Re: [Qemu-devel] virtio-rng

[Gerd Hoffmann]

On 11/16/09 13:28, Ian Molton wrote:

> I've done something similar (below)
>
> My commandline looks like:
>
> - -virtiorng dev=/dev/foo,rate=1234
>
> I added some properties to my driver which are obviously then filled in
> from the options code, and I do my init like this:

There is no need for a new -virtiorng switch.  qdev allows to set any 
property when creating devices via -device on the command line.

> VirtIODevice *virtio_rng_init(DeviceState *dev, RNGConf *rngdev)
> {
>      VirtIORng *s;
>      s = (VirtIORng *)virtio_common_init("virtio-rng",
>                                              VIRTIO_ID_RNG,
>                                              0, sizeof(VirtIORng));
>
>      if (!s)
>          return NULL;
>
>      s->vdev.get_features = virtio_rng_get_features;
>
>      s->vq = virtio_add_queue(&s->vdev, 128, virtio_rng_handle);
>      s->chr = qemu_chr_open("vrngdev", rngdev->device, NULL);
>
>      ...
>
>      return&s->vdev;
> }
>
> Is this 'not the right way' ? I think the commandline looks much simpler
> this way.

The point is to separate host and guest state.  -chardev ist host state, 
the device is guest state, and they are linked using the chardev name.

> OTOH, I can see that this might help with my next problem, which is that
> I want to write EGD support for virtio-rng, so I'm guessing my
> commandline would (in your scheme) need to be something like:
>
>
>      -tcpsocketthing $addr.of.egd.server,id=some-name-here,
>       more-egd-client-options
>      -device virtio-rng,tcpsocketthing=some-name-here,rate=$limit

Maybe ...

   -chardev socket,id=egd,host=egd.domain.tld,port=whatever
   -device virtio-rng,chardev=egd

... ?

> This doesnt, however, get around my problem that the -device option
> doesnt know how to parse OPT_SIZE types, though. (I'd like to be able to
> specify 'rate' in B/s or KB/s (or even kbits/sec, more appropriately))

Yes, there is no qdev property type (yet) which can parse those 
postfixes.  Which is no reason to not add one though ;)

cheers,
  Gerd