[PATCH 0/6] sysfs support for LightNVM

[PATCH 0/6] sysfs support for LightNVM

[Matias Bjørling]

Hi Jens and Keith,

If the patches look good to you, would you pick them up through your
tree?

The common way to expose a storage device is to export it through the
gendisk structure, which takes care of handling the visibility lifetime
of the device to user-space and carries device driver and block layer
specific sysfs entries.

For LightNVM, no gendisk is exposed, which hides these entries.

This patchset enables LightNVM to expose a subset of the gendisk entries
by implementing a new non-gendisk device that only exposes the sysfs
entries that are relevant to the device driver and LightNVM. For example
/mq and the NVMe sysfs entries.

The new non-gendisk exposes itself in place of the original gendisk.
Thus, a gendisk device exposed through /sys/class/nvme/nvme0n1, will
still be exposed through the same directory, although only with the /mq,
lightnvm, and nvme specific entries.

The LightNVM storage device can be found through sysfs by searching
for devices that has the "lightnvm" dev type.

v2:
 - Updated "expose gennvm target type" to expose its own kobject, which
   pblk and others can hook into.
 - Fixed missing disk_name in null_blk patch on lightnvm device

Thanks,
Matias

Matias Bjørling (4):
  nvme: refactor namespaces to support non-gendisk devices
  null_blk: refactor to support non-gendisk devices
  blk-mq: register device instead of disk
  lightnvm: let drivers control the lifetime of nvm_dev

Simon A. F. Lund (2):
  lightnvm: expose device geometry through sysfs
  lightnvm: expose gennvm target type through sysfs

 block/blk-mq-sysfs.c         |  13 ++-
 block/blk-sysfs.c            |   4 +-
 drivers/block/null_blk.c     | 128 +++++++++++++++++-----------
 drivers/lightnvm/Makefile    |   2 +-
 drivers/lightnvm/core.c      |  52 +++++-------
 drivers/lightnvm/gennvm.c    | 112 +++++++++++++++++++++----
 drivers/lightnvm/lightnvm.h  |  35 ++++++++
 drivers/lightnvm/sysfs.c     | 195 +++++++++++++++++++++++++++++++++++++++++++
 drivers/md/dm.c              |   2 +-
 drivers/nvme/host/core.c     | 165 ++++++++++++++++++++----------------
 drivers/nvme/host/lightnvm.c |  29 ++++++-
 drivers/nvme/host/nvme.h     |  26 ++++--
 include/linux/blk-mq.h       |   4 +-
 include/linux/lightnvm.h     |  28 +++++--
 14 files changed, 597 insertions(+), 198 deletions(-)
 create mode 100644 drivers/lightnvm/lightnvm.h
 create mode 100644 drivers/lightnvm/sysfs.c

-- 
2.1.4

[PATCH 1/6] nvme: refactor namespaces to support non-gendisk devices

[Matias Bjørling]

With LightNVM enabled namespaces, the gendisk structure is not exposed
to the user. This prevents LightNVM users from accessing the NVMe device
driver specific sysfs entries, and LightNVM namespace geometry.

Refactor the revalidation process, so that a namespace, instead of a
gendisk, is revalidated. This later allows patches to wire up the
sysfs entries up to a non-gendisk namespace.

Signed-off-by: Matias Bjørling <m@bjorling.me>
---
 drivers/nvme/host/core.c | 134 +++++++++++++++++++++++++++++------------------
 1 file changed, 82 insertions(+), 52 deletions(-)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 684062a..f615b6b 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -785,42 +785,33 @@ static void nvme_config_discard(struct nvme_ns *ns)
 	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, ns->queue);
 }
 
-static int nvme_revalidate_disk(struct gendisk *disk)
+static int nvme_revalidate_ns(struct nvme_ns *ns, struct nvme_id_ns **id)
 {
-	struct nvme_ns *ns = disk->private_data;
-	struct nvme_id_ns *id;
-	u8 lbaf, pi_type;
-	u16 old_ms;
-	unsigned short bs;
-
-	if (test_bit(NVME_NS_DEAD, &ns->flags)) {
-		set_capacity(disk, 0);
-		return -ENODEV;
-	}
-	if (nvme_identify_ns(ns->ctrl, ns->ns_id, &id)) {
+	if (nvme_identify_ns(ns->ctrl, ns->ns_id, id)) {
 		dev_warn(disk_to_dev(ns->disk), "%s: Identify failure\n",
 				__func__);
 		return -ENODEV;
 	}
-	if (id->ncap == 0) {
-		kfree(id);
-		return -ENODEV;
-	}
 
-	if (nvme_nvm_ns_supported(ns, id) && ns->type != NVME_NS_LIGHTNVM) {
-		if (nvme_nvm_register(ns->queue, disk->disk_name)) {
-			dev_warn(disk_to_dev(ns->disk),
-				"%s: LightNVM init failure\n", __func__);
-			kfree(id);
-			return -ENODEV;
-		}
-		ns->type = NVME_NS_LIGHTNVM;
+	if ((*id)->ncap == 0) {
+		kfree(*id);
+		return -ENODEV;
 	}
 
 	if (ns->ctrl->vs >= NVME_VS(1, 1))
-		memcpy(ns->eui, id->eui64, sizeof(ns->eui));
+		memcpy(ns->eui, (*id)->eui64, sizeof(ns->eui));
 	if (ns->ctrl->vs >= NVME_VS(1, 2))
-		memcpy(ns->uuid, id->nguid, sizeof(ns->uuid));
+		memcpy(ns->uuid, (*id)->nguid, sizeof(ns->uuid));
+
+	return 0;
+}
+
+static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id)
+{
+	struct nvme_ns *ns = disk->private_data;
+	u8 lbaf, pi_type;
+	u16 old_ms;
+	unsigned short bs;
 
 	old_ms = ns->ms;
 	lbaf = id->flbas & NVME_NS_FLBAS_LBA_MASK;
@@ -859,8 +850,26 @@ static int nvme_revalidate_disk(struct gendisk *disk)
 	if (ns->ctrl->oncs & NVME_CTRL_ONCS_DSM)
 		nvme_config_discard(ns);
 	blk_mq_unfreeze_queue(disk->queue);
+}
 
+static int nvme_revalidate_disk(struct gendisk *disk)
+{
+	struct nvme_ns *ns = disk->private_data;
+	struct nvme_id_ns *id = NULL;
+	int ret;
+
+	if (test_bit(NVME_NS_DEAD, &ns->flags)) {
+		set_capacity(disk, 0);
+		return -ENODEV;
+	}
+
+	ret = nvme_revalidate_ns(ns, &id);
+	if (ret)
+		return ret;
+
+	__nvme_revalidate_disk(disk, id);
 	kfree(id);
+
 	return 0;
 }
 
@@ -1430,6 +1439,8 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 {
 	struct nvme_ns *ns;
 	struct gendisk *disk;
+	struct nvme_id_ns *id;
+	char disk_name[DISK_NAME_LEN];
 	int node = dev_to_node(ctrl->dev);
 
 	lockdep_assert_held(&ctrl->namespaces_mutex);
@@ -1449,44 +1460,63 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 	ns->queue->queuedata = ns;
 	ns->ctrl = ctrl;
 
-	disk = alloc_disk_node(0, node);
-	if (!disk)
-		goto out_free_queue;
-
 	kref_init(&ns->kref);
 	ns->ns_id = nsid;
-	ns->disk = disk;
 	ns->lba_shift = 9; /* set to a default value for 512 until disk is validated */
 
-
 	blk_queue_logical_block_size(ns->queue, 1 << ns->lba_shift);
 	nvme_set_queue_limits(ctrl, ns->queue);
 
-	disk->major = nvme_major;
-	disk->first_minor = 0;
-	disk->fops = &nvme_fops;
-	disk->private_data = ns;
-	disk->queue = ns->queue;
-	disk->driverfs_dev = ctrl->device;
-	disk->flags = GENHD_FL_EXT_DEVT;
-	sprintf(disk->disk_name, "nvme%dn%d", ctrl->instance, ns->instance);
-
-	if (nvme_revalidate_disk(ns->disk))
-		goto out_free_disk;
+	if (nvme_revalidate_ns(ns, &id))
+		goto out_free_queue;
+
+	sprintf(disk_name, "nvme%dn%d", ctrl->instance, ns->instance);
+
+	if (nvme_nvm_ns_supported(ns, id)) {
+		if (nvme_nvm_register(ns->queue, disk_name)) {
+			dev_warn(ctrl->dev,
+				"%s: LightNVM init failure\n", __func__);
+			goto out_free_id;
+		}
+
+		disk = alloc_disk_node(0, node);
+		if (!disk)
+			goto out_free_id;
+		memcpy(disk->disk_name, disk_name, DISK_NAME_LEN);
+		ns->disk = disk;
+		ns->type = NVME_NS_LIGHTNVM;
+	} else {
+		disk = alloc_disk_node(0, node);
+		if (!disk)
+			goto out_free_id;
+
+		disk->major = nvme_major;
+		disk->first_minor = 0;
+		disk->fops = &nvme_fops;
+		disk->private_data = ns;
+		disk->queue = ns->queue;
+		disk->driverfs_dev = ctrl->device;
+		disk->flags = GENHD_FL_EXT_DEVT;
+		memcpy(disk->disk_name, disk_name, DISK_NAME_LEN);
+		ns->disk = disk;
+
+		__nvme_revalidate_disk(disk, id);
+
+		add_disk(ns->disk);
+
+		if (sysfs_create_group(&disk_to_dev(ns->disk)->kobj,
+							&nvme_ns_attr_group))
+			pr_warn("%s: failed to create sysfs group for identification\n",
+				ns->disk->disk_name);
+	}
 
 	list_add_tail_rcu(&ns->list, &ctrl->namespaces);
 	kref_get(&ctrl->kref);
-	if (ns->type == NVME_NS_LIGHTNVM)
-		return;
 
-	add_disk(ns->disk);
-	if (sysfs_create_group(&disk_to_dev(ns->disk)->kobj,
-					&nvme_ns_attr_group))
-		pr_warn("%s: failed to create sysfs group for identification\n",
-			ns->disk->disk_name);
+	kfree(id);
 	return;
- out_free_disk:
-	kfree(disk);
+ out_free_id:
+	kfree(id);
  out_free_queue:
 	blk_cleanup_queue(ns->queue);
  out_release_instance:
-- 
2.1.4

[PATCH 2/6] null_blk: refactor to support non-gendisk devices

[Matias Bjørling]

With LightNVM enabled devices, the gendisk structure is not exposed
to the user. This hides the device driver specific sysfs entries, and
prevents binding of LightNVM geometry information to the device.

Refactor the device registration process, so that gendisk and
non-gendisk devices are easily managed.

Signed-off-by: Matias Bjørling <m@bjorling.me>
---
 drivers/block/null_blk.c | 110 ++++++++++++++++++++++++++---------------------
 1 file changed, 61 insertions(+), 49 deletions(-)

diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index cab9759..3117df1 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -414,23 +414,6 @@ static void cleanup_queues(struct nullb *nullb)
 	kfree(nullb->queues);
 }
 
-static void null_del_dev(struct nullb *nullb)
-{
-	list_del_init(&nullb->list);
-
-	if (use_lightnvm)
-		nvm_unregister(nullb->disk_name);
-	else
-		del_gendisk(nullb->disk);
-	blk_cleanup_queue(nullb->q);
-	if (queue_mode == NULL_Q_MQ)
-		blk_mq_free_tag_set(&nullb->tag_set);
-	if (!use_lightnvm)
-		put_disk(nullb->disk);
-	cleanup_queues(nullb);
-	kfree(nullb);
-}
-
 #ifdef CONFIG_NVM
 
 static void null_lnvm_end_io(struct request *rq, int error)
@@ -564,10 +547,41 @@ static struct nvm_dev_ops null_lnvm_dev_ops = {
 	/* Simulate nvme protocol restriction */
 	.max_phys_sect		= 64,
 };
+
+static int null_nvm_register(struct nullb *nullb)
+{
+	return nvm_register(nullb->q, nullb->disk_name, &null_lnvm_dev_ops);
+}
+
+static void null_nvm_unregister(struct nullb *nullb)
+{
+	nvm_unregister(nullb->disk_name);
+}
 #else
-static struct nvm_dev_ops null_lnvm_dev_ops;
+static int null_nvm_register(struct nullb *nullb)
+{
+	return -EINVAL;
+}
+static void null_nvm_unregister(struct nullb *nullb) {}
 #endif /* CONFIG_NVM */
 
+static void null_del_dev(struct nullb *nullb)
+{
+	list_del_init(&nullb->list);
+
+	if (use_lightnvm)
+		null_nvm_unregister(nullb);
+	else
+		del_gendisk(nullb->disk);
+	blk_cleanup_queue(nullb->q);
+	if (queue_mode == NULL_Q_MQ)
+		blk_mq_free_tag_set(&nullb->tag_set);
+	if (!use_lightnvm)
+		put_disk(nullb->disk);
+	cleanup_queues(nullb);
+	kfree(nullb);
+}
+
 static int null_open(struct block_device *bdev, fmode_t mode)
 {
 	return 0;
@@ -640,11 +654,32 @@ static int init_driver_queues(struct nullb *nullb)
 	return 0;
 }
 
-static int null_add_dev(void)
+static int null_gendisk_register(struct nullb *nullb)
 {
 	struct gendisk *disk;
-	struct nullb *nullb;
 	sector_t size;
+
+	disk = nullb->disk = alloc_disk_node(1, home_node);
+	if (!disk)
+		return -ENOMEM;
+	size = gb * 1024 * 1024 * 1024ULL;
+	set_capacity(disk, size >> 9);
+
+	disk->flags |= GENHD_FL_EXT_DEVT | GENHD_FL_SUPPRESS_PARTITION_INFO;
+	disk->major		= null_major;
+	disk->first_minor	= nullb->index;
+	disk->fops		= &null_fops;
+	disk->private_data	= nullb;
+	disk->queue		= nullb->q;
+	strncpy(disk->disk_name, nullb->disk_name, DISK_NAME_LEN);
+
+	add_disk(disk);
+	return 0;
+}
+
+static int null_add_dev(void)
+{
+	struct nullb *nullb;
 	int rv;
 
 	nullb = kzalloc_node(sizeof(*nullb), GFP_KERNEL, home_node);
@@ -716,42 +751,19 @@ static int null_add_dev(void)
 
 	sprintf(nullb->disk_name, "nullb%d", nullb->index);
 
-	if (use_lightnvm) {
-		rv = nvm_register(nullb->q, nullb->disk_name,
-							&null_lnvm_dev_ops);
-		if (rv)
-			goto out_cleanup_blk_queue;
-		goto done;
-	}
+	if (use_lightnvm)
+		rv = null_nvm_register(nullb);
+	else
+		rv = null_gendisk_register(nullb);
 
-	disk = nullb->disk = alloc_disk_node(1, home_node);
-	if (!disk) {
-		rv = -ENOMEM;
-		goto out_cleanup_lightnvm;
-	}
-	size = gb * 1024 * 1024 * 1024ULL;
-	set_capacity(disk, size >> 9);
+	if (rv)
+		goto out_cleanup_blk_queue;
 
-	disk->flags |= GENHD_FL_EXT_DEVT | GENHD_FL_SUPPRESS_PARTITION_INFO;
-	disk->major		= null_major;
-	disk->first_minor	= nullb->index;
-	disk->fops		= &null_fops;
-	disk->private_data	= nullb;
-	disk->queue		= nullb->q;
-	strncpy(disk->disk_name, nullb->disk_name, DISK_NAME_LEN);
-
-	add_disk(disk);
-
-done:
 	mutex_lock(&lock);
 	list_add_tail(&nullb->list, &nullb_list);
 	mutex_unlock(&lock);
 
 	return 0;
-
-out_cleanup_lightnvm:
-	if (use_lightnvm)
-		nvm_unregister(nullb->disk_name);
 out_cleanup_blk_queue:
 	blk_cleanup_queue(nullb->q);
 out_cleanup_tags:
-- 
2.1.4

[PATCH 3/6] blk-mq: register device instead of disk

[Matias Bjørling]

Enable devices without a gendisk instance to register itself with blk-mq
and expose the associated multi-queue sysfs entries.

Signed-off-by: Matias Bjørling <m@bjorling.me>
---
 block/blk-mq-sysfs.c   | 13 +++++--------
 block/blk-sysfs.c      |  4 ++--
 drivers/md/dm.c        |  2 +-
 include/linux/blk-mq.h |  4 ++--
 4 files changed, 10 insertions(+), 13 deletions(-)

diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index 4ea4dd8..f67b5c1 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -380,9 +380,8 @@ static int blk_mq_register_hctx(struct blk_mq_hw_ctx *hctx)
 	return ret;
 }
 
-void blk_mq_unregister_disk(struct gendisk *disk)
+void blk_mq_unregister_dev(struct device *dev, struct request_queue *q)
 {
-	struct request_queue *q = disk->queue;
 	struct blk_mq_hw_ctx *hctx;
 	struct blk_mq_ctx *ctx;
 	int i, j;
@@ -402,7 +401,7 @@ void blk_mq_unregister_disk(struct gendisk *disk)
 	kobject_del(&q->mq_kobj);
 	kobject_put(&q->mq_kobj);
 
-	kobject_put(&disk_to_dev(disk)->kobj);
+	kobject_put(&dev->kobj);
 
 	q->mq_sysfs_init_done = false;
 	blk_mq_enable_hotplug();
@@ -426,10 +425,8 @@ static void blk_mq_sysfs_init(struct request_queue *q)
 	}
 }
 
-int blk_mq_register_disk(struct gendisk *disk)
+int blk_mq_register_dev(struct device *dev, struct request_queue *q)
 {
-	struct device *dev = disk_to_dev(disk);
-	struct request_queue *q = disk->queue;
 	struct blk_mq_hw_ctx *hctx;
 	int ret, i;
 
@@ -450,7 +447,7 @@ int blk_mq_register_disk(struct gendisk *disk)
 	}
 
 	if (ret)
-		blk_mq_unregister_disk(disk);
+		blk_mq_unregister_dev(dev, q);
 	else
 		q->mq_sysfs_init_done = true;
 out:
@@ -458,7 +455,7 @@ out:
 
 	return ret;
 }
-EXPORT_SYMBOL_GPL(blk_mq_register_disk);
+EXPORT_SYMBOL_GPL(blk_mq_register_dev);
 
 void blk_mq_sysfs_unregister(struct request_queue *q)
 {
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 9920596..5223c5f 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -693,7 +693,7 @@ int blk_register_queue(struct gendisk *disk)
 	kobject_uevent(&q->kobj, KOBJ_ADD);
 
 	if (q->mq_ops)
-		blk_mq_register_disk(disk);
+		blk_mq_register_dev(disk_to_dev(disk), q);
 
 	if (!q->request_fn)
 		return 0;
@@ -718,7 +718,7 @@ void blk_unregister_queue(struct gendisk *disk)
 		return;
 
 	if (q->mq_ops)
-		blk_mq_unregister_disk(disk);
+		blk_mq_unregister_dev(disk_to_dev(disk), q);
 
 	if (q->request_fn)
 		elv_unregister_queue(q);
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index aba7ed9..9de3a9c 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -2798,7 +2798,7 @@ static int dm_mq_init_request_queue(struct mapped_device *md,
 	dm_init_md_queue(md);
 
 	/* backfill 'mq' sysfs registration normally done in blk_register_queue */
-	blk_mq_register_disk(md->disk);
+	blk_mq_register_dev(disk_to_dev(md->disk), q);
 
 	return 0;
 
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 2498fdf..ab6fb32 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -176,8 +176,8 @@ enum {
 struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *);
 struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 						  struct request_queue *q);
-int blk_mq_register_disk(struct gendisk *);
-void blk_mq_unregister_disk(struct gendisk *);
+int blk_mq_register_dev(struct device *, struct request_queue *);
+void blk_mq_unregister_dev(struct device *, struct request_queue *);
 
 int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set);
 void blk_mq_free_tag_set(struct blk_mq_tag_set *set);
-- 
2.1.4

[PATCH 4/6] lightnvm: let drivers control the lifetime of nvm_dev

[Matias Bjørling]

LightNVM does not currently expose the device driver sysfs specific
entries to user-space, as the device driver skips the initialization of
gendisk.

To enable a device driver sysfs entries to be exposed, we need a struct
device to attach it to. To allow both the device driver and LightNVM to
access the same struct device, we need the device driver to track the
lifetime of the nvm_dev structure.

This patch refactors the two users of LightNVM (NVMe and null_blk),
enables them to allocate and free nvm_dev, and at last removes gendisk
usage when a LightNVM device is used.

Signed-off-by: Matias Bjørling <m@bjorling.me>
---
 drivers/block/null_blk.c     | 22 ++++++++++++++++++++--
 drivers/lightnvm/core.c      | 34 +++++++---------------------------
 drivers/nvme/host/core.c     | 34 ++++++++++++++--------------------
 drivers/nvme/host/lightnvm.c | 21 +++++++++++++++++----
 drivers/nvme/host/nvme.h     |  8 +++++---
 include/linux/lightnvm.h     | 15 +++++++++------
 6 files changed, 72 insertions(+), 62 deletions(-)

diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index 3117df1..5a0da8b 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -34,6 +34,7 @@ struct nullb {
 	unsigned int index;
 	struct request_queue *q;
 	struct gendisk *disk;
+	struct nvm_dev *ndev;
 	struct blk_mq_tag_set tag_set;
 	struct hrtimer timer;
 	unsigned int queue_depth;
@@ -550,12 +551,29 @@ static struct nvm_dev_ops null_lnvm_dev_ops = {
 
 static int null_nvm_register(struct nullb *nullb)
 {
-	return nvm_register(nullb->q, nullb->disk_name, &null_lnvm_dev_ops);
+	struct nvm_dev *dev;
+	int rv;
+
+	dev = nvm_alloc_dev(0);
+	if (!dev)
+		return -ENOMEM;
+
+	dev->q = nullb->q;
+	memcpy(dev->name, nullb->disk_name, DISK_NAME_LEN);
+	dev->ops = &null_lnvm_dev_ops;
+
+	rv = nvm_register(dev);
+	if (rv) {
+		kfree(dev);
+		return rv;
+	}
+	nullb->ndev = dev;
+	return 0;
 }
 
 static void null_nvm_unregister(struct nullb *nullb)
 {
-	nvm_unregister(nullb->disk_name);
+	nvm_unregister(nullb->ndev);
 }
 #else
 static int null_nvm_register(struct nullb *nullb)
diff --git a/drivers/lightnvm/core.c b/drivers/lightnvm/core.c
index 4574c0b..ad27768 100644
--- a/drivers/lightnvm/core.c
+++ b/drivers/lightnvm/core.c
@@ -655,23 +655,15 @@ static void nvm_exit(struct nvm_dev *dev)
 	pr_info("nvm: successfully unloaded\n");
 }
 
-int nvm_register(struct request_queue *q, char *disk_name,
-							struct nvm_dev_ops *ops)
+struct nvm_dev *nvm_alloc_dev(int node)
+{
+	return kzalloc_node(sizeof(struct nvm_dev), GFP_KERNEL, node);
+}
+
+int nvm_register(struct nvm_dev *dev)
 {
-	struct nvm_dev *dev;
 	int ret;
 
-	if (!ops->identity)
-		return -EINVAL;
-
-	dev = kzalloc(sizeof(struct nvm_dev), GFP_KERNEL);
-	if (!dev)
-		return -ENOMEM;
-
-	dev->q = q;
-	dev->ops = ops;
-	strncpy(dev->name, disk_name, DISK_NAME_LEN);
-
 	ret = nvm_init(dev);
 	if (ret)
 		goto err_init;
@@ -709,29 +701,17 @@ int nvm_register(struct request_queue *q, char *disk_name,
 	return 0;
 err_init:
 	kfree(dev->lun_map);
-	kfree(dev);
 	return ret;
 }
 EXPORT_SYMBOL(nvm_register);
 
-void nvm_unregister(char *disk_name)
+void nvm_unregister(struct nvm_dev *dev)
 {
-	struct nvm_dev *dev;
-
 	down_write(&nvm_lock);
-	dev = nvm_find_nvm_dev(disk_name);
-	if (!dev) {
-		pr_err("nvm: could not find device %s to unregister\n",
-								disk_name);
-		up_write(&nvm_lock);
-		return;
-	}
-
 	list_del(&dev->devices);
 	up_write(&nvm_lock);
 
 	nvm_exit(dev);
-	kfree(dev);
 }
 EXPORT_SYMBOL(nvm_unregister);
 
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index f615b6b..fe135d9 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -137,12 +137,14 @@ static void nvme_free_ns(struct kref *kref)
 {
 	struct nvme_ns *ns = container_of(kref, struct nvme_ns, kref);
 
-	if (ns->type == NVME_NS_LIGHTNVM)
-		nvme_nvm_unregister(ns->queue, ns->disk->disk_name);
+	if (ns->ndev)
+		nvme_nvm_unregister(ns);
 
-	spin_lock(&dev_list_lock);
-	ns->disk->private_data = NULL;
-	spin_unlock(&dev_list_lock);
+	if (ns->disk) {
+		spin_lock(&dev_list_lock);
+		ns->disk->private_data = NULL;
+		spin_unlock(&dev_list_lock);
+	}
 
 	put_disk(ns->disk);
 	ida_simple_remove(&ns->ctrl->ns_ida, ns->instance);
@@ -788,8 +790,7 @@ static void nvme_config_discard(struct nvme_ns *ns)
 static int nvme_revalidate_ns(struct nvme_ns *ns, struct nvme_id_ns **id)
 {
 	if (nvme_identify_ns(ns->ctrl, ns->ns_id, id)) {
-		dev_warn(disk_to_dev(ns->disk), "%s: Identify failure\n",
-				__func__);
+		dev_warn(ns->ctrl->dev, "%s: Identify failure\n", __func__);
 		return -ENODEV;
 	}
 
@@ -1473,18 +1474,11 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 	sprintf(disk_name, "nvme%dn%d", ctrl->instance, ns->instance);
 
 	if (nvme_nvm_ns_supported(ns, id)) {
-		if (nvme_nvm_register(ns->queue, disk_name)) {
-			dev_warn(ctrl->dev,
-				"%s: LightNVM init failure\n", __func__);
+		if (nvme_nvm_register(ns, disk_name, node)) {
+			dev_warn(ctrl->dev, "%s: LightNVM init failure\n",
+								__func__);
 			goto out_free_id;
 		}
-
-		disk = alloc_disk_node(0, node);
-		if (!disk)
-			goto out_free_id;
-		memcpy(disk->disk_name, disk_name, DISK_NAME_LEN);
-		ns->disk = disk;
-		ns->type = NVME_NS_LIGHTNVM;
 	} else {
 		disk = alloc_disk_node(0, node);
 		if (!disk)
@@ -1532,7 +1526,7 @@ static void nvme_ns_remove(struct nvme_ns *ns)
 	if (test_and_set_bit(NVME_NS_REMOVING, &ns->flags))
 		return;
 
-	if (ns->disk->flags & GENHD_FL_UP) {
+	if (ns->disk && ns->disk->flags & GENHD_FL_UP) {
 		if (blk_get_integrity(ns->disk))
 			blk_integrity_unregister(ns->disk);
 		sysfs_remove_group(&disk_to_dev(ns->disk)->kobj,
@@ -1552,7 +1546,7 @@ static void nvme_validate_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 
 	ns = nvme_find_ns(ctrl, nsid);
 	if (ns) {
-		if (revalidate_disk(ns->disk))
+		if (ns->disk && revalidate_disk(ns->disk))
 			nvme_ns_remove(ns);
 	} else
 		nvme_alloc_ns(ctrl, nsid);
@@ -1856,7 +1850,7 @@ void nvme_kill_queues(struct nvme_ctrl *ctrl)
 		 * Revalidating a dead namespace sets capacity to 0. This will
 		 * end buffered writers dirtying pages that can't be synced.
 		 */
-		if (!test_and_set_bit(NVME_NS_DEAD, &ns->flags))
+		if (ns->disk && !test_and_set_bit(NVME_NS_DEAD, &ns->flags))
 			revalidate_disk(ns->disk);
 
 		blk_set_queue_dying(ns->queue);
diff --git a/drivers/nvme/host/lightnvm.c b/drivers/nvme/host/lightnvm.c
index 97fe610..095e58e 100644
--- a/drivers/nvme/host/lightnvm.c
+++ b/drivers/nvme/host/lightnvm.c
@@ -592,14 +592,27 @@ static struct nvm_dev_ops nvme_nvm_dev_ops = {
 	.max_phys_sect		= 64,
 };
 
-int nvme_nvm_register(struct request_queue *q, char *disk_name)
+int nvme_nvm_register(struct nvme_ns *ns, char *disk_name, int node)
 {
-	return nvm_register(q, disk_name, &nvme_nvm_dev_ops);
+	struct request_queue *q = ns->queue;
+	struct nvm_dev *dev;
+
+	dev = nvm_alloc_dev(node);
+	if (!dev)
+		return -ENOMEM;
+
+	dev->q = q;
+	memcpy(dev->name, disk_name, DISK_NAME_LEN);
+	dev->ops = &nvme_nvm_dev_ops;
+	ns->ndev = dev;
+
+	return nvm_register(dev);
 }
 
-void nvme_nvm_unregister(struct request_queue *q, char *disk_name)
+void nvme_nvm_unregister(struct nvme_ns *ns)
 {
-	nvm_unregister(disk_name);
+	nvm_unregister(ns->ndev);
+	kfree(ns->ndev);
 }
 
 /* move to shared place when used in multiple places. */
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 282421f..7b49b45 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -18,6 +18,7 @@
 #include <linux/pci.h>
 #include <linux/kref.h>
 #include <linux/blk-mq.h>
+#include <linux/lightnvm.h>
 
 enum {
 	/*
@@ -122,6 +123,7 @@ struct nvme_ns {
 	struct nvme_ctrl *ctrl;
 	struct request_queue *queue;
 	struct gendisk *disk;
+	struct nvm_dev *ndev;
 	struct kref kref;
 	int instance;
 
@@ -133,7 +135,6 @@ struct nvme_ns {
 	u16 ms;
 	bool ext;
 	u8 pi_type;
-	int type;
 	unsigned long flags;
 
 #define NVME_NS_REMOVING 0
@@ -269,12 +270,13 @@ int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id);
 int nvme_nvm_register(struct request_queue *q, char *disk_name);
 void nvme_nvm_unregister(struct request_queue *q, char *disk_name);
 #else
-static inline int nvme_nvm_register(struct request_queue *q, char *disk_name)
+static inline int nvme_nvm_register(struct nvme_ns *ns, char *disk_name,
+								int node)
 {
 	return 0;
 }
 
-static inline void nvme_nvm_unregister(struct request_queue *q, char *disk_name) {};
+static inline void nvme_nvm_unregister(struct nvme_ns *ns) {};
 
 static inline int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
 {
diff --git a/include/linux/lightnvm.h b/include/linux/lightnvm.h
index ba78b83..5afc263 100644
--- a/include/linux/lightnvm.h
+++ b/include/linux/lightnvm.h
@@ -524,9 +524,9 @@ extern struct nvm_block *nvm_get_blk(struct nvm_dev *, struct nvm_lun *,
 								unsigned long);
 extern void nvm_put_blk(struct nvm_dev *, struct nvm_block *);
 
-extern int nvm_register(struct request_queue *, char *,
-						struct nvm_dev_ops *);
-extern void nvm_unregister(char *);
+extern struct nvm_dev *nvm_alloc_dev(int);
+extern int nvm_register(struct nvm_dev *);
+extern void nvm_unregister(struct nvm_dev *);
 
 void nvm_mark_blk(struct nvm_dev *dev, struct ppa_addr ppa, int type);
 
@@ -575,11 +575,14 @@ extern int nvm_dev_factory(struct nvm_dev *, int flags);
 #else /* CONFIG_NVM */
 struct nvm_dev_ops;
 
-static inline int nvm_register(struct request_queue *q, char *disk_name,
-							struct nvm_dev_ops *ops)
+static inline struct nvm_dev *nvm_alloc_dev(int node)
+{
+	return ERR_PTR(-EINVAL);
+}
+static inline int nvm_register(struct nvm_dev *dev)
 {
 	return -EINVAL;
 }
-static inline void nvm_unregister(char *disk_name) {}
+static inline void nvm_unregister(struct nvm_dev *dev) {}
 #endif /* CONFIG_NVM */
 #endif /* LIGHTNVM.H */
-- 
2.1.4

[PATCH 3/6] blk-mq: register device instead of disk

[Matias Bjørling]

Enable devices without a gendisk instance to register itself with blk-mq
and expose the associated multi-queue sysfs entries.

Signed-off-by: Matias Bjørling <m@bjorling.me>
---
 block/blk-mq-sysfs.c   | 13 +++++--------
 block/blk-sysfs.c      |  4 ++--
 drivers/md/dm.c        |  2 +-
 include/linux/blk-mq.h |  4 ++--
 4 files changed, 10 insertions(+), 13 deletions(-)

diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index 4ea4dd8..f67b5c1 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -380,9 +380,8 @@ static int blk_mq_register_hctx(struct blk_mq_hw_ctx *hctx)
 	return ret;
 }
 
-void blk_mq_unregister_disk(struct gendisk *disk)
+void blk_mq_unregister_dev(struct device *dev, struct request_queue *q)
 {
-	struct request_queue *q = disk->queue;
 	struct blk_mq_hw_ctx *hctx;
 	struct blk_mq_ctx *ctx;
 	int i, j;
@@ -402,7 +401,7 @@ void blk_mq_unregister_disk(struct gendisk *disk)
 	kobject_del(&q->mq_kobj);
 	kobject_put(&q->mq_kobj);
 
-	kobject_put(&disk_to_dev(disk)->kobj);
+	kobject_put(&dev->kobj);
 
 	q->mq_sysfs_init_done = false;
 	blk_mq_enable_hotplug();
@@ -426,10 +425,8 @@ static void blk_mq_sysfs_init(struct request_queue *q)
 	}
 }
 
-int blk_mq_register_disk(struct gendisk *disk)
+int blk_mq_register_dev(struct device *dev, struct request_queue *q)
 {
-	struct device *dev = disk_to_dev(disk);
-	struct request_queue *q = disk->queue;
 	struct blk_mq_hw_ctx *hctx;
 	int ret, i;
 
@@ -450,7 +447,7 @@ int blk_mq_register_disk(struct gendisk *disk)
 	}
 
 	if (ret)
-		blk_mq_unregister_disk(disk);
+		blk_mq_unregister_dev(dev, q);
 	else
 		q->mq_sysfs_init_done = true;
 out:
@@ -458,7 +455,7 @@ out:
 
 	return ret;
 }
-EXPORT_SYMBOL_GPL(blk_mq_register_disk);
+EXPORT_SYMBOL_GPL(blk_mq_register_dev);
 
 void blk_mq_sysfs_unregister(struct request_queue *q)
 {
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 9920596..5223c5f 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -693,7 +693,7 @@ int blk_register_queue(struct gendisk *disk)
 	kobject_uevent(&q->kobj, KOBJ_ADD);
 
 	if (q->mq_ops)
-		blk_mq_register_disk(disk);
+		blk_mq_register_dev(disk_to_dev(disk), q);
 
 	if (!q->request_fn)
 		return 0;
@@ -718,7 +718,7 @@ void blk_unregister_queue(struct gendisk *disk)
 		return;
 
 	if (q->mq_ops)
-		blk_mq_unregister_disk(disk);
+		blk_mq_unregister_dev(disk_to_dev(disk), q);
 
 	if (q->request_fn)
 		elv_unregister_queue(q);
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index aba7ed9..9de3a9c 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -2798,7 +2798,7 @@ static int dm_mq_init_request_queue(struct mapped_device *md,
 	dm_init_md_queue(md);
 
 	/* backfill 'mq' sysfs registration normally done in blk_register_queue */
-	blk_mq_register_disk(md->disk);
+	blk_mq_register_dev(disk_to_dev(md->disk), q);
 
 	return 0;
 
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 2498fdf..ab6fb32 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -176,8 +176,8 @@ enum {
 struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *);
 struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 						  struct request_queue *q);
-int blk_mq_register_disk(struct gendisk *);
-void blk_mq_unregister_disk(struct gendisk *);
+int blk_mq_register_dev(struct device *, struct request_queue *);
+void blk_mq_unregister_dev(struct device *, struct request_queue *);
 
 int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set);
 void blk_mq_free_tag_set(struct blk_mq_tag_set *set);
-- 
2.1.4