Enable threaded NAPI by default for WireGuard devices in response to low
performance behavior that we observed when multiple tunnels (and thus
multiple wg devices) are deployed on a single host. This affects any
kind of multi-tunnel deployment, regardless of whether the tunnels share
the same endpoints or not (i.e., a VPN concentrator type of gateway
would also be affected).
The problem is caused by the fact that, in case of a traffic surge that
involves multiple tunnels at the same time, the polling of the NAPI
instance of all these wg devices tends to converge onto the same core,
causing underutilization of the CPU and bottlenecking performance.
This happens because NAPI polling is hosted by default in softirq
context, but the WireGuard driver only raises this softirq after the rx
peer queue has been drained, which doesn't happen during high traffic.
In this case, the softirq already active on a core is reused instead of
raising a new one.
As a result, once two or more tunnel softirqs have been scheduled on
the same core, they remain pinned there until the surge ends.
In our experiments, this almost always leads to all tunnel NAPIs being
handled on a single core shortly after a surge begins, limiting
scalability to less than 3× the performance of a single tunnel, despite
plenty of unused CPU cores being available.
The proposed mitigation is to enable threaded NAPI for all WireGuard
devices. This moves the NAPI polling context to a dedicated per-device
kernel thread, allowing the scheduler to balance the load across all
available cores.
On our 32-core gateways, enabling threaded NAPI yields a ~4× performance
improvement with 16 tunnels, increasing throughput from ~13 Gbps to
~48 Gbps. Meanwhile, CPU usage on the receiver (which is the bottleneck)
jumps from 20% to 100%.
We have found no performance regressions in any scenario we tested.
Single-tunnel throughput remains unchanged.
More details are available in our Netdev paper.
Link: https://netdevconf.info/0x18/docs/netdev-0x18-paper23-talk-paper.pdf
Signed-off-by: Mirco Barone <mirco.barone@polito.it>
Fixes: e7096c131e ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Link: https://patch.msgid.link/20250605120616.2808744-1-Jason@zx2c4.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----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=CFYz
-----END PGP SIGNATURE-----
Merge tag 'nf-25-06-05' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following patchset contains Netfilter fixes for net:
1) Zero out the remainder in nft_pipapo AVX2 implementation, otherwise
next lookup could bogusly report a mismatch. This is followed by two
patches to update nft_pipapo selftests to cover for the previous bug.
From Florian Westphal.
2) Check for reverse tuple too in case of esoteric NAT collisions for
UDP traffic and extend selftest coverage. Also from Florian.
netfilter pull request 25-06-05
* tag 'nf-25-06-05' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
selftests: netfilter: nft_nat.sh: add test for reverse clash with nat
netfilter: nf_nat: also check reverse tuple to obtain clashing entry
selftests: netfilter: nft_concat_range.sh: add datapath check for map fill bug
selftests: netfilter: nft_concat_range.sh: prefer per element counters for testing
netfilter: nf_set_pipapo_avx2: fix initial map fill
====================
Link: https://patch.msgid.link/20250605085735.52205-1-pablo@netfilter.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Jakub Kicinski says:
====================
netlink: specs: rt-link: decode ip6gre
Adding GRE tunnels to the .config for driver tests caused
some unhappiness in YNL, as it can't decode all the link
attrs on the system. Add ip6gre support to fix the tests.
This is similar to commit 6ffdbb93a5 ("netlink: specs:
rt_link: decode ip6tnl, vti and vti6 link attrs").
====================
Link: https://patch.msgid.link/20250603135357.502626-1-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Driver tests now require GRE tunnels, while we don't configure
them with YNL, YNL will complain when it sees link types it
doesn't recognize. Teach it decoding ip6gre tunnels. The attrs
are largely the same as IPv4 GRE.
Correct the type of encap-limit, but note that this attr is
only used in ip6gre, so the mistake didn't matter until now.
Fixes: 0d0f4174f6 ("selftests: drv-net: add a simple TSO test")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20250603135357.502626-3-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
A number of fields in the ip tunnels are lacking the big-endian
designation. I suspect this is not intentional, as decoding
the ports with the right endian seems objectively beneficial.
Fixes: 6ffdbb93a5 ("netlink: specs: rt_link: decode ip6tnl, vti and vti6 link attrs")
Fixes: 077b6022d2 ("doc/netlink/specs: Add sub-message type to rt_link family")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20250603135357.502626-2-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
* dropped bogus call to setup_udp_tunnel_sock() during
cleanup, substituted by proper state unwind
* fixed race condition between peer removal (by kernel
space) and socket closing (by user space)
* fixed sleep in atomic context along TCP RX error path
* fixes for ovpn kselftests
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEmavcbPjUEuTeX8D8C3DlOqA41YcFAmg+17IACgkQC3DlOqA4
1YfGnQf7BCqweJAqPZs45DrMO/zu0mygbvzxI5x1VnBSQvtUfEEwgY3Mk9Y/8OxW
+QxD9yYl8sf30ZzGTFFYtBq0XwvcP0Nlo4WFD7+2ufvFgojIAwCsF25of+hODGLZ
5KF83VYDAMvUmK7Rq4gGyi/HC7+rWNhShfoXBh6nfQjNa75NWwIwp8wEsTMovq2Y
DiH6VzCjkUgnRjJYG5g/D6n0pHnbUkrAPi3AT+aB8PX9sikrTZhBWX4mmY8n6Xlo
KT83ByzJqc5svnbsQEGRuhXbyuKipBhcVmOohmulBB/uhXUZHDcem9L0wp1hR2SB
y/lZQxt9bNdEXfh0J7r5QxI2vkZgZQ==
=Y6+P
-----END PGP SIGNATURE-----
Merge tag 'ovpn-net-20250603' of https://github.com/OpenVPN/ovpn-net-next
Antonio Quartulli says:
====================
In this batch you can find the following bug fixes:
Patch 1: when releasing a UDP socket we were wrongly invoking
setup_udp_tunnel_sock() with an empty config. This was not
properly shutting down the UDP encap state.
With this patch we simply undo what was done during setup.
Patch 2: ovpn was holding a reference to a 'struct socket'
without increasing its reference counter. This was intended
and worked as expected until we hit a race condition where
user space tries to close the socket while kernel space is
also releasing it. In this case the (struct socket *)->sk
member would disappear under our feet leading to a null-ptr-deref.
This patch fixes this issue by having struct ovpn_socket hold
a reference directly to the sk member while also increasing
its reference counter.
Patch 3: in case of errors along the TCP RX path (softirq)
we want to immediately delete the peer, but this operation may
sleep. With this patch we move the peer deletion to a scheduled
worker.
Patch 4 and 5 are instead fixing minor issues in the ovpn
kselftests.
* tag 'ovpn-net-20250603' of https://github.com/OpenVPN/ovpn-net-next:
selftest/net/ovpn: fix missing file
selftest/net/ovpn: fix TCP socket creation
ovpn: avoid sleep in atomic context in TCP RX error path
ovpn: ensure sk is still valid during cleanup
ovpn: properly deconfigure UDP-tunnel
====================
Link: https://patch.msgid.link/20250603111110.4575-1-antonio@openvpn.net/
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Recent Qualcomm chipsets like SDX72/75 require MBIM sessionId mapping
to muxId in the range (0x70-0x8F) for the PCIe tethered use.
This has been partially addressed by the referenced commit, mapping
the default data call to muxId = 112, but the multiplexed data calls
scenario was not properly considered, mapping sessionId = 1 to muxId
1, while it should have been 113.
Fix this by moving the session_id assignment logic to mhi_mbim_newlink,
in order to map sessionId = n to muxId = n + WDS_BIND_MUX_DATA_PORT_MUX_ID.
Fixes: 65bc58c3dc ("net: wwan: mhi: make default data link id configurable")
Signed-off-by: Daniele Palmas <dnlplm@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20250603091204.2802840-1-dnlplm@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
S1G beacons are not traditional beacons but a type of extension frame.
Extension frames contain the frame control and duration fields, followed
by zero or more optional fields before the frame body. These optional
fields are distinct from the variable length elements.
The presence of optional fields is indicated in the frame control field.
To correctly locate the elements offset, the frame control must be parsed
to identify which optional fields are present. Currently, mac80211 parses
S1G beacons based on fixed assumptions about the frame layout, without
inspecting the frame control field. This can result in incorrect offsets
to the "variable" portion of the frame.
Properly parse S1G beacon frames by using the field lengths defined in
IEEE 802.11-2024, section 9.3.4.3, ensuring that the elements offset is
calculated accurately.
Fixes: 9eaffe5078 ("cfg80211: convert S1G beacon to scan results")
Fixes: cd418ba63f ("mac80211: convert S1G beacon to scan results")
Signed-off-by: Lachlan Hodges <lachlan.hodges@morsemicro.com>
Link: https://patch.msgid.link/20250603053538.468562-1-lachlan.hodges@morsemicro.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Jonas Gorski says:
====================
net: dsa: b53: fix RGMII ports
RGMII ports on BCM63xx were not really working, especially with PHYs
that support EEE and are capable of configuring their own RGMII delays.
So let's make them work, and fix additional minor rgmii related issues
found while working on it.
With a BCM96328BU-P300:
Before:
[ 3.580000] b53-switch 10700000.switch GbE3 (uninitialized): validation of rgmii with support 0000000,00000000,00000000,000062ff and advertisement 0000000,00000000,00000000,000062ff failed: -EINVAL
[ 3.600000] b53-switch 10700000.switch GbE3 (uninitialized): failed to connect to PHY: -EINVAL
[ 3.610000] b53-switch 10700000.switch GbE3 (uninitialized): error -22 setting up PHY for tree 0, switch 0, port 4
[ 3.620000] b53-switch 10700000.switch GbE1 (uninitialized): validation of rgmii with support 0000000,00000000,00000000,000062ff and advertisement 0000000,00000000,00000000,000062ff failed: -EINVAL
[ 3.640000] b53-switch 10700000.switch GbE1 (uninitialized): failed to connect to PHY: -EINVAL
[ 3.650000] b53-switch 10700000.switch GbE1 (uninitialized): error -22 setting up PHY for tree 0, switch 0, port 5
[ 3.660000] b53-switch 10700000.switch GbE4 (uninitialized): validation of rgmii with support 0000000,00000000,00000000,000062ff and advertisement 0000000,00000000,00000000,000062ff failed: -EINVAL
[ 3.680000] b53-switch 10700000.switch GbE4 (uninitialized): failed to connect to PHY: -EINVAL
[ 3.690000] b53-switch 10700000.switch GbE4 (uninitialized): error -22 setting up PHY for tree 0, switch 0, port 6
[ 3.700000] b53-switch 10700000.switch GbE5 (uninitialized): validation of rgmii with support 0000000,00000000,00000000,000062ff and advertisement 0000000,00000000,00000000,000062ff failed: -EINVAL
[ 3.720000] b53-switch 10700000.switch GbE5 (uninitialized): failed to connect to PHY: -EINVAL
[ 3.730000] b53-switch 10700000.switch GbE5 (uninitialized): error -22 setting up PHY for tree 0, switch 0, port 7
After:
[ 3.700000] b53-switch 10700000.switch GbE3 (uninitialized): PHY [mdio_mux-0.1:00] driver [Broadcom BCM54612E] (irq=POLL)
[ 3.770000] b53-switch 10700000.switch GbE1 (uninitialized): PHY [mdio_mux-0.1:01] driver [Broadcom BCM54612E] (irq=POLL)
[ 3.850000] b53-switch 10700000.switch GbE4 (uninitialized): PHY [mdio_mux-0.1:18] driver [Broadcom BCM54612E] (irq=POLL)
[ 3.920000] b53-switch 10700000.switch GbE5 (uninitialized): PHY [mdio_mux-0.1:19] driver [Broadcom BCM54612E] (irq=POLL)
====================
Link: https://patch.msgid.link/20250602193953.1010487-1-jonas.gorski@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add RGMII to supported interfaces for BCM63xx RGMII ports so they can be
actually used in RGMII mode.
Without this, phylink will fail to configure them:
[ 3.580000] b53-switch 10700000.switch GbE3 (uninitialized): validation of rgmii with support 0000000,00000000,00000000,000062ff and advertisement 0000000,00000000,00000000,000062ff failed: -EINVAL
[ 3.600000] b53-switch 10700000.switch GbE3 (uninitialized): failed to connect to PHY: -EINVAL
[ 3.610000] b53-switch 10700000.switch GbE3 (uninitialized): error -22 setting up PHY for tree 0, switch 0, port 4
Fixes: ce3bf94871 ("net: dsa: b53: add support for BCM63xx RGMIIs")
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Link: https://patch.msgid.link/20250602193953.1010487-5-jonas.gorski@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The IMP port is not a valid RGMII interface, but hard wired to internal,
so we shouldn't touch the undefined register B53_RGMII_CTRL_IMP.
While this does not seem to have any side effects, let's not touch it at
all, so limit RGMII configuration on bcm63xx to the actual RGMII ports.
Fixes: ce3bf94871 ("net: dsa: b53: add support for BCM63xx RGMIIs")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250602193953.1010487-4-jonas.gorski@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
bcm63xx's RGMII ports are always in MAC mode, never in PHY mode, so we
shouldn't enable any delays and let the PHY handle any delays as
necessary.
This fixes using RGMII ports with normal PHYs like BCM54612E, which will
handle the delay in the PHY.
Fixes: ce3bf94871 ("net: dsa: b53: add support for BCM63xx RGMIIs")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250602193953.1010487-3-jonas.gorski@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
BCM63xx internal switches do not support EEE, but provide multiple RGMII
ports where external PHYs may be connected. If one of these PHYs are EEE
capable, we may try to enable EEE for the MACs, which then hangs the
system on access of the (non-existent) EEE registers.
Fix this by checking if the switch actually supports EEE before
attempting to configure it.
Fixes: 22256b0afb ("net: dsa: b53: Move EEE functions to b53")
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Link: https://patch.msgid.link/20250602193953.1010487-2-jonas.gorski@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In MII mode, Tx lines are swapped for port0 and port1, which means
Tx port0 receives data from PRU1 and the Tx port1 receives data from
PRU0. This is an expected hardware behavior and reading the Tx stats
needs to be handled accordingly in the driver. Update the driver to
read Tx stats from the PRU1 for port0 and PRU0 for port1.
Fixes: c1e10d5dc7 ("net: ti: icssg-prueth: Add ICSSG Stats")
Signed-off-by: Meghana Malladi <m-malladi@ti.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250603052904.431203-1-m-malladi@ti.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This will fail without the previous bug fix because we erronously
believe that the clashing entry went way.
However, the clash exists in the opposite direction due to an
existing nat mapping:
PASS: IP statless for ns2-LgTIuS
ERROR: failed to test udp ns1-x4iyOW to ns2-LgTIuS with dnat rule step 2, result: ""
This is partially adapted from test instructions from the below
ubuntu tracker.
Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2109889
Signed-off-by: Florian Westphal <fw@strlen.de>
Tested-by: Shaun Brady <brady.1345@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The logic added in the blamed commit was supposed to only omit nat source
port allocation if neither the existing nor the new entry are subject to
NAT.
However, its not enough to lookup the conntrack based on the proposed
tuple, we must also check the reverse direction.
Otherwise there are esoteric cases where the collision is in the reverse
direction because that colliding connection has a port rewrite, but the
new entry doesn't. In this case, we only check the new entry and then
erronously conclude that no clash exists anymore.
The existing (udp) tuple is:
a:p -> b:P, with nat translation to s:P, i.e. pure daddr rewrite,
reverse tuple in conntrack table is s:P -> a:p.
When another UDP packet is sent directly to s, i.e. a:p->s:P, this is
correctly detected as a colliding entry: tuple is taken by existing reply
tuple in reverse direction.
But the colliding conntrack is only searched for with unreversed
direction, and we can't find such entry matching a:p->s:P.
The incorrect conclusion is that the clashing entry has timed out and
that no port address translation is required.
Such conntrack will then be discarded at nf_confirm time because the
proposed reverse direction clashes with an existing mapping in the
conntrack table.
Search for the reverse tuple too, this will then check the NAT bits of
the colliding entry and triggers port reallocation.
Followp patch extends nft_nat.sh selftest to cover this scenario.
The IPS_SEQ_ADJUST change is also a bug fix:
Instead of checking for SEQ_ADJ this tested for SEEN_REPLY and ASSURED
by accident -- _BIT is only for use with the test_bit() API.
This bug has little consequence in practice, because the sequence number
adjustments are only useful for TCP which doesn't support clash resolution.
The existing test case (conntrack_reverse_clash.sh) exercise a race
condition path (parallel conntrack creation on different CPUs), so
the colliding entries have neither SEEN_REPLY nor ASSURED set.
Thanks to Yafang Shao and Shaun Brady for an initial investigation
of this bug.
Fixes: d8f84a9bc7 ("netfilter: nf_nat: don't try nat source port reallocation for reverse dir clash")
Closes: https://bugzilla.netfilter.org/show_bug.cgi?id=1795
Reported-by: Yafang Shao <laoar.shao@gmail.com>
Reported-by: Shaun Brady <brady.1345@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Tested-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
commit 0935ee6032 ("selftests: netfilter: add test case for recent mismatch bug")
added a regression check for incorrect initial fill of the result map
that was fixed with 791a615b7a ("netfilter: nf_set_pipapo: fix initial map fill").
The test used 'nft get element', i.e., control plane checks for
match/nomatch results.
The control plane however doesn't use avx2 version, so we need to
send+match packets.
As the additional packet match/nomatch is slow, don't do this for
every element added/removed: add and use maybe_send_(no)match
helpers and use them.
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The selftest uses following rule:
... @test counter name "test"
Then sends a packet, then checks if the named counter did increment or
not.
This is fine for the 'no-match' test case: If anything matches the
counter increments and the test fails as expected.
But for the 'should match' test cases this isn't optimal.
Consider buggy matching, where the packet matches entry x, but it
should have matched entry y.
In that case the test would erronously pass.
Rework the selftest to use per-element counters to avoid this.
After sending packet that should have matched entry x, query the
relevant element via 'nft reset element' and check that its counter
had incremented.
The 'nomatch' case isn't altered, no entry should match so the named
counter must be 0, changing it to the per-element counter would then
pass if another entry matches.
The downside of this change is a slight increase in test run-time by
a few seconds.
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
If the first field doesn't cover the entire start map, then we must zero
out the remainder, else we leak those bits into the next match round map.
The early fix was incomplete and did only fix up the generic C
implementation.
A followup patch adds a test case to nft_concat_range.sh.
Fixes: 791a615b7a ("netfilter: nf_set_pipapo: fix initial map fill")
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The regulatory domain information was initialized every time the
FW was loaded and the device was restarted. This was unnecessary
and useless as at this stage the wiphy channels information was
not setup yet so while the regulatory domain was set to the wiphy,
the channel information was not updated.
In case that a specific MCC was configured during FW initialization
then following updates with this MCC are ignored, and thus the
wiphy channels information is left with information not matching
the regulatory domain.
This commit moves the regulatory domain initialization to after the
operational firmware is started, i.e., after the wiphy channels were
configured and the regulatory information is needed.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20250604061200.f138a7382093.I2fd8b3e99be13c2687da483e2cb1311ffb4fbfce@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
When reading the interrupt status after a FW reset handshake
timeout, read the actual value not the mask for the non-MSIX
case.
Fixes: ab606dea80 ("wifi: iwlwifi: pcie: add support for the reset handshake in MSI")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20250604061200.87a849a55086.I2f8571aafa55aa3b936a30b938de9d260592a584@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
In case of an error during init, in_hw_restart will be set, but it will
never get cleared.
Instead, we will retry to init again, and then we will act like we are in a
restart when we are actually not.
This causes (among others) to a NULL pointer dereference when canceling
rx_omi::finished_work, that was not even initialized, because we thought
that we are in hw_restart.
Set in_hw_restart to true only if the fw is running, then we know that
FW was loaded successfully and we are not going to the retry loop.
Fixes: 7391b2a4f7 ("wifi: iwlwifi: rework firmware error handling")
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20250604061200.e0040e0a4b09.Iae469a0abe6bfa3c26d8a88c066bad75c2e8f121@changeid
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
After using DEFINE_RAW_FLEX, cmd is a pointer to iwl_rxq_sync_cmd,
and not a variable containing both the command and notification.
Adjust hcmd->data and hcmd->len assignment as well.
Fixes: 7438843df8 ("wifi: iwlwifi: mvm: Avoid -Wflex-array-member-not-at-end warning")
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
Link: https://patch.msgid.link/20250604031321.2277481-2-miriam.rachel.korenblit@intel.com
Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com>
gve_alloc_pending_packet() can return NULL, but gve_tx_add_skb_dqo()
did not check for this case before dereferencing the returned pointer.
Add a missing NULL check to prevent a potential NULL pointer
dereference when allocation fails.
This improves robustness in low-memory scenarios.
Fixes: a57e5de476 ("gve: DQO: Add TX path")
Signed-off-by: Alok Tiwari <alok.a.tiwari@oracle.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
test-large-mtu.sh is referenced by the Makefile
but does not exist.
Add it along the other scripts.
Fixes: 944f8b6aba ("selftest/net/ovpn: extend coverage with more test cases")
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
TCP sockets cannot be created with AF_UNSPEC, but
one among the supported family must be used.
Since commit 944f8b6aba ("selftest/net/ovpn: extend
coverage with more test cases") the default address
family for all tests was changed from AF_INET to AF_UNSPEC,
thus breaking all TCP cases.
Restore AF_INET as default address family for TCP listeners.
Fixes: 944f8b6aba ("selftest/net/ovpn: extend coverage with more test cases")
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
Upon error along the TCP data_ready event path, we have
the following chain of calls:
strp_data_ready()
ovpn_tcp_rcv()
ovpn_peer_del()
ovpn_socket_release()
Since strp_data_ready() may be invoked from softirq context, and
ovpn_socket_release() may sleep, the above sequence may cause a
sleep in atomic context like the following:
BUG: sleeping function called from invalid context at ./ovpn-backports-ovpn-net-next-main-6.15.0-rc5-20250522/drivers/net/ovpn/socket.c:71
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 25, name: ksoftirqd/3
5 locks held by ksoftirqd/3/25:
#0: ffffffe000cd0580 (rcu_read_lock){....}-{1:2}, at: netif_receive_skb+0xb8/0x5b0
OpenVPN/ovpn-backports#1: ffffffe000cd0580 (rcu_read_lock){....}-{1:2}, at: netif_receive_skb+0xb8/0x5b0
OpenVPN/ovpn-backports#2: ffffffe000cd0580 (rcu_read_lock){....}-{1:2}, at: ip_local_deliver_finish+0x66/0x1e0
OpenVPN/ovpn-backports#3: ffffffe003ce9818 (slock-AF_INET/1){+.-.}-{2:2}, at: tcp_v4_rcv+0x156e/0x17a0
OpenVPN/ovpn-backports#4: ffffffe000cd0580 (rcu_read_lock){....}-{1:2}, at: ovpn_tcp_data_ready+0x0/0x1b0 [ovpn]
CPU: 3 PID: 25 Comm: ksoftirqd/3 Not tainted 5.10.104+ #0
Call Trace:
walk_stackframe+0x0/0x1d0
show_stack+0x2e/0x44
dump_stack+0xc2/0x102
___might_sleep+0x29c/0x2b0
__might_sleep+0x62/0xa0
ovpn_socket_release+0x24/0x2d0 [ovpn]
unlock_ovpn+0x6e/0x190 [ovpn]
ovpn_peer_del+0x13c/0x390 [ovpn]
ovpn_tcp_rcv+0x280/0x560 [ovpn]
__strp_recv+0x262/0x940
strp_recv+0x66/0x80
tcp_read_sock+0x122/0x410
strp_data_ready+0x156/0x1f0
ovpn_tcp_data_ready+0x92/0x1b0 [ovpn]
tcp_data_ready+0x6c/0x150
tcp_rcv_established+0xb36/0xc50
tcp_v4_do_rcv+0x25e/0x380
tcp_v4_rcv+0x166a/0x17a0
ip_protocol_deliver_rcu+0x8c/0x250
ip_local_deliver_finish+0xf8/0x1e0
ip_local_deliver+0xc2/0x2d0
ip_rcv+0x1f2/0x330
__netif_receive_skb+0xfc/0x290
netif_receive_skb+0x104/0x5b0
br_pass_frame_up+0x190/0x3f0
br_handle_frame_finish+0x3e2/0x7a0
br_handle_frame+0x750/0xab0
__netif_receive_skb_core.constprop.0+0x4c0/0x17f0
__netif_receive_skb+0xc6/0x290
netif_receive_skb+0x104/0x5b0
xgmac_dma_rx+0x962/0xb40
__napi_poll.constprop.0+0x5a/0x350
net_rx_action+0x1fe/0x4b0
__do_softirq+0x1f8/0x85c
run_ksoftirqd+0x80/0xd0
smpboot_thread_fn+0x1f0/0x3e0
kthread+0x1e6/0x210
ret_from_kernel_thread+0x8/0xc
Fix this issue by postponing the ovpn_peer_del() call to
a scheduled worker, as we already do in ovpn_tcp_send_sock()
for the very same reason.
Fixes: 11851cbd60 ("ovpn: implement TCP transport")
Reported-by: Qingfang Deng <dqfext@gmail.com>
Closes: https://github.com/OpenVPN/ovpn-net-next/issues/13
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
Removing a peer while userspace attempts to close its transport
socket triggers a race condition resulting in the following
crash:
Oops: general protection fault, probably for non-canonical address 0xdffffc0000000077: 0000 [#1] SMP KASAN
KASAN: null-ptr-deref in range [0x00000000000003b8-0x00000000000003bf]
CPU: 12 UID: 0 PID: 162 Comm: kworker/12:1 Tainted: G O 6.15.0-rc2-00635-g521139ac3840 #272 PREEMPT(full)
Tainted: [O]=OOT_MODULE
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-20240910_120124-localhost 04/01/2014
Workqueue: events ovpn_peer_keepalive_work [ovpn]
RIP: 0010:ovpn_socket_release+0x23c/0x500 [ovpn]
Code: ea 03 80 3c 02 00 0f 85 71 02 00 00 48 b8 00 00 00 00 00 fc ff df 4d 8b 64 24 18 49 8d bc 24 be 03 00 00 48 89 fa 48 c1 ea 03 <0f> b6 14 02 48 89 f8 83 e0 07 83 c0 01 38 d0 7c 08 84 d2 0f 85 30
RSP: 0018:ffffc90000c9fb18 EFLAGS: 00010217
RAX: dffffc0000000000 RBX: ffff8881148d7940 RCX: ffffffff817787bb
RDX: 0000000000000077 RSI: 0000000000000008 RDI: 00000000000003be
RBP: ffffc90000c9fb30 R08: 0000000000000000 R09: fffffbfff0d3e840
R10: ffffffff869f4207 R11: 0000000000000000 R12: 0000000000000000
R13: ffff888115eb9300 R14: ffffc90000c9fbc8 R15: 000000000000000c
FS: 0000000000000000(0000) GS:ffff8882b0151000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f37266b6114 CR3: 00000000054a8000 CR4: 0000000000750ef0
PKRU: 55555554
Call Trace:
<TASK>
unlock_ovpn+0x8b/0xe0 [ovpn]
ovpn_peer_keepalive_work+0xe3/0x540 [ovpn]
? ovpn_peers_free+0x780/0x780 [ovpn]
? lock_acquire+0x56/0x70
? process_one_work+0x888/0x1740
process_one_work+0x933/0x1740
? pwq_dec_nr_in_flight+0x10b0/0x10b0
? move_linked_works+0x12d/0x2c0
? assign_work+0x163/0x270
worker_thread+0x4d6/0xd90
? preempt_count_sub+0x4c/0x70
? process_one_work+0x1740/0x1740
kthread+0x36c/0x710
? trace_preempt_on+0x8c/0x1e0
? kthread_is_per_cpu+0xc0/0xc0
? preempt_count_sub+0x4c/0x70
? _raw_spin_unlock_irq+0x36/0x60
? calculate_sigpending+0x7b/0xa0
? kthread_is_per_cpu+0xc0/0xc0
ret_from_fork+0x3a/0x80
? kthread_is_per_cpu+0xc0/0xc0
ret_from_fork_asm+0x11/0x20
</TASK>
Modules linked in: ovpn(O)
This happens because the peer deletion operation reaches
ovpn_socket_release() while ovpn_sock->sock (struct socket *)
and its sk member (struct sock *) are still both valid.
Here synchronize_rcu() is invoked, after which ovpn_sock->sock->sk
becomes NULL, due to the concurrent socket closing triggered
from userspace.
After having invoked synchronize_rcu(), ovpn_socket_release() will
attempt dereferencing ovpn_sock->sock->sk, triggering the crash
reported above.
The reason for accessing sk is that we need to retrieve its
protocol and continue the cleanup routine accordingly.
This crash can be easily produced by running openvpn userspace in
client mode with `--keepalive 10 20`, while entirely omitting this
option on the server side.
After 20 seconds ovpn will assume the peer (server) to be dead,
will start removing it and will notify userspace. The latter will
receive the notification and close the transport socket, thus
triggering the crash.
To fix the race condition for good, we need to refactor struct ovpn_socket.
Since ovpn is always only interested in the sock->sk member (struct sock *)
we can directly hold a reference to it, raher than accessing it via
its struct socket container.
This means changing "struct socket *ovpn_socket->sock" to
"struct sock *ovpn_socket->sk".
While acquiring a reference to sk, we can increase its refcounter
without affecting the socket close()/destroy() notification
(which we rely on when userspace closes a socket we are using).
By increasing sk's refcounter we know we can dereference it
in ovpn_socket_release() without incurring in any race condition
anymore.
ovpn_socket_release() will ultimately decrease the reference
counter.
Cc: Oleksandr Natalenko <oleksandr@natalenko.name>
Fixes: 11851cbd60 ("ovpn: implement TCP transport")
Reported-by: Qingfang Deng <dqfext@gmail.com>
Closes: https://github.com/OpenVPN/ovpn-net-next/issues/1
Tested-by: Gert Doering <gert@greenie.muc.de>
Link: https://www.mail-archive.com/openvpn-devel@lists.sourceforge.net/msg31575.html
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
When deconfiguring a UDP-tunnel from a socket, we cannot
call setup_udp_tunnel_sock() with an empty config, because
this helper is expected to be invoked only during setup.
Get rid of the call to setup_udp_tunnel_sock() and just
revert what it did during socket initialization..
Note that the global udp_encap_needed_key and the GRO state
are left untouched: udp_destroy_socket() will eventually
take care of them.
Cc: Sabrina Dubroca <sd@queasysnail.net>
Cc: Oleksandr Natalenko <oleksandr@natalenko.name>
Fixes: ab66abbc76 ("ovpn: implement basic RX path (UDP)")
Reported-by: Paolo Abeni <pabeni@redhat.com>
Closes: https://lore.kernel.org/netdev/1a47ce02-fd42-4761-8697-f3f315011cc6@redhat.com
Signed-off-by: Antonio Quartulli <antonio@openvpn.net>
Lorenzo Bianconi says:
====================
net: airoha: Fix IPv6 hw acceleration
Fix IPv6 hw acceleration in bridge mode resolving ib2 and
airoha_foe_mac_info_common overwrite in
airoha_ppe_foe_commit_subflow_entry routine.
Introduce UPDMEM source-mac table used to set source mac address for
L3 IPv6 hw accelerated traffic.
====================
Link: https://patch.msgid.link/20250602-airoha-flowtable-ipv6-fix-v2-0-3287f8b55214@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Set PPE entry smac_id field to 0xf in airoha_ppe_foe_commit_subflow_entry
routine for IPv6 traffic in order to instruct the hw to keep original
source mac address for IPv6 hw accelerated traffic in bridge mode.
Fixes: cd53f62261 ("net: airoha: Add L2 hw acceleration support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250602-airoha-flowtable-ipv6-fix-v2-3-3287f8b55214@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
ib2 and airoha_foe_mac_info_common have not the same offsets in
airoha_foe_bridge and airoha_foe_ipv6 structures. Current codebase does
not accelerate IPv6 traffic in bridge mode since ib2 and l2 info are not
set properly copying airoha_foe_bridge struct into airoha_foe_ipv6 one
in airoha_ppe_foe_commit_subflow_entry routine.
Fix IPv6 hw acceleration in bridge mode resolving ib2 and
airoha_foe_mac_info_common overwrite in
airoha_ppe_foe_commit_subflow_entry() and configuring them with proper
values.
Fixes: cd53f62261 ("net: airoha: Add L2 hw acceleration support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250602-airoha-flowtable-ipv6-fix-v2-2-3287f8b55214@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
UPDMEM source-mac table is a key-value map used to store devices mac
addresses according to the port identifier. UPDMEM source mac table is
used during IPv6 traffic hw acceleration since PPE entries, for space
constraints, do not contain the full source mac address but just the
identifier in the UPDMEM source-mac table.
Configure UPDMEM source-mac table with device mac addresses and set
the source-mac ID field for PPE IPv6 entries in order to select the
proper device mac address as source mac for L3 IPv6 hw accelerated traffic.
Fixes: 00a7678310 ("net: airoha: Introduce flowtable offload support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250602-airoha-flowtable-ipv6-fix-v2-1-3287f8b55214@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
We have the logic to include net/lib automatically for net related
selftests. However, currently, this logic is only in install target
which means only `make install` will have net/lib included. This commit
adds the logic to all target so that all `make`, `make run_tests` and
`make install` will have net/lib included in net related selftests.
Signed-off-by: Bui Quang Minh <minhquangbui99@gmail.com>
Link: https://patch.msgid.link/20250601142914.13379-1-minhquangbui99@gmail.com
Fixes: b86761ff63 ("selftests: net: add scaffolding for Netlink tests in Python")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Commit 3d010c8031 ("udp: do not accept non-tunnel GSO skbs landing
in a tunnel") added checks in linux stack to not accept non-tunnel
GRO packets landing in a tunnel. This exposed an issue in vmxnet3
which was not correctly reporting GRO packets for tunnel packets.
This patch fixes this issue by setting correct GSO type for the
tunnel packets.
Currently, vmxnet3 does not support reporting inner fields for LRO
tunnel packets. The issue is not seen for egress drivers that do not
use skb inner fields. The workaround is to enable tnl-segmentation
offload on the egress interfaces if the driver supports it. This
problem pre-exists this patch fix and can be addressed as a separate
future patch.
Fixes: dacce2be33 ("vmxnet3: add geneve and vxlan tunnel offload support")
Signed-off-by: Ronak Doshi <ronak.doshi@broadcom.com>
Acked-by: Guolin Yang <guolin.yang@broadcom.com>
Link: https://patch.msgid.link/20250530152701.70354-1-ronak.doshi@broadcom.com
[pabeni@redhat.com: dropped the changelog]
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This reverts commit a571a9a1b1.
The commit in question breaks kunit for older compilers:
$ gcc --version
gcc (GCC) 11.5.0 20240719 (Red Hat 11.5.0-5)
$ ./tools/testing/kunit/kunit.py run --alltests --json --arch=x86_64
Configuring KUnit Kernel ...
Regenerating .config ...
Populating config with:
$ make ARCH=x86_64 O=.kunit olddefconfig
ERROR:root:Not all Kconfig options selected in kunitconfig were in the generated .config.
This is probably due to unsatisfied dependencies.
Missing: CONFIG_INIT_STACK_ALL_PATTERN=y
Link: https://lore.kernel.org/20250529083811.778bc31b@kernel.org
Fixes: a571a9a1b1 ("kunit: configs: Enable CONFIG_INIT_STACK_ALL_PATTERN in all_tests")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Mark Brown <broonie@kernel.org>
Acked-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://patch.msgid.link/20250530135800.13437-1-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When driver handles the napi rx polling requests, the netdev might
have been released by the dellink logic triggered by the disconnect
operation on user plane. However, in the logic of processing skb in
polling, an invalid netdev is still being used, which causes a panic.
BUG: kernel NULL pointer dereference, address: 00000000000000f1
Oops: 0000 [#1] PREEMPT SMP NOPTI
RIP: 0010:dev_gro_receive+0x3a/0x620
[...]
Call Trace:
<IRQ>
? __die_body+0x68/0xb0
? page_fault_oops+0x379/0x3e0
? exc_page_fault+0x4f/0xa0
? asm_exc_page_fault+0x22/0x30
? __pfx_t7xx_ccmni_recv_skb+0x10/0x10 [mtk_t7xx (HASH:1400 7)]
? dev_gro_receive+0x3a/0x620
napi_gro_receive+0xad/0x170
t7xx_ccmni_recv_skb+0x48/0x70 [mtk_t7xx (HASH:1400 7)]
t7xx_dpmaif_napi_rx_poll+0x590/0x800 [mtk_t7xx (HASH:1400 7)]
net_rx_action+0x103/0x470
irq_exit_rcu+0x13a/0x310
sysvec_apic_timer_interrupt+0x56/0x90
</IRQ>
Fixes: 5545b7b9f2 ("net: wwan: t7xx: Add NAPI support")
Signed-off-by: Jinjian Song <jinjian.song@fibocom.com>
Link: https://patch.msgid.link/20250530031648.5592-1-jinjian.song@fibocom.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2025-05-30 (ice, idpf)
For ice:
Michal resolves XDP issues related to Tx scheduler configuration with
large number of Tx queues.
Additional information:
https://lore.kernel.org/intel-wired-lan/20250513105529.241745-1-michal.kubiak@intel.com/
For idpf:
Brian Vazquez updates netif_subqueue_maybe_stop() condition check to
prevent possible races.
Emil shuts down virtchannel mailbox during reset to reduce timeout
delays as it's unavailable during that time.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
idpf: avoid mailbox timeout delays during reset
idpf: fix a race in txq wakeup
ice: fix rebuilding the Tx scheduler tree for large queue counts
ice: create new Tx scheduler nodes for new queues only
ice: fix Tx scheduler error handling in XDP callback
====================
Link: https://patch.msgid.link/20250530211221.2170484-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit a1e40ac5b5 ("net: gso: fix udp gso fraglist segmentation after
pull from frag_list") detected invalid geometry in frag_list skbs and
redirects them from skb_segment_list to more robust skb_segment. But some
packets with modified geometry can also hit bugs in that code. We don't
know how many such cases exist. Addressing each one by one also requires
touching the complex skb_segment code, which risks introducing bugs for
other types of skbs. Instead, linearize all these packets that fail the
basic invariants on gso fraglist skbs. That is more robust.
If only part of the fraglist payload is pulled into head_skb, it will
always cause exception when splitting skbs by skb_segment. For detailed
call stack information, see below.
Valid SKB_GSO_FRAGLIST skbs
- consist of two or more segments
- the head_skb holds the protocol headers plus first gso_size
- one or more frag_list skbs hold exactly one segment
- all but the last must be gso_size
Optional datapath hooks such as NAT and BPF (bpf_skb_pull_data) can
modify fraglist skbs, breaking these invariants.
In extreme cases they pull one part of data into skb linear. For UDP,
this causes three payloads with lengths of (11,11,10) bytes were
pulled tail to become (12,10,10) bytes.
The skbs no longer meets the above SKB_GSO_FRAGLIST conditions because
payload was pulled into head_skb, it needs to be linearized before pass
to regular skb_segment.
skb_segment+0xcd0/0xd14
__udp_gso_segment+0x334/0x5f4
udp4_ufo_fragment+0x118/0x15c
inet_gso_segment+0x164/0x338
skb_mac_gso_segment+0xc4/0x13c
__skb_gso_segment+0xc4/0x124
validate_xmit_skb+0x9c/0x2c0
validate_xmit_skb_list+0x4c/0x80
sch_direct_xmit+0x70/0x404
__dev_queue_xmit+0x64c/0xe5c
neigh_resolve_output+0x178/0x1c4
ip_finish_output2+0x37c/0x47c
__ip_finish_output+0x194/0x240
ip_finish_output+0x20/0xf4
ip_output+0x100/0x1a0
NF_HOOK+0xc4/0x16c
ip_forward+0x314/0x32c
ip_rcv+0x90/0x118
__netif_receive_skb+0x74/0x124
process_backlog+0xe8/0x1a4
__napi_poll+0x5c/0x1f8
net_rx_action+0x154/0x314
handle_softirqs+0x154/0x4b8
[118.376811] [C201134] rxq0_pus: [name:bug&]kernel BUG at net/core/skbuff.c:4278!
[118.376829] [C201134] rxq0_pus: [name:traps&]Internal error: Oops - BUG: 00000000f2000800 [#1] PREEMPT SMP
[118.470774] [C201134] rxq0_pus: [name:mrdump&]Kernel Offset: 0x178cc00000 from 0xffffffc008000000
[118.470810] [C201134] rxq0_pus: [name:mrdump&]PHYS_OFFSET: 0x40000000
[118.470827] [C201134] rxq0_pus: [name:mrdump&]pstate: 60400005 (nZCv daif +PAN -UAO)
[118.470848] [C201134] rxq0_pus: [name:mrdump&]pc : [0xffffffd79598aefc] skb_segment+0xcd0/0xd14
[118.470900] [C201134] rxq0_pus: [name:mrdump&]lr : [0xffffffd79598a5e8] skb_segment+0x3bc/0xd14
[118.470928] [C201134] rxq0_pus: [name:mrdump&]sp : ffffffc008013770
Fixes: a1e40ac5b5 ("gso: fix udp gso fraglist segmentation after pull from frag_list")
Signed-off-by: Shiming Cheng <shiming.cheng@mediatek.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Paul Chaignon says:
====================
net: Fix inet_proto_csum_replace_by_diff for IPv6
This patchset fixes a bug that causes skb->csum to hold an incorrect
value when calling inet_proto_csum_replace_by_diff for an IPv6 packet
in CHECKSUM_COMPLETE state. This bug affects BPF helper
bpf_l4_csum_replace and IPv6 ILA in adj-transport mode.
In those cases, inet_proto_csum_replace_by_diff updates the L4 checksum
field after an IPv6 address change. These two changes cancel each other
in terms of checksum, so skb->csum shouldn't be updated.
v2: https://lore.kernel.org/aCz84JU60wd8etiT@mail.gmail.com
====================
Link: https://patch.msgid.link/cover.1748509484.git.paul.chaignon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In Cilium, we use bpf_csum_diff + bpf_l4_csum_replace to, among other
things, update the L4 checksum after reverse SNATing IPv6 packets. That
use case is however not currently supported and leads to invalid
skb->csum values in some cases. This patch adds support for IPv6 address
changes in bpf_l4_csum_update via a new flag.
When calling bpf_l4_csum_replace in Cilium, it ends up calling
inet_proto_csum_replace_by_diff:
1: void inet_proto_csum_replace_by_diff(__sum16 *sum, struct sk_buff *skb,
2: __wsum diff, bool pseudohdr)
3: {
4: if (skb->ip_summed != CHECKSUM_PARTIAL) {
5: csum_replace_by_diff(sum, diff);
6: if (skb->ip_summed == CHECKSUM_COMPLETE && pseudohdr)
7: skb->csum = ~csum_sub(diff, skb->csum);
8: } else if (pseudohdr) {
9: *sum = ~csum_fold(csum_add(diff, csum_unfold(*sum)));
10: }
11: }
The bug happens when we're in the CHECKSUM_COMPLETE state. We've just
updated one of the IPv6 addresses. The helper now updates the L4 header
checksum on line 5. Next, it updates skb->csum on line 7. It shouldn't.
For an IPv6 packet, the updates of the IPv6 address and of the L4
checksum will cancel each other. The checksums are set such that
computing a checksum over the packet including its checksum will result
in a sum of 0. So the same is true here when we update the L4 checksum
on line 5. We'll update it as to cancel the previous IPv6 address
update. Hence skb->csum should remain untouched in this case.
The same bug doesn't affect IPv4 packets because, in that case, three
fields are updated: the IPv4 address, the IP checksum, and the L4
checksum. The change to the IPv4 address and one of the checksums still
cancel each other in skb->csum, but we're left with one checksum update
and should therefore update skb->csum accordingly. That's exactly what
inet_proto_csum_replace_by_diff does.
This special case for IPv6 L4 checksums is also described atop
inet_proto_csum_replace16, the function we should be using in this case.
This patch introduces a new bpf_l4_csum_replace flag, BPF_F_IPV6,
to indicate that we're updating the L4 checksum of an IPv6 packet. When
the flag is set, inet_proto_csum_replace_by_diff will skip the
skb->csum update.
Fixes: 7d672345ed ("bpf: add generic bpf_csum_diff helper")
Signed-off-by: Paul Chaignon <paul.chaignon@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://patch.msgid.link/96a6bc3a443e6f0b21ff7b7834000e17fb549e05.1748509484.git.paul.chaignon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During ILA address translations, the L4 checksums can be handled in
different ways. One of them, adj-transport, consist in parsing the
transport layer and updating any found checksum. This logic relies on
inet_proto_csum_replace_by_diff and produces an incorrect skb->csum when
in state CHECKSUM_COMPLETE.
This bug can be reproduced with a simple ILA to SIR mapping, assuming
packets are received with CHECKSUM_COMPLETE:
$ ip a show dev eth0
14: eth0@if15: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 62:ae:35:9e:0f:8d brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet6 3333:0:0:1::c078/64 scope global
valid_lft forever preferred_lft forever
inet6 fd00:10:244:1::c078/128 scope global nodad
valid_lft forever preferred_lft forever
inet6 fe80::60ae:35ff:fe9e:f8d/64 scope link proto kernel_ll
valid_lft forever preferred_lft forever
$ ip ila add loc_match fd00:10:244:1 loc 3333:0:0:1 \
csum-mode adj-transport ident-type luid dev eth0
Then I hit [fd00:10:244:1::c078]:8000 with a server listening only on
[3333:0:0:1::c078]:8000. With the bug, the SYN packet is dropped with
SKB_DROP_REASON_TCP_CSUM after inet_proto_csum_replace_by_diff changed
skb->csum. The translation and drop are visible on pwru [1] traces:
IFACE TUPLE FUNC
eth0:9 [fd00:10:244:3::3d8]:51420->[fd00:10:244:1::c078]:8000(tcp) ipv6_rcv
eth0:9 [fd00:10:244:3::3d8]:51420->[fd00:10:244:1::c078]:8000(tcp) ip6_rcv_core
eth0:9 [fd00:10:244:3::3d8]:51420->[fd00:10:244:1::c078]:8000(tcp) nf_hook_slow
eth0:9 [fd00:10:244:3::3d8]:51420->[fd00:10:244:1::c078]:8000(tcp) inet_proto_csum_replace_by_diff
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) tcp_v6_early_demux
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) ip6_route_input
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) ip6_input
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) ip6_input_finish
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) ip6_protocol_deliver_rcu
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) raw6_local_deliver
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) ipv6_raw_deliver
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) tcp_v6_rcv
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) __skb_checksum_complete
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) kfree_skb_reason(SKB_DROP_REASON_TCP_CSUM)
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) skb_release_head_state
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) skb_release_data
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) skb_free_head
eth0:9 [fd00:10:244:3::3d8]:51420->[3333:0:0:1::c078]:8000(tcp) kfree_skbmem
This is happening because inet_proto_csum_replace_by_diff is updating
skb->csum when it shouldn't. The L4 checksum is updated such that it
"cancels" the IPv6 address change in terms of checksum computation, so
the impact on skb->csum is null.
Note this would be different for an IPv4 packet since three fields
would be updated: the IPv4 address, the IP checksum, and the L4
checksum. Two would cancel each other and skb->csum would still need
to be updated to take the L4 checksum change into account.
This patch fixes it by passing an ipv6 flag to
inet_proto_csum_replace_by_diff, to skip the skb->csum update if we're
in the IPv6 case. Note the behavior of the only other user of
inet_proto_csum_replace_by_diff, the BPF subsystem, is left as is in
this patch and fixed in the subsequent patch.
With the fix, using the reproduction from above, I can confirm
skb->csum is not touched by inet_proto_csum_replace_by_diff and the TCP
SYN proceeds to the application after the ILA translation.
Link: https://github.com/cilium/pwru [1]
Fixes: 65d7ab8de5 ("net: Identifier Locator Addressing module")
Signed-off-by: Paul Chaignon <paul.chaignon@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://patch.msgid.link/b5539869e3550d46068504feb02d37653d939c0b.1748509484.git.paul.chaignon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If the ptp_rate recorded earlier in the driver happens to be 0, this
bogus value will propagate up to EST configuration, where it will
trigger a division by 0.
Prevent this division by 0 by adding the corresponding check and error
code.
Suggested-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Fixes: 8572aec3d0 ("net: stmmac: Add basic EST support for XGMAC")
Link: https://patch.msgid.link/20250529-stmmac_tstamp_div-v4-2-d73340a794d5@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The stmmac platform drivers that do not open-code the clk_ptp_rate value
after having retrieved the default one from the device-tree can end up
with 0 in clk_ptp_rate (as clk_get_rate can return 0). It will
eventually propagate up to PTP initialization when bringing up the
interface, leading to a divide by 0:
Division by zero in kernel.
CPU: 1 UID: 0 PID: 1 Comm: swapper/0 Not tainted 6.12.30-00001-g48313bd5768a #22
Hardware name: STM32 (Device Tree Support)
Call trace:
unwind_backtrace from show_stack+0x18/0x1c
show_stack from dump_stack_lvl+0x6c/0x8c
dump_stack_lvl from Ldiv0_64+0x8/0x18
Ldiv0_64 from stmmac_init_tstamp_counter+0x190/0x1a4
stmmac_init_tstamp_counter from stmmac_hw_setup+0xc1c/0x111c
stmmac_hw_setup from __stmmac_open+0x18c/0x434
__stmmac_open from stmmac_open+0x3c/0xbc
stmmac_open from __dev_open+0xf4/0x1ac
__dev_open from __dev_change_flags+0x1cc/0x224
__dev_change_flags from dev_change_flags+0x24/0x60
dev_change_flags from ip_auto_config+0x2e8/0x11a0
ip_auto_config from do_one_initcall+0x84/0x33c
do_one_initcall from kernel_init_freeable+0x1b8/0x214
kernel_init_freeable from kernel_init+0x24/0x140
kernel_init from ret_from_fork+0x14/0x28
Exception stack(0xe0815fb0 to 0xe0815ff8)
Prevent this division by 0 by adding an explicit check and error log
about the actual issue. While at it, remove the same check from
stmmac_ptp_register, which then becomes duplicate
Fixes: 19d857c903 ("stmmac: Fix calculations for ptp counters when clock input = 50Mhz.")
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Reviewed-by: Yanteng Si <si.yanteng@linux.dev>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://patch.msgid.link/20250529-stmmac_tstamp_div-v4-1-d73340a794d5@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The MANA driver's probe registers netdevice via the following call chain:
mana_probe()
register_netdev()
register_netdevice()
register_netdevice() calls notifier callback for netvsc driver,
holding the netdev mutex via netdev_lock_ops().
Further this netvsc notifier callback end up attempting to acquire the
same lock again in dev_xdp_propagate() leading to deadlock.
netvsc_netdev_event()
netvsc_vf_setxdp()
dev_xdp_propagate()
This deadlock was not observed so far because net_shaper_ops was never set,
and thus the lock was effectively a no-op in this case. Fix this by using
netif_xdp_propagate() instead of dev_xdp_propagate() to avoid recursive
locking in this path.
And, since no deadlock is observed on the other path which is via
netvsc_probe, add the lock exclusivly for that path.
Also, clean up the unregistration path by removing the unnecessary call to
netvsc_vf_setxdp(), since unregister_netdevice_many_notify() already
performs this cleanup via dev_xdp_uninstall().
Fixes: 97246d6d21 ("net: hold netdev instance lock during ndo_bpf")
Cc: stable@vger.kernel.org
Signed-off-by: Saurabh Sengar <ssengar@linux.microsoft.com>
Tested-by: Erni Sri Satya Vennela <ernis@linux.microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Subbaraya Sundeep <sbhatta@marvell.com>
Link: https://patch.msgid.link/1748513910-23963-1-git-send-email-ssengar@linux.microsoft.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix the signature of the net_devmem_bind_dmabuf API for
CONFIG_NET_DEVMEM=n.
Fixes: bd61848900 ("net: devmem: Implement TX path")
Signed-off-by: Pranjal Shrivastava <praan@google.com>
Link: https://patch.msgid.link/20250528211058.1826608-1-praan@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
