On Tue, Mar 08, 2016 at 03:50:19PM -0800, Max R.D. Parmer wrote:
> On Tue, Mar 8, 2016, at 15:06, Frank Steinmetzger wrote:
> > On Wed, Mar 09, 2016 at 12:02:23AM +0100, Frank Steinmetzger wrote:
> > 
> > > > If you would like to get rid of the /run/lvm/lvmetad.socket error just
> > > > start lvm with "service lvm start". I still get the error when starting
> > > > up but it still works.
> > > 
> > > I noticed that and quickly found /etc/init.d/lvmetad, but since I'm doing
> > > only the setup on this PC, I don't realler bother.
> > 
> > I would actually prefer a simple partition table within the luks
> > container.
> > I have no real need for the flexibility of LVM and it would only embiggen
> > the required initramfs and make the boot process more complex.
> > But folks on IRC told me was not possible.
> > 
> > -- 
> > Gruß | Greetings | Qapla’
> > Please do not share anything from, with or about me with any social
> > network.
> > 
> > There are things of which I do not even talk to myself.

> Frank, I can attest that it is possible to have an encrypted root
> without involving LVM. [...]
> You are doing things in a reasonable order it seems to me. First you
> create the partition table, then you luksFormat the partition which is
> to be encrypted (presumably leaving /boot unencrypted), and then you run
> pvcreate on the encrypted partition (although if you do not wish to use
> lvm, you should just run mkfs on the dm-crypt device in /dev/mapper).

Sounds to me you are speaking of LUKSing a single partition. That is not
what I aim at. I've been using unencrypted / and encrypted /home on my old
laptop just fine, but on an SSD, I prefer full-device encryption for
everything due to the nondeterministic nature of SSD wear leveling.

Running pvcreate on the encrypted partition is what spawned this thread in
the first place: it denies cooperation.

> LVM can be nice, though, as it lets you have a multitude of logical
> volumes all within a single encrypted disk partition

Hence my appended remark on whether it’s possible to use a partition table
inside a LUKS container.

> (otherwise maybe you would have everything on one partition and your
> system would fail if /var got full, or you would have several separately
> encrypted partitions which could cause other troubles).

Nah, I do have a partitioning scheme of /, /home and /data.
I rarely ever have space problems, especially with /. My main PC has 50 G
for /, and with all kinds of big software including debug information for
everything, distfiles for all installed packages and a kernel tree, only
uses 33 Gig of that.

> Could you send us the output of "stat `readlink -f /dev/mapper/lvm`" (or
> in your first example, "stat `readlink -f /dev/mapper/tp`")? I am
> interested to see that the file exists and has the correct attributes
> after you perform your `cryptsetup luksOpen` operation. The files in
> /dev/mapper are symlinks to /dev/dm-* devices, this will resolve the
> symlink and then run stat on the real underlying dm-* device.

It is a symlink and the corresponding dm file is there:
kern $ readlink -f /dev/mapper/tp
/dev/dm-1
kern $ ll /dev/dm-1
brw-rw---- 1 root disk 254, 1  9. Mär 01:01 /dev/dm-1

(dm-0 being the host PC’s /home)

I got a reply off-list that it is possible to create a partition table
within a LUKS container. Well, technically I tried this yesterday already
(parted /dev/mapper/tp). But I don't know how to access the separate
partitions within it for formatting and mounting. Using losetup?
-- 
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any social network.

Someone who works has no time to earn money.

Attachment: signature.asc
Description: Digital signature

Reply via email to