Merge tag 'mlx5-fixes-2022-11-24' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-fixes-2022-11-24
This series provides bug fixes to mlx5 driver.

Focusing on error handling and proper memory management in mlx5, in
general and in the newly added macsec module.

I still have few fixes left in my queue and I hope those will be the
last ones for mlx5 for this cycle.

Please pull and let me know if there is any problem.

Happy thanksgiving.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch '10GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-11-23 (ixgbevf, i40e, fm10k, iavf, e100)

This series contains updates to various Intel drivers.

Shang XiaoJing fixes init module error path stop to resource leaks for
ixgbevf and i40e.

Yuan Can also does the same for fm10k and iavf.

Wang Hai stops freeing of skb as it was causing use after free error for
e100.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: fix null-ptr-deref while probe() failed

I got a null-ptr-deref report as following when doing fault injection test:

BUG: kernel NULL pointer dereference, address: 0000000000000058
Oops: 0000 [#1] PREEMPT SMP KASAN PTI
CPU: 1 PID: 253 Comm: 507-spi-dm9051 Tainted: G    B            N 6.1.0-rc3+
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1ubuntu1.1 04/01/2014
RIP: 0010:klist_put+0x2d/0xd0
Call Trace:
 <TASK>
 klist_remove+0xf1/0x1c0
 device_release_driver_internal+0x23e/0x2d0
 bus_remove_device+0x1bd/0x240
 device_del+0x357/0x770
 phy_device_remove+0x11/0x30
 mdiobus_unregister+0xa5/0x140
 release_nodes+0x6a/0xa0
 devres_release_all+0xf8/0x150
 device_unbind_cleanup+0x19/0xd0

//probe path:
phy_device_register()
  device_add()

phy_connect
  phy_attach_direct() //set device driver
    probe() //it's failed, driver is not bound
    device_bind_driver() // probe failed, it's not called

//remove path:
phy_device_remove()
  device_del()
    device_release_driver_internal()
      __device_release_driver() //dev->drv is not NULL
        klist_remove() <- knode_driver is not added yet, cause null-ptr-deref

In phy_attach_direct(), after setting the 'dev->driver', probe() fails,
device_bind_driver() is not called, so the knode_driver->n_klist is not
set, then it causes null-ptr-deref in __device_release_driver() while
deleting device. Fix this by setting dev->driver to NULL in the error
path in phy_attach_direct().

Fixes: e13934563db0 ("[PATCH] PHY Layer fixup")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: loopback: use NET_NAME_PREDICTABLE for name_assign_type

When the name_assign_type attribute was introduced (commit
685343fc3ba6, "net: add name_assign_type netdev attribute"), the
loopback device was explicitly mentioned as one which would make use
of NET_NAME_PREDICTABLE:

    The name_assign_type attribute gives hints where the interface name of a
    given net-device comes from. These values are currently defined:
...
      NET_NAME_PREDICTABLE:
        The ifname has been assigned by the kernel in a predictable way
        that is guaranteed to avoid reuse and always be the same for a
        given device. Examples include statically created devices like
        the loopback device [...]

Switch to that so that reading /sys/class/net/lo/name_assign_type
produces something sensible instead of returning -EINVAL.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fec: don't reset irq coalesce settings to defaults on "ip link up"

Currently, when a FEC device is brought up, the irq coalesce settings
are reset to their default values (1000us, 200 frames). That's
unexpected, and breaks for example use of an appropriate .link file to
make systemd-udev apply the desired
settings (https://www.freedesktop.org/software/systemd/man/systemd.link.html),
or any other method that would do a one-time setup during early boot.

Refactor the code so that fec_restart() instead uses
fec_enet_itr_coal_set(), which simply applies the settings that are
stored in the private data, and initialize that private data with the
default values.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

octeontx2-pf: Fix pfc_alloc_status array overflow

This patch addresses pfc_alloc_status array overflow occurring for
send queue index value greater than PFC priority. Queue index can be
greater than supported PFC priority for multiple scenarios (e.g. QoS,
during non zero SMQ allocation for a PF/VF).
In those scenarios the API should return default tx scheduler '0'.
This is causing mbox errors as otx2_get_smq_idx returing invalid smq value.

Fixes: 99c969a83d82 ("octeontx2-pf: Add egress PFC support")
Signed-off-by: Suman Ghosh <sumang@marvell.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Set MAC's flow control register to reflect current settings

Currently, pause frame register GMAC_RX_FLOW_CTRL_RFE is not updated
correctly when 'ethtool -A <IFACE> autoneg off rx off tx off' command
is issued. This fix ensures the flow control change is reflected directly
in the GMAC_RX_FLOW_CTRL_RFE register.

Fixes: 46f69ded988d ("net: stmmac: Use resolved link config in mac_link_up()")
Cc: <stable@vger.kernel.org> # 5.10.x
Signed-off-by: Goh, Wei Sheng <wei.sheng.goh@intel.com>
Signed-off-by: Noor Azura Ahmad Tarmizi <noor.azura.ahmad.tarmizi@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

aquantia: Do not purge addresses when setting the number of rings

IPV6 addresses are purged when setting the number of rx/tx
rings using ethtool -G. The function aq_set_ringparam
calls dev_close, which removes the addresses. As a solution,
call an internal function (aq_ndev_close).

Fixes: c1af5427954b ("net: aquantia: Ethtool based ring size configuration")
Signed-off-by: Izabela Bakollari <ibakolla@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: avoid defines prefixed with CONFIG

Defines prefixed with "CONFIG" should be limited to proper Kconfig options,
that are introduced in a Kconfig file.

Here, constants for bitmap indices of some configs are defined and these
defines begin with the config's name, and are suffixed with BITMAP_IDX.

To avoid defines prefixed with "CONFIG", name these constants
BITMAP_IDX_FOR_CONFIG_XYZ instead of CONFIG_XYZ_BITMAP_IDX.

No functional change.

Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix sleep-in-atomic-context bugs caused by msleep

The watchdog timer is used to monitor whether the process
of transmitting data is timeout. If we use qlcnic driver,
the dev_watchdog() that is the timer handler of watchdog
timer will call qlcnic_tx_timeout() to process the timeout.
But the qlcnic_tx_timeout() calls msleep(), as a result,
the sleep-in-atomic-context bugs will happen. The processes
are shown below:

   (atomic context)
dev_watchdog
  qlcnic_tx_timeout
    qlcnic_83xx_idc_request_reset
      qlcnic_83xx_lock_driver
        msleep

---------------------------

   (atomic context)
dev_watchdog
  qlcnic_tx_timeout
    qlcnic_83xx_idc_request_reset
      qlcnic_83xx_lock_driver
        qlcnic_83xx_recover_driver_lock
          msleep

Fix by changing msleep() to mdelay(), the mdelay() is
busy-waiting and the bugs could be mitigated.

Fixes: 629263acaea3 ("qlcnic: 83xx CNA inter driver communication mechanism")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'linux-can-fixes-for-6.1-20221124' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
linux-can-fixes-for-6.1-20221124

this is a pull request of 8 patches for net/master.

Ziyang Xuan contributes a patch for the can327, fixing a potential SKB
leak when the netdev is down.

Heiko Schocher's patch for the sja1000 driver fixes the width of the
definition of the OCR_MODE_MASK.

Zhang Changzhong contributes 4 patches. In the sja1000_isa, cc770, and
m_can_pci drivers the error path in the probe() function and in case
of the etas_es58x a function that is called by probe() are fixed.

Jiasheng Jiang add a missing check for the return value of the
devm_clk_get() in the m_can driver.

Yasushi SHOJI's patch for the mcba_usb fixes setting of the external
termination resistor.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

can: mcba_usb: Fix termination command argument

Microchip USB Analyzer can activate the internal termination resistors
by setting the "termination" option ON, or OFF to to deactivate them.
As I've observed, both with my oscilloscope and captured USB packets
below, you must send "0" to turn it ON, and "1" to turn it OFF.

From the schematics in the user's guide, I can confirm that you must
drive the CAN_RES signal LOW "0" to activate the resistors.

Reverse the argument value of usb_msg.termination to fix this.

These are the two commands sequence, ON then OFF.

> No.     Time           Source                Destination           Protocol Length Info
>       1 0.000000       host                  1.3.1                 USB      46     URB_BULK out
>
> Frame 1: 46 bytes on wire (368 bits), 46 bytes captured (368 bits)
> USB URB
> Leftover Capture Data: a80000000000000000000000000000000000a8
>
> No.     Time           Source                Destination           Protocol Length Info
>       2 4.372547       host                  1.3.1                 USB      46     URB_BULK out
>
> Frame 2: 46 bytes on wire (368 bits), 46 bytes captured (368 bits)
> USB URB
> Leftover Capture Data: a80100000000000000000000000000000000a9

Signed-off-by: Yasushi SHOJI <yashi@spacecubics.com>
Link: https://lore.kernel.org/all/20221124152504.125994-1-yashi@spacecubics.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: m_can: Add check for devm_clk_get

Since the devm_clk_get may return error,
it should be better to add check for the cdev->hclk,
as same as cdev->cclk.

Fixes: f524f829b75a ("can: m_can: Create a m_can platform framework")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Link: https://lore.kernel.org/all/20221123063651.26199-1-jiasheng@iscas.ac.cn
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: m_can: pci: add missing m_can_class_free_dev() in probe/remove methods

In m_can_pci_remove() and error handling path of m_can_pci_probe(),
m_can_class_free_dev() should be called to free resource allocated by
m_can_class_allocate_dev(), otherwise there will be memleak.

Fixes: cab7ffc0324f ("can: m_can: add PCI glue driver for Intel Elkhart Lake")
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Link: https://lore.kernel.org/all/1668168684-6390-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: etas_es58x: es58x_init_netdev(): free netdev when register_candev()

In case of register_candev() fails, clear
es58x_dev->netdev[channel_idx] and add free_candev(). Otherwise
es58x_free_netdevs() will unregister the netdev that has never been
registered.

Fixes: 8537257874e9 ("can: etas_es58x: add core support for ETAS ES58X CAN USB interfaces")
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Acked-by: Arunachalam Santhanam <Arunachalam.Santhanam@in.bosch.com>
Acked-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Link: https://lore.kernel.org/all/1668413685-23354-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: cc770: cc770_isa_probe(): add missing free_cc770dev()

Add the missing free_cc770dev() before return from cc770_isa_probe()
in the register_cc770dev() error handling case.

In addition, remove blanks before goto labels.

Fixes: 7e02e5433e00 ("can: cc770: legacy CC770 ISA bus driver")
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Link: https://lore.kernel.org/all/1668168557-6024-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: sja1000_isa: sja1000_isa_probe(): add missing free_sja1000dev()

Add the missing free_sja1000dev() before return from
sja1000_isa_probe() in the register_sja1000dev() error handling case.

In addition, remove blanks before goto labels.

Fixes: 2a6ba39ad6a2 ("can: sja1000: legacy SJA1000 ISA bus driver")
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Link: https://lore.kernel.org/all/1668168521-5540-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: sja1000: fix size of OCR_MODE_MASK define

bitfield mode in ocr register has only 2 bits not 3, so correct
the OCR_MODE_MASK define.

Signed-off-by: Heiko Schocher <hs@denx.de>
Link: https://lore.kernel.org/all/20221123071636.2407823-1-hs@denx.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: can327: can327_feed_frame_to_netdev(): fix potential skb leak when netdev is down

In can327_feed_frame_to_netdev(), it did not free the skb when netdev
is down, and all callers of can327_feed_frame_to_netdev() did not free
allocated skb too. That would trigger skb leak.

Fix it by adding kfree_skb() in can327_feed_frame_to_netdev() when netdev
is down. Not tested, just compiled.

Fixes: 43da2f07622f ("can: can327: CAN/ldisc driver for ELM327 based OBD-II adapters")
Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Link: https://lore.kernel.org/all/20221110061437.411525-1-william.xuanziyang@huawei.com
Reviewed-by: Max Staudt <max@enpas.org>
Cc: stable@vger.kernel.org
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

net: thunderx: Fix the ACPI memory leak

The ACPI buffer memory (string.pointer) should be freed as the buffer is
not used after returning from bgx_acpi_match_id(), free it to prevent
memory leak.

Fixes: 46b903a01c05 ("net, thunder, bgx: Add support to get MAC address from ACPI.")
Signed-off-by: Yu Liao <liaoyu15@huawei.com>
Link: https://lore.kernel.org/r/20221123082237.1220521-1-liaoyu15@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

octeontx2-af: Fix reference count issue in rvu_sdp_init()

pci_get_device() will decrease the reference count for the *from*
parameter. So we don't need to call put_device() to decrease the
reference. Let's remove the put_device() in the loop and only decrease
the reference count of the returned 'pdev' for the last loop because it
will not be passed to pci_get_device() as input parameter. We don't need
to check if 'pdev' is NULL because it is already checked inside
pci_dev_put(). Also add pci_dev_put() for the error path.

Fixes: fe1939bb2340 ("octeontx2-af: Add SDP interface support")
Signed-off-by: Xiongfeng Wang <wangxiongfeng2@huawei.com>
Reviewed-by: Saeed Mahameed <saeed@kernel.org>
Link: https://lore.kernel.org/r/20221123065919.31499-1-wangxiongfeng2@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: altera_tse: release phylink resources in tse_shutdown()

Call phylink_disconnect_phy() in tse_shutdown() to release the
resources occupied by phylink_of_phy_connect() in the tse_open().

Fixes: fef2998203e1 ("net: altera: tse: convert to phylink")
Signed-off-by: Liu Jian <liujian56@huawei.com>
Link: https://lore.kernel.org/r/20221123011617.332302-1-liujian56@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

virtio_net: Fix probe failed when modprobe virtio_net

When doing the following test steps, an error was found:
  step 1: modprobe virtio_net succeeded
    # modprobe virtio_net        <-- OK

  step 2: fault injection in register_netdevice()
    # modprobe -r virtio_net     <-- OK
    # ...
      FAULT_INJECTION: forcing a failure.
      name failslab, interval 1, probability 0, space 0, times 0
      CPU: 0 PID: 3521 Comm: modprobe
      Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
      Call Trace:
       <TASK>
       ...
       should_failslab+0xa/0x20
       ...
       dev_set_name+0xc0/0x100
       netdev_register_kobject+0xc2/0x340
       register_netdevice+0xbb9/0x1320
       virtnet_probe+0x1d72/0x2658 [virtio_net]
       ...
       </TASK>
      virtio_net: probe of virtio0 failed with error -22

  step 3: modprobe virtio_net failed
    # modprobe virtio_net        <-- failed
      virtio_net: probe of virtio0 failed with error -2

The root cause of the problem is that the queues are not
disable on the error handling path when register_netdevice()
fails in virtnet_probe(), resulting in an error "-ENOENT"
returned in the next modprobe call in setup_vq().

virtio_pci_modern_device uses virtqueues to send or
receive message, and "queue_enable" records whether the
queues are available. In vp_modern_find_vqs(), all queues
will be selected and activated, but once queues are enabled
there is no way to go back except reset.

Fix it by reset virtio device on error handling path. This
makes error handling follow the same order as normal device
cleanup in virtnet_remove() which does: unregister, destroy
failover, then reset. And that flow is better tested than
error handling so we can be reasonably sure it works well.

Fixes: 024655555021 ("virtio_net: fix use after free on allocation failure")
Signed-off-by: Li Zetao <lizetao1@huawei.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://lore.kernel.org/r/20221122150046.3910638-1-lizetao1@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net/mlx5e: MACsec, block offload requests with encrypt off

Currently offloading MACsec with authentication only (encrypt
property set to off) is not supported, block such requests
when adding/updating a macsec device.

Fixes: 8ff0ac5be144 ("net/mlx5: Add MACsec offload Tx command support")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, fix Tx SA active field update

Currently during update Tx security association (SA) flow, the Tx SA
active state is updated only if the Tx SA in question is the same SA
that the MACsec interface is using for Tx,in consequence when the
MACsec interface chose to work with this Tx SA later, where this SA
for example should have been updated to active state and it was not,
the relevant Tx SA HW context won't be installed, hence the MACSec
flow won't be offloaded.

Fix by update Tx SA active state as part of update flow regardless
whether the SA in question is the same Tx SA used by the MACsec
interface.

Fixes: 8ff0ac5be144 ("net/mlx5: Add MACsec offload Tx command support")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, remove replay window size limitation in offload path

Currently offload path limits replay window size to 32/64/128/256 bits,
such a limitation should not exist since software allows it.
Remove such limitation.

Fixes: eb43846b43c3 ("net/mlx5e: Support MACsec offload replay window")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, fix add Rx security association (SA) rule memory leak

Currently MACsec's add Rx SA flow steering (fs) rule routine
uses a spec object which is dynamically allocated and do
not free it upon leaving. The above led to a memory leak.

Fix by freeing dynamically allocated objects.

Fixes: 3b20949cb21b ("net/mlx5e: Add MACsec RX steering rules")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, fix mlx5e_macsec_update_rxsa bail condition and functionality

Fix update Rx SA wrong bail condition, naturally update functionality
needs to check that something changed otherwise bailout currently the
active state check does just the opposite, furthermore unlike deactivate
path which remove the macsec rules to deactivate the offload, the
activation path does not include the counter part installation of the
macsec rules.

Fix by using correct bailout condition and when Rx SA changes state to
active then add the relevant macsec rules.

While at it, refine function name to reflect more precisely its role.

Fixes: aae3454e4d4c ("net/mlx5e: Add MACsec offload Rx command support")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, fix update Rx secure channel active field

The main functionality for this operation is to update the
active state of the Rx security channel (SC) if the new
active setting is different from the current active state
of this Rx SC, however the relevant active state check is
done post updating the current active state to match the
new active state, effectively blocks any offload state
update for the Rx SC in question.

Fix by delay the assignment to be post the relevant check.

Fixes: aae3454e4d4c ("net/mlx5e: Add MACsec offload Rx command support")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, fix memory leak when MACsec device is deleted

When the MACsec netdevice is deleted, all related Rx/Tx HW/SW
states should be released/deallocated, however currently part
of the Rx security channel association data is not cleaned
properly, hence the memory leaks.

Fix by make sure all related Rx Sc resources are cleaned/freed,
while at it improve code by grouping release SC context in a
function so it can be used in both delete MACsec device and
delete Rx SC operations.

Fixes: 5a39816a75e5 ("net/mlx5e: Add MACsec offload SecY support")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: MACsec, fix RX data path 16 RX security channel limit

Currently the data path metadata flow id mask wrongly limits the
number of different RX security channels (SC) to 16, whereas in
adding RX SC the limit is "2^16 - 1" this cause an overlap in
metadata flow id once more than 16 RX SCs is added, this corrupts
MACsec RX offloaded flow handling.

Fix by using the correct mask, while at it improve code to use this
mask when adding the Rx rule and improve visibility of such errors
by adding debug massage.

Fixes: b7c9400cbc48 ("net/mlx5e: Implement MACsec Rx data path using MACsec skb_metadata_dst")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Use kvfree() in mlx5e_accel_fs_tcp_create()

'accel_tcp' is allocated by kvzalloc(), which should freed by kvfree().

Fixes: f52f2faee581 ("net/mlx5e: Introduce flow steering API")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix a couple error codes

If kvzalloc() fails then return -ENOMEM.  Don't return success.

Fixes: 3b20949cb21b ("net/mlx5e: Add MACsec RX steering rules")
Fixes: e467b283ffd5 ("net/mlx5e: Add MACsec TX steering rules")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix use-after-free when reverting termination table

When having multiple dests with termination tables and second one
or afterwards fails the driver reverts usage of term tables but
doesn't reset the assignment in attr->dests[num_vport_dests].termtbl
which case a use-after-free when releasing the rule.
Fix by resetting the assignment of termtbl to null.

Fixes: 10caabdaad5a ("net/mlx5e: Use termination table for VLAN push actions")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Fix uninitialized variable bug in outlen_write()

If sscanf() return 0, outlen is uninitialized and used in kzalloc(),
this is unexpected. We should return -EINVAL if the string is invalid.

Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: E-switch, Fix duplicate lag creation

If creating bond first and then enabling sriov in switchdev mode,
will hit the following syndrome:

mlx5_core 0000:08:00.0: mlx5_cmd_out_err:778:(pid 25543): CREATE_LAG(0x840) op_mod(0x0) failed, status bad parameter(0x3), syndrome (0x7d49cb), err(-22)

The reason is because the offending patch removes eswitch mode
none. In vf lag, the checking of eswitch mode none is replaced
by checking if sriov is enabled. But when driver enables sriov,
it triggers the bond workqueue task first and then setting sriov
number in pci_enable_sriov(). So the check fails.

Fix it by checking if sriov is enabled using eswitch internal
counter that is set before triggering the bond workqueue task.

Fixes: f019679ea5f2 ("net/mlx5: E-switch, Remove dependency between sriov and eswitch mode")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: E-switch, Destroy legacy fdb table when needed

The cited commit removes eswitch mode none. But when disabling
sriov in legacy mode or changing from switchdev to legacy mode
without sriov enabled, the legacy fdb table is not destroyed.

It is not the right behavior. Destroy legacy fdb table in above
two caes.

Fixes: f019679ea5f2 ("net/mlx5: E-switch, Remove dependency between sriov and eswitch mode")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: DR, Fix uninitialized var warning

Smatch warns this:

drivers/net/ethernet/mellanox/mlx5/core/steering/dr_table.c:81
 mlx5dr_table_set_miss_action() error: uninitialized symbol 'ret'.

Initializing ret with -EOPNOTSUPP and fix missing action case.

Fixes: 7838e1725394 ("net/mlx5: DR, Expose steering table functionality")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net: wwan: t7xx: Fix the ACPI memory leak

The ACPI buffer memory (buffer.pointer) should be freed as the
buffer is not used after acpi_evaluate_object(), free it to
prevent memory leak.

Fixes: 13e920d93e37 ("net: wwan: t7xx: Add core components")
Signed-off-by: Hanjun Guo <guohanjun@huawei.com>
Link: https://lore.kernel.org/r/1669119580-28977-1-git-send-email-guohanjun@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

octeontx2-pf: Add check for devm_kcalloc

As the devm_kcalloc may return NULL pointer,
it should be better to add check for the return
value, as same as the others.

Fixes: e8e095b3b370 ("octeontx2-af: cn10k: Bandwidth profiles config support")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/20221122055449.31247-1-jiasheng@iscas.ac.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: enetc: preserve TX ring priority across reconfiguration

In the blamed commit, a rudimentary reallocation procedure for RX buffer
descriptors was implemented, for the situation when their format changes
between normal (no PTP) and extended (PTP).

enetc_hwtstamp_set() calls enetc_close() and enetc_open() in a sequence,
and this sequence loses information which was previously configured in
the TX BDR Mode Register, specifically via the enetc_set_bdr_prio() call.
The TX ring priority is configured by tc-mqprio and tc-taprio, and
affects important things for TSN such as the TX time of packets. The
issue manifests itself most visibly by the fact that isochron --txtime
reports premature packet transmissions when PTP is first enabled on an
enetc interface.

Save the TX ring priority in a new field in struct enetc_bdr (occupies a
2 byte hole on arm64) in order to make this survive a ring reconfiguration.

Fixes: 434cebabd3a2 ("enetc: Add dynamic allocation of extended Rx BD rings")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Link: https://lore.kernel.org/r/20221122130936.1704151-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: marvell: prestera: add missing unregister_netdev() in prestera_port_create()

If prestera_port_sfp_bind() fails, unregister_netdev() should be called
in error handling path.

Compile tested only.

Fixes: 52323ef75414 ("net: marvell: prestera: add phylink support")
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/r/1669115432-36841-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'nfc-st-nci-restructure-validating-logic-in-evt_transaction'

Martin Faltesek says:

====================
nfc: st-nci: Restructure validating logic in EVT_TRANSACTION

These are the same 3 patches that were applied in st21nfca here:
https://lore.kernel.org/netdev/20220607025729.1673212-1-mfaltesek@google.com
with a couple minor differences.

st-nci has nearly identical code to that of st21nfca for EVT_TRANSACTION,
except that there are two extra validation checks that are not present
in the st-nci code.

The 3/3 patch as coded for st21nfca pulls those checks in, bringing both
drivers into parity.
====================

Link: https://lore.kernel.org/r/20221122004246.4186422-1-mfaltesek@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

nfc: st-nci: fix incorrect sizing calculations in EVT_TRANSACTION

The transaction buffer is allocated by using the size of the packet buf,
and subtracting two which seems intended to remove the two tags which are
not present in the target structure. This calculation leads to under
counting memory because of differences between the packet contents and the
target structure. The aid_len field is a u8 in the packet, but a u32 in
the structure, resulting in at least 3 bytes always being under counted.
Further, the aid data is a variable length field in the packet, but fixed
in the structure, so if this field is less than the max, the difference is
added to the under counting.

To fix, perform validation checks progressively to safely reach the
next field, to determine the size of both buffers and verify both tags.
Once all validation checks pass, allocate the buffer and copy the data.
This eliminates freeing memory on the error path, as validation checks are
moved ahead of memory allocation.

Reported-by: Denis Efremov <denis.e.efremov@oracle.com>
Reviewed-by: Guenter Roeck <groeck@google.com>
Fixes: 5d1ceb7f5e56 ("NFC: st21nfcb: Add HCI transaction event support")
Signed-off-by: Martin Faltesek <mfaltesek@google.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

nfc: st-nci: fix memory leaks in EVT_TRANSACTION

Error path does not free previously allocated memory. Add devm_kfree() to
the failure path.

Reported-by: Denis Efremov <denis.e.efremov@oracle.com>
Reviewed-by: Guenter Roeck <groeck@google.com>
Fixes: 5d1ceb7f5e56 ("NFC: st21nfcb: Add HCI transaction event support")
Signed-off-by: Martin Faltesek <mfaltesek@google.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

nfc: st-nci: fix incorrect validating logic in EVT_TRANSACTION

The first validation check for EVT_TRANSACTION has two different checks
tied together with logical AND. One is a check for minimum packet length,
and the other is for a valid aid_tag. If either condition is true (fails),
then an error should be triggered. The fix is to change && to ||.

Reported-by: Denis Efremov <denis.e.efremov@oracle.com>
Reviewed-by: Guenter Roeck <groeck@google.com>
Fixes: 5d1ceb7f5e56 ("NFC: st21nfcb: Add HCI transaction event support")
Signed-off-by: Martin Faltesek <mfaltesek@google.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec

Steffen Klassert says:

====================
ipsec 2022-11-23

1) Fix "disable_policy" on ipv4 early demuxP Packets after
   the initial packet in a flow might be incorectly dropped
   on early demux if there are no matching policies.
   From Eyal Birger.

2) Fix a kernel warning in case XFRM encap type is not
   available. From Eyal Birger.

3) Fix ESN wrap around for GSO to avoid a double usage of a
    sequence number. From Christian Langrock.

4) Fix a send_acquire race with pfkey_register.
   From Herbert Xu.

5) Fix a list corruption panic in __xfrm_state_delete().
   Thomas Jarosch.

6) Fix an unchecked return value in xfrm6_init().
   Chen Zhongjin.

* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec:
  xfrm: Fix ignored return value in xfrm6_init()
  xfrm: Fix oops in __xfrm_state_delete()
  af_key: Fix send_acquire race with pfkey_register
  xfrm: replay: Fix ESN wrap around for GSO
  xfrm: lwtunnel: squelch kernel warning in case XFRM encap type is not available
  xfrm: fix "disable_policy" on ipv4 early demux
====================

Link: https://lore.kernel.org/r/20221123093117.434274-1-steffen.klassert@secunet.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

1) Fix regression in ipset hash:ip with IPv4 range, from Vishwanath Pai.
   This is fixing up a bug introduced in the 6.0 release.

2) The "netfilter: ipset: enforce documented limit to prevent allocating
   huge memory" patch contained a wrong condition which makes impossible to
   add up to 64 clashing elements to a hash:net,iface type of set while it
   is the documented feature of the set type. The patch fixes the condition
   and thus makes possible to add the elements while keeps preventing
   allocating huge memory, from Jozsef Kadlecsik. This has been broken
   for several releases.

3) Missing locking when updating the flow block list which might lead
   a reader to crash. This has been broken since the introduction of the
   flowtable hardware offload support.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: flowtable_offload: add missing locking
  netfilter: ipset: restore allowing 64 clashing elements in hash:net,iface
  netfilter: ipset: regression in ip_set_hash_ip.c
====================

Link: https://lore.kernel.org/r/20221122212814.63177-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Documentation: networking: Update generic_netlink_howto URL

The documentation refers to invalid web page under www.linuxfoundation.org
The patch refers to a working URL under wiki.linuxfoundation.org

Signed-off-by: Nir Levy <bhr166@gmail.com>
Link: https://lore.kernel.org/all/20221120220630.7443-1-bhr166@gmail.com/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

e100: Fix possible use after free in e100_xmit_prepare

In e100_xmit_prepare(), if we can't map the skb, then return -ENOMEM, so
e100_xmit_frame() will return NETDEV_TX_BUSY and the upper layer will
resend the skb. But the skb is already freed, which will cause UAF bug
when the upper layer resends the skb.

Remove the harmful free.

Fixes: 5e5d49422dfb ("e100: Release skb when DMA mapping is failed in e100_xmit_prepare")
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iavf: Fix error handling in iavf_init_module()

The iavf_init_module() won't destroy workqueue when pci_register_driver()
failed. Call destroy_workqueue() when pci_register_driver() failed to
prevent the resource leak.

Similar to the handling of u132_hcd_init in commit f276e002793c
("usb: u132-hcd: fix resource leak")

Fixes: 2803b16c10ea ("i40e/i40evf: Use private workqueue")
Signed-off-by: Yuan Can <yuancan@huawei.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

fm10k: Fix error handling in fm10k_init_module()

A problem about modprobe fm10k failed is triggered with the following log
given:

 Intel(R) Ethernet Switch Host Interface Driver
 Copyright(c) 2013 - 2019 Intel Corporation.
 debugfs: Directory 'fm10k' with parent '/' already present!

The reason is that fm10k_init_module() returns fm10k_register_pci_driver()
directly without checking its return value, if fm10k_register_pci_driver()
failed, it returns without removing debugfs and destroy workqueue,
resulting the debugfs of fm10k can never be created later and leaks the
workqueue.

 fm10k_init_module()
   alloc_workqueue()
   fm10k_dbg_init() # create debugfs
   fm10k_register_pci_driver()
     pci_register_driver()
       driver_register()
         bus_add_driver()
           priv = kzalloc(...) # OOM happened
   # return without remove debugfs and destroy workqueue

Fix by remove debugfs and destroy workqueue when
fm10k_register_pci_driver() returns error.

Fixes: 7461fd913afe ("fm10k: Add support for debugfs")
Fixes: b382bb1b3e2d ("fm10k: use separate workqueue for fm10k driver")
Signed-off-by: Yuan Can <yuancan@huawei.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

i40e: Fix error handling in i40e_init_module()

i40e_init_module() won't free the debugfs directory created by
i40e_dbg_init() when pci_register_driver() failed. Add fail path to
call i40e_dbg_exit() to remove the debugfs entries to prevent the bug.

i40e: Intel(R) Ethernet Connection XL710 Network Driver
i40e: Copyright (c) 2013 - 2019 Intel Corporation.
debugfs: Directory 'i40e' with parent '/' already present!

Fixes: 41c445ff0f48 ("i40e: main driver core")
Signed-off-by: Shang XiaoJing <shangxiaojing@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ixgbevf: Fix resource leak in ixgbevf_init_module()

ixgbevf_init_module() won't destroy the workqueue created by
create_singlethread_workqueue() when pci_register_driver() failed. Add
destroy_workqueue() in fail path to prevent the resource leak.

Similar to the handling of u132_hcd_init in commit f276e002793c
("usb: u132-hcd: fix resource leak")

Fixes: 40a13e2493c9 ("ixgbevf: Use a private workqueue to avoid certain possible hangs")
Signed-off-by: Shang XiaoJing <shangxiaojing@huawei.com>
Reviewed-by: Saeed Mahameed <saeed@kernel.org>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

net/cdc_ncm: Fix multicast RX support for CDC NCM devices with ZLP

ZLP for DisplayLink ethernet devices was enabled in 6.0:
266c0190aee3 ("net/cdc_ncm: Enable ZLP for DisplayLink ethernet devices").
The related driver_info should be the "same as cdc_ncm_info, but with
FLAG_SEND_ZLP". However, set_rx_mode that enables handling multicast
traffic was missing in the new cdc_ncm_zlp_info.

usbnet_cdc_update_filter rx mode was introduced in linux 5.9 with:
e10dcb1b6ba7 ("net: cdc_ncm: hook into set_rx_mode to admit multicast
traffic")

Without this hook, multicast, and then IPv6 SLAAC, is broken.

Fixes: 266c0190aee3 ("net/cdc_ncm: Enable ZLP for DisplayLink ethernet devices")
Signed-off-by: Santiago Ruano Rincón <santiago.ruano-rincon@imt-atlantique.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: usb: qmi_wwan: add u-blox 0x1342 composition

Add RmNet support for LARA-L6.

LARA-L6 module can be configured (by AT interface) in three different
USB modes:
* Default mode (Vendor ID: 0x1546 Product ID: 0x1341) with 4 serial
interfaces
* RmNet mode (Vendor ID: 0x1546 Product ID: 0x1342) with 4 serial
interfaces and 1 RmNet virtual network interface
* CDC-ECM mode (Vendor ID: 0x1546 Product ID: 0x1343) with 4 serial
interface and 1 CDC-ECM virtual network interface

In RmNet mode LARA-L6 exposes the following interfaces:
If 0: Diagnostic
If 1: AT parser
If 2: AT parser
If 3: AT parset/alternative functions
If 4: RMNET interface

Signed-off-by: Davide Tronchin <davide.tronchin.94@gmail.com>
Acked-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Don't sleep and disable BH under writer-side sk_callback_lock

When holding a reader-writer spin lock we cannot sleep. Calling
setup_udp_tunnel_sock() with write lock held violates this rule, because we
end up calling percpu_down_read(), which might sleep, as syzbot reports
[1]:

 __might_resched.cold+0x222/0x26b kernel/sched/core.c:9890
 percpu_down_read include/linux/percpu-rwsem.h:49 [inline]
 cpus_read_lock+0x1b/0x140 kernel/cpu.c:310
 static_key_slow_inc+0x12/0x20 kernel/jump_label.c:158
 udp_tunnel_encap_enable include/net/udp_tunnel.h:187 [inline]
 setup_udp_tunnel_sock+0x43d/0x550 net/ipv4/udp_tunnel_core.c:81
 l2tp_tunnel_register+0xc51/0x1210 net/l2tp/l2tp_core.c:1509
 pppol2tp_connect+0xcdc/0x1a10 net/l2tp/l2tp_ppp.c:723

Trim the writer-side critical section for sk_callback_lock down to the
minimum, so that it covers only operations on sk_user_data.

Also, when grabbing the sk_callback_lock, we always need to disable BH, as
Eric points out. Failing to do so leads to deadlocks because we acquire
sk_callback_lock in softirq context, which can get stuck waiting on us if:

1) it runs on the same CPU, or

       CPU0
       ----
  lock(clock-AF_INET6);
  <Interrupt>
    lock(clock-AF_INET6);

2) lock ordering leads to priority inversion

       CPU0                    CPU1
       ----                    ----
  lock(clock-AF_INET6);
                               local_irq_disable();
                               lock(&tcp_hashinfo.bhash[i].lock);
                               lock(clock-AF_INET6);
  <Interrupt>
    lock(&tcp_hashinfo.bhash[i].lock);

... as syzbot reports [2,3]. Use the _bh variants for write_(un)lock.

[1] https://lore.kernel.org/netdev/0000000000004e78ec05eda79749@google.com/
[2] https://lore.kernel.org/netdev/000000000000e38b6605eda76f98@google.com/
[3] https://lore.kernel.org/netdev/000000000000dfa31e05eda76f75@google.com/

v2:
- Check and set sk_user_data while holding sk_callback_lock for both
  L2TP encapsulation types (IP and UDP) (Tetsuo)

Cc: Tom Parkin <tparkin@katalix.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Fixes: b68777d54fac ("l2tp: Serialize access to sk_user_data with sk_callback_lock")
Reported-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot+703d9e154b3b58277261@syzkaller.appspotmail.com
Reported-by: syzbot+50680ced9e98a61f7698@syzkaller.appspotmail.com
Reported-by: syzbot+de987172bb74a381879b@syzkaller.appspotmail.com
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dm9051: Fix missing dev_kfree_skb() in dm9051_loop_rx()

The dm9051_loop_rx() returns without release skb when dm9051_stop_mrcmd()
returns error, free the skb to avoid this leak.

Fixes: 2dc95a4d30ed ("net: Add dm9051 driver")
Signed-off-by: Yuan Can <yuancan@huawei.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

arcnet: fix potential memory leak in com20020_probe()

In com20020_probe(), if com20020_config() fails, dev and info
will not be freed, which will lead to a memory leak.

This patch adds freeing dev and info after com20020_config()
fails to fix this bug.

Compile tested only.

Fixes: 15b99ac17295 ("[PATCH] pcmcia: add return value to _config() functions")
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'mlx5-fixes-2022-11-21' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2022-11-21

This series provides bug fixes to mlx5 driver.

* tag 'mlx5-fixes-2022-11-21' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5e: Fix possible race condition in macsec extended packet number update routine
  net/mlx5e: Fix MACsec update SecY
  net/mlx5e: Fix MACsec SA initialization routine
  net/mlx5e: Remove leftovers from old XSK queues enumeration
  net/mlx5e: Offload rule only when all encaps are valid
  net/mlx5e: Fix missing alignment in size of MTT/KLM entries
  net/mlx5: Fix sync reset event handler error flow
  net/mlx5: E-Switch, Set correctly vport destination
  net/mlx5: Lag, avoid lockdep warnings
  net/mlx5: Fix handling of entry refcount when command is not issued to FW
  net/mlx5: cmdif, Print info on any firmware cmd failure to tracepoint
  net/mlx5: SF: Fix probing active SFs during driver probe phase
  net/mlx5: Fix FW tracer timestamp calculation
  net/mlx5: Do not query pci info while pci disabled
====================

Link: https://lore.kernel.org/r/20221122022559.89459-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ipv4: Fix error return code in fib_table_insert()

In fib_table_insert(), if the alias was already inserted, but node not
exist, the error code should be set before return from error handling path.

Fixes: a6c76c17df02 ("ipv4: Notify route after insertion to the routing table")
Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Link: https://lore.kernel.org/r/20221120072838.2167047-1-william.xuanziyang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-ethernet-mtk_eth_soc-fix-memory-leak-in-error-path'

Yan Cangang says:

====================
net: ethernet: mtk_eth_soc: fix memory leak in error path
====================

Link: https://lore.kernel.org/r/20221120055259.224555-1-nalanzeyu@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ethernet: mtk_eth_soc: fix memory leak in error path

In mtk_ppe_init(), when dmam_alloc_coherent() or devm_kzalloc() failed,
the rhashtable ppe->l2_flows isn't destroyed. Fix it.

In mtk_probe(), when mtk_ppe_init() or mtk_eth_offload_init() or
register_netdev() failed, have the same problem. Fix it.

Fixes: 33fc42de3327 ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
Signed-off-by: Yan Cangang <nalanzeyu@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ethernet: mtk_eth_soc: fix resource leak in error path

In mtk_probe(), when mtk_ppe_init() or mtk_eth_offload_init() failed,
mtk_mdio_cleanup() isn't called. Fix it.

Fixes: ba37b7caf1ed ("net: ethernet: mtk_eth_soc: add support for initializing the PPE")
Fixes: 502e84e2382d ("net: ethernet: mtk_eth_soc: add flow offloading support")
Signed-off-by: Yan Cangang <nalanzeyu@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ethernet: mtk_eth_soc: fix potential memory leak in mtk_rx_alloc()

When fail to dma_map_single() in mtk_rx_alloc(), it returns directly.
But the memory allocated for local variable data is not freed, and
local variabel data has not been attached to ring->data[i] yet, so the
memory allocated for local variable data will not be freed outside
mtk_rx_alloc() too. Thus memory leak would occur in this scenario.

Add skb_free_frag(data) when dma_map_single() failed.

Fixes: 23233e577ef9 ("net: ethernet: mtk_eth_soc: rely on page_pool for single page buffers")
Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Acked-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Link: https://lore.kernel.org/r/20221120035405.1464341-1-william.xuanziyang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'dccp-tcp-fix-bhash2-issues-related-to-warn_on-in-inet_csk_get_port'

Kuniyuki Iwashima says:

====================
dccp/tcp: Fix bhash2 issues related to WARN_ON() in inet_csk_get_port().

syzkaller was hitting a WARN_ON() in inet_csk_get_port() in the 4th patch,
which was because we forgot to fix up bhash2 bucket when connect() for a
socket bound to a wildcard address fails in __inet_stream_connect().

There was a similar report [0], but its repro does not fire the WARN_ON() due
to inconsistent error handling.

When connect() for a socket bound to a wildcard address fails, saddr may or
may not be reset depending on where the failure happens.  When we fail in
__inet_stream_connect(), sk->sk_prot->disconnect() resets saddr.  OTOH, in
(dccp|tcp)_v[46]_connect(), if we fail after inet_hash6?_connect(), we
forget to reset saddr.

We fix this inconsistent error handling in the 1st patch, and then we'll
fix the bhash2 WARN_ON() issue.

Note that there is still an issue in that we reset saddr without checking
if there are conflicting sockets in bhash and bhash2, but this should be
another series.

See [1][2] for the previous discussion.

[0]: https://lore.kernel.org/netdev/0000000000003f33bc05dfaf44fe@google.com/
[1]: https://lore.kernel.org/netdev/20221029001249.86337-1-kuniyu@amazon.com/
[2]: https://lore.kernel.org/netdev/20221103172419.20977-1-kuniyu@amazon.com/
[3]: https://lore.kernel.org/netdev/20221118081906.053d5231@kernel.org/T/#m00aafedb29ff0b55d5e67aef0252ef1baaf4b6ee
====================

Link: https://lore.kernel.org/r/20221119014914.31792-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dccp/tcp: Fixup bhash2 bucket when connect() fails.

If a socket bound to a wildcard address fails to connect(), we
only reset saddr and keep the port.  Then, we have to fix up the
bhash2 bucket; otherwise, the bucket has an inconsistent address
in the list.

Also, listen() for such a socket will fire the WARN_ON() in
inet_csk_get_port(). [0]

Note that when a system runs out of memory, we give up fixing the
bucket and unlink sk from bhash and bhash2 by inet_put_port().

[0]:
WARNING: CPU: 0 PID: 207 at net/ipv4/inet_connection_sock.c:548 inet_csk_get_port (net/ipv4/inet_connection_sock.c:548 (discriminator 1))
Modules linked in:
CPU: 0 PID: 207 Comm: bhash2_prev_rep Not tainted 6.1.0-rc3-00799-gc8421681c845 #63
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.amzn2022.0.1 04/01/2014
RIP: 0010:inet_csk_get_port (net/ipv4/inet_connection_sock.c:548 (discriminator 1))
Code: 74 a7 eb 93 48 8b 54 24 18 0f b7 cb 4c 89 e6 4c 89 ff e8 48 b2 ff ff 49 8b 87 18 04 00 00 e9 32 ff ff ff 0f 0b e9 34 ff ff ff <0f> 0b e9 42 ff ff ff 41 8b 7f 50 41 8b 4f 54 89 fe 81 f6 00 00 ff
RSP: 0018:ffffc900003d7e50 EFLAGS: 00010202
RAX: ffff8881047fb500 RBX: 0000000000004e20 RCX: 0000000000000000
RDX: 000000000000000a RSI: 00000000fffffe00 RDI: 00000000ffffffff
RBP: ffffffff8324dc00 R08: 0000000000000001 R09: 0000000000000001
R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000000
R13: 0000000000000001 R14: 0000000000004e20 R15: ffff8881054e1280
FS:  00007f8ac04dc740(0000) GS:ffff88842fc00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000020001540 CR3: 00000001055fa003 CR4: 0000000000770ef0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
 <TASK>
 inet_csk_listen_start (net/ipv4/inet_connection_sock.c:1205)
 inet_listen (net/ipv4/af_inet.c:228)
 __sys_listen (net/socket.c:1810)
 __x64_sys_listen (net/socket.c:1819 net/socket.c:1817 net/socket.c:1817)
 do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80)
 entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120)
RIP: 0033:0x7f8ac051de5d
Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 93 af 1b 00 f7 d8 64 89 01 48
RSP: 002b:00007ffc1c177248 EFLAGS: 00000206 ORIG_RAX: 0000000000000032
RAX: ffffffffffffffda RBX: 0000000020001550 RCX: 00007f8ac051de5d
RDX: ffffffffffffff80 RSI: 0000000000000000 RDI: 0000000000000004
RBP: 00007ffc1c177270 R08: 0000000000000018 R09: 0000000000000007
R10: 0000000020001540 R11: 0000000000000206 R12: 00007ffc1c177388
R13: 0000000000401169 R14: 0000000000403e18 R15: 00007f8ac0723000
 </TASK>

Fixes: 28044fc1d495 ("net: Add a bhash2 table hashed by port and address")
Reported-by: syzbot <syzkaller@googlegroups.com>
Reported-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Joanne Koong <joannelkoong@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dccp/tcp: Update saddr under bhash's lock.

When we call connect() for a socket bound to a wildcard address, we update
saddr locklessly.  However, it could result in a data race; another thread
iterating over bhash might see a corrupted address.

Let's update saddr under the bhash bucket's lock.

Fixes: 3df80d9320bc ("[DCCP]: Introduce DCCPv6")
Fixes: 7c657876b63c ("[DCCP]: Initial implementation")
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Joanne Koong <joannelkoong@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dccp/tcp: Remove NULL check for prev_saddr in inet_bhash2_update_saddr().

When we call inet_bhash2_update_saddr(), prev_saddr is always non-NULL.
Let's remove the unnecessary test.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Joanne Koong <joannelkoong@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dccp/tcp: Reset saddr on failure after inet6?_hash_connect().

When connect() is called on a socket bound to the wildcard address,
we change the socket's saddr to a local address.  If the socket
fails to connect() to the destination, we have to reset the saddr.

However, when an error occurs after inet_hash6?_connect() in
(dccp|tcp)_v[46]_conect(), we forget to reset saddr and leave
the socket bound to the address.

From the user's point of view, whether saddr is reset or not varies
with errno.  Let's fix this inconsistent behaviour.

Note that after this patch, the repro [0] will trigger the WARN_ON()
in inet_csk_get_port() again, but this patch is not buggy and rather
fixes a bug papering over the bhash2's bug for which we need another
fix.

For the record, the repro causes -EADDRNOTAVAIL in inet_hash6_connect()
by this sequence:

  s1 = socket()
  s1.setsockopt(SOL_SOCKET, SO_REUSEADDR, 1)
  s1.bind(('127.0.0.1', 10000))
  s1.sendto(b'hello', MSG_FASTOPEN, (('127.0.0.1', 10000)))
  # or s1.connect(('127.0.0.1', 10000))

  s2 = socket()
  s2.setsockopt(SOL_SOCKET, SO_REUSEADDR, 1)
  s2.bind(('0.0.0.0', 10000))
  s2.connect(('127.0.0.1', 10000))  # -EADDRNOTAVAIL

  s2.listen(32)  # WARN_ON(inet_csk(sk)->icsk_bind2_hash != tb2);

[0]: https://syzkaller.appspot.com/bug?extid=015d756bbd1f8b5c8f09

Fixes: 3df80d9320bc ("[DCCP]: Introduce DCCPv6")
Fixes: 7c657876b63c ("[DCCP]: Initial implementation")
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Joanne Koong <joannelkoong@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netfilter: flowtable_offload: add missing locking

nf_flow_table_block_setup and the driver TC_SETUP_FT call can modify the flow
block cb list while they are being traversed elsewhere, causing a crash.
Add a write lock around the calls to protect readers

Fixes: c29f74e0df7a ("netfilter: nf_flow_table: hardware offload support")
Reported-by: Chad Monroe <chad.monroe@smartrg.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: ipset: restore allowing 64 clashing elements in hash:net,iface

The commit 510841da1fcc ("netfilter: ipset: enforce documented limit to
prevent allocating huge memory") was too strict and prevented to add up to
64 clashing elements to a hash:net,iface type of set. This patch fixes the
issue and now the type behaves as documented.

Fixes: 510841da1fcc ("netfilter: ipset: enforce documented limit to prevent allocating huge memory")
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

tsnep: Fix rotten packets

If PTP synchronisation is done every second, then sporadic the interval
is higher than one second:

ptp4l[696.582]: master offset        -17 s2 freq   -1891 path delay 573
ptp4l[697.582]: master offset        -22 s2 freq   -1901 path delay 573
ptp4l[699.368]: master offset         -1 s2 freq   -1887 path delay 573
      ^^^^^^^ Should be 698.582!

This problem is caused by rotten packets, which are received after
polling but before interrupts are enabled again. This can be fixed by
checking for pending work and rescheduling if necessary after interrupts
has been enabled again.

Fixes: 403f69bbdbad ("tsnep: Add TSN endpoint Ethernet MAC driver")
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Link: https://lore.kernel.org/r/20221119211825.81805-1-gerhard@engleder-embedded.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

octeontx2-pf: Remove duplicate MACSEC setting

Commit 4581dd480c9e ("net: octeontx2-pf: mcs: consider MACSEC setting")
has already added "depends on MACSEC || !MACSEC", so remove it.

Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20221119133616.3583538-1-zhengbin13@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

bnx2x: fix pci device refcount leak in bnx2x_vf_is_pcie_pending()

As comment of pci_get_domain_bus_and_slot() says, it returns
a pci device with refcount increment, when finish using it,
the caller must decrement the reference count by calling
pci_dev_put(). Call pci_dev_put() before returning from
bnx2x_vf_is_pcie_pending() to avoid refcount leak.

Fixes: b56e9670ffa4 ("bnx2x: Prepare device and initialize VF database")
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20221119070202.1407648-1-yangyingliang@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

NFC: nci: fix memory leak in nci_rx_data_packet()

Syzbot reported a memory leak about skb:

unreferenced object 0xffff88810e144e00 (size 240):
  comm "syz-executor284", pid 3701, jiffies 4294952403 (age 12.620s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff83ab79a9>] __alloc_skb+0x1f9/0x270 net/core/skbuff.c:497
    [<ffffffff82a5cf64>] alloc_skb include/linux/skbuff.h:1267 [inline]
    [<ffffffff82a5cf64>] virtual_ncidev_write+0x24/0xe0 drivers/nfc/virtual_ncidev.c:116
    [<ffffffff815f6503>] do_loop_readv_writev fs/read_write.c:759 [inline]
    [<ffffffff815f6503>] do_loop_readv_writev fs/read_write.c:743 [inline]
    [<ffffffff815f6503>] do_iter_write+0x253/0x300 fs/read_write.c:863
    [<ffffffff815f66ed>] vfs_writev+0xdd/0x240 fs/read_write.c:934
    [<ffffffff815f68f6>] do_writev+0xa6/0x1c0 fs/read_write.c:977
    [<ffffffff848802d5>] do_syscall_x64 arch/x86/entry/common.c:50 [inline]
    [<ffffffff848802d5>] do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
    [<ffffffff84a00087>] entry_SYSCALL_64_after_hwframe+0x63/0xcd

In nci_rx_data_packet(), if we don't get a valid conn_info, we will return
directly but forget to release the skb.

Reported-by: syzbot+cdb9a427d1bc08815104@syzkaller.appspotmail.com
Fixes: 4aeee6871e8c ("NFC: nci: Add dynamic logical connections support")
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Link: https://lore.kernel.org/r/20221118082419.239475-1-liushixin2@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sched: allow act_ct to be built without NF_NAT

In commit f11fe1dae1c4 ("net/sched: Make NET_ACT_CT depends on NF_NAT"),
it fixed the build failure when NF_NAT is m and NET_ACT_CT is y by
adding depends on NF_NAT for NET_ACT_CT. However, it would also cause
NET_ACT_CT cannot be built without NF_NAT, which is not expected. This
patch fixes it by changing to use "(!NF_NAT || NF_NAT)" as the depend.

Fixes: f11fe1dae1c4 ("net/sched: Make NET_ACT_CT depends on NF_NAT")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/b6386f28d1ba34721795fb776a91cbdabb203447.1668807183.git.lucien.xin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sparx5: fix error handling in sparx5_port_open()

If phylink_of_phy_connect() fails, the port should be disabled.
If sparx5_serdes_set()/phy_power_on() fails, the port should be
disabled and the phylink should be stopped and disconnected.

Fixes: 946e7fd5053a ("net: sparx5: add port module support")
Fixes: f3cad2611a77 ("net: sparx5: add hostmode with phylink support")
Signed-off-by: Liu Jian <liujian56@huawei.com>
Tested-by: Bjarni Jonasson <bjarni.jonasson@microchip.com>
Reviewed-by: Steen Hegelund <steen.hegelund@microchip.com>
Link: https://lore.kernel.org/r/20221117125918.203997-1-liujian56@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

sfc: fix potential memleak in __ef100_hard_start_xmit()

The __ef100_hard_start_xmit() returns NETDEV_TX_OK without freeing skb
in error handling case, add dev_kfree_skb_any() to fix it.

Fixes: 51b35a454efd ("sfc: skeleton EF100 PF driver")
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/1668671409-10909-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: wwan: iosm: use ACPI_FREE() but not kfree() in ipc_pcie_read_bios_cfg()

acpi_evaluate_dsm() should be coupled with ACPI_FREE() to free the ACPI
memory, because we need to track the allocation of acpi_object when
ACPI_DBG_TRACK_ALLOCATIONS enabled, so use ACPI_FREE() instead of kfree().

Fixes: d38a648d2d6c ("net: wwan: iosm: fix memory leak in ipc_pcie_read_bios_cfg")
Signed-off-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
Link: https://lore.kernel.org/r/20221118062447.2324881-1-bobo.shaobowang@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

xfrm: Fix ignored return value in xfrm6_init()

When IPv6 module initializing in xfrm6_init(), register_pernet_subsys()
is possible to fail but its return value is ignored.

If IPv6 initialization fails later and xfrm6_fini() is called,
removing uninitialized list in xfrm6_net_ops will cause null-ptr-deref:

KASAN: null-ptr-deref in range [0x0000000000000008-0x000000000000000f]
CPU: 1 PID: 330 Comm: insmod
RIP: 0010:unregister_pernet_operations+0xc9/0x450
Call Trace:
 <TASK>
 unregister_pernet_subsys+0x31/0x3e
 xfrm6_fini+0x16/0x30 [ipv6]
 ip6_route_init+0xcd/0x128 [ipv6]
 inet6_init+0x29c/0x602 [ipv6]
 ...

Fix it by catching the error return value of register_pernet_subsys().

Fixes: 8d068875caca ("xfrm: make gc_thresh configurable in all namespaces")
Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>

xfrm: Fix oops in __xfrm_state_delete()

Kernel 5.14 added a new "byseq" index to speed
up xfrm_state lookups by sequence number in commit
fe9f1d8779cb ("xfrm: add state hashtable keyed by seq")

While the patch was thorough, the function pfkey_send_new_mapping()
in net/af_key.c also modifies x->km.seq and never added
the current xfrm_state to the "byseq" index.

This leads to the following kernel Ooops:
    BUG: kernel NULL pointer dereference, address: 0000000000000000
    ..
    RIP: 0010:__xfrm_state_delete+0xc9/0x1c0
    ..
    Call Trace:
    <TASK>
    xfrm_state_delete+0x1e/0x40
    xfrm_del_sa+0xb0/0x110 [xfrm_user]
    xfrm_user_rcv_msg+0x12d/0x270 [xfrm_user]
    ? remove_entity_load_avg+0x8a/0xa0
    ? copy_to_user_state_extra+0x580/0x580 [xfrm_user]
    netlink_rcv_skb+0x51/0x100
    xfrm_netlink_rcv+0x30/0x50 [xfrm_user]
    netlink_unicast+0x1a6/0x270
    netlink_sendmsg+0x22a/0x480
    __sys_sendto+0x1a6/0x1c0
    ? __audit_syscall_entry+0xd8/0x130
    ? __audit_syscall_exit+0x249/0x2b0
    __x64_sys_sendto+0x23/0x30
    do_syscall_64+0x3a/0x90
    entry_SYSCALL_64_after_hwframe+0x61/0xcb

Exact location of the crash in __xfrm_state_delete():
    if (x->km.seq)
        hlist_del_rcu(&x->byseq);

The hlist_node "byseq" was never populated.

The bug only triggers if a new NAT traversal mapping (changed IP or port)
is detected in esp_input_done2() / esp6_input_done2(), which in turn
indirectly calls pfkey_send_new_mapping() *if* the kernel is compiled
with CONFIG_NET_KEY and "af_key" is active.

The PF_KEYv2 message SADB_X_NAT_T_NEW_MAPPING is not part of RFC 2367.
Various implementations have been examined how they handle
the "sadb_msg_seq" header field:

- racoon (Android): does not process SADB_X_NAT_T_NEW_MAPPING
- strongswan: does not care about sadb_msg_seq
- openswan: does not care about sadb_msg_seq

There is no standard how PF_KEYv2 sadb_msg_seq should be populated
for SADB_X_NAT_T_NEW_MAPPING and it's not used in popular
implementations either. Herbert Xu suggested we should just
use the current km.seq value as is. This fixes the root cause
of the oops since we no longer modify km.seq itself.

The update of "km.seq" looks like a copy'n'paste error
from pfkey_send_acquire(). SADB_ACQUIRE must indeed assign a unique km.seq
number according to RFC 2367. It has been verified that code paths
involving pfkey_send_acquire() don't cause the same Oops.

PF_KEYv2 SADB_X_NAT_T_NEW_MAPPING support was originally added here:
    https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git

    commit cbc3488685b20e7b2a98ad387a1a816aada569d8
    Author:     Derek Atkins <derek@ihtfp.com>
    AuthorDate: Wed Apr 2 13:21:02 2003 -0800

        [IPSEC]: Implement UDP Encapsulation framework.

        In particular, implement ESPinUDP encapsulation for IPsec
        Nat Traversal.

A note on triggering the bug: I was not able to trigger it using VMs.
There is one VPN using a high latency link on our production VPN server
that triggered it like once a day though.

Link: https://github.com/strongswan/strongswan/issues/992
Link: https://lore.kernel.org/netdev/00959f33ee52c4b3b0084d42c430418e502db554.1652340703.git.antony.antony@secunet.com/T/
Link: https://lore.kernel.org/netdev/20221027142455.3975224-1-chenzhihao@meizu.com/T/
Fixes: fe9f1d8779cb ("xfrm: add state hashtable keyed by seq")
Reported-by: Roth Mark <rothm@mail.com>
Reported-by: Zhihao Chen <chenzhihao@meizu.com>
Tested-by: Roth Mark <rothm@mail.com>
Signed-off-by: Thomas Jarosch <thomas.jarosch@intra2net.com>
Acked-by: Antony Antony <antony.antony@secunet.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>

ice: fix handling of burst Tx timestamps

Commit 1229b33973c7 ("ice: Add low latency Tx timestamp read") refactored
PTP timestamping logic to use a threaded IRQ instead of a separate kthread.

This implementation introduced ice_misc_intr_thread_fn and redefined the
ice_ptp_process_ts function interface to return a value of whether or not
the timestamp processing was complete.

ice_misc_intr_thread_fn would take the return value from ice_ptp_process_ts
and convert it into either IRQ_HANDLED if there were no more timestamps to
be processed, or IRQ_WAKE_THREAD if the thread should continue processing.

This is not correct, as the kernel does not re-schedule threaded IRQ
functions automatically. IRQ_WAKE_THREAD can only be used by the main IRQ
function.

This results in the ice_ptp_process_ts function (and in turn the
ice_ptp_tx_tstamp function) from only being called exactly once per
interrupt.

If an application sends a burst of Tx timestamps without waiting for a
response, the interrupt will trigger for the first timestamp. However,
later timestamps may not have arrived yet. This can result in dropped or
discarded timestamps. Worse, on E822 hardware this results in the interrupt
logic getting stuck such that no future interrupts will be triggered. The
result is complete loss of Tx timestamp functionality.

Fix this by modifying the ice_misc_intr_thread_fn to perform its own
polling of the ice_ptp_process_ts function. We sleep for a few microseconds
between attempts to avoid wasting significant CPU time. The value was
chosen to allow time for the Tx timestamps to complete without wasting so
much time that we overrun application wait budgets in the worst case.

The ice_ptp_process_ts function also currently returns false in the event
that the Tx tracker is not initialized. This would result in the threaded
IRQ handler never exiting if it gets started while the tracker is not
initialized.

Fix the function to appropriately return true when the tracker is not
initialized.

Note that this will not reproduce with default ptp4l behavior, as the
program always synchronously waits for a timestamp response before sending
another timestamp request.

Reported-by: Siddaraju DH <siddaraju.dh@intel.com>
Fixes: 1229b33973c7 ("ice: Add low latency Tx timestamp read")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20221118222729.1565317-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tipc: check skb_linearize() return value in tipc_disc_rcv()

If skb_linearize() fails in tipc_disc_rcv(), we need to free the skb instead of
handle it.

Fixes: 25b0b9c4e835 ("tipc: handle collisions of 32-bit node address hash values")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Link: https://lore.kernel.org/r/20221119072832.7896-1-yuehaibing@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch '40GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-11-18 (iavf)

Ivan Vecera resolves issues related to reset by adding back call to
netif_tx_stop_all_queues() and adding calls to dev_close() to ensure
device is properly closed during reset.

Stefan Assmann removes waiting for setting of MAC address as this breaks
ARP.

Slawomir adds setting of __IAVF_IN_REMOVE_TASK bit to prevent deadlock
between remove and shutdown.

* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  iavf: Fix race condition between iavf_shutdown and iavf_remove
  iavf: remove INITIAL_MAC_SET to allow gARP to work properly
  iavf: Do not restart Tx queues after reset task failure
  iavf: Fix a crash during reset task
====================

Link: https://lore.kernel.org/r/20221118222439.1565245-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'tipc-fix-two-race-issues-in-tipc_conn_alloc'

Xin Long says:

====================
tipc: fix two race issues in tipc_conn_alloc

The race exists beteen tipc_topsrv_accept() and tipc_conn_close(),
one is allocating the con while the other is freeing it and there
is no proper lock protecting it. Therefore, a null-pointer-defer
and a use-after-free may be triggered, see details on each patch.
====================

Link: https://lore.kernel.org/r/cover.1668807842.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tipc: add an extra conn_get in tipc_conn_alloc

One extra conn_get() is needed in tipc_conn_alloc(), as after
tipc_conn_alloc() is called, tipc_conn_close() may free this
con before deferencing it in tipc_topsrv_accept():

   tipc_conn_alloc();
   newsk = newsock->sk;
                                 <---- tipc_conn_close();
   write_lock_bh(&sk->sk_callback_lock);
   newsk->sk_data_ready = tipc_conn_data_ready;

Then an uaf issue can be triggered:

  BUG: KASAN: use-after-free in tipc_topsrv_accept+0x1e7/0x370 [tipc]
  Call Trace:
   <TASK>
   dump_stack_lvl+0x33/0x46
   print_report+0x178/0x4b0
   kasan_report+0x8c/0x100
   kasan_check_range+0x179/0x1e0
   tipc_topsrv_accept+0x1e7/0x370 [tipc]
   process_one_work+0x6a3/0x1030
   worker_thread+0x8a/0xdf0

This patch fixes it by holding it in tipc_conn_alloc(), then after
all accessing in tipc_topsrv_accept() releasing it. Note when does
this in tipc_topsrv_kern_subscr(), as tipc_conn_rcv_sub() returns
0 or -1 only, we don't need to check for "> 0".

Fixes: c5fa7b3cf3cb ("tipc: introduce new TIPC server infrastructure")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tipc: set con sock in tipc_conn_alloc

A crash was reported by Wei Chen:

  BUG: kernel NULL pointer dereference, address: 0000000000000018
  RIP: 0010:tipc_conn_close+0x12/0x100
  Call Trace:
   tipc_topsrv_exit_net+0x139/0x320
   ops_exit_list.isra.9+0x49/0x80
   cleanup_net+0x31a/0x540
   process_one_work+0x3fa/0x9f0
   worker_thread+0x42/0x5c0

It was caused by !con->sock in tipc_conn_close(). In tipc_topsrv_accept(),
con is allocated in conn_idr then its sock is set:

  con = tipc_conn_alloc();
  ...                    <----[1]
  con->sock = newsock;

If tipc_conn_close() is called in anytime of [1], the null-pointer-def
is triggered by con->sock->sk due to con->sock is not yet set.

This patch fixes it by moving the con->sock setting to tipc_conn_alloc()
under s->idr_lock. So that con->sock can never be NULL when getting the
con from s->conn_idr. It will be also safer to move con->server and flag
CF_CONNECTED setting under s->idr_lock, as they should all be set before
tipc_conn_alloc() is called.

Fixes: c5fa7b3cf3cb ("tipc: introduce new TIPC server infrastructure")
Reported-by: Wei Chen <harperchen1110@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: phy: at803x: fix error return code in at803x_probe()

Fix to return a negative error code from the ccr read error handling
case instead of 0, as done elsewhere in this function.

Fixes: 3265f4218878 ("net: phy: at803x: add fiber support")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20221118103635.254256-1-weiyongjun@huaweicloud.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5e: Fix possible race condition in macsec extended packet number update routine

Currenty extended packet number (EPN) update routine is accessing
macsec object without holding the general macsec lock hence facing
a possible race condition when an EPN update occurs while updating
or deleting the SA.
Fix by holding the general macsec lock before accessing the object.

Fixes: 4411a6c0abd3 ("net/mlx5e: Support MACsec offload extended packet number (EPN)")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix MACsec update SecY

Currently updating SecY destroys and re-creates RX SA objects,
the re-created RX SA objects are not identical to the destroyed
objects and it disagree on the encryption enabled property which
holds the value false after recreation, this value is not
supported with offload which leads to no traffic after an update.
Fix by recreating an identical objects.

Fixes: 5a39816a75e5 ("net/mlx5e: Add MACsec offload SecY support")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix MACsec SA initialization routine

Currently as part of MACsec SA initialization routine
extended packet number (EPN) object attribute is always
being set without checking if EPN is actually enabled,
the above could lead to a NULL dereference.
Fix by adding such a check.

Fixes: 4411a6c0abd3 ("net/mlx5e: Support MACsec offload extended packet number (EPN)")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Remove leftovers from old XSK queues enumeration

Before the cited commit, for N channels, a dedicated set of N queues was
created to support XSK, in indices [N, 2N-1], doubling the number of
queues.

In addition, changing the number of channels was prohibited, as it would
shift the indices.

Remove these two leftovers, as we moved XSK to a new queueing scheme,
starting from index 0.

Fixes: 3db4c85cde7a ("net/mlx5e: xsk: Use queue indices starting from 0 for XSK queues")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Offload rule only when all encaps are valid

The cited commit adds a for loop to support multiple encapsulations.
But it only checks if the last encap is valid.

Fix it by setting slow path flag when one of the encap is invalid.

Fixes: f493f15534ec ("net/mlx5e: Move flow attr reformat action bit to per dest flags")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: Fix missing alignment in size of MTT/KLM entries

In the cited patch, an alignment required by the HW spec was mistakenly
dropped. Bring it back to fix error completions like the below:

mlx5_core 0000:00:08.0 eth2: Error cqe on cqn 0x40b, ci 0x0, qn 0x104f, opcode 0xd, syndrome 0x2, vendor syndrome 0x68
00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000030: 00 00 00 00 86 00 68 02 25 00 10 4f 00 00 bb d2
WQE DUMP: WQ size 1024 WQ cur size 0, WQE index 0x0, len: 192
00000000: 00 00 00 25 00 10 4f 0c 00 00 00 00 00 18 2e 00
00000010: 90 00 00 00 00 02 00 00 00 00 00 00 20 00 00 00
00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000080: 08 00 00 00 48 6a 00 02 08 00 00 00 0e 10 00 02
00000090: 08 00 00 00 0c db 00 02 08 00 00 00 0e 82 00 02
000000a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
000000b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Fixes: 9f123f740428 ("net/mlx5e: Improve MTT/KSM alignment")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Fix sync reset event handler error flow

When sync reset now event handling fails on mlx5_pci_link_toggle() then
no reset was done. However, since mlx5_cmd_fast_teardown_hca() was
already done, the firmware function is closed and the driver is left
without firmware functionality.

Fix it by setting device error state and reopen the firmware resources.
Reopening is done by the thread that was called for devlink reload
fw_activate as it already holds the devlink lock.

Fixes: 5ec697446f46 ("net/mlx5: Add support for devlink reload action fw activate")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: E-Switch, Set correctly vport destination

The cited commit moved from using reformat_id integer to packet_reformat
pointer which introduced the possibility to null pointer dereference.
When setting packet reformat flag and pkt_reformat pointer must
exists so checking MLX5_ESW_DEST_ENCAP is not enough, we need
to make sure the pkt_reformat is valid and check for MLX5_ESW_DEST_ENCAP_VALID.
If the dest encap valid flag does not exists then pkt_reformat can be
either invalid address or null.
Also, to make sure we don't try to access invalid pkt_reformat set it to
null when invalidated and invalidate it before calling add flow code as
its logically more correct and to be safe.

Fixes: 2b688ea5efde ("net/mlx5: Add flow steering actions to fs_cmd shim layer")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Chris Mi <cmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Lag, avoid lockdep warnings

ldev->lock is used to serialize lag change operations. Since multiport
eswtich functionality was added, we now change the mode dynamically.
However, acquiring ldev->lock is not allowed as it could possibly lead
to a deadlock as reported by the lockdep mechanism.

[  836.154963] WARNING: possible circular locking dependency detected
[  836.155850] 5.19.0-rc5_net_56b7df2 #1 Not tainted
[  836.156549] ------------------------------------------------------
[  836.157418] handler1/12198 is trying to acquire lock:
[  836.158178] ffff888187d52b58 (&ldev->lock){+.+.}-{3:3}, at: mlx5_lag_do_mirred+0x3b/0x70 [mlx5_core]
[  836.159575]
[  836.159575] but task is already holding lock:
[  836.160474] ffff8881d4de2930 (&block->cb_lock){++++}-{3:3}, at: tc_setup_cb_add+0x5b/0x200
[  836.161669] which lock already depends on the new lock.
[  836.162905]
[  836.162905] the existing dependency chain (in reverse order) is:
[  836.164008] -> #3 (&block->cb_lock){++++}-{3:3}:
[  836.164946]        down_write+0x25/0x60
[  836.165548]        tcf_block_get_ext+0x1c6/0x5d0
[  836.166253]        ingress_init+0x74/0xa0 [sch_ingress]
[  836.167028]        qdisc_create.constprop.0+0x130/0x5e0
[  836.167805]        tc_modify_qdisc+0x481/0x9f0
[  836.168490]        rtnetlink_rcv_msg+0x16e/0x5a0
[  836.169189]        netlink_rcv_skb+0x4e/0xf0
[  836.169861]        netlink_unicast+0x190/0x250
[  836.170543]        netlink_sendmsg+0x243/0x4b0
[  836.171226]        sock_sendmsg+0x33/0x40
[  836.171860]        ____sys_sendmsg+0x1d1/0x1f0
[  836.172535]        ___sys_sendmsg+0xab/0xf0
[  836.173183]        __sys_sendmsg+0x51/0x90
[  836.173836]        do_syscall_64+0x3d/0x90
[  836.174471]        entry_SYSCALL_64_after_hwframe+0x46/0xb0
[  836.175282]

[  836.175282] -> #2 (rtnl_mutex){+.+.}-{3:3}:
[  836.176190]        __mutex_lock+0x6b/0xf80
[  836.176830]        register_netdevice_notifier+0x21/0x120
[  836.177631]        rtnetlink_init+0x2d/0x1e9
[  836.178289]        netlink_proto_init+0x163/0x179
[  836.178994]        do_one_initcall+0x63/0x300
[  836.179672]        kernel_init_freeable+0x2cb/0x31b
[  836.180403]        kernel_init+0x17/0x140
[  836.181035]        ret_from_fork+0x1f/0x30

 [  836.181687] -> #1 (pernet_ops_rwsem){+.+.}-{3:3}:
[  836.182628]        down_write+0x25/0x60
[  836.183235]        unregister_netdevice_notifier+0x1c/0xb0
[  836.184029]        mlx5_ib_roce_cleanup+0x94/0x120 [mlx5_ib]
[  836.184855]        __mlx5_ib_remove+0x35/0x60 [mlx5_ib]
[  836.185637]        mlx5_eswitch_unregister_vport_reps+0x22f/0x440 [mlx5_core]
[  836.186698]        auxiliary_bus_remove+0x18/0x30
[  836.187409]        device_release_driver_internal+0x1f6/0x270
[  836.188253]        bus_remove_device+0xef/0x160
[  836.188939]        device_del+0x18b/0x3f0
[  836.189562]        mlx5_rescan_drivers_locked+0xd6/0x2d0 [mlx5_core]
[  836.190516]        mlx5_lag_remove_devices+0x69/0xe0 [mlx5_core]
[  836.191414]        mlx5_do_bond_work+0x441/0x620 [mlx5_core]
[  836.192278]        process_one_work+0x25c/0x590
[  836.192963]        worker_thread+0x4f/0x3d0
[  836.193609]        kthread+0xcb/0xf0
[  836.194189]        ret_from_fork+0x1f/0x30

[  836.194826] -> #0 (&ldev->lock){+.+.}-{3:3}:
[  836.195734]        __lock_acquire+0x15b8/0x2a10
[  836.196426]        lock_acquire+0xce/0x2d0
[  836.197057]        __mutex_lock+0x6b/0xf80
[  836.197708]        mlx5_lag_do_mirred+0x3b/0x70 [mlx5_core]
[  836.198575]        tc_act_parse_mirred+0x25b/0x800 [mlx5_core]
[  836.199467]        parse_tc_actions+0x168/0x5a0 [mlx5_core]
[  836.200340]        __mlx5e_add_fdb_flow+0x263/0x480 [mlx5_core]
[  836.201241]        mlx5e_configure_flower+0x8a0/0x1820 [mlx5_core]
[  836.202187]        tc_setup_cb_add+0xd7/0x200
[  836.202856]        fl_hw_replace_filter+0x14c/0x1f0 [cls_flower]
[  836.203739]        fl_change+0xbbe/0x1730 [cls_flower]
[  836.204501]        tc_new_tfilter+0x407/0xd90
[  836.205168]        rtnetlink_rcv_msg+0x406/0x5a0
[  836.205877]        netlink_rcv_skb+0x4e/0xf0
[  836.206535]        netlink_unicast+0x190/0x250
[  836.207217]        netlink_sendmsg+0x243/0x4b0
[  836.207915]        sock_sendmsg+0x33/0x40
[  836.208538]        ____sys_sendmsg+0x1d1/0x1f0
[  836.209219]        ___sys_sendmsg+0xab/0xf0
[  836.209878]        __sys_sendmsg+0x51/0x90
[  836.210510]        do_syscall_64+0x3d/0x90
[  836.211137]        entry_SYSCALL_64_after_hwframe+0x46/0xb0

[  836.211954] other info that might help us debug this:
[  836.213174] Chain exists of:
[  836.213174]   &ldev->lock --> rtnl_mutex --> &block->cb_lock
   836.214650]  Possible unsafe locking scenario:
[  836.214650]
[  836.215574]        CPU0                    CPU1
[  836.216255]        ----                    ----
[  836.216943]   lock(&block->cb_lock);
[  836.217518]                                lock(rtnl_mutex);
[  836.218348]                                lock(&block->cb_lock);
[  836.219212]   lock(&ldev->lock);
[  836.219758]
[  836.219758]  *** DEADLOCK ***
[  836.219758]
 [  836.220747] 2 locks held by handler1/12198:
[  836.221390]  #0: ffff8881d4de2930 (&block->cb_lock){++++}-{3:3}, at: tc_setup_cb_add+0x5b/0x200
[  836.222646]  #1: ffff88810c9a92c0 (&esw->mode_lock){++++}-{3:3}, at: mlx5_esw_hold+0x39/0x50 [mlx5_core]

[  836.224063] stack backtrace:
[  836.224799] CPU: 6 PID: 12198 Comm: handler1 Not tainted 5.19.0-rc5_net_56b7df2 #1
[  836.225923] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
[  836.227476] Call Trace:
[  836.227929]  <TASK>
[  836.228332]  dump_stack_lvl+0x57/0x7d
[  836.228924]  check_noncircular+0x104/0x120
[  836.229562]  __lock_acquire+0x15b8/0x2a10
[  836.230201]  lock_acquire+0xce/0x2d0
[  836.230776]  ? mlx5_lag_do_mirred+0x3b/0x70 [mlx5_core]
[  836.231614]  ? find_held_lock+0x2b/0x80
[  836.232221]  __mutex_lock+0x6b/0xf80
[  836.232799]  ? mlx5_lag_do_mirred+0x3b/0x70 [mlx5_core]
[  836.233636]  ? mlx5_lag_do_mirred+0x3b/0x70 [mlx5_core]
[  836.234451]  ? xa_load+0xc3/0x190
[  836.234995]  mlx5_lag_do_mirred+0x3b/0x70 [mlx5_core]
[  836.235803]  tc_act_parse_mirred+0x25b/0x800 [mlx5_core]
[  836.236636]  ? tc_act_can_offload_mirred+0x135/0x210 [mlx5_core]
[  836.237550]  parse_tc_actions+0x168/0x5a0 [mlx5_core]
[  836.238364]  __mlx5e_add_fdb_flow+0x263/0x480 [mlx5_core]
[  836.239202]  mlx5e_configure_flower+0x8a0/0x1820 [mlx5_core]
[  836.240076]  ? lock_acquire+0xce/0x2d0
[  836.240668]  ? tc_setup_cb_add+0x5b/0x200
[  836.241294]  tc_setup_cb_add+0xd7/0x200
[  836.241917]  fl_hw_replace_filter+0x14c/0x1f0 [cls_flower]
[  836.242709]  fl_change+0xbbe/0x1730 [cls_flower]
[  836.243408]  tc_new_tfilter+0x407/0xd90
[  836.244043]  ? tc_del_tfilter+0x880/0x880
[  836.244672]  rtnetlink_rcv_msg+0x406/0x5a0
[  836.245310]  ? netlink_deliver_tap+0x7a/0x4b0
[  836.245991]  ? if_nlmsg_stats_size+0x2b0/0x2b0
[  836.246675]  netlink_rcv_skb+0x4e/0xf0
[  836.258046]  netlink_unicast+0x190/0x250
[  836.258669]  netlink_sendmsg+0x243/0x4b0
[  836.259288]  sock_sendmsg+0x33/0x40
[  836.259857]  ____sys_sendmsg+0x1d1/0x1f0
[  836.260473]  ___sys_sendmsg+0xab/0xf0
[  836.261064]  ? lock_acquire+0xce/0x2d0
[  836.261669]  ? find_held_lock+0x2b/0x80
[  836.262272]  ? __fget_files+0xb9/0x190
[  836.262871]  ? __fget_files+0xd3/0x190
[  836.263462]  __sys_sendmsg+0x51/0x90
[  836.264064]  do_syscall_64+0x3d/0x90
[  836.264652]  entry_SYSCALL_64_after_hwframe+0x46/0xb0
[  836.265425] RIP: 0033:0x7fdbe5e2677d

[  836.266012] Code: 28 89 54 24 1c 48 89 74 24 10 89 7c 24 08 e8 ba ee
ff ff 8b 54 24 1c 48 8b 74 24 10 41 89 c0 8b 7c 24 08 b8 2e 00 00 00 0f
05 <48> 3d 00 f0 ff ff 77 33 44 89 c7 48 89 44 24 08 e8 ee ee ff ff 48
[  836.268485] RSP: 002b:00007fdbe48a75a0 EFLAGS: 00000293 ORIG_RAX: 000000000000002e
[  836.269598] RAX: ffffffffffffffda RBX: 0000000000000001 RCX: 00007fdbe5e2677d
[  836.270576] RDX: 0000000000000000 RSI: 00007fdbe48a7640 RDI: 000000000000003c
[  836.271565] RBP: 00007fdbe48a8368 R08: 0000000000000000 R09: 0000000000000000
[  836.272546] R10: 00007fdbe48a84b0 R11: 0000000000000293 R12: 0000557bd17dc860
[  836.273527] R13: 0000000000000000 R14: 0000557bd17dc860 R15: 00007fdbe48a7640

[  836.274521]  </TASK>

To avoid using mode holding ldev->lock in the configure flow, we queue a
work to the lag workqueue and cease wait on a completion object.

In addition, we remove the lock from mlx5_lag_do_mirred() since it is
not really protecting anything.

It should be noted that an actual deadlock has not been observed.

Signed-off-by: Eli Cohen <elic@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: Fix handling of entry refcount when command is not issued to FW

In case command interface is down, or the command is not allowed, driver
did not increment the entry refcount, but might have decrement as part
of forced completion handling.

Fix that by always increment and decrement the refcount to make it
symmetric for all flows.

Fixes: 50b2412b7e78 ("net/mlx5: Avoid possible free of command entry while timeout comp handler")
Signed-off-by: Eran Ben Elisha <eranbe@nvidia.com>
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reported-by: Jack Wang <jinpu.wang@ionos.com>
Tested-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5: cmdif, Print info on any firmware cmd failure to tracepoint

While moving to new CMD API (quiet API), some pre-existing flows may call the new API
function that in case of error, returns the error instead of printing it as previously done.
For such flows we bring back the print but to tracepoint this time for sys admins to
have the ability to check for errors especially for commands using the new quiet API.

Tracepoint output example:
         devlink-1333    [001] .....   822.746922: mlx5_cmd: ACCESS_REG(0x805) op_mod(0x0) failed, status bad resource(0x5), syndrome (0xb06e1f), err(-22)

Fixes: f23519e542e5 ("net/mlx5: cmdif, Add new api for command execution")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>