(视频)vivi驱动层代码分析二

来源：互联网发布：js走马灯效果编辑：程序博客网时间：2024/06/05 05:03

一. V4L2框架: video for linux version 2

虚拟视频驱动vivi.c分析:
1．分配video_device
2．设置
3．注册：video_register_device

vivi_init
vivi_create_instance
v4l2_device_register // 不是主要, 只是用于初始化一些东西，比如自旋锁、引用计数
video_device_alloc
// 设置
1. vfd:
.fops = &vivi_fops,
.ioctl_ops = &vivi_ioctl_ops,
.release = video_device_release,
2.
vfd->v4l2_dev = &dev->v4l2_dev;
3. 设置"ctrl属性"(用于APP的ioctl)：
v4l2_ctrl_handler_init(hdl, 11);
dev->volume = v4l2_ctrl_new_std(hdl, &vivi_ctrl_ops,
V4L2_CID_AUDIO_VOLUME, 0, 255, 1, 200);
dev->brightness = v4l2_ctrl_new_std(hdl, &vivi_ctrl_ops,
V4L2_CID_BRIGHTNESS, 0, 255, 1, 127);
dev->contrast = v4l2_ctrl_new_std(hdl, &vivi_ctrl_ops,
V4L2_CID_CONTRAST, 0, 255, 1, 16);
video_register_device(video_device, type:VFL_TYPE_GRABBER, nr)
__video_register_device
vdev->cdev = cdev_alloc();
vdev->cdev->ops = &v4l2_fops;
cdev_add

video_device[vdev->minor] = vdev;
if (vdev->ctrl_handler == NULL)
vdev->ctrl_handler = vdev->v4l2_dev->ctrl_handler;

分析vivi.c的open,read,write,ioctl过程
1. open
app: open("/dev/video0",....)
---------------------------------------------------
drv: v4l2_fops.v4l2_open
vdev = video_devdata(filp); // 根据次设备号从数组中得到video_device
ret = vdev->fops->open(filp);
vivi_ioctl_ops.open
v4l2_fh_open

2. read
app: read ....
---------------------------------------------------
drv: v4l2_fops.v4l2_read
struct video_device *vdev = video_devdata(filp);
ret = vdev->fops->read(filp, buf, sz, off);

3. ioctl
app: ioctl
----------------------------------------------------
drv: v4l2_fops.unlocked_ioctl
v4l2_ioctl
struct video_device *vdev = video_devdata(filp);
ret = vdev->fops->unlocked_ioctl(filp, cmd, arg);
video_ioctl2
video_usercopy(file, cmd, arg, __video_do_ioctl);
__video_do_ioctl
struct video_device *vfd = video_devdata(file);
根据APP传入的cmd来获得、设置"某些属性"

v4l2_ctrl_handler的使用过程:
__video_do_ioctl
struct video_device *vfd = video_devdata(file);

case VIDIOC_QUERYCTRL:
{
struct v4l2_queryctrl *p = arg;

if (vfh && vfh->ctrl_handler)
ret = v4l2_queryctrl(vfh->ctrl_handler, p);
else if (vfd->ctrl_handler) // 在哪设置？在video_register_device
ret = v4l2_queryctrl(vfd->ctrl_handler, p);
// 根据ID在ctrl_handler里找到v4l2_ctrl，返回它的值

二、测试虚拟驱动vivi
准备工作：安装xawtv
sudo apt-get install xawtv

源码xawtv-3.95.tar.gz
http://www.kraxel.org/releases/xawtv/

在这个网站创建新的sources.list
http://repogen.simplylinux.ch/
1. 选择国家
2．选择相邻的ubuntu版本
3. 选择"Ubuntu Branches"
4. 生成sources.list
5. 把得到内容替换到/etc/apt/sources.list
6. sudo apt-get update
sudo apt-get install xawtv

测试USB摄像头：
1．让VMWAER处于前台，接上USB摄像头，可以看到生成了/dev/video0
2．执行 xawtv 即可看到图像

测试虚拟摄像头vivi：
1. 确实ubuntu的内核版本
uname -a
Linux book-desktop 2.6.31-14-generic #48-Ubuntu SMP Fri Oct 16 14:04:26 UTC 2009 i686 GNU/Linux
2. 去www.kernel.org下载同版本的内核
解压后把drivers/media/video目录取出
修改它的Makefile为：

KERN_DIR = /usr/src/linux-headers-2.6.31-14-generic

all:
make -C $(KERN_DIR) M=`pwd` modules

clean:
make -C $(KERN_DIR) M=`pwd` modules clean
rm -rf modules.order

obj-m += vivi.o
obj-m += videobuf-core.o
obj-m += videobuf-vmalloc.o
obj-m += v4l2-common.o

3. make
4. sudo modprobe vivi
sudo rmmod vivi
sudo insmod ./vivi.ko

5. ls /dev/video*
6. xawtv -c /dev/videoX

//
3. ioctl(4, VIDIOC_G_FMT
4. for()
ioctl(4, VIDIOC_ENUM_FMT
5. ioctl(4, VIDIOC_QUERYCAP // 列举性能
6. ioctl(4, VIDIOC_G_INPUT // 获得当前使用输入源
7. ioctl(4, VIDIOC_ENUMINPUT // 列举输入源
8. ioctl(4, VIDIOC_QUERYCTRL // 查询属性,比如亮度、对比度
9. ioctl(4, VIDIOC_QUERYCAP
10. ioctl(4, VIDIOC_ENUMINPUT

三、xawtv源码彻底分析
// 1~7都是在v4l2_open里调用
1. open
2. ioctl(4, VIDIOC_QUERYCAP

// 3~7 都是在get_device_capabilities里调用
3. for()
ioctl(4, VIDIOC_ENUMINPUT // 列举输入源,VIDIOC_ENUMINPUT/VIDIOC_G_INPUT/VIDIOC_S_INPUT不是必需的
4. for()
ioctl(4, VIDIOC_ENUMSTD // 列举标准(制式), 不是必需的
5. for()
ioctl(4, VIDIOC_ENUM_FMT // 列举格式

6. ioctl(4, VIDIOC_G_PARM
7. for()
ioctl(4, VIDIOC_QUERYCTRL // 查询属性(比如说亮度值最小值、最大值、默认值)

// 8~10都是通过v4l2_read_attr来调用的
8. ioctl(4, VIDIOC_G_STD // 获得当前使用的标准(制式), 不是必需的
9. ioctl(4, VIDIOC_G_INPUT
10. ioctl(4, VIDIOC_G_CTRL // 获得当前属性, 比如亮度是多少

11. ioctl(4, VIDIOC_TRY_FMT // 试试能否支持某种格式
12. ioctl(4, VIDIOC_S_FMT // 设置摄像头使用某种格式

// 13~16在v4l2_start_streaming
13. ioctl(4, VIDIOC_REQBUFS // 请求系统分配缓冲区
14. for()
ioctl(4, VIDIOC_QUERYBUF // 查询所分配的缓冲区
mmap
15. for ()
ioctl(4, VIDIOC_QBUF // 把缓冲区放入队列
16. ioctl(4, VIDIOC_STREAMON // 启动摄像头

// 17里都是通过v4l2_write_attr来调用的
17. for ()
ioctl(4, VIDIOC_S_CTRL // 设置属性
ioctl(4, VIDIOC_S_INPUT // 设置输入源
ioctl(4, VIDIOC_S_STD // 设置标准(制式), 不是必需的

// v4l2_nextframe > v4l2_waiton
18. v4l2_queue_all
v4l2_waiton
for ()
{
select(5, [4], NULL, NULL, {5, 0}) = 1 (in [4], left {4, 985979})
ioctl(4, VIDIOC_DQBUF // de-queue, 把缓冲区从队列中取出
// 处理, 之以已经通过mmap获得了缓冲区的地址, 就可以直接访问数据
ioctl(4, VIDIOC_QBUF // 把缓冲区放入队列
}

xawtv的几大函数：
1. v4l2_open
2. v4l2_read_attr/v4l2_write_attr
3. v4l2_start_streaming
4. v4l2_nextframe/v4l2_waiton

摄像头驱动程序必需的11个ioctl:
// 表示它是一个摄像头设备
.vidioc_querycap = vidioc_querycap,

/* 用于列举、获得、测试、设置摄像头的数据的格式 */
.vidioc_enum_fmt_vid_cap = vidioc_enum_fmt_vid_cap,
.vidioc_g_fmt_vid_cap = vidioc_g_fmt_vid_cap,
.vidioc_try_fmt_vid_cap = vidioc_try_fmt_vid_cap,
.vidioc_s_fmt_vid_cap = vidioc_s_fmt_vid_cap,

/* 缓冲区操作: 申请/查询/放入队列/取出队列 */
.vidioc_reqbufs = vidioc_reqbufs,
.vidioc_querybuf = vidioc_querybuf,
.vidioc_qbuf = vidioc_qbuf,
.vidioc_dqbuf = vidioc_dqbuf,

// 启动/停止
.vidioc_streamon = vidioc_streamon,
.vidioc_streamoff = vidioc_streamoff,

继续分析数据的获取过程：
1. 请求分配缓冲区: ioctl(4, VIDIOC_REQBUFS // 请求系统分配缓冲区
videobuf_reqbufs(队列, v4l2_requestbuffers) // 队列在open函数用videobuf_queue_vmalloc_init初始化
// 注意：这个IOCTL只是分配缓冲区的头部信息，真正的缓存还没有分配呢

2. 查询映射缓冲区:
ioctl(4, VIDIOC_QUERYBUF // 查询所分配的缓冲区
videobuf_querybuf // 获得缓冲区的数据格式、大小、每一行长度、高度
mmap(参数里有"大小") // 在这里才分配缓存
v4l2_mmap
vivi_mmap
videobuf_mmap_mapper
videobuf-vmalloc.c里的__videobuf_mmap_mapper
mem->vmalloc = vmalloc_user(pages); // 在这里才给缓冲区分配空间

3. 把缓冲区放入队列:
ioctl(4, VIDIOC_QBUF // 把缓冲区放入队列
videobuf_qbuf
q->ops->buf_prepare(q, buf, field); // 调用驱动程序提供的函数做些预处理
list_add_tail(&buf->stream, &q->stream); // 把缓冲区放入队列的尾部
q->ops->buf_queue(q, buf); // 调用驱动程序提供的"入队列函数"


4. 启动摄像头
ioctl(4, VIDIOC_STREAMON
videobuf_streamon
q->streaming = 1;


5. 用select查询是否有数据
// 驱动程序里必定有: 产生数据、唤醒进程
v4l2_poll
vdev->fops->poll
vivi_poll
videobuf_poll_stream
// 从队列的头部获得缓冲区
buf = list_entry(q->stream.next, struct videobuf_buffer, stream);

// 如果没有数据则休眠
poll_wait(file, &buf->done, wait);

谁来产生数据、谁来唤醒它？
内核线程vivi_thread每30MS执行一次，它调用
vivi_thread_tick
vivi_fillbuff(fh, buf); // 构造数据
wake_up(&buf->vb.done); // 唤醒进程

6. 有数据后从队列里取出缓冲区
// 有那么多缓冲区，APP如何知道哪一个缓冲区有数据？调用VIDIOC_DQBUF
ioctl(4, VIDIOC_DQBUF
vidioc_dqbuf
// 在队列里获得有数据的缓冲区
retval = stream_next_buffer(q, &buf, nonblocking);

// 把它从队列中删掉
list_del(&buf->stream);

// 把这个缓冲区的状态返回给APP
videobuf_status(q, b, buf, q->type);

7. 应用程序根据VIDIOC_DQBUF所得到缓冲区状态，知道是哪一个缓冲区有数据
就去读对应的地址(该地址来自前面的mmap)

怎么写摄像头驱动程序:
1. 分配video_device:video_device_alloc
2. 设置
.fops
.ioctl_ops (里面需要设置11项)
如果要用内核提供的缓冲区操作函数，还需要构造一个videobuf_queue_ops
3. 注册: video_register_device

四、自己写一个虚拟摄像头驱

[585955.047776] usb 1-1: new high speed USB device using ehci_hcd and address 3
[585955.269142] usb 1-1: configuration #1 chosen from 1 choice
[585955.379843] uvcvideo: Found UVC 1.00 device USB2.0 Camera (1e4e:0102)
[585955.440427] uvcvideo: UVC non compliance - GET_DEF(PROBE) not supported. Enabling workaround.
[585955.554594] input: USB2.0 Camera as /devices/pci0000:00/0000:00:11.0/0000:02:02.0/usb1/1-1/1-1:1.0/input/input5

LD_PRELOAD=/usr/lib/libv4l/v4l1compat.so camorama

sudo apt-get install cheese

多种LINUX下的摄像头工具
https://help.ubuntu.com/community/Webcam

http://blog.csdn.net/hongtao_liu/article/details/5867351
http://blog.csdn.net/hongtao_liu/article/details/5894089
基于V4L2的视频驱动开发

概念：
视频制式 : NTSC PAL V4L2_STD_NTSC V4L2_STD_PAL
视频格式：RGB YCbCr 420,422 V4L2_PIX_FMT_UYVY

[视频技术手册]中文第5版
http://ishare.iask.sina.com.cn/f/21425721.html
http://ishare.iask.sina.com.cn/f/21425722.html

http://www.360doc.com/content/08/0926/11/14148_1678949.shtml

数字视频的基本概念
http://hi.baidu.com/fengbit/item/78a161375067f1c42f8ec24e

USB协议
http://www.usb.org/developers/devclass_docs

USB Video Class Specification 笔记
http://blog.csdn.net/chinaunixj/article/details/7394315

基于嵌入式Linux的视频采集系统---UVC驱动模型介绍
http://blog.csdn.net/chinaunixj/article/details/7439870

Class-specific VC Interface Descriptor
e:\kernel_projects\linux-3.4.2\linux-3.4.2\include\linux\usb\Video.h
#define DECLARE_UVC_HEADER_DESCRIPTOR(n) \
struct UVC_HEADER_DESCRIPTOR(n) { \
__u8 bLength;\
__u8 bDescriptorType;\
__u8 bDescriptorSubType;\
__u16 bcdUVC; \
__u16 wTotalLength;\
__u32 dwClockFrequency;\
__u8 bInCollection;\
__u8 baInterfaceNr[n];\
} __attribute__ ((packed))

Input Terminal Descriptor (Camera)

驱动框架分析：
Structure of a driver
---------------------

All drivers have the following structure:

1) A struct for each device instance containing the device state.

2) A way of initializing and commanding sub-devices (if any).

3) Creating V4L2 device nodes (/dev/videoX, /dev/vbiX and /dev/radioX)
and keeping track of device-node specific data.

4) Filehandle-specific structs containing per-filehandle data;

5) video buffer handling.

This is a rough schematic of how it all relates:

device instances : v4l2_device
|
+-sub-device instances : v4l2_subdev
|
\-V4L2 device nodes : video_device, stores V4L2 device node data, this will create the character device
|
\-filehandle instances : v4l2_fh

The framework closely resembles the driver structure: it has a v4l2_device
struct for the device instance data, a v4l2_subdev struct to refer to
sub-device instances, the video_device struct stores V4L2 device node data
and in the future a v4l2_fh struct will keep track of filehandle instances
(this is not yet implemented).

The V4L2 framework also optionally integrates with the media framework. If a
driver sets the struct v4l2_device mdev field, sub-devices and video nodes
will automatically appear in the media framework as entities.

console/fs.h:2:20: error: FSlib.h: No such file or directory

No package 'fontsproto' found

error: asm/page.h: No such file or directory

把asm/page.h改为sys/user.h

sudo apt-get install libxaw7-dev

# libraries
LDLIBS := -lFS

./configure --x-includes=/usr/local/include/X11/fonts

0 0