netdev - Re: [p4tc-discussions] Re: [PATCH RFC v2 net-next 05/28] net/sched: act_api: introduce tc_lookup_action

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAAFAkD8XKBfNEQW7J7KP90eYcuA62gOD+i1ifP+xZUZD6-=Nug@mail.gmail.com>
Date: Mon, 5 Jun 2023 10:55:56 -0400
From: Jamal Hadi Salim <hadi@...atatu.com>
To: Marcelo Ricardo Leitner <mleitner@...hat.com>
Cc: Jamal Hadi Salim <jhs@...atatu.com>, netdev@...r.kernel.org, deb.chatterjee@...el.com, 
	tom@...anda.io, p4tc-discussions@...devconf.info, Mahesh.Shirshyad@....com, 
	Vipin.Jain@....com, tomasz.osinski@...el.com, xiyou.wangcong@...il.com, 
	davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org, pabeni@...hat.com, 
	khalidm@...dia.com, toke@...hat.com
Subject: Re: [p4tc-discussions] Re: [PATCH RFC v2 net-next 05/28] net/sched:
 act_api: introduce tc_lookup_action_byid()

On Mon, Jun 5, 2023 at 8:08 AM Marcelo Ricardo Leitner via
p4tc-discussions <p4tc-discussions@...devconf.info> wrote:
>
> On Sat, Jun 03, 2023 at 09:14:53AM -0400, Jamal Hadi Salim wrote:
> > On Fri, Jun 2, 2023 at 3:36 PM Marcelo Ricardo Leitner
> > <mleitner@...hat.com> wrote:
> > >
> > > On Wed, May 17, 2023 at 07:02:09AM -0400, Jamal Hadi Salim wrote:
> > > > +/* lookup by ID */
> > > > +struct tc_action_ops *tc_lookup_action_byid(struct net *net, u32 act_id)
> > > > +{
> > > > +     struct tcf_dyn_act_net *base_net;
> > > > +     struct tc_action_ops *a, *res = NULL;
> > > > +
> > > > +     if (!act_id)
> > > > +             return NULL;
> > > > +
> > > > +     read_lock(&act_mod_lock);
> > > > +
> > > > +     list_for_each_entry(a, &act_base, head) {
> > > > +             if (a->id == act_id) {
> > > > +                     if (try_module_get(a->owner)) {
> > > > +                             read_unlock(&act_mod_lock);
> > > > +                             return a;
> > > > +                     }
> > > > +                     break;
> > >
> > > It shouldn't call break here but instead already return NULL:
> > > if id matched, it cannot be present on the dyn list.
> > >
> > > Moreover, the search be optimized: now that TCA_ID_ is split between
> > > fixed and dynamic ranges (patch #3), it could jump directly into the
> > > right list. Control path performance is also important..
> >
> >
> >
> > Sorry - didnt respond to this last part: We could use standard tc
> > actions in a P4 program and we prioritize looking at them first. This
> > helper is currently only needed for us - so you could argue that we
> > should only look TCA_ID_DYN onwards but it is useful to be more
> > generic and since this is a slow path it is not critical. Unless i
>
> Yes, and no. Control path is not as critical as datapath, I agree with
> that, but it's also important. It's a small change, but oh well.. it
> should accumulate on complex datapaths.
>
> > misunderstood what you said.
>
> I meant something like this:
>
> +     if (act_id < TCA_ID_DYN) {
> +         read_lock(&act_mod_lock);
> +
> +         list_for_each_entry(a, &act_base, head) {
> +                 if (a->id == act_id) {
> +                         if (try_module_get(a->owner)) {
> +                                 read_unlock(&act_mod_lock);
> +                                 return a;
> +                         }
> +                         break; /* now break; is okay */
> +                 }
> +         }
> +         read_unlock(&act_mod_lock);
> +     } else {
> +         read_lock(&base_net->act_mod_lock);
> +
> +         base_net = net_generic(net, dyn_act_net_id);
> +         a = idr_find(&base_net->act_base, act_id);
> +         if (a && try_module_get(a->owner))
> +                 res = a;
> +
> +         read_unlock(&base_net->act_mod_lock);
> +     }

Reasonable. We will update with this approach. Thanks Marcelo.

cheers,
jamal

> >
> > cheers,
> > jamal
> >
> > > > +             }
> > > > +     }
> > > > +     read_unlock(&act_mod_lock);
> > > > +
> > > > +     read_lock(&base_net->act_mod_lock);
> > > > +
> > > > +     base_net = net_generic(net, dyn_act_net_id);
> > > > +     a = idr_find(&base_net->act_base, act_id);
> > > > +     if (a && try_module_get(a->owner))
> > > > +             res = a;
> > > > +
> > > > +     read_unlock(&base_net->act_mod_lock);
> > > > +
> > > > +     return res;
> > > > +}
> > > > +EXPORT_SYMBOL(tc_lookup_action_byid);
> > >
> >
>
> _______________________________________________
> p4tc-discussions mailing list -- p4tc-discussions@...devconf.info
> To unsubscribe send an email to p4tc-discussions-leave@...devconf.info