kpi_socketfilter.h |
Includes: |
<sys/kernel_types.h> <sys/kpi_socket.h> |
This header defines an API for intercepting communications at the socket layer.
For the most part, socket filters want to do three things: Filter data in and out, watch for state changes, and intercept a few calls for security. The number of function pointers supplied by a socket filter has been significantly reduced. The filter no longer has any knowledge of socket buffers. The filter no longer intercepts nearly every internal socket call. There are two data filters, an in filter, and an out filter. The in filter occurs before data is placed in the receive socket buffer. This is done to avoid waking the process unnecessarily. The out filter occurs before the data is appended to the send socket buffer. This should cover inbound and outbound data. For monitoring state changes, we've added a notify function that will be called when various events that the filter can not intercept occur. In addition, we've added a few functions that a filter may use to intercept common operations. These functions are: connect (inbound), connect (outbound), bind, set socket option, get socket option, and listen. Bind, listen, connect in, and connect out could be used together to build a fairly comprehensive firewall without having to do much with individual packets.
sflt_attach |
extern errno_t sflt_attach( socket_t so, sflt_handle);
socket
The socket the filter should be attached to.
handle
The handle of the registered filter to be attached.
0 on success otherwise the errno error.
Attaches a socket filter to the specified socket. A filter must be registered before it can be attached.
sflt_detach |
extern errno_t sflt_detach( socket_t so, sflt_handle);
socket
The socket the filter should be detached from.
handle
The handle of the registered filter to be detached.
0 on success otherwise the errno error.
Detaches a socket filter from a specified socket.
sflt_register |
extern errno_t sflt_register( const struct sflt_filter *filter, int domain, int type, int protocol);
filter
A structure describing the filter.
domain
The protocol domain these filters will be attached to.
type
The socket type these filters will be attached to.
protocol
The protocol these filters will be attached to.
0 on success otherwise the errno error.
Registers a socket filter. See 'man 2 socket' for a desciption of domain, type, and protocol.
sflt_unregister |
extern errno_t sflt_unregister( sflt_handle handle);
handle
The sf_handle of the socket filter to unregister.
0 on success otherwise the errno error.
Unregisters a socket filter. This will not detach the socket filter from all sockets it may be attached to at the time, it will just prevent the socket filter from being attached to any new sockets.
sock_inject_data_in |
extern errno_t sock_inject_data_in( socket_t so, const struct sockaddr *from, mbuf_t data, mbuf_t control, sflt_data_flag_t flags);
so
The socket to inject the data on.
from
The address the data is from, only necessary on un-connected sockets. A copy of the address will be made, caller is responsible for freeing the address after calling this function.
data
The data and possibly control mbufs.
control
The separate control mbufs.
flags
Flags indicating the type of data.
0 on success otherwise the errno error. If the function returns an error, the caller is responsible for freeing the mbuf.
Inject data in to the receive buffer of the socket as if it had come from the network.
sock_inject_data_out |
extern errno_t sock_inject_data_out( socket_t so, const struct sockaddr *to, mbuf_t data, mbuf_t control, sflt_data_flag_t flags);
so
The socket to inject the data on.
to
The address the data should be sent to, only necessary on un-connected sockets. The caller is responsible for freeing the to address after sock_inject_data_out returns.
data
The data and possibly control mbufs.
control
The separate control mbufs.
flags
Flags indicating the type of data.
0 on success otherwise the errno error. The data and control values are always freed regardless of return value.
Inject data in to the send buffer of the socket as if it had come from the client.
sockopt_copyin |
extern errno_t sockopt_copyin( sockopt_t sopt, void *data, size_t length);
sopt
The socket option.
data
A pointer to the buffer to copy the data in to.
length
The number of bytes to copy.
An errno error or zero upon success.
Copies the data from the socket option to a buffer.
sockopt_copyout |
extern errno_t sockopt_copyout( sockopt_t sopt, void *data, size_t length);
sopt
The socket option.
data
A pointer to the buffer to copy the data out of.
length
The number of bytes to copy.
An errno error or zero upon success.
Copies the data from a buffer to a socket option.
sockopt_direction |
extern sockopt_dir sockopt_direction( sockopt_t sopt);
sopt
The socket option.
sock_opt_get or sock_opt_set.
Retrieves the direction of the socket option (Get or Set).
sockopt_level |
extern int sockopt_level( sockopt_t sopt);
sopt
The socket option.
The socket option level. See man 2 setsockopt
Retrieves the socket option level. (SOL_SOCKET, etc).
sockopt_name |
extern int sockopt_name( sockopt_t sopt);
sopt
The socket option.
The socket option name. See man 2 setsockopt
Retrieves the socket option name. (SO_SNDBUF, etc).
sockopt_valsize |
extern size_t sockopt_valsize( sockopt_t sopt);
sopt
The socket option.
The length, in bytes, of the data.
Retrieves the size of the socket option data.
A 4 byte identifier used with the SO_NKE socket option to identify the socket filter to be attached.
sf_accept_func |
typedef errno_t ( *sf_accept_func)( void *cookie, socket_t so_listen, socket_t so, const struct sockaddr *local, const struct sockaddr *remote);
cookie
Cookie value specified when the filter attach was called.
so_listen
The listening socket.
so
The socket that is about to be accepted.
local
The local address of the about to be accepted socket.
remote
The remote address of the about to be accepted socket.
Return: 0 - The caller will continue with normal processing of accept. EJUSTRETURN - The to be accepted socket will be disconnected prior to being returned to the caller of accept. No further control or data operations on the socket will be allowed. This is the recommended return value as it has the least amount of impact, especially to applications which don't check the error value returned by accept. Anything Else - The to be accepted socket will be closed and the error will be returned to the caller of accept. Note that socket filter developers are advised to exercise caution when returning non-zero values to the caller, since some applications don't check the error value returned by accept and therefore risk breakage.
sf_accept_func is called after a socket is dequeued off the completed (incoming) connection list and before the file descriptor is associated with it. A filter can utilize this callback to intercept the accepted socket in order to examine it, prior to returning the socket to the caller of accept. Such a filter may also choose to discard the accepted socket if it wishes to do so.
sf_attach_func |
typedef errno_t ( *sf_attach_func)( void **cookie, socket_t so);
cookie
Used to allow the socket filter to set the cookie for this attachment.
so
The socket the filter is being attached to.
If you return a non-zero value, your filter will not be attached to this socket.
sf_attach_func is called to notify the filter it has been attached to a socket. The filter may allocate memory for this attachment and use the cookie to track it. This filter is called in one of two cases:
You've installed a global filter and a new socket was created.
Your non-global socket filter is being attached using the SO_NKE socket option.
sf_bind_func |
typedef errno_t ( *sf_bind_func)( void *cookie, socket_t so, const struct sockaddr *to);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
to
The local address of the socket will be bound to.
Return: 0 - The caller will continue with normal processing of the bind. Anything Else - The caller will rejecting the bind.
sf_bind_func is called before performing a bind operation on a socket.
sf_connect_in_func |
typedef errno_t ( *sf_connect_in_func)( void *cookie, socket_t so, const struct sockaddr *from);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
from
The address the incoming connection is from.
Return: 0 - The caller will continue with normal processing of the connection. Anything Else - The caller will rejecting the incoming connection.
sf_connect_in_func is called to filter inbound connections. A protocol will call this before accepting an incoming connection and placing it on the queue of completed connections. Warning: This filter is on the data path. Do not spend excesive time. Do not wait for data on another socket.
sf_connect_out_func |
typedef errno_t ( *sf_connect_out_func)( void *cookie, socket_t so, const struct sockaddr *to);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
to
The remote address of the outbound connection.
Return: 0 - The caller will continue with normal processing of the connection. Anything Else - The caller will rejecting the outbound connection.
sf_connect_out_func is called to filter outbound connections. A protocol will call this before initiating an outbound connection.
sf_data_in_func |
typedef errno_t ( *sf_data_in_func)( void *cookie, socket_t so, const struct sockaddr *from, mbuf_t *data, mbuf_t *control, sflt_data_flag_t flags);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
from
The addres the data is from, may be NULL if the socket is connected.
data
The data being received. Control data may appear in the mbuf chain, be sure to check the mbuf types to find control data.
control
Control data being passed separately from the data.
flags
Flags to indicate if this is out of band data or a record.
Return: 0 - The caller will continue with normal processing of the data. EJUSTRETURN - The caller will stop processing the data, the data will not be freed. Anything Else - The caller will free the data and stop processing.
sf_data_in_func is called to filter incoming data. If your filter intercepts data for later reinjection, it must queue all incoming data to preserve the order of the data. Use sock_inject_data_in to later reinject this data if you return EJUSTRETURN. Warning: This filter is on the data path. Do not spend excesive time. Do not wait for data on another socket.
sf_data_out_func |
typedef errno_t ( *sf_data_out_func)( void *cookie, socket_t so, const struct sockaddr *to, mbuf_t *data, mbuf_t *control, sflt_data_flag_t flags);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
from
The address the data is from, may be NULL if the socket is connected.
data
The data being received. Control data may appear in the mbuf chain, be sure to check the mbuf types to find control data.
control
Control data being passed separately from the data.
flags
Flags to indicate if this is out of band data or a record.
Return: 0 - The caller will continue with normal processing of the data. EJUSTRETURN - The caller will stop processing the data, the data will not be freed. Anything Else - The caller will free the data and stop processing.
sf_data_out_func is called to filter outbound data. If your filter intercepts data for later reinjection, it must queue all outbound data to preserve the order of the data when reinjecting. Use sock_inject_data_out to later reinject this data.
sf_detach_func |
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
If you return a non-zero value, your filter will not be attached to this socket.
sf_detach_func is called to notify the filter it has been detached from a socket. If the filter allocated any memory for this attachment, it should be freed. This function will be called when the socket is disposed of.
sf_getoption_func |
typedef errno_t ( *sf_getoption_func)( void *cookie, socket_t so, sockopt_t opt);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
opt
The socket option to get.
Return: 0 - The caller will continue with normal processing of the getsockopt. Anything Else - The caller will stop processing and return this error.
sf_getoption_func is called before performing getsockopt on a socket.
sf_getpeername_func |
typedef int ( *sf_getpeername_func)( void *cookie, socket_t so, struct sockaddr **sa);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
sa
A pointer to a socket address pointer.
If you return a non-zero value, processing will stop. If you return EJUSTRETURN, no further filters will be called but a result of zero will be returned to the caller of getpeername.
sf_getpeername_func is called to allow a filter to to intercept the getpeername function. When called, sa will point to a pointer to a socket address that was malloced in zone M_SONAME. If you want to replace this address, either modify the currenty copy or allocate a new one and free the old one.
sf_getsockname_func |
typedef int ( *sf_getsockname_func)( void *cookie, socket_t so, struct sockaddr **sa);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
sa
A pointer to a socket address pointer.
If you return a non-zero value, processing will stop. If you return EJUSTRETURN, no further filters will be called but a result of zero will be returned to the caller of getsockname.
sf_getsockname_func is called to allow a filter to to intercept the getsockname function. When called, sa will point to a pointer to a socket address that was malloced in zone M_SONAME. If you want to replace this address, either modify the currenty copy or allocate a new one and free the old one.
sf_ioctl_func |
typedef errno_t ( *sf_ioctl_func)( void *cookie, socket_t so, unsigned long request, const char *argp);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
request
The ioctl name.
argp
A pointer to the ioctl parameter.
Return: 0 - The caller will continue with normal processing of this ioctl. Anything Else - The caller will stop processing and return this error.
sf_ioctl_func is called before performing an ioctl on a socket.
All undefined ioctls are reserved for future use by Apple. If you need to communicate with your kext using an ioctl, please use SIOCSIFKPI and SIOCGIFKPI.
sf_listen_func |
typedef errno_t ( *sf_listen_func)( void *cookie, socket_t so);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
Return: 0 - The caller will continue with normal processing of listen. Anything Else - The caller will stop processing and return this error.
sf_listen_func is called before performing listen on a socket.
sf_notify_func |
typedef void ( *sf_notify_func)( void *cookie, socket_t so, sflt_event_t event, void *param);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
event
The type of event that has occurred.
param
Additional information about the event.
sf_notify_func is called to notify the filter of various state changes and other events occuring on the socket.
sf_setoption_func |
typedef errno_t ( *sf_setoption_func)( void *cookie, socket_t so, sockopt_t opt);
cookie
Cookie value specified when the filter attach was called.
so
The socket the filter is attached to.
opt
The socket option to set.
Return: 0 - The caller will continue with normal processing of the setsockopt. Anything Else - The caller will stop processing and return this error.
sf_setoption_func is called before performing setsockopt on a socket.
sf_unregistered_func |
typedef void ( *sf_unregistered_func)( sflt_handle handle);
handle
The socket filter handle used to identify this filter.
sf_unregistered_func is called to notify the filter it has been unregistered. This is the last function the stack will call and this function will only be called once all other function calls in to your filter have completed. Once this function has been called, your kext may safely unload.
sflt_handle |
A 4 byte identifier used with the SO_NKE socket option to identify the socket filter to be attached.
typedef u_int32_t sflt_handle;
sflt_filter |
struct sflt_filter { sflt_handle sf_handle; int sf_flags; char *sf_name; sf_unregistered_func sf_unregistered; sf_attach_func sf_attach; sf_detach_func sf_detach; sf_notify_func sf_notify; sf_getpeername_func sf_getpeername; sf_getsockname_func sf_getsockname; sf_data_in_func sf_data_in; sf_data_out_func sf_data_out; sf_connect_in_func sf_connect_in; sf_connect_out_func sf_connect_out; sf_bind_func sf_bind; sf_setoption_func sf_setoption; sf_getoption_func sf_getoption; sf_listen_func sf_listen; sf_ioctl_func sf_ioctl; /* * The following are valid only if SFLT_EXTENDED flag is set. * Initialize sf_ext_len to sizeof sflt_filter_ext structure. * Filters must also initialize reserved fields with zeroes. */ struct sflt_filter_ext { unsigned int sf_ext_len; sf_accept_func sf_ext_accept; void *sf_ext_rsvd[5]; /* Reserved */ } sf_ext; #define sf_len sf_ext.sf_ext_len #define sf_accept sf_ext.sf_ext_accept };
sf_handle
A value used to find socket filters by applications. An application can use this value to specify that this filter should be attached when using the SO_NKE socket option.
sf_flags
Indicate whether this filter should be attached to all new sockets or just those that request the filter be attached using the SO_NKE socket option. If this filter utilizes the socket filter extension fields, it must also set SFLT_EXTENDED.
sf_name
A name used for debug purposes.
sf_unregistered
Your function for being notified when your filter has been unregistered.
sf_attach
Your function for handling attaches to sockets.
sf_detach
Your function for handling detaches from sockets.
sf_notify
Your function for handling events. May be null.
sf_data_in
Your function for handling incoming data. May be null.
sf_data_out
Your function for handling outgoing data. May be null.
sf_connect_in
Your function for handling inbound connections. May be null.
sf_connect_out
Your function for handling outbound connections. May be null.
sf_bind
Your function for handling binds. May be null.
sf_setoption
Your function for handling setsockopt. May be null.
sf_getoption
Your function for handling getsockopt. May be null.
sf_listen
Your function for handling listen. May be null.
sf_ioctl
Your function for handling ioctls. May be null.
sf_len
Length of socket filter extension structure; developers must initialize this to sizeof sflt_filter_ext structure. This field and all fields following it will only be valid if SFLT_EXTENDED flag is set in sf_flags field.
sf_ext_accept
Your function for handling inbound connections at accept time. May be null.
sf_ext_rsvd
Reserved for future use; you must initialize the reserved fields with zeroes.
This structure is used to define a socket filter.
Inbound and outbound data filters may handle many different types of incoming and outgoing data. These flags help distinguish between normal data, out-of-band data, and records.
Events notify a filter of state changes and other various events related to the socket. These events cannot be prevented or intercepted, only observed.
Constants defining mbuf flags. Only the flags listed below can be set or retrieved.
sflt_data_flag_t |
Inbound and outbound data filters may handle many different types of incoming and outgoing data. These flags help distinguish between normal data, out-of-band data, and records.
enum { sock_data_filt_flag_oob = 1, sock_data_filt_flag_record = 2 };
sock_data_filt_flag_oob
Indicates this data is out-of-band data.
sock_data_filt_flag_record
Indicates this data is a record. This flag is only ever seen on inbound data.
sflt_event_t |
Events notify a filter of state changes and other various events related to the socket. These events cannot be prevented or intercepted, only observed.
enum { sock_evt_connecting = 1, sock_evt_connected = 2, sock_evt_disconnecting = 3, sock_evt_disconnected = 4, sock_evt_flush_read = 5, sock_evt_shutdown = 6, /* param points to an integer specifying how (read, write, or both) see man 2 shutdown */ sock_evt_cantrecvmore = 7, sock_evt_cantsendmore = 8, sock_evt_closing = 9, sock_evt_bound = 10 };
sock_evt_connected
Indicates this socket has moved to the connected state.
sock_evt_disconnected
Indicates this socket has moved to the disconnected state.
sock_evt_flush_read
The read socket buffer has been flushed.
sock_evt_shutdown
The read and or write side(s) of the connection have been shutdown. The param will point to an integer that indicates the direction that has been shutdown. See 'man 2 shutdown' for more information.
sock_evt_cantrecvmore
Indicates the socket cannot receive more data.
sock_evt_cantsendmore
Indicates the socket cannot send more data.
sock_evt_closing
Indicates the socket is closing.
sock_evt_bound
Indicates this socket has moved to the bound state (only for PF_INET/PF_INET6 domain).
sflt_flags |
Constants defining mbuf flags. Only the flags listed below can be set or retrieved.
enum { SFLT_GLOBAL = 0x01, SFLT_PROG = 0x02, SFLT_EXTENDED = 0x04 };
SFLT_GLOBAL
Indicates this socket filter should be attached to all new sockets when they're created.
SFLT_PROG
Indicates this socket filter should be attached only when request by the application using the SO_NKE socket option.
SFLT_EXTENDED
Indicates that this socket filter utilizes the extended fields within the sflt_filter structure.
Last Updated: 2010-07-29