On Thu, 15 Feb 2024 23:53:51 +0000, Anish Moorthy wrote:
> This series adds an option to cause stage-2 fault handlers to
> KVM_MEMORY_FAULT_EXIT when they would otherwise be required to fault in
> the userspace mappings. Doing so allows userspace to receive stage-2
> faults directly from KVM_RUN instead of through userfaultfd, which
> suffers from serious contention issues as the number of vCPUs scales.
>
> Support for the new option (KVM_CAP_EXIT_ON_MISSING) is added to the
> demand_paging_test, which demonstrates the scalability improvements:
> the following data was collected using [2] on an x86 machine with 256
> cores.
>
> [...]
Applied 1,2, and 4 to kvm-x86 generic, and 10-12 to kvm-x86 selftests.
I skipped all KVM_CAP_EXIT_ON_MISSING as per our decision to hold off until we
see the KVM userfault stuff. I skipped the docs patch because it would require
more massaging than I wanted to do when applying. And lastly, I skipped the
"Add memslot_flags parameter to memstress_create_vm()" patch because it would be
dead code without the exit-on-missing usage.
Please take a look at the selftests commits in particular, as I did a decent
amount of massaging when applying.
Thanks!
[01/14] KVM: Clarify meaning of hva_to_pfn()'s 'atomic' parameter
https://github.com/kvm-x86/linux/commit/ed2f049fc144
[02/14] KVM: Add function comments for __kvm_read/write_guest_page()
https://github.com/kvm-x86/linux/commit/a3bd2f7ead6d
...
[04/14] KVM: Simplify error handling in __gfn_to_pfn_memslot()
https://github.com/kvm-x86/linux/commit/f588557ac4ac
...
[10/14] KVM: selftests: Report per-vcpu demand paging rate from demand paging test
https://github.com/kvm-x86/linux/commit/2ca76c12c48b
[11/14] KVM: selftests: Allow many vCPUs and reader threads per UFFD in demand paging test
https://github.com/kvm-x86/linux/commit/df4ec5aada9d
[12/14] KVM: selftests: Use EPOLL in userfaultfd_util reader threads
https://github.com/kvm-x86/linux/commit/0cba6442e9e2
--
https://github.com/kvm-x86/linux/tree/next
