In ext_mountfs of sys/gnu/fs/ext2fs/ext2_vfsops.c it's checked
errorneously whether mounter wants a read-only mount. Therefore
this function will always act as in case of a writable mount which
yields an EROFS if the media is actually read-only.

This might also lead to unintended fs modification despite the "-o ro"
option; I haven't made experiments of such type though, it's just a
hypothetical problem at this point.

This bug is present at least in RELENG6 and HEAD branches.

Fix: Apply the follwing patch:




Csaba--Sf3BVGoCl2Lwm3R65S2ZhnEwryVMOQQDGbO1fgiVp6dpYKDA
Content-Type: text/plain; name="file.diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename="file.diff"

--- ext2_vfsops.c.orig	Fri Nov  3 01:31:42 2006
+++ ext2_vfsops.c	Fri Nov  3 01:34:10 2006
@@ -601,7 +601,7 @@ ext2_mountfs(devvp, mp, td)
 	int error;
 	int ronly;
 
-	ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
+	ronly = (mp->mnt_flag & MNT_RDONLY) ? : 1 : 0;
 	/* XXX: use VOP_ACESS to check FS perms */
 	DROP_GIANT();
 	g_topology_lock();
How-To-Repeat: - Create a sparse file in some of your filesystems.
- Format the sparse file as an ext2 filesystem by mke2fs(8).
- Remount the host filesystem read-only.
- Create an md node backed by the ext2 file using mdconfig(8).
- Try to mount the resulting md node with ext2 type. You will
  get EROFS either you use "-o ro" or not.