Kernel 4.6 vboxdrv failed to compile

[Wolfgang.Klein]

Virtualbox 5.0.20 was released at 2016-04-28
Linux Kernel 4.6 was released at 2016-05-16

I am sure there was a release candidate of kernel 4.6. *before* VB 5.0.20 was released. So, in theory, it was possible to test VB on the new upcoming kernel.

[Perryg]

From what I have seen over the years, the RC kernels may get basic support but nothing gets serious until the actual final build. It is pointless to try to tune every RC that is released since the work load would be too much.

[giaur500]

Ok, but now it's time to tune vb for stable 4.6 release. I'm waiting for update on this.

[Wolfgang.Klein]

From what I have seen over the years, the RC kernels may get basic support but nothing gets serious until the actual final build. It is pointless to try to tune every RC that is released since the work load would be too much.

That is correct, but on the other hand: if there are release candidates available on kernel.org, it never takes long for the final release to appear. With just a few more days of patience the new version of VB could have been tested on kernel 4.6 and this error and the users' frustration could have been avoided.

[towo2099]

I realy don't get your problems, vbox 5.0.22 does support kernel 4.6 fine and runs even with kernel 4.7-rc4 here.

[Wolfgang.Klein]

Sorry to disappoint you, but 5.0.22 still crashes on kernel 4.6.2.

[towo2099]

Still chrashes is saying zero since you don't provide any relevant info about your system.
Here™ all my machines, 1x Win7, 3x Android and one linux vm are working fine on kernel 4.7-rc4 and 4.6.2.

[Wolfgang.Klein]

If you take a closer look at this forum alone, there are more and more people reporting problems with kernel 4.6, people using different systems and different varieties of kernel 4.6, some using the kernels that come with their distribution, some compiling the kernel themselves. So, I think it's time to start debugging and stop telling that VM does run on kernel 4.6 and that it's just a matter which system you use.

Attached you will the log of my latest try. My system is consisting of:

- Intel QuadCore, 4GB RAM
- Linux Mint 17.3, all updates included, system totally up to date
- kernel 4.6.2, "vanilla", meaning: source code downloaded from ftp.kernel.org and compiled
- VB 5.0.22 plus extension pack, both freshly downloaded from virtualbox.org
- a VM that contains nothing special, just trying to start an ISO (Knoppix 7.7)
- results: "Guru Meditation Error", see log

The same error occurs with all other VM, no matter which guest system. When I reboot my machine into kernel 4.5.5, all the VMs run smoothly without any trouble whatsoever.

[frank]

From what I have seen over the years, the RC kernels may get basic support but nothing gets serious until the actual final build. It is pointless to try to tune every RC that is released since the work load would be too much.

That is correct, but on the other hand: if there are release candidates available on kernel.org, it never takes long for the final release to appear. With just a few more days of patience the new version of VB could have been tested on kernel 4.6 and this error and the users' frustration could have been avoided.

Changes like done in the 4.6 kernel are not the normal case. We just don't have the resources to test VirtualBox on all platforms with all available configurations and in particular we don't have time to test against RC kernels which may behave differently than the final kernel. This testing is usually done by our community and the more valuable feedback we get the more productive is the VirtualBox development team.

You seem to be convinced that the change in the 4.6 kernel is very obvious and you wonder why we didn't fix it yet. If you follow the ticket 15439 more closely then you will observe that it was a bit difficult to reproduce the issue locally. No, I do NOT have a Gentoo system available (I already said that we cannot test all and everything). With a vanilla 4.6 kernel configured with preemption enabled we didn't see the problem. Finding the responsible config option took time. We were finally able to reproduce the problem and provided a test build in the above ticket, so far I didn't see any response.

[Wolfgang.Klein]

I am sorry, if my previous article sounded a bit rude there. I didn't mean to be!

When I mentioned patience, I meant that maybe it would have been better to wait a few more days to release the new VB version, so that it could have been tested on the final kernel 4.6. Of course you are right when you point that there's no way for developers to test their product on every system and trying every option, but since this problem appears on many different systems, I believe just a few tests would have been enough to revealed it. Your mileage may differ, of course.

In the many years that I use computers and software now, and especially Linux, I have made the experience, that software and drivers, that are released just a few days before a new Linux kernel appears, are very often bound to create more new problems than they are meant to solve. So, it is my humble opinions, that it doesn't pay to rush a new release, when there are already release candidates available for a new kernel version. Once again: your mileage may differ, of course!

I will give the option mentioned in the ticket (CONFIG_PREEMPT_NOTIFIERS=y) a try, and see if it will help in my case.

[Wolfgang.Klein]

@Frank:

I've recompiled the kernel, taking into account the option that you mentioned. That was last week, and since then no virtual machine has crashed again on my system.

Thank you very much for the link to the ticket! I had not found it before, and without it, most probably, I would still be searching for the reason for this crash.

[frank]

Right, VBox 5.0.24 has this fix as well. Thanks for the feedback!