Re: [Qemu-devel] [PATCH qemu] vmstate: Define VARRAY with VMS_ALLOC

[Dr. David Alan Gilbert]

* David Gibson (address@hidden) wrote:
> On Tue, Mar 15, 2016 at 12:01:04PM +0000, Dr. David Alan Gilbert wrote:
> > * Alexey Kardashevskiy (address@hidden) wrote:
> > > This allows dynamic allocation for migrating arrays.
> > > 
> > > Already existing VMSTATE_VARRAY_UINT32 requires an array to be
> > > pre-allocated, however there are cases when the size is not known in
> > > advance and there is no real need to enforce it.
> > > 
> > > This defines another variant of VMSTATE_VARRAY_UINT32 with WMS_ALLOC
> > > flag which tells the receiving side to allocate memory for the array
> > > before receiving the data.
> > > 
> > > The first user of it is the "pseries" machine (POWER8) with
> > > dynamic DMA windows which existence and size are totally dynamic.
> > 
> > You say totally dynamic, how big do they get out of interest?
> 
> They're basically used to map all guest RAM.  Typically we'd be
> looking at one 64-bit TCE per 64K guest page, so we'd be looking at
> 1/8192th of RAM size.
> 
> Since we can in theory have guests in the 1T+ range, that might start
> getting pretty big, so we probably should look at incremental transfer
> of the TCE tables at some point.

OK, you probably need to bump up MAX_VM_CMD_PACKAGED_SIZE (sysemu.h)
which is an arbitrary size limit for postcopies device data; it's currently
16MB.  It's just there to stop anything silly, so just turn it up a bit.

However, if it is fixed at one 64-bit TCE per guest page, why do you
need to dynamically allocate it during migration load, can't you
statically allocate once you know guests memory size?

Dave

> 
> > 
> > Dave
> > 
> > > 
> > > Signed-off-by: Alexey Kardashevskiy <address@hidden>
> > > Reviewed-by: David Gibson <address@hidden>
> > > Reviewed-by: Thomas Huth <address@hidden>
> > > ---
> > > 
> > > 
> > > As David suggested, I am posting it alone (separately from the DDW 
> > > patchset).
> > > 
> > > It would be nice either to get "ack" to get this pulled via the powerpc
> > > tree or get it pulled to the upstream via some other tree. Thanks!
> > > 
> > > 
> > > 
> > > ---
> > >  include/migration/vmstate.h | 10 ++++++++++
> > >  1 file changed, 10 insertions(+)
> > > 
> > > diff --git a/include/migration/vmstate.h b/include/migration/vmstate.h
> > > index 84ee355..1622638 100644
> > > --- a/include/migration/vmstate.h
> > > +++ b/include/migration/vmstate.h
> > > @@ -386,6 +386,16 @@ extern const VMStateInfo vmstate_info_bitmap;
> > >      .offset     = vmstate_offset_pointer(_state, _field, _type),     \
> > >  }
> > >  
> > > +#define VMSTATE_VARRAY_UINT32_ALLOC(_field, _state, _field_num, 
> > > _version, _info, _type) {\
> > > +    .name       = (stringify(_field)),                               \
> > > +    .version_id = (_version),                                        \
> > > +    .num_offset = vmstate_offset_value(_state, _field_num, uint32_t),\
> > > +    .info       = &(_info),                                          \
> > > +    .size       = sizeof(_type),                                     \
> > > +    .flags      = VMS_VARRAY_UINT32|VMS_POINTER|VMS_ALLOC,           \
> > > +    .offset     = vmstate_offset_pointer(_state, _field, _type),     \
> > > +}
> > > +
> > >  #define VMSTATE_VARRAY_UINT16_UNSAFE(_field, _state, _field_num, 
> > > _version, _info, _type) {\
> > >      .name       = (stringify(_field)),                               \
> > >      .version_id = (_version),                                        \
> 
> -- 
> David Gibson                  | I'll have my music baroque, and my code
> david AT gibson.dropbear.id.au        | minimalist, thank you.  NOT _the_ 
> _other_
>                               | _way_ _around_!
> http://www.ozlabs.org/~dgibson


--
Dr. David Alan Gilbert / address@hidden / Manchester, UK