On 6/5/26 09:23, David Hildenbrand (Arm) wrote:
>>> static bool
>>> @@ -47,11 +56,26 @@ memfd_backend_memory_alloc(HostMemoryBackend *backend,
>>> Error **errp)
>>> goto have_fd;
>>> }
>>>
>>> - fd = qemu_memfd_create(TYPE_MEMORY_BACKEND_MEMFD, backend->size,
>>> - m->hugetlb, m->hugetlbsize, m->seal ?
>>> - F_SEAL_GROW | F_SEAL_SHRINK | F_SEAL_SEAL : 0,
>>> - errp);
>>> - if (fd == -1) {
>>> + if (m->guest_memfd) {
>>> + /* User choose to use fully shared guest-memfd to back the VM.. */
>>> + if (!backend->share) {
>>> + error_setg(errp, "Guest-memfd=on must be used with share=on");
>>> + return false;
>>> + }
>>> +
>>> + /* TODO: add huge page support */
>>
>> Until that's added, the related options should be disabled. m->seal as
>> well doesn't seem to be applicable for guest_memfd case.
>
> Which raises the question whether we want a new backend (I think you implement
> that in your version).
>
> guest_memfd is rather different in some aspects (e.g., resizing is already
> completely forbidden).
>
> What are the pros and cons of either approach?
>
Ah, I see a discussion on that starting here:
https://lore.kernel.org/r/lpkcfd2crgparcd64ydry3ocryx3sfc5gj5pzrrms4nwvw6j4c@ulc3wa3rmefo
--
Cheers,
David
