pyudev.monitor module

Monitor implementation.

class pyudev.monitor.Monitor(context, monitor_p)

Bases: object

A synchronous device event monitor.

A Monitor objects connects to the udev daemon and listens for changes to the device list. A monitor is created by connecting to the kernel daemon through netlink (see from_netlink()):

>>> from pyudev import Context, Monitor
>>> context = Context()
>>> monitor = Monitor.from_netlink(context)

Once the monitor is created, you can add a filter using filter_by() or filter_by_tag() to drop incoming events in subsystems, which are not of interest to the application:

>>> monitor.filter_by('input')

When the monitor is eventually set up, you can either poll for events synchronously:

>>> device = monitor.poll(timeout=3)
>>> if device:
...     print('{0.action}: {0}'.format(device))
...

Or you can monitor events asynchronously with MonitorObserver.

To integrate into various event processing frameworks, the monitor provides a selectable file description by fileno(). However, do not read or write directly on this file descriptor.

Instances of this class can directly be given as udev_monitor * to functions wrapped through ctypes.

Changed in version 0.16: Remove from_socket() which is deprecated, and even removed in recent udev versions.

enable_receiving()

Switch the monitor into listing mode.

Connect to the event source and receive incoming events. Only after calling this method, the monitor listens for incoming events.

Note

This method is implicitly called by __iter__(). You don’t need to call it explicitly, if you are iterating over the monitor.

Deprecated since version 0.16: Will be removed in 1.0. Use start() instead.

fileno()

Return the file description associated with this monitor as integer.

This is really a real file descriptor ;), which can be watched and select.select()ed.

filter_by(subsystem, device_type=None)

Filter incoming events.

subsystem is a byte or unicode string with the name of a subsystem (e.g. 'input'). Only events originating from the given subsystem pass the filter and are handed to the caller.

If given, device_type is a byte or unicode string specifying the device type. Only devices with the given device type are propagated to the caller. If device_type is not given, no additional filter for a specific device type is installed.

These filters are executed inside the kernel, and client processes will usually not be woken up for device, that do not match these filters.

Changed in version 0.15: This method can also be after start() now.

filter_by_tag(tag)

Filter incoming events by the given tag.

tag is a byte or unicode string with the name of a tag. Only events for devices which have this tag attached pass the filter and are handed to the caller.

Like with filter_by() this filter is also executed inside the kernel, so that client processes are usually not woken up for devices without the given tag.

Required udev version: 154

New in version 0.9.

Changed in version 0.15: This method can also be after start() now.

Create a monitor by connecting to the kernel daemon through netlink.

context is the Context to use. source is a string, describing the event source. Two sources are available:

'udev' (the default)

Events emitted after udev as registered and configured the device. This is the absolutely recommended source for applications.

'kernel'

Events emitted directly after the kernel has seen the device. The device has not yet been configured by udev and might not be usable at all. Never use this, unless you know what you are doing.

Return a new Monitor object, which is connected to the given source. Raise ValueError, if an invalid source has been specified. Raise EnvironmentError, if the creation of the monitor failed.

poll(timeout=None)

Poll for a device event.

You can use this method together with iter() to synchronously monitor events in the current thread:

for device in iter(monitor.poll, None):
    print('{0.action} on {0.device_path}'.format(device))

Since this method will never return None if no timeout is specified, this is effectively an endless loop. With functools.partial() you can also create a loop that only waits for a specified time:

for device in iter(partial(monitor.poll, 3), None):
    print('{0.action} on {0.device_path}'.format(device))

This loop will only wait three seconds for a new device event. If no device event occurred after three seconds, the loop will exit.

timeout is a floating point number that specifies a time-out in seconds. If omitted or None, this method blocks until a device event is available. If 0, this method just polls and will never block.

Note

This method implicitly calls start().

Return the received Device, or None if a timeout occurred. Raise EnvironmentError if event retrieval failed.

See also

Device.action

The action that created this event.

Device.sequence_number

The sequence number of this event.

New in version 0.16.

receive_device()

Receive a single device from the monitor.

Warning

You must call start() before calling this method.

The caller must make sure, that there are events available in the event queue. The call blocks, until a device is available.

If a device was available, return (action, device). device is the Device object describing the device. action is a string describing the action. Usual actions are:

'add'

A device has been added (e.g. a USB device was plugged in)

'remove'

A device has been removed (e.g. a USB device was unplugged)

'change'

Something about the device changed (e.g. a device property)

'online'

The device is online now

'offline'

The device is offline now

Raise EnvironmentError, if no device could be read.

Deprecated since version 0.16: Will be removed in 1.0. Use Monitor.poll() instead.

remove_filter()

Remove any filters installed with filter_by() or filter_by_tag() from this monitor.

Warning

Up to udev 181 (and possibly even later versions) the underlying udev_monitor_filter_remove() seems to be broken. If used with affected versions this method always raises ValueError.

Raise EnvironmentError if removal of installed filters failed.

New in version 0.15.

set_receive_buffer_size(size)

Set the receive buffer size.

size is the requested buffer size in bytes, as integer.

Note

The CAP_NET_ADMIN capability must be contained in the effective capability set of the caller for this method to succeed. Otherwise EnvironmentError will be raised, with errno set to EPERM. Unprivileged processes typically lack this capability. You can check the capabilities of the current process with the python-prctl module:

>>> import prctl
>>> prctl.cap_effective.net_admin

Raise EnvironmentError, if the buffer size could not bet set.

New in version 0.13.

start()

Start this monitor.

The monitor will not receive events until this method is called. This method does nothing if called on an already started Monitor.

Note

Typically you don’t need to call this method. It is implicitly called by poll() and __iter__().

See also

started

Changed in version 0.16: This method does nothing if the Monitor was already started.

property started

True, if this monitor was started, False otherwise. Readonly.

See also

start()

New in version 0.16.

class pyudev.monitor.MonitorObserver(monitor, event_handler=None, callback=None, *args, **kwargs)

Bases: threading.Thread

An asynchronous observer for device events.

This class subclasses Thread class to asynchronously observe a Monitor in a background thread:

>>> from pyudev import Context, Monitor, MonitorObserver
>>> context = Context()
>>> monitor = Monitor.from_netlink(context)
>>> monitor.filter_by(subsystem='input')
>>> def print_device_event(device):
...     print('background event {0.action}: {0.device_path}'.format(device))
>>> observer = MonitorObserver(monitor, callback=print_device_event, name='monitor-observer')
>>> observer.daemon
True
>>> observer.start()

In the above example, input device events will be printed in background, until stop() is called on observer.

Note

Instances of this class are always created as daemon thread. If you do not want to use daemon threads for monitoring, you need explicitly set daemon to False before invoking start().

See also

Device.action

The action that created this event.

Device.sequence_number

The sequence number of this event.

New in version 0.14.

Changed in version 0.15: Monitor.start() is implicitly called when the thread is started.

run()

Method representing the thread’s activity.

You may override this method in a subclass. The standard run() method invokes the callable object passed to the object’s constructor as the target argument, if any, with sequential and keyword arguments taken from the args and kwargs arguments, respectively.

send_stop()

Send a stop signal to the background thread.

The background thread will eventually exit, but it may still be running when this method returns. This method is essentially the asynchronous equivalent to stop().

Note

The underlying monitor is not stopped.

start()

Start the observer thread.

stop()

Synchronously stop the background thread.

Note

This method can safely be called from the observer thread. In this case it is equivalent to send_stop().

Send a stop signal to the backgroud (see send_stop()), and waits for the background thread to exit (see join()) if the current thread is not the observer thread.

After this method returns in a thread that is not the observer thread, the callback is guaranteed to not be invoked again anymore.

Note

The underlying monitor is not stopped.

Changed in version 0.16: This method can be called from the observer thread.