They declare themselves as pcm class devices, with a struct snddev_info device private structure:

          static driver_t xxx_driver = {
              "pcm",
              xxx_methods,
              sizeof(struct snddev_info)
          };

          DRIVER_MODULE(snd_xxxpci, pci, xxx_driver, pcm_devclass, 0, 0);
          MODULE_DEPEND(snd_xxxpci, snd_pcm, PCM_MINVER, PCM_PREFVER,PCM_MAXVER);

Most sound drivers need to store additional private information about their device. A private data structure is usually allocated in the attach routine. Its address is passed to pcm by the calls to pcm_register() and mixer_init(). pcm later passes back this address as a parameter in calls to the sound driver interfaces.

The sound driver attach routine should declare its MIXER or AC97 interface to pcm by calling mixer_init(). For a MIXER interface, this causes in turn a call to xxxmixer_init().

The sound driver attach routine declares its general CHANNEL configuration to pcm by calling pcm_register(dev, sc, nplay, nrec), where sc is the address for the device data structure, used in further calls from pcm, and nplay and nrec are the number of play and record channels.

The sound driver attach routine declares each of its channel objects by calls to pcm_addchan(). This sets up the channel glue in pcm and causes in turn a call to xxxchannel_init().

The sound driver detach routine should call pcm_unregister() before releasing its resources.

Use a device_identify() method (example: sound/isa/es1888.c). The device_identify() method probes for the hardware at known addresses and, if it finds a supported device, creates a new pcm device which is then passed to probe/attach.

Use a custom kernel configuration with appropriate hints for pcm devices (example: sound/isa/mss.c).