Pass originating device to drivers offloading ct connection so devices will filter the tuples and offload them more efficiently

[Bodong Wang]

* Explain the feature

    Currently, drivers register to a ct zone that can be shared by multiple
    devices. This can be inefficient for the driver to offload, as it
    needs to handle all the cases where the tuple can come from,
    instead of where it's most likely will arive from.

    For example, consider the following tc rules:
    tc filter add dev dev1 ... flower action ct commit zone 5 \
       action mirred egress redirect dev dev2

    tc filter add dev dev2 ... flower action ct zone 5 \
       action goto chain chain 2
    tc filter add dev dev2 ... flower ct_state +trk+est ... \
       action mirred egress redirect dev dev1

    Both dev2 and dev1 register to the zone 5 flow table (created
    by act_ct). A tuple originating on dev1, going to dev2, will
    be offloaded to both devices, and both will need to offload
    both directions, resulting in 4 total rules. The traffic
    will only hit originiating tuple on dev1, and reply tuple
    on dev2.

    By passing the originating device that created the connection
    with the tuple, dev1 can choose to offload only the originating
    tuple, and dev2 only the reply tuple. Resulting in a more
    efficient offload.

    The first patch adds an act_ct nf conntrack extension, to
    temporarily store the originiating device from the skb before
    offloading the connection once the connection is established.
    Once sent to offload, it fills the tuple originating device.

    The second patch get this information from tuples
    which pass in openvswitch.

    The third patch is Mellanox driver ct offload implementation using
    this information to provide a hint to firmware of where this
    offloaded tuple packets will arrive from (LOCAL or UPLINK port),
    and thus increase insertion rate.

* How to test

Create OVS bridge with 2 devices mlx5 rep devices.
Enable HW offload and configure regular connection tracking OpenFlow rules:

e.g:
    ovs-ofctl del-flows br-ovs
    ovs-ofctl add-flow br-ovs arp,actions=normal
    ovs-ofctl add-flow br-ovs "table=0, ip,ct_state=-trk actions=ct(table=1)"
    ovs-ofctl add-flow br-ovs "table=1, ip,ct_state=+trk+new actions=ct(commit),normal"
    ovs-ofctl add-flow br-ovs "table=1, ip,ct_state=+trk+est, actions=normal"

With SW steering enabled and the above commits (and up to date ofed) tuple insertion rate should be about
twice as fast. This can be seen via procfs hw offloaded count while running high traffic:
while true; do res=`sudo cat /proc/net/nf_conntrack | grep -i offload` && echo "$res" && echo "$res" | wc -l; done

* What it could break.

NA

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This bug is awaiting verification that the linux-bluefield/5.4.0-1033.36 kernel in -proposed solves the problem. Please test the kernel and update this bug with the results. If the problem is solved, change the tag 'verification-needed-focal' to 'verification-done-focal'. If the problem still exists, change the tag 'verification-needed-focal' to 'verification-failed-focal'.

If verification is not done by 5 working days from today, this fix will be dropped from the source code, and this bug will be closed.

See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you!

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux-bluefield - 5.4.0-1035.38

---------------
linux-bluefield (5.4.0-1035.38) focal; urgency=medium

  * focal/linux-bluefield: 5.4.0-1035.38 -proposed tracker (LP: #1969372)

  * mlxbf-gige: sync up with upstreamed version (LP: #1969233)
    - SAUCE: Revert "UBUNTU: SAUCE: Fix OOB handling RX packets in heavy traffic"
    - SAUCE: Revert "UBUNTU: SAUCE: mlxbf_gige: update driver version to 1.25"
    - SAUCE: Revert "UBUNTU: SAUCE: mlxbf_gige: clear valid_polarity upon open"
    - SAUCE: Revert "UBUNTU: SAUCE: mlxbf_gige: add interrupt counts to "ethtool
      -S""
    - SAUCE: Revert "UBUNTU: SAUCE: mlxbf-gige: add ethtool
      mlxbf_gige_set_ringparam"
    - SAUCE: Revert "UBUNTU: SAUCE: mlxbf-gige: add driver version"
    - mlxbf_gige: clear valid_polarity upon open
    - net: mellanox: mlxbf_gige: Replace non-standard interrupt handling
    - SAUCE: mlxbf-gige: add driver version
    - SAUCE: mlxbf_gige: add interrupt counts to "ethtool -S"
    - SAUCE: mlxbf-gige: add ethtool mlxbf_gige_set_ringparam
    - SAUCE: Fix OOB handling RX packets in heavy traffic

  * linux-bluefield: Fix build failure in mlxbf_gige (LP: #1969374)
    - gpiolib: acpi: Allow to find GpioInt() resource by name and index

linux-bluefield (5.4.0-1034.37) focal; urgency=medium

  * focal/linux-bluefield: 5.4.0-1034.37 -proposed tracker (LP: #1968766)

  * Devlink wasn't enabled from common config (LP: #1968751)
    - [Config] Bluefield: Enable CONFIG_NET_DEVLINK
    - [Config] Bluefield: Enable dummy config options NET_VENDOR_BROADCOM and
      PAGE_POOL

linux-bluefield (5.4.0-1033.36) focal; urgency=medium

  * focal/linux-bluefield: 5.4.0-1033.36 -proposed tracker (LP: #1967369)

  * Fix flow table lookup failure with no originating ifindex (LP: #1967892)
    - net/sched: act_ct: Fix flow table lookup failure with no originating ifindex

  * Fix OOB handling RX packets in heavy traffic (LP: #1964984)
    - SAUCE: Fix OOB handling RX packets in heavy traffic

  * Pass originating device to drivers offloading ct connection so devices will
    filter the tuples and offload them more efficiently (LP: #1960575)
    - net: openvswitch: Be liberal in tcp conntrack.
    - net/sched: act_ct: Fill offloading tuple iifidx
    - net: openvswitch: Fill act ct extension

  * Fix flow table lookup after ct clear or switching zones (LP: #1963948)
    - net/sched: act_ct: Fix flow table lookup after ct clear or switching zones

  * CT: Offload only ASSURED connections (LP: #1961819)
    - net/sched: act_ct: Offload only ASSURED connections

  * Sync up gpio interrupt handling with upstreamed version (LP: #1965017)
    - Revert "UBUNTU: SAUCE: gpio-mlxbf2.c: Fix setting the gpio direction to
      output"
    - Revert "UBUNTU: SAUCE: gpio-mlxbf2.c: remove phy interrupt"
    - Revert "UBUNTU: SAUCE: gpio-mlxbf2: Cleanup and use generic gpio_irq_chip
      struct"
    - Revert "UBUNTU: SAUCE: gpio-mlxbf2.c: Support soft reset gpio interrupt"
    - Revert "UBUNTU: SAUCE: gpio-mlxbf2.c: fix spinlock bug and using
      uninitialized work"
    - Revert "UBUNTU: SAUCE: gpio: Add irq support for gpio-mlxbf2"
    - gpio: mlxbf2: remove unused inclu...