[Bug 202875] ixv driver in 11.0-CURRENT doesn't pass traffic using KVM hypervisor

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=202875

            Bug ID: 202875
           Summary: ixv driver in 11.0-CURRENT doesn't pass traffic using
                    KVM hypervisor
           Product: Base System
           Version: 11.0-CURRENT
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Many People
          Priority: ---
         Component: kern
          Assignee: freebsd-bugs at FreeBSD.org
          Reporter: jeffrey.e.pieper at intel.com

ixv-1.4.0 (11.0-CURRENT snapshot 20150804) will not pass traffic using a KVM
hypervisor on a Linux host. Tested using RHEL 7.1 x86_64 with ixgbe-4.1.2 as
the host OS/PF driver. ixv-1.2.5 (10.2-RELEASE amd64) works as expected.

There were a few other commits at the same time, but this is the likely cause.

Reproduce steps
1. Load PF driver on the host and bring interface up.
2. Create 2 VF's on a single port using sysfs.
3. Attach one VF to a FreeBSD 11.0 guest and one VF to a FreeBSD 10.2 guest.
4. Boot both guests and load ixv (if not static)on both guests.
5. Assign an IP address to ixv0 and bring the interface up on both guests.
6. Attempt to pass traffic to a link partner. Both broadcast and unicast ICMP 
traffic fails on the 11.0 guest, while the 10.2 guest passes traffic.

ixv-1.4.0:

ixv0: <Intel(R) PRO/10GbE Virtual Function Network Driver, Version - 1.4.0> mem
0xfc054000-0xfc057fff,0xfc058000-0xfc05bfff at device 9.0 on pci0
ixv0: Using MSIX interrupts with 2 vectors
ixv0: MBX API 2 negotiation: 0
ixv0: Ethernet address: 72:05:dc:1c:14:2a
ixv0: netmap queues/slots: TX 1/1024, RX 1/1024
ixv0: link state changed to UP

ixv0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
   
options=405bb<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,TSO4,LRO,VLAN_HWTSO>
    ether 72:05:dc:1c:14:2a
    inet 190.0.7.20 netmask 0xffff0000 broadcast 190.0.255.255 
    nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL>
    media: Ethernet autoselect
    status: active

PING 190.0.255.255 (190.0.255.255): 56 data bytes

--- 190.0.255.255 ping statistics ---
26 packets transmitted, 0 packets received, 100.0% packet loss

ixv-1.2.5:

ixv0: <Intel(R) PRO/10GbE Virtual Function Network Driver, Version - 1.2.5> mem
0xfc054000-0xfc057fff,0xfc058000-0xfc05bfff at device 9.0 on pci0
ixv0: Using MSIX interrupts with 2 vectors
ixv0: Ethernet address: 26:a1:8c:3e:3c:fe
ixv0: link state changed to UP

ixv0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
   
options=405bb<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,TSO4,LRO,VLAN_HWTSO>
    ether 26:a1:8c:3e:3c:fe
    inet 190.0.7.21 netmask 0xffff0000 broadcast 190.0.255.255 
    nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL>
    media: Ethernet autoselect
    status: active

PING 190.0.255.255 (190.0.255.255): 56 data bytes
64 bytes from 190.0.8.13: icmp_seq=0 ttl=64 time=0.382 ms
64 bytes from 190.0.8.10: icmp_seq=0 ttl=64 time=0.397 ms (DUP!)
64 bytes from 190.0.8.7: icmp_seq=0 ttl=64 time=0.410 ms (DUP!)
64 bytes from 190.0.8.14: icmp_seq=0 ttl=64 time=0.423 ms (DUP!)
64 bytes from 190.0.8.9: icmp_seq=0 ttl=64 time=0.439 ms (DUP!)

<Snip>

--- 190.0.255.255 ping statistics ---
14 packets transmitted, 14 packets received, +196 duplicates, 0.0% packet loss
round-trip min/avg/max/stddev = 0.200/0.750/3.637/0.906 ms

I believe this commit to be the culprit:

https://svnweb.freebsd.org/base/head/sys/dev/ixgbe/ixgbe_vf.c?revision=285590&view=markup

There were a few other commits at the same time, but this is the likely cause.

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