xref: /titanic_44/usr/src/uts/common/sys/socketvar.h (revision 79b62d188314c819c0ad05bd4af8d21166941e49)
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 (c) 1996, 2010, Oracle and/or its affiliates. All rights reserved.
24  */
25 
26 /*	Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T	*/
27 /*	  All Rights Reserved  	*/
28 
29 /*
30  * University Copyright- Copyright (c) 1982, 1986, 1988
31  * The Regents of the University of California
32  * All Rights Reserved
33  *
34  * University Acknowledgment- Portions of this document are derived from
35  * software developed by the University of California, Berkeley, and its
36  * contributors.
37  */
38 
39 #ifndef _SYS_SOCKETVAR_H
40 #define	_SYS_SOCKETVAR_H
41 
42 #include <sys/types.h>
43 #include <sys/stream.h>
44 #include <sys/t_lock.h>
45 #include <sys/cred.h>
46 #include <sys/vnode.h>
47 #include <sys/file.h>
48 #include <sys/param.h>
49 #include <sys/zone.h>
50 #include <sys/sdt.h>
51 #include <sys/modctl.h>
52 #include <sys/atomic.h>
53 #include <sys/socket.h>
54 #include <sys/ksocket.h>
55 #include <sys/kstat.h>
56 
57 #ifdef _KERNEL
58 #include <sys/vfs_opreg.h>
59 #endif
60 
61 #ifdef	__cplusplus
62 extern "C" {
63 #endif
64 
65 /*
66  * Internal representation of the address used to represent addresses
67  * in the loopback transport for AF_UNIX. While the sockaddr_un is used
68  * as the sockfs layer address for AF_UNIX the pathnames contained in
69  * these addresses are not unique (due to relative pathnames) thus can not
70  * be used in the transport.
71  *
72  * The transport level address consists of a magic number (used to separate the
73  * name space for specific and implicit binds). For a specific bind
74  * this is followed by a "vnode *" which ensures that all specific binds
75  * have a unique transport level address. For implicit binds the latter
76  * part of the address is a byte string (of the same length as a pointer)
77  * that is assigned by the loopback transport.
78  *
79  * The uniqueness assumes that the loopback transport has a separate namespace
80  * for sockets in order to avoid name conflicts with e.g. TLI use of the
81  * same transport.
82  */
83 struct so_ux_addr {
84 	void	*soua_vp;	/* vnode pointer or assigned by tl */
85 	uint_t	soua_magic;	/* See below */
86 };
87 
88 #define	SOU_MAGIC_EXPLICIT	0x75787670	/* "uxvp" */
89 #define	SOU_MAGIC_IMPLICIT	0x616e6f6e	/* "anon" */
90 
91 struct sockaddr_ux {
92 	sa_family_t		sou_family;	/* AF_UNIX */
93 	struct so_ux_addr	sou_addr;
94 };
95 
96 #if defined(_KERNEL) || defined(_KMEMUSER)
97 
98 #include <sys/socket_proto.h>
99 
100 typedef struct sonodeops sonodeops_t;
101 typedef struct sonode sonode_t;
102 
103 struct sodirect_s;
104 
105 /*
106  * The sonode represents a socket. A sonode never exist in the file system
107  * name space and can not be opened using open() - only the socket, socketpair
108  * and accept calls create sonodes.
109  *
110  * The locking of sockfs uses the so_lock mutex plus the SOLOCKED and
111  * SOREADLOCKED flags in so_flag. The mutex protects all the state in the
112  * sonode. It is expected that the underlying transport protocol serializes
113  * socket operations, so sockfs will not normally not single-thread
114  * operations. However, certain sockets, including TPI based ones, can only
115  * handle one control operation at a time. The SOLOCKED flag is used to
116  * single-thread operations from sockfs users to prevent e.g. multiple bind()
117  * calls to operate on the same sonode concurrently. The SOREADLOCKED flag is
118  * used to ensure that only one thread sleeps in kstrgetmsg for a given
119  * sonode. This is needed to ensure atomic operation for things like
120  * MSG_WAITALL.
121  *
122  * The so_fallback_rwlock is used to ensure that for sockets that can
123  * fall back to TPI, the fallback is not initiated until all pending
124  * operations have completed.
125  *
126  * Note that so_lock is sometimes held across calls that might go to sleep
127  * (kmem_alloc and soallocproto*). This implies that no other lock in
128  * the system should be held when calling into sockfs; from the system call
129  * side or from strrput (in case of TPI based sockets). If locks are held
130  * while calling into sockfs the system might hang when running low on memory.
131  */
132 struct sonode {
133 	struct	vnode	*so_vnode;	/* vnode associated with this sonode */
134 
135 	sonodeops_t 	*so_ops;	/* operations vector for this sonode */
136 	void		*so_priv;	/* sonode private data */
137 
138 	krwlock_t	so_fallback_rwlock;
139 	kmutex_t	so_lock;	/* protects sonode fields */
140 
141 	kcondvar_t	so_state_cv;	/* synchronize state changes */
142 	kcondvar_t	so_single_cv;	/* wait due to SOLOCKED */
143 	kcondvar_t	so_read_cv;	/* wait due to SOREADLOCKED */
144 
145 	/* These fields are protected by so_lock */
146 
147 	uint_t		so_state;	/* internal state flags SS_*, below */
148 	uint_t		so_mode;	/* characteristics on socket. SM_* */
149 	ushort_t 	so_flag;	/* flags, see below */
150 	int		so_count;	/* count of opened references */
151 
152 	sock_connid_t	so_proto_connid; /* protocol generation number */
153 
154 	ushort_t 	so_error;	/* error affecting connection */
155 
156 	struct sockparams *so_sockparams;	/* vnode or socket module */
157 	/* Needed to recreate the same socket for accept */
158 	short	so_family;
159 	short	so_type;
160 	short	so_protocol;
161 	short	so_version;		/* From so_socket call */
162 
163 	/* Accept queue */
164 	kmutex_t	so_acceptq_lock;	/* protects accept queue */
165 	list_t		so_acceptq_list;	/* pending conns */
166 	list_t		so_acceptq_defer;	/* deferred conns */
167 	list_node_t	so_acceptq_node;	/* acceptq list node */
168 	unsigned int	so_acceptq_len;		/* # of conns (both lists) */
169 	unsigned int	so_backlog;		/* Listen backlog */
170 	kcondvar_t	so_acceptq_cv;		/* wait for new conn. */
171 	struct sonode	*so_listener;		/* parent socket */
172 
173 	/* Options */
174 	short	so_options;		/* From socket call, see socket.h */
175 	struct linger	so_linger;	/* SO_LINGER value */
176 #define	so_sndbuf	so_proto_props.sopp_txhiwat	/* SO_SNDBUF value */
177 #define	so_sndlowat	so_proto_props.sopp_txlowat	/* tx low water mark */
178 #define	so_rcvbuf	so_proto_props.sopp_rxhiwat	/* SO_RCVBUF value */
179 #define	so_rcvlowat	so_proto_props.sopp_rxlowat	/* rx low water mark */
180 #define	so_max_addr_len	so_proto_props.sopp_maxaddrlen
181 #define	so_minpsz	so_proto_props.sopp_minpsz
182 #define	so_maxpsz	so_proto_props.sopp_maxpsz
183 
184 	int	so_xpg_rcvbuf;		/* SO_RCVBUF value for XPG4 socket */
185 	clock_t	so_sndtimeo;		/* send timeout */
186 	clock_t	so_rcvtimeo;		/* recv timeout */
187 
188 	mblk_t	*so_oobmsg;		/* outofline oob data */
189 	ssize_t	so_oobmark;		/* offset of the oob data */
190 
191 	pid_t	so_pgrp;		/* pgrp for signals */
192 
193 	cred_t		*so_peercred;	/* connected socket peer cred */
194 	pid_t		so_cpid;	/* connected socket peer cached pid */
195 	zoneid_t	so_zoneid;	/* opener's zoneid */
196 
197 	struct pollhead	so_poll_list;	/* common pollhead */
198 	short		so_pollev;	/* events that should be generated */
199 
200 	/* Receive */
201 	unsigned int	so_rcv_queued;	/* # bytes on both rcv lists */
202 	mblk_t		*so_rcv_q_head;	/* processing/copyout rcv queue */
203 	mblk_t		*so_rcv_q_last_head;
204 	mblk_t		*so_rcv_head;	/* protocol prequeue */
205 	mblk_t		*so_rcv_last_head;	/* last mblk in b_next chain */
206 	kcondvar_t	so_rcv_cv;	/* wait for data */
207 	uint_t		so_rcv_wanted;	/* # of bytes wanted by app */
208 	timeout_id_t	so_rcv_timer_tid;
209 
210 #define	so_rcv_thresh	so_proto_props.sopp_rcvthresh
211 #define	so_rcv_timer_interval so_proto_props.sopp_rcvtimer
212 
213 	kcondvar_t	so_snd_cv;	/* wait for snd buffers */
214 	uint32_t
215 		so_snd_qfull: 1,	/* Transmit full */
216 		so_rcv_wakeup: 1,
217 		so_snd_wakeup: 1,
218 		so_not_str: 1,	/* B_TRUE if not streams based socket */
219 		so_pad_to_bit_31: 28;
220 
221 	/* Communication channel with protocol */
222 	sock_lower_handle_t	so_proto_handle;
223 	sock_downcalls_t 	*so_downcalls;
224 
225 	struct sock_proto_props	so_proto_props; /* protocol settings */
226 	boolean_t		so_flowctrld;	/* Flow controlled */
227 	uint_t			so_copyflag;	/* Copy related flag */
228 	kcondvar_t		so_copy_cv;	/* Copy cond variable */
229 
230 	/* kernel sockets */
231 	ksocket_callbacks_t 	so_ksock_callbacks;
232 	void			*so_ksock_cb_arg;	/* callback argument */
233 	kcondvar_t		so_closing_cv;
234 
235 	/* != NULL for sodirect enabled socket */
236 	struct sodirect_s	*so_direct;
237 
238 	/* socket filters */
239 	uint_t			so_filter_active;	/* # of active fil */
240 	uint_t			so_filter_tx;		/* pending tx ops */
241 	struct sof_instance	*so_filter_top;		/* top of stack */
242 	struct sof_instance	*so_filter_bottom;	/* bottom of stack */
243 	clock_t			so_filter_defertime;	/* time when deferred */
244 };
245 
246 #define	SO_HAVE_DATA(so)						\
247 	/*								\
248 	 * For the (tid == 0) case we must check so_rcv_{q_,}head	\
249 	 * rather than (so_rcv_queued > 0), since the latter does not	\
250 	 * take into account mblks with only control/name information.	\
251 	 */								\
252 	((so)->so_rcv_timer_tid == 0 && ((so)->so_rcv_head != NULL ||	\
253 	(so)->so_rcv_q_head != NULL)) ||				\
254 	((so)->so_state & SS_CANTRCVMORE)
255 
256 /*
257  * Events handled by the protocol (in case sd_poll is set)
258  */
259 #define	SO_PROTO_POLLEV		(POLLIN|POLLRDNORM|POLLRDBAND)
260 
261 
262 #endif /* _KERNEL || _KMEMUSER */
263 
264 /* flags */
265 #define	SOMOD		0x0001		/* update socket modification time */
266 #define	SOACC		0x0002		/* update socket access time */
267 
268 #define	SOLOCKED	0x0010		/* use to serialize open/closes */
269 #define	SOREADLOCKED	0x0020		/* serialize kstrgetmsg calls */
270 #define	SOCLONE		0x0040		/* child of clone driver */
271 #define	SOASYNC_UNBIND	0x0080		/* wait for ACK of async unbind */
272 
273 #define	SOCK_IS_NONSTR(so)	((so)->so_not_str)
274 
275 /*
276  * Socket state bits.
277  */
278 #define	SS_ISCONNECTED		0x00000001 /* socket connected to a peer */
279 #define	SS_ISCONNECTING		0x00000002 /* in process, connecting to peer */
280 #define	SS_ISDISCONNECTING	0x00000004 /* in process of disconnecting */
281 #define	SS_CANTSENDMORE		0x00000008 /* can't send more data to peer */
282 
283 #define	SS_CANTRCVMORE		0x00000010 /* can't receive more data */
284 #define	SS_ISBOUND		0x00000020 /* socket is bound */
285 #define	SS_NDELAY		0x00000040 /* FNDELAY non-blocking */
286 #define	SS_NONBLOCK		0x00000080 /* O_NONBLOCK non-blocking */
287 
288 #define	SS_ASYNC		0x00000100 /* async i/o notify */
289 #define	SS_ACCEPTCONN		0x00000200 /* listen done */
290 /*	unused			0x00000400 */	/* was SS_HASCONNIND */
291 #define	SS_SAVEDEOR		0x00000800 /* Saved MSG_EOR rcv side state */
292 
293 #define	SS_RCVATMARK		0x00001000 /* at mark on input */
294 #define	SS_OOBPEND		0x00002000 /* OOB pending or present - poll */
295 #define	SS_HAVEOOBDATA		0x00004000 /* OOB data present */
296 #define	SS_HADOOBDATA		0x00008000 /* OOB data consumed */
297 #define	SS_CLOSING		0x00010000 /* in process of closing */
298 
299 #define	SS_FIL_DEFER		0x00020000 /* filter deferred notification */
300 #define	SS_FILOP_OK		0x00040000 /* socket can attach filters */
301 #define	SS_FIL_RCV_FLOWCTRL	0x00080000 /* filter asserted rcv flow ctrl */
302 #define	SS_FIL_SND_FLOWCTRL	0x00100000 /* filter asserted snd flow ctrl */
303 /*	unused 			0x00200000 */	/* was SS_DIRECT */
304 
305 #define	SS_SODIRECT		0x00400000 /* transport supports sodirect */
306 
307 #define	SS_SENTLASTREADSIG	0x01000000 /* last rx signal has been sent */
308 #define	SS_SENTLASTWRITESIG	0x02000000 /* last tx signal has been sent */
309 
310 #define	SS_FALLBACK_DRAIN	0x20000000 /* data was/is being drained */
311 #define	SS_FALLBACK_PENDING	0x40000000 /* fallback is pending */
312 #define	SS_FALLBACK_COMP	0x80000000 /* fallback has completed */
313 
314 
315 /* Set of states when the socket can't be rebound */
316 #define	SS_CANTREBIND	(SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING|\
317 			    SS_CANTSENDMORE|SS_CANTRCVMORE|SS_ACCEPTCONN)
318 
319 /*
320  * Sockets that can fall back to TPI must ensure that fall back is not
321  * initiated while a thread is using a socket.
322  */
323 #define	SO_BLOCK_FALLBACK(so, fn)				\
324 	ASSERT(MUTEX_NOT_HELD(&(so)->so_lock));			\
325 	rw_enter(&(so)->so_fallback_rwlock, RW_READER);		\
326 	if ((so)->so_state & (SS_FALLBACK_COMP|SS_FILOP_OK)) {	\
327 		if ((so)->so_state & SS_FALLBACK_COMP) {	\
328 			rw_exit(&(so)->so_fallback_rwlock);	\
329 			return (fn);				\
330 		} else {					\
331 			mutex_enter(&(so)->so_lock);		\
332 			(so)->so_state &= ~SS_FILOP_OK;		\
333 			mutex_exit(&(so)->so_lock);		\
334 		}						\
335 	}
336 
337 #define	SO_UNBLOCK_FALLBACK(so)	{			\
338 	rw_exit(&(so)->so_fallback_rwlock);		\
339 }
340 
341 #define	SO_SND_FLOWCTRLD(so)	\
342 	((so)->so_snd_qfull || (so)->so_state & SS_FIL_SND_FLOWCTRL)
343 
344 /* Poll events */
345 #define	SO_POLLEV_IN		0x1	/* POLLIN wakeup needed */
346 #define	SO_POLLEV_ALWAYS	0x2	/* wakeups */
347 
348 /*
349  * Characteristics of sockets. Not changed after the socket is created.
350  */
351 #define	SM_PRIV			0x001	/* privileged for broadcast, raw... */
352 #define	SM_ATOMIC		0x002	/* atomic data transmission */
353 #define	SM_ADDR			0x004	/* addresses given with messages */
354 #define	SM_CONNREQUIRED		0x008	/* connection required by protocol */
355 
356 #define	SM_FDPASSING		0x010	/* passes file descriptors */
357 #define	SM_EXDATA		0x020	/* Can handle T_EXDATA_REQ */
358 #define	SM_OPTDATA		0x040	/* Can handle T_OPTDATA_REQ */
359 #define	SM_BYTESTREAM		0x080	/* Byte stream - can use M_DATA */
360 
361 #define	SM_ACCEPTOR_ID		0x100	/* so_acceptor_id is valid */
362 
363 #define	SM_KERNEL		0x200	/* kernel socket */
364 
365 /* The modes below are only for non-streams sockets */
366 #define	SM_ACCEPTSUPP		0x400	/* can handle accept() */
367 #define	SM_SENDFILESUPP		0x800	/* Private: proto supp sendfile  */
368 
369 /*
370  * Socket versions. Used by the socket library when calling _so_socket().
371  */
372 #define	SOV_STREAM	0	/* Not a socket - just a stream */
373 #define	SOV_DEFAULT	1	/* Select based on so_default_version */
374 #define	SOV_SOCKSTREAM	2	/* Socket plus streams operations */
375 #define	SOV_SOCKBSD	3	/* Socket with no streams operations */
376 #define	SOV_XPG4_2	4	/* Xnet socket */
377 
378 #if defined(_KERNEL) || defined(_KMEMUSER)
379 
380 /*
381  * sonode create and destroy functions.
382  */
383 typedef struct sonode *(*so_create_func_t)(struct sockparams *,
384     int, int, int, int, int, int *, cred_t *);
385 typedef void (*so_destroy_func_t)(struct sonode *);
386 
387 /* STREAM device information */
388 typedef struct sdev_info {
389 	char	*sd_devpath;
390 	int	sd_devpathlen; /* Is 0 if sp_devpath is a static string */
391 	vnode_t	*sd_vnode;
392 } sdev_info_t;
393 
394 #define	SOCKMOD_VERSION_1	1
395 #define	SOCKMOD_VERSION		2
396 
397 /* name of the TPI pseudo socket module */
398 #define	SOTPI_SMOD_NAME		"socktpi"
399 
400 typedef struct __smod_priv_s {
401 	so_create_func_t	smodp_sock_create_func;
402 	so_destroy_func_t	smodp_sock_destroy_func;
403 	so_proto_fallback_func_t smodp_proto_fallback_func;
404 	const char		*smodp_fallback_devpath_v4;
405 	const char		*smodp_fallback_devpath_v6;
406 } __smod_priv_t;
407 
408 /*
409  * Socket module register information
410  */
411 typedef struct smod_reg_s {
412 	int		smod_version;
413 	char		*smod_name;
414 	size_t		smod_uc_version;
415 	size_t		smod_dc_version;
416 	so_proto_create_func_t	smod_proto_create_func;
417 
418 	/* __smod_priv_data must be NULL */
419 	__smod_priv_t	*__smod_priv;
420 } smod_reg_t;
421 
422 /*
423  * Socket module information
424  */
425 typedef struct smod_info {
426 	int		smod_version;
427 	char		*smod_name;
428 	uint_t		smod_refcnt;		/* # of entries */
429 	size_t		smod_uc_version; 	/* upcall version */
430 	size_t		smod_dc_version;	/* down call version */
431 	so_proto_create_func_t	smod_proto_create_func;
432 	so_proto_fallback_func_t smod_proto_fallback_func;
433 	const char		*smod_fallback_devpath_v4;
434 	const char		*smod_fallback_devpath_v6;
435 	so_create_func_t	smod_sock_create_func;
436 	so_destroy_func_t	smod_sock_destroy_func;
437 	list_node_t	smod_node;
438 } smod_info_t;
439 
440 typedef struct sockparams_stats {
441 	kstat_named_t	sps_nfallback;	/* # of fallbacks to TPI */
442 	kstat_named_t	sps_nactive;	/* # of active sockets */
443 	kstat_named_t	sps_ncreate;	/* total # of created sockets */
444 } sockparams_stats_t;
445 
446 /*
447  * sockparams
448  *
449  * Used for mapping family/type/protocol to a socket module or STREAMS device
450  */
451 struct sockparams {
452 	/*
453 	 * The family, type, protocol, sdev_info and smod_name are
454 	 * set when the entry is created, and they will never change
455 	 * thereafter.
456 	 */
457 	int		sp_family;
458 	int		sp_type;
459 	int		sp_protocol;
460 
461 	sdev_info_t	sp_sdev_info;	/* STREAM device */
462 	char		*sp_smod_name;	/* socket module name */
463 
464 	kmutex_t	sp_lock;	/* lock for refcnt and smod_info */
465 	uint64_t	sp_refcnt;	/* entry reference count */
466 	smod_info_t	*sp_smod_info;	/* socket module */
467 
468 	sockparams_stats_t sp_stats;
469 	kstat_t		*sp_kstat;
470 
471 	/*
472 	 * The entries below are only modified while holding
473 	 * sockconf_lock as a writer.
474 	 */
475 	int		sp_flags;	/* see below */
476 	list_node_t	sp_node;
477 
478 	list_t		sp_auto_filters; /* list of automatic filters */
479 	list_t		sp_prog_filters; /* list of programmatic filters */
480 };
481 
482 struct sof_entry;
483 
484 typedef struct sp_filter {
485 	struct sof_entry *spf_filter;
486 	list_node_t	spf_node;
487 } sp_filter_t;
488 
489 
490 /*
491  * sockparams flags
492  */
493 #define	SOCKPARAMS_EPHEMERAL	0x1	/* temp. entry, not on global list */
494 
495 extern void sockparams_init(void);
496 extern struct sockparams *sockparams_hold_ephemeral_bydev(int, int, int,
497     const char *, int, int *);
498 extern struct sockparams *sockparams_hold_ephemeral_bymod(int, int, int,
499     const char *, int, int *);
500 extern void sockparams_ephemeral_drop_last_ref(struct sockparams *);
501 
502 extern struct sockparams *sockparams_create(int, int, int, char *, char *, int,
503     int, int, int *);
504 extern void 	sockparams_destroy(struct sockparams *);
505 extern int 	sockparams_add(struct sockparams *);
506 extern int	sockparams_delete(int, int, int);
507 extern int	sockparams_new_filter(struct sof_entry *);
508 extern void	sockparams_filter_cleanup(struct sof_entry *);
509 
510 extern void smod_init(void);
511 extern void smod_add(smod_info_t *);
512 extern int smod_register(const smod_reg_t *);
513 extern int smod_unregister(const char *);
514 extern smod_info_t *smod_lookup_byname(const char *);
515 
516 #define	SOCKPARAMS_HAS_DEVICE(sp)					\
517 	((sp)->sp_sdev_info.sd_devpath != NULL)
518 
519 /* Increase the smod_info_t reference count */
520 #define	SMOD_INC_REF(smodp) {						\
521 	ASSERT((smodp) != NULL);					\
522 	DTRACE_PROBE1(smodinfo__inc__ref, struct smod_info *, (smodp));	\
523 	atomic_inc_uint(&(smodp)->smod_refcnt);				\
524 }
525 
526 /*
527  * Decreace the socket module entry reference count.
528  * When no one mapping to the entry, we try to unload the module from the
529  * kernel. If the module can't unload, just leave the module entry with
530  * a zero refcnt.
531  */
532 #define	SMOD_DEC_REF(smodp, modname) {					\
533 	ASSERT((smodp) != NULL);					\
534 	ASSERT((smodp)->smod_refcnt != 0);				\
535 	atomic_dec_uint(&(smodp)->smod_refcnt);				\
536 	/*								\
537 	 * No need to atomically check the return value because the	\
538 	 * socket module framework will verify that no one is using	\
539 	 * the module before unloading. Worst thing that can happen	\
540 	 * here is multiple calls to mod_remove_by_name(), which is OK.	\
541 	 */								\
542 	if ((smodp)->smod_refcnt == 0)					\
543 		(void) mod_remove_by_name(modname);			\
544 }
545 
546 /* Increase the reference count */
547 #define	SOCKPARAMS_INC_REF(sp) {					\
548 	ASSERT((sp) != NULL);						\
549 	DTRACE_PROBE1(sockparams__inc__ref, struct sockparams *, (sp));	\
550 	mutex_enter(&(sp)->sp_lock);					\
551 	(sp)->sp_refcnt++;						\
552 	ASSERT((sp)->sp_refcnt != 0);					\
553 	mutex_exit(&(sp)->sp_lock);					\
554 }
555 
556 /*
557  * Decrease the reference count.
558  *
559  * If the sockparams is ephemeral, then the thread dropping the last ref
560  * count will destroy the entry.
561  */
562 #define	SOCKPARAMS_DEC_REF(sp) {					\
563 	ASSERT((sp) != NULL);						\
564 	DTRACE_PROBE1(sockparams__dec__ref, struct sockparams *, (sp));	\
565 	mutex_enter(&(sp)->sp_lock);					\
566 	ASSERT((sp)->sp_refcnt > 0);					\
567 	if ((sp)->sp_refcnt == 1) {					\
568 		if ((sp)->sp_flags & SOCKPARAMS_EPHEMERAL) {		\
569 			mutex_exit(&(sp)->sp_lock);			\
570 			sockparams_ephemeral_drop_last_ref((sp));	\
571 		} else {						\
572 			(sp)->sp_refcnt--;				\
573 			if ((sp)->sp_smod_info != NULL) {		\
574 				SMOD_DEC_REF((sp)->sp_smod_info,	\
575 				    (sp)->sp_smod_name);		\
576 			}						\
577 			(sp)->sp_smod_info = NULL;			\
578 			mutex_exit(&(sp)->sp_lock);			\
579 		}							\
580 	} else {							\
581 		(sp)->sp_refcnt--;					\
582 		mutex_exit(&(sp)->sp_lock);				\
583 	}								\
584 }
585 
586 /*
587  * Used to traverse the list of AF_UNIX sockets to construct the kstat
588  * for netstat(1m).
589  */
590 struct socklist {
591 	kmutex_t	sl_lock;
592 	struct sonode	*sl_list;
593 };
594 
595 extern struct socklist socklist;
596 /*
597  * ss_full_waits is the number of times the reader thread
598  * waits when the queue is full and ss_empty_waits is the number
599  * of times the consumer thread waits when the queue is empty.
600  * No locks for these as they are just indicators of whether
601  * disk or network or both is slow or fast.
602  */
603 struct sendfile_stats {
604 	uint32_t ss_file_cached;
605 	uint32_t ss_file_not_cached;
606 	uint32_t ss_full_waits;
607 	uint32_t ss_empty_waits;
608 	uint32_t ss_file_segmap;
609 };
610 
611 /*
612  * A single sendfile request is represented by snf_req.
613  */
614 typedef struct snf_req {
615 	struct snf_req	*sr_next;
616 	mblk_t		*sr_mp_head;
617 	mblk_t		*sr_mp_tail;
618 	kmutex_t	sr_lock;
619 	kcondvar_t	sr_cv;
620 	uint_t		sr_qlen;
621 	int		sr_hiwat;
622 	int		sr_lowat;
623 	int		sr_operation;
624 	struct vnode	*sr_vp;
625 	file_t 		*sr_fp;
626 	ssize_t		sr_maxpsz;
627 	u_offset_t	sr_file_off;
628 	u_offset_t	sr_file_size;
629 #define	SR_READ_DONE	0x80000000
630 	int		sr_read_error;
631 	int		sr_write_error;
632 } snf_req_t;
633 
634 /* A queue of sendfile requests */
635 struct sendfile_queue {
636 	snf_req_t	*snfq_req_head;
637 	snf_req_t	*snfq_req_tail;
638 	kmutex_t	snfq_lock;
639 	kcondvar_t	snfq_cv;
640 	int		snfq_svc_threads;	/* # of service threads */
641 	int		snfq_idle_cnt;		/* # of idling threads */
642 	int		snfq_max_threads;
643 	int		snfq_req_cnt;		/* Number of requests */
644 };
645 
646 #define	READ_OP			1
647 #define	SNFQ_TIMEOUT		(60 * 5 * hz)	/* 5 minutes */
648 
649 /* Socket network operations switch */
650 struct sonodeops {
651 	int 	(*sop_init)(struct sonode *, struct sonode *, cred_t *,
652 		    int);
653 	int	(*sop_accept)(struct sonode *, int, cred_t *, struct sonode **);
654 	int	(*sop_bind)(struct sonode *, struct sockaddr *, socklen_t,
655 		    int, cred_t *);
656 	int	(*sop_listen)(struct sonode *, int, cred_t *);
657 	int	(*sop_connect)(struct sonode *, struct sockaddr *,
658 		    socklen_t, int, int, cred_t *);
659 	int	(*sop_recvmsg)(struct sonode *, struct msghdr *,
660 		    struct uio *, cred_t *);
661 	int	(*sop_sendmsg)(struct sonode *, struct msghdr *,
662 		    struct uio *, cred_t *);
663 	int	(*sop_sendmblk)(struct sonode *, struct msghdr *, int,
664 		    cred_t *, mblk_t **);
665 	int	(*sop_getpeername)(struct sonode *, struct sockaddr *,
666 		    socklen_t *, boolean_t, cred_t *);
667 	int	(*sop_getsockname)(struct sonode *, struct sockaddr *,
668 		    socklen_t *, cred_t *);
669 	int	(*sop_shutdown)(struct sonode *, int, cred_t *);
670 	int	(*sop_getsockopt)(struct sonode *, int, int, void *,
671 		    socklen_t *, int, cred_t *);
672 	int 	(*sop_setsockopt)(struct sonode *, int, int, const void *,
673 		    socklen_t, cred_t *);
674 	int 	(*sop_ioctl)(struct sonode *, int, intptr_t, int,
675 		    cred_t *, int32_t *);
676 	int 	(*sop_poll)(struct sonode *, short, int, short *,
677 		    struct pollhead **);
678 	int 	(*sop_close)(struct sonode *, int, cred_t *);
679 };
680 
681 #define	SOP_INIT(so, flag, cr, flags)	\
682 	((so)->so_ops->sop_init((so), (flag), (cr), (flags)))
683 #define	SOP_ACCEPT(so, fflag, cr, nsop)	\
684 	((so)->so_ops->sop_accept((so), (fflag), (cr), (nsop)))
685 #define	SOP_BIND(so, name, namelen, flags, cr)	\
686 	((so)->so_ops->sop_bind((so), (name), (namelen), (flags), (cr)))
687 #define	SOP_LISTEN(so, backlog, cr)	\
688 	((so)->so_ops->sop_listen((so), (backlog), (cr)))
689 #define	SOP_CONNECT(so, name, namelen, fflag, flags, cr)	\
690 	((so)->so_ops->sop_connect((so), (name), (namelen), (fflag), (flags), \
691 	(cr)))
692 #define	SOP_RECVMSG(so, msg, uiop, cr)	\
693 	((so)->so_ops->sop_recvmsg((so), (msg), (uiop), (cr)))
694 #define	SOP_SENDMSG(so, msg, uiop, cr)	\
695 	((so)->so_ops->sop_sendmsg((so), (msg), (uiop), (cr)))
696 #define	SOP_SENDMBLK(so, msg, size, cr, mpp)	\
697 	((so)->so_ops->sop_sendmblk((so), (msg), (size), (cr), (mpp)))
698 #define	SOP_GETPEERNAME(so, addr, addrlen, accept, cr)	\
699 	((so)->so_ops->sop_getpeername((so), (addr), (addrlen), (accept), (cr)))
700 #define	SOP_GETSOCKNAME(so, addr, addrlen, cr)	\
701 	((so)->so_ops->sop_getsockname((so), (addr), (addrlen), (cr)))
702 #define	SOP_SHUTDOWN(so, how, cr)	\
703 	((so)->so_ops->sop_shutdown((so), (how), (cr)))
704 #define	SOP_GETSOCKOPT(so, level, optionname, optval, optlenp, flags, cr) \
705 	((so)->so_ops->sop_getsockopt((so), (level), (optionname),	\
706 	    (optval), (optlenp), (flags), (cr)))
707 #define	SOP_SETSOCKOPT(so, level, optionname, optval, optlen, cr)	\
708 	((so)->so_ops->sop_setsockopt((so), (level), (optionname),	\
709 	    (optval), (optlen), (cr)))
710 #define	SOP_IOCTL(so, cmd, arg, mode, cr, rvalp)	\
711 	((so)->so_ops->sop_ioctl((so), (cmd), (arg), (mode), (cr), (rvalp)))
712 #define	SOP_POLL(so, events, anyyet, reventsp, phpp) \
713 	((so)->so_ops->sop_poll((so), (events), (anyyet), (reventsp), (phpp)))
714 #define	SOP_CLOSE(so, flag, cr)	\
715 	((so)->so_ops->sop_close((so), (flag), (cr)))
716 
717 #endif /* defined(_KERNEL) || defined(_KMEMUSER) */
718 
719 #ifdef _KERNEL
720 
721 #define	ISALIGNED_cmsghdr(addr) \
722 		(((uintptr_t)(addr) & (_CMSG_HDR_ALIGNMENT - 1)) == 0)
723 
724 #define	ROUNDUP_cmsglen(len) \
725 	(((len) + _CMSG_HDR_ALIGNMENT - 1) & ~(_CMSG_HDR_ALIGNMENT - 1))
726 
727 #define	IS_NON_STREAM_SOCK(vp) \
728 	((vp)->v_type == VSOCK && (vp)->v_stream == NULL)
729 /*
730  * Macros that operate on struct cmsghdr.
731  * Used in parsing msg_control.
732  * The CMSG_VALID macro does not assume that the last option buffer is padded.
733  */
734 #define	CMSG_NEXT(cmsg)						\
735 	(struct cmsghdr *)((uintptr_t)(cmsg) +			\
736 	    ROUNDUP_cmsglen((cmsg)->cmsg_len))
737 #define	CMSG_CONTENT(cmsg)	(&((cmsg)[1]))
738 #define	CMSG_CONTENTLEN(cmsg)	((cmsg)->cmsg_len - sizeof (struct cmsghdr))
739 #define	CMSG_VALID(cmsg, start, end)					\
740 	(ISALIGNED_cmsghdr(cmsg) &&					\
741 	((uintptr_t)(cmsg) >= (uintptr_t)(start)) &&			\
742 	((uintptr_t)(cmsg) < (uintptr_t)(end)) &&			\
743 	((ssize_t)(cmsg)->cmsg_len >= sizeof (struct cmsghdr)) &&	\
744 	((uintptr_t)(cmsg) + (cmsg)->cmsg_len <= (uintptr_t)(end)))
745 
746 /*
747  * Maximum size of any argument that is copied in (addresses, options,
748  * access rights). MUST be at least MAXPATHLEN + 3.
749  * BSD and SunOS 4.X limited this to MLEN or MCLBYTES.
750  */
751 #define	SO_MAXARGSIZE	8192
752 
753 /*
754  * Convert between vnode and sonode
755  */
756 #define	VTOSO(vp)	((struct sonode *)((vp)->v_data))
757 #define	SOTOV(sp)	((sp)->so_vnode)
758 
759 /*
760  * Internal flags for sobind()
761  */
762 #define	_SOBIND_REBIND		0x01	/* Bind to existing local address */
763 #define	_SOBIND_UNSPEC		0x02	/* Bind to unspecified address */
764 #define	_SOBIND_LOCK_HELD	0x04	/* so_excl_lock held by caller */
765 #define	_SOBIND_NOXLATE		0x08	/* No addr translation for AF_UNIX */
766 #define	_SOBIND_XPG4_2		0x10	/* xpg4.2 semantics */
767 #define	_SOBIND_SOCKBSD		0x20	/* BSD semantics */
768 #define	_SOBIND_LISTEN		0x40	/* Make into SS_ACCEPTCONN */
769 #define	_SOBIND_SOCKETPAIR	0x80	/* Internal flag for so_socketpair() */
770 					/* to enable listen with backlog = 1 */
771 
772 /*
773  * Internal flags for sounbind()
774  */
775 #define	_SOUNBIND_REBIND	0x01	/* Don't clear fields - will rebind */
776 
777 /*
778  * Internal flags for soconnect()
779  */
780 #define	_SOCONNECT_NOXLATE	0x01	/* No addr translation for AF_UNIX */
781 #define	_SOCONNECT_DID_BIND	0x02	/* Unbind when connect fails */
782 #define	_SOCONNECT_XPG4_2	0x04	/* xpg4.2 semantics */
783 
784 /*
785  * Internal flags for sodisconnect()
786  */
787 #define	_SODISCONNECT_LOCK_HELD	0x01	/* so_excl_lock held by caller */
788 
789 /*
790  * Internal flags for sotpi_getsockopt().
791  */
792 #define	_SOGETSOCKOPT_XPG4_2	0x01	/* xpg4.2 semantics */
793 
794 /*
795  * Internal flags for soallocproto*()
796  */
797 #define	_ALLOC_NOSLEEP		0	/* Don't sleep for memory */
798 #define	_ALLOC_INTR		1	/* Sleep until interrupt */
799 #define	_ALLOC_SLEEP		2	/* Sleep forever */
800 
801 /*
802  * Internal structure for handling AF_UNIX file descriptor passing
803  */
804 struct fdbuf {
805 	int		fd_size;	/* In bytes, for kmem_free */
806 	int		fd_numfd;	/* Number of elements below */
807 	char		*fd_ebuf;	/* Extra buffer to free  */
808 	int		fd_ebuflen;
809 	frtn_t		fd_frtn;
810 	struct file	*fd_fds[1];	/* One or more */
811 };
812 #define	FDBUF_HDRSIZE	(sizeof (struct fdbuf) - sizeof (struct file *))
813 
814 /*
815  * Variable that can be patched to set what version of socket socket()
816  * will create.
817  */
818 extern int so_default_version;
819 
820 #ifdef DEBUG
821 /* Turn on extra testing capabilities */
822 #define	SOCK_TEST
823 #endif /* DEBUG */
824 
825 #ifdef DEBUG
826 char	*pr_state(uint_t, uint_t);
827 char	*pr_addr(int, struct sockaddr *, t_uscalar_t);
828 int	so_verify_oobstate(struct sonode *);
829 #endif /* DEBUG */
830 
831 /*
832  * DEBUG macros
833  */
834 #if defined(DEBUG)
835 #define	SOCK_DEBUG
836 
837 extern int sockdebug;
838 extern int sockprinterr;
839 
840 #define	eprint(args)	printf args
841 #define	eprintso(so, args) \
842 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) printf args; }
843 #define	eprintline(error)					\
844 {								\
845 	if (error != EINTR && (sockprinterr || sockdebug > 0))	\
846 		printf("socket error %d: line %d file %s\n",	\
847 			(error), __LINE__, __FILE__);		\
848 }
849 
850 #define	eprintsoline(so, error)					\
851 { if (sockprinterr && ((so)->so_options & SO_DEBUG))		\
852 	printf("socket(%p) error %d: line %d file %s\n",	\
853 		(void *)(so), (error), __LINE__, __FILE__);	\
854 }
855 #define	dprint(level, args)	{ if (sockdebug > (level)) printf args; }
856 #define	dprintso(so, level, args) \
857 { if (sockdebug > (level) && ((so)->so_options & SO_DEBUG)) printf args; }
858 
859 #else /* define(DEBUG) */
860 
861 #define	eprint(args)		{}
862 #define	eprintso(so, args)	{}
863 #define	eprintline(error)	{}
864 #define	eprintsoline(so, error)	{}
865 #define	dprint(level, args)	{}
866 #define	dprintso(so, level, args) {}
867 
868 #endif /* defined(DEBUG) */
869 
870 extern struct vfsops			sock_vfsops;
871 extern struct vnodeops			*socket_vnodeops;
872 extern const struct fs_operation_def	socket_vnodeops_template[];
873 
874 extern dev_t				sockdev;
875 
876 extern krwlock_t			sockconf_lock;
877 
878 /*
879  * sockfs functions
880  */
881 extern int	sock_getmsg(vnode_t *, struct strbuf *, struct strbuf *,
882 			uchar_t *, int *, int, rval_t *);
883 extern int	sock_putmsg(vnode_t *, struct strbuf *, struct strbuf *,
884 			uchar_t, int, int);
885 extern int	sogetvp(char *, vnode_t **, int);
886 extern int	sockinit(int, char *);
887 extern int	solookup(int, int, int, struct sockparams **);
888 extern void	so_lock_single(struct sonode *);
889 extern void	so_unlock_single(struct sonode *, int);
890 extern int	so_lock_read(struct sonode *, int);
891 extern int	so_lock_read_intr(struct sonode *, int);
892 extern void	so_unlock_read(struct sonode *);
893 extern void	*sogetoff(mblk_t *, t_uscalar_t, t_uscalar_t, uint_t);
894 extern void	so_getopt_srcaddr(void *, t_uscalar_t,
895 			void **, t_uscalar_t *);
896 extern int	so_getopt_unix_close(void *, t_uscalar_t);
897 extern void	fdbuf_free(struct fdbuf *);
898 extern mblk_t	*fdbuf_allocmsg(int, struct fdbuf *);
899 extern int	fdbuf_create(void *, int, struct fdbuf **);
900 extern void	so_closefds(void *, t_uscalar_t, int, int);
901 extern int	so_getfdopt(void *, t_uscalar_t, int, void **, int *);
902 t_uscalar_t	so_optlen(void *, t_uscalar_t, int);
903 extern void	so_cmsg2opt(void *, t_uscalar_t, int, mblk_t *);
904 extern t_uscalar_t
905 		so_cmsglen(mblk_t *, void *, t_uscalar_t, int);
906 extern int	so_opt2cmsg(mblk_t *, void *, t_uscalar_t, int,
907 			void *, t_uscalar_t);
908 extern void	soisconnecting(struct sonode *);
909 extern void	soisconnected(struct sonode *);
910 extern void	soisdisconnected(struct sonode *, int);
911 extern void	socantsendmore(struct sonode *);
912 extern void	socantrcvmore(struct sonode *);
913 extern void	soseterror(struct sonode *, int);
914 extern int	sogeterr(struct sonode *, boolean_t);
915 extern int	sowaitconnected(struct sonode *, int, int);
916 
917 extern ssize_t	soreadfile(file_t *, uchar_t *, u_offset_t, int *, size_t);
918 extern void	*sock_kstat_init(zoneid_t);
919 extern void	sock_kstat_fini(zoneid_t, void *);
920 extern struct sonode *getsonode(int, int *, file_t **);
921 /*
922  * Function wrappers (mostly around the sonode switch) for
923  * backward compatibility.
924  */
925 extern int	soaccept(struct sonode *, int, struct sonode **);
926 extern int	sobind(struct sonode *, struct sockaddr *, socklen_t,
927 		    int, int);
928 extern int	solisten(struct sonode *, int);
929 extern int	soconnect(struct sonode *, struct sockaddr *, socklen_t,
930 		    int, int);
931 extern int	sorecvmsg(struct sonode *, struct nmsghdr *, struct uio *);
932 extern int	sosendmsg(struct sonode *, struct nmsghdr *, struct uio *);
933 extern int	soshutdown(struct sonode *, int);
934 extern int	sogetsockopt(struct sonode *, int, int, void *, socklen_t *,
935 		    int);
936 extern int	sosetsockopt(struct sonode *, int, int, const void *,
937 		    t_uscalar_t);
938 
939 extern struct sonode	*socreate(struct sockparams *, int, int, int, int,
940 			    int *);
941 
942 extern int	so_copyin(const void *, void *, size_t, int);
943 extern int	so_copyout(const void *, void *, size_t, int);
944 
945 #endif
946 
947 /*
948  * Internal structure for obtaining sonode information from the socklist.
949  * These types match those corresponding in the sonode structure.
950  * This is not a published interface, and may change at any time.
951  */
952 struct sockinfo {
953 	uint_t		si_size;		/* real length of this struct */
954 	short		si_family;
955 	short		si_type;
956 	ushort_t	si_flag;
957 	uint_t		si_state;
958 	uint_t		si_ux_laddr_sou_magic;
959 	uint_t		si_ux_faddr_sou_magic;
960 	t_scalar_t	si_serv_type;
961 	t_uscalar_t	si_laddr_soa_len;
962 	t_uscalar_t	si_faddr_soa_len;
963 	uint16_t	si_laddr_family;
964 	uint16_t	si_faddr_family;
965 	char		si_laddr_sun_path[MAXPATHLEN + 1]; /* NULL terminated */
966 	char		si_faddr_sun_path[MAXPATHLEN + 1];
967 	boolean_t	si_faddr_noxlate;
968 	zoneid_t	si_szoneid;
969 };
970 
971 /*
972  * Subcodes for sockconf() system call
973  */
974 #define	SOCKCONFIG_ADD_SOCK		0
975 #define	SOCKCONFIG_REMOVE_SOCK		1
976 #define	SOCKCONFIG_ADD_FILTER		2
977 #define	SOCKCONFIG_REMOVE_FILTER	3
978 
979 /*
980  * Data structures for configuring socket filters.
981  */
982 
983 /*
984  * Placement hint for automatic filters
985  */
986 typedef enum {
987 	SOF_HINT_NONE,
988 	SOF_HINT_TOP,
989 	SOF_HINT_BOTTOM,
990 	SOF_HINT_BEFORE,
991 	SOF_HINT_AFTER
992 } sof_hint_t;
993 
994 /*
995  * Socket tuple. Used by sockconfig_filter_props to list socket
996  * types of interest.
997  */
998 typedef struct sof_socktuple {
999 	int	sofst_family;
1000 	int	sofst_type;
1001 	int	sofst_protocol;
1002 } sof_socktuple_t;
1003 
1004 /*
1005  * Socket filter properties used by sockconfig() system call.
1006  */
1007 struct sockconfig_filter_props {
1008 	char		*sfp_modname;
1009 	boolean_t	sfp_autoattach;
1010 	sof_hint_t	sfp_hint;
1011 	char		*sfp_hintarg;
1012 	uint_t		sfp_socktuple_cnt;
1013 	sof_socktuple_t	*sfp_socktuple;
1014 };
1015 
1016 #ifdef	_SYSCALL32
1017 
1018 typedef struct sof_socktuple32 {
1019 	int32_t	sofst_family;
1020 	int32_t	sofst_type;
1021 	int32_t	sofst_protocol;
1022 } sof_socktuple32_t;
1023 
1024 struct sockconfig_filter_props32 {
1025 	caddr32_t	sfp_modname;
1026 	boolean_t	sfp_autoattach;
1027 	sof_hint_t	sfp_hint;
1028 	caddr32_t	sfp_hintarg;
1029 	uint32_t	sfp_socktuple_cnt;
1030 	caddr32_t	sfp_socktuple;
1031 };
1032 
1033 #endif	/* _SYSCALL32 */
1034 
1035 #define	SOCKMOD_PATH	"socketmod"	/* dir where sockmods are stored */
1036 
1037 #ifdef	__cplusplus
1038 }
1039 #endif
1040 
1041 #endif	/* _SYS_SOCKETVAR_H */
1042