When the link is up, either eth_proto_oper or ext_eth_proto_oper
typically reports the active link protocol, from which both speed
and number of lanes can be retrieved. However, in certain cases,
such as when a NIC is connected via a non-standard cable, the
firmware may not report the protocol.
In such scenarios, the speed can still be obtained from the
data_rate_oper field in PTYS register. Since data_rate_oper
provides only speed information and lacks lane details, it is
incorrect to derive the number of lanes from it.
This patch corrects the behavior by setting the number of lanes to
UNKNOWN instead of incorrectly using MAX_LANES when relying on
data_rate_oper.
Fixes: 7e959797f0 ("net/mlx5e: Enable lanes configuration when auto-negotiation is off")
Signed-off-by: Shahar Shitrit <shshitrit@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250610151514.1094735-10-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Previously, a unique tunnel id was added for the matching on TC
non-zero chains, to support inner header rewrite with goto action.
Later, it was used to support VF tunnel offload for vxlan, then for
Geneve and GRE. To support VF tunnel, a temporary mlx5_flow_spec is
used to parse tunnel options. For Geneve, if there is TLV option, a
object is created, or refcnt is added if already exists. But the
temporary mlx5_flow_spec is directly freed after parsing, which causes
the leak because no information regarding the object is saved in
flow's mlx5_flow_spec, which is used to free the object when deleting
the flow.
To fix the leak, call mlx5_geneve_tlv_option_del() before free the
temporary spec if it has TLV object.
Fixes: 521933cdc4 ("net/mlx5e: Support Geneve and GRE with VF tunnel offload")
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Alex Lazar <alazar@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250610151514.1094735-9-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When there are more than one destinations, we create a FW flow
table and provide it with all the destinations. FW requires to
have wire as the last destination in the list (if it exists),
otherwise the operation fails with FW syndrome.
This patch fixes the destination array action creation: if it
contains a wire destination, it is moved to the end.
Fixes: 504e536d90 ("net/mlx5: HWS, added actions handling")
Signed-off-by: Vlad Dogaru <vdogaru@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250610151514.1094735-7-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The newly introduced mutex is only used for reformat actions, but it was
initialized for modify header instead.
The struct that contains the mutex is zero-initialized and an all-zero
mutex is valid, so the issue only shows up with CONFIG_DEBUG_MUTEXES.
Fixes: b206d9ec19 ("net/mlx5: HWS, register reformat actions with fw")
Signed-off-by: Vlad Dogaru <vdogaru@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250610151514.1094735-5-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When attempting to add a rule to an existing flow group, if a matching
flow group exists but is not active, the error code returned should be
EAGAIN, so that the rule can be added to the matching flow group once
it is active, rather than ENOENT, which indicates that no matching
flow group was found.
Fixes: bd71b08ec2 ("net/mlx5: Support multiple updates of steering rules in parallel")
Signed-off-by: Gavi Teitz <gavi@nvidia.com>
Signed-off-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Patrisious Haddad <phaddad@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250610151514.1094735-4-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When firmware asks the driver to allocate more pages, using event of
give_pages, the driver should always allocate it from same NUMA, the
original device NUMA. Current code uses dev_to_node() which can result
in different NUMA as it is changed by other driver flows, such as
mlx5_dma_zalloc_coherent_node(). Instead, use saved numa node for
allocating firmware pages.
Fixes: 311c7c71c9 ("net/mlx5e: Allocate DMA coherent memory on reader NUMA node")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Link: https://patch.msgid.link/20250610151514.1094735-2-mbloch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2025-06-10 (i40e, iavf, ice, e1000)
For i40e:
Robert Malz improves reset handling for situations where multiple reset
requests could cause some to be missed.
For iavf:
Ahmed adds detection, and handling, of reset that could occur early in
the initialization process to stop long wait/hangs.
For ice:
Anton, properly, sets missed use_nsecs value.
For e1000:
Joe Damato moves cancel_work_sync() call to avoid deadlock.
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
e1000: Move cancel_work_sync to avoid deadlock
ice/ptp: fix crosstimestamp reporting
iavf: fix reset_task for early reset event
i40e: retry VFLR handling if there is ongoing VF reset
i40e: return false from i40e_reset_vf if reset is in progress
====================
Link: https://patch.msgid.link/20250610171348.1476574-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When using publicly available tools like 'mdio-tools' to read/write data
from/to network interface and its PHY via C45 (clause 45) mdiobus,
there is no verification of parameters passed to the ioctl and
it accepts any mdio address.
Currently there is support for 32 addresses in kernel via PHY_MAX_ADDR define,
but it is possible to pass higher value than that via ioctl.
While read/write operation should generally fail in this case,
mdiobus provides stats array, where wrong address may allow out-of-bounds
read/write.
Fix that by adding address verification before C45 read/write operation.
While this excludes this access from any statistics, it improves security of
read/write operation.
Fixes: 4e4aafcddb ("net: mdio: Add dedicated C45 API to MDIO bus drivers")
Signed-off-by: Jakub Raczynski <j.raczynski@samsung.com>
Reported-by: Wenjing Shan <wenjing.shan@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When using publicly available tools like 'mdio-tools' to read/write data
from/to network interface and its PHY via mdiobus, there is no verification of
parameters passed to the ioctl and it accepts any mdio address.
Currently there is support for 32 addresses in kernel via PHY_MAX_ADDR define,
but it is possible to pass higher value than that via ioctl.
While read/write operation should generally fail in this case,
mdiobus provides stats array, where wrong address may allow out-of-bounds
read/write.
Fix that by adding address verification before read/write operation.
While this excludes this access from any statistics, it improves security of
read/write operation.
Fixes: 080bb352fa ("net: phy: Maintain MDIO device and bus statistics")
Signed-off-by: Jakub Raczynski <j.raczynski@samsung.com>
Reported-by: Wenjing Shan <wenjing.shan@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jeff Johnson says:
==================
ath.git updates for v6.16-rc2
Fix a handful of both build and stability issues across multiple drivers.
==================
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
According to 802.1AE standard, when ES and SC flags in TCI are zero,
used SCI should be the current active SC_RX. Current code uses the
header MAC address. Without this patch, when ES flag is 0 (using a
bridge or switch), header MAC will not fit the SCI and MACSec frames
will be discarted.
In order to test this issue, MACsec link should be stablished between
two interfaces, setting SC and ES flags to zero and a port identifier
different than one. For example, using ip macsec tools:
ip link add link $ETH0 macsec0 type macsec port 11 send_sci off
end_station off
ip macsec add macsec0 tx sa 0 pn 2 on key 01 $ETH1_KEY
ip macsec add macsec0 rx port 11 address $ETH1_MAC
ip macsec add macsec0 rx port 11 address $ETH1_MAC sa 0 pn 2 on key 02
ip link set dev macsec0 up
ip link add link $ETH1 macsec1 type macsec port 11 send_sci off
end_station off
ip macsec add macsec1 tx sa 0 pn 2 on key 01 $ETH0_KEY
ip macsec add macsec1 rx port 11 address $ETH0_MAC
ip macsec add macsec1 rx port 11 address $ETH0_MAC sa 0 pn 2 on key 02
ip link set dev macsec1 up
Fixes: c09440f7dc ("macsec: introduce IEEE 802.1AE driver")
Co-developed-by: Andreu Montiel <Andreu.Montiel@technica-engineering.de>
Signed-off-by: Andreu Montiel <Andreu.Montiel@technica-engineering.de>
Signed-off-by: Carlos Fernandez <carlos.fernandez@technica-engineering.de>
Reviewed-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before appending sysdata, prepare_extradata() checks if any feature is
enabled in sysdata_fields (and exits early if none is enabled).
When SYSDATA_RELEASE was introduced, we missed adding it to the list of
features being checked against sysdata_fields in prepare_extradata().
The result was that, if only SYSDATA_RELEASE is enabled in
sysdata_fields, we incorreclty exit early and fail to append the
release.
Instead of checking specific bits in sysdata_fields, check if
sysdata_fields has ALL bit zeroed and exit early if true. This fixes
case when only SYSDATA_RELEASE enabled and makes the code more general /
less error prone in future feature implementation.
Signed-off-by: Gustavo Luiz Duarte <gustavold@gmail.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Fixes: cfcc9239e7 ("netconsole: append release to sysdata")
Link: https://patch.msgid.link/20250609-netconsole-fix-v1-1-17543611ae31@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Set use_nsecs=true as timestamp is reported in ns. Lack of this result
in smaller timestamp error window which cause error during phc2sys
execution on E825 NICs:
phc2sys[1768.256]: ioctl PTP_SYS_OFFSET_PRECISE: Invalid argument
This problem was introduced in the cited commit which omitted setting
use_nsecs to true when converting the ice driver to use
convert_base_to_cs().
Testing hints (ethX is PF netdev):
phc2sys -s ethX -c CLOCK_REALTIME -O 37 -m
phc2sys[1769.256]: CLOCK_REALTIME phc offset -5 s0 freq -0 delay 0
Fixes: d4bea547eb ("ice/ptp: Remove convert_art_to_tsc()")
Signed-off-by: Anton Nadezhdin <anton.nadezhdin@intel.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Reviewed-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Tested-by: Rinitha S <sx.rinitha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
If a reset event is received from the PF early in the init cycle, the
state machine hangs for about 25 seconds.
Reproducer:
echo 1 > /sys/class/net/$PF0/device/sriov_numvfs
ip link set dev $PF0 vf 0 mac $NEW_MAC
The log shows:
[792.620416] ice 0000:5e:00.0: Enabling 1 VFs
[792.738812] iavf 0000:5e:01.0: enabling device (0000 -> 0002)
[792.744182] ice 0000:5e:00.0: Enabling 1 VFs with 17 vectors and 16 queues per VF
[792.839964] ice 0000:5e:00.0: Setting MAC 52:54:00:00:00:11 on VF 0. VF driver will be reinitialized
[813.389684] iavf 0000:5e:01.0: Failed to communicate with PF; waiting before retry
[818.635918] iavf 0000:5e:01.0: Hardware came out of reset. Attempting reinit.
[818.766273] iavf 0000:5e:01.0: Multiqueue Enabled: Queue pair count = 16
Fix it by scheduling the reset task and making the reset task capable of
resetting early in the init cycle.
Fixes: ef8693eb90 ("i40evf: refactor reset handling")
Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
Tested-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
When a VFLR interrupt is received during a VF reset initiated from a
different source, the VFLR may be not fully handled. This can
leave the VF in an undefined state.
To address this, set the I40E_VFLR_EVENT_PENDING bit again during VFLR
handling if the reset is not yet complete. This ensures the driver
will properly complete the VF reset in such scenarios.
Fixes: 52424f974b ("i40e: Fix VF hang when reset is triggered on another VF")
Signed-off-by: Robert Malz <robert.malz@canonical.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
The function i40e_vc_reset_vf attempts, up to 20 times, to handle a
VF reset request, using the return value of i40e_reset_vf as an indicator
of whether the reset was successfully triggered. Currently, i40e_reset_vf
always returns true, which causes new reset requests to be ignored if a
different VF reset is already in progress.
This patch updates the return value of i40e_reset_vf to reflect when
another VF reset is in progress, allowing the caller to properly use
the retry mechanism.
Fixes: 52424f974b ("i40e: Fix VF hang when reset is triggered on another VF")
Signed-off-by: Robert Malz <robert.malz@canonical.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
When the execution of ath12k_core_hw_group_assign() or
ath12k_core_hw_group_create() fails, the registered notifier chain is not
unregistered properly. Its memory is freed after rmmod, which may trigger
to a use-after-free (UAF) issue if there is a subsequent access to this
notifier chain.
Fixes the issue by calling ath12k_core_panic_notifier_unregister() in
failure cases.
Call trace:
notifier_chain_register+0x4c/0x1f0 (P)
atomic_notifier_chain_register+0x38/0x68
ath12k_core_init+0x50/0x4e8 [ath12k]
ath12k_pci_probe+0x5f8/0xc28 [ath12k]
pci_device_probe+0xbc/0x1a8
really_probe+0xc8/0x3a0
__driver_probe_device+0x84/0x1b0
driver_probe_device+0x44/0x130
__driver_attach+0xcc/0x208
bus_for_each_dev+0x84/0x100
driver_attach+0x2c/0x40
bus_add_driver+0x130/0x260
driver_register+0x70/0x138
__pci_register_driver+0x68/0x80
ath12k_pci_init+0x30/0x68 [ath12k]
ath12k_init+0x28/0x78 [ath12k]
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Fixes: 6f245ea0ec ("wifi: ath12k: introduce device group abstraction")
Signed-off-by: Miaoqing Pan <miaoqing.pan@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <quic_bqiang@quicinc.com>
Link: https://patch.msgid.link/20250604055250.1228501-1-miaoqing.pan@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently to get firmware stats, ath11k_mac_op_get_txpower() calls
ath11k_fw_stats_request() and ath11k_mac_op_sta_statistics() calls
ath11k_mac_get_fw_stats(). Those two helpers are basically doing
the same, except for:
1. ath11k_mac_get_fw_stats() verifies ar->state inside itself.
2. ath11k_mac_get_fw_stats() calls ath11k_mac_fw_stats_request()
which then calls ath11k_mac_fw_stats_reset() to free pdev/vdev
stats whereas only pdev stats are freed by ath11k_fw_stats_request().
3. ath11k_mac_get_fw_stats() waits for ar->fw_stats_complete and
ar->fw_stats_done, whereas ath11k_fw_stats_request() only waits for
ar->fw_stats_complete.
Change to call ath11k_mac_get_fw_stats() in ath11k_mac_op_get_txpower().
This is valid because:
1. ath11k_mac_op_get_txpower() also has the same request on ar->state.
2. it is harmless to call ath11k_fw_stats_vdevs_free() since
ar->fw_stats.vdevs should be empty and there should be no one
expecting it at that time.
3. ath11k_mac_op_get_txpower() only needs pdev stats. For pdev stats,
ar->fw_stats_done is set to true whenever ar->fw_stats_complete is
set to true in ath11k_update_stats_event(). So additional wait on
ar->fw_stats_done does not wast any time.
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250220082448.31039-8-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently ath11k_mac_get_fw_stats() is acquiring/releasing ar->conf_mutex by itself.
In order to reuse this function in a context where that lock is already taken, move
lock handling to its callers, then the function itself only has to assert it.
There is only one caller now, i.e., ath11k_mac_op_sta_statistics(), so add lock handling
there.
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250220082448.31039-7-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Commit b488c76644 ("ath11k: report rssi of each chain to mac80211 for QCA6390/WCN6855")
and commit c3b39553fc ("ath11k: add signal report to mac80211 for QCA6390 and WCN6855")
call debugfs functions in mac ops. Those functions are no-ops if CONFIG_ATH11K_DEBUGFS is
not enabled, thus cause wrong status reported.
Move them to mac.c.
Besides, since WMI_REQUEST_RSSI_PER_CHAIN_STAT and WMI_REQUEST_VDEV_STAT stats could also
be requested via mac ops, process them directly in ath11k_update_stats_event().
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Fixes: b488c76644 ("ath11k: report rssi of each chain to mac80211 for QCA6390/WCN6855")
Fixes: c3b39553fc ("ath11k: add signal report to mac80211 for QCA6390 and WCN6855")
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250220082448.31039-5-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
For WMI_REQUEST_VDEV_STAT request, firmware might split response into
multiple events dut to buffer limit, hence currently in
ath11k_debugfs_fw_stats_process() we wait until all events received.
In case there is no vdev started, this results in that below condition
would never get satisfied
((++ar->fw_stats.num_vdev_recvd) == total_vdevs_started)
finally the requestor would be blocked until wait time out.
The same applies to WMI_REQUEST_BCN_STAT request as well due to:
((++ar->fw_stats.num_bcn_recvd) == ar->num_started_vdevs)
Change to check the number of started vdev first: if it is zero, finish
wait directly; if not, follow the old way.
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Fixes: d5c65159f2 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250220082448.31039-4-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently ath11k_debugfs_fw_stats_process() is using static variables to count
firmware stat events. Taking num_vdev as an example, if for whatever reason (
say ar->num_started_vdevs is 0 or firmware bug etc.) the following condition
(++num_vdev) == total_vdevs_started
is not met, is_end is not set thus num_vdev won't be cleared. Next time when
firmware stats is requested again, even if everything is working fine, we will
fail due to the condition above will never be satisfied.
The same applies to num_bcn as well.
Change to use non-static counters so that we have a chance to clear them each
time firmware stats is requested. Currently only ath11k_fw_stats_request() and
ath11k_debugfs_fw_stats_request() are requesting firmware stats, so clear
counters there.
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Fixes: da3a9d3c15 ("ath11k: refactor debugfs code into debugfs.c")
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Acked-by: Kalle Valo <kvalo@kernel.org>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250220082448.31039-3-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
We get report [1] that CPU is running a hot loop in
ath11k_debugfs_fw_stats_request():
94.60% 0.00% i3status [kernel.kallsyms] [k] do_syscall_64
|
--94.60%--do_syscall_64
|
--94.55%--__sys_sendmsg
___sys_sendmsg
____sys_sendmsg
netlink_sendmsg
netlink_unicast
genl_rcv
netlink_rcv_skb
genl_rcv_msg
|
--94.55%--genl_family_rcv_msg_dumpit
__netlink_dump_start
netlink_dump
genl_dumpit
nl80211_dump_station
|
--94.55%--ieee80211_dump_station
sta_set_sinfo
|
--94.55%--ath11k_mac_op_sta_statistics
ath11k_debugfs_get_fw_stats
|
--94.55%--ath11k_debugfs_fw_stats_request
|
|--41.73%--_raw_spin_lock_bh
|
|--22.74%--__local_bh_enable_ip
|
|--9.22%--_raw_spin_unlock_bh
|
--6.66%--srso_alias_safe_ret
This is because, if for whatever reason ar->fw_stats_done is not set by
ath11k_update_stats_event(), ath11k_debugfs_fw_stats_request() won't yield
CPU before an up to 3s timeout.
Change to completion mechanism to avoid CPU burning.
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.37
Fixes: d5c65159f2 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Reported-by: Yury Vostrikov <mon@unformed.ru>
Closes: https://lore.kernel.org/all/7324ac7a-8b7a-42a5-aa19-de52138ff638@app.fastmail.com/ # [1]
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250220082448.31039-2-quic_bqiang@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
the wil6210 driver irq handling code is unconditionally writing
edma irq registers which are supposed to be only used on Talyn chipsets.
This however leade to a chipset hang on the older sparrow chipset
generation and firmware will not even boot.
Fix that by simply checking for edma support before handling these
registers.
Tested on Netgear R9000
Signed-off-by: Sebastian Gottschall <s.gottschall@dd-wrt.com>
Link: https://patch.msgid.link/20250304012131.25970-2-s.gottschall@dd-wrt.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
In some scenarios, the firmware may be stopped before the interface is
removed, either due to a crash or because the remoteproc (e.g., MPSS)
is shut down early during system reboot or shutdown.
This leads to a delay during interface teardown, as the driver waits for
a vdev delete response that never arrives, eventually timing out.
Example (SNOC):
$ echo stop > /sys/class/remoteproc/remoteproc0/state
[ 71.64] remoteproc remoteproc0: stopped remote processor modem
$ reboot
[ 74.84] ath10k_snoc c800000.wifi: failed to transmit packet, dropping: -108
[ 74.84] ath10k_snoc c800000.wifi: failed to submit frame: -108
[...]
[ 82.39] ath10k_snoc c800000.wifi: Timeout in receiving vdev delete response
To avoid this, skip waiting for the vdev delete response if the firmware is
already marked as unreachable (`ATH10K_FLAG_CRASH_FLUSH`), similar to how
`ath10k_mac_wait_tx_complete()` and `ath10k_vdev_setup_sync()` handle this case.
Signed-off-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20250522131704.612206-1-loic.poulain@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
In ath10k_snoc_hif_stop() we skip disabling the IRQs in the crash
recovery flow, but we still unconditionally call enable again in
ath10k_snoc_hif_start().
We can't check the ATH10K_FLAG_CRASH_FLUSH bit since it is cleared
before hif_start() is called, so instead check the
ATH10K_SNOC_FLAG_RECOVERY flag and skip enabling the IRQs during crash
recovery.
This fixes unbalanced IRQ enable splats that happen after recovering from
a crash.
Fixes: 0e622f67e0 ("ath10k: add support for WCN3990 firmware crash recovery")
Signed-off-by: Caleb Connolly <caleb.connolly@linaro.org>
Tested-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20250318205043.1043148-1-caleb.connolly@linaro.org
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
The kernel robot reported the following errors when the netc-lib driver
was compiled as a loadable module and the enetc-core driver was built-in.
ld.lld: error: undefined symbol: ntmp_init_cbdr
referenced by enetc_cbdr.c:88 (drivers/net/ethernet/freescale/enetc/enetc_cbdr.c:88)
ld.lld: error: undefined symbol: ntmp_free_cbdr
referenced by enetc_cbdr.c:96 (drivers/net/ethernet/freescale/enetc/enetc_cbdr.c:96)
Simply changing "tristate" to "bool" can fix this issue, but considering
that the netc-lib driver needs to support being compiled as a loadable
module and LS1028 does not need the netc-lib driver. Therefore, we add a
boolean symbol 'NXP_NTMP' to enable 'NXP_NETC_LIB' as needed. And when
adding NETC switch driver support in the future, there is no need to
modify the dependency, just select "NXP_NTMP" and "NXP_NETC_LIB" at the
same time.
Reported-by: Arnd Bergmann <arnd@kernel.org>
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202505220734.x6TF6oHR-lkp@intel.com/
Fixes: 4701073c3d ("net: enetc: add initial netc-lib driver to support NTMP")
Suggested-by: Arnd Bergmann <arnd@kernel.org>
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When Linux sends out untagged traffic from a port, it will enter the CPU
port without any VLAN tag, even if the port is a member of a vlan
filtering bridge with a PVID egress untagged VLAN.
This makes the CPU port's PVID take effect, and the PVID's VLAN
table entry controls if the packet will be tagged on egress.
Since commit 45e9d59d39 ("net: dsa: b53: do not allow to configure
VLAN 0") we remove bridged ports from VLAN 0 when joining or leaving a
VLAN aware bridge. But we also clear the untagged bit, causing untagged
traffic from the controller to become tagged with VID 0 (and priority
0).
Fix this by not touching the untagged map of VLAN 0. Additionally,
always keep the CPU port as a member, as the untag map is only effective
as long as there is at least one member, and we would remove it when
bridging all ports and leaving no standalone ports.
Since Linux (and the switch) treats VLAN 0 tagged traffic like untagged,
the actual impact of this is rather low, but this also prevented earlier
detection of the issue.
Fixes: 45e9d59d39 ("net: dsa: b53: do not allow to configure VLAN 0")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/20250602194914.1011890-1-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Jakub Kicinski:
"Including fixes from CAN, wireless, Bluetooth, and Netfilter.
Current release - regressions:
- Revert "kunit: configs: Enable CONFIG_INIT_STACK_ALL_PATTERN in
all_tests", makes kunit error out if compiler is old
- wifi: iwlwifi: mvm: fix assert on suspend
- rxrpc: fix return from none_validate_challenge()
Current release - new code bugs:
- ovpn: couple of fixes for socket cleanup and UDP-tunnel teardown
- can: kvaser_pciefd: refine error prone echo_skb_max handling logic
- fix net_devmem_bind_dmabuf() stub when DEVMEM not compiled
- eth: airoha: fixes for config / accel in bridge mode
Previous releases - regressions:
- Bluetooth: hci_qca: move the SoC type check to the right place, fix
GPIO integration
- prevent a NULL deref in rtnl_create_link() after locking changes
- fix udp gso skb_segment after pull from frag_list
- hv_netvsc: fix potential deadlock in netvsc_vf_setxdp()
Previous releases - always broken:
- netfilter:
- nf_nat: also check reverse tuple to obtain clashing entry
- nf_set_pipapo_avx2: fix initial map fill (zeroing)
- fix the helper for incremental update of packet checksums after
modifying the IP address, used by ILA and BPF
- eth:
- stmmac: prevent div by 0 when clock rate is misconfigured
- ice: fix Tx scheduler handling of XDP and changing queue count
- eth: fix support for the RGMII interface when delays configured"
* tag 'net-6.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (76 commits)
calipso: unlock rcu before returning -EAFNOSUPPORT
seg6: Fix validation of nexthop addresses
net: prevent a NULL deref in rtnl_create_link()
net: annotate data-races around cleanup_net_task
selftests: drv-net: tso: make bkg() wait for socat to quit
selftests: drv-net: tso: fix the GRE device name
selftests: drv-net: add configs for the TSO test
wireguard: device: enable threaded NAPI
netlink: specs: rt-link: decode ip6gre
netlink: specs: rt-link: add missing byte-order properties
net: wwan: mhi_wwan_mbim: use correct mux_id for multiplexing
wifi: cfg80211/mac80211: correctly parse S1G beacon optional elements
net: dsa: b53: do not touch DLL_IQQD on bcm53115
net: dsa: b53: allow RGMII for bcm63xx RGMII ports
net: dsa: b53: do not configure bcm63xx's IMP port interface
net: dsa: b53: do not enable RGMII delay on bcm63xx
net: dsa: b53: do not enable EEE on bcm63xx
net: ti: icssg-prueth: Fix swapped TX stats for MII interfaces.
selftests: netfilter: nft_nat.sh: add test for reverse clash with nat
netfilter: nf_nat: also check reverse tuple to obtain clashing entry
...
Tony Nguyen says:
====================
iavf: get rid of the crit lock
Przemek Kitszel says:
Fix some deadlocks in iavf, and make it less error prone for the future.
Patch 1 is simple and independent from the rest.
Patches 2, 3, 4 are strictly a refactor, but it enables the last patch
to be much smaller.
(Technically Jake given his RB tags not knowing I will send it to -net).
Patch 5 just adds annotations, this also helps prove last patch to be correct.
Patch 6 removes the crit lock, with its unusual try_lock()s.
I have more refactoring for scheduling done for -next, to be sent soon.
There is a simple test:
add VF; decrease number of queueus; remove VF
that was way too hard to pass without this series :)
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
iavf: get rid of the crit lock
iavf: sprinkle netdev_assert_locked() annotations
iavf: extract iavf_watchdog_step() out of iavf_watchdog_task()
iavf: simplify watchdog_task in terms of adminq task scheduling
iavf: centralize watchdog requeueing itself
iavf: iavf_suspend(): take RTNL before netdev_lock()
====================
Link: https://patch.msgid.link/20250603171710.2336151-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Enable threaded NAPI by default for WireGuard devices in response to low
performance behavior that we observed when multiple tunnels (and thus
multiple wg devices) are deployed on a single host. This affects any
kind of multi-tunnel deployment, regardless of whether the tunnels share
the same endpoints or not (i.e., a VPN concentrator type of gateway
would also be affected).
The problem is caused by the fact that, in case of a traffic surge that
involves multiple tunnels at the same time, the polling of the NAPI
instance of all these wg devices tends to converge onto the same core,
causing underutilization of the CPU and bottlenecking performance.
This happens because NAPI polling is hosted by default in softirq
context, but the WireGuard driver only raises this softirq after the rx
peer queue has been drained, which doesn't happen during high traffic.
In this case, the softirq already active on a core is reused instead of
raising a new one.
As a result, once two or more tunnel softirqs have been scheduled on
the same core, they remain pinned there until the surge ends.
In our experiments, this almost always leads to all tunnel NAPIs being
handled on a single core shortly after a surge begins, limiting
scalability to less than 3× the performance of a single tunnel, despite
plenty of unused CPU cores being available.
The proposed mitigation is to enable threaded NAPI for all WireGuard
devices. This moves the NAPI polling context to a dedicated per-device
kernel thread, allowing the scheduler to balance the load across all
available cores.
On our 32-core gateways, enabling threaded NAPI yields a ~4× performance
improvement with 16 tunnels, increasing throughput from ~13 Gbps to
~48 Gbps. Meanwhile, CPU usage on the receiver (which is the bottleneck)
jumps from 20% to 100%.
We have found no performance regressions in any scenario we tested.
Single-tunnel throughput remains unchanged.
More details are available in our Netdev paper.
Link: https://netdevconf.info/0x18/docs/netdev-0x18-paper23-talk-paper.pdf
Signed-off-by: Mirco Barone <mirco.barone@polito.it>
Fixes: e7096c131e ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Link: https://patch.msgid.link/20250605120616.2808744-1-Jason@zx2c4.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Antonio Quartulli says:
====================
In this batch you can find the following bug fixes:
Patch 1: when releasing a UDP socket we were wrongly invoking
setup_udp_tunnel_sock() with an empty config. This was not
properly shutting down the UDP encap state.
With this patch we simply undo what was done during setup.
Patch 2: ovpn was holding a reference to a 'struct socket'
without increasing its reference counter. This was intended
and worked as expected until we hit a race condition where
user space tries to close the socket while kernel space is
also releasing it. In this case the (struct socket *)->sk
member would disappear under our feet leading to a null-ptr-deref.
This patch fixes this issue by having struct ovpn_socket hold
a reference directly to the sk member while also increasing
its reference counter.
Patch 3: in case of errors along the TCP RX path (softirq)
we want to immediately delete the peer, but this operation may
sleep. With this patch we move the peer deletion to a scheduled
worker.
Patch 4 and 5 are instead fixing minor issues in the ovpn
kselftests.
* tag 'ovpn-net-20250603' of https://github.com/OpenVPN/ovpn-net-next:
selftest/net/ovpn: fix missing file
selftest/net/ovpn: fix TCP socket creation
ovpn: avoid sleep in atomic context in TCP RX error path
ovpn: ensure sk is still valid during cleanup
ovpn: properly deconfigure UDP-tunnel
====================
Link: https://patch.msgid.link/20250603111110.4575-1-antonio@openvpn.net/
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Recent Qualcomm chipsets like SDX72/75 require MBIM sessionId mapping
to muxId in the range (0x70-0x8F) for the PCIe tethered use.
This has been partially addressed by the referenced commit, mapping
the default data call to muxId = 112, but the multiplexed data calls
scenario was not properly considered, mapping sessionId = 1 to muxId
1, while it should have been 113.
Fix this by moving the session_id assignment logic to mhi_mbim_newlink,
in order to map sessionId = n to muxId = n + WDS_BIND_MUX_DATA_PORT_MUX_ID.
Fixes: 65bc58c3dc ("net: wwan: mhi: make default data link id configurable")
Signed-off-by: Daniele Palmas <dnlplm@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20250603091204.2802840-1-dnlplm@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add RGMII to supported interfaces for BCM63xx RGMII ports so they can be
actually used in RGMII mode.
Without this, phylink will fail to configure them:
[ 3.580000] b53-switch 10700000.switch GbE3 (uninitialized): validation of rgmii with support 0000000,00000000,00000000,000062ff and advertisement 0000000,00000000,00000000,000062ff failed: -EINVAL
[ 3.600000] b53-switch 10700000.switch GbE3 (uninitialized): failed to connect to PHY: -EINVAL
[ 3.610000] b53-switch 10700000.switch GbE3 (uninitialized): error -22 setting up PHY for tree 0, switch 0, port 4
Fixes: ce3bf94871 ("net: dsa: b53: add support for BCM63xx RGMIIs")
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Link: https://patch.msgid.link/20250602193953.1010487-5-jonas.gorski@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
