Bug #12048

Check what to do with VirtualBox on Stretch

Added by intrigeri 2016-12-20 11:16:35 . Updated 2017-07-27 07:19:50 .

Status:
Resolved
Priority:
Normal
Assignee:
intrigeri
Category:
Virtualization
Target version:
Start date:
2016-12-20
Due date:
% Done:

100%

Feature Branch:
wip/bugfix/12048-remove-virtualbox
Type of work:
Test
Blueprint:

Starter:
Affected tool:
Deliverable for:

Description

It was removed from testing due to https://bugs.debian.org/794466, so we now install it from sid (commit:acb91b93f1fee1c0865a74007efcfaa76c9f462b).


Subtasks


Related issues

Related to Tails - Feature #12104: Decide whether we can drop DKMS modules support Rejected 2017-01-01
Related to Tails - Feature #5606: Add VirtualBox host software Rejected
Related to Tails - Bug #15621: devel branch FTBFS since virtualbox 5.2.12-dfsg-2 was uploaded to sid Resolved 2018-05-26
Related to Tails - Bug #15532: devel branch FTBFS since virtualbox 5.2.8-dfsg-7 was uploaded to sid Resolved 2018-04-13

History

#1 Updated by intrigeri 2017-01-02 18:03:25

  • related to Feature #12104: Decide whether we can drop DKMS modules support added

#2 Updated by intrigeri 2017-01-02 23:35:11

It’s unlikely that VirtualBox is released in Stretch (or in any future Debian release, until Oracle fixes their policy wrt. security vulns). It’s also unclear whether it’ll be in backports (those are for packages that are in testing, and what’s in testing is meant to be for next stable).

So, what we have here is a superset of what I’ve started a discussion about on Feature #12104. While Feature #12104 is only about the kernel modules, this ticket is also about the guest additions. What’s broken with a Tails VirtualBox guest that has neither the guest additions nor the kernel modules? Help is welcome.

#3 Updated by intrigeri 2017-01-30 14:18:51

What should be tested:

  • X screen resolution:
    • what’s the one automatically configured on boot?
    • can one manually change that resolution using the GNOME Display settings?
    • does resizing the window (on the host) trigger adjustements to the screen resolution (in the guest)?
  • mouse pointer & keyboard (switching between host & guest)
  • clipboard sharing
  • shared folders

I assume none of that will work, but I’m curious how bad it is, especially the bits about screen resolution.

#4 Updated by intrigeri 2017-01-30 14:20:22

  • Feature Branch set to bugfix/12048-remove-virtualbox

#5 Updated by intrigeri 2017-01-30 14:20:43

  • Status changed from Confirmed to In Progress
  • % Done changed from 0 to 10

#6 Updated by intrigeri 2017-01-30 14:46:53

#7 Updated by intrigeri 2017-01-30 16:52:39

  • Assignee changed from intrigeri to Dr_Whax
  • Type of work changed from Research to Test

DrWhax: the ISO to test is at https://nightly.tails.boum.org/build_Tails_ISO_bugfix-12048-remove-virtualbox/lastSuccessful/archive/build-artifacts/, and what needs to be tested is listed on Bug #12048#note-3.

#8 Updated by anonym 2017-01-30 17:15:59

intrigeri wrote:
> What should be tested:
>
> * X screen resolution:
> what’s the one automatically configured on boot?
> can one manually change that resolution using the GNOME Display settings?
> does resizing the window (on the host) trigger adjustements to the screen resolution (in the guest)?
> * mouse pointer & keyboard (switching between host & guest)
> * clipboard sharing
> * shared folders
>
> I assume none of that will work, but I’m curious how bad it is, especially the bits about screen resolution.

As I have reported elsewhere (e.g. Feature #12104#note-5) all we should lose by dropping the guest modules are:

  • clipboard sharing
  • shared folders
  • host-to-guest time syncing
  • (more stuff we never cared about, like running exec (yes, it has a built-in remote shell :))

#9 Updated by intrigeri 2017-01-30 17:57:01

> As I have reported elsewhere (e.g. Feature #12104#note-5) all we should lose by dropping the guest modules are:

Note that this ticket is not only about the guest modules. It’s about the userspace utilities + kernel modules.

#10 Updated by Dr_Whax 2017-01-30 18:35:02

  • Assignee changed from Dr_Whax to intrigeri

Hello, I did some testing in Oracle’s Virtualbox (version 5.1.10 r112026) on Ubuntu 16.04, 64-bit.

These were the options I had enabled in VirtualBox:

  • 64MB of video ram
  • 3D acceleration
  • I/O APIC

I did try to boot in both normal boot and troubleshooting mode. I experienced the same on both boot modes namely: Xorg can’t detect screen with a usable configuration, after that, it reports that no screens found. it tries loading the following modules before giving up:

  • vesa
  • int10
  • fbdev
  • vbe
  • vect

Booting with: xorg-driver=modesetting didn’t make a difference.

I never reached a graphical part of the system.

#11 Updated by anonym 2017-01-30 19:16:22

I can fonfirm Dr_Whax’s results. However, I know realized that this branch removes virtualbox-guest-{utils,x11}. I thought it was only about removing the kernel modules (and the DKMS mess we’ve had to deal with regularly), so that was what my comment above was about.

So, the vesa driver is supposed to work, but it rarely does because either VirtualBox or Xorg introduces some incompatibility. So, if we want to support VirtualBox, I’m afraid we’ll need at least virtualbox-guest-x11.

#12 Updated by intrigeri 2017-01-30 19:34:53

  • Status changed from In Progress to Resolved
  • % Done changed from 10 to 100

anonym and I reached an agreement on XMPP: we’ll ship virtualbox-guest-x11 from sid as long as it’s installable on Stretch; then we’ll import the last working version in our custom APT repo. And if/when that last working version breaks (e.g. because we get a new xorg from stretch-backports and the virtualbox driver doesn’t build against it anymore, there’s no ABI compatibility between major X.Org versions, all drivers need to be rebuilt against the new one; it’s happened a few times already that whatever virtualbox backport we were shipping wasn’t compatible with the xorg from backports, etc.), then we’ll reconsider and possibly drop VirtualBox support. But perhaps then there will be some reasonable place where we can install an updated X11 driver from, or VirtualBox will have become compatible with the modesetting driver, or something. Fingers crossed.

#13 Updated by intrigeri 2017-01-30 19:36:40

Now, I wonder if it’s better to break VirtualBox support early (in 3.0) with proper warnings, early communication, pointers to alternatives, rather than breaking it later (in 3.x) without much early notice. I’ll leave this ticket closed for now, but wouldn’t mind feedback on this from our UX people. Take into account that it might be that we never have to break it later.

#14 Updated by intrigeri 2017-01-31 15:03:02

  • Feature Branch changed from bugfix/12048-remove-virtualbox to wip/bugfix/12048-remove-virtualbox

#15 Updated by intrigeri 2017-07-27 07:19:50

FWIW the vboxvideo driver made it into the staging section of Linux 4.13-rc2. But I guess that only replaces bits of the DKMS stuff, and the usespace X.Org video driver is still needed (I didn’t check though).

#16 Updated by intrigeri 2018-05-26 09:43:17

  • related to Bug #15621: devel branch FTBFS since virtualbox 5.2.12-dfsg-2 was uploaded to sid added

#17 Updated by intrigeri 2018-05-26 09:43:37

  • related to Bug #15532: devel branch FTBFS since virtualbox 5.2.8-dfsg-7 was uploaded to sid added