[Bug 270859] dmesg message: mce0: a looped back NS message is detected during DAD

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D270859

Bruce M Simpson <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|Closed                      |Open
         Resolution|Overcome By Events          |---

--- Comment #20 from Bruce M Simpson <[email protected]> ---
Unlike ix(4) and igb(4) using Mellanox hardware yielded a working native
FreeBSD SR-IOV configuration for me.

mlx5tool(8) didn't cut it. mstflint did. Assumes you have downloaded the
relevant FlexBoot images from NVidia as an extracted .zip file.

Abridged from Linux-specific instructions at:
https://static.xtremeownage.com/blog/2025/mellanox-configuration-guide/#fla=
sh-card

doas pkg install mstflint

doas /usr/local/bin/mstdevices_info
PCI devices:
------------
ConnectX4LX(rev:0)      pci0:11:0:0
ConnectX4LX(rev:0)      pci0:11:0:1

doas mstflint -d pci0:11:0:0 -i
fw-ConnectX4Lx-rel-14_32_1912-MCX4121A-ACA_Ax-UEFI-14.25.17-FlexBoot-3.6.50=
2.bin
burn

    Current FW version on flash:  14.27.1016
    New FW version:               14.32.1912

FSMST_INITIALIZE -   OK
Writing Boot image component -   OK
Restoring signature                     - OK
-I- To load new FW, issue system-level reset or use mstfwreset where
applicable.

This was insufficient, a full host reboot was required:
doas mstfwreset -d pci0:11:0:0 q
doas mstfwreset -d pci0:11:0:0 r
doas shutdown -r now

mce0 was in use as the primary NIC which complicated matters. After reboot:
sysctl dev.mlx5_core.0.hw.fw_version
dev.mlx5_core.0.hw.fw_version: 14.32.1912

This appears to have resolved the issue for the PF. However, the issue
immediately recurred in a BHyve guest bound to the ppt passed through VF,
running 16-CURRENT. The NS workaround of setting net.inet6.ip6.dad_count=3D=
0 in
/etc/sysctl.conf was still necessary.

The bright side: I can now avoid if_bridge(4) IPv4 co-assignment warnings f=
or
bridge members when using vm-bhyve vm(8).

--=20
You are receiving this mail because:
You are the assignee for the bug.=