You are not logged in.
I can reproduce this on my Fujitsu AH512, booting UEFI rEFInd with GPT partition scheme:
- linux-ck 3.7.9-2 boot's fine
- linux 3.7.9-2 and linux-ck 3.7.10-1 refuses bootI use repo-ck...
Could it be that firmware differences among manufacturers account for the fact that a 3.7.9-2 kernel boots on a Lenovo but not on a Fujitsu?
I think I read that different manufacturers are implementing the UEFI specification in slightly different fashions. I don't really know enough about this side of things to do more than speculate.
Last edited by dhave (2013-03-01 21:34:21)
Offline
Tested 3.7.9-2, and it boots just fine. ThinkPad W530, Gummiboot and UEFI-only mode.
Sometimes, when I'm trying to get any audio software or hardware working with my system, I wonder why I ditched Windows. But every time I work at a windows computer, I remember it again.
Offline
linux-ck 3.8.1-1 (from the AUR) booting fine on Thinkpad T420, rEFInd, UEFI-only mode.
Offline
Now, this is strange:
* linux 3.8.1-1 boot
* linux-ck-sandybridge 3.8.1-1 don't boot
* linux-ck-ivybridge 3.8.1-1 boot
(both ck packages from repo-ck)
For me its looks like a build/gcc optimization problem.
Some info on my system:
* UEFI only, gummiboot
* Samsung 500P4C
* i7-3630QM
$ gcc -c -Q -march=native --help=target | grep march
-march= corei7-avx
Offline
Fujitsu AH512: after upgrade to linux-ck 3.8.1-1 boot ok...
Offline
Kernel linux-ck 3.8.2-1 (from the AUR) also booting fine on Thinkpad T420, rEFInd, UEFI-only mode.
Offline
Arch kernel 3.7.10-1 does *not* boot on my Thinkpad T420, rEFInd, UEFI-only mode.
The mystery continues.
Last edited by dhave (2013-03-05 17:06:17)
Offline
Are all of these packages compiled by the same person? (My 3.7.5-1 kernel is compiled by Tobias Powalowski.) If not, it could be that differences in the build environment by two packagers could explain the variable behavior. If they're all compiled by the same person, then I suggest contacting the maintainer or filing a bug report. Nobody's mentioned doing so in this thread. I'm not affected, so I don't think it's my place to file such a bug report, but the extremely variable behavior is certainly cause for filing a bug report.
Offline
Arch kernel 3.8.1-1 does *not* boot on my Dell Inspiron 15R 5221, rEFInd, UEFI-only mode.
Arch kernel 3.8.2-1 boot..
what a strange mystery
Offline
Are all of these packages compiled by the same person? (My 3.7.5-1 kernel is compiled by Tobias Powalowski.) If not, it could be that differences in the build environment by two packagers could explain the variable behavior.
I've suspected this, but I'm not so sure about it since I've been using linux-ck and those are always built by graysky. Nevertheless, here's a helpful one-liner to query your pacman cache for the packager. I've reinstalled recently, so I don't have many kernels, so it would be interesting to have someone else run this:
$ find /var/cache/pacman/pkg/ -name 'linux-3.*' | xargs -I{} sh -c 'echo -n "{}: "; tar xf {} .PKGINFO -O | grep packager'
/var/cache/pacman/pkg/linux-3.7.8-1-x86_64.pkg.tar.xz: packager = Evangelos Foutras <evangelos@foutrelis.com>
/var/cache/pacman/pkg/linux-3.7.9-2-x86_64.pkg.tar.xz: packager = Tobias Powalowski <tpowa@archlinux.org>
/var/cache/pacman/pkg/linux-3.7.9-1-x86_64.pkg.tar.xz: packager = Evangelos Foutras <evangelos@foutrelis.com>
Offline
Add these:
/var/cache/pacman/pkg/linux-3.7.6-1-x86_64.pkg.tar.xz: packager = Tobias Powalowski <tpowa@archlinux.org>
/var/cache/pacman/pkg/linux-3.7.7-1-x86_64.pkg.tar.xz: packager = Evangelos Foutras <evangelos@foutrelis.com>
/var/cache/pacman/pkg/linux-3.7.9-1-x86_64.pkg.tar.xz: packager = Evangelos Foutras <evangelos@foutrelis.com>
/var/cache/pacman/pkg/linux-3.8.1-1-x86_64.pkg.tar.xz: packager = Tobias Powalowski <tpowa@archlinux.org>
/var/cache/pacman/pkg/linux-3.8.2-1-x86_64.pkg.tar.xz: packager = Tobias Powalowski <tpowa@archlinux.org>
Offline
FWIW: 3.7.9.2 Boots fine on my UEFI Laptop. This laptop is new -- just had it two weeks. 3.7.10.1 DOES NOT boot on this laptop. I had to roll back to 3.7.9.2. I have no idea how to help on this.
Laptop is an Acer Aspire V3-771G-9441. i7 nVidia Optimus 730M/Intel 8 GB RAM, etc.
Last edited by mrunion (2013-03-07 02:26:26)
Matt
"It is very difficult to educate the educated."
Offline
Latest Arch ISO 20130301 doesnt boot UEFI Only mode and rEFInd. Lenovo IdeaPad u510
Offline
Well ...3.7.10-1 doesn't boot with my system as well. Using rEFInd on T430.
Until now, I stayed on 3.7.7 but then I had some nvidia module problem, so I decided to upgrade to 3.7.10 but that didn't do any good. The boot problems reappeared.
What I did now is that I migrated on linux-ck (sandybridge). Currently I'm running 3.8.2-1-ck and it is all OK. I think I'm staying on CK.
Packages info:
find /var/cache/pacman/pkg/ -name 'linux-3.*' -o -name 'linux-ck-*' | xargs -I{} sh -c 'echo -n "{}: "; tar xf {} .PKGINFO -O | grep packager'
/var/cache/pacman/pkg/linux-3.7.9-1-x86_64.pkg.tar.xz: packager = Evangelos Foutras <evangelos@foutrelis.com>
/var/cache/pacman/pkg/linux-3.7.10-1-x86_64.pkg.tar.xz: packager = Tobias Powalowski <tpowa@archlinux.org>
/var/cache/pacman/pkg/linux-ck-sandybridge-3.8.2-1-x86_64.pkg.tar.xz: packager = graysky <graysky@archlinux.us>
/var/cache/pacman/pkg/linux-3.7.7-1-x86_64.pkg.tar.xz: packager = Evangelos Foutras <evangelos@foutrelis.com>
/var/cache/pacman/pkg/linux-ck-sandybridge-headers-3.8.2-1-x86_64.pkg.tar.xz: packager = graysky <graysky@archlinux.us>
Out of these only 3.7.7 and the ck-sandybridge work on my machine.
Last edited by petrov (2013-03-10 12:16:41)
Offline
I set up my PC with rEFInd yesterday. I'm using the ext4 driver and it shows me my kernel in /boot so it seems to be working. At first I had my kernel in /boot and a copy as *.efi in EFI/arch. Kernel Version is local/linux 3.7.10-1 (base)
With refind_linux.conf
## This file should be present in the same directory as the EFISTUB kernel and initramfs files
## More info at http://www.rodsbooks.com/refind/linux.html , http://www.rodsbooks.com/efi-bootloaders/efistub.html
"Boot with defaults" "root=/dev/sda4 ro rootfstype=ext4 elevator=noop add_efi_memmap"
"Boot to terminal" "root=/dev/sda4 ro rootfstype=ext4 elevator=noop add_efi_memmap systemd.unit=multi-user.target"
Both (/boot and /boot/efi/EFI/arch) boot without problems. So I removed my /boot/efi/EFI/arch directory, because I didn't see a speed penality from using the ext4 driver instead of booting from FAT32 EFI partition.
So I install 3.8.2-CK for Sandybridge and tried to boot it from /boot but it end up not being able to find /dev/sda4 and drops to root shell. Now when I copy the CK files to /boot/efi/EFI/arch and rename them to vmlinuz-linux-arch.efi etc. I can boot the CK kernel with the same refind_linux.conf.
Why doesn't it work like the normal 3.7.10 kernel from /boot which strangely seems to cause problems for a lot here but not for me?
Offline
I don't know if I've got the brass to try 3.8 yet. I do know that when I installed Arch on this new laptop, I seem to have done something that may be unusual. I followed the Beginners Guide and when it got to the part about installing a boot loader, I followed the instructions for creating an ESP partition (or whatever it's called). Because my laptop already had Windows 8 on it, I just created ANOTHER ESP partition and installed the Arch stuff there. So my machine has TWO ESP partitions. (I didn't know any better!)
As far as I'm aware, this is not "wrong", but I wonder if something in 3.7.10 doesn't like this? My machine will boot rEFInd and let me pick from my boot options with both 3.7.9.2 and 3.7.10.1. Also, when I select Arch and the next "boot" screen pops up -- the one with the blue bar at the top and the boot line that shows the boot command(s) -- 3.7.9.2 will continue booting, but 3.7.10.1 will just hang there.
Do others with the problem have a separate ESP from the "Windows 8" one? Should I "fix" this on my machine and put the Arch boot stuff on the pre-existing, Windows 8 ESP?
Sorry for all the questions.
Last edited by mrunion (2013-03-10 17:14:26)
Matt
"It is very difficult to educate the educated."
Offline
Should I "fix" this on my machine and put the Arch boot stuff on the pre-existing, Windows 8 ESP?
The EFI spec explicitly permits the use of multiple ESPs; however, the Windows 7 installer flakes out when it sees such systems, so I recommend against their use, at least on computers that dual-boot with Windows, since you may need to re-install Windows at some point. (I don't know if Windows 8 has this same problem; if not, using two ESPs shouldn't be such a risk with it.) AFAIK, the Linux kernel doesn't care about partition type codes, so you should be safe on that score.
If you've got the kernel on your second ESP but if you've got rEFInd on your first ESP, another option would be to change the type code of the second ESP, rendering it a normal FAT partition. I'd give it a type code of 8300 in gdisk, since that will keep Windows from accessing it. (OTOH, you might want Windows to be able to access it to do emergency edits on refind_linux.conf or whatnot, in which case a gdisk type code of 0700 would be better.) Such a change might or might not work if you've put rEFInd on the second ESP, though. AFAIK, most EFIs will boot from non-ESP FAT partitions, but the spec doesn't require them to do so. Therefore, changing the type code of an ESP that holds a boot loader that the EFI launches directly might render that boot loader un-bootable.
Offline
REFInd is on the second ESP. Also, the type I created was already 8300 and it is a FAT32 partition. It would be "nice" if Windows would see it as a boot option, but it's not to big of a deal to me. Also, this laptop runs Windows 8 (yuck!), and though it doesn't "flake out", it only boots Windows 8 itself. I had to rearrange my boot order in BIOS to get it to boot ReFind since Windows 8 ESP was the first bootable option by default.
I am on a STEEP learning curve here, and am trying to catch up to the EFI stuff. Thanks for all the help.
Matt
"It is very difficult to educate the educated."
Offline
REFInd is on the second ESP. Also, the type I created was already 8300 and it is a FAT32 partition.
If your "second ESP" has a gdisk type code of 8300, then it's not an ESP; it's just an ordinary FAT partition. To be an ESP, the partition must have a type code that gdisk reports as EF00, or that shows up in parted as having a "boot flag" set. If the computer can be configured to launch rEFInd from this partition, then it's obviously not beeing too fussy about booting things from a location other than the ESP.
Offline
You are correct. My fault. It was EF00, not 8300. I reread through the Beginners Guide and my notes when I installed and that's what I set it to. Again, my apologies.
Matt
"It is very difficult to educate the educated."
Offline
I just did clean install of Arch. I am on linux 3.7.10-1-ARCH and refind-efi 0.6.7-1. I followed the beginners and UEFI guide pretty closely. My system hangs during boot. It boots fine when I first choose fallback, which doesn't do anything but take back to rEFInd screen, and then stock kernel.
Offline
I just did clean install of Arch. I am on linux 3.7.10-1-ARCH and refind-efi 0.6.7-1. I followed the beginners and UEFI guide pretty closely. My system hangs during boot. It boots fine when I first choose fallback, which doesn't do anything but take back to rEFInd screen, and then stock kernel.
If you got the time to play around maybe you can try 3.8 from the Testing repo.
Last edited by petrov (2013-03-11 21:22:48)
Offline
I just did clean install of Arch. I am on linux 3.7.10-1-ARCH and refind-efi 0.6.7-1. I followed the beginners and UEFI guide pretty closely. My system hangs during boot. It boots fine when I first choose fallback, which doesn't do anything but take back to rEFInd screen, and then stock kernel.
Do you use PARTUUIDs to point to your root partition? I tested rEFInd a lot since last Friday and encountered a situation where I it would hang when showing me the kernel options used. This was when directly selecting the loader in the rEFInd mainscreen. So rebooted and selected the loader but pressed "F2" and "Boot with defaults" and it booted. Rebooted and selected the mainloader -> Hangs. Rebooted. "Boot with defaults" -> Boots. Rebooted used the mainloader. -> Hangs. This was rather strange since I thought it simply uses the default boot, when not pressing F2 to open the submenu.
After changing root=PARTUUID=somestrange number to root=/dev/sda4 (my root partiton) I never had any issues.
From my "rather short" experience with rEFInd it is best to sync the EFISTUB to /boot/efi/EFI/arch and in refind_linux.conf set the root partition as kernel name. After doing that I never had any hangs when trying to boot no matter if thourgh the submenu or not.
root=/dev/sda4 solved my issues not being able to boot the kernel from /boot without using the submenu. Using /EFI/arch directory with renamed files solved the problem of not being able to boot the 3.8.2-CK kernel from /boot (could not find root no matter how it was declared in refind_linux.conf) while 3.7.10-1-ARCH had no problems with the same config in /boot (when used through the submenu).
Last edited by blackout23 (2013-03-11 23:26:27)
Offline
blackout32, can you give a EFI-newbie like myself more info on how to do what you're saying.
donniezazen: 3.7.10.1 is not booting for many of us, but 3.7.9.2 boots fine. It's that way for me as well. I've just been staying with 3.7.9.2. I may try blackout32's advice if I can digest what he's saying to do.
Last edited by mrunion (2013-03-14 14:19:12)
Matt
"It is very difficult to educate the educated."
Offline
Kernel 3.7.10-1 --> Not booting.
This system is on an usb stick that was not upgraded since 3.6.x and this is the first time it happens to me on my thinkpad x220 laptop.
The boot procedure hangs showing the refind screen with the blue bar on top. Now i am happy with my decision to keep MBR booting on my primary archlinux.
Offline