xref: /illumos-gate/usr/src/uts/common/sys/socketvar.h (revision 445f2479fe3d7435daab18bf2cdc310b86cd6738)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 /*	Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T	*/
28 /*	  All Rights Reserved  	*/
29 
30 /*
31  * University Copyright- Copyright (c) 1982, 1986, 1988
32  * The Regents of the University of California
33  * All Rights Reserved
34  *
35  * University Acknowledgment- Portions of this document are derived from
36  * software developed by the University of California, Berkeley, and its
37  * contributors.
38  */
39 
40 #ifndef _SYS_SOCKETVAR_H
41 #define	_SYS_SOCKETVAR_H
42 
43 #pragma ident	"%Z%%M%	%I%	%E% SMI"
44 
45 #include <sys/types.h>
46 #include <sys/stream.h>
47 #include <sys/t_lock.h>
48 #include <sys/cred.h>
49 #include <sys/vnode.h>
50 #include <sys/file.h>
51 #include <sys/param.h>
52 #include <sys/zone.h>
53 #include <inet/kssl/ksslapi.h>
54 
55 #ifdef	__cplusplus
56 extern "C" {
57 #endif
58 
59 
60 
61 /*
62  * Internal representation used for addresses.
63  */
64 struct soaddr {
65 	struct sockaddr	*soa_sa;	/* Actual address */
66 	t_uscalar_t	soa_len;	/* Length in bytes for kmem_free */
67 	t_uscalar_t	soa_maxlen;	/* Allocated length */
68 };
69 /* Maximum size address for transports that have ADDR_size == 1 */
70 #define	SOA_DEFSIZE	128
71 
72 /*
73  * Internal representation of the address used to represent addresses
74  * in the loopback transport for AF_UNIX. While the sockaddr_un is used
75  * as the sockfs layer address for AF_UNIX the pathnames contained in
76  * these addresses are not unique (due to relative pathnames) thus can not
77  * be used in the transport.
78  *
79  * The transport level address consists of a magic number (used to separate the
80  * name space for specific and implicit binds). For a specific bind
81  * this is followed by a "vnode *" which ensures that all specific binds
82  * have a unique transport level address. For implicit binds the latter
83  * part of the address is a byte string (of the same length as a pointer)
84  * that is assigned by the loopback transport.
85  *
86  * The uniqueness assumes that the loopback transport has a separate namespace
87  * for sockets in order to avoid name conflicts with e.g. TLI use of the
88  * same transport.
89  */
90 struct so_ux_addr {
91 	void	*soua_vp;	/* vnode pointer or assigned by tl */
92 	uint_t	soua_magic;	/* See below */
93 };
94 
95 #define	SOU_MAGIC_EXPLICIT	0x75787670	/* "uxvp" */
96 #define	SOU_MAGIC_IMPLICIT	0x616e6f6e	/* "anon" */
97 
98 struct sockaddr_ux {
99 	sa_family_t		sou_family;	/* AF_UNIX */
100 	struct so_ux_addr	sou_addr;
101 };
102 
103 typedef struct sonodeops sonodeops_t;
104 typedef struct sonode sonode_t;
105 
106 /*
107  * The sonode represents a socket. A sonode never exist in the file system
108  * name space and can not be opened using open() - only the socket, socketpair
109  * and accept calls create sonodes.
110  *
111  * When an AF_UNIX socket is bound to a pathname the sockfs
112  * creates a VSOCK vnode in the underlying file system. However, the vnodeops
113  * etc in this VNODE remain those of the underlying file system.
114  * Sockfs uses the v_stream pointer in the underlying file system VSOCK node
115  * to find the sonode bound to the pathname. The bound pathname vnode
116  * is accessed through so_ux_vp.
117  *
118  * A socket always corresponds to a VCHR stream representing the transport
119  * provider (e.g. /dev/tcp). This information is retrieved from the kernel
120  * socket configuration table and entered into so_accessvp. sockfs uses
121  * this to perform VOP_ACCESS checks before allowing an open of the transport
122  * provider.
123  *
124  * The locking of sockfs uses the so_lock mutex plus the SOLOCKED
125  * and SOREADLOCKED flags in so_flag. The mutex protects all the state
126  * in the sonode. The SOLOCKED flag is used to single-thread operations from
127  * sockfs users to prevent e.g. multiple bind() calls to operate on the
128  * same sonode concurrently. The SOREADLOCKED flag is used to ensure that
129  * only one thread sleeps in kstrgetmsg for a given sonode. This is needed
130  * to ensure atomic operation for things like MSG_WAITALL.
131  *
132  * Note that so_lock is sometimes held across calls that might go to sleep
133  * (kmem_alloc and soallocproto*). This implies that no other lock in
134  * the system should be held when calling into sockfs; from the system call
135  * side or from strrput. If locks are held while calling into sockfs
136  * the system might hang when running low on memory.
137  */
138 struct sonode {
139 	struct	vnode	*so_vnode;	/* vnode associated with this sonode */
140 
141 	sonodeops_t	*so_ops;	/* operations vector for this sonode */
142 
143 	/*
144 	 * These fields are initialized once.
145 	 */
146 	dev_t		so_dev;		/* device the sonode represents */
147 	struct	vnode	*so_accessvp;	/* vnode for the /dev entry */
148 
149 	/* The locks themselves */
150 	kmutex_t	so_lock;	/* protects sonode fields */
151 	kmutex_t	so_plumb_lock;	/* serializes plumbs, and the related */
152 					/* fields so_version and so_pushcnt */
153 	kcondvar_t	so_state_cv;	/* synchronize state changes */
154 	kcondvar_t	so_ack_cv;	/* wait for TPI acks */
155 	kcondvar_t	so_connind_cv;	/* wait for T_CONN_IND */
156 	kcondvar_t	so_want_cv;	/* wait due to SOLOCKED */
157 
158 	/* These fields are protected by so_lock */
159 	uint_t	so_state;		/* internal state flags SS_*, below */
160 	uint_t	so_mode;		/* characteristics on socket. SM_* */
161 
162 	mblk_t	*so_ack_mp;		/* TPI ack received from below */
163 	mblk_t	*so_conn_ind_head;	/* b_next list of T_CONN_IND */
164 	mblk_t	*so_conn_ind_tail;
165 	mblk_t	*so_unbind_mp;		/* Preallocated T_UNBIND_REQ message */
166 
167 	ushort_t so_flag;		/* flags, see below */
168 	dev_t	so_fsid;		/* file system identifier */
169 	time_t  so_atime;		/* time of last access */
170 	time_t  so_mtime;		/* time of last modification */
171 	time_t  so_ctime;		/* time of last attributes change */
172 	int	so_count;		/* count of opened references */
173 
174 	/* Needed to recreate the same socket for accept */
175 	short	so_family;
176 	short	so_type;
177 	short	so_protocol;
178 	short	so_version;		/* From so_socket call */
179 	short	so_pushcnt;		/* Number of modules above "sockmod" */
180 
181 	/* Options */
182 	short	so_options;		/* From socket call, see socket.h */
183 	struct linger	so_linger;	/* SO_LINGER value */
184 	int	so_sndbuf;		/* SO_SNDBUF value */
185 	int	so_rcvbuf;		/* SO_RCVBUF value */
186 	int	so_sndlowat;		/* send low water mark */
187 	int	so_rcvlowat;		/* receive low water mark */
188 #ifdef notyet
189 	int	so_sndtimeo;		/* Not yet implemented */
190 	int	so_rcvtimeo;		/* Not yet implemented */
191 #endif /* notyet */
192 	ushort_t so_error;		/* error affecting connection */
193 	ushort_t so_delayed_error;	/* From T_uderror_ind */
194 	int	so_backlog;		/* Listen backlog */
195 
196 	/*
197 	 * The counts (so_oobcnt and so_oobsigcnt) track the number of
198 	 * urgent indicates that are (logically) queued on the stream head
199 	 * read queue. The urgent data is queued on the stream head
200 	 * as follows.
201 	 *
202 	 * In the normal case the SIGURG is not generated until
203 	 * the T_EXDATA_IND arrives at the stream head. However, transports
204 	 * that have an early indication that urgent data is pending
205 	 * (e.g. TCP receiving a "new" urgent pointer value) can send up
206 	 * an M_PCPROTO/SIGURG message to generate the signal early.
207 	 *
208 	 * The mark is indicated by either:
209 	 *  - a T_EXDATA_IND (with no M_DATA b_cont) with MSGMARK set.
210 	 *    When this message is consumed by sorecvmsg the socket layer
211 	 *    sets SS_RCVATMARK until data has been consumed past the mark.
212 	 *  - a message with MSGMARKNEXT set (indicating that the
213 	 *    first byte of the next message constitutes the mark). When
214 	 *    the last byte of the MSGMARKNEXT message is consumed in
215 	 *    the stream head the stream head sets STRATMARK. This flag
216 	 *    is cleared when at least one byte is read. (Note that
217 	 *    the MSGMARKNEXT messages can be of zero length when there
218 	 *    is no previous data to which the marknext can be attached.)
219 	 *
220 	 * While the T_EXDATA_IND method is the common case which is used
221 	 * with all TPI transports, the MSGMARKNEXT method is needed to
222 	 * indicate the mark when e.g. the TCP urgent byte has not been
223 	 * received yet but the TCP urgent pointer has made TCP generate
224 	 * the M_PCSIG/SIGURG.
225 	 *
226 	 * The signal (the M_PCSIG carrying the SIGURG) and the mark
227 	 * indication can not be delivered as a single message, since
228 	 * the signal should be delivered as high priority and any mark
229 	 * indication must flow with the data. This implies that immediately
230 	 * when the SIGURG has been delivered if the stream head queue is
231 	 * empty it is impossible to determine if this will be the position
232 	 * of the mark. This race condition is resolved by using MSGNOTMARKNEXT
233 	 * messages and the STRNOTATMARK flag in the stream head. The
234 	 * SIOCATMARK code calls the stream head to wait for either a
235 	 * non-empty queue or one of the STR*ATMARK flags being set.
236 	 * This implies that any transport that is sending M_PCSIG(SIGURG)
237 	 * should send the appropriate MSGNOTMARKNEXT message (which can be
238 	 * zero length) after sending an M_PCSIG to prevent SIOCATMARK
239 	 * from sleeping unnecessarily.
240 	 */
241 	mblk_t	*so_oobmsg;		/* outofline oob data */
242 	uint_t	so_oobsigcnt;		/* Number of SIGURG generated */
243 	uint_t	so_oobcnt;		/* Number of T_EXDATA_IND queued */
244 	pid_t	so_pgrp;		/* pgrp for signals */
245 
246 	/* From T_info_ack */
247 	t_uscalar_t	so_tsdu_size;
248 	t_uscalar_t	so_etsdu_size;
249 	t_scalar_t	so_addr_size;
250 	t_uscalar_t	so_opt_size;
251 	t_uscalar_t	so_tidu_size;
252 	t_scalar_t	so_serv_type;
253 
254 	/* From T_capability_ack */
255 	t_uscalar_t	so_acceptor_id;
256 
257 	/* Internal provider information */
258 	struct tpi_provinfo	*so_provinfo;
259 
260 	/*
261 	 * The local and remote addresses have multiple purposes
262 	 * but one of the key reasons for their existence and careful
263 	 * tracking in sockfs is to support getsockname and getpeername
264 	 * when the transport does not handle the TI_GET*NAME ioctls
265 	 * and caching when it does (signalled by valid bits in so_state).
266 	 * When all transports support the new TPI (with T_ADDR_REQ)
267 	 * we can revisit this code.
268 	 * The other usage of so_faddr is to keep the "connected to"
269 	 * address for datagram sockets.
270 	 * Finally, for AF_UNIX both local and remote addresses are used
271 	 * to record the sockaddr_un since we use a separate namespace
272 	 * in the loopback transport.
273 	 */
274 	struct soaddr so_laddr;		/* Local address */
275 	struct soaddr so_faddr;		/* Peer address */
276 #define	so_laddr_sa	so_laddr.soa_sa
277 #define	so_faddr_sa	so_faddr.soa_sa
278 #define	so_laddr_len	so_laddr.soa_len
279 #define	so_faddr_len	so_faddr.soa_len
280 #define	so_laddr_maxlen	so_laddr.soa_maxlen
281 #define	so_faddr_maxlen	so_faddr.soa_maxlen
282 	mblk_t		*so_eaddr_mp;	/* for so_delayed_error */
283 
284 	/*
285 	 * For AF_UNIX sockets:
286 	 * so_ux_laddr/faddr records the internal addresses used with the
287 	 * transport.
288 	 * so_ux_vp and v_stream->sd_vnode form the cross-
289 	 * linkage between the underlying fs vnode corresponding to
290 	 * the bound sockaddr_un and the socket node.
291 	 */
292 	struct so_ux_addr so_ux_laddr;	/* laddr bound with the transport */
293 	struct so_ux_addr so_ux_faddr;	/* temporary peer address */
294 	struct vnode	*so_ux_bound_vp; /* bound AF_UNIX file system vnode */
295 	struct sonode	*so_next;	/* next sonode on socklist	*/
296 	struct sonode	*so_prev;	/* previous sonode on socklist	*/
297 	mblk_t	*so_discon_ind_mp;	/* T_DISCON_IND received from below */
298 
299 					/* put here for delayed processing  */
300 	void		*so_priv;	/* sonode private data */
301 	cred_t		*so_peercred;	/* connected socket peer cred */
302 	pid_t		so_cpid;	/* connected socket peer cached pid */
303 	zoneid_t	so_zoneid;	/* opener's zoneid */
304 
305 	kmem_cache_t	*so_cache;	/* object cache of this "sonode". */
306 	void		*so_obj;	/* object to free */
307 
308 	/*
309 	 * For NL7C sockets:
310 	 *
311 	 * so_nl7c_flags	the NL7C state of URL processing.
312 	 *
313 	 * so_nl7c_rcv_mp	mblk_t chain of already received data to be
314 	 *			passed up to the app after NL7C gives up on
315 	 *			a socket.
316 	 *
317 	 * so_nl7c_rcv_rval	returned rval for last mblk_t from above.
318 	 *
319 	 * so_nl7c_uri		the URI currently being processed.
320 	 *
321 	 * so_nl7c_rtime	URI request gethrestime_sec().
322 	 */
323 	uint64_t	so_nl7c_flags;
324 	mblk_t		*so_nl7c_rcv_mp;
325 	int64_t		so_nl7c_rcv_rval;
326 	void		*so_nl7c_uri;
327 	time_t		so_nl7c_rtime;
328 
329 	/* For sockets acting as an in-kernel SSL proxy */
330 	kssl_endpt_type_t	so_kssl_type;	/* is proxy/is proxied/none */
331 	kssl_ent_t		so_kssl_ent;	/* SSL config entry */
332 	kssl_ctx_t		so_kssl_ctx;	/* SSL session context */
333 };
334 
335 /* flags */
336 #define	SOMOD		0x0001		/* update socket modification time */
337 #define	SOACC		0x0002		/* update socket access time */
338 
339 #define	SOLOCKED	0x0010		/* use to serialize open/closes */
340 #define	SOREADLOCKED	0x0020		/* serialize kstrgetmsg calls */
341 #define	SOWANT		0x0040		/* some process waiting on lock */
342 #define	SOCLONE		0x0080		/* child of clone driver */
343 #define	SOASYNC_UNBIND	0x0100		/* wait for ACK of async unbind */
344 
345 /*
346  * Socket state bits.
347  */
348 #define	SS_ISCONNECTED		0x00000001 /* socket connected to a peer */
349 #define	SS_ISCONNECTING		0x00000002 /* in process, connecting to peer */
350 #define	SS_ISDISCONNECTING	0x00000004 /* in process of disconnecting */
351 #define	SS_CANTSENDMORE		0x00000008 /* can't send more data to peer */
352 
353 #define	SS_CANTRCVMORE		0x00000010 /* can't receive more data */
354 #define	SS_ISBOUND		0x00000020 /* socket is bound */
355 #define	SS_NDELAY		0x00000040 /* FNDELAY non-blocking */
356 #define	SS_NONBLOCK		0x00000080 /* O_NONBLOCK non-blocking */
357 
358 #define	SS_ASYNC		0x00000100 /* async i/o notify */
359 #define	SS_ACCEPTCONN		0x00000200 /* listen done */
360 #define	SS_HASCONNIND		0x00000400 /* T_CONN_IND for poll */
361 #define	SS_SAVEDEOR		0x00000800 /* Saved MSG_EOR rcv side state */
362 
363 #define	SS_RCVATMARK		0x00001000 /* at mark on input */
364 #define	SS_OOBPEND		0x00002000 /* OOB pending or present - poll */
365 #define	SS_HAVEOOBDATA		0x00004000 /* OOB data present */
366 #define	SS_HADOOBDATA		0x00008000 /* OOB data consumed */
367 
368 #define	SS_FADDR_NOXLATE	0x00020000 /* No xlation of faddr for AF_UNIX */
369 
370 #define	SS_HASDATA		0x00040000 /* NCAfs: data available */
371 #define	SS_DONEREAD		0x00080000 /* NCAfs: all data read */
372 #define	SS_MOREDATA		0x00100000 /* NCAfs: NCA has more data */
373 
374 #define	SS_DIRECT		0x00200000 /* transport is directly below */
375 
376 #define	SS_LADDR_VALID		0x01000000	/* so_laddr valid for user */
377 #define	SS_FADDR_VALID		0x02000000	/* so_faddr valid for user */
378 
379 /* Set of states when the socket can't be rebound */
380 #define	SS_CANTREBIND	(SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING|\
381 			    SS_CANTSENDMORE|SS_CANTRCVMORE|SS_ACCEPTCONN)
382 
383 /*
384  * Characteristics of sockets. Not changed after the socket is created.
385  */
386 #define	SM_PRIV			0x001	/* privileged for broadcast, raw... */
387 #define	SM_ATOMIC		0x002	/* atomic data transmission */
388 #define	SM_ADDR			0x004	/* addresses given with messages */
389 #define	SM_CONNREQUIRED		0x008	/* connection required by protocol */
390 
391 #define	SM_FDPASSING		0x010	/* passes file descriptors */
392 #define	SM_EXDATA		0x020	/* Can handle T_EXDATA_REQ */
393 #define	SM_OPTDATA		0x040	/* Can handle T_OPTDATA_REQ */
394 #define	SM_BYTESTREAM		0x080	/* Byte stream - can use M_DATA */
395 
396 #define	SM_ACCEPTOR_ID		0x100	/* so_acceptor_id is valid */
397 
398 /*
399  * Socket versions. Used by the socket library when calling _so_socket().
400  */
401 #define	SOV_STREAM	0	/* Not a socket - just a stream */
402 #define	SOV_DEFAULT	1	/* Select based on so_default_version */
403 #define	SOV_SOCKSTREAM	2	/* Socket plus streams operations */
404 #define	SOV_SOCKBSD	3	/* Socket with no streams operations */
405 #define	SOV_XPG4_2	4	/* Xnet socket */
406 
407 #if defined(_KERNEL) || defined(_KMEMUSER)
408 /*
409  * Used for mapping family/type/protocol to vnode.
410  * Defined here so that crash can use it.
411  */
412 struct sockparams {
413 	int	sp_domain;
414 	int	sp_type;
415 	int	sp_protocol;
416 	char	*sp_devpath;
417 	int	sp_devpathlen;	/* Is 0 if sp_devpath is a static string */
418 	vnode_t	*sp_vnode;
419 	struct sockparams *sp_next;
420 };
421 
422 extern struct sockparams *sphead;
423 
424 /*
425  * Used to traverse the list of AF_UNIX sockets to construct the kstat
426  * for netstat(1m).
427  */
428 struct socklist {
429 	kmutex_t	sl_lock;
430 	struct sonode	*sl_list;
431 };
432 
433 extern struct socklist socklist;
434 /*
435  * ss_full_waits is the number of times the reader thread
436  * waits when the queue is full and ss_empty_waits is the number
437  * of times the consumer thread waits when the queue is empty.
438  * No locks for these as they are just indicators of whether
439  * disk or network or both is slow or fast.
440  */
441 struct sendfile_stats {
442 	uint32_t ss_file_cached;
443 	uint32_t ss_file_not_cached;
444 	uint32_t ss_full_waits;
445 	uint32_t ss_empty_waits;
446 	uint32_t ss_file_segmap;
447 };
448 
449 /*
450  * A single sendfile request is represented by snf_req.
451  */
452 typedef struct snf_req {
453 	struct snf_req	*sr_next;
454 	mblk_t		*sr_mp_head;
455 	mblk_t		*sr_mp_tail;
456 	kmutex_t	sr_lock;
457 	kcondvar_t	sr_cv;
458 	uint_t		sr_qlen;
459 	int		sr_hiwat;
460 	int		sr_lowat;
461 	int		sr_operation;
462 	struct vnode	*sr_vp;
463 	file_t 		*sr_fp;
464 	ssize_t		sr_maxpsz;
465 	u_offset_t	sr_file_off;
466 	u_offset_t	sr_file_size;
467 #define	SR_READ_DONE	0x80000000
468 	int		sr_read_error;
469 	int		sr_write_error;
470 } snf_req_t;
471 
472 /* A queue of sendfile requests */
473 struct sendfile_queue {
474 	snf_req_t	*snfq_req_head;
475 	snf_req_t	*snfq_req_tail;
476 	kmutex_t	snfq_lock;
477 	kcondvar_t	snfq_cv;
478 	int		snfq_svc_threads;	/* # of service threads */
479 	int		snfq_idle_cnt;		/* # of idling threads */
480 	int		snfq_max_threads;
481 	int		snfq_req_cnt;		/* Number of requests */
482 };
483 
484 #define	READ_OP			1
485 #define	SNFQ_TIMEOUT		(60 * 5 * hz)	/* 5 minutes */
486 
487 /* Socket network operations switch */
488 struct sonodeops {
489 	int	(*sop_accept)(struct sonode *, int, struct sonode **);
490 	int	(*sop_bind)(struct sonode *, struct sockaddr *, socklen_t,
491 		    int);
492 	int	(*sop_listen)(struct sonode *, int);
493 	int	(*sop_connect)(struct sonode *, const struct sockaddr *,
494 		    socklen_t, int, int);
495 	int	(*sop_recvmsg)(struct sonode *, struct msghdr *,
496 		    struct uio *);
497 	int	(*sop_sendmsg)(struct sonode *, struct msghdr *,
498 		    struct uio *);
499 	int	(*sop_getpeername)(struct sonode *);
500 	int	(*sop_getsockname)(struct sonode *);
501 	int	(*sop_shutdown)(struct sonode *, int);
502 	int	(*sop_getsockopt)(struct sonode *, int, int, void *,
503 		    socklen_t *, int);
504 	int 	(*sop_setsockopt)(struct sonode *, int, int, const void *,
505 		    socklen_t);
506 };
507 
508 #define	SOP_ACCEPT(so, fflag, nsop)	\
509 	((so)->so_ops->sop_accept((so), (fflag), (nsop)))
510 #define	SOP_BIND(so, name, namelen, flags)	\
511 	((so)->so_ops->sop_bind((so), (name), (namelen), (flags)))
512 #define	SOP_LISTEN(so, backlog)	\
513 	((so)->so_ops->sop_listen((so), (backlog)))
514 #define	SOP_CONNECT(so, name, namelen, fflag, flags)	\
515 	((so)->so_ops->sop_connect((so), (name), (namelen), (fflag), (flags)))
516 #define	SOP_RECVMSG(so, msg, uiop)	\
517 	((so)->so_ops->sop_recvmsg((so), (msg), (uiop)))
518 #define	SOP_SENDMSG(so, msg, uiop)	\
519 	((so)->so_ops->sop_sendmsg((so), (msg), (uiop)))
520 #define	SOP_GETPEERNAME(so)	\
521 	((so)->so_ops->sop_getpeername((so)))
522 #define	SOP_GETSOCKNAME(so)	\
523 	((so)->so_ops->sop_getsockname((so)))
524 #define	SOP_SHUTDOWN(so, how)	\
525 	((so)->so_ops->sop_shutdown((so), (how)))
526 #define	SOP_GETSOCKOPT(so, level, optionname, optval, optlenp, flags)	\
527 	((so)->so_ops->sop_getsockopt((so), (level), (optionname),	\
528 	    (optval), (optlenp), (flags)))
529 #define	SOP_SETSOCKOPT(so, level, optionname, optval, optlen)		\
530 	((so)->so_ops->sop_setsockopt((so), (level), (optionname),	\
531 	    (optval), (optlen)))
532 
533 #endif /* defined(_KERNEL) || defined(_KMEMUSER) */
534 
535 #ifdef _KERNEL
536 
537 #define	ISALIGNED_cmsghdr(addr) \
538 		(((uintptr_t)(addr) & (_CMSG_HDR_ALIGNMENT - 1)) == 0)
539 
540 #define	ROUNDUP_cmsglen(len) \
541 	(((len) + _CMSG_HDR_ALIGNMENT - 1) & ~(_CMSG_HDR_ALIGNMENT - 1))
542 
543 /*
544  * Used in parsing msg_control
545  */
546 #define	CMSG_NEXT(cmsg)						\
547 	(struct cmsghdr *)((uintptr_t)(cmsg) +			\
548 	    ROUNDUP_cmsglen((cmsg)->cmsg_len))
549 
550 /*
551  * Maximum size of any argument that is copied in (addresses, options,
552  * access rights). MUST be at least MAXPATHLEN + 3.
553  * BSD and SunOS 4.X limited this to MLEN or MCLBYTES.
554  */
555 #define	SO_MAXARGSIZE	8192
556 
557 /*
558  * Convert between vnode and sonode
559  */
560 #define	VTOSO(vp)	((struct sonode *)((vp)->v_data))
561 #define	SOTOV(sp)	((sp)->so_vnode)
562 
563 /*
564  * Internal flags for sobind()
565  */
566 #define	_SOBIND_REBIND		0x01	/* Bind to existing local address */
567 #define	_SOBIND_UNSPEC		0x02	/* Bind to unspecified address */
568 #define	_SOBIND_LOCK_HELD	0x04	/* so_excl_lock held by caller */
569 #define	_SOBIND_NOXLATE		0x08	/* No addr translation for AF_UNIX */
570 #define	_SOBIND_XPG4_2		0x10	/* xpg4.2 semantics */
571 #define	_SOBIND_SOCKBSD		0x20	/* BSD semantics */
572 #define	_SOBIND_LISTEN		0x40	/* Make into SS_ACCEPTCONN */
573 #define	_SOBIND_SOCKETPAIR	0x80	/* Internal flag for so_socketpair() */
574 					/* to enable listen with backlog = 1 */
575 
576 /*
577  * Internal flags for sounbind()
578  */
579 #define	_SOUNBIND_REBIND	0x01	/* Don't clear fields - will rebind */
580 
581 /*
582  * Internal flags for soconnect()
583  */
584 #define	_SOCONNECT_NOXLATE	0x01	/* No addr translation for AF_UNIX */
585 #define	_SOCONNECT_DID_BIND	0x02	/* Unbind when connect fails */
586 #define	_SOCONNECT_XPG4_2	0x04	/* xpg4.2 semantics */
587 
588 /*
589  * Internal flags for sodisconnect()
590  */
591 #define	_SODISCONNECT_LOCK_HELD	0x01	/* so_excl_lock held by caller */
592 
593 /*
594  * Internal flags for sotpi_getsockopt().
595  */
596 #define	_SOGETSOCKOPT_XPG4_2	0x01	/* xpg4.2 semantics */
597 
598 /*
599  * Internal flags for soallocproto*()
600  */
601 #define	_ALLOC_NOSLEEP		0	/* Don't sleep for memory */
602 #define	_ALLOC_INTR		1	/* Sleep until interrupt */
603 #define	_ALLOC_SLEEP		2	/* Sleep forever */
604 
605 /*
606  * Internal structure for handling AF_UNIX file descriptor passing
607  */
608 struct fdbuf {
609 	int		fd_size;	/* In bytes, for kmem_free */
610 	int		fd_numfd;	/* Number of elements below */
611 	char		*fd_ebuf;	/* Extra buffer to free  */
612 	int		fd_ebuflen;
613 	frtn_t		fd_frtn;
614 	struct file	*fd_fds[1];	/* One or more */
615 };
616 #define	FDBUF_HDRSIZE	(sizeof (struct fdbuf) - sizeof (struct file *))
617 
618 /*
619  * Variable that can be patched to set what version of socket socket()
620  * will create.
621  */
622 extern int so_default_version;
623 
624 #ifdef DEBUG
625 /* Turn on extra testing capabilities */
626 #define	SOCK_TEST
627 #endif /* DEBUG */
628 
629 #ifdef DEBUG
630 char	*pr_state(uint_t, uint_t);
631 char	*pr_addr(int, struct sockaddr *, t_uscalar_t);
632 int	so_verify_oobstate(struct sonode *);
633 #endif /* DEBUG */
634 
635 /*
636  * DEBUG macros
637  */
638 #if defined(DEBUG) && !defined(__lint)
639 #define	SOCK_DEBUG
640 
641 extern int sockdebug;
642 extern int sockprinterr;
643 
644 #define	eprint(args)	printf args
645 #define	eprintso(so, args) \
646 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) printf args; }
647 #define	eprintline(error)					\
648 {								\
649 	if (error != EINTR && (sockprinterr || sockdebug > 0))	\
650 		printf("socket error %d: line %d file %s\n",	\
651 			(error), __LINE__, __FILE__);		\
652 }
653 
654 #define	eprintsoline(so, error)					\
655 { if (sockprinterr && ((so)->so_options & SO_DEBUG))		\
656 	printf("socket(%p) error %d: line %d file %s\n",	\
657 		(so), (error), __LINE__, __FILE__);		\
658 }
659 #define	dprint(level, args)	{ if (sockdebug > (level)) printf args; }
660 #define	dprintso(so, level, args) \
661 { if (sockdebug > (level) && ((so)->so_options & SO_DEBUG)) printf args; }
662 
663 #else /* define(DEBUG) && !defined(__lint) */
664 
665 #define	eprint(args)		{}
666 #define	eprintso(so, args)	{}
667 #define	eprintline(error)	{}
668 #define	eprintsoline(so, error)	{}
669 #define	dprint(level, args)	{}
670 #define	dprintso(so, level, args) {}
671 #ifdef DEBUG
672 #undef DEBUG
673 #endif
674 
675 #endif /* defined(DEBUG) && !defined(__lint) */
676 
677 extern struct vfsops			sock_vfsops;
678 extern struct vnodeops			*socktpi_vnodeops;
679 extern const struct fs_operation_def	socktpi_vnodeops_template[];
680 
681 extern sonodeops_t			sotpi_sonodeops;
682 
683 extern dev_t				sockdev;
684 
685 /*
686  * sockfs functions
687  */
688 extern int	sock_getmsg(vnode_t *, struct strbuf *, struct strbuf *,
689 			uchar_t *, int *, int, rval_t *);
690 extern int	sock_putmsg(vnode_t *, struct strbuf *, struct strbuf *,
691 			uchar_t, int, int);
692 struct sonode	*sotpi_create(vnode_t *, int, int, int, int, struct sonode *,
693 			int *);
694 extern int	socktpi_open(struct vnode **, int, struct cred *);
695 extern int	so_sock2stream(struct sonode *);
696 extern void	so_stream2sock(struct sonode *);
697 extern int	sockinit(int, char *);
698 extern struct vnode
699 		*makesockvp(struct vnode *, int, int, int);
700 extern void	sockfree(struct sonode *);
701 extern void	so_update_attrs(struct sonode *, int);
702 extern int	soconfig(int, int, int,	char *, int);
703 extern struct vnode
704 		*solookup(int, int, int, char *, int *);
705 extern void	so_lock_single(struct sonode *);
706 extern void	so_unlock_single(struct sonode *, int);
707 extern int	so_lock_read(struct sonode *, int);
708 extern int	so_lock_read_intr(struct sonode *, int);
709 extern void	so_unlock_read(struct sonode *);
710 extern void	*sogetoff(mblk_t *, t_uscalar_t, t_uscalar_t, uint_t);
711 extern void	so_getopt_srcaddr(void *, t_uscalar_t,
712 			void **, t_uscalar_t *);
713 extern int	so_getopt_unix_close(void *, t_uscalar_t);
714 extern int	so_addr_verify(struct sonode *, const struct sockaddr *,
715 			socklen_t);
716 extern int	so_ux_addr_xlate(struct sonode *, struct sockaddr *,
717 			socklen_t, int, void **, socklen_t *);
718 extern void	fdbuf_free(struct fdbuf *);
719 extern mblk_t	*fdbuf_allocmsg(int, struct fdbuf *);
720 extern int	fdbuf_create(void *, int, struct fdbuf **);
721 extern void	so_closefds(void *, t_uscalar_t, int, int);
722 extern int	so_getfdopt(void *, t_uscalar_t, int, void **, int *);
723 t_uscalar_t	so_optlen(void *, t_uscalar_t, int);
724 extern void	so_cmsg2opt(void *, t_uscalar_t, int, mblk_t *);
725 extern t_uscalar_t
726 		so_cmsglen(mblk_t *, void *, t_uscalar_t, int);
727 extern int	so_opt2cmsg(mblk_t *, void *, t_uscalar_t, int,
728 			void *, t_uscalar_t);
729 extern void	soisconnecting(struct sonode *);
730 extern void	soisconnected(struct sonode *);
731 extern void	soisdisconnected(struct sonode *, int);
732 extern void	socantsendmore(struct sonode *);
733 extern void	socantrcvmore(struct sonode *);
734 extern void	soseterror(struct sonode *, int);
735 extern int	sogeterr(struct sonode *);
736 extern int	sogetrderr(vnode_t *, int, int *);
737 extern int	sogetwrerr(vnode_t *, int, int *);
738 extern void	so_unix_close(struct sonode *);
739 extern mblk_t	*soallocproto(size_t, int);
740 extern mblk_t	*soallocproto1(const void *, ssize_t, ssize_t, int);
741 extern void	soappendmsg(mblk_t *, const void *, ssize_t);
742 extern mblk_t	*soallocproto2(const void *, ssize_t, const void *, ssize_t,
743 			ssize_t, int);
744 extern mblk_t	*soallocproto3(const void *, ssize_t, const void *, ssize_t,
745 			const void *, ssize_t, ssize_t, int);
746 extern int	sowaitprim(struct sonode *, t_scalar_t, t_scalar_t,
747 			t_uscalar_t, mblk_t **, clock_t);
748 extern int	sowaitokack(struct sonode *, t_scalar_t);
749 extern int	sowaitack(struct sonode *, mblk_t **, clock_t);
750 extern void	soqueueack(struct sonode *, mblk_t *);
751 extern int	sowaitconnind(struct sonode *, int, mblk_t **);
752 extern void	soqueueconnind(struct sonode *, mblk_t *);
753 extern int	soflushconnind(struct sonode *, t_scalar_t);
754 extern void	so_drain_discon_ind(struct sonode *);
755 extern void	so_flush_discon_ind(struct sonode *);
756 extern int	sowaitconnected(struct sonode *, int, int);
757 
758 extern int	sostream_direct(struct sonode *, struct uio *,
759 		    mblk_t *, cred_t *);
760 extern int	sosend_dgram(struct sonode *, struct sockaddr *,
761 		    socklen_t, struct uio *, int);
762 extern int	sosend_svc(struct sonode *, struct uio *, t_scalar_t, int, int);
763 extern void	so_installhooks(struct sonode *);
764 extern int	so_strinit(struct sonode *, struct sonode *);
765 extern int	sotpi_recvmsg(struct sonode *, struct nmsghdr *,
766 		    struct uio *);
767 extern int	sotpi_getpeername(struct sonode *);
768 extern int	sotpi_getsockopt(struct sonode *, int, int, void *,
769 		    socklen_t *, int);
770 extern int	sotpi_setsockopt(struct sonode *, int, int, const void *,
771 		    socklen_t);
772 extern int	socktpi_ioctl(struct vnode *, int, intptr_t, int,
773 		    struct cred *, int *);
774 extern int	sodisconnect(struct sonode *, t_scalar_t, int);
775 extern ssize_t	soreadfile(file_t *, uchar_t *, u_offset_t, int *, size_t);
776 extern int	so_set_asyncsigs(vnode_t *, pid_t, int, int, cred_t *);
777 extern int	so_set_events(struct sonode *, vnode_t *, cred_t *);
778 extern int	so_flip_async(struct sonode *, vnode_t *, int, cred_t *);
779 extern int	so_set_siggrp(struct sonode *, vnode_t *, pid_t, int, cred_t *);
780 extern void	*sock_kstat_init(zoneid_t);
781 extern void	sock_kstat_fini(zoneid_t, void *);
782 
783 /*
784  * Function wrappers (mostly arround the sonode switch) for
785  * backward compatibility.
786  */
787 extern int	soaccept(struct sonode *, int, struct sonode **);
788 extern int	sobind(struct sonode *, struct sockaddr *, socklen_t,
789 		    int, int);
790 extern int	solisten(struct sonode *, int);
791 extern int	soconnect(struct sonode *, const struct sockaddr *, socklen_t,
792 		    int, int);
793 extern int	sorecvmsg(struct sonode *, struct nmsghdr *, struct uio *);
794 extern int	sosendmsg(struct sonode *, struct nmsghdr *, struct uio *);
795 extern int	sogetpeername(struct sonode *);
796 extern int	sogetsockname(struct sonode *);
797 extern int	soshutdown(struct sonode *, int);
798 extern int	sogetsockopt(struct sonode *, int, int, void *, socklen_t *,
799 		    int);
800 extern int	sosetsockopt(struct sonode *, int, int, const void *,
801 		    t_uscalar_t);
802 
803 extern struct sonode	*socreate(vnode_t *, int, int, int, int,
804 			    struct sonode *, int *);
805 
806 extern int	so_copyin(const void *, void *, size_t, int);
807 extern int	so_copyout(const void *, void *, size_t, int);
808 
809 extern int	socktpi_access(struct vnode *, int, int, struct cred *);
810 extern int	socktpi_fid(struct vnode *, struct fid *);
811 extern int	socktpi_fsync(struct vnode *, int, struct cred *);
812 extern int	socktpi_getattr(struct vnode *, struct vattr *, int,
813 		    struct cred *);
814 extern int	socktpi_seek(struct vnode *, offset_t, offset_t *);
815 extern int	socktpi_setattr(struct vnode *, struct vattr *, int,
816 		    struct cred *, caller_context_t *);
817 extern int	socktpi_setfl(vnode_t *, int, int, cred_t *);
818 
819 /* SCTP sockfs */
820 extern struct sonode	*sosctp_create(vnode_t *, int, int, int, int,
821 			    struct sonode *, int *);
822 extern int sosctp_init(void);
823 
824 #endif
825 
826 /*
827  * Internal structure for obtaining sonode information from the socklist.
828  * These types match those corresponding in the sonode structure.
829  * This is not a published interface, and may change at any time.
830  */
831 struct sockinfo {
832 	uint_t		si_size;		/* real length of this struct */
833 	short		si_family;
834 	short		si_type;
835 	ushort_t	si_flag;
836 	uint_t		si_state;
837 	uint_t		si_ux_laddr_sou_magic;
838 	uint_t		si_ux_faddr_sou_magic;
839 	t_scalar_t	si_serv_type;
840 	t_uscalar_t	si_laddr_soa_len;
841 	t_uscalar_t	si_faddr_soa_len;
842 	uint16_t	si_laddr_family;
843 	uint16_t	si_faddr_family;
844 	char		si_laddr_sun_path[MAXPATHLEN + 1]; /* NULL terminated */
845 	char		si_faddr_sun_path[MAXPATHLEN + 1];
846 	zoneid_t	si_szoneid;
847 };
848 
849 
850 #ifdef	__cplusplus
851 }
852 #endif
853 
854 #endif	/* _SYS_SOCKETVAR_H */
855