This document describes experiences with
running Linux on
laptops such as
X40 2371-8NG (pilchard)
(and T23 (kirki)).
The suggestion has been made that they replace Sony VAIOs as the `other' supported laptops.
It is reported that even different series IBM ThinkPads are fairly similar.
As such, it might be worth looking at reports on
and X30 machines such as
as well as X40 pages such as
This page was started the day the first machine arrived,
and is developing.
When it is working sufficiently (or some other laptop breaks),
it will be given to its user,
so I will no longer have access to it.
This report is listed at
TuxMobil - Linux on laptops, notebooks, PDAs and mobile phones
The X40 comes with a 1200MHz Pentium M
with a 64KB/1MB cache, 512MB onboard, 12.1" XGA-TFT 1024x768, 40GBi.
The machine comes with no external drive,
i.e. no floppy or CD.
It appears that it expects a USB floppy and CD.
It does a PXE boot and boots off a generic USB floppy, but not a CD.
- CPU: 1200MHz Pentium M
with 1MB of Level 2 cache.
- Memory: It comes with 512MB
It has a Hitachi DK13FA-40B 40GBi disk,
but no floppy or CD,
although it can take a USB floppy or CD (but can't CD boot).
hdparm manages 20MB/s using udma5.
- niggle: Graphics:
It uses the 82852/855GM chipset (i830) which uses main memory.
- Pointing devices:
It has an IBM ThinkPad pointer, and three mouse buttons.
- Card Slots:
It has one typeII Cardbus slots and an SD slot.
- Builtin Comms:
It has a built in
e1000 (needs 2.6.5),
802.11abg NIC (rev 01) Mini-PCI WiFi card
(not tried yet)
V.90 56k Winmodem
(not tried yet)
It measures 10.5x8.3x.94" and weighs 2.7lb.
- tcpa security chip.
They claim up to ? hours for the standard battery.
On `default' (performance?) settings, it appears to use
12W to 5W,
giving up to 10h
using a 62Wh battery.
It discharges linearly down to 0 (starts beeping at 5%),
where it remains for 14m.
It charges linearly at around 30W up to 80%, at which point it slows down for a bit, speeds up, flattens off and then charges again - some smarts going on there.
Total charge time (while running) is 3.75h.
Came with a FAT32 partition using most of the disc,
followed by an IBM "recovery" partition.
The latter can be booted by pressing F11 during boot
so long as it has original IBM MBR hasn't been overwritten.
It is said that the Restore operation resets the WHOLE DISK.
On setting up WXP, the partition is changed from FAT32 to NTFS,
with the result that Fn-F12 doesn't see a FAT FS, so cannot
hibernate (WXP uses SW hibernate, so isn't effected).
During POST, pressing F12 causes it to generate a list of boot devices,
and allows the user to select one.
I booted WXP, and installed PQMagic and shrank the NTFS FS from 39GB to 12GB.
I tried a Knoppix boot, but it failed to find the FS,
I used a PXE boot to get an NFS client running,
then did a `std Lab' NFS install.
The BIOS sets the heads to 240 rather than 255.
Left the NTFS and FAT32 FSs ASIS, and allocated a 4GB primary hda3 for root,
with the rest assigned to an extended hda4 partition.
It all `just worked'.
`lspci' knows about most of the devices.
The installed kernel could not load the e1000 driver, so I used a 3C589C PCMCIA card.
`just worked' for X40.
The T23's drive didn't boot from it.
Work with an external USB drive (Sony DRX-500UL).
A std local "FC3 image" instaleld onto md3 worked (so long as the uni kernel was booted).
UTP "just worked".
X still had "junk at top" problem.
madwifi worked (when installed)
With "ehci_hcd" HACK and "acpi_sleep=s3_bios", shutsdown OK, but on restart, TTY VCs are green boxes, and the it decided to (cleanly) shut down, rather than resume.
"acpi=off" uses APM, which works.
G3 needs work still?
The RH9 2.4.20 kernels appeared not to be able to drive the Intel Corp. 82541GI Gigabit Ethernet Controller
whereas the Fedora 2.6.5 just worked.
This also gave a usable ACPI kernel, but this breaks suspend.
kernel: Fedora 2.6.5 for ACPI and UTP
To install the driver modules, the kernel source RPM was needed,
along with a copy of the madwifi CVS module1.
gcc also needed upgrading so that the modules were built with the same version as Fedora used for the kernel,
and the Mandrake wireless-tools RPM.
Use `COPTS=-DSOFTLED' to enable the LED.
The device is ath<N> rather than the more usual eth<N>.
UNRELIABLE: WiFi: madwifi
`iwlist scanning' displays details of the available connections.
`iwpriv ath0 mode <N>' sets auto, a, b, g for <N> = 0..3.
For debugging, try
`sysctl -w dev.ath.debug=1',
`sysctl -w dev.ath.dump="hal"',
kismet 4 is needed.
Although it may work EVENTUALLY,
much of the time it reports `unable to reset channel 12 (2467MHz)',
and fails to get a DHCP address.
Matthew Garrett has a page on
how to get wireless to work.
After some sort of sleep, the screen is shifted down so that the bottom is no longer visible.
user niggle: display shifted after sleep
De- and then re- selecting the X screen fixes the problem.
This can be automated if ACPI is used, by adding a file
containing the lines
and arranging that /usr/bin/acpi-lid contained the line
`chvt 1; chvt 7' to switch away from X, and back again.
this is a show stopper for one of the user's main requirements, which is to use its
VGA output while lecturing.
Matthew Garrett suggested
which enables a separate pipe for the VGA output (e.g. `sudo i855crt on rawpipe' or `sudo i855crt on 1024x768@70' - see i855crt.conf for the known modes),
which uses the same pipe
(e.g. sudo i810switch crt on),
(built in /usr/groups/linux/extra-packages/ThinkPad).
Note that neither can manipulate the LCD, so do not boot with a CRT attached
(setting the BIOS `Config -> Display -> Boot Display Device' to `ThinkPad LCD' or `BOTH' does not work).
It's been suggested that using APM rather than ACPI fixes it - doesn't appear to work for me.
If all else fails, reboot with VGA connected.
The old APM methods rely on the BIOS to do all the work, whereas the newer ACPI methods rely on the kernel doing it all.
suspend (not yet hibernation)
When suspended, the right most "crescent" light comes on.
To wake it up, press the "on/off" button for about 3s.
By setting MODULES to ehci_hcd in /etc/sysconfig/acpi-sleep,
`sudo acpi-sleep' will unmount the ehci_hcd kernel module before writing
3 to /proc/acpi/sleep, and re-insert it when the machine wakes up.
Sometimes a machine gets into a state where the display does not come back on after waking up.
This is cleared by powering off and back on - subsequent suspends work.
Adding acpi_sleep=s3_bios to the kernel boot flags may be needed so
that the display comes back on after suspension.
If booted using APM, Fn-F4 and `apm -s' will suspend-to-RAM.
This takes a while (nothing displayed on the screen while it does it),
but once done, the `crescent moon' is displayed.
Fn-F4 (the user reports that pressing the power button makes it shut down) returns power.
Think also about
or IBM's utility.
`setkeycodes e06a 126 e069 127' to cause the extra cursor keys to
generate more normal keycodes, then
`echo keycode 126 = Decr_Console | loadkeys;
echo keycode 127 = Incr_Console | loadkeys'
to make them step through the VCs.
Using xev it appears that the extra cursor keys generate keycodes 116 and 117 (Super_R and Menu).
If of use, they can be mapped using
`xmodmap -e "keycode 116 = F19"' or similar.
I installed the Mandrake
and manually added the xinerama .so's.
tpb - access to ThinkPad buttons
`tpb -osd=on -verbose -thinkpad="xterm"&'
causes it to
display the speaker, brightness, zoom and light setup when the buttons are pressed,
log misc other events,
and start xterm when the `Access IBM' button is pressed.
Use 2.9.x from
user -> I agree -> Linux drivers -> The latest beta Linux driver
TO BE DONE: modem
Not yet tried: BT, IrDA, SD
- ... module1
cvs -d:pserver:email@example.com:/cvsroot/madwifi login;
cvs -z3 -d:pserver:firstname.lastname@example.org:/cvsroot/madwifi co madwifi