如果您从移动设备或其他发送文件

    byte[] fileData = null;
    using (var binaryReader = new BinaryReader(Request.Files[0].InputStream))
    {
        fileData = binaryReader.ReadBytes(Request.Files[0].ContentLength);
    }

你也可以尝试一次只读入部分，并展开返回的字节数组:

public byte[] StreamToByteArray(string fileName)
{
    byte[] total_stream = new byte[0];
    using (Stream input = File.Open(fileName, FileMode.Open, FileAccess.Read))
    {
        byte[] stream_array = new byte[0];
        // Setup whatever read size you want (small here for testing)
        byte[] buffer = new byte[32];// * 1024];
        int read = 0;

        while ((read = input.Read(buffer, 0, buffer.Length)) > 0)
        {
            stream_array = new byte[total_stream.Length + read];
            total_stream.CopyTo(stream_array, 0);
            Array.Copy(buffer, 0, stream_array, total_stream.Length, read);
            total_stream = stream_array;
        }
    }
    return total_stream;
}

我知道的最简单的解决方法是:

using(var memoryStream = new MemoryStream())
{
  sourceStream.CopyTo(memoryStream);
  return memoryStream.ToArray();
}

"bigEnough" array is a bit of a stretch. Sure, buffer needs to be "big ebough" but proper design of an application should include transactions and delimiters. In this configuration each transaction would have a preset length thus your array would anticipate certain number of bytes and insert it into correctly sized buffer. Delimiters would ensure transaction integrity and would be supplied within each transaction. To make your application even better, you could use 2 channels (2 sockets). One would communicate fixed length control message transactions that would include information about size and sequence number of data transaction to be transferred using data channel. Receiver would acknowledge buffer creation and only then data would be sent. If you have no control over stream sender than you need multidimensional array as a buffer. Component arrays would be small enough to be manageable and big enough to be practical based on your estimate of expected data. Process logic would seek known start delimiters and then ending delimiter in subsequent element arrays. Once ending delimiter is found, new buffer would be created to store relevant data between delimiters and initial buffer would have to be restructured to allow data disposal.

只要代码转换流字节数组是一个下面。

Stream s = yourStream;
int streamEnd = Convert.ToInt32(s.Length);
byte[] buffer = new byte[streamEnd];
s.Read(buffer, 0, streamEnd);

Byte[] Content = new BinaryReader(file.InputStream).ReadBytes(file.ContentLength);

Stream s;
int len = (int)s.Length;
byte[] b = new byte[len];
int pos = 0;
while((r = s.Read(b, pos, len - pos)) > 0) {
    pos += r;
}

一个稍微复杂一点的解决方案是s.Length超过Int32.MaxValue。但是如果需要将这么大的流读入内存，则可能需要考虑另一种解决问题的方法。

编辑:如果你的流不支持长度属性，使用Earwicker的变通方法进行修改。

public static class StreamExtensions {
    // Credit to Earwicker
    public static void CopyStream(this Stream input, Stream output) {
        byte[] b = new byte[32768];
        int r;
        while ((r = input.Read(b, 0, b.Length)) > 0) {
            output.Write(b, 0, r);
        }
    }
}

[...]

Stream s;
MemoryStream ms = new MemoryStream();
s.CopyStream(ms);
byte[] b = ms.GetBuffer();