Some notes on computer system administration

BIOS setting tips on Lenovo Thinkpads

Customize BIOS splash screen

The BIOS screen will show up with the default Lenovo logo if the Boot Mode is set to Quick (diagnose mode will disable the splash screen). For old models, you may find the ThinkWiki’s instructions under Windows OS or without access to Windows helpful. For up-to-2011 models, this instruction may be helpful as well. For later models, the process is similar, but the limitations on the image size and format are relaxed and the steps is simpler.

Take P50 for example, I search for “Lenovo BIOS Update Utility Thinkpad P50”. Here’s the download page for P50 I used: in the BIOS session, there is a Windows version and a bootable CD version of the BIOS Update Utility. Download the correct model’s BIOS Update Utility for Windows, in my case. Install it but don’t select to install the BIOS update yet. You may need to enable the Allow Rollback setting in the BIOS for the tool to install the splash image–even though you may be installing the same version of BIOS as your current one. Go to C:\Drivers\FLASH\<random string>\ where the driver is installed. Read BIOS_LOGO.TXT. The file should mention creating an image called LOGO.BMP, LOGO.JPG or LOGO.GIF.

Now I make my customized BMP splash image with a size of 768px X 432px using InkScape with the Thinkpad logo downloaded here and a few other resources downloaded from the ThinkWiki page. As stated from BIOS_LOGO.TXT that the image can be up to 40% size of the default screen resolution (1092x1080 for my P50). You may get random garbage on the splash screen if the image is too large in file size, which I don’t know the exact file size limit. To prevent errors happen, then I use MSPaint tool to open the exported LOGO.BMP file and save it as LOGO.GIF file (see below), which will reduce the file size dramatically. Notice that it may only work for images that have been edited or exported with MSPaint (not GIMP). My final version of the splash image is a little above 30KB and the resolution is the same as the BMP file. The image doesn’t have to be 16-bit color format for the earlier days.

My BIOS splash
My BIOS splash

Copy or save the LOGO.GIF image to C:\Drivers\Flash\<random string> folder.

Now run WUNUPTP.exe which is the actual BIOS Update Utility and should be in the same folder as the image. Select “Update ThinkPad BIOS” and hit next. A dialogue box should show “A custom start up image file was found. Do you apply the custom startup image file to the system?” Click Yes.

Now keep hitting next/yes (but make sure you read these instructions) until seeing “System program updates is continued by BIOS at the next reboot.” Click OK then the computer will reboot and start to flash the BIOS. If you exit out at this point the flash will still happen, but not until you restart.

Make sure you do not touch your power button at all during the reboot/flashing phase since loss of power during a firmware flash can brick your laptop. You computer will reboot afterwards, and by the end you should have a custom BIOS splash image at the beginning of the system startup.

Next time you have to flash the BIOS, the BIOS updater will detect a custom boot splash and ask you if you want to preserve it or restore the original.

Linux/Ubuntu OS

Diagnose display problems and install the default NVidia driver

The video card driver and display settings are usually tricky on Linux. I have encountered black-screen and dual-monitor display problems while installing and reinstalling Ubuntu 16.04 GNOME and Unity systems on my Lenovo P50 mobile workstation and other computers. Fortunately, I have received helps on the ubuntu forum and other places from experts and fixed those past issues.

If a black screen happened in the startup after installation, and in the TTY mode (Ctl+Alt+F1) it shows “a start job is for hold until boot finishes up…” but it never finishes, it might be related to the display settings and drivers. The following worked for me. Boot into the recovery mode with Network or run commands as root from the Advanced Ubuntu Options, and run

sudo apt-get remove plymouth
sudo apt-get remove xserver-xorg-video-intel

and restart with sudo reboot. Certainly, no need to use sudo when running as root. If it works, then reinstall xserver-xorg-video-intel. If problem persists, try sudo purge nvidia* and reboot and reinstall the default NVidia driver by sudo ubuntu-drivers autoinstall. Pay attention to the build message to see if there is any error in the process of building the driver into the Linux kernel modules. Then reconfigure the lightdm by

sudo apt-get --reinstall install gdm
sudo dpkg-reconfigure lightdm
sudo service lightdm start

It should now start a graphic window, or reboot to see the effect.

For external monitor problems, it could be related to the xserver-xorg setting and NVidia drivers. To diagnose the issue, here are some useful commands: cat .xsession-errors in the user’s home directory to see if there is any error like openConnection: connect: No such file or directory. That means ls /etc/X11/ will not show a file named as xorg.conf. This usually implies the NVidia driver and the xorg files are not correctly built. Run dpkg -l | grep linux- to see all the kernel installed on the system, and then dpkg -l | grep -i nvidia and dkms status to see what version of NVidia driver has been installed. If lsmod | grep nvidia gives nothing, if means the NVidia driver is not built into the current kernel (use uname -a to find the current kernel series number). So, the solution is to rebuild the NVidia driver using default settings by

sudo apt purge nvidia*
sudo rm /etc/X11/xorg.conf # Run only when xorg.conf is there.
sudo find / -type f 2> /dev/null | grep -i nvidia | grep -i \\.ko # If this gives any output, delete those files one by one to make sure there won't be any version mismatch problem after the installation if the driver to be installed is a different version from any previous ones.
sudo ubuntu-drivers autoinstall

The building process will be configuring all installed Linux kernels. Try to see if there is any error message for building it for each kernels. If a message shows the current kernel doesn’t support the NVidia driver, a working kernel has to be used to boot up the computer and the non-compatible kernel shouldn’t be used for the compatibility reason. After a successful rebuilding process, in the terminal run sudo service lightdm restart, and it should bring in the graphic window or try to reboot to see the effect.

For other issues, dpkg -l nvidia-prime and cat /var/log/Xorg.0.log should show some basic message to help identify where the problem is related to the video drivers. To confirm problem solved at the root, run

lspci -k | grep -EA2 'VGA|3D'
sudo lshw -C display

you should find both Intel integrated driver and the concrete NVidia display drivers are installed to the corresponding modules. The nvidia-PRIME package should also be able to run and configure how the GPUs are used on the computer.

PS: after Ubuntu 16.10, GNOME has developed better display management systems (like gnome-wayland-session) and the configuration and debugging process might be different from the above.

Ubuntu 16.04 hangs to shut down

This problem occurs when I freshly installed the Ubuntu 16.04 Gnome system. Whenever I press the Shutdown button on the Power Off menu, it will take up to 1min 30sec to respond.

Turns out, it is related to the CUPS remote printers from the cups-browsed service which automatically add network printers to the computer. I don’t need this automatic printer adding function and hence disabled this service, and then everything works fine again. Reference is here.

In most cases, a command will save the time regardless of the hanging-out issue. sudo reboot is enough to reboot the computer as soon as possible. To shutdown (not halt), here are the commands available:

sudo shutdown -h now
sudo shutdown -P now
sudo poweroff
sudo halt -p
sudo init 0

The poweroff and halt commands basically invoke shutdown (except for the poweroff -f). sudo poweroff and sudo halt -p are exactly like sudo shutdown -P now. The command sudo init 0 will take you to the runlevel 0 (shutdown).

Now what if we want to shut down forcefully, i.e., we don’t want to wait for processes to close normally? In that case you can use:

sudo poweroff -f

This will not use shutdown. Rather, it will invoke the reboot(2) system call (used for reboot, poweroff & halt) to power off the computer instantly.

Update: Without disabling the CUPS service, this bug seems having been fixed with cups-filters v1.11.4-1 yet not released in the official Ubuntu 16.04 repository. A workaround solution to install the latest version of cups-filters and its dependencies can be found in this solution.

Automatically mount partitions at startup

The graphic approach is suggested here–that is the follow. Into Disks software, select the partition that you want to mount at startup. Click its Setting, select Edit Mount Options, and then select Mount at Startup and fill in with corresponding mounting information.

If preferring to modify the configuration file, /etc/fstab is the place to play with. Filling in the mounting partition information following the instructions here, here or here. For example, one line can be

UUID=BC34947B34943A7A /media/D ntfs-3g defaults,x-gvfs-name=D 0 0

The UUID information can be found by running sudo blkid on a terminal.

Unable to mount NTFS disks automatically on startup.

When the Ubuntu 16.04 started, I got the following error message:

Unable to access “My Drive”

Error mounting /dev/sdb4 at /media/D Center: Command-line `mount -t "ntfs" -o "uhelper=udisks2,nodev,nosuid,uid=1000,gid=1000,dmask=0077,fmask=0177" "/dev/sdb4" "/media/D Center"' exited with non-zero exit status 14: The disk contains an unclean file system (0, 0).
Metadata kept in Windows cache, refused to mount.
Failed to mount '/dev/sdb4': Operation not permitted
The NTFS partition is in an unsafe state. Please resume and shutdown
Windows fully (no hibernation or fast restarting), or mount the volume
read-only with the 'ro' mount option

A solution according to this source is to run the following command, for example,

sudo ntfsfix /dev/sdb4

The root cause of this issue might be related to the fast-booting feature of Windows 10. Therefore, to fix it for all future events, the fast-booting feature should be turned off as instructed here. A detailed explanation on this issue can be found here.

Keep Linux kernel updated on a LTS Ubuntu OS

Normally, a LTS Ubuntu OS will stick to a particular kernel series to keep the OS stable. However, this doesn’t mean the Ubuntu OS doesn’t update their supported Linux kernel series on their repository. In fact, for example, Ubuntu-16.04-1 was released with kernel 4.4.X, but then Ubuntu-16.04-2 was released with kernel 4.8.X. The default behavior of a local Ubuntu-16.04-0 distribution is to stick to the initially installed kernel series before the OS is upgraded to a new distribution like Ubuntu 18.04 LTS. To change this default behavior and keep receiving updated Ubuntu’s officially supported kernel series on a LTS Ubuntu distribution, one can enable the HWE stacks by installing the following packages:

sudo apt-get install --install-recommends linux-generic-hwe-16.04 xserver-xorg-hwe-16.04

according to the official rolling LTS enablement stack page for Ubuntu 16.04 LTS distribution. The kernel update will keep rolling in the first two years after the distribution was initally released.

System files one can safely remove to save space

some files in /var

Old and big files in /var/log, /var/crash, /var/core, /tmp and /var/tmp can usually be removed safely. For an average user, the log files in /var/log ended with .gz or .old can be removed safely.

The crash files in /var/crash can be deleted safely if the dump information is no longer needed. Similar to the crash core files in /var/core or other places depending on programs (see Java, for example). One can delete core files by find . -name core -exec rm {} \; as a super admin.

For /tmp and /var/tmp files, it might be helpful to use the tmpreaper program to automatically delete old files by customizing the configure file of /etc/tmpreaper.conf. In my current configuration file, I defined TMPREAPER_TIME=30d and TMPREAPER_DIRS='/tmp/. /var/tmp/.' to automatically clean up files in both /tmp and /var/tmp directories older than 30 days.

Remove old Linux kernels to save space in /boot

One common problem of upgrading linux kernel is “no enough space” in the /boot partition. This might be due to the old kernels and the new kernel files copied to /boot while compiling. One line of command to remove those unnecessary files is

dpkg -l linux-{image,headers}-"[0-9]*" | awk '/^ii/{ print $2}' | grep -v -e `uname -r | cut -f1,2 -d"-"` | grep -e '[0-9]' | xargs sudo apt-get -y purge

Reference: the solution on Askubuntu.

Remove old backups by Deja Dup

I use Deja Dup to automatically backup my Ubuntu OS and home directory files incrementally. Sometimes, the backup folder might take a lot of space and some old backups can be removed safely.

  1. For deleting all but the last full backup, open a terminal and run this:

    duplicity remove-all-but-n-full 1 file:///media/I/Ubuntu --force
  2. Delete files from failed backup sessions:

    duplicity cleanup file:///media/I/Ubuntu --force

Mouse scrolls too fast on Chrome browser

If one scroll of the mouse on the Chrome brower can go a far distance on a page, it might have been affected by this Ubuntu bug. This only affects Chrome and a few wireless mouse brands including Microsoft. A workaround before the bug is fixed is to unplug and replug the wireless receiver of the mouse.

Make Powertop and TLP work together to save battery

I made the following configuration on my Lenovo Thinkpad P50 with Ubuntu 16.04. First, install powertop:

sudo apt install powertop

and let powertop analyze the system (don’t worry with black screens)

sudo powertop --calibrate

Make the systemd service : gksu gedit /etc/systemd/system/powertop.service paste in:

Description=Powertop tunings

ExecStart=/bin/bash /home/qxd/


More details on customizing a systemctl service can be found here. Note that my username on my computer is qxd, which you may want to adapt to your own case. Of course, I have created a file at /home/qxd/ with the following content:

# Run Powertop auto-tune.
/usr/sbin/powertop --auto-tune
# Disable USB auto-suspend for my mouse and wireless keyboard on startup.
# 1-1.1 1-1.2 1-2.1 1-2.2 1-2.4 2-1 2-1.4 2-2 2-2.4 ports are for my USB3.0 dock station.
declare -a usbs=("1-1.1" "1-1.2" "1-2" "1-2.1" "1-2.2" "1-2.4" "1-5" "1-8" "1-10" 2-1" "2-1.4" "2-2" "2-2.4" "usb1" "usb2")
sleep 5;
for i in "${usbs[@]}"
    if [ -f "$usb" ]; then
        echo 'on' > $usb;

Notice that, for the part to disable USB auto-suspension, the USB device names are found on powertop (see reference). For example, once powertop is open (sudo powertop), you can Tab to the control options page and hit ENTER to turn on or off some options like “Autosuspension of USB 1-12” and there will be a line immediately on the top of the powertop window showing the equivalent command that powertop has just committed, like echo 'on' > '/sys/bus/usb/devices/1-12/power/control'. Without the part to disable auto-suspension, the wireless devices plugged into the USB ports may really suspend after 2 sec of inactivity. If this really happens without using the full script of systemctl, you can run the echo command as root (sudo -i). For your specific case, you can modify the bash script to disable or enable power saving options in a similar way.

Now, save and enable powertop options at start:

sudo systemctl enable powertop

Reload the systemctl service since the content of powertop.service has been changed, use sudo systemctl daemon-reload on a terminal and then sudo systemctl restart powertop.service to restart the script (or replace restart with start for a first time run). To see the journal log of this powertop service, you can run journalctl -u powertop.service.
Now, install TLP:

sudo apt install tlp

TLP make nearly the same as powertop, maybe I should delete all things ever managed by powertop. But I decided to have the default lines there and only changed a few lines of parameters: My file /etc/default/tlp has the following lines for a minimum setting:

# tlp - Parameters for power save

# Hint: some features are disabled by default, remove the leading # to enable
# them.

# Set to 0 to disable, 1 to enable TLP.

# Dirty page values (timeouts in secs).

# Battery charge thresholds (ThinkPad only, tp-smapi or acpi-call kernel module
# required). Charging starts when the remaining capacity falls below the
# START_CHARGE_TRESH value and stops when exceeding the STOP_CHARGE_TRESH value.
# Main / Internal battery (values in %)
# Ultrabay / Slice / Replaceable battery (values in %)

Change MAX_LOST_WORK_SECS_ON_BAT at 15 (powertop said for VM writeback) than 60. 60 make no effect. Then enable services:

systemctl enable tlp.service  
systemctl enable tlp-sleep.service

And reboot.

A dock station can be useful if one wants to collect USB, display and audio ports in one connected device to the computer. I use the Lenovo Thinkpad USB 3.0 Docking Station as the port collector. Despite it is an old-fashion docking station, there have been stable Linux supports to the device by DisplayLink. Since I am using relative new linux kernels, I have to customize some settings to make the driver work properly.

To install the latest DisplayLink driver for the device on recent kernels, please follow the instructions on the displaylink-debian repo. After that, I added two aliases to the ~/.bashrc file:

# Define alias for the Dock station display layout.
# two
alias two="xrandr --setprovideroutputsource 1 0 && xrandr --setprovideroutputsource 2 0 && xrandr --output DVI-I-2-1 --auto --left-of eDP-1-1"

# one
alias one="xrandr --output DVI-I-2-1 --off --output eDP-1-1 --primary --pos 0x0 --rotate normal"

On the terminal, run source ~/.bashrc to update the commands. Then reboot with an external monitor connected. On the terminal, I type two to switch to two-monitor mode with the external monitor on the left-hand-side of my laptop; type one to switch back to the default laptop mode.

To make sure DisplayLink is running, use

systemctl status dlm.service

If it’s not running, start it with

sudo systemctl start dlm.service

To start DisplayLink automatically at boot, run

sudo systemctl enable dlm.service

You only need to enable the service once after installing the driver.

Note that, in writing the aliases in ~/.bashrc, I have checked the display source providers via (need to reboot after installing the driver)

xrandr --listproviders

If you find more than one provider while at least one external monitor is connected, it means the driver and the DisplayLink service are working. Then use

xrandr --setprovideroutputsource 1 0
xrandr --setprovideroutputsource 2 0

to display with two monitors for the first time. You can manually change the display resolution and relative positions in your Display setting of your desktop environment. The names of displays/monitors (shown as DVI-I-2-1 and eDP-1-1 in my ~/.bashrc script) can found by running xrandr on a terminal. You may need to adapt my script with your case. More details can be found in the post-installation setting.

Once everything is working, after every boot, you just need to run two on a terminal to enable the second monitor, or run one to switch back to one monitor display setting. A problem with audio may have not yet been solved with the dock’s DisplayLink driver.

Using LVM and btrfs filesystem to partition disks

LVM is a good partition and logical volume management tool to organize disk space, especially for the cases that one Linux directory is spanned onto two physical disks or is to be extended to extra disks in the future. On the other hand, btrfs filesystem is very flexible and convenient to make snapshot backups and secure & safe to manage system files in Linux. Combining these two, we can have a very flexible and convenient way to run Linux. Below are some common scenarios that I have encountered to use LVM with the btrfs filesystem.

I. creating LVM managed volume groups and logical volumes for installing Ubuntu along side with Windows.

To set up the partitions, I freed out ~130GB space in my case under Windows 10 using disk managing tools. As another preparation step, I went into BIOS, and disabled Secure Boot (Startup) and enbled CSM on my Thinkpad P50 laptop. Then booted in an Ubuntu USB liveCD, I used Gparted to set up SWAP (4GB), /boot for 350MB, and the rest on SSD (/dev/sdb) as LVM2 file management system with btrfs filesystem formate. Then in the command line, I created an LVM Physics Volume from GParted btrfs partition space /dev/sdb10 and two logical volumes for / and /home under the volume group ubuntuvg which only includes /dev/sdb10:

sudo pvcreate /dev/sdb10
sudo vgcreate ubuntuvg /dev/sdb10
sudo lvcreate -n root -L 25g ubuntuvg
sudo lvcreate -n home -l 100%FREE ubuntuvg

That is 25GB btrfs for / and 120GB btrfs for /home, both of which are extendable in the future event of upgrading disks. They are addressed at /dev/mapper/ubuntuvg-root and /dev/mapper/ubuntuvg-home on the disk filesystem record (run df -h to see them once mounted). On the system partition table as recognized by mount, for example, they are recognized as /dev/ubuntuvg/root and /dev/ubuntuvg/home, respectively.

Then I installed the Ubuntu OS on those partitions, and installed the bootloader to the UEFI partition (/dev/sdb7 where the Windows 10 bootloader was installed before) in my case.

To permanently add those volumes to be automatically mounted at startup of the Linux system, one can edit the /etc/fstab file to include them. My file has the following partial content:

# <file system> <mount point>   <type>  <options>       <dump>  <pass>
/dev/mapper/ubuntuvg-root /               btrfs   defaults,[email protected] 0       1
# /boot was on /dev/sdb9 during installation
UUID=5cd44ab2-7e1f-4886-9d4e-c1a6e10348ad /boot           ext4    defaults        0       2
# /boot/efi was on /dev/sdb7 during installation
UUID=6226-94B9  /boot/efi       vfat    umask=0077      0       1
/dev/mapper/ubuntuvg-home /home           btrfs   defaults,[email protected] 0       2
/dev/sdb8       none            swap    sw              0       0
/dev/disk/by-uuid/BC34947B34943A7A /media/D auto nosuid,nodev,nofail,x-gvfs-show,x-gvfs-name=D 0 0

where the UUIDs can be obtained by running sudo blkid.

II. Reallocate space between logical volumes

To resize / and /home as logical volumes defined above, I ran the following commands to reallocate a 10G space from /home to / while booted into the USB Ubuntu liveCD system. The idea is to reduce the physical directory space using btrfs filesystem tools before to use lvresize command to reduce the logical volume space for /home, and then to increase the logical volume space for / before to increase the btrfs filesystem physical space next. Using btrfs resizing tool needs to have the physical partitions mounted to a system directory, which is different from ext4 and other filesystems; while resizing logical volumes on LVM usually requires those logical volumes unmounted from the system to be safe.

sudo vgdisplay -v # to preview the physical volumes, virtual groups and logical volumes from LVM.
sudo mount -t btrfs /dev/ubuntuvg/home /mnt
sudo btrfs filesystem resize -10G /mnt
sudo umount /mnt
sudo lvresize -L -10G ubuntuvg/home
sudo lvresize -L +10G ubuntuvg/root
cd /dev
sudo mkdir media
cd media
sudo mkdir root
sudo mount -t btrfs /dev/ubuntuvg/root /dev/media/root
sudo btrfs filesystem resize +10G /dev/media/root
sudo mount /dev/ubuntuvg/home /mnt
df -h
sudo vgdisplay -v

The last step had verified that all of the commands worked out for this resizing operation between / and /home logical volumes. In case that /etc/fstab has used UUID to mount the volume groups, you may also need to update the new UUID in /etc/fstab accordingly. The new UUIDs can be found by sudo blkid.

Manage application menu

On XFCE or GNOME desktop environment of Linux OS, one can browse applications on a dropdown menu. To manage the menu structure and add shortcuts to the application menu, there are two places to look for: one is located at ~/.config/menus with files named as *.menu which define how a particular program’s shortcut is located in the application menu–not every desktop environment has this, neither all programs have a menu file defined here; the other one is located at ~/.local/share/applications with files named as *.desktop which define shortcuts with commands of running the target programs. The second type of files are called by the menu files. System-wide programs may be located at other places. For new programs, one can create the .menu and .desktop files to let the shortcut shown on the dropdown application menu.

Shorten bash terminal (shell) prompt path and show the branch name if the folder is a git repository

The first concern is that the path and computer name in the terminal may become very long when working in a deep level of directory. The second concern is that it would be nice to show the name of the current git branch if the directory is in a git repository.

For the first concern, this line of path and computer information may be greatly shortened following this instruction. The full directory path can still be seen on the top of the terminal window or typing pwd command. For the second concern, this code snippet seems serving the goal, which can show the git branch and status information in red in the terminal prompt if possible. By understanding the two code snippets in the shell prompt language, I made up the following code to replace the corresponding code session in ~/.bashrc (yes, the code is used to replace the original PS1 setting code, not to be appended at the end of the file):

# Add git branch if its present to PS1.
parse_git_branch() {
 git branch 2> /dev/null | sed -e '/^[^*]/d' -e 's/* \(.*\)/(\1)/'
# This part is to shorten the prompt path and show the git branch name in red if possible.
if [ "$color_prompt" = yes ]; then
    PS1='${debian_chroot:+($debian_chroot)}\[\033[01;32m\]\u\[\033[00m\]:\[\033[01;34m\]\W\[\033[01;31m\]$(parse_git_branch)\[\033[00m\]\$ '
    PS1='${debian_chroot:+($debian_chroot)}\u:\W$(parse_git_branch)\$ '
unset color_prompt force_color_prompt

By reopening a new terminal, you can see the path in the prompt is shortened and the branch name is shown in red if the path is in a git repository.

In fact, there have been a lot of bash prompt customization tools available on the internet. For example, the customization code snippet from the official Git Bash repo may be found here which gives the bash prompt the capability to display more git status information on the command lines. One can save the code snippet as under the user’s root folder (~). In the ~/.bashrc file, one just need to add source ~/ to replace the lines defining parse_git_branch() in my code sample, and then replace the parse_git_branch with __git_ps1 "(%s)" on the lines about PS1. The code allows the prompt to display a reminder if the local repository is in the process of merging or rebasing and other helpful information. Similarly, one can also define some alias of commands in the ~/.bashrc file as shortcuts.

Set time zone reference frame for accurate time synchronization for dual boot systems

For Ubuntu-Windows dual boot systems, the OS’s seem to mess up with each other’s time zone reference point when they are reading time from the BIOS and time server. This results in a wrong time when switching OS sometimes. To eliminate the time difference for both OS’s, this and this instructions might be helpful.

What I did is to run timedatectl set-local-rtc 0 --adjust-system-clock in Ubuntu to use the UTC time for the machine time (RTC) which could avoid the change of daylight saving time and so on. This should be the default behavior of Ubuntu 16.04 if the setting hasn’t been changed before. The time synchronization setting can be checked via command timedatectl status. Certainly, the time can be set to synchronize from a server by sudo ntpdate On Windows 10, I open a command prompt as admin and run reg add "HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\TimeZoneInformation" /v RealTimeIsUniversal /d 1 /t REG_QWORD /f. This is equivalent to add a QWORD with value 1 to a key called RealTimeIsUniversal under the HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\TimeZoneInformation register table.

Delete old XFCE sessions

If the session and startup is set to reopen apps from the last sessions in the XFCE settings, old sessions might have been stored in the computer for a while. All those sessions will be available to choose from When logging in to the XFCE desktop environment. To delete the useless ones, go to ~/.cache/sessions and edit the file named as xfce4-session-QC5-ubuntu:1 or similar with your computer’s name. There may be a section started with [Session: Default] and many others like [Session: someoneelse]. Delete the parts for [Session: someoneelse] if needed. On next login, double click the session to use, or the default one will be opened automatically if there is no other sessions available.

An issue with slow boot

I reported here and here about the issue. I have filed a bug report on launchpad. I saw people suggesting to mask the systemd-udev-settle.service service which is responsible to the biggest fraction of the slowdown, but I am using LVM which would need the service to startup. Below are some notes on debugging and solving the problem.

Notes on using some common tools

Changing Java settings

To change the default version of Java commands, one can run

sudo update-alternatives --config java

and select an installed version as default. Similarly, one can replace java with javac (the compiler), javadoc (documentation generator) and jarasigner (JAR signing tool) for configuring other java-associated settings.

After this, to change the home directory of Java, one can edit by sudo nano /etc/environment with a line of JAVA_HOME="/usr/lib/jvm/java-8-oracle", for example, where the directory path is copied from the sudo update-alternatives --config java setting. After saving the setting, run

source /etc/environment

to reload the setting. To check if the new setting has taken effect, one can use echo $JAVA_HOME and javac --version, for instance.

Automate system configuration and package installations

The software Ansible looks promising to automate some system admin jobs, which is especially useful when migrating to another computer of similar/same model. Also see project Thinkpad X1 Yoga (gen 2) for real case scenario of using it.