nvme-multipath: fix possible hang in live ns resize with ANA access
Sagi Grimberg
maintainerPATCH v2 1/1 nvme-multipath: fix possible hang in live ns resize with ANA access
When we revalidate paths as part of ns size change (as of commit e7d65803e2bb), it is possible that during the path revalidation, the only paths that is IO capable (i.e. optimized/non-optimized) are the ones that ns resize was not yet informed to the host, which will cause inflight requests to be requeued (as we have available paths but none are IO capable). These requests on the requeue list are waiting for someone to resubmit them at some point.
The IO capable paths will eventually notify the ns resize change to the host, but there is nothing that will kick the requeue list to resubmit the queued requests.
Fix this by always kicking the requeue list, and if no IO capable path exists, these requests will just end up being queued again.
A typical log that indicates that IOs are requeued:
--
nvme nvme1: creating 4 I/O queues.
nvme nvme1: new ctrl: "testnqn1"
nvme nvme2: creating 4 I/O queues.
nvme nvme2: mapped 4/0/0 default/read/poll queues. nvme nvme2: new ctrl: NQN "testnqn1", addr 127.0.0.1:8009 nvme nvme1: rescanning namespaces.
nvme1n1: detected capacity change from 2097152 to 4194304 block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O nvme nvme2: rescanning namespaces.
--
---
Changes from v1:
- fix commit msg body format
- follow reverse-xmas declaration pattern
drivers/nvme/host/multipath.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/nvme/host/multipath.c +5 -3
@@ -173,15 +173,17 @@ void nvme_mpath_revalidate_paths(struct nvme_ns *ns) {struct nvme_ns_head *head = ns->head;sector_t capacity =get_capacity (head->disk); +struct nvme_ns *n;int node; -list_for_each_entry_rcu (ns, &head->list, siblings) { -if (capacity !=get_capacity (ns->disk)) -clear_bit (NVME_NS_READY , &ns->flags); +list_for_each_entry_rcu (n, &head->list, siblings) { +if (capacity !=get_capacity (n->disk)) +clear_bit (NVME_NS_READY , &n->flags); }for_each_node (node)rcu_assign_pointer (head->current_path[node],NULL ); +nvme_kick_requeue_lists (ns->ctrl); }static bool nvme_path_is_disabled (struct nvme_ns *ns)
Signature · 1 line
2.34.1
Quoted 60 lines · Sagi Grimberg
When we revalidate paths as part of ns size change (as of commit
e7d65803e2bb), it is possible that during the path revalidation, the only paths that is IO capable (i.e. optimized/non-optimized) are the ones that ns resize was not yet informed to the host, which will cause inflight requests to be requeued (as we have available paths but none are IO capable). These requests on the requeue list are waiting for someone to resubmit them at some point.The IO capable paths will eventually notify the ns resize change to the host, but there is nothing that will kick the requeue list to resubmit the queued requests.
Fix this by always kicking the requeue list, and if no IO capable path exists, these requests will just end up being queued again.
A typical log that indicates that IOs are requeued:
--
nvme nvme1: creating 4 I/O queues.
nvme nvme1: new ctrl: "testnqn1"
nvme nvme2: creating 4 I/O queues.
nvme nvme2: mapped 4/0/0 default/read/poll queues. nvme nvme2: new ctrl: NQN "testnqn1", addr 127.0.0.1:8009 nvme nvme1: rescanning namespaces.
nvme1n1: detected capacity change from 2097152 to 4194304 block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O nvme nvme2: rescanning namespaces.
--Reported-by Yogev Cohen <[email protected]>Fixes e7d65803e2bb ("nvme-multipath: revalidate paths during rescan")Signed-off-by Sagi Grimberg <[email protected]>---
Changes from v1:
- fix commit msg body format
- follow reverse-xmas declaration patterndrivers/nvme/host/multipath.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)drivers/nvme/host/multipath.c +1
-1@@ -173,15 +173,17 @@ void nvme_mpath_revalidate_paths(struct nvme_ns *ns) {struct nvme_ns_head *head = ns->head;sector_t capacity =get_capacity (head->disk); +struct nvme_ns *n;int node; -list_for_each_entry_rcu (ns, &head->list, siblings) {
drivers/nvme/host/multipath.c void nvme_mpath_revalidate_paths(struct nvme_ns *ns) · line 186
Quoted 10 lines
drivers/nvme/host/multipath.c +4
-2-if (capacity !=get_capacity (ns->disk)) -clear_bit (NVME_NS_READY , &ns->flags); +list_for_each_entry_rcu (n, &head->list, siblings) { +if (capacity !=get_capacity (n->disk)) +clear_bit (NVME_NS_READY , &n->flags); }for_each_node (node)rcu_assign_pointer (head->current_path[node],NULL ); +nvme_kick_requeue_lists (ns->ctrl);
We just need to schedule the requeue_work of the head instead of all heads. we can do like this:
kblockd_schedule_work(&head->requeue_work);
Quoted 4 lines
drivers/nvme/host/multipath.c +0
-0}static bool nvme_path_is_disabled (struct nvme_ns *ns)
Sagi Grimberg
maintainerQuoted 67 lines · Chao Leng
Quoted 64 lines · Sagi Grimberg
When we revalidate paths as part of ns size change (as of commit
e7d65803e2bb), it is possible that during the path revalidation, the only paths that is IO capable (i.e. optimized/non-optimized) are the ones that ns resize was not yet informed to the host, which will cause inflight requests to be requeued (as we have available paths but none are IO capable). These requests on the requeue list are waiting for someone to resubmit them at some point.The IO capable paths will eventually notify the ns resize change to the host, but there is nothing that will kick the requeue list to resubmit the queued requests.
Fix this by always kicking the requeue list, and if no IO capable path exists, these requests will just end up being queued again.
A typical log that indicates that IOs are requeued:
Signature · 30 lines
nvme nvme1: creating 4 I/O queues. nvme nvme1: new ctrl: "testnqn1" nvme nvme2: creating 4 I/O queues. nvme nvme2: mapped 4/0/0 default/read/poll queues. nvme nvme2: new ctrl: NQN "testnqn1", addr 127.0.0.1:8009 nvme nvme1: rescanning namespaces. nvme1n1: detected capacity change from 2097152 to 4194304 block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O block nvme1n1: no usable path - requeuing I/O nvme nvme2: rescanning namespaces. -- Reported-by: Yogev Cohen <[email protected]> Fixes: e7d65803e2bb ("nvme-multipath: revalidate paths during rescan") Signed-off-by: Sagi Grimberg <[email protected]> --- Changes from v1: - fix commit msg body format - follow reverse-xmas declaration pattern drivers/nvme/host/multipath.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)drivers/nvme/host/multipath.c +3
-3@@ -173,15 +173,17 @@ void nvme_mpath_revalidate_paths(struct nvme_ns *ns) { struct nvme_ns_head *head = ns->head; sector_t capacity = get_capacity(head->disk); + struct nvme_ns *n; int node; - list_for_each_entry_rcu(ns, &head->list, siblings) { - if (capacity != get_capacity(ns->disk)) - clear_bit(NVME_NS_READY, &ns->flags); + list_for_each_entry_rcu(n, &head->list, siblings) { + if (capacity != get_capacity(n->disk))
Quoted 8 lines
drivers/nvme/host/multipath.c void nvme_mpath_revalidate_paths(struct nvme_ns *ns) · line 186
Quoted 5 lines
drivers/nvme/host/multipath.c +2
-0+ clear_bit(NVME_NS_READY, &n->flags); } for_each_node(node) rcu_assign_pointer(head->current_path[node], NULL); + nvme_kick_requeue_lists(ns->ctrl);We just need to schedule the
requeue_workof the head instead of all heads. we can do like this:
kblockd_schedule_work(&head->requeue_work);
Yes, you're right, that is simpler to do. I'll send a v3.