nvme-rdma: reject non-connect commands before the queue is live

If we reconncect we might have command queue up that get resent as soon
as the queue is restarted.  But until the connect command succeeded we
can't send other command.  Add a new flag that marks a queue as live when
connect finishes, and delay any non-connect command until the queue is
live based on it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
[sagig: fixes admin queue LIVE setting]
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>

diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index 5a8388177959916dc4dadbc5e6d8c366b0eb6b81..438d6948895f0dc065e8d6f21ceef55adc80921c 100644 (file)
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -83,6 +83,7 @@ enum nvme_rdma_queue_flags {
        NVME_RDMA_Q_CONNECTED = (1 << 0),
        NVME_RDMA_IB_QUEUE_ALLOCATED = (1 << 1),
        NVME_RDMA_Q_DELETING = (1 << 2),
+       NVME_RDMA_Q_LIVE = (1 << 3),
 };
 
 struct nvme_rdma_queue {
@@ -626,6 +627,7 @@ static int nvme_rdma_connect_io_queues(struct nvme_rdma_ctrl *ctrl)
                ret = nvmf_connect_io_queue(&ctrl->ctrl, i);
                if (ret)
                        break;
+               set_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[i].flags);
        }
 
        return ret;
@@ -712,6 +714,8 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
        if (ret)
                goto stop_admin_q;
 
+       set_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[0].flags);
+
        ret = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap);
        if (ret)
                goto stop_admin_q;
@@ -761,8 +765,10 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)
 
        nvme_stop_keep_alive(&ctrl->ctrl);
 
-       for (i = 0; i < ctrl->queue_count; i++)
+       for (i = 0; i < ctrl->queue_count; i++) {
                clear_bit(NVME_RDMA_Q_CONNECTED, &ctrl->queues[i].flags);
+               clear_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[i].flags);
+       }
 
        if (ctrl->queue_count > 1)
                nvme_stop_queues(&ctrl->ctrl);
@@ -1378,6 +1384,24 @@ nvme_rdma_timeout(struct request *rq, bool reserved)
        return BLK_EH_HANDLED;
 }
 
+/*
+ * We cannot accept any other command until the Connect command has completed.
+ */
+static inline bool nvme_rdma_queue_is_ready(struct nvme_rdma_queue *queue,
+               struct request *rq)
+{
+       if (unlikely(!test_bit(NVME_RDMA_Q_LIVE, &queue->flags))) {
+               struct nvme_command *cmd = (struct nvme_command *)rq->cmd;
+
+               if (rq->cmd_type != REQ_TYPE_DRV_PRIV ||
+                   cmd->common.opcode != nvme_fabrics_command ||
+                   cmd->fabrics.fctype != nvme_fabrics_type_connect)
+                       return false;
+       }
+
+       return true;
+}
+
 static int nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx,
                const struct blk_mq_queue_data *bd)
 {
@@ -1394,6 +1418,9 @@ static int nvme_rdma_queue_rq(struct blk_mq_hw_ctx *hctx,
 
        WARN_ON_ONCE(rq->tag < 0);
 
+       if (!nvme_rdma_queue_is_ready(queue, rq))
+               return BLK_MQ_RQ_QUEUE_BUSY;
+
        dev = queue->device->dev;
        ib_dma_sync_single_for_cpu(dev, sqe->dma,
                        sizeof(struct nvme_command), DMA_TO_DEVICE);
@@ -1544,6 +1571,8 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl)
        if (error)
                goto out_cleanup_queue;
 
+       set_bit(NVME_RDMA_Q_LIVE, &ctrl->queues[0].flags);
+
        error = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->cap);
        if (error) {
                dev_err(ctrl->ctrl.device,