René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Autobind of empty unix socket on Linux broken since python 3.9","articleBody":"**Bug report**\r\n\r\nWhen using empty address for unix socket on Linux, the socket is bound to a random\r\naddress in the abstract namespace.\r\n\r\nfrom unix(7):\r\n```\r\n   Autobind feature\r\n       If  a  bind(2)  call  specifies  addrlen as sizeof(sa_family_t), or the\r\n       SO_PASSCRED socket option was specified for a socket that was  not  ex‐\r\n       plicitly  bound  to  an address, then the socket is autobound to an ab‐\r\n       stract address.  The address consists of a  null  byte  followed  by  5\r\n       bytes  in  the  character set [0-9a-f].  Thus, there is a limit of 2^20\r\n       autobind addresses.  (From Linux 2.1.15, when the autobind feature  was\r\n       added,  8  bytes  were  used,  and the limit was thus 2^32 autobind ad‐\r\n       dresses.  The change to 5 bytes came in Linux 2.3.15.)\r\n```\r\n\r\nSince python 3.9, when using empty socket address (\"\"), bind() is called with\r\nwith `path.len + offsetof(struct sockaddr_un, sun_path) + 1`, so the socket is\r\nbound to the specific address \"\\0\" instead of a random address (e.g. \"\\075499\").\r\nThis breaks code assuming that multiple sockets can be bound to a random address\r\nat the same time.\r\n\r\nIn python 3.8 (not including this change):\r\n\r\n```\r\n$ python3.8\r\nPython 3.8.13 (default, Jun 10 2022, 00:00:00) \r\n[GCC 12.1.1 20220507 (Red Hat 12.1.1-1)] on linux\r\nType \"help\", \"copyright\", \"credits\" or \"license\" for more information.\r\n\u003e\u003e\u003e import socket\r\n\u003e\u003e\u003e s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)\r\n\u003e\u003e\u003e s.bind(\"\")\r\n\u003e\u003e\u003e s.getsockname()\r\nb'\\x0075499'\r\n```\r\n\r\nWith python 3.9 (including this change):\r\n```\r\n$ python3.9\r\nPython 3.9.13 (main, Jun  9 2022, 00:00:00) \r\n[GCC 12.1.1 20220507 (Red Hat 12.1.1-1)] on linux\r\nType \"help\", \"copyright\", \"credits\" or \"license\" for more information.\r\n\u003e\u003e\u003e import socket\r\n\u003e\u003e\u003e s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)\r\n\u003e\u003e\u003e s.bind(\"\")\r\n\u003e\u003e\u003e s.getsockname()\r\nb'\\x00'\r\n```\r\n\r\n**Your environment**\r\n\r\n- CPython versions tested on: python 3.9, python 3.10\r\n- Operating system and architecture: Fedora 36, Ubuntu latest (GitHub actions)\r\n\r\nThe breaking change is f6b3a07b7df60dc04d0260169ffef6e9796a2124, backported\r\nto python 3.9. I suspect that change was not needed since it did not include\r\nany test showing a real issue, but it is possible to fix handling of empty\r\nunix socket address without reverting this change. ","author":{"url":"https://github.com/nirs","@type":"Person","name":"nirs"},"datePublished":"2022-07-13T20:09:26.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/94821/cpython/issues/94821"}