Error building the graphics driver module in a RHEL 7.6 guest -> fixed after (not in) 5.2.22

[JoergK]

Hello,

I'm using VBox 5.2.20 r125813 (Qt5.6.1) on Ubuntu 16.04 Host with RHEL 7.5 guest. After updating the RHEL guest to 7.6 I tried to reinstall the guest additions but building the graphics driver module failed.

So I tried with the test build 5.2.21 revision 126213 but building the graphics driver module failed, too.

To reproduce this issue just try to install the guest additions in a RHEL 7.6 guest system.

There is a topic to this issue at: ​https://forums.virtualbox.org/viewtopic.php?f=3&t=90103

I'm going to attach related log files to this case once it is open. If I could do anything else to track this down, please let me know. I'm not experienced in tracking down and debugging possible bugs in virtualbox, yet.

[JoergK]

Log file from VBox version 5.2.20

[JoergK]

vboxadd-setup-log from version 5.2.20

[JoergK]

Log file from test build 5.2.21 126213

[JoergK]

vboxadd-setup-log from version 5.2.21 126213

[nixuser]

The failure leaves the system in a state whereby Xorg crashes on startup so you can't even use the 1024x768 (or whatever it is) default GUI until you clean it up by running the /opt/VBoxGuestAdditions-5.2.20/uninstall.sh script via a virtual console.

[HRZ d. Universität Bielefeld]

Can reproduce the problem under CentOS 7.5 as Host and RHEL 7.6 Guest with 5.2.21-126141 Guest additions. System is usable problem equal to JoergK, but not as worse as nixuser.

[wil-go]

Same issue, but system can't even run 1024x768 after uninstalling. Stuck with small window for a week now.

[Michael Thayer]

I will take a look at this as soon as OL 7.6 or CentOS 7.6 is available. Alternatively, patches welcome.

[nixuser]

It may be quicker to download an eval copy of RHEL and try it out. Not sure when 7.6 will push through to the others.

[Michael Thayer]

Actually I have already downloaded the OL 7.6 pre-release, but I need to find time to do this. If you are in a hurry, a patch may speed things up.

[rconde01]

michael - I'm going to look at doing a patch, but I think it will fix Redhat 7.6, and break others...how is that normally handled?

[Michael Thayer]

See the RHEL_7x conditionals in the code. Usually there is already a conditional on a particular kernel version, and you just need to add "|| defined(RHEL_76)" (after defining that in vbox_drv.h of course).

[rconde01]

Patch to fix build

[rconde01]

Ok - I added a first cut at a patch. FYI I'm new at working in the vbox codebase. I

• Checked out the svn repo
• Built VBox (after installing many dependencies) through
  • ./configure --disable-hardening --disable-docs
  • source ./env.sh
  • kmk all VBOX_ONLY_ADDITIONS=1
• cd ./out/linux.amd64/release/bin/additions/src/vboxvideo
• make and fix each issue until it builds

To create the patch I found the original location of the files in the repo and copied over the edited versions. I was a bit surprised when comparing because there were additional changes...not sure if there's a preprocessing step before copying to ./out/XXXX/vboxvideo. So I made the files look like the head except for my changes. Then I created an svn patch and that's what I attached. Let me know if you need any changes to those steps.

[Michael Thayer]

Thank you, I will take a look at it. Your sequence looks good to me. And you are right, there is a pre-processing stage. I will get rid of it when I get around to it. It seemed like a good idea at the time, to make what we supply look as similar as possible to what Hans put into the Linux kernel code.

[HRZ d. Universität Bielefeld]

Just for those who are interested: Problem still persists in 5.2.22.

[Michael Thayer]

@rconde01 Looks good. Could you please post a statement that you release your patch under the MIT licence, and tell me if (and how) you would like to be credited in the change log/subversion history?

[rconde01]

I Robert Conde release my patch under the MIT license. You may credit me as "Robert Conde".

[Michael Thayer]

https://www.virtualbox.org/changeset/75402/vbox

Many thanks. Test build<1> revision 126544 of the Guest Additions (5.2 branch build) should be available in an hour or so and contain the change. I see no reason not to continue using this test build as long as there is no new release of VirtualBox.

1. wiki:Testbuilds

[nixuser]

Test build VirtualBox-5.2.23-126544 not working for me, attached vboxadd-setup.log file.

[HRZ d. Universität Bielefeld]

Successfully installed Guest extensions 5.2.23-126544 without problems and it works (Host CentOS 7.5, Guest RHEL 7.6).

THX!

[nixuser]

@HRZunibi which kernel are you running, is it 3.10.0-957 ??

If so I don't understand how you are getting it to compile. I've just cleaned out yum and run latest updates and it still fails here.

[nixuser]

I was able to get the latest test build VirtualBox-5.2.23-126581 to work.

[JoergK]

Hi, Version 5.2.23-126798 fixed the issue on my site. I'm running kernel 3.10.0-957 in the guest os.

[nixuser]

Is this fixed in the 6.x stream? I see that 6.0 has just been released.

[Michael Thayer]

It should be. Did you have problems with it?

[Tomas Maroščikas]

I couldn't install 6.0.0 guest additions for Oracle Linux 7.6 with Unbreakable Enterprise Kernel 3.8.13-118.28.1.el7uek.x86_64, but installed fine with Linux 3.10.0-957.1.3.el7.x86_64 kernel.

[Tomas Maroščikas]

It also installs fine with Oracle Linux 7.6 default kernel 4.14.35-1818.3.3.el7uek.x86_64.

[nixuser]

@michael I cannot run 6.0 as I saw it was 64-bit only. I need [for now] to support 32-bit.

[Socratis]

Replying to nixuser:

I cannot run 6.0 as I saw it was 64-bit only. I need [for now] to support 32-bit.

Then your only option is to use the Test Builds, until 5.2.24 comes out, if and when it does...

[Michael Thayer]

@maroz Could you please provide the log files from the guest (/var/log/vbox*) after trying to install with 3.8.13-118.2? I have just built the OL7.3/3.8.13-118.24.2.el6uek.x86_64 combination without problems.

[Tomas Maroščikas]

logs for 3.8.13-118.28.1.el7uek.x86_64

[Michael Thayer]

@maroz Thank you. This is a different problem - see ticket #18238 - and is fixed in the current test builds. Feel free to give one a try. They are expected to be as stable as release builds for most purposes - possibly more currently, as they have several bug fixes which 6.0.0 does not have - so there is no reason not to use a test build as long as there is no new release.