qubes-doc/uefi-troubleshooting.md at b324c27fdc55bf77ea55dc805cb94f78b2fb2d2d

mirror of https://github.com/QubesOS/qubes-doc.git synced 2024-10-01 01:25:40 -04:00

2018-04-02 07:05:28 +00:00

9.0 KiB

Raw Blame History

layout	title	permalink
doc	UEFI Troubleshooting	/doc/uefi-troubleshooting/

Change installer kernel parameters in UEFI

If you've installed successfully in legacy mode but had to change some kernel parameters for it to work, you should try installing in UEFI mode with the same parameters.

Change the xen configuration on a USB media

Attach the usb disk, mount the EFI partition (second partition available on the disk)
Edit your xen config (xen.cfg/BOOTX64.cfg) changing the kernel key to add your kernel parameters on the boot entry of your choice
Install using your modified boot entry

Change xen configuration directly in an iso image

Get EFI partition boundaries parted Qubes-R4.0-rc4-x86_64.iso unit B print
Using the start address and the size of the EFI partition, setup a loop device for it sudo losetup -o 524288 --sizelimit 30562304 /dev/loop0 Qubes-R4.0-rc4-x86_64.iso
Mount the loop device sudo mount /dev/loop0 /mnt
Edit EFI/BOOT/BOOTX64.cfg to add your params to the kernel configuration key
Save your changes, unmount and dd to usb device

There is some common bug in UEFI implementation, affecting mostly Lenovo systems, but probably some others too. You can try existing workaround:

In GRUB menu¹, select "Troubleshoot", then "Boot from device", then press e.
At the end of chainloader line add /mapbs /noexitboot.
Perform installation normally, but don't reboot the system at the end yet.
Go to tty2 (Ctrl-Alt-F2).
Enable /mapbs /noexitboot on just installed system. This step differs between Qubes releases:

For Qubes 3.1:
Execute mount | grep boot/efi and note device name (first column). It should be something like /dev/sda1.
Execute efibootmgr -v, search for Qubes entry and note its number (it should be something like Boot0001 - 0001 is an entry number).
Replace existing Qubes entry with modified one. Replace XXXX with entry number from previous step, /dev/sda with your disk name and -p 1 with /boot/efi partition number):
```
efibootmgr -b XXXX -B
efibootmgr -v -c -u -L Qubes -l /EFI/qubes/xen.efi -d /dev/sda -p 1 "placeholder /mapbs /noexitboot"
```

Compare new entry with the old one (printed in step 6) - it should only differ in additional options at the end, and look probably something like this:

Boot0001* Qubes HD(1,GPT,partition-guid-here,0x800,0x64000)/File(\EFI\qubes\xen.efi)p.l.a.c.e.h.o.l.d.e.r. ./.m.a.p.b.s. ./.n.o.e.x.i.t.b.o.o.t.

If instead it looks like:

Boot0001* Qubes HD(1,0,00000000...0,0x0,0x0)/File(\EFI\qubes\xen.efi)p.l.a.c.e.h.o.l.d.e.r. ./.m.a.p.b.s. ./.n.o.e.x.i.t.b.o.o.t.

then try passing /dev/sda1 or /dev/nvme0n1p1 or whatever your EFI partition is instead of /dev/sda and -p 1.

Now you can reboot the system by issuing reboot command.

For Qubes 3.2 or later:
Edit /mnt/sysimage/boot/efi/EFI/qubes/xen.cfg (you can use vi editor) and add to every kernel section:
```
mapbs=1
noexitboot=1
```
Note: You must add these parameters on two separate new lines (one parameter on each line) at the end of each section that includes a kernel line (i.e., all sections except the first one, since it doesn't have a kernel line).
Now you can reboot the system by issuing reboot command.

System crash/restart when booting installer

Some Dell systems and probably others have another bug in UEFI firmware. And there is another workaround for it:

In GRUB menu¹ press e.
At the end of chainloader line add -- efi=attr=uc.
Perform installation normally, but don't reboot the system at the end yet.
Go to tty2 (Ctrl-Alt-F2).

Execute:

 sed -i -e 's/^options=.*/\0 efi=attr=uc/' /mnt/sysimage/boot/efi/qubes/xen.cfg

or if you're installing 3.2 execute:

 sed -i -e 's/^options=.*/\0 efi=attr=uc/' /mnt/sysimage/boot/efi/EFI/qubes/xen.cfg

Now you can reboot the system by issuing reboot command.

1 If you use rEFInd, you can see 3 options regarding the USB installer. Choose "Fallback Boot Loader" to enter the GRUB menu. ↩ ↩

Boot device not recognized after installing

Some firmware will not recognize the default Qubes EFI configuration. As such, it will have to be manually edited to be bootable (this will need to be done after every kernel and Xen update.)

Copy /boot/efi/EFI/qubes/ to /boot/efi/EFI/BOOT/.
Rename /boot/efi/EFI/BOOT/xen.efi to /boot/efi/EFI/BOOT/BOOTX64.efi.
Rename /boot/efi/EFI/BOOT/xen.cfg to /boot/efi/EFI/BOOT/BOOTX64.cfg.

Installation finished but "Qubes" boot option is missing and xen.cfg is empty

In some cases installer fails to finish EFI setup and leave the system without Qubes-specific EFI configuration. In such a case you need to finish those parts manually. You can do that just after installation (switch to tty2 with Ctrl-Alt-F2), or booting from installation media in "Rescue a Qubes system" mode.

Examine /boot/efi/EFI/qubes (if using Qubes installation media, it's in /mnt/sysimage/boot/efi/EFI/qubes). You should see 4 files there:
- xen.cfg (empty, size 0)
- xen-(xen-version).efi
- vmlinuz-(kernel-version)
- initramfs-(kernel-version).img

Copy xen-(xen-version).efi to xen.efi:

cd /mnt/sysimage/boot/efi/EFI/qubes
cp xen-*.efi xen.efi

Create xen.cfg with this content (adjust kernel version, and filesystem locations, below values are based on default installation of Qubes 3.2):

[global]
default=4.4.14-11.pvops.qubes.x86_64

[4.4.14-11.pvops.qubes.x86_64]
options=loglvl=all dom0_mem=min:1024M dom0_mem=max:4096M
kernel=vmlinuz-4.4.14-11.pvops.qubes.x86_64 root=/dev/mapper/qubes_dom0-root rd.lvm.lv=qubes_dom0/root rd.lvm.lv=qubes_dom0/swap i915.preliminary_hw_support=1 rhgb quiet
ramdisk=initramfs-4.4.14-11.pvops.qubes.x86_64.img

Create boot entry in EFI firmware (replace /dev/sda with your disk name and -p 1 with /boot/efi partition number):
```
efibootmgr -v -c -u -L Qubes -l /EFI/qubes/xen.efi -d /dev/sda -p 1 "placeholder /mapbs /noexitboot"
```

Installation freezes before getting to Anaconda

On some early, buggy UEFI implementations, you may need to disable EFI under Qubes completely. This can sometimes be done by switching to legacy mode in your BIOS/UEFI configuration. If that's not an option there, or legacy mode does not work either, you can try the following to add efi=no-rs.

Follow the steps above to edit the [qubes-verbose] section of your installer's xen.cfg. You want to modify the efi=attr=uc setting and comment out the mapbs and noexitboot lines. The end result should look like this:
```
[qubes-verbose]
options=console=vga efi=no-rs
# noexitboot=1
# mapbs=1
kernel=vmlinuz inst.stage2=hd:LABEL=Qubes-R4.0-x86_64 i915.alpha_support=1
ramdisk=initrd.img
```
Boot the installer and continue to install as normal, until towards the end when you will receive a warning about being unable to create the EFI boot entry. Click continue, but don't reboot the system at the end when prompted.
Go to tty2 (Ctrl-Alt-F2).
Use your preferred text editor (nano works) to edit /mnt/sysimage/boot/efi/EFI/qubes/xen.cfg, adding the efi=no-rs option to the end of the options= line. For example:
```
[4.14.18-1.pvops.qubes.x86_64]
options=loglvl=all dom0_mem=min:1024M dom0_mem=max:4096M iommu=no-igfx ucode=scan efi=no-rs
```

Execute the following commands:

cp -R /mnt/sysimage/boot/efi/EFI/qubes /mnt/sysimage/boot/efi/EFI/BOOT
mv /mnt/sysimage/boot/efi/EFI/BOOT/xen.efi /mnt/sysimage/boot/efi/EFI/BOOT/BOOTX64.efi
mv /mnt/sysimage/boot/efi/EFI/BOOT/xen.cfg /mnt/sysimage/boot/efi/EFI/BOOT/BOOTX64.cfg

Go back to tty6 (Ctrl-Alt-F6) and click Reboot.
Continue with setting up default templates and logging in to Qubes.

Whenever there is a kernel or Xen update for Qubes, you will need to follow these other steps above because your system is using the fallback UEFI bootloader in [...]/EFI/BOOT instead of directly booting to the Qubes entry under [...]/EFI/qubes.

9.0 KiB Raw Blame History

Troubleshooting UEFI related problems