On Wed, Apr 22, 2020 at 10:56:18AM +0200, Cornelia Huck wrote: > On Mon, 20 Apr 2020 13:17:55 +0200 > Kashyap Chamarthy <kchamart@xxxxxxxxxx> wrote: [Just noticed this today ... thanks for the review.] [...] > > +A nested guest is the ability to run a guest inside another guest (it > > +can be KVM-based or a different hypervisor). The straightforward > > +example is a KVM guest that in turn runs on KVM a guest (the rest of > > s/on KVM a guest/on a KVM guest/ Will fix in v3. [...] > > +Terminology: > > + > > +- L0 – level-0; the bare metal host, running KVM > > + > > +- L1 – level-1 guest; a VM running on L0; also called the "guest > > + hypervisor", as it itself is capable of running KVM. > > + > > +- L2 – level-2 guest; a VM running on L1, this is the "nested guest" > > + > > +.. note:: The above diagram is modelled after x86 architecture; s390x, > > s/x86 architecture/the x86 architecture/ > > > + ppc64 and other architectures are likely to have different > > s/to have/to have a/ Noted (both the above) > > + design for nesting. > > + > > + For example, s390x has an additional layer, called "LPAR > > + hypervisor" (Logical PARtition) on the baremetal, resulting in > > + "four levels" in a nested setup — L0 (bare metal, running the > > + LPAR hypervisor), L1 (host hypervisor), L2 (guest hypervisor), > > + L3 (nested guest). > > What about: > > "For example, s390x always has an LPAR (LogicalPARtition) hypervisor > running on bare metal, adding another layer and resulting in at least > four levels in a nested setup..." Yep, reads nicer; thanks. [...] > > +1. On the host hypervisor (L0), enable the ``nested`` parameter on > > + s390x:: > > + > > + $ rmmod kvm > > + $ modprobe kvm nested=1 > > + > > +.. note:: On s390x, the kernel parameter ``hpage`` parameter is mutually > > Drop one of the "parameter"? Will do. > > + exclusive with the ``nested`` paramter; i.e. to have > > + ``nested`` enabled you _must_ disable the ``hpage`` parameter. > > "i.e., in order to be able to enable ``nested``, the ``hpage`` > parameter _must_ be disabled." > > ? Yes :) > > > + > > +2. The guest hypervisor (L1) must be allowed to have ``sie`` CPU > > "must be provided with" ? > > > + feature — with QEMU, this is possible by using "host passthrough" > > s/this is possible by/this can be done by e.g./ ? > > > + (via the command-line ``-cpu host``). > > + > > +3. Now the KVM module can be enabled in the L1 (guest hypervisor):: > > s/enabled/loaded/ Will adjust the above three; thanks. > > + > > + $ modprobe kvm > > + > > + > > +Live migration with nested KVM > > +------------------------------ > > + > > +The below live migration scenarios should work as of Linux kernel 5.3 > > +and QEMU 4.2.0. In all the below cases, L1 exposes ``/dev/kvm`` in > > +it, i.e. the L2 guest is a "KVM-accelerated guest", not a "plain > > +emulated guest" (as done by QEMU's TCG). > > The 5.3/4.2 versions likely apply to x86? Should work for s390x as well > as of these version, but should have worked earlier already :) Heh, I'll specify the x86-ness of those versions :-) > > + > > +- Migrating a nested guest (L2) to another L1 guest on the *same* bare > > + metal host. > > + > > +- Migrating a nested guest (L2) to another L1 guest on a *different* > > + bare metal host. > > + > > +- Migrating an L1 guest, with an *offline* nested guest in it, to > > + another bare metal host. > > + > > +- Migrating an L1 guest, with a *live* nested guest in it, to another > > + bare metal host. > > + > > +Limitations on Linux kernel versions older than 5.3 > > +--------------------------------------------------- > > + > > +On x86 systems-only (as this does *not* apply for s390x): > > Add a "x86" marker? Or better yet, group all the x86 stuff in an x86 > section? Right, forgot here, will do. [...] > > +Reporting bugs from "nested" setups > > +----------------------------------- > > + > > +(This is written with x86 terminology in mind, but similar should apply > > +for other architectures.) > > Better to reorder it a bit (see below). [...] > > + - Kernel, libvirt, and QEMU version from L0 > > + > > + - Kernel, libvirt and QEMU version from L1 > > + > > + - QEMU command-line of L1 -- preferably full log from > > + ``/var/log/libvirt/qemu/instance.log`` > > (if you are running libvirt) > > > + > > + - QEMU command-line of L2 -- preferably full log from > > + ``/var/log/libvirt/qemu/instance.log`` > > (if you are running libvirt) Yes, I'll mention that bit. (I'm just to used to reports coming from libvirt users :-)) > > + > > + - Full ``dmesg`` output from L0 > > + > > + - Full ``dmesg`` output from L1 > > + > > + - Output of: ``x86info -a`` (& ``lscpu``) from L0 > > + > > + - Output of: ``x86info -a`` (& ``lscpu``) from L1 > > lscpu makes sense for other architectures as well. Noted. > > + > > + - Output of: ``dmidecode`` from L0 > > + > > + - Output of: ``dmidecode`` from L1 > > This looks x86 specific? Maybe have a list of things that make sense > everywhere, and list architecture-specific stuff in specific > subsections? Can do. Do you have any other specific debugging bits to look out for s390x or any other arch? Thanks for the careful review. Much appreciate it :-) -- /kashyap
