XPS M170 LED lights : howto control them under linux

[Michael E Brown]

On Tue, Mar 13, 2007 at 07:18:52PM -0700, Ed H wrote:
> Hi Michael (and Andreas)
> 
> I've attached a demo command-line program based on the dellWirelessCtl which
> seems to be doing the trick on my XPS Gen2 system, please feel free to use it however
> you see fit (I'm happy with whatever licensing the rest of the libsmbios package uses
> and hereby relinquish all rights to the attached software to Dell). Apparently the media
> LEDs aren't supported on my system so I didn't add the command-line args to program
> them. I stole the color names from the LightFX SDK and hard-coded them in as a macro
> list.

I've added libsmbios-devel into the cc-list.

Very cool. Awesome turnaround time. I dont have a system to test on yet, myself. I'll put this into the bins/ directory where I keep the "unsupported" stuff. After I get some testing in on it, I'll probably move it over to the supported-bins/.

The rest of libsmbios is GPL/OSL dual-licensed, so thanks for the license grant.

I'll add this into a beta for the 0.13.5 and post an announce on libsmbios-devel. I would be most pleased if you could join that mailing list.

> First off thanks again for releasing this info, I'm sure the linux community will make
> good use of it.
> 
> I've noticed performance isn't great (I timed it at ~4.6ms/update with a 0 ns nanosleep), I
> assume this is something with either the hardware or the dcdbas driver, right? Is there a
> reason dcdbas does things via the /sys filesystem vs. using an ioctl? No biggie for me (I wanted

It is using a SMI to set the LEDs, so I am pretty sure the performance is in the tubes because of that, not because of anything inherent in the dcdbas code. The way a SMI works is by trapping into BIOS code. The BIOS generally takes its sweet time about things.

> to be able to update the LEDs for system monitoring/debugging purposes), but I imagine
> games/visualizations are going to want somewhat better perf if possible. Assuming the
> transaction could be made asynchronously at the low level (i.e. queued update only with
> no return value) this may not be an issue.

I suppose you could do the LED update in a separate thread.

> Andreas, how do you feel about working on a /sys/class/leds driver now that Michael
> as shown us how the LEDs can be programmed?

> P.S. Michael, I had a lot of trouble with the old build/ltmain.sh, I ended up having to copy
> my system one in order to get automake etc. working for me:

My development system is sort of old, so this may be the problem. I do my development on a RHEL 3 box.

You may check to see if running "autoreconf" helps. I just converted over to autotools a couple months ago, and don't have that much experience with it, so it is entirely possible that I screwed something up.
--
Michael