Alexandra Winter [Thu, 18 Sep 2025 11:04:52 +0000 (13:04 +0200)]
dibs: Define dibs loopback
The first stage of loopback-ism was implemented as part of the
SMC module [1]. Now that we have the dibs layer, provide access to a
dibs_loopback device to all dibs clients.
This is the first step of moving loopback-ism from net/smc/smc_loopback.*
to drivers/dibs/dibs_loopback.*. One global structure lo_dev is allocated
and added to the dibs devices. Follow-on patches will move functionality.
Same as smc_loopback, dibs_loopback is provided by a config option.
Note that there is no way to dynamically add or remove the loopback
device. That could be a future improvement.
When moving code to drivers/dibs, replace ism_ prefix with dibs_ prefix.
As this is mostly a move of existing code, copyright and authors are
unchanged.
Alexandra Winter [Thu, 18 Sep 2025 11:04:51 +0000 (13:04 +0200)]
dibs: Register ism as dibs device
Register ism devices with the dibs layer. Follow-on patches will move
functionality to the dibs layer.
As DIBS is only a shim layer without any dependencies, we can depend ISM
on DIBS without adding indirect dependencies. A follow-on patch will
remove implication of SMC by ISM.
Define struct dibs_dev. Follow-on patches will move more content into
dibs_dev. The goal of follow-on patches is that ism_dev will only
contain fields that are special for this device driver. The same concept
will apply to other dibs device drivers.
Define dibs_dev_alloc(), dibs_dev_add() and dibs_dev_del() to be called
by dibs device drivers and call them from ism_drv.c
Use ism_dev.dibs for a pointer to dibs_dev.
Alexandra Winter [Thu, 18 Sep 2025 11:04:50 +0000 (13:04 +0200)]
dibs: Register smc as dibs_client
Formally register smc as dibs client. Functionality will be moved by
follow-on patches from ism_client to dibs_client until eventually
ism_client can be removed.
As DIBS is only a shim layer without any dependencies, we can depend SMC
on DIBS without adding indirect dependencies. A follow-on patch will
remove dependency of SMC on ISM.
Alexandra Winter [Thu, 18 Sep 2025 11:04:49 +0000 (13:04 +0200)]
dibs: Create drivers/dibs
Create the file structure for a 'DIBS - Direct Internal Buffer Sharing'
shim layer that will provide generic functionality and declarations for
dibs device drivers and dibs clients.
Alexandra Winter [Thu, 18 Sep 2025 11:04:48 +0000 (13:04 +0200)]
net/smc: Decouple sf and attached send_buf in smc_loopback
Before this patch there was the following assumption in
smc_loopback.c>smc_lo_move_data():
sf (signalling flag) == 0 : data is already in an attached target dmb
sf == 1 : data is not yet in the target dmb
This is true for the 2 callers in smc client
smcd_cdc_msg_send() : sf=1
smcd_tx_rdma_writes() : sf=0
but should not be a general assumption.
Add a bool to struct smc_buf_desc to indicate whether an SMC-D sndbuf_desc
is an attached buffer. Don't call move_data() for attached send_buffers,
because it is not necessary.
Move the data in smc_lo_move_data() if len != 0 and signal when requested.
====================
tcp: Update bind bucket state on port release
TL;DR
-----
This is another take on addressing the issue we already raised earlier [1].
This time around, instead of trying to relax the bind-conflict checks in
connect(), we make an attempt to fix the tcp bind bucket state accounting.
The goal of this patch set is to make the bind buckets return to "port
reusable by ephemeral connections" state when all sockets blocking the port
from reuse get unhashed.
Changelog
---------
Changes in v5:
- Fix initial port-addr bucket state on saddr update with ip_dynaddr=1
- Add Kuniyuki's tag for tests
- Link to v4: https://lore.kernel.org/r/20250913-update-bind-bucket-state-on-unhash-v4-0-33a567594df7@cloudflare.com
Changes in v4:
- Drop redundant sk_is_connect_bind helper doc comment
- Link to v3: https://lore.kernel.org/r/20250910-update-bind-bucket-state-on-unhash-v3-0-023caaf4ae3c@cloudflare.com
Changes in v3:
- Move the flag from inet_flags to sk_userlocks (Kuniyuki)
- Rename the flag from AUTOBIND to CONNECT_BIND to avoid a name clash (Kuniyuki)
- Drop unreachable code for sk_state == TCP_NEW_SYN_RECV (Kuniyuki)
- Move the helper to inet_hashtables where it's used
- Reword patch 1 description for conciseness
- Link to v2: https://lore.kernel.org/r/20250821-update-bind-bucket-state-on-unhash-v2-0-0c204543a522@cloudflare.com
Changes in v2:
- Rename the inet_sock flag from LAZY_BIND to AUTOBIND (Eric)
- Clear the AUTOBIND flag on disconnect path (Eric)
- Add a test to cover the disconnect case (Eric)
- Link to RFC v1: https://lore.kernel.org/r/20250808-update-bind-bucket-state-on-unhash-v1-0-faf85099d61b@cloudflare.com
Situation
---------
We observe the following scenario in production:
inet_bind_bucket
state for port 54321
--------------------
// Process B opens a short-lived connection:
s2 = socket(AF_INET, SOCK_STREAM)
s2.setsockopt(SO_REUSEADDR)
s2.bind(192.0.2.20, 0)
tb->fastreuse = 0
tb->fastreuseport = 0
s2.connect(192.51.100.2, 53)
s2.getsockname() -> 192.0.2.20:54321
s2.send()
s2.recv()
s2.close()
// bucket remains in this
// state even though port
// was released by s2
tb->fastreuse = 0
tb->fastreuseport = 0
// Process A attempts to open another connection
// when there is connection pressure from
// 192.0.2.30:54000..54500 to 192.51.100.1:443.
// Assume only port 54321 is still available.
We end up in a state where Process A can't reuse ephemeral port 54321 for
as long as there are sockets, like s1, that keep the bind bucket alive. The
bucket does not return to "reusable" state even when all sockets which
blocked it from reuse, like s2, are gone.
The ephemeral port becomes available for use again only after all sockets
bound to it are gone and the bind bucket is destroyed.
Programs which behave like Process B in this scenario - that is, binding to
an IP address without setting IP_BIND_ADDRESS_NO_PORT - might be considered
poorly written. However, the reality is that such implementation is not
actually uncommon. Trying to fix each and every such program is like
playing whack-a-mole.
For instance, it could be any software using Golang's net.Dialer with
LocalAddr provided:
Jakub Sitnicki [Wed, 17 Sep 2025 13:22:05 +0000 (15:22 +0200)]
selftests/net: Test tcp port reuse after unbinding a socket
Exercise the scenario described in detail in the cover letter:
1) socket A: connect() from ephemeral port X
2) socket B: explicitly bind() to port X
3) check that port X is now excluded from ephemeral ports
4) close socket B to release the port bind
5) socket C: connect() from ephemeral port X
As well as a corner case to test that the connect-bind flag is cleared:
1) connect() from ephemeral port X
2) disconnect the socket with connect(AF_UNSPEC)
3) bind() it explicitly to port X
4) check that port X is now excluded from ephemeral ports
Jakub Sitnicki [Wed, 17 Sep 2025 13:22:04 +0000 (15:22 +0200)]
tcp: Update bind bucket state on port release
Today, once an inet_bind_bucket enters a state where fastreuse >= 0 or
fastreuseport >= 0 after a socket is explicitly bound to a port, it remains
in that state until all sockets are removed and the bucket is destroyed.
In this state, the bucket is skipped during ephemeral port selection in
connect(). For applications using a reduced ephemeral port
range (IP_LOCAL_PORT_RANGE socket option), this can cause faster port
exhaustion since blocked buckets are excluded from reuse.
The reason the bucket state isn't updated on port release is unclear.
Possibly a performance trade-off to avoid scanning bucket owners, or just
an oversight.
Fix it by recalculating the bucket state when a socket releases a port. To
limit overhead, each inet_bind2_bucket stores its own (fastreuse,
fastreuseport) state. On port release, only the relevant port-addr bucket
is scanned, and the overall state is derived from these.
Jakub Kicinski [Tue, 23 Sep 2025 00:51:37 +0000 (17:51 -0700)]
Merge branch 'add-more-functionality-to-bnge'
Bhargava Marreddy says:
====================
Add more functionality to BNGE
This patch series adds the infrastructure to make the netdevice
functional. It allocates data structures for core resources,
followed by their initialisation and registration with the firmware.
The core resources include the RX, TX, AGG, CMPL, and NQ rings,
as well as the VNIC. RX/TX functionality will be introduced in the
next patch series to keep this one at a reviewable size.
====================
Add functions to add a filter to the VNIC to configure unicast
addresses. Also, add multicast, broadcast, and promiscuous settings
to the default VNIC.
Allocate the default VNIC with the firmware and configure its RSS,
HDS, and Jumbo parameters. Add related functions to support VNIC
configuration for these parameters.
Allocate the hardware statistics context with the firmware and
register DMA memory required for ring statistics. This helps the
driver to collect ring statistics provided by the firmware.
Add initial settings to all core resources, such as
the RX, AGG, TX, CQ, and NQ rings, as well as the VNIC.
This will help enable these resources in future patches.
Let's consider a nohz_full system with isolated CPUs: wq_unbound_cpumask is
set to the housekeeping CPUs, for !WQ_UNBOUND the local CPU is selected.
This leads to different scenarios if a work item is scheduled on an isolated
CPU where "delay" value is 0 or greater then 0:
schedule_delayed_work(, 0);
This will be handled by __queue_work() that will queue the work item on the
current local (isolated) CPU, while:
schedule_delayed_work(, 1);
Will move the timer on an housekeeping CPU, and schedule the work there.
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
=== Plan and future plans ===
This patchset is the first stone on a refactoring needed in order to
address the points aforementioned; it will have a positive impact also
on the cpu isolation, in the long term, moving away percpu workqueue in
favor to an unbound model.
These are the main steps:
1) API refactoring (that this patch is introducing)
- Make more clear and uniform the system wq names, both per-cpu and
unbound. This to avoid any possible confusion on what should be
used.
- Introduction of WQ_PERCPU: this flag is the complement of WQ_UNBOUND,
introduced in this patchset and used on all the callers that are not
currently using WQ_UNBOUND.
WQ_UNBOUND will be removed in a future release cycle.
Most users don't need to be per-cpu, because they don't have
locality requirements, because of that, a next future step will be
make "unbound" the default behavior.
2) Check who really needs to be per-cpu
- Remove the WQ_PERCPU flag when is not strictly required.
3) Add a new API (prefer local cpu)
- There are users that don't require a local execution, like mentioned
above; despite that, local execution yeld to performance gain.
This new API will prefer the local execution, without requiring it.
=== Introduced Changes by this series ===
1) [P 1-2] Replace use of system_wq and system_unbound_wq
system_wq is a per-CPU workqueue, but his name is not clear.
system_unbound_wq is to be used when locality is not required.
Because of that, system_wq has been renamed in system_percpu_wq, and
system_unbound_wq has been renamed in system_dfl_wq.
2) [P 3] add WQ_PERCPU to remaining alloc_workqueue() users
Every alloc_workqueue() caller should use one among WQ_PERCPU or
WQ_UNBOUND.
WQ_UNBOUND will be removed in a next release cycle.
====================
Marco Crivellari [Thu, 18 Sep 2025 14:24:27 +0000 (16:24 +0200)]
net: WQ_PERCPU added to alloc_workqueue users
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
alloc_workqueue() treats all queues as per-CPU by default, while unbound
workqueues must opt-in via WQ_UNBOUND.
This default is suboptimal: most workloads benefit from unbound queues,
allowing the scheduler to place worker threads where they’re needed and
reducing noise when CPUs are isolated.
This change adds a new WQ_PERCPU flag at the network subsystem, to explicitly
request the use of the per-CPU behavior. Both flags coexist for one release
cycle to allow callers to transition their calls.
Once migration is complete, WQ_UNBOUND can be removed and unbound will
become the implicit default.
With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU.
Marco Crivellari [Thu, 18 Sep 2025 14:24:26 +0000 (16:24 +0200)]
net: replace use of system_wq with system_percpu_wq
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
system_unbound_wq should be the default workqueue so as not to enforce
locality constraints for random work whenever it's not required.
Adding system_dfl_wq to encourage its use when unbound work should be used.
The old system_unbound_wq will be kept for a few release cycles.
Marco Crivellari [Thu, 18 Sep 2025 14:24:25 +0000 (16:24 +0200)]
net: replace use of system_unbound_wq with system_dfl_wq
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
system_unbound_wq should be the default workqueue so as not to enforce
locality constraints for random work whenever it's not required.
Adding system_dfl_wq to encourage its use when unbound work should be used.
The old system_unbound_wq will be kept for a few release cycles.
Paul adds support for Earliest TxTime First (ETF) hardware offload
for E830 devices on ice. ETF is configured per-queue using tc-etf Qdisc;
a new Tx flow mechanism utilizes a dedicated timestamp ring alongside
the standard Tx ring. The timestamp ring contains descriptors that
specify when hardware should transmit packets; up to 2048 Tx queues can
be supported.
Dave removes excess cleanup call to ice_lag_move_new_vf_nodes() in error
path.
Milena adds reporting of timestamping statistics to idpf.
Alex changes error variable type for code clarity for iavf and ixgbevf.
Brahmajit Das removes unused parameter from fm10k_unbind_hw_stats_q().
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
net: intel: fm10k: Fix parameter idx set but not used
ixgbevf: fix proper type for error code in ixgbevf_resume()
iavf: fix proper type for error code in iavf_resume()
idpf: add HW timestamping statistics
ice: Remove deprecated ice_lag_move_new_vf_nodes() call
ice: add E830 Earliest TxTime First Offload support
ice: move ice_qp_[ena|dis] for reuse
====================
net: xilinx: axienet: Fix kernel-doc warnings for missing return descriptions
Add missing "Return:" sections to kernel-doc comments for four functions:
- axienet_calc_cr()
- axienet_device_reset()
- axienet_free_tx_chain()
- axienet_dim_coalesce_count_rx()
Also standardize the return documentation format by replacing inline
"Returns" text with proper "Return:" tags as per kernel documentation
guidelines.
Fixes below kernel-doc warnings:
- Warning: No description found for return value of 'axienet_calc_cr'
- Warning: No description found for return value of 'axienet_device_reset'
- Warning: No description found for return value of 'axienet_free_tx_chain'
- Warning: No description found for return value of
'axienet_dim_coalesce_count_rx'
====================
net: dsa: microchip: Add strap description to set SPI as interface bus
At reset, the KSZ8463 uses a strap-based configuration to set SPI as
interface bus. If the required pull-ups/pull-downs are missing
(by mistake or by design to save power) the pins may float and the
configuration can go wrong preventing any communication with the switch.
This small series aims to allow to configure the KSZ8463 switch at
reset when the hardware straps are missing.
PATCH 0 and 1 add a new property to the bindings that describes the GPIOs
to be set during reset in order to configure the switch properly.
PATCH 2 implements the use of these properties in the driver.
====================
net: dsa: microchip: Set SPI as bus interface during reset for KSZ8463
At reset, the KSZ8463 uses a strap-based configuration to set SPI as
bus interface. SPI is the only bus supported by the driver. If the
required pull-ups/pull-downs are missing (by mistake or by design to
save power) the pins may float and the configuration can go wrong
preventing any communication with the switch.
Introduce a ksz8463_configure_straps_spi() function called during the
device reset. It relies on the 'straps-rxd-gpios' OF property and the
'reset' pinmux configuration to enforce SPI as bus interface.
dt-bindings: net: dsa: microchip: Add strap description to set SPI mode
At reset, KSZ8463 uses a strap-based configuration to set SPI as
interface bus. If the required pull-ups/pull-downs are missing (by
mistake or by design to save power) the pins may float and the
configuration can go wrong preventing any communication with the switch.
Add a 'reset' pinmux state
Add a KSZ8463 specific strap description that can be used by the driver
to drive the strap pins during reset. Two GPIOs are used. Users must
describe either both of them or none of them.
Jakub Kicinski [Mon, 22 Sep 2025 23:30:17 +0000 (16:30 -0700)]
Merge tag 'mlx5-next-counters' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux
Tariq Toukan says:
====================
mlx5-next updates 2025-09-21
* tag 'mlx5-next-counters' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux:
net/mlx5: Add uar access and odp page fault counters
====================
====================
net: rework SFP capability parsing and quirks
The original SPF module parsing was implemented prior to gaining any
quirks, and was designed such that the upstream calls the parsing
functions to get the translated capabilities of the module.
SFP quirks were then added to cope with modules that didn't correctly
fill out their ID EEPROM. The quirk function was called from
sfp_parse_support() to allow quirks to modify the ethtool link mode
masks.
Using just ethtool link mode masks eventually lead to difficulties
determining the correct phy_interface_t mode, so a bitmap of these
modes were added - needing both the upstream API and quirks to be
updated.
We have had significantly more SFP module quirks added since, some
which are modifying the ID EEPROM as a way of influencing the data
we provide to the upstream - for example, sfp_fixup_10gbaset_30m()
changes id.base.connector so we report PORT_TP. This could be done
more cleanly if the quirks had access to the parsed SFP port.
In order to improve flexibility, and to simplify some of the upstream
code, we group all module capabilities into a single structure that
the upstream can access via sfp_module_get_caps(). This will allow
the module capabilities to be expanded if required without reworking
all the infrastructure and upstreams again.
In this series, we rework the SFP code to use the capability structure
and then rework all the upstream implementations, finally removing the
old kernel internal APIs.
====================
net: sfp: convert sfp quirks to modify struct sfp_module_support
In order to provide extensible module support properties, arrange for
the SFP quirks to modify any member of the sfp_module_support struct,
rather than just the ethtool link modes and interfaces.
mptcp: remove unused returned value of check_data_fin
When working on a fix modifying mptcp_check_data_fin(), I noticed the
returned value was no longer used.
It looks like it was used for 3 days, between commit 7ed90803a213
("mptcp: send explicit ack on delayed ack_seq incr") and commit ea4ca586b16f ("mptcp: refine MPTCP-level ack scheduling").
mptcp: pm: netlink: only add server-side attr when true
This attribute is a boolean. No need to add it to set it to 'false'.
Indeed, the default value when this attribute is not set is naturally
'false'. A few bytes can then be saved by not adding this attribute if
the connection is not on the server side.
This prepares the future deprecation of its attribute, in favour of a
new flag.
Vivian Wang [Fri, 19 Sep 2025 12:04:33 +0000 (20:04 +0800)]
net: spacemit: Make stats_lock softirq-safe
While most of the statistics functions (emac_get_stats64() and such) are
called with softirqs enabled, emac_stats_timer() is, as its name
suggests, also called from a timer, i.e. called in softirq context.
All of these take stats_lock. Therefore, make stats_lock softirq-safe by
changing spin_lock() into spin_lock_bh() for the functions that get
statistics.
Also, instead of directly calling emac_stats_timer() in emac_up() and
emac_resume(), set the timer to trigger instead, so that
emac_stats_timer() is only called from the timer. It will keep using
spin_lock().
This fixes a lockdep warning, and potential deadlock when stats_timer is
triggered in the middle of getting statistics.
Fixes: bfec6d7f2001 ("net: spacemit: Add K1 Ethernet MAC") Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> Closes: https://lore.kernel.org/all/a52c0cf5-0444-41aa-b061-a0a1d72b02fe@samsung.com/ Signed-off-by: Vivian Wang <wangruikang@iscas.ac.cn> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://patch.msgid.link/20250919-k1-ethernet-fix-lock-v1-1-c8b700aa4954@iscas.ac.cn Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
net: enetc: improve the interface for obtaining phc_index
The first patch is to fix the issue that a sleeping function is called
in the context of rcu_read_lock(). The second patch is to use the generic
API instead of the custom API to get phc_index. In addition, the second
patch depends on the first patch to work.
net: enetc: use generic interfaces to get phc_index for ENETC v1
The commit 61f132ca8c46 ("ptp: add helpers to get the phc_index by
of_node or dev") has added two generic interfaces to get the phc_index
of the PTP clock. This eliminates the need for PTP device drivers to
provide custom APIs for consumers to retrieve the phc_index. This has
already been implemented for ENETC v4 and is also applicable to ENETC
v1. Therefore, the global variable enetc_phc_index is removed from the
driver. ENETC v1 now uses the same interface as v4 to get phc_index.
Signed-off-by: Wei Fang <wei.fang@nxp.com> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com> Tested-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://patch.msgid.link/20250919084509.1846513-3-wei.fang@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net: enetc: fix sleeping function called from rcu_read_lock() context
The rcu_read_lock() has been introduced in __ethtool_get_ts_info() since
the commit 4c61d809cf60 ("net: ethtool: Fix suspicious rcu_dereference
usage"). Therefore, the device drivers cannot use any sleeping functions
when implementing the callback of ethtool_ops::get_ts_info(). Currently,
pci_get_slot() is used in enetc_get_ts_info(), but it calls down_read()
which might sleep, so this is a potential issue. Therefore, to fix this
issue, pci_get_domain_bus_and_slot() is used to replace pci_get_slot()
in enetc_get_ts_info().
Reported-by: Vladimir Oltean <vladimir.oltean@nxp.com> Closes: https://lore.kernel.org/netdev/20250918124823.t3xlzn7w2glzkhnx@skbuf/ Fixes: f5b9a1cde0a2 ("net: enetc: add PTP synchronization support for ENETC v4") Signed-off-by: Wei Fang <wei.fang@nxp.com> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com> Tested-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://patch.msgid.link/20250919084509.1846513-2-wei.fang@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
tcp: Clean up inet_hash() and inet_unhash().
While reviewing the ehash fix series from Xuanqiang Luo [0],
I noticed that inet_twsk_hashdance_schedule() checks the
retval of __sk_nulls_del_node_init_rcu(), which looks confusing.
tcp: Remove redundant sk_unhashed() in inet_unhash().
inet_unhash() checks sk_unhashed() twice at the entry and after locking
ehash/lhash bucket.
The former was somehow added redundantly by commit 4f9bf2a2f5aa ("tcp:
Don't acquire inet_listen_hashbucket::lock with disabled BH.").
inet_unhash() is called for the full socket from 4 places, and it is
always under lock_sock() or the socket is not yet published to other
threads:
1. __sk_prot_rehash()
-> called from inet_sk_reselect_saddr(), which has
lockdep_sock_is_held()
2. sk_common_release()
-> called when inet_create() or inet6_create() fail, then the
socket is not yet published
3. tcp_set_state()
-> calls tcp_call_bpf_2arg(), and tcp_call_bpf() has
sock_owned_by_me()
4. inet_ctl_sock_create()
-> creates a kernel socket and unhashes it immediately, but TCP
socket is not hashed in sock_create_kern() (only SOCK_RAW is)
So we do not need to check sk_unhashed() twice before/after ehash/lhash
lock in inet_unhash().
David Yang [Fri, 19 Sep 2025 05:35:33 +0000 (13:35 +0800)]
selftests: forwarding: Reorder (ar)ping arguments to obey POSIX getopt
Quoted from musl wiki:
GNU getopt permutes argv to pull options to the front, ahead of
non-option arguments. musl and the POSIX standard getopt stop
processing options at the first non-option argument with no
permutation.
Thus these scripts stop working on musl since non-option arguments for
tools using getopt() (in this case, (ar)ping) do not always come last.
Fix it by reordering arguments.
====================
net: netpoll: remove dead code and speed up rtnl-locked region
This patchset introduces two minor modernizations to the netpoll
infrastructure:
The first patch removes the unused netpoll pointer from the netpoll_info
structure. This member is redundant and its presence does not benefit
multi-instance setups, as reported by Jay Vosburgh. Eliminating it cleans up
the structure and removes unnecessary code.
The second patch updates the netpoll resource cleanup routine to use
synchronize_net() instead of synchronize_rcu(). As __netpoll_free() is always
called under the RTNL lock, using synchronize_net() leverages the more
efficient synchronize_rcu_expedited() in these contexts, reducing time spent in
critical sections and improving performance.
Both changes simplify maintenance and enhance efficiency without altering
netpoll behavior.
====================
net: netpoll: use synchronize_net() instead of synchronize_rcu()
Replace synchronize_rcu() with synchronize_net() in __netpoll_free().
synchronize_net() is RTNL-aware and will use the more efficient
synchronize_rcu_expedited() when called under RTNL lock, avoiding
the potentially expensive synchronize_rcu() in RTNL critical sections.
Since __netpoll_free() is called with RTNL held (as indicated by
ASSERT_RTNL()), this change improves performance by reducing the
time spent in the RTNL critical section.
net: netpoll: remove unused netpoll pointer from netpoll_info
The netpoll_info structure contains an useless pointer back to its
associated netpoll. This field is never used, and the assignment in
__netpoll_setup() is does not comtemplate multiple instances, as
reported by Jay[1].
Drop both the member and its initialization to simplify the structure.
Antoine Tenart [Mon, 15 Sep 2025 09:19:55 +0000 (11:19 +0200)]
net: ipv4: simplify drop reason handling in ip_rcv_finish_core
Instead of setting the drop reason to SKB_DROP_REASON_NOT_SPECIFIED
early and having to reset it each time it is overridden by a function
returned value, just set the drop reason to the expected value before
returning from ip_rcv_finish_core.
Antoine Tenart [Mon, 15 Sep 2025 09:19:54 +0000 (11:19 +0200)]
net: ipv4: make udp_v4_early_demux explicitly return drop reason
udp_v4_early_demux already returns drop reasons as it either returns 0
or ip_mc_validate_source, which itself returns drop reasons. Its return
value is also already used as a drop reason itself.
Makes this explicit by making it return drop reasons.
Various network interface types make use of needed_{head,tail}room values
to efficiently reserve buffer space for additional encapsulation headers,
such as VXLAN, Geneve, IPSec, etc. However, it is not currently possible
to query these values in a generic way.
Introduce ability to query the needed_{head,tail}room values of a network
device via rtnetlink, such that applications that may wish to use these
values can do so.
For example, Cilium agent iterates over present devices based on user config
(direct routing, vxlan, geneve, wireguard etc.) and in future will configure
netkit in order to expose the needed_{head,tail}room into K8s pods. See b9ed315d3c4c ("netkit: Allow for configuring needed_{head,tail}room").
Suggested-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: Alasdair McWilliam <alasdair@mcwilliam.dev> Reviewed-by: Daniel Borkmann <daniel@iogearbox.net> Link: https://patch.msgid.link/20250917095543.14039-1-alasdair@mcwilliam.dev Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The dwmac core supports a range of interfaces, but when it comes to
SerDes interfaces, the core itself does not include the SerDes block.
Consequently, it has to provide an interface suitable to interface such
a block to, and uses TBI for this.
The driver also uses "PCS" for RGMII, even though the dwmac PCS block
is not present for this interface type - it was a convenice for the
code structure as RGMII includes inband signalling of the PHY state,
much like Cisco SGMII does at a high level.
As such, the code refers to RGMII and SGMII modes for the PCS, and
there used to be STMMAC_PCS_TBI and STMMAC_PCS_RTBI constants as well
but these were never set, although they were used in the code.
The selection of the PCS mode was from mac_interface. Thus, it seems
that the original intention was for mac_interface to describe the
interface mode used within the dwmac core, and phy_interface to
describe the external world-accessible interface (e.g. which would
connect to a PHY or SFP cage.)
It appears that many glue drivers misinterpret these. A good exmple
is socfpga. This supports SGMII and 1000BASE-X, but does not include
the dwmac PCS, relying on the Lynx PCS instead. However, it makes use
of mac_interface to configure the dwmac core to its GMII/MII mode.
So, when operating in either of these modes, the dwmac is configured
for GMII mode to communicate with the Lynx PCS which then provides
the SGMII or 1000BASE-X interface mode to the external world.
Given that phy_interface is the external world interface, and
mac_interface is the dwmac core interface, selecting the interface
mode based on mac_interface being 1000BASEX makes no sense.
Moreover, mac_interface is initialised by the stmmac core code. If
the "mac-mode" property is set in DT, this will be used. Otherwise,
it will reflect the "phy-mode" property - meaning that it defaults
to phy_interface. As no in-kernel DT makes reference to a "mac-mode"
property, we can assume that for all in-kernel platforms, these two
interface variables are the same. The exception are those platform
glues which I reviwed and suggested they use phy_interface, setting
mac_interface to PHY_INTERFACE_MODE_NA.
The conclusion to all of this is that mac_interface serves no useful
purpose, and causes confusion as the platform glue code doesn't seem
to know which one should be used.
Thus, let's get rid of mac_interface, which makes all this code more
understandable. It also helps to untangle some of the questions such
as:
- should this be using the interface passed from phylink
- should we set the range of phylink supported interfaces to be
more than one
- when we get phylink PCS support for the dwmac PCS, should we be
selecting it based on mac_interface or phy_interface, and how
should we populate the PCS' supported_interface bitmap.
Having converted socfpga to use phy_interface, this turns out to
feel like "the right way" to do this - convert the external world
"phy_interface" to whatever phy_intf_sel value that the dwmac core
needs to achieve the connection to whatever hardware blocks are
integrated inside the SoC to achieve the requested external world
interface.
As an illustration why - consider that in the case of socfpga, it
_could_ have been implemented such that the dwmac PCS was used for
SGMII, and the Lynx PCS for 1000BASE-X, or vice versa. Only the
platform glue would know which it is.
I will also note that several cores provide their currently configured
interface mode via the ACTPHYIF field of Hardware Feature 0, and thus
can be read back in the platform-independent parts of the driver to
decide whether the internal PCS or the RGMII (or actually SMII) "PCS"
should be used.
This is hot-off-the-press, and has only been build tested. As I have
none of these platforms, this series has not been run-tested, so
please test on your hardware.
====================
mac_interface has served little purpose, and has only caused confusion.
Now that we have cleaned up all platform glue drivers which should not
have been using mac_interface, there are no users remaining. Remove
mac_interface.
This results in the special dwmac specific "mac-mode" DT property
becoming redundant, and an in case, no DTS files in the kernel make use
of this property. Add a warning if the property is set, and it is
different from the "phy-mode".
dwmac-sun8i supports MII, RMII and RGMII interface modes only. It
is unclear whether the dwmac core interface is different from the
one presented to the outside world.
However, as none of the DTS files set "mac-mode", mac_interface will
be identical to phy_interface.
Convert dwmac-sun8i to use phy_interface when determining the
interface mode rather than mac_interface.
dwmac-stm32 supports MII, RMII, GMII and RGMII interface modes,
selecting the dwmac core interface mode via bits 23:21 of the
SYSCFG register. The bit combinations are identical to the
dwmac phy_intf_sel_i signals.
None of the DTS files set "mac-mode", so mac_interface will be
identical to phy_interface.
Convert dwmac-stm32 to use phy_interface when determining the
interface mode rather than mac_interface.
net: stmmac: starfive: convert to use phy_interface
dwmac-starfive uses RMII or RGMII interface modes without any PCS,
and selects the dwmac core accordingly using a register field with
the same bit encoding as the core's phy_intf_sel_i signals.
None of the DTS files set "mac-mode", so mac_interface will be
identical to phy_interface.
Convert dwmac-starfive to use phy_interface when determining the
interface mode rather than mac_interface. Also convert the error
prints to use phy_modes() so that we get a meaningful string rather
than a number for the interface mode.
net: stmmac: socfpga: convert to use phy_interface
dwmac-socfpga uses MII, RMII, GMII, RGMII, SGMII and 1000BASE-X
interface modes, and supports the Lynx PCS. The Lynx PCS will only be
used for SGMII and 1000BASE-X modes, with the MAC programmed to use
GMII or MII mode to talk to the PCS. This suggests that the Synopsys
optional dwmac PCS is not present.
None of the DTS files set "mac-mode", so mac_interface will be
identical to phy_interface.
Convert dwmac-socfpga to use phy_interface when determining the
interface mode rather than mac_interface.
net: stmmac: ingenic: convert to use phy_interface
dwmac-ingenic uses only MII, RMII, GMII or RGMII interface modes, none
of which require any kind of conversion between the MAC and external
world. Thus, mac_interface and phy_interface will be the same.
Convert dwmac-ingenic to use phy_interface when determining the
interface mode that the dwmac core should be configured to at reset,
rather than mac_interface.
Also convert the error prints to use phy_modes() and terminate with a
newline so that we get a human readable string rather than a number for
the interface mode.
Checking the IMX8MP documentation, there is no requirement for a
separate mac_interface mode definition. As mac_interface and
phy_interface will be the same, use phy_interface internally rather
than mac_interface.
Also convert the error prints to use phy_modes() so that we get a
meaningful string rather than a number for the interface mode.
net: stmmac: use phy_interface in stmmac_check_pcs_mode()
In the majority, if not all cases, mac_interface and phy_interface
are the same with the exception of some drivers that I have suggested
only use phy_interface and set mac_interface to PHY_INTERFACE_MODE_NA.
The only two that currently set mac_interface to PHY_INTERFACE_MODE_NA
are dwmac-loongson and dwmac-lpc18xx, neither of which use RGMII nor
SGMII.
In order to phase out the use of mac_interface, we need to have a path
for existing drivers so they can update to only using phy_interface
without causing regressions.
Therefore, in order to keep the "pcs" code working, we need to choose
the STMMAC integrated PCS mode based on phy_interface if mac_interface
is PHY_INTERFACE_MODE_NA.
This will allow more drivers to set mac_interface to
PHY_INTERFACE_MODE_NA without risking regressions.
net: stmmac: rework mac_interface and phy_interface documentation
Based on new research, it has come to light that the comment that I
added in a014c35556b9 ("net: stmmac: clarify difference between
"interface" and "phy_interface"") is not fully correct.
Update the comment to properly describe the difference between the two.
All of the DTS files in the kernel tree do not mention the "mac-mode"
property, which results in mac_interface and phy_interface being the
same. Also, none of the platform glue drivers set mac_interface to
anything but PHY_INTERFACE_MODE_NA. This means that for all the
platforms known to mainline, mac_interface is either the same as
phy_interface, or it is PHY_INTERFACE_MODE_NA.
Thus, updating the definition for mac_interface in stmmac.h has no
material effect on current uses known to mainline, but the change opens
the door to cleaning up all uses.
The mlx5_devlink_total_vfs_set function branches based on per_pf_support
twice. Remove the second branch as the first one exits the function when
per_pf_support is false.
Accidentally added as part of commit a4c49611cf4f ("net/mlx5: Implement
devlink total_vfs parameter").
Reported-by: Dan Carpenter <dan.carpenter@linaro.org> Closes: https://lore.kernel.org/linux-rdma/aMQWenzpdjhAX4fm@stanley.mountain/ Signed-off-by: Vlad Dumitrescu <vdumitrescu@nvidia.com> Link: https://patch.msgid.link/a6142a60-1948-439a-b0ae-ff1df26a37f8@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Zahka [Thu, 18 Sep 2025 21:27:20 +0000 (14:27 -0700)]
psp: clarify checksum behavior of psp_dev_rcv()
psp_dev_rcv() decapsulates psp headers from a received frame. This
will make any csum complete computed by the device inaccurate. Rather
than attempt to patch up skb->csum in psp_dev_rcv() just make it clear
to callers what they can expect regarding checksum complete.
Jakub Kicinski [Thu, 18 Sep 2025 18:31:19 +0000 (11:31 -0700)]
net: phy: micrel: use %pe in print format
New cocci check complains:
drivers/net/phy/micrel.c:4308:6-13: WARNING: Consider using %pe to print PTR_ERR()
drivers/net/phy/micrel.c:5742:6-13: WARNING: Consider using %pe to print PTR_ERR()
====================
address miscellaneous issues with psp_sk_get_assoc_rcu()
There were a few minor issues with psp_sk_get_assoc_rcu() identified
by Eric in his review of the initial psp series. This series addresses
them.
====================
Daniel Zahka [Thu, 18 Sep 2025 15:52:04 +0000 (08:52 -0700)]
psp: don't use flags for checking sk_state
Using flags to check sk_state only makes sense to check for a subset
of states in parallel e.g. sk_fullsock(). We are not doing that
here. Compare for individual states directly.
Daniel Zahka [Thu, 18 Sep 2025 15:52:03 +0000 (08:52 -0700)]
psp: fix preemptive inet_twsk() cast in psp_sk_get_assoc_rcu()
It is weird to cast to a timewait_sock before checking sk_state, even
if the use is after such a check. Remove the tw local variable, and
use inet_twsk() directly in the timewait branch.
ptp_ocp: make ptp_ocp driver compatible with PTP_EXTTS_REQUEST2
Originally ptp_ocp driver was not strictly checking flags for external
timestamper and was always activating rising edge timestamping as it's
the only supported mode. Recent changes to ptp made it incompatible with
PTP_EXTTS_REQUEST2 ioctl. Adjust ptp_clock_info to provide supported
mode and be compatible with new infra.
While at here remove explicit check of periodic output flags from the
driver and provide supported flags for ptp core to check.
Dan Carpenter [Thu, 18 Sep 2025 09:48:26 +0000 (12:48 +0300)]
net: ti: icssm-prueth: unwind cleanly in probe()
This error handling triggers a Smatch warning:
drivers/net/ethernet/ti/icssm/icssm_prueth.c:1574 icssm_prueth_probe()
warn: 'prueth->pru1' is an error pointer or valid
The warning is harmless because the pru_rproc_put() function has an
IS_ERR_OR_NULL() check built in. However, there is a small bug if
syscon_regmap_lookup_by_phandle() fails. In that case we should call
of_node_put() on eth0_node and eth1_node.
It's a little bit easier to re-write this code to only free things which
we know have been allocated successfully.
Fixes: 511f6c1ae093 ("net: ti: icssm-prueth: Adds ICSSM Ethernet driver") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Reviewed-by: Simon Horman <horms@kernel.org> Reviewed-by: Parvathi Pudi <parvathi@couthit.com> Link: https://patch.msgid.link/aMvVagz8aBRxMvFn@stanley.mountain Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
net/mlx5e: Support RSS for IPSec offload
The series by Jianbo uses a new firmware feature to identify the inner
protocol of decrypted packets, adding new flow groups and steering rules
to redirect them for proper L4-based RSS. This ensures traffic is spread
across multiple CPU cores.
====================
Jianbo Liu [Thu, 18 Sep 2025 07:19:23 +0000 (10:19 +0300)]
net/mlx5e: Add flow rules for the decrypted ESP packets
The previous commit introduced two new flow groups to enable L4 RSS
for decrypted IPsec traffic. This commit implements the logic to
populate these groups with the necessary steering rules.
The rules are created dynamically whenever the first IPSec offload
rule is configured via the xfrm subsystem and the decryption tables
for RX are created. Each rule matches a specific decrypted traffic
type based on its ip version (or ethertype) and outer/inner
l4_type_ext, directing it to the appropriate L4 RSS-enabled TIR.
The lifecycle of these steering rules is tied directly to the RX
tables. They are deleted when the RX tables are destroyed.
Jianbo Liu [Thu, 18 Sep 2025 07:19:22 +0000 (10:19 +0300)]
net/mlx5e: Add flow groups for the packets decrypted by crypto offload
When using IPsec crypto offload, the hardware decrypts the packet
payload but preserves the ESP header. This prevents the standard RSS
mechanism from accessing the inner L4 (TCP/UDP) headers. As a result,
the RSS hash is calculated only on the outer L3 IP headers, causing
all traffic for a given IPsec tunnel to be directed to a single queue,
leading to poor traffic distribution.
Newer firmware introduces the ability to match on l4_type_ext, which
exposes the L4 protocol type following an ESP header. This allows the
driver to create steering rules that can identify the inner protocols
of decrypted packets.
This commit leverages this new capability to improve traffic
distribution. It adds two new flow groups to steer decrypted packets
to dedicated TIRs that was configured to perform RSS on the inner L4
headers.
These groups are inserted after the standard L4 group and before the
group that handles undecrypted ESP packets added in this series. The
first new group matches decrypted packets based on the outer IP
version (or ethertype) and l4_type_ext. The second new group matches
decrypted tunneled packets based on the inner IP version and
l4_type_ext. Eight new traffic types are also defined to support this
functionality.
Jianbo Liu [Thu, 18 Sep 2025 07:19:21 +0000 (10:19 +0300)]
net/mlx5e: Recirculate decrypted packets into TTC table
In the commit 5e466345291a ("net/mlx5e: IPsec: Add IPsec steering in
local NIC RX"), the decrypted packets are handled in RX error flow
table. There is only one rule in the table, which forwards packets to
the default ESP TIR.
This patch updates the design to allow RSS after decryption. For ESP
traffic, SPI and IP addresses are the fields selected for RSS hash,
and it's common that only one SPI is configured in RX direction, so
RSS can't work properly as all the packets are hashed to one key in
this case. To take advantage of RSS and improve performance, the
decrypted packets need to be forwarded back to TTC table, where RSS
can work based on the decrypted packet types.
Jianbo Liu [Thu, 18 Sep 2025 07:19:20 +0000 (10:19 +0300)]
net/mlx5: Change TTC rules to match on undecrypted ESP packets
The TTC (Traffic Type Classifier) table classifies the traffic and
steers packet to TIRs, where RSS works based on the hash calculated
from the selected packet fields. For AH/ESP packets, SPI and IP
addresses are the fields used to calculate the hash value for RSS. So,
it's hard to distribute packets to different receiving queues as there
is usually only one SPI in that direction.
IPSec hardware offloads, crypto offload and full (packet) offload were
introduced later. For crypto offload, hardware does encryption,
decryption and authentication, kernel does the others. Kernel always
sends/receives formatted ESP packets with plaintext data instead of
the ciphertext data, all other fields are unmodified. For full
offload, hardware will take care of almost everything, kernel just
sends/receives packets without any IPSec headers.
Currently, all packets with ESP protocols are forwarded to IPSec
offload tables if IPSec rules are configured. In a downstream patch,
the decrypted packets will be recirculated to TTC table, in order to
use RSS, which does the hash on L4 fields after IPSec headers are
stripped by full offload. So those packets handled by crypto offload
must filtered out, as they still have the ESP headers, but apparently
no need to be decrypted again. To do that, ipsec_next_header is added
for the packet matching, as it is valid only after passing through
IPSec decryption.
projects / thirdparty / kernel / stable.git / log
