fs_socket.h

Go to the documentation of this file.

/* KallistiOS ##version##
 
   kos/fs_socket.h
   Copyright (C) 2006, 2009, 2010, 2012, 2013 Lawrence Sebald
 
*/
 
/** \file    kos/fs_socket.h
    \brief   Definitions for a sockets "filesystem".
    \ingroup vfs_sockets
 
    This file provides definitions to support the BSD-sockets-like filesystem
    in KallistiOS. Technically, this filesystem mounts itself on /sock, but it
    doesn't export any files there, so that point is largely irrelevant. The
    filesystem is designed to be extensible, making it possible to add
    additional socket family handlers at runtime. Currently, the kernel only
    implements UDP sockets over IPv4 and IPv6, but as mentioned, this can be
    extended in a fairly straightforward manner. In general, as a user of
    KallistiOS (someone not interested in adding additional socket family
    drivers), there's very little in this file that will be of interest.
 
    \author Lawrence Sebald
*/
 
#ifndef __KOS_FS_SOCKET_H
#define __KOS_FS_SOCKET_H
 
#include <kos/cdefs.h>
 
__BEGIN_DECLS
 
#include <kos/fs.h>
#include <kos/net.h>
#include <sys/queue.h>
#include <sys/socket.h>
#include <stdint.h>
 
/** \defgroup vfs_sockets   Sockets
    \brief                  VFS Driver for adding a BSD-sockets-like
                            filesystem.
    \ingroup  vfs_drivers
*/
 
struct fs_socket_proto;
 
/** \brief   Internal representation of a socket for fs_socket.
    \ingroup vfs_sockets
 
    This structure is the internal representation of a socket "file" that is
    used within fs_socket. A normal user will never deal with this structure
    directly (only protocol handlers and fs_socket itself ever sees this
    structure directly).
 
    \headerfile kos/fs_socket.h
*/
typedef struct net_socket {
    /** \cond */
    /* List handle */
    LIST_ENTRY(net_socket) sock_list;
    /** \endcond */
 
    /** \brief   File handle from the VFS layer. */
    file_t fd;
 
    /** \brief  The protocol handler for this socket. */
    struct fs_socket_proto *protocol;
 
    /** \brief  Protocol-specific data. */
    void *data;
} net_socket_t;
 
/** \brief   Internal sockets protocol handler.
    \ingroup vfs_sockets
 
    This structure is a protocol handler used within fs_socket. Each protocol
    that is supported has one of these registered for it within the kernel.
    Generally, users will not come in contact with this structure (unless you're
    planning on writing a protocol handler), and it can generally be ignored.
 
    For a complete list of appropriate errno values to return from any functions
    that are in here, take a look at the Single Unix Specification (aka, the
    POSIX spec), specifically the page about sys/socket.h (and all the functions
    that it defines, which is available at
    http://www.opengroup.org/onlinepubs/9699919799/basedefs/sys_socket.h.html .
 
    \headerfile kos/fs_socket.h
*/
typedef struct fs_socket_proto {
    /** \brief  Entry into the global list of protocols.
 
        Contrary to what Doxygen might think, this is <b>NOT</b> a function.
        This should be initialized with the FS_SOCKET_PROTO_ENTRY macro before
        adding the protocol to the kernel with fs_socket_proto_add().
    */
    TAILQ_ENTRY(fs_socket_proto) entry;
 
    /** \brief  Domain of support for this protocol handler.
 
        This field determines which sockets domain this protocol handler
        actually supports. This corresponds with the domain argument of the
        ::socket() function.
    */
    int domain;
 
    /** \brief  Type of support for this protocol handler.
 
        This field determines which types of sockets that this protocol handler
        pays attention to. This corresponds with the type argument of the
        ::socket() function.
    */
    int type;
 
    /** \brief  Protocol of support for this protocol handler.
 
        This field determines the protocol that this protocol handler actually
        pays attention to. This corresponds with the protocol argument of the
        ::socket() function.
    */
    int protocol;
 
    /** \brief  Create a new socket for the protocol.
 
        This function must create a new socket, initializing any data that the
        protocol might need for the socket, based on the parameters passed in.
        The socket passed in is already initialized prior to the handler being
        called, and will be cleaned up by fs_socket if an error is returned from
        the handler (a return value of -1).
 
        \param  s           The socket structure to initialize
        \param  domain      Domain of the socket
        \param  type        Type of the socket
        \param protocol     Protocol of the socket
        \retval -1          On error (errno should be set appropriately)
        \retval 0           On success
    */
    int (*socket)(net_socket_t *s, int domain, int type, int protocol);
 
    /** \brief  Close a socket that was created with the protocol.
 
        This function must do any work required to close a socket and destroy
        it. This function will be called when a socket requests to be closed
        with the close system call. There are no errors defined for this
        function.
 
        \param  s           The socket to close
    */
    void (*close)(net_socket_t *hnd);
 
    /** \brief  Accept a connection on a socket created with the protocol.
 
        This function should implement the ::accept() system call for the
        protocol. The semantics are exactly as expected for that function.
 
        \param  s           The socket to accept a connection on
        \param  addr        The address of the incoming connection
        \param  alen        The length of the address
        \return             A newly created socket for the incoming connection
                            or -1 on error (with errno set appropriately)
    */
    int (*accept)(net_socket_t *s, struct sockaddr *addr, socklen_t *alen);
 
    /** \brief  Bind a socket created with the protocol to an address.
 
        This function should implement the ::bind() system call for the
        protocol. The semantics are exactly as expected for that function.
 
        \param  s           The socket to bind to the address
        \param  addr        The address to bind to
        \param  alen        The length of the address
        \retval -1          On error (set errno appropriately)
        \retval 0           On success
    */
    int (*bind)(net_socket_t *s, const struct sockaddr *addr, socklen_t alen);
 
    /** \brief  Connect a socket created with the protocol to a remote system.
 
        This function should implement the ::connect() system call for the
        protocol. The semantics are exactly as expected for that function.
 
        \param  s           The socket to connect with
        \param  addr        The address to connect to
        \param  alen        The length of the address
        \retval -1          On error (with errno set appropriately)
        \retval 0           On success
    */
    int (*connect)(net_socket_t *s, const struct sockaddr *addr,
                   socklen_t alen);
 
    /** \brief  Listen for incoming connections on a socket created with the
                protocol.
 
        This function should implement the ::listen() system call for the
        protocol. The semantics are exactly as expected for that function.
 
        \param  s           The socket to listen on
        \param  backlog     The number of connections to queue
        \retval -1          On error (with errno set appropriately)
        \retval 0           On success
    */
    int (*listen)(net_socket_t *s, int backlog);
 
    /** \brief  Receive data on a socket created with the protocol.
 
        This function should implement the ::recvfrom() system call for the
        protocol. The semantics are exactly as expected for that function. Also,
        this function should implement the ::recv() system call, which will
        call this function with NULL for addr and alen.
 
        \param  s           The socket to receive data on
        \param  buffer      The buffer to save data in
        \param  len         The length of the buffer
        \param  flags       Flags to the function
        \param  addr        Space to store the address that data came from (NULL
                            if this was called by ::recv())
        \param  alen        Space to store the length of the address (NULL if
                            this was called by ::recv())
        \retval -1          On error (set errno appropriately)
        \retval 0           No outstanding data and the peer has disconnected
                            cleanly
        \retval n           The number of bytes received (may be less than len)
    */
    ssize_t (*recvfrom)(net_socket_t *s, void *buffer, size_t len, int flags,
                        struct sockaddr *addr, socklen_t *alen);
 
    /** \brief  Send data on a socket created with the protocol.
 
        This function should implement the ::sendto() system call for the
        protocol. The semantics are exactly as expected for that function. Also,
        this function should implement the ::send() system call, which will
        call this function with NULL for addr and 0 for alen.
 
        \param  s           The socket to send data on
        \param  msg         The data to send
        \param  len         The length of data to send
        \param  flags       Flags to the function
        \param  addr        The address to send data to (NULL if this was called
                            by ::send())
        \param  alen        The length of the address (0 if this was called by
                            ::send())
        \retval -1          On error (set errno appropriately)
        \retval n           The number of bytes actually sent (may be less than
                            len)
    */
    ssize_t (*sendto)(net_socket_t *s, const void *msg, size_t len, int flags,
                      const struct sockaddr *addr, socklen_t alen);
 
    /** \brief  Shut down a socket created with the protocol.
 
        This function should implement the ::shutdown() system call for the
        protocol. The semantics are exactly as expected for that function.
 
        \param  s           The socket to shut down
        \param  how         What should be shut down on the socket
        \retval -1          On error (set errno appropriately)
        \retval 0           On success
    */
    int (*shutdownsock)(net_socket_t *s, int how);
 
    /** \brief  Input a packet into a protocol.
 
        This function should read in the packet specified by the arguments and
        sort out what exactly to do with it. This usually involves checking if
        there is an open socket with the source address and adding it to a
        packet queue if there is.
 
        \param  src         The interface the packet was input on
        \param  domain      The low-level protocol used (AF_INET or AF_INET6)
        \param  hdr         The low-level protocol header
        \param  data        The packet itself, including any protocol headers,
                            but not any from lower-level protocols
        \param  size        The size of the packet, not including any lower-
                            level protocol headers
        \retval -1          On error (the packet is discarded)
        \retval 0           On success
    */
    int (*input)(netif_t *src, int domain, const void *hdr, const uint8_t *data,
                 size_t size);
 
    /** \brief  Get socket options.
 
        This function should implement the ::getsockopt() system call for the
        given protocol. The semantics are exactly as defined for that function.
 
        Currently all options (regardless of level) are passed onto the
        protocol handler.
 
        \param  s               The socket to get options for.
        \param  level           The protocol level to get options at.
        \param  option_name     The option to look up.
        \param  option_value    Storage for the value of the option.
        \param  option_len      The length of option_value on call, and the real
                                option length (if less than the original value)
                                on return.
        \retval -1              On error (set errno appropriately).
        \retval 0               On success.
    */
    int (*getsockopt)(net_socket_t *s, int level, int option_name,
                      void *option_value, socklen_t *option_len);
 
    /** \brief  Set socket options.
 
        This function should implement the ::setsockopt() system call for the
        given protocol. The semantics are exactly as defined for that function.
 
        Currently all options (regardless of level) are passed onto the
        protocol handler.
 
        \param  s               The socket to set options for.
        \param  level           The protocol level to set options at.
        \param  option_name     The option to set.
        \param  option_value    The value to set for the option.
        \param  option_len      The length of the option_value value.
        \retval -1              On error (set errno appropriately).
        \retval 0               On success.
    */
    int (*setsockopt)(net_socket_t *s, int level, int option_name,
                      const void *option_value, socklen_t option_len);
 
    /** \brief  Get socket name.
 
        This function should implement the ::getsockname() system call for the
        given protocol. The semantics are exactly as defined for that function.
 
        Currently all options (regardless of level) are passed onto the
        protocol handler.
 
        \param  s           The socket to get the name of.
        \param  name        Pointer to a sockaddr structure which will hold
                            the resulting address information.
        \param  name_len    The amount of space pointed to by name, in
                            bytes. On return, this is set to the actual size
                            of the returned address information.
        \retval -1          On error (with errno set appropriately).
        \retval 0           On success.
    */
    int (*getsockname)(net_socket_t *s, struct sockaddr *name, socklen_t *name_len);
 
        /** \brief  Get the name of the peer connected to a socket created with the 
                    protocol.
 
        This function should implement the ::getpeername() system call for the
        protocol. The semantics are exactly as expected for that function.
 
        \param  s           The socket from which to get the peer address.
        \param  name        Pointer to a sockaddr structure which will hold
                            the resulting address information.
        \param  name_len    The amount of space pointed to by name, in
                            bytes. On return, this is set to the actual size
                            of the returned address information.
        \retval -1          On error (with errno set appropriately).
        \retval 0           On success.
    */
    int (*getpeername)(net_socket_t *s, struct sockaddr *name, socklen_t *name_len);
 
    /** \brief  Manipulate file options.
 
        This function should implement the fcntl() system call for the given
        protocol. The semantics are exactly as defined for that function.
 
        \param  s           The socket to manipulate.
        \param  cmd         The fcntl command to run.
        \param  ap          Arguments to the command.
        \retval -1          On error (generally, set errno appropriately).
    */
    int (*fcntl)(net_socket_t *s, int cmd, va_list ap);
 
    /** \brief  Poll for events.
 
        This function should check the given socket for any events that may have
        already occurred that are specified. This is used to back the ::poll()
        system call. This function should not block to wait for any events. This
        function may be called in an interrupt.
 
        \param  s           The socket to poll.
        \param  events      The events to check for.
        \retval             A mask of any of the events specified that are
                            currently true in the socket. 0 if none are true.
    */
    short (*poll)(net_socket_t *s, short events);
} fs_socket_proto_t;
 
/** \brief   Initializer for the entry field in the fs_socket_proto_t struct. 
    \ingroup vfs_sockets
*/
#define FS_SOCKET_PROTO_ENTRY { NULL, NULL }
 
/* \cond */
/* Init/shutdown */
int fs_socket_init(void);
int fs_socket_shutdown(void);
/* \endcond */
 
/** \brief   Open a socket without calling the protocol initializer.
    \ingroup vfs_sockets
 
    This function creates a new socket, but does not call the protocol's
    socket() function. This is meant to be used for things like accepting an
    incoming connection, where calling the regular socket initializer could
    cause issues. You shouldn't really have any need to call this function
    unless you are implementing a new protocol handler.
 
    \param  proto       The protocol to use for the socket.
 
    \return             The newly created socket on success, NULL on failure.
 
    \par    Error Conditions:
    \em     EWOULDBLOCK - if the function would block in an IRQ \n
    \em     ENOMEM - out of memory \n
    \em     EMFILE - too many files open
*/
net_socket_t *fs_socket_open_sock(fs_socket_proto_t *proto);
 
/** \defgroup sock_flags                Flags
    \brief                              Flags for Socket VFS
    \ingroup  vfs_sockets
 
    These are the available flags defined for sockets.
 
    Every flag after FS_SOCKET_FAM_MAX is for internal-use only, and should
    never be passed into any functions.
    @{
*/
#define FS_SOCKET_NONBLOCK  0x00000001  /** \brief Non-blocking operations */
#define FS_SOCKET_V6ONLY    0x00000002  /** \brief IPv6 Only */
 
#define FS_SOCKET_GEN_MAX   0x00008000  /** \brief Maximum generic flag */
#define FS_SOCKET_FAM_MAX   0x00800000  /** \brief Maximum family flag */
/** @} */
 
/** \brief   Input a packet into some socket family handler.
    \ingroup vfs_sockets
 
    This function is used by the lower-level network protocol handlers to input
    packets for further processing by upper-level protocols. This will call the
    input function on the family handler, if one is found.
 
    \param  src         The network interface the packet came in on
    \param  domain      The low-level protocol used (AF_INET or AF_INET6)
    \param  protocol    The upper-level protocol that we're looking for
    \param  hdr         The low-level protocol header
    \param  data        The upper-level packet, without any lower-level protocol
                        headers, but with the upper-level ones intact
    \param  size        The size of the packet (the data parameter)
 
    \retval -2          The protocol is not known
    \retval -1          Protocol-level error processing packet
    \retval 0           On success
*/
int fs_socket_input(netif_t *src, int domain, int protocol, const void *hdr,
                    const uint8_t *data, size_t size);
 
/** \brief   Add a new protocol for use with fs_socket.
    \ingroup vfs_sockets
 
    This function registers a protocol handler with fs_socket for use when
    creating and using sockets. This protocol handler must implement all of the
    functions in the fs_socket_proto_t structure. See the code in
    kos/kernel/net/net_udp.c for an example of how to do this.
 
    \warning
    This function is NOT safe to call inside an interrupt.
 
    \param  proto       The new protocol handler to register
    
    \retval 0           On success (no error conditions are currently defined)
*/
int fs_socket_proto_add(fs_socket_proto_t *proto);
 
/** \brief   Unregister a protocol from fs_socket.
    \ingroup vfs_sockets
 
    This function does the exact opposite of fs_socket_proto_add, and removes
    a protocol from use with fs_socket. 
 
    \note
    It is the programmer's responsibility to make sure that no sockets are
    still around that are registered with the protocol to be removed (as
    they will not work properly once the handler has been removed).
 
    \param  proto       The protocol handler to remove
    
    \retval -1          On error (This function does not directly change errno)
    \retval 0           On success
*/
int fs_socket_proto_remove(fs_socket_proto_t *proto);
 
__END_DECLS
 
#endif  /* __KOS_FS_SOCKET_H */