net/smc: Remove redundant refcount increase

For passive connections, the refcount increment has been done in
smc_clcsock_accept()-->smc_sock_alloc().

Fixes: 3b2dec2603d5 ("net/smc: restructure client and server code in af_smc")
Signed-off-by: Yacan Liu <liuyacan@corp.netease.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20220830152314.838736-1-liuyacan@corp.netease.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Revert "sch_cake: Return __NET_XMIT_STOLEN when consuming enqueued skb"

This reverts commit 90fabae8a2c225c4e4936723c38857887edde5cc.

Patch was applied hastily, revert and let the v2 be reviewed.

Fixes: 90fabae8a2c2 ("sch_cake: Return __NET_XMIT_STOLEN when consuming enqueued skb")
Link: https://lore.kernel.org/all/87wnao2ha3.fsf@toke.dk/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'tcp-tcp-challenge-ack-fixes'

Eric Dumazet says:

====================
tcp: tcp challenge ack fixes

syzbot found a typical data-race addressed in the first patch.

While we are at it, second patch makes the global rate limit
per net-ns and disabled by default.
====================

Link: https://lore.kernel.org/r/20220830185656.268523-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tcp: make global challenge ack rate limitation per net-ns and default disabled

Because per host rate limiting has been proven problematic (side channel
attacks can be based on it), per host rate limiting of challenge acks ideally
should be per netns and turned off by default.

This is a long due followup of following commits:

083ae308280d ("tcp: enable per-socket rate limiting of all 'challenge acks'")
f2b2c582e824 ("tcp: mitigate ACK loops for connections as tcp_sock")
75ff39ccc1bd ("tcp: make challenge acks less predictable")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jason Baron <jbaron@akamai.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tcp: annotate data-race around challenge_timestamp

challenge_timestamp can be read an written by concurrent threads.

This was expected, but we need to annotate the race to avoid potential issues.

Following patch moves challenge_timestamp and challenge_count
to per-netns storage to provide better isolation.

Fixes: 354e4aa391ed ("tcp: RFC 5961 5.2 Blind Data Injection Attack Mitigation")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dsa: hellcreek: Print warning only once

In case the source port cannot be decoded, print the warning only once. This
still brings attention to the user and does not spam the logs at the same time.

Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220830163448.8921-1-kurt@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ip: fix triggering of 'icmp redirect'

__mkroute_input() uses fib_validate_source() to trigger an icmp redirect.
My understanding is that fib_validate_source() is used to know if the src
address and the gateway address are on the same link. For that,
fib_validate_source() returns 1 (same link) or 0 (not the same network).
__mkroute_input() is the only user of these positive values, all other
callers only look if the returned value is negative.

Since the below patch, fib_validate_source() didn't return anymore 1 when
both addresses are on the same network, because the route lookup returns
RT_SCOPE_LINK instead of RT_SCOPE_HOST. But this is, in fact, right.
Let's adapat the test to return 1 again when both addresses are on the same
link.

CC: stable@vger.kernel.org
Fixes: 747c14307214 ("ip: fix dflt addr selection for connected nexthop")
Reported-by: kernel test robot <yujie.liu@intel.com>
Reported-by: Heng Qi <hengqi@linux.alibaba.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220829100121.3821-1-nicolas.dichtel@6wind.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sch_cake: Return __NET_XMIT_STOLEN when consuming enqueued skb

When the GSO splitting feature of sch_cake is enabled, GSO superpackets
will be broken up and the resulting segments enqueued in place of the
original skb. In this case, CAKE calls consume_skb() on the original skb,
but still returns NET_XMIT_SUCCESS. This can confuse parent qdiscs into
assuming the original skb still exists, when it really has been freed. Fix
this by adding the __NET_XMIT_STOLEN flag to the return value in this case.

Fixes: 0c850344d388 ("sch_cake: Conditionally split GSO segments")
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-18231
Link: https://lore.kernel.org/r/20220831092103.442868-1-toke@toke.dk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests: net: sort .gitignore file

This is the result of `sort tools/testing/selftests/net/.gitignore`, but
preserving the comment at the top.

Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Link: https://lore.kernel.org/r/20220829184748.1535580-1-axelrasmussen@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Documentation: networking: correct possessive "its"

Change occurrences of "it's" that are possessive to "its"
so that they don't read as "it is".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20220829235414.17110-1-rdunlap@infradead.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

kcm: fix strp_init() order and cleanup

strp_init() is called just a few lines above this csk->sk_user_data
check, it also initializes strp->work etc., therefore, it is
unnecessary to call strp_done() to cancel the freshly initialized
work.

And if sk_user_data is already used by KCM, psock->strp should not be
touched, particularly strp->work state, so we need to move strp_init()
after the csk->sk_user_data check.

This also makes a lockdep warning reported by syzbot go away.

Reported-and-tested-by: syzbot+9fc084a4348493ef65d2@syzkaller.appspotmail.com
Reported-by: syzbot+e696806ef96cdd2d87cd@syzkaller.appspotmail.com
Fixes: e5571240236c ("kcm: Check if sk_user_data already set in kcm_attach")
Fixes: dff8baa26117 ("kcm: Call strp_stop before strp_done in kcm_attach")
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Link: https://lore.kernel.org/r/20220827181314.193710-1-xiyou.wangcong@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

mlxbf_gige: compute MDIO period based on i1clk

This patch adds logic to compute the MDIO period based on
the i1clk, and thereafter write the MDIO period into the YU
MDIO config register. The i1clk resource from the ACPI table
is used to provide addressing to YU bootrecord PLL registers.
The values in these registers are used to compute MDIO period.
If the i1clk resource is not present in the ACPI table, then
the current default hardcorded value of 430Mhz is used.
The i1clk clock value of 430MHz is only accurate for boards
with BF2 mid bin and main bin SoCs. The BF2 high bin SoCs
have i1clk = 500MHz, but can support a slower MDIO period.

Fixes: f92e1869d74e ("Add Mellanox BlueField Gigabit Ethernet driver")
Reviewed-by: Asmaa Mnebhi <asmaa@nvidia.com>
Signed-off-by: David Thompson <davthompson@nvidia.com>
Link: https://lore.kernel.org/r/20220826155916.12491-1-davthompson@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ethernet: rocker: fix sleep in atomic context bug in neigh_timer_handler

The function neigh_timer_handler() is a timer handler that runs in an
atomic context. When used by rocker, neigh_timer_handler() calls
"kzalloc(.., GFP_KERNEL)" that may sleep. As a result, the sleep in
atomic context bug will happen. One of the processes is shown below:

ofdpa_fib4_add()
 ...
 neigh_add_timer()

(wait a timer)

neigh_timer_handler()
 neigh_release()
  neigh_destroy()
   rocker_port_neigh_destroy()
    rocker_world_port_neigh_destroy()
     ofdpa_port_neigh_destroy()
      ofdpa_port_ipv4_neigh()
       kzalloc(sizeof(.., GFP_KERNEL) //may sleep

This patch changes the gfp_t parameter of kzalloc() from GFP_KERNEL to
GFP_ATOMIC in order to mitigate the bug.

Fixes: 00fc0c51e35b ("rocker: Change world_ops API and implementation to be switchdev independant")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: lan966x: improve error handle in lan966x_fdma_rx_get_frame()

Don't just print a warning.  Clean up and return an error as well.

Fixes: c8349639324a ("net: lan966x: Add FDMA functionality")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Link: https://lore.kernel.org/r/YwjgDm/SVd5c1tQU@kili
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

nfp: fix the access to management firmware hanging

When running `ethtool -p` with the old management firmware,
the management firmware resource is not correctly released,
which causes firmware related malfunction: all the access
to management firmware hangs.

It releases the management firmware resource when set id
mode operation is not supported.

Fixes: ccb9bc1dfa44 ("nfp: add 'ethtool --identify' support")
Signed-off-by: Gao Xiao <gao.xiao@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Link: https://lore.kernel.org/r/20220829101651.633840-1-simon.horman@corigine.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'ieee802154-for-net-2022-08-29' of git://git./linux/kernel/git/sschmidt/wpan

Stefan Schmidt says:

====================
ieee802154 for net 2022-08-29

 - repeated word fix from Jilin Yuan.
 - missed return code setting in the cc2520 driver by Li Qiong.
 - fixing a potential race in by defering the workqueue destroy
   in the adf7242 driver by Lin Ma.
 - fixing a long standing problem in the mac802154 rx path to match
   corretcly by Miquel Raynal.

* tag 'ieee802154-for-net-2022-08-29' of git://git.kernel.org/pub/scm/linux/kernel/git/sschmidt/wpan:
  ieee802154: cc2520: add rc code in cc2520_tx()
  net: mac802154: Fix a condition in the receive path
  net/ieee802154: fix repeated words in comments
  ieee802154/adf7242: defer destroy_workqueue call
====================

Link: https://lore.kernel.org/r/20220829100308.2802578-1-stefan@datenfreihafen.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: phy: micrel: Make the GPIO to be non-exclusive

The same GPIO line can be shared by multiple phys for the coma mode pin.
If that is the case then, all the other phys that share the same line
will failed to be probed because the access to the gpio line is not
non-exclusive.
Fix this by making access to the gpio line to be nonexclusive using flag
GPIOD_FLAGS_BIT_NONEXCLUSIVE. This allows all the other PHYs to be
probed.

Fixes: 738871b09250ee ("net: phy: micrel: add coma mode GPIO")
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Link: https://lore.kernel.org/r/20220830064055.2340403-1-horatiu.vultur@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: virtio_net: fix notification coalescing comments

Fix wording in comments for the notifications coalescing feature.

Signed-off-by: Alvaro Karsz <alvaro.karsz@solid-run.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://lore.kernel.org/r/20220823073947.14774-1-alvaro.karsz@solid-run.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/sched: fix netdevice reference leaks in attach_default_qdiscs()

In attach_default_qdiscs(), if a dev has multiple queues and queue 0 fails
to attach qdisc because there is no memory in attach_one_default_qdisc().
Then dev->qdisc will be noop_qdisc by default. But the other queues may be
able to successfully attach to default qdisc.

In this case, the fallback to noqueue process will be triggered. If the
original attached qdisc is not released and a new one is directly
attached, this will cause netdevice reference leaks.

The following is the bug log:

veth0: default qdisc (fq_codel) fail, fallback to noqueue
unregister_netdevice: waiting for veth0 to become free. Usage count = 32
leaked reference.
 qdisc_alloc+0x12e/0x210
 qdisc_create_dflt+0x62/0x140
 attach_one_default_qdisc.constprop.41+0x44/0x70
 dev_activate+0x128/0x290
 __dev_open+0x12a/0x190
 __dev_change_flags+0x1a2/0x1f0
 dev_change_flags+0x23/0x60
 do_setlink+0x332/0x1150
 __rtnl_newlink+0x52f/0x8e0
 rtnl_newlink+0x43/0x70
 rtnetlink_rcv_msg+0x140/0x3b0
 netlink_rcv_skb+0x50/0x100
 netlink_unicast+0x1bb/0x290
 netlink_sendmsg+0x37c/0x4e0
 sock_sendmsg+0x5f/0x70
 ____sys_sendmsg+0x208/0x280

Fix this bug by clearing any non-noop qdiscs that may have been assigned
before trying to re-attach.

Fixes: bf6dba76d278 ("net: sched: fallback to qdisc noqueue if default qdisc setup fail")
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Link: https://lore.kernel.org/r/20220826090055.24424-1-wanghai38@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sched: tbf: don't call qdisc_put() while holding tree lock

The issue is the same to commit c2999f7fb05b ("net: sched: multiq: don't
call qdisc_put() while holding tree lock"). Qdiscs call qdisc_put() while
holding sch tree spinlock, which results sleeping-while-atomic BUG.

Fixes: c266f64dbfa2 ("net: sched: protect block state with mutex")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Link: https://lore.kernel.org/r/20220826013930.340121-1-shaozhengchao@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Merge branch 'u64_stats-fixups'

Sebastian Andrzej Siewior says:

====================
net: u64_stats fixups for 32bit.

while looking at the u64-stats patch
https://lore.kernel.org/all/20220817162703.728679-10-bigeasy@linutronix.de

I noticed that u64_stats_fetch_begin() is used. That suspicious thing
about it is that network processing, including stats update, is
performed in NAPI and so I would expect to see
u64_stats_fetch_begin_irq() in order to avoid updates from NAPI during
the read. This is only needed on 32bit-UP where the seqcount is not
used. This is address in 2/2. The remaining user take some kind of
precaution and may use u64_stats_fetch_begin().

I updated the previously mentioned patch to get rid of
u64_stats_fetch_begin_irq(). If this is not considered stable patch
worthy then it can be ignored and considred fixed by the other series
which removes the special 32bit cases.

The xrs700x driver reads and writes the counter from preemptible context
so the only missing piece here is at least disable preemption on the
writer side to avoid preemption while the writer is in progress. The
possible reader would spin then until the writer completes its write
critical section which is considered bad. This is addressed in 1/2 by
using u64_stats_update_begin_irqsave() and so disable interrupts during
the write critical section.
The other closet resemblance I found is mdio_bus.c::mdiobus_stats_acct()
where preemtion is disabled unconditionally. This is something I want to
avoid since it also affects 64bit.
====================

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

net: Use u64_stats_fetch_begin_irq() for stats fetch.

On 32bit-UP u64_stats_fetch_begin() disables only preemption. If the
reader is in preemptible context and the writer side
(u64_stats_update_begin*()) runs in an interrupt context (IRQ or
softirq) then the writer can update the stats during the read operation.
This update remains undetected.

Use u64_stats_fetch_begin_irq() to ensure the stats fetch on 32bit-UP
are not interrupted by a writer. 32bit-SMP remains unaffected by this
change.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: Catherine Sullivan <csully@google.com>
Cc: David Awogbemila <awogbemila@google.com>
Cc: Dimitris Michailidis <dmichail@fungible.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Hans Ulli Kroll <ulli.kroll@googlemail.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Jeroen de Borst <jeroendb@google.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Simon Horman <simon.horman@corigine.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-wireless@vger.kernel.org
Cc: netdev@vger.kernel.org
Cc: oss-drivers@corigine.com
Cc: stable@vger.kernel.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: xrs700x: Use irqsave variant for u64 stats update

xrs700x_read_port_counters() updates the stats from a worker using the
u64_stats_update_begin() version. This is okay on 32-UP since on the
reader side preemption is disabled.
On 32bit-SMP the writer can be preempted by the reader at which point
the reader will spin on the seqcount until writer continues and
completes the update.

Assigning the mib_mutex mutex to the underlying seqcount would ensure
proper synchronisation. The API for that on the u64_stats_init() side
isn't available. Since it is the only user, just use disable interrupts
during the update.

Use u64_stats_update_begin_irqsave() on the writer side to ensure an
uninterrupted update.

Fixes: ee00b24f32eb8 ("net: dsa: add Arrow SpeedChips XRS700x driver")
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: George McCollister <george.mccollister@gmail.com>
Cc: Vivien Didelot <vivien.didelot@gmail.com>
Cc: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: George McCollister <george.mccollister@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ieee802154: cc2520: add rc code in cc2520_tx()

The rc code is 0 at the error path "status & CC2520_STATUS_TX_UNDERFLOW".
Assign rc code with '-EINVAL' at this error path to fix it.

Signed-off-by: Li Qiong <liqiong@nfschina.com>
Link: https://lore.kernel.org/r/20220829071259.18330-1-liqiong@nfschina.com
Signed-off-by: Stefan Schmidt <stefan@datenfreihafen.org>

net: mac802154: Fix a condition in the receive path

Upon reception, a packet must be categorized, either it's destination is
the host, or it is another host. A packet with no destination addressing
fields may be valid in two situations:
- the packet has no source field: only ACKs are built like that, we
  consider the host as the destination.
- the packet has a valid source field: it is directed to the PAN
  coordinator, as for know we don't have this information we consider we
  are not the PAN coordinator.

There was likely a copy/paste error made during a previous cleanup
because the if clause is now containing exactly the same condition as in
the switch case, which can never be true. In the past the destination
address was used in the switch and the source address was used in the
if, which matches what the spec says.

Cc: stable@vger.kernel.org
Fixes: ae531b9475f6 ("ieee802154: use ieee802154_addr instead of *_sa variants")
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/r/20220826142954.254853-1-miquel.raynal@bootlin.com
Signed-off-by: Stefan Schmidt <stefan@datenfreihafen.org>

r8152: add PID for the Lenovo OneLink+ Dock

The Lenovo OneLink+ Dock contains an RTL8153 controller that behaves as
a broken CDC device by default. Add the custom Lenovo PID to the r8152
driver to support it properly.

Also, systems compatible with this dock provide a BIOS option to enable
MAC address passthrough (as per Lenovo document "ThinkPad Docking
Solutions 2017"). Add the custom PID to the MAC passthrough list too.

Tested on a ThinkPad 13 1st gen with the expected results:

passthrough disabled: Invalid header when reading pass-thru MAC addr
passthrough enabled:  Using pass-thru MAC addr XX:XX:XX:XX:XX:XX

Signed-off-by: Jean-Francois Le Fillatre <jflf_kernel@gmx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

nfp: flower: fix ingress police using matchall filter

Referenced commit introduced nfp_policer_validate in the progress
installing rate limiter. This validate check the action id and will
reject police with CONTINUE, which is required to support ingress
police offload.

Fix this issue by allowing FLOW_ACTION_CONTINUE as notexceed action
id in nfp_policer_validate

Fixes: d97b4b105ce7 ("flow_offload: reject offload for all drivers with invalid police parameters")
Signed-off-by: Tianyu Yuan <tianyu.yuan@corigine.com>
Reviewed-by: Baowen Zheng <baowen.zheng@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20220825080845.507534-1-simon.horman@corigine.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

openvswitch: fix memory leak at failed datapath creation

ovs_dp_cmd_new()->ovs_dp_change()->ovs_dp_set_upcall_portids()
allocates array via kmalloc.
If for some reason new_vport() fails during ovs_dp_cmd_new()
dp->upcall_portids must be freed.
Add missing kfree.

Kmemleak example:
unreferenced object 0xffff88800c382500 (size 64):
  comm "dump_state", pid 323, jiffies 4294955418 (age 104.347s)
  hex dump (first 32 bytes):
    5e c2 79 e4 1f 7a 38 c7 09 21 38 0c 80 88 ff ff  ^.y..z8..!8.....
    03 00 00 00 0a 00 00 00 14 00 00 00 28 00 00 00  ............(...
  backtrace:
    [<0000000071bebc9f>] ovs_dp_set_upcall_portids+0x38/0xa0
    [<000000000187d8bd>] ovs_dp_change+0x63/0xe0
    [<000000002397e446>] ovs_dp_cmd_new+0x1f0/0x380
    [<00000000aa06f36e>] genl_family_rcv_msg_doit+0xea/0x150
    [<000000008f583bc4>] genl_rcv_msg+0xdc/0x1e0
    [<00000000fa10e377>] netlink_rcv_skb+0x50/0x100
    [<000000004959cece>] genl_rcv+0x24/0x40
    [<000000004699ac7f>] netlink_unicast+0x23e/0x360
    [<00000000c153573e>] netlink_sendmsg+0x24e/0x4b0
    [<000000006f4aa380>] sock_sendmsg+0x62/0x70
    [<00000000d0068654>] ____sys_sendmsg+0x230/0x270
    [<0000000012dacf7d>] ___sys_sendmsg+0x88/0xd0
    [<0000000011776020>] __sys_sendmsg+0x59/0xa0
    [<000000002e8f2dc1>] do_syscall_64+0x3b/0x90
    [<000000003243e7cb>] entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: b83d23a2a38b ("openvswitch: Introduce per-cpu upcall dispatch")
Acked-by: Aaron Conole <aconole@redhat.com>
Signed-off-by: Andrey Zhadchenko <andrey.zhadchenko@virtuozzo.com>
Link: https://lore.kernel.org/r/20220825020326.664073-1-andrey.zhadchenko@virtuozzo.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: smsc911x: Stop and start PHY during suspend and resume

Commit 744d23c71af3 ("net: phy: Warn about incorrect
mdio_bus_phy_resume() state") unveiled that the smsc911x driver was not
properly stopping and restarting the PHY during suspend/resume. Correct
that by indicating that the MAC is in charge of PHY PM operations and
ensure that all MDIO bus activity is quiescent during suspend.

Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Fixes: fba863b81604 ("net: phy: make PHY PM ops a no-op if MAC driver manages PHY PM")
Fixes: 2aa70f864955 ("net: smsc911x: Quieten netif during suspend")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20220825023951.3220-1-f.fainelli@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tg3: Disable tg3 device on system reboot to avoid triggering AER

Commit d60cd06331a3 ("PM: ACPI: reboot: Use S5 for reboot") caused a
reboot hang on one Dell servers so the commit was reverted.

Someone managed to collect the AER log and it's caused by MSI:
[ 148.762067] ACPI: Preparing to enter system sleep state S5
[ 148.794638] {1}[Hardware Error]: Hardware error from APEI Generic Hardware Error Source: 5
[ 148.803731] {1}[Hardware Error]: event severity: recoverable
[ 148.810191] {1}[Hardware Error]: Error 0, type: fatal
[ 148.816088] {1}[Hardware Error]: section_type: PCIe error
[ 148.822391] {1}[Hardware Error]: port_type: 0, PCIe end point
[ 148.829026] {1}[Hardware Error]: version: 3.0
[ 148.834266] {1}[Hardware Error]: command: 0x0006, status: 0x0010
[ 148.841140] {1}[Hardware Error]: device_id: 0000:04:00.0
[ 148.847309] {1}[Hardware Error]: slot: 0
[ 148.852077] {1}[Hardware Error]: secondary_bus: 0x00
[ 148.857876] {1}[Hardware Error]: vendor_id: 0x14e4, device_id: 0x165f
[ 148.865145] {1}[Hardware Error]: class_code: 020000
[ 148.870845] {1}[Hardware Error]: aer_uncor_status: 0x00100000, aer_uncor_mask: 0x00010000
[ 148.879842] {1}[Hardware Error]: aer_uncor_severity: 0x000ef030
[ 148.886575] {1}[Hardware Error]: TLP Header: 40000001 0000030f 90028090 00000000
[ 148.894823] tg3 0000:04:00.0: AER: aer_status: 0x00100000, aer_mask: 0x00010000
[ 148.902795] tg3 0000:04:00.0: AER: [20] UnsupReq (First)
[ 148.910234] tg3 0000:04:00.0: AER: aer_layer=Transaction Layer, aer_agent=Requester ID
[ 148.918806] tg3 0000:04:00.0: AER: aer_uncor_severity: 0x000ef030
[ 148.925558] tg3 0000:04:00.0: AER: TLP Header: 40000001 0000030f 90028090 00000000

The MSI is probably raised by incoming packets, so power down the device
and disable bus mastering to stop the traffic, as user confirmed this
approach works.

In addition to that, be extra safe and cancel reset task if it's running.

Cc: Josef Bacik <josef@toxicpanda.com>
Link: https://lore.kernel.org/all/b8db79e6857c41dab4ef08bdf826ea7c47e3bafc.1615947283.git.josef@toxicpanda.com/
BugLink: https://bugs.launchpad.net/bugs/1917471
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20220826002530.1153296-1-kai.heng.feng@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: sparx5: fix handling uneven length packets in manual extraction

Packets that are not of length divisible by 4 (e.g. 77, 78, 79) would
have the checksum included up to next multiple of 4 (a 77 bytes packet
would have 3 bytes of ethernet checksum included). The check for the
value expects it in host (Little) endian.

Fixes: f3cad2611a77 ("net: sparx5: add hostmode with phylink support")
Signed-off-by: Casper Andersson <casper.casan@gmail.com>
Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Link: https://lore.kernel.org/r/20220825084955.684637-1-casper.casan@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'for-net-2022-08-25' of git://git./linux/kernel/git/bluetooth/bluetooth

Luiz Augusto von Dentz says:

====================
bluetooth pull request for net:

 - Fix handling of duplicate connection handle
 - Fix handling of HCI vendor opcode
 - Fix suspend performance regression
 - Fix build errors
 - Fix not handling shutdown condition on ISO sockets
 - Fix double free issue

* tag 'for-net-2022-08-25' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth:
  Bluetooth: hci_sync: hold hdev->lock when cleanup hci_conn
  Bluetooth: move from strlcpy with unused retval to strscpy
  Bluetooth: hci_event: Fix checking conn for le_conn_complete_evt
  Bluetooth: ISO: Fix not handling shutdown condition
  Bluetooth: hci_sync: fix double mgmt_pending_free() in remove_adv_monitor()
  Bluetooth: MGMT: Fix Get Device Flags
  Bluetooth: L2CAP: Fix build errors in some archs
  Bluetooth: hci_sync: Fix suspend performance regression
  Bluetooth: hci_event: Fix vendor (unknown) opcode status handling
====================

Link: https://lore.kernel.org/r/20220825234559.1837409-1-luiz.dentz@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel borkmann says:

====================
The following pull-request contains BPF updates for your *net* tree.

We've added 11 non-merge commits during the last 14 day(s) which contain
a total of 13 files changed, 61 insertions(+), 24 deletions(-).

The main changes are:

1) Fix BPF verifier's precision tracking around BPF ring buffer, from Kumar Kartikeya Dwivedi.

2) Fix regression in tunnel key infra when passing FLOWI_FLAG_ANYSRC, from Eyal Birger.

3) Fix insufficient permissions for bpf_sys_bpf() helper, from YiFei Zhu.

4) Fix splat from hitting BUG when purging effective cgroup programs, from Pu Lehui.

5) Fix range tracking for array poke descriptors, from Daniel Borkmann.

6) Fix corrupted packets for XDP_SHARED_UMEM in aligned mode, from Magnus Karlsson.

7) Fix NULL pointer splat in BPF sockmap sk_msg_recvmsg(), from Liu Jian.

8) Add READ_ONCE() to bpf_jit_limit when reading from sysctl, from Kuniyuki Iwashima.

9) Add BPF selftest lru_bug check to s390x deny list, from Daniel Müller.
====================

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

Merge tag 'wireless-2022-08-26' of git://git./linux/kernel/git/wireless/wireless

Johannes Berg says:

====================
pull-request: wireless-2022-08-26

Here are a couple of fixes for the current cycle,
see the tag description below.

Just a couple of fixes:
 * two potential leaks
 * use-after-free in certain scan races
 * warning in IBSS code
 * error return from a debugfs file was wrong
 * possible NULL-ptr-deref when station lookup fails

Please pull and let me know if there's any problem.
====================

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

Bluetooth: hci_sync: hold hdev->lock when cleanup hci_conn

When disconnecting all devices, hci_conn_failed is used to cleanup
hci_conn object when the hci_conn object cannot be aborted.
The function hci_conn_failed requires the caller holds hdev->lock.

Fixes: 9b3628d79b46f ("Bluetooth: hci_sync: Cleanup hci_conn if it cannot be aborted")
Signed-off-by: Zhengping Jiang <jiangzp@google.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: move from strlcpy with unused retval to strscpy

Follow the advice of the below link and prefer 'strscpy' in this
subsystem. Conversion is 1:1 because the return value is not used.
Generated by a coccinelle script.

Link: https://lore.kernel.org/r/CAHk-=wgfRnXz0W3D37d01q3JFkr_i_uTL=V6A6G1oUZcprmknw@mail.gmail.com/
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: hci_event: Fix checking conn for le_conn_complete_evt

To prevent multiple conn complete events, we shouldn't look up the
conn with hci_lookup_le_connect, since it requires the state to be
BT_CONNECT. By the time the duplicate event is processed, the state
might have changed, so we end up processing the new event anyway.

Change the lookup function to hci_conn_hash_lookup_ba.

Fixes: d5ebaa7c5f6f6 ("Bluetooth: hci_event: Ignore multiple conn complete events")
Signed-off-by: Archie Pusaka <apusaka@chromium.org>
Reviewed-by: Sonny Sasaka <sonnysasaka@chromium.org>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: ISO: Fix not handling shutdown condition

In order to properly handle shutdown syscall the code shall not assume
that the how argument is always SHUT_RDWR resulting in SHUTDOWN_MASK as
that would result in poll to immediately report EPOLLHUP instead of
properly waiting for disconnect_cfm (Disconnect Complete) which is
rather important for the likes of BAP as the CIG may need to be
reprogrammed.

Fixes: ccf74f2390d6 ("Bluetooth: Add BTPROTO_ISO socket type")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: hci_sync: fix double mgmt_pending_free() in remove_adv_monitor()

syzbot is reporting double kfree() at remove_adv_monitor() [1], for
commit 7cf5c2978f23fdbb ("Bluetooth: hci_sync: Refactor remove Adv
Monitor") forgot to remove duplicated mgmt_pending_remove() when
merging "if (err) {" path and "if (!pending) {" path.

Link: https://syzkaller.appspot.com/bug?extid=915a8416bf15895b8e07
Reported-by: syzbot <syzbot+915a8416bf15895b8e07@syzkaller.appspotmail.com>
Fixes: 7cf5c2978f23fdbb ("Bluetooth: hci_sync: Refactor remove Adv Monitor")
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: MGMT: Fix Get Device Flags

Get Device Flags don't check if device does actually use an RPA in which
case it shall only set HCI_CONN_FLAG_REMOTE_WAKEUP if LL Privacy is
enabled.

Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: L2CAP: Fix build errors in some archs

This attempts to fix the follow errors:

In function 'memcmp',
    inlined from 'bacmp' at ./include/net/bluetooth/bluetooth.h:347:9,
    inlined from 'l2cap_global_chan_by_psm' at
    net/bluetooth/l2cap_core.c:2003:15:
./include/linux/fortify-string.h:44:33: error: '__builtin_memcmp'
specified bound 6 exceeds source size 0 [-Werror=stringop-overread]
   44 | #define __underlying_memcmp     __builtin_memcmp
      |                                 ^
./include/linux/fortify-string.h:420:16: note: in expansion of macro
'__underlying_memcmp'
  420 |         return __underlying_memcmp(p, q, size);
      |                ^~~~~~~~~~~~~~~~~~~
In function 'memcmp',
    inlined from 'bacmp' at ./include/net/bluetooth/bluetooth.h:347:9,
    inlined from 'l2cap_global_chan_by_psm' at
    net/bluetooth/l2cap_core.c:2004:15:
./include/linux/fortify-string.h:44:33: error: '__builtin_memcmp'
specified bound 6 exceeds source size 0 [-Werror=stringop-overread]
   44 | #define __underlying_memcmp     __builtin_memcmp
      |                                 ^
./include/linux/fortify-string.h:420:16: note: in expansion of macro
'__underlying_memcmp'
  420 |         return __underlying_memcmp(p, q, size);
      |                ^~~~~~~~~~~~~~~~~~~

Fixes: 332f1795ca20 ("Bluetooth: L2CAP: Fix l2cap_global_chan_by_psm regression")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: hci_sync: Fix suspend performance regression

This attempts to fix suspend performance when there is no connections by
not updating the event mask.

Fixes: ef61b6ea1544 ("Bluetooth: Always set event mask on suspend")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Bluetooth: hci_event: Fix vendor (unknown) opcode status handling

Commit c8992cffbe74 ("Bluetooth: hci_event: Use of a function table to
handle Command Complete") was (presumably) meant to only refactor things
without any functional changes.

But it does have one undesirable side-effect, before *status would always
be set to skb->data[0] and it might be overridden by some of the opcode
specific handling. While now it always set by the opcode specific handlers.
This means that if the opcode is not known *status does not get set any
more at all!

This behavior change has broken bluetooth support for BCM4343A0 HCIs,
the hci_bcm.c code tries to configure UART attached HCIs at a higher
baudraute using vendor specific opcodes. The BCM4343A0 does not
support this and this used to simply fail:

[   25.646442] Bluetooth: hci0: BCM: failed to write clock (-56)
[   25.646481] Bluetooth: hci0: Failed to set baudrate

After which things would continue with the initial baudraute. But now
that hci_cmd_complete_evt() no longer sets status for unknown opcodes
*status is left at 0. This causes the hci_bcm.c code to think the baudraute
has been changed on the HCI side and to also adjust the UART baudrate,
after which communication with the HCI is broken, leading to:

[   28.579042] Bluetooth: hci0: command 0x0c03 tx timeout
[   36.961601] Bluetooth: hci0: BCM: Reset failed (-110)

And non working bluetooth. Fix this by restoring the previous
default "*status = skb->data[0]" handling for unknown opcodes.

Fixes: c8992cffbe74 ("Bluetooth: hci_event: Use of a function table to handle Command Complete")
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

bpf: Don't use tnum_range on array range checking for poke descriptors

Hsin-Wei reported a KASAN splat triggered by their BPF runtime fuzzer which
is based on a customized syzkaller:

  BUG: KASAN: slab-out-of-bounds in bpf_int_jit_compile+0x1257/0x13f0
  Read of size 8 at addr ffff888004e90b58 by task syz-executor.0/1489
  CPU: 1 PID: 1489 Comm: syz-executor.0 Not tainted 5.19.0 #1
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
  1.13.0-1ubuntu1.1 04/01/2014
  Call Trace:
   <TASK>
   dump_stack_lvl+0x9c/0xc9
   print_address_description.constprop.0+0x1f/0x1f0
   ? bpf_int_jit_compile+0x1257/0x13f0
   kasan_report.cold+0xeb/0x197
   ? kvmalloc_node+0x170/0x200
   ? bpf_int_jit_compile+0x1257/0x13f0
   bpf_int_jit_compile+0x1257/0x13f0
   ? arch_prepare_bpf_dispatcher+0xd0/0xd0
   ? rcu_read_lock_sched_held+0x43/0x70
   bpf_prog_select_runtime+0x3e8/0x640
   ? bpf_obj_name_cpy+0x149/0x1b0
   bpf_prog_load+0x102f/0x2220
   ? __bpf_prog_put.constprop.0+0x220/0x220
   ? find_held_lock+0x2c/0x110
   ? __might_fault+0xd6/0x180
   ? lock_downgrade+0x6e0/0x6e0
   ? lock_is_held_type+0xa6/0x120
   ? __might_fault+0x147/0x180
   __sys_bpf+0x137b/0x6070
   ? bpf_perf_link_attach+0x530/0x530
   ? new_sync_read+0x600/0x600
   ? __fget_files+0x255/0x450
   ? lock_downgrade+0x6e0/0x6e0
   ? fput+0x30/0x1a0
   ? ksys_write+0x1a8/0x260
   __x64_sys_bpf+0x7a/0xc0
   ? syscall_enter_from_user_mode+0x21/0x70
   do_syscall_64+0x3b/0x90
   entry_SYSCALL_64_after_hwframe+0x63/0xcd
  RIP: 0033:0x7f917c4e2c2d

The problem here is that a range of tnum_range(0, map->max_entries - 1) has
limited ability to represent the concrete tight range with the tnum as the
set of resulting states from value + mask can result in a superset of the
actual intended range, and as such a tnum_in(range, reg->var_off) check may
yield true when it shouldn't, for example tnum_range(0, 2) would result in
00XX -> v = 0000, m = 0011 such that the intended set of {0, 1, 2} is here
represented by a less precise superset of {0, 1, 2, 3}. As the register is
known const scalar, really just use the concrete reg->var_off.value for the
upper index check.

Fixes: d2e4c1e6c294 ("bpf: Constant map key tracking for prog array pokes")
Reported-by: Hsin-Wei Hung <hsinweih@uci.edu>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/r/984b37f9fdf7ac36831d2137415a4a915744c1b6.1661462653.git.daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Merge tag 'net-6.0-rc3' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from ipsec and netfilter (with one broken Fixes tag).

  Current release - new code bugs:

   - dsa: don't dereference NULL extack in dsa_slave_changeupper()

   - dpaa: fix <1G ethernet on LS1046ARDB

   - neigh: don't call kfree_skb() under spin_lock_irqsave()

  Previous releases - regressions:

   - r8152: fix the RX FIFO settings when suspending

   - dsa: microchip: keep compatibility with device tree blobs with no
     phy-mode

   - Revert "net: macsec: update SCI upon MAC address change."

   - Revert "xfrm: update SA curlft.use_time", comply with RFC 2367

  Previous releases - always broken:

   - netfilter: conntrack: work around exceeded TCP receive window

   - ipsec: fix a null pointer dereference of dst->dev on a metadata dst
     in xfrm_lookup_with_ifid

   - moxa: get rid of asymmetry in DMA mapping/unmapping

   - dsa: microchip: make learning configurable and keep it off while
     standalone

   - ice: xsk: prohibit usage of non-balanced queue id

   - rxrpc: fix locking in rxrpc's sendmsg

  Misc:

   - another chunk of sysctl data race silencing"

* tag 'net-6.0-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (87 commits)
  net: lantiq_xrx200: restore buffer if memory allocation failed
  net: lantiq_xrx200: fix lock under memory pressure
  net: lantiq_xrx200: confirm skb is allocated before using
  net: stmmac: work around sporadic tx issue on link-up
  ionic: VF initial random MAC address if no assigned mac
  ionic: fix up issues with handling EAGAIN on FW cmds
  ionic: clear broken state on generation change
  rxrpc: Fix locking in rxrpc's sendmsg
  net: ethernet: mtk_eth_soc: fix hw hash reporting for MTK_NETSYS_V2
  MAINTAINERS: rectify file entry in BONDING DRIVER
  i40e: Fix incorrect address type for IPv6 flow rules
  ixgbe: stop resetting SYSTIME in ixgbe_ptp_start_cyclecounter
  net: Fix a data-race around sysctl_somaxconn.
  net: Fix a data-race around netdev_unregister_timeout_secs.
  net: Fix a data-race around gro_normal_batch.
  net: Fix data-races around sysctl_devconf_inherit_init_net.
  net: Fix data-races around sysctl_fb_tunnels_only_for_init_net.
  net: Fix a data-race around netdev_budget_usecs.
  net: Fix data-races around sysctl_max_skb_frags.
  net: Fix a data-race around netdev_budget.
  ...

Merge branch 'net-lantiq_xrx200-fix-errors-under-memory-pressure'

Aleksander Jan Bajkowski says:

====================
net: lantiq_xrx200: fix errors under memory pressure

This series fixes issues that can occur in the driver under memory pressure.
Situations when the system cannot allocate memory are rare, so the mentioned
bugs have been fixed recently. The patches have been tested on a BT Home
router with the Lantiq xRX200 chipset.

Changelog:
  v3: - removed netdev_err() log from the first patch
  v2:
   - the second patch has been changed, so that under memory pressure situation
     the driver will not receive packets indefinitely regardless of the NAPI budget,
   - the third patch has been added.
====================

Link: https://lore.kernel.org/r/20220824215408.4695-1-olek2@wp.pl
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lantiq_xrx200: restore buffer if memory allocation failed

In a situation where memory allocation fails, an invalid buffer address
is stored. When this descriptor is used again, the system panics in the
build_skb() function when accessing memory.

Fixes: 7ea6cd16f159 ("lantiq: net: fix duplicated skb in rx descriptor ring")
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lantiq_xrx200: fix lock under memory pressure

When the xrx200_hw_receive() function returns -ENOMEM, the NAPI poll
function immediately returns an error.
This is incorrect for two reasons:
* the function terminates without enabling interrupts or scheduling NAPI,
* the error code (-ENOMEM) is returned instead of the number of received
packets.

After the first memory allocation failure occurs, packet reception is
locked due to disabled interrupts from DMA..

Fixes: fe1a56420cf2 ("net: lantiq: Add Lantiq / Intel VRX200 Ethernet driver")
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: lantiq_xrx200: confirm skb is allocated before using

xrx200_hw_receive() assumes build_skb() always works and goes straight
to skb_reserve(). However, build_skb() can fail under memory pressure.

Add a check in case build_skb() failed to allocate and return NULL.

Fixes: e015593573b3 ("net: lantiq_xrx200: convert to build_skb")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: work around sporadic tx issue on link-up

This is a follow-up to the discussion in [0]. It seems to me that
at least the IP version used on Amlogic SoC's sometimes has a problem
if register MAC_CTRL_REG is written whilst the chip is still processing
a previous write. But that's just a guess.
Adding a delay between two writes to this register helps, but we can
also simply omit the offending second write. This patch uses the second
approach and is based on a suggestion from Qi Duan.
Benefit of this approach is that we can save few register writes, also
on not affected chip versions.

[0] https://www.spinics.net/lists/netdev/msg831526.html

Fixes: bfab27a146ed ("stmmac: add the experimental PCI support")
Suggested-by: Qi Duan <qi.duan@amlogic.com>
Suggested-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/e99857ce-bd90-5093-ca8c-8cd480b5a0a2@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

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

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-08-24 (ixgbe, i40e)

This series contains updates to ixgbe and i40e drivers.

Jake stops incorrect resetting of SYSTIME registers when starting
cyclecounter for ixgbe.

Sylwester corrects a check on source IP address when validating destination
for i40e.

* '10GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  i40e: Fix incorrect address type for IPv6 flow rules
  ixgbe: stop resetting SYSTIME in ixgbe_ptp_start_cyclecounter
====================

Link: https://lore.kernel.org/r/20220824193748.874343-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'ionic-bug-fixes'

Shannon Nelson says:

====================
ionic: bug fixes

These are a couple of maintenance bug fixes for the Pensando ionic
networking driver.

Mohamed takes care of a "plays well with others" issue where the
VF spec is a bit vague on VF mac addresses, but certain customers
have come to expect behavior based on other vendor drivers.

Shannon addresses a couple of corner cases seen in internal
stress testing.
====================

Link: https://lore.kernel.org/r/20220824165051.6185-1-snelson@pensando.io
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ionic: VF initial random MAC address if no assigned mac

Assign a random mac address to the VF interface station
address if it boots with a zero mac address in order to match
similar behavior seen in other VF drivers.  Handle the errors
where the older firmware does not allow the VF to set its own
station address.

Newer firmware will allow the VF to set the station mac address
if it hasn't already been set administratively through the PF.
Setting it will also be allowed if the VF has trust.

Fixes: fbb39807e9ae ("ionic: support sr-iov operations")
Signed-off-by: R Mohamed Shah <mohamed@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ionic: fix up issues with handling EAGAIN on FW cmds

In looping on FW update tests we occasionally see the
FW_ACTIVATE_STATUS command fail while it is in its EAGAIN loop
waiting for the FW activate step to finsh inside the FW.  The
firmware is complaining that the done bit is set when a new
dev_cmd is going to be processed.

Doing a clean on the cmd registers and doorbell before exiting
the wait-for-done and cleaning the done bit before the sleep
prevents this from occurring.

Fixes: fbfb8031533c ("ionic: Add hardware init and device commands")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ionic: clear broken state on generation change

There is a case found in heavy testing where a link flap happens just
before a firmware Recovery event and the driver gets stuck in the
BROKEN state.  This comes from the driver getting interrupted by a FW
generation change when coming back up from the link flap, and the call
to ionic_start_queues() in ionic_link_status_check() fails.  This can be
addressed by having the fw_up code clear the BROKEN bit if seen, rather
than waiting for a user to manually force the interface down and then
back up.

Fixes: 9e8eaf8427b6 ("ionic: stop watchdog when in broken state")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

rxrpc: Fix locking in rxrpc's sendmsg

Fix three bugs in the rxrpc's sendmsg implementation:

 (1) rxrpc_new_client_call() should release the socket lock when returning
     an error from rxrpc_get_call_slot().

 (2) rxrpc_wait_for_tx_window_intr() will return without the call mutex
     held in the event that we're interrupted by a signal whilst waiting
     for tx space on the socket or relocking the call mutex afterwards.

     Fix this by: (a) moving the unlock/lock of the call mutex up to
     rxrpc_send_data() such that the lock is not held around all of
     rxrpc_wait_for_tx_window*() and (b) indicating to higher callers
     whether we're return with the lock dropped.  Note that this means
     recvmsg() will not block on this call whilst we're waiting.

 (3) After dropping and regaining the call mutex, rxrpc_send_data() needs
     to go and recheck the state of the tx_pending buffer and the
     tx_total_len check in case we raced with another sendmsg() on the same
     call.

Thinking on this some more, it might make sense to have different locks for
sendmsg() and recvmsg().  There's probably no need to make recvmsg() wait
for sendmsg().  It does mean that recvmsg() can return MSG_EOR indicating
that a call is dead before a sendmsg() to that call returns - but that can
currently happen anyway.

Without fix (2), something like the following can be induced:

WARNING: bad unlock balance detected!
5.16.0-rc6-syzkaller #0 Not tainted
-------------------------------------
syz-executor011/3597 is trying to release lock (&call->user_mutex) at:
[<ffffffff885163a3>] rxrpc_do_sendmsg+0xc13/0x1350 net/rxrpc/sendmsg.c:748
but there are no more locks to release!

other info that might help us debug this:
no locks held by syz-executor011/3597.
...
Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:88 [inline]
 dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
 print_unlock_imbalance_bug include/trace/events/lock.h:58 [inline]
 __lock_release kernel/locking/lockdep.c:5306 [inline]
 lock_release.cold+0x49/0x4e kernel/locking/lockdep.c:5657
 __mutex_unlock_slowpath+0x99/0x5e0 kernel/locking/mutex.c:900
 rxrpc_do_sendmsg+0xc13/0x1350 net/rxrpc/sendmsg.c:748
 rxrpc_sendmsg+0x420/0x630 net/rxrpc/af_rxrpc.c:561
 sock_sendmsg_nosec net/socket.c:704 [inline]
 sock_sendmsg+0xcf/0x120 net/socket.c:724
 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2409
 ___sys_sendmsg+0xf3/0x170 net/socket.c:2463
 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2492
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x44/0xae

[Thanks to Hawkins Jiawei and Khalid Masum for their attempts to fix this]

Fixes: bc5e3a546d55 ("rxrpc: Use MSG_WAITALL to tell sendmsg() to temporarily ignore signals")
Reported-by: syzbot+7f0483225d0c94cb3441@syzkaller.appspotmail.com
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Marc Dionne <marc.dionne@auristor.com>
Tested-by: syzbot+7f0483225d0c94cb3441@syzkaller.appspotmail.com
cc: Hawkins Jiawei <yin31149@gmail.com>
cc: Khalid Masum <khalid.masum.92@gmail.com>
cc: Dan Carpenter <dan.carpenter@oracle.com>
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/166135894583.600315.7170979436768124075.stgit@warthog.procyon.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'Fix incorrect pruning for ARG_CONST_ALLOC_SIZE_OR_ZERO'

Kumar Kartikeya Dwivedi says:

====================

A fix for a missing mark_chain_precision call that leads to eager pruning and
loading of invalid programs when the more permissive case is in the straight
line exploration. Please see the commit log for details, and selftest for an
example.
====================

Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

selftests/bpf: Add regression test for pruning fix

Add a test to ensure we do mark_chain_precision for the argument type
ARG_CONST_ALLOC_SIZE_OR_ZERO. For other argument types, this was already
done, but propagation for missing for this case. Without the fix, this
test case loads successfully.

Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220823185500.467-1-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

bpf: Do mark_chain_precision for ARG_CONST_ALLOC_SIZE_OR_ZERO

Precision markers need to be propagated whenever we have an ARG_CONST_*
style argument, as the verifier cannot consider imprecise scalars to be
equivalent for the purposes of states_equal check when such arguments
refine the return value (in this case, set mem_size for PTR_TO_MEM). The
resultant mem_size for the R0 is derived from the constant value, and if
the verifier incorrectly prunes states considering them equivalent where
such arguments exist (by seeing that both registers have reg->precise as
false in regsafe), we can end up with invalid programs passing the
verifier which can do access beyond what should have been the correct
mem_size in that explored state.

To show a concrete example of the problem:

0000000000000000 <prog>:
       0:       r2 = *(u32 *)(r1 + 80)
       1:       r1 = *(u32 *)(r1 + 76)
       2:       r3 = r1
       3:       r3 += 4
       4:       if r3 > r2 goto +18 <LBB5_5>
       5:       w2 = 0
       6:       *(u32 *)(r1 + 0) = r2
       7:       r1 = *(u32 *)(r1 + 0)
       8:       r2 = 1
       9:       if w1 == 0 goto +1 <LBB5_3>
      10:       r2 = -1

0000000000000058 <LBB5_3>:
      11:       r1 = 0 ll
      13:       r3 = 0
      14:       call bpf_ringbuf_reserve
      15:       if r0 == 0 goto +7 <LBB5_5>
      16:       r1 = r0
      17:       r1 += 16777215
      18:       w2 = 0
      19:       *(u8 *)(r1 + 0) = r2
      20:       r1 = r0
      21:       r2 = 0
      22:       call bpf_ringbuf_submit

00000000000000b8 <LBB5_5>:
      23:       w0 = 0
      24:       exit

For the first case, the single line execution's exploration will prune
the search at insn 14 for the branch insn 9's second leg as it will be
verified first using r2 = -1 (UINT_MAX), while as w1 at insn 9 will
always be 0 so at runtime we don't get error for being greater than
UINT_MAX/4 from bpf_ringbuf_reserve. The verifier during regsafe just
sees reg->precise as false for both r2 registers in both states, hence
considers them equal for purposes of states_equal.

If we propagated precise markers using the backtracking support, we
would use the precise marking to then ensure that old r2 (UINT_MAX) was
within the new r2 (1) and this would never be true, so the verification
would rightfully fail.

The end result is that the out of bounds access at instruction 19 would
be permitted without this fix.

Note that reg->precise is always set to true when user does not have
CAP_BPF (or when subprog count is greater than 1 (i.e. use of any static
or global functions)), hence this is only a problem when precision marks
need to be explicitly propagated (i.e. privileged users with CAP_BPF).

A simplified test case has been included in the next patch to prevent
future regressions.

Fixes: 457f44363a88 ("bpf: Implement BPF ring buffer and verifier support for it")
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220823185300.406-2-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Merge tag 'cgroup-for-6.0-rc2-fixes-2' of git://git./linux/kernel/git/tj/cgroup

Pull another cgroup fix from Tejun Heo:
 "Commit 4f7e7236435c ("cgroup: Fix threadgroup_rwsem <->
  cpus_read_lock() deadlock") required the cgroup
  core to grab cpus_read_lock() before invoking ->attach().

  Unfortunately, it missed adding cpus_read_lock() in
  cgroup_attach_task_all(). Fix it"

* tag 'cgroup-for-6.0-rc2-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup: Add missing cpus_read_lock() to cgroup_attach_task_all()

cgroup: Add missing cpus_read_lock() to cgroup_attach_task_all()

syzbot is hitting percpu_rwsem_assert_held(&cpu_hotplug_lock) warning at
cpuset_attach() [1], for commit 4f7e7236435ca0ab ("cgroup: Fix
threadgroup_rwsem <-> cpus_read_lock() deadlock") missed that
cpuset_attach() is also called from cgroup_attach_task_all().
Add cpus_read_lock() like what cgroup_procs_write_start() does.

Link: https://syzkaller.appspot.com/bug?extid=29d3a3b4d86c8136ad9e
Reported-by: syzbot <syzbot+29d3a3b4d86c8136ad9e@syzkaller.appspotmail.com>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fixes: 4f7e7236435ca0ab ("cgroup: Fix threadgroup_rwsem <-> cpus_read_lock() deadlock")
Signed-off-by: Tejun Heo <tj@kernel.org>

net: ethernet: mtk_eth_soc: fix hw hash reporting for MTK_NETSYS_V2

Properly report hw rx hash for mt7986 chipset accroding to the new dma
descriptor layout.

Fixes: 197c9e9b17b11 ("net: ethernet: mtk_eth_soc: introduce support for mt7986 chipset")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/091394ea4e705fbb35f828011d98d0ba33808f69.1661257293.git.lorenzo@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

wifi: mac80211: potential NULL dereference in ieee80211_tx_control_port()

The ieee80211_lookup_ra_sta() function will sometimes set "sta" to NULL
so add this NULL check to prevent an Oops.

Fixes: 9dd1953846c7 ("wifi: nl80211/mac80211: clarify link ID in control port TX")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YuKcTAyO94YOy0Bu@kili
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wifi: cfg80211: debugfs: fix return type in ht40allow_map_read()

The return type is supposed to be ssize_t, which is signed long,
but "r" was declared as unsigned int.  This means that on 64 bit systems
we return positive values instead of negative error codes.

Fixes: 80a3511d70e8 ("cfg80211: add debugfs HT40 allow map")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YutvOQeJm0UjLhwU@kili
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wifi: mac80211: Don't finalize CSA in IBSS mode if state is disconnected

When we are not connected to a channel, sending channel "switch"
announcement doesn't make any sense.

The BSS list is empty in that case. This causes the for loop in
cfg80211_get_bss() to be bypassed, so the function returns NULL
(check line 1424 of net/wireless/scan.c), causing the WARN_ON()
in ieee80211_ibss_csa_beacon() to get triggered (check line 500
of net/mac80211/ibss.c), which was consequently reported on the
syzkaller dashboard.

Thus, check if we have an existing connection before generating
the CSA beacon in ieee80211_ibss_finish_csa().

Cc: stable@vger.kernel.org
Fixes: cd7760e62c2a ("mac80211: add support for CSA in IBSS mode")
Link: https://syzkaller.appspot.com/bug?id=05603ef4ae8926761b678d2939a3b2ad28ab9ca6
Reported-by: syzbot+b6c9fe29aefe68e4ad34@syzkaller.appspotmail.com
Signed-off-by: Siddh Raman Pant <code@siddh.me>
Tested-by: syzbot+b6c9fe29aefe68e4ad34@syzkaller.appspotmail.com
Link: https://lore.kernel.org/r/20220814151512.9985-1-code@siddh.me
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wifi: mac80211: fix possible leak in ieee80211_tx_control_port()

Add missing dev_kfree_skb() in an error path in
ieee80211_tx_control_port() to avoid a memory leak.

Fixes: dd820ed6336a ("wifi: mac80211: return error from control port TX for drops")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Link: https://lore.kernel.org/r/20220818043349.4168835-1-yangyingliang@huawei.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wifi: mac80211: always free sta in __sta_info_alloc in case of error

Free sta pointer in __sta_info_alloc routine if sta_info_alloc_link()
fails.

Fixes: 246b39e4a1ba5 ("wifi: mac80211: refactor some sta_info link handling")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/a3d079208684cddbc25289f7f7e0fed795b0cad4.1661260857.git.lorenzo@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wifi: mac80211: Fix UAF in ieee80211_scan_rx()

ieee80211_scan_rx() tries to access scan_req->flags after a
null check, but a UAF is observed when the scan is completed
and __ieee80211_scan_completed() executes, which then calls
cfg80211_scan_done() leading to the freeing of scan_req.

Since scan_req is rcu_dereference()'d, prevent the racing in
__ieee80211_scan_completed() by ensuring that from mac80211's
POV it is no longer accessed from an RCU read critical section
before we call cfg80211_scan_done().

Cc: stable@vger.kernel.org
Link: https://syzkaller.appspot.com/bug?extid=f9acff9bf08a845f225d
Reported-by: syzbot+f9acff9bf08a845f225d@syzkaller.appspotmail.com
Suggested-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Siddh Raman Pant <code@siddh.me>
Link: https://lore.kernel.org/r/20220819200340.34826-1-code@siddh.me
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

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

Pablo Neira Ayuso says:

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

1) Fix crash with malformed ebtables blob which do not provide all
   entry points, from Florian Westphal.

2) Fix possible TCP connection clogging up with default 5-days
   timeout in conntrack, from Florian.

3) Fix crash in nf_tables tproxy with unsupported chains, also from Florian.

4) Do not allow to update implicit chains.

5) Make table handle allocation per-netns to fix data race.

6) Do not truncated payload length and offset, and checksum offset.
   Instead report EINVAl.

7) Enable chain stats update via static key iff no error occurs.

8) Restrict osf expression to ip, ip6 and inet families.

9) Restrict tunnel expression to netdev family.

10) Fix crash when trying to bind again an already bound chain.

11) Flowtable garbage collector might leave behind pending work to
    delete entries. This patch comes with a previous preparation patch
    as dependency.

12) Allow net.netfilter.nf_conntrack_frag6_high_thresh to be lowered,
    from Eric Dumazet.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_defrag_ipv6: allow nf_conntrack_frag6_high_thresh increases
  netfilter: flowtable: fix stuck flows on cleanup due to pending work
  netfilter: flowtable: add function to invoke garbage collection immediately
  netfilter: nf_tables: disallow binding to already bound chain
  netfilter: nft_tunnel: restrict it to netdev family
  netfilter: nft_osf: restrict osf to ipv4, ipv6 and inet families
  netfilter: nf_tables: do not leave chain stats enabled on error
  netfilter: nft_payload: do not truncate csum_offset and csum_type
  netfilter: nft_payload: report ERANGE for too long offset and length
  netfilter: nf_tables: make table handle allocation per-netns friendly
  netfilter: nf_tables: disallow updates of implicit chain
  netfilter: nft_tproxy: restrict to prerouting hook
  netfilter: conntrack: work around exceeded receive window
  netfilter: ebtables: reject blobs that don't provide all entry points
====================

Link: https://lore.kernel.org/r/20220824220330.64283-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

MAINTAINERS: rectify file entry in BONDING DRIVER

Commit c078290a2b76 ("selftests: include bonding tests into the kselftest
infra") adds the bonding tests in the directory:

  tools/testing/selftests/drivers/net/bonding/

The file entry in MAINTAINERS for the BONDING DRIVER however refers to:

  tools/testing/selftests/net/bonding/

Hence, ./scripts/get_maintainer.pl --self-test=patterns complains about a
broken file pattern.

Repair this file entry in BONDING DRIVER.

Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Acked-by: Jonathan Toppins <jtoppins@redhat.com>
Link: https://lore.kernel.org/r/20220824072945.28606-1-lukas.bulwahn@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

i40e: Fix incorrect address type for IPv6 flow rules

It was not possible to create 1-tuple flow director
rule for IPv6 flow type. It was caused by incorrectly
checking for source IP address when validating user provided
destination IP address.

Fix this by changing ip6src to correct ip6dst address
in destination IP address validation for IPv6 flow type.

Fixes: efca91e89b67 ("i40e: Add flow director support for IPv6")
Signed-off-by: Sylwester Dziedziuch <sylwesterx.dziedziuch@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ixgbe: stop resetting SYSTIME in ixgbe_ptp_start_cyclecounter

The ixgbe_ptp_start_cyclecounter is intended to be called whenever the
cyclecounter parameters need to be changed.

Since commit a9763f3cb54c ("ixgbe: Update PTP to support X550EM_x
devices"), this function has cleared the SYSTIME registers and reset the
TSAUXC DISABLE_SYSTIME bit.

While these need to be cleared during ixgbe_ptp_reset, it is wrong to clear
them during ixgbe_ptp_start_cyclecounter. This function may be called
during both reset and link status change. When link changes, the SYSTIME
counter is still operating normally, but the cyclecounter should be updated
to account for the possibly changed parameters.

Clearing SYSTIME when link changes causes the timecounter to jump because
the cycle counter now reads zero.

Extract the SYSTIME initialization out to a new function and call this
during ixgbe_ptp_reset. This prevents the timecounter adjustment and avoids
an unnecessary reset of the current time.

This also restores the original SYSTIME clearing that occurred during
ixgbe_ptp_reset before the commit above.

Reported-by: Steve Payne <spayne@aurora.tech>
Reported-by: Ilya Evenbach <ievenbach@aurora.tech>
Fixes: a9763f3cb54c ("ixgbe: Update PTP to support X550EM_x devices")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

Merge tag 'trace-v6.0-rc2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:

 - Fix build warning for when MODULES and FTRACE_WITH_DIRECT_CALLS are
   not set. A warning happens with ops_references_rec() defined but not
   used.

* tag 'trace-v6.0-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  ftrace: Fix build warning for ops_references_rec() not used

Merge branch 'dmi-for-linus' of git://git./linux/kernel/git/jdelvare/staging

Pull dmi update from Jean Delvare.

Tiny cleanup.

* 'dmi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging:
  firmware: dmi: Use the proper accessor for the version field

Merge branch 'sysctl-data-races'

Kuniyuki Iwashima says:

====================
net: sysctl: Fix data-races around net.core.XXX

This series fixes data-races around all knobs in net_core_table and
netns_core_table except for bpf stuff.

These knobs are skipped:

  - 4 bpf knobs
  - netdev_rss_key: Written only once by net_get_random_once() and
                    read-only knob
  - rps_sock_flow_entries: Protected with sock_flow_mutex
  - flow_limit_cpu_bitmap: Protected with flow_limit_update_mutex
  - flow_limit_table_len: Protected with flow_limit_update_mutex
  - default_qdisc: Protected with qdisc_mod_lock
  - warnings: Unused
  - high_order_alloc_disable: Protected with static_key_mutex
  - skb_defer_max: Already using READ_ONCE()
  - sysctl_txrehash: Already using READ_ONCE()

Note 5th patch fixes net.core.message_cost and net.core.message_burst,
and lib/ratelimit.c does not have an explicit maintainer.

Changes:
  v3:
    * Fix build failures of CONFIG_SYSCTL=n case in 13th & 14th patches

  v2: https://lore.kernel.org/netdev/20220818035227.81567-1-kuniyu@amazon.com/
    * Remove 4 bpf knobs and added 6 knobs

  v1: https://lore.kernel.org/netdev/20220816052347.70042-1-kuniyu@amazon.com/
====================

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

net: Fix a data-race around sysctl_somaxconn.

While reading sysctl_somaxconn, 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>

net: Fix a data-race around netdev_unregister_timeout_secs.

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

Fixes: 5aa3afe107d9 ("net: make unregister netdev warning timeout configurable")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix a data-race around gro_normal_batch.

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

Fixes: 323ebb61e32b ("net: use listified RX for handling GRO_NORMAL skbs")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix data-races around sysctl_devconf_inherit_init_net.

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

Fixes: 856c395cfa63 ("net: introduce a knob to control whether to inherit devconf config")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix data-races around sysctl_fb_tunnels_only_for_init_net.

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

Fixes: 79134e6ce2c9 ("net: do not create fallback tunnels for non-default namespaces")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix a data-race around netdev_budget_usecs.

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

Fixes: 7acf8a1e8a28 ("Replace 2 jiffies with sysctl netdev_budget_usecs to enable softirq tuning")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix data-races around sysctl_max_skb_frags.

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

Fixes: 5f74f82ea34c ("net:Add sysctl_max_skb_frags")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix a data-race around netdev_budget.

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

Fixes: 51b0bdedb8e7 ("[NET]: Separate two usages of netdev_max_backlog.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix a data-race around sysctl_net_busy_read.

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

Fixes: 2d48d67fa8cd ("net: poll/select low latency socket support")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix a data-race around sysctl_net_busy_poll.

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

Fixes: 060212928670 ("net: add low latency socket poll")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix a data-race around sysctl_tstamp_allow_data.

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

Fixes: b245be1f4db1 ("net-timestamp: no-payload only sysctl")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix data-races around sysctl_optmem_max.

While reading sysctl_optmem_max, 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>

ratelimit: Fix data-races in ___ratelimit().

While reading rs->interval and rs->burst, they can be changed
concurrently via sysctl (e.g. net_ratelimit_state).  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>

net: Fix data-races around netdev_tstamp_prequeue.

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

Fixes: 3b098e2d7c69 ("net: Consistent skb timestamping")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix data-races around netdev_max_backlog.

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

While at it, we remove the unnecessary spaces in the doc.

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

net: Fix data-races around weight_p and dev_weight_[rt]x_bias.

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

Also, dev_[rt]x_weight can be read/written at the same time.  So, we
need to use READ_ONCE() and WRITE_ONCE() for its access.  Moreover, to
use the same weight_p while changing dev_[rt]x_weight, we add a mutex
in proc_do_dev_weight().

Fixes: 3d48b53fb2ae ("net: dev_weight: TX/RX orthogonality")
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix data-races around sysctl_[rw]mem_(max|default).

While reading sysctl_[rw]mem_(max|default), they 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>

net/core/skbuff: Check the return value of skb_copy_bits()

skb_copy_bits() could fail, which requires a check on the return
value.

Signed-off-by: Li Zhong <floridsleeves@gmail.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 2022-08-24

1) Fix a refcount leak in __xfrm_policy_check.
   From Xin Xiong.

2) Revert "xfrm: update SA curlft.use_time". This
   violates RFC 2367. From Antony Antony.

3) Fix a comment on XFRMA_LASTUSED.
   From Antony Antony.

4) x->lastused is not cloned in xfrm_do_migrate.
   Fix from Antony Antony.

5) Serialize the calls to xfrm_probe_algs.
   From Herbert Xu.

6) Fix a null pointer dereference of dst->dev on a metadata
   dst in xfrm_lookup_with_ifid. From Nikolay Aleksandrov.

Please pull or let me know if there are problems.
====================

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

fec: Restart PPS after link state change

On link state change, the controller gets reset,
causing PPS to drop out and the PHC to lose its
time and calibration. So we restart it if needed,
restoring calibration and time registers.

Changes since v2:
* Add `fec_ptp_save_state()`/`fec_ptp_restore_state()`
* Use `ktime_get_real_ns()`
* Use `BIT()` macro
Changes since v1:
* More ECR #define's
* Stop PPS in `fec_ptp_stop()`

Signed-off-by: Csókás Bence <csokas.bence@prolan.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: neigh: don't call kfree_skb() under spin_lock_irqsave()

It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So add all skb to
a tmp list, then free them after spin_unlock_irqrestore() at
once.

Fixes: 66ba215cb513 ("neigh: fix possible DoS due to net iface start/stop loop")
Suggested-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: nf_defrag_ipv6: allow nf_conntrack_frag6_high_thresh increases

Currently, net.netfilter.nf_conntrack_frag6_high_thresh can only be lowered.

I found this issue while investigating a probable kernel issue
causing flakes in tools/testing/selftests/net/ip_defrag.sh

In particular, these sysctl changes were ignored:
ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_high_thresh=9000000 >/dev/null 2>&1
ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_low_thresh=7000000  >/dev/null 2>&1

This change is inline with commit 836196239298 ("net/ipfrag: let ip[6]frag_high_thresh
in ns be higher than in init_net")

Fixes: 8db3d41569bb ("netfilter: nf_defrag_ipv6: use net_generic infra")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: flowtable: fix stuck flows on cleanup due to pending work

To clear the flow table on flow table free, the following sequence
normally happens in order:

  1) gc_step work is stopped to disable any further stats/del requests.
  2) All flow table entries are set to teardown state.
  3) Run gc_step which will queue HW del work for each flow table entry.
  4) Waiting for the above del work to finish (flush).
  5) Run gc_step again, deleting all entries from the flow table.
  6) Flow table is freed.

But if a flow table entry already has pending HW stats or HW add work
step 3 will not queue HW del work (it will be skipped), step 4 will wait
for the pending add/stats to finish, and step 5 will queue HW del work
which might execute after freeing of the flow table.

To fix the above, this patch flushes the pending work, then it sets the
teardown flag to all flows in the flowtable and it forces a garbage
collector run to queue work to remove the flows from hardware, then it
flushes this new pending work and (finally) it forces another garbage
collector run to remove the entry from the software flowtable.

Stack trace:
[47773.882335] BUG: KASAN: use-after-free in down_read+0x99/0x460
[47773.883634] Write of size 8 at addr ffff888103b45aa8 by task kworker/u20:6/543704
[47773.885634] CPU: 3 PID: 543704 Comm: kworker/u20:6 Not tainted 5.12.0-rc7+ #2
[47773.886745] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009)
[47773.888438] Workqueue: nf_ft_offload_del flow_offload_work_handler [nf_flow_table]
[47773.889727] Call Trace:
[47773.890214]  dump_stack+0xbb/0x107
[47773.890818]  print_address_description.constprop.0+0x18/0x140
[47773.892990]  kasan_report.cold+0x7c/0xd8
[47773.894459]  kasan_check_range+0x145/0x1a0
[47773.895174]  down_read+0x99/0x460
[47773.899706]  nf_flow_offload_tuple+0x24f/0x3c0 [nf_flow_table]
[47773.907137]  flow_offload_work_handler+0x72d/0xbe0 [nf_flow_table]
[47773.913372]  process_one_work+0x8ac/0x14e0
[47773.921325]
[47773.921325] Allocated by task 592159:
[47773.922031]  kasan_save_stack+0x1b/0x40
[47773.922730]  __kasan_kmalloc+0x7a/0x90
[47773.923411]  tcf_ct_flow_table_get+0x3cb/0x1230 [act_ct]
[47773.924363]  tcf_ct_init+0x71c/0x1156 [act_ct]
[47773.925207]  tcf_action_init_1+0x45b/0x700
[47773.925987]  tcf_action_init+0x453/0x6b0
[47773.926692]  tcf_exts_validate+0x3d0/0x600
[47773.927419]  fl_change+0x757/0x4a51 [cls_flower]
[47773.928227]  tc_new_tfilter+0x89a/0x2070
[47773.936652]
[47773.936652] Freed by task 543704:
[47773.937303]  kasan_save_stack+0x1b/0x40
[47773.938039]  kasan_set_track+0x1c/0x30
[47773.938731]  kasan_set_free_info+0x20/0x30
[47773.939467]  __kasan_slab_free+0xe7/0x120
[47773.940194]  slab_free_freelist_hook+0x86/0x190
[47773.941038]  kfree+0xce/0x3a0
[47773.941644]  tcf_ct_flow_table_cleanup_work

Original patch description and stack trace by Paul Blakey.

Fixes: c29f74e0df7a ("netfilter: nf_flow_table: hardware offload support")
Reported-by: Paul Blakey <paulb@nvidia.com>
Tested-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: flowtable: add function to invoke garbage collection immediately

Expose nf_flow_table_gc_run() to force a garbage collector run from the
offload infrastructure.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: nf_tables: disallow binding to already bound chain

Update nft_data_init() to report EINVAL if chain is already bound.

Fixes: d0e2c7de92c7 ("netfilter: nf_tables: add NFT_CHAIN_BINDING")
Reported-by: Gwangun Jung <exsociety@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>