Kexec

From ArchWiki
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

Kexec is a system call that enables you to load and boot into another kernel from the currently running kernel. This is useful for kernel developers or other people who need to reboot very quickly without waiting for the whole BIOS boot process to finish. Note that kexec may not work correctly for you due to devices not fully re-initializing when using this method, however this is rarely the case.

Installation

Install the kexec-tools package.

Rebooting using kexec

Manually

You can manually invoke kexec using:

# kexec -l /boot/vmlinuz-linux --initrd=/boot/initramfs-linux.img --reuse-cmdline
# kexec -e
Warning: Running kexec -e directly will not unmount active filesystems or terminate any running services gracefully.

It is also possible to load kernel manually and then let systemd handle service shutdown and kexec for you.

# kexec -l /boot/vmlinuz-linux --initrd=/boot/initramfs-linux.img --reuse-cmdline
# systemctl kexec

Systemd

By default, if systemd-boot is used and no kernel was loaded manually using kexec -l before, systemd will load the kernel specified in the default boot loader entry. For example, to reboot into the newer kernel after a system update, you may simply run:

# systemctl kexec

The command will refuse to execute if you have several initrd entries (e.g. for Microcode updates) which are currently not supported.

Custom unit file

If the default behavior does not work for you or you desire to conveniently load custom kernels, you may wrap the kernel loading into a service unit. Create a new unit file, kexec-load@.service, that will load the specified kernel to be kexec'ed:

/etc/systemd/system/kexec-load@.service
[Unit]
Description=load %i kernel into the current kernel
Documentation=man:kexec(8)
DefaultDependencies=no
Before=shutdown.target umount.target final.target

[Service]
Type=oneshot
ExecStart=/usr/bin/kexec -l /boot/vmlinuz-%i --initrd=/boot/initramfs-%i.img --reuse-cmdline

[Install]
WantedBy=kexec.target

Then enable the service file for the kernel you want to load, for example simply the default kernel linux:

# systemctl enable kexec-load@linux
ln -s '/etc/systemd/system/kexec-load@.service' '/etc/systemd/system/kexec.target.wants/kexec-load@linux.service'

Ensure that the shutdown hook is not part of your initramfs image by removing it from the HOOKS array in /etc/mkinitcpio.conf. If it is, remove it and rebuild your initrd image with mkinitcpio -p linux.

Then to kexec

# systemctl kexec

If you wish to load a different kernel for the next kexec, for example linux-lts, disable the service for the current kernel and enable the one for the new kernel:

# systemctl disable kexec-load@linux
# systemctl enable kexec-load@linux-lts

Separate /boot partition

The above systemd unit file will fail if /boot is not on the root file system, as systemd will likely unmount /boot before it runs the kexec-load unit file. An alternative approach is to load a "hook" unit file that does nothing on startup and invokes kexec upon termination. By making this unit file conflict with kexec.target and only kexec.target, you can ensure the new kernel gets loaded early enough and only after a "systemctl kexec" command. Here is an alternate /etc/systemd/system/kexec-load@.service file that follows this strategy:

[Unit]
Description=hook to load vmlinuz-%i kernel upon kexec
Documentation=man:kexec(8)
DefaultDependencies=no
Requires=sysinit.target
After=sysinit.target

[Service]
Type=oneshot
ExecStart=-/usr/bin/true
RemainAfterExit=yes
ExecStop=/usr/bin/kexec -l /boot/vmlinuz-%i --initrd=/boot/initramfs-%i.img --reuse-cmdline

[Install]
WantedBy=basic.target

Note that Conflicts=shutdown.target is not really needed, as it is implicitly guaranteed by strict ordering on sysinit.target which itself Conflicts= with shutdown.target.

Troubleshooting

System hangs or reboots after "kexec_core: Starting new kernel"

The troubleshooting information on General troubleshooting#Boot problems may be helpful for diagnosing the problem.

In some cases a hanging system might be an acpi related problem which can be checked on-the-fly like this:

# cmdline=$(cat /proc/cmdline)
# cmdline="$cmdline acpi_rsdp=$(grep -m1 ^ACPI /sys/firmware/efi/systab | cut -f2- -d=)"
# echo $cmdline
# ls -al /boot/
# kexec -l /boot/vmlinuz-linux-lts --initrd=/boot/initramfs-linux-lts.img --append="$cmdline"
# systemctl kexec

Please adapt the name of the initramfs image and the kernel according to your output of ls -al /boot/.

Adding the acpi_rsdp kernel parameter to the kexec command line has been suggested in [1] and may solve the issue in some cases without the need to completely disable ACPI via acpi=off.

No kernel mode-setting (Nvidia)

The graphics driver needs to be unloaded before a kexec, or the next kernel will not be able to gain exclusive control of the device. This is difficult to achieve manually because any programs which need exclusive control over the GPU (Xorg, display managers) must not be running. Below is an example systemd service that will unload the KMS driver right before kexec, which requires that you use systemctl kexec.

# systemctl edit --full --force unmodeset.service
[Unit]
Description=Unload nvidia modesetting modules from kernel
Documentation=man:modprobe(8)
DefaultDependencies=no
After=umount.target
Before=kexec.target

[Service]
Type=oneshot
ExecStart=modprobe -r nvidia_drm

[Install]
WantedBy=kexec.target

Afterwords, enable unmodeset.service

See also