Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 15 Nov 2007 08:46:47 +0100
From:      Pawel Jakub Dawidek <pjd@FreeBSD.org>
To:        Gordon Tetlow <gordon@tetlows.org>
Cc:        freebsd-geom@freebsd.org, Ivan Voras <ivoras@freebsd.org>, Eric Anderson <anderson@FreeBSD.org>
Subject:   Re: Disk mounting in recent Linuxes
Message-ID:  <20071115074647.GB80222@garage.freebsd.pl>
In-Reply-To: <61434C70-4246-47E8-9C30-5956738B611B@tetlows.org>
References:  <fel7fn$k6d$1@sea.gmane.org> <47117184.3030309@freebsd.org> <61434C70-4246-47E8-9C30-5956738B611B@tetlows.org>

next in thread | previous in thread | raw e-mail | index | archive | help

--Yylu36WmvOXNoKYn
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Nov 14, 2007 at 09:27:38PM -0800, Gordon Tetlow wrote:
>=20
> On Oct 13, 2007, at 6:31 PM, Eric Anderson wrote:
>=20
> >Ivan Voras wrote:
> >>Hi,
> >>I've installed a Linux (openSUSE) on a laptop and this is what I =20
> >>got by
> >>default in fstab:
> >>/dev/disk/by-id/scsi-SATA_Hitachi_HTS5412_HP0400BEG1922A-part2 /
> >>           ext3       acl,user_xattr        1 1
> >>/dev/disk/by-id/scsi-SATA_Hitachi_HTS5412_HP0400BEG1922A-part4 /data
> >>           ext3       acl,user_xattr        1 2
> >>/dev/disk/by-id/scsi-SATA_Hitachi_HTS5412_HP0400BEG1922A-part3 swap
> >>           swap       defaults              0 0
> >>A similar option (to use a device by id instead of location) also =20
> >>exists
> >>for network cards.
> >>(This is just a "FYI" post, I'm not complaining :) ).
> >
> >
> >I was actually wondering if we should start labeling our filesystems =20
> >at newfs time in the installer for this type of setup.  There are a =20
> >handful of potential 'gotchas' though.
>=20
> The big gotcha with GEOM_VOL_FFS (I don't know if pjd@ ever addressed =20
> this when he took it over) was that GEOM was undetermined in how it =20
> handled labels that overlapped. My original thought was to use the =20
> uuid field that I allocated in the FFS superblock (AFAIK, that's still =
=20
> unused). Whatever the uuid of the root device is would dictate the =20
> uuid for the system. Any volumes that didn't have the root uuid would =20
> be separated by namespace. ie /dev/vol/<uuid>/root if you took the =20
> root disk from another system and mounted on your system.
>=20
> Of course, with the uuid moved to the filesystem, it actually gets =20
> easier in some sense (how do you figure out which labeled filesystem =20
> is your root filesystem if you have 2 labelled filesystems with the =20
> same uuid), although the loader will probably have to send it as hint =20
> to the kernel.

The problem is that when you have two colliding providers (with the same
name) you are already in very bad position, because no matter what you
do you may end-up with system not booted properly. I decided to create
first provider and warn about all others with colliding names without
creating them at all. It fits into GEOM concept a bit better IMHO (you
never know when the next provider with colliding name comes).

All in all, the best you can do is to print a warning which demands
manual intervention.

--=20
Pawel Jakub Dawidek                       http://www.wheel.pl
pjd@FreeBSD.org                           http://www.FreeBSD.org
FreeBSD committer                         Am I Evil? Yes, I Am!

--Yylu36WmvOXNoKYn
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)

iD8DBQFHO/lnForvXbEpPzQRAl9NAKDKyBNDfdAdUXiIl4By0FZUyklCegCfTmjB
4j/r9zsojwpOtjIfIV/vUj8=
=JXUC
-----END PGP SIGNATURE-----

--Yylu36WmvOXNoKYn--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20071115074647.GB80222>