Re: [Qemu-devel] [RFC PATCH v3 06/19] Implement "-dimm" command line option

[Vasilis Liaskovitis]

On Sat, Oct 13, 2012 at 08:57:19AM +0000, Blue Swirl wrote:
> On Tue, Oct 9, 2012 at 5:04 PM, Vasilis Liaskovitis
> <address@hidden> wrote:
> >>
snip
> >> Maybe even the dimmbus device shouldn't exist by itself after all, or
> >> it should be pretty much invisible to users. On real HW, the memory
> >> controller or south bridge handles the memory. For i440fx, it's part
> >> of the same chipset. So I think we should just add qdev properties to
> >> i440fx to specify the sizes, nodes etc. Then i440fx should create the
> >> dimmbus device unconditionally using the properties. The default
> >> properties should create a sane configuration, otherwise -global
> >> i440fx.dimm_size=512M etc. could be used. Then the bus would be
> >> populated as before or with device_add.
> >
> > hmm the problem with using only i440fx properties, is that size/nodes look
> > dimm specific to me, not chipset-memcontroller specific. Unless we only 
> > allow
> > uniform size dimms. Is it possible to have a dynamic list of sizes/nodes 
> > pairs as
> > properties of a qdev device?
> 
> I don't think so, but probably there's a limit of DIMMs that real
> controllers have, something like 8 max.

In the case of i440fx specifically, do you mean that we should model the DRB
(Dram row boundary registers in section 3.2.19 of the i440fx spec) ?

The i440fx DRB registers only supports up to 8 DRAM rows (let's say 1 row
maps 1-1 to a DimmDevice for this discussion) and only supports up to 2GB of
memory afaict (bit 31 and above is ignored).

I 'd rather not model this part of the i440fx - having only 8 DIMMs seems too
restrictive. The rest of the patchset supports up to 255 DIMMs so it would be a
waste imho to model an old pc memory controller that only supports 8 DIMMs.

There was also an old discussion about i440fx modeling here:
https://lists.nongnu.org/archive/html/qemu-devel/2011-07/msg02705.html
the general direction was that i440fx is too old and we don't want to precisely
emulate the DRB registers, since they lack flexibility.

Possible solutions:

1) is there a newer and more flexible chipset that we could model?

2) model and document a generic (non-existent) i440fx that would support more
and larger DIMMs. E.g. support 255 DIMMs. If we want to use a description
similar to the i440fx DRB registers, the registers would take up a lot of space.
In i440fx there is one 8-bit DRB register per DIMM, and DRB[i] describes how
many 8MB chunks are contained in DIMMs 0...i. So, the register values are
cumulative (and total described memory cannot exceed 256x8MB = 2GB)

We could for example model: 
- an 8-bit non-cumulative register for each DIMM, denoting how many
128MB chunks it contains. This allowes 32GB for each DIMM, and with 255 DIMMs we
describe a bit less than 8TB. These registers require 255 bytes.
- a 16-bit cumulative register for each DIMM again for 128MB chunks. This allows
us to describe 8TB of memory (but the registers take up double the space, 
because
they describe cumulative memory amounts)

3) let everything be handled/abstracted by dimmbus - the chipset DRB modelling
is not done (at least for i440fx, other machines could). This is the least 
precise
in terms of emulation. On the other hand, if we are not really trying to emulate
the real (too restrictive) hardware, does it matter?

thanks,

- Vasilis