em0 watchdog timeout with nfs

[Sam Leffler]

LI Xin wrote:
> Hi, Jack,
> 
> Jack Vogel wrote:
>>> em0 at pci0:14:0:  class=0x020000 card=0x002e8086 chip=0x100e8086 rev=0x02
>>> hdr=0x00
> [...]
>> The driver in 6.2 RELEASE fixed all known problems with
>> watchdogs, other than REAL issues with the network/hardware.
>> Have you tried installing that?
> 
> A friend of mine has reported similar problem, with different em(4)
> hardware.  The server runs lighttpd on ufs, and the watchdog timeout
> occurs no matter whether there is heavy traffic.
> 
> Here is some pciconf -l output which can be interesting.
> 
> [root at me /usr/local/etc]# pciconf -l|grep em
> em0 at pci5:0:0:   class=0x020000 card=0x30a38086 chip=0x108b8086 rev=0x03
> hdr=0x00
> em1 at pci6:5:0:   class=0x020000 card=0x30a18086 chip=0x10768086 rev=0x05
> hdr=0x00
> 
> Should more debugging aid / information is needed to narrow down the
> issue please let us know, thanks!

I've reported similar problems multiple times w/o any response.  My nic
is onboard (no msi involved, no polling used):

em0 at pci2:1:0:   class=0x020000 card=0x80f71043 chip=0x10198086 rev=0x00
hdr=0x00
    vendor     = 'Intel Corporation'
    device     = '82547EI Gigabit Ethernet Controller (LOM)'
    class      = network
    subclass   = ethernet

trouble% ifconfig em0
em0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=9b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM>
        ether 00:11:2f:9e:c0:e5
        inet 10.0.0.248 netmask 0xffffff00 broadcast 10.0.0.255
        media: Ethernet autoselect (1000baseTX <full-duplex>)
        status: active

When I boot I also see massive #'s of link state transitions while
dhclient fetches a lease.  I've swapped cables, switch ports, etc.  I
can believe it might be a h/w failure but was hoping I could isolate the
issue to be certain (don't like discarding the onboard nic).

This is a very current HEAD and has been a problem for several months
(all against HEAD).

	Sam