I also can fix this by adding extra validation check in vxlan_ioctl_drvspec() to avoid renumbering.
But, this driver only uses ioctl and I've plans to add the support for netlink + vnet too.
I'm fine if anyone want me to add the extra check instead.

MFC after: 3 days

This breaks the userspace ABI. Old ifconfig binaries won't work properly with a new kernel after this change.

In D58552#1343759, @markj wrote:

This breaks the userspace ABI. Old ifconfig binaries won't work properly with a new kernel after this change.

I understand, that's why I said I can do it the other way.
I want to make sure it's needed, because users usually update their userland + kernel together.
So should I do it the other way?

In D58552#1343760, @pouria wrote:

In D58552#1343759, @markj wrote:

This breaks the userspace ABI. Old ifconfig binaries won't work properly with a new kernel after this change.

I understand, that's why I said I can do it the other way.
I want to make sure it's needed, because users usually update their userland + kernel together.
So should I do it the other way?

Yes, in general FreeBSD tries quite hard to avoid breaking the kernel<->userspace ABI. There are exceptions, but if it's easy to avoid this breakage, then we should.

It's very useful to be able to upgrade the kernel without upgrading userspace, and one might want an ifconfig binary from a jail to continue working even if the host is upgraded.