René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"create_datagram_endpoint() no longer compatible with CAN bus raw sockets ","articleBody":"# Bug report\r\n\r\n### Bug description:\r\n\r\n# Problem description\r\n\r\nWe are using `create_datagram_endpoint()` to set up an asyncio-managed connection to a native hardware CAN bus interface at `can0` under debian linux. We are using the `sock` named argument to pass in a pre-built raw socket to the FD.\r\n\r\nThis works in Python 3.5.3 but broke following a migration to 3.11.4. We found that the datagram endpoint would fail to be brought up, and the `connection_made(transport)` method on the `protocol_factory` would never be called. When I rejigged our code to await the coroutine I got:\r\n\r\n```\r\n    ValueError: A UDP Socket was expected, got \u003csocket.socket fd=7, family=29, type=3, proto=1, laddr=('can0',)\r\n```\r\n\r\nI believe this has gone unnoticed due to hardware CAN bus interfaces being rare.\r\n\r\n# Relevant history\r\n\r\nhttps://github.com/python/cpython/pull/4231\r\nhttps://github.com/python/cpython/pull/4898\r\nhttps://github.com/python/cpython/pull/4922\r\n\r\nDiscussion of underlying issues: https://bugs.python.org/issue32331\r\n\r\n# Core issue\r\n\r\nA socket connected to a CAN bus has `sock.type` `socket.SOCK_RAW`, I.E. `0b11`. This fails the comparison to `socket.SOCK_DGRAM` (`0b01`) so `create_datagram_endpoint()` raises an exception because the socket's type fails the \"is a datagram\" check, even though it is. In 3.5.3 the check was:\r\n\r\n```python\r\n    (sock.type \u0026 socket.SOCK_DGRAM) == socket.SOCK_DGRAM\r\n```\r\nThis worked because it was treating SOCK_DGRAM as a bitwise mask, which it kinda is, but kinda isn't.\r\n\r\nx86_64-linux-gnu/bits/socket_type.h:\r\n```c\r\n    ...\r\n    /* Types of sockets.  */\r\n    enum __socket_type\r\n    {\r\n    SOCK_STREAM = 1,\t\t/* Sequenced, reliable, connection-based\r\n                    byte streams.  */\r\n    #define SOCK_STREAM SOCK_STREAM\r\n    SOCK_DGRAM = 2,\t\t/* Connectionless, unreliable datagrams\r\n                    of fixed maximum length.  */\r\n    #define SOCK_DGRAM SOCK_DGRAM\r\n    SOCK_RAW = 3,\t\t\t/* Raw protocol interface.  */\r\n    #define SOCK_RAW SOCK_RAW\r\n    SOCK_RDM = 4,\t\t\t/* Reliably-delivered messages.  */\r\n    ...\r\n```\r\n\r\nI am torn on the best way to approach a fix here. I have written a PR that replaces\r\n\r\n```python\r\n    if sock.type != socket.SOCK_DGRAM:\r\n```\r\n\r\ncomparisons with\r\n\r\n```python\r\n    if not sock.type \u0026 socket.SOCK_DGRAM:\r\n```\r\nhowever this perpetuates the enum-vs-bitmask confusion. A more constrained change might be to change\r\n\r\n```python\r\n    if sock.type != socket.SOCK_DGRAM:\r\n```\r\n\r\nto\r\n\r\n```python\r\n    if not (sock.type == socket.SOCK_DGRAM || sock.type == socket.SOCK_RAW):\r\n```\r\n\r\nbut that may not catch every instance of the problem. I'd appreciate some guidance on this.\r\n\r\nCheers,\r\nTravis.\r\n\r\n### CPython versions tested on:\r\n\r\n3.11\r\n\r\n### Operating systems tested on:\r\n\r\nLinux\r\n\r\n\u003c!-- gh-linked-prs --\u003e\r\n### Linked PRs\r\n* gh-114888\r\n* gh-114893\n* gh-114977\n* gh-114978\n* gh-114979\n* gh-114980\n\u003c!-- /gh-linked-prs --\u003e\r\n","author":{"url":"https://github.com/tjhowse","@type":"Person","name":"tjhowse"},"datePublished":"2024-02-02T04:02:19.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":4},"url":"https://github.com/114887/cpython/issues/114887"}