Comment 11 for bug 1877088

Revision history for this message
bugproxy (bugproxy) wrote : Comment bridged from LTC Bugzilla

------- Comment From <email address hidden> 2020-06-02 10:15 EDT-------
"make install" also triggers initramfs creation as does manual
invocation of the installkernel script. In fact both even trigger
boot loader update/install.

This is the output:
$ sudo make install
sh -x ./arch/s390/boot/install.sh 5.4.27 arch/s390/boot/bzImage \
System.map "/boot"
+ [ -x /root/bin/installkernel ]
+ [ -x /sbin/installkernel ]
+ exec /sbin/installkernel 5.4.27 arch/s390/boot/bzImage System.map /boot
run-parts: executing /etc/kernel/postinst.d/apt-auto-removal 5.4.27 /boot/vmlinuz-5.4.27
run-parts: executing /etc/kernel/postinst.d/initramfs-tools 5.4.27 /boot/vmlinuz-5.4.27
update-initramfs: Generating /boot/initrd.img-5.4.27
Using config file '/etc/zipl.conf'
Run /lib/s390-tools/zipl_helper.device-mapper /boot
Building bootmap in '/boot'
Building menu 'menu'
Adding #1: IPL section 'ubuntu' (default)
Adding #2: IPL section 'old'
Preparing boot device: dm-0.
Done.
run-parts: executing /etc/kernel/postinst.d/unattended-upgrades 5.4.27 /boot/vmlinuz-5.4.27
run-parts: executing /etc/kernel/postinst.d/update-notifier 5.4.27 /boot/vmlinuz-5.4.27
run-parts: executing /etc/kernel/postinst.d/zz-zipl 5.4.27 /boot/vmlinuz-5.4.27
Using config file '/etc/zipl.conf'
Run /lib/s390-tools/zipl_helper.device-mapper /boot
Building bootmap in '/boot'
Building menu 'menu'
Adding #1: IPL section 'ubuntu' (default)
Adding #2: IPL section 'old'
Preparing boot device: dm-0.
Done.

I did already have /etc/kernel-img.conf
with the following content:

# Kernel image management overrides
# See kernel-img.conf(5) for details
do_symlinks = yes
do_bootloader = yes
do_initrd = yes
link_in_boot = yes

Could this depend on the legacy installer?
Our automatically installed systems are currently all installed with it
and I also used that for this test LPAR (that has PCI Physical Functions)
since the legacy installer allowed a much nicer
manual network configuration using the HMC console.
Sadly that's also why I can't currently test on another fresh install
as our automatic installation system is currently being changed
to support the new installer and I couldn't get the previous workaround
to work today.

I'm not knowledgeable enough to give a good opinion on initrd less
boot but I would suspect that the deployments at our customers are much
more varied than the average cloud so I would be hesitant about defaulting
to it. Also in my experience s390 boot times are already quite good.

Afaik with our defconfig a lot of us developers regularly boot
without initrd or even any modules, so supporting it shouldn't be
a problem. As the defconfig doesn't include the
drivers for PCI devices into the Kernel I usually use both modules
and an initrd, which might be why I stumbled over this first.