Re: In-kernel units for block numbers, etc ...

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Sat, 28 Nov 2015 19:58:19 +0000 (UTC)
    From:        mlelstv%serpens.de@localhost (Michael van Elst)
    Message-ID:  <n3d10r$l9e$1%serpens.de@localhost>

  | The correct thing is therefore to drop the dk_getdisklabel call.
  | Can you verify that this just solves the problem?

Yes, that works too.

I wondered about that, but wasn't sure if the disklabel read might be
needed for the dkwedge_discover() that immediately follows.

That I can't easily test (and I doubt almost anyone else can either) as
until we get cgd crypto validation that's based upon finding a valid GPT
(rather than, or instead of, a disklabel, or valid ffs) GPT labelled cgds
are not very useful, so I have never tried one.

Nor do I use the "convert labels to wedges" kernel option (whose name
I don't even remember... the thing that would make wd0a appear as dk0
instead).   So, if the label read is needed for that, I wouldn't observe
any ill effects.

But for a "regular" cgd, with a disklabel (I install one, even when the
entire cgd is to be one file system) you're right, that (ancient)
dk_getdisklabel() seems to be nothing more than a problem.

kre

ps: incidentally, I even though (dev_t)0 is wd0a, I doubt that the major()
part of it ever gets used, just minor(dev), which is why the PR observed
reading from cgd0 when cgd1 was configured.