Can't chainload past Fog menu (Boot hard drive) - HP zbook laptop

[Szeraax]

@sebastian-roth Just dropping the updated rEFInd files (conf and efi) gave me the default ocnfig file again (obviously). And it seems that rEFInd still hangs at “rEFInd - Initializing…” when trying to scan internal. If I do just hdbois, it has no problem (other than no boot entries). I did disable legacy boot prior to testing latest rEFInd. I did confirm windows still starts fine after disabling Begacy Boot options. I did cold boot in order to test.

I guess my next step is to try moving to working branch. I think I would LOVE it if I could figure out how to load that NVMe driver .efi that I got from the clover EFI loader… ATM, i’m trying to edit an efi file and inject it directly. But I don’t have the knowhow, lol.

[Sebastian Roth]

@szeraax said in Can’t chainload past Fog menu (Boot hard drive) - HP zbook laptop:

I think I would LOVE it if I could figure out how to load that NVMe driver .efi that I got from the clover EFI loader… ATM, i’m trying to edit an efi file and inject it directly. But I don’t have the knowhow, lol.

Sorry, haven’t done that myself yet. I had a look at the Clover project but that seems to be just another universe. Too much for me right now. Keep us posted and I am sure we can jump in with some helpful suggestions from time to time.

[Szeraax]

@sebastian-roth Just to clarify, you think that other people are booting NVMe drives via rEFInd in working without needing to do extra stuff? If so, i’d LOVE to see what their config for refind.conf is… (I’d rather avoid trying to use clover too).

[Szeraax]

@sebastian-roth HOLY CRAP. Just dropping the clover boot.efi in the place of refind.efi booted windows just fine!

[george1421]

@szeraax Interesting!! Sebastian and I have been talking about this issue in the background. My focus was to see if we can get refind to boot via usb into a command shell then we can attempt to debug why its not finding your disk.

But that may not be necessary now. Maybe we need to research what is going on here. I know the easy solution would be to just provide another exit mode and that is where we may end up. But you are not the first to have this issue and refind “should” work. FOG may not be suppling the right info to refind to get it to locate your disks. Once I understand a bit more about refind’s innards, I hope we can ask you to help with your hardware to see if we can come up with a solution.

[Szeraax]

@george1421 Absolutely. Though, i’m told that we need to return this demo laptop cause it doesn’t have enough ram and is missing the 2nd HDD. So it may be a few weeks before I have another laptop of this same model (I probably will turn it in Thursday and can do any testing before then).

[Szeraax]

@george1421 Ok, bad news. Tested twice and then tested just letting the windows boot manager try by itself and when using clover default, my keyboard and mouse do not work in the laptop once it is booted to windows. Both times I booted from clover, I had to hold down power for 7s to get it to shut off. Afterword when booting without pxe, windows came up and worked fine without any issue.

[Szeraax]

Seems to be many people that can’t use keyboard/mouse once booted up. There is some config suggested here https://www.tonymacx86.com/threads/no-mouse-or-keyboard-on-clover-instal.174311/ that might work, if the exit type for Clover were to deliver the boot.efi and config.plist I think?

It seems like the default clover is set up for hackintosh booting. If we disable or fix some of these options, it may work without issue.

[Szeraax]

Ah ha! It seems not that the keyboard and mouse stop working, but that the whole computer freezes. for about half a second, I can move the mouse, but then it stops moving and the clock never changes the minute.

[Szeraax]

And here’s a crappy video I made that shows the default options I’m seeing: https://www.youtube.com/watch?v=pj7JG0TpXEA

[Sebastian Roth]

@szeraax said in Can’t chainload past Fog menu (Boot hard drive) - HP zbook laptop:

Just to clarify, you think that other people are booting NVMe drives via rEFInd in working without needing to do extra stuff? If so, i’d LOVE to see what their config for refind.conf is… (I’d rather avoid trying to use clover too).

From my point of view (not having tested this with several different ones myself but reading about this here and there) it all depends on NVMe drivers. Some drives seem to be kind of generic and just work with the usual rEFInd binary (updated in working branch) and other just don’t. Can’t tell you which do.

As I said, this is not tested. Just the impression I get.

[Szeraax]

@george1421 Regarding your statement about booting rEFInd into a command shell via USB, I think I can probably do that fairly easily. What commands would you want the output of to get a better idea?

Of course, the fact that fog can do single part, resizable imaging on this computer means that it must not be TOO crazy of hardware otherwise I wouldn’t be able to see the disk and such.

[Sebastian Roth]

@szeraax said in Can’t chainload past Fog menu (Boot hard drive) - HP zbook laptop:

Of course, the fact that fog can do single part, resizable imaging on this computer means that it must not be TOO crazy of hardware otherwise I wouldn’t be able to see the disk and such.

No, you have to understand that those are two very different things although we packed those very closely together in the FOG project. One is the Linux Kernel and a stripped down basic Linux system doing the actual imaging work. Many drivers are part of the Kernel. Another component is iPXE that we use for booting and in some cases for exiting to boot to the local disk. In your case this does not seen to work so we have another ring in the chain called rEFInd which has to have it’s own drivers to be able to see and boot from your local disk.

[Szeraax]

@sebastian-roth @george1421 So if I put clover on a USB drive and boot it with the default config.plist, windows starts fine and seems to have no issues. It may just be that we need to transfer the config.plist with the efi file and that exit type would work great.

[george1421]

@szeraax ok, lets do a little digging here.

If you set the exit mode to refind in the webgui this is what it puts in the ipxe menu:

:fog.local
imgfetch ${boot-url}/service/ipxe/refind.conf
chain -ar ${boot-url}/service/ipxe/refind.efi || goto MENU

So in your case you swapped refind.efi for what ever the clover boot loader is. You can see that fog is also sending refind.conf.

Now as a test create a new fog menu, call it fog.clover with a title of Clover Exit or what ever.

Make the parameters look like this

imgfetch ${boot-url}/service/ipxe/config.plist
chain -ar ${boot-url}/service/ipxe/clover.efi || goto MENU

rename clover.efi to what ever the right name is and make sure both files are in /var/www/html/fog/service/ipxe directory.
Lastly in the fog menu configuration make this menu show up for all hosts.

Now pxe boot that host and from the ipxe menu pick the clover exit. The test is to see if we can create a menu entry to properly exit out of iPXE menu to boot.

[Szeraax]

@george1421 haha, I went grepping through /var/www/html and did this exact same thing via ssh before you posted. I made a new menu entry as a copy of $refind

file: /var/www/html/fog/lib/fog/bootmenu.class.php

            'refind_efi' => $refind,
            'clover_efi' => $clover,

and

        $refind = sprintf(
            'imgfetch ${boot-url}/service/ipxe/refind.conf%s'
            . 'chain -ar ${boot-url}/service/ipxe/refind.efi',
            "\n"
        );
        $clover = sprintf(
            'imgfetch ${boot-url}/service/ipxe/config.plist%s'
            . 'chain -ar ${boot-url}/service/ipxe/clover.efi',
            "\n"
        );

And edited /var/www/html/fog/lib/fog/service.class.php

            'refind_efi',
            'clover_efi',

From my test of this and via running the EFI shell manually, config.plist downloads fine. However, the clover boot menu looks totally different when booted from the USB, suggesting to me that when I exit to clover from fog, it is still not reading the config.plist file (which changes graphic modes and other settings that obviously work). imgstat looks good, etc.

I am going to try adding the exit type with a different name via the web UI as you suggest and report back.

[Szeraax]

@george1421 Looks like it still worked the exact same.

Clover menu looks the same as when I chain’d from fog. Still get kernel panic after about 1s on windows login screen. When I exit clover and return to efi shell, imgstat shows config.plist with the right number of bytes and default.ipxe at 452 bytes (script)

Going to try enabling some of the options from config.plist to see if I can avoid kernel panic. Also going to try removing extra things from USB to check if there are other external dependancies to just the clover.efi and config.plist.

I would think it best if we figure out how to get clover to read the config.plist, but I am unsure what the next steps to do that would be.

[Szeraax]

@george1421 Ok, I can remove all files from the USB drive except for these two files and have windows boot fine:

DriveRoot/EFI/BOOT/BOOTX64.efi
DriveRoot/EFI/CLOVER/CONFIG.plist

If I leave config.plist in that folder and rename it to new.plist, when I boot to USB, that config file is NOT found. Does this suggest that clover has a hardcoded path it looks for a config under? Maybe we can pass it an argument that tells it where its config is at.

Current things I want to check out:
-Run same settings as what config.plist sets and see if it’ll chainload to windows fine from fog
-Find out if I can pass arguments or get the clover.efi to find the config file that is present

[Szeraax]

@george1421 @Sebastian-Roth Ok, now that just not fair!!!

If I ONLY have DriveRoot/EFI/BOOT/BOOTX64.efi on the USB drive and boot it with the default options, it works fine. Default options means that I am not doing any ACPI DSDT flag setting, no CPU flag setting, etc. I copied that same .efi file onto fog! So all that stuff that config.plist was supposedly fixing is now out the window since the USB just works regardless of the configuration set.

So that totally chops up both of the things I was going to investigate next. Thoughts?

[Sebastian Roth]

@Szeraax Any news on this?