Sounds made using the glottalic egressive airstream mechanism (one of the six main airstream mechanisms, and also of the four that are actually phonemic in natural human languages) are known as ejectives; these are produced by pushing air upwards and outwards with the glottis. One frequently hears that ejectives can only ever be voiceless, as voicing an ejective would require opening the glottis, allowing the air above to escape through the glottis and preventing it from inducing the necessary airflow movement.
This argument would, intuitively, seem to apply to all glottalic sounds (both egressive and ingressive), not just ejectives; after all, if air can escape downwards through an upwards-moving glottis, it should also be able to escape upwards through a downwards-moving glottis.1 However, glottalic ingressives (known as implosives), far from obligatory voicelessness, are almost universally voiced, with the descending glottis being left slightly open so that a thin stream of air can pass upwards through it (often assisted by some degree of pulmonic egressive airflow coming up from the lungs) and vibrate the vocal cords.2 There seems to be no obvious reason one couldn't do that with an ejective (which is, after all, an implosive in reverse), with an almost-closed-but-slightly-open glottis moving upwards against the air column and some of the pressure above the glottis passing downwards through it to vibrate the vocal cords and voice the ejective (possibly assisted by a bit of pulmonic ingressive airflow helping to pull air down through the glottis).
Why is it that, apparently, the glottis can be slightly open for voicing and still produce sound when it's moving down, but has to be tightly closed in order to produce sound when it's moving up?
1: Several parts of the human body do have check valves that mean fluids can go one way through but not the other way; however, the respiratory system is check-valve-free.
2: In fact, with voiced implosives, most of the air movement is actually air moving upwards3 through the glottis, with only a small contribution from air being dragged downwards behind the glottis.
3: Upwards relative to the descending glottis, that is; in actuality, what happens is mostly the glottis moving downwards over a near-motionless air column.