DMA Engine API Guide — The Linux Kernel documentation
在嵌入式系统中，从需求的角度看内存，是非常简单的，可以总结为两大类：
1）CPU 有访问内存的需求，包括从内存中取指、从内存中取数据、向内存中写数据。相应的数据流为：
CPU<-------------->MMU(Optional)<----------->Memory
2）其它外部设备有访问内存的需求，包括从内存“读取”数据、向内存“写入”数据。这里的“读取”和“写入”加了引号，是因为在大部分情况下，设备不像 CPU 那样有智能，不具备直接访问内存的能力。总结来说，设备有三种途径访问内存：
a）由 CPU 从中中转，数据流如下（本质上是 CPU 访问内存）：
Device<---------------->CPU<--------------------Memory
b）由第三方具有智能的模块（如 DMA 控制器）中转，数据流如下：
Device<----------->DMA Controller<--------->Memory
c）直接访问内存，数据流如下：
Device<---------->IOMMU(Optional)--------->Memory
DMA Engine
DMA Engine可以表示两个概念，第一，一个DMA Engine实际就是一个硬件DMA设备的驱动（DMA Engine Device）。第二，DMA Engine在内核中是一个DMA设备驱动的通用框架（DMA Engine Framework）。
DMA Engine Framework跟大多数内核设备框架一样，DMA Engine Driver需要根据其硬件特性填充dma device结构体，并基于其硬件实现功能接口，注册到DMA Engine框架，由DMA Engine Framework向外提供统一功能接口。
通常DMA的使用者包括三类人：
-
DMA Engine Developer，开发DMA设备驱动，向系统注册DMA Engine Device，为系统提供DMA的搬运服务，比如dw-axi-dma。
-
DMA Client Developer，比如UART、USB、UFS的driver开发者，调用DMA Engine Framework提供的通用DMA操作接口，帮助其完成DMA搬运服务。
-
DMA Mapping User，利用DMA Alloc时可以建立SMMU页表的功能，创建用于与其他子系统共享的内存Buffer。
本文主要介绍前两种情况。
DMA Engine Device
DMA硬件是一个可以脱离CPU，独立在Memory与IO Device间做数据搬运的设备。
CPU CPU Bus
Virtual Physical Address
Address Address Space
Space Space
+-------+ +------+ +------+
| | |MMIO | Offset | |
| | Virtual |Space | applied | |
C +-------+ --------> B +------+ ----------> +------+ A
| | mapping | | by host | |
+-----+ | | | | bridge | | +--------+
| | | | +------+ | | | |
| CPU | | | | RAM | | | | Device |
| | | | | | | | | |
+-----+ +-------+ +------+ +------+ +--------+
| | Virtual |Buffer| Mapping | |
X +-------+ --------> Y +------+ <---------- +------+ Z
| | mapping | RAM | by IOMMU
| | | |
| | | |
+-------+ +------+
如图所示，DMA设备所在的BUS上有两类设备，一类是IO设备，一类是主存（DDR），DMA设备只能看到自己所在的Bus Address（ba），访问IO设备，直接使用ba，访问DDR，可能会有一个offset转换或者经过iommu的转换。也就是说，CPU这边读写DDR，是通过VA，而DMA engine device访问DDR，使用的是ba，或者叫iova（translate ba=>pa）。内核中ba用dma_addr_t类型表示。
一个DMA硬件由多条Channel，Channel的数量由硬件决定，需要在DMA Device注册时填入到DMA Device结构体中。
DMA的搬运操作由DMA Request发起，DMA Request是一个软件虚拟概念，对应到硬件上，是一组寄存器配置操作。
DMA做搬运时，支持多种方向（direction）：MEM_TO_DEV、DEV_TO_MEM、MEM_TO_MEM，具体硬件上支持哪个方向，也需要在DMA Device初始化时配入。
搬运时，硬件支持多种搬运模式（transaction_type or cap），以dw-axi-dmac为例，该DMA Device支持DMA_MEMCPY（从连续内存搬运）、DMA_SLAVE（使用scatter-gather list搬运非连续内存）、DMA_CYCLIC（搬运ringbuffer）三种搬运模式。这些硬件支持的搬运类型，在DMA Device初始化时，使用dma_cap_set来配置。（详细见provider.rst）
DMA支持的地址，不一定是64bit，所以在device注册时，需要指定src addr、dst addr的地址位宽（mask）。由于一般情况下DMA Device支持的物理地址位宽为32位，所以内核将DDR 32bit范围内的部分做成了一个DMA zone，用于给DMA设备分配可映射访问的buffer。
一次搬运操作支持的read或write字符数，叫做transfer width，但是一般DMA会支持将多次访问合并为一个burst（read、write operation collection），burst中支持合并多少个独立transfer操作，叫做burst的length，或者max_burst，同样在device初始化时需要定义。
DMA Engine Device除了需要填充上述硬件信息外，还需要实现DMA Device Operation：
device_tx_status # poll for transaction completion
device_issue_pending # push pending transactions to hardware (Do the transfer)
device_terminate_all # aborts all transfers on a channel
device_pause # pause any transfer on a channel
device_resume # resume paused transfer on a channel
device_alloc_chan_resources
device_free_chan_resources
device_synchronize
device_config
device_prep_dma_memcpy # get memcpy operation desc
device_prep_slave_sg # get slave operation desc
device_prep_dma_cyclic # get cyclic operation desc
最后通过 dmaenginem_async_device_register 注册到DMA Engine Framework中。
Reference: DMAengine controller documentation — The Linux Kernel documentation
Reference Code: dw-axi-dmac-platform.c
DMA Engine Client
DMA Engine Client是使用DMA Engine帮其从主存搬运数据的IO设备。这些设备包括支持dma的uart、pcie、usb、ufs、wifi（挂在pcie总线上）、modem（挂在pcie总线上）
Slave Device Driver在使用DMA Engine做搬移时，一般包括下面几个步骤：
- Alloc a channel
通过调用 dma_request_chan 得到一个chanel结构体
- Config the channel for transfer
构造 dma_slave_config，通过 dmaengine_slave_config 配置chanel，设置搬运方向，原地址，地址位宽。
- Prepare the transaction descriptor (desc)
根据待传输的数据类型，调用 dmaengine_prep_xxx 接口，获得传输使用的 dma_async_tx_descriptor。
设置desc的callback信息。这时，desc中包含了所有transfer的信息，包括使用的channel（源地址）、目标地址、方向、大小、传输方式、callback等信息。
- Submit transaction
调用 dmaengine_submit 将获取并准备好的desc提交。
提交后的dma desc并不会直接开始传输，只是加入到pending队列。
- Issue pending requests
调用 dma_async_issue_pending 开始chanel request队列的异步传输。
- Wait and handle the callback event
当每个desc传输完成后，会通过tasklet回调callback。
Reference: DMA Engine API Guide — The Linux Kernel documentation
Reference Code: 8250_dma.c
Files
- /drivers/dma/dmaengine.c # dma engine framework implementation
- /include/linux/dmaengine.h # dma engine framework implementation
# also interfaces for both engine and client driver
- /drivers/tty/serial/8250/8250_dma.c # dma client device driver
- /drivers/dma/dw-axi-dmac/dw-axi-dmac-platform.c # dma engine device driver
Structure
struct dma_device
dma device，在注册前初始化，前面已经有描述
struct dma_chan
dma channel，指向所属的dma device，被使用的slave device。
每个channel还对应一个供sysfs使用的dma_chan_dev结构体，在sysfs中暴露节点。（channel本身也被注册为一个device）
struct dma_slave_config
供slave device配置chanel使用的config结构体，用来为传输配置channel。
Variable
dma_device_list
dma engine device全局链表
Functions
dma_bus_init
初始化unmap pool，debugfs节点，sysfs节点
Interface（for DMA Engine）
* Reference Driver: dw-axi-dma
dma_async_device_register
DMA Engine Device注册，会将device注册到全局device链表，并注册device的每个channel
Interface（for DMA Client）
* Reference Driver: 8250_dma
Request a channel
dma_request_chan
根据dts配置拿到对应dma设备的chanel，或动态分配一个channel（调用find_candidate找一个合适channel）
dma_request_slave_channel_compat
dma_release_channel
释放一个channel
Get channel’s device caps
dma_get_slave_caps
Config the channel
dmaengine_slave_config
Pause / Resume DMA transfer
dmaengine_pause
暂停channel传输
dmaengine_resume
恢复channel传输
Prepare dma transfer descripter
dmaengine_prep_dma_memcpy
dmaengine_prep_slave_single
dmaengine_prep_dma_cyclic
Submit the transfer
dmaengine_submit
提交传输desc到channel的pending队列，返回cookie（desc的传输句柄）
Issue pending request and wait for callback
dma_async_issue_pending
开始channel的队列传输。
Check transaction finish
dma_async_is_tx_complete
通过cookie来判断submit的desc传输有没有完成
Terminate all transfer
dmaengine_terminate_async
停止channel的所有传输，异步方式
dmaengine_terminate_sync
停止channel的所有传输，同步等待
dmaengine_synchronize
如果使用async接口停止传输，可以用这个接口进入等待
Debugfs
dmaengine/summary
打印出DMA设备列表，每个DMA设备的channel数，是否正在使用。
dmaengine/<dma_device>/
调用每个DMA设备自己的dump功能打印
Sysfs
/sys/devices/system/dma/<dma_device>
- memcpy_count
- bytes_transferred
- in_use：chanels in use
Reference
How to implement a dma engine device driver:
DMAengine controller documentation — The Linux Kernel documentation
How to use dma in dma client device driver:
DMA Engine API Guide — The Linux Kernel documentation
How to test dma client device with dmatest:
DMA Test Guide — The Linux Kernel documentation
// SPDX-License-Identifier: GPL-2.0
// Copyright (C) 2019 Spreadtrum Communications Inc.
#include <linux/dma-mapping.h>
#include <linux/dmaengine.h>
#include <linux/dma/sprd-dma.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/of_reserved_mem.h>
#include <linux/platform_device.h>
#include <sound/pcm.h>
#include <sound/pcm_params.h>
#include <sound/soc.h>
#include "sprd-pcm-dma.h"
#define SPRD_PCM_DMA_LINKLIST_SIZE 64
#define SPRD_PCM_DMA_BRUST_LEN 640
struct sprd_pcm_dma_data {
struct dma_chan *chan;
struct dma_async_tx_descriptor *desc;
dma_cookie_t cookie;
dma_addr_t phys;
void *virt;
int pre_pointer;
};
struct sprd_pcm_dma_private {
struct snd_pcm_substream *substream;
struct sprd_pcm_dma_params *params;
struct sprd_pcm_dma_data data[SPRD_PCM_CHANNEL_MAX];
int hw_chan;
int dma_addr_offset;
};
static const struct snd_pcm_hardware sprd_pcm_hardware = {
.info = SNDRV_PCM_INFO_MMAP | SNDRV_PCM_INFO_MMAP_VALID |
SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_PAUSE |
SNDRV_PCM_INFO_RESUME | SNDRV_PCM_INFO_NO_PERIOD_WAKEUP,
.formats = SNDRV_PCM_FMTBIT_S16_LE | SNDRV_PCM_FMTBIT_S24_LE,
.period_bytes_min = 1,
.period_bytes_max = 64 * 1024,
.periods_min = 1,
.periods_max = PAGE_SIZE / SPRD_PCM_DMA_LINKLIST_SIZE,
.buffer_bytes_max = 64 * 1024,
};
static int sprd_pcm_open(struct snd_soc_component *component,
struct snd_pcm_substream *substream)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct device *dev = component->dev;
struct sprd_pcm_dma_private *dma_private;
int hw_chan = SPRD_PCM_CHANNEL_MAX;
int size, ret, i;
snd_soc_set_runtime_hwparams(substream, &sprd_pcm_hardware);
ret = snd_pcm_hw_constraint_step(runtime, 0,
SNDRV_PCM_HW_PARAM_PERIOD_BYTES,
SPRD_PCM_DMA_BRUST_LEN);
if (ret < 0)
return ret;
ret = snd_pcm_hw_constraint_step(runtime, 0,
SNDRV_PCM_HW_PARAM_BUFFER_BYTES,
SPRD_PCM_DMA_BRUST_LEN);
if (ret < 0)
return ret;
ret = snd_pcm_hw_constraint_integer(runtime,
SNDRV_PCM_HW_PARAM_PERIODS);
if (ret < 0)
return ret;
dma_private = devm_kzalloc(dev, sizeof(*dma_private), GFP_KERNEL);
if (!dma_private)
return -ENOMEM;
size = runtime->hw.periods_max * SPRD_PCM_DMA_LINKLIST_SIZE;
for (i = 0; i < hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
data->virt = dmam_alloc_coherent(dev, size, &data->phys,
GFP_KERNEL);
if (!data->virt) {
ret = -ENOMEM;
goto error;
}
}
dma_private->hw_chan = hw_chan;
runtime->private_data = dma_private;
dma_private->substream = substream;
return 0;
error:
for (i = 0; i < hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (data->virt)
dmam_free_coherent(dev, size, data->virt, data->phys);
}
devm_kfree(dev, dma_private);
return ret;
}
static int sprd_pcm_close(struct snd_soc_component *component,
struct snd_pcm_substream *substream)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct sprd_pcm_dma_private *dma_private = runtime->private_data;
struct device *dev = component->dev;
int size = runtime->hw.periods_max * SPRD_PCM_DMA_LINKLIST_SIZE;
int i;
for (i = 0; i < dma_private->hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
dmam_free_coherent(dev, size, data->virt, data->phys);
}
devm_kfree(dev, dma_private);
return 0;
}
static void sprd_pcm_dma_complete(void *data)
{
struct sprd_pcm_dma_private *dma_private = data;
struct snd_pcm_substream *substream = dma_private->substream;
snd_pcm_period_elapsed(substream);
}
static void sprd_pcm_release_dma_channel(struct snd_pcm_substream *substream)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct sprd_pcm_dma_private *dma_private = runtime->private_data;
int i;
for (i = 0; i < SPRD_PCM_CHANNEL_MAX; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (data->chan) {
dma_release_channel(data->chan);
data->chan = NULL;
}
}
}
static int sprd_pcm_request_dma_channel(struct snd_soc_component *component,
struct snd_pcm_substream *substream,
int channels)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct sprd_pcm_dma_private *dma_private = runtime->private_data;
struct device *dev = component->dev;
struct sprd_pcm_dma_params *dma_params = dma_private->params;
int i;
if (channels > SPRD_PCM_CHANNEL_MAX) {
dev_err(dev, "invalid dma channel number:%d\n", channels);
return -EINVAL;
}
for (i = 0; i < channels; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
data->chan = dma_request_slave_channel(dev,
dma_params->chan_name[i]);
if (!data->chan) {
dev_err(dev, "failed to request dma channel:%s\n",
dma_params->chan_name[i]);
sprd_pcm_release_dma_channel(substream);
return -ENODEV;
}
}
return 0;
}
static int sprd_pcm_hw_params(struct snd_soc_component *component,
struct snd_pcm_substream *substream,
struct snd_pcm_hw_params *params)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct sprd_pcm_dma_private *dma_private = runtime->private_data;
struct snd_soc_pcm_runtime *rtd = snd_soc_substream_to_rtd(substream);
struct sprd_pcm_dma_params *dma_params;
size_t totsize = params_buffer_bytes(params);
size_t period = params_period_bytes(params);
int channels = params_channels(params);
int is_playback = substream->stream == SNDRV_PCM_STREAM_PLAYBACK;
struct scatterlist *sg;
unsigned long flags;
int ret, i, j, sg_num;
dma_params = snd_soc_dai_get_dma_data(snd_soc_rtd_to_cpu(rtd, 0), substream);
if (!dma_params) {
dev_warn(component->dev, "no dma parameters setting\n");
dma_private->params = NULL;
return 0;
}
if (!dma_private->params) {
dma_private->params = dma_params;
ret = sprd_pcm_request_dma_channel(component,
substream, channels);
if (ret)
return ret;
}
sg_num = totsize / period;
dma_private->dma_addr_offset = totsize / channels;
sg = devm_kcalloc(component->dev, sg_num, sizeof(*sg), GFP_KERNEL);
if (!sg) {
ret = -ENOMEM;
goto sg_err;
}
for (i = 0; i < channels; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
struct dma_chan *chan = data->chan;
struct dma_slave_config config = { };
struct sprd_dma_linklist link = { };
enum dma_transfer_direction dir;
struct scatterlist *sgt = sg;
config.src_maxburst = dma_params->fragment_len[i];
config.src_addr_width = dma_params->datawidth[i];
config.dst_addr_width = dma_params->datawidth[i];
if (is_playback) {
config.src_addr = runtime->dma_addr +
i * dma_private->dma_addr_offset;
config.dst_addr = dma_params->dev_phys[i];
dir = DMA_MEM_TO_DEV;
} else {
config.src_addr = dma_params->dev_phys[i];
config.dst_addr = runtime->dma_addr +
i * dma_private->dma_addr_offset;
dir = DMA_DEV_TO_MEM;
}
sg_init_table(sgt, sg_num);
for (j = 0; j < sg_num; j++, sgt++) {
u32 sg_len = period / channels;
sg_dma_len(sgt) = sg_len;
sg_dma_address(sgt) = runtime->dma_addr +
i * dma_private->dma_addr_offset + sg_len * j;
}
/*
* Configure the link-list address for the DMA engine link-list
* mode.
*/
link.virt_addr = (unsigned long)data->virt;
link.phy_addr = data->phys;
ret = dmaengine_slave_config(chan, &config);
if (ret) {
dev_err(component->dev,
"failed to set slave configuration: %d\n", ret);
goto config_err;
}
/*
* We configure the DMA request mode, interrupt mode, channel
* mode and channel trigger mode by the flags.
*/
flags = SPRD_DMA_FLAGS(SPRD_DMA_CHN_MODE_NONE, SPRD_DMA_NO_TRG,
SPRD_DMA_FRAG_REQ, SPRD_DMA_TRANS_INT);
data->desc = chan->device->device_prep_slave_sg(chan, sg,
sg_num, dir,
flags, &link);
if (!data->desc) {
dev_err(component->dev, "failed to prepare slave sg\n");
ret = -ENOMEM;
goto config_err;
}
if (!runtime->no_period_wakeup) {
data->desc->callback = sprd_pcm_dma_complete;
data->desc->callback_param = dma_private;
}
}
devm_kfree(component->dev, sg);
return 0;
config_err:
devm_kfree(component->dev, sg);
sg_err:
sprd_pcm_release_dma_channel(substream);
return ret;
}
static int sprd_pcm_hw_free(struct snd_soc_component *component,
struct snd_pcm_substream *substream)
{
sprd_pcm_release_dma_channel(substream);
return 0;
}
static int sprd_pcm_trigger(struct snd_soc_component *component,
struct snd_pcm_substream *substream, int cmd)
{
struct sprd_pcm_dma_private *dma_private =
substream->runtime->private_data;
int ret = 0, i;
switch (cmd) {
case SNDRV_PCM_TRIGGER_START: // >=start threohold后 ，数据足够pcm_write 才trigger 搬迁。
for (i = 0; i < dma_private->hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (!data->desc)
continue;
data->cookie = dmaengine_submit(data->desc);
ret = dma_submit_error(data->cookie);
if (ret) {
dev_err(component->dev,
"failed to submit dma request: %d\n",
ret);
return ret;
}
dma_async_issue_pending(data->chan);
}
break;
case SNDRV_PCM_TRIGGER_RESUME:
case SNDRV_PCM_TRIGGER_PAUSE_RELEASE:
for (i = 0; i < dma_private->hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (data->chan)
dmaengine_resume(data->chan);
}
break;
case SNDRV_PCM_TRIGGER_STOP:
for (i = 0; i < dma_private->hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (data->chan)
dmaengine_terminate_async(data->chan);
}
break;
case SNDRV_PCM_TRIGGER_SUSPEND:
case SNDRV_PCM_TRIGGER_PAUSE_PUSH:
for (i = 0; i < dma_private->hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (data->chan)
dmaengine_pause(data->chan);
}
break;
default:
ret = -EINVAL;
}
return ret;
}
static snd_pcm_uframes_t sprd_pcm_pointer(struct snd_soc_component *component,
struct snd_pcm_substream *substream)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct sprd_pcm_dma_private *dma_private = runtime->private_data;
int pointer[SPRD_PCM_CHANNEL_MAX];
int bytes_of_pointer = 0, sel_max = 0, i;
snd_pcm_uframes_t x;
struct dma_tx_state state;
enum dma_status status;
for (i = 0; i < dma_private->hw_chan; i++) {
struct sprd_pcm_dma_data *data = &dma_private->data[i];
if (!data->chan)
continue;
status = dmaengine_tx_status(data->chan, data->cookie, &state);
if (status == DMA_ERROR) {
dev_err(component->dev,
"failed to get dma channel %d status\n", i);
return 0;
}
/*
* We just get current transfer address from the DMA engine, so
* we need convert to current pointer.
*/
pointer[i] = state.residue - runtime->dma_addr -
i * dma_private->dma_addr_offset;
if (i == 0) {
bytes_of_pointer = pointer[i];
sel_max = pointer[i] < data->pre_pointer ? 1 : 0;
} else {
sel_max ^= pointer[i] < data->pre_pointer ? 1 : 0;
if (sel_max)
bytes_of_pointer =
max(pointer[i], pointer[i - 1]) << 1;
else
bytes_of_pointer =
min(pointer[i], pointer[i - 1]) << 1;
}
data->pre_pointer = pointer[i];
}
x = bytes_to_frames(runtime, bytes_of_pointer);
if (x == runtime->buffer_size)
x = 0;
return x;
}
static int sprd_pcm_new(struct snd_soc_component *component,
struct snd_soc_pcm_runtime *rtd)
{
struct snd_card *card = rtd->card->snd_card;
struct snd_pcm *pcm = rtd->pcm;
int ret;
ret = dma_coerce_mask_and_coherent(card->dev, DMA_BIT_MASK(32));
if (ret)
return ret;
return snd_pcm_set_fixed_buffer_all(pcm, SNDRV_DMA_TYPE_DEV,
card->dev,
sprd_pcm_hardware.buffer_bytes_max);
}
static const struct snd_soc_component_driver sprd_soc_component = {
.name = DRV_NAME,
.open = sprd_pcm_open,
.close = sprd_pcm_close,
.hw_params = sprd_pcm_hw_params,
.hw_free = sprd_pcm_hw_free,
.trigger = sprd_pcm_trigger,
.pointer = sprd_pcm_pointer,
.pcm_construct = sprd_pcm_new,
.compress_ops = &sprd_platform_compress_ops,
};
static int sprd_soc_platform_probe(struct platform_device *pdev)
{
struct device_node *np = pdev->dev.of_node;
int ret;
ret = of_reserved_mem_device_init_by_idx(&pdev->dev, np, 0);
if (ret)
dev_warn(&pdev->dev,
"no reserved DMA memory for audio platform device\n");
ret = devm_snd_soc_register_component(&pdev->dev, &sprd_soc_component,
NULL, 0);
if (ret)
dev_err(&pdev->dev, "could not register platform:%d\n", ret);
return ret;
}
static const struct of_device_id sprd_pcm_of_match[] = {
{ .compatible = "sprd,pcm-platform", },
{ },
};
MODULE_DEVICE_TABLE(of, sprd_pcm_of_match);
static struct platform_driver sprd_pcm_driver = {
.driver = {
.name = "sprd-pcm-audio",
.of_match_table = sprd_pcm_of_match,
},
.probe = sprd_soc_platform_probe,
};
module_platform_driver(sprd_pcm_driver);
MODULE_DESCRIPTION("Spreadtrum ASoC PCM DMA");
MODULE_LICENSE("GPL v2");
MODULE_ALIAS("platform:sprd-audio");
扫一扫
3134
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
