On Thu, Feb 18, 2016 at 06:19:07PM +0100, Olaf Hering wrote:
> On Thu, Feb 18, Wei Liu wrote:
>
> > Sorry I don't follow. What do you mean by 1:1 copy? Why does it make the
> > update unnecessary?
>
> The current code does:
>
> libxl_device_nic_init(&nic_saved);
> libxl_device_nic_copy(CTX,
On Thu, Feb 18, Wei Liu wrote:
> Sorry I don't follow. What do you mean by 1:1 copy? Why does it make the
> update unnecessary?
The current code does:
libxl_device_nic_init(&nic_saved);
libxl_device_nic_copy(CTX, &nic_saved, nic);
nic->devid = libxl__device_nextid(gc, domid, "vif");
libx
On Thu, Feb 18, 2016 at 05:54:38PM +0100, Olaf Hering wrote:
> On Thu, Feb 18, Wei Liu wrote:
>
> > For example, user might not have specified mac address so the library
> > generates one for (s)he. You don't want mac address to regenerate after
> > save / restore or migration. But you don't want
On Thu, Feb 18, Wei Liu wrote:
> For example, user might not have specified mac address so the library
> generates one for (s)he. You don't want mac address to regenerate after
> save / restore or migration. But you don't want to preserve all
> autogenerated state, so you use the original copy as
On Thu, Feb 18, 2016 at 04:13:21PM +0100, Olaf Hering wrote:
> What is the point of libxl__update_config_nic and
> libxl__update_config_vtmp?
>
> In libxl__device_type_add (called from DEFINE_DEVICE_ADD) the input
> type is copied with libxl_device_type_copy to type_saved, which is a
> 1:1 copy. I
What is the point of libxl__update_config_nic and
libxl__update_config_vtmp?
In libxl__device_type_add (called from DEFINE_DEVICE_ADD) the input
type is copied with libxl_device_type_copy to type_saved, which is a
1:1 copy. If needed, a new devid is assigned to the input. Later the
copy is updated
