Merge tag 'nvme-6.16-2025-05-20' of git://git.infradead.org/nvme into for-6.16/block

Pull NVMe updates from Christoph:

"nvme updates for Linux 6.16

- add per-node DMA pools and use them for PRP/SGL allocations
(Caleb Sander Mateos, Keith Busch)
- nvme-fcloop refcounting fixes (Daniel Wagner)
- support delayed removal of the multipath node and optionally support
the multipath node for private namespaces (Nilay Shroff)
- support shared CQs in the PCI endpoint target code (Wilfred Mallawa)
- support admin-queue only authentication (Hannes Reinecke)
- use the crc32c library instead of the crypto API (Eric Biggers)
- misc cleanups (Christoph Hellwig, Marcelo Moreira, Hannes Reinecke,
Leon Romanovsky, Gustavo A. R. Silva)"

* tag 'nvme-6.16-2025-05-20' of git://git.infradead.org/nvme: (42 commits)
nvme: rename nvme_mpath_shutdown_disk to nvme_mpath_remove_disk
nvme: introduce multipath_always_on module param
nvme-multipath: introduce delayed removal of the multipath head node
nvme-pci: derive and better document max segments limits
nvme-pci: use struct_size for allocation struct nvme_dev
nvme-pci: add a symolic name for the small pool size
nvme-pci: use a better encoding for small prp pool allocations
nvme-pci: rename the descriptor pools
nvme-pci: remove struct nvme_descriptor
nvme-pci: store aborted state in flags variable
nvme-pci: don't try to use SGLs for metadata on the admin queue
nvme-pci: make PRP list DMA pools per-NUMA-node
nvme-pci: factor out a nvme_init_hctx_common() helper
dmapool: add NUMA affinity support
nvme-fc: do not reference lsrsp after failure
nvmet-fcloop: don't wait for lport cleanup
nvmet-fcloop: add missing fcloop_callback_host_done
nvmet-fc: take tgtport refs for portentry
nvmet-fc: free pending reqs on tgtport unregister
nvmet-fcloop: drop response if targetport is gone
...

drivers/nvme/common/auth.c

···
 {
 	const char *hmac_name;
 	struct crypto_shash *key_tfm;
-	struct shash_desc *shash;
+	SHASH_DESC_ON_STACK(shash, key_tfm);
 	struct nvme_dhchap_key *transformed_key;
 	int ret, key_len;
 
···
 	if (IS_ERR(key_tfm))
 		return ERR_CAST(key_tfm);
 
-	shash = kmalloc(sizeof(struct shash_desc) +
-			crypto_shash_descsize(key_tfm),
-			GFP_KERNEL);
-	if (!shash) {
-		ret = -ENOMEM;
-		goto out_free_key;
-	}
-
 	key_len = crypto_shash_digestsize(key_tfm);
 	transformed_key = nvme_auth_alloc_key(key_len, key->hash);
 	if (!transformed_key) {
 		ret = -ENOMEM;
-		goto out_free_shash;
+		goto out_free_key;
 	}
 
 	shash->tfm = key_tfm;
···
 	if (ret < 0)
 		goto out_free_transformed_key;
 
-	kfree(shash);
 	crypto_free_shash(key_tfm);
 
 	return transformed_key;
 
 out_free_transformed_key:
 	nvme_auth_free_key(transformed_key);
-out_free_shash:
-	kfree(shash);
 out_free_key:
 	crypto_free_shash(key_tfm);
 

drivers/nvme/host/auth.c

···
 	u32 s1;
 	u32 s2;
 	bool bi_directional;
+	bool authenticated;
 	u16 transaction;
 	u8 status;
 	u8 dhgroup_id;
···
 static void nvme_auth_free_dhchap(struct nvme_dhchap_queue_context *chap)
 {
 	nvme_auth_reset_dhchap(chap);
+	chap->authenticated = false;
 	if (chap->shash_tfm)
 		crypto_free_shash(chap->shash_tfm);
 	if (chap->dh_tfm)
···
 	}
 	if (!ret) {
 		chap->error = 0;
+		chap->authenticated = true;
 		if (ctrl->opts->concat &&
 		    (ret = nvme_auth_secure_concat(ctrl, chap))) {
 			dev_warn(ctrl->device,
 				 "%s: qid %d failed to enable secure concatenation\n",
 				 __func__, chap->qid);
 			chap->error = ret;
+			chap->authenticated = false;
 		}
 		return;
 	}
···
 		return;
 
 	for (q = 1; q < ctrl->queue_count; q++) {
-		ret = nvme_auth_negotiate(ctrl, q);
-		if (ret) {
-			dev_warn(ctrl->device,
-				 "qid %d: error %d setting up authentication\n",
-				 q, ret);
-			break;
-		}
+		struct nvme_dhchap_queue_context *chap =
+			&ctrl->dhchap_ctxs[q];
+		/*
+		 * Skip re-authentication if the queue had
+		 * not been authenticated initially.
+		 */
+		if (!chap->authenticated)
+			continue;
+		cancel_work_sync(&chap->auth_work);
+		queue_work(nvme_auth_wq, &chap->auth_work);
 	}
 
 	/*
···
 	 * the controller terminates the connection.
 	 */
 	for (q = 1; q < ctrl->queue_count; q++) {
-		ret = nvme_auth_wait(ctrl, q);
+		struct nvme_dhchap_queue_context *chap =
+			&ctrl->dhchap_ctxs[q];
+		if (!chap->authenticated)
+			continue;
+		flush_work(&chap->auth_work);
+		ret = chap->error;
+		nvme_auth_reset_dhchap(chap);
 		if (ret)
 			dev_warn(ctrl->device,
 				 "qid %d: authentication failed\n", q);
···
 		chap = &ctrl->dhchap_ctxs[i];
 		chap->qid = i;
 		chap->ctrl = ctrl;
+		chap->authenticated = false;
 		INIT_WORK(&chap->auth_work, nvme_queue_auth_work);
 	}
 

drivers/nvme/host/core.c

···
 	struct nvme_ns_head *head =
 		container_of(ref, struct nvme_ns_head, ref);
 
-	nvme_mpath_remove_disk(head);
+	nvme_mpath_put_disk(head);
 	ida_free(&head->subsys->ns_ida, head->instance);
 	cleanup_srcu_struct(&head->srcu);
 	nvme_put_subsystem(head->subsys);
···
 		 */
 		if (h->ns_id != nsid || !nvme_is_unique_nsid(ctrl, h))
 			continue;
-		if (!list_empty(&h->list) && nvme_tryget_ns_head(h))
+		if (nvme_tryget_ns_head(h))
 			return h;
 	}
 
···
 		}
 	} else {
 		ret = -EINVAL;
-		if (!info->is_shared || !head->shared) {
+		if ((!info->is_shared || !head->shared) &&
+		    !list_empty(&head->list)) {
 			dev_err(ctrl->device,
 				"Duplicate unshared namespace %d\n",
 				info->nsid);
···
 	mutex_lock(&ns->ctrl->subsys->lock);
 	list_del_rcu(&ns->siblings);
 	if (list_empty(&ns->head->list)) {
-		list_del_init(&ns->head->entry);
+		if (!nvme_mpath_queue_if_no_path(ns->head))
+			list_del_init(&ns->head->entry);
 		last_path = true;
 	}
 	mutex_unlock(&ns->ctrl->subsys->lock);
···
 	synchronize_srcu(&ns->ctrl->srcu);
 
 	if (last_path)
-		nvme_mpath_shutdown_disk(ns->head);
+		nvme_mpath_remove_disk(ns->head);
 	nvme_put_ns(ns);
 }
 

drivers/nvme/host/fc.c

···
 }
 
 static void
-nvme_fc_xmt_ls_rsp_done(struct nvmefc_ls_rsp *lsrsp)
+nvme_fc_xmt_ls_rsp_free(struct nvmefc_ls_rcv_op *lsop)
 {
-	struct nvmefc_ls_rcv_op *lsop = lsrsp->nvme_fc_private;
 	struct nvme_fc_rport *rport = lsop->rport;
 	struct nvme_fc_lport *lport = rport->lport;
 	unsigned long flags;
···
 }
 
 static void
+nvme_fc_xmt_ls_rsp_done(struct nvmefc_ls_rsp *lsrsp)
+{
+	struct nvmefc_ls_rcv_op *lsop = lsrsp->nvme_fc_private;
+
+	nvme_fc_xmt_ls_rsp_free(lsop);
+}
+
+static void
 nvme_fc_xmt_ls_rsp(struct nvmefc_ls_rcv_op *lsop)
 {
 	struct nvme_fc_rport *rport = lsop->rport;
···
 		dev_warn(lport->dev,
 			"LLDD rejected LS RSP xmt: LS %d status %d\n",
 			w0->ls_cmd, ret);
-		nvme_fc_xmt_ls_rsp_done(lsop->lsrsp);
+		nvme_fc_xmt_ls_rsp_free(lsop);
 		return;
 	}
 }

drivers/nvme/host/multipath.c

···
 #include "nvme.h"
 
 bool multipath = true;
-module_param(multipath, bool, 0444);
+static bool multipath_always_on;
+
+static int multipath_param_set(const char *val, const struct kernel_param *kp)
+{
+	int ret;
+	bool *arg = kp->arg;
+
+	ret = param_set_bool(val, kp);
+	if (ret)
+		return ret;
+
+	if (multipath_always_on && !*arg) {
+		pr_err("Can't disable multipath when multipath_always_on is configured.\n");
+		*arg = true;
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static const struct kernel_param_ops multipath_param_ops = {
+	.set = multipath_param_set,
+	.get = param_get_bool,
+};
+
+module_param_cb(multipath, &multipath_param_ops, &multipath, 0444);
 MODULE_PARM_DESC(multipath,
 	"turn on native support for multiple controllers per subsystem");
+
+static int multipath_always_on_set(const char *val,
+		const struct kernel_param *kp)
+{
+	int ret;
+	bool *arg = kp->arg;
+
+	ret = param_set_bool(val, kp);
+	if (ret < 0)
+		return ret;
+
+	if (*arg)
+		multipath = true;
+
+	return 0;
+}
+
+static const struct kernel_param_ops multipath_always_on_ops = {
+	.set = multipath_always_on_set,
+	.get = param_get_bool,
+};
+
+module_param_cb(multipath_always_on, &multipath_always_on_ops,
+		&multipath_always_on, 0444);
+MODULE_PARM_DESC(multipath_always_on,
+	"create multipath node always except for private namespace with non-unique nsid; note that this also implicitly enables native multipath support");
 
 static const char *nvme_iopolicy_names[] = {
 	[NVME_IOPOLICY_NUMA]	= "numa",
···
 			break;
 		}
 	}
-	return false;
+
+	/*
+	 * If "head->delayed_removal_secs" is configured (i.e., non-zero), do
+	 * not immediately fail I/O. Instead, requeue the I/O for the configured
+	 * duration, anticipating that if there's a transient link failure then
+	 * it may recover within this time window. This parameter is exported to
+	 * userspace via sysfs, and its default value is zero. It is internally
+	 * mapped to NVME_NSHEAD_QUEUE_IF_NO_PATH. When delayed_removal_secs is
+	 * non-zero, this flag is set to true. When zero, the flag is cleared.
+	 */
+	return nvme_mpath_queue_if_no_path(head);
 }
 
 static void nvme_ns_head_submit_bio(struct bio *bio)
···
 	}
 }
 
+static void nvme_remove_head(struct nvme_ns_head *head)
+{
+	if (test_and_clear_bit(NVME_NSHEAD_DISK_LIVE, &head->flags)) {
+		/*
+		 * requeue I/O after NVME_NSHEAD_DISK_LIVE has been cleared
+		 * to allow multipath to fail all I/O.
+		 */
+		kblockd_schedule_work(&head->requeue_work);
+
+		nvme_cdev_del(&head->cdev, &head->cdev_device);
+		synchronize_srcu(&head->srcu);
+		del_gendisk(head->disk);
+		nvme_put_ns_head(head);
+	}
+}
+
+static void nvme_remove_head_work(struct work_struct *work)
+{
+	struct nvme_ns_head *head = container_of(to_delayed_work(work),
+			struct nvme_ns_head, remove_work);
+	bool remove = false;
+
+	mutex_lock(&head->subsys->lock);
+	if (list_empty(&head->list)) {
+		list_del_init(&head->entry);
+		remove = true;
+	}
+	mutex_unlock(&head->subsys->lock);
+	if (remove)
+		nvme_remove_head(head);
+
+	module_put(THIS_MODULE);
+}
+
 int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
 {
 	struct queue_limits lim;
···
 	spin_lock_init(&head->requeue_lock);
 	INIT_WORK(&head->requeue_work, nvme_requeue_work);
 	INIT_WORK(&head->partition_scan_work, nvme_partition_scan_work);
+	INIT_DELAYED_WORK(&head->remove_work, nvme_remove_head_work);
+	head->delayed_removal_secs = 0;
 
 	/*
-	 * Add a multipath node if the subsystems supports multiple controllers.
-	 * We also do this for private namespaces as the namespace sharing flag
-	 * could change after a rescan.
+	 * If "multipath_always_on" is enabled, a multipath node is added
+	 * regardless of whether the disk is single/multi ported, and whether
+	 * the namespace is shared or private. If "multipath_always_on" is not
+	 * enabled, a multipath node is added only if the subsystem supports
+	 * multiple controllers and the "multipath" option is configured. In
+	 * either case, for private namespaces, we ensure that the NSID is
+	 * unique.
 	 */
-	if (!(ctrl->subsys->cmic & NVME_CTRL_CMIC_MULTI_CTRL) ||
-	    !nvme_is_unique_nsid(ctrl, head) || !multipath)
+	if (!multipath_always_on) {
+		if (!(ctrl->subsys->cmic & NVME_CTRL_CMIC_MULTI_CTRL) ||
+				!multipath)
+			return 0;
+	}
+
+	if (!nvme_is_unique_nsid(ctrl, head))
 		return 0;
 
 	blk_set_stacking_limits(&lim);
···
 	set_bit(GD_SUPPRESS_PART_SCAN, &head->disk->state);
 	sprintf(head->disk->disk_name, "nvme%dn%d",
 			ctrl->subsys->instance, head->instance);
+	nvme_tryget_ns_head(head);
 	return 0;
 }
 
···
 }
 DEVICE_ATTR_RO(numa_nodes);
 
+static ssize_t delayed_removal_secs_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+	struct nvme_ns_head *head = disk->private_data;
+	int ret;
+
+	mutex_lock(&head->subsys->lock);
+	ret = sysfs_emit(buf, "%u\n", head->delayed_removal_secs);
+	mutex_unlock(&head->subsys->lock);
+	return ret;
+}
+
+static ssize_t delayed_removal_secs_store(struct device *dev,
+		struct device_attribute *attr, const char *buf, size_t count)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+	struct nvme_ns_head *head = disk->private_data;
+	unsigned int sec;
+	int ret;
+
+	ret = kstrtouint(buf, 0, &sec);
+	if (ret < 0)
+		return ret;
+
+	mutex_lock(&head->subsys->lock);
+	head->delayed_removal_secs = sec;
+	if (sec)
+		set_bit(NVME_NSHEAD_QUEUE_IF_NO_PATH, &head->flags);
+	else
+		clear_bit(NVME_NSHEAD_QUEUE_IF_NO_PATH, &head->flags);
+	mutex_unlock(&head->subsys->lock);
+	/*
+	 * Ensure that update to NVME_NSHEAD_QUEUE_IF_NO_PATH is seen
+	 * by its reader.
+	 */
+	synchronize_srcu(&head->srcu);
+
+	return count;
+}
+
+DEVICE_ATTR_RW(delayed_removal_secs);
+
 static int nvme_lookup_ana_group_desc(struct nvme_ctrl *ctrl,
 		struct nvme_ana_group_desc *desc, void *data)
 {
···
 #endif
 }
 
-void nvme_mpath_shutdown_disk(struct nvme_ns_head *head)
+void nvme_mpath_remove_disk(struct nvme_ns_head *head)
 {
-	if (!head->disk)
-		return;
-	if (test_and_clear_bit(NVME_NSHEAD_DISK_LIVE, &head->flags)) {
-		nvme_cdev_del(&head->cdev, &head->cdev_device);
+	bool remove = false;
+
+	mutex_lock(&head->subsys->lock);
+	/*
+	 * We are called when all paths have been removed, and at that point
+	 * head->list is expected to be empty. However, nvme_remove_ns() and
+	 * nvme_init_ns_head() can run concurrently and so if head->delayed_
+	 * removal_secs is configured, it is possible that by the time we reach
+	 * this point, head->list may no longer be empty. Therefore, we recheck
+	 * head->list here. If it is no longer empty then we skip enqueuing the
+	 * delayed head removal work.
+	 */
+	if (!list_empty(&head->list))
+		goto out;
+
+	if (head->delayed_removal_secs) {
 		/*
-		 * requeue I/O after NVME_NSHEAD_DISK_LIVE has been cleared
-		 * to allow multipath to fail all I/O.
+		 * Ensure that no one could remove this module while the head
+		 * remove work is pending.
 		 */
-		synchronize_srcu(&head->srcu);
-		kblockd_schedule_work(&head->requeue_work);
-		del_gendisk(head->disk);
+		if (!try_module_get(THIS_MODULE))
+			goto out;
+		queue_delayed_work(nvme_wq, &head->remove_work,
+				head->delayed_removal_secs * HZ);
+	} else {
+		list_del_init(&head->entry);
+		remove = true;
 	}
+out:
+	mutex_unlock(&head->subsys->lock);
+	if (remove)
+		nvme_remove_head(head);
 }
 
-void nvme_mpath_remove_disk(struct nvme_ns_head *head)
+void nvme_mpath_put_disk(struct nvme_ns_head *head)
 {
 	if (!head->disk)
 		return;

drivers/nvme/host/nvme.h

···
 	struct work_struct	partition_scan_work;
 	struct mutex		lock;
 	unsigned long		flags;
-#define NVME_NSHEAD_DISK_LIVE	0
+	struct delayed_work	remove_work;
+	unsigned int		delayed_removal_secs;
+#define NVME_NSHEAD_DISK_LIVE		0
+#define NVME_NSHEAD_QUEUE_IF_NO_PATH	1
 	struct nvme_ns __rcu	*current_path[];
 #endif
 };
···
 void nvme_mpath_add_sysfs_link(struct nvme_ns_head *ns);
 void nvme_mpath_remove_sysfs_link(struct nvme_ns *ns);
 void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid);
-void nvme_mpath_remove_disk(struct nvme_ns_head *head);
+void nvme_mpath_put_disk(struct nvme_ns_head *head);
 int nvme_mpath_init_identify(struct nvme_ctrl *ctrl, struct nvme_id_ctrl *id);
 void nvme_mpath_init_ctrl(struct nvme_ctrl *ctrl);
 void nvme_mpath_update(struct nvme_ctrl *ctrl);
···
 bool nvme_mpath_clear_current_path(struct nvme_ns *ns);
 void nvme_mpath_revalidate_paths(struct nvme_ns *ns);
 void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl);
-void nvme_mpath_shutdown_disk(struct nvme_ns_head *head);
+void nvme_mpath_remove_disk(struct nvme_ns_head *head);
 void nvme_mpath_start_request(struct request *rq);
 void nvme_mpath_end_request(struct request *rq);
 
···
 extern struct device_attribute dev_attr_ana_state;
 extern struct device_attribute dev_attr_queue_depth;
 extern struct device_attribute dev_attr_numa_nodes;
+extern struct device_attribute dev_attr_delayed_removal_secs;
 extern struct device_attribute subsys_attr_iopolicy;
 
 static inline bool nvme_disk_is_ns_head(struct gendisk *disk)
 {
 	return disk->fops == &nvme_ns_head_ops;
+}
+static inline bool nvme_mpath_queue_if_no_path(struct nvme_ns_head *head)
+{
+	if (test_bit(NVME_NSHEAD_QUEUE_IF_NO_PATH, &head->flags))
+		return true;
+	return false;
 }
 #else
 #define multipath false
···
 static inline void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid)
 {
 }
-static inline void nvme_mpath_remove_disk(struct nvme_ns_head *head)
+static inline void nvme_mpath_put_disk(struct nvme_ns_head *head)
 {
 }
 static inline void nvme_mpath_add_sysfs_link(struct nvme_ns *ns)
···
 static inline void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl)
 {
 }
-static inline void nvme_mpath_shutdown_disk(struct nvme_ns_head *head)
+static inline void nvme_mpath_remove_disk(struct nvme_ns_head *head)
 {
 }
 static inline void nvme_trace_bio_complete(struct request *req)
···
 {
 }
 static inline bool nvme_disk_is_ns_head(struct gendisk *disk)
+{
+	return false;
+}
+static inline bool nvme_mpath_queue_if_no_path(struct nvme_ns_head *head)
 {
 	return false;
 }

drivers/nvme/host/pci.c

···
 #include <linux/mm.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
+#include <linux/nodemask.h>
 #include <linux/once.h>
 #include <linux/pci.h>
 #include <linux/suspend.h>
···
 #define SQ_SIZE(q)	((q)->q_depth << (q)->sqes)
 #define CQ_SIZE(q)	((q)->q_depth * sizeof(struct nvme_completion))
 
-#define SGES_PER_PAGE	(NVME_CTRL_PAGE_SIZE / sizeof(struct nvme_sgl_desc))
+/* Optimisation for I/Os between 4k and 128k */
+#define NVME_SMALL_POOL_SIZE	256
 
 /*
  * These can be higher, but we need to ensure that any command doesn't
  * require an sg allocation that needs more than a page of data.
  */
 #define NVME_MAX_KB_SZ	8192
-#define NVME_MAX_SEGS	128
-#define NVME_MAX_META_SEGS 15
-#define NVME_MAX_NR_ALLOCATIONS	5
+#define NVME_MAX_NR_DESCRIPTORS	5
+
+/*
+ * For data SGLs we support a single descriptors worth of SGL entries, but for
+ * now we also limit it to avoid an allocation larger than PAGE_SIZE for the
+ * scatterlist.
+ */
+#define NVME_MAX_SEGS \
+	min(NVME_CTRL_PAGE_SIZE / sizeof(struct nvme_sgl_desc), \
+	    (PAGE_SIZE / sizeof(struct scatterlist)))
+
+/*
+ * For metadata SGLs, only the small descriptor is supported, and the first
+ * entry is the segment descriptor, which for the data pointer sits in the SQE.
+ */
+#define NVME_MAX_META_SEGS \
+	((NVME_SMALL_POOL_SIZE / sizeof(struct nvme_sgl_desc)) - 1)
 
 static int use_threaded_interrupts;
 module_param(use_threaded_interrupts, int, 0444);
···
 static void nvme_delete_io_queues(struct nvme_dev *dev);
 static void nvme_update_attrs(struct nvme_dev *dev);
 
+struct nvme_descriptor_pools {
+	struct dma_pool *large;
+	struct dma_pool *small;
+};
+
 /*
  * Represents an NVM Express device.  Each nvme_dev is a PCI function.
  */
···
 	struct blk_mq_tag_set admin_tagset;
 	u32 __iomem *dbs;
 	struct device *dev;
-	struct dma_pool *prp_page_pool;
-	struct dma_pool *prp_small_pool;
 	unsigned online_queues;
 	unsigned max_qid;
 	unsigned io_queues[HCTX_MAX_TYPES];
···
 	unsigned int nr_allocated_queues;
 	unsigned int nr_write_queues;
 	unsigned int nr_poll_queues;
+	struct nvme_descriptor_pools descriptor_pools[];
 };
 
 static int io_queue_depth_set(const char *val, const struct kernel_param *kp)
···
  */
 struct nvme_queue {
 	struct nvme_dev *dev;
+	struct nvme_descriptor_pools descriptor_pools;
 	spinlock_t sq_lock;
 	void *sq_cmds;
 	 /* only used for poll queues: */
···
 	struct completion delete_done;
 };
 
-union nvme_descriptor {
-	struct nvme_sgl_desc	*sg_list;
-	__le64			*prp_list;
+/* bits for iod->flags */
+enum nvme_iod_flags {
+	/* this command has been aborted by the timeout handler */
+	IOD_ABORTED		= 1U << 0,
+
+	/* uses the small descriptor pool */
+	IOD_SMALL_DESCRIPTOR		= 1U << 1,
 };
 
 /*
  * The nvme_iod describes the data in an I/O.
- *
- * The sg pointer contains the list of PRP/SGL chunk allocations in addition
- * to the actual struct scatterlist.
  */
 struct nvme_iod {
 	struct nvme_request req;
 	struct nvme_command cmd;
-	bool aborted;
-	s8 nr_allocations;	/* PRP list pool allocations. 0 means small
-				   pool in use */
+	u8 flags;
+	u8 nr_descriptors;
 	unsigned int dma_len;	/* length of single DMA segment mapping */
 	dma_addr_t first_dma;
 	dma_addr_t meta_dma;
 	struct sg_table sgt;
 	struct sg_table meta_sgt;
-	union nvme_descriptor meta_list;
-	union nvme_descriptor list[NVME_MAX_NR_ALLOCATIONS];
+	struct nvme_sgl_desc *meta_descriptor;
+	void *descriptors[NVME_MAX_NR_DESCRIPTORS];
 };
 
 static inline unsigned int nvme_dbbuf_size(struct nvme_dev *dev)
···
 	return DIV_ROUND_UP(8 * nprps, NVME_CTRL_PAGE_SIZE - 8);
 }
 
-static int nvme_admin_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
-				unsigned int hctx_idx)
+static struct nvme_descriptor_pools *
+nvme_setup_descriptor_pools(struct nvme_dev *dev, unsigned numa_node)
+{
+	struct nvme_descriptor_pools *pools = &dev->descriptor_pools[numa_node];
+	size_t small_align = NVME_SMALL_POOL_SIZE;
+
+	if (pools->small)
+		return pools; /* already initialized */
+
+	pools->large = dma_pool_create_node("nvme descriptor page", dev->dev,
+			NVME_CTRL_PAGE_SIZE, NVME_CTRL_PAGE_SIZE, 0, numa_node);
+	if (!pools->large)
+		return ERR_PTR(-ENOMEM);
+
+	if (dev->ctrl.quirks & NVME_QUIRK_DMAPOOL_ALIGN_512)
+		small_align = 512;
+
+	pools->small = dma_pool_create_node("nvme descriptor small", dev->dev,
+			NVME_SMALL_POOL_SIZE, small_align, 0, numa_node);
+	if (!pools->small) {
+		dma_pool_destroy(pools->large);
+		pools->large = NULL;
+		return ERR_PTR(-ENOMEM);
+	}
+
+	return pools;
+}
+
+static void nvme_release_descriptor_pools(struct nvme_dev *dev)
+{
+	unsigned i;
+
+	for (i = 0; i < nr_node_ids; i++) {
+		struct nvme_descriptor_pools *pools = &dev->descriptor_pools[i];
+
+		dma_pool_destroy(pools->large);
+		dma_pool_destroy(pools->small);
+	}
+}
+
+static int nvme_init_hctx_common(struct blk_mq_hw_ctx *hctx, void *data,
+		unsigned qid)
 {
 	struct nvme_dev *dev = to_nvme_dev(data);
-	struct nvme_queue *nvmeq = &dev->queues[0];
+	struct nvme_queue *nvmeq = &dev->queues[qid];
+	struct nvme_descriptor_pools *pools;
+	struct blk_mq_tags *tags;
 
-	WARN_ON(hctx_idx != 0);
-	WARN_ON(dev->admin_tagset.tags[0] != hctx->tags);
+	tags = qid ? dev->tagset.tags[qid - 1] : dev->admin_tagset.tags[0];
+	WARN_ON(tags != hctx->tags);
+	pools = nvme_setup_descriptor_pools(dev, hctx->numa_node);
+	if (IS_ERR(pools))
+		return PTR_ERR(pools);
 
+	nvmeq->descriptor_pools = *pools;
 	hctx->driver_data = nvmeq;
 	return 0;
 }
 
-static int nvme_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
-			  unsigned int hctx_idx)
+static int nvme_admin_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
+				unsigned int hctx_idx)
 {
-	struct nvme_dev *dev = to_nvme_dev(data);
-	struct nvme_queue *nvmeq = &dev->queues[hctx_idx + 1];
+	WARN_ON(hctx_idx != 0);
+	return nvme_init_hctx_common(hctx, data, 0);
+}
 
-	WARN_ON(dev->tagset.tags[hctx_idx] != hctx->tags);
-	hctx->driver_data = nvmeq;
-	return 0;
+static int nvme_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
+			     unsigned int hctx_idx)
+{
+	return nvme_init_hctx_common(hctx, data, hctx_idx + 1);
 }
 
 static int nvme_pci_init_request(struct blk_mq_tag_set *set,
···
 	return true;
 }
 
-static void nvme_free_prps(struct nvme_dev *dev, struct request *req)
+static inline struct dma_pool *nvme_dma_pool(struct nvme_queue *nvmeq,
+		struct nvme_iod *iod)
+{
+	if (iod->flags & IOD_SMALL_DESCRIPTOR)
+		return nvmeq->descriptor_pools.small;
+	return nvmeq->descriptor_pools.large;
+}
+
+static void nvme_free_descriptors(struct nvme_queue *nvmeq, struct request *req)
 {
 	const int last_prp = NVME_CTRL_PAGE_SIZE / sizeof(__le64) - 1;
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	dma_addr_t dma_addr = iod->first_dma;
 	int i;
 
-	for (i = 0; i < iod->nr_allocations; i++) {
-		__le64 *prp_list = iod->list[i].prp_list;
+	if (iod->nr_descriptors == 1) {
+		dma_pool_free(nvme_dma_pool(nvmeq, iod), iod->descriptors[0],
+				dma_addr);
+		return;
+	}
+
+	for (i = 0; i < iod->nr_descriptors; i++) {
+		__le64 *prp_list = iod->descriptors[i];
 		dma_addr_t next_dma_addr = le64_to_cpu(prp_list[last_prp]);
 
-		dma_pool_free(dev->prp_page_pool, prp_list, dma_addr);
+		dma_pool_free(nvmeq->descriptor_pools.large, prp_list,
+				dma_addr);
 		dma_addr = next_dma_addr;
 	}
 }
 
-static void nvme_unmap_data(struct nvme_dev *dev, struct request *req)
+static void nvme_unmap_data(struct nvme_dev *dev, struct nvme_queue *nvmeq,
+			    struct request *req)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 
···
 	WARN_ON_ONCE(!iod->sgt.nents);
 
 	dma_unmap_sgtable(dev->dev, &iod->sgt, rq_dma_dir(req), 0);
-
-	if (iod->nr_allocations == 0)
-		dma_pool_free(dev->prp_small_pool, iod->list[0].sg_list,
-			      iod->first_dma);
-	else if (iod->nr_allocations == 1)
-		dma_pool_free(dev->prp_page_pool, iod->list[0].sg_list,
-			      iod->first_dma);
-	else
-		nvme_free_prps(dev, req);
+	nvme_free_descriptors(nvmeq, req);
 	mempool_free(iod->sgt.sgl, dev->iod_mempool);
 }
 
···
 	}
 }
 
-static blk_status_t nvme_pci_setup_prps(struct nvme_dev *dev,
+static blk_status_t nvme_pci_setup_prps(struct nvme_queue *nvmeq,
 		struct request *req, struct nvme_rw_command *cmnd)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
-	struct dma_pool *pool;
 	int length = blk_rq_payload_bytes(req);
 	struct scatterlist *sg = iod->sgt.sgl;
 	int dma_len = sg_dma_len(sg);
···
 	int offset = dma_addr & (NVME_CTRL_PAGE_SIZE - 1);
 	__le64 *prp_list;
 	dma_addr_t prp_dma;
-	int nprps, i;
+	int i;
 
 	length -= (NVME_CTRL_PAGE_SIZE - offset);
 	if (length <= 0) {
···
 		goto done;
 	}
 
-	nprps = DIV_ROUND_UP(length, NVME_CTRL_PAGE_SIZE);
-	if (nprps <= (256 / 8)) {
-		pool = dev->prp_small_pool;
-		iod->nr_allocations = 0;
-	} else {
-		pool = dev->prp_page_pool;
-		iod->nr_allocations = 1;
-	}
+	if (DIV_ROUND_UP(length, NVME_CTRL_PAGE_SIZE) <=
+	    NVME_SMALL_POOL_SIZE / sizeof(__le64))
+		iod->flags |= IOD_SMALL_DESCRIPTOR;
 
-	prp_list = dma_pool_alloc(pool, GFP_ATOMIC, &prp_dma);
-	if (!prp_list) {
-		iod->nr_allocations = -1;
+	prp_list = dma_pool_alloc(nvme_dma_pool(nvmeq, iod), GFP_ATOMIC,
+			&prp_dma);
+	if (!prp_list)
 		return BLK_STS_RESOURCE;
-	}
-	iod->list[0].prp_list = prp_list;
+	iod->descriptors[iod->nr_descriptors++] = prp_list;
 	iod->first_dma = prp_dma;
 	i = 0;
 	for (;;) {
 		if (i == NVME_CTRL_PAGE_SIZE >> 3) {
 			__le64 *old_prp_list = prp_list;
-			prp_list = dma_pool_alloc(pool, GFP_ATOMIC, &prp_dma);
+
+			prp_list = dma_pool_alloc(nvmeq->descriptor_pools.large,
+					GFP_ATOMIC, &prp_dma);
 			if (!prp_list)
 				goto free_prps;
-			iod->list[iod->nr_allocations++].prp_list = prp_list;
+			iod->descriptors[iod->nr_descriptors++] = prp_list;
 			prp_list[0] = old_prp_list[i - 1];
 			old_prp_list[i - 1] = cpu_to_le64(prp_dma);
 			i = 1;
···
 	cmnd->dptr.prp2 = cpu_to_le64(iod->first_dma);
 	return BLK_STS_OK;
 free_prps:
-	nvme_free_prps(dev, req);
+	nvme_free_descriptors(nvmeq, req);
 	return BLK_STS_RESOURCE;
 bad_sgl:
 	WARN(DO_ONCE(nvme_print_sgl, iod->sgt.sgl, iod->sgt.nents),
···
 	sge->type = NVME_SGL_FMT_LAST_SEG_DESC << 4;
 }
 
-static blk_status_t nvme_pci_setup_sgls(struct nvme_dev *dev,
+static blk_status_t nvme_pci_setup_sgls(struct nvme_queue *nvmeq,
 		struct request *req, struct nvme_rw_command *cmd)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
-	struct dma_pool *pool;
 	struct nvme_sgl_desc *sg_list;
 	struct scatterlist *sg = iod->sgt.sgl;
 	unsigned int entries = iod->sgt.nents;
···
 		return BLK_STS_OK;
 	}
 
-	if (entries <= (256 / sizeof(struct nvme_sgl_desc))) {
-		pool = dev->prp_small_pool;
-		iod->nr_allocations = 0;
-	} else {
-		pool = dev->prp_page_pool;
-		iod->nr_allocations = 1;
-	}
+	if (entries <= NVME_SMALL_POOL_SIZE / sizeof(*sg_list))
+		iod->flags |= IOD_SMALL_DESCRIPTOR;
 
-	sg_list = dma_pool_alloc(pool, GFP_ATOMIC, &sgl_dma);
-	if (!sg_list) {
-		iod->nr_allocations = -1;
+	sg_list = dma_pool_alloc(nvme_dma_pool(nvmeq, iod), GFP_ATOMIC,
+			&sgl_dma);
+	if (!sg_list)
 		return BLK_STS_RESOURCE;
-	}
-
-	iod->list[0].sg_list = sg_list;
+	iod->descriptors[iod->nr_descriptors++] = sg_list;
 	iod->first_dma = sgl_dma;
 
 	nvme_pci_sgl_set_seg(&cmd->dptr.sgl, sgl_dma, entries);
···
 static blk_status_t nvme_map_data(struct nvme_dev *dev, struct request *req,
 		struct nvme_command *cmnd)
 {
+	struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	blk_status_t ret = BLK_STS_RESOURCE;
 	int rc;
 
 	if (blk_rq_nr_phys_segments(req) == 1) {
-		struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
 		struct bio_vec bv = req_bvec(req);
 
 		if (!is_pci_p2pdma_page(bv.bv_page)) {
···
 	}
 
 	if (nvme_pci_use_sgls(dev, req, iod->sgt.nents))
-		ret = nvme_pci_setup_sgls(dev, req, &cmnd->rw);
+		ret = nvme_pci_setup_sgls(nvmeq, req, &cmnd->rw);
 	else
-		ret = nvme_pci_setup_prps(dev, req, &cmnd->rw);
+		ret = nvme_pci_setup_prps(nvmeq, req, &cmnd->rw);
 	if (ret != BLK_STS_OK)
 		goto out_unmap_sg;
 	return BLK_STS_OK;
···
 static blk_status_t nvme_pci_setup_meta_sgls(struct nvme_dev *dev,
 					     struct request *req)
 {
+	struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	struct nvme_rw_command *cmnd = &iod->cmd.rw;
 	struct nvme_sgl_desc *sg_list;
···
 	if (rc)
 		goto out_free_sg;
 
-	sg_list = dma_pool_alloc(dev->prp_small_pool, GFP_ATOMIC, &sgl_dma);
+	sg_list = dma_pool_alloc(nvmeq->descriptor_pools.small, GFP_ATOMIC,
+			&sgl_dma);
 	if (!sg_list)
 		goto out_unmap_sg;
 
 	entries = iod->meta_sgt.nents;
-	iod->meta_list.sg_list = sg_list;
+	iod->meta_descriptor = sg_list;
 	iod->meta_dma = sgl_dma;
 
 	cmnd->flags = NVME_CMD_SGL_METASEG;
···
 
 static blk_status_t nvme_map_metadata(struct nvme_dev *dev, struct request *req)
 {
-	if (nvme_pci_metadata_use_sgls(dev, req))
+	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
+
+	if ((iod->cmd.common.flags & NVME_CMD_SGL_METABUF) &&
+	    nvme_pci_metadata_use_sgls(dev, req))
 		return nvme_pci_setup_meta_sgls(dev, req);
 	return nvme_pci_setup_meta_mptr(dev, req);
 }
···
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	blk_status_t ret;
 
-	iod->aborted = false;
-	iod->nr_allocations = -1;
+	iod->flags = 0;
+	iod->nr_descriptors = 0;
 	iod->sgt.nents = 0;
 	iod->meta_sgt.nents = 0;
 
···
 	return BLK_STS_OK;
 out_unmap_data:
 	if (blk_rq_nr_phys_segments(req))
-		nvme_unmap_data(dev, req);
+		nvme_unmap_data(dev, req->mq_hctx->driver_data, req);
 out_free_cmd:
 	nvme_cleanup_cmd(req);
 	return ret;
···
 }
 
 static __always_inline void nvme_unmap_metadata(struct nvme_dev *dev,
+						struct nvme_queue *nvmeq,
 						struct request *req)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
···
 		return;
 	}
 
-	dma_pool_free(dev->prp_small_pool, iod->meta_list.sg_list,
-		      iod->meta_dma);
+	dma_pool_free(nvmeq->descriptor_pools.small, iod->meta_descriptor,
+			iod->meta_dma);
 	dma_unmap_sgtable(dev->dev, &iod->meta_sgt, rq_dma_dir(req), 0);
 	mempool_free(iod->meta_sgt.sgl, dev->iod_meta_mempool);
 }
···
 	struct nvme_dev *dev = nvmeq->dev;
 
 	if (blk_integrity_rq(req))
-		nvme_unmap_metadata(dev, req);
+		nvme_unmap_metadata(dev, nvmeq, req);
 
 	if (blk_rq_nr_phys_segments(req))
-		nvme_unmap_data(dev, req);
+		nvme_unmap_data(dev, nvmeq, req);
 }
 
 static void nvme_pci_complete_rq(struct request *req)
···
 	 * returned to the driver, or if this is the admin queue.
 	 */
 	opcode = nvme_req(req)->cmd->common.opcode;
-	if (!nvmeq->qid || iod->aborted) {
+	if (!nvmeq->qid || (iod->flags & IOD_ABORTED)) {
 		dev_warn(dev->ctrl.device,
 			 "I/O tag %d (%04x) opcode %#x (%s) QID %d timeout, reset controller\n",
 			 req->tag, nvme_cid(req), opcode,
···
 		atomic_inc(&dev->ctrl.abort_limit);
 		return BLK_EH_RESET_TIMER;
 	}
-	iod->aborted = true;
+	iod->flags |= IOD_ABORTED;
 
 	cmd.abort.opcode = nvme_admin_abort_cmd;
 	cmd.abort.cid = nvme_cid(req);
···
 	return 0;
 }
 
-static int nvme_setup_prp_pools(struct nvme_dev *dev)
-{
-	size_t small_align = 256;
-
-	dev->prp_page_pool = dma_pool_create("prp list page", dev->dev,
-						NVME_CTRL_PAGE_SIZE,
-						NVME_CTRL_PAGE_SIZE, 0);
-	if (!dev->prp_page_pool)
-		return -ENOMEM;
-
-	if (dev->ctrl.quirks & NVME_QUIRK_DMAPOOL_ALIGN_512)
-		small_align = 512;
-
-	/* Optimisation for I/Os between 4k and 128k */
-	dev->prp_small_pool = dma_pool_create("prp list 256", dev->dev,
-						256, small_align, 0);
-	if (!dev->prp_small_pool) {
-		dma_pool_destroy(dev->prp_page_pool);
-		return -ENOMEM;
-	}
-	return 0;
-}
-
-static void nvme_release_prp_pools(struct nvme_dev *dev)
-{
-	dma_pool_destroy(dev->prp_page_pool);
-	dma_pool_destroy(dev->prp_small_pool);
-}
-
 static int nvme_pci_alloc_iod_mempool(struct nvme_dev *dev)
 {
 	size_t meta_size = sizeof(struct scatterlist) * (NVME_MAX_META_SEGS + 1);
···
 	struct nvme_dev *dev;
 	int ret = -ENOMEM;
 
-	dev = kzalloc_node(sizeof(*dev), GFP_KERNEL, node);
+	dev = kzalloc_node(struct_size(dev, descriptor_pools, nr_node_ids),
+			GFP_KERNEL, node);
 	if (!dev)
 		return ERR_PTR(-ENOMEM);
 	INIT_WORK(&dev->ctrl.reset_work, nvme_reset_work);
···
 	if (result)
 		goto out_uninit_ctrl;
 
-	result = nvme_setup_prp_pools(dev);
-	if (result)
-		goto out_dev_unmap;
-
 	result = nvme_pci_alloc_iod_mempool(dev);
 	if (result)
-		goto out_release_prp_pools;
+		goto out_dev_unmap;
 
 	dev_info(dev->ctrl.device, "pci function %s\n", dev_name(&pdev->dev));
 
···
 out_release_iod_mempool:
 	mempool_destroy(dev->iod_mempool);
 	mempool_destroy(dev->iod_meta_mempool);
-out_release_prp_pools:
-	nvme_release_prp_pools(dev);
 out_dev_unmap:
 	nvme_dev_unmap(dev);
 out_uninit_ctrl:
···
 	nvme_free_queues(dev, 0);
 	mempool_destroy(dev->iod_mempool);
 	mempool_destroy(dev->iod_meta_mempool);
-	nvme_release_prp_pools(dev);
+	nvme_release_descriptor_pools(dev);
 	nvme_dev_unmap(dev);
 	nvme_uninit_ctrl(&dev->ctrl);
 }
···
 	BUILD_BUG_ON(sizeof(struct nvme_create_sq) != 64);
 	BUILD_BUG_ON(sizeof(struct nvme_delete_queue) != 64);
 	BUILD_BUG_ON(IRQ_AFFINITY_MAX_SETS < 2);
-	BUILD_BUG_ON(NVME_MAX_SEGS > SGES_PER_PAGE);
-	BUILD_BUG_ON(sizeof(struct scatterlist) * NVME_MAX_SEGS > PAGE_SIZE);
-	BUILD_BUG_ON(nvme_pci_npages_prp() > NVME_MAX_NR_ALLOCATIONS);
+	BUILD_BUG_ON(nvme_pci_npages_prp() > NVME_MAX_NR_DESCRIPTORS);
 
 	return pci_register_driver(&nvme_driver);
 }

drivers/nvme/host/sysfs.c

···
 	&dev_attr_ana_state.attr,
 	&dev_attr_queue_depth.attr,
 	&dev_attr_numa_nodes.attr,
+	&dev_attr_delayed_removal_secs.attr,
 #endif
 	&dev_attr_io_passthru_err_log_enabled.attr,
 	NULL,
···
 	}
 	if (a == &dev_attr_queue_depth.attr || a == &dev_attr_numa_nodes.attr) {
 		if (nvme_disk_is_ns_head(dev_to_disk(dev)))
+			return 0;
+	}
+	if (a == &dev_attr_delayed_removal_secs.attr) {
+		struct gendisk *disk = dev_to_disk(dev);
+
+		if (!nvme_disk_is_ns_head(disk))
 			return 0;
 	}
 #endif

drivers/nvme/host/tcp.c

···
 }
 
 static inline void nvme_tcp_queue_request(struct nvme_tcp_request *req,
-		bool sync, bool last)
+		bool last)
 {
 	struct nvme_tcp_queue *queue = req->queue;
 	bool empty;
···
 	 * are on the same cpu, so we don't introduce contention.
 	 */
 	if (queue->io_cpu == raw_smp_processor_id() &&
-	    sync && empty && mutex_trylock(&queue->send_mutex)) {
+	    empty && mutex_trylock(&queue->send_mutex)) {
 		nvme_tcp_send_all(queue);
 		mutex_unlock(&queue->send_mutex);
 	}
···
 	req->ttag = pdu->ttag;
 
 	nvme_tcp_setup_h2c_data_pdu(req);
-	nvme_tcp_queue_request(req, false, true);
+
+	llist_add(&req->lentry, &queue->req_list);
+	queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work);
 
 	return 0;
 }
···
 	if (ret)
 		return ret;
 
-	if (ctrl->opts && ctrl->opts->concat && !ctrl->tls_pskid) {
+	if (ctrl->opts->concat && !ctrl->tls_pskid) {
 		/* See comments for nvme_tcp_key_revoke_needed() */
 		dev_dbg(ctrl->device, "restart admin queue for secure concatenation\n");
 		nvme_stop_keep_alive(ctrl);
···
 	ctrl->async_req.curr_bio = NULL;
 	ctrl->async_req.data_len = 0;
 
-	nvme_tcp_queue_request(&ctrl->async_req, true, true);
+	nvme_tcp_queue_request(&ctrl->async_req, true);
 }
 
 static void nvme_tcp_complete_timed_out(struct request *rq)
···
 
 	nvme_start_request(rq);
 
-	nvme_tcp_queue_request(req, true, bd->last);
+	nvme_tcp_queue_request(req, bd->last);
 
 	return BLK_STS_OK;
 }

drivers/nvme/target/admin-cmd.c

···
 	if (status != NVME_SC_SUCCESS)
 		goto complete;
 
-	/*
-	 * Note: The NVMe specification allows multiple SQs to use the same CQ.
-	 * However, the target code does not really support that. So for now,
-	 * prevent this and fail the command if sqid and cqid are different.
-	 */
-	if (!cqid || cqid != sqid) {
-		pr_err("SQ %u: Unsupported CQID %u\n", sqid, cqid);
-		status = NVME_SC_CQ_INVALID | NVME_STATUS_DNR;
+	status = nvmet_check_io_cqid(ctrl, cqid, false);
+	if (status != NVME_SC_SUCCESS) {
+		pr_err("SQ %u: Invalid CQID %u\n", sqid, cqid);
 		goto complete;
 	}
 
···
 		goto complete;
 	}
 
-	status = ctrl->ops->create_sq(ctrl, sqid, sq_flags, qsize, prp1);
+	status = ctrl->ops->create_sq(ctrl, sqid, cqid, sq_flags, qsize, prp1);
 
 complete:
 	nvmet_req_complete(req, status);
···
 		goto complete;
 	}
 
-	if (!cqid) {
+	status = nvmet_check_io_cqid(ctrl, cqid, false);
+	if (status != NVME_SC_SUCCESS)
+		goto complete;
+
+	if (!ctrl->cqs[cqid] || nvmet_cq_in_use(ctrl->cqs[cqid])) {
+		/* Some SQs are still using this CQ */
 		status = NVME_SC_QID_INVALID | NVME_STATUS_DNR;
 		goto complete;
 	}
-
-	status = nvmet_check_cqid(ctrl, cqid);
-	if (status != NVME_SC_SUCCESS)
-		goto complete;
 
 	status = ctrl->ops->delete_cq(ctrl, cqid);
 
···
 		goto complete;
 	}
 
-	if (!cqid) {
-		status = NVME_SC_QID_INVALID | NVME_STATUS_DNR;
-		goto complete;
-	}
-
-	status = nvmet_check_cqid(ctrl, cqid);
+	status = nvmet_check_io_cqid(ctrl, cqid, true);
 	if (status != NVME_SC_SUCCESS)
 		goto complete;
 

drivers/nvme/target/auth.c

···
 
 bool nvmet_check_auth_status(struct nvmet_req *req)
 {
-	if (req->sq->ctrl->host_key &&
-	    !req->sq->authenticated)
-		return false;
+	if (req->sq->ctrl->host_key) {
+		if (req->sq->qid > 0)
+			return true;
+		if (!req->sq->authenticated)
+			return false;
+	}
 	return true;
 }
 
···
 			 unsigned int shash_len)
 {
 	struct crypto_shash *shash_tfm;
-	struct shash_desc *shash;
+	SHASH_DESC_ON_STACK(shash, shash_tfm);
 	struct nvmet_ctrl *ctrl = req->sq->ctrl;
 	const char *hash_name;
 	u8 *challenge = req->sq->dhchap_c1;
···
 						    req->sq->dhchap_c1,
 						    challenge, shash_len);
 		if (ret)
-			goto out_free_challenge;
+			goto out;
 	}
 
 	pr_debug("ctrl %d qid %d host response seq %u transaction %d\n",
 		 ctrl->cntlid, req->sq->qid, req->sq->dhchap_s1,
 		 req->sq->dhchap_tid);
 
-	shash = kzalloc(sizeof(*shash) + crypto_shash_descsize(shash_tfm),
-			GFP_KERNEL);
-	if (!shash) {
-		ret = -ENOMEM;
-		goto out_free_challenge;
-	}
 	shash->tfm = shash_tfm;
 	ret = crypto_shash_init(shash);
 	if (ret)
···
 		goto out;
 	ret = crypto_shash_final(shash, response);
 out:
-	kfree(shash);
-out_free_challenge:
 	if (challenge != req->sq->dhchap_c1)
 		kfree(challenge);
 out_free_response:

drivers/nvme/target/core.c

···
 }
 EXPORT_SYMBOL_GPL(nvmet_req_complete);
 
+void nvmet_cq_init(struct nvmet_cq *cq)
+{
+	refcount_set(&cq->ref, 1);
+}
+EXPORT_SYMBOL_GPL(nvmet_cq_init);
+
+bool nvmet_cq_get(struct nvmet_cq *cq)
+{
+	return refcount_inc_not_zero(&cq->ref);
+}
+EXPORT_SYMBOL_GPL(nvmet_cq_get);
+
+void nvmet_cq_put(struct nvmet_cq *cq)
+{
+	if (refcount_dec_and_test(&cq->ref))
+		nvmet_cq_destroy(cq);
+}
+EXPORT_SYMBOL_GPL(nvmet_cq_put);
+
 void nvmet_cq_setup(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq,
 		u16 qid, u16 size)
 {
 	cq->qid = qid;
 	cq->size = size;
+
+	ctrl->cqs[qid] = cq;
+}
+
+void nvmet_cq_destroy(struct nvmet_cq *cq)
+{
+	struct nvmet_ctrl *ctrl = cq->ctrl;
+
+	if (ctrl) {
+		ctrl->cqs[cq->qid] = NULL;
+		nvmet_ctrl_put(cq->ctrl);
+		cq->ctrl = NULL;
+	}
 }
 
 void nvmet_sq_setup(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq,
···
 	complete(&sq->confirm_done);
 }
 
-u16 nvmet_check_cqid(struct nvmet_ctrl *ctrl, u16 cqid)
+u16 nvmet_check_cqid(struct nvmet_ctrl *ctrl, u16 cqid, bool create)
 {
-	if (!ctrl->sqs)
+	if (!ctrl->cqs)
 		return NVME_SC_INTERNAL | NVME_STATUS_DNR;
 
 	if (cqid > ctrl->subsys->max_qid)
 		return NVME_SC_QID_INVALID | NVME_STATUS_DNR;
 
-	/*
-	 * Note: For PCI controllers, the NVMe specifications allows multiple
-	 * SQs to share a single CQ. However, we do not support this yet, so
-	 * check that there is no SQ defined for a CQ. If one exist, then the
-	 * CQ ID is invalid for creation as well as when the CQ is being
-	 * deleted (as that would mean that the SQ was not deleted before the
-	 * CQ).
-	 */
-	if (ctrl->sqs[cqid])
+	if ((create && ctrl->cqs[cqid]) || (!create && !ctrl->cqs[cqid]))
 		return NVME_SC_QID_INVALID | NVME_STATUS_DNR;
 
 	return NVME_SC_SUCCESS;
 }
+
+u16 nvmet_check_io_cqid(struct nvmet_ctrl *ctrl, u16 cqid, bool create)
+{
+	if (!cqid)
+		return NVME_SC_QID_INVALID | NVME_STATUS_DNR;
+	return nvmet_check_cqid(ctrl, cqid, create);
+}
+
+bool nvmet_cq_in_use(struct nvmet_cq *cq)
+{
+	return refcount_read(&cq->ref) > 1;
+}
+EXPORT_SYMBOL_GPL(nvmet_cq_in_use);
 
 u16 nvmet_cq_create(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq,
 		    u16 qid, u16 size)
 {
 	u16 status;
 
-	status = nvmet_check_cqid(ctrl, qid);
+	status = nvmet_check_cqid(ctrl, qid, true);
 	if (status != NVME_SC_SUCCESS)
 		return status;
 
+	if (!kref_get_unless_zero(&ctrl->ref))
+		return NVME_SC_INTERNAL | NVME_STATUS_DNR;
+	cq->ctrl = ctrl;
+
+	nvmet_cq_init(cq);
 	nvmet_cq_setup(ctrl, cq, qid, size);
 
 	return NVME_SC_SUCCESS;
···
 }
 
 u16 nvmet_sq_create(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq,
-		    u16 sqid, u16 size)
+		    struct nvmet_cq *cq, u16 sqid, u16 size)
 {
 	u16 status;
 	int ret;
···
 	if (status != NVME_SC_SUCCESS)
 		return status;
 
-	ret = nvmet_sq_init(sq);
+	ret = nvmet_sq_init(sq, cq);
 	if (ret) {
 		status = NVME_SC_INTERNAL | NVME_STATUS_DNR;
 		goto ctrl_put;
···
 	wait_for_completion(&sq->free_done);
 	percpu_ref_exit(&sq->ref);
 	nvmet_auth_sq_free(sq);
+	nvmet_cq_put(sq->cq);
 
 	/*
 	 * we must reference the ctrl again after waiting for inflight IO
···
 	complete(&sq->free_done);
 }
 
-int nvmet_sq_init(struct nvmet_sq *sq)
+int nvmet_sq_init(struct nvmet_sq *sq, struct nvmet_cq *cq)
 {
 	int ret;
+
+	if (!nvmet_cq_get(cq))
+		return -EINVAL;
 
 	ret = percpu_ref_init(&sq->ref, nvmet_sq_free, 0, GFP_KERNEL);
 	if (ret) {
 		pr_err("percpu_ref init failed!\n");
+		nvmet_cq_put(cq);
 		return ret;
 	}
 	init_completion(&sq->free_done);
 	init_completion(&sq->confirm_done);
 	nvmet_auth_sq_init(sq);
+	sq->cq = cq;
 
 	return 0;
 }
···
 	return ret;
 }
 
-bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
-		struct nvmet_sq *sq, const struct nvmet_fabrics_ops *ops)
+bool nvmet_req_init(struct nvmet_req *req, struct nvmet_sq *sq,
+		const struct nvmet_fabrics_ops *ops)
 {
 	u8 flags = req->cmd->common.flags;
 	u16 status;
 
-	req->cq = cq;
+	req->cq = sq->cq;
 	req->sq = sq;
 	req->ops = ops;
 	req->sg = NULL;
···
 	if (!ctrl->sqs)
 		goto out_free_changed_ns_list;
 
+	ctrl->cqs = kcalloc(subsys->max_qid + 1, sizeof(struct nvmet_cq *),
+			   GFP_KERNEL);
+	if (!ctrl->cqs)
+		goto out_free_sqs;
+
 	ret = ida_alloc_range(&cntlid_ida,
 			     subsys->cntlid_min, subsys->cntlid_max,
 			     GFP_KERNEL);
 	if (ret < 0) {
 		args->status = NVME_SC_CONNECT_CTRL_BUSY | NVME_STATUS_DNR;
-		goto out_free_sqs;
+		goto out_free_cqs;
 	}
 	ctrl->cntlid = ret;
 
···
 	mutex_unlock(&subsys->lock);
 	nvmet_stop_keep_alive_timer(ctrl);
 	ida_free(&cntlid_ida, ctrl->cntlid);
+out_free_cqs:
+	kfree(ctrl->cqs);
 out_free_sqs:
 	kfree(ctrl->sqs);
 out_free_changed_ns_list:
···
 
 	nvmet_async_events_free(ctrl);
 	kfree(ctrl->sqs);
+	kfree(ctrl->cqs);
 	kfree(ctrl->changed_ns_list);
 	kfree(ctrl);
 

drivers/nvme/target/discovery.c

···
 	memcpy(e->trsvcid, port->disc_addr.trsvcid, NVMF_TRSVCID_SIZE);
 	memcpy(e->traddr, traddr, NVMF_TRADDR_SIZE);
 	memcpy(e->tsas.common, port->disc_addr.tsas.common, NVMF_TSAS_SIZE);
-	strncpy(e->subnqn, subsys_nqn, NVMF_NQN_SIZE);
+	strscpy(e->subnqn, subsys_nqn, NVMF_NQN_SIZE);
 }
 
 /*

drivers/nvme/target/fabrics-cmd.c

···
 		return NVME_SC_CONNECT_CTRL_BUSY | NVME_STATUS_DNR;
 	}
 
+	kref_get(&ctrl->ref);
+	old = cmpxchg(&req->cq->ctrl, NULL, ctrl);
+	if (old) {
+		pr_warn("queue already connected!\n");
+		req->error_loc = offsetof(struct nvmf_connect_command, opcode);
+		return NVME_SC_CONNECT_CTRL_BUSY | NVME_STATUS_DNR;
+	}
+
 	/* note: convert queue size from 0's-based value to 1's-based value */
 	nvmet_cq_setup(ctrl, req->cq, qid, sqsize + 1);
 	nvmet_sq_setup(ctrl, req->sq, qid, sqsize + 1);
···
 	bool needs_auth = nvmet_has_auth(ctrl, sq);
 	key_serial_t keyid = nvmet_queue_tls_keyid(sq);
 
-	/* Do not authenticate I/O queues for secure concatenation */
-	if (ctrl->concat && sq->qid)
+	/* Do not authenticate I/O queues */
+	if (sq->qid)
 		needs_auth = false;
 
 	if (keyid)

drivers/nvme/target/fc.c

···
 
 	nvmet_fc_prep_fcp_iodlist(assoc->tgtport, queue);
 
-	ret = nvmet_sq_init(&queue->nvme_sq);
+	nvmet_cq_init(&queue->nvme_cq);
+	ret = nvmet_sq_init(&queue->nvme_sq, &queue->nvme_cq);
 	if (ret)
 		goto out_fail_iodlist;
 
···
 	return queue;
 
 out_fail_iodlist:
+	nvmet_cq_put(&queue->nvme_cq);
 	nvmet_fc_destroy_fcp_iodlist(assoc->tgtport, queue);
 	destroy_workqueue(queue->work_q);
 out_free_queue:
···
 	flush_workqueue(queue->work_q);
 
 	nvmet_sq_destroy(&queue->nvme_sq);
+	nvmet_cq_put(&queue->nvme_cq);
 
 	nvmet_fc_tgt_q_put(queue);
 }
···
 {
 	lockdep_assert_held(&nvmet_fc_tgtlock);
 
+	nvmet_fc_tgtport_get(tgtport);
 	pe->tgtport = tgtport;
 	tgtport->pe = pe;
 
···
 	unsigned long flags;
 
 	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
-	if (pe->tgtport)
+	if (pe->tgtport) {
+		nvmet_fc_tgtport_put(pe->tgtport);
 		pe->tgtport->pe = NULL;
+	}
 	list_del(&pe->pe_list);
 	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
 }
···
 
 	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
 	pe = tgtport->pe;
-	if (pe)
+	if (pe) {
+		nvmet_fc_tgtport_put(pe->tgtport);
 		pe->tgtport = NULL;
+	}
 	tgtport->pe = NULL;
 	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
 }
···
 	list_for_each_entry(pe, &nvmet_fc_portentry_list, pe_list) {
 		if (tgtport->fc_target_port.node_name == pe->node_name &&
 		    tgtport->fc_target_port.port_name == pe->port_name) {
+			if (!nvmet_fc_tgtport_get(tgtport))
+				continue;
+
 			WARN_ON(pe->tgtport);
 			tgtport->pe = pe;
 			pe->tgtport = tgtport;
···
 	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
 }
 
+static void
+nvmet_fc_free_pending_reqs(struct nvmet_fc_tgtport *tgtport)
+{
+	struct nvmet_fc_ls_req_op *lsop;
+	struct nvmefc_ls_req *lsreq;
+	struct nvmet_fc_ls_iod *iod;
+	int i;
+
+	iod = tgtport->iod;
+	for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++)
+		cancel_work(&iod->work);
+
+	/*
+	 * After this point the connection is lost and thus any pending
+	 * request can't be processed by the normal completion path. This
+	 * is likely a request from nvmet_fc_send_ls_req_async.
+	 */
+	while ((lsop = list_first_entry_or_null(&tgtport->ls_req_list,
+				struct nvmet_fc_ls_req_op, lsreq_list))) {
+		list_del(&lsop->lsreq_list);
+
+		if (!lsop->req_queued)
+			continue;
+
+		lsreq = &lsop->ls_req;
+		fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma,
+				    (lsreq->rqstlen + lsreq->rsplen),
+				    DMA_BIDIRECTIONAL);
+		nvmet_fc_tgtport_put(tgtport);
+		kfree(lsop);
+	}
+}
+
 /**
  * nvmet_fc_unregister_targetport - transport entry point called by an
  *                              LLDD to deregister/remove a previously
···
 
 	flush_workqueue(nvmet_wq);
 
-	/*
-	 * should terminate LS's as well. However, LS's will be generated
-	 * at the tail end of association termination, so they likely don't
-	 * exist yet. And even if they did, it's worthwhile to just let
-	 * them finish and targetport ref counting will clean things up.
-	 */
-
+	nvmet_fc_free_pending_reqs(tgtport);
 	nvmet_fc_tgtport_put(tgtport);
 
 	return 0;
···
 	fod->data_sg = NULL;
 	fod->data_sg_cnt = 0;
 
-	ret = nvmet_req_init(&fod->req,
-				&fod->queue->nvme_cq,
-				&fod->queue->nvme_sq,
-				&nvmet_fc_tgt_fcp_ops);
+	ret = nvmet_req_init(&fod->req, &fod->queue->nvme_sq,
+			&nvmet_fc_tgt_fcp_ops);
 	if (!ret) {
 		/* bad SQE content or invalid ctrl state */
 		/* nvmet layer has already called op done to send rsp. */
···
 	list_for_each_entry(tgtport, &nvmet_fc_target_list, tgt_list) {
 		if ((tgtport->fc_target_port.node_name == traddr.nn) &&
 		    (tgtport->fc_target_port.port_name == traddr.pn)) {
+			if (!nvmet_fc_tgtport_get(tgtport))
+				continue;
+
 			/* a FC port can only be 1 nvmet port id */
 			if (!tgtport->pe) {
 				nvmet_fc_portentry_bind(tgtport, pe, port);
 				ret = 0;
 			} else
 				ret = -EALREADY;
+
+			nvmet_fc_tgtport_put(tgtport);
 			break;
 		}
 	}
···
 nvmet_fc_remove_port(struct nvmet_port *port)
 {
 	struct nvmet_fc_port_entry *pe = port->priv;
+	struct nvmet_fc_tgtport *tgtport = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
+	if (pe->tgtport && nvmet_fc_tgtport_get(pe->tgtport))
+		tgtport = pe->tgtport;
+	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
 
 	nvmet_fc_portentry_unbind(pe);
 
-	/* terminate any outstanding associations */
-	__nvmet_fc_free_assocs(pe->tgtport);
+	if (tgtport) {
+		/* terminate any outstanding associations */
+		__nvmet_fc_free_assocs(tgtport);
+		nvmet_fc_tgtport_put(tgtport);
+	}
 
 	kfree(pe);
 }
···
 nvmet_fc_discovery_chg(struct nvmet_port *port)
 {
 	struct nvmet_fc_port_entry *pe = port->priv;
-	struct nvmet_fc_tgtport *tgtport = pe->tgtport;
+	struct nvmet_fc_tgtport *tgtport = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
+	if (pe->tgtport && nvmet_fc_tgtport_get(pe->tgtport))
+		tgtport = pe->tgtport;
+	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
+
+	if (!tgtport)
+		return;
 
 	if (tgtport && tgtport->ops->discovery_event)
 		tgtport->ops->discovery_event(&tgtport->fc_target_port);
+
+	nvmet_fc_tgtport_put(tgtport);
 }
 
 static ssize_t

drivers/nvme/target/fcloop.c

···
 struct fcloop_lport {
 	struct nvme_fc_local_port *localport;
 	struct list_head lport_list;
-	struct completion unreg_done;
 	refcount_t ref;
 };
 
 struct fcloop_lport_priv {
 	struct fcloop_lport *lport;
 };
+
+/* The port is already being removed, avoid double free */
+#define PORT_DELETED	0
 
 struct fcloop_rport {
 	struct nvme_fc_remote_port	*remoteport;
···
 	spinlock_t			lock;
 	struct list_head		ls_list;
 	struct work_struct		ls_work;
+	unsigned long			flags;
 };
 
 struct fcloop_tport {
···
 	spinlock_t			lock;
 	struct list_head		ls_list;
 	struct work_struct		ls_work;
+	unsigned long			flags;
 };
 
 struct fcloop_nport {
···
 	spinlock_t			inilock;
 };
 
+/* SLAB cache for fcloop_lsreq structures */
+static struct kmem_cache *lsreq_cache;
+
 static inline struct fcloop_lsreq *
 ls_rsp_to_lsreq(struct nvmefc_ls_rsp *lsrsp)
 {
···
 		 * callee may free memory containing tls_req.
 		 * do not reference lsreq after this.
 		 */
+		kmem_cache_free(lsreq_cache, tls_req);
 
 		spin_lock(&rport->lock);
 	}
···
 			struct nvme_fc_remote_port *remoteport,
 			struct nvmefc_ls_req *lsreq)
 {
-	struct fcloop_lsreq *tls_req = lsreq->private;
 	struct fcloop_rport *rport = remoteport->private;
+	struct fcloop_lsreq *tls_req;
 	int ret = 0;
 
+	tls_req = kmem_cache_alloc(lsreq_cache, GFP_KERNEL);
+	if (!tls_req)
+		return -ENOMEM;
 	tls_req->lsreq = lsreq;
 	INIT_LIST_HEAD(&tls_req->ls_list);
 
···
 
 	lsrsp->done(lsrsp);
 
-	if (remoteport) {
-		rport = remoteport->private;
-		spin_lock(&rport->lock);
-		list_add_tail(&tls_req->ls_list, &rport->ls_list);
-		spin_unlock(&rport->lock);
-		queue_work(nvmet_wq, &rport->ls_work);
+	if (!remoteport) {
+		kmem_cache_free(lsreq_cache, tls_req);
+		return 0;
 	}
+
+	rport = remoteport->private;
+	spin_lock(&rport->lock);
+	list_add_tail(&tls_req->ls_list, &rport->ls_list);
+	spin_unlock(&rport->lock);
+	queue_work(nvmet_wq, &rport->ls_work);
 
 	return 0;
 }
···
 		 * callee may free memory containing tls_req.
 		 * do not reference lsreq after this.
 		 */
+		kmem_cache_free(lsreq_cache, tls_req);
 
 		spin_lock(&tport->lock);
 	}
···
 fcloop_t2h_ls_req(struct nvmet_fc_target_port *targetport, void *hosthandle,
 			struct nvmefc_ls_req *lsreq)
 {
-	struct fcloop_lsreq *tls_req = lsreq->private;
 	struct fcloop_tport *tport = targetport->private;
+	struct fcloop_lsreq *tls_req;
 	int ret = 0;
 
 	/*
···
 	 * hosthandle ignored as fcloop currently is
 	 * 1:1 tgtport vs remoteport
 	 */
+
+	tls_req = kmem_cache_alloc(lsreq_cache, GFP_KERNEL);
+	if (!tls_req)
+		return -ENOMEM;
 	tls_req->lsreq = lsreq;
 	INIT_LIST_HEAD(&tls_req->ls_list);
 
···
 	ret = nvme_fc_rcv_ls_req(tport->remoteport, &tls_req->ls_rsp,
 				 lsreq->rqstaddr, lsreq->rqstlen);
 
+	if (ret)
+		kmem_cache_free(lsreq_cache, tls_req);
+
 	return ret;
 }
 
···
 	struct nvmet_fc_target_port *targetport = rport->targetport;
 	struct fcloop_tport *tport;
 
+	if (!targetport) {
+		/*
+		 * The target port is gone. The target doesn't expect any
+		 * response anymore and the ->done call is not valid
+		 * because the resources have been freed by
+		 * nvmet_fc_free_pending_reqs.
+		 *
+		 * We end up here from delete association exchange:
+		 * nvmet_fc_xmt_disconnect_assoc sends an async request.
+		 */
+		kmem_cache_free(lsreq_cache, tls_req);
+		return 0;
+	}
+
 	memcpy(lsreq->rspaddr, lsrsp->rspbuf,
 		((lsreq->rsplen < lsrsp->rsplen) ?
 				lsreq->rsplen : lsrsp->rsplen));
 	lsrsp->done(lsrsp);
 
-	if (targetport) {
-		tport = targetport->private;
-		spin_lock(&tport->lock);
-		list_add_tail(&tls_req->ls_list, &tport->ls_list);
-		spin_unlock(&tport->lock);
-		queue_work(nvmet_wq, &tport->ls_work);
-	}
+	tport = targetport->private;
+	spin_lock(&tport->lock);
+	list_add_tail(&tls_req->ls_list, &tport->ls_list);
+	spin_unlock(&tport->lock);
+	queue_work(nvmet_wq, &tport->ls_work);
 
 	return 0;
 }
···
 	}
 
 	/* release original io reference on tgt struct */
-	fcloop_tfcp_req_put(tfcp_req);
+	if (tfcp_req)
+		fcloop_tfcp_req_put(tfcp_req);
 }
 
 static bool drop_fabric_opcode;
···
 {
 	struct fcloop_fcpreq *tfcp_req =
 		container_of(work, struct fcloop_fcpreq, fcp_rcv_work);
-	struct nvmefc_fcp_req *fcpreq = tfcp_req->fcpreq;
+	struct nvmefc_fcp_req *fcpreq;
 	unsigned long flags;
 	int ret = 0;
 	bool aborted = false;
 
 	spin_lock_irqsave(&tfcp_req->reqlock, flags);
+	fcpreq = tfcp_req->fcpreq;
 	switch (tfcp_req->inistate) {
 	case INI_IO_START:
 		tfcp_req->inistate = INI_IO_ACTIVE;
···
 	}
 	spin_unlock_irqrestore(&tfcp_req->reqlock, flags);
 
-	if (unlikely(aborted))
-		ret = -ECANCELED;
-	else {
-		if (likely(!check_for_drop(tfcp_req)))
-			ret = nvmet_fc_rcv_fcp_req(tfcp_req->tport->targetport,
-				&tfcp_req->tgt_fcp_req,
-				fcpreq->cmdaddr, fcpreq->cmdlen);
-		else
-			pr_info("%s: dropped command ********\n", __func__);
+	if (unlikely(aborted)) {
+		/* the abort handler will call fcloop_call_host_done */
+		return;
 	}
+
+	if (unlikely(check_for_drop(tfcp_req))) {
+		pr_info("%s: dropped command ********\n", __func__);
+		return;
+	}
+
+	ret = nvmet_fc_rcv_fcp_req(tfcp_req->tport->targetport,
+				   &tfcp_req->tgt_fcp_req,
+				   fcpreq->cmdaddr, fcpreq->cmdlen);
 	if (ret)
 		fcloop_call_host_done(fcpreq, tfcp_req, ret);
 }
···
 	unsigned long flags;
 
 	spin_lock_irqsave(&tfcp_req->reqlock, flags);
-	fcpreq = tfcp_req->fcpreq;
 	switch (tfcp_req->inistate) {
 	case INI_IO_ABORTED:
+		fcpreq = tfcp_req->fcpreq;
+		tfcp_req->fcpreq = NULL;
 		break;
 	case INI_IO_COMPLETED:
 		completed = true;
 		break;
 	default:
 		spin_unlock_irqrestore(&tfcp_req->reqlock, flags);
+		fcloop_tfcp_req_put(tfcp_req);
 		WARN_ON(1);
 		return;
 	}
···
 	if (tfcp_req->tport->targetport)
 		nvmet_fc_rcv_fcp_abort(tfcp_req->tport->targetport,
 					&tfcp_req->tgt_fcp_req);
-
-	spin_lock_irqsave(&tfcp_req->reqlock, flags);
-	tfcp_req->fcpreq = NULL;
-	spin_unlock_irqrestore(&tfcp_req->reqlock, flags);
 
 	fcloop_call_host_done(fcpreq, tfcp_req, -ECANCELED);
 	/* call_host_done releases reference for abort downcall */
···
 
 	spin_lock(&inireq->inilock);
 	tfcp_req = inireq->tfcp_req;
-	if (tfcp_req)
-		fcloop_tfcp_req_get(tfcp_req);
+	if (tfcp_req) {
+		if (!fcloop_tfcp_req_get(tfcp_req))
+			tfcp_req = NULL;
+	}
 	spin_unlock(&inireq->inilock);
 
-	if (!tfcp_req)
+	if (!tfcp_req) {
 		/* abort has already been called */
-		return;
+		goto out_host_done;
+	}
 
 	/* break initiator/target relationship for io */
 	spin_lock_irqsave(&tfcp_req->reqlock, flags);
···
 	default:
 		spin_unlock_irqrestore(&tfcp_req->reqlock, flags);
 		WARN_ON(1);
-		return;
+		goto out_host_done;
 	}
 	spin_unlock_irqrestore(&tfcp_req->reqlock, flags);
 
···
 		 */
 		fcloop_tfcp_req_put(tfcp_req);
 	}
+
+	return;
+
+out_host_done:
+	fcloop_call_host_done(fcpreq, tfcp_req, -ECANCELED);
 }
 
 static void
···
 static void
 fcloop_nport_put(struct fcloop_nport *nport)
 {
+	unsigned long flags;
+
 	if (!refcount_dec_and_test(&nport->ref))
 		return;
+
+	spin_lock_irqsave(&fcloop_lock, flags);
+	list_del(&nport->nport_list);
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	if (nport->lport)
+		fcloop_lport_put(nport->lport);
 
 	kfree(nport);
 }
···
 	struct fcloop_lport_priv *lport_priv = localport->private;
 	struct fcloop_lport *lport = lport_priv->lport;
 
-	/* release any threads waiting for the unreg to complete */
-	complete(&lport->unreg_done);
-
 	fcloop_lport_put(lport);
 }
 
···
 fcloop_remoteport_delete(struct nvme_fc_remote_port *remoteport)
 {
 	struct fcloop_rport *rport = remoteport->private;
+	bool put_port = false;
+	unsigned long flags;
 
 	flush_work(&rport->ls_work);
-	fcloop_nport_put(rport->nport);
+
+	spin_lock_irqsave(&fcloop_lock, flags);
+	if (!test_and_set_bit(PORT_DELETED, &rport->flags))
+		put_port = true;
+	rport->nport->rport = NULL;
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	if (put_port)
+		fcloop_nport_put(rport->nport);
 }
 
 static void
 fcloop_targetport_delete(struct nvmet_fc_target_port *targetport)
 {
 	struct fcloop_tport *tport = targetport->private;
+	bool put_port = false;
+	unsigned long flags;
 
 	flush_work(&tport->ls_work);
-	fcloop_nport_put(tport->nport);
+
+	spin_lock_irqsave(&fcloop_lock, flags);
+	if (!test_and_set_bit(PORT_DELETED, &tport->flags))
+		put_port = true;
+	tport->nport->tport = NULL;
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	if (put_port)
+		fcloop_nport_put(tport->nport);
 }
 
 #define	FCLOOP_HW_QUEUES		4
···
 	/* sizes of additional private data for data structures */
 	.local_priv_sz		= sizeof(struct fcloop_lport_priv),
 	.remote_priv_sz		= sizeof(struct fcloop_rport),
-	.lsrqst_priv_sz		= sizeof(struct fcloop_lsreq),
 	.fcprqst_priv_sz	= sizeof(struct fcloop_ini_fcpreq),
 };
 
···
 	.target_features	= 0,
 	/* sizes of additional private data for data structures */
 	.target_priv_sz		= sizeof(struct fcloop_tport),
-	.lsrqst_priv_sz		= sizeof(struct fcloop_lsreq),
 };
 
 static ssize_t
···
 }
 
 static int
-__wait_localport_unreg(struct fcloop_lport *lport)
+__localport_unreg(struct fcloop_lport *lport)
 {
-	int ret;
-
-	init_completion(&lport->unreg_done);
-
-	ret = nvme_fc_unregister_localport(lport->localport);
-
-	if (!ret)
-		wait_for_completion(&lport->unreg_done);
-
-	return ret;
+	return nvme_fc_unregister_localport(lport->localport);
 }
 
+static struct fcloop_nport *
+__fcloop_nport_lookup(u64 node_name, u64 port_name)
+{
+	struct fcloop_nport *nport;
+
+	list_for_each_entry(nport, &fcloop_nports, nport_list) {
+		if (nport->node_name != node_name ||
+		    nport->port_name != port_name)
+			continue;
+
+		if (fcloop_nport_get(nport))
+			return nport;
+
+		break;
+	}
+
+	return NULL;
+}
+
+static struct fcloop_nport *
+fcloop_nport_lookup(u64 node_name, u64 port_name)
+{
+	struct fcloop_nport *nport;
+	unsigned long flags;
+
+	spin_lock_irqsave(&fcloop_lock, flags);
+	nport = __fcloop_nport_lookup(node_name, port_name);
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	return nport;
+}
+
+static struct fcloop_lport *
+__fcloop_lport_lookup(u64 node_name, u64 port_name)
+{
+	struct fcloop_lport *lport;
+
+	list_for_each_entry(lport, &fcloop_lports, lport_list) {
+		if (lport->localport->node_name != node_name ||
+		    lport->localport->port_name != port_name)
+			continue;
+
+		if (fcloop_lport_get(lport))
+			return lport;
+
+		break;
+	}
+
+	return NULL;
+}
+
+static struct fcloop_lport *
+fcloop_lport_lookup(u64 node_name, u64 port_name)
+{
+	struct fcloop_lport *lport;
+	unsigned long flags;
+
+	spin_lock_irqsave(&fcloop_lock, flags);
+	lport = __fcloop_lport_lookup(node_name, port_name);
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	return lport;
+}
 
 static ssize_t
 fcloop_delete_local_port(struct device *dev, struct device_attribute *attr,
 		const char *buf, size_t count)
 {
-	struct fcloop_lport *tlport, *lport = NULL;
+	struct fcloop_lport *lport;
 	u64 nodename, portname;
-	unsigned long flags;
 	int ret;
 
 	ret = fcloop_parse_nm_options(dev, &nodename, &portname, buf);
 	if (ret)
 		return ret;
 
-	spin_lock_irqsave(&fcloop_lock, flags);
-
-	list_for_each_entry(tlport, &fcloop_lports, lport_list) {
-		if (tlport->localport->node_name == nodename &&
-		    tlport->localport->port_name == portname) {
-			if (!fcloop_lport_get(tlport))
-				break;
-			lport = tlport;
-			break;
-		}
-	}
-	spin_unlock_irqrestore(&fcloop_lock, flags);
-
+	lport = fcloop_lport_lookup(nodename, portname);
 	if (!lport)
 		return -ENOENT;
 
-	ret = __wait_localport_unreg(lport);
+	ret = __localport_unreg(lport);
 	fcloop_lport_put(lport);
 
 	return ret ? ret : count;
···
 static struct fcloop_nport *
 fcloop_alloc_nport(const char *buf, size_t count, bool remoteport)
 {
-	struct fcloop_nport *newnport, *nport = NULL;
-	struct fcloop_lport *tmplport, *lport = NULL;
+	struct fcloop_nport *newnport, *nport;
+	struct fcloop_lport *lport;
 	struct fcloop_ctrl_options *opts;
 	unsigned long flags;
 	u32 opts_mask = (remoteport) ? RPORT_OPTS : TGTPORT_OPTS;
···
 		goto out_free_opts;
 
 	/* everything there ? */
-	if ((opts->mask & opts_mask) != opts_mask) {
-		ret = -EINVAL;
+	if ((opts->mask & opts_mask) != opts_mask)
 		goto out_free_opts;
-	}
 
 	newnport = kzalloc(sizeof(*newnport), GFP_KERNEL);
 	if (!newnport)
···
 	refcount_set(&newnport->ref, 1);
 
 	spin_lock_irqsave(&fcloop_lock, flags);
-
-	list_for_each_entry(tmplport, &fcloop_lports, lport_list) {
-		if (tmplport->localport->node_name == opts->wwnn &&
-		    tmplport->localport->port_name == opts->wwpn)
-			goto out_invalid_opts;
-
-		if (tmplport->localport->node_name == opts->lpwwnn &&
-		    tmplport->localport->port_name == opts->lpwwpn)
-			lport = tmplport;
+	lport = __fcloop_lport_lookup(opts->wwnn, opts->wwpn);
+	if (lport) {
+		/* invalid configuration */
+		fcloop_lport_put(lport);
+		goto out_free_newnport;
 	}
 
 	if (remoteport) {
-		if (!lport)
-			goto out_invalid_opts;
-		newnport->lport = lport;
-	}
-
-	list_for_each_entry(nport, &fcloop_nports, nport_list) {
-		if (nport->node_name == opts->wwnn &&
-		    nport->port_name == opts->wwpn) {
-			if ((remoteport && nport->rport) ||
-			    (!remoteport && nport->tport)) {
-				nport = NULL;
-				goto out_invalid_opts;
-			}
-
-			fcloop_nport_get(nport);
-
-			spin_unlock_irqrestore(&fcloop_lock, flags);
-
-			if (remoteport)
-				nport->lport = lport;
-			if (opts->mask & NVMF_OPT_ROLES)
-				nport->port_role = opts->roles;
-			if (opts->mask & NVMF_OPT_FCADDR)
-				nport->port_id = opts->fcaddr;
+		lport = __fcloop_lport_lookup(opts->lpwwnn, opts->lpwwpn);
+		if (!lport) {
+			/* invalid configuration */
 			goto out_free_newnport;
 		}
 	}
 
-	list_add_tail(&newnport->nport_list, &fcloop_nports);
+	nport = __fcloop_nport_lookup(opts->wwnn, opts->wwpn);
+	if (nport) {
+		if ((remoteport && nport->rport) ||
+		    (!remoteport && nport->tport)) {
+			/* invalid configuration */
+			goto out_put_nport;
+		}
 
+		/* found existing nport, discard the new nport */
+		kfree(newnport);
+	} else {
+		list_add_tail(&newnport->nport_list, &fcloop_nports);
+		nport = newnport;
+	}
+
+	if (opts->mask & NVMF_OPT_ROLES)
+		nport->port_role = opts->roles;
+	if (opts->mask & NVMF_OPT_FCADDR)
+		nport->port_id = opts->fcaddr;
+	if (lport) {
+		if (!nport->lport)
+			nport->lport = lport;
+		else
+			fcloop_lport_put(lport);
+	}
 	spin_unlock_irqrestore(&fcloop_lock, flags);
 
 	kfree(opts);
-	return newnport;
+	return nport;
 
-out_invalid_opts:
-	spin_unlock_irqrestore(&fcloop_lock, flags);
+out_put_nport:
+	if (lport)
+		fcloop_lport_put(lport);
+	fcloop_nport_put(nport);
 out_free_newnport:
+	spin_unlock_irqrestore(&fcloop_lock, flags);
 	kfree(newnport);
 out_free_opts:
 	kfree(opts);
-	return nport;
+	return NULL;
 }
 
 static ssize_t
···
 	rport->nport = nport;
 	rport->lport = nport->lport;
 	nport->rport = rport;
+	rport->flags = 0;
 	spin_lock_init(&rport->lock);
 	INIT_WORK(&rport->ls_work, fcloop_rport_lsrqst_work);
 	INIT_LIST_HEAD(&rport->ls_list);
···
 {
 	struct fcloop_rport *rport = nport->rport;
 
+	lockdep_assert_held(&fcloop_lock);
+
 	if (rport && nport->tport)
 		nport->tport->remoteport = NULL;
 	nport->rport = NULL;
-
-	list_del(&nport->nport_list);
 
 	return rport;
 }
···
 static int
 __remoteport_unreg(struct fcloop_nport *nport, struct fcloop_rport *rport)
 {
-	if (!rport)
-		return -EALREADY;
-
 	return nvme_fc_unregister_remoteport(rport->remoteport);
 }
 
···
 fcloop_delete_remote_port(struct device *dev, struct device_attribute *attr,
 		const char *buf, size_t count)
 {
-	struct fcloop_nport *nport = NULL, *tmpport;
-	static struct fcloop_rport *rport;
+	struct fcloop_nport *nport;
+	struct fcloop_rport *rport;
 	u64 nodename, portname;
 	unsigned long flags;
 	int ret;
···
 	if (ret)
 		return ret;
 
-	spin_lock_irqsave(&fcloop_lock, flags);
-
-	list_for_each_entry(tmpport, &fcloop_nports, nport_list) {
-		if (tmpport->node_name == nodename &&
-		    tmpport->port_name == portname && tmpport->rport) {
-			nport = tmpport;
-			rport = __unlink_remote_port(nport);
-			break;
-		}
-	}
-
-	spin_unlock_irqrestore(&fcloop_lock, flags);
-
+	nport = fcloop_nport_lookup(nodename, portname);
 	if (!nport)
 		return -ENOENT;
 
+	spin_lock_irqsave(&fcloop_lock, flags);
+	rport = __unlink_remote_port(nport);
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	if (!rport) {
+		ret = -ENOENT;
+		goto out_nport_put;
+	}
+
 	ret = __remoteport_unreg(nport, rport);
+
+out_nport_put:
+	fcloop_nport_put(nport);
 
 	return ret ? ret : count;
 }
···
 	tport->nport = nport;
 	tport->lport = nport->lport;
 	nport->tport = tport;
+	tport->flags = 0;
 	spin_lock_init(&tport->lock);
 	INIT_WORK(&tport->ls_work, fcloop_tport_lsrqst_work);
 	INIT_LIST_HEAD(&tport->ls_list);
···
 {
 	struct fcloop_tport *tport = nport->tport;
 
+	lockdep_assert_held(&fcloop_lock);
+
 	if (tport && nport->rport)
 		nport->rport->targetport = NULL;
 	nport->tport = NULL;
···
 static int
 __targetport_unreg(struct fcloop_nport *nport, struct fcloop_tport *tport)
 {
-	if (!tport)
-		return -EALREADY;
-
 	return nvmet_fc_unregister_targetport(tport->targetport);
 }
 
···
 fcloop_delete_target_port(struct device *dev, struct device_attribute *attr,
 		const char *buf, size_t count)
 {
-	struct fcloop_nport *nport = NULL, *tmpport;
-	struct fcloop_tport *tport = NULL;
+	struct fcloop_nport *nport;
+	struct fcloop_tport *tport;
 	u64 nodename, portname;
 	unsigned long flags;
 	int ret;
···
 	if (ret)
 		return ret;
 
-	spin_lock_irqsave(&fcloop_lock, flags);
-
-	list_for_each_entry(tmpport, &fcloop_nports, nport_list) {
-		if (tmpport->node_name == nodename &&
-		    tmpport->port_name == portname && tmpport->tport) {
-			nport = tmpport;
-			tport = __unlink_target_port(nport);
-			break;
-		}
-	}
-
-	spin_unlock_irqrestore(&fcloop_lock, flags);
-
+	nport = fcloop_nport_lookup(nodename, portname);
 	if (!nport)
 		return -ENOENT;
 
+	spin_lock_irqsave(&fcloop_lock, flags);
+	tport = __unlink_target_port(nport);
+	spin_unlock_irqrestore(&fcloop_lock, flags);
+
+	if (!tport) {
+		ret = -ENOENT;
+		goto out_nport_put;
+	}
+
 	ret = __targetport_unreg(nport, tport);
+
+out_nport_put:
+	fcloop_nport_put(nport);
 
 	return ret ? ret : count;
 }
···
 };
 static struct device *fcloop_device;
 
-
 static int __init fcloop_init(void)
 {
 	int ret;
 
+	lsreq_cache = kmem_cache_create("lsreq_cache",
+				sizeof(struct fcloop_lsreq), 0,
+				0, NULL);
+	if (!lsreq_cache)
+		return -ENOMEM;
+
 	ret = class_register(&fcloop_class);
 	if (ret) {
 		pr_err("couldn't register class fcloop\n");
-		return ret;
+		goto out_destroy_cache;
 	}
 
 	fcloop_device = device_create_with_groups(
···
 
 out_destroy_class:
 	class_unregister(&fcloop_class);
+out_destroy_cache:
+	kmem_cache_destroy(lsreq_cache);
 	return ret;
 }
 
 static void __exit fcloop_exit(void)
 {
-	struct fcloop_lport *lport = NULL;
-	struct fcloop_nport *nport = NULL;
+	struct fcloop_lport *lport;
+	struct fcloop_nport *nport;
 	struct fcloop_tport *tport;
 	struct fcloop_rport *rport;
 	unsigned long flags;
···
 	for (;;) {
 		nport = list_first_entry_or_null(&fcloop_nports,
 						typeof(*nport), nport_list);
-		if (!nport)
+		if (!nport || !fcloop_nport_get(nport))
 			break;
 
 		tport = __unlink_target_port(nport);
···
 
 		spin_unlock_irqrestore(&fcloop_lock, flags);
 
-		ret = __targetport_unreg(nport, tport);
-		if (ret)
-			pr_warn("%s: Failed deleting target port\n", __func__);
+		if (tport) {
+			ret = __targetport_unreg(nport, tport);
+			if (ret)
+				pr_warn("%s: Failed deleting target port\n",
+					__func__);
+		}
 
-		ret = __remoteport_unreg(nport, rport);
-		if (ret)
-			pr_warn("%s: Failed deleting remote port\n", __func__);
+		if (rport) {
+			ret = __remoteport_unreg(nport, rport);
+			if (ret)
+				pr_warn("%s: Failed deleting remote port\n",
+					__func__);
+		}
+
+		fcloop_nport_put(nport);
 
 		spin_lock_irqsave(&fcloop_lock, flags);
 	}
···
 
 		spin_unlock_irqrestore(&fcloop_lock, flags);
 
-		ret = __wait_localport_unreg(lport);
+		ret = __localport_unreg(lport);
 		if (ret)
 			pr_warn("%s: Failed deleting local port\n", __func__);
 
···
 
 	device_destroy(&fcloop_class, MKDEV(0, 0));
 	class_unregister(&fcloop_class);
+	kmem_cache_destroy(lsreq_cache);
 }
 
 module_init(fcloop_init);

drivers/nvme/target/loop.c

···
 
 	struct list_head	list;
 	struct blk_mq_tag_set	tag_set;
-	struct nvme_loop_iod	async_event_iod;
 	struct nvme_ctrl	ctrl;
 
 	struct nvmet_port	*port;
+
+	/* Must be last --ends in a flexible-array member. */
+	struct nvme_loop_iod	async_event_iod;
 };
 
 static inline struct nvme_loop_ctrl *to_loop_ctrl(struct nvme_ctrl *ctrl)
···
 	nvme_start_request(req);
 	iod->cmd.common.flags |= NVME_CMD_SGL_METABUF;
 	iod->req.port = queue->ctrl->port;
-	if (!nvmet_req_init(&iod->req, &queue->nvme_cq,
-			&queue->nvme_sq, &nvme_loop_ops))
+	if (!nvmet_req_init(&iod->req, &queue->nvme_sq, &nvme_loop_ops))
 		return BLK_STS_OK;
 
 	if (blk_rq_nr_phys_segments(req)) {
···
 	iod->cmd.common.command_id = NVME_AQ_BLK_MQ_DEPTH;
 	iod->cmd.common.flags |= NVME_CMD_SGL_METABUF;
 
-	if (!nvmet_req_init(&iod->req, &queue->nvme_cq, &queue->nvme_sq,
-			&nvme_loop_ops)) {
+	if (!nvmet_req_init(&iod->req, &queue->nvme_sq, &nvme_loop_ops)) {
 		dev_err(ctrl->ctrl.device, "failed async event work\n");
 		return;
 	}
···
 	nvme_unquiesce_admin_queue(&ctrl->ctrl);
 
 	nvmet_sq_destroy(&ctrl->queues[0].nvme_sq);
+	nvmet_cq_put(&ctrl->queues[0].nvme_cq);
 	nvme_remove_admin_tag_set(&ctrl->ctrl);
 }
 
···
 	for (i = 1; i < ctrl->ctrl.queue_count; i++) {
 		clear_bit(NVME_LOOP_Q_LIVE, &ctrl->queues[i].flags);
 		nvmet_sq_destroy(&ctrl->queues[i].nvme_sq);
+		nvmet_cq_put(&ctrl->queues[i].nvme_cq);
 	}
 	ctrl->ctrl.queue_count = 1;
 	/*
···
 
 	for (i = 1; i <= nr_io_queues; i++) {
 		ctrl->queues[i].ctrl = ctrl;
-		ret = nvmet_sq_init(&ctrl->queues[i].nvme_sq);
-		if (ret)
+		nvmet_cq_init(&ctrl->queues[i].nvme_cq);
+		ret = nvmet_sq_init(&ctrl->queues[i].nvme_sq,
+				&ctrl->queues[i].nvme_cq);
+		if (ret) {
+			nvmet_cq_put(&ctrl->queues[i].nvme_cq);
 			goto out_destroy_queues;
+		}
 
 		ctrl->ctrl.queue_count++;
 	}
···
 	int error;
 
 	ctrl->queues[0].ctrl = ctrl;
-	error = nvmet_sq_init(&ctrl->queues[0].nvme_sq);
-	if (error)
+	nvmet_cq_init(&ctrl->queues[0].nvme_cq);
+	error = nvmet_sq_init(&ctrl->queues[0].nvme_sq,
+			&ctrl->queues[0].nvme_cq);
+	if (error) {
+		nvmet_cq_put(&ctrl->queues[0].nvme_cq);
 		return error;
+	}
 	ctrl->ctrl.queue_count = 1;
 
 	error = nvme_alloc_admin_tag_set(&ctrl->ctrl, &ctrl->admin_tag_set,
···
 	nvme_remove_admin_tag_set(&ctrl->ctrl);
 out_free_sq:
 	nvmet_sq_destroy(&ctrl->queues[0].nvme_sq);
+	nvmet_cq_put(&ctrl->queues[0].nvme_cq);
 	return error;
 }
 

drivers/nvme/target/nvmet.h

···
 }
 
 struct nvmet_cq {
+	struct nvmet_ctrl	*ctrl;
 	u16			qid;
 	u16			size;
+	refcount_t		ref;
 };
 
 struct nvmet_sq {
 	struct nvmet_ctrl	*ctrl;
 	struct percpu_ref	ref;
+	struct nvmet_cq		*cq;
 	u16			qid;
 	u16			size;
 	u32			sqhd;
···
 struct nvmet_ctrl {
 	struct nvmet_subsys	*subsys;
 	struct nvmet_sq		**sqs;
+	struct nvmet_cq		**cqs;
 
 	void			*drvdata;
 
···
 	u16 (*get_max_queue_size)(const struct nvmet_ctrl *ctrl);
 
 	/* Operations mandatory for PCI target controllers */
-	u16 (*create_sq)(struct nvmet_ctrl *ctrl, u16 sqid, u16 flags,
+	u16 (*create_sq)(struct nvmet_ctrl *ctrl, u16 sqid, u16 cqid, u16 flags,
 			 u16 qsize, u64 prp1);
 	u16 (*delete_sq)(struct nvmet_ctrl *ctrl, u16 sqid);
 	u16 (*create_cq)(struct nvmet_ctrl *ctrl, u16 cqid, u16 flags,
···
 u16 nvmet_parse_fabrics_io_cmd(struct nvmet_req *req);
 u32 nvmet_fabrics_io_cmd_data_len(struct nvmet_req *req);
 
-bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
-		struct nvmet_sq *sq, const struct nvmet_fabrics_ops *ops);
+bool nvmet_req_init(struct nvmet_req *req, struct nvmet_sq *sq,
+		const struct nvmet_fabrics_ops *ops);
 void nvmet_req_uninit(struct nvmet_req *req);
 size_t nvmet_req_transfer_len(struct nvmet_req *req);
 bool nvmet_check_transfer_len(struct nvmet_req *req, size_t len);
···
 void nvmet_execute_get_features(struct nvmet_req *req);
 void nvmet_execute_keep_alive(struct nvmet_req *req);
 
-u16 nvmet_check_cqid(struct nvmet_ctrl *ctrl, u16 cqid);
+u16 nvmet_check_cqid(struct nvmet_ctrl *ctrl, u16 cqid, bool create);
+u16 nvmet_check_io_cqid(struct nvmet_ctrl *ctrl, u16 cqid, bool create);
+void nvmet_cq_init(struct nvmet_cq *cq);
 void nvmet_cq_setup(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq, u16 qid,
 		u16 size);
 u16 nvmet_cq_create(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq, u16 qid,
 		u16 size);
+void nvmet_cq_destroy(struct nvmet_cq *cq);
+bool nvmet_cq_get(struct nvmet_cq *cq);
+void nvmet_cq_put(struct nvmet_cq *cq);
+bool nvmet_cq_in_use(struct nvmet_cq *cq);
 u16 nvmet_check_sqid(struct nvmet_ctrl *ctrl, u16 sqid, bool create);
 void nvmet_sq_setup(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq, u16 qid,
 		u16 size);
-u16 nvmet_sq_create(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq, u16 qid,
-		u16 size);
+u16 nvmet_sq_create(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq,
+	struct nvmet_cq *cq, u16 qid, u16 size);
 void nvmet_sq_destroy(struct nvmet_sq *sq);
-int nvmet_sq_init(struct nvmet_sq *sq);
+int nvmet_sq_init(struct nvmet_sq *sq, struct nvmet_cq *cq);
 
 void nvmet_ctrl_fatal_error(struct nvmet_ctrl *ctrl);
 

drivers/nvme/target/pci-epf.c

···
 	nvmet_pci_epf_drain_queue(cq);
 	nvmet_pci_epf_remove_irq_vector(ctrl, cq->vector);
 	nvmet_pci_epf_mem_unmap(ctrl->nvme_epf, &cq->pci_map);
+	nvmet_cq_put(&cq->nvme_cq);
 
 	return NVME_SC_SUCCESS;
 }
 
 static u16 nvmet_pci_epf_create_sq(struct nvmet_ctrl *tctrl,
-		u16 sqid, u16 flags, u16 qsize, u64 pci_addr)
+		u16 sqid, u16 cqid, u16 flags, u16 qsize, u64 pci_addr)
 {
 	struct nvmet_pci_epf_ctrl *ctrl = tctrl->drvdata;
 	struct nvmet_pci_epf_queue *sq = &ctrl->sq[sqid];
+	struct nvmet_pci_epf_queue *cq = &ctrl->cq[cqid];
 	u16 status;
 
 	if (test_bit(NVMET_PCI_EPF_Q_LIVE, &sq->flags))
···
 		sq->qes = ctrl->io_sqes;
 	sq->pci_size = sq->qes * sq->depth;
 
-	status = nvmet_sq_create(tctrl, &sq->nvme_sq, sqid, sq->depth);
+	status = nvmet_sq_create(tctrl, &sq->nvme_sq, &cq->nvme_cq, sqid,
+			sq->depth);
 	if (status != NVME_SC_SUCCESS)
 		return status;
 
···
 		goto complete;
 	}
 
-	if (!nvmet_req_init(req, &iod->cq->nvme_cq, &iod->sq->nvme_sq,
-			    &nvmet_pci_epf_fabrics_ops))
+	if (!nvmet_req_init(req, &iod->sq->nvme_sq, &nvmet_pci_epf_fabrics_ops))
 		goto complete;
 
 	iod->data_len = nvmet_req_transfer_len(req);
···
 
 	qsize = aqa & 0x00000fff;
 	pci_addr = asq & GENMASK_ULL(63, 12);
-	status = nvmet_pci_epf_create_sq(ctrl->tctrl, 0, NVME_QUEUE_PHYS_CONTIG,
-					 qsize, pci_addr);
+	status = nvmet_pci_epf_create_sq(ctrl->tctrl, 0, 0,
+			NVME_QUEUE_PHYS_CONTIG, qsize, pci_addr);
 	if (status != NVME_SC_SUCCESS) {
 		dev_err(ctrl->dev, "Failed to create admin submission queue\n");
 		nvmet_pci_epf_delete_cq(ctrl->tctrl, 0);

drivers/nvme/target/rdma.c

···
 		cmd->send_sge.addr, cmd->send_sge.length,
 		DMA_TO_DEVICE);
 
-	if (!nvmet_req_init(&cmd->req, &queue->nvme_cq,
-			&queue->nvme_sq, &nvmet_rdma_ops))
+	if (!nvmet_req_init(&cmd->req, &queue->nvme_sq, &nvmet_rdma_ops))
 		return;
 
 	status = nvmet_rdma_map_sgl(cmd);
···
 	pr_debug("freeing queue %d\n", queue->idx);
 
 	nvmet_sq_destroy(&queue->nvme_sq);
+	nvmet_cq_put(&queue->nvme_cq);
 
 	nvmet_rdma_destroy_queue_ib(queue);
 	if (!queue->nsrq) {
···
 		goto out_reject;
 	}
 
-	ret = nvmet_sq_init(&queue->nvme_sq);
+	nvmet_cq_init(&queue->nvme_cq);
+	ret = nvmet_sq_init(&queue->nvme_sq, &queue->nvme_cq);
 	if (ret) {
 		ret = NVME_RDMA_CM_NO_RSC;
 		goto out_free_queue;
···
 out_destroy_sq:
 	nvmet_sq_destroy(&queue->nvme_sq);
 out_free_queue:
+	nvmet_cq_put(&queue->nvme_cq);
 	kfree(queue);
 out_reject:
 	nvmet_rdma_cm_reject(cm_id, ret);

drivers/nvme/target/tcp.c

···
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/slab.h>
+#include <linux/crc32c.h>
 #include <linux/err.h>
 #include <linux/nvme-tcp.h>
 #include <linux/nvme-keyring.h>
···
 #include <net/handshake.h>
 #include <linux/inet.h>
 #include <linux/llist.h>
-#include <crypto/hash.h>
 #include <trace/events/sock.h>
 
 #include "nvmet.h"
···
 	/* digest state */
 	bool			hdr_digest;
 	bool			data_digest;
-	struct ahash_request	*snd_hash;
-	struct ahash_request	*rcv_hash;
 
 	/* TLS state */
 	key_serial_t		tls_pskid;
···
 	return queue->data_digest ? NVME_TCP_DIGEST_LENGTH : 0;
 }
 
-static inline void nvmet_tcp_hdgst(struct ahash_request *hash,
-		void *pdu, size_t len)
+static inline void nvmet_tcp_hdgst(void *pdu, size_t len)
 {
-	struct scatterlist sg;
-
-	sg_init_one(&sg, pdu, len);
-	ahash_request_set_crypt(hash, &sg, pdu + len, len);
-	crypto_ahash_digest(hash);
+	put_unaligned_le32(~crc32c(~0, pdu, len), pdu + len);
 }
 
 static int nvmet_tcp_verify_hdgst(struct nvmet_tcp_queue *queue,
···
 	}
 
 	recv_digest = *(__le32 *)(pdu + hdr->hlen);
-	nvmet_tcp_hdgst(queue->rcv_hash, pdu, len);
+	nvmet_tcp_hdgst(pdu, len);
 	exp_digest = *(__le32 *)(pdu + hdr->hlen);
 	if (recv_digest != exp_digest) {
 		pr_err("queue %d: header digest error: recv %#x expected %#x\n",
···
 	return NVME_SC_INTERNAL;
 }
 
-static void nvmet_tcp_calc_ddgst(struct ahash_request *hash,
-		struct nvmet_tcp_cmd *cmd)
+static void nvmet_tcp_calc_ddgst(struct nvmet_tcp_cmd *cmd)
 {
-	ahash_request_set_crypt(hash, cmd->req.sg,
-		(void *)&cmd->exp_ddgst, cmd->req.transfer_len);
-	crypto_ahash_digest(hash);
+	size_t total_len = cmd->req.transfer_len;
+	struct scatterlist *sg = cmd->req.sg;
+	u32 crc = ~0;
+
+	while (total_len) {
+		size_t len = min_t(size_t, total_len, sg->length);
+
+		/*
+		 * Note that the scatterlist does not contain any highmem pages,
+		 * as it was allocated by sgl_alloc() with GFP_KERNEL.
+		 */
+		crc = crc32c(crc, sg_virt(sg), len);
+		total_len -= len;
+		sg = sg_next(sg);
+	}
+	cmd->exp_ddgst = cpu_to_le32(~crc);
 }
 
 static void nvmet_setup_c2h_data_pdu(struct nvmet_tcp_cmd *cmd)
···
 
 	if (queue->data_digest) {
 		pdu->hdr.flags |= NVME_TCP_F_DDGST;
-		nvmet_tcp_calc_ddgst(queue->snd_hash, cmd);
+		nvmet_tcp_calc_ddgst(cmd);
 	}
 
 	if (cmd->queue->hdr_digest) {
 		pdu->hdr.flags |= NVME_TCP_F_HDGST;
-		nvmet_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
+		nvmet_tcp_hdgst(pdu, sizeof(*pdu));
 	}
 }
 
 static void nvmet_setup_r2t_pdu(struct nvmet_tcp_cmd *cmd)
 {
 	struct nvme_tcp_r2t_pdu *pdu = cmd->r2t_pdu;
-	struct nvmet_tcp_queue *queue = cmd->queue;
 	u8 hdgst = nvmet_tcp_hdgst_len(cmd->queue);
 
 	cmd->offset = 0;
···
 	pdu->r2t_offset = cpu_to_le32(cmd->rbytes_done);
 	if (cmd->queue->hdr_digest) {
 		pdu->hdr.flags |= NVME_TCP_F_HDGST;
-		nvmet_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
+		nvmet_tcp_hdgst(pdu, sizeof(*pdu));
 	}
 }
 
 static void nvmet_setup_response_pdu(struct nvmet_tcp_cmd *cmd)
 {
 	struct nvme_tcp_rsp_pdu *pdu = cmd->rsp_pdu;
-	struct nvmet_tcp_queue *queue = cmd->queue;
 	u8 hdgst = nvmet_tcp_hdgst_len(cmd->queue);
 
 	cmd->offset = 0;
···
 	pdu->hdr.plen = cpu_to_le32(pdu->hdr.hlen + hdgst);
 	if (cmd->queue->hdr_digest) {
 		pdu->hdr.flags |= NVME_TCP_F_HDGST;
-		nvmet_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
+		nvmet_tcp_hdgst(pdu, sizeof(*pdu));
 	}
 }
 
···
 	smp_store_release(&queue->rcv_state, NVMET_TCP_RECV_PDU);
 }
 
-static void nvmet_tcp_free_crypto(struct nvmet_tcp_queue *queue)
-{
-	struct crypto_ahash *tfm = crypto_ahash_reqtfm(queue->rcv_hash);
-
-	ahash_request_free(queue->rcv_hash);
-	ahash_request_free(queue->snd_hash);
-	crypto_free_ahash(tfm);
-}
-
-static int nvmet_tcp_alloc_crypto(struct nvmet_tcp_queue *queue)
-{
-	struct crypto_ahash *tfm;
-
-	tfm = crypto_alloc_ahash("crc32c", 0, CRYPTO_ALG_ASYNC);
-	if (IS_ERR(tfm))
-		return PTR_ERR(tfm);
-
-	queue->snd_hash = ahash_request_alloc(tfm, GFP_KERNEL);
-	if (!queue->snd_hash)
-		goto free_tfm;
-	ahash_request_set_callback(queue->snd_hash, 0, NULL, NULL);
-
-	queue->rcv_hash = ahash_request_alloc(tfm, GFP_KERNEL);
-	if (!queue->rcv_hash)
-		goto free_snd_hash;
-	ahash_request_set_callback(queue->rcv_hash, 0, NULL, NULL);
-
-	return 0;
-free_snd_hash:
-	ahash_request_free(queue->snd_hash);
-free_tfm:
-	crypto_free_ahash(tfm);
-	return -ENOMEM;
-}
-
-
 static int nvmet_tcp_handle_icreq(struct nvmet_tcp_queue *queue)
 {
 	struct nvme_tcp_icreq_pdu *icreq = &queue->pdu.icreq;
···
 
 	queue->hdr_digest = !!(icreq->digest & NVME_TCP_HDR_DIGEST_ENABLE);
 	queue->data_digest = !!(icreq->digest & NVME_TCP_DATA_DIGEST_ENABLE);
-	if (queue->hdr_digest || queue->data_digest) {
-		ret = nvmet_tcp_alloc_crypto(queue);
-		if (ret)
-			return ret;
-	}
 
 	memset(icresp, 0, sizeof(*icresp));
 	icresp->hdr.type = nvme_tcp_icresp;
···
 	req = &queue->cmd->req;
 	memcpy(req->cmd, nvme_cmd, sizeof(*nvme_cmd));
 
-	if (unlikely(!nvmet_req_init(req, &queue->nvme_cq,
-			&queue->nvme_sq, &nvmet_tcp_ops))) {
+	if (unlikely(!nvmet_req_init(req, &queue->nvme_sq, &nvmet_tcp_ops))) {
 		pr_err("failed cmd %p id %d opcode %d, data_len: %d, status: %04x\n",
 			req->cmd, req->cmd->common.command_id,
 			req->cmd->common.opcode,
···
 {
 	struct nvmet_tcp_queue *queue = cmd->queue;
 
-	nvmet_tcp_calc_ddgst(queue->rcv_hash, cmd);
+	nvmet_tcp_calc_ddgst(cmd);
 	queue->offset = 0;
 	queue->left = NVME_TCP_DIGEST_LENGTH;
 	queue->rcv_state = NVMET_TCP_RECV_DDGST;
···
 	nvmet_sq_put_tls_key(&queue->nvme_sq);
 	nvmet_tcp_uninit_data_in_cmds(queue);
 	nvmet_sq_destroy(&queue->nvme_sq);
+	nvmet_cq_put(&queue->nvme_cq);
 	cancel_work_sync(&queue->io_work);
 	nvmet_tcp_free_cmd_data_in_buffers(queue);
 	/* ->sock will be released by fput() */
 	fput(queue->sock->file);
 	nvmet_tcp_free_cmds(queue);
-	if (queue->hdr_digest || queue->data_digest)
-		nvmet_tcp_free_crypto(queue);
 	ida_free(&nvmet_tcp_queue_ida, queue->idx);
 	page_frag_cache_drain(&queue->pf_cache);
 	kfree(queue);
···
 	if (ret)
 		goto out_ida_remove;
 
-	ret = nvmet_sq_init(&queue->nvme_sq);
+	nvmet_cq_init(&queue->nvme_cq);
+	ret = nvmet_sq_init(&queue->nvme_sq, &queue->nvme_cq);
 	if (ret)
 		goto out_free_connect;
 
···
 	mutex_unlock(&nvmet_tcp_queue_mutex);
 	nvmet_sq_destroy(&queue->nvme_sq);
 out_free_connect:
+	nvmet_cq_put(&queue->nvme_cq);
 	nvmet_tcp_free_cmd(&queue->connect);
 out_ida_remove:
 	ida_free(&nvmet_tcp_queue_ida, queue->idx);

include/linux/dmapool.h

···
 #ifndef LINUX_DMAPOOL_H
 #define	LINUX_DMAPOOL_H
 
+#include <linux/nodemask_types.h>
 #include <linux/scatterlist.h>
 #include <asm/io.h>
 
···
 
 #ifdef CONFIG_HAS_DMA
 
-struct dma_pool *dma_pool_create(const char *name, struct device *dev, 
-			size_t size, size_t align, size_t allocation);
+struct dma_pool *dma_pool_create_node(const char *name, struct device *dev,
+		size_t size, size_t align, size_t boundary, int node);
 
 void dma_pool_destroy(struct dma_pool *pool);
 
···
 void dmam_pool_destroy(struct dma_pool *pool);
 
 #else /* !CONFIG_HAS_DMA */
-static inline struct dma_pool *dma_pool_create(const char *name,
-	struct device *dev, size_t size, size_t align, size_t allocation)
-{ return NULL; }
+static inline struct dma_pool *dma_pool_create_node(const char *name,
+		struct device *dev, size_t size, size_t align, size_t boundary,
+		int node)
+{
+	return NULL;
+}
 static inline void dma_pool_destroy(struct dma_pool *pool) { }
 static inline void *dma_pool_alloc(struct dma_pool *pool, gfp_t mem_flags,
 				   dma_addr_t *handle) { return NULL; }
···
 { return NULL; }
 static inline void dmam_pool_destroy(struct dma_pool *pool) { }
 #endif /* !CONFIG_HAS_DMA */
+
+static inline struct dma_pool *dma_pool_create(const char *name,
+		struct device *dev, size_t size, size_t align, size_t boundary)
+{
+	return dma_pool_create_node(name, dev, size, align, boundary,
+				    NUMA_NO_NODE);
+}
 
 static inline void *dma_pool_zalloc(struct dma_pool *pool, gfp_t mem_flags,
 				    dma_addr_t *handle)

mm/dmapool.c

···
 	unsigned int size;
 	unsigned int allocation;
 	unsigned int boundary;
+	int node;
 	char name[32];
 	struct list_head pools;
 };
···
 
 
 /**
- * dma_pool_create - Creates a pool of consistent memory blocks, for dma.
+ * dma_pool_create_node - Creates a pool of consistent memory blocks, for dma.
  * @name: name of pool, for diagnostics
  * @dev: device that will be doing the DMA
  * @size: size of the blocks in this pool.
  * @align: alignment requirement for blocks; must be a power of two
  * @boundary: returned blocks won't cross this power of two boundary
+ * @node: optional NUMA node to allocate structs 'dma_pool' and 'dma_page' on
  * Context: not in_interrupt()
  *
  * Given one of these pools, dma_pool_alloc()
···
  * Return: a dma allocation pool with the requested characteristics, or
  * %NULL if one can't be created.
  */
-struct dma_pool *dma_pool_create(const char *name, struct device *dev,
-				 size_t size, size_t align, size_t boundary)
+struct dma_pool *dma_pool_create_node(const char *name, struct device *dev,
+		size_t size, size_t align, size_t boundary, int node)
 {
 	struct dma_pool *retval;
 	size_t allocation;
···
 
 	boundary = min(boundary, allocation);
 
-	retval = kzalloc(sizeof(*retval), GFP_KERNEL);
+	retval = kzalloc_node(sizeof(*retval), GFP_KERNEL, node);
 	if (!retval)
 		return retval;
 
···
 	retval->size = size;
 	retval->boundary = boundary;
 	retval->allocation = allocation;
+	retval->node = node;
 	INIT_LIST_HEAD(&retval->pools);
 
 	/*
···
 	mutex_unlock(&pools_reg_lock);
 	return retval;
 }
-EXPORT_SYMBOL(dma_pool_create);
+EXPORT_SYMBOL(dma_pool_create_node);
 
 static void pool_initialise_page(struct dma_pool *pool, struct dma_page *page)
 {
···
 {
 	struct dma_page *page;
 
-	page = kmalloc(sizeof(*page), mem_flags);
+	page = kmalloc_node(sizeof(*page), mem_flags, pool->node);
 	if (!page)
 		return NULL;