Debugging Lenovo Ideapad Miix 310

[Lee Rowlett]

@wayne-workman forum is deciding to fight with me… lol, added link to higher res

[Sebastian Roth]

Have been onto this with Lee for a bit already. First we thought this might be the same issue that we see with the Lenovo Miix 320 (see here). But turns out it’s not. So this is where I asked him to start a new thread and post things here.

[Sebastian Roth]

Ok again have been working on this in chat and from what it looks like so far the issue is very similar to what we had with the HP x2 210 tablet (yeah, @Matthieu-Jacquart believe it or not).

Lee is telling me that he sees the Early in __setup_efi_pci64… 1 2 3 4 5 debug output that we had with the HP book as well.

[Matthieu Jacquart]

@sebastian-roth Oh dear, so bad memories… good luck for resolution !

[Sebastian Roth]

@Lee-Rowlett I followed the video closely many times but I am still not exactly sure what’s going on. It kind of seems like we had the same hang as we saw on the HP x2 210. But using the fixed kernel I provided you were still not able to boot. So maybe there is another issue beside the PCI initialization.

So we need to start the kernel debugging from scratch again. Please try 01_bzimage as you have done with the other two kernel images as well. Use the normal/non-debug ipxe.efi and set the host’s kernel parameter to debug earlyprintk=efi loglevel=7. Schedule a job, boot it up and see what you get.

I guess 01_bzImage hangs on Early in __setup_efi_pci64… 1 2 3 4 5.

If yes, try 02_bzImage which does essentially the same thing but should also print the PCI ROM information that causes the lockup. Take a picture of the screen and post here.

Then try 03_bzImage which should fix the bug on PCI setup. Take a picture if you still see a hang. With that kernel image.

[Lee Rowlett]

@sebastian-roth as predicted 01_bzImage hangs on Early in __setup_efi_pci64… 1 2 3 4 5

testing 02_bzImage now…

[Lee Rowlett]

02_bzimage

[Lee Rowlett]

03_bzimage

[Lee Rowlett]

04_bzimage

[Sebastian Roth]

Ok, seems like this piece is another one with cranky firmware on it. The ROM size you see printed on screen is way too large to be a proper PCI expansion ROM image. The Linux kernel actually tries to allocate this much memory for the expansion ROM(s) in that very early startup and hits the wall. From the figures it looks like - and we have seen the same on that HP x2 210 tablet - ROM size is rather being initialized with some weird pointer instead of the actual size (unsigned integer).

So let’s try something just for the fun of it. 05_bzImage and 06_bzImage (the later without skipping - just in case you’re missing the download link) dereferences the ROM size “pointer”. I have a feeling that this might horribly fail but let’s give it a go.

As well I added 07_bzImage that does not do much printing but only skips the PCI ROM initialization stuff. I have a feeling that maybe the printout might be causing the later hang.

And then there is another heavy debug build in 08_bzImage. So hopefully we’ll see where exactly in exit_boot it does hang.

[Sebastian Roth]

@Lee-Rowlett Any news on this yet?