Wire up the port_bridge_join, port_bridge_leave and port_vlan_filtering
DSA callbacks to support both VLAN-unaware and VLAN-aware bridge modes.
For VLAN-unaware bridges, each bridge instance is assigned a dedicated
internal PVID via NETC_VLAN_UNAWARE_PVID(bridge.num), counting down
from VID 4095. A VFT entry is created for this PVID with hardware MAC
learning and flood-on-miss forwarding enabled. The CPU port is included
as a VFT member so frames can reach the host. The reserved VID range is
blocked in port_vlan_add to prevent user-space conflicts.
Only one VLAN-aware bridge is supported at a time; this constraint is
enforced in port_bridge_join and port_vlan_filtering. The per-port PVID
is tracked in software and written to the BPDVR register whenever VLAN
filtering is active.
When a port leaves the bridge, its dynamic FDB entries are flushed right
away in port_bridge_leave(), without waiting for the ageing cycle. When
a link down event occurs on a port, netc_mac_link_down() will also clear
the port's dynamic FDB entries via netc_port_remove_dynamic_entries().
Non-bridge ports have no dynamic FDB entries, so this call is always
safe. Additionally, .port_fast_age() callback is added to flush the
dynamic FDB entries associated to a port.
Host flood rules are removed from the ingress port filter table when a
port joins a bridge to avoid bypassing FDB lookup and MAC learning.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-9-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Implement the DSA .port_vlan_add and .port_vlan_del operations to enable
VLAN-aware bridge offloading on the NETC switch.
VLAN membership is maintained in the VLAN Filter Table (VFT). Adding the
first port to a VLAN creates a new VFT entry with hardware MAC learning
and flood-on-miss forwarding; subsequent ports update the existing
entry's membership bitmap. Removing the last port deletes the entry.
Egress tagging is handled through the Egress Treatment Table (ETT). Each
VLAN is allocated a group of ETT entries, one per available port. Ports
are assigned a sequential ett_offset during initialisation, used to
address each port's entry within the group. Untagged ports configure the
ETT to strip the outer VLAN tag; tagged ports pass frames through
unmodified. Each ETT group is optionally paired with an Egress Counter
Table (ECT) group for per-port frame counting, allocated on a best-effort
basis. When the egress rule of an ETT entry changes, the counter of the
corresponding ECT entry will be recounted to track the number of frames
that match the new egress rule.
A software shadow list serialised by vft_lock tracks active VLAN state
across both port membership and egress tagging. VID 0 is used for single
port mode and is ignored by both callbacks.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-8-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
NTMP index tables require software to allocate and manage entry IDs.
Add two bitmap helper functions to facilitate this management:
ntmp_lookup_free_eid(): finds the first zero bit in the given bitmap,
sets it to mark the entry as in-use, and returns the corresponding entry
ID. Returns NTMP_NULL_ENTRY_ID if no free entry is available.
ntmp_clear_eid_bitmap(): clears the bit associated with the given entry
ID in the bitmap to mark the entry as free. It is a no-op if the entry
ID is NTMP_NULL_ENTRY_ID.
Both functions are exported for use by other modules, such as the NETC
switch driver which needs to manage group index bitmaps for the Egress
Treatment Table (ETT) and Egress Count Table (ECT).
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-7-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The Egress Treatment Table (ETT) and Egress Count Table (ECT) are both
index tables whose entry IDs are allocated by software. Every num_ports
entries form a group, where each entry in the group corresponds to one
port. To facilitate group allocation and management, initialize the group
index bitmaps for both tables based on hardware capabilities reported by
ETTCAPR and ECTCAPR registers.
The bitmap size per table is calculated as the total number of hardware
entries divided by the number of available ports, which gives the number
of groups available for software allocation. A set bit in the bitmap
represents a group index that has been allocated.
These bitmaps will be used by subsequent patches that add VLAN support.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-6-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The egress count table is a static bounded index table, egress related
statistics are maintained in this table. The table is implemented as a
linear array of entries accessed using an index (0, 1, 2, ..., n) that
uniquely identifies an entry within the array. Egress Counter Entry ID
(EC_EID) is used as an index to an entry in this table. The EC_EID is
specified in the egress treatment table.
Egress count table entries are always present and enabled. The table
only supports access via entry ID, which is assigned by the software.
And it supports Update, Query and Query followed by Update operations.
Currently, only Update operation is supported.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-5-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Each entry in the egress treatment table contains the egress packet
processing actions to be applied to a grouping or scope of packets
exiting on a particular egress port of the switch. A scope of packets,
for example, could be the packets exiting a particular VLAN, matching
a particular 802.1Q bridge forwarding entry or belonging to a stream
identified at ingress. The egress treatment table is implemented as a
linear array of entries accessed using an index (0,1, 2, ..., n) that
uniquely identifies an entry within the array.
The egress treatment table only supports access vid entry ID, which is
assigned by the software. It supports Add, Update, Delete and Query
operations. Note that only Query operation is not supported yet.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-4-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add two interfaces to manage entries in the VLAN filter table:
ntmp_vft_update_entry(): Update the configuration element data of the
specified VLAN filter entry based on the given VLAN ID. It uses the
exact key access method to locate the entry.
ntmp_vft_delete_entry(): Delete the VLAN filter entry corresponding to
the specified VLAN ID. It also uses the exact key access method to
identify the target entry.
In addition, introduce struct vft_req_qd to describe the request data
buffer format for Query and Delete actions of the VLAN filter table,
which contains a common request data header and a VLAN access key.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-3-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add three interfaces to manage dynamic entries in the FDB table:
ntmp_fdbt_update_activity_element(): Update the activity element of all
dynamic FDB entries. For each entry, if its activity flag is not set,
which means no packet has matched this entry since the last update, the
activity counter is incremented. Otherwise, both the activity flag and
activity counter are reset. The activity counter is used to track how
long an FDB entry has been inactive, which is useful for implementing
an ageing mechanism.
ntmp_fdbt_delete_ageing_entries(): Delete all dynamic FDB entries whose
activity flag is not set and whose activity counter is greater than or
equal to the specified threshold. This is used to remove stale entries
that have been inactive for too long.
ntmp_fdbt_delete_port_dynamic_entries(): Delete all dynamic FDB entries
associated with the specified switch port. This is typically called when
a port goes down or is removed from a bridge.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://patch.msgid.link/20260611021458.2629145-2-wei.fang@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add test_action_set exercising OVS_ACTION_ATTR_SET with an ipv4 dst
rewrite. The test verifies the SET action in three steps: first
confirm normal forwarding, then apply set(ipv4(dst=10.0.0.99)) to
rewrite the destination to an address nobody owns and verify ping
fails, then restore normal forwarding and verify connectivity
recovers.
Signed-off-by: Minxi Hou <houminxi@gmail.com>
Reviewed-by: Aaron Conole <aconole@redhat.com>
Link: https://patch.msgid.link/20260612130503.311240-1-houminxi@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jonas Jelonek says:
====================
net: sfp: extend SMBus support
Today, the SFP driver only drives I2C adapters that advertise full
I2C_FUNC_I2C, or SMBus-only adapters via single-byte transfers (with
hwmon disabled). Several SoCs ship I2C/SMBus-only controllers that
support more than just byte access -- e.g. word and I2C block -- and
have SFP cages wired to them. Today, those adapters either work
poorly or not at all.
This series teaches the SFP driver to use the larger SMBus access
modes when the adapter advertises them, and along the way starts
honoring i2c_adapter quirks on read/write length so adapters that
cap below the SFP block size are handled correctly. Patch 1 is a
small prep doing only the quirks handling; patch 2 extends the
SMBus path itself.
Capability matrix supported by patch 2:
- BYTE only: single-byte access (unchanged).
- BYTE + WORD: word for >=2-byte chunks, byte tail.
- I2C_BLOCK present: block as the universal transport.
- WORD only (no BYTE/BLOCK): accepted with WARN_ONCE; works for
even-length transfers, odd-length
transfers will error at xfer time.
Adapters with asymmetric R/W capabilities (e.g. only READ_I2C_BLOCK
without WRITE_I2C_BLOCK) remain functionally correct but use the
worse-supported direction's max for both directions, since
i2c_max_block_size is a single field. No mainline I2C driver was
seen advertising such asymmetry; per-direction sizes can be added
later if needed.
====================
Link: https://patch.msgid.link/20260614133418.2068201-1-jelonek.jonas@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 7662abf4db ("net: phy: sfp: Add support for SMBus module access")
added SMBus access for SFP modules, but limited it to single-byte
transfers. As a side effect, hwmon is disabled (16-bit reads cannot be
guaranteed atomic) and a warning is printed.
Many SMBus-only I2C controllers in the wild support more than just
byte access, and SFP cages are often wired to such controllers
rather than to a full-featured I2C controller -- e.g. the SMBus
controllers in the Realtek longan and mango SoCs, which advertise
word access and I2C block reads. Today, they cannot drive an SFP at
all without falling back to the byte-only path.
Extend sfp_smbus_read()/sfp_smbus_write() so that, in addition to
the existing byte access, they also use SMBus word access and SMBus
I2C block access whenever the adapter advertises them. Both
directions are handled in a single read and a single write helper
that pick the largest supported transfer per chunk and fall back as
needed.
I2C-block is preferred unconditionally when available: the protocol
carries any length 1..32, so it can serve every chunk -- including
the 1- and 2-byte tails -- without help from word or byte access.
Note that this requires I2C_FUNC_SMBUS_I2C_BLOCK, which reads a
caller-specified number of bytes. This deviates from the official
SMBus Block Read (length is supplied by the slave) but is widely
supported by Linux I2C controllers/drivers.
Capability matrix this implementation supports:
- BYTE only: works (unchanged behaviour); 1-byte
xfers, hwmon disabled.
- BYTE + WORD: word for >=2-byte chunks, byte for
trailing odd byte.
- I2C_BLOCK present (with or
without BYTE/WORD): block as the universal transport for
every chunk.
- WORD only (no BYTE/BLOCK): accepted with WARN_ONCE. Even-length
transfers work; odd-length transfers
(e.g. the 3-byte cotsworks fixup
write) hit the BYTE branch which the
adapter does not implement, so the
xfer returns an error and the
operation is aborted. No mainline
I2C driver was found to advertise
WORD without BYTE; the warning lets
us learn about it if it ever shows
up.
Adapters with asymmetric R/W capabilities (e.g. only READ_I2C_BLOCK
but not WRITE_I2C_BLOCK) remain functionally correct -- the
per-iteration fallback uses the direction-specific bits -- but the
shared i2c_max_block_size is sized by the all-bits-set check, so a
transfer in the better-supported direction is not upgraded. None of
the mainline I2C bus drivers surveyed during review advertise such
asymmetry; promoting i2c_max_block_size to per-direction sizes can
be revisited if needed.
Signed-off-by: Jonas Jelonek <jelonek.jonas@gmail.com>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20260614133418.2068201-3-jelonek.jonas@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The SFP driver assumes all I2C adapters support reading and writing the
pre-defined block size SFP_EEPROM_BLOCK_SIZE of 16 bytes. This constant
was probably chosen based on good guesses and known limitations of a
range of I2C adapters and SFP modules.
However, I2C adapters may even support less and usually need to specify
this via I2C quirks. Theoretically, such an adapter may provide full
functionality but only support a read and write length of e.g. 8 bytes.
Currently, the SFP driver doesn't account for that.
Add handling for I2C quirks in SFP I2C configuration taking the fields
max_read_len and max_write_len in struct i2c_adapter_quirks into account
to further limit the maximum block size if needed.
Signed-off-by: Jonas Jelonek <jelonek.jonas@gmail.com>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20260614133418.2068201-2-jelonek.jonas@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----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=yhtm
-----END PGP SIGNATURE-----
Merge tag 'nf-next-26-06-14' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next
Pablo Neira Ayuso says:
====================
Netfilter/IPVS updates for net-next
The following patchset contains Netfilter/IPVS updates for net-next.
More specifically, this contains conncount rework to address AI related
reports, assorted Netfiter updates and two small incremental updates on
IPVS:
1) Replace old obsolete workqueues (system_wq, system_unbound_wq)
in IPVS, from Marco Crivellari.
2) Replace WARN_ON{_ONCE} by DEBUG_NET_WARN_ON_ONCE in nf_tables.
In the recent years, reporters say that the use of WARN_ON{_ONCE}
in conjunction with panic_on_warn=1 results in DoS. Let's replace
it by DEBUG_NET_WARN_ON_ONCE so this is only exercised by test
infrastructure and fuzzers, while also providing context to AI
agents. From Fernando F. Mancera.
Five patches from Florian Westphal to address AI reports in the conncount
infrastructures:
3) Fix missing rcu read lock section when calling
__ovs_ct_limit_get_zone_limit().
4) Add a dedicate lock per rbtree tree, this increases memory
usage but it should improve scalability.
5) Add a helper function to find the rbtree node, no functional
changes are intented.
6) Add sequence counter to detect concurrent tree modifications
and retry lookups.
7) Add locks to GC conncount walk and address other nitpicks.
Then, several assorted updates:
8) Defensive Tree-wide addition of NULL checks for ct extensions.
9) Bail out if flowtable bypass cannot be fully set up from the
flow offload expression, instead of lazy building a likely
incomplete one.
10) Fix documentation for the new conn_max sysctl toggle in IPVS.
11) Add nf_dev_xmit_recursion*() helpers and use them, to address
recent AI reports.
* tag 'nf-next-26-06-14' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next:
netfilter: nf_dup_netdev: add nf_dev_xmit_recursion*() helpers and use them
ipvs: fix doc syntax for conn_max sysctl
netfilter: flowtable: bail out if forward path cannot be discovered
netfilter: conntrack: check NULL when retrieving ct extension
netfilter: nf_conncount: gc and rcu fixes
netfilter: nf_conncount: add sequence counter to detect tree modifications
netfilter: nf_conncount: split count_tree_node rbtree walk into helper
netfilter: nf_conncount: use per nf_conncount_data spinlocks
netfilter: nf_conncount: callers must hold rcu read lock
netfilter: nf_tables: use DEBUG_NET_WARN_ON_ONCE in packet and control paths
ipvs: Replace use of system_unbound_wq with system_dfl_long_wq
====================
Link: https://patch.msgid.link/20260614114605.474783-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Dragos Tatulea says:
====================
netdev: expose page pool order via netlink
This small series exposes io_uring's high order page configuration
via the page_pool netlink interface and updates the appropriate
selftest to check this value.
====================
Link: https://patch.msgid.link/20260612211709.1456966-2-dtatulea@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Check the newly added rx_buf_len page_pool field for io_uring
in the existing large-chunks test after the receiver is up.
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Link: https://patch.msgid.link/20260612211709.1456966-4-dtatulea@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Bobby Eshleman says:
====================
selftests/vsock: improve vng version and quirk handling
As vng has continued updating, there have been two things in our
selftests that have been affected. One is that newer versions always
emit the vng version warning, and two is that we have a workaround that
is not needed in newer versions.
This series just updates the version handling to allow all newer
versions without warning and version-gates the workaround to only those
versions that don't have the commit that fixed the root cause.
Additionally, we add function for comparing major.minor versions which
is used in both patches.
-===================
Link: https://patch.msgid.link/20260612-vsock-test-update-v1-0-7d7eeed3ac8f@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The current vng version check uses a discrete allowlist of "1.33",
"1.36", and "1.37", which forces a script update on every new release
even though all post-1.36 releases work.
Replace the discrete list with: "1.33", or any version >= 1.36. 1.34
and 1.35 are skipped because they were not tested. Add a version_lt()
helper that compares MAJOR.MINOR numerically, so the check reads as a
straightforward version comparison.
Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
Link: https://patch.msgid.link/20260612-vsock-test-update-v1-1-7d7eeed3ac8f@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
An SFP cage (compatible "sff,sfp") whose MOD_DEF0 signal is not wired to a
GPIO currently falls back to sff_gpio_get_state(), which unconditionally
reports the module as present. An empty cage therefore fails its probe and
is parked in SFP_MOD_ERROR forever; because SFP_F_PRESENT never deasserts
there is no REMOVE event to recover the state machine, so a module inserted
after boot is never detected, and empty cages spam -EIO at boot.
This affects boards that route none of the cage presence signal to a
software-readable input. On the NicGiga S100-0800S-M (RTL9303, 8x SFP+) the
cage I2C bus is the switch's SMBus master; TX_DISABLE is driven via a
PCA9534 I/O expander, but no MOD_ABS/MOD_DEF0 line reaches a readable GPIO
(the RTL9303 gpio0 lines read stuck-low, the single PCA9534 is fully
consumed by TX_DISABLE, and there is no RTL8231). The Horaco ZX-SW82TS-L2P
(RTL9302D, 2x SFP+) is independently affected in the same way.
For such an SFP cage, derive presence from a throttled single-byte I2C read
of the module EEPROM instead: a successful read asserts SFP_F_PRESENT,
R_PROBE_ABSENT consecutive failures clear it (to ride out a transient error
on a live module). The existing poll then emits SFP_E_INSERT / SFP_E_REMOVE
normally, giving working hot-plug and silencing the boot-time -EIO spam on
empty cages. Presence is re-probed every T_PROBE_PRESENT, so insertion is
detected within that interval and removal within
T_PROBE_PRESENT * R_PROBE_ABSENT.
A soldered-down module (compatible "sff,sff") has no presence signal and is
genuinely always present, so it continues to use sff_gpio_get_state(); the
new path is gated on the cage type advertising SFP_F_PRESENT.
Signed-off-by: Greg Patrick <gregspatrick@hotmail.com>
Tested-by: Manuel Stocker <mensi@mensi.ch>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Tested-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20260611175341.2223184-1-gregspatrick@hotmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
David Yang says:
====================
devlink: Warn on resource ID collision with PARENT_TOP
Filter out the ambiguous case of
enum {
MY_RESOURCE_ID_A, /* == DEVLINK_RESOURCE_ID_PARENT_TOP ! */
MY_RESOURCE_ID_B,
...
};
register(..., MY_RESOURCE_ID_A, DEVLINK_RESOURCE_ID_PARENT_TOP, ...);
register(..., MY_RESOURCE_ID_B, MY_RESOURCE_ID_A, ...);
====================
Link: https://patch.msgid.link/20260611070856.889700-1-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ID 0 serves as the sentinel DEVLINK_RESOURCE_ID_PARENT_TOP to mark
top-level resources. While it is technically possible to use 0 as a real
resource ID, a user might be tempted to write:
enum {
MY_RESOURCE_ID_A, /* == DEVLINK_RESOURCE_ID_PARENT_TOP ! */
MY_RESOURCE_ID_B,
MY_RESOURCE_ID_C,
MY_RESOURCE_ID_D,
...
};
register(..., MY_RESOURCE_ID_C, DEVLINK_RESOURCE_ID_PARENT_TOP, ...);
register(..., MY_RESOURCE_ID_D, MY_RESOURCE_ID_C, ...);
/* D is a child of C */
register(..., MY_RESOURCE_ID_A, DEVLINK_RESOURCE_ID_PARENT_TOP, ...);
register(..., MY_RESOURCE_ID_B, MY_RESOURCE_ID_A, ...);
/* Is B intentionally top-level, or is it actually a child of A? */
Add a WARN_ON() to catch this and prevent confusion.
Signed-off-by: David Yang <mmyangfl@gmail.com>
Link: https://patch.msgid.link/20260611070856.889700-6-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The devlink resource ID for ATU collides with the sentinel
DEVLINK_RESOURCE_ID_PARENT_TOP (0). As a result, ATU_bin_* are
registered as in fact registered as top-level siblings, not as children
of ATU.
Whether intentional or unintentional, clarify it by keeping the real
resource IDs starting at 1. Unfortunately ATU_bin_* are already
registered at top-level, so keep their parent to PARENT_TOP.
Signed-off-by: David Yang <mmyangfl@gmail.com>
Link: https://patch.msgid.link/20260611070856.889700-5-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This might not cause real problems, but the hellcreek devlink resource
ID collides with the sentinel DEVLINK_RESOURCE_ID_PARENT_TOP (0). Avoid
it by keeping the real resource IDs starting at 1.
Signed-off-by: David Yang <mmyangfl@gmail.com>
Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
Link: https://patch.msgid.link/20260611070856.889700-4-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This might not cause real problems, but the b53 devlink resource ID
collides with the sentinel DEVLINK_RESOURCE_ID_PARENT_TOP (0). Avoid it
by keeping the real resource IDs starting at 1.
Signed-off-by: David Yang <mmyangfl@gmail.com>
Link: https://patch.msgid.link/20260611070856.889700-3-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This might not cause real problems, but the dsa_loop devlink resource ID
collides with the sentinel DEVLINK_RESOURCE_ID_PARENT_TOP (0). Avoid it
by keeping the real resource IDs starting at 1.
Signed-off-by: David Yang <mmyangfl@gmail.com>
Link: https://patch.msgid.link/20260611070856.889700-2-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Kuniyuki Iwashima says:
====================
ipv4: fib: Remove RTNL in fib_net_exit_batch().
Currently, we flush all IPv4 routes at ->exit_batch() during
netns dismantle, which requires an extra RTNL.
IPv4 routes are not added from the fast path unlike IPv6, so
we can flush routes before default_device_exit_batch().
However, there is implicit ordering between ip_fib_net_exit()
and default_device_exit_batch().
This series detangles it and moves ip_fib_net_exit() to
->exit_rtnl() to save the RTNL dance.
The same change for IPv6 will need more work.
====================
Link: https://patch.msgid.link/20260612063225.455191-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, IPv4 routes are flushed in ->exit_batch() after
all devices are unregistered.
Unlike IPv6, IPv4 routes are not added from the fast path,
so we can flush routes before default_device_exit_batch().
Let's call ip_fib_net_exit() from ->exit_rtnl() to save
one RTNL locking dance.
ip_fib_net_exit() must use list_del_rcu() for fib_table
for the fast path on dying dev.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260612063225.455191-6-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will call ip_fib_net_exit() from ->exit_rtnl().
All fib_table will be destroyed before devices are unregistered.
During device unregistration, inetdev_destroy() could call
fib_del_ifaddr(), which calls fib_magic(RTM_DELROUTE).
fib_magic() calls fib_new_table(), but we do not want to create
a new table after ip_fib_net_exit() destroys all tables.
As a prep, let's add check_net() before fib_trie_table() in
fib_new_table().
fib_trie_table() is also called from fib_trie_unmerge(), but
fib_get_table() fails first in fib_unmerge(), so the same
problem does not occur there.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260612063225.455191-5-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will call ip_fib_net_exit() from ->exit_rtnl().
However, some paths will still access net->ipv4.fib_table_hash
after ->exit_rtnl().
For example, fib_flush() is called from fib_disable_ip() for
NETDEV_UNREGISTER.
Let's move kfree(net->ipv4.fib_table_hash) and fib4_notifier_exit()
from ip_fib_net_exit() to its caller.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260612063225.455191-4-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will call ip_fib_net_exit() from ->exit_rtnl().
Since the exit callbacks are called in the following order,
1. ->pre_exit()
~~~ synchronize_rcu() ~~~
2. ->exit_rtnl() : ip_fib_net_exit()
3. ->exit() : fib_proc_exit() / nl_fib_lookup_exit()
4. ->exit_batch() : fib4_semantics_exit()
the reverse order of fib_net_init() would get messed up.
Let's move fib_proc_exit() and nl_fib_lookup_exit() to ->pre_exit().
This is fine because procfs/netlink access from userspace cannot
occur at this point and synchronize_rcu() is not needed.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260612063225.455191-3-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Even when fib_table_flush() is called with flush_all true, it does
not flush all fib_info due to this condition:
!(fi->fib_flags & RTNH_F_DEAD) && !fib_props[fa->fa_type].error)
This creates an implicit ordering between default_device_exit_batch()
and fib_net_exit_batch().
fib_table_flush(flush_all=true) must be called after all devices
are NETDEV_UNREGISTERed, which is after nexthop_flush_dev() marks
RTNH_F_DEAD.
This would cause memory leak if the order were reversed.
fib_table_flush() does not skip non-dead error routes when flush_all
is true:
!flush_all &&
!(fi->fib_flags & RTNH_F_DEAD) && fib_props[fa->fa_type].error
Let's merge the two conditions not to skip all non-dead fib_info
during netns dismantle.
Note that we could further apply !flush_all to the basic table
id check and the rtmsg_fib() call in the loop.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260612063225.455191-2-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
One fewer allocation for the priv struct.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
Link: https://patch.msgid.link/20260608045640.5172-1-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan says:
====================
net/mlx5: Add switchdev mode support for Socket Direct single netdev, part 2/2
This is part 2. Find part 1 here:
https://lore.kernel.org/all/20260531113954.395443-1-tariqt@nvidia.com/
This series enables Socket Direct single netdev to operate in switchdev
mode with shared FDB. SD single netdev combines multiple PCI functions
behind a single netdev interface. To support switchdev offloads, these
functions must participate in virtual LAG (shared FDB).
Design
Rather than introducing a separate LAG instance for SD, this series
integrates SD secondary devices into the existing LAG structure
(priv.lag) created at probe time. Each lag_func entry carries a
group_id field that identifies its SD group membership (0 means not
part of any SD group). An xarray mark (XA_MARK_PORT) distinguishes
physical port entries from SD secondaries, enabling a single unified
iterator that filters by group:
- MLX5_LAG_FILTER_PORTS: iterate port-level entries only (existing
behavior, used by bonding, FW LAG commands, v2p_map)
- MLX5_LAG_FILTER_ALL: iterate all devices including SD secondaries
(used by MPESW shared FDB across all devices)
- specific group_id: iterate only devices in that SD group (used by
per-group SD shared FDB operations)
Existing callers use mlx5_ldev_for_each() which maps to
MLX5_LAG_FILTER_PORTS, preserving current behavior for non-SD
configurations.
Lifecycle and ownership
The SD LAG lifecycle is tied to the SD group, not to bonding events:
1. At PCI probe, mlx5_lag_add_mdev() creates the LAG structure
(priv.lag) for each LAG-capable PF. e.g.: SD primary devices
2. During mlx5_sd_init(), after the SD group is fully formed (primary
and secondaries paired), sd_lag_init() registers the secondary
devices into the primary's existing priv.lag by calling
mlx5_ldev_add_mdev() with the SD group_id. The primary's lag_func
also gets its group_id set. No separate LAG instance is created.
3. After all the devices in SD group transition to switchdev,
mlx5_lag_shared_fdb_create() is invoked with the group_id to create
a software-only shared FDB scoped to that SD group. This sets
sd_fdb_active on all lag_func entries in the group. No FW LAG
commands are issued since SD devices share the same physical port.
4. If MPESW (multi-port eswitch) is enabled on top of SD groups, the
per-group SD shared FDB is torn down first, then MPESW shared FDB is
created spanning all devices (ports + SD secondaries) using
MLX5_LAG_FILTER_ALL. On MPESW disable, per-group SD shared FDB is
restored.
5. On SD teardown (mlx5_sd_cleanup or device unbind), sd_lag_cleanup()
removes secondaries from priv.lag and clears the primary's group_id.
The LAG structure itself is not destroyed.
The sd_fdb_active flag is set on all lag_func entries in a group (not
just the primary), so any device can detect the SD shared FDB state
during lag_disable_change teardown without needing to look up peer
entries.
SD shared FDB is a pure software construct -- unlike regular LAG modes
(ROCE, SRIOV, MPESW), it does not issue FW create_lag/destroy_lag
commands. The software vport LAG for SD is implemented via eswitch
egress ACL bounce rules, managed by the IB layer through
mlx5_eth_lag_init(). And the software LAG demux is implemented via
steering rules that utilize new destination, VHCA_RX.
Patches
E-Switch preparation (patch 1):
- Skip uplink IB rep load for SD secondary devices
Devcom support (patches 2-3):
- Expose locked variant of send_event
- Add DEVCOM_CANT_FAIL for non-rollback events
SD core hardening (patches 4-6):
- Make primary/secondary role determination more robust
- Add L2 table silent mode query support
- Expand vport metadata for SD secondary devices
SD switchdev transition (patches 7-8):
- Support switchdev mode transition with shared FDB
- Notify SD on eswitch disable
LAG integration (patches 9-12):
- Store demux resources per master lag_func
- Disable both regular and SD LAG on lag_disable_change
- Introduce software vport LAG implementation
- Add MPESW over SD LAG support
Deferred init (patches 13-14):
- Tie rep load/unload to SD LAG state
- Defer vport metadata init until SD is ready
Enablement (patch 15):
- Enable SD over ECPF and allow switchdev transition
v2: https://lore.kernel.org/20260608135547.482825-1-tariqt@nvidia.com
v1: https://lore.kernel.org/20260604114455.434711-1-tariqt@nvidia.com
====================
Link: https://patch.msgid.link/20260612113904.537595-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Remove the restriction blocking SD on embedded CPU PFs (ECPF), enabling
SD functionality on BlueField DPUs. Remove the blocker preventing SD
devices from transitioning to switchdev mode.
The infrastructure added in earlier patches properly handles this case.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-16-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Allow SD devices to transition to switchdev before the SD group is
fully up. Metadata allocation requires the SD group to be ready, so
defer it from esw_offloads_enable() until SD shared-FDB activation.
Add mlx5_esw_offloads_init_deferred_metadata() which allocates per-vport
metadata and refreshes the ingress ACLs that were previously programmed
with metadata=0. The helper is idempotent and can be called multiple
times.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-15-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On an SD device, vport representors are not functional until the SD
group is combined and shared FDB is active. Skip the initial load and
the reload paths in that window; reps are loaded as part of the SD LAG
activation flow once it becomes active.
In addition, explicitly unload representors when SD LAG is destroyed.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-14-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Enable MPESW LAG creation over SD LAG members, forming a composite LAG
hierarchy. This allows bonding multiple SD groups together under a
single MPESW configuration with shared FDB.
When enabling composite MPESW, the individual SD LAG shared FDB
configurations are temporarily torn down and recreated when the
composite LAG is disabled.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-13-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
SD LAG is a virtual LAG without hardware LAG support, so it cannot use
the firmware vport LAG commands. Implement a software-based vport LAG
using egress ACL bounce rules.
Add esw_set_slave_egress_rule() to create an egress ACL rule on the
slave's manager vport that bounces traffic to the master's manager
vport. This achieves the same traffic steering as hardware vport LAG.
Redirect mlx5_cmd_create_vport_lag() and mlx5_cmd_destroy_vport_lag()
to the software implementation when operating in SD LAG mode.
In addition, adjust lag_demux creation to check SD LAG mode as well.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-12-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extend mlx5_lag_disable_change() to properly disable both regular LAG
and SD LAG when requested. Each LAG type uses its own devcom component
for locking.
Use mlx5_sd_get_devcom() helper to retrieve the SD devcom component,
needed for proper locking when disabling SD LAG.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-11-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The lag demux resources (flow table, flow group, and rules xarray)
are stored on the shared ldev. With Socket Direct, multiple SD groups
each create their own demux FT/FG during their master's IB device
initialization. Since they all write to the same ldev fields, the
second group's init overwrites the first group's pointers, leaking
the first group's FT/FG.
During teardown, the cleanup uses the overwritten pointers, destroying
the wrong group's resources and leaving leaked flow tables in the LAG
namespace. These leaked tables can interfere with subsequently created
demux tables.
Move the demux resources from the shared ldev to per-master lag_func
instances. Each master device now owns its own independent demux
state. The rule_add and rule_del helpers look up the appropriate
master's lag_func via the existing filter/group infrastructure.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-10-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When eswitch is disabled, notify the SD layer so it can clean up
SD-specific resources such as the TX flow table root configuration
on secondary devices.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-9-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When the eswitch transitions, propagate the change to SD: secondaries
get their TX flow table root reconfigured for the new mode, and when
all group devices move to switchdev, the per-group shared FDB is
activated.
Shared FDB activation is best-effort - failure does not block the
eswitch transition; the next transition retries.
Note: the existing mlx5_get_sd() guard that blocks switchdev for SD
devices is intentionally retained. It will be removed once all
supporting patches are in place.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-8-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In Socket Direct configurations the primary and secondary PFs share the
same native_port_num. The eswitch vport metadata encodes pf_num in its
upper bits to distinguish vports across PFs. Without SD-awareness, both
PFs generate identical metadata, causing FDB rules to steer traffic to
the wrong representor.
Add mlx5_sd_pf_num_get() which remaps the pf_num for SD devices.
Use it so each PF in an SD group produces unique vport metadata.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-7-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add mlx5_fs_cmd_query_l2table_silent() to query the current silent mode
state from firmware. This allows detecting if firmware has already put
secondary devices into silent mode.
During SD group registration, query the silent mode of each device. If
a device is already in silent mode (set by firmware), record this in
the fw_silents_secondaries flag and use it to help determine the
primary/secondary roles.
When fw_silents_secondaries is set, skip the driver-initiated silent
mode set/unset operations since firmware manages this state. This
handles configurations where firmware persistently silences secondary
devices.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Refactor SD group registration to use devcom event-driven role
determination to ensure SD is marked as ready only after roles are fully
assigned and the group state is consistent, making outside accessors,
which will be added in downstream patches, safe to use without races.
The devcom events:
- SD_PRIMARY_SET event: each device compares bus numbers with peers
to determine which should be primary
- SD_SECONDARIES_SET event: secondaries register themselves with the
elected primary device
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Some devcom events are not expected to fail. Rather than attempting
a rollback that may not be meaningful, allow callers to pass
DEVCOM_CANT_FAIL as the rollback_event to indicate that the event
handler should not fail. If it does, emit a warning and stop
propagating to further peers, but skip the rollback path.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-4-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Factor mlx5_devcom_send_event() into two functions:
- mlx5_devcom_locked_send_event(): performs the dispatch (and
rollback) with comp->sem already held by the caller.
- mlx5_devcom_send_event(): unchanged wrapper that takes comp->sem,
calls the locked variant, and releases it.
This lets callers bracket multiple event broadcasts under a single
held write lock, eliminating the gap between consecutive dispatches
where peer state could change.
Will be used by a downstream patch.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-3-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
SD secondary devices share the primary's uplink and do not have
their own uplink representor. When reloading IB reps on secondary
devices, skip the uplink and only load VF/SF vport IB reps.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260612113904.537595-2-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
GRO_LEGACY_MAX_SIZE = 65536; total_len being 65536 is too big to fit
into a u16. As can be seen in skb_gro_receive, packets bigger or equal
to gro_max_size (or GRO_LEGACY_MAX_SIZE) are dropped with -E2BIG. Apply
the same boundary to geneve_post_decap_hint to avoid writing 65536 to a
16-bit iph->tot_len field with an overflow.
Fixes: fd0dd79657 ("geneve: use GRO hint option in the RX path")
Signed-off-by: Alice Mikityanska <alice@isovalent.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20260611192955.604661-3-alice.kernel@fastmail.im
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
