Package: dpkg Version: 1.15.5.6 I have installed some hook scripts in /etc/kernel/postinst.d and /etc/kernel/postrm.d to improve the installation and removal process for kernel image packages. I initially did this for custom kernel image packages created by the make-kpkg command of the kernel-package package, but these scripts are executed when installing and removing stock kernel image packages as well. There are currently three scripts in /etc/kernel/postrm.d. One is initramfs-tools, provided by the package of the same name, for handling kernel image packages created by "make deb-pkg". It exits without doing anything when stock kernel image packages or kernel image packages created by make-kpkg are being removed/purged, since it gets two arguments instead of the expected one. The second is initramfs, which is essentially the same as /usr/share/kernel-package/examples/etc/kernel/postrm.d/initramfs from the kernel-package package. The third is symlink_hook, which is a modified version of /usr/share/kernel-package/examples/etc/kernel/postinst.d/symlink_hook from the kernel-image package. It's purpose is to re-do the symlinks and re-run the boot loader (zipl in my case, s390 architecture) when a kernel image package is purged. The three scripts are executed in alphabetical order: initramfs, initramfs-tools, symlink_hook. Thus, symlink_hook is executed third and last. The script appears to work correctly and to exit with a return status of 0. However, I still get an error from dpkg when I purge the package. For example, in response to
dpkg --purge linux-image-2.6.32-trunk-s390x I get ... run-parts: executing /etc/kernel/postrm.d/symlink_hook 2.6.32-trunk-s390x /boot/vmlinuz-2.6.32-trunk-s390x dpkg: error processing linux-image-2.6.32-trunk-s390x (--purge): subprocess installed post-removal script returned error exit status 128 Errors were encountered while processing: linux-image-2.6.32-trunk-s390x I am running a pure "testing" system (Squeeze). There are no packages from "unstable" (Sid). I am running another (custom) kernel at the time. (I am not trying to purge a running kernel!) An exit status of 128 is particularly puzzling. An exit status is usually between 0 and 127, inclusive. According to the man page for bash, "The return value of a simple command is its exit status, or 128+n if the command is terminated by signal n." Thus, this would indicate a process terminated by "kill -n 0". Huh? What? I have searched the internet on the string "subprocess installed post-removal script returned error exit status 128". I found three other bug reports with this search: 560181, 534648, and 541275. All these bug reports were against other packages (wwwoffle, wicd, and ca-certificates, respectively). However, in each case the respective package had a postrm script that was returning a non-zero exit status. In my case, I am certain that my script is returning a zero exit status. Therefore, I am filing this bug report against dpkg itself. Here is my script: ---------- #!/bin/sh - # -*- Mode: Sh -*- # example.posthook.sh --- # Author : Manoj Srivastava ( [email protected] ) # Created On : Mon Aug 11 14:00:58 2003 # Created On Node : glaurung.green-gryphon.com # Last Modified By : Manoj Srivastava # Last Modified On : Sun Apr 12 17:16:52 2009 # Last Machine Used: anzu.internal.golden-gryphon.com # Update Count : 10 # Status : Unknown, Use with caution! # HISTORY : # Description : # # # This is an example of a script that can be run as a postrm hook, # and manages the symbolic links in a manner similar to the kernel # image default behaviour, except that the latest two versions (as # determined by ls -lct) are kept. You can modify this script # # Copyright 2003, 2004, 2005, 2006, 2007, 2008, 2009 Manoj Srivastava # Copyright 2009 Darren Salt set -e set -x # The dir where symlinks are managed SYMLINKDIR=/boot if [ $# -ne 2 ];then exit 0;fi version="$1" vmlinuz_location="$2" vmlinuz_dir="$(dirname "$2")" cd $SYMLINKDIR || exit 1 if [ -n "$DEB_MAINT_PARAMS" ]; then eval set -- "$DEB_MAINT_PARAMS" case "$1" in remove) : echo nop ;; *) exit 0 ;; esac fi rm -f vmlinuz vmlinuz.old initrd.img initrd.img.old # Create a temporary file safely if [ -x /bin/tempfile ]; then outfile=$(tempfile -p outp -m 0600); else set -e mkdir /tmp/kernel-image-$version-$$ outfile=/tmp/kernel-image-$version-$$/output fi (cd "$vmlinuz_dir" && ls -ct vmlinuz-*) > $outfile STD="$(head -n 1 $outfile | sed 's/vmlinuz-//')" OLD="$(head -n 2 $outfile | tail -n 1 | sed 's/vmlinuz-//')" if [ "X$OLD" = "X$STD" ]; then OLD= fi if [ "X$STD" = "X" ]; then exit 0 fi # echo Booting $STD, old is $OLD if [ -f "$vmlinuz_dir/"initrd.img-$STD ] ; then ln -s initrd.img-$STD initrd.img ln -s vmlinuz-$STD vmlinuz else ln -s vmlinuz-$STD vmlinuz fi if [ "X$OLD" != "X" ]; then if [ -f "$vmlinuz_dir/"initrd.img-$OLD ] ; then ln -s initrd.img-$OLD initrd.img.old ln -s vmlinuz-$OLD vmlinuz.old else ln -s vmlinuz-$OLD vmlinuz.old fi fi zipl rm -f $outfile if [ -d /tmp/kernel-image-$version-$$ ]; then rmdir /tmp/kernel-image-$version-$$ fi exit 0 ---------- (Notice the "set -x", which I have added for debugging purposes.) And here is a console log of my attempt to purge the package, which has previously been removed successfully. ---------- debian3:~# dpkg --purge linux-image-2.6.32-trunk-s390x (Reading database ... 25193 files and directories currently installed.) Removing linux-image-2.6.32-trunk-s390x ... Purging configuration files for linux-image-2.6.32-trunk-s390x ... Examining /etc/kernel/postrm.d . run-parts: executing /etc/kernel/postrm.d/initramfs 2.6.32-trunk-s390x /boot/vmlinuz-2.6.32-trunk-s390x Cannot delete /boot/initrd.img-2.6.32-trunk-s390x, doesn't exist. run-parts: executing /etc/kernel/postrm.d/initramfs-tools 2.6.32-trunk-s390x /boot/vmlinuz-2.6.32-trunk-s390x run-parts: executing /etc/kernel/postrm.d/symlink_hook 2.6.32-trunk-s390x /boot/vmlinuz-2.6.32-trunk-s390x + SYMLINKDIR=/boot + [ 2 -ne 2 ] + version=2.6.32-trunk-s390x + vmlinuz_location=/boot/vmlinuz-2.6.32-trunk-s390x + dirname /boot/vmlinuz-2.6.32-trunk-s390x + vmlinuz_dir=/boot + cd /boot + [ -n ] + rm -f vmlinuz vmlinuz.old initrd.img initrd.img.old + [ -x /bin/tempfile ] + tempfile -p outp -m 0600 + outfile=/tmp/outpB5mG9b + cd /boot + ls -ct vmlinuz-2.6.32-customtrunk-s390x + sed s/vmlinuz-// + head -n 1 /tmp/outpB5mG9b + STD=2.6.32-customtrunk-s390x + sed s/vmlinuz-// + tail -n 1 + head -n 2 /tmp/outpB5mG9b + OLD=2.6.32-customtrunk-s390x + [ X2.6.32-customtrunk-s390x = X2.6.32-customtrunk-s390x ] + OLD= + [ X2.6.32-customtrunk-s390x = X ] + [ -f /boot/initrd.img-2.6.32-customtrunk-s390x ] + ln -s initrd.img-2.6.32-customtrunk-s390x initrd.img + ln -s vmlinuz-2.6.32-customtrunk-s390x vmlinuz + [ X != X ] + zipl + rm -f /tmp/outpB5mG9b + [ -d /tmp/kernel-image-2.6.32-trunk-s390x-1071 ] + exit 0 dpkg: error processing linux-image-2.6.32-trunk-s390x (--purge): subprocess installed post-removal script returned error exit status 128 Errors were encountered while processing: linux-image-2.6.32-trunk-s390x debian3:~# ---------- By trial and error, I have found that if I comment out the invocation of the boot loader, zipl, everything works fine. But if I leave it uncommented, it fails. zipl itself works. It exits with a return status of 0. But dpkg fails. Why? There is a similar symlink_hook script in /etc/kernel/postinst.d. It invokes zipl too, and it causes no problems. Also, all three of the scripts in /etc/kernel/postrm.d can be called twice for a purge: once for a remove, if the package has not previously been removed, and then again for a purge. It only fails on the purge call. The remove call works fine. -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
