->disk)
-			kblockd_schedule_work(&ns->head->requeue_work);
+		if (!ns->head->disk)
+			continue;
+		kblockd_schedule_work(&ns->head->requeue_work);
+		if (ctrl->state == NVME_CTRL_LIVE)
+			disk_uevent(ns->head->disk, KOBJ_CHANGE);
 	}
 	up_read(&ctrl->namespaces_rwsem);
 }
Signature · 7 lines
2.29.2


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
#2 lore raw
In reply to Christoph Hellwig
Quoted 9 lines · Christoph Hellwig
Quoted 7 lines · Hannes Reinecke

In an all paths down scenario I/O will be requeued or aborted, and no further I/O will be ongoing on this namespace.
This leaves the upper layers unable to determine if the namespace becomes operational again eg. after a successful controller reset. With this patch a 'change' uevent will be sent per multipathed namespace once the underlying controller moved to LIVE and started I/O processing by calling nvme_kick_requeue_lists().

Quoted 4 lines

This looks functionally correct, but I'm a little worried about the nvme_kick_requeue_lists function name being a little off now. Can anyone on the list think of a better name?

nvme_restart_processing()?

Cheers,

Hannes

Signature · 9 lines
Dr. Hannes Reinecke		           Kernel Storage Architect
[email protected]			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
#1 lore raw
In reply to Hannes Reinecke
Quoted 7 lines · Hannes Reinecke

In an all paths down scenario I/O will be requeued or aborted, and no further I/O will be ongoing on this namespace.
This leaves the upper layers unable to determine if the namespace becomes operational again eg. after a successful controller reset. With this patch a 'change' uevent will be sent per multipathed namespace once the underlying controller moved to LIVE and started I/O processing by calling nvme_kick_requeue_lists().

This looks functionally correct, but I'm a little worried about the nvme_kick_requeue_lists function name being a little off now. Can anyone on the list think of a better name?

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

maintainer
#3 lore raw
In reply to Hannes Reinecke
Quoted 5 lines

In an all paths down scenario I/O will be requeued or aborted, and no further I/O will be ongoing on this namespace.
This leaves the upper layers unable to determine if the namespace becomes operational again eg. after a successful controller reset.

Who is upper-layers here?

Quoted 4 lines

With this patch a 'change' uevent will be sent per multipathed namespace once the underlying controller moved to LIVE and started I/O processing by calling nvme_kick_requeue_lists().

Other than that looks reasonable to me..

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

#4 lore raw
Quoted 7 lines · Sagi Grimberg
Quoted 5 lines

In an all paths down scenario I/O will be requeued or aborted, and no
further I/O will be ongoing on this namespace.
This leaves the upper layers unable to determine if the namespace becomes operational again eg. after a successful controller reset.

Who is upper-layers here?

While a notification that I/O is unlikely to fail or get queued waiting for a controller to reconnect does seem like a nice feature, it seems like without some ability to notify that we have *entered* the all paths down state, what is userspace supposed to do with a "it's working now" uevent?

Also, what about the case where there are paths available but they are all in the ANA INACCESSIBLE state? Isn't that the same thing from a user perspective? This would be a common scenario e.g. when a storage device has to change the ANA state if a array-side link fails.

-Ewan

Quoted 8 lines
Quoted 7 lines

With this patch a 'change' uevent will be sent per multipathed namespace
once the underlying controller moved to LIVE and started I/O processing
by calling nvme_kick_requeue_lists().

Quoted 7 lines

Other than that looks reasonable to me..

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme

_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme