Adapt linkstate_get_sqi() and linkstate_get_sqi_max() to take a
phy_device argument directly, enabling support for setups with
multiple PHYs. The previous assumption of a single PHY attached to
a net_device no longer holds.
Use ethnl_req_get_phydev() to identify the appropriate PHY device
for the operation. Update linkstate_prepare_data() and related
logic to accommodate this change, ensuring compatibility with
multi-PHY configurations.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
dev_deactivate_many() role is to remove the qdiscs
of a network device.
When/if a qdisc is dismantled, an rcu grace period
is needed to make sure all outstanding qdisc enqueue
are done before we proceed with a qdisc reset.
Most virtual devices do not have a qdisc.
We can call the expensive synchronize_net() only
if needed.
Note that dev_deactivate_many() does not have to deal
with qdisc-less dev_queue_xmit, as an old comment
was claiming.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250109171850.2871194-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tariq Toukan says:
====================
mlx5 HW-Managed Flow Steering in FS core level
This patchset by Moshe follows Yevgeny's patchsets [1][2] on subject
"HW-Managed Flow Steering in mlx5 driver". As introduced there in HW
managed Flow Steering mode (HWS) the driver is configuring steering
rules directly to the HW using WQs with a special new type of WQE (Work
Queue Element). This way we can reach higher rule insertion/deletion
rate with much lower CPU utilization compared to SW Managed Flow
Steering (SWS).
This patchset adds API to manage namespace, flow tables, flow groups and
prepare FTE (Flow Table Entry) rules. It also adds caching and pool
mechanisms for HWS actions to allow sharing of steering actions among
different rules. The implementation of this API in FS layer, allows FS
core to use HW Managed Flow Steering in addition to the existing FW or
SW Managed Flow Steering.
Patch 13 of this series adds support for configuring HW Managed Flow
Steering mode through devlink param, similar to configuring SW Managed
Flow Steering mode:
# devlink dev param set pci/0000:08:00.0 name flow_steering_mode \
cmode runtime value hmfs
In addition, the series contains 2 HWS patches from Yevgeny that
implement flow update support.
[1] https://lore.kernel.org/netdev/20240903031948.78006-1-saeed@kernel.org/
[2] https://lore.kernel.org/all/20250102181415.1477316-1-tariqt@nvidia.com/
====================
Link: https://patch.msgid.link/20250109160546.1733647-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch is the second part of update flow implementation.
Instead of using two action RTCs, we use the same RTC which is twice
the size of what was required before the update flow support.
This way we always allocate STEs from the same RTC (same pool),
which means that update is done similar to how create is done.
The bigger size allows us to allocate and write new STEs, and
later free the old (pre-update) STEs.
Similar to rule creation, STEs are written in reverse order:
- write action STEs, while match STE is still pointing to
the old action STEs
- overwrite the match STE with the new one, which now
is pointing to the new action STEs
Old action STEs can be freed only once we got completion on the
writing of the new match STE. To implement this we added new rule
states: UPDATING/UPDATED. Rule is moved to UPDATING state in the
beginning of the update flow. Once all completions are received,
rule is moved to UPDATED state. At this point old action STEs are
freed and rule goes back to CREATED state.
Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Vlad Dogaru <vdogaru@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250109160546.1733647-16-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch is the first part of update flow implementation.
Update flow should support rules with single STE (match STE only),
as well as rules with multiple STEs (match STE plus action STEs).
Supporting the rules with single STE is straightforward: we just
overwrite the STE, which is an atomic operation.
Supporting the rules with action STEs is a more complicated case.
The existing implementation uses two action RTCs per matcher and
alternates between the two for each update request.
This implementation was unnecessarily complex and lead to some
unhandled edge cases, so the support for rule update with multiple
STEs wasn't really functional.
This patch removes this code, and the next patch adds implementation
of a different approach.
Note that after applying this patch and before applying the next
patch we still have support for update rule with single STE (only
match STE w/o action STEs), but update will fail for rules with
action STEs.
Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Vlad Dogaru <vdogaru@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250109160546.1733647-15-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Multiple flow counters can utilize a single Hardware Steering (HWS)
action for Hardware Steering rules. Given that these counter bulks are
not exclusively created for Hardware Steering, but also serve purposes
such as statistics gathering and other steering modes, it's more
efficient to create the HWS action only when it's first needed by a
Hardware Steering rule. This approach allows for better resource
management through the use of a reference count, rather than
automatically creating an HWS action for every bulk of flow counters.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250109160546.1733647-8-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add packet reformat alloc and dealloc API functions to provide packet
reformat actions for steering rules.
Add HWS action pools for each of the following packet reformat types:
- decapl3: decapsulate l3 tunnel to l2
- encapl2: encapsulate l2 to tunnel l2
- encapl3: encapsulate l2 to tunnel l3
- insert_hdr: insert header
In addition cache remove header action for remove vlan header as this is
currently the only use case of remove header action in the driver.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250109160546.1733647-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The HW Steering actions pool will help utilize the option in HW Steering
to share steering actions among different rules.
Create pool on root namespace creation and add few HW Steering actions
that don't depend on the steering rule itself and thus can be shared
between rules, created on same namespace: tag, pop_vlan, push_vlan,
drop, decap l2.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250109160546.1733647-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
iavf uses the netdev->lock already to protect shapers.
In an upcoming series we'll try to protect NAPI instances
with netdev->lock.
We need to modify the protection a bit. All NAPI related
calls in the driver need to be consistently under the lock.
This will allow us to easily switch to a "we already hold
the lock" NAPI API later.
register_netdevice(), OTOH, must not be called under
the netdev_lock() as we do not intend to have an
"already locked" version of this call.
Link: https://patch.msgid.link/20250111071339.3709071-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
init_dummy_netdev_core() used to cater to net_devices which
did not come from alloc_netdev_mqs(). Since that's no longer
supported remove the init logic which duplicates alloc_netdev_mqs().
While at it rename back to init_dummy_netdev().
Reviewed-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250113003456.3904110-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
init_dummy_netdev() can initialize statically declared or embedded
net_devices. Such netdevs did not come from alloc_netdev_mqs().
After recent work by Breno, there are the only two cases where
we have do that.
Switch those cases to alloc_netdev_mqs() and delete init_dummy_netdev().
Dealing with static netdevs is not worth the maintenance burden.
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Link: https://patch.msgid.link/20250113003456.3904110-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a --family option to ynl to specify the spec by family name instead
of file path, with support for searching in-tree and system install
location and a --list-families option to show the available families.
./tools/net/ynl/pyynl/cli.py --family rt_addr --dump getaddr
Signed-off-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20250111154803.7496-1-donald.hunter@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When recvmsg with MSG_PEEK flag, the data will be copied to
user's buffer without advancing consume cursor and without
reducing the length of rx available data. Once the expected
peek length is larger than the value of bytes_to_rcv, in the
loop of do while in smc_rx_recvmsg, the first loop will copy
bytes_to_rcv bytes of data from the position local_tx_ctrl.cons,
the second loop will copy the min(bytes_to_rcv, read_remaining)
bytes from the position local_tx_ctrl.cons again because of the
lacking of process with advancing consume cursor and reducing
the length of available data. So do the subsequent loops. The
data copied in the second loop and the subsequent loops will
result in data error, as it should not be copied if no more data
arrives and it should be copied from the position advancing
bytes_to_rcv bytes from the local_tx_ctrl.cons if more data arrives.
This issue can be reproduce by the following python script:
server.py:
import socket
import time
server_ip = '0.0.0.0'
server_port = 12346
server_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
server_socket.bind((server_ip, server_port))
server_socket.listen(1)
print('Server is running and listening for connections...')
conn, addr = server_socket.accept()
print('Connected by', addr)
while True:
data = conn.recv(1024)
if not data:
break
print('Received request:', data.decode())
conn.sendall(b'Hello, client!\n')
time.sleep(5)
conn.sendall(b'Hello, again!\n')
conn.close()
client.py:
import socket
server_ip = '<server ip>'
server_port = 12346
resp=b'Hello, client!\nHello, again!\n'
client_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
client_socket.connect((server_ip, server_port))
request = 'Hello, server!'
client_socket.sendall(request.encode())
peek_data = client_socket.recv(len(resp),
socket.MSG_PEEK | socket.MSG_WAITALL)
print('Peeked data:', peek_data.decode())
client_socket.close()
Fixes: 952310ccf2 ("smc: receive data from RMBE")
Reported-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Link: https://patch.msgid.link/20250104143201.35529-1-guangguan.wang@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The DP83822 supports up to three configurable Light Emitting Diode (LED)
pins: LED_0, LED_1 (GPIO1), COL (GPIO2) and RX_D3 (GPIO3). Several
functions can be multiplexed onto the LEDs for different modes of
operation. LED_0 and COL (GPIO2) use the MLED function. MLED can be routed
to only one of these two pins at a time. Add minimal LED controller driver
supporting the most common uses with the 'netdev' trigger.
Signed-off-by: Dimitri Fedrau <dima.fedrau@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250107-dp83822-leds-v2-1-5b260aad874f@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Marc Kleine-Budde says:
====================
pull-request: can-next 2025-01-10
Pierre-Henry Moussay adds PIC64GX compatibility to the DT bindings for
Microchip's mpfs-can IP core.
The next 3 patches are by Sean Nyekjaer and target the tcan4x5x
driver. First the DT bindings is converted to DT schema, then nWKRQ
voltage selection is added to the driver.
Dario Binacchi's patch for the sun4i_can makes the driver more
consistent by adding a likely() to the driver.
Another patch by Sean Nyekjaer for the tcan4x5x driver gets rid of a
false error message.
Charan Pedumuru converts the atmel-can DT bindings to DT schema.
The next 2 patches are by Oliver Hartkopp. The first one maps Oliver's
former mail addresses to a dedicated CAN mail address. The second one
assigns net/sched/em_canid.c additionally to the CAN maintainers.
Ariel Otilibili's patch removes dead code from the CAN dev helper.
The next 3 patches are by Sean Nyekjaer and add HW standby support to
the tcan4x5x driver.
A patch by Dario Binacchi fixes the DT bindings for the st,stm32-bxcan
driver.
The last 4 patches are by Jimmy Assarsson and target the kvaser_usb
and the kvaser_pciefd driver: error statistics are improved and
CAN_CTRLMODE_BERR_REPORTING is added.
* tag 'linux-can-next-for-6.14-20250110' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next:
can: kvaser_pciefd: Add support for CAN_CTRLMODE_BERR_REPORTING
can: kvaser_pciefd: Update stats and state even if alloc_can_err_skb() fails
can: kvaser_usb: Add support for CAN_CTRLMODE_BERR_REPORTING
can: kvaser_usb: Update stats and state even if alloc_can_err_skb() fails
dt-bindings: can: st,stm32-bxcan: fix st,gcan property type
can: m_can: call deinit/init callback when going into suspend/resume
can: tcan4x5x: add deinit callback to set standby mode
can: m_can: add deinit callback
can: dev: can_get_state_str(): Remove dead code
MAINTAINERS: assign em_canid.c additionally to CAN maintainers
mailmap: add an entry for Oliver Hartkopp
dt-bindings: net: can: atmel: Convert to json schema
can: tcan4x5x: get rid of false clock errors
can: sun4i_can: continue to use likely() to check skb
can: tcan4x5x: add option for selecting nWKRQ voltage
dt-bindings: can: tcan4x5x: Document the ti,nwkrq-voltage-vio option
dt-bindings: can: convert tcan4x5x.txt to DT schema
dt-bindings: can: mpfs: add PIC64GX CAN compatibility
====================
Link: https://patch.msgid.link/20250110112712.3214173-1-mkl@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Russell King says:
====================
net: stmmac: clean up and fix EEE implementation
This is a rework of stmmac's EEE support in light of the addition of EEE
management to phylib. It's slightly more than 15 patches, but I think it
makes sense to be so.
Patch 1 adds configuration of the receive clock phy_eee_rx_clock_stop()
(which was part of another series, but is necessary for this patch set.)
Patch 2 converts stmmac to use phylib's tracking of tx_lpi_timer.
Patch 3 corrects the data type used for things involving the LPI
timer. The user API uses u32, so stmmac should do too, rather than
blindly converting it to "int". eee_timer is left for patch 4.
Patch 4 (new) uses an unsigned int for eee_timer.
Patch 5 makes stmmac EEE state depend on phylib's enable_tx_lpi flag,
thus using phylib's resolution of EEE state.
Patch 6 removes redundant code from the ethtool EEE operations.
Patch 7 removes some redundant code in stmmac_disable_eee_mode()
and renames it to stmmac_disable_sw_eee_mode() to better reflect its
purpose.
Patch 8 removes the driver private tx_lpi_enabled, which is managed by
phylib since patch 4.
Patch 9 removes the dependence of EEE error statistics on the EEE
enable state, instead depending on whether EEE is supported by the
hardware.
Patch 10 removes phy_init_eee(), instead using phy_eee_rx_clock_stop()
to configure whether the PHY may stop the receive clock.
Patch 11 removes priv->eee_tw_timer, which is only ever set to one
value at probe time, effectively it is a constant. Hence this is
unnecessary complexity.
Patch 12 moves priv->eee_enabled into stmmac_eee_init(), and placing
it under the protection of priv->lock, except when EEE is not
supported (where it becomes constant-false.)
Patch 13 moves priv->eee_active also into stmmac_eee_init(), so
the indication whether EEE should be enabled or not is passed in
to this function.
Since both priv->eee_enabled and priv->eee_active are assigned
true/false values, they should be typed "bool". Make it sew in
patch 14. No Singer machine required.
Patch 15 moves the initialisation of priv->eee_ctrl_timer to the
probe function - it makes no sense to re-initialise the timer each
time we want to start using it.
Patch 16 removes the unnecessary EEE handling in the driver tear-down
method. The core net code will have brought the interface down
already, meaning EEE has already been disabled.
Patch 17 reorganises the code to split the hardware LPI timer
control paths from the software LPI timer paths.
Patch 18 works on this further by eliminating
stmmac_lpi_entry_timer_config() and making direct calls to the new
functions. This reveals a potential bug where priv->eee_sw_timer_en
is set true when EEE is disabled. This is not addressed in this
series, but will be in a future separate patch - so that if fixing
that causes a regression, it can be handled separately.
====================
Link: https://patch.msgid.link/Z36sHIlnExQBuFJE@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
phylink_stop() will cause phylink to call the mac_link_down() operation
before phylink_stop() returns. As mac_link_down() will call
stmmac_eee_init(false), this will set both priv->eee_active and
priv->eee_enabled to be false, deleting the eee_ctrl_timer if
priv->eee_enabled was previously set.
As stmmac_release() calls phylink_stop() before checking whether
priv->eee_enabled is true, this is a condition that can never be
satisfied, and thus the code within this if() block will never be
executed. Remove it.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Choong Yong Liang <yong.liang.choong@linux.intel.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tVZEg-0002LE-HH@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All call sites for stmmac_eee_init() assign the return code to
priv->eee_enabled. Rather than having this coded at each call site,
move the assignment inside stmmac_eee_init().
Since stmmac_init_eee() takes priv->lock before checking the state of
priv->eee_enabled, move the assignment within the locked region. Also,
stmmac_suspend() checks the state of this member under the lock. While
two concurrent calls to stmmac_init_eee() aren't possible, there is
a possibility that stmmac_suspend() may run concurrently with a change
of priv->eee_enabled unless we modify it under the lock.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Choong Yong Liang <yong.liang.choong@linux.intel.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tVZEM-0002Kq-2Z@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
