Archive for October, 2013

The WYSE S30 is a thin client. It’s CPU is an AMD Geode GX500. (It identifies itself as National Semiconductor). It’s based on an CPU design by Cyrix. Anyhow, the default kernel doesn’t boot on this CPU. I’ve tried to run the installation image on this machine. In this situation, the kernel gives a message about an unknown CPU manufacturer and says the system might be unstable. After this, the system hangs. The bootable ArchLinux installation I’ve just created, it gets stuck in a reboot loop.

I have tried to compile a kernel optimised for this specific CPU, but I still get inside this reboot loop. I am wondering if the problem is caused by the bootloader. As said, I’ve been using SYSLINUX. In stead I’ve tried GRUB2, but that doesn’t even display the menu. Seeing these symptoms, it appears there are some more incompatibilities about this particular device.

Looking at the CPU, from what I’ve read, it’s basically an 686-class CPU with some AMD extentions such as 3dnow! However, it is lacking one i686 instruction. However, as I’ve created a custom kernel, that should not be the problem. Therefore I am blaming the bootloader now. So… I might have to look at some alternative boot loaders. But…. not today…. been working at this for hours. It’s enough for today.

To install a new ArchLinux installation from an existing installation, one would generally use the method described on their wiki.

However, I am running an x86_64 installation, and I would like to create an i686 bootable USB stick. My first idea was to replace the x86_64 bootstrap image by the i686 bootstrap image. However, the resulting USB stick turned out to be x86_64. So, that’s not quite what I was looking for.

In order to create an i686 installation from an existing x86_64, one needs to adjust the /tmp/root.i686/etc/pacman.conf file ( or /tmp/root.x86_64/etc/pacman.conf for that matter) and change the Architecture line.

Architecture = auto


Architecture = i686

then run the pacstrap command as described on the wiki.

After pacstrap is finished, chroot into the installed system, and install the bootloader. I’m using syslinux for this.

sh-4.2# arch-chroot /mnt
sh-4.2# pacman -S syslinux
error: failed to prepare transaction (package architecture is not valid)
:: package syslinux-6.02-4-i686 does not have a valid architecture

The “Architecture” must also be set in the target installation. In the “base” installation, nano is the available editor. Run nano /etc/pacman.conf and modify the file, then we can install syslinux.

sh-4.2# pacman -S syslinux
resolving dependencies...
looking for inter-conflicts...

Packages (1): syslinux-6.02-4

Total Download Size:    1.09 MiB
Total Installed Size:   3.24 MiB

:: Proceed with installation? [Y/n] 
:: Retrieving packages ...

etc. etc. the installation continues normally.

Installing the bootloader can take a few minutes. Be patient.

sh-4.2# syslinux-install_update -i -a -m
Syslinux BIOS install successful
Boot Flag Set - /dev/sdc1
Installed MBR (/usr/lib/syslinux/bios/mbr.bin) to /dev/sdc

Now, we need a little configuration. We still have to create an fstab, and configure the bootloader. Since we’re going to boot from USB, I will be using UUID’s to identify the right volume.

[root@8570w andre]# blkid /dev/sdc1
/dev/sdc1: UUID="a7d9a26e-abb3-461b-aacd-1ae4fe96f2e4" TYPE="ext2" PARTUUID="b71a2cfb-01" 

/etc/fstab becomes

UUID=a7d9a26e-abb3-461b-aacd-1ae4fe96f2e4	/         	ext2      	rw,relatime,data=ordered	0 1

in /boot/syslinux/syslinux.cfg, set the right root device, so it looks like

    MENU LABEL Arch Linux
    LINUX ../vmlinuz-linux
    APPEND root=UUID=a7d9a26e-abb3-461b-aacd-1ae4fe96f2e4 rw
    INITRD ../initramfs-linux.img

I have tried to boot the resulting stick in qemu-system-i386, which fails, but it boots on real hardware. (To test this, I have enabled CSM temporary on my laptop)

Now, we are this far, we can start the second phase of the project.


Yaourt is a wrapper for the pacman package manager, which allows installing from the Arch User Repository, AUR. It also supports installing from a normal user, and then prompts for a password to install through su or sudo. Yaourt is not available in the standard repositories. However, it’s available in the ArchLinux France repository:
Adding the archlinuxfr repository and install yaourt.


In order to have a dynamic network configuration, for example, to easily connect to a WiFi, NetworkManager is a convenient tool.

[andre@8570w ~]$ yaourt -S openntpd networkmanager dnsmasq bluez network-manager-applet gnome-keyring
[root@8570w andre]# systemctl enable NetworkManager 
[root@8570w andre]# systemctl enable NetworkManager-wait-online

Network was previously configures to use DHCP on the ethernet connection, disable this to prevent conflicts:

[root@8570w andre]# systemctl disable dhcpcd@enp0s25

In order to start openntpd once the connection has been established, look here.


Installing NFS support and enabling the daemons:

[andre@8570w ~]$ yaourt -S nfs-utils
[root@8570w andre]# systemctl enable rpc-statd
[root@8570w andre]# systemctl enable rpc-gssd

NFS3 vs NFS4

On the old laptop, I was running NFS3 as NFS4 gave me some problems. There was a detail I’ve overlooked. Back then, I was like, I’ll come back to this issue later, and never looked into it again. The difference between mounting NFS3 and NFS4 is the fact NFS3 required the full path on the server, while NFS4 requires a relative path to the NFS root. Therefore the corresponding lines in the fstab are

NFS3:   /mnt/1000   nfs defaults,user,noauto 0 0

NFS4:   /mnt/1000   nfs4 defaults,user,noauto 0 0

Mounting from thunar

In order to mount removable media, for example, an SD card, “USB stick” or external hard disk through Thunar, we need the gvfs

[andre@8570w ~]$ yaourt -S gvfs

Before I installed gvfs, I had my user mountable NFS mounts from my fstab listed in Thunar, after installing they were gone. To get them back, add x-gvfs-show to the mount options in fstab   /mnt/1000   nfs4 defaults,user,noauto,x-gvfs-show 0 0


This laptop has a huge touchpad. By default it was set to “two finger scrolling”. Really annoying. This is not a Mac for God’s sake! After a day, I’m getting RSI pains. No good… but fortunately, this is just a configuration option, and can be set to normal edge scrolling.

woohoo --- i can use edge scrolling


At first, I didn’t configure a language. Just running in English was fine, but what I didn’t realise is that this meant I was running in “C”, meaning it wasn’t using UTF-8 encoding, which prevented me from listening to Russian music, as their file names contain non-ASCII characters.

[andre@8570w ~]$ echo $LANG

Before setting any locale, it has to be generated. Uncomment the desired locales in /etc/locale-gen and run locale-gen

[root@8570w andre]# locale-gen

So, I am going to configure a global setting of UK English with UTF-8,

[root@8570w andre]#  localectl set-locale LANG=en_GB.UTF-8

and set my own profile to Dutch by adding the following to ~/.bashrc

export LANG=nl_NL.UTF-8

So… I am a little very frustrated at gitosis. But first things first…

Yesterday I’ve been finishing the configuration on my new laptop. Installing some software and such. I will post a report on that later. This afternoon, I’ve been backupping my old laptop. I took an imagine of the Windows partition, and made a tarball of the linux root file system. Earlier I’ve already made a backup of the home partition.

So, I can start opening the old laptop, and remove all the dust from it. When that’s done, it will be completely reinstalled. Actually, I was thinking about just putting a copy of the root file system from my new laptop on the old one, and then just make two required modifications: setting the hostname and creating a new fstab (as it’s UUID based)

But that’s not what my frustrations are about. This afternoon, I wanted to work on my OAuth WordPress plugin. Still a few things to be done before it can be released. Well… in order to work on my source I need to access the git. I am using gitosis to manage access to my repository. Therefore I need to add the SSH public key from my new laptop into it’s access control list. Since all my computers are in the gitosis system, I logged into my raspberry pi server, and attempted to add the said key.

[andre@rpi-server keydir]$ git push
Counting objects: 5, done.
Compressing objects: 100% (3/3), done.
Writing objects: 100% (3/3), 338 bytes | 0 bytes/s, done.
Total 3 (delta 1), reused 0 (delta 0)
remote: WARNING:gitosis.ssh:Unsafe SSH username in keyfile: ''
   345de48..f110208  master -> master

Unsafe SSH username, and the key is not added. What the fuck. I started googling, and it suggested the “username” gitosis complains about is the past part of the keyfile. (This is just an arbitrary string, a comment, to identify the key to a user looking at the string, and it not used by the machine.)

ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQCbkYGFFpgrYZKy2eJIdSIebvuqaBhW96owRD7WT5MgwUXVtv76NeTJHgoZbUM7/zex+DV4tZwxkTZMQ8yXUmmqdrjYdHSlkxekcE7uvX0Xul0sRywm/Qfbdlu/DQeuTorhY5zb/Zx6TNXR+zw5KKlTPbnTEIXDwqh9kpenQR0auWL24/UQjoC9hTZavxY3cQlMEABGyfyTx7xnxQFZZGZ7h5x+OOWraXY6eYw2eHV53M+eSEIqFJz3LTD4tSl4svWNHTMz7EEfKpO5FmGjwJKIz8RYQiutCmz4bfziIbZvGYJGnUl/u/UlgtdkxaMjvPkZCmnWB+nqtLre2RFKQOWz andre@8570w

But even changing that to, for example or whatever I tried. (Trying a FQDN formatted hostname as last attempt) Everything I tried, it kept complaining. As last resort I simply disabled the checking for the name. This is done in /usr/share/pyshared/gitosis/
I have replaced

def isSafeUsername(user):
    match = _ACCEPTABLE_USER_RE.match(user)
    return (match is not None)


def isSafeUsername(user):
    return (True)

So, my key would be accepted regardless this “username”. But really, I’ve tried anything… I don’t get it.

Well… looking at what _ACCEPTABLE_USER_RE.match(user) is:

_ACCEPTABLE_USER_RE = re.compile(r'^[a-zA-Z][a-zA-Z0-9_.-]*(@[a-zA-Z][a-zA-Z0-9.-]*)?$')

I am not familiar with python, but, it looks like a regluar expression saying the hostname should start with a letter, which was not the case for my hostname. But, I’ve tried many variants, including plenty that should just pass this little test. So… I don’t get the problem, and I’ve been pulling my hair out. Well… I just disabled the silly little test and got access to my reposiroty.

The crazy thing is, I’ve not only added my laptop, but also another ssh account at my server, which I use for development, right in the shell, which got accepted without a complaint. Really…. what the fuck.

Update: ran this test, and expected, is accepted just fine by this test.

andre@blaatschaap:~$ cat 
import re

_ACCEPTABLE_USER_RE = re.compile(r'^[a-zA-Z][a-zA-Z0-9_.-]*(@[a-zA-Z][a-zA-Z0-9.-]*)?$')

def isSafeUsername(user):
    match = _ACCEPTABLE_USER_RE.match(user)
    return (match is not None)

if isSafeUsername(""):
    print "ok"
    print "nak"

if isSafeUsername(""):
    print "ok"
    print "nak"

if isSafeUsername("andre@8570w"):
    print "ok"
    print "nak"

andre@blaatschaap:~$ python 

Therefore I wonder, is this “username” it complains about something else? Encoded in the base64 string? I thought that was the key only….

Well, last post I’ve discussed the problems I’ve encountered while trying to boot the ArchLinux installation CD in UEFI mode. As mentioned, it booted fine in legacy (BIOS) mode. Anyhow, let’s have a look at the installation procedure on an UEFI/GPT system.

IMG_1324Since the hard disk was previously partitioned using the MBR partition scheme and booting in legacy (BIOS) mode, we need to start from scratch. This includes creating the UEFI system partition.

When we start the GPT version of fdisk, gdisk, it will give a message about the fact the hard disk is currently using  the MBR partitioning scheme, and converting could possibly be destructive. We are going to repartition the whole disk, so it’s a destructive operation anyways.

So, we create an empty parition table, with the o command, just the way we would do in fdisk, we create a new partition with the n command. Here we notice a little difference, we can enter a number between 1 and 128 here. We’ll proceed with creating a 512 MB partition, and enter HEX code EF00 for EFI BOOT PARTITION.

Since the new laptop got 8 GB of RAM, the size of the Swap partition will be 16 GB. This is the next partition I’m going to create. I will enter 8200 for Linux Swap Parition.

The next two partitions will be the root file system and the home partition, 50 and 200 GB. I have kept left an amount of unpartitioned disk space for potential use with other operating systems. gdisk defaults to code 8300, which is ext2/3/4, so I don’t need to change this. All the same as with fdisk.

Next, will be formatting the partitions. However, attempting to format the partitions immediately after closing gdisk will result in an in-use error message. Re-reading the partition table doesn’t make a difference, a REBOOT is required to proceed. After rebooting, formatting works like a charm. The UEFI BOOT partition MUST be formatted FAT32. The other partitions will be formatted ext4, and the swap partition will have to be initialised with mkswap.

# mkfs.vfat -F32 /dev/sda1
# mkfs.ext4 /dev/sda3
# mkfs.ext4 /dev/sda4
# mkswap /dev/sda2

After formatting the partitions, we can mount them and start the ArchLinux installation process.

# swapon /dev/sda2
# mount /dev/sda3 /mnt
# mkdir /mnt/home
# mount /dev/sda4 /mnt/home
# mkdir -p /mnt/boot/efi
# mount /dev/sda1 /mnt/boot/efi
# pacstrap base /mnt
# genfstab -U -p /mnt >> /mnt/etc/fstab
# arch-chroot /mnt
# ln -s /usr/share/zoneinfo/Europe/Amsterdam /etc/localtime
# mkinitcpio -p linux

So far, the installation procedure itself, has been pretty standard. The next step, setting up the boot loader, has caused me some problems. And this is the step that differentiates UEFI from BIOS booting, the boot process.

First of all, the EFI variables must be mounted. Please note the grub-install is executed inside the arch-chroot.

# mount -t efivarfs efivarfs /sys/firmware/efi/efivars
# pacman -S grub efibootmgr
# grub-install –target=x86_64-efi –efi-directory=/boot/efi –bootloader-id=grub

However, this resulted in an error message show_boot_order() "no such file or directory". The result is the system doesn’t boot automatically, but when I choose to browse the file system, I can load GRUB2 and boot correctly.

Next, I’ve tried to install Gummiboot, and “chainload” GRUB2. This works fine. However, gummiboot assumes the EFI boot partition to be mounted at /boot.

# pacman -S gummiboot
# gummiboot –path=/boot/efi

Now, setting it up to chainload GRUB, in /boot/efi/loader/entries/grub.conf:

title GRUB
efi /EFI/grub/grubx64.efi

And in /boot/efi/loader/loader.conf

default grub

makes the system bootable.

UEFI boot

As I am about to install my new laptop, I am considering using UEFI. The laptop supports UEFI but came with legacy booting enabled as default option. Whether this is the university setting or the factory default I’m not sure, since my father’s HP laptop got UEFI enables as default option.

I haven’t been working with UEFI systems before. Well, I’ve installed Ubuntu on my father’s laptop, on an external (USB) hard disk, so he could dual boot Windows and Linux. And that’s one of the advantages of UEFI. No boot manager is required. This is built in into UEFI.

My new laptop, the HP EliteBook 8570w has three options, legacy boot, uefi hybrid (both legacy and uefi / with csm), and native uefi (without csm).

I have selected the compatible mode. Now, as I try to boot from the ArchLinux installation CD. I am still using my external (USB) optical drive, since the laptop was delivered without internal optical drive. Well… so I boot from the CD, a menu shows, where I can choose between booting Arch, an UEFI shell (version 1 and 2) and to return to the firmware.

So, I select to boot Arch. The screen goes black. The USB optical drive appears to be reading slowly, but nothing happens. When I reboot, the uefi firmware does not detect the USB optical drive. Something is not going right. ArchLinux boot media are a hybrid case. The ISO contains both SYSLINUX and UEFI booting, but can also be dd’d to an USB flash medium. I believe this is the reason why I have had experienced booting ArchLinux installation media on my EliteBook 8530w. I have made an UEFI bootable USB flash drive according to the instructions on the ArchWiki.Using this I can boot normally. This appears to confirm my hypothesis the hybrid boot medium confuses the HP firmware.

So, creating partitions. Since I am using GPT, I will use gdisk in stead of fdisk. Creating partitions will be similar to fdisk. Partitions have numbers from 1 to 128. There are no concepts like extended and logical partitions. They’re a dirty hack of the MBR partitioning scheme anyways. I am still overthinking how I need to partition this system. This is the first time I am working an UEFI based system and GPT style partitioning. I still haven’t decided what boot loader I will be using, either SYSLINUX or GRUB2.


Oké mensjes, ik heb vanmiddag mijn nieuwe laptop van de universiteit opgehaald. Het is een HP EliteBook 8570w. Hij ziet er goed uit, en zo’n verlicht toetsenbord is best cool (en handig als je in het donker wil typen)IMG_1297

Natuurlijk staat hier de universiteitsimage op met allerlei meuk waar ik niet op zit te wachten. Het is trouwens een Windows 7 systeem. Thank God is not Windows 8, maar boeiend, het gaat er toch af. Op de vorige universiteitslaptop (HP EliteBook 8530w) werd ik gek van die “anti-rsi” software die zeurt dat ik te snel typ. Maar ff serieus, kijk eens hoeveel ikoontjes er langs de klok staan! En waar is het allemaal goed voor? Argh… ik haat voorgeïnstalleerde software.  IMG_1299

En kijk hoe ik “begroet” wordt? Muisgedrag? Opvolging interventies? Ja weer van die “anti-rsi” software (andere dan de vorige keer) die me constant lastig gaat vallen . Nee…. daar gaan we dus niet aan beginnen. We gaan die hele laptop leeg gooien en opnieuw installeren. En als we dat dan toch moeten doen, er natuurlijk gelijk Linux op zetten. Geen Microsoft meuk! Afin. Het plan was dus eerst een image te trekken van de huidige inhoud van die harde schijf, om er vervolgens ArchLinux op te zetten.

startup_screemHet eerste issue daarbij is dat er blijkbaar geen optische drive in deze laptop zit. Wegbezuinigd? Afin, er lijkt een blindklep op te zitten, dus waarschijnlijk is deze gewoon te bestellen. (zie hiero) Dusch…. dan maar externe DVD drive aangesloten… doet die drive het niet meer. Arrgghh…. Dat schiet dus niet op. Nou ja… eerst maar eens navragen of ik via de universiteit nog een optische drive kan aanschaffen, anders maar kijken waar ik die kan bestellen, want ik heb geen zin altijd een USB DVD drive te moeten meezeulen. (Trouwens ik betwijfel of ik een nieuwe USB DVD drive ga aanschaffen aangezien ik geen netbook meer heb)

Het zou in principe mogelijk zijn om vanaf een USB stick te booten en toch door te gaan met de image en installatieprocedure, maar ik heb daar vandaag mooi geen zin meer in.

Edit: It’s the power adaptor, it’s always the power adaptor. Ik had dus twee gelijke voedingen (precies zelfde type) voor de desbetreffende USB DVD drive. Het probleem lag dus aan de voeding. Met de andere voeding werkt hij wel. (alsof ik niet al genoeg ellende met voedingen heb… het in “Power for Pi” project is ook alweer gestrand in verband met een crappy chinese voeding)