bubblewrap/0001-Add-bind-fd-and-ro-bind-fd-to-let-you-bind-a-O_PATH-.patch

From 52d5f7c3ba5c8e10b3a992304fd28cd4d18caeeb Mon Sep 17 00:00:00 2001
From: Alexander Larsson <alexl@redhat.com>
Date: Tue, 18 Jun 2024 10:20:36 +0200
Subject: [PATCH] Add --bind-fd and --ro-bind-fd to let you bind a O_PATH fd.

This is useful for example if you for some reason don't have the real
path. It is also a way to make bind-mounts race-free (i.e. to have the
mount actually be the thing you wanted to be mounted, avoiding issues
where some other process replaces the target in parallel with the bwrap
launch.

Unfortunately due to some technical details we can't actually directly
mount the dirfd, as they come from different user namespace which is not
permitted, but at least we can delay resolving the fd to a path as much as
possible, and then validate after mount that we actually mounted the right
thing.

Signed-off-by: Alexander Larsson <alexl@redhat.com>
(cherry picked from commit a253257cd298892da43e15201d83f9a02c9b58b5)
[kalev: Backport to 0.4.x]
Signed-off-by: Kalev Lember <klember@redhat.com>
---
 bubblewrap.c      | 50 +++++++++++++++++++++++++++++++++++++++++++++++
 tests/test-run.sh |  7 ++++++-
 2 files changed, 56 insertions(+), 1 deletion(-)

diff --git a/bubblewrap.c b/bubblewrap.c
index b3d52bc..38b3646 100644
--- a/bubblewrap.c
+++ b/bubblewrap.c
@@ -250,6 +250,8 @@ usage (int ecode, FILE *out)
            "    --dev-bind-try SRC DEST      Equal to --dev-bind but ignores non-existent SRC\n"
            "    --ro-bind SRC DEST           Bind mount the host path SRC readonly on DEST\n"
            "    --ro-bind-try SRC DEST       Equal to --ro-bind but ignores non-existent SRC\n"
+           "    --bind-fd FD DEST            Bind open directory or path fd on DEST\n"
+           "    --ro-bind-fd FD DEST         Bind open directory or path fd read-only on DEST\n"
            "    --remount-ro DEST            Remount DEST as readonly; does not recursively remount\n"
            "    --exec-label LABEL           Exec label for the sandbox\n"
            "    --file-label LABEL           File label for temporary sandbox content\n"
@@ -1111,6 +1113,30 @@ setup_newroot (bool unshare_pid,
                          (op->type == SETUP_RO_BIND_MOUNT ? BIND_READONLY : 0) |
                          (op->type == SETUP_DEV_BIND_MOUNT ? BIND_DEVICES : 0),
                          source, dest);
+
+          if (op->fd >= 0)
+            {
+              struct stat fd_st, mount_st;
+
+              /* When using bind-fd, there is a race condition between resolving the fd as a magic symlink
+               * and mounting it, where someone could replace what is at the symlink target. Ideally
+               * we would not even resolve the symlink and directly bind-mount from the fd, but unfortunately
+               * we can't do that, because its not permitted to bind mount a fd from another user namespace.
+               * So, we resolve, mount and then compare fstat+stat to detect the race. */
+
+              if (fstat(op->fd, &fd_st) != 0)
+                die_with_error("Can't stat fd %d", op->fd);
+              if (lstat(dest, &mount_st) != 0)
+                die_with_error("Can't stat mount at %s", dest);
+
+              if (fd_st.st_ino != mount_st.st_ino ||
+                  fd_st.st_dev != mount_st.st_dev)
+                die_with_error("Race condition binding dirfd");
+
+              close(op->fd);
+              op->fd = -1;
+            }
+
           break;
 
         case SETUP_REMOUNT_RO_NO_RECURSIVE:
@@ -1648,6 +1674,30 @@ parse_args_recurse (int          *argcp,
           if (strcmp(arg, "--dev-bind-try") == 0)
             op->flags = ALLOW_NOTEXIST;
 
+          argv += 2;
+          argc -= 2;
+        }
+      else if (strcmp (arg, "--bind-fd") == 0 ||
+               strcmp (arg, "--ro-bind-fd") == 0)
+        {
+          int src_fd;
+          char *endptr;
+
+          if (argc < 3)
+            die ("--bind-fd takes two arguments");
+
+          src_fd = strtol (argv[1], &endptr, 10);
+          if (argv[1][0] == 0 || endptr[0] != 0 || src_fd < 0)
+            die ("Invalid fd: %s", argv[1]);
+
+          if (strcmp(arg, "--ro-bind-fd") == 0)
+            op = setup_op_new (SETUP_RO_BIND_MOUNT);
+          else
+            op = setup_op_new (SETUP_BIND_MOUNT);
+          op->source = xasprintf ("/proc/self/fd/%d", src_fd);
+          op->fd = src_fd;
+          op->dest = argv[2];
+
           argv += 2;
           argc -= 2;
         }
diff --git a/tests/test-run.sh b/tests/test-run.sh
index 702c480..ce1eaf6 100755
--- a/tests/test-run.sh
+++ b/tests/test-run.sh
@@ -80,7 +80,7 @@ if ! $RUN true; then
     skip Seems like bwrap is not working at all. Maybe setuid is not working
 fi
 
-echo "1..49"
+echo "1..50"
 
 # Test help
 ${BWRAP} --help > help.txt
@@ -382,5 +382,10 @@ else
     echo "ok - Test --pidns"
 fi
 
+echo "foobar" > file-data
+$RUN --proc /proc --dev /dev --bind / / --bind-fd 100 /tmp cat /tmp/file-data 100< . > stdout
+assert_file_has_content stdout foobar
+
+echo "ok - bind-fd"
 
 echo "ok - End of test"
-- 
2.46.0