Ubuntu mainline 7.1.x linux-image postinst trigger regression skips /etc/kernel/postinst.d

Robert Hardy rhardy at webcon.ca
Sat Jul 4 16:03:52 UTC 2026


Ubuntu mainline kernel packages 7.1.x appear to have a generated 
linux-image postinst regression.

Issue with reproducer and diff:
https://github.com/bkw777/mainline/issues/374

This is not a mainline installer GUI problem and not a DKMS/NVIDIA build 
failure. The generated linux-image-unsigned postinst script for 7.1.x 
changed the trigger generation logic.

Known-good 7.0.13 postinst writes one trigger file containing:

   run-parts ... /etc/kernel/postinst.d /usr/share/kernel/postinst.d

Broken 7.1.x postinst loops over both directories but writes to the same 
trigger file inside the loop:

   for d in /etc/kernel/postinst.d /usr/share/kernel/postinst.d; do
       cat - >/usr/lib/linux/triggers/$version <<EOF
       ...
       run-parts ... $d
EOF
       dpkg-trigger --no-await linux-update-$version
   done

Because /usr/share/kernel/postinst.d is processed after 
/etc/kernel/postinst.d, the second write overwrites the first trigger 
body. The dpkg trigger then runs only /usr/share/kernel/postinst.d and 
skips /etc/kernel/postinst.d entirely.

That skips the normal hooks that create initramfs and run 
DKMS/update-grub on affected systems. Result: kernel installs without 
errors but no initrd is generated until update-initramfs is run manually.

The packaging fix should be to collect the existing hook directories 
first and write exactly one trigger file containing one run-parts 
invocation over the full directory list, rather than writing 
/usr/lib/linux/triggers/$version repeatedly inside the loop.

Regards,
Rob

-- 
---------------------"Happiness is understanding."----------------------
Robert Hardy, B.Eng Computer Systems    Owner, IT and Security Architect
rhardy at webcon.ca           GPG Key available              (613) 276-7327
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20260704/9a7cc324/attachment.html>


More information about the kernel-team mailing list