On Fri, 25 Jan 2019 at 21:04, Takashi Iwai <ti...@suse.de> wrote: > > > > > > Erm, obviously it's not enough. Each attach / detach needs to manage > > > the refcount, too, for covering the cases above. It can re-use the > > > PCM mmap_refount, though. > > > > But we've used the DMA buffer file's refcounting to manage the DMA > > buffer. So is this not enough? > > Unless you manage the PCM substream refcount (or block the state > change), the PCM stream itself can be released (or re-setup) freely. >
OK. Thanks for your comments. -- Baolin Wang Best Regards