Debian on a Fujitsu Lifebook T2010

Debian 'unstable' (as of Dec 26th 2008) seems to have everything needed and there is no longer the need to by hand compile kernels and what-not in the general case. Worth bearing in mind, is that I opted for the 64bit install and so far had no problems.

Not much to say about this. Of course the laptop does not come with any kind of DVD or CD drive which defeats the usual Windoze, Ubuntu and Gentoo weenies. Fortunately being a clever lad, having had the experience and enjoyed watching the pansies cry online “I don't know how to install an OS without a CD drive” I just smiled and PXE booted the thing.

Whining Weiners….you should all be ashamed. For those wishing to learn what real computing hardware is about, buy a secondhard Sun, or Alpha, boxen from fleabay and install Linux on that. Once you have done that you will be able to join the rest of us in our intrinsic hatred for x86….anyone want to give me a Tadpole laptop? :)

Why people still burn CD's when they could just PXE boot in 95% of situations, especially so when you can get even Windoze to bootstrap, is beyond me… I guess you could USB boot, however that really is wimping out. :)

This is actually a serial port device, and is automatically configured via ACPI's PNP.

To use it you will need a patched version of inputattach that supports baud/w8001 and to have wacom_w8001 (CONFIG_TOUCHSCREEN_WACOM_W8001) support compiled into your kernel.

# cat <<'EOF' > /lib/udev/inputattach
#!/bin/sh -e
#
# udev script for inputattach

(
        . /lib/udev/hotplug.functions
        wait_for_file /usr/bin/inputattach
        exec /usr/bin/inputattach $@
) &
EOF
# chmod +x /lib/udev/inputattach

# cat <<'EOF' >> /etc/udev/rules.d/99-local.rules
SUBSYSTEM=="tty", KERNEL=="ttyS[0-9]*", ATTRS{id}=="FUJ02e5", ACTION=="add", RUN+="inputattach --daemon --baud 19200 --w8001 /dev/%k"
EOF

It's worth having a look at Debian Bug #616443.

N.B. the pen fails to align properly with the bottom corners of the screen and so needs some adjusting. When I find it becomes annoying, I first have to find a use for the pen, I'll post up what I did to fix the problem.

'xbacklight' works generally fine, however before it starts funtioning you will need to put the following into your '`~/.xsession`' file or you will be blinded:

xbacklight -set 30

The screen is very bright at 100% I would recommend you drop it to aroun the 30 or 40% mark.

All the hard work has been done already, it's lurking in the fjbtndrv package. To fetch it type:

$ svn co http://fjbtndrv.svn.sourceforge.net/svnroot/fjbtndrv/trunk fjbtndrv 

This downloads the latest version of the whole lot, however the whole thing is pretty sucky except for '/usr/src/fjbtndrv/src/linux/' bit, go into that directory, type 'make' and then as root type 'make install'. Afterwards you can slip 'fjbtndrv' into your '/etc/modules' and xorg will start seeing keyboard events.

We need a few udev rules though, so in '`/etc/udev/rules.d/99-local.rules`' add:

# cat <<'EOF' >> /etc/udev/rules.d/99-local.rules
SUBSYSTEM=="input", KERNEL=="event[0-9]*", ATTRS{name}=="fsc tablet buttons", SYMLINK+="input/by-path/platform-fsc_btns-event-kbd"
# as xorg does not support sw events we use evrouter which means we need global readable 
# N.B. no a risk as the most people can do is read when we go into tablet mode :) 
SUBSYSTEM=="input", KERNEL=="event[0-9]*", ATTRS{name}=="fsc tablet switch", SYMLINK+="input/by-path/platform-fsc_tab-event-sw", MODE="644" 
EOF

To make use of the buttons I would recommend you use 'xbindkeys' and evrouter to give you the support for the 'tablet mode' switch (the bit that tells your computer that you have tabletised the screen). I originally rolled the patch to add switch event support to evrouter as xorg does not support switch events and fortunately the author has incorporated this into his application.

Once compiled (with 'debuild') and installed you will need to create ~/.evrouterrc (similar to below) and then you should start being able to run scripts on this event:

"fsc tablet switch" "/dev/input/by-path/platform-fsc_tab-event-sw" none sw/1/1 "Shell//usr/local/bin/tabletRotateScreen normal"
"fsc tablet switch" "/dev/input/by-path/platform-fsc_tab-event-sw" none sw/1/0 "Shell//usr/local/bin/tabletRotateScreen right"

Annoyingly the screen brightness hotkeys are non-functional here, they pop up as ACPI events so you will need to add to '`/etc/rc.local`' (you will need 'acpi-support' installed too but thats a good thing anyway):

# fujitsu t2010 brightness keys 
/usr/bin/setkeycodes e0e0 224 # brightness down 
/usr/bin/setkeycodes e0e1 225 # brightness up 

'xbindkeys' will run a script when you press the screen rotate button and also for the volume hotkeys when you add to your '`~/.xbindkeysrc`':

alex@berk:~$ grep "^[^#].*" ~/.xbindkeysrc 
"tabletRotateScreen"
  XF86RotateWindows
"amixer sset Master,0 mute"
  XF86AudioMute
"amixer sset Master,0 unmute; amixer sset Master,0 5%-"
  XF86AudioLowerVolume
"amixer sset Master,0 unmute; amixer sset Master,0 5%+"
  XF86AudioRaiseVolume
"flipper"
  XF86Display

'flipper' is a handy script that lets you flip between the TFT and external monitor:

$ cat <<'EOF' > /usr/local/bin/flipper
#!/bin/sh -e

#set -x

SCREEN_DEF=LVDS

XRANDR=$(xrandr --query)

SCREEN_DEF=$(echo "$XRANDR" | grep "^$SCREEN_DEF" | head -n1 | cut -d' ' -f 1)

SCREENS=$(echo "$XRANDR" | grep ' connected ' | cut -d' ' -f 1)

if [ $(echo "$XRANDR" | grep '*' | wc -l) -ne 1 ]; then
        xrandr --output $SCREEN_DEF --auto --preferred --primary

        for SCREEN in $SCREENS
        do
                [ "$SCREEN" = "$SCREEN_DEF" ] && continue

                xrandr --output $SCREEN --off
        done

        exit 0
fi

[ $(echo "$XRANDR" | grep ' connected ' | wc -l) -eq 1 ] && exit 0

SCREEN_CUR=$(echo "$XRANDR" | grep -m 1 -B 1 '*' | head -n1 | cut -d' ' -f 1)

FLAG=0
for SCREEN in $SCREENS
do
        [ $FLAG -eq 1 ] && break

        [ "$SCREEN" = "$SCREEN_CUR" ] && FLAG=1
done

[ "$SCREEN" = "$SCREEN_CUR" ] && SCREEN=$(echo $SCREENS | cut -d' ' -f1)

xrandr --output $SCREEN --auto --preferred --primary
xrandr --output $SCREEN_CUR --off

exit 0
EOF

$ chmod +x /usr/local/bin/flipper

'tabletRotateScreen' is a handy script to deal with screen rotation:

$ cat <<'EOF' > /usr/local/bin/tabletRotateScreen
#!/bin/sh -e

set -x

SCREEN=LVDS
STYLUS='Wacom Serial Penabled Touchscreen'

XRANDR=$(xrandr)
SCREEN=$(echo "$XRANDR" | grep "^$SCREEN" | head -n1 | cut -d' ' -f 1)

STATE=$(xrandr | grep "^$SCREEN connected" | awk '{ split($3,res,/x|\+/); print (res[1] > res[2]? "wide": "tall") }')

# we probably are in VGA with the LCD turned off so exit
[ -z "$STATE" ] && exit 0

# we munge the Fn/Alt tablet key to be more useful also
# Fn  -> wayv/stuff
# Alt -> window manager
if [ "$STATE" = "wide" ]; then
  # go right
  xrandr --output $SCREEN --rotate right &
  xinput set-prop "$STYLUS" 'Evdev Axis Inversion' 0, 1
  xinput set-prop "$STYLUS" 'Evdev Axes Swap' 1

  xmodmap -e 'remove Control = Control_L'
  xmodmap -e 'add Mod3 = Control_L'
  xmodmap -e 'remove Mod1 = Alt_L'
  xmodmap -e 'add Mod4 = Alt_L'
else
  # go normal
  xrandr --output $SCREEN --rotate normal &
  xinput set-prop "$STYLUS" 'Evdev Axis Inversion' 0, 0
  xinput set-prop "$STYLUS" 'Evdev Axes Swap' 0

  xmodmap -e 'remove Mod4 = Alt_L'
  xmodmap -e 'add Mod1 = Alt_L'
  xmodmap -e 'remove Mod3 = Control_L'
  xmodmap -e 'add Control = Control_L'
fi

exit 0
EOF

$ chmod +x /usr/local/bin/tabletRotateScreen

You will need to fixup the keymap slightly to make the onscreen buttons more useful than they are by default, best to put in your '`~/.xsession`' file:

# sort out the tablet keys to be sane 
xmodmap -e 'keycode 161 = XF86RotateWindows' 
xmodmap -e 'keycode 185 = Prior' # XF86ScrollUp 
xmodmap -e 'keycode 186 = Next'  # XF86ScrollDown 

It works, but the 3D acceleration is not great, glxgears comes back with a unstunning 415fps. I feel it should be faster but others are reporting a similar value…and it seems to be linked to that the Intel driver for the chipset is not really 3D 'enabled'. Not bothered, I'm not a gamer or an eye candy kind of person.

alex@berk:~$ glxinfo 
name of display: :0.0
display: :0  screen: 0
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.2
server glx extensions:
[snipped]
72 GLX Visuals
   visual  x  bf lv rg d st colorbuffer ax dp st accumbuffer  ms  cav
 id dep cl sp sz l  ci b ro  r  g  b  a bf th cl  r  g  b  a ns b eat
----------------------------------------------------------------------
0x21 24 tc  0 32  0 r  y  .  8  8  8  8  0 24  8  0  0  0  0  0 0 None
0x22 24 dc  0 32  0 r  y  .  8  8  8  8  0 24  8  0  0  0  0  0 0 None
0xcd 24 tc  0 24  0 r  .  .  8  8  8  0  0  0  0  0  0  0  0  0 0 None
0xce 24 tc  0 24  0 r  .  .  8  8  8  0  0  0  0 16 16 16  0  0 0 Slow
0xcf 24 tc  0 24  0 r  y  .  8  8  8  0  0  0  0  0  0  0  0  0 0 None
0xd0 24 tc  0 24  0 r  y  .  8  8  8  0  0  0  0 16 16 16  0  0 0 Slow
0xd1 24 tc  0 24  0 r  y  .  8  8  8  0  0  0  0  0  0  0  0  0 0 None
0xd2 24 tc  0 24  0 r  y  .  8  8  8  0  0  0  0 16 16 16  0  0 0 Slow
0xd3 24 tc  0 24  0 r  .  .  8  8  8  0  0 24  0  0  0  0  0  0 0 None
0xd4 24 tc  0 24  0 r  .  .  8  8  8  0  0 24  0 16 16 16  0  0 0 Slow
[snipped more 'None' and 'Slow' :( ]

Nice to get to the good old British pound symbol on my keyboard :) Add to your '`~/.xsession`' file something like:

# give me sterling for Shift-Alt-3 
xmodmap -e 'keycode 12 = 3 numbersign sterling' 
xmodmap -e 'keycode 26 = e E [[EuroSign]] cent' 

The fingerprint reader is an “AuthenTec, Inc. AES2501 Fingerprint Sensor” (USB-ID: 08ff:2580) and is supported via the userspace debian tool 'aes2501-wy'. It works fine when you call it via 'su -c aes2501' creating PNM files…fingerprint readers are really just scanners anyway. Apparently if you use the Debian experimental bits then you get 'libpam-fprint' which will give you an easy way to do fingerprint authentication.

I personally think fingerprint authentication is a joke (you could use it combined with some other method though, or maybe as a quick xscreensaver unlocker…) and should be avoided, two main reasons:

• you cannot revoke fingerprints when they are compromised. Any security system that uses unrevokable authentication tokens is fundementally flawed
• a fingerprint reader…next to a keyboard…covered in the users fingerprints…yeah, great idea!

One thing that I might look into using it for is as an extended keyboard. Some ideas:

• when the laptop is in tablet mode you only get the pen, with gestures to learn and slow character input, and five physical not so useful keys on the screen. The fingerprint reader however is also on the screen and conviently placed to be potentially persuaded to act as a nifty shortcut feature. For example I could get it to open an xterm when I run my index finger on the scanner
• when not in tablet mode it is a generally useless place for gestures. However, there is no suspend-to-ram or hibernate hotkeys on the laptop so instead I could swipe my finger to hibernate the laptop; before I only did a suspend-to-ram when closing the lid
• annoying Fn-F10 (screen off) does not work so I cannot use this to instantly lock the machine, again finger reader action here

The ath5x driver has worked generally fine for me since 2.6.29. With more recent kernels, if you have rfkill support enabled, your hardswitch on the front left controls both bluetooth (works with btusb) and wifi (without rfkill it simply 'unplugs' the internal USB bluetooth dongle).

The only issue I had to overcome was the braindeadness of the HAL refusing to accept my region settings, as I'm a UK dweller. To deal with this you type the following as root:

ath_info -v -g 3:0 -w 0xfc200000 regdomain 0x60

For more details on this have a read of the madwifi ticket 2156, ath_info and country codes.

Now in your '/etc/network/interfaces' file you should have something like:

#allow-hotplug wlan0
iface wlan0 inet manual
        wpa-roam /etc/wpa_supplicant/wlan0.conf

        # if you do not rmmod the driver, the hardswitch is latched on after a 'ip link set wlan0 down'
        pre-up    modprobe -r ath5k
        pre-up    modprobe ath5k

        pre-up    sysctl -q -w net.ipv6.conf.wlan0.use_tempaddr=2
        up        ip link set dev wlan0 txqueuelen 10
        post-down ip addr flush dev wlan0

iface default inet dhcp
        # unfortunately ifmetric no longer works with wpa-roam
        #up        [ "x$IFACE" = "xwlan0" ] && ifmetric wlan0 5

Your '/etc/wpa_supplicant/wlan0.conf' file should look like:

ctrl_interface=/var/run/wpa_supplicant.wlan0
ctrl_interface_group=netdev
update_config=1

ap_scan=1
fast_reauth=1

network={
        [snipped]
}

network={
        [snipped]
}

You should now be able to ifup/ifdown to your hearts content, those chasing after a GUI should add themselves to the 'netdev' group and install 'wpagui'.

With older kernels you used to have to pass 'model=fujitsu' to the 'snd_hda_intel' module, however now-a-days it seems to just work fine; if you of course overlook the fact Fujitsu put a naff tinny mono speaker on it

Works fine however there is no DMA mode, just PIO mode (not surprising though).

sdhci:slot0: Unknown controller version (2). You may experience problems. 
mmc0: SDHCI at 0xfc402800 irq 18 PIO 
mmc0: new SD card at address 1234 
mmcblk0: mmc0:1234 SD02G 1992704KiB 
mmcblk0: p1 

Some speed statistics, bear in mind the CPU becomes busy and even the mouse jerks around the screen when I do the following:

berk:/home/alex# dd if=/dev/mmcblk0 of=/dev/null bs=1M count=100 
100+0 records in 
100+0 records out 
104857600 bytes (105 MB) copied, 42.258 s, 2.5 MB/s 

I had a brief go at forcing DMA mode in the sdhci.c (using `SDHCI_QUIRK_FORCE_DMA`, `SDHCI_QUIRK_32BIT_DMA_ADDR` and `SDHCI_QUIRK_32BIT_DMA_SIZE`) source code for the controller but had no joy; let me know if anyone finds out something.

The smartcard reader works with the o2scr driver, however there are some problems with it. I spoke to the authors on the mailing list and the following patches make the thing functional:

$ svn co http://svn.gna.org/svn/o2scr/trunk o2scr 

If you then patch up the code as per the following thread on the o2scr-main mailing list.

• missing archived emails
  • patches
    • first fix
    • second fix

Bear in mind, that smartcards under Linux is currently not too great. Brace yourself for some work on this, once I have fully gotten a working setup into place, I will slap some instructions on my site under a seperate section.

A list on some useful smartcard related pages:

• Using an OpenPGP SmartCard
• Using the OpenPGP card with subkeys

Hibernate and Suspend-to-RAM generally works fine.

N.B. the BIOS that came with my laptop (1.04) had a bug so that about 30% of the time on a resume after a Suspend-to-RAM event would fire off the LID event once every second; effectively killing any ACPI functionality afterwards. A reboot would be needed to fix it. When I upgraded to 1.08, things started to work fine.

I like to Suspend-to-RAM when I close my laptop lid, so I use the following:

# cat <<'EOF' > /etc/acpi/local/lid.sh.post 
#!/bin/sh

# dependencies: acpi-support-base
. /usr/share/acpi-support/power-funcs

RUNLEVEL=$(/sbin/runlevel | awk '{ print $2 }')

if [ "$RUNLEVEL" = "0" ] | [ "$RUNLEVEL" = "5" ]; then
  exit 0
fi

if [ $(cat /proc/acpi/button/lid/LID/state | awk '{ print $2 }') != "open" ]; then
  pkill ^ssh$

  for x in /tmp/.X11-unix/*; do
    displaynum=`echo $x | sed s#/tmp/.X11-unix/X##`
    getXuser;
    if [ x"$XAUTHORITY" != x"" ]; then
      export DISPLAY=":$displaynum"

      LVDS=$(xrandr --dryrun --query | grep '^LVDS' | cut -d\  -f 1)
      VGA=$(xrandr --dryrun --query | grep '^VGA' | cut -d\  -f 1)

      su $user -c xrandr -- --output ${LVDS} --auto --preferred --primary
      su $user -c xrandr -- --output ${VGA} --off

      su $user -c xautolock -enable
    fi
  done

  for x in `cat /proc/net/wireless | awk 'BEGIN { FS="[: ]" } /:/ { print $2 }'`; do
    ifdown $x &> /dev/null | true
  done

  sync

  echo mem > /sys/power/state
fi

exit 0
EOF

# chmod +x /etc/acpi/local/lid.sh.post

Okay, if we overlook the bouncing keyboard and the crappy tinny speaker, the only other problem with the laptop is that the ACPI bits suck:

• CPU temperature is hardcoded to 300K
• if the CPU is even slightly busy and on AC power, the fan just will not shut up

So to fix up the DSDT, you need to compile a custom kernel (I actually could not be bothered with the Debian kernel and now have gone with the vanilla/git kernels) annoyingly as Debian refuses to add the very helpful feature of loading the DSDT from initramfs:

You will need the ACPI DSDT in initrd patch.

N.B. provide fixed up DSDT once I have finished fiddling :)

Now the DSDT playing:

# aptitude install iasl
$ mkdir /usr/src/dsdt
$ cd /usr/src/dsdt
$ cat /sys/firmware/acpi/tables/DSDT DSDT.aml
$ iasl -d DSDT.aml
$ patch < fujitsu-t2010-dsdt-fixup.diff
$ iasl DSDT.dsl
# cp dsdt-v1.04.aml /etc/initramfs-tools/DSDT.aml

PowerTOP alas will only show me getting down to about 16 (no-X11) to 26 (with X11) wakeup's per second which is a bit naff, however to get this you need to do the following (remember to turn off bluetooth with the physical switch too when not using it).

Tweak your '`/etc/rc.local`' file to include:

/sbin/sysctl -q -w vm.dirty_writeback_centisecs=1500
/sbin/sysctl -q -w fs.xfs.xfsbufd_centisecs=3000
echo 1 > /sys/devices/system/cpu/sched_mc_power_savings 
echo 10 > /sys/module/snd_hda_intel/parameters/power_save 

N.B. ignore the powertop usbcore autosuspend stuff, it does not know what it is talking about.

Annoying, according to PowerTOP, XFS adds a wakeup for every XFS mounted filesystem which shows up as the process xfsaild. I annoyingly have 19 LV, excluding my root filesystem, XFS filesystems so I'm annoyed, but what can I do.

I used the automounter ('autofs5') to reduce this problem so I could still abuse LVM to my liking. I added the following to my `/etc/auto.master` file:

/mnt/auto.xfs /etc/auto.xfs --timeout=300

My `/etc/auto.xfs` file looks like:

backupninja     -fstype=xfs,rw,relatime,nodev,nosuid,noexec             :LABEL=backupninja

src             -fstype=xfs,rw,relatime,nodev,nosuid                    :LABEL=src
ts7800          -fstype=xfs,rw,relatime,nodev,nosuid                    :LABEL=ts7800
wag54g          -fstype=xfs,rw,relatime,nodev,nosuid                    :LABEL=wag54g

bitbucket       -fstype=xfs,rw,relatime,nodev,nosuid,noexec             :LABEL=bitbucket
photos          -fstype=xfs,rw,relatime,nodev,nosuid,noexec             :LABEL=photos

old             -fstype=xfs,ro,relatime,nodev,nosuid,noexec             :LABEL=old

You then can comment out the entries from your `/etc/fstab` file that you have added to your `/etc/auto.xfs` file and then replace all your mount points with softlinks to `/mnt/auto.xfs/<NAME>`.

All seems to work fine with the tpm_infineon module:

berk:/home/alex# dmesg | grep -i tpm 
pnp: the driver 'tpm_inf_pnp' has been registered 
tpm_inf_pnp 00:03: Found TPM with ID IFX0102 
tpm_inf_pnp 00:03: TPM found: config base 0x4e, data base 0xfd00, chip
version 0x000b, vendor id 0x15d1 (Infineon), product id 0x000b (SLB 9635 TT 1.2) 
tpm_inf_pnp 00:03: driver attached 

Using the TPM is outside the scope of this page. The Infineon TPM chip series supposedly generates True RNG which means we could use it as an entropy source through rng-tools.

For the sake of completeness, the single nagging device that I'm yet to find is the accelerometer on the laptop. The last place I can think of is the I2C bus however I guess I should spend more time digging through the ACPI DSDT.

An initial nosey on the i801 smbus (SMBus I801 adapter at 18e0) gives me:

berk:/home/alex# i2cdetect -y 0 
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f 
00:          -- -- -- -- -- 08 09 0a 0b -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- 19 -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- 44 -- -- -- -- -- -- -- -- -- -- -- 
50: 50 51 52 -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- 69 -- -- -- -- -- -- 
70: -- -- -- -- -- -- -- -- 

From my guess work I have found by using “i2cdump -y 0 0x??” where '??' is the address I'm probing and also noseying in '/sys/bus/i2c/devices/':

• 0x19 : the Fujitsu Application Panel - module 'apanel'
• 0x50 : the TFT screen (BOE HYDIS HV121WX6-100) - module 'eeprom'
• 0x51 : one of my 1GB DIMM's (70T2953EZ3-CE6) - module 'eeprom'
• 0x52 : the other of my 1GB DIMM's (70T2953EZ3-CE6) - module 'eeprom'

As for the rest, no idea what they are however the lm-sensors website helps fill in some of the blanks potentially:

• 0x09 : battery charger
• 0x0a : battery manager
• 0x0b : smart battery
• 0x69 : clock

• Misc files regarding the Fujitsu T2010 under Linux:
  • DSDT AML Files
    • BIOS version 1.04
    • BIOS version 1.08
• Quieting The Fan - this works amazingly!
  • you will need to purchase a Fujitsu T4220 Fan
  • you need to also buy some thermal grease, I actually went all Artic Silver 5
  • now you can go ahead and disassemble the Fujitsu T2010
• Fujitsu 3D Shock Sensor
  • Fujitsu Shock Sensor Reverse Engineering
  • after replacing the fan on my T2010, I noticed the accelerometer is a HAAM-325B
  • Fujitsu LifeBook Shock Sensor on Linux
• Really there was only one site that was interesting in setting this all up, Linux on the FSC Lifebook E8210, for the smartcard section