The ACPI Daemon (acpid) With Netlink Support

Introduction

In recent linux kernels, the /proc/acpi/event interface has been deprecated. The same information (and more) is available via netlink (a way for the kernel to communicate with userspace that is usually used for networking) and the input layer (mouse, keyboard, power button, etc...). This version of acpid supports netlink and the input layer.

Current Issues

LOG_CONS

acpid opens the syslog with the LOG_CONS option. This can cause acpid related logging to appear on the console. If you are seeing acpid related messages pop up on your screen, take a look at your startup scripts (usually in /etc/rc*.d) and make sure syslogd is started prior to acpid and stopped after acpid. If anyone has any opinions on whether LOG_CONS is appropriate for acpid, I'd love to hear them. I hesitate to remove this until I get a better understanding.

Also

Be sure to check the Resolved Issues section below to see if you are experiencing a problem that has been fixed.

Latest Version

acpid-2.0.14.tar.gz - Latest version with netlink and input layer support.

• The format of the event strings is slightly different from /proc/acpi/event. I took liberties with the lid switch where we have additional info with the input layer. It would be pretty easy to make this 100% compatible with the old style if needed. This should be compatible enough that it won't break any existing conf files.
• Switched from poll() to select(). Can easily switch back if there's a good reason for poll() that I'm not aware of. select() is older and should be more compatible. poll() is easier to use and can be faster, though probably not in acpid's case.
• I cleaned up the #include's. This might cause trouble in other OS's.
• Input layer and netlink tie us to linux. Should we conditionally compile them out on non-linux systems? Yi Yang's version of this did that.

kacpimon - Kernel ACPI Monitor. This is now included in the acpid source, so there's no need to download this. It's a test program that connects to /proc/acpi/event, the input layer (/dev/input/event*), and netlink. It then displays any events it receives on the console. Be sure to run this as root. Kill acpid ("killall acpid") to make sure kacpimon can open /proc/acpi/event.

evtest - Vojtech Pavlik's input layer monitor tool.

Subscribe For Updates

If you're a package maintainer for a distro, please drop me a line and I'll add you to my distribution list. I update acpid on the 15th of each month.

Previous Versions

acpid-2.0.13.tar.gz
acpid-2.0.12.tar.gz
acpid-2.0.11.tar.gz
acpid-2.0.10.tar.gz
acpid-2.0.9.tar.gz
acpid-2.0.8.tar.gz
acpid-2.0.7.tar.gz
acpid-2.0.6.tar.gz
acpid-2.0.5.tar.gz
acpid-2.0.4.tar.gz
acpid-2.0.3.tar.gz
acpid-2.0.2.tar.gz
acpid-2.0.1.tar.gz
acpid-2.0.0.tar.gz
acpid-1.0.10-netlink6.tar.gz
acpid-1.0.10-netlink5.tar.gz
acpid-1.0.10-netlink4.tar.bz2
acpid-1.0.10-netlink3.tar.bz2
acpid-1.0.10-netlink2.tar.bz2
acpid-1.0.8ted1.tar.bz2

Links

acpid in the Debian Package Tracking System (PTS) - I've been working closely with Debian and have received a lot of excellent feedback. acpid 2.x would not be as stable as it is without Debian's help.
kacpimon in the Debian PTS
acpid in Fedora git
acpid SourceForge Project - Covers up to 1.0.10 so far.

Resolved Issues

kacpimon

kacpimon has been added as a separate project in Debian, and I need to sync up with changes that are in there.

Endless Loop

The acpi-support package in Debian has files that cause an endless loop condition. The following files should be removed from acpi-support:

• /etc/acpi/events/videobtn (note that the videobtn.sh is fine, don't delete it)
• /etc/acpi/events/video_brightnessdown
• /etc/acpi/video_brightnessdown.sh
• /etc/acpi/events/video_brightnessup
• /etc/acpi/video_brightnessup.sh

Upstream has apparently removed these already, so they will be removed from Debian soon. See Debian 556261.

Crash on Device Disconnect

Debian 522756 (HIGH) - acpid crashes when an input layer device goes away. Mikhail Krivtsov has provided a patch to handle this. However, how do we discover the connection of new devices?

Vojtech Pavlik provided the following ideas 11/6/2009:

• select() on /proc/bus/input/devices (too early, though)
• udev (seems perfect, but difficult)
• hal (seems too heavy)

After a bit of poking around, I've decided to go with inotify(7) on /dev/input. It appears to work fine (for me). I'll be releasing soon...

Missing Thinkpad Events

Debian 521280 (MED) - acpid doesn't cover all the ThinkPad events. I knew this one was coming. My original stance on this was that we'll cover those that the original kernel thinkpad support covered. I'm not sure how many that actually was. Regardless, it looks like a more generic system might be a better idea. Maybe just put out the event numbers for those we don't recognize. Ugly, but it works. There's a nice huge list of events in this bug report to start with.

<- Back to my Linux page.