From: Vladimir Oltean <vladimir.oltean@nxp.com>
To: Andrew Lunn <andrew@lunn.ch>
Cc: netdev <netdev@vger.kernel.org>,
Nikolay Aleksandrov <razor@blackwall.org>,
Ido Schimmel <idosch@mellanox.com>,
"bridge@lists.linux-foundation.org"
<bridge@lists.linux-foundation.org>
Subject: Re: [PATCH RFC] net: bridge: Clear offload_fwd_mark when passing frame up bridge interface.
Date: Fri, 6 May 2022 14:36:45 +0000 [thread overview]
Message-ID: <20220506143644.mzfffht44t3glwci@skbuf> (raw)
In-Reply-To: <20220505225904.342388-1-andrew@lunn.ch>
Hi Andrew,
On Fri, May 06, 2022 at 12:59:04AM +0200, Andrew Lunn wrote:
> It is possible to stack bridges on top of each other. Consider the
> following which makes use of an Ethernet switch:
>
> br1
> / \
> / \
> / \
> br0.11 wlan0
> |
> br0
> / | \
> p1 p2 p3
>
> br0 is offloaded to the switch. Above br0 is a vlan interface, for
> vlan 11. This vlan interface is then a slave of br1. br1 also has
> wireless interface as a slave. This setup trunks wireless lan traffic
> over the copper network inside a VLAN.
>
> A frame received on p1 which is passed up to the bridge has the
> skb->offload_fwd_mark flag set to true, indicating it that the switch
> has dealt with forwarding the frame out ports p2 and p3 as
> needed. This flag instructs the software bridge it does not need to
> pass the frame back down again. However, the flag is not getting reset
> when the frame is passed upwards. As a result br1 sees the flag,
> wrongly interprets it, and fails to forward the frame to wlan0.
>
> When passing a frame upwards, clear the flag.
>
> RFC because i don't know the bridge code well enough if this is the
> correct place to do this, and if there are any side effects, could the
> skb be a clone, etc.
Each skb has its own offload_fwd_mark, so clearing it for this skb does
not affect a clone. And when a packet is simultaneously forwarded and
locally received, the order is first forward/flood it, then receive it.
Cloning takes place during forwarding using deliver_clone(), so it
shouldn't be the case that you are clearing the offload_fwd_mark for a
yet-to-be-forwarded packet, either. So I think we're good there.
>
> Fixes: f1c2eddf4cb6 ("bridge: switchdev: Use an helper to clear forward mark")
> Signed-off-by: Andrew Lunn <andrew@lunn.ch>
> ---
> net/bridge/br_input.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/net/bridge/br_input.c b/net/bridge/br_input.c
> index 196417859c4a..9327a5fad1df 100644
> --- a/net/bridge/br_input.c
> +++ b/net/bridge/br_input.c
> @@ -39,6 +39,13 @@ static int br_pass_frame_up(struct sk_buff *skb)
> dev_sw_netstats_rx_add(brdev, skb->len);
>
> vg = br_vlan_group_rcu(br);
> +
> + /* Reset the offload_fwd_mark because there could be a stacked
> + * bridge above, and it should not think this bridge it doing
> + * that bridges work forward out its ports.
"this bridge is doing that bridge's work forwarding out its ports"
> + */
> + br_switchdev_frame_unmark(skb);
> +
> /* Bridge is just like any other port. Make sure the
> * packet is allowed except in promisc mode when someone
> * may be running packet capture.
> --
> 2.36.0
>
The good thing with this patch is that it avoids conditionals.
The bad thing is that it prevents true offloading of this configuration
from being possible (when "wlan0" is "p4").
I don't know what hardware is capable of doing this, but I think it's
cautious to not exclude it, either.
Some safer alternatives to this patch are based on the idea that we
could ignore skb->offload_fwd_mark coming from an unoffloaded bridge
port (i.e. treat this condition at br1, not at br0). We could:
- clear skb->offload_fwd_mark in br_handle_frame_finish(), if p->hwdom is 0
- change nbp_switchdev_allowed_egress() to return true if cb->src_hwdom == 0
next prev parent reply other threads:[~2022-05-06 14:36 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-05 22:59 [PATCH RFC] net: bridge: Clear offload_fwd_mark when passing frame up bridge interface Andrew Lunn
2022-05-05 23:07 ` Stephen Hemminger
2022-05-06 1:18 ` Andrew Lunn
2022-05-06 15:05 ` Stephen Hemminger
2022-05-06 14:36 ` Vladimir Oltean [this message]
2022-05-06 16:58 ` Andrew Lunn
2022-05-08 7:52 ` Ido Schimmel
2022-05-12 20:38 ` Andrew Lunn
2022-05-13 12:47 ` Ido Schimmel
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=20220506143644.mzfffht44t3glwci@skbuf \
--to=vladimir.oltean@nxp.com \
--cc=andrew@lunn.ch \
--cc=bridge@lists.linux-foundation.org \
--cc=idosch@mellanox.com \
--cc=netdev@vger.kernel.org \
--cc=razor@blackwall.org \
/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).