GIO is striving to provide a modern, easy-to-use VFS API that sits at the right level in the library stack, as well as other generally useful APIs for desktop applications (such as networking and D-Bus support). The goal is to overcome the shortcomings of GnomeVFS and provide an API that is so good that developers prefer it over raw POSIX calls. Among other things that means using GObject. It also means not cloning the POSIX API, but providing higher-level, document-centric interfaces. The abstract file system model of GIO consists of a number of interfaces and base classes for I/O and files: GFile reference to a file GFileInfo information about a file or filesystem GFileEnumerator list files in directories GDrive represents a drive GVolume represents a file system in an abstract way GMount represents a mounted file system Then there is a number of stream classes, similar to the input and output stream hierarchies that can be found in frameworks like Java: GInputStream read data GOutputStream write data GIOStream read and write data GSeekable interface optionally implemented by streams to support seeking There are interfaces related to applications and the types of files they handle: GAppInfo information about an installed application GIcon abstract type for file and application icons There is a framework for storing and retrieving application settings: GSettings stores and retrieves application settings There is support for network programming, including connectivity monitoring, name resolution, lowlevel socket APIs and highlevel client and server helper classes: GSocket lowlevel platform independent socket object GResolver asynchronous and cancellable DNS resolver GSocketClient high-level network client helper GSocketService high-level network server helper GSocketConnection network connection stream GNetworkMonitor network connectivity monitoring There is support for connecting to D-Bus, sending and receiving messages, owning and watching bus names, and making objects available on the bus: GDBusConnection a D-Bus connection GDBusMethodInvocation for handling remote calls GDBusServer helper for accepting connections GDBusProxy proxy to access D-Bus interfaces on a remote object Beyond these, GIO provides facilities for file monitoring, asynchronous I/O and filename completion. In addition to the interfaces, GIO provides implementations for the local case. Implementations for various network file systems are provided by the GVFS package as loadable modules. Other design choices which consciously break with the GnomeVFS design are to move backends out-of-process, which minimizes the dependency bloat and makes the whole system more robust. The backends are not included in GIO, but in the separate GVFS package. The GVFS package also contains the GVFS daemon, which spawn further mount daemons for each individual connection.

The GIO model of I/O is stateful: if an application establishes e.g. a SFTP connection to a server, it becomes available to all applications in the session; the user does not have to enter his password over and over again. One of the big advantages of putting the VFS in the GLib layer is that GTK+ can directly use it, e.g. in the filechooser. The information in the GLib documentation about writing GLib applications is generally applicable when writing GIO applications. GDBus has its own private worker thread, so applications using GDBus have at least 3 threads. GIO makes heavy use of the concept of a thread-default main context to execute callbacks of asynchronous methods in the same context in which the operation was started. Many GIO functions come in two versions: synchronous and asynchronous, denoted by an _async suffix. It is important to use these appropriately: synchronous calls should not be used from within a main loop which is shared with other code, such as one in the application’s main thread. Synchronous calls block until they complete, and I/O operations can take noticeable amounts of time (even on ‘fast’ SSDs). Blocking a main loop iteration while waiting for I/O means that other sources in the main loop will not be dispatched, such as input and redraw handlers for the application’s UI. This can cause the application to ‘freeze’ until I/O completes. A few self-contained groups of functions, such as code generated by gdbus-codegen, use a different convention: functions are asynchronous default, and it is the synchronous version which has a _sync suffix. Aside from naming differences, they should be treated the same way as functions following the normal convention above. The asynchronous (_async) versions of functions return control to the caller immediately, after scheduling the I/O in the kernel and adding a callback for it to the main loop. This callback will be invoked when the operation has completed. From the callback, the paired _finish function should be called to retrieve the return value of the I/O operation, and any errors which occurred. For more information on using and implementing asynchronous functions, see GAsyncResult and GTask. By starting multiple asynchronous operations in succession, they will be executed in parallel (up to an arbitrary limit imposed by GIO’s internal worker thread pool). The synchronous versions of functions can be used early in application startup when there is no main loop to block, for example to load initial configuration files. They can also be used for I/O on files which are guaranteed to be small and on the local disk. Note that the user’s home directory is not guaranteed to be on the local disk. When your program needs to carry out some privileged operation (say, create a new user account), there are various ways in which you can go about this: Implement a daemon that offers the privileged operation. A convenient way to do this is as a D-Bus system-bus service. The daemon will probably need ways to check the identity and authorization of the caller before executing the operation. polkit is a framework that allows this. Use a small helper that is executed with elevated privileges via pkexec. pkexec is a small program launcher that is part of polkit. Use a small helper that is executed with elevated privileges by being suid root. None of these approaches is the clear winner, they all have their advantages and disadvantages. When writing code that runs with elevated privileges, it is important to follow some basic rules of secure programming. David Wheeler has an excellent book on this topic, Secure Programming for Linux and Unix HOWTO. When using GIO in code that runs with elevated privileges, you have to be careful. GIO has extension points whose implementations get loaded from modules (executable code in shared objects), which could allow an attacker to sneak his own code into your application by tricking it into loading the code as a module. However, GIO will never load modules from your home directory except when explictly asked to do so via an environment variable. In most cases, your helper program should be so small that you don't need GIO, whose APIs are largely designed to support full-blown desktop applications. If you can't resist the convenience of these APIs, here are some steps you should take: Clear the environment, e.g. using the clearenv() function. David Wheeler has a good explanation for why it is important to sanitize the environment. See for a list of all environment variables affecting GIO. In particular, PATH (used to locate binaries), GIO_EXTRA_MODULES (used to locate loadable modules) and DBUS_{SYSTEM,SESSION}_BUS_ADDRESS (used to locate the D-Bus system and session bus) are important. Don't use GVfs, by setting GIO_USE_VFS=local in the environment. The reason to avoid GVfs in security-sensitive programs is that it uses many libraries which have not necessarily been audited for security problems. Gvfs is also heavily distributed and relies on a session bus to be present. GIO comes with a gio-2.0.pc file that you should use together with pkg-config to obtain the necessary information about header files and libraries. See the pkg-config man page or the GLib documentation for more information on how to use pkg-config to compile your application. If you are using GIO on UNIX-like systems, you may want to use UNIX-specific GIO interfaces such as #GUnixInputStream, #GUnixOutputStream, #GUnixMount or #GDesktopAppInfo. To do so, use the gio-unix-2.0.pc file instead of gio-2.0.pc GIO inspects a few of environment variables in addition to the ones used by GLib. GIO uses these environment variables to locate MIME information. For more information, see the Shared MIME-info Database and the Base Directory Specification. This variable can be set to keep #Gvfs from starting the fuse backend, which may be unwanted or unnecessary in certain situations. The following environment variables are only useful for debugging GIO itself or modules that it loads. They should not be set in a production environment. This environment variable can be set to the name of a #GVfs implementation to override the default for debugging purposes. The #GVfs implementation for local files that is included in GIO has the name "local", the implementation in the gvfs module has the name "gvfs". This variable can be set to the name of a #GFileMonitor implementation to override the default for debugging purposes. The #GFileMonitor implementation for local files that is included in GIO on Linux has the name "inotify", others that are built are built as modules (depending on the platform) are called "fam" and "fen". This variable can be set to the name of a #GVolumeMonitor implementation to override the default for debugging purposes. The #GVolumeMonitor implementation for local files that is included in GIO has the name "unix", the hal-based implementation in the gvfs module has the name "hal". This variable can be set to the name of a #GTlsBackend implementation to override the default for debugging purposes. GIO does not include a #GTlsBackend implementation, the gnutls-based implementation in the glib-networking module has the name "gnutls". When this environment variable is set to a path, GIO will load modules from this alternate directory instead of the directory built into GIO. This is useful when running tests, for example. When this environment variable is set to a path, or a set of paths separated by a colon, GIO will attempt to load additional modules from within the path. This variable can be set to the name of a #GSettingsBackend implementation to override the default for debugging purposes. The memory-based implementation that is included in GIO has the name "memory", the one in dconf has the name "dconf-settings". This variable can be set to the name of a directory that is considered in addition to the glib-2.0/schemas subdirectories of the XDG system data dirs when looking for compiled schemas for #GSettings. This variable is consulted to find the address of the D-Bus system bus. For the format of D-Bus addresses, see the D-Bus specification. Setting this variable overrides platform-specific ways of determining the system bus address. This variable is consulted to find the address of the D-Bus session bus. Setting this variable overrides platform-specific ways of determining the session bus address. This variable is consulted to find out the 'starter' bus for an application that has been started via D-Bus activation. The possible values are 'system' or 'session'. This variable can be set to a list of debug options, which cause GLib to print out different types of debugging information when using the D-Bus routines. transport Show IO activity (e.g. reads and writes) message Show all sent and received D-Bus messages payload Show payload for all sent and received D-Bus messages (implies message) call Trace g_dbus_connection_call() and g_dbus_connection_call_sync() API usage signal Show when a D-Bus signal is received incoming Show when an incoming D-Bus method call is received return Show when a reply is returned via the #GDBusMethodInvocation API emission Trace g_dbus_connection_emit_signal() API usage authentication Show information about connection authentication address Show information about D-Bus address lookups and autolaunching The special value all can be used to turn on all debug options. The special value help can be used to print a list of supported options to standard output. Can be used to override the directory used to store the keyring used in the DBUS_COOKIE_SHA1 authentication mechanism. Normally the directory used is .dbus-keyrings in the user's home directory. If set, the permissions of the directory used to store the keyring used in the DBUS_COOKIE_SHA1 authentication mechanism won't be checked. Normally the directory must be readable only by the user. A lot of the functionality that is accessible through GIO is implemented in loadable modules, and modules provide a convenient way to extend GIO. In addition to the #GIOModule API which supports writing such modules, GIO has a mechanism to define extension points, and register implementations thereof, see #GIOExtensionPoint. The following extension points are currently defined by GIO: Allows to override the functionality of the #GVfs class. Implementations of this extension point must be derived from #GVfs. GIO uses the implementation with the highest priority that is active, see g_vfs_is_active(). GIO implements this extension point for local files, gvfs contains an implementation that supports all the backends in gvfs. Allows to add more volume monitors. Implementations of this extension point must be derived from #GVolumeMonitor. GIO uses all registered extensions. gvfs contains an implementation that works together with the #GVfs implementation in gvfs. Allows to override the 'native' volume monitor. Implementations of this extension point must be derived from #GNativeVolumeMonitor. GIO uses the implementation with the highest priority that is supported, as determined by the is_supported() vfunc in #GVolumeMonitorClass. GIO implements this extension point for local mounts, gvfs contains a hal-based implementation. Allows to override the file monitor implementation for local files. Implementations of this extension point must be derived from #GLocalFileMonitor. GIO uses the implementation with the highest priority that is supported, as determined by the is_supported() vfunc in #GLocalFileMonitorClass. GIO uses this extension point internally, to switch between its fam-based and inotify-based file monitoring implementations. Allows to override the directory monitor implementation for local files. Implementations of this extension point must be derived from #GLocalDirectoryMonitor. GIO uses the implementation with the highest priority that is supported, as determined by the is_supported() vfunc in #GLocalDirectoryMonitorClass. GIO uses this extension point internally, to switch between its fam-based and inotify-based directory monitoring implementations. Unix-only. Allows to provide a way to associate default handlers with URI schemes. Implementations of this extension point must implement the #GDesktopAppInfoLookup interface. GIO uses the implementation with the highest priority. This extension point has been discontinued in GLib 2.28. It is still available to keep API and ABI stability, but GIO is no longer using it for default handlers. Instead, the mime handler mechanism is used, together with x-scheme-handler pseudo-mimetypes. Allows to provide an alternative storage for #GSettings. Implementations of this extension point must derive from the #GSettingsBackend type. GIO contains a keyfile-based implementation of this extension point, another one is provided by dconf. Allows to provide implementations for network proxying. Implementations of this extension point must provide the #GProxy interface, and must be named after the network protocol they are proxying. glib-networking contains an implementation of this extension point based on libproxy. Allows to provide implementations for TLS support. Implementations of this extension point must implement the #GTlsBackend interface. glib-networking contains an implementation of this extension point. Allows to provide implementations for network connectivity monitoring. Implementations of this extension point must implement the #GNetworkMonitorInterface interface. GIO contains an implementation of this extension point that is using the netlink interface of the Linux kernel.