bce(4) network wierdness

Hi

Periodically (initially I thought it was related to bgfsck) but
I've just had the problem again after 6 days of uptime with no
provocation.

I get massive packet loss and out of order delivery as evidenced below.

64 bytes from 41.161.56.2: icmp_seq=2 ttl=64 time=1351.206 ms
64 bytes from 41.161.56.2: icmp_seq=5 ttl=64 time=2214.913 ms
64 bytes from 41.161.56.2: icmp_seq=4 ttl=64 time=3616.881 ms
64 bytes from 41.161.56.2: icmp_seq=9 ttl=64 time=0.090 ms
64 bytes from 41.161.56.2: icmp_seq=8 ttl=64 time=1417.588 ms
64 bytes from 41.161.56.2: icmp_seq=12 ttl=64 time=924.168 ms
64 bytes from 41.161.56.2: icmp_seq=13 ttl=64 time=1001.294 ms
64 bytes from 41.161.56.2: icmp_seq=16 ttl=64 time=117.199 ms
64 bytes from 41.161.56.2: icmp_seq=15 ttl=64 time=1372.514 ms
64 bytes from 41.161.56.2: icmp_seq=17 ttl=64 time=943.103 ms
64 bytes from 41.161.56.2: icmp_seq=21 ttl=64 time=4041.519 ms
64 bytes from 41.161.56.2: icmp_seq=24 ttl=64 time=1430.088 ms
64 bytes from 41.161.56.2: icmp_seq=26 ttl=64 time=1770.405 ms
64 bytes from 41.161.56.2: icmp_seq=28 ttl=64 time=426.317 ms
64 bytes from 41.161.56.2: icmp_seq=31 ttl=64 time=748.205 ms
64 bytes from 41.161.56.2: icmp_seq=66 ttl=64 time=3623.306 ms
64 bytes from 41.161.56.2: icmp_seq=104 ttl=64 time=1653.222 ms

Kernel messages at the time:

Aug  6 08:06:04 firewall1 kernel: bce0: discard frame w/o leading ethernet header (len 4294967292 pkt len 4294967292)
Aug  6 08:06:16 firewall1 last message repeated 190 times

Hardware is:

bce0_at_pci0:4:0:0:        class=0x020000 card=0x02231028 chip=0x164c14e4 rev=0x12 hdr=0x00
    vendor     = 'Broadcom Corporation'
    device     = '5708C Broadcom NetXtreme II Gigabit Ethernet Adapter'
    class      = network
    subclass   = ethernet

I'm running vlans and carp with polling on top of this.

This is the builtin NIC on a Del R905.

Ian

--
Ian Freislich