__team_port_change_send() calls __ethtool_get_link_ksettings() on
the port, which will soon take the port's ops lock. The notifier
caller already holds it while the slave-add/del callers do not,
so the function would either deadlock or run unprotected depending
on the path.
Make __team_port_change_send() expect the port's ops lock held and
switch to netif_get_link_ksettings(). team_device_event()'s NETDEV_UP /
NETDEV_CHANGE already arrive with the port's ops lock held.
team_port_add() now take it explicitly.
Note that NETDEV_DOWN and team_port_del() will pass false as @linkup
so they will not execute netif_get_link_ksettings(). This is fortunate
as NETDEV_DOWN has somewhat mixed locking right now.
Link: https://patch.msgid.link/20260603012840.2254293-7-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
bond_update_speed_duplex() calls __ethtool_get_link_ksettings() on
the slave, which will soon take the slave's ops lock. One of its
callers already holds it and the other three don't, so the function
would either deadlock or run unprotected depending on the path.
Make the helper expect the slave's ops lock held and switch to
netif_get_link_ksettings(). Wrap the three call sites that don't
already hold it:
* bond_enslave() (rtnl held; core drops the lower's ops lock
around ->ndo_add_slave).
* bond_miimon_commit() (rtnl_trylock'd from the mii workqueue).
* bond_ethtool_get_link_ksettings() (rtnl held via ethtool layer,
bond device itself is not ops locked).
The call site which does already hold the ops lock is
bond_slave_netdev_event() via NETDEV_UP / NETDEV_CHANGE notifiers,
so it stays as-is.
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Link: https://patch.msgid.link/20260603012840.2254293-6-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
__ethtool_get_link_ksettings() is exported and called from sysfs
and many drivers. It invokes ethtool_ops->get_link_ksettings
so by our own docs it should be holding netdev lock for ops locked
devices. Looks like commit 2bcf4772e4 ("net: ethtool:
try to protect all callback with netdev instance lock")
missed adding the ops lock here.
There's a number of callers we need to fix up so let's add the
netif_get_link_ksettings() helper first, without any actual
locking changes (this commit is a nop).
Not treating this as a fix because I don't think any driver cares
at this point, but if we want to remove the rtnl_lock protection
this will become critical.
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20260603012840.2254293-5-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
NETDEV_CHANGENAME is only emitted from netif_change_name().
netif_change_name() has two callers both of which hold netdev_lock_ops()
around the call site:
- dev_change_name()
- do_setlink()
Document NETDEV_CHANGENAME as always ops locked.
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20260603012840.2254293-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
FW module flashing was written so that the flashing happens
without holding rtnl_lock. This allows flashing multiple modules
at once. Current drivers can handle that well, but we should
let drivers depend on the netdev instance lock. Instance lock
is per netdev, and so is the module so we won't break parallel
updates.
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20260603012840.2254293-3-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jakub suggests renaming the existing assert to match
the netdev_lock_ops_compat() semantics.
We want netdev_assert_locked_ops() to mean - if the driver
is ops locked - check that it's holding the device lock.
The existing helper check for either ops lock or rtnl_lock,
which is the locking behavior of netdev_lock_ops_compat().
The reason for naming divergence is likely that
netdev_ops_assert_locked() predated the _compat() helpers.
Suggested-by: Jakub Sitnicki <jakub@cloudflare.com>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20260603012840.2254293-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dr_icm_buddy_init_ste_cache() allocates the per-buddy miss_list using
the open-coded kvmalloc(n * sizeof(*p), ...) form. The neighbouring
allocations in the same function already use the kvcalloc()/
kvzalloc_objs() forms; switch this last one to kvmalloc_array() for
consistency and for the size_mul overflow check that kvmalloc_array()
performs.
The semantics are unchanged: kvmalloc_array() returns a non-zeroed
buffer, just like the previous kvmalloc() call. Existing callers of
buddy->miss_list initialise each list_head before use.
Signed-off-by: William Theesfeld <william@theesfeld.net>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260601193758.626537-1-william@theesfeld.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The dmaengine path shares ethtool_ops with the legacy AXI DMA path,
including .get_coalesce/.set_coalesce that poke XAXIDMA_*_CR_OFFSET
directly. In dmaengine mode lp->dma_regs is not mapped by axienet, so
those ethtool calls touch unmapped/unrelated memory and report values
unrelated to the channel actually in use.
.get_ringparam/.set_ringparam only touch lp->rx_bd_num/lp->tx_bd_num,
fields used only by the legacy path for BD ring sizing. In dmaengine
mode the descriptor ring is owned by the dmaengine provider and these
fields are not consulted, so reporting them is misleading.
No dmaengine API exists today to query or program either coalescing
or ring size on behalf of the client, so neither can be exposed
meaningfully in dmaengine mode.
Add axienet_ethtool_dmaengine_ops without the coalesce and ringparam
hooks. Also move the ethtool_ops assignment from early probe into the
if/else alongside netdev_ops, so the legacy and dmaengine paths pick
their respective ops in one place. No functional change for the
legacy DMA path.
Signed-off-by: Suraj Gupta <suraj.gupta2@amd.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20260601124454.3384601-1-suraj.gupta2@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are some typos in comments and while they are harmless and not visible,
there is no reason not to fix them. Fix the ones that are not register related,
which might have intentional naming convention.
Signed-off-by: Jakub Raczynski <j.raczynski@samsung.com>
Link: https://patch.msgid.link/20260601163727.554364-1-j.raczynski@samsung.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
rx-copybreak requires an extra slab allocation. Since bnxt uses
page pool frags and HDS by default, the rx-copybreak doesn't
buy us anything. The extra pressure on slab causes overload
on pre-sheaves kernels on modern AMD platforms.
In synthetic testing on net-next this patch shows little difference
but I think copybreak is "obvious waste" at this point.
Default rx-copybreak threshold to 0 / disabled.
The "copybreak" defines are really the size bounds for the Rx header
buffer. Rename them.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20260602003759.1545645-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In ip6_parse_tlv(), recompute the network header pointer once regardless
of the option processed (Hbh or Dest), as missing recomputation for
specific options has caused issues in the past.
Signed-off-by: Justin Iurman <justin.iurman@gmail.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260602213033.12244-1-justin.iurman@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
platform_get_irq() returns a negative errno on failure.
Commit a598f66d91 replaced irq_of_parse_and_map() (which returns 0
on failure) with platform_get_irq() but dropped the error check.
Without it, a negative IRQ number is passed to devm_request_irq(),
which fails with -EINVAL instead of propagating the real error
from platform_get_irq().
Add the missing error check and goto err_gone.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20260601040201.103481-1-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Instead of passing opt_size address to mptcp_established_options(),
change this function to return it by value.
This removes the need for an expensive stack canary in
tcp_established_options() when CONFIG_STACKPROTECTOR_STRONG=y.
$ scripts/bloat-o-meter -t vmlinux.old vmlinux.new
add/remove: 0/0 grow/shrink: 0/3 up/down: 0/-92 (-92)
Function old new delta
tcp_options_write.isra 1423 1407 -16
mptcp_established_options 2746 2720 -26
tcp_established_options 553 503 -50
Total: Before=22110750, After=22110658, chg -0.00%
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260602125138.2317015-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
IPv6 can not be a module anymore, we no longer need to export:
- raw_hash_sk()
- raw_unhash_sk()
- raw_abort()
- raw_seq_start()
- raw_seq_next()
- raw_seq_stop()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20260602165036.2712408-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Kuniyuki Iwashima says:
====================
geneve: Allow binding UDP socket to a specific address.
By default, a GENEVE device bind()s its underlying UDP socket(s) to
the IPv4 or IPv6 wildcard address because there is no way to specify
a specific local IP address to bind() to.
This prevents deploying multiple GENEVE devices on a multi-homed host
where each device should be isolated and bound to a different local IP
address on the same UDP port.
This series introduces two options to specify local IPv4 or IPv6
addresses for a GENEVE device.
====================
Link: https://patch.msgid.link/20260602190436.139591-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
By default, a GENEVE device bind()s its underlying UDP socket(s) to
the IPv4 or IPv6 wildcard address because there is no way to specify
a specific local IP address to bind() to.
This prevents deploying multiple GENEVE devices on a multi-homed host
where each device should be isolated and bound to a different local IP
address on the same UDP port.
Let's introduce new options, IFLA_GENEVE_LOCAL and IFLA_GENEVE_LOCAL6,
to allow specifying a local IPv4/IPv6 address for the backend UDP
socket.
By default, when collect metadata mode (IFLA_GENEVE_COLLECT_METADATA)
is enabled, both IPv4 and IPv6 sockets are created. However, if a
source address is specified via the new attributes, only a single
socket corresponding to that specific address family is created.
Accordingly, geneve_find_sock() and geneve_find_dev() are updated to
take the source address into account, ensuring that multiple devices
and sockets configured with different source addresses can coexist
without conflict.
In addition, the source address is validated in geneve_xmit_skb()
and geneve6_xmit_skb(), so the BPF prog must set it in bpf_tunnel_key.
With this change, multiple GENEVE devices can be successfully created
and bound to their respective local IP addresses:
(*) "local" is the keyword for IFLA_GENEVE_LOCAL / IFLA_GENEVE_LOCAL6
# for i in $(seq 1 2);
do
ip link add geneve4_${i} type geneve local 192.168.0.${i} external
ip addr add 192.168.0.${i}/24 dev geneve4_${i}
ip link set geneve4_${i} up
ip link add geneve6_${i} type geneve local 2001:9292::${i} external
ip addr add 2001:9292::${i}/64 dev geneve6_${i} nodad
ip link set geneve6_${i} up
done
# ip -d l | grep geneve
9: geneve4_1: <BROADCAST,MULTICAST,UP,LOWER_UP> ...
geneve external id 0 local 192.168.0.1 ...
10: geneve6_1: <BROADCAST,MULTICAST,UP,LOWER_UP> ...
geneve external id 0 local 2001:9292::1 ...
11: geneve4_2: <BROADCAST,MULTICAST,UP,LOWER_UP> ...
geneve external id 0 local 192.168.0.2 ...
12: geneve6_2: <BROADCAST,MULTICAST,UP,LOWER_UP> ...
geneve external id 0 local 2001:9292::2 ...
# ss -ua | grep geneve
UNCONN 0 0 192.168.0.2:geneve 0.0.0.0:*
UNCONN 0 0 192.168.0.1:geneve 0.0.0.0:*
UNCONN 0 0 [2001:9292::2]:geneve *:*
UNCONN 0 0 [2001:9292::1]:geneve *:*
Note that even if the local address is explicitly configured with
the wildcard address, kernel does not dump it except for devices with
IFLA_GENEVE_COLLECT_METADATA. This is consistent with the behaviour
of is_tnl_info_zero(), which treats the wildcard remote address as not
configured.
## ynl example.
# ./tools/net/ynl/pyynl/cli.py \
--spec ./Documentation/netlink/specs/rt-link.yaml \
--do newlink --create \
--json '{"ifname": "geneve0",
"linkinfo": {"kind":"geneve",
"data": {"local": "0.0.0.0",
"collect-metadata": true}}}'
# ./tools/net/ynl/pyynl/cli.py \
--spec ./Documentation/netlink/specs/rt-link.yaml \
--do getlink \
--json '{"ifname": "geneve0"}' --output-json | \
jq .linkinfo.data.local
"0.0.0.0"
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260602190436.139591-6-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When collect metadata mode (IFLA_GENEVE_COLLECT_METADATA) is
enabled, the GENEVE device creates both IPv4 and IPv6 sockets
and bind()s them to wildcard addresses.
The next patch allows creating only one socket bound to a
specific address even when the collect metadata mode is
enabled.
Then, we need a flag to distinguish dualstack GENEVE devices
to detect local address conflict.
Let's add the dualstack flag to struct geneve_config.
IFLA_GENEVE_COLLECT_METADATA processing is moved up in
geneve_nl2info() for the next patch to overwrite dualstack
to false while keeping collect_md true.
Note that IFLA_GENEVE_REMOTE and IFLA_GENEVE_REMOTE6 does not
set cfg->dualstack to false since is_tnl_info_zero() ignores
the wildcard remote address:
# ip link add geneve0 type geneve external remote 0.0.0.1
Error: Device is externally controlled, so attributes (VNI, Port, and so on) must not be specified.
# ip link add geneve0 type geneve external remote 0.0.0.0
# ss -ua | grep geneve
UNCONN 0 0 0.0.0.0:geneve 0.0.0.0:*
UNCONN 0 0 *:geneve *:*
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260602190436.139591-5-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is a prep patch to make a subsequent patch clean.
We will need to access geneve_dev->cfg.info.key.u.{ipv4,ipv6}.src
in geneve_find_sock() later and extend conditions there.
Let's pass down struct geneve from geneve_sock_add() to
geneve_find_sock() and flatten the conditional logic.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260602190436.139591-4-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is a prep patch to make a subsequent patch clean.
We will need to access geneve_dev->cfg.info.key.u.{ipv4,ipv6}.src
in geneve_create_sock() later.
Let's pass down struct geneve_dev from geneve_sock_add() to
geneve_create_sock() instead of individual config fields.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260602190436.139591-3-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
geneve_nl2info() calls ipv6_addr_type() to check if the remote
IPv6 address is link-local.
Then, it also calls ipv6_addr_is_multicast() for the same address.
Let's not call ipv6_addr_is_multicast() and reuse ipv6_addr_type().
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260602190436.139591-2-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There's a subtle error with the memcpy here, where b44_gstrings should
not be dereferenced. Dereferening causes the following error with W=1:
In file included from drivers/net/ethernet/broadcom/b44.c:17:
In file included from ./include/linux/module.h:18:
In file included from ./include/linux/kmod.h:9:
In file included from ./include/linux/umh.h:4:
In file included from ./include/linux/gfp.h:7:
In file included from ./include/linux/mmzone.h:8:
In file included from ./include/linux/spinlock.h:56:
In file included from ./include/linux/preempt.h:79:
In file included from ./arch/powerpc/include/asm/preempt.h:5:
In file included from ./include/asm-generic/preempt.h:5:
In file included from ./include/linux/thread_info.h:23:
In file included from ./arch/powerpc/include/asm/current.h:13:
In file included from ./arch/powerpc/include/asm/paca.h:16:
In file included from ./include/linux/string.h:386:
./include/linux/fortify-string.h:578:4: error: call to
'__read_overflow2_field' declared with 'warning' attribute: detected read
beyond size of field (2nd parameter); maybe use>
578 | __read_overflow2_field(q_size_field, size);
| ^
Instead of fixing the memcpy, use ethtool_puts, which is the proper
helper for printing ethtool gstrings.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20260531000334.388351-1-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 1/2
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
Infrastructure (patches 1, 5-6):
- Factor out shared FDB code into a dedicated file
- Extend lag_func with group_id and sd_fdb_active fields;
add XA_MARK_PORT and unified iterator with group_id filter
- Extend shared FDB API with group_id parameter
E-Switch preparation (patches 2-3):
- Align eswitch disable sequence ordering
- Move devcom init from TC to eswitch layer
SD group management (patches 4, 7-9):
- Replace peer count check with direct peer lookup
- Register SD secondaries in the existing LAG at SD init time
- Block RoCE and VF LAG for SD devices
- Block multipath LAG for SD devices
Switchdev integration (patch 10):
- Keep netdev resources local in switchdev mode
Steering (patches 11-12):
- Track peer flow slots with bitmap for selective peer flow deletion
- Enable TC flow steering for SD LAG
Enablement (patch 13):
- Verify unique vhca_id count for cross-VHCA RQT
====================
Link: https://patch.msgid.link/20260531113954.395443-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Change verify_num_vhca_ids() to count the number of unique vhca_ids
and verify this count doesn't exceed max_num_vhca_id, rather than
validating individual vhca_id values are within a specific range.
The previous implementation checked if each vhca_id was in the range
[0, max_num_vhca_id - 1], which is overly restrictive. The hardware
capability max_rqt_vhca_id represents the maximum number of unique
vhca_ids that can be used, not a range constraint on individual IDs.
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/20260531113954.395443-14-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Enable TC flow steering for SD LAG mode by extending multiport
eligibility checks and peer flow handling.
SD LAG operates similarly to MPESW for TC offloads - flows on
secondary devices need peer flow creation on the primary, and
multiport forwarding rules are eligible when either MPESW or SD LAG
is active.
Add mlx5_lag_is_sd() helper to query SD LAG mode, and
mlx5_sd_is_primary() to identify the primary device. Redirect uplink
priv/proto_dev queries to the primary device's eswitch in SD
configurations.
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/20260531113954.395443-13-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
With SD devices joining the LAG, peer flows are not created for all
devcom peers - SD devices skip peers that belong to a different SD
group. However, the delete path iterated all devcom peers
unconditionally, attempting to delete from slots that were never
populated.
Track which peer slots are populated using a bitmap in mlx5e_tc_flow.
The delete path now iterates only set bits, matching exactly the slots
that were set up during flow creation.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260531113954.395443-12-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In SD switchdev mode, network device resources such as channels and
completion vectors must remain on the same PF rather than being
distributed across SD group members.
Modify mlx5_sd_ch_ix_get_dev_ix() to return 0 and
mlx5_sd_ch_ix_get_vec_ix() to return the channel index directly when
in switchdev mode, keeping resources local to the requesting PF.
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/20260531113954.395443-11-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
SD devices are not compatible with multipath LAG since they use
dedicated SD LAG for cross-socket connectivity. Add an SD check
to the multipath prereq validation to prevent multipath LAG
activation on SD-configured ports.
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/20260531113954.395443-10-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Socket Direct devices manage their own LAG via SD LAG infrastructure.
Block the standard netdev-event-driven LAG path (RoCE LAG and VF LAG)
for SD devices to prevent conflicting LAG configurations.
Expose mlx5_sd_is_supported() as a public helper that encapsulates all
SD eligibility checks. Use it in mlx5_lag_dev_alloc() to skip netdev
notifier registration for SD-capable devices at alloc time. Some sd
code is reordered to expose the new function, no logic is changed.
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/20260531113954.395443-9-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Register SD secondary devices with the existing LAG structure by
adding them to the primary's ldev xarray with a shared group_id.
This ties the SD LAG lifecycle to the SD group lifecycle.
Add sd_lag_state debugfs entry for LAG state visibility. To avoid
race between this entry and LAG deletion, have debugfs creation
and deletion done last on SD init and first on SD cleanup.
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/20260531113954.395443-8-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a group_id parameter to mlx5_lag_shared_fdb_create() and
mlx5_lag_shared_fdb_destroy() to scope shared FDB operations to a
specific SD group.
When group_id is U32_MAX, the functions operate on all LAG devices. When
group_id is non-zero, they operate only on devices in that SD group
without issuing FW LAG commands, since SD LAG is a pure software
construct.
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/20260531113954.395443-7-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Socket Direct (SD) secondaries devices will participate in LAG, even
though they are silent. SD secondary devices share the same physical
port as their primary but are separate PCI functions that need to be
tracked alongside regular LAG ports.
Extend lag_func with a group_id field to identify SD group membership
and introduce a unified iterator that can filter by group. Add APIs
for registering SD secondary devices in an existing 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/20260531113954.395443-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Replace mlx5_eswitch_get_npeers() count-based check with a new
mlx5_eswitch_is_peer() function that directly verifies the peer
relationship between two eswitches.
This change prepares for SD LAG support, which is a virtual LAG that
does not have num_lag_ports capability and cannot use the count-based
peer validation.
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/20260531113954.395443-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move the E-swtich devcom component management from TC layer to ESW
layer.
This refactoring places devcom lifecycle management at the appropriate
layer and prepares for SD LAG which needs devcom registration
independent of the TC/representor initialization.
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/20260531113954.395443-4-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch align the eswitch disable sequence with the
switchdev-to-legacy mode transition, where eswitch must be disabled
before device detachment. The consistent ordering is required for proper
SD LAG cleanup which depends on eswitch state during teardown.
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/20260531113954.395443-3-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Refactor shared FDB LAG logic into a new lag/shared_fdb.c file to
improve code organization and enable reuse. Move shared FDB specific
functions from lag.c and introduce consolidated APIs:
- mlx5_lag_shared_fdb_create() handles LAG activation with shared FDB
- mlx5_lag_shared_fdb_destroy() handles LAG deactivation with shared FDB
Update mlx5_do_bond(), mlx5_disable_lag() and mpesw.c to use the new
APIs, which simplifies the shared FDB code paths.
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/20260531113954.395443-2-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The getaddrinfo() call in bind_socket() dynamically allocates memory
for the result linked list that must be freed with freeaddrinfo().
However, none of the code paths after a successful getaddrinfo() call
free this memory, causing a leak in every invocation of bind_socket().
Signed-off-by: longlong yan <yanlonglong@kylinos.cn>
Link: https://patch.msgid.link/20260601013927.1835-1-yanlonglong@kylinos.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Calling hrtimer_start() on an already-active txtimer is unnecessary
and expensive. Skip the restart if the timer is already active by
adding an hrtimer_active() check before hrtimer_start().
Previously, each packet reset the timer to tx_coal_timer in the future,
acting as a sliding window that delayed NAPI under burst traffic. With
this change, an already-active timer is left to fire sooner, scheduling
NAPI within tx_coal_timer of the first packet and freeing TX descriptors
earlier.
There is no race concern: hrtimer_start() is internally serialized and
safe to call on an active timer. In the event of a race between
hrtimer_active() and hrtimer_start(), the worst case is calling
hrtimer_start() on an already-active timer, which is identical to the
pre-patch behaviour.
Performance on Cyclone V with dwmac-socfpga (iperf3 -u -b 0 -l 64):
Before: ~45200 pps
After: ~52300 pps (~15% improvement)
Additionally, ~10% improvement in UDP throughput observed on Agilex5,
with hrtimer CPU usage reduced from ~8% to ~0.6%.
Signed-off-by: Rohan G Thomas <rohan.g.thomas@altera.com>
Tested-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Nazim Amirul <muhammad.nazim.amirul.nazle.asmade@altera.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20260529064659.32287-1-muhammad.nazim.amirul.nazle.asmade@altera.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Ioana Ciornei says:
====================
dpaa2-switch: various improvements
This patch set is a comprised of improvements and fixes for
long-standing bugs which were only caught by sashiko while reviewing the
LAG support patches for the dpaa2-switch:
https://lore.kernel.org/all/20260512131554.952971-1-ioana.ciornei@nxp.com/
In order to not just add to the already big set, I am submitting these
before any v3 of the LAG support patches.
The individual patches tackle FDB and VLAN management in the
dpaa2-switch driver as well as removal of some duplicated code. The
error path of the dpaa2_switch_rx() is also improved.
====================
Link: https://patch.msgid.link/20260528173452.1953102-1-ioana.ciornei@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All the NAPI instances for a DPSW device are attached to the first
switch port's net_device but shared by all ports. The NAPI instances get
disabled only once the last port goes down.
This causes an issue on the .remove() path where each port is
unregistered and freed one at a time, causing the NAPI instances to be
deleted even though they are not disabled.
In order to avoid this, split up the unregister_netdev() calls from the
free_netdev() so that we make sure all ports go down before we attempt
a deletion of NAPI instances. Also, make the netif_napi_del() explicit
as it is on the .probe() path.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Link: https://patch.msgid.link/20260528173452.1953102-6-ioana.ciornei@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The switchdev core notifies the driver on VLAN flag changes on existing
VIDs through the changed field of the switchdev_obj_port_vlan structure.
Without this patch, the driver was erroring out from the start if the
same VID was inserted twice, from its perspective, even though the
second call was actually a flag change.
$ bridge vlan add dev eth2 vid 30 untagged
$ bridge vlan add dev eth2 vid 30
[ 458.589534] fsl_dpaa2_switch dpsw.0 eth2: VLAN 30 already configured
This patch fixes the above behavior by, first of all, removing the
checks and return of errors on a VLAN already being installed. The patch
also moves the sequence of code which checks if there is space for a new
VLAN so that the verification is being done only for VLANs not know by
the switch and not flag changes.
A new parameter is added to the dpaa2_switch_port_add_vlan() function so
that we pass the vlan->changed necessary information. Based on this new
parameter and the flags value, the untagged flag will be added or
removed from a VLAN installed on a port. The same thing is also extended
to the PVID configuration.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Link: https://patch.msgid.link/20260528173452.1953102-5-ioana.ciornei@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The check for the maximum number of VLANs is exactly duplicated twice in
the dpaa2_switch_port_vlans_add() function. Remove one of the instances
so that we do not have dead code.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Link: https://patch.msgid.link/20260528173452.1953102-4-ioana.ciornei@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In case of an error in dpaa2_switch_rx(), the dpaa2_switch_free_fd()
function is called in order to free the FD. This is incorrect since the
dpaa2_switch_free_fd() is intended to be used on Tx frame descriptors,
meaning that it expects in the software annotation area of the FD data
to find a valid skb pointer on which to call dev_kfree_skb().
Fix this by extracting the dma_unmap_page() from
dpaa2_switch_build_linear_skb() directly into the dpaa2_switch_rx()
function. This allows us to directly use free_pages() in case of an
error before an SKB was created and kfree_skb() afterwards.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Link: https://patch.msgid.link/20260528173452.1953102-3-ioana.ciornei@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On bridge leave, the dpaa2_switch_port_set_fdb() function always
allocates a new FDB for the port which is becoming standalone. In case
no FDB is found, then the port leaving a bridge will continue to use the
current one.
The above logic does not cover the case in which there are multiple
bridges which have ports from the same DPSW instance. In this case, when
the last port leaves bridge #1, it finds an unused FDB to switch to, but
the old FDB is not marked as unused. Since the number of FDBs is equal
to the number of DPSW interfaces, this will eventually lead to multiple
ports sharing the same FDB.
Fix this by changing how we are managing the FDBs on the leave path.
Instead of directly allocating a new FDB, first verify if the current
port is the last one to leave a bridge. If this is the case, then
continue to use the current FDB and only allocate another FDB if there
are other ports remaining in the bridge.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Link: https://patch.msgid.link/20260528173452.1953102-2-ioana.ciornei@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
dsa_port_from_netdev() may return a valid port from a different switch
chip. Programming another chip's port index into the local hardware
causes redirection to the wrong port, or an out-of-bounds access if the
index exceeds the local chip's port count.
Apply a minimal fix that adds a check to catch this case and adjusts the
extack message. When cls->common.skip_sw is not set, the operation could
instead redirect to the upstream port and let the software or upstream
switch(es) handle the forward, but that is not addressed here.
Signed-off-by: David Yang <mmyangfl@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/20260530003940.2000994-1-mmyangfl@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add the missing @may_sleep parameter description to the
mpc52xx_fec_stop kernel-doc comment.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: https://patch.msgid.link/20260531000042.369043-1-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add dec_ttl action support to the OVS kernel datapath selftest
framework:
- Add dec_ttl nested NLA class to ovs-dpctl.py with proper
OVS_DEC_TTL_ATTR_ACTION sub-attribute handling
- Add parse support for dec_ttl(le_1(<inner_actions>)) action
string, consistent with the odp-util.c format where le_1()
holds the actions taken when TTL reaches 1
- Add dpstr output formatting for dec_ttl actions
- Add test_dec_ttl() to openvswitch.sh that verifies:
* Normal TTL packets are forwarded after decrement
* TTL=1 packets are dropped (TTL expiry)
* Graceful skip via ksft_skip if kernel lacks dec_ttl support
The dec_ttl class uses late-binding type resolution to reference
ovsactions for its inner action list, avoiding circular references
at class definition time.
Signed-off-by: Minxi Hou <houminxi@gmail.com>
Reviewed-by: Aaron Conole <aconole@redhat.com>
Link: https://patch.msgid.link/20260530021443.1734484-1-houminxi@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan says:
====================
net/mlx5: Avoid payload in skb's linear part for better GRO-processing
This is V7 of a series originally submitted by Christoph.
When LRO is enabled on the MLX, mlx5e_skb_from_cqe_mpwrq_nonlinear
copies parts of the payload to the linear part of the skb.
This triggers suboptimal processing in GRO, causing slow throughput.
This patch series addresses this by using eth_get_headlen to compute the
size of the protocol headers and only copy those bits. This results in a
significant throughput improvement (detailed results in the specific
patch).
====================
Link: https://patch.msgid.link/20260601061522.398044-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
mlx5e_skb_from_cqe_mpwrq_nonlinear() copies MLX5E_RX_MAX_HEAD (256)
bytes from the page-pool to the skb's linear part. Those 256 bytes
include part of the payload.
When attempting to do GRO in skb_gro_receive, if headlen > data_offset
(and skb->head_frag is not set), we end up aggregating packets in the
frag_list.
This is of course not good when we are CPU-limited. Also causes a worse
skb->len/truesize ratio,...
So, let's avoid copying parts of the payload to the linear part. We use
eth_get_headlen() to parse the headers and compute the length of the
protocol headers, which will be used to copy the relevant bits of the
skb's linear part.
We still allocate MLX5E_RX_MAX_HEAD for the skb so that if the networking
stack needs to call pskb_may_pull() later on, we don't need to reallocate
memory.
This gives a nice throughput increase (ARM Neoverse-V2 with CX-7 NIC and
LRO enabled):
BEFORE:
=======
(netserver pinned to core receiving interrupts)
$ netperf -H 10.221.81.118 -T 80,9 -P 0 -l 60 -- -m 256K -M 256K
87380 16384 262144 60.01 32547.82
(netserver pinned to adjacent core receiving interrupts)
$ netperf -H 10.221.81.118 -T 80,10 -P 0 -l 60 -- -m 256K -M 256K
87380 16384 262144 60.00 52531.67
AFTER:
======
(netserver pinned to core receiving interrupts)
$ netperf -H 10.221.81.118 -T 80,9 -P 0 -l 60 -- -m 256K -M 256K
87380 16384 262144 60.00 52896.06
(netserver pinned to adjacent core receiving interrupts)
$ netperf -H 10.221.81.118 -T 80,10 -P 0 -l 60 -- -m 256K -M 256K
87380 16384 262144 60.00 85094.90
Additional tests across a larger range of parameters w/ and w/o LRO, w/
and w/o IPv6-encapsulation, different MTUs (1500, 4096, 9000), different
TCP read/write-sizes as well as UDP benchmarks, all have shown equal or
better performance with this patch.
For XDP pull at most ETH_HLEN bytes in the linear area so that XDP_PASS
can also benefit from this improvement and keep things simple when
dealing with skb geometry changes from the XDP program.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Christoph Paasch <cpaasch@openai.com>
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260601061522.398044-3-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Doing the call to dma_sync_single_for_cpu() earlier will allow us to
adjust headlen based on the actual size of the protocol headers.
Doing this earlier means that we don't need to call
mlx5e_copy_skb_header() anymore and rather can call
skb_copy_to_linear_data() directly.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Christoph Paasch <cpaasch@openai.com>
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260601061522.398044-2-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Lorenzo Bianconi says:
====================
net: airoha: Preliminary patches to support multiple net_devices connected to the same GDM port
EN7581 or AN7583 SoCs support connecting multiple external SerDes (e.g.
Ethernet or USB SerDes) to GDM3 or GDM4 ports via a hw arbiter that
manages the traffic in a TDM manner. As a result multiple net_devices can
connect to the same GDM{3,4} port and there is a theoretical "1:n"
relation between GDM ports and net_devices.
┌─────────────────────────────────┐
│ │ ┌──────┐
│ P1 GDM1 ├────►MT7530│
│ │ └──────┘
│ │ ETH0 (DSA conduit)
│ │
│ PSE/FE │
│ │
│ │
│ │ ┌─────┐
│ P0 CDM1 ├────►QDMA0│
│ P4 P9 GDM4 │ └─────┘
└──┬─────────────────────────┬────┘
│ │
┌──▼──┐ ┌────▼────┐
│ PPE │ │ ARB │
└─────┘ └─┬─────┬─┘
│ │
┌──▼──┐┌─▼───┐
│ ETH ││ USB │
└─────┘└─────┘
ETH1 ETH2
This is a preliminary series to introduce support for multiple net_devices
connected to the same Frame Engine (FE) GDM port (GDM3 or GDM4) via an
external hw arbiter.
====================
Link: https://patch.msgid.link/20260527-airoha-eth-multi-serdes-preliminary-v1-0-ec6ed73ef7fc@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
