PXE Boot HP X2 210 (Hybrid tablet Windows 10 Pro)

[Matthieu Jacquart]

@Sebastian-Roth ok done, but I saw no difference (I put loglevel to 7 in fog interface, just in case)

USB

iPXE

[Tom Elliott]

@Sebastian-Roth Very true. I just was hoping to see default behavior play a little bit better.

[Sebastian Roth]

@Matthieu-Jacquart Ok, got some news. Hope you are still keen to keep on trying out things. Will be several tests and I will post a short message for each to not mix things up.

First please move your original ipxe.efi binary, download this (DEBUG=efi_wrap with some added dump code within HandleProtocol function) and put in place. This is with full embedded script so you don’t have to type all the commands in iPXE shell. Hope we still see the output! Boot your tablet on this using any of the kernels. Doesn’t really matter which one as they all fail on the same issue. Please take a picture of the screen when it freezes. Don’t need a video or several pictures of the whole process.

[Sebastian Roth]

Ok, then I started to think about quick fixing the issue right in the kernel. The first try is to entirely skip the EFI PCI setup thing. I had the idea because when EFI boot was added to the kernel this PCI setup was not part of it but added later. From what I understand about this the setup code is mostly just loading optional PCI ROM BARs. Skipping this some PCI devices won’t work in the worst case I hope.

You might want to leave the debug ipxe.efi in place - shouldn’t hurt - and use bzImage_sp. There might be other things happening. I have no idea. I booted this kernel in my virtual QEMU test environment fine and hopefully it will boot up on your tablet just as well.

[Matthieu Jacquart]

@Sebastian-Roth Ok, I’ll test this on monday ! Great thanks

[Sebastian Roth]

And number three. Another kernel fix that might address the issue a bit better than just skipping the whole EFI PCI setup. Again, I am not sure what’s the outcome of this. But it can’t be much worse than what we have now. Possibly this check can prevent the issue on your tablet. Try bzImage_rb that you find in the same shared gdrive folder.

[Sebastian Roth]

@Matthieu-Jacquart Seams like we are pushing this quite far. Michael from iPXE now contacted people at HP (responsible for that particular UEFI firmware build) to see what they might know about this. Might be a firmware bug. We will see.

From the things I posted the ipxe.efi binary with more dump code added is the most important! This might be helpful information. Please get into this first on Monday.

And there is one more thing that just came to my mind. Have you tried all these things on one of your tablets only? Could you please do the USB stick test on at least two or three other devices to make sure we see the same issue and address/size values on all of them!!!

[Matthieu Jacquart]

@Sebastian-Roth From beginning I always made each test on 2 tablets, I hope it’s good enough

ok first one with ipxe file (DEBUG=efi_wrap…)

[Matthieu Jacquart]

@Sebastian-Roth Second test, with bzimage_sp. Tested with ipxe and usb key (both tests on both tab)

iPXE

USB

[Matthieu Jacquart]

@Sebastian-Roth And third test with bzImage_rb (once again on both tablets with iPXE and USB)

iPXE

USB

[Sebastian Roth]

@Matthieu-Jacquart Nice, thanks for all the pictures and testing. Got a first answer from HP but seams like they are on holiday till the 15th and won’t be able to look into this earlier.

So my quick fixes only worked partly. Seams like it fixed the hang when booting via USB. The kernel panic is fine because we cannot easily add kernel parameters and initrd when booting the kernel straight as bootx64.efi. Unfortunately the same kernel seams to hang when trying to ExitBootServices. One obvious difference is the efi_wrap debug outputs in iPXE. So you might want to go back to the original ipxe.efi and give that a try with bzImage_rb… keeping my fingers crossed.

[Matthieu Jacquart]

@Sebastian-Roth POPOPO !!! Much better !

Upload task with USB 2.0/100Mb adapter…

Many thanks Seb, that’s great !

[Sebastian Roth]

Great!!! Let’s see if this quick fix will be enough for you to capture and deploy those devices. I really hope it will as getting it fixed by HP or the BIOS vendor (Insyde) may take weeks. Please keep us posted as you try deployment as well.

[Matthieu Jacquart]

@Sebastian-Roth Just done, I’ve uploaded image from first tablet I have prepared and deploy it on second one, everything seems fine
That’s cool ^^

[Wayne Workman]

Adding the HP X2 210 to the wiki of working devices, I will link it to this thread.

@Matthieu-Jacquart @Sebastian What was the boot file used? The version of FOG? The kernel? Exactly what model of USB adapter did you use? What was done to make this work?

[Matthieu Jacquart]

@Wayne-Workman For now it seems to work only with modified bzImage given by Sebastian (bzImage_rb). No other change (stock ipxe.efi file), last svn used

@Wayne-Workman And network adapter is ASIX AX88772C. I’ll try to buy one USB 3.0, this one is just USB 2.0 an limited to 100Mb…

[Wayne Workman]

@Matthieu-Jacquart said:

I’ll try to buy one USB 3.0

Please let us know how that goes. We really need to get into the USB 3.0 stuff.

[Sebastian Roth]

@Wayne-Workman As Matthieu said this is only possible with the patched bzImage_rb. But it’s a quite ugly hack I think and I have absolutely no idea if this will cause issues on other devices. Up to now it looks like the freeze is caused by a bug in the UEFI firmware but we don’t have a clear confirmation yet. Will take some time. iPXE is definitely not an issue on these HP X2 210 tablets and therefore trunk version is fine except the linux kernel.

[Matthieu Jacquart]

@Wayne-Workman I think we are going to buy this one (Realtek RTL8153).
http://www.tp-link.com/en/products/details/cat-5062_UE300.html
Any know issue with this chipset ?

[Sebastian Roth]

Try searching the forum for RTL8153. From what I know several people have trouble using RTL8153 based USB NICs even in BIOS legacy mode. I don’t think UEFI will be easier. Probably best to start a new discussion on this as it might be a long way to go just as this one was.