netdev - Re: [PATCH net-next v10 1/3] act_vlan: Change stats update to use per-core stats

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: <CAJrKpLYJRqtMbiA7w8i3d+BUEtHMkoB649=N-aAHXd1vjboCGw@mail.gmail.com>
Date:   Wed, 8 Nov 2017 10:45:41 -0500
From:   Manish Kurup <kurup.manish@...il.com>
To:     Or Gerlitz <gerlitz.or@...il.com>
Cc:     Jamal Hadi Salim <jhs@...atatu.com>,
        Cong Wang <xiyou.wangcong@...il.com>,
        Jiri Pirko <jiri@...nulli.us>,
        David Miller <davem@...emloft.net>,
        Jakub Kicinski <jakub.kicinski@...ronome.com>,
        Pieter Jansen van Vuuren 
        <pieter.jansenvanvuuren@...ronome.com>,
        Simon Horman <simon.horman@...ronome.com>,
        John Hurley <john.hurley@...ronome.com>,
        oss-drivers@...ronome.com,
        Linux Netdev List <netdev@...r.kernel.org>,
        Alexander Aring <aring@...atatu.com>,
        Roman Mashak <mrv@...atatu.com>,
        Manish Kurup <manish.kurup@...izon.com>
Subject: Re: [PATCH net-next v10 1/3] act_vlan: Change stats update to use
 per-core stats

Hi Or,

On Wed, Nov 8, 2017 at 9:20 AM, Or Gerlitz <gerlitz.or@...il.com> wrote:
> On Wed, Nov 8, 2017 at 9:03 PM, Manish Kurup <kurup.manish@...il.com> wrote:
>> The VLAN action maintains one set of stats across all cores, and uses a
>> spinlock to synchronize updates to it from the same. Changed this to use a
>> per-CPU stats context instead.
>> This change will result in better performance.
>>
>> Acked-by: Jamal Hadi Salim <jhs@...atatu.com>
>> Acked-by: Jiri Pirko <jiri@...lanox.com>
>> Signed-off-by: Manish Kurup <manish.kurup@...izon.com>
>> ---
>>  net/sched/act_vlan.c | 10 ++++++----
>>  1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/net/sched/act_vlan.c b/net/sched/act_vlan.c
>> index 115fc33..8a35efe 100644
>> --- a/net/sched/act_vlan.c
>> +++ b/net/sched/act_vlan.c
>> @@ -30,9 +30,10 @@ static int tcf_vlan(struct sk_buff *skb, const struct tc_action *a,
>>         int err;
>>         u16 tci;
>>
>> -       spin_lock(&v->tcf_lock);
>>         tcf_lastuse_update(&v->tcf_tm);
>> -       bstats_update(&v->tcf_bstats, skb);
>> +       bstats_cpu_update(this_cpu_ptr(v->common.cpu_bstats), skb);
>> +
>> +       spin_lock(&v->tcf_lock);
>>         action = v->tcf_action;
>
> (if this was asked && answered in earlier Vs, sorry for that, if not and I got
> some small real problem here && you're @ netdev, maybe buy me Korean beer?)
>
> before your changes the spin lock also protected the lastuse update call but
> now it doesn't, why?
Phase I of my changes, was to get rid of spin_locks, and convert the
stats to a per-cpu stats model to get better forwarding performance.
While doing this, I looked at a few 'model TC actions' within
net/sched (tcf_mirred for example). Neither of them protected the
tcf_lastuse_update(). I assumed that this was the case because this
was a 'display-only' field, and as long as it changed to a latest
timestamp based on packets received, it was OK.

I tested this using our suite of traffic tests, and verified that the
last-use field did update, and did not cause any other problems.

Do you envision any issues that could be caused due to this?

Thanks,

-Manish