net: phy: respect cached advertising when re-enabling EEE

If we remove modes from EEE advertisement and disable / re-enable EEE,
then advertisement is set to all supported modes. I don't think this is
what the user expects. So respect the cached advertisement and just fall
back to all supported modes if cached advertisement is empty.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://patch.msgid.link/c75f7f8b-5571-429f-abd3-ce682d178a4b@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-add-debug-checks-to-skb_reset_xxx_header'

Eric Dumazet says:

====================
net: add debug checks to skb_reset_xxx_header()

Add debug checks (only enabled for CONFIG_DEBUG_NET=y builds),
to catch bugs earlier.
====================

Link: https://patch.msgid.link/20241105174403.850330-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: add debug check in skb_reset_mac_header()

Make sure (skb->data - skb->head) can fit in skb->mac_header

This needs CONFIG_DEBUG_NET=y.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-8-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: add debug check in skb_reset_network_header()

Make sure (skb->data - skb->head) can fit in skb->network_header

This needs CONFIG_DEBUG_NET=y.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: add debug check in skb_reset_transport_header()

Make sure (skb->data - skb->head) can fit in skb->transport_header

This needs CONFIG_DEBUG_NET=y.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-6-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: add debug check in skb_reset_inner_mac_header()

Make sure (skb->data - skb->head) can fit in skb->inner_mac_header

This needs CONFIG_DEBUG_NET=y.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: add debug check in skb_reset_inner_network_header()

Make sure (skb->data - skb->head) can fit in skb->inner_network_header

This needs CONFIG_DEBUG_NET=y.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: add debug check in skb_reset_inner_transport_header()

Make sure (skb->data - skb->head) can fit in skb->inner_transport_header

This needs CONFIG_DEBUG_NET=y.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: skb_reset_mac_len() must check if mac_header was set

Recent discussions show that skb_reset_mac_len() should be more careful.

We expect the MAC header being set.

If not, clear skb->mac_len and fire a warning for CONFIG_DEBUG_NET=y builds.

If after investigations we find that not having a MAC header was okay,
we can remove the warning.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/netdev/CANn89iJZGH+yEfJxfPWa3Hm7jxb-aeY2Up4HufmLMnVuQXt38A@mail.gmail.com/T/
Cc: En-Wei Wu <en-wei.wu@canonical.com>
Reviewed-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://patch.msgid.link/20241105174403.850330-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'ipv6-fix-hangup-on-device-removal'

Paolo Abeni says:

====================
ipv6: fix hangup on device removal

This addresses the infamous unregister_netdevice splat in net selftests;
the actual fix is carried by the first patch, while the 2nd one
addresses a related problem in the relevant test that was patially
hiding the problem.

Targeting net-next as the issue is quite old and I feel a little lost
in the fib info/nh jungle.
====================

Link: https://patch.msgid.link/cover.1730828007.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests: net: really check for bg process completion

A recent refactor transformed the check for process completion
in a true statement, due to a typo.

As a result, the relevant test-case is unable to catch the
regression it was supposed to detect.

Restore the correct condition.

Fixes: 691bb4e49c98 ("selftests: net: avoid just another constant wait")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/0e6f213811f8e93a235307e683af8225cc6277ae.1730828007.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ipv6: release nexthop on device removal

The CI is hitting some aperiodic hangup at device removal time in the
pmtu.sh self-test:

unregister_netdevice: waiting for veth_A-R1 to become free. Usage count = 6
ref_tracker: veth_A-R1@ffff888013df15d8 has 1/5 users at
dst_init+0x84/0x4a0
dst_alloc+0x97/0x150
ip6_dst_alloc+0x23/0x90
ip6_rt_pcpu_alloc+0x1e6/0x520
ip6_pol_route+0x56f/0x840
fib6_rule_lookup+0x334/0x630
ip6_route_output_flags+0x259/0x480
ip6_dst_lookup_tail.constprop.0+0x5c2/0x940
ip6_dst_lookup_flow+0x88/0x190
udp_tunnel6_dst_lookup+0x2a7/0x4c0
vxlan_xmit_one+0xbde/0x4a50 [vxlan]
vxlan_xmit+0x9ad/0xf20 [vxlan]
dev_hard_start_xmit+0x10e/0x360
__dev_queue_xmit+0xf95/0x18c0
arp_solicit+0x4a2/0xe00
neigh_probe+0xaa/0xf0

While the first suspect is the dst_cache, explicitly tracking the dst
owing the last device reference via probes proved such dst is held by
the nexthop in the originating fib6_info.

Similar to commit f5b51fe804ec ("ipv6: route: purge exception on
removal"), we need to explicitly release the originating fib info when
disconnecting a to-be-removed device from a live ipv6 dst: move the
fib6_info cleanup into ip6_dst_ifdown().

Tested running:

./pmtu.sh cleanup_ipv6_exception

in a tight loop for more than 400 iterations with no spat, running an
unpatched kernel  I observed a splat every ~10 iterations.

Fixes: f88d8ea67fbd ("ipv6: Plumb support for nexthop object in a fib6_info")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/604c45c188c609b732286b47ac2a451a40f6cf6d.1730828007.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests: netfilter: nft_queue.sh: fix warnings with socat 1.8.0.0

Updated to a more recent socat release and saw this:
 socat E xioopen_ipdgram_listen(): unknown address family 0
 socat W address is opened in read-write mode but only supports read-only

First error is avoided via pf=ipv4 option, second one via -u
(unidirectional) mode.

Signed-off-by: Florian Westphal <fw@strlen.de>
Link: https://patch.msgid.link/20241104142821.2608-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests: netfilter: run conntrack_dump_flush in netns

This test will fail if the initial namespace has conntrack
active due to unexpected number of flows returned on dump:

  conntrack_dump_flush.c:451:test_flush_by_zone:Expected ret (7) == 2 (2)
  test_flush_by_zone: Test failed
  FAIL  conntrack_dump_flush.test_flush_by_zone
  not ok 2 conntrack_dump_flush.test_flush_by_zone

Add a wrapper that unshares this program to avoid this problem.

Signed-off-by: Florian Westphal <fw@strlen.de>
Link: https://patch.msgid.link/20241104142529.2352-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

mptcp: remove unneeded lock when listing scheds

mptcp_get_available_schedulers() needs to iterate over the schedulers'
list only to read the names: it doesn't modify anything there.

In this case, it is enough to hold the RCU read lock, no need to combine
this with the associated spin lock as it was done since its introduction
in commit 73c900aa3660 ("mptcp: add net.mptcp.available_schedulers").

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Geliang Tang <geliang@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20241104-net-next-mptcp-sched-unneeded-lock-v2-1-2ccc1e0c750c@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'add-the-dwmac-driver-support-for-t-head-th1520-soc'

Drew Fustini says:

====================
Add the dwmac driver support for T-HEAD TH1520 SoC

This series adds support for dwmac gigabit ethernet in the T-Head TH1520
RISC-V SoC used on boards like BeagleV Ahead and the LicheePi 4A.

The gigabit ethernet on these boards does need pinctrl support to mux
the necessary pads. The pinctrl-th1520 driver, pinctrl binding, and
related dts patches are in linux-next. However, they are not yet in
net-next/main.

Therefore, I am dropping the dts patch for v5 as it will not build on
net-next/main due to the lack of the padctrl0_apsys pin controller node
in next-next/main version th1520.dtsi. It does exist in linux-next [1]
and the two patches in this series allow the ethernet ports to work
correctly on the LPi4A and Ahead when applied to linux-next.

The dwmac-thead driver in this series does not need the pinctrl-th1520
driver to build. Nor does the thead,th1520-gmac.yaml binding need the
pinctrl binding to pass the schema check.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/arch/riscv/boot/dts/thead/th1520.dtsi
====================

Link: https://patch.msgid.link/20241103-th1520-gmac-v7-0-ef094a30169c@tenstorrent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: Add glue layer for T-HEAD TH1520 SoC

Add dwmac glue driver to support the DesignWare-based GMAC controllers
on the T-HEAD TH1520 SoC.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
Signed-off-by: Drew Fustini <dfustini@tenstorrent.com>
Link: https://patch.msgid.link/20241103-th1520-gmac-v7-2-ef094a30169c@tenstorrent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dt-bindings: net: Add T-HEAD dwmac support

Add documentation to describe the DesginWare-based GMAC controllers in
the T-HEAD TH1520 SoC.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
Signed-off-by: Drew Fustini <dfustini@tenstorrent.com>
Link: https://patch.msgid.link/20241103-th1520-gmac-v7-1-ef094a30169c@tenstorrent.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

r8169: remove leftover locks after reverted change

After e31a9fedc7d8 ("Revert "r8169: disable ASPM during NAPI poll"")
these locks aren't needed any longer.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://patch.msgid.link/680f2606-ac7d-4ced-8694-e5033855da9b@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'add-support-for-synopsis-designware-version-3-72a'

Lothar Rubusch says:

====================
Add support for Synopsis DesignWare version 3.72a

Add compatibility and dt-binding for Synopsis DesignWare version 3.72a.
The dwmac is used on some older Altera/Intel SoCs such as Arria10.
Updating compatibles in the driver and bindings for the DT improves the
binding check coverage for such SoCs.
====================

Link: https://patch.msgid.link/20241102114122.4631-1-l.rubusch@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dt-bindings: net: snps,dwmac: add support for Arria10

The hard processor system (HPS) on the Intel/Altera Arria10 provides
three Ethernet Media Access Controller (EMAC) peripherals. Each EMAC
can be used to transmit and receive data at 10/100/1000 Mbps over
ethernet connections in compliance with the IEEE 802.3 specification.
The EMACs on the Arria10 are instances of the Synopsis DesignWare
Universal 10/100/1000 Ethernet MAC, version 3.72a.

Support the Synopsis DesignWare version 3.72a, which is used in Intel's
Arria10 SoC, since it was missing.

Signed-off-by: Lothar Rubusch <l.rubusch@gmail.com>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://patch.msgid.link/20241102114122.4631-3-l.rubusch@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: add support for dwmac 3.72a

The dwmac 3.72a is an ip version that can be found on Intel/Altera Arria10
SoCs. Going by the hardware features "snps,multicast-filter-bins" and
"snps,perfect-filter-entries" shall be supported. Thus add a
compatibility flag, and extend coverage of the driver for the 3.72a.

Signed-off-by: Lothar Rubusch <l.rubusch@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20241102114122.4631-2-l.rubusch@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: hisilicon: hns: use ethtool string helpers

The latter is the preferred way to copy ethtool strings.

Avoids manually incrementing the pointer. Cleans up the code quite well.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Reviewed-by: Jijie Shao <shaojijie@huawei.com>
Link: https://patch.msgid.link/20241101220023.290926-1-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

openvswitch: Pass on secpath details for internal port rx.

Clearing the secpath for internal ports will cause packet drops when
ipsec offload or early SW ipsec decrypt are used.  Systems that rely
on these will not be able to actually pass traffic via openvswitch.

There is still an open issue for a flow miss packet - this is because
we drop the extensions during upcall and there is no facility to
restore such data (and it is non-trivial to add such functionality
to the upcall interface).  That means that when a flow miss occurs,
there will still be packet drops.  With this patch, when a flow is
found then traffic which has an associated xfrm extension will
properly flow.

Signed-off-by: Aaron Conole <aconole@redhat.com>
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Link: https://patch.msgid.link/20241101204732.183840-1-aconole@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

r8169: improve initialization of RSS registers on RTL8125/RTL8126

Replace the register addresses with the names used in r8125/r8126
vendor driver, and consider that RSS_CTRL_8125 is a 32 bit register.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://patch.msgid.link/3bf2f340-b369-4174-97bf-fd38d4217492@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'a-pile-of-sfc-deadcode'

Dr. David Alan Gilbert says:

====================
A pile of sfc deadcode

This is a collection of deadcode removal in the sfc
drivers;  the split is vaguely where I found them in
the tree, with some left over.

This has been build tested and booted on an x86 VM,
but I fon't have the hardware to test; however
it's all full function removal.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
====================

Link: https://patch.msgid.link/20241102151625.39535-1-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: Remove more unused functions

efx_ticks_to_usecs(), efx_reconfigure_port(), efx_ptp_get_mode(), and
efx_tx_get_copy_buffer_limited() are unused.
They seem to be partially due to the later splits to Siena, but
some seem unused for longer.

Remove them.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://patch.msgid.link/20241102151625.39535-5-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: Remove unused mcdi functions

efx_mcdi_flush_rxqs(), efx_mcdi_rpc_async_quiet(),
efx_mcdi_rpc_finish_quiet(), and efx_mcdi_wol_filter_get_magic()
are unused.
I think these are fall out from the split into Siena
that happened in
commit 4d49e5cd4b09 ("sfc/siena: Rename functions in mcdi headers to avoid
conflicts with sfc")
and
commit d48523cb88e0 ("sfc: Copy shared files needed for Siena (part 2)")

Remove them.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://patch.msgid.link/20241102151625.39535-4-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: Remove unused efx_mae_mport_vf

efx_mae_mport_vf() has been unused since
commit 5227adff37af ("sfc: add mport lookup based on driver's mport data")

Remove it.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://patch.msgid.link/20241102151625.39535-3-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: Remove falcon deadcode

ef4_farch_dimension_resources(), ef4_nic_fix_nodesc_drop_stat(),
ef4_ticks_to_usecs() and ef4_tx_get_copy_buffer_limited() were
copied over from efx_ equivalents in 2016 but never used by
commit 5a6681e22c14 ("sfc: separate out SFC4000 ("Falcon") support into new
sfc-falcon driver")

EF4_MAX_FLUSH_TIME is also unused.

Remove them.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://patch.msgid.link/20241102151625.39535-2-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netlink: typographical error in nlmsg_type constants definition

This commit fix a typographical error in netlink nlmsg_type constants definition in the include/uapi/linux/rtnetlink.h at line 177. The definition is RTM_NEWNVLAN RTM_NEWVLAN instead of RTM_NEWVLAN RTM_NEWVLAN.

Signed-off-by: Maurice Lambert <mauricelambert434@gmail.com>
Fixes: 8dcea187088b ("net: bridge: vlan: add rtm definitions and dump support")
Link: https://patch.msgid.link/20241103223950.230300-1-mauricelambert434@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: replace PTP spinlock with seqlock

We can see high contention on ptp_lock while doing RX timestamping
on high packet rates over several queues. Spinlock is not effecient
to protect timecounter for RX timestamps when reads are the most
usual operations and writes are only occasional. It's better to use
seqlock in such cases.

Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
Link: https://patch.msgid.link/20241103215108.557531-2-vadfed@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: cache only 24 bits of hw counter

This hardware can provide only 48 bits of cycle counter. We can leave
only 24 bits in the cache to extend RX timestamps from 32 bits to 48
bits. Lower 8 bits of the cached value will be used to check for
roll-over while extending to full 48 bits.
This change makes cache writes atomic even on 32 bit platforms and we
can simply use READ_ONCE()/WRITE_ONCE() pair and remove spinlock. The
configuration structure will be also reduced by 4 bytes.

Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
Link: https://patch.msgid.link/20241103215108.557531-1-vadfed@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests: net: include lib/sh/*.sh with lib.sh

Recently, the net/lib.sh file has been modified to include defer.sh from
net/lib/sh/ directory. The Makefile from net/lib has been modified
accordingly, but not the ones from the sub-targets using net/lib.sh.

Because of that, the new file is not installed as expected when
installing the Forwarding, MPTCP, and Netfilter targets, e.g.

  # make -C tools/testing/selftests TARGETS=net/mptcp install \
        INSTALL_PATH=/tmp/kself
  # cd /tmp/kself/
  # ./run_kselftest.sh -c net/mptcp
    TAP version 13
    1..7
    # timeout set to 1800
    # selftests: net/mptcp: mptcp_connect.sh
    # ./../lib.sh: line 5: /tmp/kself/net/lib/sh/defer.sh: No such file
      or directory
    # (...)

This can be fixed simply by adding all the .sh files from net/lib/sh
directory to the TEST_INCLUDES variable in the different Makefile's.

Fixes: a6e263f125cd ("selftests: net: lib: Introduce deferred commands")
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Link: https://patch.msgid.link/20241104-net-next-selftests-lib-sh-deps-v1-1-7c9f7d939fc2@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

mlx5_en: use read sequence for gettimex64

The gettimex64() doesn't modify values in timecounter, that's why there
is no need to update sequence counter. Reduce the contention on sequence
lock for multi-thread PHC reading use-case.

Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
Reviewed-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Acked-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20241014170103.2473580-1-vadfed@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-lan969x-add-vcap-functionality'

Daniel Machon says:

====================
net: lan969x: add VCAP functionality

== Description:

This series is the third of a multi-part series, that prepares and adds
support for the new lan969x switch driver.

The upstreaming efforts is split into multiple series (might change a
bit as we go along):

        1) Prepare the Sparx5 driver for lan969x (merged)

        2) Add support for lan969x (same basic features as Sparx5
           provides excl. FDMA and VCAP, merged).

    --> 3) Add lan969x VCAP functionality.

        4) Add RGMII and FDMA functionality.

== VCAP support:

The Versatile Content-Aware Processor (VCAP) is a content-aware packet
processor that allows wirespeed packet inspection for rich
implementation of, for example, advanced VLAN and QoS classification and
manipulations, IP source guarding, longest prefix matching for Layer-3
routing, and security features for wireline and wireless applications.
This is all achieved by programming rules into the VCAP.

When a VCAP is enabled, every frame passing through the switch is
analyzed and multiple keys are created based on the contents of the
frame. The frame is examined to determine the frame type (for example,
IPv4 TCP frame), so that the frame information is extracted according to
the frame type, port-specific configuration, and classification results
from the basic classification. Keys are applied to the VCAP and when
there is a match between a key and a rule in the VCAP, the rule is then
applied to the frame from which the key was extracted.

After this series is applied, the lan969x driver will support the same
VCAP functionality as Sparx5.

== Patch breakdown:

Patch #1 exposes some VCAP symbols for lan969x.

Patch #2 replaces VCAP uses of SPX5_PORTS with n_ports from the match
data.

Patch #3 adds new VCAP constants to match data

Patch #4 removes the is_sparx5() check to now initialize the VCAP API on
lan969x.

Patch #5 adds the auto-generated VCAP data for lan969x.

Patch #6 adds the VCAP configuration data for lan969x.

Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
====================

Link: https://patch.msgid.link/20241101-sparx5-lan969x-switch-driver-3-v1-0-3c76f22f4bfa@microchip.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: lan969x: add VCAP configuration data

Add configuration data (for consumption by the VCAP API) for the four
VCAP's that we are going to support. The following VCAP's will be
supported:

 - VCAP CLM: (also known as IS0) is part of the analyzer and enables
   frame classification using VCAP functionality.

 - VCAP IS2: is part of ANA_ACL and enables access control lists, using
   VCAP functionality.

 - VCAP ES0: is part of the rewriter and enables rewriting of frames
   using VCAP functionality.

 - VCAP ES2: is part of EACL and enables egress access control lists
   using VCAP functionality

The two VCAP's: CLM and IS2 use shared resources from the SUPER VCAP.
The SUPER VCAP is a shared pool of 6 blocks that can be distributed
freely among CLM and IS2.  Each block in the pool has 3,072 addresses
with entries, actions, and counters. ES0 and ES2 does not use shared
resources.

In the configuration data for lan969x CLM uses blocks 2-4 with a total
of 6 lookups. IS2 uses blocks 0-1 with a total of 4 lookups.

Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Reviewed-by: Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com>
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: lan969x: add autogenerated VCAP information

Platform VCAP data for each VCAP instance is auto-generated using an
internal Microchip tool. The generated VCAP data contains information
about keyfields, keyfield sets, actionfields, actionfield sets and
typegroups, which in combination are used to encode and decode rules in
the VCAP.

Add the auto-generated VCAP file lan969x_vcap_ag_api.c and assign the
two structs: lan969x_vcaps and lan969x_vcap_stats to the match data.

Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Reviewed-by: Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com>
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sparx5: execute sparx5_vcap_init() on lan969x

The is_sparx5() check was introduced in an earlier series, to make sure
the sparx5_vcap_init() was not executed on lan969x, as it was not
implemented there yet. Now that it is, remove that check.

Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Reviewed-by: Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com>
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sparx5: add new VCAP constants to match data

In preparation for lan969x VCAP support, add the following three new
VCAP constants to match data:

    - vcaps_cfg (contains configuration data for each VCAP).

    - vcaps (contains auto-generated information about VCAP keys and
      actions).

    - vcap_stats: (contains auto-generated string names of all the keys
      and actions)

Add these constants to the Sparx5 match data constants and use them to
initialize the VCAP's in sparx5_vcap_init().

Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Reviewed-by: Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com>
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sparx5: replace SPX5_PORTS with n_ports

The Sparx5 VCAP implementation uses the SPX5_PORTS symbol to iterate over
the 65 front ports of Sparx5. Replace the use with the n_ports constant
from the match data, which translates to 65 of Sparx5 and 30 on lan969x.

Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Reviewed-by: Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com>
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

net: sparx5: expose some sparx5 VCAP symbols

In preparation for lan969x VCAP support, expose the following symbols for
use by the lan969x VCAP implementation:

- The symbols SPARX5_*_LOOKUPS defines the number of lookups in each
  VCAP instance.  These are the same for lan969x. Move them to the
  header file.

- The struct sparx5_vcap_inst encapsulates information about a single
  VCAP instance. Move this struct to the header file and declare the
  sparx5_vcap_inst_cfg as extern.

Reviewed-by: Steen Hegelund <Steen.Hegelund@microchip.com>
Reviewed-by: Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com>
Signed-off-by: Daniel Machon <daniel.machon@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Merge branch 'virtio_net-enable-premapped-mode-by-default'

Xuan Zhuo says:

====================
virtio_net: enable premapped mode by default

v1:
    1. fix some small problems
    2. remove commit "virtio_net: introduce vi->mode"

In the last linux version, we disabled this feature to fix the
regress[1].

The patch set is try to fix the problem and re-enable it.

More info: http://lore.kernel.org/all/20240820071913.68004-1-xuanzhuo@linux.alibaba.com

[1]: http://lore.kernel.org/all/8b20cc28-45a9-4643-8e87-ba164a540c0a@oracle.com
====================

Link: https://patch.msgid.link/20241029084615.91049-1-xuanzhuo@linux.alibaba.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

virtio_net: rx remove premapped failover code

Now, the premapped mode can be enabled unconditionally.

So we can remove the failover code for merge and small mode.

The virtnet_rq_xxx() helper would be only used if the mode is using pre
mapping. A check is added to prevent misusing of these API.

Tested-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

virtio_net: enable premapped mode for merge and small by default

Currently, the virtio core will perform a dma operation for each
buffer. Although, the same page may be operated multiple times.

In premapped mod, we can perform only one dma operation for the pages of
the alloc frag. This is beneficial for the iommu device.

kernel command line: intel_iommu=on iommu.passthrough=0

       |  strict=0  | strict=1
Before |  775496pps | 428614pps
After  | 1109316pps | 742853pps

In the 6.11, we disabled this feature because a regress [1].

Now, we fix the problem and re-enable it.

[1]: http://lore.kernel.org/all/8b20cc28-45a9-4643-8e87-ba164a540c0a@oracle.com

Tested-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

virtio_net: big mode skip the unmap check

The virtio-net big mode did not enable premapped mode,
so we did not need to check the unmap. And the subsequent
commit will remove the failover code for failing enable
premapped for merge and small mode. So we need to remove
the checking do_dma code in the big mode path.

Tested-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

virtio-net: fix overflow inside virtnet_rq_alloc

When the frag just got a page, then may lead to regression on VM.
Specially if the sysctl net.core.high_order_alloc_disable value is 1,
then the frag always get a page when do refill.

Which could see reliable crashes or scp failure (scp a file 100M in size
to VM).

The issue is that the virtnet_rq_dma takes up 16 bytes at the beginning
of a new frag. When the frag size is larger than PAGE_SIZE,
everything is fine. However, if the frag is only one page and the
total size of the buffer and virtnet_rq_dma is larger than one page, an
overflow may occur.

The commit f9dac92ba908 ("virtio_ring: enable premapped mode whatever
use_dma_api") introduced this problem. And we reverted some commits to
fix this in last linux version. Now we try to enable it and fix this
bug directly.

Here, when the frag size is not enough, we reduce the buffer len to fix
this problem.

Reported-by: "Si-Wei Liu" <si-wei.liu@oracle.com>
Tested-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Merge branch 'fix-sparse-warnings-in-dpaa_eth-driver'

Vladimir Oltean says:

====================
Fix sparse warnings in dpaa_eth driver

This is a follow-up of the discussion at:
https://lore.kernel.org/oe-kbuild-all/20241028-sticky-refined-lionfish-b06c0c@leitao/
where I said I would take care of the sparse warnings uncovered by
Breno's COMPILE_TEST change for the dpaa_eth driver.

There was one warning that I decided to treat as an actual bug:
https://lore.kernel.org/netdev/20241029163105.44135-1-vladimir.oltean@nxp.com/
and what remains here are those warnings which I consider harmless.

I would like Christophe to ack the entire series to be taken through
netdev. I find it weird that the qbman driver, whose major API consumer
is netdev, is maintained by a different group. In this case, the buggy
qm_sg_entry_get_off() function is defined in qbman but exclusively
called in netdev.
====================

Link: https://patch.msgid.link/20241029164317.50182-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dpaa_eth: extract hash using __be32 pointer in rx_default_dqrr()

Sparse provides the following output:

warning: cast to restricted __be32

This is a harmless warning due to the fact that we dereference the hash
stored in the FD using an incorrect type annotation. Suppress the
warning by using the correct __be32 type instead of u32. No functional
change.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Link: https://patch.msgid.link/20241029164317.50182-4-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dpaa_eth: add assertions about SGT entry offsets in sg_fd_to_skb()

Multi-buffer frame descriptors (FDs) point to a buffer holding a
scatter/gather table (SGT), which is a finite array of fixed-size
entries, the last of which has qm_sg_entry_is_final(&sgt[i]) == true.

Each SGT entry points to a buffer holding pieces of the frame.
DPAARM.pdf explains in the figure called "Internal and External Margins,
Scatter/Gather Frame Format" that the SGT table is located within its
buffer at the same offset as the frame data start is located within the
first packet buffer.

                                 +------------------------+
    Scatter/Gather Buffer        |        First Buffer    |   Last Buffer
      ^ +------------+ ^       +-|---->^ +------------+   +->+------------+
      | |            | | ICEOF | |     | |            |      |////////////|
      | +------------+ v       | |     | |            |      |////////////|
 BSM  | |/ part of //|         | |BSM  | |            |      |////////////|
      | |/ Internal /|         | |     | |            |      |////////////|
      | |/ Context //|         | |     | |            |      |// Frame ///|
      | +------------+         | |     | |            | ...  |/ content //|
      | |            |         | |     | |            |      |////////////|
      | |            |         | |     | |            |      |////////////|
      v +------------+         | |     v +------------+      |////////////|
        | Scatter/ //| sgt[0]--+ |       |// Frame ///|      |////////////|
        | Gather List| ...       |       |/ content //|      +------------+ ^
        |////////////| sgt[N]----+       |////////////|      |            | | BEM
        |////////////|                   |////////////|      |            | |
        +------------+                   +------------+      +------------+ v

BSM = Buffer Start Margin, BEM = Buffer End Margin, both are configured
by dpaa_eth_init_rx_port() for the RX FMan port relevant here.

sg_fd_to_skb() runs in the calling context of rx_default_dqrr() -
the NAPI receive callback - which only expects to receive contiguous
(qm_fd_contig) or scatter/gather (qm_fd_sg) frame descriptors.
Everything else is irrelevant codewise.

The processing done by sg_fd_to_skb() is weird because it does not
conform to the expectations laid out by the aforementioned figure.
Namely, it parses the OFFSET field only for SGT entries with i != 0
(codewise, skb != NULL). In those cases, OFFSET should always be 0.
Also, it does not parse the OFFSET field for the sgt[0] case, the only
case where the buffer offset is meaningful in this context. There, it
uses the fd_off, aka the offset to the Scatter/Gather List in the
Scatter/Gather Buffer from the figure. By equivalence, they should both
be equal to the BSM (in turn, equal to priv->rx_headroom).

This can actually be explained due to the bug which we had in
qm_sg_entry_get_off() until the previous change:

- qm_sg_entry_get_off() did not actually _work_ for sgt[0]. It returned
  zero even with a non-zero offset, so fd_off had to be used as a fill-in.

- qm_sg_entry_get_off() always returned zero for sgt[i>0], and that
  resulted in no user-visible bug, because the buffer offset _was
  supposed_ to be zero for those buffers. So remove it from calculations.

Add assertions about the OFFSET field in both cases (first or subsequent
SGT entries) to make it absolutely obvious when something is not well
handled.

Similar logic can be seen in the driver for the architecturally similar
DPAA2, where dpaa2_eth_build_frag_skb() calls dpaa2_sg_get_offset() only
for i == 0. For the rest, there is even a comment stating the same thing:

 * Data in subsequent SG entries is stored from the
 * beginning of the buffer, so we don't need to add the
 * sg_offset.

Tested on LS1046A.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Link: https://patch.msgid.link/20241029164317.50182-3-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

soc: fsl_qbman: use be16_to_cpu() in qm_sg_entry_get_off()

struct qm_sg_entry :: offset is a 13-bit field, declared as __be16.

When using be32_to_cpu(), a wrong value will be calculated on little
endian systems (Arm), because type promotion from 16-bit to 32-bit,
which is done before the byte swap and always in the CPU native
endianness, changes the value of the scatter/gather list entry offset in
big-endian interpretation (adds two zero bytes in the LSB interpretation).
The result of the byte swap is ANDed with GENMASK(12, 0), so the result
is always zero, because only those bytes added by type promotion remain
after the application of the bit mask.

The impact of the bug is that scatter/gather frames with a non-zero
offset into the buffer are treated by the driver as if they had a zero
offset. This is all in theory, because in practice, qm_sg_entry_get_off()
has a single caller, where the bug is inconsequential, because at that
call site the buffer offset will always be zero, as will be explained in
the subsequent change.

Flagged by sparse:

warning: cast to restricted __be32
warning: cast from restricted __be16

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Link: https://patch.msgid.link/20241029164317.50182-2-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ena: remove devm from ethtool

There's no need for devm bloat here. In addition, these are freed right
before the function exits.

Also swapped kcalloc order for consistency.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Reviewed-by: Shay Agroskin <shayagr@amazon.com>
Link: https://patch.msgid.link/20241101214828.289752-2-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ptp: Remove 'default y' for VMCLOCK PTP device

The VMCLOCK device gives support for accurate timekeeping even across
live migration, unlike the KVM PTP clock. To help ensure that users can
always use ptp_vmclock where it's available in preference to ptp_kvm,
set it to 'default PTP_1588_CLOCK_VMCLOCK' instead of 'default y'.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Link: https://patch.msgid.link/89955b74d225129d6e3d79b53aa8d81d1b50560f.camel@infradead.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ena: Remove deadcode

ena_com_get_dev_basic_stats() has been unused since 2017's
commit d81db2405613 ("net/ena: refactor ena_get_stats64 to be atomic
context safe")

ena_com_get_offload_settings() has been unused since the original
commit of ENA back in 2016 in
commit 1738cd3ed342 ("net: ena: Add a driver for Amazon Elastic
Network Adapters (ENA)")

Remove them.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Reviewed-by: David Arinzon <darinzon@amazon.com>
Link: https://patch.msgid.link/20241102220142.80285-1-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ena: Remove autopolling mode

This manually reverts
commit a4e262cde3cd ("net: ena: allow automatic fallback to polling mode")

which is unused.

(I did it manually because there are other minor comment
and function changes surrounding it).
Build tested only.

Suggested-by: David Arinzon <darinzon@amazon.com>
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Link: https://patch.msgid.link/20241103194149.293456-1-linux@treblig.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tools: ynl-gen: de-kdocify enums with no doc for entries

Sometimes the names of the enum entries are self-explanatory
or come from standards. Forcing authors to write trivial kdoc
for each of such entries seems unreasonable, but kdoc would
complain about undocumented entries.

Detect enums which only have documentation for the entire
type and no documentation for entries. Render their doc
as a plain comment.

Link: https://patch.msgid.link/20241103165314.1631237-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: tcp: replace the document for "lsndtime" in tcp_sock

Commit d5fed5addb2b ("tcp: reorganize tcp_sock fast path variables")
moved the fields around and misplaced the documentation for "lsndtime".
So, let's replace it in the proper place.

Signed-off-by: Menglong Dong <dongml2@chinatelecom.cn>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241104070041.64302-1-dongml2@chinatelecom.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'mx95-netc-support'

Wei Fang says:

====================
net: add basic support for i.MX95 NETC

This is first time that the NETC IP is applied on i.MX MPU platform.
Its revision has been upgraded to 4.1, which is very different from
the NETC of LS1028A (its revision is 1.0). Therefore, some existing
drivers of NETC devices in the Linux kernel are not compatible with
the current hardware. For example, the fsl-enetc driver is used to
drive the ENETC PF of LS1028A, but for i.MX95 ENETC PF, its registers
and tables configuration are very different from those of LS1028A,
and only the station interface (SI) part remains basically the same.
For the SI part, Vladimir has separated the fsl-enetc-core driver, so
we can reuse this driver on i.MX95. However, for other parts of PF,
the fsl-enetc driver cannot be reused, so the nxp-enetc4 driver is
added to support revision 4.1 and later.

During the development process, we found that the two PF drivers have
some interfaces with basically the same logic, and the only difference
is the hardware configuration. So in order to reuse these interfaces
and reduce code redundancy, we extracted these interfaces and compiled
them into a separate nxp-enetc-pf-common driver for use by the two PF
drivers.

In addition, we have developed the nxp-netc-blk-ctrl driver, which
is used to control three blocks, namely Integrated Endpoint Register
Block (IERB), Privileged Register Block (PRB) and NETCMIX block. The
IERB contains registers that are used for pre-boot initialization,
debug, and non-customer configuration. The PRB controls global reset
and global error handling for NETC. The NETCMIX block is mainly used
to set MII protocol and PCS protocol of the links, it also contains
settings for some other functions.

---
Link: https://lore.kernel.org/imx/20241009095116.147412-1-wei.fang@nxp.com/
Link: https://lore.kernel.org/imx/20241015125841.1075560-1-wei.fang@nxp.com/
Link: https://lore.kernel.org/imx/20241017074637.1265584-1-wei.fang@nxp.com/
Link: https://lore.kernel.org/imx/20241022055223.382277-1-wei.fang@nxp.com/
Link: https://lore.kernel.org/imx/20241024065328.521518-1-wei.fang@nxp.com/
====================

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

MAINTAINERS: update ENETC driver files and maintainers

Add related YAML documentation and header files. Also, add maintainers
from the i.MX side as ENETC starts to be used on i.MX platforms.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: add preliminary support for i.MX95 ENETC PF

The i.MX95 ENETC has been upgraded to revision 4.1, which is different
from the LS1028A ENETC (revision 1.0) except for the SI part. Therefore,
the fsl-enetc driver is incompatible with i.MX95 ENETC PF. So add new
nxp-enetc4 driver to support i.MX95 ENETC PF, and this driver will be
used to support the ENETC PF with major revision 4 for other SoCs in the
future.

Currently, the nxp-enetc4 driver only supports basic transmission feature
for i.MX95 ENETC PF, the more basic and advanced features will be added
in the subsequent patches. In addition, PCS support has not been added
yet, so 10G ENETC (ENETC instance 2) is not supported now.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: optimize the allocation of tx_bdr

There is a situation where num_tx_rings cannot be divided by bdr_int_num.
For example, num_tx_rings is 8 and bdr_int_num is 3. According to the
previous logic, this results in two tx_bdr corresponding memories not
being allocated, so when sending packets to tx ring 6 or 7, wild pointers
will be accessed. Of course, this issue doesn't exist on LS1028A, because
its num_tx_rings is 8, and bdr_int_num is either 1 or 2. However, there
is a risk for the upcoming i.MX95. Therefore, it is necessary to ensure
that each tx_bdr can be allocated to the corresponding memory.

Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Claudiu Manoil <claudiu.manoil@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: extract enetc_int_vector_init/destroy() from enetc_alloc_msix()

Extract enetc_int_vector_init() and enetc_int_vector_destroy() from
enetc_alloc_msix() so that the code is more concise and readable.

Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: add i.MX95 EMDIO support

The verdor ID and device ID of i.MX95 EMDIO are different from LS1028A
EMDIO, so add new vendor ID and device ID to pci_device_id table to
support i.MX95 EMDIO.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: remove ERR050089 workaround for i.MX95

The ERR050089 workaround causes performance degradation and potential
functional issues (e.g., RCU stalls) under certain workloads. Since
new SoCs like i.MX95 do not require this workaround, use a static key
to compile out enetc_lock_mdio() and enetc_unlock_mdio() at runtime,
improving performance and avoiding unnecessary logic.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: build enetc_pf_common.c as a separate module

Compile enetc_pf_common.c as a standalone module to allow shared usage
between ENETC v1 and v4 PF drivers. Add struct enetc_pf_ops to register
different hardware operation interfaces for both ENETC v1 and v4 PF
drivers.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: extract common ENETC PF parts for LS1028A and i.MX95 platforms

The ENETC PF driver of LS1028A (rev 1.0) is incompatible with the version
used on the i.MX95 platform (rev 4.1), except for the station interface
(SI) part. To reduce code redundancy and prepare for a new driver for rev
4.1 and later, extract shared interfaces from enetc_pf.c and move them to
enetc_pf_common.c. This refactoring lays the groundwork for compiling
enetc_pf_common.c into a shared driver for both platforms' PF drivers.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enetc: add initial netc-blk-ctrl driver support

The netc-blk-ctrl driver is used to configure Integrated Endpoint
Register Block (IERB) and Privileged Register Block (PRB) of NETC.
For i.MX platforms, it is also used to configure the NETCMIX block.

The IERB contains registers that are used for pre-boot initialization,
debug, and non-customer configuration. The PRB controls global reset
and global error handling for NETC. The NETCMIX block is mainly used
to set MII protocol and PCS protocol of the links, it also contains
settings for some other functions.

Note the IERB configuration registers can only be written after being
unlocked by PRB, otherwise, all write operations are inhibited. A warm
reset is performed when the IERB is unlocked, and it results in an FLR
to all NETC devices. Therefore, all NETC device drivers must be probed
or initialized after the warm reset is finished.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: add bindings for NETC blocks control

Add bindings for NXP NETC blocks control. Usually, NETC has 2 blocks of
64KB registers, integrated endpoint register block (IERB) and privileged
register block (PRB). IERB is used for pre-boot initialization for all
NETC devices, such as ENETC, Timer, EMDIO and so on. And PRB controls
global reset and global error handling for NETC. Moreover, for the i.MX
platform, there is also a NETCMIX block for link configuration, such as
MII protocol, PCS protocol, etc.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: add i.MX95 ENETC support

The ENETC of i.MX95 has been upgraded to revision 4.1, and the vendor
ID and device ID have also changed, so add the new compatible strings
for i.MX95 ENETC. In addition, i.MX95 supports configuration of RGMII
or RMII reference clock.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: add compatible string for i.MX95 EMDIO

The EMDIO of i.MX95 has been upgraded to revision 4.1, and the vendor
ID and device ID have also changed, so add the new compatible strings
for i.MX95 EMDIO.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'mlx5-misc-patches-2024-10-31'

Tariq Toukan says:

====================
mlx5 misc patches 2024-10-31

First patch by Cosmin fixes an issue in a recent commit.

Followed by 2 patches by Yevgeny that organize and rename the files
under the steering directory.

Finally, 2 patches by William that save the creation of the unused
egress-XDP_REDIRECT send queue on non-uplink representor.
====================

Link: https://patch.msgid.link/20241031125856.530927-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5e: do not create xdp_redirect for non-uplink rep

XDP and XDP socket require extra SQ/RQ/CQs. Most of these resources
are dynamically created: no XDP program loaded, no resources are
created. One exception is the SQ/CQ created for XDP_REDRIECT, used
for other netdev to forward packet to mlx5 for transmit. The patch
disables creation of SQ and CQ used for egress XDP_REDIRECT, by
checking whether ndo_xdp_xmit is set or not.

For netdev without XDP support such as non-uplink representor, this
saves around 0.35MB of memory, per representor netdevice per channel.

Signed-off-by: William Tu <witu@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20241031125856.530927-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5e: move XDP_REDIRECT sq to dynamic allocation

Dynamically allocating xdpsq, used by egress side XDP_REDIRECT.
mlx5 has multiple XDP sqs. Under struct mlx5e_channel:
1. rx_xdpsq: used for XDP_TX, an XDP prog handles the rx packet and
transmits using the same queue as rx.
2. xdpsq: used by egress side XDP_REDIRECT. This is for another interface
to redirect packet to the mlx5 interface, using ndo_xdp_xmit .
3. xsksq: used by XSK. XSK has its own dedicated channel, and it also
has resources of 1 and 2.

The patch changes only the 2. xdpsq.

Signed-off-by: William Tu <witu@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20241031125856.530927-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5: HWS, renamed the files in accordance with naming convention

Removed the 'mlx5hws_' file name prefix from the internal HWS files.

Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20241031125856.530927-4-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5: DR, moved all the SWS code into a separate directory

After adding HWS support in a separate folder, moving all the SWS
code into its own folder as well.
Now SWS and HWS implementation are located in their appropriate
folders:
 - steering/sws/
 - steering/hws/

Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20241031125856.530927-3-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5: Rework esw qos domain init and cleanup

The first approach was flawed, because there are situations where the
esw mode change fails, leaving the qos domain as NULL. Various calls
into the QoS infra then trigger a NULL pointer access and unhappiness.

Improve that by a combination of:
- Allocating the QoS domain on esw init and cleaning it up on teardown.
- Refactoring mode change to only call qos domain init but not cleanup.
- Making qos domain init idempotent - not change anything if nothing
  needs changing.

Together, these should guarantee that, as long as the memory allocations
succeed, there should always be a valid qos domain until the esw
cleanup, no matter what mode changes happen (or failures thereof).

Fixes: 107a034d5c1e ("net/mlx5: qos: Store rate groups in a qos domain")
Signed-off-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Carolina Jubran <cjubran@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20241031125856.530927-2-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests/tc-testing: add tests for qdisc_tree_reduce_backlog

Add 3 tests to check for the expected behaviour of
qdisc_tree_reduce_backlog in special scenarios.

- The first test checks if the qdisc class is notified of deletion for
major handle 'ffff:'.
- The second test checks the same as the first test but with 'ffff:' as the root
qdisc.
- The third test checks if everything works if ingress is active.

Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Cong Wang <cong.wang@bytedance.com>
Link: https://patch.msgid.link/20241101143148.1218890-1-pctammela@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-stmmac-refactor-fpe-as-a-separate-module'

Furong Xu says:

====================
net: stmmac: Refactor FPE as a separate module

Refactor FPE implementation by moving common code for DWMAC4 and
DWXGMAC into a separate FPE module.

FPE implementation for DWMAC4 and DWXGMAC differs only for:
1) Offset address of MAC_FPE_CTRL_STS and MTL_FPE_CTRL_STS
2) FPRQ(Frame Preemption Residue Queue) field in MAC_RxQ_Ctrl1
3) Bit offset of Frame Preemption Interrupt Enable

Tested on DWMAC CORE 5.20a and DWXGMAC CORE 3.20a
====================

Link: https://patch.msgid.link/cover.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: xgmac: Enable FPE for tc-mqprio/tc-taprio

The FPE on XGMAC is ready, it is time to update dwxgmac_tc_ops to
let user configure FPE via tc-mqprio/tc-taprio.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/0575ef1553d572b7c8bc1baafa3fb7ac641073e0.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: xgmac: Complete FPE support

Implement the necessary fpe_map_preemption_class callback for xgmac.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/d0347f2b8a71fee372e53293fe26a6538775ec5d.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: xgmac: Rename XGMAC_RQ to XGMAC_FPRQ

Synopsys XGMAC Databook defines MAC_RxQ_Ctrl1 register:
RQ: Frame Preemption Residue Queue

XGMAC_FPRQ is more readable and more consistent with GMAC4.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/611991edf9e9d6fac8b29c3fe952791b193ca179.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: Get the TC number of net_device by netdev_get_num_tc()

netdev_get_num_tc() is the right method, we should not access
net_device.num_tc directly.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/6298463f4655a76faf94e4273a4205c13ca17c77.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: Refactor FPE functions to generic version

FPE implementation for DWMAC4 and DWXGMAC differs only for:
1) Offset address of MAC_FPE_CTRL_STS and MTL_FPE_CTRL_STS
2) FPRQ(Frame Preemption Residue Queue) field in MAC_RxQ_Ctrl1
3) Bit offset of Frame Preemption Interrupt Enable

Refactor FPE functions to avoid code duplication and
to simplify the code flow by avoiding the use of
function pointers.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/49de4607bae69ffe751b13329a3c07a990b82419.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: Introduce stmmac_fpe_supported()

A single "priv->dma_cap.fpesel" checks HW capability only,
while both HW capability and driver capability shall be
checked by later refactoring to prevent unexpected behavior
for FPE on unsupported MAC cores and keep FPE as an optional
implementation for current and new MAC cores.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/01e9cd13aedd38cb0e9a5d9875c475ce35250188.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: Rework macro definitions for gmac4 and xgmac

Rename and add macro definitions to better reuse them in common code.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/510b85288b13aa2cce5adf849291009c6f29a84a.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: stmmac: Introduce separate files for FPE implementation

By moving FPE related code info separate files, FPE implementation
becomes a separate module initially.
No functional change intended.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/e9ddf4fbf0fc053ae30592aa6c4363e72a4d8e62.1730449003.git.0x1207@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'r8169-align-rtl8125-rtl8126-phy-config-with-vendor-driver'

Heiner Kallweit says:

====================
r8169: align RTL8125/RTL8126 PHY config with vendor driver

This series aligns the RTL8125/RTL8126 PHY config with vendor drivers
r8125 and r8126 respectively.
====================

Link: https://patch.msgid.link/7a849c7c-50ff-4a9b-9a1c-a963b0561c79@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

r8169: align RTL8126 EEE config with vendor driver

Align the EEE config for RTL8126A with vendor driver r8126 to avoid
compatibility issues.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://patch.msgid.link/71e4859e-4cd0-4b6b-b7fa-621d7721992f@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

r8169: align RTL8125/RTL8126 PHY config with vendor driver

This aligns some parameters with vendor driver r8125/r8126 to avoid
compatibility issues. Note that for RTL8125B there's no functional
change, just the open-coded version of the function is replaced.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://patch.msgid.link/a8a9d896-fbe6-41f2-bf87-666567d3cdb3@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

r8169: align RTL8125 EEE config with vendor driver

Align the EEE config for RTL8125A/RTL8125B with vendor driver r8125.
This should help to avoid compatibility issues.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://patch.msgid.link/044c925e-8669-4b98-87df-95b4056f4f5f@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2024-10-31

We've added 13 non-merge commits during the last 16 day(s) which contain
a total of 16 files changed, 710 insertions(+), 668 deletions(-).

The main changes are:

1) Optimize and homogenize bpf_csum_diff helper for all archs and also
   add a batch of new BPF selftests for it, from Puranjay Mohan.

2) Rewrite and migrate the test_tcp_check_syncookie.sh BPF selftest
   into test_progs so that it can be run in BPF CI, from Alexis Lothoré.

3) Two BPF sockmap selftest fixes, from Zijian Zhang.

4) Small XDP synproxy BPF selftest cleanup to remove IP_DF check,
   from Vincent Li.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next:
  selftests/bpf: Add a selftest for bpf_csum_diff()
  selftests/bpf: Don't mask result of bpf_csum_diff() in test_verifier
  bpf: bpf_csum_diff: Optimize and homogenize for all archs
  net: checksum: Move from32to16() to generic header
  selftests/bpf: remove xdp_synproxy IP_DF check
  selftests/bpf: remove test_tcp_check_syncookie
  selftests/bpf: test MSS value returned with bpf_tcp_gen_syncookie
  selftests/bpf: add ipv4 and dual ipv4/ipv6 support in btf_skc_cls_ingress
  selftests/bpf: get rid of global vars in btf_skc_cls_ingress
  selftests/bpf: add missing ns cleanups in btf_skc_cls_ingress
  selftests/bpf: factorize conn and syncookies tests in a single runner
  selftests/bpf: Fix txmsg_redir of test_txmsg_pull in test_sockmap
  selftests/bpf: Fix msg_verify_data in test_sockmap

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

Link: https://patch.msgid.link/20241031221543.108853-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'ibm-emac-cleanup-modules-to-use-devm'

Rosen Penev says:

====================
ibm: emac: cleanup modules to use devm

simplifies probe and removes remove functions. These drivers are small.
====================

Link: https://patch.msgid.link/20241030203727.6039-1-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: mal: move irq maps down

Moves the handling right before they are used and allows merging a
branch.

Also get rid of the error handling as devm_request_irq can handle that.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-13-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: mal: use devm for request_irq

Avoids manual frees. Also replaced irq_of_parse_and_map with
platform_get_irq since it's simpler and does the same thing.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-12-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: mal: use devm for kzalloc

Simplifies the probe function by removing gotos.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-11-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: zmii: devm_platform_get_resource

Simplifies the probe function by a bit and allows removing the _remove
function such that devm now handles all cleanup.

printk gets converted to dev_err as np is now gone.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-10-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: zmii: use devm for mutex_init

It seems that since inception, this driver never called mutex_destroy in
_remove. Use devm to handle this automatically.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-9-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: zmii: use devm for kzalloc

Simplifies the probe function by removing gotos.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-8-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: rgmii: devm_platform_get_resource

Simplifies the probe function by a bit and allows removing the _remove
function such that devm now handles all cleanup.

printk gets converted to dev_err as np is now gone.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-7-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ibm: emac: rgmii: use devm for mutex_init

It seems that since inception, this driver never called mutex_destroy in
_remove. Use devm to handle this automatically.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20241030203727.6039-6-rosenp@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>