[PATCH net-next v2 0/2] tcp: document and test MTU probing black hole fallback
From: Yongjian Wang via B4 Relay
Date: Wed Nov 18 2026 - 03:11:04 EST
ip-sysctl.rst does not say when TCP detects a black hole with
tcp_mtu_probing=1 or how the MSS drops afterwards. Patch 1 documents
both. Patch 2 adds a packetdrill test for the part that had no
coverage: each further retransmission timeout halves the MSS, down to
tcp_mtu_probe_floor.
On x86_64 KVM, the new test passed in IPv4, IPv6 and IPv4-mapped IPv6.
Variants expecting no floor or no halving failed in all three modes.
Changes in v2:
- Spell out the 200 ms RTO, say when the time starts and when it is
checked, and give the resulting detection time (Eric, patch 1).
- Say that the MSS is only lowered, and that each further
retransmission timeout lowers it again (Eric, patch 1).
- Add a test for tcp_mtu_probe_floor (patch 2).
- Link to v1: https://lore.kernel.org/r/20261117-tcp-mtu-probing-doc-v1-v1-1-552ae04de4ef@xxxxxxxxx
Signed-off-by: Yongjian Wang <echoowall@xxxxxxxxx>
---
Yongjian Wang (2):
tcp: document black hole detection for tcp_mtu_probing
selftests/net: packetdrill: test tcp_mtu_probe_floor
Documentation/networking/ip-sysctl.rst | 21 ++++++++++++
.../tcp_mtu_probing_blackhole_floor.pkt | 37 ++++++++++++++++++++++
2 files changed, 58 insertions(+)
---
base-commit: 6ff4c5942a301675eb12e5b49d53a81e604624a1
change-id: 20261117-tcp-mtu-probing-doc-v1-351cb754fa7b
Best regards,
--
Yongjian Wang <echoowall@xxxxxxxxx>