CODEC :音频芯片的控制,比如静音、打开(关闭)ADC(DAC)、设置ADC(DAC)的增益、耳机模式的检测等操作。
I2S :数字音频接口,用于CPU和Codec之间的数字音频流raw data的传输。每当有playback或record操作时,snd_soc_dai_ops.prepare()会被调用,启动I2S总线。
PCM :我不知道为什么会取这个模块名,它其实是定义DMA操作的,用于将音频数据通过DMA传到I2S控制器的FIFO中。
这里的PCM实际是就是更新和管理音频数据流的地址,分配DMA等等,将RAM中存放的音频数据的地址传给I2S,不是PCM协议。
音频数据流向:
| DMA | | I2S/PCM/AC97 |
RAM --------> I2SControllerFIFO -----------------> CODEC ----> SPK/Headset
PCM模块初始化:
struct snd_soc_platform rk29_soc_platform = {
.name = "rockchip-audio",
.pcm_ops = &rockchip_pcm_ops,
.pcm_new = rockchip_pcm_new,
.pcm_free = rockchip_pcm_free_dma_buffers,
};
EXPORT_SYMBOL_GPL(rk29_soc_platform);
static int __init rockchip_soc_platform_init(void)
{
DBG("Enter::%s, %d\n", __FUNCTION__, __LINE__);
return snd_soc_register_platform(&rk29_soc_platform);
}
module_init(rockchip_soc_platform_init);
static void __exit rockchip_soc_platform_exit(void)
{
snd_soc_unregister_platform(&rk29_soc_platform);
}
调用snd_soc_register_platform()向ALSA core注册一个snd_soc_platform结构体。
成员pcm_new需要调用dma_alloc_writecombine()给DMA分配一块write-combining的内存空间,并把这块缓冲区的相关信息保存到substream->dma_buffer中,相当于构造函数。pcm_free则相反。这些成员函数都还算简单,看看代码即可以理解其流程。
snd_pcm_ops
接着我们看一下snd_pcm_ops结构体,该结构体的操作函数集的实现是本模块的主体。
struct snd_pcm_ops {
int (*open)(struct snd_pcm_substream *substream);
int (*close)(struct snd_pcm_substream *substream);
int (*ioctl)(struct snd_pcm_substream * substream,
unsigned int cmd, void *arg);
int (*hw_params)(struct snd_pcm_substream *substream,
struct snd_pcm_hw_params *params);
int (*hw_free)(struct snd_pcm_substream *substream);
int (*prepare)(struct snd_pcm_substream *substream);
int (*trigger)(struct snd_pcm_substream *substream, int cmd);
snd_pcm_uframes_t (*pointer)(struct snd_pcm_substream *substream);
int (*copy)(struct snd_pcm_substream *substream, int channel,
snd_pcm_uframes_t pos,
void __user *buf, snd_pcm_uframes_t count);
int (*silence)(struct snd_pcm_substream *substream, int channel,
snd_pcm_uframes_t pos, snd_pcm_uframes_t count);
struct page *(*page)(struct snd_pcm_substream *substream,
unsigned long offset);
int (*mmap)(struct snd_pcm_substream *substream, struct vm_area_struct *vma);
int (*ack)(struct snd_pcm_substream *substream);
};
open
open函数为PCM模块设定支持的传输模式、数据格式、通道数、period等参数,并为playback/capture stream分配相应的DMA通道。其一般实现如下:
static int rockchip_pcm_open(struct snd_pcm_substream *substream)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct rockchip_runtime_data *prtd;
DBG("Enter::%s----%d\n",__FUNCTION__,__LINE__);
snd_soc_set_runtime_hwparams(substream, &rockchip_pcm_hardware);
prtd = kzalloc(sizeof(struct rockchip_runtime_data), GFP_KERNEL);
if (prtd == NULL)
return -ENOMEM;
spin_lock_init(&prtd->lock);
runtime->private_data = prtd;
return 0;
}
其中硬件参数要根据芯片的数据手册来定义,如:
int snd_soc_set_runtime_hwparams(struct snd_pcm_substream *substream,
const struct snd_pcm_hardware *hw)
{
struct snd_pcm_runtime *runtime = substream->runtime;
runtime->hw.info = hw->info;
runtime->hw.formats = hw->formats;
runtime->hw.period_bytes_min = hw->period_bytes_min;
runtime->hw.period_bytes_max = hw->period_bytes_max;
runtime->hw.periods_min = hw->periods_min;
runtime->hw.periods_max = hw->periods_max;
runtime->hw.buffer_bytes_max = hw->buffer_bytes_max;
runtime->hw.fifo_size = hw->fifo_size;
return 0;
}
关于peroid的概念有这样的描述:The “period” is a term that corresponds to a fragment in the OSS world. The period defines the size at which a PCM interrupt is generated. peroid的概念很重要,建议去alsa官网找相关详细说明了解一下。
上层ALSA lib可以通过接口来获得这些参数的,如snd_pcm_hw_params_get_buffer_size_max()来取得buffer_bytes_max。
hw_free是hw_params的相反操作,调用snd_pcm_set_runtime_buffer(substream, NULL)即可。
注:代码中的dma_buffer是DMA缓冲区,它通过4个字段定义:dma_area、dma_addr、dma_bytes和dma_private。其中dma_area是缓冲区逻辑地址,dma_addr是缓冲区的物理地址,dma_bytes是缓冲区的大小,dma_private是ALSA的DMA管理用到的。dma_buffer是在pcm_new()中初始化的;当然也可以把分配dma缓冲区的工作放到这部分来实现,但考虑到减少碎片,故还是在pcm_new中以最大size(即buffer_bytes_max)来分配。
关于DMA的中断处理
另外留意open函数中的audio_dma_request(&s[0], audio_dma_callback);中的audio_dma_callback,这是dma的中断函数,这里以callback的形式存在,其实到dma的底层还是这样的形式:static irqreturn_t dma_irq_handler(int irq, void *dev_id),在DMA中断处理dma_irq_handler()中调用callback。这些跟具体硬件平台的DMA实现相关,如果没有类似的机制,那么还是要在pcm模块中实现这个中断。
void rockchip_pcm_dma_irq(s32 ch, void *data)
{
struct snd_pcm_substream *substream = data;
struct rockchip_runtime_data *prtd;
unsigned long flags;
DBG("Enter::%s----%d\n",__FUNCTION__,__LINE__);
prtd = substream->runtime->private_data;
if (substream)
snd_pcm_period_elapsed(substream);
spin_lock(&prtd->lock);
prtd->dma_loaded--;
if (prtd->state & ST_RUNNING) {
rockchip_pcm_enqueue(substream);
}
spin_unlock(&prtd->lock);
local_irq_save(flags);
if (prtd->state & ST_RUNNING) {
if (prtd->dma_loaded) {
if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
audio_start_dma(substream, DMA_MODE_WRITE);
else
audio_start_dma(substream, DMA_MODE_READ);
}
}
local_irq_restore(flags);
}
prepare
当pcm“准备好了”调用该函数。在这里根据channels、buffer_bytes等来设定DMA传输参数,跟具体硬件平台相关。注:每次调用snd_pcm_prepare()的时候均会调用prepare函数。
trigger
当pcm开始、停止、暂停的时候都会调用trigger函数。
static int rockchip_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
{
struct rockchip_runtime_data *prtd = substream->runtime->private_data;
int ret = 0;
/**************add by qiuen for volume*****/
struct snd_soc_pcm_runtime *rtd = substream->private_data;
struct snd_soc_dai *pCodec_dai = rtd->dai->codec_dai;
int vol = 0;
int streamType = 0;
DBG("Enter::%s----%d\n",__FUNCTION__,__LINE__);
if(cmd==SNDRV_PCM_TRIGGER_VOLUME){
vol = substream->number % 100;
streamType = (substream->number / 100) % 100;
DBG("enter:vol=%d,streamType=%d\n",vol,streamType);
if(pCodec_dai->ops->set_volume)
pCodec_dai->ops->set_volume(streamType, vol);
}
/****************************************************/
spin_lock(&prtd->lock);
switch (cmd) {
case SNDRV_PCM_TRIGGER_START:
DBG(" START \n");
prtd->state |= ST_RUNNING;
rk29_dma_ctrl(prtd->params->channel, RK29_DMAOP_START);
/*
if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
audio_start_dma(substream, DMA_MODE_WRITE);
} else {
audio_start_dma(substream, DMA_MODE_READ);
}
*/
#ifdef CONFIG_ANDROID_POWER
android_lock_suspend(&audio_lock);
DBG("%s::start audio , lock system suspend\n" , __func__ );
#endif
break;
case SNDRV_PCM_TRIGGER_RESUME:
DBG(" RESUME \n");
break;
case SNDRV_PCM_TRIGGER_PAUSE_RELEASE:
DBG(" RESTART \n");
break;
case SNDRV_PCM_TRIGGER_STOP:
case SNDRV_PCM_TRIGGER_SUSPEND:
case SNDRV_PCM_TRIGGER_PAUSE_PUSH:
DBG(" STOPS \n");
prtd->state &= ~ST_RUNNING;
rk29_dma_ctrl(prtd->params->channel, RK29_DMAOP_STOP);
//disable_dma(prtd->params->channel);
#ifdef CONFIG_ANDROID_POWER
android_unlock_suspend(&audio_lock );
DBG("%s::stop audio , unlock system suspend\n" , __func__ );
#endif
break;
default:
ret = -EINVAL;
break;
}
spin_unlock(&prtd->lock);
return ret;
}
Trigger函数里面的操作应该是原子的,不要在调用这些操作时进入睡眠,trigger函数应尽量小,甚至仅仅是触发DMA。
pointer
static snd_pcm_uframes_t wmt_pcm_pointer(struct snd_pcm_substream *substream)
PCM中间层通过调用这个函数来获取缓冲区的位置。一般情况下,在中断函数中调用snd_pcm_period_elapsed()或在pcm中间层更新buffer的时候调用它。然后pcm中间层会更新指针位置和计算缓冲区可用空间,唤醒那些在等待的线程。这个函数也是原子的。
snd_pcm_runtime
我们会留意到ops各成员函数均需要取得一个snd_pcm_runtime结构体指针,这个指针可以通过substream->runtime来获得。snd_pcm_runtime是pcm运行时的信息。当打开一个pcm子流时,pcm运行时实例就会分配给这个子流。它拥有很多多种信息:hw_params和sw_params配置拷贝,缓冲区指针,mmap记录,自旋锁等。snd_pcm_runtime对于驱动程序操作集函数是只读的,仅pcm中间层可以改变或更新这些信息。