kernel_optimize_test/Documentation/sound/oss/README.modules

Building a modular sound driver
================================

  The following information is current as of linux-2.1.85. Check the other
readme files, especially README.OSS, for information not specific to
making sound modular.

  First, configure your kernel. This is an idea of what you should be
setting in the sound section:

<M> Sound card support 

<M> 100% Sound Blaster compatibles (SB16/32/64, ESS, Jazz16) support 

  I have SoundBlaster. Select your card from the list.

<M> Generic OPL2/OPL3 FM synthesizer support 
<M> FM synthesizer (YM3812/OPL-3) support 

  If you don't set these, you will probably find you can play .wav files
but not .midi. As the help for them says, set them unless you know your
card does not use one of these chips for FM support.

  Once you are configured, make zlilo, modules, modules_install; reboot.
Note that it is no longer necessary or possible to configure sound in the
drivers/sound dir. Now one simply configures and makes one's kernel and
modules in the usual way.

 Then, add to your /etc/modprobe.conf something like:

alias char-major-14-* sb
install sb /sbin/modprobe -i sb && /sbin/modprobe adlib_card
options sb io=0x220 irq=7 dma=1 dma16=5 mpu_io=0x330
options adlib_card io=0x388     # FM synthesizer

 Alternatively, if you have compiled in kernel level ISAPnP support:

alias char-major-14 sb
post-install sb /sbin/modprobe "-k" "adlib_card"
options adlib_card io=0x388

  The effect of this is that the sound driver and all necessary bits and
pieces autoload on demand, assuming you use kerneld (a sound choice) and
autoclean when not in use. Also, options for the device drivers are
set. They will not work without them. Change as appropriate for your card.
If you are not yet using the very cool kerneld, you will have to "modprobe
-k sb" yourself to get things going. Eventually things may be fixed so
that this kludgery is not necessary; for the time being, it seems to work
well.

  Replace 'sb' with the driver for your card, and give it the right
options. To find the filename of the driver, look in
/lib/modules/<kernel-version>/misc. Mine looks like:

adlib_card.o # This is the generic OPLx driver
opl3.o # The OPL3 driver
sb.o # <<The SoundBlaster driver. Yours may differ.>>
sound.o # The sound driver
uart401.o # Used by sb, maybe other cards

 Whichever card you have, try feeding it the options that would be the
default if you were making the driver wired, not as modules. You can
look at function referred to by module_init() for the card to see what
args are expected.

 Note that at present there is no way to configure the io, irq and other
parameters for the modular drivers as one does for the wired drivers.. One
needs to pass the modules the necessary parameters as arguments, either
with /etc/modprobe.conf or with command-line args to modprobe, e.g.

modprobe sb io=0x220 irq=7 dma=1 dma16=5 mpu_io=0x330
modprobe adlib_card io=0x388

 recommend using /etc/modprobe.conf.

Persistent DMA Buffers:

The sound modules normally allocate DMA buffers during open() and
deallocate them during close(). Linux can often have problems allocating
DMA buffers for ISA cards on machines with more than 16MB RAM. This is
because ISA DMA buffers must exist below the 16MB boundary and it is quite
possible that we can't find a large enough free block in this region after
the machine has been running for any amount of time. The way to avoid this
problem is to allocate the DMA buffers during module load and deallocate
them when the module is unloaded. For this to be effective we need to load
the sound modules right after the kernel boots, either manually or by an
init script, and keep them around until we shut down. This is a little
wasteful of RAM, but it guarantees that sound always works.

To make the sound driver use persistent DMA buffers we need to pass the
sound.o module a "dmabuf=1" command-line argument. This is normally done
in /etc/modprobe.conf like so:

options sound		dmabuf=1

If you have 16MB or less RAM or a PCI sound card, this is wasteful and
unnecessary. It is possible that machine with 16MB or less RAM will find
this option useful, but if your machine is so memory-starved that it
cannot find a 64K block free, you will be wasting even more RAM by keeping
the sound modules loaded and the DMA buffers allocated when they are not
needed. The proper solution is to upgrade your RAM. But you do also have
this improper solution as well. Use it wisely.

  I'm afraid I know nothing about anything but my setup, being more of a
text-mode guy anyway. If you have options for other cards or other helpful
hints, send them to me, Jim Bray, jb@as220.org, http://as220.org/jb.