On 06.03.2020 10:22, Ilia Zykov wrote: > Hello, > > this page - https://blog.netbsd.org/tnf/entry/from_zero_to_nvmm says: > > “One thing you may have noticed from Fig. A, is that the complex emulation > machinery is not in the kernel, but in USERLAND. This is an excellent > security property of NVMM, because it reduces the risk for the host in case > of bug or vulnerability – the host kernel remains unaffected –, and also has > the advantage of making the machinery easily fuzzable. Currently, this > property is NOT FOUND in other hypervisors such as KVM, HAXM or Bhyve, and I > hope we'll be able to preserve it as we move forward with more backends.” > > I don't understand what does it mean “emulation machinery is not in the > kernel, but in USERLAND ... this property is NOT FOUND in other hypervisors > such as KVM ..."? > After all, qemu-kvm also runs in userland and does the same thing. What is > the difference between qemu-kvm(without virtio) and qemu-nvmm? > I'm more interested there are architectural differences, not implementation > differences. > > Kvm's guest code start point is in kernel space (but it is isolated from the > main part by using VT-x / AMD-V). Take a look at slide 9 on this page: > http://www.linux-kongress.org/2010/slides/KVM-Architecture-LK2010.pdf > But nvmm, does the same. Or nvmm returns in userspace before execute the main > guest code? I don't undestand. > > The kvm mane loop is: > open("/dev/kvm") > ioctl(KVM_CREATE_VM) > ioctl(KVM_CREATE_VCPU) > for (;;) { > ioctl(KVM_RUN) > switch (exit_reason) { > case KVM_EXIT_IO: /* ... */ > case KVM_EXIT_HLT: /* ... */ > } > } > But nvmm, does the same. > struct nvmm_exit exit; > while (1) { > nvmm_vcpu_run(&mach, 0, &exit); > switch (exit.reason) { > case NVMM_EXIT_NONE: > break; /* nothing to do */ > case ... /* completed as needed */ > } > } > Tell me, please, where am I mistaken? > I would really appreciate any clarification. > Thank you. >
A hypervisor backend shall implement instruction decoder for MMIO/PIO operations. NVMM performs this emulation in userspace, while others like HAXM perform this inside the kernel. There are pros and cons but it is a distinct property of NVMM, but it is definitely a more secure approach.
signature.asc
Description: OpenPGP digital signature
