Python asyncio SSL 傳輸未關閉

Question

我正在嘗試使用 python 異步協議建立與 Windows 10 VM 的 RDP 連接。 在第一次消息交換協商 TLS 連接之后，必須將連接升級到 TLS，為此我使用 start_tls。 一切正常，直到我嘗試通過調用 transport.close() 在第二個數據包之后關閉連接。 文檔說協議實例的“connection_lost”方法在所有 output 緩沖區都被刷新后調用為 None （ Source ）。 在我的例子中， close() 被調用並且調用完成，但 connection_lost() 從未被調用，這使得我的應用程序等待直到服務器在超時后發送一個 RST，這會正確觸發 connection_lost。 如果服務器協商 TLS 1.2，則確實會發生這種情況，但對於 TLS 1.3 則不會。 我檢查了 output 緩沖區是否為空，我查看了 cpython 源但無法弄清楚它掛在哪里。 使用 Wireshark，我看到調用 close() 時沒有任何反應，我的客戶端沒有向服務器發送任何內容。 我在 Ubuntu 20.04 上使用 python 3.8.5。

有誰知道為什么不調用 connection_lost ？

請參閱下面的最小工作示例：

#!/usr/bin/env python3

import asyncio
import ssl
import binascii

# first two packets for RDP, tested thoroughly
PACKET_NEGO = binascii.unhexlify("0300002d28e00000000000436f6f6b69653a206d737473686173683d6c696f6e6c656c0d0a0100080001000000")
PACKET_CONN = binascii.unhexlify("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")

class RDPConnection(asyncio.Protocol):
    def __init__(self, on_con_close, on_tls, loop, ip):
        self.on_tls = on_tls
        self.on_con_close = on_con_close
        self.loop = loop
        self.state = 0
        self.ip = ip

    def connection_made(self, transport):
        self.transport = transport

        if self.state == 0:
            transport.write(PACKET_NEGO) # write first packet on initial connection
        else:
            transport.write(PACKET_CONN) # write second packet on TLS connection

    def data_received(self, data):
        if self.state == 0:
            self.state += 1
            self.on_tls.set_result(True) # start TLS connection
        else:
            print("closing connection")
            self.transport.close()
            print("closed connection")
        
    def connection_lost(self, exc):
        print("connection lost called")
        self.on_con_close.set_result(True)
        print("{}: {}".format(self.ip, exc))

    def _timeout(self):
        if self.transport:
            self.transport.close()
        
async def handle_connection(loop, ip, timeout):
    on_tls = loop.create_future() # called when connection is ready to be upgraded to TLS
    con_close = loop.create_future() # called when connection is closed

    try:
        transport, protocol = await asyncio.wait_for(loop.create_connection(lambda: RDPConnection(con_close, on_tls, loop, ip), ip, 3389), timeout=timeout)
    except:
        return
    await on_tls # wait for first response

    ssl_ctx = ssl._create_unverified_context() # VM uses self signed certificate
    try:
        transport2 = await asyncio.wait_for(loop.start_tls(transport, protocol, ssl_ctx), timeout=timeout) # upgrade to TLS
        protocol.connection_made(transport2) # set TLS transport
    except:
        return
    await con_close # wait for connection_lost to be called
    return

async def main():
    loop = asyncio.get_running_loop()
    await handle_connection(loop, "my VM", 10)

asyncio.run(main())

Output 在我的機器上：

closing connection
closed connection
connection lost called
my VM: [Errno 104] Connection reset by peer

請注意，“連接丟失調用”僅在服務器重置連接時調用（約 30 秒后）

Answer 1

transport.abort()是你的朋友。

參考