Fedora Core on Thinkpad X40 Customization HOWTO

A band of garbage appears at the top of the display when you close and open the lid. Keith Packard tells me this probably means the BIOS is restoring the video start address incorrectly.

The workaround is simple; Ctrl-Alt-F1 followed by Ctrl-Alt-F7 to switch to a text display and back. But the real remedy is to back down to APM (see below). This problem only manifests under ACPI.

The underlying problem is that stock FC2 enables ACPI support, which is not yet quite ready for prime time. Backing down to APM works around this problem.

To accomplish this, do the following steps:

Note that this is suspend-to-RAM, like ACPI state S3. Software suspend to disk (ACPI state S4, like the Windows hibernation feature) is a trickier problem. There are no fewer than three implementations of this (see Software Suspend For Linux for a comparison). The default swsusp feature in 2.6 does not yet work on the X40.

There are three different NICs inside X40s — the Intel Wireless Pro 2100 (g only), Intel Wireless Pro 2200 (bg), or the Atheros (abg), You can tell which you have by looking at the output of lspci(1). Below I give a recipe for enabling the Atheros 802.11abg card.

I have the Atheros 802.11abg NIC. It's supported by the MADWIFI project. Here are the steps I took to get it working:

You may find that ath0 does not reliably come up at boot time, failing with a message about the ath_pci device not being there yet and deferring initialization. If so, you'll either have to do ifconfig ath0 up moanually or put it in /etc/rc.

A program called tpb (for ThinkPad Buttons) allows you to bind Thinkpad special keys to programs. It installs and works fine on the X40. You may have either to configure it without xosd or build xosd from source; I found that some strange bug in the packaging of tpb makes it believe the xosd library is out of date even when it is not (this is in tpb 0.6.2, it could be fixed in later versions).

Running tpb -v shows that tpb is able to see the “Access IBM” key being pressed, and believes it to be the “ThinkPad” key that you bind with the --thinkpad= option.

There is an open-source tool especially for tweaking Thinkpad settings. It's called tpctl, and consists of a kernel module and two front ends. There is even a GUI front end for this program.

I was able to build and install the thinkpad.ko kernel module without difficulty by following the directions in the package README of version 4.14. However, I was not able to get the module to autoload, though I could load it manually with modprobe. Also, most tpctl requests I tried then returned an error message like this:

SMAPI BIOS error 086 ("function  not supported") -- exiting

There are a few useful exceptions. The --info-CMOS-RAM, --info-display, --info-processor, --info-UltraBay, and --info-docking options appear to return valid data. So do --info-pm and --setup-all.

But all the probes that depend on querying the SMAPI BIOS fail. I don't know whether the X40 simply lacks one or the driver can't see it. It may be that the SMAPI BIOS isn't there at all, having been replaced by ACPI. The main thing you'd normally want to do with it (suspend/resume) is supported by APM.

ACPI is partly working. Hitting the power-off button while Linux is at runlevel 3 or 5 dispatches through acpid(8) and runs shutdown -h now as it should.

Suspend/resume doesn't yet work ynder ACPI, though. By running acpi_listen(8). I was able to determine that Linux's ACPI support does see an event when the F4 button is pressed. Here is what the even notification looks like:

button/sleep SLPB 00000080 00000001

The last field is a sequence number that will increment each time. None of the other special keys seem to generate a visible event. Lid opens and closes both generate an event that looks like this:

button/lid LID 00000080 00000001

I wrote the the following script as an attempt to do a software suspend to RAM under ACPI:

rmmod uhci_hcd ehci_hcd
echo 3 >/proc/acpi/sleep && hwclock ——hctosys
echo "Resuming processes...."
modprobe ehci_hcd
modprobe uhci_hcd

Unloading the USB modules is necessary, otherwise the attempt to sleep the machine will fail because the USB driver won't suspend. With the unload, the suspend apparently succeeds (the crescent-moon suspend light on the X40 goes on) but I can't figure out how to wake the machine up afterwards. Opening the lid doesn't do it. I'm told that pressing the power button is supposed to, but I have not found this reliable.

I've heard a report that the 2.6.8rc3 kernel has much better ACPI support. Paul Ionescu says “I use kernel 2.6.8-rc3 with latest acpica patch from 17-july-2004. Prior to this patch, I had a lot of problems with restoring interrupts after suspend. And with this patch, you can resume by opening the lid or pressing the Fn button for a couple of secs.”

He continues: “For some models of thinkpads I have to add acpi_sleep=s3_bios kernel parameter in order to get proper video resume from S3. In this new kernel I only have to remove ehci-hcd to make it suspend, and I don't really need to change to textmode. An echo 3 > /proc/acpi/sleep is enough for me (T41/R50p).”

I plan to try this when 2.6.8-final comes out. For now, APM is good enough