2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_param.h"
37 #include <sys/param.h>
38 #include <sys/aio.h> /* for aio_swake proto */
39 #include <sys/kernel.h>
42 #include <sys/mutex.h>
44 #include <sys/protosw.h>
45 #include <sys/resourcevar.h>
46 #include <sys/signalvar.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
50 #include <sys/sysctl.h>
53 * Function pointer set by the AIO routines so that the socket buffer code
54 * can call back into the AIO module if it is loaded.
56 void (*aio_swake)(struct socket *, struct sockbuf *);
59 * Primitive routines for operating on socket buffers
62 u_long sb_max = SB_MAX;
64 SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
66 static u_long sb_efficiency = 8; /* parameter for sbreserve() */
68 static void sbdrop_internal(struct sockbuf *sb, int len);
69 static void sbflush_internal(struct sockbuf *sb);
72 * Socantsendmore indicates that no more data will be sent on the socket; it
73 * would normally be applied to a socket when the user informs the system
74 * that no more data is to be sent, by the protocol code (in case
75 * PRU_SHUTDOWN). Socantrcvmore indicates that no more data will be
76 * received, and will normally be applied to the socket by a protocol when it
77 * detects that the peer will send no more data. Data queued for reading in
78 * the socket may yet be read.
81 socantsendmore_locked(struct socket *so)
84 SOCKBUF_LOCK_ASSERT(&so->so_snd);
86 so->so_snd.sb_state |= SBS_CANTSENDMORE;
88 mtx_assert(SOCKBUF_MTX(&so->so_snd), MA_NOTOWNED);
92 socantsendmore(struct socket *so)
95 SOCKBUF_LOCK(&so->so_snd);
96 socantsendmore_locked(so);
97 mtx_assert(SOCKBUF_MTX(&so->so_snd), MA_NOTOWNED);
101 socantrcvmore_locked(struct socket *so)
104 SOCKBUF_LOCK_ASSERT(&so->so_rcv);
106 so->so_rcv.sb_state |= SBS_CANTRCVMORE;
107 sorwakeup_locked(so);
108 mtx_assert(SOCKBUF_MTX(&so->so_rcv), MA_NOTOWNED);
112 socantrcvmore(struct socket *so)
115 SOCKBUF_LOCK(&so->so_rcv);
116 socantrcvmore_locked(so);
117 mtx_assert(SOCKBUF_MTX(&so->so_rcv), MA_NOTOWNED);
121 * Wait for data to arrive at/drain from a socket buffer.
124 sbwait(struct sockbuf *sb)
127 SOCKBUF_LOCK_ASSERT(sb);
129 sb->sb_flags |= SB_WAIT;
130 return (msleep(&sb->sb_cc, &sb->sb_mtx,
131 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
136 sblock(struct sockbuf *sb, int flags)
139 KASSERT((flags & SBL_VALID) == flags,
140 ("sblock: flags invalid (0x%x)", flags));
142 if (flags & SBL_WAIT) {
143 if ((sb->sb_flags & SB_NOINTR) ||
144 (flags & SBL_NOINTR)) {
145 sx_xlock(&sb->sb_sx);
148 return (sx_xlock_sig(&sb->sb_sx));
150 if (sx_try_xlock(&sb->sb_sx) == 0)
151 return (EWOULDBLOCK);
157 sbunlock(struct sockbuf *sb)
160 sx_xunlock(&sb->sb_sx);
164 * Wakeup processes waiting on a socket buffer. Do asynchronous notification
165 * via SIGIO if the socket has the SS_ASYNC flag set.
167 * Called with the socket buffer lock held; will release the lock by the end
168 * of the function. This allows the caller to acquire the socket buffer lock
169 * while testing for the need for various sorts of wakeup and hold it through
170 * to the point where it's no longer required. We currently hold the lock
171 * through calls out to other subsystems (with the exception of kqueue), and
172 * then release it to avoid lock order issues. It's not clear that's
176 sowakeup(struct socket *so, struct sockbuf *sb)
179 SOCKBUF_LOCK_ASSERT(sb);
181 selwakeuppri(&sb->sb_sel, PSOCK);
182 if (!SEL_WAITING(&sb->sb_sel))
183 sb->sb_flags &= ~SB_SEL;
184 if (sb->sb_flags & SB_WAIT) {
185 sb->sb_flags &= ~SB_WAIT;
188 KNOTE_LOCKED(&sb->sb_sel.si_note, 0);
190 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
191 pgsigio(&so->so_sigio, SIGIO, 0);
192 if (sb->sb_flags & SB_UPCALL)
193 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
194 if (sb->sb_flags & SB_AIO)
196 mtx_assert(SOCKBUF_MTX(sb), MA_NOTOWNED);
200 * Socket buffer (struct sockbuf) utility routines.
202 * Each socket contains two socket buffers: one for sending data and one for
203 * receiving data. Each buffer contains a queue of mbufs, information about
204 * the number of mbufs and amount of data in the queue, and other fields
205 * allowing select() statements and notification on data availability to be
208 * Data stored in a socket buffer is maintained as a list of records. Each
209 * record is a list of mbufs chained together with the m_next field. Records
210 * are chained together with the m_nextpkt field. The upper level routine
211 * soreceive() expects the following conventions to be observed when placing
212 * information in the receive buffer:
214 * 1. If the protocol requires each message be preceded by the sender's name,
215 * then a record containing that name must be present before any
216 * associated data (mbuf's must be of type MT_SONAME).
217 * 2. If the protocol supports the exchange of ``access rights'' (really just
218 * additional data associated with the message), and there are ``rights''
219 * to be received, then a record containing this data should be present
220 * (mbuf's must be of type MT_RIGHTS).
221 * 3. If a name or rights record exists, then it must be followed by a data
222 * record, perhaps of zero length.
224 * Before using a new socket structure it is first necessary to reserve
225 * buffer space to the socket, by calling sbreserve(). This should commit
226 * some of the available buffer space in the system buffer pool for the
227 * socket (currently, it does nothing but enforce limits). The space should
228 * be released by calling sbrelease() when the socket is destroyed.
231 soreserve(struct socket *so, u_long sndcc, u_long rcvcc)
233 struct thread *td = curthread;
235 SOCKBUF_LOCK(&so->so_snd);
236 SOCKBUF_LOCK(&so->so_rcv);
237 if (sbreserve_locked(&so->so_snd, sndcc, so, td) == 0)
239 if (sbreserve_locked(&so->so_rcv, rcvcc, so, td) == 0)
241 if (so->so_rcv.sb_lowat == 0)
242 so->so_rcv.sb_lowat = 1;
243 if (so->so_snd.sb_lowat == 0)
244 so->so_snd.sb_lowat = MCLBYTES;
245 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
246 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
247 SOCKBUF_UNLOCK(&so->so_rcv);
248 SOCKBUF_UNLOCK(&so->so_snd);
251 sbrelease_locked(&so->so_snd, so);
253 SOCKBUF_UNLOCK(&so->so_rcv);
254 SOCKBUF_UNLOCK(&so->so_snd);
259 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
262 u_long tmp_sb_max = sb_max;
264 error = sysctl_handle_long(oidp, &tmp_sb_max, arg2, req);
265 if (error || !req->newptr)
267 if (tmp_sb_max < MSIZE + MCLBYTES)
270 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
275 * Allot mbufs to a sockbuf. Attempt to scale mbmax so that mbcnt doesn't
276 * become limiting if buffering efficiency is near the normal case.
279 sbreserve_locked(struct sockbuf *sb, u_long cc, struct socket *so,
284 SOCKBUF_LOCK_ASSERT(sb);
287 * When a thread is passed, we take into account the thread's socket
288 * buffer size limit. The caller will generally pass curthread, but
289 * in the TCP input path, NULL will be passed to indicate that no
290 * appropriate thread resource limits are available. In that case,
291 * we don't apply a process limit.
296 PROC_LOCK(td->td_proc);
297 sbsize_limit = lim_cur(td->td_proc, RLIMIT_SBSIZE);
298 PROC_UNLOCK(td->td_proc);
300 sbsize_limit = RLIM_INFINITY;
301 if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
304 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
305 if (sb->sb_lowat > sb->sb_hiwat)
306 sb->sb_lowat = sb->sb_hiwat;
311 sbreserve(struct sockbuf *sb, u_long cc, struct socket *so,
317 error = sbreserve_locked(sb, cc, so, td);
323 * Free mbufs held by a socket, and reserved mbuf space.
326 sbrelease_internal(struct sockbuf *sb, struct socket *so)
329 sbflush_internal(sb);
330 (void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
336 sbrelease_locked(struct sockbuf *sb, struct socket *so)
339 SOCKBUF_LOCK_ASSERT(sb);
341 sbrelease_internal(sb, so);
345 sbrelease(struct sockbuf *sb, struct socket *so)
349 sbrelease_locked(sb, so);
354 sbdestroy(struct sockbuf *sb, struct socket *so)
357 sbrelease_internal(sb, so);
361 * Routines to add and remove data from an mbuf queue.
363 * The routines sbappend() or sbappendrecord() are normally called to append
364 * new mbufs to a socket buffer, after checking that adequate space is
365 * available, comparing the function sbspace() with the amount of data to be
366 * added. sbappendrecord() differs from sbappend() in that data supplied is
367 * treated as the beginning of a new record. To place a sender's address,
368 * optional access rights, and data in a socket receive buffer,
369 * sbappendaddr() should be used. To place access rights and data in a
370 * socket receive buffer, sbappendrights() should be used. In either case,
371 * the new data begins a new record. Note that unlike sbappend() and
372 * sbappendrecord(), these routines check for the caller that there will be
373 * enough space to store the data. Each fails if there is not enough space,
374 * or if it cannot find mbufs to store additional information in.
376 * Reliable protocols may use the socket send buffer to hold data awaiting
377 * acknowledgement. Data is normally copied from a socket send buffer in a
378 * protocol with m_copy for output to a peer, and then removing the data from
379 * the socket buffer with sbdrop() or sbdroprecord() when the data is
380 * acknowledged by the peer.
384 sblastrecordchk(struct sockbuf *sb, const char *file, int line)
386 struct mbuf *m = sb->sb_mb;
388 SOCKBUF_LOCK_ASSERT(sb);
390 while (m && m->m_nextpkt)
393 if (m != sb->sb_lastrecord) {
394 printf("%s: sb_mb %p sb_lastrecord %p last %p\n",
395 __func__, sb->sb_mb, sb->sb_lastrecord, m);
396 printf("packet chain:\n");
397 for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt)
399 panic("%s from %s:%u", __func__, file, line);
404 sblastmbufchk(struct sockbuf *sb, const char *file, int line)
406 struct mbuf *m = sb->sb_mb;
409 SOCKBUF_LOCK_ASSERT(sb);
411 while (m && m->m_nextpkt)
414 while (m && m->m_next)
417 if (m != sb->sb_mbtail) {
418 printf("%s: sb_mb %p sb_mbtail %p last %p\n",
419 __func__, sb->sb_mb, sb->sb_mbtail, m);
420 printf("packet tree:\n");
421 for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt) {
423 for (n = m; n != NULL; n = n->m_next)
427 panic("%s from %s:%u", __func__, file, line);
430 #endif /* SOCKBUF_DEBUG */
432 #define SBLINKRECORD(sb, m0) do { \
433 SOCKBUF_LOCK_ASSERT(sb); \
434 if ((sb)->sb_lastrecord != NULL) \
435 (sb)->sb_lastrecord->m_nextpkt = (m0); \
437 (sb)->sb_mb = (m0); \
438 (sb)->sb_lastrecord = (m0); \
439 } while (/*CONSTCOND*/0)
442 * Append mbuf chain m to the last record in the socket buffer sb. The
443 * additional space associated the mbuf chain is recorded in sb. Empty mbufs
444 * are discarded and mbufs are compacted where possible.
447 sbappend_locked(struct sockbuf *sb, struct mbuf *m)
451 SOCKBUF_LOCK_ASSERT(sb);
462 if (n->m_flags & M_EOR) {
463 sbappendrecord_locked(sb, m); /* XXXXXX!!!! */
466 } while (n->m_next && (n = n->m_next));
469 * XXX Would like to simply use sb_mbtail here, but
470 * XXX I need to verify that I won't miss an EOR that
473 if ((n = sb->sb_lastrecord) != NULL) {
475 if (n->m_flags & M_EOR) {
476 sbappendrecord_locked(sb, m); /* XXXXXX!!!! */
479 } while (n->m_next && (n = n->m_next));
482 * If this is the first record in the socket buffer,
483 * it's also the last record.
485 sb->sb_lastrecord = m;
488 sbcompress(sb, m, n);
493 * Append mbuf chain m to the last record in the socket buffer sb. The
494 * additional space associated the mbuf chain is recorded in sb. Empty mbufs
495 * are discarded and mbufs are compacted where possible.
498 sbappend(struct sockbuf *sb, struct mbuf *m)
502 sbappend_locked(sb, m);
507 * This version of sbappend() should only be used when the caller absolutely
508 * knows that there will never be more than one record in the socket buffer,
509 * that is, a stream protocol (such as TCP).
512 sbappendstream_locked(struct sockbuf *sb, struct mbuf *m)
514 SOCKBUF_LOCK_ASSERT(sb);
516 KASSERT(m->m_nextpkt == NULL,("sbappendstream 0"));
517 KASSERT(sb->sb_mb == sb->sb_lastrecord,("sbappendstream 1"));
521 sbcompress(sb, m, sb->sb_mbtail);
523 sb->sb_lastrecord = sb->sb_mb;
528 * This version of sbappend() should only be used when the caller absolutely
529 * knows that there will never be more than one record in the socket buffer,
530 * that is, a stream protocol (such as TCP).
533 sbappendstream(struct sockbuf *sb, struct mbuf *m)
537 sbappendstream_locked(sb, m);
543 sbcheck(struct sockbuf *sb)
547 u_long len = 0, mbcnt = 0;
549 SOCKBUF_LOCK_ASSERT(sb);
551 for (m = sb->sb_mb; m; m = n) {
553 for (; m; m = m->m_next) {
556 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
557 mbcnt += m->m_ext.ext_size;
560 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
561 printf("cc %ld != %u || mbcnt %ld != %u\n", len, sb->sb_cc,
562 mbcnt, sb->sb_mbcnt);
569 * As above, except the mbuf chain begins a new record.
572 sbappendrecord_locked(struct sockbuf *sb, struct mbuf *m0)
576 SOCKBUF_LOCK_ASSERT(sb);
585 * Put the first mbuf on the queue. Note this permits zero length
590 SBLINKRECORD(sb, m0);
597 if (m && (m0->m_flags & M_EOR)) {
598 m0->m_flags &= ~M_EOR;
601 sbcompress(sb, m, m0);
605 * As above, except the mbuf chain begins a new record.
608 sbappendrecord(struct sockbuf *sb, struct mbuf *m0)
612 sbappendrecord_locked(sb, m0);
617 * Append address and data, and optionally, control (ancillary) data to the
618 * receive queue of a socket. If present, m0 must include a packet header
619 * with total length. Returns 0 if no space in sockbuf or insufficient
623 sbappendaddr_locked(struct sockbuf *sb, const struct sockaddr *asa,
624 struct mbuf *m0, struct mbuf *control)
626 struct mbuf *m, *n, *nlast;
627 int space = asa->sa_len;
629 SOCKBUF_LOCK_ASSERT(sb);
631 if (m0 && (m0->m_flags & M_PKTHDR) == 0)
632 panic("sbappendaddr_locked");
634 space += m0->m_pkthdr.len;
635 space += m_length(control, &n);
637 if (space > sbspace(sb))
640 if (asa->sa_len > MLEN)
643 MGET(m, M_DONTWAIT, MT_SONAME);
646 m->m_len = asa->sa_len;
647 bcopy(asa, mtod(m, caddr_t), asa->sa_len);
649 n->m_next = m0; /* concatenate data to control */
653 for (n = m; n->m_next != NULL; n = n->m_next)
659 sb->sb_mbtail = nlast;
667 * Append address and data, and optionally, control (ancillary) data to the
668 * receive queue of a socket. If present, m0 must include a packet header
669 * with total length. Returns 0 if no space in sockbuf or insufficient
673 sbappendaddr(struct sockbuf *sb, const struct sockaddr *asa,
674 struct mbuf *m0, struct mbuf *control)
679 retval = sbappendaddr_locked(sb, asa, m0, control);
685 sbappendcontrol_locked(struct sockbuf *sb, struct mbuf *m0,
686 struct mbuf *control)
688 struct mbuf *m, *n, *mlast;
691 SOCKBUF_LOCK_ASSERT(sb);
694 panic("sbappendcontrol_locked");
695 space = m_length(control, &n) + m_length(m0, NULL);
697 if (space > sbspace(sb))
699 n->m_next = m0; /* concatenate data to control */
703 for (m = control; m->m_next; m = m->m_next)
707 SBLINKRECORD(sb, control);
709 sb->sb_mbtail = mlast;
717 sbappendcontrol(struct sockbuf *sb, struct mbuf *m0, struct mbuf *control)
722 retval = sbappendcontrol_locked(sb, m0, control);
728 * Append the data in mbuf chain (m) into the socket buffer sb following mbuf
729 * (n). If (n) is NULL, the buffer is presumed empty.
731 * When the data is compressed, mbufs in the chain may be handled in one of
734 * (1) The mbuf may simply be dropped, if it contributes nothing (no data, no
735 * record boundary, and no change in data type).
737 * (2) The mbuf may be coalesced -- i.e., data in the mbuf may be copied into
738 * an mbuf already in the socket buffer. This can occur if an
739 * appropriate mbuf exists, there is room, and no merging of data types
742 * (3) The mbuf may be appended to the end of the existing mbuf chain.
744 * If any of the new mbufs is marked as M_EOR, mark the last mbuf appended as
748 sbcompress(struct sockbuf *sb, struct mbuf *m, struct mbuf *n)
753 SOCKBUF_LOCK_ASSERT(sb);
756 eor |= m->m_flags & M_EOR;
759 (((o = m->m_next) || (o = n)) &&
760 o->m_type == m->m_type))) {
761 if (sb->sb_lastrecord == m)
762 sb->sb_lastrecord = m->m_next;
766 if (n && (n->m_flags & M_EOR) == 0 &&
768 ((sb->sb_flags & SB_NOCOALESCE) == 0) &&
769 m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
770 m->m_len <= M_TRAILINGSPACE(n) &&
771 n->m_type == m->m_type) {
772 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
774 n->m_len += m->m_len;
775 sb->sb_cc += m->m_len;
776 if (m->m_type != MT_DATA && m->m_type != MT_OOBDATA)
777 /* XXX: Probably don't need.*/
778 sb->sb_ctl += m->m_len;
789 m->m_flags &= ~M_EOR;
794 KASSERT(n != NULL, ("sbcompress: eor && n == NULL"));
801 * Free all mbufs in a sockbuf. Check that all resources are reclaimed.
804 sbflush_internal(struct sockbuf *sb)
807 while (sb->sb_mbcnt) {
809 * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
810 * we would loop forever. Panic instead.
812 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
814 sbdrop_internal(sb, (int)sb->sb_cc);
816 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
817 panic("sbflush_internal: cc %u || mb %p || mbcnt %u",
818 sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
822 sbflush_locked(struct sockbuf *sb)
825 SOCKBUF_LOCK_ASSERT(sb);
826 sbflush_internal(sb);
830 sbflush(struct sockbuf *sb)
839 * Drop data from (the front of) a sockbuf.
842 sbdrop_internal(struct sockbuf *sb, int len)
847 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
856 if (m->m_len > len) {
860 if (sb->sb_sndptroff != 0)
861 sb->sb_sndptroff -= len;
862 if (m->m_type != MT_DATA && m->m_type != MT_OOBDATA)
870 while (m && m->m_len == 0) {
880 * First part is an inline SB_EMPTY_FIXUP(). Second part makes sure
881 * sb_lastrecord is up-to-date if we dropped part of the last record.
885 sb->sb_mbtail = NULL;
886 sb->sb_lastrecord = NULL;
887 } else if (m->m_nextpkt == NULL) {
888 sb->sb_lastrecord = m;
893 * Drop data from (the front of) a sockbuf.
896 sbdrop_locked(struct sockbuf *sb, int len)
899 SOCKBUF_LOCK_ASSERT(sb);
901 sbdrop_internal(sb, len);
905 sbdrop(struct sockbuf *sb, int len)
909 sbdrop_locked(sb, len);
914 * Maintain a pointer and offset pair into the socket buffer mbuf chain to
915 * avoid traversal of the entire socket buffer for larger offsets.
918 sbsndptr(struct sockbuf *sb, u_int off, u_int len, u_int *moff)
920 struct mbuf *m, *ret;
922 KASSERT(sb->sb_mb != NULL, ("%s: sb_mb is NULL", __func__));
923 KASSERT(off + len <= sb->sb_cc, ("%s: beyond sb", __func__));
924 KASSERT(sb->sb_sndptroff <= sb->sb_cc, ("%s: sndptroff broken", __func__));
927 * Is off below stored offset? Happens on retransmits.
928 * Just return, we can't help here.
930 if (sb->sb_sndptroff > off) {
935 /* Return closest mbuf in chain for current offset. */
936 *moff = off - sb->sb_sndptroff;
937 m = ret = sb->sb_sndptr ? sb->sb_sndptr : sb->sb_mb;
939 /* Advance by len to be as close as possible for the next transmit. */
940 for (off = off - sb->sb_sndptroff + len - 1;
941 off > 0 && m != NULL && off >= m->m_len;
943 sb->sb_sndptroff += m->m_len;
946 if (off > 0 && m == NULL)
947 panic("%s: sockbuf %p and mbuf %p clashing", __func__, sb, ret);
954 * Drop a record off the front of a sockbuf and move the next record to the
958 sbdroprecord_locked(struct sockbuf *sb)
962 SOCKBUF_LOCK_ASSERT(sb);
966 sb->sb_mb = m->m_nextpkt;
976 * Drop a record off the front of a sockbuf and move the next record to the
980 sbdroprecord(struct sockbuf *sb)
984 sbdroprecord_locked(sb);
989 * Create a "control" mbuf containing the specified data with the specified
990 * type for presentation on a socket buffer.
993 sbcreatecontrol(caddr_t p, int size, int type, int level)
998 if (CMSG_SPACE((u_int)size) > MCLBYTES)
999 return ((struct mbuf *) NULL);
1000 if (CMSG_SPACE((u_int)size) > MLEN)
1001 m = m_getcl(M_DONTWAIT, MT_CONTROL, 0);
1003 m = m_get(M_DONTWAIT, MT_CONTROL);
1005 return ((struct mbuf *) NULL);
1006 cp = mtod(m, struct cmsghdr *);
1008 KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
1009 ("sbcreatecontrol: short mbuf"));
1011 (void)memcpy(CMSG_DATA(cp), p, size);
1012 m->m_len = CMSG_SPACE(size);
1013 cp->cmsg_len = CMSG_LEN(size);
1014 cp->cmsg_level = level;
1015 cp->cmsg_type = type;
1020 * This does the same for socket buffers that sotoxsocket does for sockets:
1021 * generate an user-format data structure describing the socket buffer. Note
1022 * that the xsockbuf structure, since it is always embedded in a socket, does
1023 * not include a self pointer nor a length. We make this entry point public
1024 * in case some other mechanism needs it.
1027 sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
1030 xsb->sb_cc = sb->sb_cc;
1031 xsb->sb_hiwat = sb->sb_hiwat;
1032 xsb->sb_mbcnt = sb->sb_mbcnt;
1033 xsb->sb_mcnt = sb->sb_mcnt;
1034 xsb->sb_ccnt = sb->sb_ccnt;
1035 xsb->sb_mbmax = sb->sb_mbmax;
1036 xsb->sb_lowat = sb->sb_lowat;
1037 xsb->sb_flags = sb->sb_flags;
1038 xsb->sb_timeo = sb->sb_timeo;
1041 /* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
1043 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
1044 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_ULONG|CTLFLAG_RW,
1045 &sb_max, 0, sysctl_handle_sb_max, "LU", "Maximum socket buffer size");
1046 SYSCTL_ULONG(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
1047 &sb_efficiency, 0, "");