X1 AIO Desktop - i7 vPro network issue with Intel I219-LM [was: Make new bzImage...]

[Tom Elliott]

@george1421 I had another person who had a very similar issue with this card.

What if you, for these systems, use the ipxe.efi file? Or the intel.efi file?

[george1421]

@Tom-Elliott said in Make new bzImage...:

What if you, for these systems, use the ipxe.efi file? Or the intel.efi file?

FWIW: Unless I missed something the OP is getting past the iPXE kernel, its the bzImage that’s not detecting the network adapter. That is why he wants to build his own kernel. Its still early and I’m low on coffee, so I might have missed the intent.

[Tom Elliott]

@george1421 I think, if it were to be booted into a debug mode, we’d most likely actually see the NIC without issue. It’s just failing to allow the NIC to get a valid IP address. Then again, It IS indeed still early.

[mandrade]

@Tom-Elliott said in Make new bzImage...:

C to get a valid

Correct george1421 exactly my intention.

[george1421]

@mandrade I would still like to know if that hardware is already in the current bzImage. Could you get the vendor id and subsystem id? I know that many of my Dell Optiplex systems use that same (number wise) network adapter and they boot just fine.

[mandrade]

Sure can will have to be later on though at this stage I don’t have access to the machine. Important to note though, when I compile my kernel to version 4.6 and create my own bzImage file the network then works however FOG no longer can detect a HDD,

[mandrade]

Hi Guys,

so I have the subsys as requested in the screenshot below:

[mandrade]

Also here are the results of the compatibility test:

[george1421]

@mandrade Just for clarity ven_8086&dev_156F (8086:156F) comes back to a Intel I219-LM NIC.

From the disk controllers standpoint, the devs may be including additional disk controllers drivers that are not part of the core linux distribution or the config file from github doesn’t have all of the necessary disk controllers enabled.

[mandrade]

@george1421 ,

Right so, when I do a make menuconfig I could be missing a step somewhere that is not including the necessary controller drivers?

[george1421]

@mandrade Did you start with the config file from the github site? I think you can just rename it as the default config then run menuconfig program picking your additional network drivers. It will use those settings as the defaults as you run through the menuconfig. Understand its been over 10 years since I built linux kernels but that is what I remember.

Ref: https://github.com/FOGProject/fogproject/tree/dev-branch/kernel You probably want the TomElliott.config.64 one.

[edit] It would be interesting to know what network adapter option you are loading to get this network adapter to work, so the devs can include it in the official kernel. That way you wouldn’t have to keep rebuilding the kernel as new official releases come out.

[mandrade]

@george1421 said in Make new bzImage...:

think you can just rename it as the default config then run menuconfig progra

I used the standard Kitchensink one that is available when you download FOG_1.2.0 so no I did not. I will give that a try.

Thank you very much for the suggestions and help thus far!

[Sebastian Roth]

@george1421 said:

Just for clarity ven_8086&dev_156F (8086:156F) comes back to a Intel I219-LM NIC.

Sure but not every I219-LM NIC is the same. Check out this listing to see when support for the different NICs (PCI IDs) was added. That’s why I always ask people about PCI IDs. Support for the 8086:156f was added with kernel 4.1 (while 15e3 seems to have been added only very recently - kernel version 4.6).

So I guess your NIC is fully supported using the 4.5 kernel from https://fogproject.org/kernels/. Guess this is a network issue. Spanning tree, 802.1x/MAB/NAC, EEE/802.3az, auto-negotiation!? @mandrade To see if I stand correct can you please use a dumb mini switch to connect between client and your main network. Then try compatibility test again.

@Tom-Elliott I think we should add lspci output (or simply PCI IDs) to the compatibility test screen when failing! This way people don’t need to extra-boot and find out about this information. Shouldn’t take much effort! What do you think?

[mandrade]

Hey Guys,

So I’m now able to image my new machines, it works a treat!

In the end I decided to scrap attempting to build my own Kernel and went with a Trunk version of FOG. I am aware that it isn’t stable but after deploying the new bleeding edge version of FOG all my machines are supported and are imaging without any issue.

We’re about to do a bulk deploy of 21 Machines, will let everybody know how that went. Experiences thus far, although frustrating in the beginning, are good. Thanks again for all the help supplied.

[george1421]

@mandrade I’m glad you settled on a solid choice here.

The thing you have to remember is the fog 1.2.0 is almost 3 years old. The devs have put great effort in getting the trunk build to the point it is today at the expense of delaying the official release of 1.3.0. They felt it was vital to build support for Win10, uefi support, and gpt disks into the trunk build since 1.3.0 is the last (intended) release of FOG until FOG 2.0 comes out.

[Wayne Workman]

@george1421 Intended.

[mandrade]

Hi guys,

So the feedback is two fold. Success on the all in one ThinkCentre machines. We are now deploying without issue to those.

However, we have three and soon to be more Lenovo ThinkPad X1 Carbons to deploy. They have no onboard NIC and use an external Onelink+ Dock. I can PXE boot but when it starts loading the undionly.kpxe image it hangs.

Now I have tried boot to undionly.kkpxe and it gets further. However, it stop again at 55% as can be seen below:

The NIC is a Realtek USB RTL8135

[Sebastian Roth]

@mandrade I’d be willing (and really interested) to find out what’s going on there. To properly diagnose this I need a packet dump of this. Please install the package tcpdump on your FOG server and run sudo tcpdump -i eth0 -w hang.pcap host x.x.x.x (where x.x.x.x is the IP of the client you are trying to boot). Just leave that command sitting there and boot up your client. When it hangs stop tcpdump (ctrl+c) and send me that hang.pcap file via mail (will send you a private message with my mail address).

Does it always hang at 55%??

[mandrade]

@Sebastian-Roth ,

Sure will see what I can do. Yep always at 55%.

[Tom Elliott]

@mandrade It stops at 55% on all three machines?