Describe the bug
Python-can’s SocketcanBus uses Python’s select.select() to wait for socket read and write availability. This function is an interface to the Unix select system call (ref). There is a known issue where the file descriptor limit FD_SETSIZE is hardcoded to 1024 in the glibc implementation (ref).
This means that even on systems where the file descriptor limit is > 1023, when the socket file descriptor is > 1023, SocketcanBus propagates a ValueError raised by select.select() with the message, filedescriptor out of range in select().
The recommended fix is to use poll() or epoll() instead.
To Reproduce
- Set the system file descriptor limit above 1024 (e.g., ulimit -n 8192)
- Open > 1024 files to simulate a high number of open files (e.g., for i in {1..2000}; do exec {fd}>/dev/null; done
- Use a SocketcanBus instance to send or receive a CAN message
Expected behavior
I'd expect SocketcanBus to just work when socket fd > 1023
Additional context
OS and version: Ubuntu 22.04.5 LTS
Python version: 3.10
python-can version: first seen on 4.5.0, verified also an issue on 4.6.1 with unit tests below
python-can interface/s (if applicable): SocketcanBus
Traceback and logs
Traceback from repro on python-can 4.5.0:
self = <can.interfaces.socketcan.socketcan.SocketcanBus object at 0x758060538730>
msg = can.Message(timestamp=0.0, arbitration_id=0x18ebff17, is_extended_id=True, dlc=8, data=[0x1, 0x31, 0x58, 0x50, 0x42, 0x44, 0x34, 0x39])
timeout = 0
def send(self, msg: Message, timeout: Optional[float] = None) -> None:
"""Transmit a message to the CAN bus.
:param msg: A message object.
:param timeout:
Wait up to this many seconds for the transmit queue to be ready.
If not given, the call may fail immediately.
:raises ~can.exceptions.CanError:
if the message could not be written.
"""
log.debug("We've been asked to write a message to the bus")
logger_tx = log.getChild("tx")
logger_tx.debug("sending: %s", msg)
started = time.time()
# If no timeout is given, poll for availability
if timeout is None:
timeout = 0
time_left = timeout
data = build_can_frame(msg)
while time_left >= 0:
# Wait for write availability
> ready = select.select([], [self.socket], [], time_left)[1]
E ValueError: filedescriptor out of range in select()
../python_can+/can/interfaces/socketcan/socketcan.py:788: ValueError
Unit tests that also reproduce the error:
#!/usr/bin/env python
"""
Test that SocketcanBus.send() and recv() work with file descriptors > 1023.
"""
import select
import unittest
from unittest.mock import MagicMock, patch
import can
from can import Message
from can.interfaces.socketcan.socketcan import build_can_frame
from .config import IS_LINUX
HIGH_FD = 1024
@unittest.skipUnless(IS_LINUX, "socketcan is only available on Linux")
class TestSocketcanHighFd(unittest.TestCase):
"""Verify SocketcanBus works when the underlying socket fd exceeds 1023."""
def setUp(self):
patcher_create = patch("can.interfaces.socketcan.socketcan.create_socket")
patcher_bind = patch("can.interfaces.socketcan.socketcan.bind_socket")
self.mock_create_socket = patcher_create.start()
self.mock_bind_socket = patcher_bind.start()
self.mock_socket = MagicMock()
self.mock_socket.fileno.return_value = HIGH_FD
self.mock_create_socket.return_value = self.mock_socket
self.bus = can.Bus(interface="socketcan", channel="can0")
self.addCleanup(patcher_create.stop)
self.addCleanup(patcher_bind.stop)
def tearDown(self):
self.bus.shutdown()
def test_send_high_fd(self):
"""send() succeeds when the socket fd > 1023."""
msg = Message(arbitration_id=0x123, data=[1, 2, 3, 4, 5, 6, 7, 8])
frame_data = build_can_frame(msg)
self.mock_socket.send.return_value = len(frame_data)
self.bus.send(msg)
self.mock_socket.send.assert_called_once_with(frame_data)
@patch("can.interfaces.socketcan.socketcan.capture_message")
def test_recv_high_fd(self, mock_capture):
"""recv() succeeds when the socket fd > 1023."""
expected_msg = Message(
arbitration_id=0x123,
data=[1, 2, 3, 4, 5, 6, 7, 8],
channel="can0",
timestamp=1000.0,
)
mock_capture.return_value = expected_msg
msg = self.bus.recv(timeout=1.0)
self.assertIsNotNone(msg)
self.assertEqual(msg.arbitration_id, 0x123)
self.assertEqual(msg.data, bytearray([1, 2, 3, 4, 5, 6, 7, 8]))
mock_capture.assert_called_once_with(self.mock_socket, False)
if __name__ == "__main__":
unittest.main()Unit test failure output on 4.6.1:
FAIL [ 9.983ms] test/test_high_fd.py::TestSocketcanHighFd::test_recv_high_fd
stdout ───
stderr ───
test/test_high_fd.py:66 in test_recv_high_fd
self = <test.test_high_fd.TestSocketcanHighFd testMethod=test_recv_high_fd>, mock_capture = <MagicMock name='capture_message' id='140435170767152'>
63 │ │ )
64 │ │ mock_capture.return_value = expected_msg
65 │ │
❱ 66 │ │ msg = self.bus.recv(timeout=1.0)
67 │ │
68 │ │ self.assertIsNotNone(msg)
69 │ │ self.assertEqual(msg.arbitration_id, 0x123)
can/bus.py:121 in recv
118 │ │
119 │ │ while True:
120 │ │ │ # try to get a message
❱ 121 │ │ │ msg, already_filtered = self._recv_internal(timeout=time_left)
122 │ │ │
123 │ │ │ # return it, if it matches
124 │ │ │ if msg and (already_filtered or self._matches_filters(msg)):
121: in recv
❱
can/interfaces/socketcan/socketcan.py:827 in _recv_internal
self = <can.interfaces.socketcan.socketcan.SocketcanBus object at 0x7fb99c7661d0>, timeout = 1.0
824 │ │ try:
825 │ │ │ # get all sockets that are ready (can be a list with a single value
826 │ │ │ # being self.socket or an empty list if self.socket is not ready)
❱ 827 │ │ │ ready_receive_sockets, _, _ = select.select([self.socket], [], [], timeout)
828 │ │ except OSError as error:
829 │ │ │ # something bad happened (e.g. the interface went down)
830 │ │ │ raise can.CanOperationError(
827: ValueError
❱ ready_receive_sockets, _, _ = select.select([self.socket], [], [], timeout)
E ValueError: filedescriptor out of range in select()
FAIL [ 4.712ms] test/test_high_fd.py::TestSocketcanHighFd::test_send_high_fd
stdout ───
stderr ───
test/test_high_fd.py:50 in test_send_high_fd
self = <test.test_high_fd.TestSocketcanHighFd testMethod=test_send_high_fd>
47 │ │ frame_data = build_can_frame(msg)
48 │ │ self.mock_socket.send.return_value = len(frame_data)
49 │ │
❱ 50 │ │ self.bus.send(msg)
51 │ │
52 │ │ self.mock_socket.send.assert_called_once_with(frame_data)
53
can/interfaces/socketcan/socketcan.py:869 in send
self = <can.interfaces.socketcan.socketcan.SocketcanBus object at 0x7fb99c71bf70>, msg = can.Message(timestamp=0.0, arbitration_id=0x123,
is_extended_id=True, dlc=8, data=[0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8]), timeout = 0
866 │ │
867 │ │ while time_left >= 0:
868 │ │ │ # Wait for write availability
❱ 869 │ │ │ ready = select.select([], [self.socket], [], time_left)[1]
870 │ │ │ if not ready:
871 │ │ │ │ # Timeout
872 │ │ │ │ break
869: ValueError
❱ ready = select.select([], [self.socket], [], time_left)[1]
E ValueError: filedescriptor out of range in select()
Describe the bug
Python-can’s SocketcanBus uses Python’s select.select() to wait for socket read and write availability. This function is an interface to the Unix select system call (ref). There is a known issue where the file descriptor limit FD_SETSIZE is hardcoded to 1024 in the glibc implementation (ref).
This means that even on systems where the file descriptor limit is > 1023, when the socket file descriptor is > 1023, SocketcanBus propagates a ValueError raised by select.select() with the message, filedescriptor out of range in select().
The recommended fix is to use poll() or epoll() instead.
To Reproduce
Expected behavior
I'd expect SocketcanBus to just work when socket fd > 1023
Additional context
OS and version: Ubuntu 22.04.5 LTS
Traceback and logsPython version: 3.10
python-can version: first seen on 4.5.0, verified also an issue on 4.6.1 with unit tests below
python-can interface/s (if applicable): SocketcanBus
Traceback from repro on python-can 4.5.0:
self = <can.interfaces.socketcan.socketcan.SocketcanBus object at 0x758060538730> msg = can.Message(timestamp=0.0, arbitration_id=0x18ebff17, is_extended_id=True, dlc=8, data=[0x1, 0x31, 0x58, 0x50, 0x42, 0x44, 0x34, 0x39]) timeout = 0 def send(self, msg: Message, timeout: Optional[float] = None) -> None: """Transmit a message to the CAN bus. :param msg: A message object. :param timeout: Wait up to this many seconds for the transmit queue to be ready. If not given, the call may fail immediately. :raises ~can.exceptions.CanError: if the message could not be written. """ log.debug("We've been asked to write a message to the bus") logger_tx = log.getChild("tx") logger_tx.debug("sending: %s", msg) started = time.time() # If no timeout is given, poll for availability if timeout is None: timeout = 0 time_left = timeout data = build_can_frame(msg) while time_left >= 0: # Wait for write availability > ready = select.select([], [self.socket], [], time_left)[1] E ValueError: filedescriptor out of range in select() ../python_can+/can/interfaces/socketcan/socketcan.py:788: ValueErrorUnit tests that also reproduce the error:
Unit test failure output on 4.6.1:
FAIL [ 9.983ms] test/test_high_fd.py::TestSocketcanHighFd::test_recv_high_fd stdout ─── stderr ─── test/test_high_fd.py:66 in test_recv_high_fd self = <test.test_high_fd.TestSocketcanHighFd testMethod=test_recv_high_fd>, mock_capture = <MagicMock name='capture_message' id='140435170767152'> 63 │ │ ) 64 │ │ mock_capture.return_value = expected_msg 65 │ │ ❱ 66 │ │ msg = self.bus.recv(timeout=1.0) 67 │ │ 68 │ │ self.assertIsNotNone(msg) 69 │ │ self.assertEqual(msg.arbitration_id, 0x123) can/bus.py:121 in recv 118 │ │ 119 │ │ while True: 120 │ │ │ # try to get a message ❱ 121 │ │ │ msg, already_filtered = self._recv_internal(timeout=time_left) 122 │ │ │ 123 │ │ │ # return it, if it matches 124 │ │ │ if msg and (already_filtered or self._matches_filters(msg)): 121: in recv ❱ can/interfaces/socketcan/socketcan.py:827 in _recv_internal self = <can.interfaces.socketcan.socketcan.SocketcanBus object at 0x7fb99c7661d0>, timeout = 1.0 824 │ │ try: 825 │ │ │ # get all sockets that are ready (can be a list with a single value 826 │ │ │ # being self.socket or an empty list if self.socket is not ready) ❱ 827 │ │ │ ready_receive_sockets, _, _ = select.select([self.socket], [], [], timeout) 828 │ │ except OSError as error: 829 │ │ │ # something bad happened (e.g. the interface went down) 830 │ │ │ raise can.CanOperationError( 827: ValueError ❱ ready_receive_sockets, _, _ = select.select([self.socket], [], [], timeout) E ValueError: filedescriptor out of range in select() FAIL [ 4.712ms] test/test_high_fd.py::TestSocketcanHighFd::test_send_high_fd stdout ─── stderr ─── test/test_high_fd.py:50 in test_send_high_fd self = <test.test_high_fd.TestSocketcanHighFd testMethod=test_send_high_fd> 47 │ │ frame_data = build_can_frame(msg) 48 │ │ self.mock_socket.send.return_value = len(frame_data) 49 │ │ ❱ 50 │ │ self.bus.send(msg) 51 │ │ 52 │ │ self.mock_socket.send.assert_called_once_with(frame_data) 53 can/interfaces/socketcan/socketcan.py:869 in send self = <can.interfaces.socketcan.socketcan.SocketcanBus object at 0x7fb99c71bf70>, msg = can.Message(timestamp=0.0, arbitration_id=0x123, is_extended_id=True, dlc=8, data=[0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8]), timeout = 0 866 │ │ 867 │ │ while time_left >= 0: 868 │ │ │ # Wait for write availability ❱ 869 │ │ │ ready = select.select([], [self.socket], [], time_left)[1] 870 │ │ │ if not ready: 871 │ │ │ │ # Timeout 872 │ │ │ │ break 869: ValueError ❱ ready = select.select([], [self.socket], [], time_left)[1] E ValueError: filedescriptor out of range in select()