Re: [RFC PATCH 0/4] KVM: ioctl for populating guest_memfd

[Paolo Bonzini]

On 10/24/24 11:54, Nikita Kalyazin wrote:
> Firecracker currently allows to populate guest memory from a separate
> process via UserfaultFD [1].  This helps keep the VMM codebase and
> functionality concise and generic, while offloading the logic of
> obtaining guest memory to another process.  UserfaultFD is currently not
> supported for guest_memfd, because it binds to a VMA, while guest_memfd
> does not need to (or cannot) be necessarily mapped to userspace,
> especially for private memory.  [2] proposes an alternative to
> UserfaultFD for intercepting stage-2 faults, while this series
> conceptually compliments it with the ability to populate guest memory
> backed by guest_memfd for `KVM_X86_SW_PROTECTED_VM` VMs.
>
> Patches 1-3 add a new ioctl, `KVM_GUEST_MEMFD_POPULATE`, that uses a
> vendor-agnostic implementation of `post_populate` callback.
>
> Patch 4 allows to call the ioctl from a separate (non-VMM) process.  It
> has been prohibited by [3], but I have not been able to locate the exact
> justification for the requirement.

The justification is that the "struct kvm" has a long-lived tie to a 
host process's address space.

Invoking ioctls like KVM_SET_USER_MEMORY_REGION and KVM_RUN from 
different processes would make things very messy, because it is not 
clear which mm you are working with: the MMU notifier is registered for 
kvm->mm, but some functions such as get_user_pages do not take an mm for 
example and always operate on current->mm.

In your case, it should be enough to add a ioctl on the guestmemfd 
instead?  But the real question is, what are you using 
KVM_X86_SW_PROTECTED_VM for?

Paolo

> Questions:
>   - Does exposing a generic population interface via ioctl look
>     sensible in this form?
>   - Is there a path where "only VMM can call KVM API" requirement is
>     relaxed? If not, what is the recommended efficient alternative for
>     populating guest memory from outside the VMM?
>
> [1]: https://github.com/firecracker-microvm/firecracker/blob/main/docs/snapshotting/handling-page-faults-on-snapshot-resume.md
> [2]: https://lore.kernel.org/kvm/CADrL8HUHRMwUPhr7jLLBgD9YLFAnVHc=N-C=8er-x6GUtV97pQ@xxxxxxxxxxxxxx/T/
> [3]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=6d4e4c4fca5be806b888d606894d914847e82d78
>
> Nikita
>
> Nikita Kalyazin (4):
>    KVM: guest_memfd: add generic post_populate callback
>    KVM: add KVM_GUEST_MEMFD_POPULATE ioctl for guest_memfd
>    KVM: allow KVM_GUEST_MEMFD_POPULATE in another mm
>    KVM: document KVM_GUEST_MEMFD_POPULATE ioctl
>
>   Documentation/virt/kvm/api.rst | 23 +++++++++++++++++++++++
>   include/linux/kvm_host.h       |  3 +++
>   include/uapi/linux/kvm.h       |  9 +++++++++
>   virt/kvm/guest_memfd.c         | 28 ++++++++++++++++++++++++++++
>   virt/kvm/kvm_main.c            | 19 ++++++++++++++++++-
>   5 files changed, 81 insertions(+), 1 deletion(-)
>
>
> base-commit: c8d430db8eec7d4fd13a6bea27b7086a54eda6da