mirror of
https://gitlab.nic.cz/labs/bird.git
synced 2024-11-10 05:08:42 +00:00
adc2fd4825
Support for active UNIX sockets is added. UNIX socket are now created with sk_open. The socket name/path is passes in host, the same way as SSH address. For passive UNIX socket the filesystem entry is considered as part of the resource and hence is unlinked in rfree.
190 lines
7.7 KiB
C
190 lines
7.7 KiB
C
/*
|
|
* BIRD Socket Interface
|
|
*
|
|
* (c) 1998--2004 Martin Mares <mj@ucw.cz>
|
|
*
|
|
* Can be freely distributed and used under the terms of the GNU GPL.
|
|
*/
|
|
|
|
#ifndef _BIRD_SOCKET_H_
|
|
#define _BIRD_SOCKET_H_
|
|
|
|
#include <errno.h>
|
|
|
|
#include "lib/resource.h"
|
|
#ifdef HAVE_LIBSSH
|
|
#define LIBSSH_LEGACY_0_4
|
|
#include <libssh/libssh.h>
|
|
#endif
|
|
|
|
#ifdef HAVE_LIBSSH
|
|
struct ssh_sock {
|
|
const char *username; /* (Required) SSH user name */
|
|
const char *server_hostkey_path; /* (Optional) Filepath to the SSH public key of remote side, can be knownhost file */
|
|
const char *client_privkey_path; /* (Optional) Filepath to the SSH private key of BIRD */
|
|
const char *subsystem; /* (Optional) Name of SSH subsytem */
|
|
ssh_session session; /* Internal */
|
|
ssh_channel channel; /* Internal */
|
|
int state; /* Internal */
|
|
#define SK_SSH_CONNECT 0 /* Start state */
|
|
#define SK_SSH_SERVER_KNOWN 1 /* Internal */
|
|
#define SK_SSH_USERAUTH 2 /* Internal */
|
|
#define SK_SSH_CHANNEL 3 /* Internal */
|
|
#define SK_SSH_SESSION 4 /* Internal */
|
|
#define SK_SSH_SUBSYSTEM 5 /* Internal */
|
|
#define SK_SSH_ESTABLISHED 6 /* Final state */
|
|
};
|
|
#endif
|
|
|
|
typedef struct birdsock {
|
|
resource r;
|
|
pool *pool; /* Pool where incoming connections should be allocated (for SK_xxx_PASSIVE) */
|
|
int type; /* Socket type */
|
|
int subtype; /* Socket subtype */
|
|
void *data; /* User data */
|
|
ip_addr saddr, daddr; /* IPA_NONE = unspecified */
|
|
const char *host; /* Alternative to daddr especially for UNIX sockets, NULL = unspecified */
|
|
uint sport, dport; /* 0 = unspecified (for IP: protocol type) */
|
|
int tos; /* TOS / traffic class, -1 = default */
|
|
int priority; /* Local socket priority, -1 = default */
|
|
int ttl; /* Time To Live, -1 = default */
|
|
u32 flags;
|
|
struct iface *iface; /* Interface; specify this for broad/multicast sockets */
|
|
struct iface *vrf; /* Related VRF instance, NULL if global */
|
|
|
|
byte *rbuf, *rpos; /* NULL=allocate automatically */
|
|
uint fast_rx; /* RX has higher priority in event loop */
|
|
uint rbsize;
|
|
int (*rx_hook)(struct birdsock *, uint size); /* NULL=receiving turned off, returns 1 to clear rx buffer */
|
|
|
|
byte *tbuf, *tpos; /* NULL=allocate automatically */
|
|
byte *ttx; /* Internal */
|
|
uint tbsize;
|
|
void (*tx_hook)(struct birdsock *);
|
|
|
|
void (*err_hook)(struct birdsock *, int); /* errno or zero if EOF */
|
|
|
|
/* Information about received datagrams (UDP, RAW), valid in rx_hook */
|
|
ip_addr faddr, laddr; /* src (From) and dst (Local) address of the datagram */
|
|
uint fport; /* src port of the datagram */
|
|
uint lifindex; /* local interface that received the datagram */
|
|
/* laddr and lifindex are valid only if SKF_LADDR_RX flag is set to request it */
|
|
|
|
int af; /* System-dependend adress family (e.g. AF_INET) */
|
|
int fd; /* System-dependent data */
|
|
int index; /* Index in poll buffer */
|
|
int rcv_ttl; /* TTL of last received datagram */
|
|
node n;
|
|
void *rbuf_alloc, *tbuf_alloc;
|
|
const char *password; /* Password for MD5 authentication */
|
|
const char *err; /* Error message */
|
|
struct ssh_sock *ssh; /* Used in SK_SSH */
|
|
} sock;
|
|
|
|
sock *sock_new(pool *); /* Allocate new socket */
|
|
#define sk_new(X) sock_new(X) /* Wrapper to avoid name collision with OpenSSL */
|
|
|
|
int sk_open(sock *); /* Open socket */
|
|
int sk_rx_ready(sock *s);
|
|
int sk_send(sock *, uint len); /* Send data, <0=err, >0=ok, 0=sleep */
|
|
int sk_send_to(sock *, uint len, ip_addr to, uint port); /* sk_send to given destination */
|
|
void sk_reallocate(sock *); /* Free and allocate tbuf & rbuf */
|
|
void sk_set_rbsize(sock *s, uint val); /* Resize RX buffer */
|
|
void sk_set_tbsize(sock *s, uint val); /* Resize TX buffer, keeping content */
|
|
void sk_set_tbuf(sock *s, void *tbuf); /* Switch TX buffer, NULL-> return to internal */
|
|
void sk_dump_all(void);
|
|
|
|
int sk_is_ipv4(sock *s); /* True if socket is IPv4 */
|
|
int sk_is_ipv6(sock *s); /* True if socket is IPv6 */
|
|
int sk_is_unix(sock *s); /* True if socket is UNIX socket */
|
|
|
|
static inline int sk_tx_buffer_empty(sock *sk)
|
|
{ return sk->tbuf == sk->tpos; }
|
|
|
|
int sk_setup_multicast(sock *s); /* Prepare UDP or IP socket for multicasting */
|
|
int sk_join_group(sock *s, ip_addr maddr); /* Join multicast group on sk iface */
|
|
int sk_leave_group(sock *s, ip_addr maddr); /* Leave multicast group on sk iface */
|
|
int sk_setup_broadcast(sock *s);
|
|
int sk_set_ttl(sock *s, int ttl); /* Set transmit TTL for given socket */
|
|
int sk_set_min_ttl(sock *s, int ttl); /* Set minimal accepted TTL for given socket */
|
|
int sk_set_md5_auth(sock *s, ip_addr local, ip_addr remote, int pxlen, struct iface *ifa, const char *passwd, int setkey);
|
|
int sk_set_ipv6_checksum(sock *s, int offset);
|
|
int sk_set_icmp6_filter(sock *s, int p1, int p2);
|
|
int sk_check_unix(const char *path);
|
|
void sk_log_error(sock *s, const char *p);
|
|
|
|
byte * sk_rx_buffer(sock *s, int *len); /* Temporary */
|
|
|
|
extern int sk_priority_control; /* Suggested priority for control traffic, should be sysdep define */
|
|
|
|
|
|
/* Socket flags */
|
|
|
|
#define SKF_V6ONLY 0x02 /* Use IPV6_V6ONLY socket option */
|
|
#define SKF_LADDR_RX 0x04 /* Report local address for RX packets */
|
|
#define SKF_TTL_RX 0x08 /* Report TTL / Hop Limit for RX packets */
|
|
#define SKF_BIND 0x10 /* Bind datagram socket to given source address */
|
|
#define SKF_HIGH_PORT 0x20 /* Choose port from high range if possible */
|
|
#define SKF_FREEBIND 0x40 /* Allow socket to bind to a nonlocal address */
|
|
#define SKF_CONNECT 0x80 /* Connect datagram socket to given dst address/port */
|
|
|
|
#define SKF_THREAD 0x100 /* Socked used in thread, Do not add to main loop */
|
|
#define SKF_TRUNCATED 0x200 /* Received packet was truncated, set by IO layer */
|
|
#define SKF_HDRINCL 0x400 /* Used internally */
|
|
#define SKF_PKTINFO 0x800 /* Used internally */
|
|
|
|
#define SKF_UDP6_NO_CSUM_RX 0x1000 /* Accept zero checksums for received UDPv6 packets */
|
|
|
|
/*
|
|
* Socket types SA SP DA DP IF TTL SendTo (?=may, -=must not, *=must)
|
|
*/
|
|
|
|
#define SK_TCP_PASSIVE 0 /* ? * - - - ? - */
|
|
#define SK_TCP_ACTIVE 1 /* ? ? * * - ? - */
|
|
#define SK_TCP 2
|
|
#define SK_UDP 3 /* ? ? ? ? ? ? ? */
|
|
#define SK_IP 5 /* ? - ? * ? ? ? */
|
|
#define SK_MAGIC 7 /* Internal use by sysdep code */
|
|
#define SK_UNIX_PASSIVE 8 /* - - * - - - - DA = host */
|
|
#define SK_UNIX_ACTIVE 9 /* - - * - - - - DA = host */
|
|
#define SK_UNIX 10
|
|
#define SK_SSH_ACTIVE 11 /* - - * * - ? - DA = host */
|
|
#define SK_SSH 12
|
|
|
|
/*
|
|
* Socket subtypes
|
|
*/
|
|
|
|
#define SK_IPV4 1
|
|
#define SK_IPV6 2
|
|
|
|
/*
|
|
* For TCP/IP sockets, Address family (IPv4 or IPv6) can be specified either
|
|
* explicitly (SK_IPV4 or SK_IPV6) or implicitly (based on saddr, daddr). But
|
|
* these specifications must be consistent.
|
|
*
|
|
* For SK_UDP or SK_IP sockets setting DA/DP allows to use sk_send(), otherwise
|
|
* sk_send_to() must be used.
|
|
*
|
|
* For SK_IP sockets setting DP specifies protocol number, which is used for
|
|
* both receiving and sending.
|
|
*
|
|
* For multicast on SK_UDP or SK_IP sockets set IF and TTL, call
|
|
* sk_setup_multicast() to enable multicast on that socket, and then use
|
|
* sk_join_group() and sk_leave_group() to manage a set of received multicast
|
|
* groups.
|
|
*
|
|
* For datagram (SK_UDP, SK_IP) sockets, there are two ways to handle source
|
|
* address. The socket could be bound to it using bind() syscall, but that also
|
|
* forbids the reception of multicast packets, or the address could be set on
|
|
* per-packet basis using platform dependent options (but these are not
|
|
* available in some corner cases). The first way is used when SKF_BIND is
|
|
* specified, the second way is used otherwise.
|
|
*
|
|
* For UNIX sockets (SK_UNIX_PASSIVE, SK_UNIX_ACTIVE), path is passed in host
|
|
* string. If the path does not fit into the sockaddr_un sun_path array, the
|
|
* sk_open will fail. One can check path length with sk_check_unix.
|
|
*/
|
|
|
|
#endif
|