Posts made by george1421

@Wayne-Workman said in Serve different types of .efi boot files:

Well that would be amazing. We should work on using that for ISC-DHCP too.

It would be wonderful if all hardware manufacturers did this too. But as with the Lenovo the OP has, its all zeros. As long as that is model unique then we can work with it, but I suspect all lenovos have that set to zero.

From what I understand isc-dhcp can do this too like dnsmasq make run time decisions based on what client advertises. I just haven’t dug into the isc-dhcp settings as of now to say for absolute.

We use the intel nucs on our campus, not the 6i but the 5i5 models.

On the intel nucs, have you gone into the advanced firmware settings and confirmed you are in uefi or legacy(bios) mode? Also some of the nucs have a “personality” option, where you can set it for win7, win8, win10, unix make sure you have the right personality selected on the nuc. Also be sure you have the latest firmware installed on the nucs they have updates every quarter for new versions. Make sure you are up to date.

I just looked up error EC-8

PXE-EC8: !PXE structure was not found in UNDI driver code segment.
The UNDI ROM image has probably been corrupted, or has not been initialized by the BIOS. This error is most often caused by one of three things:
A .NIC image was programmed into a BIOS when a .LOM image should have been used.
The memory allocated by the POST Memory Manager ($PMM) during PXE option ROM initialization has been corrupted or erased before PXE option ROM boot.
The UNDI_Loader structure was not properly initialized during option ROM initialization.

Just a stab in the dark here, maybe a undi boot kernel was sent to a uefi based system??

@mageta52 If I remember pfsense there is a field for bios (legacy) mode and that is typically undionly,kpxe and there is also a uefi field where you might puth ipxe.efi.

@Tom-Elliott said in Serve different types of .efi boot files:

Oct 12 12:59:32 dureal99d-Precision-WorkStation-T5400 dnsmasq-dhcp[1280]: 3272245650 vendor class: PXEClient:Arch:00007:UNDI:003016

That is what I see:

Not ignoring you on this. The vendor class (option 60) doesn’t have anything that is unique to describe this machine. While the arch does tell you if it is a uefi or bios machine that is all that can be gleaned from that option. The UNDI part is only what version of UNDI the client supports. The Dell 6230 I had also had a undi value of 003 016 and that one used ipxe.efi or snp.efi for a uefi boot.

The UUID (option 97) is suppose to be unique on a model (first half of uuid) and device (second half of uuid). on a per machine basis. My hope is that we can use the uuid for this. The only other thing that is unique in the dhcp request is the mac address, but then you will need a match filter for every mac address you want a unique iPXE boot kernel for. That (IMO) is not very sustainable if you have more than about 20 devices on your campus.

@dureal99d This one may be harder than with the Dell computers I’ve tested on. The uuid is actually made up of two parts the IAID which should indicate the model and DUID which should should be device specific. In this case the model identifier is ‘00:11:00:00:00:00:00:00:00’ We can filter on that, but it doesn’t look too unique.

@dureal99d Ok that is a bit what I saw when I was working on the tutorial at home. There has to be an easier way, I just haven’t learned it yet.

But here is what I did, I mashed the two values together.

# From wireshark.
00000011-0000-0000-0000-50465d986f3c
# Translated into 2 nibble hex numbers
00:00:00:11:00:00:00:00:00:00:50:46:5d:98:6f:3c

# From dnsmasq log
00:11:00:00:00:00:00:00:00:00:00:50:46:5d...

#Side by side aligned
   00:00:00:11:00:00:00:00:00:00:50:46:5d:98:6f:3c
00:11:00:00:00:00:00:00:00:00:00:50:46:5d...

#Missing bits filled into the dnsmasq one
00:11:00:00:00:00:00:00:00:00:00:50:46:5d:98:6f:3c

End results to create the unique uuid for this asus computer. 00:11:00:00:00:00:00:00:00:00:00:50:46:5d:98:6f:3c

So now we need to make one of those magical dnsmasq filters that will identify this computer for us.

Edit (duh, I just realized that the 50:46:5d:98:6f:3c of that uuid is the mac address of that computer. That won’t do for us exactly. but we are close)

@dureal99d Like it enough to make that your default kernel and then the ones that don’t like it you make exceptions for? Use the 80/20 rule what do you have the most of on your campus and then make that your default kernel and the others be the exception.

@dureal99d Which computer needs the special kernel the ASUS or Lenovo?

@dureal99d said in Serve different types of .efi boot files:

@george1421 said in Serve different types of .efi boot files:

might also want to learn about capture filters in wireshark

for asus 00000011-0000-0000-0000-50465d986f3c / udp.port == 68

Bonus points for you!!

@dureal99d The asus one has something we can work with. That dhcp option 97 is ‘00000011-0000-0000-0000-50465d986f3c’

I can tell you if you boot this computer in bios mode (just for this test) you will probably see this listed on the initial pxe boot screen as GUID.

This number should also be visible in the /var/log/syslog or where ever dnsmasq is writing its log. In my testing the value from wireshark did not match up exactly with what was found in the dnsmasq log. You need the value that looks similar from the dnsmasq log. But with that information we can take the next step to create a match test.

@dureal99d well on the first one I didn’t see any systems that transmitted their uuid in dhcp option 97. That sux a bit.

You might also want to learn about capture filters in wireshark. They are similar to the capture filters for tcpdump.

‘udp.port == 67 or udp.port == 68 or udp.port == 69 or udp.port == 4011’ this one is equiv to the tcpdump one we were using. Let me look at the ASUS computer.

@dureal99d nuts, can you try another?

You can capture this with wireshark on a computer and not have to go the tcpdump route and then copy the file, etc. Any computer on the same subnet as the booting computer can detect these broadcast messages if you are running wireshark.

@dureal99d Well that is a unique pattern too. But I would expect something like 44454c4c45440-038-3610-804ec4-c04f4a5831 the numbers are not important just the uniqueness of the UUID field.

Did that uuid (dhcp option 97) come from a vm or a real computer?

@dureal99d Oh man, I spent 4 hours writing that and its still confusing?

Actually it is a hard subject to master.

Basically what you need is a new match test.

I want you to go look at any of the pcaps you took yesterday and in the initial “hello I’m here” (probably frame 1 or 2) from the target computer to the world, look at the options, specifically option #97 and post the answer here.

@educapole said in FOG 1.3RC13 TFTP error:

@Tom-Elliott Indeed, i’ve found a second DHCP server on my network. An user has plug a damn wifi acces point…

Bingo!! and those dang wifi access points with built in dhcp will point to them selves as the next hop server so the clients will try to download the boot image from them.

Good find.

@Wayne-Workman

Actually I’d like to see a pcap from the point of view of the FOG server as long as the fog server, dhcp server and target computer is on the same subnet. I would have the OP run the tcpdump program (which may need to be installed) with a command like. tcpdump -w output.pcap port 67 or port 68 or port 69 or port 4011 Then pxe boot the target computer, once the target computer reaches the error (or question in this case) press control-C on the tcpdump program. From there upload the pcap to this forum so we can look at it.

@Wayne-Workman I agree. You can’t play with the format (sizeable vs non-resizable) once the image has been captured. If you capture it resizable you must deploy resizable. The same for non-resizable capture.

@Tom-Elliott This is a Dell 380 (circa 2007 ish) This system doesn’t know anything about uefi. That didn’t come into the picture until the 390/790 series). So this is mbr / bios system.

From the ls -la I can see partition 1 is about 11GB in size. Does that sound about right for your data size on the source image?

@Sozen Will you log into the FOG server /images/(image name) and do a ls -la and post the results here. Lets make sure what is uploaded looks complete. We may need to get one of the developers to help here since I don’t remember 1.2.0 very well.

@Sozen The size on the fog server doesn’t sound right. I could understand that if you have a windows computer with a 200GB hard drive and you had selected single disk resizeable or even nonresizable and the image size was 20GB on the FOG server that would be logical, since FOG only uploads data not empty space on the hard drive. So if you look on the golden image there may be 20GB of data used. Also understand that the FOG compresses the image as it uploads to the fog server so that 20GB of raw data may end up 15GB in the image file.