The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/uipc_socket2.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    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.
   16  *
   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
   27  * SUCH DAMAGE.
   28  *
   29  *      @(#)uipc_socket2.c      8.1 (Berkeley) 6/10/93
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD: releng/6.0/sys/kern/uipc_socket2.c 150420 2005-09-21 15:30:54Z rwatson $");
   34 
   35 #include "opt_mac.h"
   36 #include "opt_param.h"
   37 
   38 #include <sys/param.h>
   39 #include <sys/aio.h> /* for aio_swake proto */
   40 #include <sys/domain.h>
   41 #include <sys/event.h>
   42 #include <sys/file.h>   /* for maxfiles */
   43 #include <sys/kernel.h>
   44 #include <sys/lock.h>
   45 #include <sys/mac.h>
   46 #include <sys/malloc.h>
   47 #include <sys/mbuf.h>
   48 #include <sys/mutex.h>
   49 #include <sys/proc.h>
   50 #include <sys/protosw.h>
   51 #include <sys/resourcevar.h>
   52 #include <sys/signalvar.h>
   53 #include <sys/socket.h>
   54 #include <sys/socketvar.h>
   55 #include <sys/stat.h>
   56 #include <sys/sysctl.h>
   57 #include <sys/systm.h>
   58 
   59 int     maxsockets;
   60 
   61 void (*aio_swake)(struct socket *, struct sockbuf *);
   62 
   63 /*
   64  * Primitive routines for operating on sockets and socket buffers
   65  */
   66 
   67 u_long  sb_max = SB_MAX;
   68 static  u_long sb_max_adj =
   69     SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
   70 
   71 static  u_long sb_efficiency = 8;       /* parameter for sbreserve() */
   72 
   73 /*
   74  * Procedures to manipulate state flags of socket
   75  * and do appropriate wakeups.  Normal sequence from the
   76  * active (originating) side is that soisconnecting() is
   77  * called during processing of connect() call,
   78  * resulting in an eventual call to soisconnected() if/when the
   79  * connection is established.  When the connection is torn down
   80  * soisdisconnecting() is called during processing of disconnect() call,
   81  * and soisdisconnected() is called when the connection to the peer
   82  * is totally severed.  The semantics of these routines are such that
   83  * connectionless protocols can call soisconnected() and soisdisconnected()
   84  * only, bypassing the in-progress calls when setting up a ``connection''
   85  * takes no time.
   86  *
   87  * From the passive side, a socket is created with
   88  * two queues of sockets: so_incomp for connections in progress
   89  * and so_comp for connections already made and awaiting user acceptance.
   90  * As a protocol is preparing incoming connections, it creates a socket
   91  * structure queued on so_incomp by calling sonewconn().  When the connection
   92  * is established, soisconnected() is called, and transfers the
   93  * socket structure to so_comp, making it available to accept().
   94  *
   95  * If a socket is closed with sockets on either
   96  * so_incomp or so_comp, these sockets are dropped.
   97  *
   98  * If higher level protocols are implemented in
   99  * the kernel, the wakeups done here will sometimes
  100  * cause software-interrupt process scheduling.
  101  */
  102 
  103 void
  104 soisconnecting(so)
  105         register struct socket *so;
  106 {
  107 
  108         SOCK_LOCK(so);
  109         so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
  110         so->so_state |= SS_ISCONNECTING;
  111         SOCK_UNLOCK(so);
  112 }
  113 
  114 void
  115 soisconnected(so)
  116         struct socket *so;
  117 {
  118         struct socket *head;
  119 
  120         ACCEPT_LOCK();
  121         SOCK_LOCK(so);
  122         so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
  123         so->so_state |= SS_ISCONNECTED;
  124         head = so->so_head;
  125         if (head != NULL && (so->so_qstate & SQ_INCOMP)) {
  126                 if ((so->so_options & SO_ACCEPTFILTER) == 0) {
  127                         SOCK_UNLOCK(so);
  128                         TAILQ_REMOVE(&head->so_incomp, so, so_list);
  129                         head->so_incqlen--;
  130                         so->so_qstate &= ~SQ_INCOMP;
  131                         TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
  132                         head->so_qlen++;
  133                         so->so_qstate |= SQ_COMP;
  134                         ACCEPT_UNLOCK();
  135                         sorwakeup(head);
  136                         wakeup_one(&head->so_timeo);
  137                 } else {
  138                         ACCEPT_UNLOCK();
  139                         so->so_upcall =
  140                             head->so_accf->so_accept_filter->accf_callback;
  141                         so->so_upcallarg = head->so_accf->so_accept_filter_arg;
  142                         so->so_rcv.sb_flags |= SB_UPCALL;
  143                         so->so_options &= ~SO_ACCEPTFILTER;
  144                         SOCK_UNLOCK(so);
  145                         so->so_upcall(so, so->so_upcallarg, M_DONTWAIT);
  146                 }
  147                 return;
  148         }
  149         SOCK_UNLOCK(so);
  150         ACCEPT_UNLOCK();
  151         wakeup(&so->so_timeo);
  152         sorwakeup(so);
  153         sowwakeup(so);
  154 }
  155 
  156 void
  157 soisdisconnecting(so)
  158         register struct socket *so;
  159 {
  160 
  161         /*
  162          * XXXRW: This code assumes that SOCK_LOCK(so) and
  163          * SOCKBUF_LOCK(&so->so_rcv) are the same.
  164          */
  165         SOCKBUF_LOCK(&so->so_rcv);
  166         so->so_state &= ~SS_ISCONNECTING;
  167         so->so_state |= SS_ISDISCONNECTING;
  168         so->so_rcv.sb_state |= SBS_CANTRCVMORE;
  169         sorwakeup_locked(so);
  170         SOCKBUF_LOCK(&so->so_snd);
  171         so->so_snd.sb_state |= SBS_CANTSENDMORE;
  172         sowwakeup_locked(so);
  173         wakeup(&so->so_timeo);
  174 }
  175 
  176 void
  177 soisdisconnected(so)
  178         register struct socket *so;
  179 {
  180 
  181         /*
  182          * XXXRW: This code assumes that SOCK_LOCK(so) and
  183          * SOCKBUF_LOCK(&so->so_rcv) are the same.
  184          */
  185         SOCKBUF_LOCK(&so->so_rcv);
  186         so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
  187         so->so_state |= SS_ISDISCONNECTED;
  188         so->so_rcv.sb_state |= SBS_CANTRCVMORE;
  189         sorwakeup_locked(so);
  190         SOCKBUF_LOCK(&so->so_snd);
  191         so->so_snd.sb_state |= SBS_CANTSENDMORE;
  192         sbdrop_locked(&so->so_snd, so->so_snd.sb_cc);
  193         sowwakeup_locked(so);
  194         wakeup(&so->so_timeo);
  195 }
  196 
  197 /*
  198  * When an attempt at a new connection is noted on a socket
  199  * which accepts connections, sonewconn is called.  If the
  200  * connection is possible (subject to space constraints, etc.)
  201  * then we allocate a new structure, propoerly linked into the
  202  * data structure of the original socket, and return this.
  203  * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
  204  *
  205  * note: the ref count on the socket is 0 on return
  206  */
  207 struct socket *
  208 sonewconn(head, connstatus)
  209         register struct socket *head;
  210         int connstatus;
  211 {
  212         register struct socket *so;
  213         int over;
  214 
  215         ACCEPT_LOCK();
  216         over = (head->so_qlen > 3 * head->so_qlimit / 2);
  217         ACCEPT_UNLOCK();
  218         if (over)
  219                 return (NULL);
  220         so = soalloc(M_NOWAIT);
  221         if (so == NULL)
  222                 return (NULL);
  223         if ((head->so_options & SO_ACCEPTFILTER) != 0)
  224                 connstatus = 0;
  225         so->so_head = head;
  226         so->so_type = head->so_type;
  227         so->so_options = head->so_options &~ SO_ACCEPTCONN;
  228         so->so_linger = head->so_linger;
  229         so->so_state = head->so_state | SS_NOFDREF;
  230         so->so_proto = head->so_proto;
  231         so->so_timeo = head->so_timeo;
  232         so->so_cred = crhold(head->so_cred);
  233 #ifdef MAC
  234         SOCK_LOCK(head);
  235         mac_create_socket_from_socket(head, so);
  236         SOCK_UNLOCK(head);
  237 #endif
  238         knlist_init(&so->so_rcv.sb_sel.si_note, SOCKBUF_MTX(&so->so_rcv),
  239             NULL, NULL, NULL);
  240         knlist_init(&so->so_snd.sb_sel.si_note, SOCKBUF_MTX(&so->so_snd),
  241             NULL, NULL, NULL);
  242         if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
  243             (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
  244                 sodealloc(so);
  245                 return (NULL);
  246         }
  247         so->so_state |= connstatus;
  248         ACCEPT_LOCK();
  249         if (connstatus) {
  250                 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
  251                 so->so_qstate |= SQ_COMP;
  252                 head->so_qlen++;
  253         } else {
  254                 /*
  255                  * Keep removing sockets from the head until there's room for
  256                  * us to insert on the tail.  In pre-locking revisions, this
  257                  * was a simple if(), but as we could be racing with other
  258                  * threads and soabort() requires dropping locks, we must
  259                  * loop waiting for the condition to be true.
  260                  */
  261                 while (head->so_incqlen > head->so_qlimit) {
  262                         struct socket *sp;
  263                         sp = TAILQ_FIRST(&head->so_incomp);
  264                         TAILQ_REMOVE(&so->so_incomp, sp, so_list);
  265                         head->so_incqlen--;
  266                         sp->so_qstate &= ~SQ_INCOMP;
  267                         sp->so_head = NULL;
  268                         ACCEPT_UNLOCK();
  269                         (void) soabort(sp);
  270                         ACCEPT_LOCK();
  271                 }
  272                 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
  273                 so->so_qstate |= SQ_INCOMP;
  274                 head->so_incqlen++;
  275         }
  276         ACCEPT_UNLOCK();
  277         if (connstatus) {
  278                 sorwakeup(head);
  279                 wakeup_one(&head->so_timeo);
  280         }
  281         return (so);
  282 }
  283 
  284 /*
  285  * Socantsendmore indicates that no more data will be sent on the
  286  * socket; it would normally be applied to a socket when the user
  287  * informs the system that no more data is to be sent, by the protocol
  288  * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
  289  * will be received, and will normally be applied to the socket by a
  290  * protocol when it detects that the peer will send no more data.
  291  * Data queued for reading in the socket may yet be read.
  292  */
  293 void
  294 socantsendmore_locked(so)
  295         struct socket *so;
  296 {
  297 
  298         SOCKBUF_LOCK_ASSERT(&so->so_snd);
  299 
  300         so->so_snd.sb_state |= SBS_CANTSENDMORE;
  301         sowwakeup_locked(so);
  302         mtx_assert(SOCKBUF_MTX(&so->so_snd), MA_NOTOWNED);
  303 }
  304 
  305 void
  306 socantsendmore(so)
  307         struct socket *so;
  308 {
  309 
  310         SOCKBUF_LOCK(&so->so_snd);
  311         socantsendmore_locked(so);
  312         mtx_assert(SOCKBUF_MTX(&so->so_snd), MA_NOTOWNED);
  313 }
  314 
  315 void
  316 socantrcvmore_locked(so)
  317         struct socket *so;
  318 {
  319 
  320         SOCKBUF_LOCK_ASSERT(&so->so_rcv);
  321 
  322         so->so_rcv.sb_state |= SBS_CANTRCVMORE;
  323         sorwakeup_locked(so);
  324         mtx_assert(SOCKBUF_MTX(&so->so_rcv), MA_NOTOWNED);
  325 }
  326 
  327 void
  328 socantrcvmore(so)
  329         struct socket *so;
  330 {
  331 
  332         SOCKBUF_LOCK(&so->so_rcv);
  333         socantrcvmore_locked(so);
  334         mtx_assert(SOCKBUF_MTX(&so->so_rcv), MA_NOTOWNED);
  335 }
  336 
  337 /*
  338  * Wait for data to arrive at/drain from a socket buffer.
  339  */
  340 int
  341 sbwait(sb)
  342         struct sockbuf *sb;
  343 {
  344 
  345         SOCKBUF_LOCK_ASSERT(sb);
  346 
  347         sb->sb_flags |= SB_WAIT;
  348         return (msleep(&sb->sb_cc, &sb->sb_mtx,
  349             (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
  350             sb->sb_timeo));
  351 }
  352 
  353 /*
  354  * Lock a sockbuf already known to be locked;
  355  * return any error returned from sleep (EINTR).
  356  */
  357 int
  358 sb_lock(sb)
  359         register struct sockbuf *sb;
  360 {
  361         int error;
  362 
  363         SOCKBUF_LOCK_ASSERT(sb);
  364 
  365         while (sb->sb_flags & SB_LOCK) {
  366                 sb->sb_flags |= SB_WANT;
  367                 error = msleep(&sb->sb_flags, &sb->sb_mtx,
  368                     (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
  369                     "sblock", 0);
  370                 if (error)
  371                         return (error);
  372         }
  373         sb->sb_flags |= SB_LOCK;
  374         return (0);
  375 }
  376 
  377 /*
  378  * Wakeup processes waiting on a socket buffer.  Do asynchronous
  379  * notification via SIGIO if the socket has the SS_ASYNC flag set.
  380  *
  381  * Called with the socket buffer lock held; will release the lock by the end
  382  * of the function.  This allows the caller to acquire the socket buffer lock
  383  * while testing for the need for various sorts of wakeup and hold it through
  384  * to the point where it's no longer required.  We currently hold the lock
  385  * through calls out to other subsystems (with the exception of kqueue), and
  386  * then release it to avoid lock order issues.  It's not clear that's
  387  * correct.
  388  */
  389 void
  390 sowakeup(so, sb)
  391         register struct socket *so;
  392         register struct sockbuf *sb;
  393 {
  394 
  395         SOCKBUF_LOCK_ASSERT(sb);
  396 
  397         selwakeuppri(&sb->sb_sel, PSOCK);
  398         sb->sb_flags &= ~SB_SEL;
  399         if (sb->sb_flags & SB_WAIT) {
  400                 sb->sb_flags &= ~SB_WAIT;
  401                 wakeup(&sb->sb_cc);
  402         }
  403         KNOTE_LOCKED(&sb->sb_sel.si_note, 0);
  404         SOCKBUF_UNLOCK(sb);
  405         if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
  406                 pgsigio(&so->so_sigio, SIGIO, 0);
  407         if (sb->sb_flags & SB_UPCALL)
  408                 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
  409         if (sb->sb_flags & SB_AIO)
  410                 aio_swake(so, sb);
  411         mtx_assert(SOCKBUF_MTX(sb), MA_NOTOWNED);
  412 }
  413 
  414 /*
  415  * Socket buffer (struct sockbuf) utility routines.
  416  *
  417  * Each socket contains two socket buffers: one for sending data and
  418  * one for receiving data.  Each buffer contains a queue of mbufs,
  419  * information about the number of mbufs and amount of data in the
  420  * queue, and other fields allowing select() statements and notification
  421  * on data availability to be implemented.
  422  *
  423  * Data stored in a socket buffer is maintained as a list of records.
  424  * Each record is a list of mbufs chained together with the m_next
  425  * field.  Records are chained together with the m_nextpkt field. The upper
  426  * level routine soreceive() expects the following conventions to be
  427  * observed when placing information in the receive buffer:
  428  *
  429  * 1. If the protocol requires each message be preceded by the sender's
  430  *    name, then a record containing that name must be present before
  431  *    any associated data (mbuf's must be of type MT_SONAME).
  432  * 2. If the protocol supports the exchange of ``access rights'' (really
  433  *    just additional data associated with the message), and there are
  434  *    ``rights'' to be received, then a record containing this data
  435  *    should be present (mbuf's must be of type MT_RIGHTS).
  436  * 3. If a name or rights record exists, then it must be followed by
  437  *    a data record, perhaps of zero length.
  438  *
  439  * Before using a new socket structure it is first necessary to reserve
  440  * buffer space to the socket, by calling sbreserve().  This should commit
  441  * some of the available buffer space in the system buffer pool for the
  442  * socket (currently, it does nothing but enforce limits).  The space
  443  * should be released by calling sbrelease() when the socket is destroyed.
  444  */
  445 
  446 int
  447 soreserve(so, sndcc, rcvcc)
  448         register struct socket *so;
  449         u_long sndcc, rcvcc;
  450 {
  451         struct thread *td = curthread;
  452 
  453         SOCKBUF_LOCK(&so->so_snd);
  454         SOCKBUF_LOCK(&so->so_rcv);
  455         if (sbreserve_locked(&so->so_snd, sndcc, so, td) == 0)
  456                 goto bad;
  457         if (sbreserve_locked(&so->so_rcv, rcvcc, so, td) == 0)
  458                 goto bad2;
  459         if (so->so_rcv.sb_lowat == 0)
  460                 so->so_rcv.sb_lowat = 1;
  461         if (so->so_snd.sb_lowat == 0)
  462                 so->so_snd.sb_lowat = MCLBYTES;
  463         if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
  464                 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
  465         SOCKBUF_UNLOCK(&so->so_rcv);
  466         SOCKBUF_UNLOCK(&so->so_snd);
  467         return (0);
  468 bad2:
  469         sbrelease_locked(&so->so_snd, so);
  470 bad:
  471         SOCKBUF_UNLOCK(&so->so_rcv);
  472         SOCKBUF_UNLOCK(&so->so_snd);
  473         return (ENOBUFS);
  474 }
  475 
  476 static int
  477 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
  478 {
  479         int error = 0;
  480         u_long old_sb_max = sb_max;
  481 
  482         error = SYSCTL_OUT(req, arg1, sizeof(u_long));
  483         if (error || !req->newptr)
  484                 return (error);
  485         error = SYSCTL_IN(req, arg1, sizeof(u_long));
  486         if (error)
  487                 return (error);
  488         if (sb_max < MSIZE + MCLBYTES) {
  489                 sb_max = old_sb_max;
  490                 return (EINVAL);
  491         }
  492         sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
  493         return (0);
  494 }
  495         
  496 /*
  497  * Allot mbufs to a sockbuf.
  498  * Attempt to scale mbmax so that mbcnt doesn't become limiting
  499  * if buffering efficiency is near the normal case.
  500  */
  501 int
  502 sbreserve_locked(sb, cc, so, td)
  503         struct sockbuf *sb;
  504         u_long cc;
  505         struct socket *so;
  506         struct thread *td;
  507 {
  508         rlim_t sbsize_limit;
  509 
  510         SOCKBUF_LOCK_ASSERT(sb);
  511 
  512         /*
  513          * td will only be NULL when we're in an interrupt
  514          * (e.g. in tcp_input())
  515          */
  516         if (cc > sb_max_adj)
  517                 return (0);
  518         if (td != NULL) {
  519                 PROC_LOCK(td->td_proc);
  520                 sbsize_limit = lim_cur(td->td_proc, RLIMIT_SBSIZE);
  521                 PROC_UNLOCK(td->td_proc);
  522         } else
  523                 sbsize_limit = RLIM_INFINITY;
  524         if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
  525             sbsize_limit))
  526                 return (0);
  527         sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
  528         if (sb->sb_lowat > sb->sb_hiwat)
  529                 sb->sb_lowat = sb->sb_hiwat;
  530         return (1);
  531 }
  532 
  533 int
  534 sbreserve(sb, cc, so, td)
  535         struct sockbuf *sb;
  536         u_long cc;
  537         struct socket *so;
  538         struct thread *td;
  539 {
  540         int error;
  541 
  542         SOCKBUF_LOCK(sb);
  543         error = sbreserve_locked(sb, cc, so, td);
  544         SOCKBUF_UNLOCK(sb);
  545         return (error);
  546 }
  547 
  548 /*
  549  * Free mbufs held by a socket, and reserved mbuf space.
  550  */
  551 void
  552 sbrelease_locked(sb, so)
  553         struct sockbuf *sb;
  554         struct socket *so;
  555 {
  556 
  557         SOCKBUF_LOCK_ASSERT(sb);
  558 
  559         sbflush_locked(sb);
  560         (void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
  561             RLIM_INFINITY);
  562         sb->sb_mbmax = 0;
  563 }
  564 
  565 void
  566 sbrelease(sb, so)
  567         struct sockbuf *sb;
  568         struct socket *so;
  569 {
  570 
  571         SOCKBUF_LOCK(sb);
  572         sbrelease_locked(sb, so);
  573         SOCKBUF_UNLOCK(sb);
  574 }
  575 /*
  576  * Routines to add and remove
  577  * data from an mbuf queue.
  578  *
  579  * The routines sbappend() or sbappendrecord() are normally called to
  580  * append new mbufs to a socket buffer, after checking that adequate
  581  * space is available, comparing the function sbspace() with the amount
  582  * of data to be added.  sbappendrecord() differs from sbappend() in
  583  * that data supplied is treated as the beginning of a new record.
  584  * To place a sender's address, optional access rights, and data in a
  585  * socket receive buffer, sbappendaddr() should be used.  To place
  586  * access rights and data in a socket receive buffer, sbappendrights()
  587  * should be used.  In either case, the new data begins a new record.
  588  * Note that unlike sbappend() and sbappendrecord(), these routines check
  589  * for the caller that there will be enough space to store the data.
  590  * Each fails if there is not enough space, or if it cannot find mbufs
  591  * to store additional information in.
  592  *
  593  * Reliable protocols may use the socket send buffer to hold data
  594  * awaiting acknowledgement.  Data is normally copied from a socket
  595  * send buffer in a protocol with m_copy for output to a peer,
  596  * and then removing the data from the socket buffer with sbdrop()
  597  * or sbdroprecord() when the data is acknowledged by the peer.
  598  */
  599 
  600 #ifdef SOCKBUF_DEBUG
  601 void
  602 sblastrecordchk(struct sockbuf *sb, const char *file, int line)
  603 {
  604         struct mbuf *m = sb->sb_mb;
  605 
  606         SOCKBUF_LOCK_ASSERT(sb);
  607 
  608         while (m && m->m_nextpkt)
  609                 m = m->m_nextpkt;
  610 
  611         if (m != sb->sb_lastrecord) {
  612                 printf("%s: sb_mb %p sb_lastrecord %p last %p\n",
  613                         __func__, sb->sb_mb, sb->sb_lastrecord, m);
  614                 printf("packet chain:\n");
  615                 for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt)
  616                         printf("\t%p\n", m);
  617                 panic("%s from %s:%u", __func__, file, line);
  618         }
  619 }
  620 
  621 void
  622 sblastmbufchk(struct sockbuf *sb, const char *file, int line)
  623 {
  624         struct mbuf *m = sb->sb_mb;
  625         struct mbuf *n;
  626 
  627         SOCKBUF_LOCK_ASSERT(sb);
  628 
  629         while (m && m->m_nextpkt)
  630                 m = m->m_nextpkt;
  631 
  632         while (m && m->m_next)
  633                 m = m->m_next;
  634 
  635         if (m != sb->sb_mbtail) {
  636                 printf("%s: sb_mb %p sb_mbtail %p last %p\n",
  637                         __func__, sb->sb_mb, sb->sb_mbtail, m);
  638                 printf("packet tree:\n");
  639                 for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt) {
  640                         printf("\t");
  641                         for (n = m; n != NULL; n = n->m_next)
  642                                 printf("%p ", n);
  643                         printf("\n");
  644                 }
  645                 panic("%s from %s:%u", __func__, file, line);
  646         }
  647 }
  648 #endif /* SOCKBUF_DEBUG */
  649 
  650 #define SBLINKRECORD(sb, m0) do {                                       \
  651         SOCKBUF_LOCK_ASSERT(sb);                                        \
  652         if ((sb)->sb_lastrecord != NULL)                                \
  653                 (sb)->sb_lastrecord->m_nextpkt = (m0);                  \
  654         else                                                            \
  655                 (sb)->sb_mb = (m0);                                     \
  656         (sb)->sb_lastrecord = (m0);                                     \
  657 } while (/*CONSTCOND*/0)
  658 
  659 /*
  660  * Append mbuf chain m to the last record in the
  661  * socket buffer sb.  The additional space associated
  662  * the mbuf chain is recorded in sb.  Empty mbufs are
  663  * discarded and mbufs are compacted where possible.
  664  */
  665 void
  666 sbappend_locked(sb, m)
  667         struct sockbuf *sb;
  668         struct mbuf *m;
  669 {
  670         register struct mbuf *n;
  671 
  672         SOCKBUF_LOCK_ASSERT(sb);
  673 
  674         if (m == 0)
  675                 return;
  676 
  677         SBLASTRECORDCHK(sb);
  678         n = sb->sb_mb;
  679         if (n) {
  680                 while (n->m_nextpkt)
  681                         n = n->m_nextpkt;
  682                 do {
  683                         if (n->m_flags & M_EOR) {
  684                                 sbappendrecord_locked(sb, m); /* XXXXXX!!!! */
  685                                 return;
  686                         }
  687                 } while (n->m_next && (n = n->m_next));
  688         } else {
  689                 /*
  690                  * XXX Would like to simply use sb_mbtail here, but
  691                  * XXX I need to verify that I won't miss an EOR that
  692                  * XXX way.
  693                  */
  694                 if ((n = sb->sb_lastrecord) != NULL) {
  695                         do {
  696                                 if (n->m_flags & M_EOR) {
  697                                         sbappendrecord_locked(sb, m); /* XXXXXX!!!! */
  698                                         return;
  699                                 }
  700                         } while (n->m_next && (n = n->m_next));
  701                 } else {
  702                         /*
  703                          * If this is the first record in the socket buffer,
  704                          * it's also the last record.
  705                          */
  706                         sb->sb_lastrecord = m;
  707                 }
  708         }
  709         sbcompress(sb, m, n);
  710         SBLASTRECORDCHK(sb);
  711 }
  712 
  713 /*
  714  * Append mbuf chain m to the last record in the
  715  * socket buffer sb.  The additional space associated
  716  * the mbuf chain is recorded in sb.  Empty mbufs are
  717  * discarded and mbufs are compacted where possible.
  718  */
  719 void
  720 sbappend(sb, m)
  721         struct sockbuf *sb;
  722         struct mbuf *m;
  723 {
  724 
  725         SOCKBUF_LOCK(sb);
  726         sbappend_locked(sb, m);
  727         SOCKBUF_UNLOCK(sb);
  728 }
  729 
  730 /*
  731  * This version of sbappend() should only be used when the caller
  732  * absolutely knows that there will never be more than one record
  733  * in the socket buffer, that is, a stream protocol (such as TCP).
  734  */
  735 void
  736 sbappendstream_locked(struct sockbuf *sb, struct mbuf *m)
  737 {
  738         SOCKBUF_LOCK_ASSERT(sb);
  739 
  740         KASSERT(m->m_nextpkt == NULL,("sbappendstream 0"));
  741         KASSERT(sb->sb_mb == sb->sb_lastrecord,("sbappendstream 1"));
  742 
  743         SBLASTMBUFCHK(sb);
  744 
  745         sbcompress(sb, m, sb->sb_mbtail);
  746 
  747         sb->sb_lastrecord = sb->sb_mb;
  748         SBLASTRECORDCHK(sb);
  749 }
  750 
  751 /*
  752  * This version of sbappend() should only be used when the caller
  753  * absolutely knows that there will never be more than one record
  754  * in the socket buffer, that is, a stream protocol (such as TCP).
  755  */
  756 void
  757 sbappendstream(struct sockbuf *sb, struct mbuf *m)
  758 {
  759 
  760         SOCKBUF_LOCK(sb);
  761         sbappendstream_locked(sb, m);
  762         SOCKBUF_UNLOCK(sb);
  763 }
  764 
  765 #ifdef SOCKBUF_DEBUG
  766 void
  767 sbcheck(sb)
  768         struct sockbuf *sb;
  769 {
  770         struct mbuf *m;
  771         struct mbuf *n = 0;
  772         u_long len = 0, mbcnt = 0;
  773 
  774         SOCKBUF_LOCK_ASSERT(sb);
  775 
  776         for (m = sb->sb_mb; m; m = n) {
  777             n = m->m_nextpkt;
  778             for (; m; m = m->m_next) {
  779                 len += m->m_len;
  780                 mbcnt += MSIZE;
  781                 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
  782                         mbcnt += m->m_ext.ext_size;
  783             }
  784         }
  785         if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
  786                 printf("cc %ld != %u || mbcnt %ld != %u\n", len, sb->sb_cc,
  787                     mbcnt, sb->sb_mbcnt);
  788                 panic("sbcheck");
  789         }
  790 }
  791 #endif
  792 
  793 /*
  794  * As above, except the mbuf chain
  795  * begins a new record.
  796  */
  797 void
  798 sbappendrecord_locked(sb, m0)
  799         register struct sockbuf *sb;
  800         register struct mbuf *m0;
  801 {
  802         register struct mbuf *m;
  803 
  804         SOCKBUF_LOCK_ASSERT(sb);
  805 
  806         if (m0 == 0)
  807                 return;
  808         m = sb->sb_mb;
  809         if (m)
  810                 while (m->m_nextpkt)
  811                         m = m->m_nextpkt;
  812         /*
  813          * Put the first mbuf on the queue.
  814          * Note this permits zero length records.
  815          */
  816         sballoc(sb, m0);
  817         SBLASTRECORDCHK(sb);
  818         SBLINKRECORD(sb, m0);
  819         if (m)
  820                 m->m_nextpkt = m0;
  821         else
  822                 sb->sb_mb = m0;
  823         m = m0->m_next;
  824         m0->m_next = 0;
  825         if (m && (m0->m_flags & M_EOR)) {
  826                 m0->m_flags &= ~M_EOR;
  827                 m->m_flags |= M_EOR;
  828         }
  829         sbcompress(sb, m, m0);
  830 }
  831 
  832 /*
  833  * As above, except the mbuf chain
  834  * begins a new record.
  835  */
  836 void
  837 sbappendrecord(sb, m0)
  838         register struct sockbuf *sb;
  839         register struct mbuf *m0;
  840 {
  841 
  842         SOCKBUF_LOCK(sb);
  843         sbappendrecord_locked(sb, m0);
  844         SOCKBUF_UNLOCK(sb);
  845 }
  846 
  847 /*
  848  * As above except that OOB data
  849  * is inserted at the beginning of the sockbuf,
  850  * but after any other OOB data.
  851  */
  852 void
  853 sbinsertoob_locked(sb, m0)
  854         register struct sockbuf *sb;
  855         register struct mbuf *m0;
  856 {
  857         register struct mbuf *m;
  858         register struct mbuf **mp;
  859 
  860         SOCKBUF_LOCK_ASSERT(sb);
  861 
  862         if (m0 == 0)
  863                 return;
  864         for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
  865             m = *mp;
  866             again:
  867                 switch (m->m_type) {
  868 
  869                 case MT_OOBDATA:
  870                         continue;               /* WANT next train */
  871 
  872                 case MT_CONTROL:
  873                         m = m->m_next;
  874                         if (m)
  875                                 goto again;     /* inspect THIS train further */
  876                 }
  877                 break;
  878         }
  879         /*
  880          * Put the first mbuf on the queue.
  881          * Note this permits zero length records.
  882          */
  883         sballoc(sb, m0);
  884         m0->m_nextpkt = *mp;
  885         *mp = m0;
  886         m = m0->m_next;
  887         m0->m_next = 0;
  888         if (m && (m0->m_flags & M_EOR)) {
  889                 m0->m_flags &= ~M_EOR;
  890                 m->m_flags |= M_EOR;
  891         }
  892         sbcompress(sb, m, m0);
  893 }
  894 
  895 /*
  896  * As above except that OOB data
  897  * is inserted at the beginning of the sockbuf,
  898  * but after any other OOB data.
  899  */
  900 void
  901 sbinsertoob(sb, m0)
  902         register struct sockbuf *sb;
  903         register struct mbuf *m0;
  904 {
  905 
  906         SOCKBUF_LOCK(sb);
  907         sbinsertoob_locked(sb, m0);
  908         SOCKBUF_UNLOCK(sb);
  909 }
  910 
  911 /*
  912  * Append address and data, and optionally, control (ancillary) data
  913  * to the receive queue of a socket.  If present,
  914  * m0 must include a packet header with total length.
  915  * Returns 0 if no space in sockbuf or insufficient mbufs.
  916  */
  917 int
  918 sbappendaddr_locked(sb, asa, m0, control)
  919         struct sockbuf *sb;
  920         const struct sockaddr *asa;
  921         struct mbuf *m0, *control;
  922 {
  923         struct mbuf *m, *n, *nlast;
  924         int space = asa->sa_len;
  925 
  926         SOCKBUF_LOCK_ASSERT(sb);
  927 
  928         if (m0 && (m0->m_flags & M_PKTHDR) == 0)
  929                 panic("sbappendaddr_locked");
  930         if (m0)
  931                 space += m0->m_pkthdr.len;
  932         space += m_length(control, &n);
  933 
  934         if (space > sbspace(sb))
  935                 return (0);
  936 #if MSIZE <= 256
  937         if (asa->sa_len > MLEN)
  938                 return (0);
  939 #endif
  940         MGET(m, M_DONTWAIT, MT_SONAME);
  941         if (m == 0)
  942                 return (0);
  943         m->m_len = asa->sa_len;
  944         bcopy(asa, mtod(m, caddr_t), asa->sa_len);
  945         if (n)
  946                 n->m_next = m0;         /* concatenate data to control */
  947         else
  948                 control = m0;
  949         m->m_next = control;
  950         for (n = m; n->m_next != NULL; n = n->m_next)
  951                 sballoc(sb, n);
  952         sballoc(sb, n);
  953         nlast = n;
  954         SBLINKRECORD(sb, m);
  955 
  956         sb->sb_mbtail = nlast;
  957         SBLASTMBUFCHK(sb);
  958 
  959         SBLASTRECORDCHK(sb);
  960         return (1);
  961 }
  962 
  963 /*
  964  * Append address and data, and optionally, control (ancillary) data
  965  * to the receive queue of a socket.  If present,
  966  * m0 must include a packet header with total length.
  967  * Returns 0 if no space in sockbuf or insufficient mbufs.
  968  */
  969 int
  970 sbappendaddr(sb, asa, m0, control)
  971         struct sockbuf *sb;
  972         const struct sockaddr *asa;
  973         struct mbuf *m0, *control;
  974 {
  975         int retval;
  976 
  977         SOCKBUF_LOCK(sb);
  978         retval = sbappendaddr_locked(sb, asa, m0, control);
  979         SOCKBUF_UNLOCK(sb);
  980         return (retval);
  981 }
  982 
  983 int
  984 sbappendcontrol_locked(sb, m0, control)
  985         struct sockbuf *sb;
  986         struct mbuf *control, *m0;
  987 {
  988         struct mbuf *m, *n, *mlast;
  989         int space;
  990 
  991         SOCKBUF_LOCK_ASSERT(sb);
  992 
  993         if (control == 0)
  994                 panic("sbappendcontrol_locked");
  995         space = m_length(control, &n) + m_length(m0, NULL);
  996 
  997         if (space > sbspace(sb))
  998                 return (0);
  999         n->m_next = m0;                 /* concatenate data to control */
 1000 
 1001         SBLASTRECORDCHK(sb);
 1002 
 1003         for (m = control; m->m_next; m = m->m_next)
 1004                 sballoc(sb, m);
 1005         sballoc(sb, m);
 1006         mlast = m;
 1007         SBLINKRECORD(sb, control);
 1008 
 1009         sb->sb_mbtail = mlast;
 1010         SBLASTMBUFCHK(sb);
 1011 
 1012         SBLASTRECORDCHK(sb);
 1013         return (1);
 1014 }
 1015 
 1016 int
 1017 sbappendcontrol(sb, m0, control)
 1018         struct sockbuf *sb;
 1019         struct mbuf *control, *m0;
 1020 {
 1021         int retval;
 1022 
 1023         SOCKBUF_LOCK(sb);
 1024         retval = sbappendcontrol_locked(sb, m0, control);
 1025         SOCKBUF_UNLOCK(sb);
 1026         return (retval);
 1027 }
 1028 
 1029 /*
 1030  * Append the data in mbuf chain (m) into the socket buffer sb following mbuf
 1031  * (n).  If (n) is NULL, the buffer is presumed empty.
 1032  *
 1033  * When the data is compressed, mbufs in the chain may be handled in one of
 1034  * three ways:
 1035  *
 1036  * (1) The mbuf may simply be dropped, if it contributes nothing (no data, no
 1037  *     record boundary, and no change in data type).
 1038  *
 1039  * (2) The mbuf may be coalesced -- i.e., data in the mbuf may be copied into
 1040  *     an mbuf already in the socket buffer.  This can occur if an
 1041  *     appropriate mbuf exists, there is room, and no merging of data types
 1042  *     will occur.
 1043  *
 1044  * (3) The mbuf may be appended to the end of the existing mbuf chain.
 1045  *
 1046  * If any of the new mbufs is marked as M_EOR, mark the last mbuf appended as
 1047  * end-of-record.
 1048  */
 1049 void
 1050 sbcompress(sb, m, n)
 1051         register struct sockbuf *sb;
 1052         register struct mbuf *m, *n;
 1053 {
 1054         register int eor = 0;
 1055         register struct mbuf *o;
 1056 
 1057         SOCKBUF_LOCK_ASSERT(sb);
 1058 
 1059         while (m) {
 1060                 eor |= m->m_flags & M_EOR;
 1061                 if (m->m_len == 0 &&
 1062                     (eor == 0 ||
 1063                      (((o = m->m_next) || (o = n)) &&
 1064                       o->m_type == m->m_type))) {
 1065                         if (sb->sb_lastrecord == m)
 1066                                 sb->sb_lastrecord = m->m_next;
 1067                         m = m_free(m);
 1068                         continue;
 1069                 }
 1070                 if (n && (n->m_flags & M_EOR) == 0 &&
 1071                     M_WRITABLE(n) &&
 1072                     m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
 1073                     m->m_len <= M_TRAILINGSPACE(n) &&
 1074                     n->m_type == m->m_type) {
 1075                         bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
 1076                             (unsigned)m->m_len);
 1077                         n->m_len += m->m_len;
 1078                         sb->sb_cc += m->m_len;
 1079                         if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
 1080                             m->m_type != MT_OOBDATA)
 1081                                 /* XXX: Probably don't need.*/
 1082                                 sb->sb_ctl += m->m_len;
 1083                         m = m_free(m);
 1084                         continue;
 1085                 }
 1086                 if (n)
 1087                         n->m_next = m;
 1088                 else
 1089                         sb->sb_mb = m;
 1090                 sb->sb_mbtail = m;
 1091                 sballoc(sb, m);
 1092                 n = m;
 1093                 m->m_flags &= ~M_EOR;
 1094                 m = m->m_next;
 1095                 n->m_next = 0;
 1096         }
 1097         if (eor) {
 1098                 KASSERT(n != NULL, ("sbcompress: eor && n == NULL"));
 1099                 n->m_flags |= eor;
 1100         }
 1101         SBLASTMBUFCHK(sb);
 1102 }
 1103 
 1104 /*
 1105  * Free all mbufs in a sockbuf.
 1106  * Check that all resources are reclaimed.
 1107  */
 1108 void
 1109 sbflush_locked(sb)
 1110         register struct sockbuf *sb;
 1111 {
 1112 
 1113         SOCKBUF_LOCK_ASSERT(sb);
 1114 
 1115         if (sb->sb_flags & SB_LOCK)
 1116                 panic("sbflush_locked: locked");
 1117         while (sb->sb_mbcnt) {
 1118                 /*
 1119                  * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
 1120                  * we would loop forever. Panic instead.
 1121                  */
 1122                 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
 1123                         break;
 1124                 sbdrop_locked(sb, (int)sb->sb_cc);
 1125         }
 1126         if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
 1127                 panic("sbflush_locked: cc %u || mb %p || mbcnt %u", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
 1128 }
 1129 
 1130 void
 1131 sbflush(sb)
 1132         register struct sockbuf *sb;
 1133 {
 1134 
 1135         SOCKBUF_LOCK(sb);
 1136         sbflush_locked(sb);
 1137         SOCKBUF_UNLOCK(sb);
 1138 }
 1139 
 1140 /*
 1141  * Drop data from (the front of) a sockbuf.
 1142  */
 1143 void
 1144 sbdrop_locked(sb, len)
 1145         register struct sockbuf *sb;
 1146         register int len;
 1147 {
 1148         register struct mbuf *m;
 1149         struct mbuf *next;
 1150 
 1151         SOCKBUF_LOCK_ASSERT(sb);
 1152 
 1153         next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
 1154         while (len > 0) {
 1155                 if (m == 0) {
 1156                         if (next == 0)
 1157                                 panic("sbdrop");
 1158                         m = next;
 1159                         next = m->m_nextpkt;
 1160                         continue;
 1161                 }
 1162                 if (m->m_len > len) {
 1163                         m->m_len -= len;
 1164                         m->m_data += len;
 1165                         sb->sb_cc -= len;
 1166                         if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
 1167                             m->m_type != MT_OOBDATA)
 1168                                 sb->sb_ctl -= len;
 1169                         break;
 1170                 }
 1171                 len -= m->m_len;
 1172                 sbfree(sb, m);
 1173                 m = m_free(m);
 1174         }
 1175         while (m && m->m_len == 0) {
 1176                 sbfree(sb, m);
 1177                 m = m_free(m);
 1178         }
 1179         if (m) {
 1180                 sb->sb_mb = m;
 1181                 m->m_nextpkt = next;
 1182         } else
 1183                 sb->sb_mb = next;
 1184         /*
 1185          * First part is an inline SB_EMPTY_FIXUP().  Second part
 1186          * makes sure sb_lastrecord is up-to-date if we dropped
 1187          * part of the last record.
 1188          */
 1189         m = sb->sb_mb;
 1190         if (m == NULL) {
 1191                 sb->sb_mbtail = NULL;
 1192                 sb->sb_lastrecord = NULL;
 1193         } else if (m->m_nextpkt == NULL) {
 1194                 sb->sb_lastrecord = m;
 1195         }
 1196 }
 1197 
 1198 /*
 1199  * Drop data from (the front of) a sockbuf.
 1200  */
 1201 void
 1202 sbdrop(sb, len)
 1203         register struct sockbuf *sb;
 1204         register int len;
 1205 {
 1206 
 1207         SOCKBUF_LOCK(sb);
 1208         sbdrop_locked(sb, len);
 1209         SOCKBUF_UNLOCK(sb);
 1210 }
 1211 
 1212 /*
 1213  * Drop a record off the front of a sockbuf
 1214  * and move the next record to the front.
 1215  */
 1216 void
 1217 sbdroprecord_locked(sb)
 1218         register struct sockbuf *sb;
 1219 {
 1220         register struct mbuf *m;
 1221 
 1222         SOCKBUF_LOCK_ASSERT(sb);
 1223 
 1224         m = sb->sb_mb;
 1225         if (m) {
 1226                 sb->sb_mb = m->m_nextpkt;
 1227                 do {
 1228                         sbfree(sb, m);
 1229                         m = m_free(m);
 1230                 } while (m);
 1231         }
 1232         SB_EMPTY_FIXUP(sb);
 1233 }
 1234 
 1235 /*
 1236  * Drop a record off the front of a sockbuf
 1237  * and move the next record to the front.
 1238  */
 1239 void
 1240 sbdroprecord(sb)
 1241         register struct sockbuf *sb;
 1242 {
 1243 
 1244         SOCKBUF_LOCK(sb);
 1245         sbdroprecord_locked(sb);
 1246         SOCKBUF_UNLOCK(sb);
 1247 }
 1248 
 1249 /*
 1250  * Create a "control" mbuf containing the specified data
 1251  * with the specified type for presentation on a socket buffer.
 1252  */
 1253 struct mbuf *
 1254 sbcreatecontrol(p, size, type, level)
 1255         caddr_t p;
 1256         register int size;
 1257         int type, level;
 1258 {
 1259         register struct cmsghdr *cp;
 1260         struct mbuf *m;
 1261 
 1262         if (CMSG_SPACE((u_int)size) > MCLBYTES)
 1263                 return ((struct mbuf *) NULL);
 1264         if (CMSG_SPACE((u_int)size) > MLEN)
 1265                 m = m_getcl(M_DONTWAIT, MT_CONTROL, 0);
 1266         else
 1267                 m = m_get(M_DONTWAIT, MT_CONTROL);
 1268         if (m == NULL)
 1269                 return ((struct mbuf *) NULL);
 1270         cp = mtod(m, struct cmsghdr *);
 1271         m->m_len = 0;
 1272         KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
 1273             ("sbcreatecontrol: short mbuf"));
 1274         if (p != NULL)
 1275                 (void)memcpy(CMSG_DATA(cp), p, size);
 1276         m->m_len = CMSG_SPACE(size);
 1277         cp->cmsg_len = CMSG_LEN(size);
 1278         cp->cmsg_level = level;
 1279         cp->cmsg_type = type;
 1280         return (m);
 1281 }
 1282 
 1283 /*
 1284  * Some routines that return EOPNOTSUPP for entry points that are not
 1285  * supported by a protocol.  Fill in as needed.
 1286  */
 1287 int
 1288 pru_abort_notsupp(struct socket *so)
 1289 {
 1290         return EOPNOTSUPP;
 1291 }
 1292 
 1293 int
 1294 pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
 1295 {
 1296         return EOPNOTSUPP;
 1297 }
 1298 
 1299 int
 1300 pru_attach_notsupp(struct socket *so, int proto, struct thread *td)
 1301 {
 1302         return EOPNOTSUPP;
 1303 }
 1304 
 1305 int
 1306 pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
 1307 {
 1308         return EOPNOTSUPP;
 1309 }
 1310 
 1311 int
 1312 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
 1313 {
 1314         return EOPNOTSUPP;
 1315 }
 1316 
 1317 int
 1318 pru_connect2_notsupp(struct socket *so1, struct socket *so2)
 1319 {
 1320         return EOPNOTSUPP;
 1321 }
 1322 
 1323 int
 1324 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
 1325         struct ifnet *ifp, struct thread *td)
 1326 {
 1327         return EOPNOTSUPP;
 1328 }
 1329 
 1330 int
 1331 pru_detach_notsupp(struct socket *so)
 1332 {
 1333         return EOPNOTSUPP;
 1334 }
 1335 
 1336 int
 1337 pru_disconnect_notsupp(struct socket *so)
 1338 {
 1339         return EOPNOTSUPP;
 1340 }
 1341 
 1342 int
 1343 pru_listen_notsupp(struct socket *so, struct thread *td)
 1344 {
 1345         return EOPNOTSUPP;
 1346 }
 1347 
 1348 int
 1349 pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam)
 1350 {
 1351         return EOPNOTSUPP;
 1352 }
 1353 
 1354 int
 1355 pru_rcvd_notsupp(struct socket *so, int flags)
 1356 {
 1357         return EOPNOTSUPP;
 1358 }
 1359 
 1360 int
 1361 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
 1362 {
 1363         return EOPNOTSUPP;
 1364 }
 1365 
 1366 int
 1367 pru_send_notsupp(struct socket *so, int flags, struct mbuf *m,
 1368         struct sockaddr *addr, struct mbuf *control, struct thread *td)
 1369 {
 1370         return EOPNOTSUPP;
 1371 }
 1372 
 1373 /*
 1374  * This isn't really a ``null'' operation, but it's the default one
 1375  * and doesn't do anything destructive.
 1376  */
 1377 int
 1378 pru_sense_null(struct socket *so, struct stat *sb)
 1379 {
 1380         sb->st_blksize = so->so_snd.sb_hiwat;
 1381         return 0;
 1382 }
 1383 
 1384 int
 1385 pru_shutdown_notsupp(struct socket *so)
 1386 {
 1387         return EOPNOTSUPP;
 1388 }
 1389 
 1390 int
 1391 pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam)
 1392 {
 1393         return EOPNOTSUPP;
 1394 }
 1395 
 1396 int
 1397 pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio,
 1398         struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
 1399 {
 1400         return EOPNOTSUPP;
 1401 }
 1402 
 1403 int
 1404 pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr,
 1405         struct uio *uio, struct mbuf **mp0, struct mbuf **controlp,
 1406         int *flagsp)
 1407 {
 1408         return EOPNOTSUPP;
 1409 }
 1410 
 1411 int
 1412 pru_sopoll_notsupp(struct socket *so, int events, struct ucred *cred,
 1413         struct thread *td)
 1414 {
 1415         return EOPNOTSUPP;
 1416 }
 1417 
 1418 /*
 1419  * For protocol types that don't keep cached copies of labels in their
 1420  * pcbs, provide a null sosetlabel that does a NOOP.
 1421  */
 1422 void
 1423 pru_sosetlabel_null(struct socket *so)
 1424 {
 1425 
 1426 }
 1427 
 1428 /*
 1429  * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
 1430  */
 1431 struct sockaddr *
 1432 sodupsockaddr(const struct sockaddr *sa, int mflags)
 1433 {
 1434         struct sockaddr *sa2;
 1435 
 1436         sa2 = malloc(sa->sa_len, M_SONAME, mflags);
 1437         if (sa2)
 1438                 bcopy(sa, sa2, sa->sa_len);
 1439         return sa2;
 1440 }
 1441 
 1442 /*
 1443  * Create an external-format (``xsocket'') structure using the information
 1444  * in the kernel-format socket structure pointed to by so.  This is done
 1445  * to reduce the spew of irrelevant information over this interface,
 1446  * to isolate user code from changes in the kernel structure, and
 1447  * potentially to provide information-hiding if we decide that
 1448  * some of this information should be hidden from users.
 1449  */
 1450 void
 1451 sotoxsocket(struct socket *so, struct xsocket *xso)
 1452 {
 1453         xso->xso_len = sizeof *xso;
 1454         xso->xso_so = so;
 1455         xso->so_type = so->so_type;
 1456         xso->so_options = so->so_options;
 1457         xso->so_linger = so->so_linger;
 1458         xso->so_state = so->so_state;
 1459         xso->so_pcb = so->so_pcb;
 1460         xso->xso_protocol = so->so_proto->pr_protocol;
 1461         xso->xso_family = so->so_proto->pr_domain->dom_family;
 1462         xso->so_qlen = so->so_qlen;
 1463         xso->so_incqlen = so->so_incqlen;
 1464         xso->so_qlimit = so->so_qlimit;
 1465         xso->so_timeo = so->so_timeo;
 1466         xso->so_error = so->so_error;
 1467         xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
 1468         xso->so_oobmark = so->so_oobmark;
 1469         sbtoxsockbuf(&so->so_snd, &xso->so_snd);
 1470         sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
 1471         xso->so_uid = so->so_cred->cr_uid;
 1472 }
 1473 
 1474 /*
 1475  * This does the same for sockbufs.  Note that the xsockbuf structure,
 1476  * since it is always embedded in a socket, does not include a self
 1477  * pointer nor a length.  We make this entry point public in case
 1478  * some other mechanism needs it.
 1479  */
 1480 void
 1481 sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
 1482 {
 1483         xsb->sb_cc = sb->sb_cc;
 1484         xsb->sb_hiwat = sb->sb_hiwat;
 1485         xsb->sb_mbcnt = sb->sb_mbcnt;
 1486         xsb->sb_mbmax = sb->sb_mbmax;
 1487         xsb->sb_lowat = sb->sb_lowat;
 1488         xsb->sb_flags = sb->sb_flags;
 1489         xsb->sb_timeo = sb->sb_timeo;
 1490 }
 1491 
 1492 /*
 1493  * Here is the definition of some of the basic objects in the kern.ipc
 1494  * branch of the MIB.
 1495  */
 1496 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
 1497 
 1498 /* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
 1499 static int dummy;
 1500 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
 1501 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_ULONG|CTLFLAG_RW, 
 1502     &sb_max, 0, sysctl_handle_sb_max, "LU", "Maximum socket buffer size");
 1503 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RDTUN, 
 1504     &maxsockets, 0, "Maximum number of sockets avaliable");
 1505 SYSCTL_ULONG(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
 1506     &sb_efficiency, 0, "");
 1507 
 1508 /*
 1509  * Initialise maxsockets 
 1510  */
 1511 static void init_maxsockets(void *ignored)
 1512 {
 1513         TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
 1514         maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
 1515 }
 1516 SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);

Cache object: a793392d07aee166db409c30262fabef


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.