Linux-3.10.10 fails silently when booting with UEFI

HalosGhost · 2013-09-02 02:32:21

This problem has occured for me on several occasions in the past (with 3.10.2 and 3.10.8) but that has been with kernels in [testing]. With 3.10.10 dropping to [core], this is the first time this has happened for me with an official release.

Though the linked-to thread is marked as "Solved", the only solution found was downgrading and waiting for a new kernel to be released that did not have the issue (3.10.3 and 3.10.9). I am using UEFI with gummiboot, but I have no idea why this is happening (and since the boot fails silently it's rather difficult for me to provide helpful logs). Is anyone else experiencing this issue? Have you found any workable solution that isn't just waiting? Can anyone else guide me in some troubleshooting?

All the best,

-HG

jasonwryan · 2013-09-02 02:40:10

I can confirm it with UEFI/gummiboot. It works fine on MBR/syslinux boxes.

As a potential comparison: this kernel config boots fine on the same UEFI/gummiboot machine https://bitbucket.org/jasonwryan/shiv/s … onfig.last

srs5694 · 2013-09-02 04:19:41

I haven't read the whole thread to which you referred, but I believe it's the problem discussed at length in the following thread:

https://bbs.archlinux.org/viewtopic.php?id=156670

Unfortunately, there's no real fix at the moment except to stick with a working kernel or switch your boot manager/loader.

HalosGhost · 2013-09-02 14:08:22

srs5694 wrote:

I haven't read the whole thread to which you referred, but I believe it's the problem discussed at length in the following thread:
https://bbs.archlinux.org/viewtopic.php?id=156670
Unfortunately, there's no real fix at the moment except to stick with a working kernel or switch your boot manager/loader.

I've been contemplating switching back to syslinux anyway since UEFI support for it is in [testing]. Having said that, this isn't booting EFISTUB from the firmware, this is booting from gummiboot (which, if I understand the UEFI boot process correctly, are different). That thread was quite long and I stopped following it around page five, do you happen to know if they ever discovered what the exact issue was that caused booting to silently fail?

Update: I just tried switching to syslinux, and though it doesn't silently fail, it appears to hang during the boot process and I cannot solve it, so it looks like I'll have to stick with gummiboot and 3.10.9 for now.

All the best,

-HG

Last edited by HalosGhost (2013-09-02 16:20:05)

srs5694 · 2013-09-02 16:44:05

Booting directly from the firmware and via gummiboot are basically the same, in the sense that they both rely on the EFI stub loader. The problem under discussion in the thread to which I linked appears to be a bug in the EFI stub loader that interacts with the firmware version -- it affects many Lenovo computers, but few others. It appears when using the EFI stub loader directly, from rEFInd, or from gummiboot; but it's a highly inconsistent bug. It can go away by switching from gummiboot to rEFInd, from rEFInd to gummiboot, from one version of rEFInd to another, from either of these to a direct boot from the firmware, etc. It also comes and goes with the exact kernel version, with different people reporting problems with different kernels.

As I said, there's no real fix. The precise cause hasn't been nailed down; if it had been, chances are there'd be a fix by now.

If SYSLINUX is getting you further into the boot process, then I recommend you try to debug what's happening with SYSLINUX. Try posting details of any error messages you see.

HalosGhost · 2013-09-02 18:14:54

srs5694 wrote:

Booting directly from the firmware and via gummiboot are basically the same, in the sense that they both rely on the EFI stub loader. The problem under discussion in the thread to which I linked appears to be a bug in the EFI stub loader that interacts with the firmware version -- it affects many Lenovo computers, but few others. It appears when using the EFI stub loader directly, from rEFInd, or from gummiboot; but it's a highly inconsistent bug. It can go away by switching from gummiboot to rEFInd, from rEFInd to gummiboot, from one version of rEFInd to another, from either of these to a direct boot from the firmware, etc. It also comes and goes with the exact kernel version, with different people reporting problems with different kernels.
As I said, there's no real fix. The precise cause hasn't been nailed down; if it had been, chances are there'd be a fix by now.

Fascinating. I wonder what's going wrong with it.

If SYSLINUX is getting you further into the boot process, then I recommend you try to debug what's happening with SYSLINUX. Try posting details of any error messages you see.

Fair enough. Then, here's my plan. Given that this thread appears to be a duplicate of the thread srs5694 linked to, it should probably either be merged in or closed (I don't necessarily have any preference, and since there has been no new discovery made in it, it doesn't seem as though there is any inherent advantage one way or another). And, I will shortly open a new thread on troubleshooting syslinux.

Thank you for your guidance srs! You've been a huge help.

All the best,

-HG

WonderWoofy · 2013-09-02 19:48:23

FWIW, I have syslinux-efi from [testing] set up for a bit now (and also tested the.ridikulus.rat's syslinux-efi AUR package for a while before that), and it is fabulous! I still rely primarily on gummiboot/efistub as my primary method of booting, but I always keep a backup bootloader set up just in case. I had been bitten by this bug back in 3.7 or 3.8, and it sucked. So at that point, I set up elilo and grub-efi as a "just in case" kind of a thing. So now I have grub-efi, elilo, syslinux-efi, gummiboot, and rEFInd all set up on my machine. So unless there is something grossly wrong with the kernel itself or the initramfs, I should be able to get my machine going.

BTW, a bit OT, but I want to point out that although I really dislike grub-efi, it allows for the booting of iso's in a loop mount. So that is why I keep it around. I just put a custom archiso in my ESP and then have it set up to boot that. I wanted to use syslinux and memdisk, but unfortunately, the memdisk functionality is bios specific, with no real plans on porting a memdisk.efi. I guess as a btrfs/linux-git user who likes to tinker with things, I feel as though I should have a backup solution in the event that i manage to bork the shit out of my machine.

jasonwryan · 2013-09-03 22:21:43

For the record, 3.11 in [testing] fails as well. 3.10.9 is the last vanilla kernel that UEFI boots successfully.

HalosGhost · 2013-09-03 23:08:12

jasonwryan wrote:

For the record, 3.11 in [testing] fails as well. 3.10.9 is the last vanilla kernel that UEFI boots successfully.

Really? 3.11.0 is booting correctly for me with Gummiboot (though it's not quite booting correctly with syslinux yet—I'm working on that).

All the best,

-HG

Arch Linux

#1 2013-09-02 02:32:21

Linux-3.10.10 fails silently when booting with UEFI

#2 2013-09-02 02:40:10

Re: Linux-3.10.10 fails silently when booting with UEFI

#3 2013-09-02 04:19:41

Re: Linux-3.10.10 fails silently when booting with UEFI

#4 2013-09-02 14:08:22

Re: Linux-3.10.10 fails silently when booting with UEFI

#5 2013-09-02 16:44:05

Re: Linux-3.10.10 fails silently when booting with UEFI

#6 2013-09-02 18:14:54

Re: Linux-3.10.10 fails silently when booting with UEFI

#7 2013-09-02 19:48:23

Re: Linux-3.10.10 fails silently when booting with UEFI

#8 2013-09-03 22:21:43

Re: Linux-3.10.10 fails silently when booting with UEFI

#9 2013-09-03 23:08:12

Re: Linux-3.10.10 fails silently when booting with UEFI

Board footer