On Mar 19, 11:43, Warner Losh wrote:
>: VFS: Mounted root (ext2 filesystem) readonly.
>: EXT2-fs error (device 02:00): ext2_find_entry: bad entry in directory #12:
>: rec_len is smaller than minimal - offset=40, inode=0, rec_len=0, name_len=0
>
>Same problem here (except mine was due to the fs code thinking the
>directory spanned multiple blocks). I've discovered so far that
>either the same data is transferred twice, or that the second trafer
>doesn't happen, even though the kernel thinks that it happened. I've
>not had the time to go in and figure out the exact reasons for this.
>I do know that the kernel will checksum the blocks to the same thing
>(well, I added that code). Are you using a 720k floppy, or a 1.4M
>job? I'm using a 1.4M that I created from scratch on my FreeBSD
>system, but that mounts and seems to work correctly from the Linux
>1.1.59 system that I have laying around on another partition of my
>disk.
I'm using a 1.44MB floppy disk drive, and I tried both a self-made
(on Linux/x86) ext2fs 1.44MB file system, and the rhello-720k.image
(just cat'ed to a formatted 1.44MB floppy). Same problem on both. From
what I can find out using debugfs it looks like the 'directory #' (which is
different for my two floppies) in both cases points to the /dev directory on
the floppy file systems -- given that I interpret the debugfs 'ls' output
correctly of course.. I'm not sure what Linux tries to read first from a
root mounted file system, for all I know the /dev directory will always be
the first one? If it really finds the /dev directory then the problem seems
to be limited to reading the entries in the directory, not total corruption.
But this is pure speculation, I should look more into this. I have limited
time available at the moment, unfortunately.
Tor
|