Date:   Mon, 15 Feb 2021 10:48:41 -0600
From:   George McCollister <george.mccollister@...il.com>
To:     "Wenzel, Marco" <Marco.Wenzel@...berle.de>
Cc:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: HSR/PRP sequence counter issue with Cisco Redbox

On Mon, Feb 15, 2021 at 6:30 AM Wenzel, Marco <Marco.Wenzel@...berle.de> wrote:
>
> > On Wed, Jan 27, 2021 at 6:32 AM Wenzel, Marco <Marco.Wenzel@a-
> > eberle.de> wrote:
> > >
> > > Hi,
> > >
> > > we have figured out an issue with the current PRP driver when trying to
> > communicate with Cisco IE 2000 industrial Ethernet switches in Redbox
> > mode. The Cisco always resets the HSR/PRP sequence counter to "1" at low
> > traffic (<= 1 frame in 400 ms). It can be reproduced by a simple ICMP echo
> > request with 1 s interval between a Linux box running with PRP and a VDAN
> > behind the Cisco Redbox. The Linux box then always receives frames with
> > sequence counter "1" and drops them. The behavior is not configurable at
> > the Cisco Redbox.
> > >
> > > I fixed it by ignoring sequence counters with value "1" at the sequence
> > counter check in hsr_register_frame_out ():
> > >
> > > diff --git a/net/hsr/hsr_framereg.c b/net/hsr/hsr_framereg.c index
> > > 5c97de459905..630c238e81f0 100644
> > > --- a/net/hsr/hsr_framereg.c
> > > +++ b/net/hsr/hsr_framereg.c
> > > @@ -411,7 +411,7 @@ void hsr_register_frame_in(struct hsr_node *node,
> > > struct hsr_port *port,  int hsr_register_frame_out(struct hsr_port *port,
> > struct hsr_node *node,
> > >                            u16 sequence_nr)  {
> > > -       if (seq_nr_before_or_eq(sequence_nr, node->seq_out[port->type]))
> > > +       if (seq_nr_before_or_eq(sequence_nr,
> > > + node->seq_out[port->type]) && (sequence_nr != 1))
> > >                 return 1;
> > >
> > >         node->seq_out[port->type] = sequence_nr;
> > >
> > >
> > > Do you think this could be a solution? Should this patch be officially applied
> > in order to avoid other users running into these communication issues?
> >
> > This isn't the correct way to solve the problem. IEC 62439-3 defines
> > EntryForgetTime as "Time after which an entry is removed from the duplicate
> > table" with a value of 400ms and states devices should usually be configured
> > to keep entries in the table for a much shorter time. hsr_framereg.c needs to
> > be reworked to handle this according to the specification.
>
> Sorry for the delay but I did not have the time to take a closer look at the problem until now.
>
> My suggestion for the EntryForgetTime feature would be the following: A time_out element will be added to the hsr_node structure, which always stores the current time when entering hsr_register_frame_out(). If the last stored time is older than EntryForgetTime (400 ms) the sequence number check will be ignored.
>
> diff --git a/net/hsr/hsr_framereg.c b/net/hsr/hsr_framereg.c
> index 5c97de459905..a97bffbd2581 100644
> --- a/net/hsr/hsr_framereg.c
> +++ b/net/hsr/hsr_framereg.c
> @@ -164,8 +164,10 @@ static struct hsr_node *hsr_add_node(struct hsr_priv *hsr,
>          * as initialization. (0 could trigger an spurious ring error warning).
>          */
>         now = jiffies;
> -       for (i = 0; i < HSR_PT_PORTS; i++)
> +       for (i = 0; i < HSR_PT_PORTS; i++) {
>                 new_node->time_in[i] = now;
> +               new_node->time_out[i] = now;
> +       }
>         for (i = 0; i < HSR_PT_PORTS; i++)
>                 new_node->seq_out[i] = seq_out;
>
> @@ -411,9 +413,12 @@ void hsr_register_frame_in(struct hsr_node *node, struct hsr_port *port,
>  int hsr_register_frame_out(struct hsr_port *port, struct hsr_node *node,
>                            u16 sequence_nr)
>  {
> -       if (seq_nr_before_or_eq(sequence_nr, node->seq_out[port->type]))
> +       if (seq_nr_before_or_eq(sequence_nr, node->seq_out[port->type]) &&
> +                time_is_after_jiffies(node->time_out[port->type] + msecs_to_jiffies(HSR_ENTRY_FORGET_TIME))) {
>                 return 1;
> +       }
>
> +       node->time_out[port->type] = jiffies;
>         node->seq_out[port->type] = sequence_nr;
>         return 0;
>  }
> diff --git a/net/hsr/hsr_framereg.h b/net/hsr/hsr_framereg.h
> index 86b43f539f2c..d9628e7a5f05 100644
> --- a/net/hsr/hsr_framereg.h
> +++ b/net/hsr/hsr_framereg.h
> @@ -75,6 +75,7 @@ struct hsr_node {
>         enum hsr_port_type      addr_B_port;
>         unsigned long           time_in[HSR_PT_PORTS];
>         bool                    time_in_stale[HSR_PT_PORTS];
> +       unsigned long           time_out[HSR_PT_PORTS];
>         /* if the node is a SAN */
>         bool                    san_a;
>         bool                    san_b;
> diff --git a/net/hsr/hsr_main.h b/net/hsr/hsr_main.h
> index 7dc92ce5a134..f79ca55d6986 100644
> --- a/net/hsr/hsr_main.h
> +++ b/net/hsr/hsr_main.h
> @@ -21,6 +21,7 @@
>  #define HSR_LIFE_CHECK_INTERVAL                 2000 /* ms */
>  #define HSR_NODE_FORGET_TIME           60000 /* ms */
>  #define HSR_ANNOUNCE_INTERVAL            100 /* ms */
> +#define HSR_ENTRY_FORGET_TIME            400 /* ms */
>
>  /* By how much may slave1 and slave2 timestamps of latest received frame from
>   * each node differ before we notify of communication problem?
>
>
> This approach works fine with the Cisco IE 2000 and I think it implements the correct way to handle sequence numbers as defined in IEC 62439-3.

Looks good to me. Can you send an official patch? If so I'll try it
out. Even if I can't replicate the Cisco situation I can try it with
my setups and make sure it doesn't break anything.

Regards,
George McCollister

>
> Regards,
> Marco Wenzel
>
> > >
> > > Thanks
> > > Marco Wenzel
> >
> > Regards,
> > George McCollister

Powered by blists - more mailing lists