Re: [Qemu-devel] RAMBlocks and memory_region_init_ram_nomigrate

Re: [Qemu-devel] RAMBlocks and memory_region_init_ram_nomigrate

[Paolo Bonzini]

On 13/08/2018 18:16, Frank Yang wrote:
> Hi Paolo,
> 
> I see that migration/ram.c saves RAMBlocks associated with memory
> regions initialized with nomigrate. Is this intended?

Probably the name and size of the RAMBlocks must match but the contents
need not (but honestly I haven't looked at the code to find the answer).
 CCing the qemu mailing list (always a good idea) and a couple people
that might know.

Paolo

Re: [Qemu-devel] RAMBlocks and memory_region_init_ram_nomigrate

[Peter Maydell]

On 13 August 2018 at 17:27, Paolo Bonzini <pbonzini@redhat.com> wrote:
> On 13/08/2018 18:16, Frank Yang wrote:
>> Hi Paolo,
>>
>> I see that migration/ram.c saves RAMBlocks associated with memory
>> regions initialized with nomigrate. Is this intended?

Which regions, and which version of QEMU? In general for
current-head-of-master this shouldn't happen, unless the
code has subsequently manually registered the ramblock
for migration by calling vmstate_register_ram() itself
at some point after the MR is initialized.

thanks
-- PMM

Re: [Qemu-devel] RAMBlocks and memory_region_init_ram_nomigrate

[Dr. David Alan Gilbert]

* Paolo Bonzini (pbonzini@redhat.com) wrote:
> On 13/08/2018 18:16, Frank Yang wrote:
> > Hi Paolo,
> > 
> > I see that migration/ram.c saves RAMBlocks associated with memory
> > regions initialized with nomigrate. Is this intended?
> 
> Probably the name and size of the RAMBlocks must match but the contents
> need not (but honestly I haven't looked at the code to find the answer).
>  CCing the qemu mailing list (always a good idea) and a couple people
> that might know.

All the migration code should now be using RAMBLOCK_FOREACH_MIGRATABLE and
qemu_ram_is_migratable whenever it's iterating the ramblock list,
so that *shouldn't* happen these days.
Of course we could have messed it up somewhere; what are you seeing?

Dave
> Paolo
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK

Re: [Qemu-devel] RAMBlocks and memory_region_init_ram_nomigrate

[Frank Yang]

Ah got it, thanks for the replies / info!

We're using a modified QEMU 2.12, and I don't see the migratable-only loops
and field, so it either got missed in the rebase or was added after 2.12.

Frank

On Mon, Aug 13, 2018 at 9:45 AM Dr. David Alan Gilbert <dgilbert@redhat.com>
wrote:

> * Paolo Bonzini (pbonzini@redhat.com) wrote:
> > On 13/08/2018 18:16, Frank Yang wrote:
> > > Hi Paolo,
> > >
> > > I see that migration/ram.c saves RAMBlocks associated with memory
> > > regions initialized with nomigrate. Is this intended?
> >
> > Probably the name and size of the RAMBlocks must match but the contents
> > need not (but honestly I haven't looked at the code to find the answer).
> >  CCing the qemu mailing list (always a good idea) and a couple people
> > that might know.
>
> All the migration code should now be using RAMBLOCK_FOREACH_MIGRATABLE and
> qemu_ram_is_migratable whenever it's iterating the ramblock list,
> so that *shouldn't* happen these days.
> Of course we could have messed it up somewhere; what are you seeing?
>
> Dave
> > Paolo
> --
> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
>

Re: [Qemu-devel] RAMBlocks and memory_region_init_ram_nomigrate

[Dr. David Alan Gilbert]

* Frank Yang (lfy@google.com) wrote:
> Ah got it, thanks for the replies / info!
> 
> We're using a modified QEMU 2.12, and I don't see the migratable-only loops
> and field, so it either got missed in the rebase or was added after 2.12.

It went in after 2.12; it's b895de50 (followed by some cleanups to get
some we missed).

Dave

> Frank
> 
> On Mon, Aug 13, 2018 at 9:45 AM Dr. David Alan Gilbert <dgilbert@redhat.com>
> wrote:
> 
> > * Paolo Bonzini (pbonzini@redhat.com) wrote:
> > > On 13/08/2018 18:16, Frank Yang wrote:
> > > > Hi Paolo,
> > > >
> > > > I see that migration/ram.c saves RAMBlocks associated with memory
> > > > regions initialized with nomigrate. Is this intended?
> > >
> > > Probably the name and size of the RAMBlocks must match but the contents
> > > need not (but honestly I haven't looked at the code to find the answer).
> > >  CCing the qemu mailing list (always a good idea) and a couple people
> > > that might know.
> >
> > All the migration code should now be using RAMBLOCK_FOREACH_MIGRATABLE and
> > qemu_ram_is_migratable whenever it's iterating the ramblock list,
> > so that *shouldn't* happen these days.
> > Of course we could have messed it up somewhere; what are you seeing?
> >
> > Dave
> > > Paolo
> > --
> > Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
> >
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK