nvme-rdma: centralize controller setup sequence
authorSagi Grimberg <sagi@grimberg.me>
Mon, 9 Jul 2018 09:49:06 +0000 (12:49 +0300)
committerChristoph Hellwig <hch@lst.de>
Tue, 24 Jul 2018 13:55:49 +0000 (15:55 +0200)
Centralize controller sequence to a single routine that correctly cleans
up after failures instead of having multiple apperances in several flows
(create, reset, reconnect).

One thing that we also gain here are the sanity/boundary checks also
when connecting back to a dynamic controller.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
drivers/nvme/host/rdma.c

index 2b683b8d47637c3ebb24782a0f7091f3f8f41714..c22125c5661bdc546e7eb98d80d699224bc64a73 100644 (file)
@@ -917,24 +917,44 @@ static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl)
        }
 }
 
-static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
+static int nvme_rdma_setup_ctrl(struct nvme_rdma_ctrl *ctrl, bool new)
 {
-       struct nvme_rdma_ctrl *ctrl = container_of(to_delayed_work(work),
-                       struct nvme_rdma_ctrl, reconnect_work);
+       int ret = -EINVAL;
        bool changed;
-       int ret;
-
-       ++ctrl->ctrl.nr_reconnects;
 
-       ret = nvme_rdma_configure_admin_queue(ctrl, false);
+       ret = nvme_rdma_configure_admin_queue(ctrl, new);
        if (ret)
-               goto requeue;
+               return ret;
+
+       if (ctrl->ctrl.icdoff) {
+               dev_err(ctrl->ctrl.device, "icdoff is not supported!\n");
+               goto destroy_admin;
+       }
+
+       if (!(ctrl->ctrl.sgls & (1 << 2))) {
+               dev_err(ctrl->ctrl.device,
+                       "Mandatory keyed sgls are not supported!\n");
+               goto destroy_admin;
+       }
+
+       if (ctrl->ctrl.opts->queue_size > ctrl->ctrl.sqsize + 1) {
+               dev_warn(ctrl->ctrl.device,
+                       "queue_size %zu > ctrl sqsize %u, clamping down\n",
+                       ctrl->ctrl.opts->queue_size, ctrl->ctrl.sqsize + 1);
+       }
+
+       if (ctrl->ctrl.sqsize + 1 > ctrl->ctrl.maxcmd) {
+               dev_warn(ctrl->ctrl.device,
+                       "sqsize %u > ctrl maxcmd %u, clamping down\n",
+                       ctrl->ctrl.sqsize + 1, ctrl->ctrl.maxcmd);
+               ctrl->ctrl.sqsize = ctrl->ctrl.maxcmd - 1;
+       }
 
        if (ctrl->ctrl.sgls & (1 << 20))
                ctrl->use_inline_data = true;
 
        if (ctrl->ctrl.queue_count > 1) {
-               ret = nvme_rdma_configure_io_queues(ctrl, false);
+               ret = nvme_rdma_configure_io_queues(ctrl, new);
                if (ret)
                        goto destroy_admin;
        }
@@ -943,10 +963,31 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
        if (!changed) {
                /* state change failure is ok if we're in DELETING state */
                WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING);
-               return;
+               ret = -EINVAL;
+               goto destroy_io;
        }
 
        nvme_start_ctrl(&ctrl->ctrl);
+       return 0;
+
+destroy_io:
+       if (ctrl->ctrl.queue_count > 1)
+               nvme_rdma_destroy_io_queues(ctrl, new);
+destroy_admin:
+       nvme_rdma_stop_queue(&ctrl->queues[0]);
+       nvme_rdma_destroy_admin_queue(ctrl, new);
+       return ret;
+}
+
+static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
+{
+       struct nvme_rdma_ctrl *ctrl = container_of(to_delayed_work(work),
+                       struct nvme_rdma_ctrl, reconnect_work);
+
+       ++ctrl->ctrl.nr_reconnects;
+
+       if (nvme_rdma_setup_ctrl(ctrl, false))
+               goto requeue;
 
        dev_info(ctrl->ctrl.device, "Successfully reconnected (%d attempts)\n",
                        ctrl->ctrl.nr_reconnects);
@@ -955,9 +996,6 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
 
        return;
 
-destroy_admin:
-       nvme_rdma_stop_queue(&ctrl->queues[0]);
-       nvme_rdma_destroy_admin_queue(ctrl, false);
 requeue:
        dev_info(ctrl->ctrl.device, "Failed reconnect attempt %d\n",
                        ctrl->ctrl.nr_reconnects);
@@ -1786,8 +1824,6 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
 {
        struct nvme_rdma_ctrl *ctrl =
                container_of(work, struct nvme_rdma_ctrl, ctrl.reset_work);
-       int ret;
-       bool changed;
 
        nvme_stop_ctrl(&ctrl->ctrl);
        nvme_rdma_shutdown_ctrl(ctrl, false);
@@ -1798,25 +1834,9 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
                return;
        }
 
-       ret = nvme_rdma_configure_admin_queue(ctrl, false);
-       if (ret)
+       if (nvme_rdma_setup_ctrl(ctrl, false))
                goto out_fail;
 
-       if (ctrl->ctrl.queue_count > 1) {
-               ret = nvme_rdma_configure_io_queues(ctrl, false);
-               if (ret)
-                       goto out_fail;
-       }
-
-       changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
-       if (!changed) {
-               /* state change failure is ok if we're in DELETING state */
-               WARN_ON_ONCE(ctrl->ctrl.state != NVME_CTRL_DELETING);
-               return;
-       }
-
-       nvme_start_ctrl(&ctrl->ctrl);
-
        return;
 
 out_fail:
@@ -1979,49 +1999,10 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev,
        changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_CONNECTING);
        WARN_ON_ONCE(!changed);
 
-       ret = nvme_rdma_configure_admin_queue(ctrl, true);
+       ret = nvme_rdma_setup_ctrl(ctrl, true);
        if (ret)
                goto out_uninit_ctrl;
 
-       /* sanity check icdoff */
-       if (ctrl->ctrl.icdoff) {
-               dev_err(ctrl->ctrl.device, "icdoff is not supported!\n");
-               ret = -EINVAL;
-               goto out_remove_admin_queue;
-       }
-
-       /* sanity check keyed sgls */
-       if (!(ctrl->ctrl.sgls & (1 << 2))) {
-               dev_err(ctrl->ctrl.device,
-                       "Mandatory keyed sgls are not supported!\n");
-               ret = -EINVAL;
-               goto out_remove_admin_queue;
-       }
-
-       /* only warn if argument is too large here, will clamp later */
-       if (opts->queue_size > ctrl->ctrl.sqsize + 1) {
-               dev_warn(ctrl->ctrl.device,
-                       "queue_size %zu > ctrl sqsize %u, clamping down\n",
-                       opts->queue_size, ctrl->ctrl.sqsize + 1);
-       }
-
-       /* warn if maxcmd is lower than sqsize+1 */
-       if (ctrl->ctrl.sqsize + 1 > ctrl->ctrl.maxcmd) {
-               dev_warn(ctrl->ctrl.device,
-                       "sqsize %u > ctrl maxcmd %u, clamping down\n",
-                       ctrl->ctrl.sqsize + 1, ctrl->ctrl.maxcmd);
-               ctrl->ctrl.sqsize = ctrl->ctrl.maxcmd - 1;
-       }
-
-       if (opts->nr_io_queues) {
-               ret = nvme_rdma_configure_io_queues(ctrl, true);
-               if (ret)
-                       goto out_remove_admin_queue;
-       }
-
-       changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
-       WARN_ON_ONCE(!changed);
-
        dev_info(ctrl->ctrl.device, "new ctrl: NQN \"%s\", addr %pISpcs\n",
                ctrl->ctrl.opts->subsysnqn, &ctrl->addr);
 
@@ -2031,13 +2012,8 @@ static struct nvme_ctrl *nvme_rdma_create_ctrl(struct device *dev,
        list_add_tail(&ctrl->list, &nvme_rdma_ctrl_list);
        mutex_unlock(&nvme_rdma_ctrl_mutex);
 
-       nvme_start_ctrl(&ctrl->ctrl);
-
        return &ctrl->ctrl;
 
-out_remove_admin_queue:
-       nvme_rdma_stop_queue(&ctrl->queues[0]);
-       nvme_rdma_destroy_admin_queue(ctrl, true);
 out_uninit_ctrl:
        nvme_uninit_ctrl(&ctrl->ctrl);
        nvme_put_ctrl(&ctrl->ctrl);