5.0.5 on SGI UV2000 is unable to boot VMs, even newly created VMs

[emanuele lombardi]

Hi, I ran virtualbox on my Silicon Graphich UV2000 (SLES11SP3) a few months ago, but now after upgrading kernel from 3.0.101-0.40 to 3.0.101-0.47.55 virtualbox is unable start VM and also to create a new VM and install it from its iso.

The problem doesn't happen on a HP-DL585G7 running the same OS and kernel version of the UV2000 and this lets me think the problem is somehow due to the "strange" topology of the UV2000 (256 cores and 1TB RAM) But the problem didn't exist with kernel 3.0.101-0.40

I'm aware the machine is quite unusual, but I'll be glad to help as much as possible in solving the problem.

I must say that the installation of VirtualBox has always been and still is errorles and kernel modules are properly compiled.

When the kernel was 3.0.101-0.40 VirtualBox-4.3-4.3.28_100309_sles11.0-1.x86_64.rpm worked properly

Now with kernel 3.0.101-0.47.55 Virtualbox is unable to start ANY VM. I tried the following VirtualBoxes:

VirtualBox-4.3.29-101478-Linux_amd64.run
VirtualBox-5.0.0_RC3-101495-Linux_amd64.run
VirtualBox-5.0-5.0.0_101573_sles11.0-1.x86_64.rpm

None of the above works properly even if when vboxdrv are loaded in /var/log/messages I get "nice" messages :

Jul  9 18:50:20 giotto kernel: [ 3356.708359] vboxdrv: Found 256 processor cores
Jul  9 18:50:20 giotto kernel: [ 3356.732960] vboxdrv: TSC mode is Invariant, tentative frequency 2400019782 Hz
Jul  9 18:50:20 giotto kernel: [ 3356.732964] vboxdrv: Successfully loaded version 5.0.0 (interface 0x00230001)
Jul  9 18:50:20 giotto kernel: [ 3356.958906] VBoxNetFlt: Successfully started.
Jul  9 18:50:20 giotto kernel: [ 3356.965830] VBoxNetAdp: Successfully started.
Jul  9 18:50:20 giotto kernel: [ 3356.971906] VBoxPciLinuxInit
Jul  9 18:50:20 giotto kernel: [ 3356.971933] vboxpci: IOMMU not found (not registered)

Errors depend upon virtrualbox version infact booting a VM

VirtualBox-4.3.29-101478 gives these kind of errors and the boot freezes:

    BUG: soft lockup - CPU#3 stuck for 22s! [watchdog/3:21]

VirtualBox-5.0.0* things are worst since the booting phase not even starts and I get these messages:

    RTR3InitEx failed with rc=1018 please try reinstalling Virtualbox

    The virtual machine "lavoro7" has terminated unexpectedly during startup with exit code 1 (0x1)

    Result Code: NS_ERROR_FAILURE (0x800004005) Component: MachineWrap Interface: Machine {.....}

Unfortunately with 5.0.* the error happens so "early" that no log is written, but in /var/log/messages I see:

Jul  9 18:53:44 giotto kernel: [ 3561.109602] supdrvGipDetectGetGipCpuCallback: idCpu=0xd6 iCpuSet=214 idApic=0x4c - duplicate APIC ID.
...
...
...
Jul  9 18:53:44 giotto kernel: [ 3561.109602] supdrvGipDetectGetGipCpuCallback: idCpu=0xd7 iCpuSet=215 idApic=0x4e - duplicate APIC ID.
Jul  9 18:53:44 giotto kernel: [ 3561.109528] supdrvGipDetectGetGipCpuCallback: idCpu=0x44 iCpuSet=68 idApic=0x8 - duplicate APIC ID.
Jul  9 18:53:44 giotto kernel: [ 3561.109528] supdrvGipDetectGetGipCpuCallback: idCpu=0x42 iCpuSet=66 idApic=0x4 - duplicate APIC ID.
Jul  9 18:53:44 giotto kernel: [ 3561.116419] SUPR0GipMap: APIC ID, CPU ID or CPU set index problem detected on CPU #72 (0x48)!
Jul  9 18:53:44 giotto kernel: [ 3561.116434] SUPR0GipMap: failed rc=-1018

The above are the same errors I reported yesterday in ticket n. 14249 but I now send a new ticked since I realized errors are not related to any pre-existing VM since they occur also for new VM created from scratch.

Hoping this helps, regards, Emanuele

[emanuele lombardi]

Unfortunately I'm still unable to run VB on my UV2000 (256 cores SLES11SP3) also with VirtualBox-5.0-5.0.4_102546_sles11.0-1.x86_64.rpm

I just noticed that in the VB gui the max number of CPUs is 32. While 32 is usually a large enough number, in my case it is not. Is it possible the limit to 32 is somehow the reason of my problem which occurs also for newly created VM?

Thanks a lot, Emanuele