VirtualBox

Opened 5 years ago

Last modified 5 years ago

#18191 new defect

VirtualBox VM start crashes Win10 host inside a Linux KVM

Reported by: tomty89 Owned by:
Component: other Version: VirtualBox 5.2.22
Keywords: kvm Cc:
Guest type: other Host type: other

Description

When I start a VirtualBox VM in a Win10 host, which is in turn a Linux KVM guest, the Win10 host crashes right away (it hard resets itself without even showing a BSOD). It doesn't seem to matter if I am doing nested virtualization (i.e. kvm_intel loaded with nested=1). It appears to be a regression introduced in 5.2.8.

Change History (6)

comment:1 by Socratis, 5 years ago

Nested virtualization is not supported. You were told that over the IRC, yet four hours later you open a ticket? Nice...

Last edited 5 years ago by Socratis (previous) (diff)

comment:2 by tomty89, 5 years ago

You were *also* told that 5.2.8 or older works fine (and told here again) in a KVM when doing either virtualization or emulation. Also the bug isn't even "nested virtualization"-specific (as I also told twice), it also happen when I do pure emulation (32-bit only) without loading kvm_intel with nested=1.

There was even a similar (but probably not the same in nature, as that one was apparently nested virtualization specific) issue three years ago for 5.0.12, which was "fixed" (instead of "closed: not supported" or something like that): https://www.virtualbox.org/ticket/14965

Are you even a developer here? If not please just go away with your outdated knowledge.

in reply to:  2 comment:3 by Socratis, 5 years ago

Replying to tomty89:

You were *also* told that 5.2.8 or older works fine (and told here again) in a KVM when doing either virtualization or emulation. Also the bug isn't even "nested virtualization"-specific (as I also told twice), it also happen when I do pure emulation (32-bit only) without loading kvm_intel with nested=1.

And I said, and I really want you to try hard and understand it this time, that it's not supported. Do you realize what that means? Do you realize the difference between "supported" and "not supported"? The latter means if it works, good for you. If it breaks, that's too bad, but no one is going to lose their sleep over it.

There was even a similar (but probably not the same in nature, as that one was apparently nested virtualization specific) issue three years ago for 5.0.12, which was "fixed" (instead of "closed: not supported" or something like that): https://www.virtualbox.org/ticket/14965

Ever heard of the expression "Past results do not guarantee future performance"? Well, things like this fall into this category. It might work, it might not.

Are you even a developer here? If not please just go away with your outdated knowledge.

Oh, I'm terribly sorry, I didn't realize this was a closed club for developers only! Mea culpa! </sarcasm>

Go away? I'm sorry, you must have something wrong in your book, or even the wrong book altogether. You can't (and shouldn't) shoo away anyone that has a different approach than your idealistic one. All I'm trying is to tell you what to expect and what not to expect... (hint: not a high priority issue, if an issue at all).

Finally, please explain to me why is this a VirtualBox issue and not a KVM one. VirtualBox may have had the need to adjust some things, access some registers which are not working properly in the nested virtualization setup from the KVM side. Should VirtualBox back pedal and be denied access to its own features just because it's not working in your specific, not supported scenario? How do you envision such a reversal? Just out of curiosity...

Maybe the "raw mode" feature (a.k.a. no VT-x on the host) of VirtualBox isn't getting as much attention as it was getting in the past. There's a reason for that; the vast majority (90+%) of the computers that have shipped for the last 10 years has VT-x. Should the developers go out of their way and fix something from the Jurassic era? Next thing you're going to start complaining about 6.0.0 dropping support for Win* 32-bit hosts...

comment:4 by tomty89, 5 years ago

Yeah yeah yeah, you don't find it necessary / like it and think it should not be supported so it is unsupported. Sure. Feel free to spam even more just because you are too idle and think that your preference matters to the world. Who even asked *you* to lose *your* sleep over it? Just because you subscribed this bugtracker and thought you were a lord or whatever here?

If you aren't an official developer or representative from Oracle, what even gives you the authority to say it's unsupported? (You are not even quoting a line from the latest official docs or anything.)

Sure, it "may" be a KVM problem. If you desired so much on the closure of this issue, why don't you just find me a commit with message explicitly stating that "running in kvm will not be supported anymore and unsupported means it will just crash the OS if you do that"? Or even identify the problem in KVM (instead of just saying, it "may" be KVM's problem, even when an older version works perfectly, a.k.a. there's a regression?)

If you don't want to "lose your sleep over it", why don't you go away (yes, again) and get your sleep?

No one is trying to "get attention" for "raw mode" or whatever that you think should be dropped or out of your delicate sight. I am merely stating the facts of the issue I am experiencing.

P.S. To whom it may concern, I don't mind my comments being deleted along with the ones above, as I fail to see any relevance of it to the issue. I do wish that they all get deleted so that we can focus back onto the issue again, if anyone would like to put effort in fixing it. I actually made a typo in my last one: 5.2.6, but not 5.2.8, is the last working version; 5.2.8 is the one with the regression introduced.

comment:5 by tomty89, 5 years ago

Also, did any authority ever say that "unsupported" means I cannot and/or should not open a ticket for it?

comment:6 by Michael Thayer, 5 years ago

I am happy to delete comments if that suits everyone involved. As Socratis already pointed out, this is not a use case we support, even if it happened to work before. We support running on certain hardware configurations, but KVM is not one of those. You are probably better talking to the KVM people, who are of course free to talk to us if they are interested and think that we can help them. In the meantime I will close this ticket unless someone points out a good reason for keeping it open, as this is not something we indent to work on. (If someone else is interested in working on this that might be a reason, but they will probably find themselves working on KVM to make it work, not on VirtualBox.)

Note: See TracTickets for help on using tickets.

© 2023 Oracle
ContactPrivacy policyTerms of Use