Date: Mon, 26 Apr 2021 19:48:25 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 255432] pf fragment reassembly leads to invalid IP checksum since 13.0-RELEASE Message-ID: <bug-255432-227@https.bugs.freebsd.org/bugzilla/>
next in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D255432 Bug ID: 255432 Summary: pf fragment reassembly leads to invalid IP checksum since 13.0-RELEASE Product: Base System Version: 13.0-RELEASE Hardware: amd64 OS: Any Status: New Severity: Affects Some People Priority: --- Component: kern Assignee: bugs@FreeBSD.org Reporter: topical@gmx.net Scenario: I have 2 lagg interfaces: one with mtu 1500 another with mtu 9000. Both have several VLANs. Fragmented UDP packets are received by VLAN of first interface, reassembled= by pf ("scrub in all fragment reassemble"), and then sent out as jumbo frame by VLAN of second interface. All networks cards are mellanox ("mce"). All H/W offload features but LRO a= re enabled. Before 13.0 everything worked fine. Since 13.0, the resulting jumbo frame is sent out with invalid IP checksum and thus discarded by receiver. Disabling TXCSUM on outgoing VLAN interface doesn't change anything. I tried to disable VLAN_HWCSUM on outgoing LAGG interface, but this seems t= o be impossible without reboot - ifconfig still shows this feature as enabled. As this is a production system, rebooting for further tests is not an option.= =20 But: if I decrease mtu of outgoing VLAN interface to 1500, the packet gets fragmented again and IP checksums are correct.=20 I know that there are (cheap) network cards with broken/limited H/W offload support and thus you should disable all these features. But AFAIK mellanox cards are not affected by this and actually they worked until 13.0 --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-255432-227>