Crashkernel changes

crashkernel/crash-recovery.sh

@@ -0,0 +1,77 @@
+# We are called from initramfs:/init ( /project/admin/initramfs/init )
+#
+# Execution environment:
+#   kernel is the crashkernel
+#   we are pid 1
+#   original root is mounted at /
+#   root might be mounted "ro" (if we have that in the kernel command line of the crash kernel)
+#   /dev is mounted (prepopulated by devtmpfs)
+#   we are executed by bash
+
+# graphics console is not yet working, because the replaced kernel
+# most probably used a framebuffer driver which reinitialized the graphics adapter
+# so it no longer runs in a mode initialized by bios/grub.
+# These bios-settings can not be restored easily, because this would require
+# real mode(true?), which kexec doesn't go through.
+#
+# We can start the framebuffer driver and get a working console (fbcon) if
+# we just load the right driver for the graphics adapter.
+#
+# However, nouveau currently requires to much memory (because of debugging data) and
+# would just create a out of memory kernel panic.
+#
+#modprobe bochs_drm   # DRM Support for bochs dispi vga interface (qemu stdvga)
+#modprobe nouveau     # nvidia : requires 512MB memory....
+#
+# We can't use /dev/kmsg (kernel ring buffer), because it only accepts a very limited
+# number of messages per time slot....
+#
+#exec >/dev/kmsg 2>&1
+#
+# So for now just use our serial console /dev/ttyS1 which hopefully is connected to another system.
+#
+exec >/dev/ttyS1 2>&1
+
+reboot() {
+	test "$1" && echo "$S $1"
+	echo "+++++++++++++++++++++++++++++++++++++ reboot in 10s"
+	sleep 10
+	/sbin/reboot -f
+}
+
+free_space() {
+	local path="$1"
+	echo $(( $(stat -f -c %a "$path") *  $(stat -f -c %S "$path") ))
+}
+
+set -x
+echo "+++++++++++++++++++++++++++++++++++++ crash recovery"
+
+mount -n -t proc     proc     /proc
+mount -n -t sysfs    sysfs    /sys
+
+test -e /proc/vmcore || reboot "no crashdump"
+core_size=$(stat -c %s /proc/vmcore)
+
+mount -o remount,rw / /
+
+if mount -L CRASHDUMP /mnt; then
+	test -e /mnt/crash.vmcore && rm /mnt/crash.vmcore
+	if (( $(free_space /mnt) - $core_size > 1024*1024*1024 )); then
+		echo "+++++++++++++++++++++++++++++++++++++ saving crash dump to CRASHDUMP volume"
+		cp /proc/vmcore /mnt/crash.vmcore
+		umount /mnt
+	else
+		echo "+++++++++++++++++++++++++++++++++++++ not enough free space on CRASHDUMP volume"
+	fi
+	reboot
+fi;
+
+test -e /var/crash.vmcore && rm /var/crash.vmcore
+if (( $(free_space /var) - $core_size > 1024*1024*1024 )); then
+	echo "+++++++++++++++++++++++++++++++++++++ saving crash dump to /var"
+	cp /proc/vmcore /var/crash.vmcore
+else
+	echo "+++++++++++++++++++++++++++++++++++++ not enough free space in /var"
+fi
+reboot

misc_systemd_units/crashkernel.service → crashkernel/crashkernel.service

@@ -6,7 +6,7 @@ ConditionPathExists=/boot/bzImage.crash
 StandardOutput=kmsg
 StandardError=kmsg
 Type=oneshot
-ExecStart=/usr/sbin/kexec -p /boot/bzImage.crash --initrd=/boot/grub/initramfs.igz --command-line="root=LABEL=root ro console=ttyS1,115200n8 console=tty0 irqpoll maxcpus=1 reset_devices CRASH"
+ExecStart=/usr/sbin/kexec -p /boot/bzImage.crash --initrd=/boot/grub/initramfs.igz --command-line="root=LABEL=root ro console=ttyS1,115200n8 console=tty0 irqpoll maxcpus=1 reset_devices panic=5 CRASH"
 ExecStop=/usr/sbin/kexec -p -u
 RemainAfterExit=yes
 

install.sh

@@ -149,5 +149,6 @@ install_data misc_systemd_units/mxvlan.service            "$DESTDIR$systemdunitd
 mkdir -p "$DESTDIR$usr_prefix/share/applications"
 install_data applications-defaults/defaults.list          "$DESTDIR$usr_prefix/share/applications/defaults.list"
 install_data misc_etc_files/modprobe.d/nfs_slotlimit.conf "$DESTDIR$sysconfdir/modprobe.d/nfs_slotlimit.conf"
-install_data misc_systemd_units/crashkernel.service       "$DESTDIR$systemdunitdir/crashkernel.service"
+install_data crashkernel/crashkernel.service              "$DESTDIR$systemdunitdir/crashkernel.service"
+install_exec crashkernel/crash-recovery.sh                "$DESTDIR$root_sbindir/crash-recovery.sh"
 exit

mxgrub/mxgrub

@@ -239,7 +239,7 @@ sub scan_mariux {
 	}
 }
 
-our $KERNEL_PARAMETER="ro crashkernel=256M console=ttyS0,115200n8 console=ttyS1,115200n8 console=tty0 init=/bin/systemd audit=0";
+our $KERNEL_PARAMETER="ro crashkernel=512M console=ttyS1,115200n8 console=tty0 init=/bin/systemd audit=0";
 
 sub update_grub_cfg {
 	my $kernellist='';
@@ -248,7 +248,7 @@ sub update_grub_cfg {
 
 	for my $label (@MARIUX) {
 		my $image=label_to_image($label);
-		$kernellist.="\tmenuentry \"$label\" --unrestricted { save_env chosen ; linux /boot/$image crashkernel=256M root=LABEL=root ro console=ttyS1,115200n8 console=tty0 init=/bin/systemd audit=0 ; initrd /boot/grub/initramfs.igz }\n";
+		$kernellist.="\tmenuentry \"$label\" --unrestricted { save_env chosen ; linux /boot/$image root=LABEL=root $KERNEL_PARAMETER ; initrd /boot/grub/initramfs.igz }\n";
 	}
 
 	my $GRUB_CFG_NEW=<<"EOF";