如何解决VIDIOC_QBUF:无效参数 V4l2 MEMORY USERPTR python 实现
我正在尝试使用用户指针 v4l2 驱动程序。 目前我正在使用 v4l2 MMAP 并且它工作正常,但由于读取性能,我想将其更改为用户指针。
工作示例(使用 mmap):
def init_mmap(self,width=8192,height=256):
req = v4l2_requestbuffers()
req.type = V4L2_BUF_TYPE_VIDEO_CAPTURE
req.memory = V4L2_MEMORY_MMAP # V4L2_MEMORY_MMAP V4L2_MEMORY_USERPTR
req.count = 1 # number of buffer frames
self.height = max(self.height,32)
# get device capabilities
cp = v4l2_capability()
fcntl.ioctl(self.vid.fileno(),VIDIOC_QUERYCAP,cp)
# Configure v4l2 format
fmt = v4l2_format()
fmt.type = V4L2_BUF_TYPE_VIDEO_CAPTURE
fmt.fmt.pix.width = self.width # 8192
fmt.fmt.pix.height = self.height # 256
fmt.fmt.pix.pixelformat = V4L2_PIX_FMT_RG12
fmt.fmt.pix.field = V4L2_FIELD_NONE
fmt.fmt.pix.bytesperline = 0
fcntl.ioctl(self.vid.fileno(),VIDIOC_S_FMT,fmt)
# init mmap capture
fcntl.ioctl(self.vid.fileno(),VIDIOC_REQBUFS,req) # tell the driver that we want some buffers
for ind in range(self.req.count):
print(f"allocate MMAP buffer {ind}")
buf = v4l2_buffer()
buf.type = V4L2_BUF_TYPE_VIDEO_CAPTURE
buf.memory = V4L2_MEMORY_MMAP
buf.index = ind
buf.reserved = 0
buf.flags = 0
# queue the buffer for capture
ret = fcntl.ioctl(self.vid.fileno(),VIDIOC_QBUF,buf) # return 0 on success.
if ret != 0:
print("Could not allocate buffers!")
return -1
fcntl.ioctl(self.vid.fileno(),VIDIOC_QUERYBUF,buf)
return 0
我想要实现的目标(userptr):
def init_userptr(self,height=256,buffer=None):
req = v4l2_requestbuffers()
req.type = V4L2_BUF_TYPE_VIDEO_CAPTURE
req.memory = V4L2_MEMORY_USERPTR # V4L2_MEMORY_MMAP V4L2_MEMORY_USERPTR
req.count = 1 # number of buffer frames
self.height = max(self.height,req) # tell the driver that we want some buffers
for ind in range(self.req.count):
print(f"allocate USERPTR buffer {ind}")
buf = v4l2_buffer()
buf.type = V4L2_BUF_TYPE_VIDEO_CAPTURE
buf.memory = V4L2_MEMORY_USERPTR
buf.index = ind
buf.m.userptr = ctypes.c_ulong(ctypes.addressof(buffer[ind])) # Tried id(buffer) as well and few more similar options
buf.length = len(buffer) # ctypes.c_uint32(len(buffer))
# queue the buffer for capture
ret = fcntl.ioctl(self.vid.fileno(),buf) # ## FAILS HERE ## #
if ret != 0:
print("Could not allocate buffers!")
return -1
fcntl.ioctl(self.vid.fileno(),buf)
return 0
上一节中的缓冲区是这样分配的:
arr = ctypes.ARRAY(ctypes.ARRAY(ctypes.c_float,4194304),3)
buffer = arr()
代码在 ret = fcntl.ioctl(self.vid.fileno(),buf)
行失败并显示错误:
“OSError:[Errno 22] 无效参数”
在做了一些阅读之后,我遇到了这个: "不支持缓冲区类型,或索引越界,或尚未分配缓冲区,或用户指针或长度无效。"
这不是很有帮助。
我发现了一些用户指针 v4l2 读数的 C 实现,但我无法让它在我的 Python 代码上运行。
我尝试使用 ctypes 和 bytearray 分配缓冲区,但每次都得到相同的结果。
相关信息:
C 实现 - https://gist.github.com/maxlapshin/1253534
videodev2.h - https://elixir.bootlin.com/linux/latest/source/include/uapi/linux/videodev2.h#L1114
ioctl VIDIOC_QBUF、VIDIOC_DQBUF - https://www.kernel.org/doc/html/v4.15/media/uapi/v4l/vidioc-qbuf.html
编辑:
好的,经过一些阅读,我设法做到了
ret = fcntl.ioctl(self.vid.fileno(),buf)
通过改变我分配缓冲区的方式来工作。
缓冲区分配:
b_address = ctypes.c_void_p()
buf.length = fmt.fmt.pix.width * fmt.fmt.pix.height * 2
posix_memalign(ctypes.byref(b_address),getpagesize(),buf.length)
buf.m.userptr = b_address.value
buf.type = self.req.type
buf.memory = self.req.memory
ret = fcntl.ioctl(self.vid.fileno(),buf) # this line is now working!
我使用这篇文章作为参考 - user pointer in python
现在我面临着 select()
的新问题。当我尝试读取帧时,它超时并挂起 VIDIOC_DQBUF。
示例:
def get_frame(self):
dbuf = v4l2_buffer()
dbuf.type = V4L2_BUF_TYPE_VIDEO_CAPTURE
dbuf.memory = V4L2_MEMORY_USERPTR # V4L2_MEMORY_MMAP
# wait till buffer is ready for reading
t0 = time.time()
max_t = 5
ready_to_read,ready_to_write,in_error = ([],[],[])
while len(ready_to_read) == 0 and time.time() - t0 < max_t:
ready_to_read,in_error = select.select([self.vid.fileno()],max_t) # Reaches time out here
# read (de-queue) buffer
ret = fcntl.ioctl(self.vid.fileno(),VIDIOC_DQBUF,dbuf) # Hangs here.
我知道这个问题与缓冲区大小或可能不正确的分配有关,但我真的无法指出具体的事情。
任何帮助将不胜感激!
先谢谢大家!
版权声明:本文内容由互联网用户自发贡献,该文观点与技术仅代表作者本人。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容, 请发送邮件至 dio@foxmail.com 举报,一经查实,本站将立刻删除。