Re: patch: MFSv3 support (libsa) for boot2 (i386)

[Antoine LECA <antoine.leca.1%gmail.com@localhost>]

Hi,

Sorry to enter the discussion without former presentations. I am
following Evgeniy's effort, and was following this thread albeit I know
next to nothing to NetBSD.

David Laight wrote:
> The actual 'pc' boot sequence is:
> Stages 1-3 (and maybe 4) are common to all OS.
<...>
> 3) The pbr code has to determine where it was loaded from, it could:
>    a) reread sector zero and look for a partition of the relevant type
>    b) have the sector number previously written into the sector data.

     c)
>    The netbsd bootselect mbr passes the sector number in a register
>    (non-standard) and the pbr code scans the partitions looking for
>    one that starts in the correct place.

     d) have the partition table entry (from the MBR) been passed in
     well-known registers, and use that information "blindly".

     e) (actually an extension of d) have a faked partition table entry
     (e.g. from the "hybrid MBR/GPT") been passed in well-known
     registers, and use that information "blindly".

...speaking in the view of "all OS" here, of course.

I am not sure whether NetBSD currently uses a or c (or a combination of
both); b is the method used with FAT and generally Microsoft file
systems, among many others; d is the method pushed by the Phoenix et
alii Standard Committee (Bios boot specifications); e is also supposed
to be a Standard (T13 EDD-4 annex 4) and is apparently induced into
NetBSD by the new (August 17th) sys/arch/i386/stand/mbr/gptmbr*


Antoine