.net - 为什么在尝试使用 Marshal.PtrToStructure 将字节数组转换为类实例时会出现访问冲突？

Question

在 C++ 中，我有这个:

struct BasePacketProto
{
    unsigned short PACKET_OPCODE;
    unsigned short PACKET_MAGIC_NUMBER;
    unsigned short PACKET_REMAIN_DATA_LENGTH;
};

struct CM_Request_AcceptAccount : BasePacketProto
{
    unsigned char ACCOUNT_LOGIN[16];
    unsigned char ACCOUNT_PASSWORD[16];
};

static void SendPacket()
{
    CM_Request_AcceptAccount packet;
    packet.PACKET_OPCODE = opCM_Request_AcceptAccount;
    packet.PACKET_MAGIC_NUMBER = 123;
    packet.PACKET_REMAIN_DATA_LENGTH = sizeof(CM_Request_AcceptAccount) -
                                       sizeof(BasePacketProto);
    memcpy(packet.ACCOUNT_LOGIN, "asd", sizeof("asd") * sizeof(char));
    memcpy(packet.ACCOUNT_PASSWORD, "asd_pass", sizeof("asd_pass") * sizeof(char));

    //Send the packet to the server.
    int lLength = send(lhSocket, (const char*)&packet, sizeof(CM_Request_AcceptAccount), 0);
}

在 C# ， 这个:

[StructLayout(LayoutKind.Sequential)]
class BasePacketProto
{
    public System.UInt16 PACKET_OPCODE;
    public System.UInt16 PACKET_MAGIC_NUMBER;
    public System.UInt16 PACKET_REMAIN_DATA_LENGTH;
}

[StructLayout(LayoutKind.Sequential)]
class CM_Request_AcceptAccount : BasePacketProto
{
    public byte[] ACCOUNT_LOGIN = new byte[16];
    public byte[] ACCOUNT_PASSWORD = new byte[16];
}

拆分数据包的类:

public class PacketProcessor
{
    static List<byte> raw_packet = new List<byte>();
    static int PACKET_HEADER_SIZE = Marshal.SizeOf(typeof(BasePacketProto));

    static public void ProcessPacketBytes(byte[] bytes, int size)
    {
        for (int i = 0; i < size; i++)
            raw_packet.Add(bytes[i]); //Adding bytes to own storage

        if (raw_packet.Count < PACKET_HEADER_SIZE) //If we don't have enough bytes
                                                   //to build base packet, we will
                                                   //return and wait for more.
            return;

        //This packet building works fine!
        BasePacketProto bpp =
            ConvertBytesTo<BasePacketProto>(raw_packet.GetRange(
                0, PACKET_HEADER_SIZE).ToArray());

        if (raw_packet.Count >= (PACKET_HEADER_SIZE +
                                 bpp.PACKET_REMAIN_DATA_LENGTH)) //If we have enough
                                           bytes in storage to restore child packet.
        {
            switch ((ClientPacketOpcodes)bpp.PACKET_OPCODE)
            {
                case ClientPacketOpcodes.opCM_Request_AcceptAccount:
                    //But this one fails
                    bpp = ConvertBytesTo<CM_Request_AcceptAccount>(raw_packet.GetRange(
                        0, PACKET_HEADER_SIZE + bpp.PACKET_REMAIN_DATA_LENGTH).ToArray());

                    PacketHandler.Handle_opCM_Request_AcceptAccount((CM_Request_AcceptAccount)bpp);
                    break;

                default:
                    break;
            }

            raw_packet.RemoveRange(0, PACKET_HEADER_SIZE + bpp.PACKET_REMAIN_DATA_LENGTH);
        }
    }

    static T ConvertBytesTo<T>(byte[] data)
    {
        unsafe
        {
            fixed(byte *ptr = data)
            {
                //I am getting an access violation here when trying to
                //build child packet :(
                return (T)Marshal.PtrToStructure(new IntPtr(ptr), typeof(T));
            }
        }
    }
}

在接收线程中的某处:

while (clientStream.CanRead)
{
    byte[] temp_buff = new byte[1024];
    int received = 0;

    while ((received = clientStream.Read(temp_buff, 0, 1024)) > 0) //Here we receive 38 bytes
    {
        //Passing it to the packet splitter.
        PacketProcessor.ProcessPacketBytes(temp_buff, received);
    }
}

客户端向服务器发送数据包时的结果:

System.AccessViolationException was not handled
Attempt to read or write to protected memory. Most likely it points that other memory is corrupted.

为什么无法将 38 个字节转换为 CM_Request_AcceptAccount？我该怎么做才能让它发挥作用？

Answer 1

当你声明一个结构体时

struct CM_Request_AcceptAccount : BasePacketProto
{
    unsigned char ACCOUNT_LOGIN[16];
    unsigned char ACCOUNT_PASSWORD[16];
};

在 C++ 中，数组是“内联”和“固定”长度的，换句话说，每个数组为结构的“大小”贡献 16 个字节。

但是在 C# 中，您重新声明了与以下相同的结构:

[StructLayout(LayoutKind.Sequential)]
class CM_Request_AcceptAccount : BasePacketProto
{
    public byte[] ACCOUNT_LOGIN = new byte[16];
    public byte[] ACCOUNT_PASSWORD = new byte[16];
}

在这里，您没有提供前 16 个字节属于 ACCOUNT_LOGIN 的信息。数组和接下来的 16 到 ACCOUNT_PASSWORD .

线
字节[] ACCOUNT_LOGIN = 新字节[16]

没有告诉编码员任何事情。它只会导致 CLR当 CM_Request_AcceptAccount 的实例时，在堆上分配一个 16 字节的数组是在代码中创建的。

为了正确编码结构，请将 C# 声明更改为:

[StructLayout(LayoutKind.Sequential)]
class CM_Request_AcceptAccount : BasePacketProto
{
    [MarshalAs(UnmanagedType.ByValArray, SizeConst=16)]
    public byte[] ACCOUNT_LOGIN;
    [MarshalAs(UnmanagedType.ByValArray, SizeConst=16)]
    public byte[] ACCOUNT_PASSWORD;
}

附加信息:由于您的字符数组旨在保存 C 样式字符串，您还可以使用

[MarshalAs(UnmanagedType.ByValTStr, SizeConst=16)]
public string ACCOUNT_LOGIN;

这里要记住的一点是，编码器将在您在 C++ 代码中设置的 16 个字符中期望一个空终止字符。