Am 05.03.2014 02:23, schrieb Alex Williamson:
On Tue, 2014-03-04 at 18:24 -0600, Kim Phillips wrote:
On Fri, 28 Feb 2014 11:03:18 -0700
Alex Williamson alex.william...@redhat.com wrote:
This all looks reasonable, but I suspect it would be cleaner if
vfio_find_get_group() was in a common file
Il 05/03/2014 12:28, Andreas Färber ha scritto:
Am 05.03.2014 02:23, schrieb Alex Williamson:
On Tue, 2014-03-04 at 18:24 -0600, Kim Phillips wrote:
On Fri, 28 Feb 2014 11:03:18 -0700
Alex Williamson alex.william...@redhat.com wrote:
This all looks reasonable, but I suspect it would be
On Fri, 28 Feb 2014 11:03:18 -0700
Alex Williamson alex.william...@redhat.com wrote:
On Tue, 2014-02-25 at 20:37 -0600, Kim Phillips wrote:
- support allocating a variable number of regions
- VFIODevice's bars[] become dynamically allocated *regions
- VFIOBAR's device fd replaced with
On Tue, 2014-03-04 at 18:24 -0600, Kim Phillips wrote:
On Fri, 28 Feb 2014 11:03:18 -0700
Alex Williamson alex.william...@redhat.com wrote:
On Tue, 2014-02-25 at 20:37 -0600, Kim Phillips wrote:
- support allocating a variable number of regions
- VFIODevice's bars[] become
Hi Kim,
1) I agree with Alex on the fact I would prefer to have the qemu-side
platform device code separated from PCI device one, both using a separate
generic helper code. Indeed calling functions referencing BAR in the
platform device does not look natural to me; although I understand the
code
On Tue, 2014-02-25 at 20:37 -0600, Kim Phillips wrote:
We basically add support for the SysBusDevice type in addition to the
existing PCIDevice support. This involves taking common code from the
existing vfio_initfn(), and putting it under a new vfio_find_get_group(),
that both vfio_initfn()
We basically add support for the SysBusDevice type in addition to the
existing PCIDevice support. This involves taking common code from the
existing vfio_initfn(), and putting it under a new vfio_find_get_group(),
that both vfio_initfn() and the new vfio_platform_realize() call.
Since realize()