From: Florian Fainelli <florian.fainelli@broadcom.com>
To: netdev-bot+sashiko@kernel.org
Cc: netdev@vger.kernel.org, rafal@milecki.pl,
bcm-kernel-feedback-list@broadcom.com, andrew+netdev@lunn.ch,
davem@davemloft.net, edumazet@kernel.org, kuba@kernel.org,
pabeni@redhat.com, robh@kernel.org, krzk+dt@kernel.org,
conor+dt@kernel.org, horms@kernel.org, corbet@lwn.net,
jonas.gorski@gmail.com, olteanv@gmail.com, pablo@netfilter.org,
fw@strlen.de, phil@nwl.cc, devicetree@vger.kernel.org,
linux-kernel@vger.kernel.org, linux-doc@vger.kernel.org,
netfilter-devel@vger.kernel.org, coreteam@netfilter.org
Subject: Re: [PATCH net-next 9/9] docs: net: document Northstar FA setup and NAT qualification
Date: Wed, 18 Nov 2026 10:25:48 -0800 [thread overview]
Message-ID: <563a68b7-c300-4ec8-9594-9a79fe4c37d5@broadcom.com> (raw)
In-Reply-To: <179495817095.3136.15574702882079067890@kernel.org>
On 11/17/26 15:29, netdev-bot+sashiko@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 6 potential
> issue(s) to consider.
>
> Critical: 0 · High: 0 · Medium: 4 · Low: 2
>
> If you believe none of the issues are legitimate, please reply to each
> issue of Medium severity or higher and explain why the AI is wrong.
> If any of them are legitimate and you plan to respin, please let patchwork
> know by including "pw-bot: cr" as a separate line at the end of your reply
> (one such reply per series is enough).
>
> - [Medium] The reference nftables ruleset in
> Documentation/networking/device_drivers/ethernet/broadcom/bcm-ns-…
> - [Medium] bcm-ns-fa.rst (lines 130-131) says "Every IPv4 FIB route, rule
> and nexthop change flushes all FA-owned flows".
> - [Medium] B53 and the FA backend decide independently whether FA
> metadata mode is active, and they use different tests.
> - [Medium] The FA block is at 0x18027c00 (reg <0x18027c00 0xc0>), which
> is offset 0xc00 inside the GMAC3 BCMA core aperture (gmac3:…
> - [Low] The 'Automated validation' section runs kunit.py with the suite
> filter `'*fa*'` against drivers/net/ethernet/broadcom/.kunitconfig.
> - [Low] The commit message for "docs: net: document Northstar FA setup
> and NAT qualification" ends with "Keep campaign history and review…
>
> --
>
> docs: net: document Northstar FA setup and NAT qualification
>
> This adds documentation for the Broadcom Northstar flow accelerator. It
> covers the supported GMAC2/GMAC0 topology, the device tree and kernel
> configuration, and how route and neighbour invalidation works. It also
> describes a three-host NAT qualification procedure, fault injection tests
> and KUnit validation, and adds the new file to the index and MAINTAINERS.
>
>> validation from the driver's incomplete FIB eligibility probe. Keep
>> campaign history and review status in the accompanying package.
>
> [Severity: Low]
> This isn't a bug, but what does "the accompanying package" refer to? The
> patch only adds bcm-ns-fa.rst, an index entry and a MAINTAINERS line.
> Nothing in the tree matches that description.
>
> Could this sentence go in the cover letter or below the --- line instead
> of the permanent changelog?
>
>> diff --git a/Documentation/networking/device_drivers/ethernet/broadcom/bcm-ns-fa.rst b/Documentation/networking/device_drivers/ethernet/broadcom/bcm-ns-fa.rst
>> new file mode 100644
>> index 0000000000000..2bc6cc5d94cea
>> --- /dev/null
>> +++ b/Documentation/networking/device_drivers/ethernet/broadcom/bcm-ns-fa.rst
>> @@ -0,0 +1,507 @@
>
> [ ... ]
>
>> +Netdevice shutdown, address/MTU changes, new active switch ports, and upper
>> +interface changes flush entries. Every IPv4 FIB route, rule and nexthop change
>> +flushes all FA-owned flows, including unrelated prefixes and local-table
>> +entries. This is intentional: the rule does not carry enough routing-policy
>> +context to prove that an event cannot affect an installed path. Work is
>> +coalesced, but even unrelated routing churn can reduce offload residency.
>
> [Severity: Medium]
> Can the IPv4 FIB notifier actually provide this guarantee? fa_fib_event()
> in drivers/net/ethernet/broadcom/bcm-ns-fa.c is the only route
> invalidation path. The same promise appears in its comment. However,
> fib_table_insert() only notifies when the new alias is the first one for
> that prefix and table:
>
> net/ipv4/fib_trie.c:fib_table_insert() {
> ...
> if (fib_find_alias(&l->leaf, new_fa->fa_slen, 0, 0, tb->tb_id, true) ==
> new_fa) {
> enum fib_event_type fib_event;
> ...
> }
>
> The NLM_F_REPLACE branch and the delete path work the same way:
>
> net/ipv4/fib_trie.c:fib_notify_alias_delete() {
> ...
> /* Do not notify if we do not care about the route. */
> if (fib_find_alias(fah, slen, 0, 0, tb_id, true) != fa_to_delete)
> return;
> ...
> }
>
> Aliases are sorted by descending DSCP. If a prefix also has a
> DSCP-specific alias (ip route ... tos X), the DSCP-0 alias that ordinary
> traffic uses is not first. Adding, replacing or deleting that alias raises
> no notifier event. As a result, neither fa_invalidate_locked() nor
> fa_mtu_route_update() runs.
>
> Flowtable GC doesn't catch this for hardware flows either, because
> validation-only references are skipped:
>
> include/net/netfilter/nf_flow_table.h:nf_flow_dst_check() {
> /* A driver-only reference must not change direct-path/GC semantics. */
> if (!tuple->dst_cache || tuple->dst_validation_only)
> return true;
> ...
> }
>
> nf_flow_offload_validate() only runs at admission.
>
> Suppose the DSCP-0 route is replaced with a blackhole, moved to another
> next hop, or given an MTU below 1500 as a non-first alias. Would the
> hardware rows keep forwarding on the old path while counter activity keeps
> refreshing them? The same gap applies to the later claim that flows stay
> in software while any route has an MTU below 1500. The blackhole and MTU
> 1200 qualification steps only use first-alias routes, so they would pass
> anyway.
>
> A minor point: fa_fib_event() also handles FIB_EVENT_ENTRY_ADD and
> FIB_EVENT_ENTRY_APPEND, which IPv4 never emits.
>
> [ ... ]
>
>> +For a BCM53012 board, add the following node under the root address space
>> +(adjust the compatible to ``brcm,bcm53011-fa`` for BCM53011)::
>> +
>> + / {
>> + fa: flow-accelerator@18027c00 {
>> + compatible = "brcm,bcm53012-fa";
>> + reg = <0x18027c00 0xc0>;
>> + };
>> + };
>
> [ ... ]
>
>> +Remove any other conflicting CPU assignments for this test. B53 selects ``brcm-fa``
>> +when it finds the enabled accelerator through this phandle. Do not set
>> +``dsa-tag-protocol``: B53 does not implement the runtime protocol-change
>> +operation. No existing board is enabled automatically by this series.
>
> [Severity: Medium]
> Is this the whole condition? B53 and the FA backend each decide on their
> own whether FA metadata mode is active, and they use different tests:
>
> drivers/net/dsa/b53/b53_common.c:b53_uses_fa_tag() {
> ...
> enabled = IS_ENABLED(CONFIG_BCM_NS_FA) && of_device_is_available(fa) &&
> (dev->chip_id == BCM53011_DEVICE_ID ||
> dev->chip_id == BCM53012_DEVICE_ID ||
> (is5301x(dev) && of_device_is_compatible(fa, "brcm,bcm4708-fa")));
> ...
> }
>
> drivers/net/ethernet/broadcom/bcm-ns-fa.c:bcm_ns_fa_init() {
> ...
> if (!of_device_is_compatible(np, "brcm,bcm4708-fa") &&
> !of_device_is_compatible(np, "brcm,bcm53011-fa") &&
> !of_device_is_compatible(np, "brcm,bcm53012-fa")) {
> of_node_put(np);
> return ERR_PTR(-ENODEV);
> }
> ...
> }
>
> bcm_ns_fa_init() ignores the chip. It then always enables FA_BRCM_HDR
> processing and clears FA_CTL_BYPASS.
>
> Take the brcm,bcm53012-fa node from the example above on a 5301x switch
> whose runtime chip_id is BCM53010, BCM53018 or BCM53019. The backend would
> switch GMAC2 RX to the FA descriptor format. B53 would keep the ordinary
> brcm tagger on port 8 and skip the CFP exception guards. Would conduit RX
> on port 8 then be misparsed or dropped?
>
> An available FA node with any other compatible makes bcm_ns_fa_init()
> return -ENODEV. That fails the whole GMAC2 probe:
>
> drivers/net/ethernet/broadcom/bgmac.c:bgmac_enet_probe() {
> ...
> bgmac->fa = bcm_ns_fa_init(net_dev);
> if (IS_ERR(bgmac->fa)) {
> err = PTR_ERR(bgmac->fa);
> bgmac->fa = NULL;
> ...
> }
>
> Also, "brcm,bcm4708-fa" turns on FA mode for any 5301x chip. That includes
> parts the binding, the Kconfig help and this document do not list as
> supported.
>
> Could B53 and the backend share one predicate, and could the document
> state the actual conditions?
>
>> +
>> +The FA resource is outside the GMAC3 DMA register window. It is managed by
>> +the GMAC2 consumer; it does not need a separate platform driver or character
>> +device. Leave the accelerator disabled or omit the GMAC2 phandle to retain
>> +the existing board configuration.
>
> [Severity: Medium]
> What state does the GMAC3 core need to be in? The FA block at 0x18027c00
> is at offset 0xc00 inside the GMAC3 BCMA core aperture (gmac3:
> ethernet@27000 in bcm-ns.dtsi). bcm_ns_fa_init() maps and programs it from
> GMAC2 without checking or enabling the core that hosts it.
>
> If the FA shares GMAC3's reset or clock domain, two cases look like
> problems.
>
> First, when GMAC3 is disabled in DT, nothing releases the core's reset:
>
> drivers/bcma/main.c:bcma_register_devices() {
> ...
> np = core->dev.of_node;
> if (np && !of_device_is_available(np))
> continue;
> ...
> }
>
> Could FA register accesses then abort, or hit the FA_INIT_DONE poll
> timeout, depending on what the bootloader left behind?
>
> Second, when GMAC3 is enabled, bgmac_probe() in bgmac-bcma.c only rejects
> the phandle on a core other than unit 2. GMAC3 therefore still probes, and
> bgmac_enet_probe() unconditionally calls:
>
> bgmac_clk_enable(bgmac, 0);
>
> This reaches bcma_core_enable(), which pulses the GMAC3 core reset. If
> that happens after GMAC2 has initialized the FA, does it wipe FA_CTL and
> the table state? The GMAC2 side would still believe its flows[] and
> next-hop bookkeeping match the hardware. BGMAC_FEAT_NO_RESET suppresses
> the reset in open/close, but not this probe-time enable.
>
> The setup steps only cover GMAC0 and GMAC2. Should they say whether GMAC3
> must be enabled, disabled or left alone?
>
> [ ... ]
>
>> + table inet fa_test {
>> + flowtable ft {
>> + hook ingress priority 0;
>> + devices = { lan1, wan };
>> + flags offload;
>> + counter;
>> + }
>> + chain forward {
>> + type filter hook forward priority 0; policy drop;
>> + meta l4proto udp ct state established flow add @ft
>> + ct state established,related counter accept
>> + iifname "lan1" oifname "wan" udp dport 5001 counter accept
>> + iifname "wan" oifname "lan1" udp dport 5001 ct status dnat counter accept
>> + }
>> + chain prerouting {
>> + type nat hook prerouting priority dstnat; policy accept;
>> + iifname "wan" udp dport 15001 dnat ip to 192.0.2.2:5001
>> + }
>> + chain postrouting {
>> + type nat hook postrouting priority srcnat; policy accept;
>> + iifname "lan1" oifname "wan" ip saddr 192.0.2.2 snat ip to 198.51.100.1:45000
>> + }
>> + }
>
> [Severity: Medium]
> Will this table load with nft -f? The snat rule in the postrouting chain
> asks for a port mapping. However, iifname, oifname and ip saddr don't set
> a transport protocol. nft rejects such a rule with "transport protocol
> mapping is only valid after transport protocol match". nft -f loads the
> file as one transaction, so the whole fa_test table would be rejected.
>
> The dnat rule is fine because udp dport 15001 comes first. Would a meta
> l4proto (or tcp/udp) match before the snat statement fix this? The
> in-tree nft_flowtable.sh and nft_nat.sh selftests do that.
>
> Separately, the forward chain only handles UDP:
>
> meta l4proto udp ct state established flow add @ft
> iifname "lan1" oifname "wan" udp dport 5001 counter accept
>
> With policy drop, wouldn't a new TCP SYN match no accept rule and be
> dropped? The text below says "run a UDP or TCP echo server". Later
> sections ask for the SNAT, DNAT, pressure, fallback and aging tests to be
> repeated with TCP, but no TCP ruleset is given.
>
> [ ... ]
>
>> + ./tools/testing/kunit/kunit.py run --arch=x86_64 \
>> + --kunitconfig=drivers/net/ethernet/broadcom/.kunitconfig \
>> + --build_dir=/tmp/bcm-ns-fa-kunit '*fa*'
>
> [Severity: Low]
> Does the '*fa*' filter skip suites that this kunitconfig builds?
> drivers/net/ethernet/broadcom/.kunitconfig also sets
> CONFIG_NF_FLOW_TABLE_KUNIT_TEST=y. That builds the nf-flow-lifetime suite
> (net/netfilter/nf_flow_table_core_test.c) and the nf-flow-path suite
> (net/netfilter/nf_flow_table_offload_test.c). Neither name contains "fa",
> so the documented command builds them but never runs them.
>
> nf-flow-path covers the cached route and neighbour validation in
> nf_flow_offload_validate() that the commit message relies on.
>
> Should the filter include these suites? Should the text below also
> mention CONFIG_NF_FLOW_TABLE_KUNIT_TEST next to the three suites it lists?
>
> [ ... ]
>
[Medium] FIB notifications
Agreed: only the first alias of a prefix is notified, so changes to
non-first aliases (DSCP routes) are not seen. v2 documents the limit and
that DSCP-specific routes must not be used with FA.
[Medium] B53 and the backend deciding differently
Both now accept the same compatibles (patch 6), on a BCM5301x switch. v2
documents the actual conditions and what happens for an enabled node
with another compatible.
[Medium] GMAC3
GMAC3 is not probed by bgmac without a MAC address, so its core is not
reset by the driver. Whether the accelerator registers are accessible
depends on the bootloader state; v2 documents it, and makes reading them
the first step of the hardware validation.
[Medium] nftables ruleset
Agreed, v2 matches the protocol before the NAT statements and covers TCP
as well as UDP.
[Low] kunit filter, commit message
Fixed in v2.
pw-bot: cr
--
Florian
next prev parent reply other threads:[~2026-11-18 18:25 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-16 22:25 [PATCH net-next 0/9] net: bgmac: add Northstar IPv4 TCP/UDP NAT flowtable offload Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 1/9] net: ethernet: bgmac: validate RX length against the DMA packet capacity Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-17 23:43 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 2/9] netfilter: flowtable: reject non-host IPv4 packets before lookup Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-17 23:51 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 3/9] netfilter: flowtable: retain offload ownership until deletion completes Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-17 23:57 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 4/9] dt-bindings: net: describe the Northstar flow accelerator Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-18 17:22 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 5/9] net: dsa: add Broadcom FA metadata and IPv4 exception paths Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-18 18:08 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 6/9] net: flow_offload: allow drivers to validate and invalidate cached paths Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-18 18:10 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 7/9] net: bgmac: add Northstar IPv4 TCP/UDP NAT flowtable offload Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-18 18:18 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 8/9] net: add KUnit coverage for FA and flowtable path validation Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-18 18:22 ` Florian Fainelli
2026-11-16 22:25 ` [PATCH net-next 9/9] docs: net: document Northstar FA setup and NAT qualification Florian Fainelli
2026-11-17 23:29 ` netdev-bot+sashiko
2026-11-18 18:25 ` Florian Fainelli [this message]
2026-11-16 22:38 ` [PATCH net-next 0/9] net: bgmac: add Northstar IPv4 TCP/UDP NAT flowtable offload netdev-bot+sinfo
2026-11-16 22:45 ` Florian Fainelli
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=563a68b7-c300-4ec8-9594-9a79fe4c37d5@broadcom.com \
--to=florian.fainelli@broadcom.com \
--cc=andrew+netdev@lunn.ch \
--cc=bcm-kernel-feedback-list@broadcom.com \
--cc=conor+dt@kernel.org \
--cc=corbet@lwn.net \
--cc=coreteam@netfilter.org \
--cc=davem@davemloft.net \
--cc=devicetree@vger.kernel.org \
--cc=edumazet@kernel.org \
--cc=fw@strlen.de \
--cc=horms@kernel.org \
--cc=jonas.gorski@gmail.com \
--cc=krzk+dt@kernel.org \
--cc=kuba@kernel.org \
--cc=linux-doc@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev-bot+sashiko@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=netfilter-devel@vger.kernel.org \
--cc=olteanv@gmail.com \
--cc=pabeni@redhat.com \
--cc=pablo@netfilter.org \
--cc=phil@nwl.cc \
--cc=rafal@milecki.pl \
--cc=robh@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox