On Fri, 13 Nov 2020 16:08:39 +0100 Karsten Graul wrote:
> On 11/11/2020 23:34, Jakub Kicinski wrote:
> > On Mon, 9 Nov 2020 16:18:08 +0100 Karsten Graul wrote:
> >> @@ -295,6 +377,14 @@ static int smc_diag_dump(struct sk_buff *skb, struct
> >> netlink_callback *cb)
> >>
> >> static int smc_diag_dump_ext(struct sk_buff *skb, struct netlink_callback
> >> *cb)
> >> {
> >> + struct smc_diag_req_v2 *req = nlmsg_data(cb->nlh);
> >> +
> >> + if (req->cmd == SMC_DIAG_GET_LGR_INFO) {
> >> + if ((req->cmd_ext & (1 << (SMC_DIAG_LGR_INFO_SMCR - 1))))
> >> + smc_diag_fill_lgr_list(smc_diag_ops->get_lgr_list(),
> >> + skb, cb, req);
> >> + }
> >> +
> >> return skb->len;
> >> }
> >
> > IDK if this is appropriate for socket diag handler.
> >
> > Is there precedent for funneling commands through socket diag instead
> > of just creating a genetlink family?
>
> Thank you for your valuable comments. We are looking into a better way
> to retrieve the various information from the kernel into user space,
> and we will come up with a v5 for that.
Thanks, but do double check that no other socket type is doing this,
I'm far from a socket layer expert.
