1 files changed, 335 insertions, 22 deletions

diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
index 2a7635565083..3da980dc60d9 100644
--- a/drivers/nvme/host/multipath.c
+++ b/drivers/nvme/host/multipath.c
@@ -10,10 +10,61 @@
 #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",
 	[NVME_IOPOLICY_RR]	= "round-robin",
@@ -427,7 +478,7 @@ static bool nvme_available_path(struct nvme_ns_head *head)
 	struct nvme_ns *ns;
 
 	if (!test_bit(NVME_NSHEAD_DISK_LIVE, &head->flags))
-		return NULL;
+		return false;
 
 	list_for_each_entry_srcu(ns, &head->list, siblings,
 				 srcu_read_lock_held(&head->srcu)) {
@@ -442,7 +493,17 @@ static bool nvme_available_path(struct nvme_ns_head *head)
 			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)
@@ -617,6 +678,40 @@ static void nvme_requeue_work(struct work_struct *work)
 	}
 }
 
+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;
@@ -626,19 +721,31 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
 	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);
 	lim.dma_alignment = 3;
-	lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT | BLK_FEAT_POLL;
+	lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT |
+		BLK_FEAT_POLL | BLK_FEAT_ATOMIC_WRITES;
 	if (head->ids.csi == NVME_CSI_ZNS)
 		lim.features |= BLK_FEAT_ZONED;
 
@@ -653,12 +760,13 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
 	 * controller's scan_work context. If a path error occurs here, the IO
 	 * will wait until a path becomes available or all paths are torn down,
 	 * but that action also occurs within scan_work, so it would deadlock.
-	 * Defer the partion scan to a different context that does not block
+	 * Defer the partition scan to a different context that does not block
 	 * scan_work.
 	 */
 	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;
 }
 
@@ -686,6 +794,8 @@ static void nvme_mpath_set_live(struct nvme_ns *ns)
 		kblockd_schedule_work(&head->partition_scan_work);
 	}
 
+	nvme_mpath_add_sysfs_link(ns->head);
+
 	mutex_lock(&head->lock);
 	if (nvme_path_is_optimized(ns)) {
 		int node, srcu_idx;
@@ -768,6 +878,25 @@ static void nvme_update_ns_ana_state(struct nvme_ana_group_desc *desc,
 	if (nvme_state_is_live(ns->ana_state) &&
 	    nvme_ctrl_state(ns->ctrl) == NVME_CTRL_LIVE)
 		nvme_mpath_set_live(ns);
+	else {
+		/*
+		 * Add sysfs link from multipath head gendisk node to path
+		 * device gendisk node.
+		 * If path's ana state is live (i.e. state is either optimized
+		 * or non-optimized) while we alloc the ns then sysfs link would
+		 * be created from nvme_mpath_set_live(). In that case we would
+		 * not fallthrough this code path. However for the path's ana
+		 * state other than live, we call nvme_mpath_set_live() only
+		 * after ana state transitioned to the live state. But we still
+		 * want to create the sysfs link from head node to a path device
+		 * irrespctive of the path's ana state.
+		 * If we reach through here then it means that path's ana state
+		 * is not live but still create the sysfs link to this path from
+		 * head node if head node of the path has already come alive.
+		 */
+		if (test_bit(NVME_NSHEAD_DISK_LIVE, &ns->head->flags))
+			nvme_mpath_add_sysfs_link(ns->head);
+	}
 }
 
 static int nvme_update_ana_state(struct nvme_ctrl *ctrl,
@@ -839,7 +968,7 @@ static int nvme_read_ana_log(struct nvme_ctrl *ctrl)
 	if (nr_change_groups)
 		mod_timer(&ctrl->anatt_timer, ctrl->anatt * HZ * 2 + jiffies);
 	else
-		del_timer_sync(&ctrl->anatt_timer);
+		timer_delete_sync(&ctrl->anatt_timer);
 out_unlock:
 	mutex_unlock(&ctrl->ana_lock);
 	return error;
@@ -869,7 +998,7 @@ void nvme_mpath_update(struct nvme_ctrl *ctrl)
 
 static void nvme_anatt_timeout(struct timer_list *t)
 {
-	struct nvme_ctrl *ctrl = from_timer(ctrl, t, anatt_timer);
+	struct nvme_ctrl *ctrl = timer_container_of(ctrl, t, anatt_timer);
 
 	dev_info(ctrl->device, "ANATT timeout, resetting controller.\n");
 	nvme_reset_ctrl(ctrl);
@@ -879,7 +1008,7 @@ void nvme_mpath_stop(struct nvme_ctrl *ctrl)
 {
 	if (!nvme_ctrl_use_ana(ctrl))
 		return;
-	del_timer_sync(&ctrl->anatt_timer);
+	timer_delete_sync(&ctrl->anatt_timer);
 	cancel_work_sync(&ctrl->ana_work);
 }
 
@@ -955,6 +1084,88 @@ static ssize_t ana_state_show(struct device *dev, struct device_attribute *attr,
 }
 DEVICE_ATTR_RO(ana_state);
 
+static ssize_t queue_depth_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct nvme_ns *ns = nvme_get_ns_from_dev(dev);
+
+	if (ns->head->subsys->iopolicy != NVME_IOPOLICY_QD)
+		return 0;
+
+	return sysfs_emit(buf, "%d\n", atomic_read(&ns->ctrl->nr_active));
+}
+DEVICE_ATTR_RO(queue_depth);
+
+static ssize_t numa_nodes_show(struct device *dev, struct device_attribute *attr,
+		char *buf)
+{
+	int node, srcu_idx;
+	nodemask_t numa_nodes;
+	struct nvme_ns *current_ns;
+	struct nvme_ns *ns = nvme_get_ns_from_dev(dev);
+	struct nvme_ns_head *head = ns->head;
+
+	if (head->subsys->iopolicy != NVME_IOPOLICY_NUMA)
+		return 0;
+
+	nodes_clear(numa_nodes);
+
+	srcu_idx = srcu_read_lock(&head->srcu);
+	for_each_node(node) {
+		current_ns = srcu_dereference(head->current_path[node],
+				&head->srcu);
+		if (ns == current_ns)
+			node_set(node, numa_nodes);
+	}
+	srcu_read_unlock(&head->srcu, srcu_idx);
+
+	return sysfs_emit(buf, "%*pbl\n", nodemask_pr_args(&numa_nodes));
+}
+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)
 {
@@ -967,6 +1178,85 @@ static int nvme_lookup_ana_group_desc(struct nvme_ctrl *ctrl,
 	return -ENXIO; /* just break out of the loop */
 }
 
+void nvme_mpath_add_sysfs_link(struct nvme_ns_head *head)
+{
+	struct device *target;
+	int rc, srcu_idx;
+	struct nvme_ns *ns;
+	struct kobject *kobj;
+
+	/*
+	 * Ensure head disk node is already added otherwise we may get invalid
+	 * kobj for head disk node
+	 */
+	if (!test_bit(GD_ADDED, &head->disk->state))
+		return;
+
+	kobj = &disk_to_dev(head->disk)->kobj;
+
+	/*
+	 * loop through each ns chained through the head->list and create the
+	 * sysfs link from head node to the ns path node
+	 */
+	srcu_idx = srcu_read_lock(&head->srcu);
+
+	list_for_each_entry_srcu(ns, &head->list, siblings,
+				 srcu_read_lock_held(&head->srcu)) {
+		/*
+		 * Ensure that ns path disk node is already added otherwise we
+		 * may get invalid kobj name for target
+		 */
+		if (!test_bit(GD_ADDED, &ns->disk->state))
+			continue;
+
+		/*
+		 * Avoid creating link if it already exists for the given path.
+		 * When path ana state transitions from optimized to non-
+		 * optimized or vice-versa, the nvme_mpath_set_live() is
+		 * invoked which in truns call this function. Now if the sysfs
+		 * link already exists for the given path and we attempt to re-
+		 * create the link then sysfs code would warn about it loudly.
+		 * So we evaluate NVME_NS_SYSFS_ATTR_LINK flag here to ensure
+		 * that we're not creating duplicate link.
+		 * The test_and_set_bit() is used because it is protecting
+		 * against multiple nvme paths being simultaneously added.
+		 */
+		if (test_and_set_bit(NVME_NS_SYSFS_ATTR_LINK, &ns->flags))
+			continue;
+
+		target = disk_to_dev(ns->disk);
+		/*
+		 * Create sysfs link from head gendisk kobject @kobj to the
+		 * ns path gendisk kobject @target->kobj.
+		 */
+		rc = sysfs_add_link_to_group(kobj, nvme_ns_mpath_attr_group.name,
+				&target->kobj, dev_name(target));
+		if (unlikely(rc)) {
+			dev_err(disk_to_dev(ns->head->disk),
+					"failed to create link to %s\n",
+					dev_name(target));
+			clear_bit(NVME_NS_SYSFS_ATTR_LINK, &ns->flags);
+		}
+	}
+
+	srcu_read_unlock(&head->srcu, srcu_idx);
+}
+
+void nvme_mpath_remove_sysfs_link(struct nvme_ns *ns)
+{
+	struct device *target;
+	struct kobject *kobj;
+
+	if (!test_bit(NVME_NS_SYSFS_ATTR_LINK, &ns->flags))
+		return;
+
+	target = disk_to_dev(ns->disk);
+	kobj = &disk_to_dev(ns->head->disk)->kobj;
+	sysfs_remove_link_from_group(kobj, nvme_ns_mpath_attr_group.name,
+			dev_name(target));
+	clear_bit(NVME_NS_SYSFS_ATTR_LINK, &ns->flags);
+}
+
 void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid)
 {
 	if (nvme_ctrl_use_ana(ns->ctrl)) {
@@ -998,23 +1288,46 @@ void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid)
 #endif
 }
 
-void nvme_mpath_shutdown_disk(struct nvme_ns_head *head)
+void nvme_mpath_remove_disk(struct nvme_ns_head *head)
 {
+	bool remove = false;
+
 	if (!head->disk)
 		return;
-	if (test_and_clear_bit(NVME_NSHEAD_DISK_LIVE, &head->flags)) {
-		nvme_cdev_del(&head->cdev, &head->cdev_device);
+
+	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;
+		mod_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;