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/5.3/sys/kern/uipc_socket2.c 137026 2004-10-28 20:04:14Z maxim $");
   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         SOCK_LOCK(so);
  121         so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
  122         so->so_state |= SS_ISCONNECTED;
  123         SOCK_UNLOCK(so);
  124         ACCEPT_LOCK();
  125         head = so->so_head;
  126         if (head != NULL && (so->so_qstate & SQ_INCOMP)) {
  127                 if ((so->so_options & SO_ACCEPTFILTER) == 0) {
  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                         SOCK_LOCK(so);
  140                         so->so_upcall =
  141                             head->so_accf->so_accept_filter->accf_callback;
  142                         so->so_upcallarg = head->so_accf->so_accept_filter_arg;
  143                         so->so_rcv.sb_flags |= SB_UPCALL;
  144                         so->so_options &= ~SO_ACCEPTFILTER;
  145                         SOCK_UNLOCK(so);
  146                         so->so_upcall(so, so->so_upcallarg, M_TRYWAIT);
  147                 }
  148                 return;
  149         }
  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 separately acquires SOCK_LOCK(so) and
  163          * SOCKBUF_LOCK(&so->so_rcv) even though they are the same mutex to
  164          * avoid introducing the assumption  that they are the same.
  165          */
  166         SOCK_LOCK(so);
  167         so->so_state &= ~SS_ISCONNECTING;
  168         so->so_state |= SS_ISDISCONNECTING;
  169         SOCK_UNLOCK(so);
  170         SOCKBUF_LOCK(&so->so_rcv);
  171         so->so_rcv.sb_state |= SBS_CANTRCVMORE;
  172         sorwakeup_locked(so);
  173         SOCKBUF_LOCK(&so->so_snd);
  174         so->so_snd.sb_state |= SBS_CANTSENDMORE;
  175         sowwakeup_locked(so);
  176         wakeup(&so->so_timeo);
  177 }
  178 
  179 void
  180 soisdisconnected(so)
  181         register struct socket *so;
  182 {
  183 
  184         /*
  185          * XXXRW: This code separately acquires SOCK_LOCK(so) and
  186          * SOCKBUF_LOCK(&so->so_rcv) even though they are the same mutex to
  187          * avoid introducing the assumption  that they are the same.
  188          */
  189         /* XXXRW: so_state locking? */
  190         SOCK_LOCK(so);
  191         so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
  192         so->so_state |= SS_ISDISCONNECTED;
  193         SOCK_UNLOCK(so);
  194         SOCKBUF_LOCK(&so->so_rcv);
  195         so->so_rcv.sb_state |= SBS_CANTRCVMORE;
  196         sorwakeup_locked(so);
  197         SOCKBUF_LOCK(&so->so_snd);
  198         so->so_snd.sb_state |= SBS_CANTSENDMORE;
  199         sbdrop_locked(&so->so_snd, so->so_snd.sb_cc);
  200         sowwakeup_locked(so);
  201         wakeup(&so->so_timeo);
  202 }
  203 
  204 /*
  205  * When an attempt at a new connection is noted on a socket
  206  * which accepts connections, sonewconn is called.  If the
  207  * connection is possible (subject to space constraints, etc.)
  208  * then we allocate a new structure, propoerly linked into the
  209  * data structure of the original socket, and return this.
  210  * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
  211  *
  212  * note: the ref count on the socket is 0 on return
  213  */
  214 struct socket *
  215 sonewconn(head, connstatus)
  216         register struct socket *head;
  217         int connstatus;
  218 {
  219         register struct socket *so;
  220         int over;
  221 
  222         ACCEPT_LOCK();
  223         over = (head->so_qlen > 3 * head->so_qlimit / 2);
  224         ACCEPT_UNLOCK();
  225         if (over)
  226                 return ((struct socket *)0);
  227         so = soalloc(M_NOWAIT);
  228         if (so == NULL)
  229                 return ((struct socket *)0);
  230         if ((head->so_options & SO_ACCEPTFILTER) != 0)
  231                 connstatus = 0;
  232         so->so_head = head;
  233         so->so_type = head->so_type;
  234         so->so_options = head->so_options &~ SO_ACCEPTCONN;
  235         so->so_linger = head->so_linger;
  236         so->so_state = head->so_state | SS_NOFDREF;
  237         so->so_proto = head->so_proto;
  238         so->so_timeo = head->so_timeo;
  239         so->so_cred = crhold(head->so_cred);
  240 #ifdef MAC
  241         SOCK_LOCK(head);
  242         mac_create_socket_from_socket(head, so);
  243         SOCK_UNLOCK(head);
  244 #endif
  245         knlist_init(&so->so_rcv.sb_sel.si_note, SOCKBUF_MTX(&so->so_rcv));
  246         knlist_init(&so->so_snd.sb_sel.si_note, SOCKBUF_MTX(&so->so_snd));
  247         if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
  248             (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
  249                 sodealloc(so);
  250                 return ((struct socket *)0);
  251         }
  252         ACCEPT_LOCK();
  253         if (connstatus) {
  254                 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
  255                 so->so_qstate |= SQ_COMP;
  256                 head->so_qlen++;
  257         } else {
  258                 /*
  259                  * XXXRW: Keep removing sockets from the head until there's
  260                  * room for us to insert on the tail.  In pre-locking
  261                  * revisions, this was a simple if(), but as we could be
  262                  * racing with other threads and soabort() requires dropping
  263                  * locks, we must loop waiting for the condition to be true.
  264                  */
  265                 while (head->so_incqlen > head->so_qlimit) {
  266                         struct socket *sp;
  267                         sp = TAILQ_FIRST(&head->so_incomp);
  268                         TAILQ_REMOVE(&so->so_incomp, sp, so_list);
  269                         head->so_incqlen--;
  270                         sp->so_qstate &= ~SQ_INCOMP;
  271                         sp->so_head = NULL;
  272                         ACCEPT_UNLOCK();
  273                         (void) soabort(sp);
  274                         ACCEPT_LOCK();
  275                 }
  276                 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
  277                 so->so_qstate |= SQ_INCOMP;
  278                 head->so_incqlen++;
  279         }
  280         ACCEPT_UNLOCK();
  281         if (connstatus) {
  282                 so->so_state |= connstatus;
  283                 sorwakeup(head);
  284                 wakeup_one(&head->so_timeo);
  285         }
  286         return (so);
  287 }
  288 
  289 /*
  290  * Socantsendmore indicates that no more data will be sent on the
  291  * socket; it would normally be applied to a socket when the user
  292  * informs the system that no more data is to be sent, by the protocol
  293  * code (in case PRU_SHUTDOWN).  Socantrcvmore indicates that no more data
  294  * will be received, and will normally be applied to the socket by a
  295  * protocol when it detects that the peer will send no more data.
  296  * Data queued for reading in the socket may yet be read.
  297  */
  298 void
  299 socantsendmore_locked(so)
  300         struct socket *so;
  301 {
  302 
  303         SOCKBUF_LOCK_ASSERT(&so->so_snd);
  304 
  305         so->so_snd.sb_state |= SBS_CANTSENDMORE;
  306         sowwakeup_locked(so);
  307         mtx_assert(SOCKBUF_MTX(&so->so_snd), MA_NOTOWNED);
  308 }
  309 
  310 void
  311 socantsendmore(so)
  312         struct socket *so;
  313 {
  314 
  315         SOCKBUF_LOCK(&so->so_snd);
  316         socantsendmore_locked(so);
  317         mtx_assert(SOCKBUF_MTX(&so->so_snd), MA_NOTOWNED);
  318 }
  319 
  320 void
  321 socantrcvmore_locked(so)
  322         struct socket *so;
  323 {
  324 
  325         SOCKBUF_LOCK_ASSERT(&so->so_rcv);
  326 
  327         so->so_rcv.sb_state |= SBS_CANTRCVMORE;
  328         sorwakeup_locked(so);
  329         mtx_assert(SOCKBUF_MTX(&so->so_rcv), MA_NOTOWNED);
  330 }
  331 
  332 void
  333 socantrcvmore(so)
  334         struct socket *so;
  335 {
  336 
  337         SOCKBUF_LOCK(&so->so_rcv);
  338         socantrcvmore_locked(so);
  339         mtx_assert(SOCKBUF_MTX(&so->so_rcv), MA_NOTOWNED);
  340 }
  341 
  342 /*
  343  * Wait for data to arrive at/drain from a socket buffer.
  344  */
  345 int
  346 sbwait(sb)
  347         struct sockbuf *sb;
  348 {
  349 
  350         SOCKBUF_LOCK_ASSERT(sb);
  351 
  352         sb->sb_flags |= SB_WAIT;
  353         return (msleep(&sb->sb_cc, &sb->sb_mtx,
  354             (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
  355             sb->sb_timeo));
  356 }
  357 
  358 /*
  359  * Lock a sockbuf already known to be locked;
  360  * return any error returned from sleep (EINTR).
  361  */
  362 int
  363 sb_lock(sb)
  364         register struct sockbuf *sb;
  365 {
  366         int error;
  367 
  368         SOCKBUF_LOCK_ASSERT(sb);
  369 
  370         while (sb->sb_flags & SB_LOCK) {
  371                 sb->sb_flags |= SB_WANT;
  372                 error = msleep(&sb->sb_flags, &sb->sb_mtx,
  373                     (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
  374                     "sblock", 0);
  375                 if (error)
  376                         return (error);
  377         }
  378         sb->sb_flags |= SB_LOCK;
  379         return (0);
  380 }
  381 
  382 /*
  383  * Wakeup processes waiting on a socket buffer.  Do asynchronous
  384  * notification via SIGIO if the socket has the SS_ASYNC flag set.
  385  *
  386  * Called with the socket buffer lock held; will release the lock by the end
  387  * of the function.  This allows the caller to acquire the socket buffer lock
  388  * while testing for the need for various sorts of wakeup and hold it through
  389  * to the point where it's no longer required.  We currently hold the lock
  390  * through calls out to other subsystems (with the exception of kqueue), and
  391  * then release it to avoid lock order issues.  It's not clear that's
  392  * correct.
  393  */
  394 void
  395 sowakeup(so, sb)
  396         register struct socket *so;
  397         register struct sockbuf *sb;
  398 {
  399 
  400         SOCKBUF_LOCK_ASSERT(sb);
  401 
  402         selwakeuppri(&sb->sb_sel, PSOCK);
  403         sb->sb_flags &= ~SB_SEL;
  404         if (sb->sb_flags & SB_WAIT) {
  405                 sb->sb_flags &= ~SB_WAIT;
  406                 wakeup(&sb->sb_cc);
  407         }
  408         KNOTE_LOCKED(&sb->sb_sel.si_note, 0);
  409         SOCKBUF_UNLOCK(sb);
  410         if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
  411                 pgsigio(&so->so_sigio, SIGIO, 0);
  412         if (sb->sb_flags & SB_UPCALL)
  413                 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
  414         if (sb->sb_flags & SB_AIO)
  415                 aio_swake(so, sb);
  416         mtx_assert(SOCKBUF_MTX(sb), MA_NOTOWNED);
  417 }
  418 
  419 /*
  420  * Socket buffer (struct sockbuf) utility routines.
  421  *
  422  * Each socket contains two socket buffers: one for sending data and
  423  * one for receiving data.  Each buffer contains a queue of mbufs,
  424  * information about the number of mbufs and amount of data in the
  425  * queue, and other fields allowing select() statements and notification
  426  * on data availability to be implemented.
  427  *
  428  * Data stored in a socket buffer is maintained as a list of records.
  429  * Each record is a list of mbufs chained together with the m_next
  430  * field.  Records are chained together with the m_nextpkt field. The upper
  431  * level routine soreceive() expects the following conventions to be
  432  * observed when placing information in the receive buffer:
  433  *
  434  * 1. If the protocol requires each message be preceded by the sender's
  435  *    name, then a record containing that name must be present before
  436  *    any associated data (mbuf's must be of type MT_SONAME).
  437  * 2. If the protocol supports the exchange of ``access rights'' (really
  438  *    just additional data associated with the message), and there are
  439  *    ``rights'' to be received, then a record containing this data
  440  *    should be present (mbuf's must be of type MT_RIGHTS).
  441  * 3. If a name or rights record exists, then it must be followed by
  442  *    a data record, perhaps of zero length.
  443  *
  444  * Before using a new socket structure it is first necessary to reserve
  445  * buffer space to the socket, by calling sbreserve().  This should commit
  446  * some of the available buffer space in the system buffer pool for the
  447  * socket (currently, it does nothing but enforce limits).  The space
  448  * should be released by calling sbrelease() when the socket is destroyed.
  449  */
  450 
  451 int
  452 soreserve(so, sndcc, rcvcc)
  453         register struct socket *so;
  454         u_long sndcc, rcvcc;
  455 {
  456         struct thread *td = curthread;
  457 
  458         SOCKBUF_LOCK(&so->so_snd);
  459         SOCKBUF_LOCK(&so->so_rcv);
  460         if (sbreserve_locked(&so->so_snd, sndcc, so, td) == 0)
  461                 goto bad;
  462         if (sbreserve_locked(&so->so_rcv, rcvcc, so, td) == 0)
  463                 goto bad2;
  464         if (so->so_rcv.sb_lowat == 0)
  465                 so->so_rcv.sb_lowat = 1;
  466         if (so->so_snd.sb_lowat == 0)
  467                 so->so_snd.sb_lowat = MCLBYTES;
  468         if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
  469                 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
  470         SOCKBUF_UNLOCK(&so->so_rcv);
  471         SOCKBUF_UNLOCK(&so->so_snd);
  472         return (0);
  473 bad2:
  474         sbrelease_locked(&so->so_snd, so);
  475 bad:
  476         SOCKBUF_UNLOCK(&so->so_rcv);
  477         SOCKBUF_UNLOCK(&so->so_snd);
  478         return (ENOBUFS);
  479 }
  480 
  481 static int
  482 sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
  483 {
  484         int error = 0;
  485         u_long old_sb_max = sb_max;
  486 
  487         error = SYSCTL_OUT(req, arg1, sizeof(u_long));
  488         if (error || !req->newptr)
  489                 return (error);
  490         error = SYSCTL_IN(req, arg1, sizeof(u_long));
  491         if (error)
  492                 return (error);
  493         if (sb_max < MSIZE + MCLBYTES) {
  494                 sb_max = old_sb_max;
  495                 return (EINVAL);
  496         }
  497         sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
  498         return (0);
  499 }
  500         
  501 /*
  502  * Allot mbufs to a sockbuf.
  503  * Attempt to scale mbmax so that mbcnt doesn't become limiting
  504  * if buffering efficiency is near the normal case.
  505  */
  506 int
  507 sbreserve_locked(sb, cc, so, td)
  508         struct sockbuf *sb;
  509         u_long cc;
  510         struct socket *so;
  511         struct thread *td;
  512 {
  513         rlim_t sbsize_limit;
  514 
  515         SOCKBUF_LOCK_ASSERT(sb);
  516 
  517         /*
  518          * td will only be NULL when we're in an interrupt
  519          * (e.g. in tcp_input())
  520          */
  521         if (cc > sb_max_adj)
  522                 return (0);
  523         if (td != NULL) {
  524                 PROC_LOCK(td->td_proc);
  525                 sbsize_limit = lim_cur(td->td_proc, RLIMIT_SBSIZE);
  526                 PROC_UNLOCK(td->td_proc);
  527         } else
  528                 sbsize_limit = RLIM_INFINITY;
  529         if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
  530             sbsize_limit))
  531                 return (0);
  532         sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
  533         if (sb->sb_lowat > sb->sb_hiwat)
  534                 sb->sb_lowat = sb->sb_hiwat;
  535         return (1);
  536 }
  537 
  538 int
  539 sbreserve(sb, cc, so, td)
  540         struct sockbuf *sb;
  541         u_long cc;
  542         struct socket *so;
  543         struct thread *td;
  544 {
  545         int error;
  546 
  547         SOCKBUF_LOCK(sb);
  548         error = sbreserve_locked(sb, cc, so, td);
  549         SOCKBUF_UNLOCK(sb);
  550         return (error);
  551 }
  552 
  553 /*
  554  * Free mbufs held by a socket, and reserved mbuf space.
  555  */
  556 void
  557 sbrelease_locked(sb, so)
  558         struct sockbuf *sb;
  559         struct socket *so;
  560 {
  561 
  562         SOCKBUF_LOCK_ASSERT(sb);
  563 
  564         sbflush_locked(sb);
  565         (void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
  566             RLIM_INFINITY);
  567         sb->sb_mbmax = 0;
  568 }
  569 
  570 void
  571 sbrelease(sb, so)
  572         struct sockbuf *sb;
  573         struct socket *so;
  574 {
  575 
  576         SOCKBUF_LOCK(sb);
  577         sbrelease_locked(sb, so);
  578         SOCKBUF_UNLOCK(sb);
  579 }
  580 /*
  581  * Routines to add and remove
  582  * data from an mbuf queue.
  583  *
  584  * The routines sbappend() or sbappendrecord() are normally called to
  585  * append new mbufs to a socket buffer, after checking that adequate
  586  * space is available, comparing the function sbspace() with the amount
  587  * of data to be added.  sbappendrecord() differs from sbappend() in
  588  * that data supplied is treated as the beginning of a new record.
  589  * To place a sender's address, optional access rights, and data in a
  590  * socket receive buffer, sbappendaddr() should be used.  To place
  591  * access rights and data in a socket receive buffer, sbappendrights()
  592  * should be used.  In either case, the new data begins a new record.
  593  * Note that unlike sbappend() and sbappendrecord(), these routines check
  594  * for the caller that there will be enough space to store the data.
  595  * Each fails if there is not enough space, or if it cannot find mbufs
  596  * to store additional information in.
  597  *
  598  * Reliable protocols may use the socket send buffer to hold data
  599  * awaiting acknowledgement.  Data is normally copied from a socket
  600  * send buffer in a protocol with m_copy for output to a peer,
  601  * and then removing the data from the socket buffer with sbdrop()
  602  * or sbdroprecord() when the data is acknowledged by the peer.
  603  */
  604 
  605 #ifdef SOCKBUF_DEBUG
  606 void
  607 sblastrecordchk(struct sockbuf *sb, const char *file, int line)
  608 {
  609         struct mbuf *m = sb->sb_mb;
  610 
  611         SOCKBUF_LOCK_ASSERT(sb);
  612 
  613         while (m && m->m_nextpkt)
  614                 m = m->m_nextpkt;
  615 
  616         if (m != sb->sb_lastrecord) {
  617                 printf("%s: sb_mb %p sb_lastrecord %p last %p\n",
  618                         __func__, sb->sb_mb, sb->sb_lastrecord, m);
  619                 printf("packet chain:\n");
  620                 for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt)
  621                         printf("\t%p\n", m);
  622                 panic("%s from %s:%u", __func__, file, line);
  623         }
  624 }
  625 
  626 void
  627 sblastmbufchk(struct sockbuf *sb, const char *file, int line)
  628 {
  629         struct mbuf *m = sb->sb_mb;
  630         struct mbuf *n;
  631 
  632         SOCKBUF_LOCK_ASSERT(sb);
  633 
  634         while (m && m->m_nextpkt)
  635                 m = m->m_nextpkt;
  636 
  637         while (m && m->m_next)
  638                 m = m->m_next;
  639 
  640         if (m != sb->sb_mbtail) {
  641                 printf("%s: sb_mb %p sb_mbtail %p last %p\n",
  642                         __func__, sb->sb_mb, sb->sb_mbtail, m);
  643                 printf("packet tree:\n");
  644                 for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt) {
  645                         printf("\t");
  646                         for (n = m; n != NULL; n = n->m_next)
  647                                 printf("%p ", n);
  648                         printf("\n");
  649                 }
  650                 panic("%s from %s:%u", __func__, file, line);
  651         }
  652 }
  653 #endif /* SOCKBUF_DEBUG */
  654 
  655 #define SBLINKRECORD(sb, m0) do {                                       \
  656         SOCKBUF_LOCK_ASSERT(sb);                                        \
  657         if ((sb)->sb_lastrecord != NULL)                                \
  658                 (sb)->sb_lastrecord->m_nextpkt = (m0);                  \
  659         else                                                            \
  660                 (sb)->sb_mb = (m0);                                     \
  661         (sb)->sb_lastrecord = (m0);                                     \
  662 } while (/*CONSTCOND*/0)
  663 
  664 /*
  665  * Append mbuf chain m to the last record in the
  666  * socket buffer sb.  The additional space associated
  667  * the mbuf chain is recorded in sb.  Empty mbufs are
  668  * discarded and mbufs are compacted where possible.
  669  */
  670 void
  671 sbappend_locked(sb, m)
  672         struct sockbuf *sb;
  673         struct mbuf *m;
  674 {
  675         register struct mbuf *n;
  676 
  677         SOCKBUF_LOCK_ASSERT(sb);
  678 
  679         if (m == 0)
  680                 return;
  681 
  682         SBLASTRECORDCHK(sb);
  683         n = sb->sb_mb;
  684         if (n) {
  685                 while (n->m_nextpkt)
  686                         n = n->m_nextpkt;
  687                 do {
  688                         if (n->m_flags & M_EOR) {
  689                                 sbappendrecord_locked(sb, m); /* XXXXXX!!!! */
  690                                 return;
  691                         }
  692                 } while (n->m_next && (n = n->m_next));
  693         } else {
  694                 /*
  695                  * XXX Would like to simply use sb_mbtail here, but
  696                  * XXX I need to verify that I won't miss an EOR that
  697                  * XXX way.
  698                  */
  699                 if ((n = sb->sb_lastrecord) != NULL) {
  700                         do {
  701                                 if (n->m_flags & M_EOR) {
  702                                         sbappendrecord_locked(sb, m); /* XXXXXX!!!! */
  703                                         return;
  704                                 }
  705                         } while (n->m_next && (n = n->m_next));
  706                 } else {
  707                         /*
  708                          * If this is the first record in the socket buffer,
  709                          * it's also the last record.
  710                          */
  711                         sb->sb_lastrecord = m;
  712                 }
  713         }
  714         sbcompress(sb, m, n);
  715         SBLASTRECORDCHK(sb);
  716 }
  717 
  718 /*
  719  * Append mbuf chain m to the last record in the
  720  * socket buffer sb.  The additional space associated
  721  * the mbuf chain is recorded in sb.  Empty mbufs are
  722  * discarded and mbufs are compacted where possible.
  723  */
  724 void
  725 sbappend(sb, m)
  726         struct sockbuf *sb;
  727         struct mbuf *m;
  728 {
  729 
  730         SOCKBUF_LOCK(sb);
  731         sbappend_locked(sb, m);
  732         SOCKBUF_UNLOCK(sb);
  733 }
  734 
  735 /*
  736  * This version of sbappend() should only be used when the caller
  737  * absolutely knows that there will never be more than one record
  738  * in the socket buffer, that is, a stream protocol (such as TCP).
  739  */
  740 void
  741 sbappendstream_locked(struct sockbuf *sb, struct mbuf *m)
  742 {
  743         SOCKBUF_LOCK_ASSERT(sb);
  744 
  745         KASSERT(m->m_nextpkt == NULL,("sbappendstream 0"));
  746         KASSERT(sb->sb_mb == sb->sb_lastrecord,("sbappendstream 1"));
  747 
  748         SBLASTMBUFCHK(sb);
  749 
  750         sbcompress(sb, m, sb->sb_mbtail);
  751 
  752         sb->sb_lastrecord = sb->sb_mb;
  753         SBLASTRECORDCHK(sb);
  754 }
  755 
  756 /*
  757  * This version of sbappend() should only be used when the caller
  758  * absolutely knows that there will never be more than one record
  759  * in the socket buffer, that is, a stream protocol (such as TCP).
  760  */
  761 void
  762 sbappendstream(struct sockbuf *sb, struct mbuf *m)
  763 {
  764 
  765         SOCKBUF_LOCK(sb);
  766         sbappendstream_locked(sb, m);
  767         SOCKBUF_UNLOCK(sb);
  768 }
  769 
  770 #ifdef SOCKBUF_DEBUG
  771 void
  772 sbcheck(sb)
  773         struct sockbuf *sb;
  774 {
  775         struct mbuf *m;
  776         struct mbuf *n = 0;
  777         u_long len = 0, mbcnt = 0;
  778 
  779         SOCKBUF_LOCK_ASSERT(sb);
  780 
  781         for (m = sb->sb_mb; m; m = n) {
  782             n = m->m_nextpkt;
  783             for (; m; m = m->m_next) {
  784                 len += m->m_len;
  785                 mbcnt += MSIZE;
  786                 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
  787                         mbcnt += m->m_ext.ext_size;
  788             }
  789         }
  790         if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
  791                 printf("cc %ld != %u || mbcnt %ld != %u\n", len, sb->sb_cc,
  792                     mbcnt, sb->sb_mbcnt);
  793                 panic("sbcheck");
  794         }
  795 }
  796 #endif
  797 
  798 /*
  799  * As above, except the mbuf chain
  800  * begins a new record.
  801  */
  802 void
  803 sbappendrecord_locked(sb, m0)
  804         register struct sockbuf *sb;
  805         register struct mbuf *m0;
  806 {
  807         register struct mbuf *m;
  808 
  809         SOCKBUF_LOCK_ASSERT(sb);
  810 
  811         if (m0 == 0)
  812                 return;
  813         m = sb->sb_mb;
  814         if (m)
  815                 while (m->m_nextpkt)
  816                         m = m->m_nextpkt;
  817         /*
  818          * Put the first mbuf on the queue.
  819          * Note this permits zero length records.
  820          */
  821         sballoc(sb, m0);
  822         SBLASTRECORDCHK(sb);
  823         SBLINKRECORD(sb, m0);
  824         if (m)
  825                 m->m_nextpkt = m0;
  826         else
  827                 sb->sb_mb = m0;
  828         m = m0->m_next;
  829         m0->m_next = 0;
  830         if (m && (m0->m_flags & M_EOR)) {
  831                 m0->m_flags &= ~M_EOR;
  832                 m->m_flags |= M_EOR;
  833         }
  834         sbcompress(sb, m, m0);
  835 }
  836 
  837 /*
  838  * As above, except the mbuf chain
  839  * begins a new record.
  840  */
  841 void
  842 sbappendrecord(sb, m0)
  843         register struct sockbuf *sb;
  844         register struct mbuf *m0;
  845 {
  846 
  847         SOCKBUF_LOCK(sb);
  848         sbappendrecord_locked(sb, m0);
  849         SOCKBUF_UNLOCK(sb);
  850 }
  851 
  852 /*
  853  * As above except that OOB data
  854  * is inserted at the beginning of the sockbuf,
  855  * but after any other OOB data.
  856  */
  857 void
  858 sbinsertoob_locked(sb, m0)
  859         register struct sockbuf *sb;
  860         register struct mbuf *m0;
  861 {
  862         register struct mbuf *m;
  863         register struct mbuf **mp;
  864 
  865         SOCKBUF_LOCK_ASSERT(sb);
  866 
  867         if (m0 == 0)
  868                 return;
  869         for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
  870             m = *mp;
  871             again:
  872                 switch (m->m_type) {
  873 
  874                 case MT_OOBDATA:
  875                         continue;               /* WANT next train */
  876 
  877                 case MT_CONTROL:
  878                         m = m->m_next;
  879                         if (m)
  880                                 goto again;     /* inspect THIS train further */
  881                 }
  882                 break;
  883         }
  884         /*
  885          * Put the first mbuf on the queue.
  886          * Note this permits zero length records.
  887          */
  888         sballoc(sb, m0);
  889         m0->m_nextpkt = *mp;
  890         *mp = m0;
  891         m = m0->m_next;
  892         m0->m_next = 0;
  893         if (m && (m0->m_flags & M_EOR)) {
  894                 m0->m_flags &= ~M_EOR;
  895                 m->m_flags |= M_EOR;
  896         }
  897         sbcompress(sb, m, m0);
  898 }
  899 
  900 /*
  901  * As above except that OOB data
  902  * is inserted at the beginning of the sockbuf,
  903  * but after any other OOB data.
  904  */
  905 void
  906 sbinsertoob(sb, m0)
  907         register struct sockbuf *sb;
  908         register struct mbuf *m0;
  909 {
  910 
  911         SOCKBUF_LOCK(sb);
  912         sbinsertoob_locked(sb, m0);
  913         SOCKBUF_UNLOCK(sb);
  914 }
  915 
  916 /*
  917  * Append address and data, and optionally, control (ancillary) data
  918  * to the receive queue of a socket.  If present,
  919  * m0 must include a packet header with total length.
  920  * Returns 0 if no space in sockbuf or insufficient mbufs.
  921  */
  922 int
  923 sbappendaddr_locked(sb, asa, m0, control)
  924         struct sockbuf *sb;
  925         const struct sockaddr *asa;
  926         struct mbuf *m0, *control;
  927 {
  928         struct mbuf *m, *n, *nlast;
  929         int space = asa->sa_len;
  930 
  931         SOCKBUF_LOCK_ASSERT(sb);
  932 
  933         if (m0 && (m0->m_flags & M_PKTHDR) == 0)
  934                 panic("sbappendaddr_locked");
  935         if (m0)
  936                 space += m0->m_pkthdr.len;
  937         space += m_length(control, &n);
  938 
  939         if (space > sbspace(sb))
  940                 return (0);
  941 #if MSIZE <= 256
  942         if (asa->sa_len > MLEN)
  943                 return (0);
  944 #endif
  945         MGET(m, M_DONTWAIT, MT_SONAME);
  946         if (m == 0)
  947                 return (0);
  948         m->m_len = asa->sa_len;
  949         bcopy(asa, mtod(m, caddr_t), asa->sa_len);
  950         if (n)
  951                 n->m_next = m0;         /* concatenate data to control */
  952         else
  953                 control = m0;
  954         m->m_next = control;
  955         for (n = m; n->m_next != NULL; n = n->m_next)
  956                 sballoc(sb, n);
  957         sballoc(sb, n);
  958         nlast = n;
  959         SBLINKRECORD(sb, m);
  960 
  961         sb->sb_mbtail = nlast;
  962         SBLASTMBUFCHK(sb);
  963 
  964         SBLASTRECORDCHK(sb);
  965         return (1);
  966 }
  967 
  968 /*
  969  * Append address and data, and optionally, control (ancillary) data
  970  * to the receive queue of a socket.  If present,
  971  * m0 must include a packet header with total length.
  972  * Returns 0 if no space in sockbuf or insufficient mbufs.
  973  */
  974 int
  975 sbappendaddr(sb, asa, m0, control)
  976         struct sockbuf *sb;
  977         const struct sockaddr *asa;
  978         struct mbuf *m0, *control;
  979 {
  980         int retval;
  981 
  982         SOCKBUF_LOCK(sb);
  983         retval = sbappendaddr_locked(sb, asa, m0, control);
  984         SOCKBUF_UNLOCK(sb);
  985         return (retval);
  986 }
  987 
  988 int
  989 sbappendcontrol_locked(sb, m0, control)
  990         struct sockbuf *sb;
  991         struct mbuf *control, *m0;
  992 {
  993         struct mbuf *m, *n, *mlast;
  994         int space;
  995 
  996         SOCKBUF_LOCK_ASSERT(sb);
  997 
  998         if (control == 0)
  999                 panic("sbappendcontrol_locked");
 1000         space = m_length(control, &n) + m_length(m0, NULL);
 1001 
 1002         if (space > sbspace(sb))
 1003                 return (0);
 1004         n->m_next = m0;                 /* concatenate data to control */
 1005 
 1006         SBLASTRECORDCHK(sb);
 1007 
 1008         for (m = control; m->m_next; m = m->m_next)
 1009                 sballoc(sb, m);
 1010         sballoc(sb, m);
 1011         mlast = m;
 1012         SBLINKRECORD(sb, control);
 1013 
 1014         sb->sb_mbtail = mlast;
 1015         SBLASTMBUFCHK(sb);
 1016 
 1017         SBLASTRECORDCHK(sb);
 1018         return (1);
 1019 }
 1020 
 1021 int
 1022 sbappendcontrol(sb, m0, control)
 1023         struct sockbuf *sb;
 1024         struct mbuf *control, *m0;
 1025 {
 1026         int retval;
 1027 
 1028         SOCKBUF_LOCK(sb);
 1029         retval = sbappendcontrol_locked(sb, m0, control);
 1030         SOCKBUF_UNLOCK(sb);
 1031         return (retval);
 1032 }
 1033 
 1034 /*
 1035  * Compress mbuf chain m into the socket
 1036  * buffer sb following mbuf n.  If n
 1037  * is null, the buffer is presumed empty.
 1038  */
 1039 void
 1040 sbcompress(sb, m, n)
 1041         register struct sockbuf *sb;
 1042         register struct mbuf *m, *n;
 1043 {
 1044         register int eor = 0;
 1045         register struct mbuf *o;
 1046 
 1047         SOCKBUF_LOCK_ASSERT(sb);
 1048 
 1049         while (m) {
 1050                 eor |= m->m_flags & M_EOR;
 1051                 if (m->m_len == 0 &&
 1052                     (eor == 0 ||
 1053                      (((o = m->m_next) || (o = n)) &&
 1054                       o->m_type == m->m_type))) {
 1055                         if (sb->sb_lastrecord == m)
 1056                                 sb->sb_lastrecord = m->m_next;
 1057                         m = m_free(m);
 1058                         continue;
 1059                 }
 1060                 if (n && (n->m_flags & M_EOR) == 0 &&
 1061                     M_WRITABLE(n) &&
 1062                     m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
 1063                     m->m_len <= M_TRAILINGSPACE(n) &&
 1064                     n->m_type == m->m_type) {
 1065                         bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
 1066                             (unsigned)m->m_len);
 1067                         n->m_len += m->m_len;
 1068                         sb->sb_cc += m->m_len;
 1069                         if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
 1070                             m->m_type != MT_OOBDATA)
 1071                                 /* XXX: Probably don't need.*/
 1072                                 sb->sb_ctl += m->m_len;
 1073                         m = m_free(m);
 1074                         continue;
 1075                 }
 1076                 if (n)
 1077                         n->m_next = m;
 1078                 else
 1079                         sb->sb_mb = m;
 1080                 sb->sb_mbtail = m;
 1081                 sballoc(sb, m);
 1082                 n = m;
 1083                 m->m_flags &= ~M_EOR;
 1084                 m = m->m_next;
 1085                 n->m_next = 0;
 1086         }
 1087         if (eor) {
 1088                 if (n)
 1089                         n->m_flags |= eor;
 1090                 else
 1091                         printf("semi-panic: sbcompress\n");
 1092         }
 1093         SBLASTMBUFCHK(sb);
 1094 }
 1095 
 1096 /*
 1097  * Free all mbufs in a sockbuf.
 1098  * Check that all resources are reclaimed.
 1099  */
 1100 void
 1101 sbflush_locked(sb)
 1102         register struct sockbuf *sb;
 1103 {
 1104 
 1105         SOCKBUF_LOCK_ASSERT(sb);
 1106 
 1107         if (sb->sb_flags & SB_LOCK)
 1108                 panic("sbflush_locked: locked");
 1109         while (sb->sb_mbcnt) {
 1110                 /*
 1111                  * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
 1112                  * we would loop forever. Panic instead.
 1113                  */
 1114                 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
 1115                         break;
 1116                 sbdrop_locked(sb, (int)sb->sb_cc);
 1117         }
 1118         if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
 1119                 panic("sbflush_locked: cc %u || mb %p || mbcnt %u", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
 1120 }
 1121 
 1122 void
 1123 sbflush(sb)
 1124         register struct sockbuf *sb;
 1125 {
 1126 
 1127         SOCKBUF_LOCK(sb);
 1128         sbflush_locked(sb);
 1129         SOCKBUF_UNLOCK(sb);
 1130 }
 1131 
 1132 /*
 1133  * Drop data from (the front of) a sockbuf.
 1134  */
 1135 void
 1136 sbdrop_locked(sb, len)
 1137         register struct sockbuf *sb;
 1138         register int len;
 1139 {
 1140         register struct mbuf *m;
 1141         struct mbuf *next;
 1142 
 1143         SOCKBUF_LOCK_ASSERT(sb);
 1144 
 1145         next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
 1146         while (len > 0) {
 1147                 if (m == 0) {
 1148                         if (next == 0)
 1149                                 panic("sbdrop");
 1150                         m = next;
 1151                         next = m->m_nextpkt;
 1152                         continue;
 1153                 }
 1154                 if (m->m_len > len) {
 1155                         m->m_len -= len;
 1156                         m->m_data += len;
 1157                         sb->sb_cc -= len;
 1158                         if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
 1159                             m->m_type != MT_OOBDATA)
 1160                                 sb->sb_ctl -= len;
 1161                         break;
 1162                 }
 1163                 len -= m->m_len;
 1164                 sbfree(sb, m);
 1165                 m = m_free(m);
 1166         }
 1167         while (m && m->m_len == 0) {
 1168                 sbfree(sb, m);
 1169                 m = m_free(m);
 1170         }
 1171         if (m) {
 1172                 sb->sb_mb = m;
 1173                 m->m_nextpkt = next;
 1174         } else
 1175                 sb->sb_mb = next;
 1176         /*
 1177          * First part is an inline SB_EMPTY_FIXUP().  Second part
 1178          * makes sure sb_lastrecord is up-to-date if we dropped
 1179          * part of the last record.
 1180          */
 1181         m = sb->sb_mb;
 1182         if (m == NULL) {
 1183                 sb->sb_mbtail = NULL;
 1184                 sb->sb_lastrecord = NULL;
 1185         } else if (m->m_nextpkt == NULL) {
 1186                 sb->sb_lastrecord = m;
 1187         }
 1188 }
 1189 
 1190 /*
 1191  * Drop data from (the front of) a sockbuf.
 1192  */
 1193 void
 1194 sbdrop(sb, len)
 1195         register struct sockbuf *sb;
 1196         register int len;
 1197 {
 1198 
 1199         SOCKBUF_LOCK(sb);
 1200         sbdrop_locked(sb, len);
 1201         SOCKBUF_UNLOCK(sb);
 1202 }
 1203 
 1204 /*
 1205  * Drop a record off the front of a sockbuf
 1206  * and move the next record to the front.
 1207  */
 1208 void
 1209 sbdroprecord_locked(sb)
 1210         register struct sockbuf *sb;
 1211 {
 1212         register struct mbuf *m;
 1213 
 1214         SOCKBUF_LOCK_ASSERT(sb);
 1215 
 1216         m = sb->sb_mb;
 1217         if (m) {
 1218                 sb->sb_mb = m->m_nextpkt;
 1219                 do {
 1220                         sbfree(sb, m);
 1221                         m = m_free(m);
 1222                 } while (m);
 1223         }
 1224         SB_EMPTY_FIXUP(sb);
 1225 }
 1226 
 1227 /*
 1228  * Drop a record off the front of a sockbuf
 1229  * and move the next record to the front.
 1230  */
 1231 void
 1232 sbdroprecord(sb)
 1233         register struct sockbuf *sb;
 1234 {
 1235 
 1236         SOCKBUF_LOCK(sb);
 1237         sbdroprecord_locked(sb);
 1238         SOCKBUF_UNLOCK(sb);
 1239 }
 1240 
 1241 /*
 1242  * Create a "control" mbuf containing the specified data
 1243  * with the specified type for presentation on a socket buffer.
 1244  */
 1245 struct mbuf *
 1246 sbcreatecontrol(p, size, type, level)
 1247         caddr_t p;
 1248         register int size;
 1249         int type, level;
 1250 {
 1251         register struct cmsghdr *cp;
 1252         struct mbuf *m;
 1253 
 1254         if (CMSG_SPACE((u_int)size) > MCLBYTES)
 1255                 return ((struct mbuf *) NULL);
 1256         if (CMSG_SPACE((u_int)size) > MLEN)
 1257                 m = m_getcl(M_DONTWAIT, MT_CONTROL, 0);
 1258         else
 1259                 m = m_get(M_DONTWAIT, MT_CONTROL);
 1260         if (m == NULL)
 1261                 return ((struct mbuf *) NULL);
 1262         cp = mtod(m, struct cmsghdr *);
 1263         m->m_len = 0;
 1264         KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
 1265             ("sbcreatecontrol: short mbuf"));
 1266         if (p != NULL)
 1267                 (void)memcpy(CMSG_DATA(cp), p, size);
 1268         m->m_len = CMSG_SPACE(size);
 1269         cp->cmsg_len = CMSG_LEN(size);
 1270         cp->cmsg_level = level;
 1271         cp->cmsg_type = type;
 1272         return (m);
 1273 }
 1274 
 1275 /*
 1276  * Some routines that return EOPNOTSUPP for entry points that are not
 1277  * supported by a protocol.  Fill in as needed.
 1278  */
 1279 int
 1280 pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
 1281 {
 1282         return EOPNOTSUPP;
 1283 }
 1284 
 1285 int
 1286 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
 1287 {
 1288         return EOPNOTSUPP;
 1289 }
 1290 
 1291 int
 1292 pru_connect2_notsupp(struct socket *so1, struct socket *so2)
 1293 {
 1294         return EOPNOTSUPP;
 1295 }
 1296 
 1297 int
 1298 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
 1299                     struct ifnet *ifp, struct thread *td)
 1300 {
 1301         return EOPNOTSUPP;
 1302 }
 1303 
 1304 int
 1305 pru_listen_notsupp(struct socket *so, struct thread *td)
 1306 {
 1307         return EOPNOTSUPP;
 1308 }
 1309 
 1310 int
 1311 pru_rcvd_notsupp(struct socket *so, int flags)
 1312 {
 1313         return EOPNOTSUPP;
 1314 }
 1315 
 1316 int
 1317 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
 1318 {
 1319         return EOPNOTSUPP;
 1320 }
 1321 
 1322 /*
 1323  * This isn't really a ``null'' operation, but it's the default one
 1324  * and doesn't do anything destructive.
 1325  */
 1326 int
 1327 pru_sense_null(struct socket *so, struct stat *sb)
 1328 {
 1329         sb->st_blksize = so->so_snd.sb_hiwat;
 1330         return 0;
 1331 }
 1332 
 1333 /*
 1334  * For protocol types that don't keep cached copies of labels in their
 1335  * pcbs, provide a null sosetlabel that does a NOOP.
 1336  */
 1337 void
 1338 pru_sosetlabel_null(struct socket *so)
 1339 {
 1340 
 1341 }
 1342 
 1343 /*
 1344  * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
 1345  */
 1346 struct sockaddr *
 1347 sodupsockaddr(const struct sockaddr *sa, int mflags)
 1348 {
 1349         struct sockaddr *sa2;
 1350 
 1351         sa2 = malloc(sa->sa_len, M_SONAME, mflags);
 1352         if (sa2)
 1353                 bcopy(sa, sa2, sa->sa_len);
 1354         return sa2;
 1355 }
 1356 
 1357 /*
 1358  * Create an external-format (``xsocket'') structure using the information
 1359  * in the kernel-format socket structure pointed to by so.  This is done
 1360  * to reduce the spew of irrelevant information over this interface,
 1361  * to isolate user code from changes in the kernel structure, and
 1362  * potentially to provide information-hiding if we decide that
 1363  * some of this information should be hidden from users.
 1364  */
 1365 void
 1366 sotoxsocket(struct socket *so, struct xsocket *xso)
 1367 {
 1368         xso->xso_len = sizeof *xso;
 1369         xso->xso_so = so;
 1370         xso->so_type = so->so_type;
 1371         xso->so_options = so->so_options;
 1372         xso->so_linger = so->so_linger;
 1373         xso->so_state = so->so_state;
 1374         xso->so_pcb = so->so_pcb;
 1375         xso->xso_protocol = so->so_proto->pr_protocol;
 1376         xso->xso_family = so->so_proto->pr_domain->dom_family;
 1377         xso->so_qlen = so->so_qlen;
 1378         xso->so_incqlen = so->so_incqlen;
 1379         xso->so_qlimit = so->so_qlimit;
 1380         xso->so_timeo = so->so_timeo;
 1381         xso->so_error = so->so_error;
 1382         xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
 1383         xso->so_oobmark = so->so_oobmark;
 1384         sbtoxsockbuf(&so->so_snd, &xso->so_snd);
 1385         sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
 1386         xso->so_uid = so->so_cred->cr_uid;
 1387 }
 1388 
 1389 /*
 1390  * This does the same for sockbufs.  Note that the xsockbuf structure,
 1391  * since it is always embedded in a socket, does not include a self
 1392  * pointer nor a length.  We make this entry point public in case
 1393  * some other mechanism needs it.
 1394  */
 1395 void
 1396 sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
 1397 {
 1398         xsb->sb_cc = sb->sb_cc;
 1399         xsb->sb_hiwat = sb->sb_hiwat;
 1400         xsb->sb_mbcnt = sb->sb_mbcnt;
 1401         xsb->sb_mbmax = sb->sb_mbmax;
 1402         xsb->sb_lowat = sb->sb_lowat;
 1403         xsb->sb_flags = sb->sb_flags;
 1404         xsb->sb_timeo = sb->sb_timeo;
 1405 }
 1406 
 1407 /*
 1408  * Here is the definition of some of the basic objects in the kern.ipc
 1409  * branch of the MIB.
 1410  */
 1411 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
 1412 
 1413 /* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
 1414 static int dummy;
 1415 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
 1416 SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_ULONG|CTLFLAG_RW, 
 1417     &sb_max, 0, sysctl_handle_sb_max, "LU", "Maximum socket buffer size");
 1418 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RDTUN, 
 1419     &maxsockets, 0, "Maximum number of sockets avaliable");
 1420 SYSCTL_ULONG(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
 1421     &sb_efficiency, 0, "");
 1422 
 1423 /*
 1424  * Initialise maxsockets 
 1425  */
 1426 static void init_maxsockets(void *ignored)
 1427 {
 1428         TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
 1429         maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
 1430 }
 1431 SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);

Cache object: c1cb94c2dcc90a7b741f5d2d8aa12bb5


[ 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.