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 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
40 */
41
42 #ifndef _SYS_SOCKETVAR_H
43 #define _SYS_SOCKETVAR_H
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 <sys/sdt.h>
54 #include <sys/modctl.h>
55 #include <sys/atomic.h>
56 #include <sys/socket.h>
57 #include <sys/ksocket.h>
58 #include <sys/kstat.h>
59
60 #ifdef _KERNEL
61 #include <sys/vfs_opreg.h>
62 #endif
63
64 #ifdef __cplusplus
65 extern "C" {
66 #endif
67
68 /*
69 * Internal representation of the address used to represent addresses
70 * in the loopback transport for AF_UNIX. While the sockaddr_un is used
71 * as the sockfs layer address for AF_UNIX the pathnames contained in
72 * these addresses are not unique (due to relative pathnames) thus can not
73 * be used in the transport.
74 *
75 * The transport level address consists of a magic number (used to separate the
76 * name space for specific and implicit binds). For a specific bind
77 * this is followed by a "vnode *" which ensures that all specific binds
78 * have a unique transport level address. For implicit binds the latter
79 * part of the address is a byte string (of the same length as a pointer)
80 * that is assigned by the loopback transport.
81 *
82 * The uniqueness assumes that the loopback transport has a separate namespace
83 * for sockets in order to avoid name conflicts with e.g. TLI use of the
84 * same transport.
85 */
86 struct so_ux_addr {
87 void *soua_vp; /* vnode pointer or assigned by tl */
88 uint_t soua_magic; /* See below */
89 };
90
91 #define SOU_MAGIC_EXPLICIT 0x75787670 /* "uxvp" */
92 #define SOU_MAGIC_IMPLICIT 0x616e6f6e /* "anon" */
93
94 struct sockaddr_ux {
95 sa_family_t sou_family; /* AF_UNIX */
96 struct so_ux_addr sou_addr;
97 };
98
99 #if defined(_KERNEL) || defined(_KMEMUSER)
100
101 #include <sys/socket_proto.h>
102
103 typedef struct sonodeops sonodeops_t;
104 typedef struct sonode sonode_t;
105
106 struct sodirect_s;
107
108 /*
109 * The sonode represents a socket. A sonode never exist in the file system
110 * name space and can not be opened using open() - only the socket, socketpair
111 * and accept calls create sonodes.
112 *
113 * The locking of sockfs uses the so_lock mutex plus the SOLOCKED and
114 * SOREADLOCKED flags in so_flag. The mutex protects all the state in the
115 * sonode. It is expected that the underlying transport protocol serializes
116 * socket operations, so sockfs will not normally not single-thread
117 * operations. However, certain sockets, including TPI based ones, can only
118 * handle one control operation at a time. The SOLOCKED flag is used to
119 * single-thread operations from sockfs users to prevent e.g. multiple bind()
120 * calls to operate on the same sonode concurrently. The SOREADLOCKED flag is
121 * used to ensure that only one thread sleeps in kstrgetmsg for a given
122 * sonode. This is needed to ensure atomic operation for things like
123 * MSG_WAITALL.
124 *
125 * The so_fallback_rwlock is used to ensure that for sockets that can
126 * fall back to TPI, the fallback is not initiated until all pending
127 * operations have completed.
128 *
129 * Note that so_lock is sometimes held across calls that might go to sleep
130 * (kmem_alloc and soallocproto*). This implies that no other lock in
131 * the system should be held when calling into sockfs; from the system call
132 * side or from strrput (in case of TPI based sockets). If locks are held
133 * while calling into sockfs the system might hang when running low on memory.
134 */
135 struct sonode {
136 struct vnode *so_vnode; /* vnode associated with this sonode */
137
138 sonodeops_t *so_ops; /* operations vector for this sonode */
139 void *so_priv; /* sonode private data */
140
141 krwlock_t so_fallback_rwlock;
142 kmutex_t so_lock; /* protects sonode fields */
143
144 kcondvar_t so_state_cv; /* synchronize state changes */
145 kcondvar_t so_single_cv; /* wait due to SOLOCKED */
146 kcondvar_t so_read_cv; /* wait due to SOREADLOCKED */
147
148 /* These fields are protected by so_lock */
149
150 uint_t so_state; /* internal state flags SS_*, below */
151 uint_t so_mode; /* characteristics on socket. SM_* */
152 ushort_t so_flag; /* flags, see below */
153 int so_count; /* count of opened references */
154
155 sock_connid_t so_proto_connid; /* protocol generation number */
156
157 ushort_t so_error; /* error affecting connection */
158
159 struct sockparams *so_sockparams; /* vnode or socket module */
160 /* Needed to recreate the same socket for accept */
161 short so_family;
162 short so_type;
163 short so_protocol;
164 short so_version; /* From so_socket call */
165
166 /* Accept queue */
167 kmutex_t so_acceptq_lock; /* protects accept queue */
168 list_t so_acceptq_list; /* pending conns */
169 list_t so_acceptq_defer; /* deferred conns */
170 list_node_t so_acceptq_node; /* acceptq list node */
171 unsigned int so_acceptq_len; /* # of conns (both lists) */
172 unsigned int so_backlog; /* Listen backlog */
173 kcondvar_t so_acceptq_cv; /* wait for new conn. */
174 struct sonode *so_listener; /* parent socket */
175
176 /* Options */
177 short so_options; /* From socket call, see socket.h */
178 struct linger so_linger; /* SO_LINGER value */
179 #define so_sndbuf so_proto_props.sopp_txhiwat /* SO_SNDBUF value */
180 #define so_sndlowat so_proto_props.sopp_txlowat /* tx low water mark */
181 #define so_rcvbuf so_proto_props.sopp_rxhiwat /* SO_RCVBUF value */
182 #define so_rcvlowat so_proto_props.sopp_rxlowat /* rx low water mark */
183 #define so_max_addr_len so_proto_props.sopp_maxaddrlen
184 #define so_minpsz so_proto_props.sopp_minpsz
185 #define so_maxpsz so_proto_props.sopp_maxpsz
186
187 int so_xpg_rcvbuf; /* SO_RCVBUF value for XPG4 socket */
188 clock_t so_sndtimeo; /* send timeout */
189 clock_t so_rcvtimeo; /* recv timeout */
190
191 mblk_t *so_oobmsg; /* outofline oob data */
192 ssize_t so_oobmark; /* offset of the oob data */
193
194 pid_t so_pgrp; /* pgrp for signals */
195
196 cred_t *so_peercred; /* connected socket peer cred */
197 pid_t so_cpid; /* connected socket peer cached pid */
198 zoneid_t so_zoneid; /* opener's zoneid */
199
200 struct pollhead so_poll_list; /* common pollhead */
201 short so_pollev; /* events that should be generated */
202
203 /* Receive */
204 unsigned int so_rcv_queued; /* # bytes on both rcv lists */
205 mblk_t *so_rcv_q_head; /* processing/copyout rcv queue */
206 mblk_t *so_rcv_q_last_head;
207 mblk_t *so_rcv_head; /* protocol prequeue */
208 mblk_t *so_rcv_last_head; /* last mblk in b_next chain */
209 kcondvar_t so_rcv_cv; /* wait for data */
210 uint_t so_rcv_wanted; /* # of bytes wanted by app */
211 timeout_id_t so_rcv_timer_tid;
212
213 #define so_rcv_thresh so_proto_props.sopp_rcvthresh
214 #define so_rcv_timer_interval so_proto_props.sopp_rcvtimer
215
216 kcondvar_t so_snd_cv; /* wait for snd buffers */
217 uint32_t
218 so_snd_qfull: 1, /* Transmit full */
219 so_rcv_wakeup: 1,
220 so_snd_wakeup: 1,
221 so_not_str: 1, /* B_TRUE if not streams based socket */
222 so_pad_to_bit_31: 28;
223
224 /* Communication channel with protocol */
225 sock_lower_handle_t so_proto_handle;
226 sock_downcalls_t *so_downcalls;
227
228 struct sock_proto_props so_proto_props; /* protocol settings */
229 boolean_t so_flowctrld; /* Flow controlled */
230 uint_t so_copyflag; /* Copy related flag */
231 kcondvar_t so_copy_cv; /* Copy cond variable */
232
233 /* kernel sockets */
234 ksocket_callbacks_t so_ksock_callbacks;
235 void *so_ksock_cb_arg; /* callback argument */
236 kcondvar_t so_closing_cv;
237
238 /* != NULL for sodirect enabled socket */
239 struct sodirect_s *so_direct;
240
241 /* socket filters */
242 uint_t so_filter_active; /* # of active fil */
243 uint_t so_filter_tx; /* pending tx ops */
244 struct sof_instance *so_filter_top; /* top of stack */
245 struct sof_instance *so_filter_bottom; /* bottom of stack */
246 clock_t so_filter_defertime; /* time when deferred */
247 };
248
249 #define SO_HAVE_DATA(so) \
250 /* \
251 * For the (tid == 0) case we must check so_rcv_{q_,}head \
252 * rather than (so_rcv_queued > 0), since the latter does not \
253 * take into account mblks with only control/name information. \
254 */ \
255 ((so)->so_rcv_timer_tid == 0 && ((so)->so_rcv_head != NULL || \
256 (so)->so_rcv_q_head != NULL)) || \
257 ((so)->so_state & SS_CANTRCVMORE)
258
259 /*
260 * Events handled by the protocol (in case sd_poll is set)
261 */
262 #define SO_PROTO_POLLEV (POLLIN|POLLRDNORM|POLLRDBAND)
263
264
265 #endif /* _KERNEL || _KMEMUSER */
266
267 /* flags */
268 #define SOMOD 0x0001 /* update socket modification time */
269 #define SOACC 0x0002 /* update socket access time */
270
271 #define SOLOCKED 0x0010 /* use to serialize open/closes */
272 #define SOREADLOCKED 0x0020 /* serialize kstrgetmsg calls */
273 #define SOCLONE 0x0040 /* child of clone driver */
274 #define SOASYNC_UNBIND 0x0080 /* wait for ACK of async unbind */
275
276 #define SOCK_IS_NONSTR(so) ((so)->so_not_str)
277
278 /*
279 * Socket state bits.
280 */
281 #define SS_ISCONNECTED 0x00000001 /* socket connected to a peer */
282 #define SS_ISCONNECTING 0x00000002 /* in process, connecting to peer */
283 #define SS_ISDISCONNECTING 0x00000004 /* in process of disconnecting */
284 #define SS_CANTSENDMORE 0x00000008 /* can't send more data to peer */
285
286 #define SS_CANTRCVMORE 0x00000010 /* can't receive more data */
287 #define SS_ISBOUND 0x00000020 /* socket is bound */
288 #define SS_NDELAY 0x00000040 /* FNDELAY non-blocking */
289 #define SS_NONBLOCK 0x00000080 /* O_NONBLOCK non-blocking */
290
291 #define SS_ASYNC 0x00000100 /* async i/o notify */
292 #define SS_ACCEPTCONN 0x00000200 /* listen done */
293 /* unused 0x00000400 */ /* was SS_HASCONNIND */
294 #define SS_SAVEDEOR 0x00000800 /* Saved MSG_EOR rcv side state */
295
296 #define SS_RCVATMARK 0x00001000 /* at mark on input */
297 #define SS_OOBPEND 0x00002000 /* OOB pending or present - poll */
298 #define SS_HAVEOOBDATA 0x00004000 /* OOB data present */
299 #define SS_HADOOBDATA 0x00008000 /* OOB data consumed */
300 #define SS_CLOSING 0x00010000 /* in process of closing */
301
302 #define SS_FIL_DEFER 0x00020000 /* filter deferred notification */
303 #define SS_FILOP_OK 0x00040000 /* socket can attach filters */
304 #define SS_FIL_RCV_FLOWCTRL 0x00080000 /* filter asserted rcv flow ctrl */
305 #define SS_FIL_SND_FLOWCTRL 0x00100000 /* filter asserted snd flow ctrl */
306 #define SS_FIL_STOP 0x00200000 /* no more filter actions */
307
308 #define SS_SODIRECT 0x00400000 /* transport supports sodirect */
309
310 #define SS_SENTLASTREADSIG 0x01000000 /* last rx signal has been sent */
311 #define SS_SENTLASTWRITESIG 0x02000000 /* last tx signal has been sent */
312
313 #define SS_FALLBACK_DRAIN 0x20000000 /* data was/is being drained */
314 #define SS_FALLBACK_PENDING 0x40000000 /* fallback is pending */
315 #define SS_FALLBACK_COMP 0x80000000 /* fallback has completed */
316
317
318 /* Set of states when the socket can't be rebound */
319 #define SS_CANTREBIND (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING|\
320 SS_CANTSENDMORE|SS_CANTRCVMORE|SS_ACCEPTCONN)
321
322 /*
323 * Sockets that can fall back to TPI must ensure that fall back is not
324 * initiated while a thread is using a socket.
325 */
326 #define SO_BLOCK_FALLBACK(so, fn) \
327 ASSERT(MUTEX_NOT_HELD(&(so)->so_lock)); \
328 rw_enter(&(so)->so_fallback_rwlock, RW_READER); \
329 if ((so)->so_state & (SS_FALLBACK_COMP|SS_FILOP_OK)) { \
330 if ((so)->so_state & SS_FALLBACK_COMP) { \
331 rw_exit(&(so)->so_fallback_rwlock); \
332 return (fn); \
333 } else { \
334 mutex_enter(&(so)->so_lock); \
335 (so)->so_state &= ~SS_FILOP_OK; \
336 mutex_exit(&(so)->so_lock); \
337 } \
338 }
339
340 #define SO_UNBLOCK_FALLBACK(so) { \
341 rw_exit(&(so)->so_fallback_rwlock); \
342 }
343
344 #define SO_SND_FLOWCTRLD(so) \
345 ((so)->so_snd_qfull || (so)->so_state & SS_FIL_SND_FLOWCTRL)
346
347 /* Poll events */
348 #define SO_POLLEV_IN 0x1 /* POLLIN wakeup needed */
349 #define SO_POLLEV_ALWAYS 0x2 /* wakeups */
350
351 /*
352 * Characteristics of sockets. Not changed after the socket is created.
353 */
354 #define SM_PRIV 0x001 /* privileged for broadcast, raw... */
355 #define SM_ATOMIC 0x002 /* atomic data transmission */
356 #define SM_ADDR 0x004 /* addresses given with messages */
357 #define SM_CONNREQUIRED 0x008 /* connection required by protocol */
358
359 #define SM_FDPASSING 0x010 /* passes file descriptors */
360 #define SM_EXDATA 0x020 /* Can handle T_EXDATA_REQ */
361 #define SM_OPTDATA 0x040 /* Can handle T_OPTDATA_REQ */
362 #define SM_BYTESTREAM 0x080 /* Byte stream - can use M_DATA */
363
364 #define SM_ACCEPTOR_ID 0x100 /* so_acceptor_id is valid */
365
366 #define SM_KERNEL 0x200 /* kernel socket */
367
368 /* The modes below are only for non-streams sockets */
369 #define SM_ACCEPTSUPP 0x400 /* can handle accept() */
370 #define SM_SENDFILESUPP 0x800 /* Private: proto supp sendfile */
371
372 /*
373 * Socket versions. Used by the socket library when calling _so_socket().
374 */
375 #define SOV_STREAM 0 /* Not a socket - just a stream */
376 #define SOV_DEFAULT 1 /* Select based on so_default_version */
377 #define SOV_SOCKSTREAM 2 /* Socket plus streams operations */
378 #define SOV_SOCKBSD 3 /* Socket with no streams operations */
379 #define SOV_XPG4_2 4 /* Xnet socket */
380
381 #if defined(_KERNEL) || defined(_KMEMUSER)
382
383 /*
384 * sonode create and destroy functions.
385 */
386 typedef struct sonode *(*so_create_func_t)(struct sockparams *,
387 int, int, int, int, int, int *, cred_t *);
388 typedef void (*so_destroy_func_t)(struct sonode *);
389
390 /* STREAM device information */
391 typedef struct sdev_info {
392 char *sd_devpath;
393 int sd_devpathlen; /* Is 0 if sp_devpath is a static string */
394 vnode_t *sd_vnode;
395 } sdev_info_t;
396
397 #define SOCKMOD_VERSION_1 1
398 #define SOCKMOD_VERSION 2
399
400 /* name of the TPI pseudo socket module */
401 #define SOTPI_SMOD_NAME "socktpi"
402
403 typedef struct __smod_priv_s {
404 so_create_func_t smodp_sock_create_func;
405 so_destroy_func_t smodp_sock_destroy_func;
406 so_proto_fallback_func_t smodp_proto_fallback_func;
407 const char *smodp_fallback_devpath_v4;
408 const char *smodp_fallback_devpath_v6;
409 } __smod_priv_t;
410
411 /*
412 * Socket module register information
413 */
414 typedef struct smod_reg_s {
415 int smod_version;
416 char *smod_name;
417 size_t smod_uc_version;
418 size_t smod_dc_version;
419 so_proto_create_func_t smod_proto_create_func;
420
421 /* __smod_priv_data must be NULL */
422 __smod_priv_t *__smod_priv;
423 } smod_reg_t;
424
425 /*
426 * Socket module information
427 */
428 typedef struct smod_info {
429 int smod_version;
430 char *smod_name;
431 uint_t smod_refcnt; /* # of entries */
432 size_t smod_uc_version; /* upcall version */
433 size_t smod_dc_version; /* down call version */
434 so_proto_create_func_t smod_proto_create_func;
435 so_proto_fallback_func_t smod_proto_fallback_func;
436 const char *smod_fallback_devpath_v4;
437 const char *smod_fallback_devpath_v6;
438 so_create_func_t smod_sock_create_func;
439 so_destroy_func_t smod_sock_destroy_func;
440 list_node_t smod_node;
441 } smod_info_t;
442
443 typedef struct sockparams_stats {
444 kstat_named_t sps_nfallback; /* # of fallbacks to TPI */
445 kstat_named_t sps_nactive; /* # of active sockets */
446 kstat_named_t sps_ncreate; /* total # of created sockets */
447 } sockparams_stats_t;
448
449 /*
450 * sockparams
451 *
452 * Used for mapping family/type/protocol to a socket module or STREAMS device
453 */
454 struct sockparams {
455 /*
456 * The family, type, protocol, sdev_info and smod_name are
457 * set when the entry is created, and they will never change
458 * thereafter.
459 */
460 int sp_family;
461 int sp_type;
462 int sp_protocol;
463
464 sdev_info_t sp_sdev_info; /* STREAM device */
465 char *sp_smod_name; /* socket module name */
466
467 kmutex_t sp_lock; /* lock for refcnt and smod_info */
468 uint64_t sp_refcnt; /* entry reference count */
469 smod_info_t *sp_smod_info; /* socket module */
470
471 sockparams_stats_t sp_stats;
472 kstat_t *sp_kstat;
473
474 /*
475 * The entries below are only modified while holding
476 * sockconf_lock as a writer.
477 */
478 int sp_flags; /* see below */
479 list_node_t sp_node;
480
481 list_t sp_auto_filters; /* list of automatic filters */
482 list_t sp_prog_filters; /* list of programmatic filters */
483 };
484
485 struct sof_entry;
486
487 typedef struct sp_filter {
488 struct sof_entry *spf_filter;
489 list_node_t spf_node;
490 } sp_filter_t;
491
492
493 /*
494 * sockparams flags
495 */
496 #define SOCKPARAMS_EPHEMERAL 0x1 /* temp. entry, not on global list */
497
498 extern void sockparams_init(void);
499 extern struct sockparams *sockparams_hold_ephemeral_bydev(int, int, int,
500 const char *, int, int *);
501 extern struct sockparams *sockparams_hold_ephemeral_bymod(int, int, int,
502 const char *, int, int *);
503 extern void sockparams_ephemeral_drop_last_ref(struct sockparams *);
504
505 extern struct sockparams *sockparams_create(int, int, int, char *, char *, int,
506 int, int, int *);
507 extern void sockparams_destroy(struct sockparams *);
508 extern int sockparams_add(struct sockparams *);
509 extern int sockparams_delete(int, int, int);
510 extern int sockparams_new_filter(struct sof_entry *);
511 extern void sockparams_filter_cleanup(struct sof_entry *);
512 extern int sockparams_copyout_socktable(uintptr_t);
513
514 extern void smod_init(void);
515 extern void smod_add(smod_info_t *);
516 extern int smod_register(const smod_reg_t *);
517 extern int smod_unregister(const char *);
518 extern smod_info_t *smod_lookup_byname(const char *);
519
520 #define SOCKPARAMS_HAS_DEVICE(sp) \
521 ((sp)->sp_sdev_info.sd_devpath != NULL)
522
523 /* Increase the smod_info_t reference count */
524 #define SMOD_INC_REF(smodp) { \
525 ASSERT((smodp) != NULL); \
526 DTRACE_PROBE1(smodinfo__inc__ref, struct smod_info *, (smodp)); \
527 atomic_inc_uint(&(smodp)->smod_refcnt); \
528 }
529
530 /*
531 * Decreace the socket module entry reference count.
532 * When no one mapping to the entry, we try to unload the module from the
533 * kernel. If the module can't unload, just leave the module entry with
534 * a zero refcnt.
535 */
536 #define SMOD_DEC_REF(smodp, modname) { \
537 ASSERT((smodp) != NULL); \
538 ASSERT((smodp)->smod_refcnt != 0); \
539 atomic_dec_uint(&(smodp)->smod_refcnt); \
540 /* \
541 * No need to atomically check the return value because the \
542 * socket module framework will verify that no one is using \
543 * the module before unloading. Worst thing that can happen \
544 * here is multiple calls to mod_remove_by_name(), which is OK. \
545 */ \
546 if ((smodp)->smod_refcnt == 0) \
547 (void) mod_remove_by_name(modname); \
548 }
549
550 /* Increase the reference count */
551 #define SOCKPARAMS_INC_REF(sp) { \
552 ASSERT((sp) != NULL); \
553 DTRACE_PROBE1(sockparams__inc__ref, struct sockparams *, (sp)); \
554 mutex_enter(&(sp)->sp_lock); \
555 (sp)->sp_refcnt++; \
556 ASSERT((sp)->sp_refcnt != 0); \
557 mutex_exit(&(sp)->sp_lock); \
558 }
559
560 /*
561 * Decrease the reference count.
562 *
563 * If the sockparams is ephemeral, then the thread dropping the last ref
564 * count will destroy the entry.
565 */
566 #define SOCKPARAMS_DEC_REF(sp) { \
567 ASSERT((sp) != NULL); \
568 DTRACE_PROBE1(sockparams__dec__ref, struct sockparams *, (sp)); \
569 mutex_enter(&(sp)->sp_lock); \
570 ASSERT((sp)->sp_refcnt > 0); \
571 if ((sp)->sp_refcnt == 1) { \
572 if ((sp)->sp_flags & SOCKPARAMS_EPHEMERAL) { \
573 mutex_exit(&(sp)->sp_lock); \
574 sockparams_ephemeral_drop_last_ref((sp)); \
575 } else { \
576 (sp)->sp_refcnt--; \
577 if ((sp)->sp_smod_info != NULL) { \
578 SMOD_DEC_REF((sp)->sp_smod_info, \
579 (sp)->sp_smod_name); \
580 } \
581 (sp)->sp_smod_info = NULL; \
582 mutex_exit(&(sp)->sp_lock); \
583 } \
584 } else { \
585 (sp)->sp_refcnt--; \
586 mutex_exit(&(sp)->sp_lock); \
587 } \
588 }
589
590 /*
591 * Used to traverse the list of AF_UNIX sockets to construct the kstat
592 * for netstat(1m).
593 */
594 struct socklist {
595 kmutex_t sl_lock;
596 struct sonode *sl_list;
597 };
598
599 extern struct socklist socklist;
600 /*
601 * ss_full_waits is the number of times the reader thread
602 * waits when the queue is full and ss_empty_waits is the number
603 * of times the consumer thread waits when the queue is empty.
604 * No locks for these as they are just indicators of whether
605 * disk or network or both is slow or fast.
606 */
607 struct sendfile_stats {
608 uint32_t ss_file_cached;
609 uint32_t ss_file_not_cached;
610 uint32_t ss_full_waits;
611 uint32_t ss_empty_waits;
612 uint32_t ss_file_segmap;
613 };
614
615 /*
616 * A single sendfile request is represented by snf_req.
617 */
618 typedef struct snf_req {
619 struct snf_req *sr_next;
620 mblk_t *sr_mp_head;
621 mblk_t *sr_mp_tail;
622 kmutex_t sr_lock;
623 kcondvar_t sr_cv;
624 uint_t sr_qlen;
625 int sr_hiwat;
626 int sr_lowat;
627 int sr_operation;
628 struct vnode *sr_vp;
629 file_t *sr_fp;
630 ssize_t sr_maxpsz;
631 u_offset_t sr_file_off;
632 u_offset_t sr_file_size;
633 #define SR_READ_DONE 0x80000000
634 int sr_read_error;
635 int sr_write_error;
636 } snf_req_t;
637
638 /* A queue of sendfile requests */
639 struct sendfile_queue {
640 snf_req_t *snfq_req_head;
641 snf_req_t *snfq_req_tail;
642 kmutex_t snfq_lock;
643 kcondvar_t snfq_cv;
644 int snfq_svc_threads; /* # of service threads */
645 int snfq_idle_cnt; /* # of idling threads */
646 int snfq_max_threads;
647 int snfq_req_cnt; /* Number of requests */
648 };
649
650 #define READ_OP 1
651 #define SNFQ_TIMEOUT (60 * 5 * hz) /* 5 minutes */
652
653 /* Socket network operations switch */
654 struct sonodeops {
655 int (*sop_init)(struct sonode *, struct sonode *, cred_t *,
656 int);
657 int (*sop_accept)(struct sonode *, int, cred_t *, struct sonode **);
658 int (*sop_bind)(struct sonode *, struct sockaddr *, socklen_t,
659 int, cred_t *);
660 int (*sop_listen)(struct sonode *, int, cred_t *);
661 int (*sop_connect)(struct sonode *, struct sockaddr *,
662 socklen_t, int, int, cred_t *);
663 int (*sop_recvmsg)(struct sonode *, struct msghdr *,
664 struct uio *, cred_t *);
665 int (*sop_sendmsg)(struct sonode *, struct msghdr *,
666 struct uio *, cred_t *);
667 int (*sop_sendmblk)(struct sonode *, struct msghdr *, int,
668 cred_t *, mblk_t **);
669 int (*sop_getpeername)(struct sonode *, struct sockaddr *,
670 socklen_t *, boolean_t, cred_t *);
671 int (*sop_getsockname)(struct sonode *, struct sockaddr *,
672 socklen_t *, cred_t *);
673 int (*sop_shutdown)(struct sonode *, int, cred_t *);
674 int (*sop_getsockopt)(struct sonode *, int, int, void *,
675 socklen_t *, int, cred_t *);
676 int (*sop_setsockopt)(struct sonode *, int, int, const void *,
677 socklen_t, cred_t *);
678 int (*sop_ioctl)(struct sonode *, int, intptr_t, int,
679 cred_t *, int32_t *);
680 int (*sop_poll)(struct sonode *, short, int, short *,
681 struct pollhead **);
682 int (*sop_close)(struct sonode *, int, cred_t *);
683 };
684
685 #define SOP_INIT(so, flag, cr, flags) \
686 ((so)->so_ops->sop_init((so), (flag), (cr), (flags)))
687 #define SOP_ACCEPT(so, fflag, cr, nsop) \
688 ((so)->so_ops->sop_accept((so), (fflag), (cr), (nsop)))
689 #define SOP_BIND(so, name, namelen, flags, cr) \
690 ((so)->so_ops->sop_bind((so), (name), (namelen), (flags), (cr)))
691 #define SOP_LISTEN(so, backlog, cr) \
692 ((so)->so_ops->sop_listen((so), (backlog), (cr)))
693 #define SOP_CONNECT(so, name, namelen, fflag, flags, cr) \
694 ((so)->so_ops->sop_connect((so), (name), (namelen), (fflag), (flags), \
695 (cr)))
696 #define SOP_RECVMSG(so, msg, uiop, cr) \
697 ((so)->so_ops->sop_recvmsg((so), (msg), (uiop), (cr)))
698 #define SOP_SENDMSG(so, msg, uiop, cr) \
699 ((so)->so_ops->sop_sendmsg((so), (msg), (uiop), (cr)))
700 #define SOP_SENDMBLK(so, msg, size, cr, mpp) \
701 ((so)->so_ops->sop_sendmblk((so), (msg), (size), (cr), (mpp)))
702 #define SOP_GETPEERNAME(so, addr, addrlen, accept, cr) \
703 ((so)->so_ops->sop_getpeername((so), (addr), (addrlen), (accept), (cr)))
704 #define SOP_GETSOCKNAME(so, addr, addrlen, cr) \
705 ((so)->so_ops->sop_getsockname((so), (addr), (addrlen), (cr)))
706 #define SOP_SHUTDOWN(so, how, cr) \
707 ((so)->so_ops->sop_shutdown((so), (how), (cr)))
708 #define SOP_GETSOCKOPT(so, level, optionname, optval, optlenp, flags, cr) \
709 ((so)->so_ops->sop_getsockopt((so), (level), (optionname), \
710 (optval), (optlenp), (flags), (cr)))
711 #define SOP_SETSOCKOPT(so, level, optionname, optval, optlen, cr) \
712 ((so)->so_ops->sop_setsockopt((so), (level), (optionname), \
713 (optval), (optlen), (cr)))
714 #define SOP_IOCTL(so, cmd, arg, mode, cr, rvalp) \
715 ((so)->so_ops->sop_ioctl((so), (cmd), (arg), (mode), (cr), (rvalp)))
716 #define SOP_POLL(so, events, anyyet, reventsp, phpp) \
717 ((so)->so_ops->sop_poll((so), (events), (anyyet), (reventsp), (phpp)))
718 #define SOP_CLOSE(so, flag, cr) \
719 ((so)->so_ops->sop_close((so), (flag), (cr)))
720
721 #endif /* defined(_KERNEL) || defined(_KMEMUSER) */
722
723 #ifdef _KERNEL
724
725 #define ISALIGNED_cmsghdr(addr) \
726 (((uintptr_t)(addr) & (_CMSG_HDR_ALIGNMENT - 1)) == 0)
727
728 #define ROUNDUP_cmsglen(len) \
729 (((len) + _CMSG_HDR_ALIGNMENT - 1) & ~(_CMSG_HDR_ALIGNMENT - 1))
730
731 #define IS_NON_STREAM_SOCK(vp) \
732 ((vp)->v_type == VSOCK && (vp)->v_stream == NULL)
733 /*
734 * Macros that operate on struct cmsghdr.
735 * Used in parsing msg_control.
736 * The CMSG_VALID macro does not assume that the last option buffer is padded.
737 */
738 #define CMSG_NEXT(cmsg) \
739 (struct cmsghdr *)((uintptr_t)(cmsg) + \
740 ROUNDUP_cmsglen((cmsg)->cmsg_len))
741 #define CMSG_CONTENT(cmsg) (&((cmsg)[1]))
742 #define CMSG_CONTENTLEN(cmsg) ((cmsg)->cmsg_len - sizeof (struct cmsghdr))
743 #define CMSG_VALID(cmsg, start, end) \
744 (ISALIGNED_cmsghdr(cmsg) && \
745 ((uintptr_t)(cmsg) >= (uintptr_t)(start)) && \
746 ((uintptr_t)(cmsg) < (uintptr_t)(end)) && \
747 ((ssize_t)(cmsg)->cmsg_len >= sizeof (struct cmsghdr)) && \
748 ((uintptr_t)(cmsg) + (cmsg)->cmsg_len <= (uintptr_t)(end)))
749
750 /*
751 * Maximum size of any argument that is copied in (addresses, options,
752 * access rights). MUST be at least MAXPATHLEN + 3.
753 * BSD and SunOS 4.X limited this to MLEN or MCLBYTES.
754 */
755 #define SO_MAXARGSIZE 8192
756
757 /*
758 * Convert between vnode and sonode
759 */
760 #define VTOSO(vp) ((struct sonode *)((vp)->v_data))
761 #define SOTOV(sp) ((sp)->so_vnode)
762
763 /*
764 * Internal flags for sobind()
765 */
766 #define _SOBIND_REBIND 0x01 /* Bind to existing local address */
767 #define _SOBIND_UNSPEC 0x02 /* Bind to unspecified address */
768 #define _SOBIND_LOCK_HELD 0x04 /* so_excl_lock held by caller */
769 #define _SOBIND_NOXLATE 0x08 /* No addr translation for AF_UNIX */
770 #define _SOBIND_XPG4_2 0x10 /* xpg4.2 semantics */
771 #define _SOBIND_SOCKBSD 0x20 /* BSD semantics */
772 #define _SOBIND_LISTEN 0x40 /* Make into SS_ACCEPTCONN */
773 #define _SOBIND_SOCKETPAIR 0x80 /* Internal flag for so_socketpair() */
774 /* to enable listen with backlog = 1 */
775
776 /*
777 * Internal flags for sounbind()
778 */
779 #define _SOUNBIND_REBIND 0x01 /* Don't clear fields - will rebind */
780
781 /*
782 * Internal flags for soconnect()
783 */
784 #define _SOCONNECT_NOXLATE 0x01 /* No addr translation for AF_UNIX */
785 #define _SOCONNECT_DID_BIND 0x02 /* Unbind when connect fails */
786 #define _SOCONNECT_XPG4_2 0x04 /* xpg4.2 semantics */
787
788 /*
789 * Internal flags for sodisconnect()
790 */
791 #define _SODISCONNECT_LOCK_HELD 0x01 /* so_excl_lock held by caller */
792
793 /*
794 * Internal flags for sotpi_getsockopt().
795 */
796 #define _SOGETSOCKOPT_XPG4_2 0x01 /* xpg4.2 semantics */
797
798 /*
799 * Internal flags for soallocproto*()
800 */
801 #define _ALLOC_NOSLEEP 0 /* Don't sleep for memory */
802 #define _ALLOC_INTR 1 /* Sleep until interrupt */
803 #define _ALLOC_SLEEP 2 /* Sleep forever */
804
805 /*
806 * Internal structure for handling AF_UNIX file descriptor passing
807 */
808 struct fdbuf {
809 int fd_size; /* In bytes, for kmem_free */
810 int fd_numfd; /* Number of elements below */
811 char *fd_ebuf; /* Extra buffer to free */
812 int fd_ebuflen;
813 frtn_t fd_frtn;
814 struct file *fd_fds[1]; /* One or more */
815 };
816 #define FDBUF_HDRSIZE (sizeof (struct fdbuf) - sizeof (struct file *))
817
818 /*
819 * Variable that can be patched to set what version of socket socket()
820 * will create.
821 */
822 extern int so_default_version;
823
824 #ifdef DEBUG
825 /* Turn on extra testing capabilities */
826 #define SOCK_TEST
827 #endif /* DEBUG */
828
829 #ifdef DEBUG
830 char *pr_state(uint_t, uint_t);
831 char *pr_addr(int, struct sockaddr *, t_uscalar_t);
832 int so_verify_oobstate(struct sonode *);
833 #endif /* DEBUG */
834
835 /*
836 * DEBUG macros
837 */
838 #if defined(DEBUG)
839 #define SOCK_DEBUG
840
841 extern int sockdebug;
842 extern int sockprinterr;
843
844 #define eprint(args) printf args
845 #define eprintso(so, args) \
846 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) printf args; }
847 #define eprintline(error) \
848 { \
849 if (error != EINTR && (sockprinterr || sockdebug > 0)) \
850 printf("socket error %d: line %d file %s\n", \
851 (error), __LINE__, __FILE__); \
852 }
853
854 #define eprintsoline(so, error) \
855 { if (sockprinterr && ((so)->so_options & SO_DEBUG)) \
856 printf("socket(%p) error %d: line %d file %s\n", \
857 (void *)(so), (error), __LINE__, __FILE__); \
858 }
859 #define dprint(level, args) { if (sockdebug > (level)) printf args; }
860 #define dprintso(so, level, args) \
861 { if (sockdebug > (level) && ((so)->so_options & SO_DEBUG)) printf args; }
862
863 #else /* define(DEBUG) */
864
865 #define eprint(args) {}
866 #define eprintso(so, args) {}
867 #define eprintline(error) {}
868 #define eprintsoline(so, error) {}
869 #define dprint(level, args) {}
870 #define dprintso(so, level, args) {}
871
872 #endif /* defined(DEBUG) */
873
874 extern struct vfsops sock_vfsops;
875 extern struct vnodeops *socket_vnodeops;
876 extern const struct fs_operation_def socket_vnodeops_template[];
877
878 extern dev_t sockdev;
879
880 extern krwlock_t sockconf_lock;
881
882 /*
883 * sockfs functions
884 */
885 extern int sock_getmsg(vnode_t *, struct strbuf *, struct strbuf *,
886 uchar_t *, int *, int, rval_t *);
887 extern int sock_putmsg(vnode_t *, struct strbuf *, struct strbuf *,
888 uchar_t, int, int);
889 extern int sogetvp(char *, vnode_t **, int);
890 extern int sockinit(int, char *);
891 extern int solookup(int, int, int, struct sockparams **);
892 extern void so_lock_single(struct sonode *);
893 extern void so_unlock_single(struct sonode *, int);
894 extern int so_lock_read(struct sonode *, int);
895 extern int so_lock_read_intr(struct sonode *, int);
896 extern void so_unlock_read(struct sonode *);
897 extern void *sogetoff(mblk_t *, t_uscalar_t, t_uscalar_t, uint_t);
898 extern void so_getopt_srcaddr(void *, t_uscalar_t,
899 void **, t_uscalar_t *);
900 extern int so_getopt_unix_close(void *, t_uscalar_t);
901 extern void fdbuf_free(struct fdbuf *);
902 extern mblk_t *fdbuf_allocmsg(int, struct fdbuf *);
903 extern int fdbuf_create(void *, int, struct fdbuf **);
904 extern void so_closefds(void *, t_uscalar_t, int, int);
905 extern int so_getfdopt(void *, t_uscalar_t, int, void **, int *);
906 t_uscalar_t so_optlen(void *, t_uscalar_t, int);
907 extern void so_cmsg2opt(void *, t_uscalar_t, int, mblk_t *);
908 extern t_uscalar_t
909 so_cmsglen(mblk_t *, void *, t_uscalar_t, int);
910 extern int so_opt2cmsg(mblk_t *, void *, t_uscalar_t, int,
911 void *, t_uscalar_t);
912 extern void soisconnecting(struct sonode *);
913 extern void soisconnected(struct sonode *);
914 extern void soisdisconnected(struct sonode *, int);
915 extern void socantsendmore(struct sonode *);
916 extern void socantrcvmore(struct sonode *);
917 extern void soseterror(struct sonode *, int);
918 extern int sogeterr(struct sonode *, boolean_t);
919 extern int sowaitconnected(struct sonode *, int, int);
920
921 extern ssize_t soreadfile(file_t *, uchar_t *, u_offset_t, int *, size_t);
922 extern void *sock_kstat_init(zoneid_t);
923 extern void sock_kstat_fini(zoneid_t, void *);
924 extern struct sonode *getsonode(int, int *, file_t **);
925 /*
926 * Function wrappers (mostly around the sonode switch) for
927 * backward compatibility.
928 */
929 extern int soaccept(struct sonode *, int, struct sonode **);
930 extern int sobind(struct sonode *, struct sockaddr *, socklen_t,
931 int, int);
932 extern int solisten(struct sonode *, int);
933 extern int soconnect(struct sonode *, struct sockaddr *, socklen_t,
934 int, int);
935 extern int sorecvmsg(struct sonode *, struct nmsghdr *, struct uio *);
936 extern int sosendmsg(struct sonode *, struct nmsghdr *, struct uio *);
937 extern int soshutdown(struct sonode *, int);
938 extern int sogetsockopt(struct sonode *, int, int, void *, socklen_t *,
939 int);
940 extern int sosetsockopt(struct sonode *, int, int, const void *,
941 t_uscalar_t);
942
943 extern struct sonode *socreate(struct sockparams *, int, int, int, int,
944 int *);
945
946 extern int so_copyin(const void *, void *, size_t, int);
947 extern int so_copyout(const void *, void *, size_t, int);
948
949 #endif
950
951 /*
952 * Internal structure for obtaining sonode information from the socklist.
953 * These types match those corresponding in the sonode structure.
954 * This is not a published interface, and may change at any time.
955 */
956 struct sockinfo {
957 uint_t si_size; /* real length of this struct */
958 short si_family;
959 short si_type;
960 ushort_t si_flag;
961 uint_t si_state;
962 uint_t si_ux_laddr_sou_magic;
963 uint_t si_ux_faddr_sou_magic;
964 t_scalar_t si_serv_type;
965 t_uscalar_t si_laddr_soa_len;
966 t_uscalar_t si_faddr_soa_len;
967 uint16_t si_laddr_family;
968 uint16_t si_faddr_family;
969 char si_laddr_sun_path[MAXPATHLEN + 1]; /* NULL terminated */
970 char si_faddr_sun_path[MAXPATHLEN + 1];
971 boolean_t si_faddr_noxlate;
972 zoneid_t si_szoneid;
973 };
974
975 /*
976 * Subcodes for sockconf() system call
977 */
978 #define SOCKCONFIG_ADD_SOCK 0
979 #define SOCKCONFIG_REMOVE_SOCK 1
980 #define SOCKCONFIG_ADD_FILTER 2
981 #define SOCKCONFIG_REMOVE_FILTER 3
982 #define SOCKCONFIG_GET_SOCKTABLE 4
983
984 /*
985 * Data structures for configuring socket filters.
986 */
987
988 /*
989 * Placement hint for automatic filters
990 */
991 typedef enum {
992 SOF_HINT_NONE,
993 SOF_HINT_TOP,
994 SOF_HINT_BOTTOM,
995 SOF_HINT_BEFORE,
996 SOF_HINT_AFTER
997 } sof_hint_t;
998
999 /*
1000 * Socket tuple. Used by sockconfig_filter_props to list socket
1001 * types of interest.
1002 */
1003 typedef struct sof_socktuple {
1004 int sofst_family;
1005 int sofst_type;
1006 int sofst_protocol;
1007 } sof_socktuple_t;
1008
1009 /*
1010 * Socket filter properties used by sockconfig() system call.
1011 */
1012 struct sockconfig_filter_props {
1013 char *sfp_modname;
1014 boolean_t sfp_autoattach;
1015 sof_hint_t sfp_hint;
1016 char *sfp_hintarg;
1017 uint_t sfp_socktuple_cnt;
1018 sof_socktuple_t *sfp_socktuple;
1019 };
1020
1021 /*
1022 * Data structures for the in-kernel socket configuration table.
1023 */
1024 typedef struct sockconfig_socktable_entry {
1025 int se_family;
1026 int se_type;
1027 int se_protocol;
1028 int se_refcnt;
1029 int se_flags;
1030 char se_modname[MODMAXNAMELEN];
1031 char se_strdev[MAXPATHLEN];
1032 } sockconfig_socktable_entry_t;
1033
1034 typedef struct sockconfig_socktable {
1035 uint_t num_of_entries;
1036 sockconfig_socktable_entry_t *st_entries;
1037 } sockconfig_socktable_t;
1038
1039 #ifdef _SYSCALL32
1040
1041 typedef struct sof_socktuple32 {
1042 int32_t sofst_family;
1043 int32_t sofst_type;
1044 int32_t sofst_protocol;
1045 } sof_socktuple32_t;
1046
1047 struct sockconfig_filter_props32 {
1048 caddr32_t sfp_modname;
1049 boolean_t sfp_autoattach;
1050 sof_hint_t sfp_hint;
1051 caddr32_t sfp_hintarg;
1052 uint32_t sfp_socktuple_cnt;
1053 caddr32_t sfp_socktuple;
1054 };
1055
1056 typedef struct sockconfig_socktable32 {
1057 uint_t num_of_entries;
1058 caddr32_t st_entries;
1059 } sockconfig_socktable32_t;
1060
1061 #endif /* _SYSCALL32 */
1062
1063 #define SOCKMOD_PATH "socketmod" /* dir where sockmods are stored */
1064
1065 #ifdef __cplusplus
1066 }
1067 #endif
1068
1069 #endif /* _SYS_SOCKETVAR_H */