Intel 1000/PT not working

[rdhickman]

I am testing VI3 on a system with initially just a single Intel Pro 1000/PT server adapter, however this led to the service console not being accessible and the machine could not ping anything, even the gateway (the vmnic \*was* assigned to the service console). The error message when pinging was 'destination host unreachable'.

Plugging in an (old) Intel EtherExpress Pro/100 and linking that to the service console allowed me to start remotely configuring the system. Strangely, the VI client considers the 1000/PT to be 'up', and if I create a new vSwitch with a service console for the 1000/PT on a different IP address I can log in via that IP. Pinging that IP seems to make the Pro/100 respond however rather than the 1000/PT (looking at the activity lights on the NICs anyway). If I change the service console gateway from vswif0 to vswif1 (which I created) and then remove the old/default service console the server is no longer accessible at all.

The drivers are e1000 for the 1000/PT and e100 for the Pro 100. I don't suppose I can swap the 1000/PT over to the e100 driver or something equally strange? Is there some kind of issue with PCI express? I'm really at a loss regarding this issue. Any help is appreciated.

[wobbly1]

first things first ensure that your are up to date on your firmware for card and bios for server

[Paul_Lalonde]

Can you post the output of lspci, lspci -v, and cat /proc/vmware/pci

Also, anything unusual showing up in /var/log/vmkernel ?

Thanks

Paul

[einstein-a-go-g]

check your logs for TX Unit Hang?

[jonw757]

Not that its going to help in your situation but I can confirm these work, I just put 6 total in 2 different ESX 3 servers. Is it possible that its just looking for more then 1 card?

[virtech]

rdhickman what hardware do you have your card installed in?

I have a few servers configured now using this card , no problems after the patch install.

[rdhickman]

Thanks jonw757/mooiehoek for at least confirming these cards do work. I've already flashed them with proboot from Intel's website.

I don't see anything in vmkernel that's obviously out of order. The e1000 driver claims the device and the initialisation apparently "succeeded".

The portion of lspci:

00:18.0 Host bridge: Advanced Micro Devices \[AMD] K8 \[Athlon64/Opteron] HyperTransport Technology Configuration

00:18.1 Host bridge: Advanced Micro Devices \[AMD] K8 \[Athlon64/Opteron] Address MapK8 NorthBridge

00:18.2 Host bridge: Advanced Micro Devices \[AMD] K8 \[Athlon64/Opteron] DRAM Controller

00:18.3 Host bridge: Advanced Micro Devices \[AMD] K8 \[Athlon64/Opteron] Miscellaneous Control

01:06.0 Ethernet controller: Intel Corporation 82557/8/9 \[Ethernet Pro 100] (rev 08)

01:07.0 RAID bus controller: Adaptec Adaptec SCSI (rev 01)

03:00.0 Ethernet controller: Intel Corporation 82572EI Gigabit Ethernet Controller (rev 06)

Looking at lspci -v for the gigabit controller gives this

03:00.0 Ethernet controller: Intel Corporation 82572EI Gigabit Ethernet Controller (rev 06)

Subsystem: Intel Corporation: Unknown device 1082

Flags: bus master, fast devsel, latency 0, IRQ 16

Memory at e7020000 (32-bit, non-prefetchable) \[size=128K]

Memory at e7000000 (32-bit, non-prefetchable) \[size=128K]

I/O ports at b000 \[size=32]

Expansion ROM at \[disabled] \[size=128K]

Capabilities: \[c8] Power Management version 2

Capabilities: \[d0] Message Signalled Interrupts: 64bit+ Queue=0/0 Enable-

Capabilities: \[e0] #10 \[0001]

I have to say the "unknown device bit" isn't exactly encouraging. It's definitely a retail Intel Pro/1000 PT. Do I need to make a device map here?

Lastly, from the cat...pci command:

003:00.0 8086:107d 8086:1082 Ethernet Intel 5/ 16/0x81 A V e1000 vmnic2

It's vmnic2 because I've moved it around in ports a few times and added the 100 mbit card too.

[rdhickman]

I have just found this, which sounds awfully similar to the problem I'm getting. Recompiling the driver is probably impossible in this case though.

https://launchpad.net/ubuntu/source/linux-source-2.6.15/bug/56885

[Yps]

Have you created a support request on this problem?

Best Regards, Magnus

[rdhickman]

I know this is a slightly late response - I only just remembered I had made this thread. This long line of misery was compounded when I was setting up a set of linux workstations with e1000's and starting having disturbingly similar problems. In this situation I could build a new e1000 driver - which I did - but things did not help. If I forced the card to 10Mbit half-duplex I could scratch an almightly 10KB/s on FTP over LAN (yes, 10 KILObytes per sec). A check of verbose logs showed invalid checksums on over 90% of the packets.

In a fit of inspiration (read: desperation), I chainlinked another switch off of the one currently in use and plugged the machines into that instead. All fixed.

Some people have had luck with ethtool and disabling various offloading features on the card. But it seems some cheap switches are really not worth the cash. Something to keep in mind, folks!

(A final aside: the Windows e1000 driver works fine with both switches, so I'll let you decide where to lay the blame at)