tcp: Fix data-races around sysctl_tcp_recovery.

While reading sysctl_tcp_recovery, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 4f41b1c58a32 ("tcp: use RACK to detect losses")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix a data-race around sysctl_tcp_early_retrans.

While reading sysctl_tcp_early_retrans, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: eed530b6c676 ("tcp: early retransmit")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl knobs related to SYN option.

While reading these knobs, they can be changed concurrently.
Thus, we need to add READ_ONCE() to their readers.

  - tcp_sack
  - tcp_window_scaling
  - tcp_timestamps

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

udp: Fix a data-race around sysctl_udp_l3mdev_accept.

While reading sysctl_udp_l3mdev_accept, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 63a6fff353d0 ("net: Avoid receiving packets with an l3mdev on unbound UDP sockets")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix data-races around sysctl_ip_prot_sock.

sysctl_ip_prot_sock is accessed concurrently, and there is always a chance
of data-race.  So, all readers and writers need some basic protection to
avoid load/store-tearing.

Fixes: 4548b683b781 ("Introduce a sysctl that modifies the value of PROT_SOCK.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Fix data-races around sysctl_fib_multipath_hash_fields.

While reading sysctl_fib_multipath_hash_fields, it can be changed
concurrently.  Thus, we need to add READ_ONCE() to its readers.

Fixes: ce5c9c20d364 ("ipv4: Add a sysctl to control multipath hash fields")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Fix data-races around sysctl_fib_multipath_hash_policy.

While reading sysctl_fib_multipath_hash_policy, it can be changed
concurrently.  Thus, we need to add READ_ONCE() to its readers.

Fixes: bf4e0a3db97e ("net: ipv4: add support for ECMP hash policy choice")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Fix a data-race around sysctl_fib_multipath_use_neigh.

While reading sysctl_fib_multipath_use_neigh, it can be changed
concurrently.  Thus, we need to add READ_ONCE() to its reader.

Fixes: a6db4494d218 ("net: ipv4: Consider failed nexthops in multipath routes")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Steffen Klassert says:

====================
pull request (net): ipsec 2022-07-20

1) Fix a policy refcount imbalance in xfrm_bundle_lookup.
   From Hangyu Hua.

2) Fix some clang -Wformat warnings.
   Justin Stitt
====================

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

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

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-07-18

This series contains updates to iavf driver only.

Przemyslaw fixes handling of multiple VLAN requests to account for
individual errors instead of rejecting them all. He removes incorrect
implementations of ETHTOOL_COALESCE_MAX_FRAMES and
ETHTOOL_COALESCE_MAX_FRAMES_IRQ.

He also corrects an issue with NULL pointer caused by improper handling of
dummy receive descriptors. Finally, he corrects debug prints reporting an
unknown state.

* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  iavf: Fix missing state logs
  iavf: Fix handling of dummy receive descriptors
  iavf: Disallow changing rx/tx-frames and rx/tx-frames-irq
  iavf: Fix VLAN_V2 addition/rejection
====================

Link: https://lore.kernel.org/r/20220718174807.4113582-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Documentation: fix udp_wmem_min in ip-sysctl.rst

UDP doesn't support tx memory accounting, and sysctl udp_wmem_min
is not really used anywhere. So we should fix the description in
ip-sysctl.rst accordingly.

Fixes: 95766fff6b9a ("[UDP]: Add memory accounting.")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/c880a963d9b1fb5f442ae3c9e4dfa70d45296a16.1658167019.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ethernet: mtk_ppe: fix possible NULL pointer dereference in mtk_flow_get_wdma_info

odev pointer can be NULL in mtk_flow_offload_replace routine according
to the flower action rules. Fix possible NULL pointer dereference in
mtk_flow_get_wdma_info.

Fixes: a333215e10cb5 ("net: ethernet: mtk_eth_soc: implement flow offloading to WED devices")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/4e1685bc4976e21e364055f6bee86261f8f9ee93.1658137753.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

r8152: fix a WOL issue

This fixes that the platform is waked by an unexpected packet. The
size and range of FIFO is different when the device enters S3 state,
so it is necessary to correct some settings when suspending.

Regardless of jumbo frame, set RMS to 1522 and MTPS to MTPS_DEFAULT.
Besides, enable MCU_BORW_EN to update the method of calculating the
pointer of data. Then, the hardware could get the correct data.

Fixes: 195aae321c82 ("r8152: support new chips")
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Link: https://lore.kernel.org/r/20220718082120.10957-391-nic_swsd@realtek.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'amt-fix-validation-and-synchronization-bugs'

Taehee Yoo says:

====================
amt: fix validation and synchronization bugs

There are some synchronization issues in the amt module.
Especially, an amt gateway doesn't well synchronize its own variables
and status(amt->status).
It tries to use a workqueue for handles in a single thread.
A global lock is also good, but it would occur complex locking complex.

In this patchset, only the gateway uses workqueue.
The reason why only gateway interface uses workqueue is that gateway
should manage its own states and variables a little bit statefully.
But relay doesn't need to manage tunnels statefully, stateless is okay.
So, relay side message handlers are okay to be called concurrently.
But it doesn't mean that no lock is needed.

Only amt multicast data message type will not be processed by the work
queue because It contains actual multicast data.
So, it should be processed immediately.

When any amt gateway events are triggered(sending discovery message by
delayed_work, sending request message by delayed_work and receiving
messages), it stores event and skb into the event queue(amt->events[16]).
Then, workqueue processes these events one by one.

The first patch is to use the work queue.

The second patch is to remove unnecessary lock due to a previous patch.

The third patch is to use READ_ONCE() in the amt module.
Even if the amt module uses a single thread, some variables (ready4,
ready6, amt->status) can be accessed concurrently.

The fourth patch is to add missing nonce generation logic when it sends a
new request message.

The fifth patch is to drop unexpected advertisement messages.
advertisement message should be received only after the gateway sends
a discovery message first.
So, the gateway should drop advertisement messages if it has never
sent a discovery message and it also should drop duplicate advertisement
messages.
Using nonce is good to distinguish whether a received message is an
expected message or not.

The sixth patch is to drop unexpected query messages.
This is the same behavior as the fourth patch.
Query messages should be received only after the gateway sends a request
message first.
The nonce variable is used to distinguish whether it is a reply to a
previous request message or not.
amt->ready4 and amt->ready6 are used to distinguish duplicate messages.

The seventh patch is to drop unexpected multicast data.
AMT gateway should not receive multicast data message type before
establish between gateway and relay.
In order to drop unexpected multicast data messages, it checks amt->status.

The last patch is to fix a locking problem on the relay side.
amt->nr_tunnels variable is protected by amt->lock.
But amt_request_handler() doesn't protect this variable.

v2:
 - Use local_bh_disable() instead of rcu_read_lock_bh() in
   amt_membership_query_handler.
 - Fix using uninitialized variables.
 - Fix unexpectedly start the event_wq after stopping.
 - Fix possible deadlock in amt_event_work().
 - Add a limit variable in amt_event_work() to prevent infinite working.
 - Rename amt_queue_events() to amt_queue_event().
====================

Link: https://lore.kernel.org/r/20220717160910.19156-1-ap420073@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: do not use amt->nr_tunnels outside of lock

amt->nr_tunnels is protected by amt->lock.
But, amt_request_handler() has been using this variable without the
amt->lock.
So, it expands context of amt->lock in the amt_request_handler() to
protect amt->nr_tunnels variable.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: drop unexpected multicast data

AMT gateway interface should not receive unexpected multicast data.
Multicast data message type should be received after sending an update
message, which means all establishment between gateway and relay is
finished.
So, amt_multicast_data_handler() checks amt->status.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: drop unexpected query message

AMT gateway interface should not receive unexpected query messages.
In order to drop unexpected query messages, it checks nonce.
And it also checks ready4 and ready6 variables to drop duplicated messages.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: drop unexpected advertisement message

AMT gateway interface should not receive unexpected advertisement messages.
In order to drop these packets, it should check nonce and amt->status.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: add missing regeneration nonce logic in request logic

When AMT gateway starts sending a new request message, it should
regenerate the nonce variable.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: use READ_ONCE() in amt module

There are some data races in the amt module.
amt->ready4, amt->ready6, and amt->status can be accessed concurrently
without locks.
So, it uses READ_ONCE() and WRITE_ONCE().

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: remove unnecessary locks

By the previous patch, amt gateway handlers are changed to worked by
a single thread.
So, most locks for gateway are not needed.
So, it removes.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

amt: use workqueue for gateway side message handling

There are some synchronization issues(amt->status, amt->req_cnt, etc)
if the interface is in gateway mode because gateway message handlers
are processed concurrently.
This applies a work queue for processing these messages instead of
expanding the locking context.

So, the purposes of this patch are to fix exist race conditions and to make
gateway to be able to validate a gateway status more correctly.

When the AMT gateway interface is created, it tries to establish to relay.
The establishment step looks stateless, but it should be managed well.
In order to handle messages in the gateway, it saves the current
status(i.e. AMT_STATUS_XXX).
This patch makes gateway code to be worked with a single thread.

Now, all messages except the multicast are triggered(received or
delay expired), and these messages will be stored in the event
queue(amt->events).
Then, the single worker processes stored messages asynchronously one
by one.
The multicast data message type will be still processed immediately.

Now, amt->lock is only needed to access the event queue(amt->events)
if an interface is the gateway mode.

Fixes: cbc21dc1cfe9 ("amt: add data plane of amt interface")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: dsa: vitesse-vsc73xx: silent spi_device_id warnings

Add spi_device_id entries to silent SPI warnings.

Fixes: 5fa6863ba692 ("spi: Check we have a spi_device_id for each DT compatible")
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220717135831.2492844-2-o.rempel@pengutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: dsa: sja1105: silent spi_device_id warnings

Add spi_device_id entries to silent following warnings:
 SPI driver sja1105 has no spi_device_id for nxp,sja1105e
 SPI driver sja1105 has no spi_device_id for nxp,sja1105t
 SPI driver sja1105 has no spi_device_id for nxp,sja1105p
 SPI driver sja1105 has no spi_device_id for nxp,sja1105q
 SPI driver sja1105 has no spi_device_id for nxp,sja1105r
 SPI driver sja1105 has no spi_device_id for nxp,sja1105s
 SPI driver sja1105 has no spi_device_id for nxp,sja1110a
 SPI driver sja1105 has no spi_device_id for nxp,sja1110b
 SPI driver sja1105 has no spi_device_id for nxp,sja1110c
 SPI driver sja1105 has no spi_device_id for nxp,sja1110d

Fixes: 5fa6863ba692 ("spi: Check we have a spi_device_id for each DT compatible")
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20220717135831.2492844-1-o.rempel@pengutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

be2net: Fix buffer overflow in be_get_module_eeprom

be_cmd_read_port_transceiver_data assumes that it is given a buffer that
is at least PAGE_DATA_LEN long, or twice that if the module supports SFF
8472. However, this is not always the case.

Fix this by passing the desired offset and length to
be_cmd_read_port_transceiver_data so that we only copy the bytes once.

Fixes: e36edd9d26cf ("be2net: add ethtool "-m" option support")
Signed-off-by: Hristo Venev <hristo@venev.name>
Link: https://lore.kernel.org/r/20220716085134.6095-1-hristo@venev.name
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: stmmac: remove redunctant disable xPCS EEE call

Disable is done in stmmac_init_eee() on the event of MAC link down.
Since setting enable/disable EEE via ethtool will eventually trigger
a MAC down, removing this redunctant call in stmmac_ethtool.c to avoid
calling xpcs_config_eee() twice.

Fixes: d4aeaed80b0e ("net: stmmac: trigger PCS EEE to turn off on link down")
Signed-off-by: Wong Vee Khee <vee.khee.wong@linux.intel.com>
Link: https://lore.kernel.org/r/20220715122402.1017470-1-vee.khee.wong@linux.intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'fix-2-dsa-issues-with-vlan_filtering_is_global'

Vladimir Oltean says:

====================
Fix 2 DSA issues with vlan_filtering_is_global

This patch set fixes 2 issues with vlan_filtering_is_global switches.

Both are regressions introduced by refactoring commit d0004a020bb5
("net: dsa: remove the "dsa_to_port in a loop" antipattern from the
core"), which wasn't tested on a wide enough variety of switches.

Tested on the sja1105 driver.
====================

Link: https://lore.kernel.org/r/20220715151659.780544-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dsa: fix NULL pointer dereference in dsa_port_reset_vlan_filtering

The "ds" iterator variable used in dsa_port_reset_vlan_filtering() ->
dsa_switch_for_each_port() overwrites the "dp" received as argument,
which is later used to call dsa_port_vlan_filtering() proper.

As a result, switches which do enter that code path (the ones with
vlan_filtering_is_global=true) will dereference an invalid dp in
dsa_port_reset_vlan_filtering() after leaving a VLAN-aware bridge.

Use a dedicated "other_dp" iterator variable to avoid this from
happening.

Fixes: d0004a020bb5 ("net: dsa: remove the "dsa_to_port in a loop" antipattern from the core")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dsa: fix dsa_port_vlan_filtering when global

The blamed refactoring commit changed a "port" iterator with "other_dp",
but still looked at the slave_dev of the dp outside the loop, instead of
other_dp->slave from the loop.

As a result, dsa_port_vlan_filtering() would not call
dsa_slave_manage_vlan_filtering() except for the port in cause, and not
for all switch ports as expected.

Fixes: d0004a020bb5 ("net: dsa: remove the "dsa_to_port in a loop" antipattern from the core")
Reported-by: Lucian Banu <Lucian.Banu@westermo.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ixgbe: Add locking to prevent panic when setting sriov_numvfs to zero

It is possible to disable VFs while the PF driver is processing requests
from the VF driver.  This can result in a panic.

BUG: unable to handle kernel paging request at 000000000000106c
PGD 0 P4D 0
Oops: 0000 [#1] SMP NOPTI
CPU: 8 PID: 0 Comm: swapper/8 Kdump: loaded Tainted: G I      --------- -
Hardware name: Dell Inc. PowerEdge R740/06WXJT, BIOS 2.8.2 08/27/2020
RIP: 0010:ixgbe_msg_task+0x4c8/0x1690 [ixgbe]
Code: 00 00 48 8d 04 40 48 c1 e0 05 89 7c 24 24 89 fd 48 89 44 24 10 83 ff
01 0f 84 b8 04 00 00 4c 8b 64 24 10 4d 03 a5 48 22 00 00 <41> 80 7c 24 4c
00 0f 84 8a 03 00 00 0f b7 c7 83 f8 08 0f 84 8f 0a
RSP: 0018:ffffb337869f8df8 EFLAGS: 00010002
RAX: 0000000000001020 RBX: 0000000000000000 RCX: 000000000000002b
RDX: 0000000000000002 RSI: 0000000000000008 RDI: 0000000000000006
RBP: 0000000000000006 R08: 0000000000000002 R09: 0000000000029780
R10: 00006957d8f42832 R11: 0000000000000000 R12: 0000000000001020
R13: ffff8a00e8978ac0 R14: 000000000000002b R15: ffff8a00e8979c80
FS:  0000000000000000(0000) GS:ffff8a07dfd00000(0000) knlGS:00000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000000000000106c CR3: 0000000063e10004 CR4: 00000000007726e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
 <IRQ>
 ? ttwu_do_wakeup+0x19/0x140
 ? try_to_wake_up+0x1cd/0x550
 ? ixgbevf_update_xcast_mode+0x71/0xc0 [ixgbevf]
 ixgbe_msix_other+0x17e/0x310 [ixgbe]
 __handle_irq_event_percpu+0x40/0x180
 handle_irq_event_percpu+0x30/0x80
 handle_irq_event+0x36/0x53
 handle_edge_irq+0x82/0x190
 handle_irq+0x1c/0x30
 do_IRQ+0x49/0xd0
 common_interrupt+0xf/0xf

This can be eventually be reproduced with the following script:

while :
do
    echo 63 > /sys/class/net/<devname>/device/sriov_numvfs
    sleep 1
    echo 0 > /sys/class/net/<devname>/device/sriov_numvfs
    sleep 1
done

Add lock when disabling SR-IOV to prevent process VF mailbox communication.

Fixes: d773d1310625 ("ixgbe: Fix memory leak when SR-IOV VFs are direct assigned")
Signed-off-by: Piotr Skajewski <piotrx.skajewski@intel.com>
Tested-by: Marek Szlosek <marek.szlosek@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20220715214456.2968711-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

i40e: Fix erroneous adapter reinitialization during recovery process

Fix an issue when driver incorrectly detects state
of recovery process and erroneously reinitializes interrupts,
which results in a kernel error and call trace message.

The issue was caused by a combination of two factors:
1. Assuming the EMP reset issued after completing
firmware recovery means the whole recovery process is complete.
2. Erroneous reinitialization of interrupt vector after detecting
the above mentioned EMP reset.

Fixes (1) by changing how recovery state change is detected
and (2) by adjusting the conditional expression to ensure using proper
interrupt reinitialization method, depending on the situation.

Fixes: 4ff0ee1af016 ("i40e: Introduce recovery mode support")
Signed-off-by: Dawid Lukwinski <dawid.lukwinski@intel.com>
Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20220715214542.2968762-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ethernet: mtk_eth_soc: fix off by one check of ARRAY_SIZE

In mtk_wed_tx_ring_setup(.., int idx, ..), idx is used as an index here
  struct mtk_wed_ring *ring = &dev->tx_ring[idx];

The bounds of idx are checked here
  BUG_ON(idx > ARRAY_SIZE(dev->tx_ring));

If idx is the size of the array, it will pass this check and overflow.
So change the check to >= .

Fixes: 804775dfc288 ("net: ethernet: mtk_eth_soc: add support for Wireless Ethernet Dispatch (WED)")
Signed-off-by: Tom Rix <trix@redhat.com>
Link: https://lore.kernel.org/r/20220716214654.1540240-1-trix@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-lan966x-fix-issues-with-mac-table'

Horatiu Vultur says:

====================
net: lan966x: Fix issues with MAC table

The patch series fixes 2 issues:
- when an entry was forgotten the irq thread was holding a spin lock and then
  was talking also rtnl_lock.
- the access to the HW MAC table is indirect, so the access to the HW MAC
  table was not synchronized, which means that there could be race conditions.
====================

Link: https://lore.kernel.org/r/20220714194040.231651-1-horatiu.vultur@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lan966x: Fix usage of lan966x->mac_lock when used by FDB

When the SW bridge was trying to add/remove entries to/from HW, the
access to HW was not protected by any lock. In this way, it was
possible to have race conditions.
Fix this by using the lan966x->mac_lock to protect parallel access to HW
for this cases.

Fixes: 25ee9561ec622 ("net: lan966x: More MAC table functionality")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lan966x: Fix usage of lan966x->mac_lock inside lan966x_mac_irq_handler

The problem with this spin lock is that it was just protecting the list
of the MAC entries in SW and not also the access to the MAC entries in HW.
Because the access to HW is indirect, then it could happen to have race
conditions.
For example when SW introduced an entry in MAC table and the irq mac is
trying to read something from the MAC.
Update such that also the access to MAC entries in HW is protected by
this lock.

Fixes: 5ccd66e01cbef ("net: lan966x: add support for interrupts from analyzer")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lan966x: Fix usage of lan966x->mac_lock when entry is removed

To remove an entry to the MAC table, it is required first to setup the
entry and then issue a command for the MAC to forget the entry.
So if it happens for two threads to remove simultaneously an entry
in MAC table then it would be a race condition.
Fix this by using lan966x->mac_lock to protect the HW access.

Fixes: e18aba8941b40 ("net: lan966x: add mactable support")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lan966x: Fix usage of lan966x->mac_lock when entry is added

To add an entry to the MAC table, it is required first to setup the
entry and then issue a command for the MAC to learn the entry.
So if it happens for two threads to add simultaneously an entry in MAC
table then it would be a race condition.
Fix this by using lan966x->mac_lock to protect the HW access.

Fixes: fc0c3fe7486f2 ("net: lan966x: Add function lan966x_mac_ip_learn()")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lan966x: Fix taking rtnl_lock while holding spin_lock

When the HW deletes an entry in MAC table then it generates an
interrupt. The SW will go through it's own list of MAC entries and if it
is not found then it would notify the listeners about this. The problem
is that when the SW will go through it's own list it would take a spin
lock(lan966x->mac_lock) and when it notifies that the entry is deleted.
But to notify the listeners it taking the rtnl_lock which is illegal.

This is fixed by instead of notifying right away that the entry is
deleted, move the entry on a temp list and once, it checks all the
entries then just notify that the entries from temp list are deleted.

Fixes: 5ccd66e01cbe ("net: lan966x: add support for interrupts from analyzer")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

iavf: Fix missing state logs

Fix debug prints, by adding missing state prints.

Extend iavf_state_str by strings for __IAVF_INIT_EXTENDED_CAPS and
__IAVF_INIT_CONFIG_ADAPTER.

Without this patch, when enabling debug prints for iavf.h, user will
see:
iavf 0000:06:0e.0: state transition from:__IAVF_INIT_GET_RESOURCES to:__IAVF_UNKNOWN_STATE
iavf 0000:06:0e.0: state transition from:__IAVF_UNKNOWN_STATE to:__IAVF_UNKNOWN_STATE

Fixes: 605ca7c5c670 ("iavf: Fix kernel BUG in free_msi_irqs")
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Jun Zhang <xuejun.zhang@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iavf: Fix handling of dummy receive descriptors

Fix memory leak caused by not handling dummy receive descriptor properly.
iavf_get_rx_buffer now sets the rx_buffer return value for dummy receive
descriptors. Without this patch, when the hardware writes a dummy
descriptor, iavf would not free the page allocated for the previous receive
buffer. This is an unlikely event but can still happen.

[Jesse: massaged commit message]

Fixes: efa14c398582 ("iavf: allow null RX descriptors")
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iavf: Disallow changing rx/tx-frames and rx/tx-frames-irq

Remove from supported_coalesce_params ETHTOOL_COALESCE_MAX_FRAMES
and ETHTOOL_COALESCE_MAX_FRAMES_IRQ. As tx-frames-irq allowed
user to change budget for iavf_clean_tx_irq, remove work_limit
and use define for budget.

Without this patch there would be possibility to change rx/tx-frames
and rx/tx-frames-irq, which for rx/tx-frames did nothing, while for
rx/tx-frames-irq it changed rx/tx-frames and only changed budget
for cleaning NAPI poll.

Fixes: fbb7ddfef253 ("i40evf: core ethtool functionality")
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Jun Zhang <xuejun.zhang@intel.com>
Tested-by: Marek Szlosek <marek.szlosek@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iavf: Fix VLAN_V2 addition/rejection

Fix VLAN addition, so that PF driver does not reject whole VLAN batch.
Add VLAN reject handling, so rejected VLANs, won't litter VLAN filter
list. Fix handling of active_(c/s)vlans, so it will be possible to
re-add VLAN filters for user.
Without this patch, after changing trust to off, with VLAN filters
saturated, no VLAN is added, due to PF rejecting addition.

Fixes: 92fc50859872 ("iavf: Restrict maximum VLAN filters for VIRTCHNL_VF_OFFLOAD_VLAN_V2")
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

Merge branch 'dsa-docs'

Vladimir Oltean says:

====================
Update DSA documentation

These are some updates of dsa.rst, since it hasn't kept up with
development (in some cases, even since 2017). I've added Fixes: tags as
I thought was appropriate.
====================

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

docs: net: dsa: mention that VLANs are now refcounted on shared ports

The blamed commit updated the way in which VLANs are handled at the
cross-chip notifier layer and didn't update the documentation to say
that. Fix it.

Fixes: 134ef2388e7f ("net: dsa: add explicit support for host bridge VLANs")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: delete misinformation about -EOPNOTSUPP for FDB/MDB/VLAN

Returning -EOPNOTSUPP does *NOT* mean anything special.

port_vlan_add() is actually called from 2 code paths, one is
vlan_vid_add() from 8021q module and the other is
br_switchdev_port_vlan_add() from switchdev.

The bridge has a wrapper __vlan_vid_add() which first tries via
switchdev, then if that returns -EOPNOTSUPP, tries again via the VLAN RX
filters in the 8021q module. But DSA doesn't distinguish between one
call path and the other when calling the driver's port_vlan_add(), so if
the driver returns -EOPNOTSUPP to switchdev, it also returns -EOPNOTSUPP
to the 8021q module. And the latter is a hard error.

port_fdb_add() is called from the deferred dsa_owq only, so obviously
its return code isn't propagated anywhere, and cannot be interpreted in
any way.

The return code from port_mdb_add() is propagated to the bridge, but
again, this doesn't do anything special when -EOPNOTSUPP is returned,
but rather, br_switchdev_mdb_notify() returns void.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: re-explain what port_fdb_dump actually does

Switchdev has changed radically from its initial implementation, and the
currently provided definition is incorrect and very confusing.

Rewrite it in light of what it actually does.

Fixes: 2bedde1abbef ("net: dsa: Move FDB dump implementation inside DSA")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: add a section for address databases

The given definition for what VID 0 represents in the current
port_fdb_add and port_mdb_add is blatantly wrong. Delete it and explain
the concepts surrounding DSA's understanding of FDB isolation.

Fixes: c26933639b54 ("net: dsa: request drivers to perform FDB isolation")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: delete port_mdb_dump

This was deleted in 2017, stop documenting it.

Fixes: dc0cbff3ff9f ("net: dsa: Remove redundant MDB dump support")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: remove port_vlan_dump

This was deleted in 2017, delete the obsolete documentation.

Fixes: c069fcd82c57 ("net: dsa: Remove support for bypass bridge port attributes/vlan set")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: remove port_bridge_tx_fwd_offload

We've changed the API through which we can offload the bridge TX
forwarding process. Update the documentation in light of the removal of
2 DSA switch ops.

Fixes: b079922ba2ac ("net: dsa: add a "tx_fwd_offload" argument to ->port_bridge_join")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: document port_fast_age

The provided information about FDB flushing is not really up to date.
The DSA core automatically calls port_fast_age() when necessary, and
drivers should just implement that rather than hooking it to
port_bridge_leave, port_stp_state_set and others.

Fixes: 732f794c1baf ("net: dsa: add port fast ageing")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: document port_setup and port_teardown

These methods were added without being documented, fix that.

Fixes: fd292c189a97 ("net: dsa: tear down devlink port regions when tearing down the devlink port on error")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: document the teardown method

A teardown method was added to dsa_switch_ops without being documented.
Do so now.

Fixes: 5e3f847a02aa ("net: dsa: Add teardown callback for drivers")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: document change_tag_protocol

Support for changing the tagging protocol was added without this
operation being documented; do so now.

Fixes: 53da0ebaad10 ("net: dsa: allow changing the tag protocol via the "tagging" device attribute")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: add more info about the other arguments to get_tag_protocol

Changes were made to the prototype of get_tag_protocol without
describing at a high level what they are about. Update the documentation
to explain that.

Fixes: 5ed4e3eb0217 ("net: dsa: Pass a port to get_tag_protocol()")
Fixes: 4d776482ecc6 ("net: dsa: Get information about stacked DSA protocol")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: rename tag_protocol to get_tag_protocol

Since the blamed commit, the enum was turned into a function pointer and
also renamed. Update the documentation.

Fixes: 7b314362a234 ("net: dsa: Allow the DSA driver to indicate the tag protocol")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: document the shutdown behavior

Document the changes that took place in the DSA core in the blamed
commit.

Fixes: 0650bf52b31f ("net: dsa: be compatible with masters which unregister on shutdown")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docs: net: dsa: update probing documentation

Since the blamed commit we don't have register_switch_driver() and
unregister_switch_driver() anymore. Additionally, the expected
dsa_register_switch() and dsa_unregister_switch() calls aren't
documented.

Update the probing section with the details of how things are currently
done.

Fixes: 93e86b3bc842 ("net: dsa: Remove legacy probing support")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'net-ipv4-sysctl-races-part-3'

Kuniyuki Iwashima says:

====================
sysctl: Fix data-races around ipv4_net_table (Round 3).

This series fixes data-races around 21 knobs after
igmp_link_local_mcast_reports in ipv4_net_table.

These 4 knobs are skipped because they are safe.

  - tcp_congestion_control: Safe with RCU and xchg().
  - tcp_available_congestion_control: Read only.
  - tcp_allowed_congestion_control: Safe with RCU and spinlock().
  - tcp_fastopen_key: Safe with RCU and xchg()

So, round 4 will start with fib_multipath_use_neigh.
====================

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

tcp: Fix data-races around sysctl_tcp_fastopen_blackhole_timeout.

While reading sysctl_tcp_fastopen_blackhole_timeout, it can be changed
concurrently.  Thus, we need to add READ_ONCE() to its readers.

Fixes: cf1ef3f0719b ("net/tcp_fastopen: Disable active side TFO in certain scenarios")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_fastopen.

While reading sysctl_tcp_fastopen, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 2100c8d2d9db ("net-tcp: Fast Open base")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_max_syn_backlog.

While reading sysctl_max_syn_backlog, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix a data-race around sysctl_tcp_tw_reuse.

While reading sysctl_tcp_tw_reuse, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix a data-race around sysctl_tcp_notsent_lowat.

While reading sysctl_tcp_notsent_lowat, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: c9bee3b7fdec ("tcp: TCP_NOTSENT_LOWAT socket option")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around some timeout sysctl knobs.

While reading these sysctl knobs, they can be changed concurrently.
Thus, we need to add READ_ONCE() to their readers.

  - tcp_retries1
  - tcp_retries2
  - tcp_orphan_retries
  - tcp_fin_timeout

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_reordering.

While reading sysctl_tcp_reordering, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_migrate_req.

While reading sysctl_tcp_migrate_req, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: f9ac779f881c ("net: Introduce net.ipv4.tcp_migrate_req.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_syncookies.

While reading sysctl_tcp_syncookies, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_syn(ack)?_retries.

While reading sysctl_tcp_syn(ack)?_retries, they can be changed
concurrently.  Thus, we need to add READ_ONCE() to their readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around keepalive sysctl knobs.

While reading sysctl_tcp_keepalive_(time|probes|intvl), they can be changed
concurrently.  Thus, we need to add READ_ONCE() to their readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igmp: Fix data-races around sysctl_igmp_qrv.

While reading sysctl_igmp_qrv, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

This test can be packed into a helper, so such changes will be in the
follow-up series after net is merged into net-next.

  qrv ?: READ_ONCE(net->ipv4.sysctl_igmp_qrv);

Fixes: a9fe8e29945d ("ipv4: implement igmp_qrv sysctl to tune igmp robustness variable")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igmp: Fix data-races around sysctl_igmp_max_msf.

While reading sysctl_igmp_max_msf, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igmp: Fix a data-race around sysctl_igmp_max_memberships.

While reading sysctl_igmp_max_memberships, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igmp: Fix data-races around sysctl_igmp_llm_reports.

While reading sysctl_igmp_llm_reports, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

This test can be packed into a helper, so such changes will be in the
follow-up series after net is merged into net-next.

  if (ipv4_is_local_multicast(pmc->multiaddr) &&
      !READ_ONCE(net->ipv4.sysctl_igmp_llm_reports))

Fixes: df2cf4a78e48 ("IGMP: Inhibit reports for local multicast groups")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: prestera: acl: use proper mask for port selector

Adjusted as per packet processor documentation.
This allows to properly match 'indev' for clsact rules.

Fixes: 47327e198d42 ("net: prestera: acl: migrate to new vTCAM api")
Signed-off-by: Maksym Glubokiy <maksym.glubokiy@plvision.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/tls: Fix race in TLS device down flow

Socket destruction flow and tls_device_down function sync against each
other using tls_device_lock and the context refcount, to guarantee the
device resources are freed via tls_dev_del() by the end of
tls_device_down.

In the following unfortunate flow, this won't happen:
- refcount is decreased to zero in tls_device_sk_destruct.
- tls_device_down starts, skips the context as refcount is zero, going
  all the way until it flushes the gc work, and returns without freeing
  the device resources.
- only then, tls_device_queue_ctx_destruction is called, queues the gc
  work and frees the context's device resources.

Solve it by decreasing the refcount in the socket's destruction flow
under the tls_device_lock, for perfect synchronization.  This does not
slow down the common likely destructor flow, in which both the refcount
is decreased and the spinlock is acquired, anyway.

Fixes: e8f69799810c ("net/tls: Add generic NIC offload infrastructure")
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: fix dma queue left shift overflow issue

When queue number is > 4, left shift overflows due to 32 bits
integer variable. Mask calculation is wrong for MTL_RXQ_DMA_MAP1.

If CONFIG_UBSAN is enabled, kernel dumps below warning:
[   10.363842] ==================================================================
[   10.363882] UBSAN: shift-out-of-bounds in /build/linux-intel-iotg-5.15-8e6Tf4/
linux-intel-iotg-5.15-5.15.0/drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:224:12
[   10.363929] shift exponent 40 is too large for 32-bit type 'unsigned int'
[   10.363953] CPU: 1 PID: 599 Comm: NetworkManager Not tainted 5.15.0-1003-intel-iotg
[   10.363956] Hardware name: ADLINK Technology Inc. LEC-EL/LEC-EL, BIOS 0.15.11 12/22/2021
[   10.363958] Call Trace:
[   10.363960]  <TASK>
[   10.363963]  dump_stack_lvl+0x4a/0x5f
[   10.363971]  dump_stack+0x10/0x12
[   10.363974]  ubsan_epilogue+0x9/0x45
[   10.363976]  __ubsan_handle_shift_out_of_bounds.cold+0x61/0x10e
[   10.363979]  ? wake_up_klogd+0x4a/0x50
[   10.363983]  ? vprintk_emit+0x8f/0x240
[   10.363986]  dwmac4_map_mtl_dma.cold+0x42/0x91 [stmmac]
[   10.364001]  stmmac_mtl_configuration+0x1ce/0x7a0 [stmmac]
[   10.364009]  ? dwmac410_dma_init_channel+0x70/0x70 [stmmac]
[   10.364020]  stmmac_hw_setup.cold+0xf/0xb14 [stmmac]
[   10.364030]  ? page_pool_alloc_pages+0x4d/0x70
[   10.364034]  ? stmmac_clear_tx_descriptors+0x6e/0xe0 [stmmac]
[   10.364042]  stmmac_open+0x39e/0x920 [stmmac]
[   10.364050]  __dev_open+0xf0/0x1a0
[   10.364054]  __dev_change_flags+0x188/0x1f0
[   10.364057]  dev_change_flags+0x26/0x60
[   10.364059]  do_setlink+0x908/0xc40
[   10.364062]  ? do_setlink+0xb10/0xc40
[   10.364064]  ? __nla_validate_parse+0x4c/0x1a0
[   10.364068]  __rtnl_newlink+0x597/0xa10
[   10.364072]  ? __nla_reserve+0x41/0x50
[   10.364074]  ? __kmalloc_node_track_caller+0x1d0/0x4d0
[   10.364079]  ? pskb_expand_head+0x75/0x310
[   10.364082]  ? nla_reserve_64bit+0x21/0x40
[   10.364086]  ? skb_free_head+0x65/0x80
[   10.364089]  ? security_sock_rcv_skb+0x2c/0x50
[   10.364094]  ? __cond_resched+0x19/0x30
[   10.364097]  ? kmem_cache_alloc_trace+0x15a/0x420
[   10.364100]  rtnl_newlink+0x49/0x70

This change fixes MTL_RXQ_DMA_MAP1 mask issue and channel/queue
mapping warning.

Fixes: d43042f4da3e ("net: stmmac: mapping mtl rx to dma channel")
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=216195
Reported-by: Cedric Wassenaar <cedric@bytespeed.nl>
Signed-off-by: Junxiao Chang <junxiao.chang@intel.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: switch to use interrupt for hw crosstimestamping

Using current implementation of polling mode, there is high chances we
will hit into timeout error when running phc2sys. Hence, update the
implementation of hardware crosstimestamping to use the MAC interrupt
service routine instead of polling for TSIS bit in the MAC Timestamp
Interrupt Status register to be set.

Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Wong Vee Khee <vee.khee.wong@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp/udp: Make early_demux back namespacified.

Commit e21145a9871a ("ipv4: namespacify ip_early_demux sysctl knob") made
it possible to enable/disable early_demux on a per-netns basis.  Then, we
introduced two knobs, tcp_early_demux and udp_early_demux, to switch it for
TCP/UDP in commit dddb64bcb346 ("net: Add sysctl to toggle early demux for
tcp and udp").  However, the .proc_handler() was wrong and actually
disabled us from changing the behaviour in each netns.

We can execute early_demux if net.ipv4.ip_early_demux is on and each proto
.early_demux() handler is not NULL.  When we toggle (tcp|udp)_early_demux,
the change itself is saved in each netns variable, but the .early_demux()
handler is a global variable, so the handler is switched based on the
init_net's sysctl variable.  Thus, netns (tcp|udp)_early_demux knobs have
nothing to do with the logic.  Whether we CAN execute proto .early_demux()
is always decided by init_net's sysctl knob, and whether we DO it or not is
by each netns ip_early_demux knob.

This patch namespacifies (tcp|udp)_early_demux again.  For now, the users
of the .early_demux() handler are TCP and UDP only, and they are called
directly to avoid retpoline.  So, we can remove the .early_demux() handler
from inet6?_protos and need not dereference them in ip6?_rcv_finish_core().
If another proto needs .early_demux(), we can restore it at that time.

Fixes: dddb64bcb346 ("net: Add sysctl to toggle early demux for tcp and udp")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20220713175207.7727-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch '1GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-07-14

This series contains updates to e1000e and igc drivers.

Sasha re-enables GPT clock when exiting s0ix to prevent hardware unit
hang and reverts a workaround for this issue on e1000e.

Lennert Buytenhek restores checks for removed device while accessing
registers to prevent NULL pointer dereferences for igc.

* '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  igc: Reinstate IGC_REMOVED logic and implement it properly
  Revert "e1000e: Fix possible HW unit hang after an s0ix exit"
  e1000e: Enable GPT clock before sending message to CSME
====================

Link: https://lore.kernel.org/r/20220714175857.933537-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dsa: microchip: ksz_common: Fix refcount leak bug

In ksz_switch_register(), we should call of_node_put() for the
reference returned by of_get_child_by_name() which has increased
the refcount.

Fixes: 912aae27c6af ("net: dsa: microchip: really look for phy-mode in port nodes")
Signed-off-by: Liang He <windhl@126.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220714153138.375919-1-windhl@126.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'stmmac-dwmac-mediatec-clock-fix'

Biao Huang says:

====================
stmmac: dwmac-mediatek: fix clock issue

changes in v5:
1. add reivewd-by as Matthias's comments.
2. fix "warning: unused variable 'ret' [-Wunused-variable]" as Jakub's comments

changes in v4:
1. improve commit message and test ko insertion/remove as Matthias's comments.
2. add patch "net: stmmac: fix pm runtime issue in stmmac_dvr_remove()" to
   fix vlan filter deletion issue.
3. add patch "net: stmmac: fix unbalanced ptp clock issue in suspend/resume flow"
   to fix unbalanced ptp clock issue in suspend/resume flow.

changes in v3:
1. delete mediatek_dwmac_exit() since there is no operation in it,
as Matthias's comments.

changes in v2:
1. clock configuration is still needed in probe,
and invoke mediatek_dwmac_clks_config() instead.
2. update commit message.

v1:
remove duplicated clock configuration in init/exit.
====================

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

net: stmmac: fix unbalanced ptp clock issue in suspend/resume flow

Current stmmac driver will prepare/enable ptp_ref clock in
stmmac_init_tstamp_counter().

The stmmac_pltfr_noirq_suspend will disable it once in suspend flow.

But in resume flow,
stmmac_pltfr_noirq_resume --> stmmac_init_tstamp_counter
stmmac_resume --> stmmac_hw_setup --> stmmac_init_ptp --> stmmac_init_tstamp_counter
ptp_ref clock reference counter increases twice, which leads to unbalance
ptp clock when resume back.

Move ptp_ref clock prepare/enable out of stmmac_init_tstamp_counter to fix it.

Fixes: 0735e639f129d ("net: stmmac: skip only stmmac_ptp_register when resume from suspend")
Signed-off-by: Biao Huang <biao.huang@mediatek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: fix pm runtime issue in stmmac_dvr_remove()

If netif is running when stmmac_dvr_remove is invoked,
the unregister_netdev will call ndo_stop(stmmac_release) and
vlan_kill_rx_filter(stmmac_vlan_rx_kill_vid).

Currently, stmmac_dvr_remove() will disable pm runtime before
unregister_netdev. When stmmac_vlan_rx_kill_vid is invoked,
pm_runtime_resume_and_get in it returns EACCESS error number,
and reports:

dwmac-mediatek 11021000.ethernet eth0: stmmac_dvr_remove: removing driver
dwmac-mediatek 11021000.ethernet eth0: FPE workqueue stop
dwmac-mediatek 11021000.ethernet eth0: failed to kill vid 0081/0

Move the pm_runtime_disable to the end of stmmac_dvr_remove
to fix this issue.

Fixes: 6449520391dfc ("net: stmmac: properly handle with runtime pm in stmmac_dvr_remove()")
Signed-off-by: Biao Huang <biao.huang@mediatek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

stmmac: dwmac-mediatek: fix clock issue

The pm_runtime takes care of the clock handling in current
stmmac drivers, and dwmac-mediatek implement the
mediatek_dwmac_clks_config() as the callback for pm_runtime.

Then, stripping duplicated clocks handling in old init()/exit()
to fix clock issue in suspend/resume test.

As to clocks in probe/remove, vendor need symmetric handling to
ensure clocks balance.

Test pass, including suspend/resume and ko insertion/remove.

Fixes: 3186bdad97d5 ("stmmac: dwmac-mediatek: add platform level clocks management")
Signed-off-by: Biao Huang <biao.huang@mediatek.com>
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'net-sysctl-races-round2'

Kuniyuki Iwashima says:

====================
sysctl: Fix data-races around ipv4_net_table (Round 2).

This series fixes data-races around 15 knobs after ip_default_ttl in
ipv4_net_table.

These two knobs are skipped.
  - ip_local_port_range is safe with its own lock.
  - ip_local_reserved_ports uses proc_do_large_bitmap(), which will need
    an additional lock and can be fixed later.

So, the next round will start with igmp_link_local_mcast_reports.
====================

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

tcp: Fix a data-race around sysctl_tcp_probe_interval.

While reading sysctl_tcp_probe_interval, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 05cbc0db03e8 ("ipv4: Create probe timer for tcp PMTU as per RFC4821")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix a data-race around sysctl_tcp_probe_threshold.

While reading sysctl_tcp_probe_threshold, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 6b58e0a5f32d ("ipv4: Use binary search to choose tcp PMTU probe_size")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix a data-race around sysctl_tcp_mtu_probe_floor.

While reading sysctl_tcp_mtu_probe_floor, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: c04b79b6cfd7 ("tcp: add new tcp_mtu_probe_floor sysctl")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_min_snd_mss.

While reading sysctl_tcp_min_snd_mss, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 5f3e2bf008c2 ("tcp: add tcp_min_snd_mss sysctl")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_base_mss.

While reading sysctl_tcp_base_mss, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 5d424d5a674f ("[TCP]: MTU probing")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_mtu_probing.

While reading sysctl_tcp_mtu_probing, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 5d424d5a674f ("[TCP]: MTU probing")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Fix data-races around sysctl_tcp_l3mdev_accept.

While reading sysctl_tcp_l3mdev_accept, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 6dd9a14e92e5 ("net: Allow accepted sockets to be bound to l3mdev domain")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp/dccp: Fix a data-race around sysctl_tcp_fwmark_accept.

While reading sysctl_tcp_fwmark_accept, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 84f39b08d786 ("net: support marking accepting TCP sockets")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix a data-race around sysctl_fwmark_reflect.

While reading sysctl_fwmark_reflect, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: e110861f8609 ("net: add a sysctl to reflect the fwmark on replies")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix a data-race around sysctl_ip_autobind_reuse.

While reading sysctl_ip_autobind_reuse, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.

Fixes: 4b01a9674231 ("tcp: bind(0) remove the SO_REUSEADDR restriction when ephemeral ports are exhausted.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix data-races around sysctl_ip_nonlocal_bind.

While reading sysctl_ip_nonlocal_bind, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix data-races around sysctl_ip_fwd_update_priority.

While reading sysctl_ip_fwd_update_priority, it can be changed
concurrently.  Thus, we need to add READ_ONCE() to its readers.

Fixes: 432e05d32892 ("net: ipv4: Control SKB reprioritization after forwarding")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix data-races around sysctl_ip_fwd_use_pmtu.

While reading sysctl_ip_fwd_use_pmtu, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: f87c10a8aa1e ("ipv4: introduce ip_dst_mtu_maybe_forward and protect forwarding path against pmtu spoofing")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip: Fix data-races around sysctl_ip_no_pmtu_disc.

While reading sysctl_ip_no_pmtu_disc, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>