Add support for bind-mount on case-insensitive filesystems

If we are using a case-insensitive filesystem the bind-mount operation
might fail when `/proc/self/mountinfo` is checked.

In a case-insensitive filesystem, if we ask to mount a certain
directory, e.g. '/CI_fs/foo', the kernel might add its entry in
`mountinfo` as '/CI_fs/FOO'. This happens because the kernel populates
`mountinfo` with whatever case combination first appeared in the dcache.

With this patch we open the requested path and look at its
`/proc/self/fd`, using readlink(), to get the path case combination that
the kernel is also expected to be using.

Signed-off-by: Ludovico de Nittis <ludovico.denittis@collabora.com>

4 files changed, 59 insertions, 10 deletions

diff --git a/bind-mount.c b/bind-mount.c
index 4bd187d..e692762 100644
--- a/bind-mount.c
+++ b/bind-mount.c
@@ -385,6 +385,10 @@ bind_mount (int           proc_fd,
   unsigned long current_flags, new_flags;
   cleanup_mount_tab MountTab mount_tab = NULL;
   cleanup_free char *resolved_dest = NULL;
+  cleanup_free char *dest_proc = NULL;
+  cleanup_free char *oldroot_dest_proc = NULL;
+  cleanup_free char *kernel_case_combination = NULL;
+  cleanup_fd int dest_fd = -1;
   int i;
 
   if (src)
@@ -399,14 +403,34 @@ bind_mount (int           proc_fd,
   if (resolved_dest == NULL)
     return 2;
 
-  mount_tab = parse_mountinfo (proc_fd, resolved_dest);
+  dest_fd = open (resolved_dest, O_PATH | O_CLOEXEC);
+  if (dest_fd < 0)
+    return 2;
+
+  /* If we are in a case-insensitive filesystem, mountinfo might contain a
+   * different case combination of the path we requested to mount.
+   * This is due to the fact that the kernel, as of the beginning of 2021,
+   * populates mountinfo with whatever case combination first appeared in the
+   * dcache; kernel developers plan to change this in future so that it
+   * reflects the on-disk encoding instead.
+   * To avoid throwing an error when this happens, we use readlink() result
+   * instead of the provided @root_mount, so that we can compare the mountinfo
+   * entries with the same case combination that the kernel is expected to
+   * use. */
+  dest_proc = xasprintf ("/proc/self/fd/%d", dest_fd);
+  oldroot_dest_proc = get_oldroot_path (dest_proc);
+  kernel_case_combination = readlink_malloc (oldroot_dest_proc);
+  if (kernel_case_combination == NULL)
+    die_with_error ("Can't read the link in %s", oldroot_dest_proc);
+
+  mount_tab = parse_mountinfo (proc_fd, kernel_case_combination);
   if (mount_tab[0].mountpoint == NULL)
     {
       errno = EINVAL;
       return 2; /* No mountpoint at dest */
     }
 
-  assert (path_equal (mount_tab[0].mountpoint, resolved_dest));
+  assert (path_equal (mount_tab[0].mountpoint, kernel_case_combination));
   current_flags = mount_tab[0].options;
   new_flags = current_flags | (devices ? 0 : MS_NODEV) | MS_NOSUID | (readonly ? MS_RDONLY : 0);
   if (new_flags != current_flags &&
diff --git a/bubblewrap.c b/bubblewrap.c
index efca03c..2172904 100644
--- a/bubblewrap.c
+++ b/bubblewrap.c
@@ -861,14 +861,6 @@ get_newroot_path (const char *path)
   return strconcat ("/newroot/", path);
 }
 
-static char *
-get_oldroot_path (const char *path)
-{
-  while (*path == '/')
-    path++;
-  return strconcat ("/oldroot/", path);
-}
-
 static void
 write_uid_gid_map (uid_t sandbox_uid,
                    uid_t parent_uid,
diff --git a/utils.c b/utils.c
index a99a865..e45ae56 100644
--- a/utils.c
+++ b/utils.c
@@ -764,6 +764,37 @@ read_pid_from_socket (int socket)
   die ("No pid returned on socket");
 }
 
+/* Sets errno on error (== NULL),
+ * Always ensures terminating zero */
+char *
+readlink_malloc (const char *pathname)
+{
+  size_t size = 50;
+  ssize_t n;
+  cleanup_free char *value = NULL;
+
+  do
+    {
+      size *= 2;
+      value = xrealloc (value, size);
+      n = readlink (pathname, value, size - 1);
+      if (n < 0)
+        return NULL;
+    }
+  while (size - 2 < n);
+
+  value[n] = 0;
+  return steal_pointer (&value);
+}
+
+char *
+get_oldroot_path (const char *path)
+{
+  while (*path == '/')
+    path++;
+  return strconcat ("/oldroot/", path);
+}
+
 int
 raw_clone (unsigned long flags,
            void         *child_stack)
diff --git a/utils.h b/utils.h
index 8c4db61..c90c8d7 100644
--- a/utils.h
+++ b/utils.h
@@ -112,6 +112,8 @@ int   mkdir_with_parents (const char *pathname,
 void create_pid_socketpair (int sockets[2]);
 void send_pid_on_socket (int socket);
 int  read_pid_from_socket (int socket);
+char *get_oldroot_path (const char *path);
+char *readlink_malloc (const char *pathname);
 
 /* syscall wrappers */
 int   raw_clone (unsigned long flags,