nsfs: iterate through mount namespaces

It is already possible to list mounts in other mount namespaces and to
retrieve namespace file descriptors without having to go through procfs
by deriving them from pidfds.

Augment these abilities by adding the ability to retrieve information
about a mount namespace via NS_MNT_GET_INFO. This will return the mount
namespace id and the number of mounts currently in the mount namespace.
The number of mounts can be used to size the buffer that needs to be
used for listmount() and is in general useful without having to actually
iterate through all the mounts. The structure is extensible.

And add the ability to iterate through all mount namespaces over which
the caller holds privilege returning the file descriptor for the next or
previous mount namespace.

To retrieve a mount namespace the caller must be privileged wrt to it's
owning user namespace. This means that PID 1 on the host can list all
mounts in all mount namespaces or that a container can list all mounts
of its nested containers.

Optionally pass a structure for NS_MNT_GET_INFO with
NS_MNT_GET_{PREV,NEXT} to retrieve information about the mount namespace
in one go. Both ioctls can be implemented for other namespace types
easily.

Together with recent api additions this means one can iterate through
all mounts in all mount namespaces without ever touching procfs.

Link: https://lore.kernel.org/r/20240719-work-mount-namespace-v1-5-834113cab0d2@kernel.org
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>

3 files changed, 144 insertions, 6 deletions

diff --git a/fs/mount.h b/fs/mount.h
index ad4b1ddebb54..c1db0c709c6a 100644
--- a/fs/mount.h
+++ b/fs/mount.h
@@ -155,3 +155,16 @@ static inline void move_from_ns(struct mount *mnt, struct list_head *dt_list)
 
 extern void mnt_cursor_del(struct mnt_namespace *ns, struct mount *cursor);
 bool has_locked_children(struct mount *mnt, struct dentry *dentry);
+struct mnt_namespace *__lookup_next_mnt_ns(struct mnt_namespace *mnt_ns, bool previous);
+static inline struct mnt_namespace *lookup_next_mnt_ns(struct mnt_namespace *mntns)
+{
+	return __lookup_next_mnt_ns(mntns, false);
+}
+static inline struct mnt_namespace *lookup_prev_mnt_ns(struct mnt_namespace *mntns)
+{
+	return __lookup_next_mnt_ns(mntns, true);
+}
+static inline struct mnt_namespace *to_mnt_ns(struct ns_common *ns)
+{
+	return container_of(ns, struct mnt_namespace, ns);
+}
diff --git a/fs/namespace.c b/fs/namespace.c
index 1a0b9e3089a2..3dabe49f7376 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -2060,14 +2060,41 @@ static bool is_mnt_ns_file(struct dentry *dentry)
 	       dentry->d_fsdata == &mntns_operations;
 }
 
-static struct mnt_namespace *to_mnt_ns(struct ns_common *ns)
+struct ns_common *from_mnt_ns(struct mnt_namespace *mnt)
 {
-	return container_of(ns, struct mnt_namespace, ns);
+	return &mnt->ns;
 }
 
-struct ns_common *from_mnt_ns(struct mnt_namespace *mnt)
+struct mnt_namespace *__lookup_next_mnt_ns(struct mnt_namespace *mntns, bool previous)
 {
-	return &mnt->ns;
+	guard(read_lock)(&mnt_ns_tree_lock);
+	for (;;) {
+		struct rb_node *node;
+
+		if (previous)
+			node = rb_prev(&mntns->mnt_ns_tree_node);
+		else
+			node = rb_next(&mntns->mnt_ns_tree_node);
+		if (!node)
+			return ERR_PTR(-ENOENT);
+
+		mntns = node_to_mnt_ns(node);
+		node = &mntns->mnt_ns_tree_node;
+
+		if (!ns_capable_noaudit(mntns->user_ns, CAP_SYS_ADMIN))
+			continue;
+
+		/*
+		 * Holding mnt_ns_tree_lock prevents the mount namespace from
+		 * being freed but it may well be on it's deathbed. We want an
+		 * active reference, not just a passive one here as we're
+		 * persisting the mount namespace.
+		 */
+		if (!refcount_inc_not_zero(&mntns->ns.count))
+			continue;
+
+		return mntns;
+	}
 }
 
 static bool mnt_ns_loop(struct dentry *dentry)
diff --git a/fs/nsfs.c b/fs/nsfs.c
index 97c37a9631e5..67ee176b8824 100644
--- a/fs/nsfs.c
+++ b/fs/nsfs.c
@@ -12,6 +12,7 @@
 #include <linux/user_namespace.h>
 #include <linux/nsfs.h>
 #include <linux/uaccess.h>
+#include <linux/mnt_namespace.h>
 
 #include "mount.h"
 #include "internal.h"
@@ -128,6 +129,30 @@ int open_related_ns(struct ns_common *ns,
 }
 EXPORT_SYMBOL_GPL(open_related_ns);
 
+static int copy_ns_info_to_user(const struct mnt_namespace *mnt_ns,
+				struct mnt_ns_info __user *uinfo, size_t usize,
+				struct mnt_ns_info *kinfo)
+{
+	/*
+	 * If userspace and the kernel have the same struct size it can just
+	 * be copied. If userspace provides an older struct, only the bits that
+	 * userspace knows about will be copied. If userspace provides a new
+	 * struct, only the bits that the kernel knows aobut will be copied and
+	 * the size value will be set to the size the kernel knows about.
+	 */
+	kinfo->size		= min(usize, sizeof(*kinfo));
+	kinfo->mnt_ns_id	= mnt_ns->seq;
+	kinfo->nr_mounts	= READ_ONCE(mnt_ns->nr_mounts);
+	/* Subtract the root mount of the mount namespace. */
+	if (kinfo->nr_mounts)
+		kinfo->nr_mounts--;
+
+	if (copy_to_user(uinfo, kinfo, kinfo->size))
+		return -EFAULT;
+
+	return 0;
+}
+
 static long ns_ioctl(struct file *filp, unsigned int ioctl,
 			unsigned long arg)
 {
@@ -135,6 +160,8 @@ static long ns_ioctl(struct file *filp, unsigned int ioctl,
 	struct pid_namespace *pid_ns;
 	struct task_struct *tsk;
 	struct ns_common *ns = get_proc_ns(file_inode(filp));
+	struct mnt_namespace *mnt_ns;
+	bool previous = false;
 	uid_t __user *argp;
 	uid_t uid;
 	int ret;
@@ -156,7 +183,6 @@ static long ns_ioctl(struct file *filp, unsigned int ioctl,
 		uid = from_kuid_munged(current_user_ns(), user_ns->owner);
 		return put_user(uid, argp);
 	case NS_GET_MNTNS_ID: {
-		struct mnt_namespace *mnt_ns;
 		__u64 __user *idp;
 		__u64 id;
 
@@ -211,7 +237,79 @@ static long ns_ioctl(struct file *filp, unsigned int ioctl,
 
 		if (!ret)
 			ret = -ESRCH;
-		break;
+		return ret;
+	}
+	}
+
+	/* extensible ioctls */
+	switch (_IOC_NR(ioctl)) {
+	case _IOC_NR(NS_MNT_GET_INFO): {
+		struct mnt_ns_info kinfo = {};
+		struct mnt_ns_info __user *uinfo = (struct mnt_ns_info __user *)arg;
+		size_t usize = _IOC_SIZE(ioctl);
+
+		if (ns->ops->type != CLONE_NEWNS)
+			return -EINVAL;
+
+		if (!uinfo)
+			return -EINVAL;
+
+		if (usize < MNT_NS_INFO_SIZE_VER0)
+			return -EINVAL;
+
+		return copy_ns_info_to_user(to_mnt_ns(ns), uinfo, usize, &kinfo);
+	}
+	case _IOC_NR(NS_MNT_GET_PREV):
+		previous = true;
+		fallthrough;
+	case _IOC_NR(NS_MNT_GET_NEXT): {
+		struct mnt_ns_info kinfo = {};
+		struct mnt_ns_info __user *uinfo = (struct mnt_ns_info __user *)arg;
+		struct path path __free(path_put) = {};
+		struct file *f __free(fput) = NULL;
+		size_t usize = _IOC_SIZE(ioctl);
+
+		if (ns->ops->type != CLONE_NEWNS)
+			return -EINVAL;
+
+		if (usize < MNT_NS_INFO_SIZE_VER0)
+			return -EINVAL;
+
+		if (previous)
+			mnt_ns = lookup_prev_mnt_ns(to_mnt_ns(ns));
+		else
+			mnt_ns = lookup_next_mnt_ns(to_mnt_ns(ns));
+		if (IS_ERR(mnt_ns))
+			return PTR_ERR(mnt_ns);
+
+		ns = to_ns_common(mnt_ns);
+		/* Transfer ownership of @mnt_ns reference to @path. */
+		ret = path_from_stashed(&ns->stashed, nsfs_mnt, ns, &path);
+		if (ret)
+			return ret;
+
+		CLASS(get_unused_fd, fd)(O_CLOEXEC);
+		if (fd < 0)
+			return fd;
+
+		f = dentry_open(&path, O_RDONLY, current_cred());
+		if (IS_ERR(f))
+			return PTR_ERR(f);
+
+		if (uinfo) {
+			/*
+			 * If @uinfo is passed return all information about the
+			 * mount namespace as well.
+			 */
+			ret = copy_ns_info_to_user(to_mnt_ns(ns), uinfo, usize, &kinfo);
+			if (ret)
+				return ret;
+		}
+
+		/* Transfer reference of @f to caller's fdtable. */
+		fd_install(fd, no_free_ptr(f));
+		/* File descriptor is live so hand it off to the caller. */
+		return take_fd(fd);
 	}
 	default:
 		ret = -ENOTTY;