The refcount wr_tx_refcnt may cause cache thrashing problems among
cores and we can use percpu ref to mitigate this issue here. We
gain some performance improvement with percpu ref here on our
customized smc-r verion. Applying cache alignment may also mitigate
this problem but it seem more reasonable to use percpu ref here.
We can also replace wr_reg_refcnt with one percpu reference like
wr_tx_refcnt.

redis-benchmark on smc-r with atomic wr_tx_refcnt:
SET: 525707.06 requests per second, p50=0.087 msec
GET: 554877.38 requests per second, p50=0.087 msec

redis-benchmark on the percpu_ref version:
SET: 540482.06 requests per second, p50=0.087 msec
GET: 570711.12 requests per second, p50=0.079 msec

Cases are like "redis-benchmark -h x.x.x.x -q -t set,get -P 1 -n
5000000 -c 50 -d 10 --threads 4".
Signed-off-by: Kai Shen <KaiShen@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Eric Dumazet says:

====================
inet: better const qualifier awareness

inet_sk() can be changed to propagate const qualifier,
thanks to container_of_const()

Following patches in this series add more const qualifiers.

Other helpers like tcp_sk(), udp_sk(), raw_sk(), ... will be handled
in following series.
====================
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Now both raw_v4_match() and raw_v6_match() accept a const socket,
raw_lookup() can do the same to clarify its role.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This clarifies raw_v4_match() intent.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This clarifies raw_v6_match() intent.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This clarifies __udp_v6_is_mcast_sock() intent.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

inet6_mc_check() is essentially a read-only function.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This clarifies __udp_is_mcast_sock() intent.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This clarifies ip_mc_sf_allow() intent.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can change inet_sk() to propagate const qualifier of its argument.

This should avoid some potential errors caused by accidental
(const -> not_const) promotion.

Other helpers like tcp_sk(), udp_sk(), raw_sk() will be handled
in separate patch series.

v2: use container_of_const() as advised by Jakub and Linus
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/netdev/20230315142841.3a2ac99a@kernel.org/
Link: https://lore.kernel.org/netdev/CAHk-=wiOf12nrYEF2vJMcucKjWPN-Ns_SW9fA7LwST_2Dzp7rw@mail.gmail.com/Signed-off-by: David S. Miller <davem@davemloft.net>

Chuck wanted to put the UAPI header in linux/net/ which seems
reasonable, allow genetlink families to choose the location.
It doesn't really matter for non-C-like languages.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Devlink is quite complex but put in the very basics so we can
incrementally fill in the commands as needed.

$ ./tools/net/ynl/cli.py --spec Documentation/netlink/specs/devlink.yaml \
    --dump get

[{'bus-name': 'netdevsim',
  'dev-name': 'netdevsim1',
  'dev-stats': {'reload-stats': {'reload-action-info': {'reload-action': 1,
                                                        'reload-action-stats': {'reload-stats-entry': [{'reload-stats-limit': 0,
                                                                                                        'reload-stats-value': 0}]}}},
                'remote-reload-stats': {'reload-action-info': {'reload-action': 2,
                                                               'reload-action-stats': {'reload-stats-entry': [{'reload-stats-limit': 0,
                                                                                                               'reload-stats-value': 0},
                                                                                                              {'reload-stats-limit': 1,
                                                                                                               'reload-stats-value': 0}]}}}},
  'reload-failed': 0}]
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Eric Dumazet says:

====================
net/packet: KCSAN awareness

This series is based on one syzbot report [1]

Seven 'flags/booleans' are converted to atomic bit variant.

po->xmit and po->tp_tstamp accesses get annotations.

[1]
BUG: KCSAN: data-race in packet_rcv / packet_setsockopt

read-write to 0xffff88813dbe84e4 of 1 bytes by task 12312 on cpu 0:
packet_setsockopt+0xb77/0xe60 net/packet/af_packet.c:3900
__sys_setsockopt+0x212/0x2b0 net/socket.c:2252
__do_sys_setsockopt net/socket.c:2263 [inline]
__se_sys_setsockopt net/socket.c:2260 [inline]
__x64_sys_setsockopt+0x62/0x70 net/socket.c:2260
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x2b/0x70 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

read to 0xffff88813dbe84e4 of 1 bytes by task 1911 on cpu 1:
packet_rcv+0x4b1/0xa40 net/packet/af_packet.c:2187
deliver_skb net/core/dev.c:2189 [inline]
dev_queue_xmit_nit+0x3a9/0x620 net/core/dev.c:2259
xmit_one+0x71/0x2a0 net/core/dev.c:3586
dev_hard_start_xmit+0x72/0x120 net/core/dev.c:3606
__dev_queue_xmit+0x91c/0x11c0 net/core/dev.c:4256
dev_queue_xmit include/linux/netdevice.h:3008 [inline]
neigh_hh_output include/net/neighbour.h:530 [inline]
neigh_output include/net/neighbour.h:544 [inline]
ip6_finish_output2+0x9e9/0xc30 net/ipv6/ip6_output.c:134
__ip6_finish_output net/ipv6/ip6_output.c:195 [inline]
ip6_finish_output+0x395/0x4f0 net/ipv6/ip6_output.c:206
NF_HOOK_COND include/linux/netfilter.h:291 [inline]
ip6_output+0x10e/0x210 net/ipv6/ip6_output.c:227
dst_output include/net/dst.h:445 [inline]
ip6_local_out+0x60/0x80 net/ipv6/output_core.c:161
ip6tunnel_xmit include/net/ip6_tunnel.h:161 [inline]
udp_tunnel6_xmit_skb+0x321/0x4a0 net/ipv6/ip6_udp_tunnel.c:109
send6+0x2ed/0x3b0 drivers/net/wireguard/socket.c:152
wg_socket_send_skb_to_peer+0xbb/0x120 drivers/net/wireguard/socket.c:178
wg_packet_create_data_done drivers/net/wireguard/send.c:251 [inline]
wg_packet_tx_worker+0x142/0x360 drivers/net/wireguard/send.c:276
process_one_work+0x3d3/0x720 kernel/workqueue.c:2289
worker_thread+0x618/0xa70 kernel/workqueue.c:2436
kthread+0x1a9/0x1e0 kernel/kthread.c:376
ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:306
====================
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Not only this removes some READ_ONCE()/WRITE_ONCE(),
this also removes one integer.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Instead of consuming 32 bits for po->running, use
one available bit in po->flags.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

po->has_vnet_hdr can be read locklessly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tp_loss can be read locklessly.

Convert it to an atomic flag to avoid races.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is to use existing space in po->flags, and reclaim
the storage used by the non atomic bit fields.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tp_tstamp is read locklessly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

po->auxdata can be read while another thread
is changing its value, potentially raising KCSAN splat.

Convert it to PACKET_SOCK_AUXDATA flag.

Fixes: 8dc41944 ("[PACKET]: Add optional checksum computation for recvmsg")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

syzbot/KCAN reported that po->origdev can be read
while another thread is changing its value.

We can avoid this splat by converting this field
to an actual bit.

Following patches will convert remaining 1bit fields.

Fixes: 80feaacb ("[AF_PACKET]: Add option to return orig_dev to userspace.")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

po->xmit can be set from setsockopt(PACKET_QDISC_BYPASS),
while read locklessly.

Use READ_ONCE()/WRITE_ONCE() to avoid potential load/store
tearing issues.

Fixes: d346a3fa ("packet: introduce PACKET_QDISC_BYPASS socket option")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Praveen Kaligineedi says:

====================
gve: Add XDP support for GQI-QPL format

Adding support for XDP DROP, PASS, TX, REDIRECT for GQI QPL format.
Add AF_XDP zero-copy support.

When an XDP program is installed, dedicated TX queues are created to
handle XDP traffic. The user needs to ensure that the number of
configured TX queues is equal to the number of configured RX queues; and
the number of TX/RX queues is less than or equal to half the maximum
number of TX/RX queues.

The XDP traffic from AF_XDP sockets and from other NICs (arriving via
XDP_REDIRECT) will also egress through the dedicated XDP TX queues.

Although these changes support AF_XDP socket in zero-copy mode, there is
still a copy happening within the driver between XSK buffer pool and QPL
bounce buffers in GQI-QPL format.

The following example demonstrates how the XDP packets are mapped to
TX queues:

Example configuration:
Max RX queues : 2N, Max TX queues : 2N
Configured RX queues : N, Configured TX queues : N

TX queue mapping:
TX queues with queue id 0,...,N-1 will handle traffic from the stack.
TX queues with queue id N,...,2N-1 will handle XDP traffic.

For the XDP packets transmitted using XDP_TX action:
<Egress TX queue id> = N + <Ingress RX queue id>

For the XDP packets that arrive from other NICs via XDP_REDIRECT action:
<Egress TX queue id> = N + ( smp_processor_id % N )

For AF_XDP zero-copy mode:
<Egress TX queue id> = N + <AF_XDP TX queue id>

Changes in v2:
- Removed gve_close/gve_open when adding XDP dedicated queues. Instead
we add and register additional TX queues when the XDP program is
installed. If the allocation/registration fails we return error and do
not install the XDP program. Added a new patch to enable adding TX queues
without gve_close/gve_open
- Removed xdp tx spin lock from this patch. It is needed for XDP_REDIRECT
support as both XDP_REDIRECT and XDP_TX traffic share the dedicated XDP
queues. Moved the code to add xdp tx spinlock to the subsequent patch
that adds XDP_REDIRECT support.
- Added netdev_err when the user tries to set rx/tx queues to the values
not supported when XDP is enabled.
- Removed rcu annotation for xdp_prog. We disable the napi prior to
adding/removing the xdp_prog and reenable it after the program has
been installed for all the queues.
- Ring the tx doorbell once for napi instead of every XDP TX packet.
- Added a new helper function for freeing the FIFO buffer
- Unregister xdp rxq for all the queues when the registration
fails during XDP program installation
- Register xsk rxq only when XSK buff pool is enabled
- Removed code accessing internal xsk_buff_pool fields
- Removed sleep driven code when disabling XSK buff pool. Disable
napi and re-enable it after disabling XSK pool.
- Make sure that we clean up dma mappings on XSK pool disable
- Use napi_if_scheduled_mark_missed to avoid unnecessary napi move
to the CPU calling ndo_xsk_wakeup()

Changes in v3:
- Padding bytes are used if the XDP TX packet headers do not
fit at tail of TX FIFO. Taking these padding bytes into account
while checking if enough space is available in TX FIFO.

Changes in v4:
- Turn on the carrier based on the link status synchronously rather
than asynchronously when XDP is installed/uninstalled
- Set the supported flags in net_device.xdp_features
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Adding AF_XDP zero-copy support.

Note: Although these changes support AF_XDP socket in zero-copy
mode, there is still a copy happening within the driver between
XSK buffer pool and QPL bounce buffers in GQI-QPL format.
In GQI-QPL queue format, the driver needs to allocate a fixed size
memory, the size specified by vNIC device, for RX/TX and register this
memory as a bounce buffer with the vNIC device when a queue is
created. The number of pages in the bounce buffer is limited and the
pages need to be made available to the vNIC by copying the RX data out
to prevent head-of-line blocking. Therefore, we cannot pass the XSK
buffer pool to the vNIC.

The number of copies on RX path from the bounce buffer to XSK buffer is 2
for AF_XDP copy mode (bounce buffer -> allocated page frag -> XSK buffer)
and 1 for AF_XDP zero-copy mode (bounce buffer -> XSK buffer).

This patch contains the following changes:
1) Enable and disable XSK buffer pool
2) Copy XDP packets from QPL bounce buffers to XSK buffer on rx
3) Copy XDP packets from XSK buffer to QPL bounce buffers and
   ring the doorbell as part of XDP TX napi poll
4) ndo_xsk_wakeup callback support
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch contains the following changes:
1) Support for XDP REDIRECT action on rx
2) ndo_xdp_xmit callback support

In GQI-QPL queue format, the driver needs to allocate a fixed size
memory, the size specified by vNIC device, for RX/TX and register this
memory as a bounce buffer with the vNIC device when a queue is created.
The number of pages in the bounce buffer is limited and the pages need to
be made available to the vNIC by copying the RX data out to prevent
head-of-line blocking. The XDP_REDIRECT packets are therefore immediately
copied to a newly allocated page.
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for XDP PASS, DROP and TX actions.

This patch contains the following changes:
1) Support installing/uninstalling XDP program
2) Add dedicated XDP TX queues
3) Add support for XDP DROP action
4) Add support for XDP TX action
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Changes to enable adding and removing TX queues without calling
gve_close() and gve_open().

Made the following changes:
1) priv->tx, priv->rx and priv->qpls arrays are allocated based on
   max tx queues and max rx queues
2) Changed gve_adminq_create_tx_queues(), gve_adminq_destroy_tx_queues(),
gve_tx_alloc_rings() and gve_tx_free_rings() functions to add/remove a
subset of TX queues rather than all the TX queues.
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds/modifies helper functions needed to add XDP
support.
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Jeroen de Borst <jeroendb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Eric Dumazet says:

====================
net: annotate lockless accesses to sk_err[_soft]

This patch series is inspired by yet another syzbot report.

Most poll() handlers are lockless and read sk->sk_err
while other cpus can change it.

Add READ_ONCE/WRITE_ONCE() to major/usual offenders.

More to come later.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

unix_poll() and unix_dgram_poll() read sk->sk_err
without any lock held.

Add relevant READ_ONCE()/WRITE_ONCE() annotations.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mptcp_poll() reads sk->sk_err without socket lock held/owned.

Add READ_ONCE() and WRITE_ONCE() to avoid load/store tearing.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_poll() reads sk->sk_err without socket lock held/owned.

We should used READ_ONCE() here, and update writers
to use WRITE_ONCE().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This field can be read/written without lock synchronization.

tcp and dccp have been handled in different patches.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This field can be read/written without lock synchronization.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This field can be read/written without lock synchronization.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Setting timestamp filter was explicitly disabled on vlan devices in
containers because it might affect other processes on the host. But it's
absolutely legit in case when real device is in the same namespace.

Fixes: 873017af ("vlan: disable SIOCSHWTSTAMP in container")
Signed-off-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>

Russell King (Oracle) says:

====================
Minor fixes for pcs_get_state() implementations

This series contains a number fixes for minor issues with some
pcs_get_state() implementations, particualrly for the phylink
state->an_enabled member. As they are minor, I'm suggesting we
queue them in net-next as there is follow-on work for these, and
there is no urgency for them to be in -rc.

Just like phylib, state->advertising's Autoneg bit is a copy of
state->an_enabled, and thus it is my intention to remove
state->an_enabled from phylink to simplify things.

This series gets rid of state->an_enabled assignments or
reporting that should never have been there.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

an_enabled will be going away, and in any case, pcs_get_state() should
not be updating this member. Remove the print.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Phylink does not want the current state of the link when reading the
PCS link state - it wants the latched state. Don't double-read the
MII status register. Phylink will re-read as necessary to capture
transient link-down events as of dbae3388 ("net: phylink: Force
retrigger in case of latched link-fail indicator").

The above referenced commit is a dependency for this change, and thus
this change should not be backported to any kernel that does not
contain the above referenced commit.

Fixes: fcb26bd2 ("net: phy: Add Synopsys DesignWare XPCS MDIO module")
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Ido Schimmel says:

====================
vxlan: Add MDB support

tl;dr
=====

This patchset implements MDB support in the VXLAN driver, allowing it to
selectively forward IP multicast traffic to VTEPs with interested
receivers instead of flooding it to all the VTEPs as BUM. The motivating
use case is intra and inter subnet multicast forwarding using EVPN
[1][2], which means that MDB entries are only installed by the user
space control plane and no snooping is implemented, thereby avoiding a
lot of unnecessary complexity in the kernel.

Background
==========

Both the bridge and VXLAN drivers have an FDB that allows them to
forward Ethernet frames based on their destination MAC addresses and
VLAN/VNI. These FDBs are managed using the same PF_BRIDGE/RTM_*NEIGH
netlink messages and bridge(8) utility.

However, only the bridge driver has an MDB that allows it to selectively
forward IP multicast packets to bridge ports with interested receivers
behind them, based on (S, G) and (*, G) MDB entries. When these packets
reach the VXLAN driver they are flooded using the "all-zeros" FDB entry
(00:00:00:00:00:00). The entry either includes the list of all the VTEPs
in the tenant domain (when ingress replication is used) or the multicast
address of the BUM tunnel (when P2MP tunnels are used), to which all the
VTEPs join.

Networks that make heavy use of multicast in the overlay can benefit
from a solution that allows them to selectively forward IP multicast
traffic only to VTEPs with interested receivers. Such a solution is
described in the next section.

Motivation
==========

RFC 7432 [3] defines a "MAC/IP Advertisement route" (type 2) [4] that
allows VTEPs in the EVPN network to advertise and learn reachability
information for unicast MAC addresses. Traffic destined to a unicast MAC
address can therefore be selectively forwarded to a single VTEP behind
which the MAC is located.

The same is not true for IP multicast traffic. Such traffic is simply
flooded as BUM to all VTEPs in the broadcast domain (BD) / subnet,
regardless if a VTEP has interested receivers for the multicast stream
or not. This is especially problematic for overlay networks that make
heavy use of multicast.

The issue is addressed by RFC 9251 [1] that defines a "Selective
Multicast Ethernet Tag Route" (type 6) [5] which allows VTEPs in the
EVPN network to advertise multicast streams that they are interested in.
This is done by having each VTEP suppress IGMP/MLD packets from being
transmitted to the NVE network and instead communicate the information
over BGP to other VTEPs.

The draft in [2] further extends RFC 9251 with procedures to allow
efficient forwarding of IP multicast traffic not only in a given subnet,
but also between different subnets in a tenant domain.

The required changes in the bridge driver to support the above were
already merged in merge commit 8150f0cf ("Merge branch
'bridge-mcast-extensions-for-evpn'"). However, full support entails MDB
support in the VXLAN driver so that it will be able to selectively
forward IP multicast traffic only to VTEPs with interested receivers.
The implementation of this MDB is described in the next section.

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

The user interface is extended to allow user space to specify the
destination VTEP(s) and related parameters. Example usage:

 # bridge mdb add dev vxlan0 port vxlan0 grp 239.1.1.1 permanent dst 198.51.100.1
 # bridge mdb add dev vxlan0 port vxlan0 grp 239.1.1.1 permanent dst 192.0.2.1

 $ bridge -d -s mdb show
 dev vxlan0 port vxlan0 grp 239.1.1.1 permanent filter_mode exclude proto static dst 192.0.2.1    0.00
 dev vxlan0 port vxlan0 grp 239.1.1.1 permanent filter_mode exclude proto static dst 198.51.100.1    0.00

Since the MDB is fully managed by user space and since snooping is not
implemented, only permanent entries can be installed and temporary
entries are rejected by the kernel.

The netlink interface is extended with a few new attributes in the
RTM_NEWMDB / RTM_DELMDB request messages:

[ struct nlmsghdr ]
[ struct br_port_msg ]
[ MDBA_SET_ENTRY ]
	struct br_mdb_entry
[ MDBA_SET_ENTRY_ATTRS ]
	[ MDBE_ATTR_SOURCE ]
		struct in_addr / struct in6_addr
	[ MDBE_ATTR_SRC_LIST ]
		[ MDBE_SRC_LIST_ENTRY ]
			[ MDBE_SRCATTR_ADDRESS ]
				struct in_addr / struct in6_addr
		[ ...]
	[ MDBE_ATTR_GROUP_MODE ]
		u8
	[ MDBE_ATTR_RTPORT ]
		u8
	[ MDBE_ATTR_DST ]	// new
		struct in_addr / struct in6_addr
	[ MDBE_ATTR_DST_PORT ]	// new
		u16
	[ MDBE_ATTR_VNI ]	// new
		u32
	[ MDBE_ATTR_IFINDEX ]	// new
		s32
	[ MDBE_ATTR_SRC_VNI ]	// new
		u32

RTM_NEWMDB / RTM_DELMDB responses and notifications are extended with
corresponding attributes.

One MDB entry that can be installed in the VXLAN MDB, but not in the
bridge MDB is the catchall entry (0.0.0.0 / ::). It is used to transmit
unregistered multicast traffic that is not link-local and is especially
useful when inter-subnet multicast forwarding is required. See patch #12
for a detailed explanation and motivation. It is similar to the
"all-zeros" FDB entry that can be installed in the VXLAN FDB, but not
the bridge FDB.

"added_by_star_ex" entries
--------------------------

The bridge driver automatically installs (S, G) MDB port group entries
marked as "added_by_star_ex" whenever it detects that an (S, G) entry
can prevent traffic from being forwarded via a port associated with an
EXCLUDE (*, G) entry. The bridge will add the port to the port group of
the (S, G) entry, thereby creating a new port group entry. The
complexity associated with these entries is not trivial, but it needs to
reside in the bridge driver because it automatically installs MDB
entries in response to snooped IGMP / MLD packets.

The same in not true for the VXLAN MDB which is entirely managed by user
space who is fully capable of forming the correct replication lists on
its own. In addition, the complexity associated with the
"added_by_star_ex" entries in the VXLAN driver is higher compared to the
bridge: Whenever a remote VTEP is added to the catchall entry, it needs
to be added to all the existing MDB entries, as such a remote requested
all the multicast traffic to be forwarded to it. Similarly, whenever an
(*, G) or (S, G) entry is added, all the remotes associated with the
catchall entry need to be added to it.

Given the above, this patchset does not implement support for such
entries.  One argument against this decision can be that in the future
someone might want to populate the VXLAN MDB in response to decapsulated
IGMP / MLD packets and not according to EVPN routes. Regardless of my
doubts regarding this possibility, it can be implemented using a new
VXLAN device knob that will also enable the "added_by_star_ex"
functionality.

Testing
=======

Tested using existing VXLAN and MDB selftests under "net/" and
"net/forwarding/". Added a dedicated selftest in the last patch.

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

Patches #1-#3 are small preparations in the bridge driver. I plan to
submit them separately together with an MDB dump test case.

Patches #4-#6 are additional preparations centered around the extraction
of the MDB netlink handlers from the bridge driver to the common
rtnetlink code. This allows reusing the existing MDB netlink messages
for the configuration of the VXLAN MDB.

Patches #7-#9 include more small preparations in the common rtnetlink
code and the VXLAN driver.

Patch #10 implements the MDB control path in the VXLAN driver, which
will allow user space to create, delete, replace and dump MDB entries.

Patches #11-#12 implement the MDB data path in the VXLAN driver,
allowing it to selectively forward IP multicast traffic according to the
matched MDB entry.

Patch #13 finally enables MDB support in the VXLAN driver.

iproute2 patches can be found here [6].

Note that in order to fully support the specifications in [1] and [2],
additional functionality is required from the data path. However, it can
be achieved using existing kernel interfaces which is why it is not
described here.

Changelog
=========

Since v1 [7]:

Patch #9: Use htons() in 'case' instead of ntohs() in 'switch'.

Since RFC [8]:

Patch #3: Use NL_ASSERT_DUMP_CTX_FITS().
Patch #3: memset the entire context when moving to the next device.
Patch #3: Reset sequence counters when moving to the next device.
Patch #3: Use NL_SET_ERR_MSG_ATTR() in rtnl_validate_mdb_entry().
Patch #7: Remove restrictions regarding mixing of multicast and unicast
remote destination IPs in an MDB entry. While such configuration does
not make sense to me, it is no forbidden by the VXLAN FDB code and does
not crash the kernel.
Patch #7: Fix check regarding all-zeros MDB entry and source.
Patch #11: New patch.

[1] https://datatracker.ietf.org/doc/html/rfc9251
[2] https://datatracker.ietf.org/doc/html/draft-ietf-bess-evpn-irb-mcast
[3] https://datatracker.ietf.org/doc/html/rfc7432
[4] https://datatracker.ietf.org/doc/html/rfc7432#section-7.2
[5] https://datatracker.ietf.org/doc/html/rfc9251#section-9.1
[6] https://github.com/idosch/iproute2/commits/submit/mdb_vxlan_rfc_v1
[7] https://lore.kernel.org/netdev/20230313145349.3557231-1-idosch@nvidia.com/
[8] https://lore.kernel.org/netdev/20230204170801.3897900-1-idosch@nvidia.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>