This can be worked around in user space as well by using a nonblocking

drain and

sleeping based on how many samples are queued. Would that be the right

thing to

do or is this (or something similar) worth fixing here.

To test this I opened a pcm and set no period wakeups, played a few

samples and

then called drain, which hung indefinitely. Using a USB headset with a max period size of 128k but only 15ms of the buffer filled, having the drain time shorter without losing samples makes a blocking call feasible.

Do you mean that you tested on a patched version of USB audio driver since only three drivers support SNDRV_PCM_INFO_NO_PERIOD_WAKEUP

http://git.kernel.org/cgit/linux/kernel/git/tiwai/sound.git/log/sound?qt=gre...