On Fri, 21 Jul 2023 at 08:30, Philippe Mathieu-Daudé <philmd@xxxxxxxxxx> wrote: > > Hi Akihiko, > > On 21/7/23 08:24, Akihiko Odaki wrote: > > libvirt uses "none" machine type to test KVM availability. Before this > > change, QEMU used to pass 0 as machine type when calling KVM_CREATE_VM. > > > > The kernel documentation says: > >> On arm64, the physical address size for a VM (IPA Size limit) is > >> limited to 40bits by default. The limit can be configured if the host > >> supports the extension KVM_CAP_ARM_VM_IPA_SIZE. When supported, use > >> KVM_VM_TYPE_ARM_IPA_SIZE(IPA_Bits) to set the size in the machine type > >> identifier, where IPA_Bits is the maximum width of any physical > >> address used by the VM. The IPA_Bits is encoded in bits[7-0] of the > >> machine type identifier. > >> > >> e.g, to configure a guest to use 48bit physical address size:: > >> > >> vm_fd = ioctl(dev_fd, KVM_CREATE_VM, KVM_VM_TYPE_ARM_IPA_SIZE(48)); > >> > >> The requested size (IPA_Bits) must be: > >> > >> == ========================================================= > >> 0 Implies default size, 40bits (for backward compatibility) > >> N Implies N bits, where N is a positive integer such that, > >> 32 <= N <= Host_IPA_Limit > >> == ========================================================= > > > >> Host_IPA_Limit is the maximum possible value for IPA_Bits on the host > >> and is dependent on the CPU capability and the kernel configuration. > >> The limit can be retrieved using KVM_CAP_ARM_VM_IPA_SIZE of the > >> KVM_CHECK_EXTENSION ioctl() at run-time. > >> > >> Creation of the VM will fail if the requested IPA size (whether it is > >> implicit or explicit) is unsupported on the host. > > https://docs.kernel.org/virt/kvm/api.html#kvm-create-vm > > > > So if Host_IPA_Limit < 40, such KVM_CREATE_VM will fail, and libvirt > > incorrectly thinks KVM is not available. This actually happened on M2 > > MacBook Air. > > > > Fix this by specifying 32 for IPA_Bits as any arm64 system should > > support the value according to the documentation. > > > > Signed-off-by: Akihiko Odaki <akihiko.odaki@xxxxxxxxxx> > > --- > > V1 -> V2: Introduced an arch hook > > > > include/sysemu/kvm.h | 1 + > > accel/kvm/kvm-all.c | 2 +- > > target/arm/kvm.c | 2 ++ > > target/i386/kvm/kvm.c | 2 ++ > > target/mips/kvm.c | 2 ++ > > target/ppc/kvm.c | 2 ++ > > target/riscv/kvm.c | 2 ++ > > target/s390x/kvm/kvm.c | 2 ++ > > 8 files changed, 14 insertions(+), 1 deletion(-) > > My understanding of Peter's suggestion would be smth like: > > -- >8 -- > diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h > index 115f0cca79..c0af15eb6c 100644 > --- a/include/sysemu/kvm.h > +++ b/include/sysemu/kvm.h > @@ -201,10 +201,15 @@ typedef struct KVMCapabilityInfo { > > struct KVMState; > > +struct KVMClass { > + AccelClass parent_class; > + > + int default_vm_type; The kernel docs say you need to check for the KVM_CAP_ARM_VM_IPA_SIZE before you can pass something other than zero to the KVM_CREATE_VM ioctl, so this needs to be a method, not just a value. (kvm_arm_get_max_vm_ipa_size() will do this bit for you.) If the machine doesn't provide a kvm_type method, we should default to "largest the host supports", I think. I was wondering if we could have one per-arch method for "actually create the VM" that both was a place for arm to set the default vm type and also let us get the TARGET_S390X and TARGET_PPC ifdefs out of this bit of kvm-all.c, but maybe that would look just a bit too awkward: if (kc->create_vm(s, board_sets_kvm_type, board_kvm_type) < 0) { goto err; } where board_sets_kvm_type is a bool, true if board_kvm_type is valid, and board_kvm_type is whatever the board's mc->kvm_type method told us. (Default impl of the method: call KVM_CREATE_VM ioctl with retry-on-eintr, printing the simple error message; PPC and s390 versions similar but with their arch specific extra messages; arm version has a different default type if board_sets_kvm_type is false.) Not trying to do both of those things with one method would result in a simpler type = kc->get_default_kvm_type(s); API. thanks -- PMM
