On Fri, May 01, 2015 at 08:42:57AM +0200, Maxime Villard wrote: > Yes, ffs_oldfscompat_read() should theoretically be called before > ffs_superblock_validate(), otherwise, several fields being overwritten, > you either end up with unsanitized values or your disk gets kicked out.
I am not sure what ffs_oldfscompat_read() does, but several people (me included) have now unmountable disks that used to work fine in production before, and it is not quite clear what causes the inconsistency, see the thread at https://mail-index.netbsd.org/current-users/2015/04/30/msg027310.html I am pretty sure (but not 100%) that my root filesystem in question has been created past the newfs change cited in that thread, but it is clear that a -current newfs would not trigger the issue. Martin
