mirror of
https://github.com/owntone/owntone-server.git
synced 2024-12-28 08:05:56 -05:00
1476 lines
39 KiB
C
1476 lines
39 KiB
C
/*
|
|
* Copyright (C) 2015-2019 Espen Jürgensen <espenjurgensen@gmail.com>
|
|
* Copyright (C) 2010 Julien BLACHE <jb@jblache.org>
|
|
*
|
|
* Copyright (c) 2010 Clemens Ladisch <clemens@ladisch.de>
|
|
* from alsa-utils/alsamixer/volume_mapping.c
|
|
* use_linear_dB_scale()
|
|
* lrint_dir()
|
|
* volume_normalized_set()
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write to the Free Software
|
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
*/
|
|
|
|
#ifdef HAVE_CONFIG_H
|
|
# include <config.h>
|
|
#endif
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <unistd.h>
|
|
#include <string.h>
|
|
#include <errno.h>
|
|
#include <stdint.h>
|
|
#include <inttypes.h>
|
|
#include <math.h>
|
|
|
|
#include <alsa/asoundlib.h>
|
|
|
|
#include "misc.h"
|
|
#include "conffile.h"
|
|
#include "logger.h"
|
|
#include "player.h"
|
|
#include "outputs.h"
|
|
|
|
|
|
// For setting volume, treat everything below this as linear scale
|
|
#define MAX_LINEAR_DB_SCALE 24
|
|
|
|
// We measure latency each second, and after a number of measurements determined
|
|
// by adjust_period_seconds we try to determine drift and latency. If both are
|
|
// below the two thresholds set by the below, we don't do anything. Otherwise we
|
|
// may attempt compensation by resampling. Latency is measured in samples, and
|
|
// drift is change of latency per second. Both are floats.
|
|
#define ALSA_MAX_LATENCY 480.0
|
|
#define ALSA_MAX_DRIFT 16.0
|
|
// If latency is jumping up and down we don't do compensation since we probably
|
|
// wouldn't do a good job. We use linear regression to determine the trend, but
|
|
// if r2 is below this value we won't attempt to correct sync.
|
|
#define ALSA_MAX_VARIANCE 0.3
|
|
|
|
// We correct latency by adjusting the sample rate in steps. However, if the
|
|
// latency keeps drifting we give up after reaching this step.
|
|
#define ALSA_RESAMPLE_STEP_MAX 8
|
|
// The sample rate gets adjusted by a multiple of this number. The number of
|
|
// multiples depends on the sample rate, i.e. a low sample rate may get stepped
|
|
// by 16, while high one would get stepped by 4 x 16
|
|
#define ALSA_RESAMPLE_STEP_MULTIPLE 2
|
|
|
|
#define ALSA_ERROR_WRITE -1
|
|
#define ALSA_ERROR_UNDERRUN -2
|
|
#define ALSA_ERROR_SESSION -3
|
|
#define ALSA_ERROR_DEVICE -4
|
|
#define ALSA_ERROR_DEVICE_BUSY -5
|
|
|
|
enum alsa_sync_state
|
|
{
|
|
ALSA_SYNC_OK,
|
|
ALSA_SYNC_AHEAD,
|
|
ALSA_SYNC_BEHIND,
|
|
};
|
|
|
|
struct alsa_mixer
|
|
{
|
|
snd_mixer_t *hdl;
|
|
snd_mixer_elem_t *vol_elem;
|
|
|
|
long vol_min;
|
|
long vol_max;
|
|
};
|
|
|
|
struct alsa_playback_session
|
|
{
|
|
snd_pcm_t *pcm;
|
|
|
|
int buffer_nsamp;
|
|
|
|
uint32_t pos;
|
|
|
|
uint32_t last_pos;
|
|
uint32_t last_buflen;
|
|
|
|
struct media_quality quality;
|
|
struct timespec last_pts;
|
|
|
|
// Used for syncing with the clock
|
|
struct timespec stamp_pts;
|
|
uint64_t stamp_pos;
|
|
|
|
// Array of latency calculations, where latency_counter tells how many are
|
|
// currently in the array
|
|
double *latency_history;
|
|
int latency_counter;
|
|
|
|
int sync_resample_step;
|
|
|
|
// Here we buffer samples during startup
|
|
struct ringbuffer prebuf;
|
|
|
|
struct alsa_playback_session *next;
|
|
};
|
|
|
|
// Info about the device, which is not required by the player, only internally
|
|
struct alsa_extra
|
|
{
|
|
const char *card_name;
|
|
const char *mixer_name;
|
|
const char *mixer_device_name;
|
|
int offset_ms;
|
|
};
|
|
|
|
struct alsa_session
|
|
{
|
|
enum output_device_state state;
|
|
|
|
uint64_t device_id;
|
|
int callback_id;
|
|
|
|
const char *devname;
|
|
const char *mixer_name;
|
|
const char *mixer_device_name;
|
|
|
|
struct alsa_mixer mixer;
|
|
|
|
int offset_ms;
|
|
|
|
// A session will have multiple playback sessions when the quality changes
|
|
struct alsa_playback_session *pb;
|
|
|
|
struct alsa_session *next;
|
|
};
|
|
|
|
static struct alsa_session *sessions;
|
|
|
|
static bool alsa_sync_disable;
|
|
static int alsa_latency_history_size;
|
|
|
|
// We will try to play the music with the source quality, but if the card
|
|
// doesn't support that we resample to the fallback quality
|
|
static struct media_quality alsa_fallback_quality = { 44100, 16, 2, 0 };
|
|
static struct media_quality alsa_last_quality;
|
|
|
|
|
|
/* -------------------------------- FORWARDS -------------------------------- */
|
|
|
|
static void
|
|
alsa_status(struct alsa_session *as);
|
|
|
|
|
|
/* ------------------------------- MISC HELPERS ----------------------------- */
|
|
|
|
static void
|
|
dump_config(snd_pcm_t *pcm)
|
|
{
|
|
snd_output_t *output;
|
|
char *debug_pcm_cfg;
|
|
int ret;
|
|
|
|
// Dump PCM config data for E_DBG logging
|
|
ret = snd_output_buffer_open(&output);
|
|
if (ret == 0)
|
|
{
|
|
if (snd_pcm_dump_setup(pcm, output) == 0)
|
|
{
|
|
snd_output_buffer_string(output, &debug_pcm_cfg);
|
|
DPRINTF(E_DBG, L_LAUDIO, "Dump of sound device config:\n%s\n", debug_pcm_cfg);
|
|
}
|
|
|
|
snd_output_close(output);
|
|
}
|
|
}
|
|
|
|
static void
|
|
dump_card(int card, snd_ctl_card_info_t *info)
|
|
{
|
|
char hwdev[14]; // 'hw:' (3) + max_uint (10)
|
|
snd_ctl_t *hdl;
|
|
snd_mixer_t *mixer;
|
|
snd_mixer_elem_t *elem;
|
|
char mixerstr[256];
|
|
int err;
|
|
|
|
snprintf(hwdev, sizeof(hwdev), "hw:%d", card);
|
|
|
|
err = snd_ctl_open(&hdl, hwdev, 0);
|
|
if (err < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Failed to probe ALSA card=%d - %s\n", card, snd_strerror(err));
|
|
return;
|
|
}
|
|
|
|
err = snd_ctl_card_info(hdl, info);
|
|
if (err < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Failed to probe ALSA (info) card=%d - %s\n", card, snd_strerror(err));
|
|
goto error;
|
|
}
|
|
|
|
err = snd_mixer_open(&mixer, 0);
|
|
if (err < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Failed to probe ALSA (mixer open) card=%d - %s\n", card, snd_strerror(err));
|
|
goto error;
|
|
}
|
|
|
|
err = snd_mixer_attach(mixer, hwdev);
|
|
if (err < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Failed to probe ALSA (mixer attach) card=%d - %s\n", card, snd_strerror(err));
|
|
goto errormixer;
|
|
}
|
|
|
|
err = snd_mixer_selem_register(mixer, NULL, NULL);
|
|
if (err < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Failed to probe ALSA (mixer setup) card=%d - %s\n", card, snd_strerror(err));
|
|
goto errormixer;
|
|
}
|
|
|
|
err = snd_mixer_load(mixer);
|
|
if (err < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Failed to probe ALSA (mixer setup) card=%d - %s\n", card, snd_strerror(err));
|
|
goto errormixer;
|
|
}
|
|
|
|
memset(mixerstr, 0, sizeof(mixerstr));
|
|
for (elem = snd_mixer_first_elem(mixer); elem; elem = snd_mixer_elem_next(elem))
|
|
{
|
|
if (snd_mixer_selem_has_common_volume(elem) || !snd_mixer_selem_has_playback_volume(elem))
|
|
continue;
|
|
|
|
safe_snprintf_cat(mixerstr, sizeof(mixerstr), " '%s'", snd_mixer_selem_get_name(elem));
|
|
}
|
|
|
|
if (mixerstr[0] == '\0')
|
|
sprintf(mixerstr, " (no mixers found)");
|
|
|
|
DPRINTF(E_INFO, L_LAUDIO, "Available ALSA playback mixer(s) on %s CARD=%s (%s):%s\n", hwdev, snd_ctl_card_info_get_id(info), snd_ctl_card_info_get_name(info), mixerstr);
|
|
|
|
errormixer:
|
|
snd_mixer_close(mixer);
|
|
error:
|
|
snd_ctl_close(hdl);
|
|
}
|
|
|
|
// Walk all the alsa devices here and log valid playback mixers
|
|
static void
|
|
cards_list()
|
|
{
|
|
snd_ctl_card_info_t *info = NULL;
|
|
int card = 0;
|
|
|
|
snd_ctl_card_info_alloca(&info);
|
|
if (!info)
|
|
return;
|
|
|
|
while (card >= 0)
|
|
{
|
|
dump_card(card, info);
|
|
|
|
if (snd_card_next(&card) < 0)
|
|
break;
|
|
}
|
|
}
|
|
|
|
static snd_pcm_format_t
|
|
bps2format(int bits_per_sample)
|
|
{
|
|
if (bits_per_sample == 16)
|
|
return SND_PCM_FORMAT_S16_LE;
|
|
else if (bits_per_sample == 24)
|
|
return SND_PCM_FORMAT_S24_3LE;
|
|
else if (bits_per_sample == 32)
|
|
return SND_PCM_FORMAT_S32_LE;
|
|
else
|
|
return SND_PCM_FORMAT_UNKNOWN;
|
|
}
|
|
|
|
|
|
/* from alsa-utils/alsamixer/volume_mapping.c
|
|
*
|
|
* The mapping is designed so that the position in the interval is proportional
|
|
* to the volume as a human ear would perceive it (i.e., the position is the
|
|
* cubic root of the linear sample multiplication factor). For controls with
|
|
* a small range (24 dB or less), the mapping is linear in the dB values so
|
|
* that each step has the same size visually. Only for controls without dB
|
|
* information, a linear mapping of the hardware volume register values is used
|
|
* (this is the same algorithm as used in the old alsamixer).
|
|
*
|
|
* When setting the volume, 'dir' is the rounding direction:
|
|
* -1/0/1 = down/nearest/up.
|
|
*/
|
|
static inline bool
|
|
use_linear_dB_scale(long dBmin, long dBmax)
|
|
{
|
|
return dBmax - dBmin <= MAX_LINEAR_DB_SCALE * 100;
|
|
}
|
|
|
|
static long
|
|
lrint_dir(double x, int dir)
|
|
{
|
|
if (dir > 0)
|
|
return lrint(ceil(x));
|
|
else if (dir < 0)
|
|
return lrint(floor(x));
|
|
else
|
|
return lrint(x);
|
|
}
|
|
|
|
// from alsamixer/volume-mapping.c, sets volume in line with human perception
|
|
static int
|
|
volume_normalized_set(snd_mixer_elem_t *elem, double volume, int dir)
|
|
{
|
|
long min, max, value;
|
|
double min_norm;
|
|
int err;
|
|
|
|
err = snd_mixer_selem_get_playback_dB_range(elem, &min, &max);
|
|
if (err < 0 || min >= max)
|
|
{
|
|
err = snd_mixer_selem_get_playback_volume_range(elem, &min, &max);
|
|
if (err < 0)
|
|
return err;
|
|
|
|
value = lrint_dir(volume * (max - min), dir) + min;
|
|
return snd_mixer_selem_set_playback_volume_all(elem, value);
|
|
}
|
|
|
|
// Corner case from mpd - log10() expects non-zero
|
|
if (volume <= 0)
|
|
return snd_mixer_selem_set_playback_dB_all(elem, min, dir);
|
|
else if (volume >= 1)
|
|
return snd_mixer_selem_set_playback_dB_all(elem, max, dir);
|
|
|
|
if (use_linear_dB_scale(min, max))
|
|
{
|
|
value = lrint_dir(volume * (max - min), dir) + min;
|
|
return snd_mixer_selem_set_playback_dB_all(elem, value, dir);
|
|
}
|
|
|
|
if (min != SND_CTL_TLV_DB_GAIN_MUTE)
|
|
{
|
|
min_norm = pow(10, (min - max) / 6000.0);
|
|
volume = volume * (1 - min_norm) + min_norm;
|
|
}
|
|
|
|
value = lrint_dir(6000.0 * log10(volume), dir) + max;
|
|
return snd_mixer_selem_set_playback_dB_all(elem, value, dir);
|
|
}
|
|
|
|
static int
|
|
volume_set(struct alsa_mixer *mixer, int volume)
|
|
{
|
|
int ret;
|
|
|
|
snd_mixer_handle_events(mixer->hdl);
|
|
|
|
if (!snd_mixer_selem_is_active(mixer->vol_elem))
|
|
return -1;
|
|
|
|
DPRINTF(E_DBG, L_LAUDIO, "Setting ALSA volume to %d\n", volume);
|
|
|
|
ret = volume_normalized_set(mixer->vol_elem, volume >= 0 && volume <= 100 ? volume/100.0 : 0.75, 0);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to set ALSA volume to %d\n: %s", volume, snd_strerror(ret));
|
|
return -1;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
mixer_open(struct alsa_mixer *mixer, const char *mixer_device_name, const char *mixer_name)
|
|
{
|
|
snd_mixer_t *mixer_hdl;
|
|
snd_mixer_elem_t *vol_elem;
|
|
snd_mixer_elem_t *elem;
|
|
snd_mixer_elem_t *master;
|
|
snd_mixer_elem_t *pcm;
|
|
snd_mixer_elem_t *custom;
|
|
snd_mixer_selem_id_t *sid;
|
|
long vol_min;
|
|
long vol_max;
|
|
int ret;
|
|
|
|
ret = snd_mixer_open(&mixer_hdl, 0);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to open mixer: %s\n", snd_strerror(ret));
|
|
return -1;
|
|
}
|
|
|
|
ret = snd_mixer_attach(mixer_hdl, mixer_device_name);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to attach mixer '%s': %s\n", mixer_device_name, snd_strerror(ret));
|
|
goto out_close;
|
|
}
|
|
|
|
ret = snd_mixer_selem_register(mixer_hdl, NULL, NULL);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to register mixer '%s': %s\n", mixer_device_name, snd_strerror(ret));
|
|
goto out_detach;
|
|
}
|
|
|
|
ret = snd_mixer_load(mixer_hdl);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to load mixer '%s': %s\n", mixer_device_name, snd_strerror(ret));
|
|
goto out_detach;
|
|
}
|
|
|
|
// Grab interesting elements
|
|
snd_mixer_selem_id_alloca(&sid);
|
|
|
|
pcm = NULL;
|
|
master = NULL;
|
|
custom = NULL;
|
|
for (elem = snd_mixer_first_elem(mixer_hdl); elem; elem = snd_mixer_elem_next(elem))
|
|
{
|
|
snd_mixer_selem_get_id(elem, sid);
|
|
|
|
if (mixer_name && (strcmp(snd_mixer_selem_id_get_name(sid), mixer_name) == 0))
|
|
{
|
|
custom = elem;
|
|
break;
|
|
}
|
|
else if (strcmp(snd_mixer_selem_id_get_name(sid), "PCM") == 0)
|
|
pcm = elem;
|
|
else if (strcmp(snd_mixer_selem_id_get_name(sid), "Master") == 0)
|
|
master = elem;
|
|
}
|
|
|
|
if (mixer_name)
|
|
{
|
|
if (custom)
|
|
vol_elem = custom;
|
|
else
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to open configured mixer element '%s'\n", mixer_name);
|
|
goto out_detach;
|
|
}
|
|
}
|
|
else if (pcm)
|
|
vol_elem = pcm;
|
|
else if (master)
|
|
vol_elem = master;
|
|
else
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Failed to open PCM or Master mixer element\n");
|
|
goto out_detach;
|
|
}
|
|
|
|
// Get min & max volume
|
|
snd_mixer_selem_get_playback_volume_range(vol_elem, &vol_min, &vol_max);
|
|
|
|
// All done, export
|
|
mixer->hdl = mixer_hdl;
|
|
mixer->vol_elem = vol_elem;
|
|
mixer->vol_min = vol_min;
|
|
mixer->vol_max = vol_max;
|
|
|
|
return 0;
|
|
|
|
out_detach:
|
|
snd_mixer_detach(mixer_hdl, mixer_device_name);
|
|
out_close:
|
|
snd_mixer_close(mixer_hdl);
|
|
|
|
return -1;
|
|
}
|
|
|
|
static void
|
|
mixer_close(struct alsa_mixer *mixer, const char *mixer_device_name)
|
|
{
|
|
if (!mixer || !mixer->hdl)
|
|
return;
|
|
|
|
snd_mixer_detach(mixer->hdl, mixer_device_name);
|
|
snd_mixer_close(mixer->hdl);
|
|
}
|
|
|
|
static int
|
|
pcm_open(snd_pcm_t **pcm, const char *device_name, struct media_quality *quality)
|
|
{
|
|
snd_pcm_t *hdl;
|
|
snd_pcm_hw_params_t *hw_params;
|
|
snd_pcm_uframes_t bufsize;
|
|
int ret;
|
|
|
|
ret = snd_pcm_open(&hdl, device_name, SND_PCM_STREAM_PLAYBACK, 0);
|
|
if (ret < 0)
|
|
{
|
|
if (ret == -EBUSY)
|
|
return ALSA_ERROR_DEVICE_BUSY;
|
|
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not open playback device '%s': %s\n", device_name, snd_strerror(ret));
|
|
return ALSA_ERROR_DEVICE;
|
|
}
|
|
|
|
// HW params
|
|
ret = snd_pcm_hw_params_malloc(&hw_params);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not allocate hw params: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params_any(hdl, hw_params);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not retrieve hw params: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params_set_access(hdl, hw_params, SND_PCM_ACCESS_RW_INTERLEAVED);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set access method: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params_set_format(hdl, hw_params, bps2format(quality->bits_per_sample));
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set format (bits per sample %d): %s\n", quality->bits_per_sample, snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params_set_channels(hdl, hw_params, quality->channels);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set stereo output: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params_set_rate(hdl, hw_params, quality->sample_rate, 0);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Hardware doesn't support %u Hz: %s\n", quality->sample_rate, snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params_get_buffer_size_max(hw_params, &bufsize);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not get max buffer size: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
// Enable this line to simulate devices with low buffer size
|
|
//bufsize = 32768;
|
|
|
|
ret = snd_pcm_hw_params_set_buffer_size_max(hdl, hw_params, &bufsize);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set buffer size to max: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_hw_params(hdl, hw_params);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set hw params in pcm_open(): %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
snd_pcm_hw_params_free(hw_params);
|
|
|
|
*pcm = hdl;
|
|
|
|
return 0;
|
|
|
|
out_fail:
|
|
snd_pcm_hw_params_free(hw_params);
|
|
snd_pcm_close(hdl);
|
|
|
|
return ALSA_ERROR_DEVICE;
|
|
}
|
|
|
|
static void
|
|
pcm_close(snd_pcm_t *hdl)
|
|
{
|
|
if (!hdl)
|
|
return;
|
|
|
|
snd_pcm_close(hdl);
|
|
}
|
|
|
|
static int
|
|
pcm_configure(snd_pcm_t *hdl)
|
|
{
|
|
snd_pcm_sw_params_t *sw_params;
|
|
int ret;
|
|
|
|
ret = snd_pcm_sw_params_malloc(&sw_params);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not allocate sw params: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_sw_params_current(hdl, sw_params);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not retrieve current sw params: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_sw_params_set_tstamp_type(hdl, sw_params, SND_PCM_TSTAMP_TYPE_MONOTONIC);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set tstamp type: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_sw_params_set_tstamp_mode(hdl, sw_params, SND_PCM_TSTAMP_ENABLE);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set tstamp mode: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
ret = snd_pcm_sw_params(hdl, sw_params);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not set sw params: %s\n", snd_strerror(ret));
|
|
goto out_fail;
|
|
}
|
|
|
|
snd_pcm_sw_params_free(sw_params);
|
|
|
|
return 0;
|
|
|
|
out_fail:
|
|
snd_pcm_sw_params_free(sw_params);
|
|
|
|
return -1;
|
|
}
|
|
|
|
static void
|
|
playback_session_free(struct alsa_playback_session *pb)
|
|
{
|
|
if (!pb)
|
|
return;
|
|
|
|
// Unsubscribe from qualities that sync_correct() might have requested
|
|
if (pb->sync_resample_step != 0)
|
|
outputs_quality_unsubscribe(&pb->quality);
|
|
|
|
pcm_close(pb->pcm);
|
|
|
|
ringbuffer_free(&pb->prebuf, 1);
|
|
|
|
free(pb->latency_history);
|
|
free(pb);
|
|
}
|
|
|
|
static void
|
|
playback_session_remove(struct alsa_session *as, struct alsa_playback_session *pb)
|
|
{
|
|
struct alsa_playback_session *s;
|
|
|
|
DPRINTF(E_DBG, L_LAUDIO, "Removing playback session (quality %d/%d/%d) from ALSA device '%s'\n",
|
|
pb->quality.sample_rate, pb->quality.bits_per_sample, pb->quality.channels, as->devname);
|
|
|
|
if (pb == as->pb)
|
|
as->pb = as->pb->next;
|
|
else
|
|
{
|
|
for (s = as->pb; s && (s->next != pb); s = s->next)
|
|
; /* EMPTY */
|
|
|
|
if (!s)
|
|
DPRINTF(E_WARN, L_LAUDIO, "WARNING: struct alsa_playback_session not found in list; BUG!\n");
|
|
else
|
|
s->next = pb->next;
|
|
}
|
|
|
|
playback_session_free(pb);
|
|
}
|
|
|
|
static void
|
|
playback_session_remove_all(struct alsa_session *as)
|
|
{
|
|
struct alsa_playback_session *s;
|
|
|
|
for (s = as->pb; s; s = as->pb)
|
|
{
|
|
as->pb = s->next;
|
|
playback_session_free(s);
|
|
}
|
|
}
|
|
|
|
static int
|
|
playback_session_add(struct alsa_session *as, struct media_quality *quality, struct timespec pts)
|
|
{
|
|
struct alsa_playback_session *pb;
|
|
struct alsa_playback_session *tail_pb;
|
|
struct timespec ts;
|
|
snd_pcm_sframes_t offset_nsamp;
|
|
size_t size;
|
|
int ret;
|
|
|
|
DPRINTF(E_DBG, L_LAUDIO, "Adding playback session (quality %d/%d/%d) to ALSA device '%s'\n",
|
|
quality->sample_rate, quality->bits_per_sample, quality->channels, as->devname);
|
|
|
|
CHECK_NULL(L_LAUDIO, pb = calloc(1, sizeof(struct alsa_playback_session)));
|
|
CHECK_NULL(L_LAUDIO, pb->latency_history = calloc(alsa_latency_history_size, sizeof(double)));
|
|
|
|
ret = pcm_open(&pb->pcm, as->devname, quality);
|
|
if (ret == ALSA_ERROR_DEVICE_BUSY)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "ALSA device '%s' won't open due to existing session (no support for concurrent audio), truncating audio\n", as->devname);
|
|
playback_session_remove_all(as);
|
|
ret = pcm_open(&pb->pcm, as->devname, quality);
|
|
if (ret == ALSA_ERROR_DEVICE_BUSY)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "ALSA device '%s' failed: Device still busy after closing previous sessions\n", as->devname);
|
|
goto error;
|
|
}
|
|
}
|
|
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Device '%s' does not support quality (%d/%d/%d), falling back to default\n", as->devname, quality->sample_rate, quality->bits_per_sample, quality->channels);
|
|
ret = pcm_open(&pb->pcm, as->devname, &alsa_fallback_quality);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "ALSA device failed setting fallback quality\n");
|
|
goto error;
|
|
}
|
|
|
|
pb->quality = alsa_fallback_quality;
|
|
}
|
|
else
|
|
pb->quality = *quality;
|
|
|
|
// If this fails it just means we won't get timestamps, which we can handle
|
|
pcm_configure(pb->pcm);
|
|
|
|
dump_config(pb->pcm);
|
|
|
|
// Time stamps used for syncing, here we set when playback should start
|
|
ts.tv_sec = OUTPUTS_BUFFER_DURATION;
|
|
ts.tv_nsec = (uint64_t)as->offset_ms * 1000000UL;
|
|
pb->stamp_pts = timespec_add(pts, ts);
|
|
|
|
// The difference between pos and start pos should match the 2 second buffer
|
|
// that AirPlay uses (OUTPUTS_BUFFER_DURATION) + user configured offset_ms. We
|
|
// will not use alsa's buffer for the initial buffering, because my sound
|
|
// card's start_threshold is not to be counted on. Instead we allocate our own
|
|
// buffer, and when it is time to play we write as much as we can to alsa's
|
|
// buffer.
|
|
offset_nsamp = (as->offset_ms * pb->quality.sample_rate / 1000);
|
|
|
|
pb->buffer_nsamp = OUTPUTS_BUFFER_DURATION * pb->quality.sample_rate + offset_nsamp;
|
|
size = STOB(pb->buffer_nsamp, pb->quality.bits_per_sample, pb->quality.channels);
|
|
ringbuffer_init(&pb->prebuf, size);
|
|
|
|
// Add to the end of the list, because when we iterate through it in
|
|
// alsa_write() we want to write data from the oldest playback session first
|
|
if (as->pb)
|
|
{
|
|
for (tail_pb = as->pb; tail_pb->next; tail_pb = tail_pb->next)
|
|
; // Fast forward
|
|
tail_pb->next = pb;
|
|
}
|
|
else
|
|
as->pb = pb;
|
|
|
|
return 0;
|
|
|
|
error:
|
|
playback_session_free(pb);
|
|
|
|
return -1;
|
|
}
|
|
|
|
// This function writes the sample buf into either the prebuffer or directly to
|
|
// ALSA, depending on how much room there is in ALSA, and whether we are
|
|
// prebuffering or not. It also transfers from the the prebuffer to ALSA, if
|
|
// needed. Returns 0 on success, negative on error.
|
|
static int
|
|
buffer_write(struct alsa_playback_session *pb, struct output_data *odata, snd_pcm_sframes_t avail)
|
|
{
|
|
uint8_t *buf;
|
|
ssize_t bufsize;
|
|
size_t wrote;
|
|
snd_pcm_sframes_t nsamp;
|
|
snd_pcm_sframes_t ret;
|
|
|
|
// Prebuffering, no actual writing
|
|
if (avail == 0)
|
|
{
|
|
wrote = ringbuffer_write(&pb->prebuf, odata->buffer, odata->bufsize);
|
|
if (wrote < odata->bufsize)
|
|
DPRINTF(E_WARN, L_LAUDIO, "Bug! Partial prebuf write %zu/%zu\n", wrote, odata->bufsize);
|
|
|
|
nsamp = snd_pcm_bytes_to_frames(pb->pcm, wrote);
|
|
return nsamp;
|
|
}
|
|
|
|
// Read from prebuffer if it has data and write to device
|
|
if (pb->prebuf.read_avail != 0)
|
|
{
|
|
// Maximum amount of bytes we want to read
|
|
bufsize = snd_pcm_frames_to_bytes(pb->pcm, avail);
|
|
|
|
bufsize = ringbuffer_read(&buf, bufsize, &pb->prebuf);
|
|
if (bufsize == 0)
|
|
return 0;
|
|
|
|
// DPRINTF(E_DBG, L_LAUDIO, "Writing prebuffer (read_avail=%zu, bufsize=%zu, avail=%li)\n", pb->prebuf.read_avail, bufsize, avail);
|
|
|
|
nsamp = snd_pcm_bytes_to_frames(pb->pcm, bufsize);
|
|
|
|
ret = snd_pcm_writei(pb->pcm, buf, nsamp);
|
|
if (ret < 0)
|
|
return ret;
|
|
|
|
avail -= ret;
|
|
}
|
|
|
|
// Write to prebuffer if device buffer does not have availability or if we are
|
|
// still prebuffering. Note that if the prebuffer doesn't have enough room,
|
|
// which can happen if avail stays low, i.e. device buffer is overrunning,
|
|
// then the extra samples get dropped
|
|
if (odata->samples > avail || pb->prebuf.read_avail != 0)
|
|
{
|
|
wrote = ringbuffer_write(&pb->prebuf, odata->buffer, odata->bufsize);
|
|
if (wrote < odata->bufsize)
|
|
DPRINTF(E_WARN, L_LAUDIO, "Dropped %zu bytes of audio - device is overrunning!\n", odata->bufsize - wrote);
|
|
|
|
return odata->samples;
|
|
}
|
|
|
|
nsamp = snd_pcm_bytes_to_frames(pb->pcm, odata->bufsize);
|
|
|
|
ret = snd_pcm_writei(pb->pcm, odata->buffer, nsamp);
|
|
if (ret < 0)
|
|
return ret;
|
|
|
|
if (ret != odata->samples)
|
|
DPRINTF(E_WARN, L_LAUDIO, "ALSA partial write detected\n");
|
|
|
|
return ret;
|
|
}
|
|
|
|
static enum alsa_sync_state
|
|
sync_check(double *drift, double *latency, struct alsa_playback_session *pb, snd_pcm_sframes_t delay)
|
|
{
|
|
enum alsa_sync_state sync;
|
|
struct timespec ts;
|
|
int elapsed;
|
|
uint64_t cur_pos;
|
|
uint64_t exp_pos;
|
|
int32_t diff;
|
|
double r2;
|
|
int ret;
|
|
|
|
// Would be nice to use snd_pcm_status_get_audio_htstamp here, but it doesn't
|
|
// seem to be supported on my computer
|
|
clock_gettime(CLOCK_MONOTONIC, &ts);
|
|
|
|
// Here we calculate elapsed time since last reference position (which is
|
|
// equal to playback start time, unless we have reset due to sync correction),
|
|
// taking into account buffer time and configuration of offset_ms. We then
|
|
// calculate our expected position based on elapsed time, and if different
|
|
// from where we are + what is in the buffers then ALSA is out of sync.
|
|
elapsed = (ts.tv_sec - pb->stamp_pts.tv_sec) * 1000L + (ts.tv_nsec - pb->stamp_pts.tv_nsec) / 1000000;
|
|
if (elapsed < 0)
|
|
return ALSA_SYNC_OK;
|
|
|
|
cur_pos = (uint64_t)pb->pos - pb->stamp_pos - (delay + BTOS(pb->prebuf.read_avail, pb->quality.bits_per_sample, pb->quality.channels));
|
|
exp_pos = (uint64_t)elapsed * pb->quality.sample_rate / 1000;
|
|
diff = cur_pos - exp_pos;
|
|
|
|
DPRINTF(E_SPAM, L_LAUDIO, "counter %d/%d, stamp %lu:%lu, now %lu:%lu, elapsed is %d ms, cur_pos=%" PRIu64 ", exp_pos=%" PRIu64 ", diff=%d\n",
|
|
pb->latency_counter, alsa_latency_history_size, pb->stamp_pts.tv_sec, pb->stamp_pts.tv_nsec / 1000000, ts.tv_sec, ts.tv_nsec / 1000000, elapsed, cur_pos, exp_pos, diff);
|
|
|
|
// Add the latency to our measurement history
|
|
pb->latency_history[pb->latency_counter] = (double)diff;
|
|
pb->latency_counter++;
|
|
|
|
// Haven't collected enough samples for sync evaluation yet, so just return
|
|
if (pb->latency_counter < alsa_latency_history_size)
|
|
return ALSA_SYNC_OK;
|
|
|
|
pb->latency_counter = 0;
|
|
|
|
ret = linear_regression(drift, latency, &r2, NULL, pb->latency_history, alsa_latency_history_size);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "Linear regression of collected latency samples failed\n");
|
|
return ALSA_SYNC_OK;
|
|
}
|
|
|
|
// Set *latency to the "average" within the period
|
|
*latency = (*drift) * alsa_latency_history_size / 2 + (*latency);
|
|
|
|
if (fabs(*latency) < ALSA_MAX_LATENCY && fabs(*drift) < ALSA_MAX_DRIFT)
|
|
sync = ALSA_SYNC_OK; // If both latency and drift are within thresholds -> no action
|
|
else if (*latency > 0 && *drift > 0)
|
|
sync = ALSA_SYNC_AHEAD;
|
|
else if (*latency < 0 && *drift < 0)
|
|
sync = ALSA_SYNC_BEHIND;
|
|
else
|
|
sync = ALSA_SYNC_OK; // Drift is counteracting latency -> no action
|
|
|
|
if (sync != ALSA_SYNC_OK && r2 < ALSA_MAX_VARIANCE)
|
|
{
|
|
DPRINTF(E_DBG, L_LAUDIO, "Too much variance in latency measurements (r2=%f/%f), won't try to compensate\n", r2, ALSA_MAX_VARIANCE);
|
|
sync = ALSA_SYNC_OK;
|
|
}
|
|
|
|
DPRINTF(E_DBG, L_LAUDIO, "Sync check result: drift=%f, latency=%f, r2=%f, sync=%d\n", *drift, *latency, r2, sync);
|
|
|
|
return sync;
|
|
}
|
|
|
|
static void
|
|
sync_correct(struct alsa_playback_session *pb, double drift, double latency, struct timespec pts, snd_pcm_sframes_t delay)
|
|
{
|
|
int step;
|
|
int sign;
|
|
int ret;
|
|
|
|
// We change the sample_rate in steps that are a multiple of 50. So we might
|
|
// step 44100 -> 44000 -> 40900 -> 44000 -> 44100. If we used percentages to
|
|
// to step, we would have to deal with rounding; we don't want to step 44100
|
|
// -> 39996 -> 44099.
|
|
step = ALSA_RESAMPLE_STEP_MULTIPLE * (pb->quality.sample_rate / 20000);
|
|
|
|
sign = (drift < 0) ? -1 : 1;
|
|
|
|
if (abs(pb->sync_resample_step) == ALSA_RESAMPLE_STEP_MAX)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "The sync of ALSA device cannot be corrected (drift=%f, latency=%f)\n", drift, latency);
|
|
pb->sync_resample_step += sign;
|
|
return;
|
|
}
|
|
else if (abs(pb->sync_resample_step) > ALSA_RESAMPLE_STEP_MAX)
|
|
return; // Don't do anything, we have given up
|
|
|
|
// Step 0 is the original audio quality (or the fallback quality), which we
|
|
// will just keep receiving
|
|
if (pb->sync_resample_step != 0)
|
|
outputs_quality_unsubscribe(&pb->quality);
|
|
|
|
pb->sync_resample_step += sign;
|
|
pb->quality.sample_rate += sign * step;
|
|
|
|
if (pb->sync_resample_step != 0)
|
|
{
|
|
ret = outputs_quality_subscribe(&pb->quality);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Error adjusting sample rate to %d to maintain sync\n", pb->quality.sample_rate);
|
|
return;
|
|
}
|
|
}
|
|
|
|
// Reset position so next sync_correct latency correction is only based on
|
|
// what has elapsed since our correction
|
|
pb->stamp_pos = (uint64_t)pb->pos - (delay + BTOS(pb->prebuf.read_avail, pb->quality.bits_per_sample, pb->quality.channels));;
|
|
pb->stamp_pts = pts;
|
|
|
|
DPRINTF(E_INFO, L_LAUDIO, "Adjusted sample rate to %d to sync ALSA device (drift=%f, latency=%f)\n", pb->quality.sample_rate, drift, latency);
|
|
}
|
|
|
|
static int
|
|
playback_drain(struct alsa_playback_session *pb)
|
|
{
|
|
uint8_t *buf;
|
|
ssize_t bufsize;
|
|
snd_pcm_state_t state;
|
|
snd_pcm_sframes_t avail;
|
|
snd_pcm_sframes_t delay;
|
|
snd_pcm_sframes_t nsamp;
|
|
int ret;
|
|
|
|
state = snd_pcm_state(pb->pcm);
|
|
if (state == SND_PCM_STATE_DRAINING)
|
|
return 0;
|
|
else if (state != SND_PCM_STATE_RUNNING)
|
|
return ALSA_ERROR_SESSION; // We are probably done draining, so this makes the caller close the pb session
|
|
|
|
// If the prebuffer is empty we are done writing to this pcm
|
|
if (pb->prebuf.read_avail == 0)
|
|
{
|
|
snd_pcm_drain(pb->pcm); // Plays pending frames and then stops the pcm
|
|
return 0;
|
|
}
|
|
|
|
ret = snd_pcm_avail_delay(pb->pcm, &avail, &delay);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Error getting avail/delay: %s\n", snd_strerror(ret));
|
|
return ALSA_ERROR_SESSION;
|
|
}
|
|
|
|
// Maximum amount of bytes we want to read
|
|
bufsize = snd_pcm_frames_to_bytes(pb->pcm, avail);
|
|
|
|
bufsize = ringbuffer_read(&buf, bufsize, &pb->prebuf);
|
|
if (bufsize == 0)
|
|
return 0; // avail too low to actually write anything
|
|
|
|
// DPRINTF(E_DBG, L_LAUDIO, "Draining prebuffer (read_avail=%zu, bufsize=%zu, avail=%li)\n", pb->prebuf.read_avail / 4, bufsize, avail);
|
|
|
|
nsamp = snd_pcm_bytes_to_frames(pb->pcm, bufsize);
|
|
|
|
ret = snd_pcm_writei(pb->pcm, buf, nsamp);
|
|
|
|
return ((ret < 0) ? ALSA_ERROR_SESSION : 0);
|
|
}
|
|
|
|
static int
|
|
playback_write(struct alsa_playback_session *pb, struct output_buffer *obuf)
|
|
{
|
|
snd_pcm_sframes_t avail;
|
|
snd_pcm_sframes_t delay;
|
|
enum alsa_sync_state sync;
|
|
double drift;
|
|
double latency;
|
|
bool prebuffering;
|
|
int ret;
|
|
int i;
|
|
|
|
// Find the quality we want
|
|
for (i = 0; obuf->data[i].buffer; i++)
|
|
{
|
|
if (quality_is_equal(&pb->quality, &obuf->data[i].quality))
|
|
break;
|
|
}
|
|
|
|
if (!obuf->data[i].buffer)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Output not delivering required data quality, aborting\n");
|
|
return -1;
|
|
}
|
|
|
|
prebuffering = (pb->pos + obuf->data[i].bufsize <= pb->buffer_nsamp);
|
|
if (prebuffering)
|
|
{
|
|
// Can never fail since we don't actually write to the device
|
|
pb->pos += buffer_write(pb, &obuf->data[i], 0);
|
|
return 0;
|
|
}
|
|
|
|
ret = snd_pcm_avail_delay(pb->pcm, &avail, &delay);
|
|
if (ret < 0)
|
|
goto alsa_error;
|
|
|
|
// Check sync each second (or if this is first write where last_pts is zero)
|
|
if (!alsa_sync_disable && (obuf->pts.tv_sec != pb->last_pts.tv_sec))
|
|
{
|
|
sync = sync_check(&drift, &latency, pb, delay);
|
|
if (sync != ALSA_SYNC_OK)
|
|
sync_correct(pb, drift, latency, obuf->pts, delay);
|
|
|
|
pb->last_pts = obuf->pts;
|
|
}
|
|
|
|
ret = buffer_write(pb, &obuf->data[i], avail);
|
|
if (ret < 0)
|
|
goto alsa_error;
|
|
|
|
pb->pos += ret;
|
|
|
|
return 0;
|
|
|
|
alsa_error:
|
|
if (ret == -EPIPE)
|
|
{
|
|
DPRINTF(E_WARN, L_LAUDIO, "ALSA buffer underrun, restarting session\n");
|
|
return ALSA_ERROR_UNDERRUN;
|
|
}
|
|
|
|
DPRINTF(E_LOG, L_LAUDIO, "ALSA write error: %s\n", snd_strerror(ret));
|
|
return ALSA_ERROR_WRITE;
|
|
}
|
|
|
|
|
|
/* ---------------------------- SESSION HANDLING ---------------------------- */
|
|
|
|
static void
|
|
alsa_session_free(struct alsa_session *as)
|
|
{
|
|
if (!as)
|
|
return;
|
|
|
|
outputs_quality_unsubscribe(&alsa_fallback_quality);
|
|
|
|
playback_session_remove_all(as);
|
|
|
|
mixer_close(&as->mixer, as->mixer_device_name);
|
|
|
|
free(as);
|
|
}
|
|
|
|
static void
|
|
alsa_session_cleanup(struct alsa_session *as)
|
|
{
|
|
struct alsa_session *s;
|
|
|
|
if (as == sessions)
|
|
sessions = sessions->next;
|
|
else
|
|
{
|
|
for (s = sessions; s && (s->next != as); s = s->next)
|
|
; /* EMPTY */
|
|
|
|
if (!s)
|
|
DPRINTF(E_WARN, L_LAUDIO, "WARNING: struct alsa_session not found in list; BUG!\n");
|
|
else
|
|
s->next = as->next;
|
|
}
|
|
|
|
outputs_device_session_remove(as->device_id);
|
|
|
|
alsa_session_free(as);
|
|
}
|
|
|
|
static struct alsa_session *
|
|
alsa_session_make(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as;
|
|
struct alsa_extra *ae;
|
|
int ret;
|
|
|
|
ae = device->extra_device_info;
|
|
|
|
CHECK_NULL(L_LAUDIO, as = calloc(1, sizeof(struct alsa_session)));
|
|
|
|
as->device_id = device->id;
|
|
as->callback_id = callback_id;
|
|
|
|
as->devname = ae->card_name;
|
|
as->mixer_name = ae->mixer_name;
|
|
as->mixer_device_name = ae->mixer_device_name;
|
|
as->offset_ms = ae->offset_ms;
|
|
|
|
ret = mixer_open(&as->mixer, as->mixer_device_name, as->mixer_name);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not open mixer '%s' ('%s')\n", as->mixer_device_name, as->mixer_name);
|
|
goto error_free_session;
|
|
}
|
|
|
|
ret = outputs_quality_subscribe(&alsa_fallback_quality);
|
|
if (ret < 0)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "Could not subscribe to fallback audio quality\n");
|
|
goto error_mixer_close;
|
|
}
|
|
|
|
as->state = OUTPUT_STATE_CONNECTED;
|
|
as->next = sessions;
|
|
sessions = as;
|
|
|
|
// as is now the official device session
|
|
outputs_device_session_add(device->id, as);
|
|
|
|
return as;
|
|
|
|
error_mixer_close:
|
|
mixer_close(&as->mixer, as->mixer_device_name);
|
|
error_free_session:
|
|
free(as);
|
|
return NULL;
|
|
}
|
|
|
|
static void
|
|
alsa_status(struct alsa_session *as)
|
|
{
|
|
outputs_cb(as->callback_id, as->device_id, as->state);
|
|
as->callback_id = -1;
|
|
|
|
if (as->state == OUTPUT_STATE_FAILED || as->state == OUTPUT_STATE_STOPPED)
|
|
alsa_session_cleanup(as);
|
|
}
|
|
|
|
|
|
/* ------------------ INTERFACE FUNCTIONS CALLED BY OUTPUTS.C --------------- */
|
|
|
|
static int
|
|
alsa_device_start(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as;
|
|
|
|
as = alsa_session_make(device, callback_id);
|
|
if (!as)
|
|
return -1;
|
|
|
|
volume_set(&as->mixer, device->volume);
|
|
|
|
as->state = OUTPUT_STATE_CONNECTED;
|
|
alsa_status(as);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
alsa_device_stop(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as = device->session;
|
|
|
|
as->callback_id = callback_id;
|
|
as->state = OUTPUT_STATE_STOPPED;
|
|
alsa_status(as); // Will terminate the session since the state is STOPPED
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
alsa_device_flush(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as = device->session;
|
|
|
|
playback_session_remove_all(as);
|
|
|
|
as->callback_id = callback_id;
|
|
as->state = OUTPUT_STATE_CONNECTED;
|
|
alsa_status(as);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
alsa_device_probe(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as;
|
|
|
|
as = alsa_session_make(device, callback_id);
|
|
if (!as)
|
|
return -1;
|
|
|
|
as->state = OUTPUT_STATE_STOPPED;
|
|
alsa_status(as); // Will terminate the session since the state is STOPPED
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
alsa_device_volume_set(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as = device->session;
|
|
|
|
if (!as)
|
|
return 0;
|
|
|
|
volume_set(&as->mixer, device->volume);
|
|
|
|
as->callback_id = callback_id;
|
|
alsa_status(as);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static void
|
|
alsa_device_cb_set(struct output_device *device, int callback_id)
|
|
{
|
|
struct alsa_session *as = device->session;
|
|
|
|
as->callback_id = callback_id;
|
|
}
|
|
|
|
static void
|
|
alsa_device_free_extra(struct output_device *device)
|
|
{
|
|
struct alsa_extra *ae = device->extra_device_info;
|
|
|
|
free(ae);
|
|
}
|
|
|
|
static void
|
|
alsa_write(struct output_buffer *obuf)
|
|
{
|
|
struct alsa_session *as;
|
|
struct alsa_session *as_next;
|
|
struct alsa_playback_session *pb;
|
|
struct alsa_playback_session *pb_next;
|
|
bool quality_changed;
|
|
int ret;
|
|
|
|
quality_changed = !quality_is_equal(&obuf->data[0].quality, &alsa_last_quality);
|
|
alsa_last_quality = obuf->data[0].quality;
|
|
|
|
for (as = sessions; as; as = as->next)
|
|
{
|
|
if (quality_changed || as->state == OUTPUT_STATE_CONNECTED)
|
|
{
|
|
ret = playback_session_add(as, &obuf->data[0].quality, obuf->pts);
|
|
if (ret < 0)
|
|
{
|
|
as->state = OUTPUT_STATE_FAILED;
|
|
continue;
|
|
}
|
|
|
|
as->state = OUTPUT_STATE_STREAMING;
|
|
}
|
|
|
|
for (pb = as->pb; pb; pb = pb_next)
|
|
{
|
|
pb_next = pb->next;
|
|
|
|
// If !pb_next then it means that it is the most recent session, so it
|
|
// is setup with the quality level that matches obuf. The other pb's
|
|
// may still have data that needs to be written before removal.
|
|
if (!pb_next)
|
|
ret = playback_write(pb, obuf);
|
|
else
|
|
ret = playback_drain(pb);
|
|
|
|
if (ret < 0)
|
|
{
|
|
playback_session_remove(as, pb); // pb becomes invalid
|
|
if (ret == ALSA_ERROR_WRITE)
|
|
as->state = OUTPUT_STATE_FAILED;
|
|
else if (ret == ALSA_ERROR_UNDERRUN)
|
|
as->state = OUTPUT_STATE_CONNECTED;
|
|
}
|
|
}
|
|
}
|
|
|
|
// Cleanup failed sessions
|
|
for (as = sessions; as; as = as_next)
|
|
{
|
|
as_next = as->next;
|
|
|
|
if (as->state == OUTPUT_STATE_FAILED)
|
|
alsa_status(as); // as becomes invalid
|
|
}
|
|
}
|
|
|
|
static void
|
|
alsa_device_add(cfg_t* cfg_audio, int id)
|
|
{
|
|
struct output_device *device;
|
|
struct alsa_extra *ae;
|
|
const char *nickname;
|
|
int ret;
|
|
|
|
CHECK_NULL(L_LAUDIO, device = calloc(1, sizeof(struct output_device)));
|
|
CHECK_NULL(L_LAUDIO, ae = calloc(1, sizeof(struct alsa_extra)));
|
|
|
|
device->id = id;
|
|
device->type = OUTPUT_TYPE_ALSA;
|
|
device->type_name = outputs_name(device->type);
|
|
device->extra_device_info = ae;
|
|
|
|
// The audio section will have no title, so there we get the value from the
|
|
// "card" option
|
|
ae->card_name = cfg_title(cfg_audio);
|
|
if (!ae->card_name)
|
|
ae->card_name = cfg_getstr(cfg_audio, "card");
|
|
|
|
nickname = cfg_getstr(cfg_audio, "nickname");
|
|
device->name = strdup(nickname ? nickname : ae->card_name);
|
|
|
|
ae->mixer_name = cfg_getstr(cfg_audio, "mixer");
|
|
ae->mixer_device_name = cfg_getstr(cfg_audio, "mixer_device");
|
|
if (!ae->mixer_device_name || strlen(ae->mixer_device_name) == 0)
|
|
ae->mixer_device_name = ae->card_name;
|
|
|
|
ae->offset_ms = cfg_getint(cfg_audio, "offset_ms");
|
|
if (abs(ae->offset_ms) > 1000)
|
|
{
|
|
DPRINTF(E_LOG, L_LAUDIO, "The ALSA offset_ms (%d) set in the configuration is out of bounds\n", ae->offset_ms);
|
|
ae->offset_ms = 1000 * (ae->offset_ms/abs(ae->offset_ms));
|
|
}
|
|
|
|
DPRINTF(E_INFO, L_LAUDIO, "Adding ALSA device '%s' with name '%s'\n", ae->card_name, device->name);
|
|
|
|
ret = player_device_add(device);
|
|
if (ret < 0)
|
|
outputs_device_free(device);
|
|
}
|
|
|
|
static int
|
|
alsa_init(void)
|
|
{
|
|
cfg_t *cfg_audio;
|
|
cfg_t *cfg_alsasec;
|
|
const char *type;
|
|
int i;
|
|
int alsa_cfg_secn;
|
|
|
|
// Is ALSA enabled in config?
|
|
cfg_audio = cfg_getsec(cfg, "audio");
|
|
type = cfg_getstr(cfg_audio, "type");
|
|
if (type && (strcasecmp(type, "alsa") != 0))
|
|
return -1;
|
|
|
|
cards_list();
|
|
|
|
alsa_sync_disable = cfg_getbool(cfg_audio, "sync_disable");
|
|
alsa_latency_history_size = cfg_getint(cfg_audio, "adjust_period_seconds");
|
|
|
|
alsa_cfg_secn = cfg_size(cfg, "alsa");
|
|
if (alsa_cfg_secn == 0)
|
|
{
|
|
alsa_device_add(cfg_audio, 0);
|
|
}
|
|
else
|
|
{
|
|
for (i = 0; i < alsa_cfg_secn; ++i)
|
|
{
|
|
cfg_alsasec = cfg_getnsec(cfg, "alsa", i);
|
|
alsa_device_add(cfg_alsasec, i);
|
|
}
|
|
}
|
|
|
|
snd_lib_error_set_handler(logger_alsa);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void
|
|
alsa_deinit(void)
|
|
{
|
|
struct alsa_session *as;
|
|
|
|
snd_lib_error_set_handler(NULL);
|
|
|
|
for (as = sessions; sessions; as = sessions)
|
|
{
|
|
sessions = as->next;
|
|
alsa_session_free(as);
|
|
}
|
|
}
|
|
|
|
struct output_definition output_alsa =
|
|
{
|
|
.name = "ALSA",
|
|
.type = OUTPUT_TYPE_ALSA,
|
|
.priority = 3,
|
|
.disabled = 0,
|
|
.init = alsa_init,
|
|
.deinit = alsa_deinit,
|
|
.device_start = alsa_device_start,
|
|
.device_stop = alsa_device_stop,
|
|
.device_flush = alsa_device_flush,
|
|
.device_probe = alsa_device_probe,
|
|
.device_volume_set = alsa_device_volume_set,
|
|
.device_cb_set = alsa_device_cb_set,
|
|
.device_free_extra = alsa_device_free_extra,
|
|
.write = alsa_write,
|
|
};
|