IFLA_NUM_VF not set or improperly parsed

[Jef Oliver]

On 03/13/2014 05:02 AM, Thomas Graf wrote:
> Extending rtnl_link_build_get_request() will fix
> rtnl_link_get_kernel() as you have confirmed. In order to also
> fix it for cache operations, the function link_request_update()
> needs to be extended to something like this:
> 
> static int link_request_update(struct nl_cache *cache, struct nl_sock 
> *sk)
> {
>         struct ifinfomsg ifi;
>         struct nl_msg *msg;
> 	int err = 0;
> 
>         memset(&ifi, 0, sizeof(ifi));
> 
>         if (!(msg = nlmsg_alloc_simple(RTM_GETLINK, NLM_F_DUMP)))
>                 return -NLE_NOMEM;
> 
> 	ifi.ifi_family = cache->c_iarg1;
> 
>         if (nlmsg_append(msg, &ifi, sizeof(ifi), NLMSG_ALIGNTO) < 0) {
> 		err = -NLE_MSGSIZE;
>                 goto nla_put_failure;
> 	}
> 
> 	NLA_PUT_U32(msg, IFLA_EXT_MASK, RTEXT_FILTER_VF);
> 
> 	nl_send_auto(sk, msg);
> 
> nla_put_failure:
>         nlmsg_free(msg);
>         return err;
> }

I am bringing together a patch that includes this and the previous 
suggestion, as well as functionality to get information for a VF. I 
appreciate the information and answers.

Jef Oliver