Again, failure inside guest leads to failure of host

[MikhailRokhin]

Again, failure inside guest leads to failure of host. It was prev. with 4.x.x via BSD guest either.

It's not acceptable, that guest is still able to ruin host.

[Mihai Hanor]

In what manner is the host affected? A crashed VM only affects that VM instance. How is this problem affecting the host?

[MikhailRokhin]

Replying to mhanor:

In what manner is the host affected? A crashed VM only affects that VM instance. How is this problem affecting the host?

Guest terminates the host with the error message shown in screen-shots attached. Would it be just guest failure it couldn't crash & terminate host.

[Socratis]

All I see is an "Application Crash", a VirtualBox crash, not a host crash. I can't read Russian, but it seems like a Memory Protection error, yes or no? Can you translate the error message?

Also, in one of the screenshots (https://www.virtualbox.org/attachment/ticket/18033/zfs_kills-vbox.JPG) I see that the VirtualBox installed is 5.1.14. Is that the case? Because in the ticket summary you have filed 5.2.18. Which one is it? If it's 5.1.14, you should definitely update to the latest 5.2.18, as of this writing.

[aeichner]

This is not reproducible for me locally. Please provide us with a core dump (see here) and the matching VBox.log file when this happens again. You can send the core dump to "alexander (dot) eichner (at) oracle (dot) com".

[MikhailRokhin]

Replying to socratis:

All I see is an "Application Crash", a VirtualBox crash, not a host crash. I can't read Russian, but it seems like a Memory Protection error, yes or no? Can you translate the error message?

Also, in one of the screenshots (https://www.virtualbox.org/attachment/ticket/18033/zfs_kills-vbox.JPG) I see that the VirtualBox installed is 5.1.14. Is that the case? Because in the ticket summary you have filed 5.2.18. Which one is it? If it's 5.1.14, you should definitely update to the latest 5.2.18, as of this writing.

zfs_kills-vbox.JPG is screen-shot from 5.1.14
guest_fails_host_via_zfs_failure.JPG - 5.2.18

Which means problem still exists (and it was in 4.xx version either, don't have screen-shot).

Error message trans: VirtualBox.exe Application Error: "Instruction at address _xxx_ pointed to memory address _xxx_ . Memory can not be read. "

Probably, the problem addresses to HDD I/O system of VB - LSI Logic SAS.

[MikhailRokhin]

Replying to aeichner:

This is not reproducible for me locally. Please provide us with a core dump (see here) and the matching VBox.log file when this happens again. You can send the core dump to "alexander (dot) eichner (at) oracle (dot) com".

Logs are attached.

The problem is sudden & spontaneous, so I can't reproduce it on regular basis either.

Probably, the problem addresses to HDD I/O system of VB - LSI Logic SAS.

There might be just VB driver memory management error or VB+Windows memory management system error.

There was earlier (for VirtualBox 5.0.10 ; patched for 5.1.xx+) VB memory management problem, discovered by FBSD-11-current-i386 kernel compiled with specially crafted options:

options KSTACK_PAGES=64 options KVA_PAGES=1024

, which led to VB host failure (commonly "memory can not be read" problem).

[MikhailRokhin]

In addition:

Host is Windows x64 10 Home (1803 , 17134.320 ; latest)

Guest is FreeBSD 12-current-amd64 (latest)

[MikhailRokhin]

Might the "Memory can not be read" failure relate to improper interact of VB subsystem with Microsoft Address Space Load Randomization ?

[aeichner]

Can't reproduce anymore with a recent VirtualBox release, please reopen if still an issue and provide updated logs.