Re: [RFC PATCH 4/5] Add migration support for KVM guest with MTE

[Haibo Xu]

On Tue, 16 Feb 2021 at 23:31, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 2/7/21 7:20 PM, Haibo Xu wrote:
> > +    if (kvm_physical_memory_addr_from_host(kvm_state, addr, &ipa)) {
> > +        /* Buffer for the page tags(one byte per tag) */
> > +        tag_buf = g_try_malloc0(size);
> > +        if (!tag_buf) {
> > +            error_report("%s: Error allocating MTE tag_buf", __func__);
> > +            return 0;
> > +        }
> > +
> > +        if (kvm_arm_mte_get_tags(ipa, TARGET_PAGE_SIZE, tag_buf) < 0) {
> > +            error_report("%s: Can't get MTE tags from guest", __func__);
> > +            g_free(tag_buf);
> > +            return 0;
> > +        }
> > +
> > +        qemu_put_buffer(f, tag_buf, size);
> > +
> > +        g_free(tag_buf);
> > +
> > +        return size;
> > +    }
>
> So, in patch 2 you disabled the allocation of tag-memory.  Now you're
> allocating new memory (admittedly quite a small amount -- 1/16th of a page,
> small enough to just be a local variable).
>

Hi Richard!

Thanks so much for the comments!

Yes, the allocated memory here is for temporary tag storage. As you
mentioned, it can be
defined as a local variable which is reserved for temporary tag buffers.

> Why don't you allocate tag-memory, copy the data into it, and then let
> migration proceed as normal.  Then you don't have to have a random data block
> that happens to follow each ram page.
>

As I mentioned in the cover later, the reason to let the tag go with
the memory data together
is to make it easier to sync with each other. I think if we migratie
them separately, it would be
hard to keep the tags to sync with the data.

Saying if we migration all the data first, then the tags. If the data
got dirty during the migration
of the tag memory, we may need to send the data again, or freeze the
source VM after data
migration? What's more, the KVM_GET_DIRTY_LOG API may not be able to
differentiate
between a tag and data changes.

I'm not sure whether the aforementioned situation does exist. Please
correct me if something goes wrong!

> I'm concerned that what you're doing here makes it impossible to migrate
> between kvm and tcg.
>
>

You mean to migrate from a KVM mode VM to a TCG mode VM?

> r~

Regards,
Haibo