Re: [PATCH (repost) net-next] sched: add extack for tfilter_notify

[Jakub Kicinski <kuba@kernel.org>]

On Thu, 10 Nov 2022 14:29:29 +0800 Hangbin Liu wrote:
> On Wed, Nov 09, 2022 at 06:20:53PM -0800, Jakub Kicinski wrote:
> > Unless we want to create a separate netlink multicast channel for 
> > just ext acks of a family. That's fine by me, I guess. I'm mostly
> > objecting to pretending notifications are multi-msg just to reuse
> > NLMSG_DONE, and forcing all notification listeners to deal with it.  
> 
> Actually I'm a little curious about how should we use NLMSG_DONE.
> Does a normal nlmsg(with NLM_F_MULTI flag) + a NLMSG_DONE msg illegal?
> Should we need at least  2 nlmsgs + a NLMSG_DONE message.
> 
> Because when I wrote this patch, I saw some functions, like
> team_nl_send_options_get(), team_nl_send_port_list_get() in team driver,
> devlink_dpipe_tables_fill() in netlink.c, even netlink_dump_done(), could
> *possible* only have 1 nlmsg + 1 NLMSG_DONE message.
> 
> In my understand, we can send only 1 nlmsg without NLM_F_MULTI flag. But if
> there is 1 nlmsg + 1 NLMSG_DONE message. It should be considered as multi
> message, and the first nlmsg need to add NLM_F_MULTI flag. Maybe there is
> a little abuse of using NLMSG_DONE, but should be legal.
> 
> What do you think? Did I miss something?

No, I mean, it's perfectly legal to send a single message with MULTI
and NLMSG_DONE, but it's hard to deserialize that. You can hand parse
anything, but how would you describe that in terms of abstract objects
that a normal high level language can consume directly?