]> git.ipfire.org Git - thirdparty/linux.git/commitdiff
nstree: switch to new structures
authorChristian Brauner <brauner@kernel.org>
Mon, 10 Nov 2025 15:08:17 +0000 (16:08 +0100)
committerChristian Brauner <brauner@kernel.org>
Tue, 11 Nov 2025 09:01:30 +0000 (10:01 +0100)
Switch the nstree management to the new combined structures.

Link: https://patch.msgid.link/20251110-work-namespace-nstree-fixes-v1-5-e8a9264e0fb9@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
fs/namespace.c
include/linux/ns/ns_common_types.h
include/linux/ns/nstree_types.h
include/linux/ns_common.h
include/linux/nstree.h
kernel/nscommon.c
kernel/nstree.c

index eded33eeb6473ee2ef8ad9c75872f3c3c50cf906..ad19530a13b225fc79c8716286cb84e152b1670a 100644 (file)
@@ -138,7 +138,7 @@ static inline struct mnt_namespace *node_to_mnt_ns(const struct rb_node *node)
 
        if (!node)
                return NULL;
-       ns = rb_entry(node, struct ns_common, ns_tree_node);
+       ns = rb_entry(node, struct ns_common, ns_tree_node.ns_node);
        return container_of(ns, struct mnt_namespace, ns);
 }
 
index ccd1d1e116f644a9783b995567e0444a797a42c4..b332b019b29cb8bc15f7d5fba6a945cbe5129b3a 100644 (file)
@@ -3,6 +3,7 @@
 #define _LINUX_NS_COMMON_TYPES_H
 
 #include <linux/atomic.h>
+#include <linux/ns/nstree_types.h>
 #include <linux/rbtree.h>
 #include <linux/refcount.h>
 #include <linux/types.h>
@@ -98,6 +99,13 @@ extern const struct proc_ns_operations utsns_operations;
  * Initial namespaces:
  *   Boot-time namespaces (init_net, init_pid_ns, etc.) start with
  *   __ns_ref_active = 1 and remain active forever.
+ *
+ * @ns_type: type of namespace (e.g., CLONE_NEWNET)
+ * @stashed: cached dentry to be used by the vfs
+ * @ops: namespace operations
+ * @inum: namespace inode number (quickly recycled for non-initial namespaces)
+ * @__ns_ref: main reference count (do not use directly)
+ * @ns_tree: namespace tree nodes and active reference count
  */
 struct ns_common {
        u32 ns_type;
@@ -106,24 +114,7 @@ struct ns_common {
        unsigned int inum;
        refcount_t __ns_ref; /* do not use directly */
        union {
-               struct {
-                       u64 ns_id;
-                       struct /* global namespace rbtree and list */ {
-                               struct rb_node ns_unified_tree_node;
-                               struct list_head ns_unified_list_node;
-                       };
-                       struct /* per type rbtree and list */ {
-                               struct rb_node ns_tree_node;
-                               struct list_head ns_list_node;
-                       };
-                       struct /* namespace ownership rbtree and list */ {
-                               struct rb_root ns_owner_tree; /* rbtree of namespaces owned by this namespace */
-                               struct list_head ns_owner; /* list of namespaces owned by this namespace */
-                               struct rb_node ns_owner_tree_node; /* node in the owner namespace's rbtree */
-                               struct list_head ns_owner_entry; /* node in the owner namespace's ns_owned list */
-                       };
-                       atomic_t __ns_ref_active; /* do not use directly */
-               };
+               struct ns_tree;
                struct rcu_head ns_rcu;
        };
 };
index 6ee0c39686f8429b56d1403c416c74a2feff6256..2fb28ee31efbb620d5ba3f11d92589799070645d 100644 (file)
@@ -33,4 +33,23 @@ struct ns_tree_node {
        struct list_head ns_list_entry;
 };
 
+/**
+ * struct ns_tree - Namespace tree nodes and active reference count
+ * @ns_id: Unique namespace identifier
+ * @__ns_ref_active: Active reference count (do not use directly)
+ * @ns_unified_node: Node in the global namespace tree
+ * @ns_tree_node: Node in the per-type namespace tree
+ * @ns_owner_node: Node in the owner namespace's tree of owned namespaces
+ * @ns_owner_root: Root of the tree of namespaces owned by this namespace
+ *                 (only used when this namespace is an owner)
+ */
+struct ns_tree {
+       u64 ns_id;
+       atomic_t __ns_ref_active;
+       struct ns_tree_node ns_unified_node;
+       struct ns_tree_node ns_tree_node;
+       struct ns_tree_node ns_owner_node;
+       struct ns_tree_root ns_owner_root;
+};
+
 #endif /* _LINUX_NSTREE_TYPES_H */
index 6a4ca8c3b9c4243e81d38123ba44bceaf6ba88ce..f90509ee09007572f956d57e5af8b6d2147e165e 100644 (file)
@@ -26,20 +26,19 @@ static __always_inline bool is_ns_init_id(const struct ns_common *ns)
        return ns->ns_id <= NS_LAST_INIT_ID;
 }
 
-
-#define NS_COMMON_INIT(nsname, refs)                                                   \
-{                                                                                      \
-       .ns_type                = ns_common_type(&nsname),                              \
-       .ns_id                  = ns_init_id(&nsname),                                  \
-       .inum                   = ns_init_inum(&nsname),                                \
-       .ops                    = to_ns_operations(&nsname),                            \
-       .stashed                = NULL,                                                 \
-       .__ns_ref               = REFCOUNT_INIT(refs),                                  \
-       .__ns_ref_active        = ATOMIC_INIT(1),                                       \
-       .ns_list_node           = LIST_HEAD_INIT(nsname.ns.ns_list_node),               \
-       .ns_owner_entry         = LIST_HEAD_INIT(nsname.ns.ns_owner_entry),             \
-       .ns_owner               = LIST_HEAD_INIT(nsname.ns.ns_owner),                   \
-       .ns_unified_list_node   = LIST_HEAD_INIT(nsname.ns.ns_unified_list_node),       \
+#define NS_COMMON_INIT(nsname, refs)                                                                   \
+{                                                                                                      \
+       .ns_type                        = ns_common_type(&nsname),                                      \
+       .ns_id                          = ns_init_id(&nsname),                                          \
+       .inum                           = ns_init_inum(&nsname),                                        \
+       .ops                            = to_ns_operations(&nsname),                                    \
+       .stashed                        = NULL,                                                         \
+       .__ns_ref                       = REFCOUNT_INIT(refs),                                          \
+       .__ns_ref_active                = ATOMIC_INIT(1),                                               \
+       .ns_unified_node.ns_list_entry  = LIST_HEAD_INIT(nsname.ns.ns_unified_node.ns_list_entry),      \
+       .ns_tree_node.ns_list_entry     = LIST_HEAD_INIT(nsname.ns.ns_tree_node.ns_list_entry),         \
+       .ns_owner_node.ns_list_entry    = LIST_HEAD_INIT(nsname.ns.ns_owner_node.ns_list_entry),        \
+       .ns_owner_root.ns_list_head     = LIST_HEAD_INIT(nsname.ns.ns_owner_root.ns_list_head),         \
 }
 
 #define ns_common_init(__ns)                     \
index 98b848cf2f1ce77d160d2775d1952a18a292cdad..175e4625bfa6d9db0992bf1b6c8b9540cf9bf1f8 100644 (file)
 
 struct ns_common;
 
-extern struct ns_tree cgroup_ns_tree;
-extern struct ns_tree ipc_ns_tree;
-extern struct ns_tree mnt_ns_tree;
-extern struct ns_tree net_ns_tree;
-extern struct ns_tree pid_ns_tree;
-extern struct ns_tree time_ns_tree;
-extern struct ns_tree user_ns_tree;
-extern struct ns_tree uts_ns_tree;
+extern struct ns_tree_root cgroup_ns_tree;
+extern struct ns_tree_root ipc_ns_tree;
+extern struct ns_tree_root mnt_ns_tree;
+extern struct ns_tree_root net_ns_tree;
+extern struct ns_tree_root pid_ns_tree;
+extern struct ns_tree_root time_ns_tree;
+extern struct ns_tree_root user_ns_tree;
+extern struct ns_tree_root uts_ns_tree;
 
 void ns_tree_node_init(struct ns_tree_node *node);
 void ns_tree_root_init(struct ns_tree_root *root);
@@ -46,14 +46,14 @@ void ns_tree_node_del(struct ns_tree_node *node, struct ns_tree_root *root);
                         (((__ns) == ns_init_ns(__ns)) ? ns_init_id(__ns) : 0))
 
 u64 __ns_tree_gen_id(struct ns_common *ns, u64 id);
-void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree *ns_tree);
-void __ns_tree_remove(struct ns_common *ns, struct ns_tree *ns_tree);
+void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree_root *ns_tree);
+void __ns_tree_remove(struct ns_common *ns, struct ns_tree_root *ns_tree);
 struct ns_common *ns_tree_lookup_rcu(u64 ns_id, int ns_type);
 struct ns_common *__ns_tree_adjoined_rcu(struct ns_common *ns,
-                                        struct ns_tree *ns_tree,
+                                        struct ns_tree_root *ns_tree,
                                         bool previous);
 
-static inline void __ns_tree_add(struct ns_common *ns, struct ns_tree *ns_tree, u64 id)
+static inline void __ns_tree_add(struct ns_common *ns, struct ns_tree_root *ns_tree, u64 id)
 {
        __ns_tree_gen_id(ns, id);
        __ns_tree_add_raw(ns, ns_tree);
@@ -91,6 +91,6 @@ static inline void __ns_tree_add(struct ns_common *ns, struct ns_tree *ns_tree,
 #define ns_tree_adjoined_rcu(__ns, __previous) \
        __ns_tree_adjoined_rcu(to_ns_common(__ns), to_ns_tree(__ns), __previous)
 
-#define ns_tree_active(__ns) (!RB_EMPTY_NODE(&to_ns_common(__ns)->ns_tree_node))
+#define ns_tree_active(__ns) (!RB_EMPTY_NODE(&to_ns_common(__ns)->ns_tree_node.ns_node))
 
 #endif /* _LINUX_NSTREE_H */
index c910b979e4336c3a017eb21ddd1c21a2a17dc57e..88f70baccb75ecf13fc58b8a28f31727c5e42e69 100644 (file)
@@ -2,6 +2,7 @@
 /* Copyright (c) 2025 Christian Brauner <brauner@kernel.org> */
 
 #include <linux/ns_common.h>
+#include <linux/nstree.h>
 #include <linux/proc_ns.h>
 #include <linux/user_namespace.h>
 #include <linux/vfsdebug.h>
@@ -61,14 +62,10 @@ int __ns_common_init(struct ns_common *ns, u32 ns_type, const struct proc_ns_ope
        ns->ops = ops;
        ns->ns_id = 0;
        ns->ns_type = ns_type;
-       RB_CLEAR_NODE(&ns->ns_tree_node);
-       RB_CLEAR_NODE(&ns->ns_unified_tree_node);
-       RB_CLEAR_NODE(&ns->ns_owner_tree_node);
-       INIT_LIST_HEAD(&ns->ns_list_node);
-       INIT_LIST_HEAD(&ns->ns_unified_list_node);
-       ns->ns_owner_tree = RB_ROOT;
-       INIT_LIST_HEAD(&ns->ns_owner);
-       INIT_LIST_HEAD(&ns->ns_owner_entry);
+       ns_tree_node_init(&ns->ns_tree_node);
+       ns_tree_node_init(&ns->ns_unified_node);
+       ns_tree_node_init(&ns->ns_owner_node);
+       ns_tree_root_init(&ns->ns_owner_root);
 
 #ifdef CONFIG_DEBUG_VFS
        ns_debug(ns, ops);
index fe71ff943f703a55190e7b20a817135d4a6b85e8..6c7ec9fbf25f2e706d75a7fe87cd1980c017750b 100644 (file)
@@ -9,68 +9,51 @@
 #include <linux/user_namespace.h>
 
 static __cacheline_aligned_in_smp DEFINE_SEQLOCK(ns_tree_lock);
-static struct rb_root ns_unified_tree = RB_ROOT; /* protected by ns_tree_lock */
-static LIST_HEAD(ns_unified_list); /* protected by ns_tree_lock */
 
-/**
- * struct ns_tree - Namespace tree
- * @ns_tree: Rbtree of namespaces of a particular type
- * @ns_list: Sequentially walkable list of all namespaces of this type
- * @type: type of namespaces in this tree
- */
-struct ns_tree {
-       struct rb_root ns_tree;
-       struct list_head ns_list;
-       int type;
+static struct ns_tree_root ns_unified_root = { /* protected by ns_tree_lock */
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(ns_unified_root.ns_list_head),
 };
 
-struct ns_tree mnt_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(mnt_ns_tree.ns_list),
-       .type = CLONE_NEWNS,
+struct ns_tree_root mnt_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(mnt_ns_tree.ns_list_head),
 };
 
-struct ns_tree net_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(net_ns_tree.ns_list),
-       .type = CLONE_NEWNET,
+struct ns_tree_root net_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(net_ns_tree.ns_list_head),
 };
 EXPORT_SYMBOL_GPL(net_ns_tree);
 
-struct ns_tree uts_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(uts_ns_tree.ns_list),
-       .type = CLONE_NEWUTS,
+struct ns_tree_root uts_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(uts_ns_tree.ns_list_head),
 };
 
-struct ns_tree user_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(user_ns_tree.ns_list),
-       .type = CLONE_NEWUSER,
+struct ns_tree_root user_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(user_ns_tree.ns_list_head),
 };
 
-struct ns_tree ipc_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(ipc_ns_tree.ns_list),
-       .type = CLONE_NEWIPC,
+struct ns_tree_root ipc_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(ipc_ns_tree.ns_list_head),
 };
 
-struct ns_tree pid_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(pid_ns_tree.ns_list),
-       .type = CLONE_NEWPID,
+struct ns_tree_root pid_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(pid_ns_tree.ns_list_head),
 };
 
-struct ns_tree cgroup_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(cgroup_ns_tree.ns_list),
-       .type = CLONE_NEWCGROUP,
+struct ns_tree_root cgroup_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(cgroup_ns_tree.ns_list_head),
 };
 
-struct ns_tree time_ns_tree = {
-       .ns_tree = RB_ROOT,
-       .ns_list = LIST_HEAD_INIT(time_ns_tree.ns_list),
-       .type = CLONE_NEWTIME,
+struct ns_tree_root time_ns_tree = {
+       .ns_rb = RB_ROOT,
+       .ns_list_head = LIST_HEAD_INIT(time_ns_tree.ns_list_head),
 };
 
 /**
@@ -162,21 +145,21 @@ static inline struct ns_common *node_to_ns(const struct rb_node *node)
 {
        if (!node)
                return NULL;
-       return rb_entry(node, struct ns_common, ns_tree_node);
+       return rb_entry(node, struct ns_common, ns_tree_node.ns_node);
 }
 
 static inline struct ns_common *node_to_ns_unified(const struct rb_node *node)
 {
        if (!node)
                return NULL;
-       return rb_entry(node, struct ns_common, ns_unified_tree_node);
+       return rb_entry(node, struct ns_common, ns_unified_node.ns_node);
 }
 
 static inline struct ns_common *node_to_ns_owner(const struct rb_node *node)
 {
        if (!node)
                return NULL;
-       return rb_entry(node, struct ns_common, ns_owner_tree_node);
+       return rb_entry(node, struct ns_common, ns_owner_node.ns_node);
 }
 
 static int ns_id_cmp(u64 id_a, u64 id_b)
@@ -203,35 +186,22 @@ static int ns_cmp_owner(struct rb_node *a, const struct rb_node *b)
        return ns_id_cmp(node_to_ns_owner(a)->ns_id, node_to_ns_owner(b)->ns_id);
 }
 
-void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree *ns_tree)
+void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree_root *ns_tree)
 {
-       struct rb_node *node, *prev;
+       struct rb_node *node;
        const struct proc_ns_operations *ops = ns->ops;
 
        VFS_WARN_ON_ONCE(!ns->ns_id);
-       VFS_WARN_ON_ONCE(ns->ns_type != ns_tree->type);
 
        write_seqlock(&ns_tree_lock);
 
-       node = rb_find_add_rcu(&ns->ns_tree_node, &ns_tree->ns_tree, ns_cmp);
-       /*
-        * If there's no previous entry simply add it after the
-        * head and if there is add it after the previous entry.
-        */
-       prev = rb_prev(&ns->ns_tree_node);
-       if (!prev)
-               list_add_rcu(&ns->ns_list_node, &ns_tree->ns_list);
-       else
-               list_add_rcu(&ns->ns_list_node, &node_to_ns(prev)->ns_list_node);
+       /* Add to per-type tree and list */
+       node = ns_tree_node_add(&ns->ns_tree_node, ns_tree, ns_cmp);
 
        /* Add to unified tree and list */
-       rb_find_add_rcu(&ns->ns_unified_tree_node, &ns_unified_tree, ns_cmp_unified);
-       prev = rb_prev(&ns->ns_unified_tree_node);
-       if (!prev)
-               list_add_rcu(&ns->ns_unified_list_node, &ns_unified_list);
-       else
-               list_add_rcu(&ns->ns_unified_list_node, &node_to_ns_unified(prev)->ns_unified_list_node);
+       ns_tree_node_add(&ns->ns_unified_node, &ns_unified_root, ns_cmp_unified);
 
+       /* Add to owner's tree if applicable */
        if (ops) {
                struct user_namespace *user_ns;
 
@@ -241,15 +211,8 @@ void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree *ns_tree)
                        struct ns_common *owner = &user_ns->ns;
                        VFS_WARN_ON_ONCE(owner->ns_type != CLONE_NEWUSER);
 
-                       /* Insert into owner's rbtree */
-                       rb_find_add_rcu(&ns->ns_owner_tree_node, &owner->ns_owner_tree, ns_cmp_owner);
-
-                       /* Insert into owner's list in sorted order */
-                       prev = rb_prev(&ns->ns_owner_tree_node);
-                       if (!prev)
-                               list_add_rcu(&ns->ns_owner_entry, &owner->ns_owner);
-                       else
-                               list_add_rcu(&ns->ns_owner_entry, &node_to_ns_owner(prev)->ns_owner_entry);
+                       /* Insert into owner's tree and list */
+                       ns_tree_node_add(&ns->ns_owner_node, &owner->ns_owner_root, ns_cmp_owner);
                } else {
                        /* Only the initial user namespace doesn't have an owner. */
                        VFS_WARN_ON_ONCE(ns != to_ns_common(&init_user_ns));
@@ -260,36 +223,29 @@ void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree *ns_tree)
        VFS_WARN_ON_ONCE(node);
 }
 
-void __ns_tree_remove(struct ns_common *ns, struct ns_tree *ns_tree)
+void __ns_tree_remove(struct ns_common *ns, struct ns_tree_root *ns_tree)
 {
        const struct proc_ns_operations *ops = ns->ops;
        struct user_namespace *user_ns;
 
-       VFS_WARN_ON_ONCE(RB_EMPTY_NODE(&ns->ns_tree_node));
-       VFS_WARN_ON_ONCE(list_empty(&ns->ns_list_node));
-       VFS_WARN_ON_ONCE(ns->ns_type != ns_tree->type);
+       VFS_WARN_ON_ONCE(ns_tree_node_empty(&ns->ns_tree_node));
+       VFS_WARN_ON_ONCE(list_empty(&ns->ns_tree_node.ns_list_entry));
 
        write_seqlock(&ns_tree_lock);
-       rb_erase(&ns->ns_tree_node, &ns_tree->ns_tree);
-       RB_CLEAR_NODE(&ns->ns_tree_node);
-
-       list_bidir_del_rcu(&ns->ns_list_node);
 
-       rb_erase(&ns->ns_unified_tree_node, &ns_unified_tree);
-       RB_CLEAR_NODE(&ns->ns_unified_tree_node);
+       /* Remove from per-type tree and list */
+       ns_tree_node_del(&ns->ns_tree_node, ns_tree);
 
-       list_bidir_del_rcu(&ns->ns_unified_list_node);
+       /* Remove from unified tree and list */
+       ns_tree_node_del(&ns->ns_unified_node, &ns_unified_root);
 
-       /* Remove from owner's rbtree if this namespace has an owner */
+       /* Remove from owner's tree if applicable */
        if (ops) {
                user_ns = ops->owner(ns);
                if (user_ns) {
                        struct ns_common *owner = &user_ns->ns;
-                       rb_erase(&ns->ns_owner_tree_node, &owner->ns_owner_tree);
-                       RB_CLEAR_NODE(&ns->ns_owner_tree_node);
+                       ns_tree_node_del(&ns->ns_owner_node, &owner->ns_owner_root);
                }
-
-               list_bidir_del_rcu(&ns->ns_owner_entry);
        }
 
        write_sequnlock(&ns_tree_lock);
@@ -320,7 +276,7 @@ static int ns_find_unified(const void *key, const struct rb_node *node)
        return 0;
 }
 
-static struct ns_tree *ns_tree_from_type(int ns_type)
+static struct ns_tree_root *ns_tree_from_type(int ns_type)
 {
        switch (ns_type) {
        case CLONE_NEWCGROUP:
@@ -351,7 +307,7 @@ static struct ns_common *__ns_unified_tree_lookup_rcu(u64 ns_id)
 
        do {
                seq = read_seqbegin(&ns_tree_lock);
-               node = rb_find_rcu(&ns_id, &ns_unified_tree, ns_find_unified);
+               node = rb_find_rcu(&ns_id, &ns_unified_root.ns_rb, ns_find_unified);
                if (node)
                        break;
        } while (read_seqretry(&ns_tree_lock, seq));
@@ -361,7 +317,7 @@ static struct ns_common *__ns_unified_tree_lookup_rcu(u64 ns_id)
 
 static struct ns_common *__ns_tree_lookup_rcu(u64 ns_id, int ns_type)
 {
-       struct ns_tree *ns_tree;
+       struct ns_tree_root *ns_tree;
        struct rb_node *node;
        unsigned int seq;
 
@@ -371,7 +327,7 @@ static struct ns_common *__ns_tree_lookup_rcu(u64 ns_id, int ns_type)
 
        do {
                seq = read_seqbegin(&ns_tree_lock);
-               node = rb_find_rcu(&ns_id, &ns_tree->ns_tree, ns_find);
+               node = rb_find_rcu(&ns_id, &ns_tree->ns_rb, ns_find);
                if (node)
                        break;
        } while (read_seqretry(&ns_tree_lock, seq));
@@ -399,22 +355,20 @@ struct ns_common *ns_tree_lookup_rcu(u64 ns_id, int ns_type)
  * there is no next/previous namespace, -ENOENT is returned.
  */
 struct ns_common *__ns_tree_adjoined_rcu(struct ns_common *ns,
-                                        struct ns_tree *ns_tree, bool previous)
+                                        struct ns_tree_root *ns_tree, bool previous)
 {
        struct list_head *list;
 
        RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "suspicious ns_tree_adjoined_rcu() usage");
 
        if (previous)
-               list = rcu_dereference(list_bidir_prev_rcu(&ns->ns_list_node));
+               list = rcu_dereference(list_bidir_prev_rcu(&ns->ns_tree_node.ns_list_entry));
        else
-               list = rcu_dereference(list_next_rcu(&ns->ns_list_node));
-       if (list_is_head(list, &ns_tree->ns_list))
+               list = rcu_dereference(list_next_rcu(&ns->ns_tree_node.ns_list_entry));
+       if (list_is_head(list, &ns_tree->ns_list_head))
                return ERR_PTR(-ENOENT);
 
-       VFS_WARN_ON_ONCE(list_entry_rcu(list, struct ns_common, ns_list_node)->ns_type != ns_tree->type);
-
-       return list_entry_rcu(list, struct ns_common, ns_list_node);
+       return list_entry_rcu(list, struct ns_common, ns_tree_node.ns_list_entry);
 }
 
 /**
@@ -508,7 +462,7 @@ static struct ns_common *lookup_ns_owner_at(u64 ns_id, struct ns_common *owner)
        VFS_WARN_ON_ONCE(owner->ns_type != CLONE_NEWUSER);
 
        read_seqlock_excl(&ns_tree_lock);
-       node = owner->ns_owner_tree.rb_node;
+       node = owner->ns_owner_root.ns_rb.rb_node;
 
        while (node) {
                struct ns_common *ns;
@@ -638,16 +592,15 @@ static ssize_t do_listns_userns(struct klistns *kls)
        }
 
        ret = 0;
-       head = &to_ns_common(kls->user_ns)->ns_owner;
+       head = &to_ns_common(kls->user_ns)->ns_owner_root.ns_list_head;
        kls->userns_capable = ns_capable_noaudit(kls->user_ns, CAP_SYS_ADMIN);
 
        rcu_read_lock();
 
        if (!first_ns)
-               first_ns = list_entry_rcu(head->next, typeof(*ns), ns_owner_entry);
-
-       for (ns = first_ns; &ns->ns_owner_entry != head && nr_ns_ids;
-            ns = list_entry_rcu(ns->ns_owner_entry.next, typeof(*ns), ns_owner_entry)) {
+               first_ns = list_entry_rcu(head->next, typeof(*ns), ns_owner_node.ns_list_entry);
+       for (ns = first_ns; &ns->ns_owner_node.ns_list_entry != head && nr_ns_ids;
+            ns = list_entry_rcu(ns->ns_owner_node.ns_list_entry.next, typeof(*ns), ns_owner_node.ns_list_entry)) {
                struct ns_common *valid;
 
                valid = legitimize_ns(kls, ns);
@@ -682,7 +635,7 @@ static ssize_t do_listns_userns(struct klistns *kls)
 static struct ns_common *lookup_ns_id_at(u64 ns_id, int ns_type)
 {
        struct ns_common *ret = NULL;
-       struct ns_tree *ns_tree = NULL;
+       struct ns_tree_root *ns_tree = NULL;
        struct rb_node *node;
 
        if (ns_type) {
@@ -693,9 +646,9 @@ static struct ns_common *lookup_ns_id_at(u64 ns_id, int ns_type)
 
        read_seqlock_excl(&ns_tree_lock);
        if (ns_tree)
-               node = ns_tree->ns_tree.rb_node;
+               node = ns_tree->ns_rb.rb_node;
        else
-               node = ns_unified_tree.rb_node;
+               node = ns_unified_root.ns_rb.rb_node;
 
        while (node) {
                struct ns_common *ns;
@@ -725,28 +678,28 @@ static struct ns_common *lookup_ns_id_at(u64 ns_id, int ns_type)
 }
 
 static inline struct ns_common *first_ns_common(const struct list_head *head,
-                                               struct ns_tree *ns_tree)
+                                               struct ns_tree_root *ns_tree)
 {
        if (ns_tree)
-               return list_entry_rcu(head->next, struct ns_common, ns_list_node);
-       return list_entry_rcu(head->next, struct ns_common, ns_unified_list_node);
+               return list_entry_rcu(head->next, struct ns_common, ns_tree_node.ns_list_entry);
+       return list_entry_rcu(head->next, struct ns_common, ns_unified_node.ns_list_entry);
 }
 
 static inline struct ns_common *next_ns_common(struct ns_common *ns,
-                                              struct ns_tree *ns_tree)
+                                              struct ns_tree_root *ns_tree)
 {
        if (ns_tree)
-               return list_entry_rcu(ns->ns_list_node.next, struct ns_common, ns_list_node);
-       return list_entry_rcu(ns->ns_unified_list_node.next, struct ns_common, ns_unified_list_node);
+               return list_entry_rcu(ns->ns_tree_node.ns_list_entry.next, struct ns_common, ns_tree_node.ns_list_entry);
+       return list_entry_rcu(ns->ns_unified_node.ns_list_entry.next, struct ns_common, ns_unified_node.ns_list_entry);
 }
 
 static inline bool ns_common_is_head(struct ns_common *ns,
                                     const struct list_head *head,
-                                    struct ns_tree *ns_tree)
+                                    struct ns_tree_root *ns_tree)
 {
        if (ns_tree)
-               return &ns->ns_list_node == head;
-       return &ns->ns_unified_list_node == head;
+               return &ns->ns_tree_node.ns_list_entry == head;
+       return &ns->ns_unified_node.ns_list_entry == head;
 }
 
 static ssize_t do_listns(struct klistns *kls)
@@ -754,7 +707,7 @@ static ssize_t do_listns(struct klistns *kls)
        u64 __user *ns_ids = kls->uns_ids;
        size_t nr_ns_ids = kls->nr_ns_ids;
        struct ns_common *ns, *first_ns = NULL, *prev = NULL;
-       struct ns_tree *ns_tree = NULL;
+       struct ns_tree_root *ns_tree = NULL;
        const struct list_head *head;
        u32 ns_type;
        ssize_t ret;
@@ -779,9 +732,9 @@ static ssize_t do_listns(struct klistns *kls)
 
        ret = 0;
        if (ns_tree)
-               head = &ns_tree->ns_list;
+               head = &ns_tree->ns_list_head;
        else
-               head = &ns_unified_list;
+               head = &ns_unified_root.ns_list_head;
 
        rcu_read_lock();