diff --git a/fs/binfmt_misc.c b/fs/binfmt_misc.c
index cf5ed5cd4102..deacc105119d 100644
--- a/fs/binfmt_misc.c
+++ b/fs/binfmt_misc.c
@@ -40,9 +40,6 @@ enum {
 	VERBOSE_STATUS = 1 /* make it zero to save 400 bytes kernel memory */
 };
 
-static LIST_HEAD(entries);
-static int enabled = 1;
-
 enum {Enabled, Magic};
 #define MISC_FMT_PRESERVE_ARGV0 (1UL << 31)
 #define MISC_FMT_OPEN_BINARY (1UL << 30)
@@ -63,7 +60,6 @@ typedef struct {
 	refcount_t users;		/* sync removal with load_misc_binary() */
 } Node;
 
-static DEFINE_RWLOCK(entries_lock);
 static struct file_system_type bm_fs_type;
 
 /*
@@ -91,13 +87,14 @@ static struct file_system_type bm_fs_type;
  *
  * Return: binary type list entry on success, NULL on failure
  */
-static Node *search_binfmt_handler(struct linux_binprm *bprm)
+static Node *search_binfmt_handler(struct binfmt_misc *misc,
+				   struct linux_binprm *bprm)
 {
 	char *p = strrchr(bprm->interp, '.');
 	Node *e;
 
 	/* Walk all the registered handlers. */
-	list_for_each_entry(e, &entries, list) {
+	list_for_each_entry(e, &misc->entries, list) {
 		char *s;
 		int j;
 
@@ -140,15 +137,16 @@ static Node *search_binfmt_handler(struct linux_binprm *bprm)
  *
  * Return: binary type list entry on success, NULL on failure
  */
-static Node *get_binfmt_handler(struct linux_binprm *bprm)
+static Node *get_binfmt_handler(struct binfmt_misc *misc,
+				struct linux_binprm *bprm)
 {
 	Node *e;
 
-	read_lock(&entries_lock);
-	e = search_binfmt_handler(bprm);
+	read_lock(&misc->entries_lock);
+	e = search_binfmt_handler(misc, bprm);
 	if (e)
 		refcount_inc(&e->users);
-	read_unlock(&entries_lock);
+	read_unlock(&misc->entries_lock);
 	return e;
 }
 
@@ -169,6 +167,35 @@ static void put_binfmt_handler(Node *e)
 	}
 }
 
+/**
+ * load_binfmt_misc - load the binfmt_misc of the caller's user namespace
+ *
+ * To be called in load_misc_binary() to load the relevant struct binfmt_misc.
+ * If a user namespace doesn't have its own binfmt_misc mount it can make use
+ * of its ancestor's binfmt_misc handlers. This mimicks the behavior of
+ * pre-namespaced binfmt_misc where all registered binfmt_misc handlers where
+ * available to all user and user namespaces on the system.
+ *
+ * Return: the binfmt_misc instance of the caller's user namespace
+ */
+static struct binfmt_misc *load_binfmt_misc(void)
+{
+	const struct user_namespace *user_ns;
+	struct binfmt_misc *misc;
+
+	user_ns = current_user_ns();
+	while (user_ns) {
+		/* Pairs with smp_store_release() in bm_fill_super(). */
+		misc = smp_load_acquire(&user_ns->binfmt_misc);
+		if (misc)
+			return misc;
+
+		user_ns = user_ns->parent;
+	}
+
+	return &init_binfmt_misc;
+}
+
 /*
  * the loader itself
  */
@@ -176,13 +203,14 @@ static int load_misc_binary(struct linux_binprm *bprm)
 {
 	Node *fmt;
 	struct file *interp_file = NULL;
-	int retval;
+	int retval = -ENOEXEC;
+	struct binfmt_misc *misc;
 
-	retval = -ENOEXEC;
-	if (!enabled)
+	misc = load_binfmt_misc();
+	if (!misc->enabled)
 		return retval;
 
-	fmt = get_binfmt_handler(bprm);
+	fmt = get_binfmt_handler(misc, bprm);
 	if (!fmt)
 		return retval;
 
@@ -240,9 +268,9 @@ ret:
 	/*
 	 * If we actually put the node here all concurrent calls to
 	 * load_misc_binary() will have finished. We also know
-	 * that for the refcount to be zero ->evict_inode() must have removed
-	 * the node to be deleted from the list. All that is left for us is to
-	 * close and free.
+	 * that for the refcount to be zero someone must have concurently
+	 * removed the binary type handler from the list and it's our job to
+	 * free it.
 	 */
 	put_binfmt_handler(fmt);
 
@@ -334,7 +362,7 @@ static Node *create_entry(const char __user *buffer, size_t count)
 
 	err = -ENOMEM;
 	memsize = sizeof(Node) + count + 8;
-	e = kmalloc(memsize, GFP_KERNEL);
+	e = kmalloc(memsize, GFP_KERNEL_ACCOUNT);
 	if (!e)
 		goto out;
 
@@ -446,7 +474,7 @@ static Node *create_entry(const char __user *buffer, size_t count)
 
 			if (e->mask) {
 				int i;
-				char *masked = kmalloc(e->size, GFP_KERNEL);
+				char *masked = kmalloc(e->size, GFP_KERNEL_ACCOUNT);
 
 				print_hex_dump_bytes(
 					KBUILD_MODNAME ": register:  mask[decoded]: ",
@@ -599,6 +627,22 @@ static struct inode *bm_get_inode(struct super_block *sb, int mode)
 	return inode;
 }
 
+/**
+ * i_binfmt_misc - retrieve struct binfmt_misc from a binfmt_misc inode
+ * @inode: inode of the relevant binfmt_misc instance
+ *
+ * This helper retrieves struct binfmt_misc from a binfmt_misc inode. This can
+ * be done without any memory barriers because we are guaranteed that
+ * user_ns->binfmt_misc is fully initialized. It was fully initialized when the
+ * binfmt_misc mount was first created.
+ *
+ * Return: struct binfmt_misc of the relevant binfmt_misc instance
+ */
+static struct binfmt_misc *i_binfmt_misc(struct inode *inode)
+{
+	return inode->i_sb->s_user_ns->binfmt_misc;
+}
+
 /**
  * bm_evict_inode - cleanup data associated with @inode
  * @inode: inode to which the data is attached
@@ -619,10 +663,13 @@ static void bm_evict_inode(struct inode *inode)
 	clear_inode(inode);
 
 	if (e) {
-		write_lock(&entries_lock);
+		struct binfmt_misc *misc;
+
+		misc = i_binfmt_misc(inode);
+		write_lock(&misc->entries_lock);
 		if (!list_empty(&e->list))
 			list_del_init(&e->list);
-		write_unlock(&entries_lock);
+		write_unlock(&misc->entries_lock);
 		put_binfmt_handler(e);
 	}
 }
@@ -677,11 +724,11 @@ static void unlink_binfmt_dentry(struct dentry *dentry)
  * to use writes to files in order to delete binary type handlers. But it has
  * worked for so long that it's not a pressing issue.
  */
-static void remove_binfmt_handler(Node *e)
+static void remove_binfmt_handler(struct binfmt_misc *misc, Node *e)
 {
-	write_lock(&entries_lock);
+	write_lock(&misc->entries_lock);
 	list_del_init(&e->list);
-	write_unlock(&entries_lock);
+	write_unlock(&misc->entries_lock);
 	unlink_binfmt_dentry(e->dentry);
 }
 
@@ -737,7 +784,7 @@ static ssize_t bm_entry_write(struct file *file, const char __user *buffer,
 		 * actually remove the entry from the list.
 		 */
 		if (!list_empty(&e->list))
-			remove_binfmt_handler(e);
+			remove_binfmt_handler(i_binfmt_misc(inode), e);
 
 		inode_unlock(inode);
 		break;
@@ -763,6 +810,7 @@ static ssize_t bm_register_write(struct file *file, const char __user *buffer,
 	struct inode *inode;
 	struct super_block *sb = file_inode(file)->i_sb;
 	struct dentry *root = sb->s_root, *dentry;
+	struct binfmt_misc *misc;
 	int err = 0;
 	struct file *f = NULL;
 
@@ -772,7 +820,18 @@ static ssize_t bm_register_write(struct file *file, const char __user *buffer,
 		return PTR_ERR(e);
 
 	if (e->flags & MISC_FMT_OPEN_FILE) {
+		const struct cred *old_cred;
+
+		/*
+		 * Now that we support unprivileged binfmt_misc mounts make
+		 * sure we use the credentials that the register @file was
+		 * opened with to also open the interpreter. Before that this
+		 * didn't matter much as only a privileged process could open
+		 * the register file.
+		 */
+		old_cred = override_creds(file->f_cred);
 		f = open_exec(e->interpreter);
+		revert_creds(old_cred);
 		if (IS_ERR(f)) {
 			pr_notice("register: failed to install interpreter file %s\n",
 				 e->interpreter);
@@ -804,9 +863,10 @@ static ssize_t bm_register_write(struct file *file, const char __user *buffer,
 	inode->i_fop = &bm_entry_operations;
 
 	d_instantiate(dentry, inode);
-	write_lock(&entries_lock);
-	list_add(&e->list, &entries);
-	write_unlock(&entries_lock);
+	misc = i_binfmt_misc(inode);
+	write_lock(&misc->entries_lock);
+	list_add(&e->list, &misc->entries);
+	write_unlock(&misc->entries_lock);
 
 	err = 0;
 out2:
@@ -833,26 +893,31 @@ static const struct file_operations bm_register_operations = {
 static ssize_t
 bm_status_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
 {
-	char *s = enabled ? "enabled\n" : "disabled\n";
+	struct binfmt_misc *misc;
+	char *s;
 
+	misc = i_binfmt_misc(file_inode(file));
+	s = misc->enabled ? "enabled\n" : "disabled\n";
 	return simple_read_from_buffer(buf, nbytes, ppos, s, strlen(s));
 }
 
 static ssize_t bm_status_write(struct file *file, const char __user *buffer,
 		size_t count, loff_t *ppos)
 {
+	struct binfmt_misc *misc;
 	int res = parse_command(buffer, count);
 	Node *e, *next;
 	struct inode *inode;
 
+	misc = i_binfmt_misc(file_inode(file));
 	switch (res) {
 	case 1:
 		/* Disable all handlers. */
-		enabled = 0;
+		misc->enabled = false;
 		break;
 	case 2:
 		/* Enable all handlers. */
-		enabled = 1;
+		misc->enabled = true;
 		break;
 	case 3:
 		/* Delete all handlers. */
@@ -868,8 +933,8 @@ static ssize_t bm_status_write(struct file *file, const char __user *buffer,
 		 * read-only. So we only need to take the write lock when we
 		 * actually remove the entry from the list.
 		 */
-		list_for_each_entry_safe(e, next, &entries, list)
-			remove_binfmt_handler(e);
+		list_for_each_entry_safe(e, next, &misc->entries, list)
+			remove_binfmt_handler(misc, e);
 
 		inode_unlock(inode);
 		break;
@@ -888,32 +953,100 @@ static const struct file_operations bm_status_operations = {
 
 /* Superblock handling */
 
+static void bm_put_super(struct super_block *sb)
+{
+	struct user_namespace *user_ns = sb->s_fs_info;
+
+	sb->s_fs_info = NULL;
+	put_user_ns(user_ns);
+}
+
 static const struct super_operations s_ops = {
 	.statfs		= simple_statfs,
 	.evict_inode	= bm_evict_inode,
+	.put_super	= bm_put_super,
 };
 
 static int bm_fill_super(struct super_block *sb, struct fs_context *fc)
 {
 	int err;
+	struct user_namespace *user_ns = sb->s_user_ns;
+	struct binfmt_misc *misc;
 	static const struct tree_descr bm_files[] = {
 		[2] = {"status", &bm_status_operations, S_IWUSR|S_IRUGO},
 		[3] = {"register", &bm_register_operations, S_IWUSR},
 		/* last one */ {""}
 	};
 
+	if (WARN_ON(user_ns != current_user_ns()))
+		return -EINVAL;
+
+	/*
+	 * Lazily allocate a new binfmt_misc instance for this namespace, i.e.
+	 * do it here during the first mount of binfmt_misc. We don't need to
+	 * waste memory for every user namespace allocation. It's likely much
+	 * more common to not mount a separate binfmt_misc instance than it is
+	 * to mount one.
+	 *
+	 * While multiple superblocks can exist they are keyed by userns in
+	 * s_fs_info for binfmt_misc. Hence, the vfs guarantees that
+	 * bm_fill_super() is called exactly once whenever a binfmt_misc
+	 * superblock for a userns is created. This in turn lets us conclude
+	 * that when a binfmt_misc superblock is created for the first time for
+	 * a userns there's no one racing us. Therefore we don't need any
+	 * barriers when we dereference binfmt_misc.
+	 */
+	misc = user_ns->binfmt_misc;
+	if (!misc) {
+		/*
+		 * If it turns out that most user namespaces actually want to
+		 * register their own binary type handler and therefore all
+		 * create their own separate binfm_misc mounts we should
+		 * consider turning this into a kmem cache.
+		 */
+		misc = kzalloc(sizeof(struct binfmt_misc), GFP_KERNEL);
+		if (!misc)
+			return -ENOMEM;
+
+		INIT_LIST_HEAD(&misc->entries);
+		rwlock_init(&misc->entries_lock);
+
+		/* Pairs with smp_load_acquire() in load_binfmt_misc(). */
+		smp_store_release(&user_ns->binfmt_misc, misc);
+	}
+
+	/*
+	 * When the binfmt_misc superblock for this userns is shutdown
+	 * ->enabled might have been set to false and we don't reinitialize
+	 * ->enabled again in put_super() as someone might already be mounting
+	 * binfmt_misc again. It also would be pointless since by the time
+	 * ->put_super() is called we know that the binary type list for this
+	 * bintfmt_misc mount is empty making load_misc_binary() return
+	 * -ENOEXEC independent of whether ->enabled is true. Instead, if
+	 * someone mounts binfmt_misc for the first time or again we simply
+	 * reset ->enabled to true.
+	 */
+	misc->enabled = true;
+
 	err = simple_fill_super(sb, BINFMTFS_MAGIC, bm_files);
 	if (!err)
 		sb->s_op = &s_ops;
 	return err;
 }
 
+static void bm_free(struct fs_context *fc)
+{
+	if (fc->s_fs_info)
+		put_user_ns(fc->s_fs_info);
+}
+
 static int bm_get_tree(struct fs_context *fc)
 {
-	return get_tree_single(fc, bm_fill_super);
+	return get_tree_keyed(fc, bm_fill_super, get_user_ns(fc->user_ns));
 }
 
 static const struct fs_context_operations bm_context_ops = {
+	.free		= bm_free,
 	.get_tree	= bm_get_tree,
 };
 
@@ -932,6 +1065,7 @@ static struct file_system_type bm_fs_type = {
 	.owner		= THIS_MODULE,
 	.name		= "binfmt_misc",
 	.init_fs_context = bm_init_fs_context,
+	.fs_flags	= FS_USERNS_MOUNT,
 	.kill_sb	= kill_litter_super,
 };
 MODULE_ALIAS_FS("binfmt_misc");
diff --git a/include/linux/binfmts.h b/include/linux/binfmts.h
index 8d51f69f9f5e..70f97f685bff 100644
--- a/include/linux/binfmts.h
+++ b/include/linux/binfmts.h
@@ -90,6 +90,16 @@ struct linux_binfmt {
 #endif
 } __randomize_layout;
 
+#if IS_ENABLED(CONFIG_BINFMT_MISC)
+struct binfmt_misc {
+	struct list_head entries;
+	rwlock_t entries_lock;
+	bool enabled;
+} __randomize_layout;
+
+extern struct binfmt_misc init_binfmt_misc;
+#endif
+
 extern void __register_binfmt(struct linux_binfmt *fmt, int insert);
 
 /* Registration of default binfmt handlers */
diff --git a/include/linux/user_namespace.h b/include/linux/user_namespace.h
index 45f09bec02c4..6030a8235617 100644
--- a/include/linux/user_namespace.h
+++ b/include/linux/user_namespace.h
@@ -65,6 +65,10 @@ enum rlimit_type {
 	UCOUNT_RLIMIT_COUNTS,
 };
 
+#if IS_ENABLED(CONFIG_BINFMT_MISC)
+struct binfmt_misc;
+#endif
+
 struct user_namespace {
 	struct uid_gid_map	uid_map;
 	struct uid_gid_map	gid_map;
@@ -102,6 +106,10 @@ struct user_namespace {
 	struct ucounts		*ucounts;
 	long ucount_max[UCOUNT_COUNTS];
 	long rlimit_max[UCOUNT_RLIMIT_COUNTS];
+
+#if IS_ENABLED(CONFIG_BINFMT_MISC)
+	struct binfmt_misc *binfmt_misc;
+#endif
 } __randomize_layout;
 
 struct ucounts {
diff --git a/kernel/user.c b/kernel/user.c
index d667debeafd6..03cedc366dc9 100644
--- a/kernel/user.c
+++ b/kernel/user.c
@@ -18,8 +18,18 @@
 #include <linux/interrupt.h>
 #include <linux/export.h>
 #include <linux/user_namespace.h>
+#include <linux/binfmts.h>
 #include <linux/proc_ns.h>
 
+#if IS_ENABLED(CONFIG_BINFMT_MISC)
+struct binfmt_misc init_binfmt_misc = {
+	.entries = LIST_HEAD_INIT(init_binfmt_misc.entries),
+	.enabled = true,
+	.entries_lock = __RW_LOCK_UNLOCKED(init_binfmt_misc.entries_lock),
+};
+EXPORT_SYMBOL_GPL(init_binfmt_misc);
+#endif
+
 /*
  * userns count is 1 for root user, 1 for init_uts_ns,
  * and 1 for... ?
@@ -67,6 +77,9 @@ struct user_namespace init_user_ns = {
 	.keyring_name_list = LIST_HEAD_INIT(init_user_ns.keyring_name_list),
 	.keyring_sem = __RWSEM_INITIALIZER(init_user_ns.keyring_sem),
 #endif
+#if IS_ENABLED(CONFIG_BINFMT_MISC)
+	.binfmt_misc = &init_binfmt_misc,
+#endif
 };
 EXPORT_SYMBOL_GPL(init_user_ns);
 
diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c
index 1d8e47bed3f1..d52a894ecf57 100644
--- a/kernel/user_namespace.c
+++ b/kernel/user_namespace.c
@@ -213,6 +213,9 @@ static void free_user_ns(struct work_struct *work)
 			kfree(ns->projid_map.forward);
 			kfree(ns->projid_map.reverse);
 		}
+#if IS_ENABLED(CONFIG_BINFMT_MISC)
+		kfree(ns->binfmt_misc);
+#endif
 		retire_userns_sysctls(ns);
 		key_free_user_ns(ns);
 		ns_free_inum(&ns->ns);