Linux host running "perf top" kills Virtualbox vm's (VERR_PGM_INVALID_CR3_ADDR)

[dimaqq]

If I run Linux tool "perf top" on host and Virtualbox VM at the same time, VM crashes. It doesn't matter which is started first.

Host versions tested:

• Linux 4.3.3-3-ARCH #1 SMP PREEMPT x86_64
• perf version 4.4.gafd2ff
• VirtualBox 5.0.14_OSE r105127

Guest versions tested:

• Linux 4.4.0-4-ARCH #1 SMP PREEMPT x86_64
• Linux 2.6.18-164el5PAE #1 SMP i686
• Windows10 64-bit modern.ie without guest additions

Host dmesg:

[  313.192046] capability: warning: `VirtualBox' uses 32-bit capabilities (legacy support in use)
[  371.880286] SUPR0GipMap: fGetGipCpu=0x3
[  372.327223] vboxdrv: ffffffffa08b4020 VMMR0.r0
[  372.522422] vboxdrv: ffffffffa09af020 VBoxDDR0.r0
[  372.529397] vboxdrv: ffffffffa00c6020 VBoxDD2R0.r0
[  372.617800] VBoxNetFlt: attached to 'eth0' / 74:d0:2b:95:1b:92
[  372.624009] device eth0 entered promiscuous mode
[  374.602722] HPET: Using timer above configured range: 2
[  374.603501] HPET: Using timer above configured range: 2
[  374.603507] HPET: Using timer above configured range: 2
[  374.603511] HPET: Using timer above configured range: 2
[  374.603515] HPET: Using timer above configured range: 2
[  374.603519] HPET: Using timer above configured range: 2
[  374.603523] HPET: Using timer above configured range: 2
[  374.603527] HPET: Using timer above configured range: 2
[  374.603531] HPET: Using timer above configured range: 2
[  374.603535] HPET: Using timer above configured range: 2
[  374.603539] HPET: Using timer above configured range: 2
[  374.603543] HPET: Using timer above configured range: 2
[  389.380249] device eth0 left promiscuous mode
[  390.099756] vboxnetflt: 0 out of 4 packets were not sent (directed to host)

Attaching VBox.log's in a second.