device-mapper blocks udev, but needed for cryptsetup

f.kater · Guru Joined: 23 May 2002 Posts: 342 Location: Berlin

Hi,

just in case nobody noticed: The recent udev update to 146-r1 is blocked by the latest version of device-mapper in portage.
However, I cannot uninstall device-mapper since it is needed by cryptsetup.

I've found this bug https://bugs.gentoo.org/show_bug.cgi?id=284222, however it does NOT cover my case where cryptsetup needs the device-mapper.

Thanks for hints and pointers...
Felix

VoidMage · Watchman Joined: 14 Oct 2006 Posts: 6196

Read that bug a bit more, focusing on comment 5,
ignoring mostly invalid comments 7,8,9.
Also, :evil:

SEARCH THE BOARD :evil:

anonybosh · Guru Joined: 20 Nov 2005 Posts: 324

So basically the solution is:

honeymak · Guru Joined: 30 Dec 2002 Posts: 537

if emerge -1 can be counted as a solution
why would guys need any update?

:lol:

_________________
hackers - make sth real
academics - read sth said to be real

FrancoisVal · Posted: Wed Nov 11, 2009 10:24 am Post subject:

In my case, this upgrade seems to be a nightmare. I am using LVM and the root partition is on a lvm volume. I use genkernel to build the initramfs but the lvm volumes aren't found. So, I can't boot my computer if I rebuild a new initramfs image.
Does anybody know a solution to this problem ?
_________________
François Valenduc

f.kater · Guru Joined: 23 May 2002 Posts: 342 Location: Berlin

derdestiller · n00b Joined: 08 Jan 2009 Posts: 17

Updating udev screwed up my entire configuration (my /dev) was empty.

First i thought compiling CONFIG_SYSFS_DEPRECATED_V2 not set in kernel will fix the problem. It did not.

I then tried to downgrading udev (somehow updateing to udev-146 didn't pull in baselayout-2,which is ok, because i want to have my sys tools stable)

I emerge udev-124 again and enabled CONFIG_DEV_TARBALL="yes". Didn't solve the problem.

Finally i copied an old backup to /dev, had to recreate /dev/null and /dev/console and chown root:tty some files.

Xserver didn't work, so i downgraded to my old nohal configuraton (i tried reemerging evdev)

So now it works again,but maybe someone can give me a hint why my system screwed up?

honeymak · Guru Joined: 30 Dec 2002 Posts: 537

so any response/help from genkernel team?
_________________
hackers - make sth real
academics - read sth said to be real

rufnut · Apprentice Joined: 16 May 2005 Posts: 247

derdestiller · n00b Joined: 08 Jan 2009 Posts: 17

Moriah · Posted: Tue Nov 17, 2009 8:33 pm Post subject:

So I was busy with a hardware upgrade on a windoze box this weekend and only started to do my weekly updates today. :oops:

I saw that lvm and devicemapper and udev were in a full fledged bar room brawl, so I did not do that update yet, deciding to search the forums *BEFORE* I messed everything up.

I have 18 boxes here, some x86, some x-64, and they all run lvm on the root filesystem. One of them runs lvm on raid-1 on the root filesystem.

What am I to do with this latest update mess? :?:

Thanks!

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Mike Hunt · Watchman Joined: 19 Jul 2009 Posts: 5287

Generally, it's as simple as unmerging device-mapper, then doing your update as usual, and lvm2 will be pulled in correctly.

Lvm2 contains device-mapper now. That's the reason for this block.

Only if you use baselayout2 then you need to add lvm to the boot runlevel. You will see a postinst elog about that.

Moriah · Posted: Tue Nov 17, 2009 8:56 pm Post subject:

Are you sure there is no danger in doing this when your root filesystem uses LVM? I used to make my own initdisk images, but lately I have been using genkernel to do this for me. Are there any special considerations I should keep in mind?

I do have 2 boxes running x86 32 bit gentoo that could be "sacrificial" if need be, as they are not presently doeing anything critical, but I would prefer not to scrog them, but better them than any other systems.
_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Mike Hunt · Watchman Joined: 19 Jul 2009 Posts: 5287

Good point. I've never actually put the / partition on the logical volume so I am uncertain about that.
Let me look around.

Mike Hunt · Watchman Joined: 19 Jul 2009 Posts: 5287

EDIT @ 5:00 pm: Found nothing specific.

I suspect that you would need to rebuild your initial ramdisk to contain the new lvm2.

If that works, please add 10 geek points to my score.

Moriah · Posted: Wed Nov 18, 2009 12:54 am Post subject:

As I suspected: great tribulation with this. :evil:

I have a server that is built from Pappy's Seeds, but it has a flakey ethernet on the NIC that connects to my LAN. It works fine with the LiveCD, so it is either a kernel or driver issue. I have pretty well eliminated the driver as a contender, so I need to play with kernels here. For now, it is expendable, software-wise. This box will eventually by my NIDS and packet logger, but it has a 5-slot SATA hot-swap chassis, so I can dd the boot drive for instant recovery from trouble, and hot-swap from the front of the case without opening the box.

I will make an image of the boot drive, then try this stunt. It is also an AMD quad-core running 2.3 GHz (or something close to that), so the builds will go quickly.

If it flops, I just pull the clobbered boot drive out and stuff in the copy.

Hopefully, I wil have a chance to try this tonight. If all goes smoothly (probably after several tries...) then I will upgrade the other boxes.

Only trouble is, the drives are all 1.5 GB, so it takes a while to clone a copy.

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Moriah · Posted: Wed Nov 18, 2009 12:57 am Post subject:

!!! Idea !!! :idea:

Could I boot the LiveCD, then chroot and do the update from there? Would that be safer than doing the update using the running normal system? :?:

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

honeymak · Guru Joined: 30 Dec 2002 Posts: 537

i would not advice any update should be done
IF and only IF u r still out of the mess.......
:oops:

seems lvm and mdadm and genkernel are NOT catching up....

u will run into lvm2 compilation problem....becoz "dm is included in lvm2".......compilation deadlock......and NO fix yet
_________________
hackers - make sth real
academics - read sth said to be real

Moriah · Posted: Wed Nov 18, 2009 3:47 am Post subject:

OK, updates are on hold for now.

I guess I could lock the present versions of these components and then do the updates, but not tonight. I just found out earlier this evening that I have to be at a client's tomorow morning, and its a long drive, so I have to get up at 5:00 AM tomorow morning.

Maybe tomorow afternoon...

PS I just realized that I need an intel x86 platform to accurately test this update, as I did not get complaints out of x-64 boxes. The box I was going to use for a test is an amd x-64 box. I have no hot-swap chassis on any intel 32-bit boxes.

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Moriah · Posted: Wed Nov 18, 2009 4:27 am Post subject:

More grief! :evil:

I did emerge --update system on a gentoo system running on a virtual machine where I could use vmware snapshot to cover my butt.

In the config files that needed updating, I noticed:

FrancoisVal · Posted: Wed Nov 18, 2009 5:51 pm Post subject:

As I previously said, this update turn's out to be a nightmare if you use a LVM volume for the root partition and if you don't enable CONFIG_SYSFS_DEPRECATED_V2 in the kernel configuration (in general setup). The solution is to use a newer version of genkernel which is not yet stable (see bugs 255196 and 225249).
_________________
François Valenduc

Moriah · Posted: Thu Nov 19, 2009 12:15 am Post subject:

Since I run LVM on the root filesystem of every linux box on my network, this would have definately been a total disaster if I had just "let 'er rip", instead of being careful and paranoid.

I run LVM on all my linux filesystems unless there is a very good reason not to. This is because I use the LVM snapshot facility to freeze a view of the filesystem when it is being backed up by my backup server every night.

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Raffi · l33t Joined: 17 Mar 2003 Posts: 731 Location: Moscow, Id.

I run lvm as root on many of my boxes. I have just finished fighting this issue on one of them. The update itself seems to have done the right thing followed by much pain getting the system to boot again. The things that killed me were related to my initrd. I will lay out what I did, though I suspect not all of it was necessary.

Changed my initrd to mount sysfs. My old version only mounted sysfs if it was trying to do a restore from suspend.

Changed the root= in my kernel line in grub from /dev/vgname/lvname to /dev/mapper/vgname-lvname. Looks like the new mapper stuff does not add the symbolic links in /dev for mapper/... to vgname/... I did the same for /etc/fstab. These changes were probably all that was really needed.

Re-created the initrd with the latest lvm on it.

VoidMage · Watchman Joined: 14 Oct 2006 Posts: 6196

Well, in the end it seems that the real problem lies in genkernel
being not up to date with these changes.

Solour · Tux's lil' helper Joined: 01 Jul 2006 Posts: 107

Hi,

an emerge -u world gives me