From f52fe6efd61f54c5cb0e19ef1fde96cf23048a70 Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: Sat, 26 Apr 2025 17:48:57 +0300
Subject: [PATCH 01/16] selftests: net: tsn_lib: add window_size argument to
 isochron_do()

Make out-of-band testing (send a packet when its traffic class gate is
closed, expecting it to be delayed) more predictable by allowing the
window size to be customized by isochron_do().

From man isochron-send, the window size alters the advance time (the
delta between the transmission time of the packet, and its expected TX
time when using SO_TXTIME or tc-taprio on the sender). In absence of the
argument, isochron-send defaults to maximizing the advance time (making
it equal to the cycle length).

The default behavior is exactly what is problematic. An advance time
that is too large will make packets intended to be out-of-band still be
potentially in-band with an open gate from the schedule's previous cycle.
We need to allow that advance time to be reduced.

Perhaps a bit confusingly, isochron_do() has a shift_time argument
currently, but that does not help here. The shift time shifts both the
user space wakeup time and the expected TX time by equal amounts, it is
unable of bringing them closer to one another.

Set the window size properly for the Ocelot PSFP selftest as well.
That used to work due to a very carefully chosen SHIFT_TIME_NS.
I've re-tested that the test still works properly.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20250426144859.3128352-4-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 tools/testing/selftests/drivers/net/ocelot/psfp.sh | 1 +
 tools/testing/selftests/net/forwarding/tsn_lib.sh  | 5 +++++
 2 files changed, 6 insertions(+)

diff --git a/tools/testing/selftests/drivers/net/ocelot/psfp.sh b/tools/testing/selftests/drivers/net/ocelot/psfp.sh
index f96a4bc7120f..8972f42dfe03 100755
--- a/tools/testing/selftests/drivers/net/ocelot/psfp.sh
+++ b/tools/testing/selftests/drivers/net/ocelot/psfp.sh
@@ -266,6 +266,7 @@ run_test()
 		"${base_time}" \
 		"${CYCLE_TIME_NS}" \
 		"${SHIFT_TIME_NS}" \
+		"${GATE_DURATION_NS}" \
 		"${NUM_PKTS}" \
 		"${STREAM_VID}" \
 		"${STREAM_PRIO}" \
diff --git a/tools/testing/selftests/net/forwarding/tsn_lib.sh b/tools/testing/selftests/net/forwarding/tsn_lib.sh
index 19da1ccceac8..bcee7960a39f 100644
--- a/tools/testing/selftests/net/forwarding/tsn_lib.sh
+++ b/tools/testing/selftests/net/forwarding/tsn_lib.sh
@@ -182,6 +182,7 @@ isochron_do()
 	local base_time=$1; shift
 	local cycle_time=$1; shift
 	local shift_time=$1; shift
+	local window_size=$1; shift
 	local num_pkts=$1; shift
 	local vid=$1; shift
 	local priority=$1; shift
@@ -212,6 +213,10 @@ isochron_do()
 		extra_args="${extra_args} --shift-time=${shift_time}"
 	fi
 
+	if ! [ -z "${window_size}" ]; then
+		extra_args="${extra_args} --window-size=${window_size}"
+	fi
+
 	if [ "${use_l2}" = "true" ]; then
 		extra_args="${extra_args} --l2 --etype=0xdead ${vid}"
 		receiver_extra_args="--l2 --etype=0xdead"
-- 
2.51.0


From 4eb9da050f005fbbb7d301e8e99cfdb6e4771a0d Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: Sat, 26 Apr 2025 17:48:58 +0300
Subject: [PATCH 02/16] selftests: net: tc_taprio: new test

Add a forwarding path test for tc-taprio, based on isochron. This is
specifically intended for NICs with an offloaded data path (switchdev/DSA)
and requires taprio 'flags 2'. Also, $h1 and $h2 must support hardware
timestamping, and $h1 tc-etf offload, for isochron to work.

Packets received by a switch while the egress port has a taprio schedule
with an open gate for the traffic class must be sent right away.

Packets received by the switch while the traffic class gate must be
delayed until it opens.

Packets received by the switch must be dropped if the gate for the
traffic class never opens.

Packets should pass if the maximum SDU for the traffic class allows it,
and should be dropped otherwise.

The schedule should auto-update itself if clock jumps take place while
taprio is installed. Repeat most of the above tests after forcing two
clock jumps, one backwards (in Jan 1970) and one back into the present.

Symlink it from tools/testing/selftests/drivers/net/dsa, because usually
DSA ports have the same MAC address, and we need STABLE_MAC_ADDRS=yes
from its forwarding.config for the test to run successfully.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20250426144859.3128352-5-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 .../selftests/drivers/net/dsa/tc_taprio.sh    |   1 +
 .../selftests/net/forwarding/tc_taprio.sh     | 421 ++++++++++++++++++
 .../selftests/net/forwarding/tsn_lib.sh       |  10 +
 3 files changed, 432 insertions(+)
 create mode 120000 tools/testing/selftests/drivers/net/dsa/tc_taprio.sh
 create mode 100755 tools/testing/selftests/net/forwarding/tc_taprio.sh

diff --git a/tools/testing/selftests/drivers/net/dsa/tc_taprio.sh b/tools/testing/selftests/drivers/net/dsa/tc_taprio.sh
new file mode 120000
index 000000000000..d16a65e7595d
--- /dev/null
+++ b/tools/testing/selftests/drivers/net/dsa/tc_taprio.sh
@@ -0,0 +1 @@
+run_net_forwarding_test.sh
\ No newline at end of file
diff --git a/tools/testing/selftests/net/forwarding/tc_taprio.sh b/tools/testing/selftests/net/forwarding/tc_taprio.sh
new file mode 100755
index 000000000000..8992aeabfe0b
--- /dev/null
+++ b/tools/testing/selftests/net/forwarding/tc_taprio.sh
@@ -0,0 +1,421 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+ALL_TESTS=" \
+	test_clock_jump_backward \
+	test_taprio_after_ptp \
+	test_max_sdu \
+	test_clock_jump_backward_forward \
+"
+NUM_NETIFS=4
+source tc_common.sh
+source lib.sh
+source tsn_lib.sh
+
+require_command python3
+
+# The test assumes the usual topology from the README, where h1 is connected to
+# swp1, h2 to swp2, and swp1 and swp2 are together in a bridge.
+# Additional assumption: h1 and h2 use the same PHC, and so do swp1 and swp2.
+# By synchronizing h1 to swp1 via PTP, h2 is also implicitly synchronized to
+# swp1 (and both to CLOCK_REALTIME).
+h1=${NETIFS[p1]}
+swp1=${NETIFS[p2]}
+swp2=${NETIFS[p3]}
+h2=${NETIFS[p4]}
+
+UDS_ADDRESS_H1="/var/run/ptp4l_h1"
+UDS_ADDRESS_SWP1="/var/run/ptp4l_swp1"
+
+H1_IPV4="192.0.2.1"
+H2_IPV4="192.0.2.2"
+H1_IPV6="2001:db8:1::1"
+H2_IPV6="2001:db8:1::2"
+
+# Tunables
+NUM_PKTS=100
+STREAM_VID=10
+STREAM_PRIO_1=6
+STREAM_PRIO_2=5
+STREAM_PRIO_3=4
+# PTP uses TC 0
+ALL_GATES=$((1 << 0 | 1 << STREAM_PRIO_1 | 1 << STREAM_PRIO_2))
+# Use a conservative cycle of 10 ms to allow the test to still pass when the
+# kernel has some extra overhead like lockdep etc
+CYCLE_TIME_NS=10000000
+# Create two Gate Control List entries, one OPEN and one CLOSE, of equal
+# durations
+GATE_DURATION_NS=$((CYCLE_TIME_NS / 2))
+# Give 2/3 of the cycle time to user space and 1/3 to the kernel
+FUDGE_FACTOR=$((CYCLE_TIME_NS / 3))
+# Shift the isochron base time by half the gate time, so that packets are
+# always received by swp1 close to the middle of the time slot, to minimize
+# inaccuracies due to network sync
+SHIFT_TIME_NS=$((GATE_DURATION_NS / 2))
+
+path_delay=
+
+h1_create()
+{
+	simple_if_init $h1 $H1_IPV4/24 $H1_IPV6/64
+}
+
+h1_destroy()
+{
+	simple_if_fini $h1 $H1_IPV4/24 $H1_IPV6/64
+}
+
+h2_create()
+{
+	simple_if_init $h2 $H2_IPV4/24 $H2_IPV6/64
+}
+
+h2_destroy()
+{
+	simple_if_fini $h2 $H2_IPV4/24 $H2_IPV6/64
+}
+
+switch_create()
+{
+	local h2_mac_addr=$(mac_get $h2)
+
+	ip link set $swp1 up
+	ip link set $swp2 up
+
+	ip link add br0 type bridge vlan_filtering 1
+	ip link set $swp1 master br0
+	ip link set $swp2 master br0
+	ip link set br0 up
+
+	bridge vlan add dev $swp2 vid $STREAM_VID
+	bridge vlan add dev $swp1 vid $STREAM_VID
+	bridge fdb add dev $swp2 \
+		$h2_mac_addr vlan $STREAM_VID static master
+}
+
+switch_destroy()
+{
+	ip link del br0
+}
+
+ptp_setup()
+{
+	# Set up swp1 as a master PHC for h1, synchronized to the local
+	# CLOCK_REALTIME.
+	phc2sys_start $UDS_ADDRESS_SWP1
+	ptp4l_start $h1 true $UDS_ADDRESS_H1
+	ptp4l_start $swp1 false $UDS_ADDRESS_SWP1
+}
+
+ptp_cleanup()
+{
+	ptp4l_stop $swp1
+	ptp4l_stop $h1
+	phc2sys_stop
+}
+
+txtime_setup()
+{
+	local if_name=$1
+
+	tc qdisc add dev $if_name clsact
+	# Classify PTP on TC 7 and isochron on TC 6
+	tc filter add dev $if_name egress protocol 0x88f7 \
+		flower action skbedit priority 7
+	tc filter add dev $if_name egress protocol 802.1Q \
+		flower vlan_ethtype 0xdead action skbedit priority 6
+	tc qdisc add dev $if_name handle 100: parent root mqprio num_tc 8 \
+		queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
+		map 0 1 2 3 4 5 6 7 \
+		hw 1
+	# Set up TC 5, 6, 7 for SO_TXTIME. tc-mqprio queues count from 1.
+	tc qdisc replace dev $if_name parent 100:$((STREAM_PRIO_1 + 1)) etf \
+		clockid CLOCK_TAI offload delta $FUDGE_FACTOR
+	tc qdisc replace dev $if_name parent 100:$((STREAM_PRIO_2 + 1)) etf \
+		clockid CLOCK_TAI offload delta $FUDGE_FACTOR
+	tc qdisc replace dev $if_name parent 100:$((STREAM_PRIO_3 + 1)) etf \
+		clockid CLOCK_TAI offload delta $FUDGE_FACTOR
+}
+
+txtime_cleanup()
+{
+	local if_name=$1
+
+	tc qdisc del dev $if_name clsact
+	tc qdisc del dev $if_name root
+}
+
+taprio_replace()
+{
+	local if_name="$1"; shift
+	local extra_args="$1"; shift
+
+	# STREAM_PRIO_1 always has an open gate.
+	# STREAM_PRIO_2 has a gate open for GATE_DURATION_NS (half the cycle time)
+	# STREAM_PRIO_3 always has a closed gate.
+	tc qdisc replace dev $if_name root stab overhead 24 taprio num_tc 8 \
+		queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
+		map 0 1 2 3 4 5 6 7 \
+		sched-entry S $(printf "%x" $ALL_GATES) $GATE_DURATION_NS \
+		sched-entry S $(printf "%x" $((ALL_GATES & ~(1 << STREAM_PRIO_2)))) $GATE_DURATION_NS \
+		base-time 0 flags 0x2 $extra_args
+	taprio_wait_for_admin $if_name
+}
+
+taprio_cleanup()
+{
+	local if_name=$1
+
+	tc qdisc del dev $if_name root
+}
+
+probe_path_delay()
+{
+	local isochron_dat="$(mktemp)"
+	local received
+
+	log_info "Probing path delay"
+
+	isochron_do "$h1" "$h2" "$UDS_ADDRESS_H1" "" 0 \
+		"$CYCLE_TIME_NS" "" "" "$NUM_PKTS" \
+		"$STREAM_VID" "$STREAM_PRIO_1" "" "$isochron_dat"
+
+	received=$(isochron_report_num_received "$isochron_dat")
+	if [ "$received" != "$NUM_PKTS" ]; then
+		echo "Cannot establish basic data path between $h1 and $h2"
+		exit $ksft_fail
+	fi
+
+	printf "pdelay = {}\n" > isochron_data.py
+	isochron report --input-file "$isochron_dat" \
+		--printf-format "pdelay[%u] = %d - %d\n" \
+		--printf-args "qRT" \
+		>> isochron_data.py
+	cat <<-'EOF' > isochron_postprocess.py
+	#!/usr/bin/env python3
+
+	from isochron_data import pdelay
+	import numpy as np
+
+	w = np.array(list(pdelay.values()))
+	print("{}".format(np.max(w)))
+	EOF
+	path_delay=$(python3 ./isochron_postprocess.py)
+
+	log_info "Path delay from $h1 to $h2 estimated at $path_delay ns"
+
+	if [ "$path_delay" -gt "$GATE_DURATION_NS" ]; then
+		echo "Path delay larger than gate duration, aborting"
+		exit $ksft_fail
+	fi
+
+	rm -f ./isochron_data.py 2> /dev/null
+	rm -f ./isochron_postprocess.py 2> /dev/null
+	rm -f "$isochron_dat" 2> /dev/null
+}
+
+setup_prepare()
+{
+	vrf_prepare
+
+	h1_create
+	h2_create
+	switch_create
+
+	txtime_setup $h1
+
+	# Temporarily set up PTP just to probe the end-to-end path delay.
+	ptp_setup
+	probe_path_delay
+	ptp_cleanup
+}
+
+cleanup()
+{
+	pre_cleanup
+
+	isochron_recv_stop
+	txtime_cleanup $h1
+
+	switch_destroy
+	h2_destroy
+	h1_destroy
+
+	vrf_cleanup
+}
+
+run_test()
+{
+	local base_time=$1; shift
+	local stream_prio=$1; shift
+	local expected_delay=$1; shift
+	local should_fail=$1; shift
+	local test_name=$1; shift
+	local isochron_dat="$(mktemp)"
+	local received
+	local median_delay
+
+	RET=0
+
+	# Set the shift time equal to the cycle time, which effectively
+	# cancels the default advance time. Packets won't be sent early in
+	# software, which ensures that they won't prematurely enter through
+	# the open gate in __test_out_of_band(). Also, the gate is open for
+	# long enough that this won't cause a problem in __test_in_band().
+	isochron_do "$h1" "$h2" "$UDS_ADDRESS_H1" "" "$base_time" \
+		"$CYCLE_TIME_NS" "$SHIFT_TIME_NS" "$GATE_DURATION_NS" \
+		"$NUM_PKTS" "$STREAM_VID" "$stream_prio" "" "$isochron_dat"
+
+	received=$(isochron_report_num_received "$isochron_dat")
+	[ "$received" = "$NUM_PKTS" ]
+	check_err_fail $should_fail $? "Reception of $NUM_PKTS packets"
+
+	if [ $should_fail = 0 ] && [ "$received" = "$NUM_PKTS" ]; then
+		printf "pdelay = {}\n" > isochron_data.py
+		isochron report --input-file "$isochron_dat" \
+			--printf-format "pdelay[%u] = %d - %d\n" \
+			--printf-args "qRT" \
+			>> isochron_data.py
+		cat <<-'EOF' > isochron_postprocess.py
+		#!/usr/bin/env python3
+
+		from isochron_data import pdelay
+		import numpy as np
+
+		w = np.array(list(pdelay.values()))
+		print("{}".format(int(np.median(w))))
+		EOF
+		median_delay=$(python3 ./isochron_postprocess.py)
+
+		# If the condition below is true, packets were delayed by a closed gate
+		[ "$median_delay" -gt $((path_delay + expected_delay)) ]
+		check_fail $? "Median delay $median_delay is greater than expected delay $expected_delay plus path delay $path_delay"
+
+		# If the condition below is true, packets were sent expecting them to
+		# hit a closed gate in the switch, but were not delayed
+		[ "$expected_delay" -gt 0 ] && [ "$median_delay" -lt "$expected_delay" ]
+		check_fail $? "Median delay $median_delay is less than expected delay $expected_delay"
+	fi
+
+	log_test "$test_name"
+
+	rm -f ./isochron_data.py 2> /dev/null
+	rm -f ./isochron_postprocess.py 2> /dev/null
+	rm -f "$isochron_dat" 2> /dev/null
+}
+
+__test_always_open()
+{
+	run_test 0.000000000 $STREAM_PRIO_1 0 0 "Gate always open"
+}
+
+__test_always_closed()
+{
+	run_test 0.000000000 $STREAM_PRIO_3 0 1 "Gate always closed"
+}
+
+__test_in_band()
+{
+	# Send packets in-band with the OPEN gate entry
+	run_test 0.000000000 $STREAM_PRIO_2 0 0 "In band with gate"
+}
+
+__test_out_of_band()
+{
+	# Send packets in-band with the CLOSE gate entry
+	run_test 0.005000000 $STREAM_PRIO_2 \
+		$((GATE_DURATION_NS - SHIFT_TIME_NS)) 0 \
+		"Out of band with gate"
+}
+
+run_subtests()
+{
+	__test_always_open
+	__test_always_closed
+	__test_in_band
+	__test_out_of_band
+}
+
+test_taprio_after_ptp()
+{
+	log_info "Setting up taprio after PTP"
+	ptp_setup
+	taprio_replace $swp2
+	run_subtests
+	taprio_cleanup $swp2
+	ptp_cleanup
+}
+
+__test_under_max_sdu()
+{
+	# Limit max-sdu for STREAM_PRIO_1
+	taprio_replace "$swp2" "max-sdu 0 0 0 0 0 0 100 0"
+	run_test 0.000000000 $STREAM_PRIO_1 0 0 "Under maximum SDU"
+}
+
+__test_over_max_sdu()
+{
+	# Limit max-sdu for STREAM_PRIO_1
+	taprio_replace "$swp2" "max-sdu 0 0 0 0 0 0 20 0"
+	run_test 0.000000000 $STREAM_PRIO_1 0 1 "Over maximum SDU"
+}
+
+test_max_sdu()
+{
+	ptp_setup
+	__test_under_max_sdu
+	__test_over_max_sdu
+	taprio_cleanup $swp2
+	ptp_cleanup
+}
+
+# Perform a clock jump in the past without synchronization running, so that the
+# time base remains where it was set by phc_ctl.
+test_clock_jump_backward()
+{
+	# This is a more complex schedule specifically crafted in a way that
+	# has been problematic on NXP LS1028A. Not much to test with it other
+	# than the fact that it passes traffic.
+	tc qdisc replace dev $swp2 root stab overhead 24 taprio num_tc 8 \
+		queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 map 0 1 2 3 4 5 6 7 \
+		base-time 0 sched-entry S 20 300000 sched-entry S 10 200000 \
+		sched-entry S 20 300000 sched-entry S 48 200000 \
+		sched-entry S 20 300000 sched-entry S 83 200000 \
+		sched-entry S 40 300000 sched-entry S 00 200000 flags 2
+
+	log_info "Forcing a backward clock jump"
+	phc_ctl $swp1 set 0
+
+	ping_test $h1 192.0.2.2
+	taprio_cleanup $swp2
+}
+
+# Test that taprio tolerates clock jumps.
+# Since ptp4l and phc2sys are running, it is expected for the time to
+# eventually recover (through yet another clock jump). Isochron waits
+# until that is the case.
+test_clock_jump_backward_forward()
+{
+	log_info "Forcing a backward and a forward clock jump"
+	taprio_replace $swp2
+	phc_ctl $swp1 set 0
+	ptp_setup
+	ping_test $h1 192.0.2.2
+	run_subtests
+	ptp_cleanup
+	taprio_cleanup $swp2
+}
+
+tc_offload_check
+if [[ $? -ne 0 ]]; then
+	log_test_skip "Could not test offloaded functionality"
+	exit $EXIT_STATUS
+fi
+
+trap cleanup EXIT
+
+setup_prepare
+setup_wait
+tests_run
+
+exit $EXIT_STATUS
diff --git a/tools/testing/selftests/net/forwarding/tsn_lib.sh b/tools/testing/selftests/net/forwarding/tsn_lib.sh
index bcee7960a39f..08c044ff6689 100644
--- a/tools/testing/selftests/net/forwarding/tsn_lib.sh
+++ b/tools/testing/selftests/net/forwarding/tsn_lib.sh
@@ -2,6 +2,8 @@
 # SPDX-License-Identifier: GPL-2.0
 # Copyright 2021-2022 NXP
 
+tc_testing_scripts_dir=$(dirname $0)/../../tc-testing/scripts
+
 REQUIRE_ISOCHRON=${REQUIRE_ISOCHRON:=yes}
 REQUIRE_LINUXPTP=${REQUIRE_LINUXPTP:=yes}
 
@@ -18,6 +20,7 @@ fi
 if [[ "$REQUIRE_LINUXPTP" = "yes" ]]; then
 	require_command phc2sys
 	require_command ptp4l
+	require_command phc_ctl
 fi
 
 phc2sys_start()
@@ -263,3 +266,10 @@ isochron_report_num_received()
 		--printf-format "%u\n" --printf-args "R" | \
 		grep -w -v '0' | wc -l
 }
+
+taprio_wait_for_admin()
+{
+	local if_name="$1"; shift
+
+	"$tc_testing_scripts_dir/taprio_wait_for_admin.sh" "$(which tc)" "$if_name"
+}
-- 
2.51.0


From b936a9b8d4a585ccb6d454921c36286bfe63e01d Mon Sep 17 00:00:00 2001
From: Felix Fietkau <nbd@nbd.name>
Date: Sat, 26 Apr 2025 17:32:09 +0200
Subject: [PATCH 03/16] net: ipv6: fix UDPv6 GSO segmentation with NAT

If any address or port is changed, update it in all packets and recalculate
checksum.

Fixes: 9fd1ff5d2ac7 ("udp: Support UDP fraglist GRO/GSO.")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20250426153210.14044-1-nbd@nbd.name
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 net/ipv4/udp_offload.c | 61 +++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 60 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
index 2c0725583be3..9a8142ccbabe 100644
--- a/net/ipv4/udp_offload.c
+++ b/net/ipv4/udp_offload.c
@@ -247,6 +247,62 @@ static struct sk_buff *__udpv4_gso_segment_list_csum(struct sk_buff *segs)
 	return segs;
 }
 
+static void __udpv6_gso_segment_csum(struct sk_buff *seg,
+				     struct in6_addr *oldip,
+				     const struct in6_addr *newip,
+				     __be16 *oldport, __be16 newport)
+{
+	struct udphdr *uh = udp_hdr(seg);
+
+	if (ipv6_addr_equal(oldip, newip) && *oldport == newport)
+		return;
+
+	if (uh->check) {
+		inet_proto_csum_replace16(&uh->check, seg, oldip->s6_addr32,
+					  newip->s6_addr32, true);
+
+		inet_proto_csum_replace2(&uh->check, seg, *oldport, newport,
+					 false);
+		if (!uh->check)
+			uh->check = CSUM_MANGLED_0;
+	}
+
+	*oldip = *newip;
+	*oldport = newport;
+}
+
+static struct sk_buff *__udpv6_gso_segment_list_csum(struct sk_buff *segs)
+{
+	const struct ipv6hdr *iph;
+	const struct udphdr *uh;
+	struct ipv6hdr *iph2;
+	struct sk_buff *seg;
+	struct udphdr *uh2;
+
+	seg = segs;
+	uh = udp_hdr(seg);
+	iph = ipv6_hdr(seg);
+	uh2 = udp_hdr(seg->next);
+	iph2 = ipv6_hdr(seg->next);
+
+	if (!(*(const u32 *)&uh->source ^ *(const u32 *)&uh2->source) &&
+	    ipv6_addr_equal(&iph->saddr, &iph2->saddr) &&
+	    ipv6_addr_equal(&iph->daddr, &iph2->daddr))
+		return segs;
+
+	while ((seg = seg->next)) {
+		uh2 = udp_hdr(seg);
+		iph2 = ipv6_hdr(seg);
+
+		__udpv6_gso_segment_csum(seg, &iph2->saddr, &iph->saddr,
+					 &uh2->source, uh->source);
+		__udpv6_gso_segment_csum(seg, &iph2->daddr, &iph->daddr,
+					 &uh2->dest, uh->dest);
+	}
+
+	return segs;
+}
+
 static struct sk_buff *__udp_gso_segment_list(struct sk_buff *skb,
 					      netdev_features_t features,
 					      bool is_ipv6)
@@ -259,7 +315,10 @@ static struct sk_buff *__udp_gso_segment_list(struct sk_buff *skb,
 
 	udp_hdr(skb)->len = htons(sizeof(struct udphdr) + mss);
 
-	return is_ipv6 ? skb : __udpv4_gso_segment_list_csum(skb);
+	if (is_ipv6)
+		return __udpv6_gso_segment_list_csum(skb);
+	else
+		return __udpv4_gso_segment_list_csum(skb);
 }
 
 struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb,
-- 
2.51.0


From 9ab7a709c926c16b4433cf02d04fcbcf35aaab2b Mon Sep 17 00:00:00 2001
From: Shravya KN <shravya.k-n@broadcom.com>
Date: Mon, 28 Apr 2025 15:58:56 -0700
Subject: [PATCH 04/16] bnxt_en: Fix error handling path in bnxt_init_chip()

WARN_ON() is triggered in __flush_work() if bnxt_init_chip() fails
because we call cancel_work_sync() on dim work that has not been
initialized.

WARNING: CPU: 37 PID: 5223 at kernel/workqueue.c:4201 __flush_work.isra.0+0x212/0x230

The driver relies on the BNXT_STATE_NAPI_DISABLED bit to check if dim
work has already been cancelled.  But in the bnxt_open() path,
BNXT_STATE_NAPI_DISABLED is not set and this causes the error
path to think that it needs to cancel the uninitalized dim work.
Fix it by setting BNXT_STATE_NAPI_DISABLED during initialization.
The bit will be cleared when we enable NAPI and initialize dim work.

Fixes: 40452969a506 ("bnxt_en: Fix DIM shutdown")
Suggested-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Shravya KN <shravya.k-n@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 2c8e2c19d854..c4bccc683597 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -11602,6 +11602,9 @@ static void bnxt_init_napi(struct bnxt *bp)
 		poll_fn = bnxt_poll_p5;
 	else if (BNXT_CHIP_TYPE_NITRO_A0(bp))
 		cp_nr_rings--;
+
+	set_bit(BNXT_STATE_NAPI_DISABLED, &bp->state);
+
 	for (i = 0; i < cp_nr_rings; i++) {
 		bnapi = bp->bnapi[i];
 		netif_napi_add_config_locked(bp->dev, &bnapi->napi, poll_fn,
-- 
2.51.0


From 8e6cc9045380f3f0c48ebda2bda5e1abe263388d Mon Sep 17 00:00:00 2001
From: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Date: Mon, 28 Apr 2025 15:58:57 -0700
Subject: [PATCH 05/16] bnxt_en: Fix ethtool selftest output in one of the
 failure cases

When RDMA driver is loaded, running offline self test is not
supported and driver returns failure early. But it is not clearing
the input buffer and hence the application prints some junk
characters for individual test results.

Fix it by clearing the buffer before returning.

Fixes: 895621f1c816 ("bnxt_en: Don't support offline self test when RoCE driver is loaded")
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 48dd5922e4dd..7be37976f3e4 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -4991,6 +4991,7 @@ static void bnxt_self_test(struct net_device *dev, struct ethtool_test *etest,
 	if (!bp->num_tests || !BNXT_PF(bp))
 		return;
 
+	memset(buf, 0, sizeof(u64) * bp->num_tests);
 	if (etest->flags & ETH_TEST_FL_OFFLINE &&
 	    bnxt_ulp_registered(bp->edev)) {
 		etest->flags |= ETH_TEST_FL_FAILED;
@@ -4998,7 +4999,6 @@ static void bnxt_self_test(struct net_device *dev, struct ethtool_test *etest,
 		return;
 	}
 
-	memset(buf, 0, sizeof(u64) * bp->num_tests);
 	if (!netif_running(dev)) {
 		etest->flags |= ETH_TEST_FL_FAILED;
 		return;
-- 
2.51.0


From a63db07e4ecd45b027718168faf7d798bb47bf58 Mon Sep 17 00:00:00 2001
From: Somnath Kotur <somnath.kotur@broadcom.com>
Date: Mon, 28 Apr 2025 15:58:58 -0700
Subject: [PATCH 06/16] bnxt_en: Add missing skb_mark_for_recycle() in
 bnxt_rx_vlan()

If bnxt_rx_vlan() fails because the VLAN protocol ID is invalid,
the SKB is freed but we're missing the call to recycle it.  This
may cause the warning:

"page_pool_release_retry() stalled pool shutdown"

Add the missing skb_mark_for_recycle() in bnxt_rx_vlan().

Fixes: 86b05508f775 ("bnxt_en: Use the unified RX page pool buffers for XDP and non-XDP")
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index c4bccc683597..cfc9ccab39bf 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -2015,6 +2015,7 @@ static struct sk_buff *bnxt_rx_vlan(struct sk_buff *skb, u8 cmp_type,
 	}
 	return skb;
 vlan_err:
+	skb_mark_for_recycle(skb);
 	dev_kfree_skb(skb);
 	return NULL;
 }
-- 
2.51.0


From 1ae04e489dd757e1e61999362f33e7c554c3b9e3 Mon Sep 17 00:00:00 2001
From: Kashyap Desai <kashyap.desai@broadcom.com>
Date: Mon, 28 Apr 2025 15:58:59 -0700
Subject: [PATCH 07/16] bnxt_en: call pci_alloc_irq_vectors() after
 bnxt_reserve_rings()

On some architectures (e.g. ARM), calling pci_alloc_irq_vectors()
will immediately cause the MSIX table to be written.  This will not
work if we haven't called bnxt_reserve_rings() to properly map
the MSIX table to the MSIX vectors reserved by FW.

Fix the FW error recovery path to delay the bnxt_init_int_mode() ->
pci_alloc_irq_vectors() call by removing it from bnxt_hwrm_if_change().
bnxt_request_irq() later in the code path will call it and by then the
MSIX table is properly mapped.

Fixes: 4343838ca5eb ("bnxt_en: Replace deprecated PCI MSIX APIs")
Suggested-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index cfc9ccab39bf..3b2ea36f25a2 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -12399,13 +12399,8 @@ static int bnxt_hwrm_if_change(struct bnxt *bp, bool up)
 				set_bit(BNXT_STATE_ABORT_ERR, &bp->state);
 				return rc;
 			}
+			/* IRQ will be initialized later in bnxt_request_irq()*/
 			bnxt_clear_int_mode(bp);
-			rc = bnxt_init_int_mode(bp);
-			if (rc) {
-				clear_bit(BNXT_STATE_FW_RESET_DET, &bp->state);
-				netdev_err(bp->dev, "init int mode failed\n");
-				return rc;
-			}
 		}
 		rc = bnxt_cancel_reservations(bp, fw_reset);
 	}
-- 
2.51.0


From c2d20a3814d1b57dea6db82229edd702bde9c878 Mon Sep 17 00:00:00 2001
From: Kashyap Desai <kashyap.desai@broadcom.com>
Date: Mon, 28 Apr 2025 15:59:00 -0700
Subject: [PATCH 08/16] bnxt_en: delay pci_alloc_irq_vectors() in the AER path

This patch is similar to the last patch to delay the
pci_alloc_irq_vectors() call in the AER path until after calling
bnxt_reserve_rings().  bnxt_reserve_rings() needs to properly map
the MSIX table first before we call pci_alloc_irq_vectors() which
may immediately write to the MSIX table in some architectures.

Move the bnxt_init_int_mode() call from bnxt_io_slot_reset() to
bnxt_io_resume() after calling bnxt_reserve_rings().

With this change, the AER path may call bnxt_open() ->
bnxt_hwrm_if_change() with bp->irq_tbl set to NULL.  bp->irq_tbl is
cleared when we call bnxt_clear_int_mode() in bnxt_io_slot_reset().
So we cannot use !bp->irq_tbl to detect aborted FW reset.  Add a
new BNXT_FW_RESET_STATE_ABORT to detect aborted FW reset in
bnxt_hwrm_if_change().

Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 17 +++++++++++------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  1 +
 2 files changed, 12 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 3b2ea36f25a2..78e496b0ec26 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -12325,12 +12325,15 @@ static int bnxt_hwrm_if_change(struct bnxt *bp, bool up)
 {
 	struct hwrm_func_drv_if_change_output *resp;
 	struct hwrm_func_drv_if_change_input *req;
-	bool fw_reset = !bp->irq_tbl;
 	bool resc_reinit = false;
 	bool caps_change = false;
 	int rc, retry = 0;
+	bool fw_reset;
 	u32 flags = 0;
 
+	fw_reset = (bp->fw_reset_state == BNXT_FW_RESET_STATE_ABORT);
+	bp->fw_reset_state = 0;
+
 	if (!(bp->fw_cap & BNXT_FW_CAP_IF_CHANGE))
 		return 0;
 
@@ -14833,7 +14836,7 @@ static void bnxt_fw_reset_abort(struct bnxt *bp, int rc)
 	clear_bit(BNXT_STATE_IN_FW_RESET, &bp->state);
 	if (bp->fw_reset_state != BNXT_FW_RESET_STATE_POLL_VF)
 		bnxt_dl_health_fw_status_update(bp, false);
-	bp->fw_reset_state = 0;
+	bp->fw_reset_state = BNXT_FW_RESET_STATE_ABORT;
 	netif_close(bp->dev);
 }
 
@@ -16931,10 +16934,9 @@ static pci_ers_result_t bnxt_io_slot_reset(struct pci_dev *pdev)
 		if (!err)
 			result = PCI_ERS_RESULT_RECOVERED;
 
+		/* IRQ will be initialized later in bnxt_io_resume */
 		bnxt_ulp_irq_stop(bp);
 		bnxt_clear_int_mode(bp);
-		err = bnxt_init_int_mode(bp);
-		bnxt_ulp_irq_restart(bp, err);
 	}
 
 reset_exit:
@@ -16963,10 +16965,13 @@ static void bnxt_io_resume(struct pci_dev *pdev)
 
 	err = bnxt_hwrm_func_qcaps(bp);
 	if (!err) {
-		if (netif_running(netdev))
+		if (netif_running(netdev)) {
 			err = bnxt_open(netdev);
-		else
+		} else {
 			err = bnxt_reserve_rings(bp, true);
+			if (!err)
+				err = bnxt_init_int_mode(bp);
+		}
 	}
 
 	if (!err)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 21726cf56586..bc8b3b7e915d 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -2614,6 +2614,7 @@ struct bnxt {
 #define BNXT_FW_RESET_STATE_POLL_FW	4
 #define BNXT_FW_RESET_STATE_OPENING	5
 #define BNXT_FW_RESET_STATE_POLL_FW_DOWN	6
+#define BNXT_FW_RESET_STATE_ABORT	7
 
 	u16			fw_reset_min_dsecs;
 #define BNXT_DFLT_FW_RST_MIN_DSECS	20
-- 
2.51.0


From ea9376cf68230e05492f22ca45d329f16e262c7b Mon Sep 17 00:00:00 2001
From: Shruti Parab <shruti.parab@broadcom.com>
Date: Mon, 28 Apr 2025 15:59:01 -0700
Subject: [PATCH 09/16] bnxt_en: Fix coredump logic to free allocated buffer

When handling HWRM_DBG_COREDUMP_LIST FW command in
bnxt_hwrm_dbg_dma_data(), the allocated buffer info->dest_buf is
not freed in the error path.  In the normal path, info->dest_buf
is assigned to coredump->data and it will eventually be freed after
the coredump is collected.

Free info->dest_buf immediately inside bnxt_hwrm_dbg_dma_data() in
the error path.

Fixes: c74751f4c392 ("bnxt_en: Return error if FW returns more data than dump length")
Reported-by: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Shruti Parab <shruti.parab@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c
index 5576e7cf8463..9a74793648f6 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c
@@ -116,6 +116,11 @@ static int bnxt_hwrm_dbg_dma_data(struct bnxt *bp, void *msg,
 				memcpy(info->dest_buf + off, dma_buf, len);
 			} else {
 				rc = -ENOBUFS;
+				if (cmn_req->req_type ==
+				    cpu_to_le16(HWRM_DBG_COREDUMP_LIST)) {
+					kfree(info->dest_buf);
+					info->dest_buf = NULL;
+				}
 				break;
 			}
 		}
-- 
2.51.0


From 6b87bd94f34370bbf1dfa59352bed8efab5bf419 Mon Sep 17 00:00:00 2001
From: Shruti Parab <shruti.parab@broadcom.com>
Date: Mon, 28 Apr 2025 15:59:02 -0700
Subject: [PATCH 10/16] bnxt_en: Fix out-of-bound memcpy() during ethtool -w

When retrieving the FW coredump using ethtool, it can sometimes cause
memory corruption:

BUG: KFENCE: memory corruption in __bnxt_get_coredump+0x3ef/0x670 [bnxt_en]
Corrupted memory at 0x000000008f0f30e8 [ ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ] (in kfence-#45):
__bnxt_get_coredump+0x3ef/0x670 [bnxt_en]
ethtool_get_dump_data+0xdc/0x1a0
__dev_ethtool+0xa1e/0x1af0
dev_ethtool+0xa8/0x170
dev_ioctl+0x1b5/0x580
sock_do_ioctl+0xab/0xf0
sock_ioctl+0x1ce/0x2e0
__x64_sys_ioctl+0x87/0xc0
do_syscall_64+0x5c/0xf0
entry_SYSCALL_64_after_hwframe+0x78/0x80

...

This happens when copying the coredump segment list in
bnxt_hwrm_dbg_dma_data() with the HWRM_DBG_COREDUMP_LIST FW command.
The info->dest_buf buffer is allocated based on the number of coredump
segments returned by the FW.  The segment list is then DMA'ed by
the FW and the length of the DMA is returned by FW.  The driver then
copies this DMA'ed segment list to info->dest_buf.

In some cases, this DMA length may exceed the info->dest_buf length
and cause the above BUG condition.  Fix it by capping the copy
length to not exceed the length of info->dest_buf.  The extra
DMA data contains no useful information.

This code path is shared for the HWRM_DBG_COREDUMP_LIST and the
HWRM_DBG_COREDUMP_RETRIEVE FW commands.  The buffering is different
for these 2 FW commands.  To simplify the logic, we need to move
the line to adjust the buffer length for HWRM_DBG_COREDUMP_RETRIEVE
up, so that the new check to cap the copy length will work for both
commands.

Fixes: c74751f4c392 ("bnxt_en: Return error if FW returns more data than dump length")
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Shruti Parab <shruti.parab@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 .../net/ethernet/broadcom/bnxt/bnxt_coredump.c    | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c
index 9a74793648f6..a000d3f630bd 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c
@@ -110,10 +110,19 @@ static int bnxt_hwrm_dbg_dma_data(struct bnxt *bp, void *msg,
 			}
 		}
 
+		if (cmn_req->req_type ==
+				cpu_to_le16(HWRM_DBG_COREDUMP_RETRIEVE))
+			info->dest_buf_size += len;
+
 		if (info->dest_buf) {
 			if ((info->seg_start + off + len) <=
 			    BNXT_COREDUMP_BUF_LEN(info->buf_len)) {
-				memcpy(info->dest_buf + off, dma_buf, len);
+				u16 copylen = min_t(u16, len,
+						    info->dest_buf_size - off);
+
+				memcpy(info->dest_buf + off, dma_buf, copylen);
+				if (copylen < len)
+					break;
 			} else {
 				rc = -ENOBUFS;
 				if (cmn_req->req_type ==
@@ -125,10 +134,6 @@ static int bnxt_hwrm_dbg_dma_data(struct bnxt *bp, void *msg,
 			}
 		}
 
-		if (cmn_req->req_type ==
-				cpu_to_le16(HWRM_DBG_COREDUMP_RETRIEVE))
-			info->dest_buf_size += len;
-
 		if (!(cmn_resp->flags & HWRM_DBG_CMN_FLAGS_MORE))
 			break;
 
-- 
2.51.0


From 02e8be5a032cae0f4ca33c6053c44d83cf4acc93 Mon Sep 17 00:00:00 2001
From: Michael Chan <michael.chan@broadcom.com>
Date: Mon, 28 Apr 2025 15:59:03 -0700
Subject: [PATCH 11/16] bnxt_en: Fix ethtool -d byte order for 32-bit values

For version 1 register dump that includes the PCIe stats, the existing
code incorrectly assumes that all PCIe stats are 64-bit values.  Fix it
by using an array containing the starting and ending index of the 32-bit
values.  The loop in bnxt_get_regs() will use the array to do proper
endian swap for the 32-bit values.

Fixes: b5d600b027eb ("bnxt_en: Add support for 'ethtool -d'")
Reviewed-by: Shruti Parab <shruti.parab@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 38 ++++++++++++++++---
 1 file changed, 32 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 7be37976f3e4..f5d490bf997e 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -2062,6 +2062,17 @@ static int bnxt_get_regs_len(struct net_device *dev)
 	return reg_len;
 }
 
+#define BNXT_PCIE_32B_ENTRY(start, end)			\
+	 { offsetof(struct pcie_ctx_hw_stats, start),	\
+	   offsetof(struct pcie_ctx_hw_stats, end) }
+
+static const struct {
+	u16 start;
+	u16 end;
+} bnxt_pcie_32b_entries[] = {
+	BNXT_PCIE_32B_ENTRY(pcie_ltssm_histogram[0], pcie_ltssm_histogram[3]),
+};
+
 static void bnxt_get_regs(struct net_device *dev, struct ethtool_regs *regs,
 			  void *_p)
 {
@@ -2094,12 +2105,27 @@ static void bnxt_get_regs(struct net_device *dev, struct ethtool_regs *regs,
 	req->pcie_stat_host_addr = cpu_to_le64(hw_pcie_stats_addr);
 	rc = hwrm_req_send(bp, req);
 	if (!rc) {
-		__le64 *src = (__le64 *)hw_pcie_stats;
-		u64 *dst = (u64 *)(_p + BNXT_PXP_REG_LEN);
-		int i;
-
-		for (i = 0; i < sizeof(*hw_pcie_stats) / sizeof(__le64); i++)
-			dst[i] = le64_to_cpu(src[i]);
+		u8 *dst = (u8 *)(_p + BNXT_PXP_REG_LEN);
+		u8 *src = (u8 *)hw_pcie_stats;
+		int i, j;
+
+		for (i = 0, j = 0; i < sizeof(*hw_pcie_stats); ) {
+			if (i >= bnxt_pcie_32b_entries[j].start &&
+			    i <= bnxt_pcie_32b_entries[j].end) {
+				u32 *dst32 = (u32 *)(dst + i);
+
+				*dst32 = le32_to_cpu(*(__le32 *)(src + i));
+				i += 4;
+				if (i > bnxt_pcie_32b_entries[j].end &&
+				    j < ARRAY_SIZE(bnxt_pcie_32b_entries) - 1)
+					j++;
+			} else {
+				u64 *dst64 = (u64 *)(dst + i);
+
+				*dst64 = le64_to_cpu(*(__le64 *)(src + i));
+				i += 8;
+			}
+		}
 	}
 	hwrm_req_drop(bp, req);
 }
-- 
2.51.0


From 927069d5c40c1cfa7b2d13cfc6d7d58bc6f85c50 Mon Sep 17 00:00:00 2001
From: Vadim Fedorenko <vadfed@meta.com>
Date: Wed, 30 Apr 2025 10:03:43 -0700
Subject: [PATCH 12/16] bnxt_en: fix module unload sequence

Recent updates to the PTP part of bnxt changed the way PTP FIFO is
cleared, skbs waiting for TX timestamps are now cleared during
ndo_close() call. To do clearing procedure, the ptp structure must
exist and point to a valid address. Module destroy sequence had ptp
clear code running before netdev close causing invalid memory access and
kernel crash. Change the sequence to destroy ptp structure after device
close.

Fixes: 8f7ae5a85137 ("bnxt_en: improve TX timestamping FIFO configuration")
Reported-by: Taehee Yoo <ap420073@gmail.com>
Closes: https://lore.kernel.org/netdev/CAMArcTWDe2cd41=ub=zzvYifaYcYv-N-csxfqxUvejy_L0D6UQ@mail.gmail.com/
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Tested-by: Taehee Yoo <ap420073@gmail.com>
Link: https://patch.msgid.link/20250430170343.759126-1-vadfed@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 78e496b0ec26..86a5de44b6f3 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -16006,8 +16006,8 @@ static void bnxt_remove_one(struct pci_dev *pdev)
 
 	bnxt_rdma_aux_device_del(bp);
 
-	bnxt_ptp_clear(bp);
 	unregister_netdev(dev);
+	bnxt_ptp_clear(bp);
 
 	bnxt_rdma_aux_device_uninit(bp);
 
-- 
2.51.0


From f920436a44295ca791ebb6dae3f4190142eec703 Mon Sep 17 00:00:00 2001
From: Jibin Zhang <jibin.zhang@mediatek.com>
Date: Tue, 29 Apr 2025 09:59:48 +0800
Subject: [PATCH 13/16] net: use sock_gen_put() when sk_state is TCP_TIME_WAIT

It is possible for a pointer of type struct inet_timewait_sock to be
returned from the functions __inet_lookup_established() and
__inet6_lookup_established(). This can cause a crash when the
returned pointer is of type struct inet_timewait_sock and
sock_put() is called on it. The following is a crash call stack that
shows sk->sk_wmem_alloc being accessed in sk_free() during the call to
sock_put() on a struct inet_timewait_sock pointer. To avoid this issue,
use sock_gen_put() instead of sock_put() when sk->sk_state
is TCP_TIME_WAIT.

mrdump.ko        ipanic() + 120
vmlinux          notifier_call_chain(nr_to_call=-1, nr_calls=0) + 132
vmlinux          atomic_notifier_call_chain(val=0) + 56
vmlinux          panic() + 344
vmlinux          add_taint() + 164
vmlinux          end_report() + 136
vmlinux          kasan_report(size=0) + 236
vmlinux          report_tag_fault() + 16
vmlinux          do_tag_recovery() + 16
vmlinux          __do_kernel_fault() + 88
vmlinux          do_bad_area() + 28
vmlinux          do_tag_check_fault() + 60
vmlinux          do_mem_abort() + 80
vmlinux          el1_abort() + 56
vmlinux          el1h_64_sync_handler() + 124
vmlinux        > 0xFFFFFFC080011294()
vmlinux          __lse_atomic_fetch_add_release(v=0xF2FFFF82A896087C)
vmlinux          __lse_atomic_fetch_sub_release(v=0xF2FFFF82A896087C)
vmlinux          arch_atomic_fetch_sub_release(i=1, v=0xF2FFFF82A896087C)
+ 8
vmlinux          raw_atomic_fetch_sub_release(i=1, v=0xF2FFFF82A896087C)
+ 8
vmlinux          atomic_fetch_sub_release(i=1, v=0xF2FFFF82A896087C) + 8
vmlinux          __refcount_sub_and_test(i=1, r=0xF2FFFF82A896087C,
oldp=0) + 8
vmlinux          __refcount_dec_and_test(r=0xF2FFFF82A896087C, oldp=0) + 8
vmlinux          refcount_dec_and_test(r=0xF2FFFF82A896087C) + 8
vmlinux          sk_free(sk=0xF2FFFF82A8960700) + 28
vmlinux          sock_put() + 48
vmlinux          tcp6_check_fraglist_gro() + 236
vmlinux          tcp6_gro_receive() + 624
vmlinux          ipv6_gro_receive() + 912
vmlinux          dev_gro_receive() + 1116
vmlinux          napi_gro_receive() + 196
ccmni.ko         ccmni_rx_callback() + 208
ccmni.ko         ccmni_queue_recv_skb() + 388
ccci_dpmaif.ko   dpmaif_rxq_push_thread() + 1088
vmlinux          kthread() + 268
vmlinux          0xFFFFFFC08001F30C()

Fixes: c9d1d23e5239 ("net: add heuristic for enabling TCP fraglist GRO")
Signed-off-by: Jibin Zhang <jibin.zhang@mediatek.com>
Signed-off-by: Shiming Cheng <shiming.cheng@mediatek.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20250429020412.14163-1-shiming.cheng@mediatek.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 net/ipv4/tcp_offload.c   | 2 +-
 net/ipv6/tcpv6_offload.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
index 934f777f29d3..d293087b426d 100644
--- a/net/ipv4/tcp_offload.c
+++ b/net/ipv4/tcp_offload.c
@@ -439,7 +439,7 @@ static void tcp4_check_fraglist_gro(struct list_head *head, struct sk_buff *skb,
 				       iif, sdif);
 	NAPI_GRO_CB(skb)->is_flist = !sk;
 	if (sk)
-		sock_put(sk);
+		sock_gen_put(sk);
 }
 
 INDIRECT_CALLABLE_SCOPE
diff --git a/net/ipv6/tcpv6_offload.c b/net/ipv6/tcpv6_offload.c
index d9b11fe41bf0..a8a04f441e78 100644
--- a/net/ipv6/tcpv6_offload.c
+++ b/net/ipv6/tcpv6_offload.c
@@ -42,7 +42,7 @@ static void tcp6_check_fraglist_gro(struct list_head *head, struct sk_buff *skb,
 					iif, sdif);
 	NAPI_GRO_CB(skb)->is_flist = !sk;
 	if (sk)
-		sock_put(sk);
+		sock_gen_put(sk);
 #endif /* IS_ENABLED(CONFIG_IPV6) */
 }
 
-- 
2.51.0


From e98386d79a23c57cf179fe4138322e277aa3aa74 Mon Sep 17 00:00:00 2001
From: Sagi Maimon <sagi.maimon@adtran.com>
Date: Tue, 29 Apr 2025 10:33:20 +0300
Subject: [PATCH 14/16] ptp: ocp: Fix NULL dereference in Adva board SMA sysfs
 operations

On Adva boards, SMA sysfs store/get operations can call
__handle_signal_outputs() or __handle_signal_inputs() while the `irig`
and `dcf` pointers are uninitialized, leading to a NULL pointer
dereference in __handle_signal() and causing a kernel crash. Adva boards
don't use `irig` or `dcf` functionality, so add Adva-specific callbacks
`ptp_ocp_sma_adva_set_outputs()` and `ptp_ocp_sma_adva_set_inputs()` that
avoid invoking `irig` or `dcf` input/output routines.

Fixes: ef61f5528fca ("ptp: ocp: add Adva timecard support")
Signed-off-by: Sagi Maimon <maimon.sagi@gmail.com>
Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20250429073320.33277-1-maimon.sagi@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 drivers/ptp/ptp_ocp.c | 52 +++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 50 insertions(+), 2 deletions(-)

diff --git a/drivers/ptp/ptp_ocp.c b/drivers/ptp/ptp_ocp.c
index faf6e027f89a..2ccdca4f6960 100644
--- a/drivers/ptp/ptp_ocp.c
+++ b/drivers/ptp/ptp_ocp.c
@@ -2578,12 +2578,60 @@ static const struct ocp_sma_op ocp_fb_sma_op = {
 	.set_output	= ptp_ocp_sma_fb_set_output,
 };
 
+static int
+ptp_ocp_sma_adva_set_output(struct ptp_ocp *bp, int sma_nr, u32 val)
+{
+	u32 reg, mask, shift;
+	unsigned long flags;
+	u32 __iomem *gpio;
+
+	gpio = sma_nr > 2 ? &bp->sma_map1->gpio2 : &bp->sma_map2->gpio2;
+	shift = sma_nr & 1 ? 0 : 16;
+
+	mask = 0xffff << (16 - shift);
+
+	spin_lock_irqsave(&bp->lock, flags);
+
+	reg = ioread32(gpio);
+	reg = (reg & mask) | (val << shift);
+
+	iowrite32(reg, gpio);
+
+	spin_unlock_irqrestore(&bp->lock, flags);
+
+	return 0;
+}
+
+static int
+ptp_ocp_sma_adva_set_inputs(struct ptp_ocp *bp, int sma_nr, u32 val)
+{
+	u32 reg, mask, shift;
+	unsigned long flags;
+	u32 __iomem *gpio;
+
+	gpio = sma_nr > 2 ? &bp->sma_map2->gpio1 : &bp->sma_map1->gpio1;
+	shift = sma_nr & 1 ? 0 : 16;
+
+	mask = 0xffff << (16 - shift);
+
+	spin_lock_irqsave(&bp->lock, flags);
+
+	reg = ioread32(gpio);
+	reg = (reg & mask) | (val << shift);
+
+	iowrite32(reg, gpio);
+
+	spin_unlock_irqrestore(&bp->lock, flags);
+
+	return 0;
+}
+
 static const struct ocp_sma_op ocp_adva_sma_op = {
 	.tbl		= { ptp_ocp_adva_sma_in, ptp_ocp_adva_sma_out },
 	.init		= ptp_ocp_sma_fb_init,
 	.get		= ptp_ocp_sma_fb_get,
-	.set_inputs	= ptp_ocp_sma_fb_set_inputs,
-	.set_output	= ptp_ocp_sma_fb_set_output,
+	.set_inputs	= ptp_ocp_sma_adva_set_inputs,
+	.set_output	= ptp_ocp_sma_adva_set_output,
 };
 
 static int
-- 
2.51.0


From 2d52e2e38b85c8b7bc00dca55c2499f46f8c8198 Mon Sep 17 00:00:00 2001
From: Thangaraj Samynathan <thangaraj.s@microchip.com>
Date: Tue, 29 Apr 2025 10:55:27 +0530
Subject: [PATCH 15/16] net: lan743x: Fix memleak issue when GSO enabled

Always map the `skb` to the LS descriptor. Previously skb was
mapped to EXT descriptor when the number of fragments is zero with
GSO enabled. Mapping the skb to EXT descriptor prevents it from
being freed, leading to a memory leak

Fixes: 23f0703c125b ("lan743x: Add main source files for new lan743x driver")
Signed-off-by: Thangaraj Samynathan <thangaraj.s@microchip.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20250429052527.10031-1-thangaraj.s@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 drivers/net/ethernet/microchip/lan743x_main.c | 8 ++++++--
 drivers/net/ethernet/microchip/lan743x_main.h | 1 +
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
index 23760b613d3e..e2d6bfb5d693 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.c
+++ b/drivers/net/ethernet/microchip/lan743x_main.c
@@ -1815,6 +1815,7 @@ static void lan743x_tx_frame_add_lso(struct lan743x_tx *tx,
 	if (nr_frags <= 0) {
 		tx->frame_data0 |= TX_DESC_DATA0_LS_;
 		tx->frame_data0 |= TX_DESC_DATA0_IOC_;
+		tx->frame_last = tx->frame_first;
 	}
 	tx_descriptor = &tx->ring_cpu_ptr[tx->frame_tail];
 	tx_descriptor->data0 = cpu_to_le32(tx->frame_data0);
@@ -1884,6 +1885,7 @@ static int lan743x_tx_frame_add_fragment(struct lan743x_tx *tx,
 		tx->frame_first = 0;
 		tx->frame_data0 = 0;
 		tx->frame_tail = 0;
+		tx->frame_last = 0;
 		return -ENOMEM;
 	}
 
@@ -1924,16 +1926,18 @@ static void lan743x_tx_frame_end(struct lan743x_tx *tx,
 	    TX_DESC_DATA0_DTYPE_DATA_) {
 		tx->frame_data0 |= TX_DESC_DATA0_LS_;
 		tx->frame_data0 |= TX_DESC_DATA0_IOC_;
+		tx->frame_last = tx->frame_tail;
 	}
 
-	tx_descriptor = &tx->ring_cpu_ptr[tx->frame_tail];
-	buffer_info = &tx->buffer_info[tx->frame_tail];
+	tx_descriptor = &tx->ring_cpu_ptr[tx->frame_last];
+	buffer_info = &tx->buffer_info[tx->frame_last];
 	buffer_info->skb = skb;
 	if (time_stamp)
 		buffer_info->flags |= TX_BUFFER_INFO_FLAG_TIMESTAMP_REQUESTED;
 	if (ignore_sync)
 		buffer_info->flags |= TX_BUFFER_INFO_FLAG_IGNORE_SYNC;
 
+	tx_descriptor = &tx->ring_cpu_ptr[tx->frame_tail];
 	tx_descriptor->data0 = cpu_to_le32(tx->frame_data0);
 	tx->frame_tail = lan743x_tx_next_index(tx, tx->frame_tail);
 	tx->last_tail = tx->frame_tail;
diff --git a/drivers/net/ethernet/microchip/lan743x_main.h b/drivers/net/ethernet/microchip/lan743x_main.h
index 7f73d66854be..db5fc73e41cc 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.h
+++ b/drivers/net/ethernet/microchip/lan743x_main.h
@@ -980,6 +980,7 @@ struct lan743x_tx {
 	u32		frame_first;
 	u32		frame_data0;
 	u32		frame_tail;
+	u32		frame_last;
 
 	struct lan743x_tx_buffer_info *buffer_info;
 
-- 
2.51.0


From a179aad12badc43201cbf45d1e8ed2c1383c76b9 Mon Sep 17 00:00:00 2001
From: Mattias Barthel <mattias.barthel@atlascopco.com>
Date: Tue, 29 Apr 2025 11:08:26 +0200
Subject: [PATCH 16/16] net: fec: ERR007885 Workaround for conventional TX

Activate TX hang workaround also in
fec_enet_txq_submit_skb() when TSO is not enabled.

Errata: ERR007885

Symptoms: NETDEV WATCHDOG: eth0 (fec): transmit queue 0 timed out

commit 37d6017b84f7 ("net: fec: Workaround for imx6sx enet tx hang when enable three queues")
There is a TDAR race condition for mutliQ when the software sets TDAR
and the UDMA clears TDAR simultaneously or in a small window (2-4 cycles).
This will cause the udma_tx and udma_tx_arbiter state machines to hang.

So, the Workaround is checking TDAR status four time, if TDAR cleared by
    hardware and then write TDAR, otherwise don't set TDAR.

Fixes: 53bb20d1faba ("net: fec: add variable reg_desc_active to speed things up")
Signed-off-by: Mattias Barthel <mattias.barthel@atlascopco.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250429090826.3101258-1-mattiasbarthel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
 drivers/net/ethernet/freescale/fec_main.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index a86cfebedaa8..17e9bddb9ddd 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -714,7 +714,12 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
-	writel(0, txq->bd.reg_desc_active);
+	if (!(fep->quirks & FEC_QUIRK_ERR007885) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active))
+		writel(0, txq->bd.reg_desc_active);
 
 	return 0;
 }
-- 
2.51.0