documentation of GEOM data structures needed

[symbolics at gmx.com]

On Tue, Nov 12, 2013 at 11:16:58AM +0800, Erich Dollansky wrote:
> Hi,
>
[...]
> > 
> > In which case run `gpart backup da1' (assuming da1 is the name of one
> > of the intact discs). You can store the output text to a file and
> > then edit it to fit. Then use `gpart restore da0' to apply it to the
> > new disc.
> > 
> > First of all, take backups of your existing MBRs, just in case:
> > 
> I copied already the identified data structures into files.
> 
> > # gpart backup da1 > ~/da1.gpart
> 
> This is what I would like to avoid. I would like to understand how it
> works and do it manually.

The code is in sys/geom and sbin/geom. The structure definitions can be
found in the header files in sys/geom. The backup command itself can be
found in sbin/geom/class/part/geom_part.c.

> > 
> > If you like you can try to fit the first five partitions first and
> > then worry about the last one later. If da1 is an intact disc then
> > running the following piped command should reset all but the last
> > partiton on da0:
> > 
> > # gpart backup da1 | sed \$d | gpart restore da0
> > 
> > Try fscking and mounting the first five partitions and see how you get
> > on.  Hopefully they'll just work.
> > 
> The hopefully is the point why I would like to do it by hand. I do not
> think that there is a huge secret behind. It is just that I did not
> find the documentation. Or doesn't it exist?

It depends on what you mean by documentation I suppose. Quite a lot of
the GEOM API is documented to varying levels (something I intend to
improve as my time allows). You can look at the sys/sys/diskmbr.h header
file for some useful information though. The MBR format is documented on
Wikipedia. 

> > > > > > disc, you could reconstruct things that way. What does `gpart
> > > > > > show' look like at the moment?
> > > > > 
> > > > > It does not come that far
> > > > > 
> > > > > gpart list da0
> > > > > gpart: No such geom: da0.
> > > > > 
> > > > > is all I get.
> > > > > 
> > > > > My luck is that I have three disks which are the type but
> > > > > manufactured with some months between. But their sizes differ a
> > > > > bit. I think that I should be able to recover much by just
> > > > > comparing the entries.
> > > > > 
> > > > 
> > > > You can try looking at diskinfo -v da0 to see the numbers.
> > > >
> > >          512             # sectorsize
> > >         500107860480    # mediasize in bytes (466G)
> > >         976773165       # mediasize in sectors
> > >         0               # stripesize
> > >         0               # stripeoffset
> > >         60801           # Cylinders according to firmware.
> > >         255             # Heads according to firmware.
> > >         63              # Sectors according to firmware.
> > >         0000000000006121        # Disk ident.
> > > 
> > > One other disk shows the same data while the third one shows this:
> > > 
> > >         512             # sectorsize
> > >         500107862016    # mediasize in bytes (466G)
> > >         976773168       # mediasize in sectors
> > >         4096            # stripesize
> > >         0               # stripeoffset
> > >         15504336        # Cylinders according to firmware.
> > >         1               # Heads according to firmware.
> > >         63              # Sectors according to firmware.
> > >         TF0504YS02ZPBP  # Disk ident.
> > > 
> > 
> > These look really different. Are you using other GEOM classes? What
> > else is different about these discs?
> 
> The media size is already different. The drives are all Hitachi's with
> 4k sectors. They should be all from the same factory but obviously
> there is a difference.
> 

Ok, good luck.

--sym