nvme-rdma: use the tagset alloc/free helpers
authorChristoph Hellwig <hch@lst.de>
Tue, 20 Sep 2022 15:14:53 +0000 (17:14 +0200)
committerChristoph Hellwig <hch@lst.de>
Tue, 27 Sep 2022 12:44:16 +0000 (14:44 +0200)
Use the common helpers to allocate and free the tagsets.  To make this
work the generic nvme_ctrl now needs to be stored in the hctx private
data instead of the nvme_rdma_ctrl.

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

index 8bc2930..5ad0ab2 100644 (file)
@@ -788,64 +788,21 @@ out_free_queues:
        return ret;
 }
 
-static int nvme_rdma_alloc_admin_tag_set(struct nvme_ctrl *nctrl)
+static int nvme_rdma_alloc_tag_set(struct nvme_ctrl *ctrl)
 {
-       struct nvme_rdma_ctrl *ctrl = to_rdma_ctrl(nctrl);
-       struct blk_mq_tag_set *set = &ctrl->admin_tag_set;
-       int ret;
+       unsigned int cmd_size = sizeof(struct nvme_rdma_request) +
+                               NVME_RDMA_DATA_SGL_SIZE;
 
-       memset(set, 0, sizeof(*set));
-       set->ops = &nvme_rdma_admin_mq_ops;
-       set->queue_depth = NVME_AQ_MQ_TAG_DEPTH;
-       set->reserved_tags = NVMF_RESERVED_TAGS;
-       set->numa_node = nctrl->numa_node;
-       set->cmd_size = sizeof(struct nvme_rdma_request) +
-                       NVME_RDMA_DATA_SGL_SIZE;
-       set->driver_data = &ctrl->ctrl;
-       set->nr_hw_queues = 1;
-       set->timeout = NVME_ADMIN_TIMEOUT;
-       set->flags = BLK_MQ_F_NO_SCHED;
-       ret = blk_mq_alloc_tag_set(set);
-       if (!ret)
-               ctrl->ctrl.admin_tagset = set;
-       return ret;
-}
-
-static int nvme_rdma_alloc_tag_set(struct nvme_ctrl *nctrl)
-{
-       struct nvme_rdma_ctrl *ctrl = to_rdma_ctrl(nctrl);
-       struct blk_mq_tag_set *set = &ctrl->tag_set;
-       int ret;
+       if (ctrl->max_integrity_segments)
+               cmd_size += sizeof(struct nvme_rdma_sgl) +
+                           NVME_RDMA_METADATA_SGL_SIZE;
 
-       memset(set, 0, sizeof(*set));
-       set->ops = &nvme_rdma_mq_ops;
-       set->queue_depth = nctrl->sqsize + 1;
-       set->reserved_tags = NVMF_RESERVED_TAGS;
-       set->numa_node = nctrl->numa_node;
-       set->flags = BLK_MQ_F_SHOULD_MERGE;
-       set->cmd_size = sizeof(struct nvme_rdma_request) +
-                       NVME_RDMA_DATA_SGL_SIZE;
-       if (nctrl->max_integrity_segments)
-               set->cmd_size += sizeof(struct nvme_rdma_sgl) +
-                                NVME_RDMA_METADATA_SGL_SIZE;
-       set->driver_data = &ctrl->ctrl;
-       set->nr_hw_queues = nctrl->queue_count - 1;
-       set->timeout = NVME_IO_TIMEOUT;
-       set->nr_maps = nctrl->opts->nr_poll_queues ? HCTX_MAX_TYPES : 2;
-       ret = blk_mq_alloc_tag_set(set);
-       if (!ret)
-               ctrl->ctrl.tagset = set;
-       return ret;
+       return nvme_alloc_io_tag_set(ctrl, &to_rdma_ctrl(ctrl)->tag_set,
+                       &nvme_rdma_mq_ops, BLK_MQ_F_SHOULD_MERGE, cmd_size);
 }
 
-static void nvme_rdma_destroy_admin_queue(struct nvme_rdma_ctrl *ctrl,
-               bool remove)
+static void nvme_rdma_destroy_admin_queue(struct nvme_rdma_ctrl *ctrl)
 {
-       if (remove) {
-               blk_mq_destroy_queue(ctrl->ctrl.admin_q);
-               blk_mq_destroy_queue(ctrl->ctrl.fabrics_q);
-               blk_mq_free_tag_set(ctrl->ctrl.admin_tagset);
-       }
        if (ctrl->async_event_sqe.data) {
                cancel_work_sync(&ctrl->ctrl.async_event_work);
                nvme_rdma_free_qe(ctrl->device->dev, &ctrl->async_event_sqe,
@@ -887,26 +844,19 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl,
                goto out_free_queue;
 
        if (new) {
-               error = nvme_rdma_alloc_admin_tag_set(&ctrl->ctrl);
+               error = nvme_alloc_admin_tag_set(&ctrl->ctrl,
+                               &ctrl->admin_tag_set, &nvme_rdma_admin_mq_ops,
+                               BLK_MQ_F_NO_SCHED,
+                               sizeof(struct nvme_rdma_request) +
+                               NVME_RDMA_DATA_SGL_SIZE);
                if (error)
                        goto out_free_async_qe;
 
-               ctrl->ctrl.fabrics_q = blk_mq_init_queue(&ctrl->admin_tag_set);
-               if (IS_ERR(ctrl->ctrl.fabrics_q)) {
-                       error = PTR_ERR(ctrl->ctrl.fabrics_q);
-                       goto out_free_tagset;
-               }
-
-               ctrl->ctrl.admin_q = blk_mq_init_queue(&ctrl->admin_tag_set);
-               if (IS_ERR(ctrl->ctrl.admin_q)) {
-                       error = PTR_ERR(ctrl->ctrl.admin_q);
-                       goto out_cleanup_fabrics_q;
-               }
        }
 
        error = nvme_rdma_start_queue(ctrl, 0);
        if (error)
-               goto out_cleanup_queue;
+               goto out_remove_admin_tag_set;
 
        error = nvme_enable_ctrl(&ctrl->ctrl);
        if (error)
@@ -933,15 +883,9 @@ out_quiesce_queue:
 out_stop_queue:
        nvme_rdma_stop_queue(&ctrl->queues[0]);
        nvme_cancel_admin_tagset(&ctrl->ctrl);
-out_cleanup_queue:
+out_remove_admin_tag_set:
        if (new)
-               blk_mq_destroy_queue(ctrl->ctrl.admin_q);
-out_cleanup_fabrics_q:
-       if (new)
-               blk_mq_destroy_queue(ctrl->ctrl.fabrics_q);
-out_free_tagset:
-       if (new)
-               blk_mq_free_tag_set(ctrl->ctrl.admin_tagset);
+               nvme_remove_admin_tag_set(&ctrl->ctrl);
 out_free_async_qe:
        if (ctrl->async_event_sqe.data) {
                nvme_rdma_free_qe(ctrl->device->dev, &ctrl->async_event_sqe,
@@ -953,16 +897,6 @@ out_free_queue:
        return error;
 }
 
-static void nvme_rdma_destroy_io_queues(struct nvme_rdma_ctrl *ctrl,
-               bool remove)
-{
-       if (remove) {
-               blk_mq_destroy_queue(ctrl->ctrl.connect_q);
-               blk_mq_free_tag_set(ctrl->ctrl.tagset);
-       }
-       nvme_rdma_free_io_queues(ctrl);
-}
-
 static int nvme_rdma_configure_io_queues(struct nvme_rdma_ctrl *ctrl, bool new)
 {
        int ret, nr_queues;
@@ -975,10 +909,6 @@ static int nvme_rdma_configure_io_queues(struct nvme_rdma_ctrl *ctrl, bool new)
                ret = nvme_rdma_alloc_tag_set(&ctrl->ctrl);
                if (ret)
                        goto out_free_io_queues;
-
-               ret = nvme_ctrl_init_connect_q(&(ctrl->ctrl));
-               if (ret)
-                       goto out_free_tag_set;
        }
 
        /*
@@ -989,7 +919,7 @@ static int nvme_rdma_configure_io_queues(struct nvme_rdma_ctrl *ctrl, bool new)
        nr_queues = min(ctrl->tag_set.nr_hw_queues + 1, ctrl->ctrl.queue_count);
        ret = nvme_rdma_start_io_queues(ctrl, 1, nr_queues);
        if (ret)
-               goto out_cleanup_connect_q;
+               goto out_cleanup_tagset;
 
        if (!new) {
                nvme_start_queues(&ctrl->ctrl);
@@ -1022,13 +952,10 @@ out_wait_freeze_timed_out:
        nvme_stop_queues(&ctrl->ctrl);
        nvme_sync_io_queues(&ctrl->ctrl);
        nvme_rdma_stop_io_queues(ctrl);
-out_cleanup_connect_q:
+out_cleanup_tagset:
        nvme_cancel_tagset(&ctrl->ctrl);
        if (new)
-               blk_mq_destroy_queue(ctrl->ctrl.connect_q);
-out_free_tag_set:
-       if (new)
-               blk_mq_free_tag_set(ctrl->ctrl.tagset);
+               nvme_remove_io_tag_set(&ctrl->ctrl);
 out_free_io_queues:
        nvme_rdma_free_io_queues(ctrl);
        return ret;
@@ -1041,9 +968,11 @@ static void nvme_rdma_teardown_admin_queue(struct nvme_rdma_ctrl *ctrl,
        blk_sync_queue(ctrl->ctrl.admin_q);
        nvme_rdma_stop_queue(&ctrl->queues[0]);
        nvme_cancel_admin_tagset(&ctrl->ctrl);
-       if (remove)
+       if (remove) {
                nvme_start_admin_queue(&ctrl->ctrl);
-       nvme_rdma_destroy_admin_queue(ctrl, remove);
+               nvme_remove_admin_tag_set(&ctrl->ctrl);
+       }
+       nvme_rdma_destroy_admin_queue(ctrl);
 }
 
 static void nvme_rdma_teardown_io_queues(struct nvme_rdma_ctrl *ctrl,
@@ -1055,9 +984,11 @@ static void nvme_rdma_teardown_io_queues(struct nvme_rdma_ctrl *ctrl,
                nvme_sync_io_queues(&ctrl->ctrl);
                nvme_rdma_stop_io_queues(ctrl);
                nvme_cancel_tagset(&ctrl->ctrl);
-               if (remove)
+               if (remove) {
                        nvme_start_queues(&ctrl->ctrl);
-               nvme_rdma_destroy_io_queues(ctrl, remove);
+                       nvme_remove_io_tag_set(&ctrl->ctrl);
+               }
+               nvme_rdma_free_io_queues(ctrl);
        }
 }
 
@@ -1179,14 +1110,18 @@ destroy_io:
                nvme_sync_io_queues(&ctrl->ctrl);
                nvme_rdma_stop_io_queues(ctrl);
                nvme_cancel_tagset(&ctrl->ctrl);
-               nvme_rdma_destroy_io_queues(ctrl, new);
+               if (new)
+                       nvme_remove_io_tag_set(&ctrl->ctrl);
+               nvme_rdma_free_io_queues(ctrl);
        }
 destroy_admin:
        nvme_stop_admin_queue(&ctrl->ctrl);
        blk_sync_queue(ctrl->ctrl.admin_q);
        nvme_rdma_stop_queue(&ctrl->queues[0]);
        nvme_cancel_admin_tagset(&ctrl->ctrl);
-       nvme_rdma_destroy_admin_queue(ctrl, new);
+       if (new)
+               nvme_remove_admin_tag_set(&ctrl->ctrl);
+       nvme_rdma_destroy_admin_queue(ctrl);
        return ret;
 }