Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 05 Dec 2023 20:02:11 +0000
From:      Benoit Chesneau <benoitc@enki-multimedia.eu>
To:        Kristof Provost <kp@FreeBSD.org>
Cc:        freebsd-net@freebsd.org
Subject:   Re: vxlan with IPv6 underlay ?
Message-ID:  <MtPP9z0-kP65O109ZU7zmjeTXThi9XL2k36RlRMxXks-yniimXQ4loev6RIGMgCgh4TFgYlYiAVN3ANZ9UXkDmqxljm3CvJPfVNHQCiiuKc=@enki-multimedia.eu>
In-Reply-To: <3E2D29F7-2B64-45DD-A13B-7BEB799AF9BA@FreeBSD.org>
References:  <bmkPdpqYkXzCqs5CYfBIZO6sXwmR1hU38pCr5cRlBeTxM4VIbzTj_A0YU6yFP2CbVxXHoOLeqHpdBY1reGJS_VvNBesk3J1HBWblpjijl2o=@enki-multimedia.eu> <793EA8DF-BD26-414A-A12D-866356D339DB@FreeBSD.org> <Q4lNxBr17sBg-pHCFTyIugCDvSpd0oXXH6UBv5yEkJtq2lJKIXG_K4MSDarhwkFt7PAFvaLTLdm-rj6MR-ix7B8BfG-7aINXUg5h2a5VkPg=@enki-multimedia.eu> <3E2D29F7-2B64-45DD-A13B-7BEB799AF9BA@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help
OK Thanks for the answer. I will create a ticket. Found that the option `vx=
lanhwcsum` is the culprit.



Beno=C3=AEt
On Tuesday, December 5th, 2023 at 15:24, Kristof Provost <kp@FreeBSD.org> w=
rote:


> On 5 Dec 2023, at 15:21, Benoit Chesneau wrote:
>=20
> > Changing the mtu wasn't enough. To make it works i had to disable vxlan=
 offload features:
> >=20
> > `ifconfig mce0 -vxlanhwcsum -vxlanhwtso`
> >=20
> > Then suddenly all started to work. The card is the following:
> >=20
> > `mlx5_core1@pci0:17:0:1: class=3D0x020000 rev=3D0x00 hdr=3D0x00 vendor=
=3D0x15b3 device=3D0x1015 subvendor=3D0x1590 subdevice=3D0x00d4 vendor =3D =
'Mellanox Technologies' device =3D 'MT27710 Family [ConnectX-4 Lx]' class =
=3D network subclass =3D ethernet PCI-e errors =3D Correctable Error Detect=
ed Unsupported Request Detected Corrected =3D Advisory Non-Fatal Error`
> >=20
> > Which is odd. Wonder why it's needed. Thoughts?
>=20
> Driver bug?
>=20
> I=E2=80=99m not familiar with the mlx5 driver (or hardware), but that=
=E2=80=99d be my guess.
> The first step in debugging that would be to figure out which one of thos=
e offload features is responsible, and probably also capturing traffic (out=
side of this machine) to see if anything is sent and if so how it=E2=80=
=99s broken.
>=20
> Best regards,
> Kristof



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?MtPP9z0-kP65O109ZU7zmjeTXThi9XL2k36RlRMxXks-yniimXQ4loev6RIGMgCgh4TFgYlYiAVN3ANZ9UXkDmqxljm3CvJPfVNHQCiiuKc=>