With 3.17-git (x64, self compiled) I'm now getting the newly introduced "setup_efi_pci() failed!" message on every boot, both on my laptop and desktop (each using gummiboot). But that's it, the boot just continues as before.
(On my desktop. On the laptop, it actually aborts with "initramfs unpacking failed", but I'm inclined to blame that on something else.)
Correction, after bisecting it turns out that it actually is the recent x86-efi-for-linus branch merge that causes the laptop to fail to boot:
Attached is a picture of the message it prints.
In case it matters, laptop is a Clevo W840SU, mobo of the desktop a Gigabyte G1.Sniper M5 (bios F9a).
Created attachment 145621 [details]
Andreas, it would be helpful if you could track down exactly which commit introduces this regression.
The commit that introduced the "setup_efi_pci() failed" message is this one, but I doubt that it's directly responsible for your issues,
Alright, the commit that causes my laptop to fail is "x86/efi: Support initrd loaded above 4G", checked both by bisection and reverting it from the current HEAD.
Still getting the "setup_efi_pci() failed!".
Laptop has 8G RAM and the desktop 16G. Also, the laptop uses dm-crypt drive encryption.
Andreas, thanks for doing the bisect. Could you please provide dmesgs for both of your machine with a working kernel?
Also, please let me know if the following commit helps (you'll need to add efi=nochunk to your EFI boot stub params),
Created attachment 146651 [details]
Yes, the commit together with efi=nochunk works (kernel at 899552d6). Attached is the resulting dmesg.
Created attachment 146681 [details]
Ups, forgot about the second dmesg. Here it is.
I pulled from linus' repo yesterday (commit 899552d6e84babd24611fd36ac7051068cb1eb2d) and received this same error message when trying to boot into the kernel. I was not actually able to boot into the kernel due to kernel panic about not being able to mount the root fs (I'm a newb so am not sure if this is relevant to the specified error). I did not know how to obtain log data from the boot, so I attached a picture of the kernel panic core dump. Sorry I am not more help!
Created attachment 146751 [details]
kernel panic core dump after "setup_efi_pci() failed"
I should note that it says to 'Please append a correct "root=" boot option', but I do have this kernel parameter set (to root=/dev/sda3).
Jaret, could you try building a kernel with the following patch?
Hopefully that should fix your boot failure.
Can't speak for Jaret, but it's working for my laptop, unsurprisingly.
Gah, I feel like a jerk. I have actually re-installed Windows on my PC in order to practice .NET programming (I am currently not a programmer but trying to find a programming position). Odds are I will switch back to Linux in the near future, but I won't be able to test out this patch for a little while. Thank you so much for responding to my post though. =/
(In reply to Matt Fleming from comment #11)
> Jaret, could you try building a kernel with the following patch?
> Hopefully that should fix your boot failure.
with that patch my computer boots however showing the "setup_efi_pci() failed" message. without it i get something mentioning initramfs containing junk and stops there.
this is on a desktop with the motherboard MSI Z77A-G45 Socket-1155, with kernel 3.17-rc5
When running on x86_64 3.17-rc6 compiled from git, my Asus laptop displays "setup_efi_pci() failed" and then immediately reboots. The system boots fine on 3.16. I'll attach dmesg from booting into 3.16. There's no additional debug data displayed. It looks like 3.17-rc6 contains Matt's patch. I'll try and figure out what the last 3.17 rc I can boot into is and bisect, but if there's anything additional I can do to assist debugging please let me know.
Created attachment 151981 [details]
3.17-rc6 dmesg from Acer Aspire E5-571-563B
Maybe try with the revert contained in rc7:
Sounds like the same issue.
Rebuilt on the latest and it works. Still displays the setup_pci_efi error, but it continues to boot. Thanks.
Given the fixes for both the boot failures and the "failed" message (with "x86/efi: Delete misleading efi_printk() error message" in rc7), I'll take the liberty of closing the bug. Thanks, devs!