Skip to contentSkip to conversation

nft_flowtable.sh selftest failures

DISCUSSION Paolo Abeni 4 participants 6 messages Aug 12, 2025 – Aug 13, 2025

Subsystems (4)
lore
Share & tools
lore

Paolo Abeni

author maintainer
#0 lore raw

Hi,

the mentioned self test failed in the last 2 CI iterations, on both metal and debug build, with the following output:

# PASS: flow offload for ns1/ns2 with dnat and pmtu discovery ns1 <- ns2 # Error: Requested AUTH algorithm not found.
# Error: Requested AUTH algorithm not found.
# Error: Requested AUTH algorithm not found.
# Error: Requested AUTH algorithm not found.
# FAIL: file mismatch for ns1 -> ns2
# -rw------- 1 root root 2097152 Aug 11 20:23 /tmp/tmp.x1oVr3mu0P # -rw------- 1 root root 0 Aug 11 20:23 /tmp/tmp.77gElv9oit # FAIL: file mismatch for ns1 <- ns2
# -rw------- 1 root root 2097152 Aug 11 20:23 /tmp/tmp.x1oVr3mu0P # -rw------- 1 root root 0 Aug 11 20:23 /tmp/tmp.ogDiTh8ZXf # FAIL: ipsec tunnel mode for ns1/ns2

see, i.e.:
https://netdev-3.bots.linux.dev/vmksft-nf/results/249461/14-nft-flowtable-sh/

I don't see relevant patches landing in the relevant builds, I suspect the relevant kernel config knob (CONFIG_CRYPTO_SHA1 ?) was always missing in the ST config, pulled in by NIPA due to some CI setup tweak possibly changed recently (Jakub could possibly have a better idea/view about the latter). Could you please have a look?

NIPA generates the kernel config and the kernel build itself with something alike:

rm -f .config
vng --build --config tools/testing/selftests/net/forwarding/config

/P

author maintainer
#1 lore raw
Quoted 25 lines · Paolo Abeni

the mentioned self test failed in the last 2 CI iterations, on both metal and debug build, with the following output:

# PASS: flow offload for ns1/ns2 with dnat and pmtu discovery ns1 <- ns2 # Error: Requested AUTH algorithm not found.
# Error: Requested AUTH algorithm not found.
# Error: Requested AUTH algorithm not found.
# Error: Requested AUTH algorithm not found.
# FAIL: file mismatch for ns1 -> ns2
# -rw------- 1 root root 2097152 Aug 11 20:23 /tmp/tmp.x1oVr3mu0P # -rw------- 1 root root 0 Aug 11 20:23 /tmp/tmp.77gElv9oit # FAIL: file mismatch for ns1 <- ns2
# -rw------- 1 root root 2097152 Aug 11 20:23 /tmp/tmp.x1oVr3mu0P # -rw------- 1 root root 0 Aug 11 20:23 /tmp/tmp.ogDiTh8ZXf # FAIL: ipsec tunnel mode for ns1/ns2

see, i.e.:
https://netdev-3.bots.linux.dev/vmksft-nf/results/249461/14-nft-flowtable-sh/

I don't see relevant patches landing in the relevant builds, I suspect the relevant kernel config knob (CONFIG_CRYPTO_SHA1 ?) was always missing in the ST config, pulled in by NIPA due to some CI setup tweak possibly changed recently (Jakub could possibly have a better idea/view about the latter). Could you please have a look?

Quoted 5 lines

NIPA generates the kernel config and the kernel build itself with something alike:

rm -f .config
vng --build --config tools/testing/selftests/net/forwarding/config

Addendum: others (not nft-related) tests (vrf-xfrm-tests.sh, xfrm_policy.sh) are failing apparently due to the same root cause (missing sha1 knob), so I guess it's really a NIPA issue.

/P

maintainer
#2 lore raw
Quoted 5 lines · Paolo Abeni
Quoted 5 lines

I don't see relevant patches landing in the relevant builds, I suspect the relevant kernel config knob (CONFIG_CRYPTO_SHA1 ?) was always missing in the ST config, pulled in by NIPA due to some CI setup tweak possibly changed recently (Jakub could possibly have a better idea/view about the latter). Could you please have a look?

Can't reproduce this here.

Latest net tree:
vng --build --config tools/testing/selftests/net/netfilter/config
grep SHA1 .config
# CONFIG_SCTP_DEFAULT_COOKIE_HMAC_SHA1 is not set
# CONFIG_SCTP_COOKIE_HMAC_SHA1 is not set
CONFIG_CRYPTO_SHA1=m
make -C tools/testing/selftests/ TARGETS=net/netfilter
vng -v --run . --user root --cpus 4 -- \
make -C tools/testing/selftests TARGETS=net/netfilter run_tests TEST_PROGS=nft_flowtable.sh
[..]
# PASS: ipsec tunnel mode for ns1/ns2
ok 1 selftests: net/netfilter: nft_flowtable.sh

Quoted 9 lines
Quoted 5 lines

NIPA generates the kernel config and the kernel build itself with something alike:

rm -f .config
vng --build --config tools/testing/selftests/net/forwarding/config

Addendum: others (not nft-related) tests (vrf-xfrm-tests.sh, xfrm_policy.sh) are failing apparently due to the same root cause (missing sha1 knob), so I guess it's really a NIPA issue.

Looks like it, I mean, I can't repro it here.

Let me know if I missed anything or if there is something I can do to help debugging this.

#3 lore raw

Hi Florian,

(+ Eric in Cc)

Quoted 10 lines · Florian Westphal
Quoted 5 lines · Paolo Abeni
Quoted 5 lines

I don't see relevant patches landing in the relevant builds, I suspect the relevant kernel config knob (CONFIG_CRYPTO_SHA1 ?) was always missing in the ST config, pulled in by NIPA due to some CI setup tweak possibly changed recently (Jakub could possibly have a better idea/view about the latter). Could you please have a look?

Can't reproduce this here.

Latest net tree:

I don't know if it can help, but did you try to reproduce it on top of the branch used by the CI?

https://github.com/linux-netdev/testing/tree/net-next-2025-08-12--06-00

This branch is on top of net-next, where 'net' has been merged, all pending patches listed on Patchwork have been applied, plus a few additional patches are there to either fix some temp issues or improve the CI somehow. Maybe one of these patches caused the removal of CONFIG_CRYPTO_SHA1.

I guess that's the case, because when looking at the diff [1] when the issue got introduced, I see some patches [2] from Eric Biggers modifying some sctp's Kconfig file. They probably cause the issue, but the fix should be to add CONFIG_CRYPTO_SHA1 in the ST config as mentioned by Paolo.

[1]
https://netdev.bots.linux.dev/static/nipa/branch_deltas/net-next-2025-08-12--03-00.html
[2] https://patchwork.kernel.org/project/netdevbpf/list/?series=990201

Cheers,
Matt

Signature · 1 line
Sponsored by the NGI0 Core fund.
maintainer
#4 lore raw
Quoted 10 lines · Matthieu Baerts

I don't know if it can help, but did you try to reproduce it on top of the branch used by the CI?

https://github.com/linux-netdev/testing/tree/net-next-2025-08-12--06-00

This branch is on top of net-next, where 'net' has been merged, all pending patches listed on Patchwork have been applied, plus a few additional patches are there to either fix some temp issues or improve the CI somehow. Maybe one of these patches caused the removal of CONFIG_CRYPTO_SHA1.

Yes:
sctp: Use HMAC-SHA1 and HMAC-SHA256 library for chunk authentication

removes it.

Quoted 4 lines

I guess that's the case, because when looking at the diff [1] when the issue got introduced, I see some patches [2] from Eric Biggers modifying some sctp's Kconfig file. They probably cause the issue, but the fix should be to add CONFIG_CRYPTO_SHA1 in the ST config as mentioned by Paolo.

seems like these two are the only ones that need it. at least xfrm_policy.sh passes again after this change.

tools/testing/selftests/net/config +1 -0
@@ -115,6 +115,7 @@ CONFIG_VXLAN=m
 CONFIG_IP_SCTP=m
 CONFIG_NETFILTER_XT_MATCH_POLICY=m
 CONFIG_CRYPTO_ARIA=y
+CONFIG_CRYPTO_SHA1=y
 CONFIG_XFRM_INTERFACE=m
 CONFIG_XFRM_USER=m
 CONFIG_IP_NF_MATCH_RPFILTER=m
tools/testing/selftests/net/netfilter/config +1 -0
@@ -98,3 +98,4 @@ CONFIG_NET_PKTGEN=m
 CONFIG_TUN=m
 CONFIG_INET_DIAG=m
 CONFIG_INET_SCTP_DIAG=m
+CONFIG_CRYPTO_SHA1=y
#5 lore raw
Quoted 40 lines · Florian Westphal
Quoted 10 lines · Matthieu Baerts

I don't know if it can help, but did you try to reproduce it on top of the branch used by the CI?

https://github.com/linux-netdev/testing/tree/net-next-2025-08-12--06-00

This branch is on top of net-next, where 'net' has been merged, all pending patches listed on Patchwork have been applied, plus a few additional patches are there to either fix some temp issues or improve the CI somehow. Maybe one of these patches caused the removal of CONFIG_CRYPTO_SHA1.

Yes:
sctp: Use HMAC-SHA1 and HMAC-SHA256 library for chunk authentication

removes it.

Quoted 4 lines

I guess that's the case, because when looking at the diff [1] when the issue got introduced, I see some patches [2] from Eric Biggers modifying some sctp's Kconfig file. They probably cause the issue, but the fix should be to add CONFIG_CRYPTO_SHA1 in the ST config as mentioned by Paolo.

seems like these two are the only ones that need it. at least xfrm_policy.sh passes again after this change.

tools/testing/selftests/net/config +1 -0
@@ -115,6 +115,7 @@ CONFIG_VXLAN=m
 CONFIG_IP_SCTP=m
 CONFIG_NETFILTER_XT_MATCH_POLICY=m
 CONFIG_CRYPTO_ARIA=y
+CONFIG_CRYPTO_SHA1=y
 CONFIG_XFRM_INTERFACE=m
 CONFIG_XFRM_USER=m
 CONFIG_IP_NF_MATCH_RPFILTER=m

tools/testing/selftests/net/netfilter/config CONFIG_NET_PKTGEN=m · line 101

Quoted 4 lines
tools/testing/selftests/net/netfilter/config +1 -0
@@ -98,3 +98,4 @@ CONFIG_NET_PKTGEN=m
 CONFIG_TUN=m
 CONFIG_INET_DIAG=m
 CONFIG_INET_SCTP_DIAG=m
+CONFIG_CRYPTO_SHA1=y

Yes that's correct. I've included a fix for this in v2 of the series
(https://lore.kernel.org/netdev/[email protected]/).
Thanks for finding this!

- Eric