<html>
<head>
<title>linuxassembly.org - Using the audio device under Linux</title>
</head>
<body bgcolor="#ffffff" text="#000000">
<h1 align="center">Using the audio device under Linux</h1>
<p align="center"><font size="-1">by <a href="mailto:linuxassembly@unusedino.de">Karsten Scheibler</a></font></p>
<dl><dt><b>Remark (a):</b></dt>
<dd>Save this page as <tt>audio.html</tt> and use the following command to extract the source code with a sample Makefile:<br>
<tt>sh -c "( mkdir audio && cd audio && awk '/^<!--eof/{d=0}{if(d)print>f}/^<!--sof/{f=\$2;d=1}' ) < audio.html"</tt>
</dd><dt><b>Remark (b):</b></dt>
<dd>Some browsers try to be super smart and reformat the html code
so you may have trouble to extract the files. This is the right time to learn more
about wget, a tool to download files via http or ftp ;-)
</dd><dt><b>Remark (c):</b></dt>
<dd>Some versions of nasm are known to generate broken code.
I have used nasm-0.98 for this example which worked for me.
</dd>
</dl>
<p align="justify">This is a short explanation how you can use the OSS interface of the kernel
to output audio samples. If you find mistakes or have suggestions, mail
<a href="mailto:linuxassembly@unusedino.de">me</a>.</p>
<p align="justify">The example code below is very simple. It doesn't check if your soundcard
supports the options required (44.1 kHz mono/stereo 16 Bits signed). This
causes wrong output if your soundcard doesn't support this. Of course your kernel
has to have loaded the correct driver for your soundcard in order to get this work.</p>
<p>short procedure:<ol type="1">
<li>open the device via <b>sys_open</b> (usually <b>/dev/dsp</b>)</li>
<li>set your audio options via <b>sys_ioctl</b></li>
<li>use <b>sys_write</b> to output the audio samples</li>
</ol></p>
<p align="justify">Mostly all of the ioctl's used here are read/write this means the address
given in edx must be writable, because the kernel writes the real parameter
for your sound options back to this address (if you want to set 48 kHz but
your soundcard only supports 44.1 kHz you will find this value in your
variable addressed with edx). Therefore placing such a variable in
<b>section .text</b> is a bad idea. Notice: if you look at the kernel sources
(especially <b>/usr/src/linux/include/linux/soundcard.h</b>) you will find multiple
names for the same ioctl (example: <b>SNDCTL_DSP_CHANNELS</b> is equal to
<b>SOUND_PCM_WRITE_CHANNELS</b>).</p>
<p align="justify">If you want to use sound output in your program you should do something more
than this sample code. Here some more interesting ioctl's:<ul>
<li>check <b>SNDCTL_DSP_GETBLKSIZE</b>
how many data you have to write to your
soundcard, before playing starts (the value should be in bytes).
If you write less than this to the audio device no output
will be earable.
Remember: samples already in the buffer also count.</li>
<li>use <b>SNDCTL_DSP_GETODELAY</b> to check how many bytes (NOT SAMPLES!!) are
currently in the output buffer</li>
<li>perhaps you have to use <b>SNDCTL_DSP_NONBLOCK</b> to make writes (smaller than
the DMA buffer) nonblocking.</li>
<li>while using <b>SNDCTL_DSP_SETFMT</b>, <b>SNDCTL_DSP_CHANNELS</b> and
<b>SNDCTL_DSP_SPEED</b> don't forget to check the values the kernel
gives you back.</li>
<li>use <b>SNDCTL_DSP_SETFRAGMENT</b> to set the granularity of DMA buffer.
If i remember right, this influences also the amount of generated IRQs by the
soundcard. The higher the granularity the higher the latency and the lower the number
of generated IRQs.</li>
</ul></p>
<p align="justify">It is always a good idea to write as much data as possible at once, because
every syscall consumes extra cpu cycles.</p>
<p align="justify">The sample code below makes 2 outputs. First a simple <a href="#sawtooth">sawtooth waveform</a> and
second it tries to open the file <a href="#audio_raw">audio.raw</a> and if it exists it is played.
This file should be in the format 44.1 kHz, stereo, 16 Bit signed.</p>
<center><table><tr><td><pre>
<!--sof audio.n -->
;****************************************************************************
;****************************************************************************
;*
;* USING THE AUDIO DEVICE UNDER LINUX
;*
;* written by Karsten Scheibler, 2003-DEC-08
;*
;****************************************************************************
;****************************************************************************
global audio_start
;****************************************************************************
;* some assign's
;****************************************************************************
%assign SYS_READ 3
%assign SYS_WRITE 4
%assign SYS_OPEN 5
%assign SYS_IOCTL 54
%assign O_RDONLY 0
%assign O_RDWR 2
%assign SNDCTL_DSP_RESET 0x00005000
%assign SNDCTL_DSP_SPEED 0xc0045002
%assign SNDCTL_DSP_SETFMT 0xc0045005
%assign AFMT_S16_LE 0x00000010
%assign SNDCTL_DSP_CHANNELS 0xc0045006
%assign AUDIO_DATA_SIZE 0x100000 ;1 MB
;****************************************************************************
;* data
;****************************************************************************
section .data
align 4
audio_handle: dd 0
ioctl_table: dd SNDCTL_DSP_RESET, 0
dd SNDCTL_DSP_SPEED, 44100
dd SNDCTL_DSP_CHANNELS, 1
dd SNDCTL_DSP_SETFMT, AFMT_S16_LE
section .bss
alignb 4
audio_data: resb AUDIO_DATA_SIZE
;****************************************************************************
;* audio_start
;****************************************************************************
section .text
audio_start:
;open the device
mov dword eax, SYS_OPEN
mov dword ebx, .device
mov dword ecx, O_RDWR
int byte 0x80
test dword eax, eax
js near audio_error
mov dword [audio_handle], eax
;set the options: 44.1 kHz, mono, 16 Bit signed
xor dword ebp, ebp
.do_ioctl:
mov dword eax, SYS_IOCTL
mov dword ebx, [audio_handle]
mov dword ecx, [ioctl_table + 8 * ebp]
lea dword edx, [ioctl_table + 8 * ebp + 4]
int byte 0x80
test dword eax, eax
js near audio_error
inc dword ebp
cmp dword ebp, byte 4
jb short .do_ioctl
;<a name="sawtooth"></a>output something very simple (sawtooth waveform)
;not recommended: write every single sample like here
mov dword esi, 0x40000
xor dword ebp, ebp
.loop:
mov word [audio_data], bp
mov dword eax, SYS_WRITE
mov dword ebx, [audio_handle]
mov dword ecx, audio_data
mov dword edx, 2
int byte 0x80
test dword eax, eax
js near audio_error
add dword ebp, 0x0100
dec dword esi
jnz short .loop
;<a name="audio_raw"></a>open a raw audio file
;format should be 44.1 kHz, stereo, 16 Bit signed
mov dword eax, SYS_OPEN
mov dword ebx, .file
mov dword ecx, O_RDONLY
int byte 0x80
mov dword ebx, eax
test dword eax, eax
js short audio_error
mov dword eax, SYS_READ
mov dword ecx, audio_data
mov dword edx, AUDIO_DATA_SIZE
int byte 0x80
test dword eax, eax
js short audio_error
;set output to stereo
lea dword ebp, [ioctl_table + 4 * 6]
mov dword eax, SYS_IOCTL
mov dword ebx, [audio_handle]
mov dword ecx, SNDCTL_DSP_CHANNELS
mov dword [ebp], 2
mov dword edx, ebp
int byte 0x80
test dword eax, eax
js short audio_error
;dump it to the audio device
mov dword eax, SYS_WRITE
mov dword ebx, [audio_handle]
mov dword ecx, audio_data
mov dword edx, AUDIO_DATA_SIZE
int byte 0x80
test dword eax, eax
js short audio_error
;exit nicely
jmp short audio_end
.device: db "/dev/dsp", 0
.file: db "audio.raw", 0
;****************************************************************************
;* audio_error
;****************************************************************************
section .text
audio_error:
xor dword eax, eax
inc dword eax
mov dword ebx, eax
int byte 0x80
;****************************************************************************
;* audio_end
;****************************************************************************
section .text
audio_end:
xor dword eax, eax
xor dword ebx, ebx
inc dword eax
int byte 0x80
;*********************************************** linuxassembly@unusedino.de *
<!--eof-->
</pre></td></tr></table></center>
<!--sof Makefile
NASM=nasm -w+orphan-labels -w+macro-params -w+number-overflow -f elf
STRIP=strip -R .note -R .comment
LD=ld -s
RM=rm -f
.PHONY: all clean
all: audio
audio: audio.n
${NASM} -o audio.o audio.n
${LD} -e audio_start -o audio audio.o
${STRIP} audio
clean:
${RM} *.bak *~ audio audio.o core
<!--eof-->
</body>
</html>