From: James Smart <jsmart2021@xxxxxxxxx>
[ Upstream commit a7d139145a6640172516b193abf6d2398620aa14 ]
The __nvmf_check_ready() routine used to bounce all filesystem io if the
controller state isn't LIVE. However, a later patch changed the logic so
that it rejection ends up being based on the Q live check. The FC
transport has a slightly different sequence from rdma and tcp for
shutting down queues/marking them non-live. FC marks its queue non-live
after aborting all ios and waiting for their termination, leaving a
rather large window for filesystem io to continue to hit the transport.
Unfortunately this resulted in filesystem I/O or applications seeing I/O
errors.
Change the FC transport to mark the queues non-live at the first sign of
teardown for the association (when I/O is initially terminated).
Fixes: 73a5379937ec ("nvme-fabrics: allow to queue requests for live queues")
Signed-off-by: James Smart <jsmart2021@xxxxxxxxx>
Reviewed-by: Sagi Grimberg <sagi@xxxxxxxxxxx>
Reviewed-by: Himanshu Madhani <himanshu.madhani@xxxxxxxxxx>
Reviewed-by: Hannes Reinecke <hare@xxxxxxx>
Signed-off-by: Christoph Hellwig <hch@xxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
drivers/nvme/host/fc.c | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c
index 41257daf7464..a0bcec33b020 100644
--- a/drivers/nvme/host/fc.c
+++ b/drivers/nvme/host/fc.c
@@ -2460,6 +2460,18 @@ nvme_fc_terminate_exchange(struct request *req, void *data, bool reserved)
static void
__nvme_fc_abort_outstanding_ios(struct nvme_fc_ctrl *ctrl, bool start_queues)
{
+ int q;
+
+ /*
+ * if aborting io, the queues are no longer good, mark them
+ * all as not live.
+ */
+ if (ctrl->ctrl.queue_count > 1) {
+ for (q = 1; q < ctrl->ctrl.queue_count; q++)
+ clear_bit(NVME_FC_Q_LIVE, &ctrl->queues[q].flags);
+ }
+ clear_bit(NVME_FC_Q_LIVE, &ctrl->queues[0].flags);
+
/*
* If io queues are present, stop them and terminate all outstanding
* ios on them. As FC allocates FC exchange for each io, the
--
2.30.2
