[Bug]: GNet对于accpet的到的socket没有应用用户的设置

[CloudGuan]

Actions I've taken before I'm here

• [X] I've thoroughly read the documentations on this issue but still have no clue.
• [X] I've searched the Github Issues but didn't find any duplicate issues that have been resolved.
• [X] I've searched the internet for this issue but didn't find anything helpful.

What happened?

在linux系上，用户初始化gnet时，如果配置了socket相关配置，如：接受缓冲区长度，nodely, 关闭timewait等设置 在accept之后，是否应该对于新的文件描述服也要应用对应设置？ 代码如下：

// 用户设置
err := gnet.Serve(gl.sEH, protoAddr, gnet.WithReuseAddr(true), gnet.WithTCPKeepAlive(0),
          gnet.WithTCPNoDelay(gnet.TCPNoDelay), gnet.WithSocketRecvBuffer(gl.socketBufferSize),
          gnet.WithSocketSendBuffer(gl.socketBufferSize),
)

accept 代码

func (svr *server) accept(fd int, _ netpoll.IOEvent) error {
    nfd, sa, err := unix.Accept(fd)
    if err != nil {
        if err == unix.EAGAIN || err == unix.ECONNABORTED {
            return nil
        }
        svr.opts.Logger.Errorf("Accept() fails due to error: %v", err)
        return errors.ErrAcceptSocket
    }
    if err = os.NewSyscallError("fcntl nonblock", unix.SetNonblock(nfd, true)); err != nil {
        return err
    }

    remoteAddr := socket.SockaddrToTCPOrUnixAddr(sa)
    if svr.opts.TCPKeepAlive > 0 && svr.ln.network == "tcp" {
        err = socket.SetKeepAlive(nfd, int(svr.opts.TCPKeepAlive/time.Second))
        logging.Error(err)
    }

    el := svr.lb.next(remoteAddr)
    c := newTCPConn(nfd, el, sa, svr.opts.Codec, el.ln.addr, remoteAddr)

    err = el.poller.UrgentTrigger(el.register, c)
    if err != nil {
        _ = unix.Close(nfd)
        c.releaseTCP()
    }
    return nil
}

Major version of gnet

v1

Specific version of gnet

v1.6.7

Operating system

Linux

Relevant log output

无

Code snippets (optional)

建议是否可以修改为这样，我认为用户的配置应该对于accept到的socket也要生效

func (svr *server) accept(fd int, _ netpoll.IOEvent) error {
    nfd, sa, err := unix.Accept(fd)
    if err != nil {
        if err == unix.EAGAIN || err == unix.ECONNABORTED {
            return nil
        }
        svr.opts.Logger.Errorf("Accept() fails due to error: %v", err)
        return errors.ErrAcceptSocket
    }
    if err = os.NewSyscallError("fcntl nonblock", unix.SetNonblock(nfd, true)); err != nil {
        return err
    }

    remoteAddr := socket.SockaddrToTCPOrUnixAddr(sa)
    if svr.opts.TCPKeepAlive > 0 && svr.ln.network == "tcp" {
        err = socket.SetKeepAlive(nfd, int(svr.opts.TCPKeepAlive/time.Second))
        logging.Error(err)
    }

    // Set RecvBufferSize
    if svr.opts.SocketRecvBuffer > 0 {
        err = socket.SetRecvBuffer(nfd, svr.opts.SocketRecvBuffer)
        logging.Error(err)
    }

    // SetSendBufferSize
    if svr.opts.SocketSendBuffer > 0 {
        err = socket.SetSendBuffer(nfd, svr.opts.SocketRecvBuffer)
        logging.Error(err)
    }

    // Disabling TIME_WAIT state
    l := unix.Linger{
        Onoff:  0,
        Linger: 0,
    }
    err = unix.SetsockoptLinger(nfd, unix.SOL_SOCKET, unix.SO_LINGER, &l)
    if err != nil {
        logging.Error(err)
    }

    el := svr.lb.next(remoteAddr)
    c := newTCPConn(nfd, el, sa, svr.opts.Codec, el.ln.addr, remoteAddr)

    err = el.poller.UrgentTrigger(el.register, c)
    if err != nil {
        _ = unix.Close(nfd)
        c.releaseTCP()
    }
    return nil
}

### How to Reproduce

Steps to reproduce the behavior:
1. Go to '....'
2. Click on '....'
3. Do '....'
4. See '....'

### Does this issue reproduce with the latest release?

It can reproduce with the latest release

[panjf2000]

根据 TCP 协议标准，通过 accept 返回的新 sockets 会从 listening socket 继承以下 socket options：

• SO_DEBUG
• SO_DONTROUTE
• SO_KEEPALIVE
• SO_LINGER
• SO_OOBINLINE
• SO_RCVBUF
• SO_RCVLOWAT
• SO_SNDBUF
• SO_SNDLOWAT
• TCP_MAXSEG
• TCP_NODELAY

在 Linux 系统上，accpet 返回的 socket 不会继承 file flags 比如 O_NONBLOCK/O_ASYNC，而 BSD 系统则会连 file flags 也一起继承下来，所以并不需要再次显式设置。