nvme_fc: add controller reset support

This patch actually does quite a few things.  When looking to add
controller reset support, the organization modeled after rdma was
very fragmented. rdma duplicates the reset and teardown paths and does
different things to the block layer on the two paths. The code to build
up the controller is also duplicated between the initial creation and
the reset/error recovery paths. So I decided to make this sane.

I reorganized the controller creation and teardown so that there is a
connect path and a disconnect path.  Initial creation obviously uses
the connect path.  Controller teardown will use the disconnect path,
followed last access code. Controller reset will use the disconnect
path to stop operation, and then the connect path to re-establish
the controller.

Along the way, several things were fixed
- aens were not properly set up. They are allocated differently from
  the per-request structure on the blk queues.
- aens were oddly torn down. the prior patch corrected to abort, but
  we still need to dma unmap and free relative elements.
- missed a few ref counting points: in aen completion and on i/o's
  that fail
- controller initial create failure paths were still confused vs teardown
  before converting to ref counting vs after we convert to refcounting.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

1 files changed, 616 insertions, 315 deletions

diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c
index b7118db11ea7..5aa52863ba54 100644
--- a/drivers/nvme/host/fc.c
+++ b/drivers/nvme/host/fc.c
@@ -19,6 +19,7 @@
 #include <linux/parser.h>
 #include <uapi/scsi/fc/fc_fs.h>
 #include <uapi/scsi/fc/fc_els.h>
+#include <linux/delay.h>
 
 #include "nvme.h"
 #include "fabrics.h"
@@ -44,6 +45,8 @@ enum nvme_fc_queue_flags {
 
 #define NVMEFC_QUEUE_DELAY	3		/* ms units */
 
+#define NVME_FC_MAX_CONNECT_ATTEMPTS	1
+
 struct nvme_fc_queue {
 	struct nvme_fc_ctrl	*ctrl;
 	struct device		*dev;
@@ -137,19 +140,17 @@ struct nvme_fc_rport {
 	struct kref			ref;
 } __aligned(sizeof(u64));	/* alignment for other things alloc'd with */
 
-enum nvme_fcctrl_state {
-	FCCTRL_INIT		= 0,
-	FCCTRL_ACTIVE		= 1,
+enum nvme_fcctrl_flags {
+	FCCTRL_TERMIO		= (1 << 0),
 };
 
 struct nvme_fc_ctrl {
 	spinlock_t		lock;
 	struct nvme_fc_queue	*queues;
-	u32			queue_count;
-
 	struct device		*dev;
 	struct nvme_fc_lport	*lport;
 	struct nvme_fc_rport	*rport;
+	u32			queue_count;
 	u32			cnum;
 
 	u64			association_id;
@@ -162,8 +163,14 @@ struct nvme_fc_ctrl {
 	struct blk_mq_tag_set	tag_set;
 
 	struct work_struct	delete_work;
+	struct work_struct	reset_work;
+	struct delayed_work	connect_work;
+	int			reconnect_delay;
+	int			connect_attempts;
+
 	struct kref		ref;
-	int			state;
+	u32			flags;
+	u32			iocnt;
 
 	struct nvme_fc_fcp_op	aen_ops[NVME_FC_NR_AEN_COMMANDS];
 
@@ -1204,7 +1211,10 @@ nvme_fc_abort_aen_ops(struct nvme_fc_ctrl *ctrl)
 			continue;
 
 		spin_lock_irqsave(&ctrl->lock, flags);
-		aen_op->flags |= FCOP_FLAGS_TERMIO;
+		if (ctrl->flags & FCCTRL_TERMIO) {
+			ctrl->iocnt++;
+			aen_op->flags |= FCOP_FLAGS_TERMIO;
+		}
 		spin_unlock_irqrestore(&ctrl->lock, flags);
 
 		ret = __nvme_fc_abort_op(ctrl, aen_op);
@@ -1217,6 +1227,8 @@ nvme_fc_abort_aen_ops(struct nvme_fc_ctrl *ctrl)
 
 			/* back out the flags/counters */
 			spin_lock_irqsave(&ctrl->lock, flags);
+			if (ctrl->flags & FCCTRL_TERMIO)
+				ctrl->iocnt--;
 			aen_op->flags &= ~FCOP_FLAGS_TERMIO;
 			spin_unlock_irqrestore(&ctrl->lock, flags);
 			return;
@@ -1232,6 +1244,10 @@ __nvme_fc_fcpop_chk_teardowns(struct nvme_fc_ctrl *ctrl,
 	bool complete_rq = false;
 
 	spin_lock_irqsave(&ctrl->lock, flags);
+	if (unlikely(op->flags & FCOP_FLAGS_TERMIO)) {
+		if (ctrl->flags & FCCTRL_TERMIO)
+			ctrl->iocnt--;
+	}
 	if (op->flags & FCOP_FLAGS_RELEASED)
 		complete_rq = true;
 	else
@@ -1447,19 +1463,29 @@ nvme_fc_init_aen_ops(struct nvme_fc_ctrl *ctrl)
 	struct nvme_fc_fcp_op *aen_op;
 	struct nvme_fc_cmd_iu *cmdiu;
 	struct nvme_command *sqe;
+	void *private;
 	int i, ret;
 
 	aen_op = ctrl->aen_ops;
 	for (i = 0; i < NVME_FC_NR_AEN_COMMANDS; i++, aen_op++) {
+		private = kzalloc(ctrl->lport->ops->fcprqst_priv_sz,
+						GFP_KERNEL);
+		if (!private)
+			return -ENOMEM;
+
 		cmdiu = &aen_op->cmd_iu;
 		sqe = &cmdiu->sqe;
 		ret = __nvme_fc_init_request(ctrl, &ctrl->queues[0],
 				aen_op, (struct request *)NULL,
 				(AEN_CMDID_BASE + i));
-		if (ret)
+		if (ret) {
+			kfree(private);
 			return ret;
+		}
 
 		aen_op->flags = FCOP_FLAGS_AEN;
+		aen_op->fcp_req.first_sgl = NULL; /* no sg list */
+		aen_op->fcp_req.private = private;
 
 		memset(sqe, 0, sizeof(*sqe));
 		sqe->common.opcode = nvme_admin_async_event;
@@ -1469,6 +1495,23 @@ nvme_fc_init_aen_ops(struct nvme_fc_ctrl *ctrl)
 	return 0;
 }
 
+static void
+nvme_fc_term_aen_ops(struct nvme_fc_ctrl *ctrl)
+{
+	struct nvme_fc_fcp_op *aen_op;
+	int i;
+
+	aen_op = ctrl->aen_ops;
+	for (i = 0; i < NVME_FC_NR_AEN_COMMANDS; i++, aen_op++) {
+		if (!aen_op->fcp_req.private)
+			continue;
+
+		__nvme_fc_exit_request(ctrl, aen_op);
+
+		kfree(aen_op->fcp_req.private);
+		aen_op->fcp_req.private = NULL;
+	}
+}
 
 static inline void
 __nvme_fc_init_hctx(struct blk_mq_hw_ctx *hctx, struct nvme_fc_ctrl *ctrl,
@@ -1568,15 +1611,6 @@ __nvme_fc_delete_hw_queue(struct nvme_fc_ctrl *ctrl,
 }
 
 static void
-nvme_fc_destroy_admin_queue(struct nvme_fc_ctrl *ctrl)
-{
-	__nvme_fc_delete_hw_queue(ctrl, &ctrl->queues[0], 0);
-	blk_cleanup_queue(ctrl->ctrl.admin_q);
-	blk_mq_free_tag_set(&ctrl->admin_tag_set);
-	nvme_fc_free_queue(&ctrl->queues[0]);
-}
-
-static void
 nvme_fc_free_io_queues(struct nvme_fc_ctrl *ctrl)
 {
 	int i;
@@ -1663,17 +1697,24 @@ nvme_fc_ctrl_free(struct kref *ref)
 		container_of(ref, struct nvme_fc_ctrl, ref);
 	unsigned long flags;
 
-	if (ctrl->state != FCCTRL_INIT) {
-		/* remove from rport list */
-		spin_lock_irqsave(&ctrl->rport->lock, flags);
-		list_del(&ctrl->ctrl_list);
-		spin_unlock_irqrestore(&ctrl->rport->lock, flags);
+	if (ctrl->ctrl.tagset) {
+		blk_cleanup_queue(ctrl->ctrl.connect_q);
+		blk_mq_free_tag_set(&ctrl->tag_set);
 	}
 
+	/* remove from rport list */
+	spin_lock_irqsave(&ctrl->rport->lock, flags);
+	list_del(&ctrl->ctrl_list);
+	spin_unlock_irqrestore(&ctrl->rport->lock, flags);
+
+	blk_cleanup_queue(ctrl->ctrl.admin_q);
+	blk_mq_free_tag_set(&ctrl->admin_tag_set);
+
+	kfree(ctrl->queues);
+
 	put_device(ctrl->dev);
 	nvme_fc_rport_put(ctrl->rport);
 
-	kfree(ctrl->queues);
 	ida_simple_remove(&nvme_fc_ctrl_cnt, ctrl->cnum);
 	nvmf_free_options(ctrl->ctrl.opts);
 	kfree(ctrl);
@@ -1696,32 +1737,35 @@ nvme_fc_ctrl_get(struct nvme_fc_ctrl *ctrl)
  * controller. Called after last nvme_put_ctrl() call
  */
 static void
-nvme_fc_free_nvme_ctrl(struct nvme_ctrl *nctrl)
+nvme_fc_nvme_ctrl_freed(struct nvme_ctrl *nctrl)
 {
 	struct nvme_fc_ctrl *ctrl = to_fc_ctrl(nctrl);
 
 	WARN_ON(nctrl != &ctrl->ctrl);
 
-	/*
-	 * Tear down the association, which will generate link
-	 * traffic to terminate connections
-	 */
-
-	if (ctrl->state != FCCTRL_INIT) {
-		/* send a Disconnect(association) LS to fc-nvme target */
-		nvme_fc_xmt_disconnect_assoc(ctrl);
+	nvme_fc_ctrl_put(ctrl);
+}
 
-		if (ctrl->ctrl.tagset) {
-			blk_cleanup_queue(ctrl->ctrl.connect_q);
-			blk_mq_free_tag_set(&ctrl->tag_set);
-			nvme_fc_delete_hw_io_queues(ctrl);
-			nvme_fc_free_io_queues(ctrl);
-		}
+static void
+nvme_fc_error_recovery(struct nvme_fc_ctrl *ctrl, char *errmsg)
+{
+	dev_warn(ctrl->ctrl.device,
+		"NVME-FC{%d}: transport association error detected: %s\n",
+		ctrl->cnum, errmsg);
+	dev_info(ctrl->ctrl.device,
+		"NVME-FC{%d}: resetting controller\n", ctrl->cnum);
 
-		nvme_fc_destroy_admin_queue(ctrl);
+	if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RECONNECTING)) {
+		dev_err(ctrl->ctrl.device,
+			"NVME-FC{%d}: error_recovery: Couldn't change state "
+			"to RECONNECTING\n", ctrl->cnum);
+		return;
 	}
 
-	nvme_fc_ctrl_put(ctrl);
+	if (!queue_work(nvme_fc_wq, &ctrl->reset_work))
+		dev_err(ctrl->ctrl.device,
+			"NVME-FC{%d}: error_recovery: Failed to schedule "
+			"reset work\n", ctrl->cnum);
 }
 
 enum blk_eh_timer_return
@@ -1740,11 +1784,13 @@ nvme_fc_timeout(struct request *rq, bool reserved)
 		return BLK_EH_HANDLED;
 
 	/*
-	 * TODO: force a controller reset
-	 *   when that happens, queues will be torn down and outstanding
-	 *   ios will be terminated, and the above abort, on a single io
-	 *   will no longer be needed.
+	 * we can't individually ABTS an io without affecting the queue,
+	 * thus killing the queue, adn thus the association.
+	 * So resolve by performing a controller reset, which will stop
+	 * the host/io stack, terminate the association on the link,
+	 * and recreate an association on the link.
 	 */
+	nvme_fc_error_recovery(ctrl, "io timeout error");
 
 	return BLK_EH_HANDLED;
 }
@@ -1838,6 +1884,13 @@ nvme_fc_start_fcp_op(struct nvme_fc_ctrl *ctrl, struct nvme_fc_queue *queue,
 	u32 csn;
 	int ret;
 
+	/*
+	 * before attempting to send the io, check to see if we believe
+	 * the target device is present
+	 */
+	if (ctrl->rport->remoteport.port_state != FC_OBJSTATE_ONLINE)
+		return BLK_MQ_RQ_QUEUE_ERROR;
+
 	if (!nvme_fc_ctrl_get(ctrl))
 		return BLK_MQ_RQ_QUEUE_ERROR;
 
@@ -1885,8 +1938,6 @@ nvme_fc_start_fcp_op(struct nvme_fc_ctrl *ctrl, struct nvme_fc_queue *queue,
 	if (!(op->flags & FCOP_FLAGS_AEN)) {
 		ret = nvme_fc_map_data(ctrl, op->rq, op);
 		if (ret < 0) {
-			dev_err(queue->ctrl->ctrl.device,
-			     "Failed to map data (%d)\n", ret);
 			nvme_cleanup_cmd(op->rq);
 			nvme_fc_ctrl_put(ctrl);
 			return (ret == -ENOMEM || ret == -EAGAIN) ?
@@ -1907,9 +1958,6 @@ nvme_fc_start_fcp_op(struct nvme_fc_ctrl *ctrl, struct nvme_fc_queue *queue,
 					queue->lldd_handle, &op->fcp_req);
 
 	if (ret) {
-		dev_err(ctrl->dev,
-			"Send nvme command failed - lldd returned %d.\n", ret);
-
 		if (op->rq) {			/* normal request */
 			nvme_fc_unmap_data(ctrl, op->rq, op);
 			nvme_cleanup_cmd(op->rq);
@@ -1979,12 +2027,8 @@ nvme_fc_poll(struct blk_mq_hw_ctx *hctx, unsigned int tag)
 	struct nvme_fc_fcp_op *op;
 
 	req = blk_mq_tag_to_rq(nvme_fc_tagset(queue), tag);
-	if (!req) {
-		dev_err(queue->ctrl->ctrl.device,
-			 "tag 0x%x on QNum %#x not found\n",
-			tag, queue->qnum);
+	if (!req)
 		return 0;
-	}
 
 	op = blk_mq_rq_to_pdu(req);
 
@@ -2001,11 +2045,21 @@ nvme_fc_submit_async_event(struct nvme_ctrl *arg, int aer_idx)
 {
 	struct nvme_fc_ctrl *ctrl = to_fc_ctrl(arg);
 	struct nvme_fc_fcp_op *aen_op;
+	unsigned long flags;
+	bool terminating = false;
 	int ret;
 
 	if (aer_idx > NVME_FC_NR_AEN_COMMANDS)
 		return;
 
+	spin_lock_irqsave(&ctrl->lock, flags);
+	if (ctrl->flags & FCCTRL_TERMIO)
+		terminating = true;
+	spin_unlock_irqrestore(&ctrl->lock, flags);
+
+	if (terminating)
+		return;
+
 	aen_op = &ctrl->aen_ops[aer_idx];
 
 	ret = nvme_fc_start_fcp_op(ctrl, aen_op->queue, aen_op, 0,
@@ -2059,6 +2113,57 @@ nvme_fc_complete_rq(struct request *rq)
 		__nvme_fc_final_op_cleanup(rq);
 }
 
+/*
+ * This routine is used by the transport when it needs to find active
+ * io on a queue that is to be terminated. The transport uses
+ * blk_mq_tagset_busy_itr() to find the busy requests, which then invoke
+ * this routine to kill them on a 1 by 1 basis.
+ *
+ * As FC allocates FC exchange for each io, the transport must contact
+ * the LLDD to terminate the exchange, thus releasing the FC exchange.
+ * After terminating the exchange the LLDD will call the transport's
+ * normal io done path for the request, but it will have an aborted
+ * status. The done path will return the io request back to the block
+ * layer with an error status.
+ */
+static void
+nvme_fc_terminate_exchange(struct request *req, void *data, bool reserved)
+{
+	struct nvme_ctrl *nctrl = data;
+	struct nvme_fc_ctrl *ctrl = to_fc_ctrl(nctrl);
+	struct nvme_fc_fcp_op *op = blk_mq_rq_to_pdu(req);
+	unsigned long flags;
+	int status;
+
+	if (!blk_mq_request_started(req))
+		return;
+
+	spin_lock_irqsave(&ctrl->lock, flags);
+	if (ctrl->flags & FCCTRL_TERMIO) {
+		ctrl->iocnt++;
+		op->flags |= FCOP_FLAGS_TERMIO;
+	}
+	spin_unlock_irqrestore(&ctrl->lock, flags);
+
+	status = __nvme_fc_abort_op(ctrl, op);
+	if (status) {
+		/*
+		 * if __nvme_fc_abort_op failed the io wasn't
+		 * active. Thus this call path is running in
+		 * parallel to the io complete. Treat as non-error.
+		 */
+
+		/* back out the flags/counters */
+		spin_lock_irqsave(&ctrl->lock, flags);
+		if (ctrl->flags & FCCTRL_TERMIO)
+			ctrl->iocnt--;
+		op->flags &= ~FCOP_FLAGS_TERMIO;
+		spin_unlock_irqrestore(&ctrl->lock, flags);
+		return;
+	}
+}
+
+
 static const struct blk_mq_ops nvme_fc_mq_ops = {
 	.queue_rq	= nvme_fc_queue_rq,
 	.complete	= nvme_fc_complete_rq,
@@ -2070,152 +2175,275 @@ static const struct blk_mq_ops nvme_fc_mq_ops = {
 	.timeout	= nvme_fc_timeout,
 };
 
-static const struct blk_mq_ops nvme_fc_admin_mq_ops = {
-	.queue_rq	= nvme_fc_queue_rq,
-	.complete	= nvme_fc_complete_rq,
-	.init_request	= nvme_fc_init_admin_request,
-	.exit_request	= nvme_fc_exit_request,
-	.reinit_request	= nvme_fc_reinit_request,
-	.init_hctx	= nvme_fc_init_admin_hctx,
-	.timeout	= nvme_fc_timeout,
-};
-
 static int
-nvme_fc_configure_admin_queue(struct nvme_fc_ctrl *ctrl)
+nvme_fc_create_io_queues(struct nvme_fc_ctrl *ctrl)
 {
-	u32 segs;
-	int error;
+	struct nvmf_ctrl_options *opts = ctrl->ctrl.opts;
+	int ret;
 
-	nvme_fc_init_queue(ctrl, 0, NVME_FC_AQ_BLKMQ_DEPTH);
+	ret = nvme_set_queue_count(&ctrl->ctrl, &opts->nr_io_queues);
+	if (ret) {
+		dev_info(ctrl->ctrl.device,
+			"set_queue_count failed: %d\n", ret);
+		return ret;
+	}
 
-	error = nvme_fc_connect_admin_queue(ctrl, &ctrl->queues[0],
-				NVME_FC_AQ_BLKMQ_DEPTH,
-				(NVME_FC_AQ_BLKMQ_DEPTH / 4));
-	if (error)
-		return error;
+	ctrl->queue_count = opts->nr_io_queues + 1;
+	if (!opts->nr_io_queues)
+		return 0;
 
-	memset(&ctrl->admin_tag_set, 0, sizeof(ctrl->admin_tag_set));
-	ctrl->admin_tag_set.ops = &nvme_fc_admin_mq_ops;
-	ctrl->admin_tag_set.queue_depth = NVME_FC_AQ_BLKMQ_DEPTH;
-	ctrl->admin_tag_set.reserved_tags = 2; /* fabric connect + Keep-Alive */
-	ctrl->admin_tag_set.numa_node = NUMA_NO_NODE;
-	ctrl->admin_tag_set.cmd_size = sizeof(struct nvme_fc_fcp_op) +
+	dev_info(ctrl->ctrl.device, "creating %d I/O queues.\n",
+			opts->nr_io_queues);
+
+	nvme_fc_init_io_queues(ctrl);
+
+	memset(&ctrl->tag_set, 0, sizeof(ctrl->tag_set));
+	ctrl->tag_set.ops = &nvme_fc_mq_ops;
+	ctrl->tag_set.queue_depth = ctrl->ctrl.opts->queue_size;
+	ctrl->tag_set.reserved_tags = 1; /* fabric connect */
+	ctrl->tag_set.numa_node = NUMA_NO_NODE;
+	ctrl->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
+	ctrl->tag_set.cmd_size = sizeof(struct nvme_fc_fcp_op) +
 					(SG_CHUNK_SIZE *
 						sizeof(struct scatterlist)) +
 					ctrl->lport->ops->fcprqst_priv_sz;
-	ctrl->admin_tag_set.driver_data = ctrl;
-	ctrl->admin_tag_set.nr_hw_queues = 1;
-	ctrl->admin_tag_set.timeout = ADMIN_TIMEOUT;
+	ctrl->tag_set.driver_data = ctrl;
+	ctrl->tag_set.nr_hw_queues = ctrl->queue_count - 1;
+	ctrl->tag_set.timeout = NVME_IO_TIMEOUT;
 
-	error = blk_mq_alloc_tag_set(&ctrl->admin_tag_set);
-	if (error)
-		goto out_free_queue;
+	ret = blk_mq_alloc_tag_set(&ctrl->tag_set);
+	if (ret)
+		return ret;
 
-	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_free_tagset;
+	ctrl->ctrl.tagset = &ctrl->tag_set;
+
+	ctrl->ctrl.connect_q = blk_mq_init_queue(&ctrl->tag_set);
+	if (IS_ERR(ctrl->ctrl.connect_q)) {
+		ret = PTR_ERR(ctrl->ctrl.connect_q);
+		goto out_free_tag_set;
+	}
+
+	ret = nvme_fc_create_hw_io_queues(ctrl, ctrl->ctrl.opts->queue_size);
+	if (ret)
+		goto out_cleanup_blk_queue;
+
+	ret = nvme_fc_connect_io_queues(ctrl, ctrl->ctrl.opts->queue_size);
+	if (ret)
+		goto out_delete_hw_queues;
+
+	return 0;
+
+out_delete_hw_queues:
+	nvme_fc_delete_hw_io_queues(ctrl);
+out_cleanup_blk_queue:
+	nvme_stop_keep_alive(&ctrl->ctrl);
+	blk_cleanup_queue(ctrl->ctrl.connect_q);
+out_free_tag_set:
+	blk_mq_free_tag_set(&ctrl->tag_set);
+	nvme_fc_free_io_queues(ctrl);
+
+	/* force put free routine to ignore io queues */
+	ctrl->ctrl.tagset = NULL;
+
+	return ret;
+}
+
+static int
+nvme_fc_reinit_io_queues(struct nvme_fc_ctrl *ctrl)
+{
+	struct nvmf_ctrl_options *opts = ctrl->ctrl.opts;
+	int ret;
+
+	ret = nvme_set_queue_count(&ctrl->ctrl, &opts->nr_io_queues);
+	if (ret) {
+		dev_info(ctrl->ctrl.device,
+			"set_queue_count failed: %d\n", ret);
+		return ret;
 	}
 
-	error = __nvme_fc_create_hw_queue(ctrl, &ctrl->queues[0], 0,
+	/* check for io queues existing */
+	if (ctrl->queue_count == 1)
+		return 0;
+
+	dev_info(ctrl->ctrl.device, "Recreating %d I/O queues.\n",
+			opts->nr_io_queues);
+
+	nvme_fc_init_io_queues(ctrl);
+
+	ret = blk_mq_reinit_tagset(&ctrl->tag_set);
+	if (ret)
+		goto out_free_io_queues;
+
+	ret = nvme_fc_create_hw_io_queues(ctrl, ctrl->ctrl.opts->queue_size);
+	if (ret)
+		goto out_free_io_queues;
+
+	ret = nvme_fc_connect_io_queues(ctrl, ctrl->ctrl.opts->queue_size);
+	if (ret)
+		goto out_delete_hw_queues;
+
+	return 0;
+
+out_delete_hw_queues:
+	nvme_fc_delete_hw_io_queues(ctrl);
+out_free_io_queues:
+	nvme_fc_free_io_queues(ctrl);
+	return ret;
+}
+
+/*
+ * This routine restarts the controller on the host side, and
+ * on the link side, recreates the controller association.
+ */
+static int
+nvme_fc_create_association(struct nvme_fc_ctrl *ctrl)
+{
+	struct nvmf_ctrl_options *opts = ctrl->ctrl.opts;
+	u32 segs;
+	int ret;
+	bool changed;
+
+	ctrl->connect_attempts++;
+
+	/*
+	 * Create the admin queue
+	 */
+
+	nvme_fc_init_queue(ctrl, 0, NVME_FC_AQ_BLKMQ_DEPTH);
+
+	ret = __nvme_fc_create_hw_queue(ctrl, &ctrl->queues[0], 0,
 				NVME_FC_AQ_BLKMQ_DEPTH);
-	if (error)
-		goto out_cleanup_queue;
+	if (ret)
+		goto out_free_queue;
 
-	error = nvmf_connect_admin_queue(&ctrl->ctrl);
-	if (error)
+	ret = nvme_fc_connect_admin_queue(ctrl, &ctrl->queues[0],
+				NVME_FC_AQ_BLKMQ_DEPTH,
+				(NVME_FC_AQ_BLKMQ_DEPTH / 4));
+	if (ret)
 		goto out_delete_hw_queue;
 
-	error = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->cap);
-	if (error) {
+	if (ctrl->ctrl.state != NVME_CTRL_NEW)
+		blk_mq_start_stopped_hw_queues(ctrl->ctrl.admin_q, true);
+
+	ret = nvmf_connect_admin_queue(&ctrl->ctrl);
+	if (ret)
+		goto out_disconnect_admin_queue;
+
+	/*
+	 * Check controller capabilities
+	 *
+	 * todo:- add code to check if ctrl attributes changed from
+	 * prior connection values
+	 */
+
+	ret = nvmf_reg_read64(&ctrl->ctrl, NVME_REG_CAP, &ctrl->cap);
+	if (ret) {
 		dev_err(ctrl->ctrl.device,
 			"prop_get NVME_REG_CAP failed\n");
-		goto out_delete_hw_queue;
+		goto out_disconnect_admin_queue;
 	}
 
 	ctrl->ctrl.sqsize =
 		min_t(int, NVME_CAP_MQES(ctrl->cap) + 1, ctrl->ctrl.sqsize);
 
-	error = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap);
-	if (error)
-		goto out_delete_hw_queue;
+	ret = nvme_enable_ctrl(&ctrl->ctrl, ctrl->cap);
+	if (ret)
+		goto out_disconnect_admin_queue;
 
 	segs = min_t(u32, NVME_FC_MAX_SEGMENTS,
 			ctrl->lport->ops->max_sgl_segments);
 	ctrl->ctrl.max_hw_sectors = (segs - 1) << (PAGE_SHIFT - 9);
 
-	error = nvme_init_identify(&ctrl->ctrl);
-	if (error)
-		goto out_delete_hw_queue;
+	ret = nvme_init_identify(&ctrl->ctrl);
+	if (ret)
+		goto out_disconnect_admin_queue;
+
+	/* sanity checks */
+
+	/* FC-NVME does not have other data in the capsule */
+	if (ctrl->ctrl.icdoff) {
+		dev_err(ctrl->ctrl.device, "icdoff %d is not supported!\n",
+				ctrl->ctrl.icdoff);
+		goto out_disconnect_admin_queue;
+	}
 
 	nvme_start_keep_alive(&ctrl->ctrl);
 
-	return 0;
+	/* FC-NVME supports normal SGL Data Block Descriptors */
+
+	if (opts->queue_size > ctrl->ctrl.maxcmd) {
+		/* warn if maxcmd is lower than queue_size */
+		dev_warn(ctrl->ctrl.device,
+			"queue_size %zu > ctrl maxcmd %u, reducing "
+			"to queue_size\n",
+			opts->queue_size, ctrl->ctrl.maxcmd);
+		opts->queue_size = ctrl->ctrl.maxcmd;
+	}
 
+	ret = nvme_fc_init_aen_ops(ctrl);
+	if (ret)
+		goto out_term_aen_ops;
+
+	/*
+	 * Create the io queues
+	 */
+
+	if (ctrl->queue_count > 1) {
+		if (ctrl->ctrl.state == NVME_CTRL_NEW)
+			ret = nvme_fc_create_io_queues(ctrl);
+		else
+			ret = nvme_fc_reinit_io_queues(ctrl);
+		if (ret)
+			goto out_term_aen_ops;
+	}
+
+	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
+	WARN_ON_ONCE(!changed);
+
+	ctrl->connect_attempts = 0;
+
+	kref_get(&ctrl->ctrl.kref);
+
+	if (ctrl->queue_count > 1) {
+		nvme_start_queues(&ctrl->ctrl);
+		nvme_queue_scan(&ctrl->ctrl);
+		nvme_queue_async_events(&ctrl->ctrl);
+	}
+
+	return 0;	/* Success */
+
+out_term_aen_ops:
+	nvme_fc_term_aen_ops(ctrl);
+	nvme_stop_keep_alive(&ctrl->ctrl);
+out_disconnect_admin_queue:
+	/* send a Disconnect(association) LS to fc-nvme target */
+	nvme_fc_xmt_disconnect_assoc(ctrl);
 out_delete_hw_queue:
 	__nvme_fc_delete_hw_queue(ctrl, &ctrl->queues[0], 0);
-out_cleanup_queue:
-	blk_cleanup_queue(ctrl->ctrl.admin_q);
-out_free_tagset:
-	blk_mq_free_tag_set(&ctrl->admin_tag_set);
 out_free_queue:
 	nvme_fc_free_queue(&ctrl->queues[0]);
-	return error;
+
+	return ret;
 }
 
 /*
- * This routine is used by the transport when it needs to find active
- * io on a queue that is to be terminated. The transport uses
- * blk_mq_tagset_busy_itr() to find the busy requests, which then invoke
- * this routine to kill them on a 1 by 1 basis.
- *
- * As FC allocates FC exchange for each io, the transport must contact
- * the LLDD to terminate the exchange, thus releasing the FC exchange.
- * After terminating the exchange the LLDD will call the transport's
- * normal io done path for the request, but it will have an aborted
- * status. The done path will return the io request back to the block
- * layer with an error status.
+ * This routine stops operation of the controller on the host side.
+ * On the host os stack side: Admin and IO queues are stopped,
+ *   outstanding ios on them terminated via FC ABTS.
+ * On the link side: the association is terminated.
  */
 static void
-nvme_fc_terminate_exchange(struct request *req, void *data, bool reserved)
+nvme_fc_delete_association(struct nvme_fc_ctrl *ctrl)
 {
-	struct nvme_ctrl *nctrl = data;
-	struct nvme_fc_ctrl *ctrl = to_fc_ctrl(nctrl);
-	struct nvme_fc_fcp_op *op = blk_mq_rq_to_pdu(req);
 	unsigned long flags;
-	int status;
 
-	if (!blk_mq_request_started(req))
-		return;
+	nvme_stop_keep_alive(&ctrl->ctrl);
 
 	spin_lock_irqsave(&ctrl->lock, flags);
-	op->flags |= FCOP_FLAGS_TERMIO;
+	ctrl->flags |= FCCTRL_TERMIO;
+	ctrl->iocnt = 0;
 	spin_unlock_irqrestore(&ctrl->lock, flags);
 
-	status = __nvme_fc_abort_op(ctrl, op);
-	if (status) {
-		/*
-		 * if __nvme_fc_abort_op failed the io wasn't
-		 * active. Thus this call path is running in
-		 * parallel to the io complete. Treat as non-error.
-		 */
-
-		/* back out the flags/counters */
-		spin_lock_irqsave(&ctrl->lock, flags);
-		op->flags &= ~FCOP_FLAGS_TERMIO;
-		spin_unlock_irqrestore(&ctrl->lock, flags);
-		return;
-	}
-}
-
-/*
- * This routine stops operation of the controller. Admin and IO queues
- * are stopped, outstanding ios on them terminated, and the nvme ctrl
- * is shutdown.
- */
-static void
-nvme_fc_shutdown_ctrl(struct nvme_fc_ctrl *ctrl)
-{
 	/*
 	 * If io queues are present, stop them and terminate all outstanding
 	 * ios on them. As FC allocates FC exchange for each io, the
@@ -2234,11 +2462,20 @@ nvme_fc_shutdown_ctrl(struct nvme_fc_ctrl *ctrl)
 				nvme_fc_terminate_exchange, &ctrl->ctrl);
 	}
 
-	if (ctrl->ctrl.state == NVME_CTRL_LIVE)
-		nvme_shutdown_ctrl(&ctrl->ctrl);
+	/*
+	 * Other transports, which don't have link-level contexts bound
+	 * to sqe's, would try to gracefully shutdown the controller by
+	 * writing the registers for shutdown and polling (call
+	 * nvme_shutdown_ctrl()). Given a bunch of i/o was potentially
+	 * just aborted and we will wait on those contexts, and given
+	 * there was no indication of how live the controlelr is on the
+	 * link, don't send more io to create more contexts for the
+	 * shutdown. Let the controller fail via keepalive failure if
+	 * its still present.
+	 */
 
 	/*
-	 * now clean up the admin queue. Same thing as above.
+	 * clean up the admin queue. Same thing as above.
 	 * use blk_mq_tagset_busy_itr() and the transport routine to
 	 * terminate the exchanges.
 	 */
@@ -2248,24 +2485,56 @@ nvme_fc_shutdown_ctrl(struct nvme_fc_ctrl *ctrl)
 
 	/* kill the aens as they are a separate path */
 	nvme_fc_abort_aen_ops(ctrl);
+
+	/* wait for all io that had to be aborted */
+	spin_lock_irqsave(&ctrl->lock, flags);
+	while (ctrl->iocnt) {
+		spin_unlock_irqrestore(&ctrl->lock, flags);
+		msleep(1000);
+		spin_lock_irqsave(&ctrl->lock, flags);
+	}
+	ctrl->flags &= ~FCCTRL_TERMIO;
+	spin_unlock_irqrestore(&ctrl->lock, flags);
+
+	nvme_fc_term_aen_ops(ctrl);
+
+	/*
+	 * send a Disconnect(association) LS to fc-nvme target
+	 * Note: could have been sent at top of process, but
+	 * cleaner on link traffic if after the aborts complete.
+	 * Note: if association doesn't exist, association_id will be 0
+	 */
+	if (ctrl->association_id)
+		nvme_fc_xmt_disconnect_assoc(ctrl);
+
+	if (ctrl->ctrl.tagset) {
+		nvme_fc_delete_hw_io_queues(ctrl);
+		nvme_fc_free_io_queues(ctrl);
+	}
+
+	__nvme_fc_delete_hw_queue(ctrl, &ctrl->queues[0], 0);
+	nvme_fc_free_queue(&ctrl->queues[0]);
 }
 
-/*
- * Called to teardown an association.
- * May be called with association fully in place or partially in place.
- */
 static void
-__nvme_fc_remove_ctrl(struct nvme_fc_ctrl *ctrl)
+nvme_fc_delete_ctrl_work(struct work_struct *work)
 {
-	nvme_stop_keep_alive(&ctrl->ctrl);
+	struct nvme_fc_ctrl *ctrl =
+		container_of(work, struct nvme_fc_ctrl, delete_work);
 
-	/* stop and terminate ios on admin and io queues */
-	nvme_fc_shutdown_ctrl(ctrl);
+	cancel_work_sync(&ctrl->reset_work);
+	cancel_delayed_work_sync(&ctrl->connect_work);
+
+	/*
+	 * kill the association on the link side.  this will block
+	 * waiting for io to terminate
+	 */
+	nvme_fc_delete_association(ctrl);
 
 	/*
 	 * tear down the controller
 	 * This will result in the last reference on the nvme ctrl to
-	 * expire, calling the transport nvme_fc_free_nvme_ctrl() callback.
+	 * expire, calling the transport nvme_fc_nvme_ctrl_freed() callback.
 	 * From there, the transport will tear down it's logical queues and
 	 * association.
 	 */
@@ -2274,15 +2543,6 @@ __nvme_fc_remove_ctrl(struct nvme_fc_ctrl *ctrl)
 	nvme_put_ctrl(&ctrl->ctrl);
 }
 
-static void
-nvme_fc_del_ctrl_work(struct work_struct *work)
-{
-	struct nvme_fc_ctrl *ctrl =
-			container_of(work, struct nvme_fc_ctrl, delete_work);
-
-	__nvme_fc_remove_ctrl(ctrl);
-}
-
 static int
 __nvme_fc_del_ctrl(struct nvme_fc_ctrl *ctrl)
 {
@@ -2302,25 +2562,85 @@ static int
 nvme_fc_del_nvme_ctrl(struct nvme_ctrl *nctrl)
 {
 	struct nvme_fc_ctrl *ctrl = to_fc_ctrl(nctrl);
-	struct nvme_fc_rport *rport = ctrl->rport;
-	unsigned long flags;
 	int ret;
 
-	spin_lock_irqsave(&rport->lock, flags);
+	if (!kref_get_unless_zero(&ctrl->ctrl.kref))
+		return -EBUSY;
+
 	ret = __nvme_fc_del_ctrl(ctrl);
-	spin_unlock_irqrestore(&rport->lock, flags);
-	if (ret)
-		return ret;
 
-	flush_work(&ctrl->delete_work);
+	if (!ret)
+		flush_workqueue(nvme_fc_wq);
 
-	return 0;
+	nvme_put_ctrl(&ctrl->ctrl);
+
+	return ret;
 }
 
+static void
+nvme_fc_reset_ctrl_work(struct work_struct *work)
+{
+	struct nvme_fc_ctrl *ctrl =
+			container_of(work, struct nvme_fc_ctrl, reset_work);
+	int ret;
+
+	/* will block will waiting for io to terminate */
+	nvme_fc_delete_association(ctrl);
+
+	ret = nvme_fc_create_association(ctrl);
+	if (ret) {
+		dev_warn(ctrl->ctrl.device,
+			"NVME-FC{%d}: reset: Reconnect attempt failed (%d)\n",
+			ctrl->cnum, ret);
+		if (ctrl->connect_attempts >= NVME_FC_MAX_CONNECT_ATTEMPTS) {
+			dev_warn(ctrl->ctrl.device,
+				"NVME-FC{%d}: Max reconnect attempts (%d) "
+				"reached. Removing controller\n",
+				ctrl->cnum, ctrl->connect_attempts);
+
+			if (!nvme_change_ctrl_state(&ctrl->ctrl,
+				NVME_CTRL_DELETING)) {
+				dev_err(ctrl->ctrl.device,
+					"NVME-FC{%d}: failed to change state "
+					"to DELETING\n", ctrl->cnum);
+				return;
+			}
+
+			WARN_ON(!queue_work(nvme_fc_wq, &ctrl->delete_work));
+			return;
+		}
+
+		dev_warn(ctrl->ctrl.device,
+			"NVME-FC{%d}: Reconnect attempt in %d seconds.\n",
+			ctrl->cnum, ctrl->reconnect_delay);
+		queue_delayed_work(nvme_fc_wq, &ctrl->connect_work,
+				ctrl->reconnect_delay * HZ);
+	} else
+		dev_info(ctrl->ctrl.device,
+			"NVME-FC{%d}: controller reset complete\n", ctrl->cnum);
+}
+
+/*
+ * called by the nvme core layer, for sysfs interface that requests
+ * a reset of the nvme controller
+ */
 static int
 nvme_fc_reset_nvme_ctrl(struct nvme_ctrl *nctrl)
 {
-	return -EIO;
+	struct nvme_fc_ctrl *ctrl = to_fc_ctrl(nctrl);
+
+	dev_warn(ctrl->ctrl.device,
+		"NVME-FC{%d}: admin requested controller reset\n", ctrl->cnum);
+
+	if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_RESETTING))
+		return -EBUSY;
+
+	if (!queue_work(nvme_fc_wq, &ctrl->reset_work))
+		return -EBUSY;
+
+	flush_work(&ctrl->reset_work);
+
+	return 0;
 }
 
 static const struct nvme_ctrl_ops nvme_fc_ctrl_ops = {
@@ -2331,95 +2651,75 @@ static const struct nvme_ctrl_ops nvme_fc_ctrl_ops = {
 	.reg_read64		= nvmf_reg_read64,
 	.reg_write32		= nvmf_reg_write32,
 	.reset_ctrl		= nvme_fc_reset_nvme_ctrl,
-	.free_ctrl		= nvme_fc_free_nvme_ctrl,
+	.free_ctrl		= nvme_fc_nvme_ctrl_freed,
 	.submit_async_event	= nvme_fc_submit_async_event,
 	.delete_ctrl		= nvme_fc_del_nvme_ctrl,
 	.get_subsysnqn		= nvmf_get_subsysnqn,
 	.get_address		= nvmf_get_address,
 };
 
-static int
-nvme_fc_create_io_queues(struct nvme_fc_ctrl *ctrl)
+static void
+nvme_fc_connect_ctrl_work(struct work_struct *work)
 {
-	struct nvmf_ctrl_options *opts = ctrl->ctrl.opts;
 	int ret;
 
-	ret = nvme_set_queue_count(&ctrl->ctrl, &opts->nr_io_queues);
-	if (ret) {
-		dev_info(ctrl->ctrl.device,
-			"set_queue_count failed: %d\n", ret);
-		return ret;
-	}
-
-	ctrl->queue_count = opts->nr_io_queues + 1;
-	if (!opts->nr_io_queues)
-		return 0;
-
-	dev_info(ctrl->ctrl.device, "creating %d I/O queues.\n",
-			opts->nr_io_queues);
-
-	nvme_fc_init_io_queues(ctrl);
-
-	memset(&ctrl->tag_set, 0, sizeof(ctrl->tag_set));
-	ctrl->tag_set.ops = &nvme_fc_mq_ops;
-	ctrl->tag_set.queue_depth = ctrl->ctrl.opts->queue_size;
-	ctrl->tag_set.reserved_tags = 1; /* fabric connect */
-	ctrl->tag_set.numa_node = NUMA_NO_NODE;
-	ctrl->tag_set.flags = BLK_MQ_F_SHOULD_MERGE;
-	ctrl->tag_set.cmd_size = sizeof(struct nvme_fc_fcp_op) +
-					(SG_CHUNK_SIZE *
-						sizeof(struct scatterlist)) +
-					ctrl->lport->ops->fcprqst_priv_sz;
-	ctrl->tag_set.driver_data = ctrl;
-	ctrl->tag_set.nr_hw_queues = ctrl->queue_count - 1;
-	ctrl->tag_set.timeout = NVME_IO_TIMEOUT;
-
-	ret = blk_mq_alloc_tag_set(&ctrl->tag_set);
-	if (ret)
-		return ret;
-
-	ctrl->ctrl.tagset = &ctrl->tag_set;
-
-	ctrl->ctrl.connect_q = blk_mq_init_queue(&ctrl->tag_set);
-	if (IS_ERR(ctrl->ctrl.connect_q)) {
-		ret = PTR_ERR(ctrl->ctrl.connect_q);
-		goto out_free_tag_set;
-	}
-
-	ret = nvme_fc_create_hw_io_queues(ctrl, ctrl->ctrl.opts->queue_size);
-	if (ret)
-		goto out_cleanup_blk_queue;
+	struct nvme_fc_ctrl *ctrl =
+			container_of(to_delayed_work(work),
+				struct nvme_fc_ctrl, connect_work);
 
-	ret = nvme_fc_connect_io_queues(ctrl, ctrl->ctrl.opts->queue_size);
-	if (ret)
-		goto out_delete_hw_queues;
+	ret = nvme_fc_create_association(ctrl);
+	if (ret) {
+		dev_warn(ctrl->ctrl.device,
+			"NVME-FC{%d}: Reconnect attempt failed (%d)\n",
+			ctrl->cnum, ret);
+		if (ctrl->connect_attempts >= NVME_FC_MAX_CONNECT_ATTEMPTS) {
+			dev_warn(ctrl->ctrl.device,
+				"NVME-FC{%d}: Max reconnect attempts (%d) "
+				"reached. Removing controller\n",
+				ctrl->cnum, ctrl->connect_attempts);
+
+			if (!nvme_change_ctrl_state(&ctrl->ctrl,
+				NVME_CTRL_DELETING)) {
+				dev_err(ctrl->ctrl.device,
+					"NVME-FC{%d}: failed to change state "
+					"to DELETING\n", ctrl->cnum);
+				return;
+			}
 
-	return 0;
+			WARN_ON(!queue_work(nvme_fc_wq, &ctrl->delete_work));
+			return;
+		}
 
-out_delete_hw_queues:
-	nvme_fc_delete_hw_io_queues(ctrl);
-out_cleanup_blk_queue:
-	nvme_stop_keep_alive(&ctrl->ctrl);
-	blk_cleanup_queue(ctrl->ctrl.connect_q);
-out_free_tag_set:
-	blk_mq_free_tag_set(&ctrl->tag_set);
-	nvme_fc_free_io_queues(ctrl);
+		dev_warn(ctrl->ctrl.device,
+			"NVME-FC{%d}: Reconnect attempt in %d seconds.\n",
+			ctrl->cnum, ctrl->reconnect_delay);
+		queue_delayed_work(nvme_fc_wq, &ctrl->connect_work,
+				ctrl->reconnect_delay * HZ);
+	} else
+		dev_info(ctrl->ctrl.device,
+			"NVME-FC{%d}: controller reconnect complete\n",
+			ctrl->cnum);
+}
 
-	/* force put free routine to ignore io queues */
-	ctrl->ctrl.tagset = NULL;
 
-	return ret;
-}
+static const struct blk_mq_ops nvme_fc_admin_mq_ops = {
+	.queue_rq	= nvme_fc_queue_rq,
+	.complete	= nvme_fc_complete_rq,
+	.init_request	= nvme_fc_init_admin_request,
+	.exit_request	= nvme_fc_exit_request,
+	.reinit_request	= nvme_fc_reinit_request,
+	.init_hctx	= nvme_fc_init_admin_hctx,
+	.timeout	= nvme_fc_timeout,
+};
 
 
 static struct nvme_ctrl *
-__nvme_fc_create_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,
+nvme_fc_init_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,
 	struct nvme_fc_lport *lport, struct nvme_fc_rport *rport)
 {
 	struct nvme_fc_ctrl *ctrl;
 	unsigned long flags;
 	int ret, idx;
-	bool changed;
 
 	ctrl = kzalloc(sizeof(*ctrl), GFP_KERNEL);
 	if (!ctrl) {
@@ -2438,17 +2738,15 @@ __nvme_fc_create_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,
 	ctrl->lport = lport;
 	ctrl->rport = rport;
 	ctrl->dev = lport->dev;
-	ctrl->state = FCCTRL_INIT;
 	ctrl->cnum = idx;
 
-	ret = nvme_init_ctrl(&ctrl->ctrl, dev, &nvme_fc_ctrl_ops, 0);
-	if (ret)
-		goto out_free_ida;
-
 	get_device(ctrl->dev);
 	kref_init(&ctrl->ref);
 
-	INIT_WORK(&ctrl->delete_work, nvme_fc_del_ctrl_work);
+	INIT_WORK(&ctrl->delete_work, nvme_fc_delete_ctrl_work);
+	INIT_WORK(&ctrl->reset_work, nvme_fc_reset_ctrl_work);
+	INIT_DELAYED_WORK(&ctrl->connect_work, nvme_fc_connect_ctrl_work);
+	ctrl->reconnect_delay = opts->reconnect_delay;
 	spin_lock_init(&ctrl->lock);
 
 	/* io queue count */
@@ -2465,87 +2763,86 @@ __nvme_fc_create_ctrl(struct device *dev, struct nvmf_ctrl_options *opts,
 	ctrl->queues = kcalloc(ctrl->queue_count, sizeof(struct nvme_fc_queue),
 				GFP_KERNEL);
 	if (!ctrl->queues)
-		goto out_uninit_ctrl;
-
-	ret = nvme_fc_configure_admin_queue(ctrl);
-	if (ret)
-		goto out_uninit_ctrl;
-
-	/* sanity checks */
-
-	/* FC-NVME does not have other data in the capsule */
-	if (ctrl->ctrl.icdoff) {
-		dev_err(ctrl->ctrl.device, "icdoff %d is not supported!\n",
-				ctrl->ctrl.icdoff);
-		goto out_remove_admin_queue;
-	}
-
-	/* FC-NVME supports normal SGL Data Block Descriptors */
+		goto out_free_ida;
 
-	if (opts->queue_size > ctrl->ctrl.maxcmd) {
-		/* warn if maxcmd is lower than queue_size */
-		dev_warn(ctrl->ctrl.device,
-			"queue_size %zu > ctrl maxcmd %u, reducing "
-			"to queue_size\n",
-			opts->queue_size, ctrl->ctrl.maxcmd);
-		opts->queue_size = ctrl->ctrl.maxcmd;
-	}
+	memset(&ctrl->admin_tag_set, 0, sizeof(ctrl->admin_tag_set));
+	ctrl->admin_tag_set.ops = &nvme_fc_admin_mq_ops;
+	ctrl->admin_tag_set.queue_depth = NVME_FC_AQ_BLKMQ_DEPTH;
+	ctrl->admin_tag_set.reserved_tags = 2; /* fabric connect + Keep-Alive */
+	ctrl->admin_tag_set.numa_node = NUMA_NO_NODE;
+	ctrl->admin_tag_set.cmd_size = sizeof(struct nvme_fc_fcp_op) +
+					(SG_CHUNK_SIZE *
+						sizeof(struct scatterlist)) +
+					ctrl->lport->ops->fcprqst_priv_sz;
+	ctrl->admin_tag_set.driver_data = ctrl;
+	ctrl->admin_tag_set.nr_hw_queues = 1;
+	ctrl->admin_tag_set.timeout = ADMIN_TIMEOUT;
 
-	ret = nvme_fc_init_aen_ops(ctrl);
+	ret = blk_mq_alloc_tag_set(&ctrl->admin_tag_set);
 	if (ret)
-		goto out_stop_keep_alive;
+		goto out_free_queues;
 
-	if (ctrl->queue_count > 1) {
-		ret = nvme_fc_create_io_queues(ctrl);
-		if (ret)
-			goto out_stop_keep_alive;
+	ctrl->ctrl.admin_q = blk_mq_init_queue(&ctrl->admin_tag_set);
+	if (IS_ERR(ctrl->ctrl.admin_q)) {
+		ret = PTR_ERR(ctrl->ctrl.admin_q);
+		goto out_free_admin_tag_set;
 	}
 
-	spin_lock_irqsave(&ctrl->lock, flags);
-	ctrl->state = FCCTRL_ACTIVE;
-	spin_unlock_irqrestore(&ctrl->lock, flags);
-
-	changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
-	WARN_ON_ONCE(!changed);
+	/*
+	 * Would have been nice to init io queues tag set as well.
+	 * However, we require interaction from the controller
+	 * for max io queue count before we can do so.
+	 * Defer this to the connect path.
+	 */
 
-	dev_info(ctrl->ctrl.device,
-		"NVME-FC{%d}: new ctrl: NQN \"%s\"\n",
-		ctrl->cnum, ctrl->ctrl.opts->subsysnqn);
+	ret = nvme_init_ctrl(&ctrl->ctrl, dev, &nvme_fc_ctrl_ops, 0);
+	if (ret)
+		goto out_cleanup_admin_q;
 
-	kref_get(&ctrl->ctrl.kref);
+	/* at this point, teardown path changes to ref counting on nvme ctrl */
 
 	spin_lock_irqsave(&rport->lock, flags);
 	list_add_tail(&ctrl->ctrl_list, &rport->ctrl_list);
 	spin_unlock_irqrestore(&rport->lock, flags);
 
-	if (opts->nr_io_queues) {
-		nvme_queue_scan(&ctrl->ctrl);
-		nvme_queue_async_events(&ctrl->ctrl);
+	ret = nvme_fc_create_association(ctrl);
+	if (ret) {
+		/* initiate nvme ctrl ref counting teardown */
+		nvme_uninit_ctrl(&ctrl->ctrl);
+		nvme_put_ctrl(&ctrl->ctrl);
+
+		/* as we're past the point where we transition to the ref
+		 * counting teardown path, if we return a bad pointer here,
+		 * the calling routine, thinking it's prior to the
+		 * transition, will do an rport put. Since the teardown
+		 * path also does a rport put, we do an extra get here to
+		 * so proper order/teardown happens.
+		 */
+		nvme_fc_rport_get(rport);
+
+		if (ret > 0)
+			ret = -EIO;
+		return ERR_PTR(ret);
 	}
 
-	return &ctrl->ctrl;
+	dev_info(ctrl->ctrl.device,
+		"NVME-FC{%d}: new ctrl: NQN \"%s\"\n",
+		ctrl->cnum, ctrl->ctrl.opts->subsysnqn);
 
-out_stop_keep_alive:
-	nvme_stop_keep_alive(&ctrl->ctrl);
-out_remove_admin_queue:
-	/* send a Disconnect(association) LS to fc-nvme target */
-	nvme_fc_xmt_disconnect_assoc(ctrl);
-	nvme_stop_keep_alive(&ctrl->ctrl);
-	nvme_fc_destroy_admin_queue(ctrl);
-out_uninit_ctrl:
-	nvme_uninit_ctrl(&ctrl->ctrl);
-	nvme_put_ctrl(&ctrl->ctrl);
-	if (ret > 0)
-		ret = -EIO;
-	/* exit via here will follow ctlr ref point callbacks to free */
-	return ERR_PTR(ret);
+	return &ctrl->ctrl;
 
+out_cleanup_admin_q:
+	blk_cleanup_queue(ctrl->ctrl.admin_q);
+out_free_admin_tag_set:
+	blk_mq_free_tag_set(&ctrl->admin_tag_set);
+out_free_queues:
+	kfree(ctrl->queues);
 out_free_ida:
+	put_device(ctrl->dev);
 	ida_simple_remove(&nvme_fc_ctrl_cnt, ctrl->cnum);
 out_free_ctrl:
 	kfree(ctrl);
 out_fail:
-	nvme_fc_rport_put(rport);
 	/* exit via here doesn't follow ctlr ref points */
 	return ERR_PTR(ret);
 }
@@ -2617,6 +2914,7 @@ nvme_fc_create_ctrl(struct device *dev, struct nvmf_ctrl_options *opts)
 {
 	struct nvme_fc_lport *lport;
 	struct nvme_fc_rport *rport;
+	struct nvme_ctrl *ctrl;
 	struct nvmet_fc_traddr laddr = { 0L, 0L };
 	struct nvmet_fc_traddr raddr = { 0L, 0L };
 	unsigned long flags;
@@ -2648,7 +2946,10 @@ nvme_fc_create_ctrl(struct device *dev, struct nvmf_ctrl_options *opts)
 
 			spin_unlock_irqrestore(&nvme_fc_lock, flags);
 
-			return __nvme_fc_create_ctrl(dev, opts, lport, rport);
+			ctrl = nvme_fc_init_ctrl(dev, opts, lport, rport);
+			if (IS_ERR(ctrl))
+				nvme_fc_rport_put(rport);
+			return ctrl;
 		}
 	}
 	spin_unlock_irqrestore(&nvme_fc_lock, flags);