Add test cases for VXLAN MDB, testing the control and data paths. Two
different sets of namespaces (i.e., ns{1,2}_v4 and ns{1,2}_v6) are used
in order to test VXLAN MDB with both IPv4 and IPv6 underlays,
respectively.
Example truncated output:
# ./test_vxlan_mdb.sh
[...]
Tests passed: 620
Tests failed: 0
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that the VXLAN MDB control and data paths are in place we can expose
the VXLAN MDB functionality to user space.
Set the VXLAN MDB net device operations to the appropriate functions,
thereby allowing the rtnetlink code to reach the VXLAN driver.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Integrate MDB support into the Tx path of the VXLAN driver, allowing it
to selectively forward IP multicast traffic according to the matched MDB
entry.
If MDB entries are configured (i.e., 'VXLAN_F_MDB' is set) and the
packet is an IP multicast packet, perform up to three different lookups
according to the following priority:
1. For an (S, G) entry, using {Source VNI, Source IP, Destination IP}.
2. For a (*, G) entry, using {Source VNI, Destination IP}.
3. For the catchall MDB entry (0.0.0.0 or ::), using the source VNI.
The catchall MDB entry is similar to the catchall FDB entry
(00:00:00:00:00:00) that is currently used to transmit BUM (broadcast,
unknown unicast and multicast) traffic. However, unlike the catchall FDB
entry, this entry is only used to transmit unregistered IP multicast
traffic that is not link-local. Therefore, when configured, the catchall
FDB entry will only transmit BULL (broadcast, unknown unicast,
link-local multicast) traffic.
The catchall MDB entry is useful in deployments where inter-subnet
multicast forwarding is used and not all the VTEPs in a tenant domain
are members in all the broadcast domains. In such deployments it is
advantageous to transmit BULL (broadcast, unknown unicast and link-local
multicast) and unregistered IP multicast traffic on different tunnels.
If the same tunnel was used, a VTEP only interested in IP multicast
traffic would also pull all the BULL traffic and drop it as it is not a
member in the originating broadcast domain [1].
If the packet did not match an MDB entry (or if the packet is not an IP
multicast packet), return it to the Tx path, allowing it to be forwarded
according to the FDB.
If the packet did match an MDB entry, forward it to the associated
remote VTEPs. However, if the entry is a (*, G) entry and the associated
remote is in INCLUDE mode, then skip over it as the source IP is not in
its source list (otherwise the packet would have matched on an (S, G)
entry). Similarly, if the associated remote is marked as BLOCKED (can
only be set on (S, G) entries), then skip over it as well as the remote
is in EXCLUDE mode and the source IP is in its source list.
[1] https://datatracker.ietf.org/doc/html/draft-ietf-bess-evpn-irb-mcast#section-2.6
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add an internal flag to indicate whether MDB entries are configured or
not. Set the flag after installing the first MDB entry and clear it
before deleting the last one.
The flag will be consulted by the data path which will only perform an
MDB lookup if the flag is set, thereby keeping the MDB overhead to a
minimum when the MDB is not used.
Another option would have been to use a static key, but it is global and
not per-device, unlike the current approach.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement MDB control path support, enabling the creation, deletion,
replacement and dumping of MDB entries in a similar fashion to the
bridge driver. Unlike the bridge driver, each entry stores a list of
remote VTEPs to which matched packets need to be replicated to and not a
list of bridge ports.
The motivating use case is the installation of MDB entries by a user
space control plane in response to received EVPN routes. As such, only
allow permanent MDB entries to be installed and do not implement
snooping functionality, avoiding a lot of unnecessary complexity.
Since entries can only be modified by user space under RTNL, use RTNL as
the write lock. Use RCU to ensure that MDB entries and remotes are not
freed while being accessed from the data path during transmission.
In terms of uAPI, reuse the existing MDB netlink interface, but add a
few new attributes to request and response messages:
* IP address of the destination VXLAN tunnel endpoint where the
multicast receivers reside.
* UDP destination port number to use to connect to the remote VXLAN
tunnel endpoint.
* VXLAN VNI Network Identifier to use to connect to the remote VXLAN
tunnel endpoint. Required when Ingress Replication (IR) is used and
the remote VTEP is not a member of originating broadcast domain
(VLAN/VNI) [1].
* Source VNI Network Identifier the MDB entry belongs to. Used only when
the VXLAN device is in external mode.
* Interface index of the outgoing interface to reach the remote VXLAN
tunnel endpoint. This is required when the underlay destination IP is
multicast (P2MP), as the multicast routing tables are not consulted.
All the new attributes are added under the 'MDBA_SET_ENTRY_ATTRS' nest
which is strictly validated by the bridge driver, thereby automatically
rejecting the new attributes.
[1] https://datatracker.ietf.org/doc/html/draft-ietf-bess-evpn-irb-mcast#section-3.2.2
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Given a packet and a remote destination, the function will take care of
encapsulating the packet and transmitting it to the destination.
Expose it so that it could be used in subsequent patches by the MDB code
to transmit a packet to the remote destination(s) stored in the MDB
entry.
It will allow us to keep the MDB code self-contained, not exposing its
data structures to the rest of the VXLAN driver.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the helpers out of the core C file to the private header so that
they could be used by the upcoming MDB code.
While at it, constify the second argument of vxlan_nla_get_addr().
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the upcoming VXLAN MDB implementation, the 0.0.0.0 and :: MDB entries
will act as catchall entries for unregistered IP multicast traffic in a
similar fashion to the 00:00:00:00:00:00 VXLAN FDB entry that is used to
transmit BUM traffic.
In deployments where inter-subnet multicast forwarding is used, not all
the VTEPs in a tenant domain are members in all the broadcast domains.
It is therefore advantageous to transmit BULL (broadcast, unknown
unicast and link-local multicast) and unregistered IP multicast traffic
on different tunnels. If the same tunnel was used, a VTEP only
interested in IP multicast traffic would also pull all the BULL traffic
and drop it as it is not a member in the originating broadcast domain
[1].
Prepare for this change by allowing the 0.0.0.0 group address in the
common rtnetlink MDB code and forbid it in the bridge driver. A similar
change is not needed for IPv6 because the common code only validates
that the group address is not the all-nodes address.
[1] https://datatracker.ietf.org/doc/html/draft-ietf-bess-evpn-irb-mcast#section-2.6
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the bridge driver registers handlers for MDB netlink
messages, making it impossible for other drivers to implement MDB
support.
As a preparation for VXLAN MDB support, move the MDB handlers out of the
bridge driver to the core rtnetlink code. The rtnetlink code will call
into individual drivers by invoking their previously added MDB net
device operations.
Note that while the diffstat is large, the change is mechanical. It
moves code out of the bridge driver to rtnetlink code. Also note that a
similar change was made in 2012 with commit 77162022ab ("net: add
generic PF_BRIDGE:RTM_ FDB hooks") that moved FDB handlers out of the
bridge driver to the core rtnetlink code.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement the previously added MDB net device operations in the bridge
driver so that they could be invoked by core rtnetlink code in the next
patch.
The operations are identical to the existing br_mdb_{dump,add,del}
functions. The '_new' suffix will be removed in the next patch. The
functions are re-implemented in this patch to make the conversion in the
next patch easier to review.
Add dummy implementations when 'CONFIG_BRIDGE_IGMP_SNOOPING' is
disabled, so that an error will be returned to user space when it is
trying to add or delete an MDB entry. This is consistent with existing
behavior where the bridge driver does not even register rtnetlink
handlers for RTM_{NEW,DEL,GET}MDB messages when this Kconfig option is
disabled.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add MDB net device operations that will be invoked by rtnetlink code in
response to received RTM_{NEW,DEL,GET}MDB messages. Subsequent patches
will implement these operations in the bridge and VXLAN drivers.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Siddharth Vadapalli says:
====================
Add J784S4 CPSW9G NET Bindings
This series cleans up the bindings by reordering the compatibles, followed
by adding the bindings for CPSW9G instance of CPSW Ethernet Switch on TI's
J784S4 SoC.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Update bindings for TI K3 J784S4 SoC which contains 9 ports (8 external
ports) CPSW9G module and add compatible for it.
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reorder compatibles to follow alphanumeric order.
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extended performance counter stats in 'ethtool -S <interface>' output
for MANA VF to facilitate troubleshooting.
Tested-on: Ubuntu22
Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The rtl8365mb was using a fixed MTU size of 1536, which was probably
inspired by the rtl8366rb's initial frame size. However, unlike that
family, the rtl8365mb family can specify the max frame size in bytes,
rather than in fixed steps.
DSA calls change_mtu for the CPU port once the max MTU value among the
ports changes. As the max frame size is defined globally, the switch
is configured only when the call affects the CPU port.
The available specifications do not directly define the max supported
frame size, but it mentions a 16k limit. This driver will use the 0x3FFF
limit as it is used in the vendor API code. However, the switch sets the
max frame size to 16368 bytes (0x3FF0) after it resets.
change_mtu uses MTU size, or ethernet payload size, while the switch
works with frame size. The frame size is calculated considering the
ethernet header (14 bytes), a possible 802.1Q tag (4 bytes), the payload
size (MTU), and the Ethernet FCS (4 bytes). The CPU tag (8 bytes) is
consumed before the switch enforces the limit.
During setup, the driver will use the default 1500-byte MTU of DSA to
set the maximum frame size. The current sum will be
VLAN_ETH_HLEN+1500+ETH_FCS_LEN, which results in 1522 bytes. Although
it is lower than the previous initial value of 1536 bytes, the driver
will increase the frame size for a larger MTU. However, if something
requires more space without increasing the MTU, such as QinQ, we would
need to add the extra length to the rtl8365mb_port_change_mtu() formula.
MTU was tested up to 2018 (with 802.1Q) as that is as far as mt7620
(where rtl8367s is stacked) can go. The register was manually
manipulated byte-by-byte to ensure the MTU to frame size conversion was
correct. For frames without 802.1Q tag, the frame size limit will be 4
bytes over the required size.
There is a jumbo register, enabled by default at 6k frame size.
However, the jumbo settings do not seem to limit nor expand the maximum
tested MTU (2018), even when jumbo is disabled. More tests are needed
with a device that can handle larger frames.
Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is preferred to use typed property access functions (i.e.
of_property_read_<type> functions) rather than low-level
of_get_property/of_find_property functions for reading properties.
Convert reading boolean properties to of_property_read_bool().
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are no users of nfcmrvl platform_data struct outside of the
driver and none will be added, so move it into the driver.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
GPY2xx devices need 3 seconds to fully switch out of loopback mode
before it can safely re-enter loopback mode. Implement timeout mechanism
to guarantee 3 seconds waited before re-enter loopback mode.
Signed-off-by: Xu Liang <lxu@maxlinear.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Louis Peens says:
====================
nfp: flower: add support for multi-zone conntrack
This series add changes to support offload of connection tracking across
multiple zones. Previously the driver only supported offloading of a
single goto_chain, spanning a single zone. This was implemented by
merging a pre_ct rule, post_ct rule and the nft rule. This series
provides updates to let the original post_ct rule act as the new pre_ct
rule for a next set of merges if it contains another goto and
conntrack action. In pseudo-tc rule format this adds support for:
ingress chain 0 proto ip flower
action ct zone 1 pipe action goto 1
ingress chain 1 proto ip flower ct_state +tr+new ct_zone 1
action ct_clear pipe action ct zone 2 pipe action goto 2
ingress chain 1 proto ip flower ct_state +tr+est ct_zone 1
action ct_clear pipe action ct zone 2 pipe action goto 2
ingress chain 2 proto ip flower ct_state +tr+new ct_zone 2
action mirred egress redirect dev ...
ingress chain 2 proto ip flower ct_state +tr+est ct_zone 2
action mirred egress redirect dev ...
This can continue for up to a maximum of 4 zone recirculations.
The first few patches are some smaller preparation patches while the
last one introduces the functionality.
====================
Link: https://lore.kernel.org/r/20230314063610.10544-1-louis.peens@corigine.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If goto_chain action present in the post ct flow rule, merge flow rules
in this ct-zone, create a new pre_ct entry as the pre ct flow rule of
next ct-zone, but do not offload merged flow rules to firmware. Repeat
the process in the next ct-zone until no goto_chain action present in
the post ct flow rule in a certain ct-zone, merged all the flow rules.
Offload to firmware finally.
Signed-off-by: Wentao Jia <wentao.jia@corigine.com>
Acked-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The fixed number of offload flow rule is only supported scenario of one
ct zone, in the scenario of multiple ct zones, dynamic number and more
number of offload flow rules are required. In order to support scenario
of multiple ct zones, parameter num_rules is added for to offload flow
rules
Signed-off-by: Wentao Jia <wentao.jia@corigine.com>
Acked-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The chain_index has different means in pre ct entry and post ct entry.
In pre ct entry, it means chain index, but in post ct entry, it means
goto chain index, it is confused.
chain_index and goto_chain_index may be present in one flow rule, It
cannot be distinguished by one field chain_index, both chain_index
and goto_chain_index are required in the follow-up patch to support
multiple ct zones
Another field goto_chain_index is added to record the goto chain index.
If no goto action in post ct entry, goto_chain_index is 0.
Signed-off-by: Wentao Jia <wentao.jia@corigine.com>
Acked-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
'ct_clear' action only or no ct action is supported for 'post_ct_flow'.
But in scenario of multiple ct zones, one non 'ct_clear' ct action or
more ct actions, including 'ct_clear action', may be present in one flow
rule. If ct state match key is 'ct_established', the flow rule is still
expected to be classified as 'post_ct_flow'. Check ct status first in
function "is_post_ct_flow" to achieve this.
Signed-off-by: Wentao Jia <wentao.jia@corigine.com>
Acked-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In the scenario of multiple ct zones, ct state key match and ct action
is present in one flow rule, the flow rule is classified to post_ct_flow
in design.
There is no ct state key match for pre ct flow, the judging condition
is added to function "is_pre_ct_flow".
Chain_index is another field for judging which flows are pre ct flow
If chain_index not 0, the flow is not pre ct flow.
Signed-off-by: Wentao Jia <wentao.jia@corigine.com>
Acked-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
CT action is a special case different from other actions, CT clear action
is not required when get ct action, but this case is not considered.
If CT clear action in the flow rule, skip the CT clear action when get ct
action, return the first ct action that is not a CT clear action
Signed-off-by: Wentao Jia <wentao.jia@corigine.com>
Acked-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Saeed Mahameed says:
====================
mlx5-updates-2023-03-13
1) Trivial cleanup patches
2) By Sandipan Patra: Implement thermal zone to report NIC temperature
3) Adham Faris, Improves devlink health diagnostics for netdev objects
4) From Maor, Enable TC offload for egress and engress MACVLAN over bond
5) From Gal, add devlink hairpin queues parameters to replace debugfs
as was discussed in [1]:
[1] https://lore.kernel.org/all/20230111194608.7f15b9a1@kernel.org/
====================
Link: https://lore.kernel.org/all/20230314054234.267365-1-saeed@kernel.org/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently RQ health diagnostics doesn't inform the user whether an RQ
is an XSK RQ or not.
Address this, by adding XSK state flag to RQ SW state enum in core/en.h.
XSK will be '1' if current RQ is an XSK RQ, and it will be '0' if it's
not.
In this example below, it can be seen that XSK field value is '1' since
xdpsock program have been attached to channel 0 before issuing the
devlink query command:
$ devlink health diagnose auxiliary/mlx5_core.eth.0/65535 reporter rx
Output:
=======================================================================
Common config:
RQ:
type: 2 stride size: 4096 size: 16 ts_format: FRC
CQ:
stride size: 64 size: 1024
RQs:
channel ix: 0 rqn: 4236 HW state: 1 WQE counter: 15 posted WQEs: 15 cc: 15
SW State:
enabled: 1 recovering: 0 am: 1 no_csum_complete: 1 csum_full: 0 mini_cqe_hw_stridx: 1 shampo: 0 mini_cqe_enhanced: 0 xsk: 1
CQ:
cqn: 1085 HW status: 0 ci: 0 size: 1024
EQ:
eqn: 7 irqn: 32 vecidx: 0 ci: 5 size: 2048
ICOSQ:
sqn: 4229 HW state: 1 cc: 158 pc: 158 WQE size: 2048
CQ:
cqn: 1080 cc: 1 size: 2048
Signed-off-by: Adham Faris <afaris@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20230314054234.267365-10-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Dynamic interrupt moderation RQ and SQ feature represented by
MLX5E_RQ_STATE_AM and MLX5E_SQ_STATE_AM enums respectively, is not
consistent with the feature naming in the driver, and with the formal
feature and library names.
Hence, change MLX5E_RQ_STATE_AM and MLX5E_SQ_STATE_AM enum type names in
core/en.h to MLX5E_RQ_STATE_DIM and MLX5E_SQ_STATE_DIM respectively.
Signed-off-by: Adham Faris <afaris@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20230314054234.267365-7-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If driver teardown is called while PCI is turned off, there is a race
between health recovery and teardown. If health recovery already started
it will wait 60 sec trying to see if PCI gets back and it can recover,
but actually there is no need to wait anymore once teardown was called.
Use the MLX5_BREAK_FW_WAIT flag which is set on driver teardown to break
waiting for PCI up.
Signed-off-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20230314054234.267365-3-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When shutdown or remove callbacks are called the driver sets the flag
MLX5_BREAK_FW_WAIT, to stop waiting for FW as teardown was called. There
is no need to clear the bit as once shutdown or remove were called as
there is no way back, the driver is going down. Furthermore, if not
cleared the flag can be used also in other loops where we may wait while
teardown was already called.
Use test_bit() instead of test_and_clear_bit() as there is no need to
clear the flag.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20230314054234.267365-2-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
