[Bug 194923] New: kvm-intel.nested=1 does not enable nested KVM

[Bug 194923] New: kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

            Bug ID: 194923
           Summary: kvm-intel.nested=1 does not enable nested KVM
           Product: Virtualization
           Version: unspecified
    Kernel Version: 4.8.17
          Hardware: Intel
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: blocking
          Priority: P1
         Component: kvm
          Assignee: virtualization_kvm@kernel-bugs.osdl.org
          Reporter: R030t1@gmail.com
        Regression: No

Kernel is booted with kvm-intel.nested=1 in arguments. `cat
/sys/module/kvm_intel/parameters/nested` is "N".

These steps work for kernel 3.18.9. I apologize if the severity is too high,
but this is a serious regression and prevents usage of previously available
functionality.

Bug 64801 and 76141 may be related.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

R030t1@gmail.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Regression|No                          |Yes

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #1 from R030t1@gmail.com ---
Nested kvm works at least until 4.1.7.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #2 from R030t1@gmail.com ---
Correction - kvm nesting is displayed as a feature in 4.1.7 but does not work
guest-side. Will test other kernels time permitting.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

Kashyap Chamarthy (kashyap.cv@gmail.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kashyap.cv@gmail.com

--- Comment #3 from Kashyap Chamarthy (kashyap.cv@gmail.com) ---
(In reply to R030t1 from comment #2)
> Correction - kvm nesting is displayed as a feature in 4.1.7 but does not
> work guest-side. Will test other kernels time permitting.

Based on your comment#2, there's not much useful information to diagnose this
issue further.  You don't mention what kind of guest problem this is, or any
other details.

Please provide details about your guest OS, and related details -- potentially
`dmidecode`, `x86info -a`, `dmesg` from your physical host (level-0), and your
level-1 guest (guest hypervisor).

Side note: A colleague who runs 4.9.13 Kernel reports that a Linux guest
hypervisor was able to detect /dev/kvm just fine.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #4 from R030t1@gmail.com ---
(In reply to Kashyap Chamarthy from comment #3)
> Based on your comment#2, there's not much useful information to diagnose
> this issue further.  You don't mention what kind of guest problem this is,
> or any other details.
> 

This problem was reported based on the behavior of a Windows guest, but I will
see what I can do in a Linux guest.

I run qemu as follows:

QEMU_AUDIO_DRV=pa sudo qemu-system-x86_64 \
        -machine accel=kvm \
        -cpu host,hv_relaxed,hv_spinlocks=0x1fff,hv_vapic,hv_time,+vmx,kvm=on \
        -smp 4 \
        -m 8G \
        -drive if=pflash,format=raw,readonly,file=OVMF_CODE-pure-efi.fd \
        -drive if=pflash,format=raw,file=OVMF_VARS.fd \
        -net tap,ifname=tap0,script=no,downscript=no -net nic,model=virtio \
        -drive file=/dev/sdc,format=raw,cache=none,if=virtio \
        -drive file=windows_10_pro.iso,media=cdrom \
        -drive file=../virtio-win-0.1.126.iso,media=cdrom \
        -usbdevice tablet \
        -vga qxl \
        -soundhw hda 

I found instructions
(https://ladipro.wordpress.com/2017/02/24/running-hyperv-in-kvm-guest/) which
gave me the options passed to the cpu flag. Using what the linked article gives
does not work; I have to pass kvm=on to have Hyper-V in the Windows guest
detect the ability to host VMs. It seems like passing kvm=on instructs qemu to
lie about functionality, but when using my most recent kernel which does not
support KVM nesting it properly reports that it is unable to pass
virtualization extensions to the guest.

When Hyper-V can detect VT-x on my processor it will let me create guests but
will be unable to run them:

    Failed to start the virtual machine 'WindowsServer' because one of the
Hyper-V components is not running.

It is my guess that the Windows service which acts like qemu is unable to start
due to being unable to use virtualization extensions due to some
incompatibility with real hardware.

> Please provide details about your guest OS, and related details --
> potentially `dmidecode`, `x86info -a`, `dmesg` from your physical host
> (level-0), and your level-1 guest (guest hypervisor).
> 

I'll do my best to attach what you've requested. I don't have very good guest
integration set up so getting some of the information out may take some time.

> Side note: A colleague who runs 4.9.13 Kernel reports that a Linux guest
> hypervisor was able to detect /dev/kvm just fine.

It may be hardware specific, but I am unsure why that would be. Considering the
type of relationship Windows has with the hypervisor and the virtual hardware I
am not sure it is right to blame it.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #5 from R030t1@gmail.com ---
Created attachment 255379
  --> https://bugzilla.kernel.org/attachment.cgi?id=255379&action=edit
Output of dmidecode on host OS.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #6 from R030t1@gmail.com ---
Created attachment 255381
  --> https://bugzilla.kernel.org/attachment.cgi?id=255381&action=edit
Output of x86info -a.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #7 from R030t1@gmail.com ---
Created attachment 255383
  --> https://bugzilla.kernel.org/attachment.cgi?id=255383&action=edit
Output of dmesg.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #8 from R030t1@gmail.com ---
Created attachment 255385
  --> https://bugzilla.kernel.org/attachment.cgi?id=255385&action=edit
Output of dmidecode on guest.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #9 from R030t1@gmail.com ---
Created attachment 255387
  --> https://bugzilla.kernel.org/attachment.cgi?id=255387&action=edit
Output of x86info -a (guest).

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #10 from R030t1@gmail.com ---
Created attachment 255389
  --> https://bugzilla.kernel.org/attachment.cgi?id=255389&action=edit
Output of dmesg (guest).

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #11 from R030t1@gmail.com ---
Created attachment 255391
  --> https://bugzilla.kernel.org/attachment.cgi?id=255391&action=edit
Output of dmidecode on host OS, kernel 4.8.17.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #12 from R030t1@gmail.com ---
Created attachment 255393
  --> https://bugzilla.kernel.org/attachment.cgi?id=255393&action=edit
Output of x86info -a (4.8.17).

-- 
You are receiving this mail because:
You are watching the assignee of the bug.

[Bug 194923] kvm-intel.nested=1 does not enable nested KVM

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=194923

--- Comment #13 from R030t1@gmail.com ---
Created attachment 255395
  --> https://bugzilla.kernel.org/attachment.cgi?id=255395&action=edit
Output of dmesg (4.8.17).

-- 
You are receiving this mail because:
You are watching the assignee of the bug.