It's not necessarily a current flow mechanism -- electron current works just fine in metals, after all.
The trick is that a semiconductor's electrons are stuck in the valence band, immobile. When you give them some minimum energy (the band gap energy), they are temporarily free to move (the energy promotes them to the conduction band). As it happens, the holes thus formed in the valence band, are also free to move, though not quite as easily (hole mobility in silicon is about 2.5 times poorer than electron mobility; most materials have a much worse ratio, like GaAs where hole mobility is something like 20 times worse -- which is why GaAs FETs are always N-channel).
Dopants add new energy levels within the band gap. This allows thermal energy (~26meV, versus gap ~1eV) to promote electrons away from the valence band, thus creating holes -- P type doping. This also works for dopant levels near the conduction band, giving free electrons -- N type doping (though I forget why you don't need free electrons up there already?).
Holes and electrons can coexist just fine. There is a tendency for them to recombine, of course; this occurs spontaneously (when both are present), or at dopant atoms or defects (where the energy barrier is smaller = more probable).
If nothing else, there is an intrinsic doping. This is due to thermal energy very rarely kicking electrons to the conduction band, without dopants. Rare, as in, a chance on the order of 1e-10 per atom.
This effect, and the diffusion rate of free charges, limits the distance they can travel, to about a dozen micrometers in silicon. This means you can't make a BJT (or at least, a worthwhile one) over longer distances -- it's no surprise that it took so long for transistors to be developed, as there was no good way to position point-contact whiskers this precisely (you can make a schottky transistor with metal whiskers and a homogeneous doped crystal).
There's also field effect behavior, where an electric field draws charges to the surface. This is absolutely normal, for instance in metals it causes some electrons to move away from the positive end and towards the negative end; but metals have a huge excess of free electrons, so the surface is always just as conductive as ever. Only with semiconductors, where the charge density is small enough to be useful, can you see the field influencing material conductivity. But only a very thin surface layer (within the Debye scattering length, give or take) is affected. Surfaces are also sensitive to contamination, and chemical purity (not just the bulk material, but every single thing that ever touches that surface!) has to be very good. This is why it took until the 70s to commercialize power MOS and CMOS ICs -- it took decades for the chemistry, materials science, and processing equipment to develop to the point where such fine structures could be built up to practical, macroscopic devices.
Tim