netdev - Re: [PATCH net-next 05/14] net_sched: sch_codel: implement lockless codel

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: <CANn89iLKtpszMyzmOvj0QdQCvsyDv7S_R04SrfEdOMZb5HQexQ@mail.gmail.com>
Date: Wed, 17 Apr 2024 18:05:46 +0200
From: Eric Dumazet <edumazet@...gle.com>
To: Simon Horman <horms@...nel.org>
Cc: "David S. Miller" <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>, 
	Paolo Abeni <pabeni@...hat.com>, Jamal Hadi Salim <jhs@...atatu.com>, 
	Cong Wang <xiyou.wangcong@...il.com>, Jiri Pirko <jiri@...nulli.us>, netdev@...r.kernel.org, 
	eric.dumazet@...il.com
Subject: Re: [PATCH net-next 05/14] net_sched: sch_codel: implement lockless codel_dump()

On Wed, Apr 17, 2024 at 5:59 PM Simon Horman <horms@...nel.org> wrote:
>
> On Mon, Apr 15, 2024 at 01:20:45PM +0000, Eric Dumazet wrote:
> > Instead of relying on RTNL, codel_dump() can use READ_ONCE()
> > annotations, paired with WRITE_ONCE() ones in codel_change().
> >
> > Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> > ---
> >  net/sched/sch_codel.c | 29 ++++++++++++++++++-----------
> >  1 file changed, 18 insertions(+), 11 deletions(-)
> >
> > diff --git a/net/sched/sch_codel.c b/net/sched/sch_codel.c
> > index ecb3f164bb25b33bd662c8ee07dc1b5945fd882d..3e8d4fe4d91e3ef2b7715640f6675aa5e8e2a326 100644
> > --- a/net/sched/sch_codel.c
> > +++ b/net/sched/sch_codel.c
> > @@ -118,26 +118,31 @@ static int codel_change(struct Qdisc *sch, struct nlattr *opt,
> >       if (tb[TCA_CODEL_TARGET]) {
> >               u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);
> >
> > -             q->params.target = ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT;
> > +             WRITE_ONCE(q->params.target,
> > +                        ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT);
> >       }
> >
> >       if (tb[TCA_CODEL_CE_THRESHOLD]) {
> >               u64 val = nla_get_u32(tb[TCA_CODEL_CE_THRESHOLD]);
> >
> > -             q->params.ce_threshold = (val * NSEC_PER_USEC) >> CODEL_SHIFT;
> > +             WRITE_ONCE(q->params.ce_threshold,
> > +                        (val * NSEC_PER_USEC) >> CODEL_SHIFT);
> >       }
> >
> >       if (tb[TCA_CODEL_INTERVAL]) {
> >               u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);
> >
> > -             q->params.interval = ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT;
> > +             WRITE_ONCE(q->params.interval,
> > +                        ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT);
> >       }
> >
> >       if (tb[TCA_CODEL_LIMIT])
> > -             sch->limit = nla_get_u32(tb[TCA_CODEL_LIMIT]);
> > +             WRITE_ONCE(sch->limit,
> > +                        nla_get_u32(tb[TCA_CODEL_LIMIT]));
> >
>
> Hi Eric,
>
> Sorry to be so bothersome.
>
> As a follow-up to our discussion of patch 4/14 (net_choke),
> I'm wondering if reading sch->limit in codel_qdisc_enqueue()
> should be updated to use READ_ONCE().

No worries !

A READ_ONCE() in codel_qdisc_enqueue() is not needed
because codel_change() writes  all the fields  under the protection of
qdisc spinlock.

sch_tree_lock() / ... / sch_tree_unlock()

Note that I removed the READ_ONCE() in choke enqueue() in V2,
for the same reason.

>
> >       if (tb[TCA_CODEL_ECN])
> > -             q->params.ecn = !!nla_get_u32(tb[TCA_CODEL_ECN]);
> > +             WRITE_ONCE(q->params.ecn,
> > +                        !!nla_get_u32(tb[TCA_CODEL_ECN]));
> >
> >       qlen = sch->q.qlen;
> >       while (sch->q.qlen > sch->limit) {
>
> ...