Ubuntu boot in Oracle VirtualBox ends up with a corrupted display

[Ｖ字龍(Vdragon)]

(Content copied from "Ubuntu 14.10 ISO live boot ends up with a corrupted display" #13615 VirtualBox bug ticket, by piotrjurkiewicz)

Booting up live session of Ubuntu MATE 14.10 and Ubuntu Gnome 14.10 from ISO ends up with a corrupted screen. See the screenshot:
https://launchpadlibrarian.net/186864347/ubu.png

Switching back and forth to the tty7 (ctrl+alt+F1 and then alt+F7) fixes the display.

I think that this might be a problem with resolution setting:

```
00:00:31.226379 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=0000000000000000 w=1152 h=400 bpp=0 cbLine=0x0, flags=0x1
00:00:31.226399 UIFrameBuffer::RequestResize: Screen=0, Format=0, BitsPerPixel=0, BytesPerLine=0, Size=1152x400, Sending to async-handler..
00:00:31.226457 UIFrameBufferQImage::resizeEvent: Format=0, BitsPerPixel=0, BytesPerLine=0, Size=1152x400
00:00:31.226467 UIFrameBufferQImage::resizeEvent: Resizing to FALLBACK buffer due to format is invalid..
```

After switching back and forth to the tty7 (what fixes the display) the log says:

```
00:00:31.226379 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=0000000000000000 w=1152 h=400 bpp=0 cbLine=0x0, flags=0x1
00:00:31.226399 UIFrameBuffer::RequestResize: Screen=0, Format=0, BitsPerPixel=0, BytesPerLine=0, Size=1152x400, Sending to async-handler..
00:00:31.226457 UIFrameBufferQImage::resizeEvent: Format=0, BitsPerPixel=0, BytesPerLine=0, Size=1152x400
00:00:31.226467 UIFrameBufferQImage::resizeEvent: Resizing to FALLBACK buffer due to format is invalid..
00:01:37.584616 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=0000000000000000 w=720 h=400 bpp=0 cbLine=0x0, flags=0x1
00:01:37.584634 UIFrameBuffer::RequestResize: Screen=0, Format=0, BitsPerPixel=0, BytesPerLine=0, Size=720x400, Sending to async-handler..
00:01:37.584680 UIFrameBufferQImage::resizeEvent: Format=0, BitsPerPixel=0, BytesPerLine=0, Size=720x400
00:01:37.584692 UIFrameBufferQImage::resizeEvent: Resizing to FALLBACK buffer due to format is invalid..
00:01:39.055475 Display::handleDisplayResize(): uScreenId = 0, pvVRAM=000000000a7f0000 w=1024 h=768 bpp=32 cbLine=0x1000, flags=0x1
00:01:39.055495 UIFrameBuffer::RequestResize: Screen=0, Format=843204434, BitsPerPixel=32, BytesPerLine=4096, Size=1024x768, Sending to async-handler..
00:01:39.055559 UIFrameBufferQImage::resizeEvent: Format=843204434, BitsPerPixel=32, BytesPerLine=4096, Size=1024x768
00:01:39.055568 UIFrameBufferQImage::resizeEvent: Resizing to directly use VGA device content..
```

I have tested 4.3.18 and also the testbuild 4.3.19 r96825, it happens on both of them.

You can download the ISO images of these two systems here:

 https://ubuntu-mate.r.worldssl.net/download/ubuntu-mate-14.10-desktop-amd64.iso

 http://cdimage.ubuntu.com/ubuntu-gnome/releases/14.10/release/ubuntu-gnome-14.10-desktop-amd64.iso

[Michael Thayer (michael-thayer)]

Please see comment 1[1] and 14[2] on that ticket. We believe that this is a problem in setfont, not in VirtualBox (or Ubuntu for that matter).

[1] https://www.virtualbox.org/ticket/13615#comment:1
[2] https://www.virtualbox.org/ticket/13615#comment:14

[Stéphane Charette (stephanecharette)]

Lots of people suffering from this: http://askubuntu.com/questions/541006/ubuntu-14-10-does-not-install-in-virtualbox

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[rogerdpack (rogerpack2005)]

I just got this with 16.04 and virtualbox 5.1.2

So whose bug is this exactly? It results in a poor installation experience in virtualbox, unfortunately. Never run into it before now.

[Gianfranco Costamagna (costamagnagianfranco)]

Unfortunately I'm trying to trace it down, which host/guest iso do you have? which architecture?

[Gianfranco Costamagna (costamagnagianfranco)]

according to upstream vbox, it is not a vbox bug, just it happens more often on VM

[Michael Thayer (michael-thayer)]

I think that in this day and age EGA console fonts are meaningless for 99% of the user base or more. I also think that if you remove the functionality the remaining 1% will complain very loudly. However, I suspect that many of those will have set up custom fonts to be loaded. So my suggested fix is to comment out everything in /etc/default/console-setup. Then the code will just not execute for users who have not set up custom fonts.

Someone who actually understands the details of this should check whether this makes sense before implementing it.

[Michael Thayer (michael-thayer)]

If you do actually find someone at Ubuntu who knows about this, perhaps you can ask them to comment on this thread:

https://www.virtualbox.org/pipermail/vbox-dev/2016-July/013926.html

I note that the person asking the question seemingly never bothered to create an Ubuntu bug, so it can't have been that important to them.

[Daniel Lange (dlange)]

Work around:
# use the following command to switch to VT1 from outside the VM:
VBoxManage controlvm "Ubuntu 16.04" keyboardputscancode 1D 38 3B 9D B8 CB
# inside the VM, log in and then insert # in front of all variable assignments as to @michael-thayer above:
sudo nano -w /etc/default/console-setup
# finally reboot
sudo reboot

[Jure Sah (dustwolfy)]

Also affects Ubuntu 16.04 LTS.

[S. McCandlish (smccandlish)]

Re: "I note that the person asking the question seemingly never bothered to create an Ubuntu bug, so it can't have been that important to them." That's not actually a safe assumption at all. Many people are on the clock and are not paid to file bug reports for the rest of the world. As soon they find a work-around that deals with the problem temporarily and locally, they move on. This problem has been reported world-wide in multiple circumstances, and is still happening with 16.04. It just happened to me right now (with the Xubunutu 16.04.1 ISO in a VirtualBox VM). The Right-Ctrl F1 and Right-Ctrl F7 (or "Host" F1, "Host" F7, if you've redefined the "Host" key) trick worked for me in this case, so I'm done with it and getting back to work. It is possible I (or someone in my situation) might remember to try to replicate and debug this problem further for everyone, on a VM at home, on our own time. But Canonical cannot depend on this happening. It's a "bad administrator experience" issue that should be dealt with like any other serious bug.

[Michael Thayer (michael-thayer)]

Hello,

Short answer (mobile mail). On the whole a bug with a workaround is usually (understandably) less important to people than one without. I am slightly surprised that Canonical have not looked at this since I have provided a fix which may well be acceptable as is, but of course I do not presume to try to tell them what their priorities should be.

Regards,

Michael

On 30. August 2016 02:58:14 GMT+03:00, "S. McCandlish" <email address hidden> wrote:
>Re: "I note that the person asking the question seemingly never
>bothered
>to create an Ubuntu bug, so it can't have been that important to them."
>That's not actually a safe assumption at all. Many people are on the
>clock and are not paid to file bug reports for the rest of the world.
>As
>soon they find a work-around that deals with the problem temporarily
>and
>locally, they move on. This problem has been reported world-wide in
>multiple circumstances, and is still happening with 16.04. It just
>happened to me right now (with the Xubunutu 16.04.1 ISO in a VirtualBox
>VM). The Right-Ctrl F1 and Right-Ctrl F7 (or "Host" F1, "Host" F7, if
>you've redefined the "Host" key) trick worked for me in this case, so
>I'm done with it and getting back to work. It is possible I (or
>someone
>in my situation) might remember to try to replicate and debug this
>problem further for everyone, on a VM at home, on our own time. But
>Canonical cannot depend on this happening. It's a "bad administrator
>experience" issue that should be dealt with like any other serious bug.
>
>--
>You received this bug notification because you are subscribed to the
>bug
>report.
>https://bugs.launchpad.net/bugs/1443853
>
>Title:
> Ubuntu 14.10 ISO live boot in Oracle VirtualBox ends up with a
> corrupted display
>
>Status in Virtualbox:
> Invalid
>Status in console-setup package in Ubuntu:
> Confirmed
>Status in kbd package in Ubuntu:
> Confirmed
>Status in virtualbox package in Ubuntu:
> Incomplete
>
>Bug description:
> (Content copied from "Ubuntu 14.10 ISO live boot ends up with a
> corrupted display" #13615 VirtualBox bug ticket, by piotrjurkiewicz)
>
>Booting up live session of Ubuntu MATE 14.10 and Ubuntu Gnome 14.10
>from ISO ends up with a corrupted screen. See the screenshot:
> https://launchpadlibrarian.net/186864347/ubu.png
>
> Switching back and forth to the tty7 (ctrl+alt+F1 and then alt+F7)
> fixes the display.
>
> I think that this might be a problem with resolution setting:
>
> ```
>00:00:31.226379 Display::handleDisplayResize(): uScreenId = 0,
>pvVRAM=0000000000000000 w=1152 h=400 bpp=0 cbLine=0x0, flags=0x1
>00:00:31.226399 UIFrameBuffer::RequestResize: Screen=0, Format=0,
>BitsPerPixel=0, BytesPerLine=0, Size=1152x400, Sending to
>async-handler..
>00:00:31.226457 UIFrameBufferQImage::resizeEvent: Format=0,
>BitsPerPixel=0, BytesPerLine=0, Size=1152x400
>00:00:31.226467 UIFrameBufferQImage::resizeEvent: Resizing to FALLBACK
>buffer due to format is invalid..
> ```
>
> After switching back and forth to the tty7 (what fixes the display)
> the log says:
>
> ```
>00:00:31.226379 Display::handleDisplayResize(): uScreenId = 0,
>pvVRAM=0000000000000000 w=1152 h=400 bpp=0 cbLine=0x0, flags=0x...

[Gianfranco Costamagna (costamagnagianfranco)]

setting the severity to critical, I really think we should fix this bug, if we can find somebody understanding the console-setup bits

[Gianfranco Costamagna (costamagnagianfranco)]

I'm setting it to incomplete, because with the vboxvideo driver loaded in the kernel, I can't reproduce this issue anymore
(I might even not able to reproduce it because of my new laptop with different hw configuration)

[Alexander Stohr (alexander-stohr)]

this problem again confirmed for Ubuntu 16.04.1 on i386 during the install into virtual box 5.1.6

the amount of people affected is assumed to be high and thus even if there is a work around (but not anybody gets it fast enough how to solve that but rather assume its just broken) it still would serve the Linux community much better if those issue is simply fixed and thus never seen.

[Gianfranco Costamagna (costamagnagianfranco)]

I *think* we fixed this for the upcoming ubuntu zesty, I backported the fix, can you please try it?

https://launchpad.net/~costamagnagianfranco/+archive/ubuntu/costamagnagianfranco-ppa

thanks

[Michael Thayer (michael-thayer)]

Gianfranco, now I am interested - could you point me to the change you mean?

[Gianfranco Costamagna (costamagnagianfranco)]

Hi Michael, I mean this change
https://anonscm.debian.org/cgit/pkg-virtualbox/virtualbox.git/commit/?id=17ee2def5d1b193f63fc378eda32d1bc9ead73cb

but you are right, I commented on the wrong bug report. silly me

[Norbert (nrbrtx)]

Get this issue on Ubuntu 12.04.5 amd64 host using VirtualBox 5.1.10 with Ubuntu 16.04.1 x86 guest.
The screen shows rainbow of pixels, not the desktop. Turning off 3D acceleration does not solve the problem.
Pressing right Ctrl+F1, then Ctrl+F7 solved the problem.

Dear Ubuntu developers, please fix this problem or contact VirtualBox developers. VirtualBox-way is the easiest to get first impression of Ubuntu. So it is expected that Ubuntu will work on VirtualBox without special actions.

[Peter Eijlander (peter-eijlander)]

It affects Ubuntu-MATE 16.04.1, both 32 and 64bit but it seems to be more Ubuntu and less MATE for in Mint MATE 18.1 (=16.04.1) this problem does not occur. At least I have never seen this in Mint. I need right-ctrl-F1, then right-ctrl F7 (a black screen with blinking cursor appears) and then right-ctrl-F1 again to finally see the desktop allowing to either run the "Try" or "Install" choices.

[Peter Eijlander (peter-eijlander)]

Interesting to notice: In Linux Mint MATE the "Live-ISO" understands the correct sequence Right-crtl-F7 for the desktop and Right-ctrl-F1 for the first terminal right away, where Ubuntu-MATE does not do so. Any Ctrl-F other than F1 gives a blank "terminal" with just a cursor and no login prompt at all. After installing Ubuntu-MATE this problem is gone and that too follows the correct sequence and shows the login on the available terminals.

[Michael Thayer (michael-thayer)]

This issue should be fixed in versions of Ubuntu which provide the VirtualBox graphics driver. However the person who re-opened the bug discovered another issue: at least the Lubuntu live ISO seems to load the VirtualBox driver rather late in the boot process, which can result in X.Org loading with the VESA driver (because it does not find the VirtualBox one) at the same time as the VirtualBox driver, and both trying to access the graphics device at the same time. Not sure if this deserves a new bug.

[Gianfranco Costamagna (costamagnagianfranco)]

Since now we use wayland, lets wait for an eventual new bug report :

[Michael Thayer (michael-thayer)]

I would just point out that the problem with Lubuntu live which was reported and which I reproduced would force a fall-back from Wayland to X.Org.

[Michael Thayer (michael-thayer)]

Actually Lubuntu probably doesn't support Wayland at all.

[Michael Thayer (michael-thayer)]

Ah, it is the old problem with lightdm not waiting for the udev notification that the driver is loaded.

[Ubuntu QA Website (ubuntuqa)]

This bug has been reported on the Ubuntu ISO testing tracker.

A list of all reports related to this bug can be found here:
http://iso.qa.ubuntu.com/qatracker/reports/bugs/1443853