When copying a namespace we won't have added the new copy into the namespace rbtree until after the copy succeeded. Calling free_mnt_ns() will try to remove the copy from the rbtree which is invalid. Simply free the namespace skeleton directly.
Fixes: 1901c92497bd ("fs: keep an index of current mount namespaces") Cc: stable@vger.kernel.org # v6.11+ Reported-by: Brad Spengler spender@grsecurity.net Tested-by: Brad Spengler spender@grsecurity.net Suggested-by: Brad Spengler spender@grsecurity.net Signed-off-by: Christian Brauner brauner@kernel.org --- In vfs.fixes unless I hear objections. --- fs/namespace.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/fs/namespace.c b/fs/namespace.c index 93c377816d75..d26f5e6d2ca3 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -3944,7 +3944,9 @@ struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, new = copy_tree(old, old->mnt.mnt_root, copy_flags); if (IS_ERR(new)) { namespace_unlock(); - free_mnt_ns(new_ns); + ns_free_inum(&new_ns->ns); + dec_mnt_namespaces(new_ns->ucounts); + mnt_ns_release(new_ns); return ERR_CAST(new); } if (user_ns != ns->user_ns) {