1.. SPDX-License-Identifier: GPL-2.0
2
3==============================
4Running nested guests with KVM
5==============================
6
7A nested guest is the ability to run a guest inside another guest (it
8can be KVM-based or a different hypervisor).  The straightforward
9example is a KVM guest that in turn runs on a KVM guest (the rest of
10this document is built on this example)::
11
12              .----------------.  .----------------.
13              |                |  |                |
14              |      L2        |  |      L2        |
15              | (Nested Guest) |  | (Nested Guest) |
16              |                |  |                |
17              |----------------'--'----------------|
18              |                                    |
19              |       L1 (Guest Hypervisor)        |
20              |          KVM (/dev/kvm)            |
21              |                                    |
22      .------------------------------------------------------.
23      |                 L0 (Host Hypervisor)                 |
24      |                    KVM (/dev/kvm)                    |
25      |------------------------------------------------------|
26      |        Hardware (with virtualization extensions)     |
27      '------------------------------------------------------'
28
29Terminology:
30
31- L0 – level-0; the bare metal host, running KVM
32
33- L1 – level-1 guest; a VM running on L0; also called the "guest
34  hypervisor", as it itself is capable of running KVM.
35
36- L2 – level-2 guest; a VM running on L1, this is the "nested guest"
37
38.. note:: The above diagram is modelled after the x86 architecture;
39          s390x, ppc64 and other architectures are likely to have
40          a different design for nesting.
41
42          For example, s390x always has an LPAR (LogicalPARtition)
43          hypervisor running on bare metal, adding another layer and
44          resulting in at least four levels in a nested setup — L0 (bare
45          metal, running the LPAR hypervisor), L1 (host hypervisor), L2
46          (guest hypervisor), L3 (nested guest).
47
48          This document will stick with the three-level terminology (L0,
49          L1, and L2) for all architectures; and will largely focus on
50          x86.
51
52
53Use Cases
54---------
55
56There are several scenarios where nested KVM can be useful, to name a
57few:
58
59- As a developer, you want to test your software on different operating
60  systems (OSes).  Instead of renting multiple VMs from a Cloud
61  Provider, using nested KVM lets you rent a large enough "guest
62  hypervisor" (level-1 guest).  This in turn allows you to create
63  multiple nested guests (level-2 guests), running different OSes, on
64  which you can develop and test your software.
65
66- Live migration of "guest hypervisors" and their nested guests, for
67  load balancing, disaster recovery, etc.
68
69- VM image creation tools (e.g. ``virt-install``,  etc) often run
70  their own VM, and users expect these to work inside a VM.
71
72- Some OSes use virtualization internally for security (e.g. to let
73  applications run safely in isolation).
74
75
76Enabling "nested" (x86)
77-----------------------
78
79From Linux kernel v4.20 onwards, the ``nested`` KVM parameter is enabled
80by default for Intel and AMD.  (Though your Linux distribution might
81override this default.)
82
83In case you are running a Linux kernel older than v4.19, to enable
84nesting, set the ``nested`` KVM module parameter to ``Y`` or ``1``.  To
85persist this setting across reboots, you can add it in a config file, as
86shown below:
87
881. On the bare metal host (L0), list the kernel modules and ensure that
89   the KVM modules::
90
91    $ lsmod | grep -i kvm
92    kvm_intel             133627  0
93    kvm                   435079  1 kvm_intel
94
952. Show information for ``kvm_intel`` module::
96
97    $ modinfo kvm_intel | grep -i nested
98    parm:           nested:bool
99
1003. For the nested KVM configuration to persist across reboots, place the
101   below in ``/etc/modprobed/kvm_intel.conf`` (create the file if it
102   doesn't exist)::
103
104    $ cat /etc/modprobe.d/kvm_intel.conf
105    options kvm-intel nested=y
106
1074. Unload and re-load the KVM Intel module::
108
109    $ sudo rmmod kvm-intel
110    $ sudo modprobe kvm-intel
111
1125. Verify if the ``nested`` parameter for KVM is enabled::
113
114    $ cat /sys/module/kvm_intel/parameters/nested
115    Y
116
117For AMD hosts, the process is the same as above, except that the module
118name is ``kvm-amd``.
119
120
121Additional nested-related kernel parameters (x86)
122-------------------------------------------------
123
124If your hardware is sufficiently advanced (Intel Haswell processor or
125higher, which has newer hardware virt extensions), the following
126additional features will also be enabled by default: "Shadow VMCS
127(Virtual Machine Control Structure)", APIC Virtualization on your bare
128metal host (L0).  Parameters for Intel hosts::
129
130    $ cat /sys/module/kvm_intel/parameters/enable_shadow_vmcs
131    Y
132
133    $ cat /sys/module/kvm_intel/parameters/enable_apicv
134    Y
135
136    $ cat /sys/module/kvm_intel/parameters/ept
137    Y
138
139.. note:: If you suspect your L2 (i.e. nested guest) is running slower,
140          ensure the above are enabled (particularly
141          ``enable_shadow_vmcs`` and ``ept``).
142
143
144Starting a nested guest (x86)
145-----------------------------
146
147Once your bare metal host (L0) is configured for nesting, you should be
148able to start an L1 guest with::
149
150    $ qemu-kvm -cpu host [...]
151
152The above will pass through the host CPU's capabilities as-is to the
153guest, or for better live migration compatibility, use a named CPU
154model supported by QEMU. e.g.::
155
156    $ qemu-kvm -cpu Haswell-noTSX-IBRS,vmx=on
157
158then the guest hypervisor will subsequently be capable of running a
159nested guest with accelerated KVM.
160
161
162Enabling "nested" (s390x)
163-------------------------
164
1651. On the host hypervisor (L0), enable the ``nested`` parameter on
166   s390x::
167
168    $ rmmod kvm
169    $ modprobe kvm nested=1
170
171.. note:: On s390x, the kernel parameter ``hpage`` is mutually exclusive
172          with the ``nested`` paramter — i.e. to be able to enable
173          ``nested``, the ``hpage`` parameter *must* be disabled.
174
1752. The guest hypervisor (L1) must be provided with the ``sie`` CPU
176   feature — with QEMU, this can be done by using "host passthrough"
177   (via the command-line ``-cpu host``).
178
1793. Now the KVM module can be loaded in the L1 (guest hypervisor)::
180
181    $ modprobe kvm
182
183
184Live migration with nested KVM
185------------------------------
186
187Migrating an L1 guest, with a  *live* nested guest in it, to another
188bare metal host, works as of Linux kernel 5.3 and QEMU 4.2.0 for
189Intel x86 systems, and even on older versions for s390x.
190
191On AMD systems, once an L1 guest has started an L2 guest, the L1 guest
192should no longer be migrated or saved (refer to QEMU documentation on
193"savevm"/"loadvm") until the L2 guest shuts down.  Attempting to migrate
194or save-and-load an L1 guest while an L2 guest is running will result in
195undefined behavior.  You might see a ``kernel BUG!`` entry in ``dmesg``, a
196kernel 'oops', or an outright kernel panic.  Such a migrated or loaded L1
197guest can no longer be considered stable or secure, and must be restarted.
198Migrating an L1 guest merely configured to support nesting, while not
199actually running L2 guests, is expected to function normally even on AMD
200systems but may fail once guests are started.
201
202Migrating an L2 guest is always expected to succeed, so all the following
203scenarios should work even on AMD systems:
204
205- Migrating a nested guest (L2) to another L1 guest on the *same* bare
206  metal host.
207
208- Migrating a nested guest (L2) to another L1 guest on a *different*
209  bare metal host.
210
211- Migrating a nested guest (L2) to a bare metal host.
212
213Reporting bugs from nested setups
214-----------------------------------
215
216Debugging "nested" problems can involve sifting through log files across
217L0, L1 and L2; this can result in tedious back-n-forth between the bug
218reporter and the bug fixer.
219
220- Mention that you are in a "nested" setup.  If you are running any kind
221  of "nesting" at all, say so.  Unfortunately, this needs to be called
222  out because when reporting bugs, people tend to forget to even
223  *mention* that they're using nested virtualization.
224
225- Ensure you are actually running KVM on KVM.  Sometimes people do not
226  have KVM enabled for their guest hypervisor (L1), which results in
227  them running with pure emulation or what QEMU calls it as "TCG", but
228  they think they're running nested KVM.  Thus confusing "nested Virt"
229  (which could also mean, QEMU on KVM) with "nested KVM" (KVM on KVM).
230
231Information to collect (generic)
232~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
233
234The following is not an exhaustive list, but a very good starting point:
235
236  - Kernel, libvirt, and QEMU version from L0
237
238  - Kernel, libvirt and QEMU version from L1
239
240  - QEMU command-line of L1 -- when using libvirt, you'll find it here:
241    ``/var/log/libvirt/qemu/instance.log``
242
243  - QEMU command-line of L2 -- as above, when using libvirt, get the
244    complete libvirt-generated QEMU command-line
245
246  - ``cat /sys/cpuinfo`` from L0
247
248  - ``cat /sys/cpuinfo`` from L1
249
250  - ``lscpu`` from L0
251
252  - ``lscpu`` from L1
253
254  - Full ``dmesg`` output from L0
255
256  - Full ``dmesg`` output from L1
257
258x86-specific info to collect
259~~~~~~~~~~~~~~~~~~~~~~~~~~~~
260
261Both the below commands, ``x86info`` and ``dmidecode``, should be
262available on most Linux distributions with the same name:
263
264  - Output of: ``x86info -a`` from L0
265
266  - Output of: ``x86info -a`` from L1
267
268  - Output of: ``dmidecode`` from L0
269
270  - Output of: ``dmidecode`` from L1
271
272s390x-specific info to collect
273~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
274
275Along with the earlier mentioned generic details, the below is
276also recommended:
277
278  - ``/proc/sysinfo`` from L1; this will also include the info from L0
279