netdev - Re: [PATCH net-next] net: lan966x: Add support for TC flower filter statistics

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: <20230206095227.25jh3cpix5k55qv3@soft-dev3-1>
Date:   Mon, 6 Feb 2023 10:52:27 +0100
From:   Horatiu Vultur <horatiu.vultur@...rochip.com>
To:     Simon Horman <simon.horman@...igine.com>
CC:     <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <davem@...emloft.net>, <edumazet@...gle.com>, <kuba@...nel.org>,
        <pabeni@...hat.com>, <UNGLinuxDriver@...rochip.com>
Subject: Re: [PATCH net-next] net: lan966x: Add support for TC flower filter
 statistics

The 02/04/2023 18:12, Simon Horman wrote:
> 
> On Fri, Feb 03, 2023 at 02:53:49PM +0100, Horatiu Vultur wrote:
> > Add flower filter packet statistics. This will just read the TCAM
> > counter of the rule, which mention how many packages were hit by this
> > rule.
> 
> I am curious to know how HW stats only updating the packet count
> interacts with SW stats also incrementing other values, such as the byte
> count.

First, our HW can count only the packages and not also the bytes,
unfortunately. Also we use the flag 'skip_sw' when we add the rules in
this case the statistics look OK.
If the user doesn't use the skip_sw then the statistics will look
something like this (using command: tc -s filter show dev eth0 ingress):

        Action statistics:
        Sent 92 bytes 4 pkt (dropped 0, overlimits 0 requeues 0)
        Sent software 92 bytes 2 pkt
        Sent hardware 0 bytes 2 pkt
        backlog 0b 0p requeues 0
        used_hw_stats immediate

As you see there are different counters for SW and Hw statistics.

> 
> > Signed-off-by: Horatiu Vultur <horatiu.vultur@...rochip.com>
> > ---
> >  .../microchip/lan966x/lan966x_tc_flower.c     | 22 +++++++++++++++++++
> >  1 file changed, 22 insertions(+)
> >
> > diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.c b/drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.c
> > index 88c655d6318fa..aac3d7c87f1d5 100644
> > --- a/drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.c
> > +++ b/drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.c
> > @@ -234,6 +234,26 @@ static int lan966x_tc_flower_del(struct lan966x_port *port,
> >       return err;
> >  }
> >
> > +static int lan966x_tc_flower_stats(struct lan966x_port *port,
> > +                                struct flow_cls_offload *f,
> > +                                struct vcap_admin *admin)
> > +{
> > +     struct vcap_counter count;
> > +     int err;
> > +
> > +     memset(&count, 0, sizeof(count));
> 
> nit: As was pointed out to me recently it's simpler to declare
>      count as follows and skip the memset entirely.
>      No need to respin for this!
> 
>      struct vcap_counter count = {};

Good to know this.

> 
> > +
> > +     err = vcap_get_rule_count_by_cookie(port->lan966x->vcap_ctrl,
> > +                                         &count, f->cookie);
> > +     if (err)
> > +             return err;
> > +
> > +     flow_stats_update(&f->stats, 0x0, count.value, 0, 0,
> > +                       FLOW_ACTION_HW_STATS_IMMEDIATE);
> > +
> > +     return err;
> > +}
> > +
> >  int lan966x_tc_flower(struct lan966x_port *port,
> >                     struct flow_cls_offload *f,
> >                     bool ingress)
> > @@ -252,6 +272,8 @@ int lan966x_tc_flower(struct lan966x_port *port,
> >               return lan966x_tc_flower_add(port, f, admin, ingress);
> >       case FLOW_CLS_DESTROY:
> >               return lan966x_tc_flower_del(port, f, admin);
> > +     case FLOW_CLS_STATS:
> > +             return lan966x_tc_flower_stats(port, f, admin);
> >       default:
> >               return -EOPNOTSUPP;
> >       }
> > --
> 
> Also, not strictly related, but could you consider, as a favour to
> reviewers, fixing the driver so that the following doesn't fail:
> 
> $ make drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.o
>   DESCEND objtool
>   CALL    scripts/checksyscalls.sh
>   CC      drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.o
> In file included from drivers/net/ethernet/microchip/lan966x/lan966x_tc_flower.c:3:
> drivers/net/ethernet/microchip/lan966x/lan966x_main.h:18:10: fatal error: vcap_api.h: No such file or directory
>    18 | #include <vcap_api.h>
>       |          ^~~~~~~~~~~~
> compilation terminated.

I will try to have a look at this.


-- 
/Horatiu