Radar.

I guess you could make use of the GPS or GSM clocking, though I don't know how well these are exposed to us as end users.

Also you have the problem of being able to get a half decent duty cycle from the IR transmitter. Can its control circuitry switch it on and off fast enough to give a nice short pulse width? That would mostly determine your minimum range, I have a feeling this particular figure would turn out to be greater than the actual range of the IR diode, so probably we would be dead in the water right there.

tjsgls mentions using the audio block as the receive channel, I can't imagine the audio block being fast or reliable enough to give us a good count between emitted and reflected pulses.

It's a cool idea, but I think we couldn't pull it off due to hardware limitations. I'd love to be wrong about that though!

In short - ranging will never work.
I've got a half-designed laser ranger module that I want to get selling on ebay, and have looked into the schematics of the n900.

There is no IR reciever of any sort that can resolve time accurately enough.

One billionth of a second is 30cm error.
The LED is probably not detectable by a sensitive detector when bouncing off something 10 meters away, so you need to have an error of maybe a hundred millionth of a second to make it useful at all.

The CPU does not really have the ability to easily time events like this. At least not without extra hardware.

Sonar can in principle work somewhat.

The IR emitter is designed in functionality - it did not come along 'for free'.
It's a seperate IR LED unrelated to anything else, and is connected via a transitor or two to a IO pin on the CPU.

30 centimeters would be good enough for avoiding colliding with stuff

Btw, what about interfacing with the DSP? Is it flexible enough for such uses?