Re: [PATCH v4 11/11] KVM: selftests: Add guest_memfd_preservation_test

On Thu, Jul 30 2026, Ackerley Tng wrote:

> Tarun Sahu <[email protected]> writes:
>
>> Add guest_memfd_preservation_test to verify that KVM VM state and guest
>> memory backed by guest_memfd are preserved across a kexec reboot via Live
>> Update Orchestrator (LUO).
>>
>> This test includes multiple tests
>> - (A) Test to preserve non-guest_memfd, non-memfd which must fail (/dev/null)
>> - (B) Test to preserve guest_memfd with no INIT_SHARED flag which must fail
>> - (C) Test to allocate the guest_memfd memory after preservation because
>>   preservation makes the guest_memfd frozen. Which must fail as well.
>> - (D) Test to verify guest_memfd with INIT_SHARED flag preservation across
>>   kexec: Below is the details of stages in this test.
[...]
>> +++ b/tools/testing/selftests/kvm/guest_memfd_preservation_test.c
>> @@ -0,0 +1,357 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +/*
>> + * Copyright (C) 2026, Google LLC.
>> + *
>> + * Author: Tarun Sahu <[email protected]>
>> + *
>> + * Test for VM and guest_memfd preservation across kexec (Live Update) via LUO.
>> + *
>> + * NOTE: This is a MANUAL test and is excluded from automated CI/testing
>> + * frameworks because Stage 1 daemonizes into the background to pin resources
>> + * and requires a human operator to manually trigger kexec before Stage 2
>> + * is executed. Running Stage 1 automatically would leak the background daemon
>> + * and cause CI runners to falsely interpret it as a passed test.
>> + *

I think this is fine as a first step but we really should automate
these.

We have some infrastructure to automate other live update tests across
both stages already by launching a VM under qemu and making the test
process as the init task. See tools/testing/selftests/kho/vmtest.sh. For
KVM tests you'd need the runner to be able to run nested VMs.

The script right now is pretty bare bones, but I really think we should
spend time improving it to the point where we can automate all our tests
using it.

>
> Usually there aren't many printfs in tests, but since this is a MANUAL
> test, maybe that's ok?
>
> Would like to know what Sean thinks of this, and printfs in general.

-- 
Regards,
Pratyush Yadav