modesetting driver does not support being an offload source, should not advertise cap ?

[Dave Airlie]

On 7 July 2016 at 02:08, Hans de Goede <hdegoede at redhat.com> wrote:
> Hi all,
>
> Currently the modesetting driver sets a cap flag claiming that it can
> be an offload source, yet using DRI_PRIME=1 with the modesetting driver
> being the driver for slave nvidia gpu, does not work unless the master
> gpu driver (intel) has dri3 enabled, at which point the whole xserver
> no longer gets involved (AFAICT).
>
> For the modesetting driver to really support being an offload
> source in the dri2 case, it needs to have a version of 9
> or higher in its DRI2Info struct and have the CreateBuffer2 /
> DestroyBuffer2 callbacks, which the modesetting driver does not have.
>
> So now I'm wondering am I missing something, or should the modesetting
> driver not export the offloadsource cap (or actually implement
> a version 9 DRI2Inof) ?
>

It should implement the version 9 I suspect, until it does it
shouldn't advertise offload source.
Dave.