Very Computer

I can't believe i'm the only person who wants to do this!  Am I in the
wrong group?

        Allen

Quote:> I did a "newfs /dev/rfd0a floppy" on my SUN workstation to create a fs
> on my 3.5 HDD floppy.  I mounted this fs successfully, and copied
> to/from it ok.  When I take the floppy home to use on my FreeBSD and
> try to mount, I get the following error:
>    mount failed : bogus super-block.
> This is no verbatim, but close enough.  This is the advice I got from
> a co-worker:
> "Your problem with the newfs command is probably from an item in the
> superblock called the "magic number".  The superblocks have undergone
> so many format changes from vendor to vendor that not all superblocks
> can be read by a particular OS.  To prevent corruption from an
> "incorrectly" formatted superblock from another vendor, most vendors
> come up with a "unique" magic number for their superblock.  You can
> think of the "magic number" as a serial number.  If the serial number
> doesn't match, then we can't read this format.
> This is only a guess, but your simptoms match my experience.  I was
> not suprised you would have problems doing this.  You may be able to
> write a superblock conversion routine though.  The header file for
> describing a superblock is contained in the /usr/include
> subdirectories.  If the required data is compatible you could simply
> translate the Sun superblock format to a BSD superblock format and
> overwrite the Sun superblock.  Then the disk could be mountable on a
> BSD system.  (Some times all this translation requires is to change
> the magic number.)"
> The question is... how do I mount a SUN newfs'd file system on a BSD
> box (and vice versa)?
>            Thanks,
>            Allen

Quote:>   mount failed : bogus super-block.

>This is no verbatim, but close enough.  This is the advice I got from
>a co-worker:

Quote:>The question is... how do I mount a SUN newfs'd file system on a BSD
>box (and vice versa)?

But the byte order is different, so datum written as more than single
byte units (ie: inode numbers and file system type identifiers, etc.)
don't have the same values.

As an exercise to prove I'm not lying, write a program to fill out a
struct containing one of every type of object to a file as a unit,
and then ftp (or tar) the file over to the other machine and read it
out.

something like:

struct foo {
        long    l;
        char    c;
        short   s;
        int     i;      /* happen to konw it's 32 bits*/

Quote:};

Quote:};

main()
{
        int fd;

#ifdef READIT
        fd = open( "my_foo, 0, 0);

        read( fd, readme, sizeof(readme));
        printf( "long    %x\n", readme.l);
        printf( "char    %x\n", readme.c);
        printf( "short   %x\n", readme.s);
        printf( "int     %x\n", readme.i);
#else
        fd = open( "my_foo", 2, 0666);

        write( fd, writeme, sizeof(writeme));
#endif

Quote:}

Basically, you'd have to put byte order adjustemnets into the FS source
on one machine and compile it up on the other and link it into the kernel
to mount one systems disk on the other system.  This is non-trivial,
mostly because the underlying kernel structure that the FS on each machine
depends on is very different, and it simply won't compile wtihout a lot
of work and a lot of knowledge.

You should either use tar or another archiver, or use the DOS FS on both
machines instead (the Sun pcfs type has the byte order adjustments in it
already).

                                        Terry Lambert

---
Any opinions in this posting are my own and not those of my present
or previous employers.