|
|
| (9 intermediate revisions by 2 users not shown) |
| Line 1: |
Line 1: |
| − | == ThinkPad T43 2687-DDU setup for Debian Etch == | + | == Who is Hmh? == |
| | | | |
| − | This is an ongoing, spotty guide for an advanced {{Debian}} setup of a {{T43}}, focused on using absolutely all of its capabilities that can be used. It is not an installation guide, but rather a configuration guide. Refer to {{Debian}} for a list of pages with initial installation instructions.
| + | I am a Debian Developer for quite a few years now, and the current maintainer of [[ibm-acpi]]/[[thinkpad-acpi]], aka "ThinkPad ACPI Extras" Linux kernel driver for ThinkPad laptops. |
| | | | |
| − | When it is really usable, I will move it to some other far more public place in [[ThinkWiki]].
| + | I am also involved with the [[tp_smapi]] and [[HDAPS]] Linux drivers, as a beta-tester and power-user. I don't list myself as a developer for those projects because I have never contributed a single line of code to them, just suggestions and bug reports ;-) |
| | | | |
| − | Installing should be done using the latest Debian Etch debian-installer release. Anything else simply will not run on the ThinkPad.
| + | == My ThinkPads == |
| | | | |
| − | === Kernel ===
| + | I own just a simple [[:Category:T43|T43]] model 2687-DDU. It is a great machine: a very solid ThinkPad, with its own set of annoying quirks just like any other IBM-era ThinkPad. The SATA-PATA bridge, tg3 noise at full network load, and lack of support for AHCI being the worst ones IMHO. |
| | | | |
| − | Baseline kernel: Debian's latest 2.6.16 (2.6.16.17 with stabilization patches) {{NOTE|Thinkpad support is an ongoing, bleeding edge effort. One ends up needing to apply a ton of patches and to use the latest available kernel, sometimes even kernels not yet released}} {{HINT|The [http://members.optusnet.com.au/ckolivas/kernel/ Con Kolivas (ck) desktop kernel] patches might be very applicable to a ThinkPad Kernel}}
| + | == ping == |
| | | | |
| − | ==== Patches employed ====
| + | This is a ping edit, to let Akw know I am still alive and to not purge my account ;-) |
| − | * Latest ipw2200 driver and ieee80211 layer:
| |
| − | :There are two ways to go about this: either '''disable''' IEEE 802.11 support (and thus ipw2200) in the kernel and build both modules out-of-tree (failure to do so will break the out-of-tree build), or update the kernel subsystems with more up-to-date files.
| |
| − | :The advantage of updating the kernel drivers instead of doing an out-of-tree build is that should there be any configuration options that require ieee80211 support, those will be available. On the other hand, it is a bit more difficult, as you will have to mess with Kconfig.
| |
| − | :* copied files over the in-kernel files in {{path|include/net}}, {{path|drivers/net/wireless}}, {{path|Documentation/networking}}, {{path|net/ieee80211}}, thus replacing the old drivers with the new ones.
| |
| − | :* ipw2200 often have extra options you need to enable, look in the toplevel Makefile and add them to {{path|drivers/net/wireless/Kconfig}}. They are usually bool types, so it is trivial to copy, e.g. IPW_DEBUG and rename to the new required options.
| |
| − | * [[Problems with SATA and Linux#Hang on resume from suspend to RAM|SATA power-management]]
| |
| − | * [[HDAPS|freeze-block-queue patch (for HDAPS)]]
| |
| − | * [http://acpi.sourceforge.net/download.html ACPI update patches]
| |
| − | * [[Problems with ACPI suspend-to-ram#Troubles on resume|ACPI shutdown on resume patch]]
| |
| − | * [[Tp smapi|TP SMAPI support]]
| |
| − | * [[Software Suspend 2]]
| |
| − | * [[Conexant HSF modem drivers#The Linuxant hsf driver|Linuxant Conexant HSF modem driver]]
| |
| − | * [[How to make use of IrDA#ISA PnP patch|nsc-ircc ISA PNP patch]]
| |
| | | | |
| − | ==== Kernel setup tasks ==== | + | == Links == |
| − | # Make sure the following subsystems are modules (for suspend/resume)
| |
| − | ## USB (especially EHCI_HCD and UHCI_HCD)
| |
| − | ## HDAPS (may work compiled in, as well)
| |
| − | # Make sure the following subsystems are enabled and not modules (safety net)
| |
| − | ## ThinkPad SMAPI
| |
| − | # Place the following modules on /etc/modules:
| |
| − | ## hdaps
| |
| − | ## tg3
| |
| − | ## ipw2200
| |
| − | # [[How to make use of IrDA#Kernel configuration|Configure IrDA module options]]
| |
| − | # ipw2200 firmware needs to go in {{path|/lib/firmware}}. The module should be given the led=1 module parameter option
| |
| − | # If you do not have a dock, you can and should get rid of the legacy floppy support in the kernel. Disable CONFIG_BLK_DEV_FD for a ~0.5s speedup on boot time.
| |
| | | | |
| − | === Userland ===
| + | * [[ibm-acpi|The Linux kernel ibm-acpi driver, which I currently maintain]] |
| − | | + | * [http://people.debian.org/~hmh/ My Debian.org homepage with some old stuff] |
| − | ==== Complex configuration issues ====
| |
| − | | |
| − | You want to install and configure to your liking:
| |
| − | *smartmontools
| |
| − | *kde/gnome power management thingies
| |
| − | *X.org power management on the video driver
| |
| − | | |
| − | ==== Simple configuration issues ====
| |
| − | | |
| − | These are fairly simple minor things you might want to do:
| |
| − | | |
| − | ===== /etc/sysctl.conf =====
| |
| − | | |
| − | install the procps package. Add the following lines to /etc/sysctl.conf:
| |
| − | | |
| − | {{FIXME}}
| |
| − | | |
| − | ===== /etc/sysfs.conf =====
| |
| − | | |
| − | install the sysfsutils package. Add the following lines to /etc/sysfs.conf
| |
| − | | |
| − | # ThinkPad battery charger defaults
| |
| − | devices/platform/smapi/BAT0/start_charge_tresh=40
| |
| − | devices/platform/smapi/BAT0/stop_charge_tresh=85
| |
| − | devices/platform/smapi/BAT1/start_charge_tresh=70
| |
| − | devices/platform/smapi/BAT1/stop_charge_tresh=90
| |
| − |
| |
| − | # Default frequency scalling governor
| |
| − | devices/system/cpu/cpu0/cpufreq/scaling_governor=conservative
| |
| − | | |
| − | {{NOTE|Using the '''ondemand''' governor can do a number on your CPU power regulator, the '''conservative''' governor is a much better choice overall, and in fact is good enough at what it does that no userspace power governor daemon is needed, or even desired}}
| |
| − | | |
| − | ==== udev rules ====
| |
| − | | |
| − | # udev extended network interface naming needs to be configured, order of tg3/ipw2200 won't be respected if {{path|/etc/modules.conf}} is read after udev did the PCI coldplugging. Debian does this automatically, if you are lucky... check {{path|/etc/udev/rules.d/*persistent-net.rules}}.
| |
| − | # The thinkpad driver takes a long time to load and times-out udev every time. To shed off this timeout from startup time, edit {{path|/etc/udev/rules.d/020_permission.rules}} and change the line containing '''WAIT_FOR_SYSFS="bus"''' to also have the following conditions: '''BUS!="serio", SUBSYSTEM!="serio"''''
| |
| − | | |
| − | ==== Kernel/module configuration ====
| |
| − | | |
| − | This is done in /etc/modprobe.d/ files. Install the setserial package.
| |
| − | | |
| − | Add a /etc/modprobe.d/thinkpad-t43 file there with this content:
| |
| − | # Intel Wireless, default to radio offline for safety and security
| |
| − | options ipw2200 disable=1 led=1 hwcrypto=1
| |
| − |
| |
| − | # IrDA
| |
| − | alias irda0 nsc-ircc
| |
| − | options nsc-ircc dongle_id=0x09 io=0x2f8 irq=3 dma=3
| |
| − | install nsc-ircc /bin/setserial /dev/ttyS1 uart none port 0 irq 0; /sbin/modprobe --ignore-install nsc-ircc
| |
| − | | |
| − | References:
| |
| − | * [[How to make use of IrDA]] | |
| − | | |
| − | ==== laptop-mode-tools ====
| |
| − | | |
| − | Install and configure laptop-mode-tools to your liking. If you really want to protect your HD when moving the ThinkPad around, laptop-mode-tools can help that dramatically by leaving the drive spun-down and parked most of the time.
| |
| − | | |
| − | ==== Suspend2 ====
| |
| − | | |
| − | On top of what is usually done for ThinkPads, remember to set the acpi_sleep=s3_bios kernel option if you have an ATI GPU. Suspend to RAM should be done using the default mode in ram.conf (RediSafe-like suspend-to-ram-with-backup-on-disk is possible, using Suspend2... but it takes 10 times more to suspend).
| |
| − | | |
| − | If you can do the suspend to ram without unloading any modules, it is much much faster. This heavily depends on the kernel, as a lot of drivers are still quite buggy.
| |
| − | | |
| − | # Suspend to RAM mode reminders
| |
| − | UseSysfsPowerState mem
| |
| − | SaveClock restore-only
| |
| − | Unmount /media/*
| |
| − | IbmAcpi yes
| |
| − | | |
| − | # Suspend to disk mode reminders
| |
| − | UseSuspend2 yes
| |
| − | Unmount /media/*
| |
| − | UnmountFSTypes vfat msdos cifs smbfs ntfs nfs
| |
| − | IbmAcpi yes
| |
| − | | |
| − | ==== acpi events ====
| |
| − | | |
| − | To setup the hotkey mask, {{path|/etc/default/acpi}} can be abused (it is a shell script fragment).
| |
| − | Just append this fragment to that file:
| |
| − | # ThinkPad ACPI setup
| |
| − | [ -w /proc/acpi/ibm/hotkey ] && echo enable,0x81c >/proc/acpi/ibm/hotkey
| |
| − | :
| |
| − | | |
| − | ===== For suspend-to-ram (sleep) =====
| |
| − | | |
| − | {{path|/etc/acpi/actions/sleep.sh}}
| |
| − | #!/bin/sh
| |
| − |
| |
| − | # Check the reason we are running
| |
| − | case "$1" in
| |
| − | button/lid)
| |
| − | # do not run with the lid open
| |
| − | grep -q open /proc/acpi/$1/$2/state && exit 0
| |
| − | ;;
| |
| − | esac
| |
| − |
| |
| − | exec /usr/sbin/hibernate -F /etc/hibernate/ram.conf
| |
| − | | |
| − | To sleep when the lid is closed:
| |
| − | | |
| − | {{path|/etc/acpi/events/lid}}
| |
| − | event=button[ /]lid
| |
| − | action=/etc/acpi/actions/sleep.sh %e
| |
| − | | |
| − | To sleep when fn+f4 is pressed:
| |
| − | | |
| − | {{path|/etc/acpi/events/ibm_fn_sleep}}
| |
| − | event=button[ /]sleep
| |
| − | action=/etc/acpi/actions/sleep.sh %e
| |
| − | {{path|/etc/acpi/events/ibm_fn_f4}}
| |
| − | event=ibm[ /]hotkey HKEY 0+80 0+1004
| |
| − | action=/etc/acpi/actions/sleep.sh %e
| |
| − | | |
| − | ===== For hibernation (suspend-to-disk) =====
| |
| − | | |
| − | Requires that ibm-acpi has hotkeys enabled, mask 0x0800.
| |
| − | | |
| − | {{path|/etc/acpi/events/ibm-fn-suspend}}
| |
| − | event=ibm[ /]hotkey HKEY 0+80 0+100[cC]
| |
| − | action=/usr/sbin/hibernate
| |
| − | | |
| − | ===== Radio software switch (rf_kill) =====
| |
| − | | |
| − | If your radios have rf_kill support, the following script can be used to toggle them on or off:
| |
| − | | |
| − | {{path|/etc/acpi/actions/toggle_radios.sh}}
| |
| − | #!/bin/sh
| |
| − |
| |
| − | set -e
| |
| − |
| |
| − | cd /sys/class/net
| |
| − | state=0
| |
| − |
| |
| − | # Find all radios, set state to the first one we find
| |
| − | for i in * ; do
| |
| − | if [ -r "$i/device/rf_kill" ] ; then
| |
| − | state=$(cat "$i/device/rf_kill")
| |
| − | break
| |
| − | fi
| |
| − | done
| |
| − |
| |
| − | if [ $state != "0" ] ; then
| |
| − | state=0
| |
| − | txtstate="Enabling radio"
| |
| − | echo 3 >/proc/acpi/ibm/beep || true
| |
| − | else
| |
| − | state=1
| |
| − | txtstate="Disabling radio"
| |
| − | echo 12 >/proc/acpi/ibm/beep || true
| |
| − | fi
| |
| − |
| |
| − | for i in * ; do
| |
| − | if [ -w "$i/device/rf_kill" ] ; then
| |
| − | echo $state > "$i/device/rf_kill"
| |
| − | echo $txtstate $i
| |
| − | fi
| |
| − | done
| |
| − |
| |
| − | :
| |
| − | | |
| − | You could also improve the script to manipulate the bluetooth RF switch, see [[ibm-acpi]] for more information.
| |
| − | | |
| − | To bind that script to fn+f5 (requires ibm-acpi hotkeys enabled, and mask 0x0010):
| |
| − | | |
| − | {{path|/etc/acpi/events/ibm_fn_f5}}
| |
| − | event=ibm/hotkey HKEY 0+80 0+1005
| |
| − | action=/etc/acpi/actions/toggle_radios.sh
| |
| − | | |
| − | === Software that needs packaging ===
| |
| − | {{NOTE|If nobody beats me to it, I will probably package them for official upload to Debian, and official Debian Sarge backports. This requires that the software be properly licensed first, of course}}
| |
| − | * hdapsd (a bit crude, needs a lot of work still)
| |
| − | * ipw2200/ieee80211 need adopting, and since they clash terribly with in-kernel support, they will only work right if we either convert them to be patch-based, or to detect that the user built with in-kernel ieee80211/ipw2200 enabled and abort the out-of-tree compilation.
| |
| − | | |
| − | === Software that could benefit from re-packaging ===
| |
| − | * hsfmodem: should be easy to build out-of-tree in another machine other than the one it will be installed at
| |
| − | | |
| − | == External Links ==
| |
| − | * [http://www.debianhelp.co.uk/udev.htm udev interface naming guide] | |
| − | * [http://members.optusnet.com.au/ckolivas/kernel/ ck desktop-kernel patches]
| |
