Re: [PATCH 2/6] vxlan: Group Policy extension

On Mon, Jan 12, 2015 at 5:03 PM, Thomas Graf [off-list ref] wrote:

On 01/12/15 at 10:14am, Tom Herbert wrote:

quoted

quoted

diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index f7d0d2d..9f07bf5 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h

@@ -370,10 +370,18 @@ enum {
        IFLA_VXLAN_UDP_CSUM,
        IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
        IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
+       IFLA_VXLAN_EXTENSION,
        __IFLA_VXLAN_MAX
 };
 #define IFLA_VXLAN_MAX (__IFLA_VXLAN_MAX - 1)

+enum {
+       IFLA_VXLAN_EXT_UNSPEC,
+       IFLA_VXLAN_EXT_GBP,
+       __IFLA_VXLAN_EXT_MAX,
+};
+#define IFLA_VXLAN_EXT_MAX (__IFLA_VXLAN_EXT_MAX - 1)
+

Creating a level of indirection for extensions seems overly
complicated to me. Why not just define IFLA_VXLAN_GBP as just another
enum above?

I think it's cleaner to group them in a nested attribute.
It clearly separates the optional extensions from the base
attributes. RCO, GPE, GBP can all live in there.

This is inconsistent with similar things in GRE and GUE. For instance,
GRE keyid is set as its own attribute. It just seems like this adding
more code to the driver than is necessary for the functionality
needed.