Inspired from struct flow_cls_offload :: cmd, in order for taprio to be
able to report statistics (which is future work), it seems that we need
to drill one step further with the ndo_setup_tc(TC_SETUP_QDISC_TAPRIO)
multiplexing, and pass the command as part of the common portion of the
muxed structure.

Since we already have an "enable" variable in tc_taprio_qopt_offload,
refactor all drivers to check for "cmd" instead of "enable", and reject
every other command except "replace" and "destroy" - to be future proof.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com> # for lan966x
Acked-by: Kurt Kanzenbach <kurt@linutronix.de> # hellcreek
Reviewed-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Reviewed-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In taprio_dump_class_stats() we don't need a reference to the root Qdisc
once we get the reference to the child corresponding to this traffic
class, so it's okay to overwrite "sch". But in a future patch we will
need the root Qdisc too, so create a dedicated "child" pointer variable
to hold the child reference. This also makes the code adhere to a more
conventional coding style.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Lukasz Majewski says:

====================
dsa: marvell: Add support for mv88e6071 and 6020  switches

After the commit (SHA1: 7e951737);
"net: dsa: mv88e6xxx: fix max_mtu of 1492 on 6165, 6191, 6220, 6250, 6290" the
error when mv88e6020 or mv88e6071 is used is not present anymore.

As a result patches for adding max frame size are not required to provide
working setup with aforementioned switches.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

A mv88e6250 family switch with 5 internal PHYs, 2 RMIIs
and no PTP support.
Signed-off-by: Lukasz Majewski <lukma@denx.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

A mv88e6250 family switch with 2 PHY and RMII ports and
no PTP support.
Signed-off-by: Matthias Schiffer <matthias.schiffer@ew.tq-group.com>
Signed-off-by: Lukasz Majewski <lukma@denx.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Switches from mv88e6250 family (including mv88e6020 and mv88e6071) need
the possibility to setup the maximal frame size, as they support frames
up to 2048 bytes.
Signed-off-by: Lukasz Majewski <lukma@denx.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

After commit b8a1a4cd ("i2c: Provide a temporary .probe_new()
call-back type"), all drivers being converted to .probe_new() and then
03c835f4 ("i2c: Switch .probe() to not take an id parameter") convert
back to (the new) .probe() to be able to eventually drop .probe_new() from
struct i2c_driver.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_gro_complete() function only updates the skb fields related to GRO
and it always returns zero. All the 3 drivers which are using it
do not check for the return value either.

Change it to return void instead which simplifies its callers as
error handing becomes unnecessary.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Christian Marangi says:

====================
leds: introduce new LED hw control APIs

Since this series is cross subsystem between LED and netdev,
a stable branch was created to facilitate merging process.

This is based on top of branch ib-leds-netdev-v6.5 present here [1]
and rebased on top of net-next since the LED stable branch got merged.

This is a continue of [2]. It was decided to take a more gradual
approach to implement LEDs support for switch and phy starting with
basic support and then implementing the hw control part when we have all
the prereq done.

This is the main part of the series, the one that actually implement the
hw control API.

Some history about this feature and why
=======================================

This proposal is highly requested by the entire net community but the API
is not strictly designed for net usage but for a more generic usage.

Initial version were very flexible and designed to try to support every
aspect of the LED driver with many complex function that served multiple
purpose. There was an idea to have sw only and hw only LEDs and sw only
and hw only LEDs.

With some heads up from Andrew from the net mailing list, it was suggested
to implement a more basic yet easy to implement system.

These API strictly work with a designated trigger to offload their
function.
This may be confused with hw blink offload but LED may have an even more
advanced configuration where the entire aspect of the trigger is
offloaded and completely handled by the hardware.

An example of this usage are PHY or switch port LEDs. Almost every of
these kind of device have multiple LED attached and provide info of the
current port state.

Currently we lack any support of them but these device always provide a
way to configure them, from basic feature like turning the LED off or no
(implemented in previous series related to this feature) or even entirely
driven by the hw and power on/off/blink based on some events, like tx/rx
traffic, ethernet cable attached, link speed of 10mbps, 100mbps, 1000mbps
or more. They can also support multiple logic like blink with traffic only
if a particular link speed is attached. (an example of this is when a LED
is designated to be turned on only with 100mbps link speed and configured
to blink on traffic and a secondary LED of a different color is present to
serve the same function but only when the link speed is 1000mbps)

These case are very common for a PHY or a switch but they were never
standardized so OEM support all kind of variant and configuration.

Again with Andrew we compared some feature and we reached a common set
of modes that are for sure present in every kind of devices.

And this concludes history and why.

What is present in this series
==============================

This patch contain the required API to support this feature, I decided on
the name of hw control to quickly describe this feature.

I documented each require API in the related Documentation for leds-class
so I think it might me redundant to expose them here. Feel free to tell me
how to improve it if anything is not clear.

On an abstract idea, this feature require this:

    - The trigger needs to make use of it, this is currently implemented
      for the netdev trigger but other trigger can be expanded if the
      device expose these function. An idea might be a anything that
      handle a storage disk and have the LED configurable to blink when
      there is any activity to the disk.

    - The LED driver needs to expose and implement these new API.

Currently a LED driver supports only a trigger. The trigger should use
the related helper to check if the LED can be driven hy hardware.

The different modes a trigger support are exposed in the kernel include
leds.h header and are used by the LED driver to understand what to do.

From a user standpoint, he should enable modes as usual from sysfs and if
anything is not supported warned.

Final words and missing piece from this series
==============================================

I honestly hope this feature can finally be implemented.

This series originally had also additional modes and logic to add to the
netdev trigger, but I decided to strip them and implement only the API
and support basic tx and rx. After this is merged, I will quickly propose
these additional modes.

Currently this is limited to tx and rx and this is what the current user
qca8k use. Marvell PHY support link and a generic blink with any kind of
traffic (both rx and tx). qca8k switch supports keeping the LED on based on
link speed.

The next series will add the concept of hw control only modes to the netdev
trigger and support for these additional modes:
- link_10
- link_100
- link_1000
- activity

The current implementation is voluntary basic and limited to put the ground
work and have something easy to implement and usable. 99% part of the logic
is done on the trigger side, leaving to the LED driver only the validating
and the apply part.

As shown for the PHY led binding, people are really intrested in this
feature as quickly after they were merged, people were already working on
adding support for it.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/lee/leds.git/?h=ib-leds-netdev-6.5
[2] https://lore.kernel.org/lkml/20230216013230.22978-1-ansuelsmth@gmail.com/

Changes in v4:
- Added review tag from Andrew.
- Move default interval to a define to keep them synced.
- Apply suggested reword to improve Documentation rst.

Changes in v3:
- Rebased on top of net-next

Changes in v2:
- Drop helper as currently used only by one trigger
- Improve Documentation and document return error of some functions
- Squash some patch to reduce series size
- Drop trigger mode mask as currently not used
- Rework hw control validating function to a simple implementation

Changes from previous v8 series:
- Rewrite Documentation from scratch and move to separate commit
- Strip additional trigger modes (to propose in a different series)
- Strip from qca8k driver additional modes (to implement in the different
  series)
- Split the netdev chages to smaller piece to permit easier review

Changelog in the previous v8 series: (stripped of unrelated changes)
v8:
- Improve the documentation of the new feature
- Rename to a more symbolic name
- Fix some bug in netdev trigger (not using BIT())
- Add more define for qca8k-leds driver
- Drop interval support
- Fix many bugs in the validate option in the netdev trigger
v7:
- Fix qca8k leds documentation warning
- Remove RFC tag
v6:
- Back to RFC.
- Drop additional trigger
- Rework netdev trigger to support common modes used by switch and
  hardware only triggers
- Refresh qca8k leds logic and driver
v5:
- Move out of RFC. (no comments from Andrew this is the right path?)
- Fix more spelling mistake (thx Randy)
- Fix error reported by kernel test bot
- Drop the additional HW_CONTROL flag. It does simplify CONFIG
  handling and hw control should be available anyway to support
  triggers as module.
v4:
- Rework implementation and drop hw_configure logic.
  We now expand blink_set.
- Address even more spelling mistake. (thx a lot Randy)
- Drop blink option and use blink_set delay.
v3:
- Rework start/stop as Andrew asked.
- Use test_bit API to check flag passed to hw_control_configure.
- Added a new cmd to hw_control_configure to reset any active blink_mode.
- Refactor all the patches to follow this new implementation.
v2:
- Fix spelling mistake (sorry)
- Drop patch 02 "permit to declare supported offload triggers".
  Change the logic, now the LED driver declare support for them
  using the configure_offload with the cmd TRIGGER_SUPPORTED.
- Rework code to follow this new implementation.
- Update Documentation to better describe how this offload
  implementation work.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

In order that the LED trigger can blink the switch MAC ports LED, it
needs to know the netdev associated to the port. Add the callback to
return the struct device of the netdev.

Add an helper function qca8k_phy_to_port() to convert the phy back to
dsa_port index, as we reference LED port based on the internal PHY
index and needs to be converted back.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Implement hw_control ops to drive Switch LEDs based on hardware events.

Netdev trigger is the declared supported trigger for hw control
operation and supports the following mode:
- tx
- rx

When hw_control_set is called, LEDs are set to follow the requested
mode.
Each LEDs will blink at 4Hz by default.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Expose netdev trigger modes to make them accessible by LED driver that
will support netdev trigger for hw control.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

On netdev trigger activation, hw control may be already active by
default. If this is the case and a device is actually provided by
hw_control_get_device(), init the already active mode and set the
bool to hw_control bool to true to reflect the already set mode in the
trigger_data.
Co-developed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The netdev which the LED should blink for is configurable in
/sys/class/led/foo/device_name. Ensure when offloading that the
configured netdev is the same as the netdev the LED is associated
with. If it is not, only perform software blinking.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for LED hw control for the netdev trigger.

The trigger on calling set_baseline_state to configure a new mode, will
do various check to verify if hw control can be used for the requested
mode in can_hw_control() function.

It will first check if the LED driver supports hw control for the netdev
trigger, then will use hw_control_is_supported() and finally will call
hw_control_set() to apply the requested mode.

To use such mode, interval MUST be set to the default value and net_dev
MUST be set. If one of these 2 value are not valid, hw control will
never be used and normal software fallback is used.

The default interval value is moved to a define to make sure they are
always synced.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Reject interval store with hw_control enabled. It's are currently not
supported and MUST be set to the default value with hw control enabled.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add basic check for hw control support. Check if the required API are
defined and check if the defined trigger supported in hw control for the
LED driver match netdev.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Introduce function to check if the requested mode can use hw control in
preparation for hw control support. Currently everything is handled in
software so can_hw_control will always return false.

Add knob with the new value hw_control in trigger_data struct to
set hw control possible. Useful for future implementation to implement
in set_baseline_state() the required function to set the requested mode
using LEDs hw control ops and in other function to reject set if hw
control is currently active.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move the code into a helper, ready for it to be called at
other times. No intended behaviour change.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Document new Hardware driven LEDs APIs.

Some LEDs can be programmed to be driven by hardware. This is not
limited to blink but also to turn off or on autonomously.
To support this feature, a LED needs to implement various additional
ops and needs to declare specific support for the supported triggers.

Add documentation for each required value and API to make hw control
possible and implementable by both LEDs and triggers.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some specific LED triggers blink the LED based on events from a device
or subsystem.
For example, an LED could be blinked to indicate a network device is
receiving packets, or a disk is reading blocks. To correctly enable and
request the hw control of the LED, the trigger has to check if the
network interface or block device configured via a /sys/class/led file
match the one the LED driver provide for hw control for.

Provide an API call to get the device which the LED blinks for.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add an option to permit LED driver to declare support for a specific
trigger to use hw control and setup the LED to blink based on specific
provided modes.

Add APIs for LEDs hw control. These functions will be used to activate
hardware control where a LED will use the provided flags, from an
unique defined supported trigger, to setup the LED to be driven by
hardware.

Add hw_control_is_supported() to ask the LED driver if the requested
mode by the trigger are supported and the LED can be setup to follow
the requested modes.

Deactivate hardware blink control by setting brightness to LED_OFF via
the brightness_set() callback.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since commit a4dfa72d ("tipc: set default MTU for UDP media"), it's
been no longer using dev->mtu for b->mtu, and the issue described in
commit 3de81b75 ("tipc: check minimum bearer MTU") doesn't exist
in UDP bearer any more.

Besides, dev->mtu can still be changed to a too small mtu after the UDP
bearer is created even with tipc_mtu_bad() check in tipc_udp_enable().
Note that NETDEV_CHANGEMTU event processing in tipc_l2_device_event()
doesn't really work for UDP bearer.

So this patch deletes the unnecessary tipc_mtu_bad from tipc_udp_enable.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
Link: https://lore.kernel.org/r/282f1f5cc40e6cad385aa1c60569e6c5b70e2fb3.1685371933.git.lucien.xin@gmail.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Alexis Lothoré says:

====================
net: dsa: mv88e6xxx: add 88E6361 support

This series brings initial support for Marvell 88E6361 switch.

MV88E6361 is a 8 ports switch with 5 integrated Gigabit PHYs and 3
2.5Gigabit SerDes interfaces. It is in fact a new variant in the
88E639X/88E6193X/88E6191X family with a subset of existing features:
- port 0: MII, RMII, RGMII, 1000BaseX, 2500BaseX
- port 3 to 7: triple speed internal phys
- port 9 and 10: 1000BaseX, 25000BaseX

Since said family is already well supported in mv88e6xxx driver, adding
initial support for this new switch mostly consists in finding the ID
exposed in its identification register, adding a proper description
in switch description tables in mv88e6xxx driver, and enforcing 88E6361
specificities in mv88e6393x_XXX methods.

- first 4 commits introduce an internal phy offset field for switches which
  have internal phys but not starting from port 0
- 5th commit is a fix on existing switches based on first commits
- 6th commit is a slight modification to prepare 886361 support
- last commit introduces 88E6361 support in 88E6393X family

This initial support has been tested with two samples of a custom board
with the following hardware configuration:
- a main CPU connected to MV88E6361 using port 0 as CPU port
- port 9 wired to a SFP cage
- port 10 wired to a G.Hn transceiver

The following setup was used:
PC <-ethernet-> (copper SFP) - Board 1 - (G.hn) <-phone line(RJ11)-> (G.hn) Board 2

The unit 1 has been configured to bridge SFP port and G.hn port together,
which allowed to successfully ping Board 2 from PC.
====================

Link: https://lore.kernel.org/r/20230529080246.82953-1-alexis.lothore@bootlin.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Marvell 88E6361 is an 8-port switch derived from the
88E6393X/88E9193X/88E6191X switches family. It can benefit from the
existing mv88e6xxx driver by simply adding the proper switch description in
the driver. Main differences with other switches from this
family are:
- 8 ports exposed (instead of 11): ports 1, 2 and 8 not available
- No 5GBase-x nor SFI/USXGMII support
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Some switches families have minor differences on supported link speed for
ports. Instead of redefining a new port_max_speed_mode for each different
configuration, allow to pass mv88e6xxx_chip structure to allow
differentiating those chips by known chip id
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

88E6393X/88E6193X/88E6191X switches have in fact 8 internal PHYs, but those
are not present starting at port 0: supported ports go from 1 to 8
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

mv88e6xxx currently assumes that switch equipped with internal phys have
those phys mapped contiguously starting from port 0 (see
mv88e6xxx_phy_is_internal). However, some switches have internal PHYs but
NOT starting from port 0. For example 88e6393X, 88E6193X and 88E6191X have
integrated PHYs available on ports 1 to 8
To properly support this offset, add a new field to allow specifying an
internal PHYs layout. If field is not set, default layout is assumed (start
at port 0)
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Make sure to use existing helper to get internal PHYs count instead of
redoing it manually
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Since this function is a simple helper, we do not need to pass a full
dsa_switch structure, we can directly pass the mv88e6xxx_chip structure.
Doing so will allow to share this function with any other function
not manipulating dsa_switch structure but needing info about number of
internal phys
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Marvell MV88E6361 is an 8-port switch derived from the
88E6393X/88E9193X/88E6191X switches family. Since its functional behavior
is very close to switches from this family, it can benefit from existing
drivers for this family, so add it to the list of compatible switches
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Ido Schimmel says:

====================
Add layer 2 miss indication and filtering

tl;dr
=====

This patchset adds a single bit to the tc skb extension to indicate that
a packet encountered a layer 2 miss in the bridge and extends flower to
match on this metadata. This is required for non-DF (Designated
Forwarder) filtering in EVPN multi-homing which prevents decapsulated
BUM packets from being forwarded multiple times to the same multi-homed
host.

Background
==========

In a typical EVPN multi-homing setup each host is multi-homed using a
set of links called ES (Ethernet Segment, i.e., LAG) to multiple leaf
switches in a rack. These switches act as VTEPs and are not directly
connected (as opposed to MLAG), but can communicate with each other (as
well as with VTEPs in remote racks) via spine switches over L3.

When a host sends a BUM packet over ES1 to VTEP1, the VTEP will flood it
to other VTEPs in the network, including those connected to the host
over ES1. The receiving VTEPs must drop the packet and not forward it
back to the host. This is called "split-horizon filtering" (SPH) [1].

FRR configures SPH filtering using two tc filters. The first, an ingress
filter that matches on packets received from VTEP1 and marks them using
a fwmark (firewall mark). The second, an egress filter configured on the
LAG interface connected to the host that matches on the fwmark and drops
the packets. Example:

 # tc filter add dev vxlan0 ingress pref 1 proto all flower enc_src_ip $VTEP1_IP action skbedit mark 101
 # tc filter add dev bond0 egress pref 1 handle 101 fw action drop

Motivation
==========

For each ES, only one VTEP is elected by the control plane as the DF.
The DF is responsible for forwarding decapsulated BUM traffic to the
host over the ES. The non-DF VTEPs must drop such traffic as otherwise
the host will receive multiple copies of BUM traffic. This is called
"non-DF filtering" [2].

Filtering of multicast and broadcast traffic can be achieved using the
following flower filter:

 # tc filter add dev bond0 egress pref 1 proto all flower indev vxlan0 dst_mac 01:00:00:00:00:00/01:00:00:00:00:00 action drop

Unlike broadcast and multicast traffic, it is not currently possible to
filter unknown unicast traffic. The classification into unknown unicast
is performed by the bridge driver, but is not visible to other layers.

Implementation
==============

The proposed solution is to add a single bit to the tc skb extension
that is set by the bridge for packets that encountered an FDB or MDB
miss. The flower classifier is extended to be able to match on this new
metadata bit in a similar fashion to existing metadata options such as
'indev'.

A bit that is set for every flooded packet would also work, but it does
not allow us to differentiate between registered and unregistered
multicast traffic which might be useful in the future.

A relatively generic name is chosen for this bit - 'l2_miss' - to allow
its use to be extended to other layer 2 devices such as VXLAN, should a
use case arise.

With the above, the control plane can implement a non-DF filter using
the following tc filters:

 # tc filter add dev bond0 egress pref 1 proto all flower indev vxlan0 dst_mac 01:00:00:00:00:00/01:00:00:00:00:00 action drop
 # tc filter add dev bond0 egress pref 2 proto all flower indev vxlan0 l2_miss true action drop

The first drops broadcast and multicast traffic and the second drops
unknown unicast traffic.

Testing
=======

A test exercising the different permutations of the 'l2_miss' bit is
added in patch #8.

Patchset overview
=================

Patch #1 adds the new bit to the tc skb extension and sets it in the
bridge driver for packets that encountered a miss. The marking of the
packets and the use of this extension is protected by the
'tc_skb_ext_tc' static key in order to keep performance impact to a
minimum when the feature is not in use.

Patch #2 extends the flow dissector to dissect this information from the
tc skb extension into the 'FLOW_DISSECTOR_KEY_META' key.

Patch #3 extends the flower classifier to be able to match on the new
layer 2 miss metadata. The classifier enables the 'tc_skb_ext_tc' static
key upon the installation of the first filter that matches on 'l2_miss'
and disables the key upon the removal of the last filter that matches on
it.

Patch #4 rejects matching on the new metadata in drivers that already
support the 'FLOW_DISSECTOR_KEY_META' key.

Patches #5-#6 are small preparations in mlxsw.

Patch #7 extends mlxsw to be able to match on layer 2 miss.

Patch #8 adds a selftest.

iproute2 patches can be found here [3].

[1] https://datatracker.ietf.org/doc/html/rfc7432#section-8.3
[2] https://datatracker.ietf.org/doc/html/rfc7432#section-8.5
[3] https://github.com/idosch/iproute2/tree/submit/non_df_filter_v1
[4] https://lore.kernel.org/netdev/20230518113328.1952135-1-idosch@nvidia.com/
[5] https://lore.kernel.org/netdev/20230509070446.246088-1-idosch@nvidia.com/
====================

Link: https://lore.kernel.org/r/20230529114835.372140-1-idosch@nvidia.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Add test cases to verify that the bridge driver correctly marks layer 2
misses only when it should and that the flower classifier can match on
this metadata.

Example output:

 # ./tc_flower_l2_miss.sh
 TEST: L2 miss - Unicast                                             [ OK ]
 TEST: L2 miss - Multicast (IPv4)                                    [ OK ]
 TEST: L2 miss - Multicast (IPv6)                                    [ OK ]
 TEST: L2 miss - Link-local multicast (IPv4)                         [ OK ]
 TEST: L2 miss - Link-local multicast (IPv6)                         [ OK ]
 TEST: L2 miss - Broadcast                                           [ OK ]
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Add the 'fdb_miss' key element to supported key blocks and make use of
it to match on layer 2 miss.

The key is only supported on Spectrum-{2,3,4}. An error is returned for
Spectrum-1 since the key element is not present in any of its key
blocks.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Currently, mlxsw only supports the 'ingress_ifindex' field in the
'FLOW_DISSECTOR_KEY_META' key, but subsequent patches are going to add
support for the 'l2_miss' field as well. It is valid to only match on
'l2_miss' without 'ingress_ifindex', so do not force matching on it.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Currently, mlxsw only supports the 'ingress_ifindex' field in the
'FLOW_DISSECTOR_KEY_META' key, but subsequent patches are going to add
support for the 'l2_miss' field as well. Split the parsing of the
'ingress_ifindex' field to a separate function to avoid nesting. No
functional changes intended.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Adjust drivers that support the 'FLOW_DISSECTOR_KEY_META' key to reject
filters that try to match on the newly added layer 2 miss field. Add an
extack message to clearly communicate the failure reason to user space.

The following users were not patched:

1. mtk_flow_offload_replace(): Only checks that the key is present, but
   does not do anything with it.
2. mlx5_tc_ct_set_tuple_match(): Used as part of netfilter offload,
   which does not make use of the new field, unlike tc.
3. get_netdev_from_rule() in nfp: Likewise.

Example:

 # tc filter add dev swp1 egress pref 1 proto all flower skip_sw l2_miss true action drop
 Error: mlxsw_spectrum: Can't match on "l2_miss".
 We have an error talking to the kernel
Acked-by: Elad Nachman <enachman@marvell.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Add the 'TCA_FLOWER_L2_MISS' netlink attribute that allows user space to
match on packets that encountered a layer 2 miss. The miss indication is
set as metadata in the tc skb extension by the bridge driver upon FDB or
MDB lookup miss and dissected by the flow dissector to the
'FLOW_DISSECTOR_KEY_META' key.

The use of this skb extension is guarded by the 'tc_skb_ext_tc' static
key. As such, enable / disable this key when filters that match on layer
2 miss are added / deleted.

Tested:

 # cat tc_skb_ext_tc.py
 #!/usr/bin/env -S drgn -s vmlinux

 refcount = prog["tc_skb_ext_tc"].key.enabled.counter.value_()
 print(f"tc_skb_ext_tc reference count is {refcount}")

 # ./tc_skb_ext_tc.py
 tc_skb_ext_tc reference count is 0

 # tc filter add dev swp1 egress proto all handle 101 pref 1 flower src_mac 00:11:22:33:44:55 action drop
 # tc filter add dev swp1 egress proto all handle 102 pref 2 flower src_mac 00:11:22:33:44:55 l2_miss true action drop
 # tc filter add dev swp1 egress proto all handle 103 pref 3 flower src_mac 00:11:22:33:44:55 l2_miss false action drop

 # ./tc_skb_ext_tc.py
 tc_skb_ext_tc reference count is 2

 # tc filter replace dev swp1 egress proto all handle 102 pref 2 flower src_mac 00:01:02:03:04:05 l2_miss false action drop

 # ./tc_skb_ext_tc.py
 tc_skb_ext_tc reference count is 2

 # tc filter del dev swp1 egress proto all handle 103 pref 3 flower
 # tc filter del dev swp1 egress proto all handle 102 pref 2 flower
 # tc filter del dev swp1 egress proto all handle 101 pref 1 flower

 # ./tc_skb_ext_tc.py
 tc_skb_ext_tc reference count is 0
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Extend the 'FLOW_DISSECTOR_KEY_META' key with a new 'l2_miss' field and
populate it from a field with the same name in the tc skb extension.
This field is set by the bridge driver for packets that incur an FDB or
MDB miss.

The next patch will extend the flower classifier to be able to match on
layer 2 misses.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

For EVPN non-DF (Designated Forwarder) filtering we need to be able to
prevent decapsulated traffic from being flooded to a multi-homed host.
Filtering of multicast and broadcast traffic can be achieved using the
following flower filter:

 # tc filter add dev bond0 egress pref 1 proto all flower indev vxlan0 dst_mac 01:00:00:00:00:00/01:00:00:00:00:00 action drop

Unlike broadcast and multicast traffic, it is not currently possible to
filter unknown unicast traffic. The classification into unknown unicast
is performed by the bridge driver, but is not visible to other layers
such as tc.

Solve this by adding a new 'l2_miss' bit to the tc skb extension. Clear
the bit whenever a packet enters the bridge (received from a bridge port
or transmitted via the bridge) and set it if the packet did not match an
FDB or MDB entry. If there is no skb extension and the bit needs to be
cleared, then do not allocate one as no extension is equivalent to the
bit being cleared. The bit is not set for broadcast packets as they
never perform a lookup and therefore never incur a miss.

A bit that is set for every flooded packet would also work for the
current use case, but it does not allow us to differentiate between
registered and unregistered multicast traffic, which might be useful in
the future.

To keep the performance impact to a minimum, the marking of packets is
guarded by the 'tc_skb_ext_tc' static key. When 'false', the skb is not
touched and an skb extension is not allocated. Instead, only a
5 bytes nop is executed, as demonstrated below for the call site in
br_handle_frame().

Before the patch:

```
        memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  c37b09:       49 c7 44 24 28 00 00    movq   $0x0,0x28(%r12)
  c37b10:       00 00

        p = br_port_get_rcu(skb->dev);
  c37b12:       49 8b 44 24 10          mov    0x10(%r12),%rax
        memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  c37b17:       49 c7 44 24 30 00 00    movq   $0x0,0x30(%r12)
  c37b1e:       00 00
  c37b20:       49 c7 44 24 38 00 00    movq   $0x0,0x38(%r12)
  c37b27:       00 00
```

After the patch (when static key is disabled):

```
        memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
  c37c29:       49 c7 44 24 28 00 00    movq   $0x0,0x28(%r12)
  c37c30:       00 00
  c37c32:       49 8d 44 24 28          lea    0x28(%r12),%rax
  c37c37:       48 c7 40 08 00 00 00    movq   $0x0,0x8(%rax)
  c37c3e:       00
  c37c3f:       48 c7 40 10 00 00 00    movq   $0x0,0x10(%rax)
  c37c46:       00

#ifdef CONFIG_HAVE_JUMP_LABEL_HACK

static __always_inline bool arch_static_branch(struct static_key *key, bool branch)
{
        asm_volatile_goto("1:"
  c37c47:       0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)
        br_tc_skb_miss_set(skb, false);

        p = br_port_get_rcu(skb->dev);
  c37c4c:       49 8b 44 24 10          mov    0x10(%r12),%rax
```

Subsequent patches will extend the flower classifier to be able to match
on the new 'l2_miss' bit and enable / disable the static key when
filters that match on it are added / deleted.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>