4.Technical Description

  This project is phase 1 of the umbrella project described in
  "Open Sound System (OSS) for Solaris" (PSARC 2007/238)
  There are two main goals of this phase. One is to integrate OSS high
  definition audio driver "hdaudio" into Solaris to support the most
  popular audio hardware. The other goal is to keep SADA
  (Solaris Audio Driver Architecture) compatibility so that there will
  be no regression for the current SADA applications.

  Theoretically, all the HD audio hardware will be supported. We will test
  the ones supported by SADA HD audio driver to prevent regression. Several
  other popular codecs will also be covered.

  The SADA compatibility is accomplished by an emulation layer named
  "sadasupport". It is a pseudo SADA audio driver which provides SADA
  interface to upper level. From the applications' point of view, it
  is not different from other SADA drivers. On the other hand, it does
  not operate the hardware directly. Instead, it accesses the device file
  provided by the lower level OSS drivers through ldi interface. The
  following diagram illustrates the relations between SADA framework,
  sadasupport, OSS framework and OSS driver where there is a HD audio
  hardware in the system

                   |-------------|
  User level       |app1,app2 ...|
                   |------+------|
                          |(1)
                          |
                    |-----+----|              |----------------|
 -------------------|/dev/audio|--------------|/dev/private_dsp|--------
                    |-----+----|              |----+--+--------|
                          |(2)          (4)        |  |
  Kernel level            |   +--------------------+  |(5)
                          |   |                       |
 |--------------|(3)|-----+---+-|                 |---+---|(6)|-------------|
 |SADA framework+---+sadasupport|                 |hdaudio+---+OSS framework|
 |--------------|   |-----------|                 |---+---|   |-------------|
                                                      |(7)
                                                  |---+----|
                                                  |HD audio|
                                                  |hardware|
                                                  |--------|

      * (1),(2) Applications make audio requests through /dev/audio,
        which is exported by sadasupport.
      * (3) Sadasupport turns to SADA framework to fulfill the requests
      * (4),(5) Sadasupport access hdaudio driver by ldi interface when
        it needs service from hdaudio driver to complete the requests
      * (6),(7) Hdaudio turn to OSS framework to fulfill the requests
        and operate the hardware when needed.

  In phase 1, the device file "/dev/private_dsp" provided by OSS drivers is a
  project private interface and only root user is able to access it.
  No new interface will be published in this phase. Although HD audio
  driver is replaced by OSS hdaudio driver, other SADA drivers will not
  be replaced in phase 1. Drivers of the two different frameworks can work
  together. SADA framework will not be removed, either. Detailed information
  of SADA can be found in man pages: audio(7I), mixer(7I) and audio_support(7I).

  Summarily, we use OSS hdaudio driver to support HD audio hardware
  and use sadasupport to maintain SADA compatibility.

  During the discussion of the umbrella case, the issue of latency has
  been brought up. In this context, latency means the time gap between
  the time hardware/driver is changed by other applications or by external
  events and the time when applications sense it. And it should be
  achieved without high CPU usage. It can be achieved by leverage
  feature provided by SADA interface. Since SADA interface will be kept,
  latency issues is addressed naturally. Consensus was reached during
  the offline discussion.

  Another concern is about power management(PM). This project will cause
  no regression in PM. Since only HD audio driver will be delivered, there
  will be no binary released on SPARC. Therefore, SPARC systems will not
  be affected at all. For the x86 systems from Sun, the ones with non-HD
  audio hardware will not be affected by this project. And the ones with
  HD audio hardware are all supported by SADA HD audio driver which does not
  support PM. Therefore, there will be no regression, either.

  Here is the list of all kernel module and driver files:
  osscommon
  osscore
  osscore.conf
  sadasupport
  sadasupport.conf
  vmix
  vmix.conf
  hdaudio
  hdaudio.conf

  Some of the tunables provided by the .conf files are for changing
  some aspect of the OSS system. But the default value is suitable
  in most case and they rarely need to be changed. Other tunable are
  for debugging.

  All these files and the tunables are of project private stability in this
  phase. They might be changed in the following phases.

5. Reference Documents:

  Open Sound System (OSS) for Solaris (PSARC 2007/238).
  http://opensolaris.org/os/project/opensound