From: Vladimir Oltean <olteanv@gmail.com>
To: Linus Walleij <linus.walleij@linaro.org>
Cc: Andrew Lunn <andrew@lunn.ch>,
Vivien Didelot <vivien.didelot@gmail.com>,
Florian Fainelli <f.fainelli@gmail.com>,
"David S . Miller" <davem@davemloft.net>,
Jakub Kicinski <kuba@kernel.org>, netdev <netdev@vger.kernel.org>,
DENG Qingfang <dqfext@gmail.com>,
Mauri Sandberg <sandberg@mailfence.com>
Subject: Re: [PATCH net] net: dsa: tag_rtl4_a: Fix egress tags
Date: Tue, 31 Aug 2021 22:05:59 +0300 [thread overview]
Message-ID: <20210831190559.3kiwaxyeyxn2733p@skbuf> (raw)
In-Reply-To: <CACRpkdbX4XErV-7UCezobF4jLX-HvjMHE=dnYYLqD5Sb8LkCpw@mail.gmail.com>
On Tue, Aug 31, 2021 at 08:35:05PM +0200, Linus Walleij wrote:
> On Tue, Aug 31, 2021 at 12:20 AM Vladimir Oltean <olteanv@gmail.com> wrote:
>
> > > > Does it get broadcast, or forwarded by MAC DA/VLAN ID as you'd expect
> > > > for a regular data packet?
> > >
> > > It gets broadcast :/
> >
> > Okay, so a packet sent to a port mask of zero behaves just the same as a
> > packet sent to a port mask of all ones is what you're saying?
> > Sounds a bit... implausible?
> >
> > When I phrased the question whether it gets "forwarded by MAC DA/VLAN ID",
> > obviously this includes the possibility of _flooding_, if the MAC
> > DA/VLAN ID is unknown to the FDB. The behavior of flooding a packet due
> > to unknown destination can be practically indistinguishable from a
> > "broadcast" (the latter having the sense that "you've told the switch to
> > broadcast this packet to all ports", at least this is what is implied by
> > the context of your commit message).
> >
> > The point is that if the destination is not unknown, the packet is not
> > flooded (or "broadcast" as you say). So "broadcast" would be effectively
> > a mischaracterization of the behavior.
>
> Oh OK sorry what I mean is that the packet appears on all ports of
> the switch. Not sent to the broadcast address.
Yes, but why (due to which hardware decision does this behavior take place)?
I was not hung up on the "broadcast" word. That was used a bit imprecisely,
but I got over that. I was curious as to _why_ would the packets be
delivered to all ports of the switch. After all, you told the switch to
send the packet to _no_ port :-/
The reason why I'm so interested about this is because other switches
(mt7530) treat a destination port mask of 0x0 as "look up the FDB"
(reported by Qingfang here):
https://patchwork.kernel.org/project/netdevbpf/patch/20210825083832.2425886-3-dqfext@gmail.com/#24407683
This means it would be possible to implement the bridge TX forwarding
offload feature:
https://patchwork.kernel.org/project/netdevbpf/cover/20210722155542.2897921-1-vladimir.oltean@nxp.com/
I just wanted to know what type of packets were you testing with. If you
were testing with a unidirectional stream (where the switch has no
opportunity to learn the destination MAC on a particular port), then it
is much more likely that what's happening in your case is that the
packets were flooded, and not simply "broadcast". Pick a different MAC
DA, which _is_ learned in the FDB, and the packets would not be
"broadcast" (actually flooded) at all.
This is still my hypothesis about what was going on.
> > Just want to make sure that the switch does indeed "broadcast" packets
> > with a destination port mask of zero. Also curious if by "all ports",
> > the CPU port itself is also included (effectively looping back the packet)?
>
> It does not seem to appear at the CPU port. It appear on ports
> 0..4.
Which again would be consistent with my theory.
> > > > > - out = (RTL4_A_PROTOCOL_RTL8366RB << 12) | (2 << 8);
> > > >
> > > > What was 2 << 8? This patch changes that part.
> > >
> > > It was a bit set in the ingress packets, we don't really know
> > > what egress tag bits there are so first I just copied this
> > > and since it turns out the bits in the lower order are not
> > > correct I dropped this too and it works fine.
> > >
> > > Do you want me to clarify in the commit message and
> > > resend?
> >
> > Well, it is definitely not a logical part of the change. Also, a bug fix
> > patch that goes to stable kernels seems like the last place to me where
> > you'd want to change something that you don't really know what it does...
> > In net-next, this extra change is more than welcome. Possibly has
> > something to do with hardware address learning on the CPU port, but this
> > is just a very wild guess based on some other Realtek tagging protocol
> > drivers I've looked at recently. Anyway, more than likely not just a
> > random number with no effect.
>
> Yeah but I don't know anything else about it than that it appear
> in the ingress packets which are known to have a different
> format than the egress packets, the assumption that they were
> the same format was wrong ... so it seems best to drop it as well.
> But if you insist I can defer that to a separate patch for next.
> I just can't see that it has any effect at all.
I was about to say that you can do as you wish, but I see you've resent
already. In general it is good to keep a change to the point, and
documented well and clear.
next prev parent reply other threads:[~2021-08-31 19:06 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-28 23:56 [PATCH net] net: dsa: tag_rtl4_a: Fix egress tags Linus Walleij
2021-08-30 7:29 ` Vladimir Oltean
2021-08-30 21:56 ` Linus Walleij
2021-08-30 22:20 ` Vladimir Oltean
2021-08-31 18:35 ` Linus Walleij
2021-08-31 19:05 ` Vladimir Oltean [this message]
-- strict thread matches above, loose matches on Subject: below --
2021-02-28 17:08 [PATCH net] net: dsa: tag_rtl4_a: fix " DENG Qingfang
2021-02-28 17:54 ` Florian Fainelli
2021-03-01 13:58 ` Linus Walleij
2021-03-01 14:01 ` Vladimir Oltean
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210831190559.3kiwaxyeyxn2733p@skbuf \
--to=olteanv@gmail.com \
--cc=andrew@lunn.ch \
--cc=davem@davemloft.net \
--cc=dqfext@gmail.com \
--cc=f.fainelli@gmail.com \
--cc=kuba@kernel.org \
--cc=linus.walleij@linaro.org \
--cc=netdev@vger.kernel.org \
--cc=sandberg@mailfence.com \
--cc=vivien.didelot@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).