So I figured I would update the list on what I've found so far, and since I'm not at a loss, maybe someone else can pick up and get it going. So with 5.1 being released, I decided to move my trial to that version. Besides now having to load more modules, it happens to be just about the same. First, you'll want to extract the entire ISO to a directory on your http/tftp/netboot server. I did this by mounting it via loop on the command line, and using cp -R to also grab the directories (although I don't think these are needed). Here: http://node4.smart-serv.net/~brandon/esxi51.ipxe is the file (also attached) that I chain to in my iPXE menu to load the installer. Few things to note here: 1. I had to use imgfree here since I did have pxelinux.0 embedded into iPXE for my vesamenu (which actually isn't used anymore, but it doesn't hurt to have it anyway).
2. The tboot.b00 file was gunzip'd (gunzip -S b00 tboot.b00; mv tboot. tboot) and renamed to just tboot. The option "runweasel" it set on the kernelopts line in the esxi5 boot.cfg file (as noted previously on the mailing list) 3. All of the modules are actually gzip'd. I used this script (http://node4.smart-serv.net/~brandon/esxi-script.sh also attached) to extract almost all of the modules, and rename them back to what they were called. The ones I don't gunzip are the ones ending in .gz or similar format, as iPXE does handle them, and it seems the kernel wants to unpack them itself. From what I was finding, the kernel would never boot for me when I included all of these as-shipped from VMWare. 4. The "useropts.gz" module is missing from the list. I found this is actually a file of 0 size, that was gzip'd to 20bytes, and I found by including it (either compressed, or not), ipxe either fails to boot the image, or it keeps trying to download a 0byte file (which obviously there is nothing to download). 5. chardevs is just that in the file, while it is actually named chardevs.b00, this is explained below. On the iPXE side, the following needs to be changed to work: * src/arch/i386/image/multiboot.c * Line 55 (on current GIT): #define MAX_MODULES needs to be changed to a much higher value. I set mine to 100. * Line 65 (on current GIT): #define MB_MAX_CMDLINE needs to be set to a _much_ higher value. I set mine to 4096, as it includes all the characters of each module. So, with all of these changes, I at least get the ESXi kernel to boot! One step in the right direction! However, it gets the Pink Screen, when it tries to start loading modules: http://node4.smart-serv.net/~brandon/esxi-pink.png (also attached). If you hit ESC, then open l (lowercase L), to display the VMKernel log, you can see what the issue is: http://node4.smart-serv.net/~brandon/esxi-moderr.png (again, also attached). The line that peaked my interest was: "BootModule: 133: unknown boot module type for chardevs". This brings us to issue 5 above. It seems no matter what I try, including the raw chardevs.b00 (gzip'd, but does still boot), or unpacking it to just the chardevs file that Nautilus identifies as an Object Code file, I get the above error. If I do not include this module at all, when it tries to load the user.b00 file, it fails on the chardevs namespace, indicating to me that it is required. So.. this is where I got to trying to boot ESXi's installer via iPXE's multiboot image functionality. I did come up with a workaround for a person in the IRC channel that chainloads to pxelinux.0 to load the installer via it's mboot.c32 file, so they could at least get ESXi installing. I can post that if there is interest. Hopefully this can be useful to someone trying to debug further, to actually get ESXi to boot. Brandon Penglase
esxi51.ipxe
Description: Binary data
esxi-script.sh
Description: application/shellscript
<<attachment: esxi-pink.png>>
<<attachment: esxi-moderr.png>>
_______________________________________________ ipxe-devel mailing list ipxe-devel@lists.ipxe.org https://lists.ipxe.org/mailman/listinfo.cgi/ipxe-devel
