8a962f23f2
the original ad-audiomp branch notes: Add MP locking to the audio drivers. Making the audio drivers MP safe is necessary before efforts can be made to make the VM system MP safe. The are two locks per device instance, an ISR lock and a character device lock. The ISR lock replaces calls to splaudio()/splx(), and will be held across calls to device methods which were called at splaudio() before (e.g. trigger_output). The character device lock is held across calls to nearly all of the methods, excluding some only used for initialization, e.g. get_locks. Welcome to 5.99.57.
1174 lines
33 KiB
C
1174 lines
33 KiB
C
/* $NetBSD: auconv.c,v 1.25 2011/11/23 23:07:31 jmcneill Exp $ */
|
|
|
|
/*
|
|
* Copyright (c) 1996 The NetBSD Foundation, Inc.
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
* 3. All advertising materials mentioning features or use of this software
|
|
* must display the following acknowledgement:
|
|
* This product includes software developed by the Computer Systems
|
|
* Engineering Group at Lawrence Berkeley Laboratory.
|
|
* 4. Neither the name of the University nor of the Laboratory may be used
|
|
* to endorse or promote products derived from this software without
|
|
* specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
*
|
|
*/
|
|
|
|
#include <sys/cdefs.h>
|
|
__KERNEL_RCSID(0, "$NetBSD: auconv.c,v 1.25 2011/11/23 23:07:31 jmcneill Exp $");
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/audioio.h>
|
|
#include <sys/device.h>
|
|
#include <sys/errno.h>
|
|
#include <sys/malloc.h>
|
|
#include <sys/null.h>
|
|
#include <sys/systm.h>
|
|
#include <dev/audio_if.h>
|
|
#include <dev/auconv.h>
|
|
#include <dev/mulaw.h>
|
|
#include <machine/limits.h>
|
|
#ifndef _KERNEL
|
|
#include <stddef.h>
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <stdbool.h>
|
|
#endif
|
|
|
|
#include <aurateconv.h> /* generated by config(8) */
|
|
#include <mulaw.h> /* generated by config(8) */
|
|
|
|
/* #define AUCONV_DEBUG */
|
|
#ifdef AUCONV_DEBUG
|
|
# define DPRINTF(x) printf x
|
|
#else
|
|
# define DPRINTF(x)
|
|
#endif
|
|
|
|
#if NAURATECONV > 0
|
|
static int auconv_rateconv_supportable(u_int, u_int, u_int);
|
|
static int auconv_rateconv_check_channels(const struct audio_format *, int,
|
|
int, const audio_params_t *,
|
|
stream_filter_list_t *);
|
|
static int auconv_rateconv_check_rates(const struct audio_format *, int,
|
|
int, const audio_params_t *,
|
|
audio_params_t *,
|
|
stream_filter_list_t *);
|
|
#endif
|
|
#ifdef AUCONV_DEBUG
|
|
static void auconv_dump_formats(const struct audio_format *, int);
|
|
#endif
|
|
static void auconv_dump_params(const audio_params_t *);
|
|
static int auconv_exact_match(const struct audio_format *, int, int,
|
|
const struct audio_params *);
|
|
static u_int auconv_normalize_encoding(u_int, u_int);
|
|
static int auconv_is_supported_rate(const struct audio_format *, u_int);
|
|
static int auconv_add_encoding(int, int, int, struct audio_encoding_set **,
|
|
int *);
|
|
|
|
#ifdef _KERNEL
|
|
#define AUCONV_MALLOC(size) malloc(size, M_DEVBUF, M_NOWAIT)
|
|
#define AUCONV_REALLOC(p, size) realloc(p, size, M_DEVBUF, M_NOWAIT)
|
|
#define AUCONV_FREE(p) free(p, M_DEVBUF)
|
|
#else
|
|
#define AUCONV_MALLOC(size) malloc(size)
|
|
#define AUCONV_REALLOC(p, size) realloc(p, size)
|
|
#define AUCONV_FREE(p) free(p)
|
|
#endif
|
|
|
|
struct audio_encoding_set {
|
|
int size;
|
|
audio_encoding_t items[1];
|
|
};
|
|
#define ENCODING_SET_SIZE(n) (offsetof(struct audio_encoding_set, items) \
|
|
+ sizeof(audio_encoding_t) * (n))
|
|
|
|
struct conv_table {
|
|
u_int encoding;
|
|
u_int validbits;
|
|
u_int precision;
|
|
stream_filter_factory_t *play_conv;
|
|
stream_filter_factory_t *rec_conv;
|
|
};
|
|
/*
|
|
* SLINEAR-16 or SLINEAR-24 should precede in a table because
|
|
* aurateconv supports only SLINEAR.
|
|
*/
|
|
static const struct conv_table s8_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
linear8_to_linear16, linear16_to_linear8},
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
linear8_to_linear16, linear16_to_linear8},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 8, 8,
|
|
change_sign8, change_sign8},
|
|
{0, 0, 0, NULL, NULL}};
|
|
static const struct conv_table u8_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
linear8_to_linear16, linear16_to_linear8},
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
linear8_to_linear16, linear16_to_linear8},
|
|
{AUDIO_ENCODING_SLINEAR_LE, 8, 8,
|
|
change_sign8, change_sign8},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 16, 16,
|
|
linear8_to_linear16, linear16_to_linear8},
|
|
{AUDIO_ENCODING_ULINEAR_BE, 16, 16,
|
|
linear8_to_linear16, linear16_to_linear8},
|
|
{0, 0, 0, NULL, NULL}};
|
|
static const struct conv_table s16le_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
swap_bytes, swap_bytes},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 16, 16,
|
|
change_sign16, change_sign16},
|
|
{AUDIO_ENCODING_ULINEAR_BE, 16, 16,
|
|
swap_bytes_change_sign16, swap_bytes_change_sign16},
|
|
{0, 0, 0, NULL, NULL}};
|
|
static const struct conv_table s16be_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
swap_bytes, swap_bytes},
|
|
{AUDIO_ENCODING_ULINEAR_BE, 16, 16,
|
|
change_sign16, change_sign16},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 16, 16,
|
|
swap_bytes_change_sign16, swap_bytes_change_sign16},
|
|
{0, 0, 0, NULL, NULL}};
|
|
static const struct conv_table u16le_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
change_sign16, change_sign16},
|
|
{AUDIO_ENCODING_ULINEAR_BE, 16, 16,
|
|
swap_bytes, swap_bytes},
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
swap_bytes_change_sign16, swap_bytes_change_sign16},
|
|
{0, 0, 0, NULL, NULL}};
|
|
static const struct conv_table u16be_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
change_sign16, change_sign16},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 16, 16,
|
|
swap_bytes, swap_bytes},
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
swap_bytes_change_sign16, swap_bytes_change_sign16},
|
|
{0, 0, 0, NULL, NULL}};
|
|
#if NMULAW > 0
|
|
static const struct conv_table mulaw_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
mulaw_to_linear16, linear16_to_mulaw},
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
mulaw_to_linear16, linear16_to_mulaw},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 16, 16,
|
|
mulaw_to_linear16, linear16_to_mulaw},
|
|
{AUDIO_ENCODING_ULINEAR_BE, 16, 16,
|
|
mulaw_to_linear16, linear16_to_mulaw},
|
|
{AUDIO_ENCODING_SLINEAR_LE, 8, 8,
|
|
mulaw_to_linear8, linear8_to_mulaw},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 8, 8,
|
|
mulaw_to_linear8, linear8_to_mulaw},
|
|
{0, 0, 0, NULL, NULL}};
|
|
static const struct conv_table alaw_table[] = {
|
|
{AUDIO_ENCODING_SLINEAR_LE, 16, 16,
|
|
alaw_to_linear16, linear16_to_alaw},
|
|
{AUDIO_ENCODING_SLINEAR_BE, 16, 16,
|
|
alaw_to_linear16, linear16_to_alaw},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 16, 16,
|
|
alaw_to_linear16, linear16_to_alaw},
|
|
{AUDIO_ENCODING_ULINEAR_BE, 16, 16,
|
|
alaw_to_linear16, linear16_to_alaw},
|
|
{AUDIO_ENCODING_SLINEAR_LE, 8, 8,
|
|
alaw_to_linear8, linear8_to_alaw},
|
|
{AUDIO_ENCODING_ULINEAR_LE, 8, 8,
|
|
alaw_to_linear8, linear8_to_alaw},
|
|
{0, 0, 0, NULL, NULL}};
|
|
#endif
|
|
#ifdef AUCONV_DEBUG
|
|
static const char *encoding_dbg_names[] = {
|
|
"none", AudioEmulaw, AudioEalaw, "pcm16",
|
|
"pcm8", AudioEadpcm, AudioEslinear_le, AudioEslinear_be,
|
|
AudioEulinear_le, AudioEulinear_be,
|
|
AudioEslinear, AudioEulinear,
|
|
AudioEmpeg_l1_stream, AudioEmpeg_l1_packets,
|
|
AudioEmpeg_l1_system, AudioEmpeg_l2_stream,
|
|
AudioEmpeg_l2_packets, AudioEmpeg_l2_system,
|
|
AudioEac3
|
|
};
|
|
#endif
|
|
|
|
void
|
|
stream_filter_set_fetcher(stream_filter_t *this, stream_fetcher_t *p)
|
|
{
|
|
this->prev = p;
|
|
}
|
|
|
|
void
|
|
stream_filter_set_inputbuffer(stream_filter_t *this, audio_stream_t *stream)
|
|
{
|
|
this->src = stream;
|
|
}
|
|
|
|
stream_filter_t *
|
|
auconv_nocontext_filter_factory(
|
|
int (*fetch_to)(struct audio_softc *, stream_fetcher_t *,
|
|
audio_stream_t *, int))
|
|
{
|
|
stream_filter_t *this;
|
|
|
|
this = AUCONV_MALLOC(sizeof(stream_filter_t));
|
|
if (this == NULL)
|
|
return NULL;
|
|
this->base.fetch_to = fetch_to;
|
|
this->dtor = auconv_nocontext_filter_dtor;
|
|
this->set_fetcher = stream_filter_set_fetcher;
|
|
this->set_inputbuffer = stream_filter_set_inputbuffer;
|
|
this->prev = NULL;
|
|
this->src = NULL;
|
|
return this;
|
|
}
|
|
|
|
void
|
|
auconv_nocontext_filter_dtor(struct stream_filter *this)
|
|
{
|
|
if (this != NULL)
|
|
AUCONV_FREE(this);
|
|
}
|
|
|
|
#define DEFINE_FILTER(name) \
|
|
static int \
|
|
name##_fetch_to(struct audio_softc *, stream_fetcher_t *, audio_stream_t *, int); \
|
|
stream_filter_t * \
|
|
name(struct audio_softc *sc, const audio_params_t *from, \
|
|
const audio_params_t *to) \
|
|
{ \
|
|
return auconv_nocontext_filter_factory(name##_fetch_to); \
|
|
} \
|
|
static int \
|
|
name##_fetch_to(struct audio_softc *sc, stream_fetcher_t *self, \
|
|
audio_stream_t *dst, int max_used)
|
|
|
|
DEFINE_FILTER(change_sign8)
|
|
{
|
|
stream_filter_t *this;
|
|
int m, err;
|
|
|
|
this = (stream_filter_t *)self;
|
|
if ((err = this->prev->fetch_to(sc, this->prev, this->src, max_used)))
|
|
return err;
|
|
m = dst->end - dst->start;
|
|
m = min(m, max_used);
|
|
FILTER_LOOP_PROLOGUE(this->src, 1, dst, 1, m) {
|
|
*d = *s ^ 0x80;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
return 0;
|
|
}
|
|
|
|
DEFINE_FILTER(change_sign16)
|
|
{
|
|
stream_filter_t *this;
|
|
int m, err, enc;
|
|
|
|
this = (stream_filter_t *)self;
|
|
max_used = (max_used + 1) & ~1; /* round up to even */
|
|
if ((err = this->prev->fetch_to(sc, this->prev, this->src, max_used)))
|
|
return err;
|
|
m = (dst->end - dst->start) & ~1;
|
|
m = min(m, max_used);
|
|
enc = dst->param.encoding;
|
|
if (enc == AUDIO_ENCODING_SLINEAR_LE
|
|
|| enc == AUDIO_ENCODING_ULINEAR_LE) {
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 2, m) {
|
|
d[0] = s[0];
|
|
d[1] = s[1] ^ 0x80;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else {
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 2, m) {
|
|
d[0] = s[0] ^ 0x80;
|
|
d[1] = s[1];
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
DEFINE_FILTER(swap_bytes)
|
|
{
|
|
stream_filter_t *this;
|
|
int m, err;
|
|
|
|
this = (stream_filter_t *)self;
|
|
max_used = (max_used + 1) & ~1; /* round up to even */
|
|
if ((err = this->prev->fetch_to(sc, this->prev, this->src, max_used)))
|
|
return err;
|
|
m = (dst->end - dst->start) & ~1;
|
|
m = min(m, max_used);
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 2, m) {
|
|
d[0] = s[1];
|
|
d[1] = s[0];
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
return 0;
|
|
}
|
|
|
|
DEFINE_FILTER(swap_bytes_change_sign16)
|
|
{
|
|
stream_filter_t *this;
|
|
int m, err, enc;
|
|
|
|
this = (stream_filter_t *)self;
|
|
max_used = (max_used + 1) & ~1; /* round up to even */
|
|
if ((err = this->prev->fetch_to(sc, this->prev, this->src, max_used)))
|
|
return err;
|
|
m = (dst->end - dst->start) & ~1;
|
|
m = min(m, max_used);
|
|
enc = dst->param.encoding;
|
|
if (enc == AUDIO_ENCODING_SLINEAR_LE
|
|
|| enc == AUDIO_ENCODING_ULINEAR_LE) {
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 2, m) {
|
|
d[0] = s[1];
|
|
d[1] = s[0] ^ 0x80;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else {
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 2, m) {
|
|
d[0] = s[1] ^ 0x80;
|
|
d[1] = s[0];
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
DEFINE_FILTER(linear8_to_linear16)
|
|
{
|
|
stream_filter_t *this;
|
|
int m, err, enc_dst, enc_src;
|
|
|
|
this = (stream_filter_t *)self;
|
|
max_used = (max_used + 1) & ~1; /* round up to even */
|
|
if ((err = this->prev->fetch_to(sc, this->prev, this->src, max_used / 2)))
|
|
return err;
|
|
m = (dst->end - dst->start) & ~1;
|
|
m = min(m, max_used);
|
|
enc_dst = dst->param.encoding;
|
|
enc_src = this->src->param.encoding;
|
|
if ((enc_src == AUDIO_ENCODING_SLINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_SLINEAR_LE)
|
|
|| (enc_src == AUDIO_ENCODING_ULINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_ULINEAR_LE)) {
|
|
/*
|
|
* slinear8 -> slinear16_le
|
|
* ulinear8 -> ulinear16_le
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 1, dst, 2, m) {
|
|
d[0] = 0;
|
|
d[1] = s[0];
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else if ((enc_src == AUDIO_ENCODING_SLINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_SLINEAR_BE)
|
|
|| (enc_src == AUDIO_ENCODING_ULINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_ULINEAR_BE)) {
|
|
/*
|
|
* slinear8 -> slinear16_be
|
|
* ulinear8 -> ulinear16_be
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 1, dst, 2, m) {
|
|
d[0] = s[0];
|
|
d[1] = 0;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else if ((enc_src == AUDIO_ENCODING_SLINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_ULINEAR_LE)
|
|
|| (enc_src == AUDIO_ENCODING_ULINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_SLINEAR_LE)) {
|
|
/*
|
|
* slinear8 -> ulinear16_le
|
|
* ulinear8 -> slinear16_le
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 1, dst, 2, m) {
|
|
d[0] = 0;
|
|
d[1] = s[0] ^ 0x80;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else {
|
|
/*
|
|
* slinear8 -> ulinear16_be
|
|
* ulinear8 -> slinear16_be
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 1, dst, 2, m) {
|
|
d[0] = s[0] ^ 0x80;
|
|
d[1] = 0;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
DEFINE_FILTER(linear16_to_linear8)
|
|
{
|
|
stream_filter_t *this;
|
|
int m, err, enc_src, enc_dst;
|
|
|
|
this = (stream_filter_t *)self;
|
|
if ((err = this->prev->fetch_to(sc, this->prev, this->src, max_used * 2)))
|
|
return err;
|
|
m = dst->end - dst->start;
|
|
m = min(m, max_used);
|
|
enc_dst = dst->param.encoding;
|
|
enc_src = this->src->param.encoding;
|
|
if ((enc_src == AUDIO_ENCODING_SLINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_SLINEAR_LE)
|
|
|| (enc_src == AUDIO_ENCODING_ULINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_ULINEAR_LE)) {
|
|
/*
|
|
* slinear16_le -> slinear8
|
|
* ulinear16_le -> ulinear8
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 1, m) {
|
|
d[0] = s[1];
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else if ((enc_src == AUDIO_ENCODING_SLINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_ULINEAR_LE)
|
|
|| (enc_src == AUDIO_ENCODING_ULINEAR_LE
|
|
&& enc_dst == AUDIO_ENCODING_SLINEAR_LE)) {
|
|
/*
|
|
* slinear16_le -> ulinear8
|
|
* ulinear16_le -> slinear8
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 1, m) {
|
|
d[0] = s[1] ^ 0x80;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else if ((enc_src == AUDIO_ENCODING_SLINEAR_BE
|
|
&& enc_dst == AUDIO_ENCODING_SLINEAR_LE)
|
|
|| (enc_src == AUDIO_ENCODING_ULINEAR_BE
|
|
&& enc_dst == AUDIO_ENCODING_ULINEAR_LE)) {
|
|
/*
|
|
* slinear16_be -> slinear8
|
|
* ulinear16_be -> ulinear8
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 1, m) {
|
|
d[0] = s[0];
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
} else {
|
|
/*
|
|
* slinear16_be -> ulinear8
|
|
* ulinear16_be -> slinear8
|
|
*/
|
|
FILTER_LOOP_PROLOGUE(this->src, 2, dst, 1, m) {
|
|
d[0] = s[0] ^ 0x80;
|
|
} FILTER_LOOP_EPILOGUE(this->src, dst);
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* Set appropriate parameters in `param,' and return the index in
|
|
* the hardware capability array `formats.'
|
|
*
|
|
* @param formats [IN] An array of formats which a hardware can support.
|
|
* @param nformats [IN] The number of elements of the array.
|
|
* @param mode [IN] Either AUMODE_PLAY or AUMODE_RECORD.
|
|
* @param param [IN] Requested format. param->sw_code may be set.
|
|
* @param rateconv [IN] true if aurateconv may be used.
|
|
* @param list [OUT] stream_filters required for param.
|
|
* @return The index of selected audio_format entry. -1 if the device
|
|
* can not support the specified param.
|
|
*/
|
|
int
|
|
auconv_set_converter(const struct audio_format *formats, int nformats,
|
|
int mode, const audio_params_t *param, int rateconv,
|
|
stream_filter_list_t *list)
|
|
{
|
|
audio_params_t work;
|
|
const struct conv_table *table;
|
|
stream_filter_factory_t *conv;
|
|
int enc;
|
|
int i, j;
|
|
|
|
#ifdef AUCONV_DEBUG
|
|
DPRINTF(("%s: ENTER rateconv=%d\n", __func__, rateconv));
|
|
auconv_dump_formats(formats, nformats);
|
|
#endif
|
|
enc = auconv_normalize_encoding(param->encoding, param->precision);
|
|
|
|
/* check support by native format */
|
|
i = auconv_exact_match(formats, nformats, mode, param);
|
|
if (i >= 0) {
|
|
DPRINTF(("%s: LEAVE with %d (exact)\n", __func__, i));
|
|
return i;
|
|
}
|
|
|
|
#if NAURATECONV > 0
|
|
/* native format with aurateconv */
|
|
DPRINTF(("%s: native with aurateconv\n", __func__));
|
|
if (rateconv
|
|
&& auconv_rateconv_supportable(enc, param->precision,
|
|
param->validbits)) {
|
|
i = auconv_rateconv_check_channels(formats, nformats,
|
|
mode, param, list);
|
|
if (i >= 0) {
|
|
DPRINTF(("%s: LEAVE with %d (aurateconv1)\n", __func__, i));
|
|
return i;
|
|
}
|
|
}
|
|
#endif
|
|
|
|
/* check for emulation */
|
|
DPRINTF(("%s: encoding emulation\n", __func__));
|
|
table = NULL;
|
|
switch (enc) {
|
|
case AUDIO_ENCODING_SLINEAR_LE:
|
|
if (param->precision == 8)
|
|
table = s8_table;
|
|
else if (param->precision == 16)
|
|
table = s16le_table;
|
|
break;
|
|
case AUDIO_ENCODING_SLINEAR_BE:
|
|
if (param->precision == 8)
|
|
table = s8_table;
|
|
else if (param->precision == 16)
|
|
table = s16be_table;
|
|
break;
|
|
case AUDIO_ENCODING_ULINEAR_LE:
|
|
if (param->precision == 8)
|
|
table = u8_table;
|
|
else if (param->precision == 16)
|
|
table = u16le_table;
|
|
break;
|
|
case AUDIO_ENCODING_ULINEAR_BE:
|
|
if (param->precision == 8)
|
|
table = u8_table;
|
|
else if (param->precision == 16)
|
|
table = u16be_table;
|
|
break;
|
|
#if NMULAW > 0
|
|
case AUDIO_ENCODING_ULAW:
|
|
table = mulaw_table;
|
|
break;
|
|
case AUDIO_ENCODING_ALAW:
|
|
table = alaw_table;
|
|
break;
|
|
#endif
|
|
}
|
|
if (table == NULL) {
|
|
DPRINTF(("%s: LEAVE with -1 (no-emultable)\n", __func__));
|
|
return -1;
|
|
}
|
|
work = *param;
|
|
for (j = 0; table[j].precision != 0; j++) {
|
|
work.encoding = table[j].encoding;
|
|
work.precision = table[j].precision;
|
|
work.validbits = table[j].validbits;
|
|
i = auconv_exact_match(formats, nformats, mode, &work);
|
|
if (i >= 0) {
|
|
conv = mode == AUMODE_PLAY
|
|
? table[j].play_conv : table[j].rec_conv;
|
|
list->append(list, conv, &work);
|
|
DPRINTF(("%s: LEAVE with %d (emultable)\n", __func__, i));
|
|
return i;
|
|
}
|
|
}
|
|
/* not found */
|
|
|
|
#if NAURATECONV > 0
|
|
/* emulation with aurateconv */
|
|
DPRINTF(("%s: encoding emulation with aurateconv\n", __func__));
|
|
if (!rateconv) {
|
|
DPRINTF(("%s: LEAVE with -1 (no-rateconv)\n", __func__));
|
|
return -1;
|
|
}
|
|
work = *param;
|
|
for (j = 0; table[j].precision != 0; j++) {
|
|
if (!auconv_rateconv_supportable(table[j].encoding,
|
|
table[j].precision,
|
|
table[j].validbits))
|
|
continue;
|
|
work.encoding = table[j].encoding;
|
|
work.precision = table[j].precision;
|
|
work.validbits = table[j].validbits;
|
|
i = auconv_rateconv_check_channels(formats, nformats,
|
|
mode, &work, list);
|
|
if (i >= 0) {
|
|
/* work<=>hw conversion is already registered */
|
|
conv = mode == AUMODE_PLAY
|
|
? table[j].play_conv : table[j].rec_conv;
|
|
/* register userland<=>work conversion */
|
|
list->append(list, conv, &work);
|
|
DPRINTF(("%s: LEAVE with %d (rateconv2)\n", __func__, i));
|
|
return i;
|
|
}
|
|
}
|
|
|
|
#endif
|
|
DPRINTF(("%s: LEAVE with -1 (bottom)\n", __func__));
|
|
return -1;
|
|
}
|
|
|
|
#if NAURATECONV > 0
|
|
static int
|
|
auconv_rateconv_supportable(u_int encoding, u_int precision, u_int validbits)
|
|
{
|
|
if (encoding != AUDIO_ENCODING_SLINEAR_LE
|
|
&& encoding != AUDIO_ENCODING_SLINEAR_BE)
|
|
return false;
|
|
if (precision != 16 && precision != 24 && precision != 32)
|
|
return false;
|
|
if (precision < validbits)
|
|
return false;
|
|
return true;
|
|
}
|
|
|
|
static int
|
|
auconv_rateconv_check_channels(const struct audio_format *formats, int nformats,
|
|
int mode, const audio_params_t *param,
|
|
stream_filter_list_t *list)
|
|
{
|
|
audio_params_t hw_param;
|
|
int ind, n;
|
|
|
|
hw_param = *param;
|
|
/* check for the specified number of channels */
|
|
ind = auconv_rateconv_check_rates(formats, nformats, mode, param,
|
|
&hw_param, list);
|
|
if (ind >= 0)
|
|
return ind;
|
|
|
|
/* check for larger numbers */
|
|
for (n = param->channels + 1; n <= AUDIO_MAX_CHANNELS; n++) {
|
|
hw_param.channels = n;
|
|
ind = auconv_rateconv_check_rates(formats, nformats, mode,
|
|
param, &hw_param, list);
|
|
if (ind >= 0)
|
|
return ind;
|
|
}
|
|
|
|
/* check for stereo:monaural conversion */
|
|
if (param->channels == 2) {
|
|
hw_param.channels = 1;
|
|
ind = auconv_rateconv_check_rates(formats, nformats, mode,
|
|
param, &hw_param, list);
|
|
if (ind >= 0)
|
|
return ind;
|
|
}
|
|
return -1;
|
|
}
|
|
|
|
static int
|
|
auconv_rateconv_check_rates(const struct audio_format *formats, int nformats,
|
|
int mode, const audio_params_t *param,
|
|
audio_params_t *hw_param, stream_filter_list_t *list)
|
|
{
|
|
int ind, i, j, enc, f_enc;
|
|
u_int rate, minrate, maxrate, orig_rate;
|
|
|
|
/* exact match */
|
|
ind = auconv_exact_match(formats, nformats, mode, hw_param);
|
|
if (ind >= 0)
|
|
goto found;
|
|
|
|
/* determine min/max of specified encoding/precision/channels */
|
|
minrate = UINT_MAX;
|
|
maxrate = 0;
|
|
enc = auconv_normalize_encoding(param->encoding,
|
|
param->precision);
|
|
for (i = 0; i < nformats; i++) {
|
|
if (!AUFMT_IS_VALID(&formats[i]))
|
|
continue;
|
|
if ((formats[i].mode & mode) == 0)
|
|
continue;
|
|
f_enc = auconv_normalize_encoding(formats[i].encoding,
|
|
formats[i].precision);
|
|
if (f_enc != enc)
|
|
continue;
|
|
if (formats[i].validbits != hw_param->validbits)
|
|
continue;
|
|
if (formats[i].precision != hw_param->precision)
|
|
continue;
|
|
if (formats[i].channels != hw_param->channels)
|
|
continue;
|
|
if (formats[i].frequency_type == 0) {
|
|
if (formats[i].frequency[0] < minrate)
|
|
minrate = formats[i].frequency[0];
|
|
if (formats[i].frequency[1] > maxrate)
|
|
maxrate = formats[i].frequency[1];
|
|
} else {
|
|
for (j = 0; j < formats[i].frequency_type; j++) {
|
|
if (formats[i].frequency[j] < minrate)
|
|
minrate = formats[i].frequency[j];
|
|
if (formats[i].frequency[j] > maxrate)
|
|
maxrate = formats[i].frequency[j];
|
|
}
|
|
}
|
|
}
|
|
if (maxrate == 0)
|
|
return -1;
|
|
|
|
/* try multiples of sample_rate */
|
|
orig_rate = hw_param->sample_rate;
|
|
for (i = 2; (rate = param->sample_rate * i) <= maxrate; i++) {
|
|
hw_param->sample_rate = rate;
|
|
ind = auconv_exact_match(formats, nformats, mode, hw_param);
|
|
if (ind >= 0)
|
|
goto found;
|
|
}
|
|
|
|
hw_param->sample_rate = param->sample_rate >= minrate
|
|
? maxrate : minrate;
|
|
ind = auconv_exact_match(formats, nformats, mode, hw_param);
|
|
if (ind >= 0)
|
|
goto found;
|
|
hw_param->sample_rate = orig_rate;
|
|
return -1;
|
|
|
|
found:
|
|
list->append(list, aurateconv, hw_param);
|
|
return ind;
|
|
}
|
|
#endif /* NAURATECONV */
|
|
|
|
#ifdef AUCONV_DEBUG
|
|
static void
|
|
auconv_dump_formats(const struct audio_format *formats, int nformats)
|
|
{
|
|
const struct audio_format *f;
|
|
int i, j;
|
|
|
|
for (i = 0; i < nformats; i++) {
|
|
f = &formats[i];
|
|
printf("[%2d]: mode=", i);
|
|
if (!AUFMT_IS_VALID(f)) {
|
|
printf("INVALID");
|
|
} else if (f->mode == AUMODE_PLAY) {
|
|
printf("PLAY");
|
|
} else if (f->mode == AUMODE_RECORD) {
|
|
printf("RECORD");
|
|
} else if (f->mode == (AUMODE_PLAY | AUMODE_RECORD)) {
|
|
printf("PLAY|RECORD");
|
|
} else {
|
|
printf("0x%x", f->mode);
|
|
}
|
|
printf(" enc=%s", encoding_dbg_names[f->encoding]);
|
|
printf(" %u/%ubit", f->validbits, f->precision);
|
|
printf(" %uch", f->channels);
|
|
|
|
printf(" channel_mask=");
|
|
if (f->channel_mask == AUFMT_MONAURAL) {
|
|
printf("MONAURAL");
|
|
} else if (f->channel_mask == AUFMT_STEREO) {
|
|
printf("STEREO");
|
|
} else if (f->channel_mask == AUFMT_SURROUND4) {
|
|
printf("SURROUND4");
|
|
} else if (f->channel_mask == AUFMT_DOLBY_5_1) {
|
|
printf("DOLBY5.1");
|
|
} else {
|
|
printf("0x%x", f->channel_mask);
|
|
}
|
|
|
|
if (f->frequency_type == 0) {
|
|
printf(" %uHz-%uHz", f->frequency[0],
|
|
f->frequency[1]);
|
|
} else {
|
|
printf(" %uHz", f->frequency[0]);
|
|
for (j = 1; j < f->frequency_type; j++)
|
|
printf(",%uHz", f->frequency[j]);
|
|
}
|
|
printf("\n");
|
|
}
|
|
}
|
|
|
|
static void
|
|
auconv_dump_params(const audio_params_t *p)
|
|
{
|
|
printf("enc=%s", encoding_dbg_names[p->encoding]);
|
|
printf(" %u/%ubit", p->validbits, p->precision);
|
|
printf(" %uch", p->channels);
|
|
printf(" %uHz", p->sample_rate);
|
|
printf("\n");
|
|
}
|
|
#else
|
|
static void
|
|
auconv_dump_params(const audio_params_t *p)
|
|
{
|
|
}
|
|
#endif /* AUCONV_DEBUG */
|
|
|
|
/**
|
|
* a sub-routine for auconv_set_converter()
|
|
*/
|
|
static int
|
|
auconv_exact_match(const struct audio_format *formats, int nformats,
|
|
int mode, const audio_params_t *param)
|
|
{
|
|
int i, enc, f_enc;
|
|
|
|
DPRINTF(("%s: ENTER: mode=0x%x target:", __func__, mode));
|
|
auconv_dump_params(param);
|
|
enc = auconv_normalize_encoding(param->encoding,
|
|
param->precision);
|
|
DPRINTF(("%s: target normalized: %s\n", __func__,
|
|
encoding_dbg_names[enc]));
|
|
for (i = 0; i < nformats; i++) {
|
|
if (!AUFMT_IS_VALID(&formats[i]))
|
|
continue;
|
|
if ((formats[i].mode & mode) == 0)
|
|
continue;
|
|
f_enc = auconv_normalize_encoding(formats[i].encoding,
|
|
formats[i].precision);
|
|
DPRINTF(("%s: format[%d] normalized: %s\n",
|
|
__func__, i, encoding_dbg_names[f_enc]));
|
|
if (f_enc != enc)
|
|
continue;
|
|
/**
|
|
* XXX we need encoding-dependent check.
|
|
* XXX Is to check precision/channels meaningful for
|
|
* MPEG encodings?
|
|
*/
|
|
if (enc != AUDIO_ENCODING_AC3) {
|
|
if (formats[i].validbits != param->validbits)
|
|
continue;
|
|
if (formats[i].precision != param->precision)
|
|
continue;
|
|
if (formats[i].channels != param->channels)
|
|
continue;
|
|
}
|
|
if (!auconv_is_supported_rate(&formats[i],
|
|
param->sample_rate))
|
|
continue;
|
|
return i;
|
|
}
|
|
return -1;
|
|
}
|
|
|
|
/**
|
|
* a sub-routine for auconv_set_converter()
|
|
* SLINEAR ==> SLINEAR_<host-endian>
|
|
* ULINEAR ==> ULINEAR_<host-endian>
|
|
* SLINEAR_BE 8bit ==> SLINEAR_LE 8bit
|
|
* ULINEAR_BE 8bit ==> ULINEAR_LE 8bit
|
|
* This should be the same rule as audio_check_params()
|
|
*/
|
|
static u_int
|
|
auconv_normalize_encoding(u_int encoding, u_int precision)
|
|
{
|
|
int enc;
|
|
|
|
enc = encoding;
|
|
if (enc == AUDIO_ENCODING_SLINEAR_LE)
|
|
return enc;
|
|
if (enc == AUDIO_ENCODING_ULINEAR_LE)
|
|
return enc;
|
|
#if BYTE_ORDER == LITTLE_ENDIAN
|
|
if (enc == AUDIO_ENCODING_SLINEAR)
|
|
return AUDIO_ENCODING_SLINEAR_LE;
|
|
else if (enc == AUDIO_ENCODING_ULINEAR)
|
|
return AUDIO_ENCODING_ULINEAR_LE;
|
|
#else
|
|
if (enc == AUDIO_ENCODING_SLINEAR)
|
|
enc = AUDIO_ENCODING_SLINEAR_BE;
|
|
else if (enc == AUDIO_ENCODING_ULINEAR)
|
|
enc = AUDIO_ENCODING_ULINEAR_BE;
|
|
#endif
|
|
if (precision == 8 && enc == AUDIO_ENCODING_SLINEAR_BE)
|
|
return AUDIO_ENCODING_SLINEAR_LE;
|
|
if (precision == 8 && enc == AUDIO_ENCODING_ULINEAR_BE)
|
|
return AUDIO_ENCODING_ULINEAR_LE;
|
|
return enc;
|
|
}
|
|
|
|
/**
|
|
* a sub-routine for auconv_set_converter()
|
|
*/
|
|
static int
|
|
auconv_is_supported_rate(const struct audio_format *format, u_int rate)
|
|
{
|
|
u_int i;
|
|
|
|
if (format->frequency_type == 0) {
|
|
return format->frequency[0] <= rate
|
|
&& rate <= format->frequency[1];
|
|
}
|
|
for (i = 0; i < format->frequency_type; i++) {
|
|
if (format->frequency[i] == rate)
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
/**
|
|
* Create an audio_encoding_set besed on hardware capability represented
|
|
* by audio_format.
|
|
*
|
|
* Usage:
|
|
* foo_attach(...) {
|
|
* :
|
|
* if (auconv_create_encodings(formats, nformats,
|
|
* &sc->sc_encodings) != 0) {
|
|
* // attach failure
|
|
* }
|
|
*
|
|
* @param formats [IN] An array of formats which a hardware can support.
|
|
* @param nformats [IN] The number of elements of the array.
|
|
* @param encodings [OUT] receives an address of an audio_encoding_set.
|
|
* @return errno; 0 for success.
|
|
*/
|
|
int
|
|
auconv_create_encodings(const struct audio_format *formats, int nformats,
|
|
struct audio_encoding_set **encodings)
|
|
{
|
|
struct audio_encoding_set *buf;
|
|
int capacity;
|
|
int i;
|
|
int err;
|
|
|
|
#define ADD_ENCODING(enc, prec, flags) do { \
|
|
err = auconv_add_encoding(enc, prec, flags, &buf, &capacity); \
|
|
if (err != 0) goto err_exit; \
|
|
} while (/*CONSTCOND*/0)
|
|
|
|
capacity = 10;
|
|
buf = AUCONV_MALLOC(ENCODING_SET_SIZE(capacity));
|
|
if (buf == NULL) {
|
|
err = ENOMEM;
|
|
goto err_exit;
|
|
}
|
|
buf->size = 0;
|
|
for (i = 0; i < nformats; i++) {
|
|
if (!AUFMT_IS_VALID(&formats[i]))
|
|
continue;
|
|
switch (formats[i].encoding) {
|
|
case AUDIO_ENCODING_SLINEAR_LE:
|
|
ADD_ENCODING(formats[i].encoding,
|
|
formats[i].precision, 0);
|
|
ADD_ENCODING(AUDIO_ENCODING_SLINEAR_BE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ULINEAR_LE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ULINEAR_BE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
#if NMULAW > 0
|
|
if (formats[i].precision == 8
|
|
|| formats[i].precision == 16) {
|
|
ADD_ENCODING(AUDIO_ENCODING_ULAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ALAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
}
|
|
#endif
|
|
break;
|
|
case AUDIO_ENCODING_SLINEAR_BE:
|
|
ADD_ENCODING(formats[i].encoding,
|
|
formats[i].precision, 0);
|
|
ADD_ENCODING(AUDIO_ENCODING_SLINEAR_LE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ULINEAR_LE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ULINEAR_BE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
#if NMULAW > 0
|
|
if (formats[i].precision == 8
|
|
|| formats[i].precision == 16) {
|
|
ADD_ENCODING(AUDIO_ENCODING_ULAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ALAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
}
|
|
#endif
|
|
break;
|
|
case AUDIO_ENCODING_ULINEAR_LE:
|
|
ADD_ENCODING(formats[i].encoding,
|
|
formats[i].precision, 0);
|
|
ADD_ENCODING(AUDIO_ENCODING_SLINEAR_BE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_SLINEAR_LE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ULINEAR_BE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
#if NMULAW > 0
|
|
if (formats[i].precision == 8
|
|
|| formats[i].precision == 16) {
|
|
ADD_ENCODING(AUDIO_ENCODING_ULAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ALAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
}
|
|
#endif
|
|
break;
|
|
case AUDIO_ENCODING_ULINEAR_BE:
|
|
ADD_ENCODING(formats[i].encoding,
|
|
formats[i].precision, 0);
|
|
ADD_ENCODING(AUDIO_ENCODING_SLINEAR_BE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ULINEAR_LE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_SLINEAR_LE,
|
|
formats[i].precision,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
#if NMULAW > 0
|
|
if (formats[i].precision == 8
|
|
|| formats[i].precision == 16) {
|
|
ADD_ENCODING(AUDIO_ENCODING_ULAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
ADD_ENCODING(AUDIO_ENCODING_ALAW, 8,
|
|
AUDIO_ENCODINGFLAG_EMULATED);
|
|
}
|
|
#endif
|
|
break;
|
|
|
|
case AUDIO_ENCODING_ULAW:
|
|
case AUDIO_ENCODING_ALAW:
|
|
case AUDIO_ENCODING_ADPCM:
|
|
case AUDIO_ENCODING_MPEG_L1_STREAM:
|
|
case AUDIO_ENCODING_MPEG_L1_PACKETS:
|
|
case AUDIO_ENCODING_MPEG_L1_SYSTEM:
|
|
case AUDIO_ENCODING_MPEG_L2_STREAM:
|
|
case AUDIO_ENCODING_MPEG_L2_PACKETS:
|
|
case AUDIO_ENCODING_MPEG_L2_SYSTEM:
|
|
case AUDIO_ENCODING_AC3:
|
|
ADD_ENCODING(formats[i].encoding,
|
|
formats[i].precision, 0);
|
|
break;
|
|
|
|
case AUDIO_ENCODING_SLINEAR:
|
|
case AUDIO_ENCODING_ULINEAR:
|
|
case AUDIO_ENCODING_LINEAR:
|
|
case AUDIO_ENCODING_LINEAR8:
|
|
default:
|
|
printf("%s: invalid encoding value "
|
|
"for audio_format: %d\n",
|
|
__func__, formats[i].encoding);
|
|
break;
|
|
}
|
|
}
|
|
*encodings = buf;
|
|
return 0;
|
|
|
|
err_exit:
|
|
if (buf != NULL)
|
|
AUCONV_FREE(buf);
|
|
*encodings = NULL;
|
|
return err;
|
|
}
|
|
|
|
/**
|
|
* a sub-routine for auconv_create_encodings()
|
|
*/
|
|
static int
|
|
auconv_add_encoding(int enc, int prec, int flags,
|
|
struct audio_encoding_set **buf, int *capacity)
|
|
{
|
|
static const char *encoding_names[] = {
|
|
NULL, AudioEmulaw, AudioEalaw, NULL,
|
|
NULL, AudioEadpcm, AudioEslinear_le, AudioEslinear_be,
|
|
AudioEulinear_le, AudioEulinear_be,
|
|
AudioEslinear, AudioEulinear,
|
|
AudioEmpeg_l1_stream, AudioEmpeg_l1_packets,
|
|
AudioEmpeg_l1_system, AudioEmpeg_l2_stream,
|
|
AudioEmpeg_l2_packets, AudioEmpeg_l2_system,
|
|
AudioEac3
|
|
};
|
|
struct audio_encoding_set *set;
|
|
struct audio_encoding_set *new_buf;
|
|
audio_encoding_t *e;
|
|
int i;
|
|
|
|
set = *buf;
|
|
/* already has the same encoding? */
|
|
e = set->items;
|
|
for (i = 0; i < set->size; i++, e++) {
|
|
if (e->encoding == enc && e->precision == prec) {
|
|
/* overwrite EMULATED flag */
|
|
if ((e->flags & AUDIO_ENCODINGFLAG_EMULATED)
|
|
&& (flags & AUDIO_ENCODINGFLAG_EMULATED) == 0) {
|
|
e->flags &= ~AUDIO_ENCODINGFLAG_EMULATED;
|
|
}
|
|
return 0;
|
|
}
|
|
}
|
|
/* We don't have the specified one. */
|
|
|
|
if (set->size >= *capacity) {
|
|
new_buf = AUCONV_REALLOC(set,
|
|
ENCODING_SET_SIZE(*capacity + 10));
|
|
if (new_buf == NULL)
|
|
return ENOMEM;
|
|
*buf = new_buf;
|
|
set = new_buf;
|
|
*capacity += 10;
|
|
}
|
|
|
|
e = &set->items[set->size];
|
|
e->index = 0;
|
|
strlcpy(e->name, encoding_names[enc], MAX_AUDIO_DEV_LEN);
|
|
e->encoding = enc;
|
|
e->precision = prec;
|
|
e->flags = flags;
|
|
set->size += 1;
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* Delete an audio_encoding_set created by auconv_create_encodings().
|
|
*
|
|
* Usage:
|
|
* foo_detach(...) {
|
|
* :
|
|
* auconv_delete_encodings(sc->sc_encodings);
|
|
* :
|
|
* }
|
|
*
|
|
* @param encodings [IN] An audio_encoding_set which was created by
|
|
* auconv_create_encodings().
|
|
* @return errno; 0 for success.
|
|
*/
|
|
int auconv_delete_encodings(struct audio_encoding_set *encodings)
|
|
{
|
|
if (encodings != NULL)
|
|
AUCONV_FREE(encodings);
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* Copy the element specified by aep->index.
|
|
*
|
|
* Usage:
|
|
* int foo_query_encoding(void *v, audio_encoding_t *aep) {
|
|
* struct foo_softc *sc = (struct foo_softc *)v;
|
|
* return auconv_query_encoding(sc->sc_encodings, aep);
|
|
* }
|
|
*
|
|
* @param encodings [IN] An audio_encoding_set created by
|
|
* auconv_create_encodings().
|
|
* @param aep [IN/OUT] resultant audio_encoding_t.
|
|
*/
|
|
int
|
|
auconv_query_encoding(const struct audio_encoding_set *encodings,
|
|
audio_encoding_t *aep)
|
|
{
|
|
if (aep->index >= encodings->size)
|
|
return EINVAL;
|
|
strlcpy(aep->name, encodings->items[aep->index].name,
|
|
MAX_AUDIO_DEV_LEN);
|
|
aep->encoding = encodings->items[aep->index].encoding;
|
|
aep->precision = encodings->items[aep->index].precision;
|
|
aep->flags = encodings->items[aep->index].flags;
|
|
return 0;
|
|
}
|