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_syscalls.c

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1982, 1986, 1989, 1990, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * sendfile(2) and related extensions:
    6  * Copyright (c) 1998, David Greenman. All rights reserved.
    7  *
    8  * Redistribution and use in source and binary forms, with or without
    9  * modification, are permitted provided that the following conditions
   10  * are met:
   11  * 1. Redistributions of source code must retain the above copyright
   12  *    notice, this list of conditions and the following disclaimer.
   13  * 2. Redistributions in binary form must reproduce the above copyright
   14  *    notice, this list of conditions and the following disclaimer in the
   15  *    documentation and/or other materials provided with the distribution.
   16  * 4. Neither the name of the University nor the names of its contributors
   17  *    may be used to endorse or promote products derived from this software
   18  *    without specific prior written permission.
   19  *
   20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   30  * SUCH DAMAGE.
   31  *
   32  *      @(#)uipc_syscalls.c     8.4 (Berkeley) 2/21/94
   33  */
   34 
   35 #include <sys/cdefs.h>
   36 __FBSDID("$FreeBSD: releng/10.1/sys/kern/uipc_syscalls.c 300085 2016-05-17 22:28:11Z glebius $");
   37 
   38 #include "opt_capsicum.h"
   39 #include "opt_inet.h"
   40 #include "opt_inet6.h"
   41 #include "opt_sctp.h"
   42 #include "opt_compat.h"
   43 #include "opt_ktrace.h"
   44 
   45 #include <sys/param.h>
   46 #include <sys/systm.h>
   47 #include <sys/capability.h>
   48 #include <sys/condvar.h>
   49 #include <sys/kernel.h>
   50 #include <sys/lock.h>
   51 #include <sys/mutex.h>
   52 #include <sys/sysproto.h>
   53 #include <sys/malloc.h>
   54 #include <sys/filedesc.h>
   55 #include <sys/event.h>
   56 #include <sys/proc.h>
   57 #include <sys/fcntl.h>
   58 #include <sys/file.h>
   59 #include <sys/filio.h>
   60 #include <sys/jail.h>
   61 #include <sys/mman.h>
   62 #include <sys/mount.h>
   63 #include <sys/mbuf.h>
   64 #include <sys/protosw.h>
   65 #include <sys/rwlock.h>
   66 #include <sys/sf_buf.h>
   67 #include <sys/sysent.h>
   68 #include <sys/socket.h>
   69 #include <sys/socketvar.h>
   70 #include <sys/signalvar.h>
   71 #include <sys/syscallsubr.h>
   72 #include <sys/sysctl.h>
   73 #include <sys/uio.h>
   74 #include <sys/vnode.h>
   75 #ifdef KTRACE
   76 #include <sys/ktrace.h>
   77 #endif
   78 #ifdef COMPAT_FREEBSD32
   79 #include <compat/freebsd32/freebsd32_util.h>
   80 #endif
   81 
   82 #include <net/vnet.h>
   83 
   84 #include <security/audit/audit.h>
   85 #include <security/mac/mac_framework.h>
   86 
   87 #include <vm/vm.h>
   88 #include <vm/vm_param.h>
   89 #include <vm/vm_object.h>
   90 #include <vm/vm_page.h>
   91 #include <vm/vm_pager.h>
   92 #include <vm/vm_kern.h>
   93 #include <vm/vm_extern.h>
   94 
   95 #if defined(INET) || defined(INET6)
   96 #ifdef SCTP
   97 #include <netinet/sctp.h>
   98 #include <netinet/sctp_peeloff.h>
   99 #endif /* SCTP */
  100 #endif /* INET || INET6 */
  101 
  102 /*
  103  * Flags for accept1() and kern_accept4(), in addition to SOCK_CLOEXEC
  104  * and SOCK_NONBLOCK.
  105  */
  106 #define ACCEPT4_INHERIT 0x1
  107 #define ACCEPT4_COMPAT  0x2
  108 
  109 static int sendit(struct thread *td, int s, struct msghdr *mp, int flags);
  110 static int recvit(struct thread *td, int s, struct msghdr *mp, void *namelenp);
  111 
  112 static int accept1(struct thread *td, int s, struct sockaddr *uname,
  113                    socklen_t *anamelen, int flags);
  114 static int do_sendfile(struct thread *td, struct sendfile_args *uap,
  115                    int compat);
  116 static int getsockname1(struct thread *td, struct getsockname_args *uap,
  117                         int compat);
  118 static int getpeername1(struct thread *td, struct getpeername_args *uap,
  119                         int compat);
  120 
  121 counter_u64_t sfstat[sizeof(struct sfstat) / sizeof(uint64_t)];
  122 
  123 /*
  124  * sendfile(2)-related variables and associated sysctls
  125  */
  126 static SYSCTL_NODE(_kern_ipc, OID_AUTO, sendfile, CTLFLAG_RW, 0,
  127     "sendfile(2) tunables");
  128 static int sfreadahead = 1;
  129 SYSCTL_INT(_kern_ipc_sendfile, OID_AUTO, readahead, CTLFLAG_RW,
  130     &sfreadahead, 0, "Number of sendfile(2) read-ahead MAXBSIZE blocks");
  131 
  132 
  133 static void
  134 sfstat_init(const void *unused)
  135 {
  136 
  137         COUNTER_ARRAY_ALLOC(sfstat, sizeof(struct sfstat) / sizeof(uint64_t),
  138             M_WAITOK);
  139 }
  140 SYSINIT(sfstat, SI_SUB_MBUF, SI_ORDER_FIRST, sfstat_init, NULL);
  141 
  142 static int
  143 sfstat_sysctl(SYSCTL_HANDLER_ARGS)
  144 {
  145         struct sfstat s;
  146 
  147         COUNTER_ARRAY_COPY(sfstat, &s, sizeof(s) / sizeof(uint64_t));
  148         if (req->newptr)
  149                 COUNTER_ARRAY_ZERO(sfstat, sizeof(s) / sizeof(uint64_t));
  150         return (SYSCTL_OUT(req, &s, sizeof(s)));
  151 }
  152 SYSCTL_PROC(_kern_ipc, OID_AUTO, sfstat, CTLTYPE_OPAQUE | CTLFLAG_RW,
  153     NULL, 0, sfstat_sysctl, "I", "sendfile statistics");
  154 
  155 /*
  156  * Convert a user file descriptor to a kernel file entry and check if required
  157  * capability rights are present.
  158  * A reference on the file entry is held upon returning.
  159  */
  160 static int
  161 getsock_cap(struct filedesc *fdp, int fd, cap_rights_t *rightsp,
  162     struct file **fpp, u_int *fflagp)
  163 {
  164         struct file *fp;
  165         int error;
  166 
  167         error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
  168         if (error != 0)
  169                 return (error);
  170         if (fp->f_type != DTYPE_SOCKET) {
  171                 fdrop(fp, curthread);
  172                 return (ENOTSOCK);
  173         }
  174         if (fflagp != NULL)
  175                 *fflagp = fp->f_flag;
  176         *fpp = fp;
  177         return (0);
  178 }
  179 
  180 /*
  181  * System call interface to the socket abstraction.
  182  */
  183 #if defined(COMPAT_43)
  184 #define COMPAT_OLDSOCK
  185 #endif
  186 
  187 int
  188 sys_socket(td, uap)
  189         struct thread *td;
  190         struct socket_args /* {
  191                 int     domain;
  192                 int     type;
  193                 int     protocol;
  194         } */ *uap;
  195 {
  196         struct socket *so;
  197         struct file *fp;
  198         int fd, error, type, oflag, fflag;
  199 
  200         AUDIT_ARG_SOCKET(uap->domain, uap->type, uap->protocol);
  201 
  202         type = uap->type;
  203         oflag = 0;
  204         fflag = 0;
  205         if ((type & SOCK_CLOEXEC) != 0) {
  206                 type &= ~SOCK_CLOEXEC;
  207                 oflag |= O_CLOEXEC;
  208         }
  209         if ((type & SOCK_NONBLOCK) != 0) {
  210                 type &= ~SOCK_NONBLOCK;
  211                 fflag |= FNONBLOCK;
  212         }
  213 
  214 #ifdef MAC
  215         error = mac_socket_check_create(td->td_ucred, uap->domain, type,
  216             uap->protocol);
  217         if (error != 0)
  218                 return (error);
  219 #endif
  220         error = falloc(td, &fp, &fd, oflag);
  221         if (error != 0)
  222                 return (error);
  223         /* An extra reference on `fp' has been held for us by falloc(). */
  224         error = socreate(uap->domain, &so, type, uap->protocol,
  225             td->td_ucred, td);
  226         if (error != 0) {
  227                 fdclose(td->td_proc->p_fd, fp, fd, td);
  228         } else {
  229                 finit(fp, FREAD | FWRITE | fflag, DTYPE_SOCKET, so, &socketops);
  230                 if ((fflag & FNONBLOCK) != 0)
  231                         (void) fo_ioctl(fp, FIONBIO, &fflag, td->td_ucred, td);
  232                 td->td_retval[0] = fd;
  233         }
  234         fdrop(fp, td);
  235         return (error);
  236 }
  237 
  238 /* ARGSUSED */
  239 int
  240 sys_bind(td, uap)
  241         struct thread *td;
  242         struct bind_args /* {
  243                 int     s;
  244                 caddr_t name;
  245                 int     namelen;
  246         } */ *uap;
  247 {
  248         struct sockaddr *sa;
  249         int error;
  250 
  251         error = getsockaddr(&sa, uap->name, uap->namelen);
  252         if (error == 0) {
  253                 error = kern_bind(td, uap->s, sa);
  254                 free(sa, M_SONAME);
  255         }
  256         return (error);
  257 }
  258 
  259 static int
  260 kern_bindat(struct thread *td, int dirfd, int fd, struct sockaddr *sa)
  261 {
  262         struct socket *so;
  263         struct file *fp;
  264         cap_rights_t rights;
  265         int error;
  266 
  267         AUDIT_ARG_FD(fd);
  268         AUDIT_ARG_SOCKADDR(td, dirfd, sa);
  269         error = getsock_cap(td->td_proc->p_fd, fd,
  270             cap_rights_init(&rights, CAP_BIND), &fp, NULL);
  271         if (error != 0)
  272                 return (error);
  273         so = fp->f_data;
  274 #ifdef KTRACE
  275         if (KTRPOINT(td, KTR_STRUCT))
  276                 ktrsockaddr(sa);
  277 #endif
  278 #ifdef MAC
  279         error = mac_socket_check_bind(td->td_ucred, so, sa);
  280         if (error == 0) {
  281 #endif
  282                 if (dirfd == AT_FDCWD)
  283                         error = sobind(so, sa, td);
  284                 else
  285                         error = sobindat(dirfd, so, sa, td);
  286 #ifdef MAC
  287         }
  288 #endif
  289         fdrop(fp, td);
  290         return (error);
  291 }
  292 
  293 int
  294 kern_bind(struct thread *td, int fd, struct sockaddr *sa)
  295 {
  296 
  297         return (kern_bindat(td, AT_FDCWD, fd, sa));
  298 }
  299 
  300 /* ARGSUSED */
  301 int
  302 sys_bindat(td, uap)
  303         struct thread *td;
  304         struct bindat_args /* {
  305                 int     fd;
  306                 int     s;
  307                 caddr_t name;
  308                 int     namelen;
  309         } */ *uap;
  310 {
  311         struct sockaddr *sa;
  312         int error;
  313 
  314         error = getsockaddr(&sa, uap->name, uap->namelen);
  315         if (error == 0) {
  316                 error = kern_bindat(td, uap->fd, uap->s, sa);
  317                 free(sa, M_SONAME);
  318         }
  319         return (error);
  320 }
  321 
  322 /* ARGSUSED */
  323 int
  324 sys_listen(td, uap)
  325         struct thread *td;
  326         struct listen_args /* {
  327                 int     s;
  328                 int     backlog;
  329         } */ *uap;
  330 {
  331         struct socket *so;
  332         struct file *fp;
  333         cap_rights_t rights;
  334         int error;
  335 
  336         AUDIT_ARG_FD(uap->s);
  337         error = getsock_cap(td->td_proc->p_fd, uap->s,
  338             cap_rights_init(&rights, CAP_LISTEN), &fp, NULL);
  339         if (error == 0) {
  340                 so = fp->f_data;
  341 #ifdef MAC
  342                 error = mac_socket_check_listen(td->td_ucred, so);
  343                 if (error == 0)
  344 #endif
  345                         error = solisten(so, uap->backlog, td);
  346                 fdrop(fp, td);
  347         }
  348         return(error);
  349 }
  350 
  351 /*
  352  * accept1()
  353  */
  354 static int
  355 accept1(td, s, uname, anamelen, flags)
  356         struct thread *td;
  357         int s;
  358         struct sockaddr *uname;
  359         socklen_t *anamelen;
  360         int flags;
  361 {
  362         struct sockaddr *name;
  363         socklen_t namelen;
  364         struct file *fp;
  365         int error;
  366 
  367         if (uname == NULL)
  368                 return (kern_accept4(td, s, NULL, NULL, flags, NULL));
  369 
  370         error = copyin(anamelen, &namelen, sizeof (namelen));
  371         if (error != 0)
  372                 return (error);
  373 
  374         error = kern_accept4(td, s, &name, &namelen, flags, &fp);
  375 
  376         /*
  377          * return a namelen of zero for older code which might
  378          * ignore the return value from accept.
  379          */
  380         if (error != 0) {
  381                 (void) copyout(&namelen, anamelen, sizeof(*anamelen));
  382                 return (error);
  383         }
  384 
  385         if (error == 0 && uname != NULL) {
  386 #ifdef COMPAT_OLDSOCK
  387                 if (flags & ACCEPT4_COMPAT)
  388                         ((struct osockaddr *)name)->sa_family =
  389                             name->sa_family;
  390 #endif
  391                 error = copyout(name, uname, namelen);
  392         }
  393         if (error == 0)
  394                 error = copyout(&namelen, anamelen,
  395                     sizeof(namelen));
  396         if (error != 0)
  397                 fdclose(td->td_proc->p_fd, fp, td->td_retval[0], td);
  398         fdrop(fp, td);
  399         free(name, M_SONAME);
  400         return (error);
  401 }
  402 
  403 int
  404 kern_accept(struct thread *td, int s, struct sockaddr **name,
  405     socklen_t *namelen, struct file **fp)
  406 {
  407         return (kern_accept4(td, s, name, namelen, ACCEPT4_INHERIT, fp));
  408 }
  409 
  410 int
  411 kern_accept4(struct thread *td, int s, struct sockaddr **name,
  412     socklen_t *namelen, int flags, struct file **fp)
  413 {
  414         struct filedesc *fdp;
  415         struct file *headfp, *nfp = NULL;
  416         struct sockaddr *sa = NULL;
  417         struct socket *head, *so;
  418         cap_rights_t rights;
  419         u_int fflag;
  420         pid_t pgid;
  421         int error, fd, tmp;
  422 
  423         if (name != NULL)
  424                 *name = NULL;
  425 
  426         AUDIT_ARG_FD(s);
  427         fdp = td->td_proc->p_fd;
  428         error = getsock_cap(fdp, s, cap_rights_init(&rights, CAP_ACCEPT),
  429             &headfp, &fflag);
  430         if (error != 0)
  431                 return (error);
  432         head = headfp->f_data;
  433         if ((head->so_options & SO_ACCEPTCONN) == 0) {
  434                 error = EINVAL;
  435                 goto done;
  436         }
  437 #ifdef MAC
  438         error = mac_socket_check_accept(td->td_ucred, head);
  439         if (error != 0)
  440                 goto done;
  441 #endif
  442         error = falloc(td, &nfp, &fd, (flags & SOCK_CLOEXEC) ? O_CLOEXEC : 0);
  443         if (error != 0)
  444                 goto done;
  445         ACCEPT_LOCK();
  446         if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->so_comp)) {
  447                 ACCEPT_UNLOCK();
  448                 error = EWOULDBLOCK;
  449                 goto noconnection;
  450         }
  451         while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) {
  452                 if (head->so_rcv.sb_state & SBS_CANTRCVMORE) {
  453                         head->so_error = ECONNABORTED;
  454                         break;
  455                 }
  456                 error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH,
  457                     "accept", 0);
  458                 if (error != 0) {
  459                         ACCEPT_UNLOCK();
  460                         goto noconnection;
  461                 }
  462         }
  463         if (head->so_error) {
  464                 error = head->so_error;
  465                 head->so_error = 0;
  466                 ACCEPT_UNLOCK();
  467                 goto noconnection;
  468         }
  469         so = TAILQ_FIRST(&head->so_comp);
  470         KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP"));
  471         KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP"));
  472 
  473         /*
  474          * Before changing the flags on the socket, we have to bump the
  475          * reference count.  Otherwise, if the protocol calls sofree(),
  476          * the socket will be released due to a zero refcount.
  477          */
  478         SOCK_LOCK(so);                  /* soref() and so_state update */
  479         soref(so);                      /* file descriptor reference */
  480 
  481         TAILQ_REMOVE(&head->so_comp, so, so_list);
  482         head->so_qlen--;
  483         if (flags & ACCEPT4_INHERIT)
  484                 so->so_state |= (head->so_state & SS_NBIO);
  485         else
  486                 so->so_state |= (flags & SOCK_NONBLOCK) ? SS_NBIO : 0;
  487         so->so_qstate &= ~SQ_COMP;
  488         so->so_head = NULL;
  489 
  490         SOCK_UNLOCK(so);
  491         ACCEPT_UNLOCK();
  492 
  493         /* An extra reference on `nfp' has been held for us by falloc(). */
  494         td->td_retval[0] = fd;
  495 
  496         /* connection has been removed from the listen queue */
  497         KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
  498 
  499         if (flags & ACCEPT4_INHERIT) {
  500                 pgid = fgetown(&head->so_sigio);
  501                 if (pgid != 0)
  502                         fsetown(pgid, &so->so_sigio);
  503         } else {
  504                 fflag &= ~(FNONBLOCK | FASYNC);
  505                 if (flags & SOCK_NONBLOCK)
  506                         fflag |= FNONBLOCK;
  507         }
  508 
  509         finit(nfp, fflag, DTYPE_SOCKET, so, &socketops);
  510         /* Sync socket nonblocking/async state with file flags */
  511         tmp = fflag & FNONBLOCK;
  512         (void) fo_ioctl(nfp, FIONBIO, &tmp, td->td_ucred, td);
  513         tmp = fflag & FASYNC;
  514         (void) fo_ioctl(nfp, FIOASYNC, &tmp, td->td_ucred, td);
  515         sa = 0;
  516         error = soaccept(so, &sa);
  517         if (error != 0) {
  518                 /*
  519                  * return a namelen of zero for older code which might
  520                  * ignore the return value from accept.
  521                  */
  522                 if (name)
  523                         *namelen = 0;
  524                 goto noconnection;
  525         }
  526         if (sa == NULL) {
  527                 if (name)
  528                         *namelen = 0;
  529                 goto done;
  530         }
  531         AUDIT_ARG_SOCKADDR(td, AT_FDCWD, sa);
  532         if (name) {
  533                 /* check sa_len before it is destroyed */
  534                 if (*namelen > sa->sa_len)
  535                         *namelen = sa->sa_len;
  536 #ifdef KTRACE
  537                 if (KTRPOINT(td, KTR_STRUCT))
  538                         ktrsockaddr(sa);
  539 #endif
  540                 *name = sa;
  541                 sa = NULL;
  542         }
  543 noconnection:
  544         free(sa, M_SONAME);
  545 
  546         /*
  547          * close the new descriptor, assuming someone hasn't ripped it
  548          * out from under us.
  549          */
  550         if (error != 0)
  551                 fdclose(fdp, nfp, fd, td);
  552 
  553         /*
  554          * Release explicitly held references before returning.  We return
  555          * a reference on nfp to the caller on success if they request it.
  556          */
  557 done:
  558         if (fp != NULL) {
  559                 if (error == 0) {
  560                         *fp = nfp;
  561                         nfp = NULL;
  562                 } else
  563                         *fp = NULL;
  564         }
  565         if (nfp != NULL)
  566                 fdrop(nfp, td);
  567         fdrop(headfp, td);
  568         return (error);
  569 }
  570 
  571 int
  572 sys_accept(td, uap)
  573         struct thread *td;
  574         struct accept_args *uap;
  575 {
  576 
  577         return (accept1(td, uap->s, uap->name, uap->anamelen, ACCEPT4_INHERIT));
  578 }
  579 
  580 int
  581 sys_accept4(td, uap)
  582         struct thread *td;
  583         struct accept4_args *uap;
  584 {
  585 
  586         if (uap->flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
  587                 return (EINVAL);
  588 
  589         return (accept1(td, uap->s, uap->name, uap->anamelen, uap->flags));
  590 }
  591 
  592 #ifdef COMPAT_OLDSOCK
  593 int
  594 oaccept(td, uap)
  595         struct thread *td;
  596         struct accept_args *uap;
  597 {
  598 
  599         return (accept1(td, uap->s, uap->name, uap->anamelen,
  600             ACCEPT4_INHERIT | ACCEPT4_COMPAT));
  601 }
  602 #endif /* COMPAT_OLDSOCK */
  603 
  604 /* ARGSUSED */
  605 int
  606 sys_connect(td, uap)
  607         struct thread *td;
  608         struct connect_args /* {
  609                 int     s;
  610                 caddr_t name;
  611                 int     namelen;
  612         } */ *uap;
  613 {
  614         struct sockaddr *sa;
  615         int error;
  616 
  617         error = getsockaddr(&sa, uap->name, uap->namelen);
  618         if (error == 0) {
  619                 error = kern_connect(td, uap->s, sa);
  620                 free(sa, M_SONAME);
  621         }
  622         return (error);
  623 }
  624 
  625 static int
  626 kern_connectat(struct thread *td, int dirfd, int fd, struct sockaddr *sa)
  627 {
  628         struct socket *so;
  629         struct file *fp;
  630         cap_rights_t rights;
  631         int error, interrupted = 0;
  632 
  633         AUDIT_ARG_FD(fd);
  634         AUDIT_ARG_SOCKADDR(td, dirfd, sa);
  635         error = getsock_cap(td->td_proc->p_fd, fd,
  636             cap_rights_init(&rights, CAP_CONNECT), &fp, NULL);
  637         if (error != 0)
  638                 return (error);
  639         so = fp->f_data;
  640         if (so->so_state & SS_ISCONNECTING) {
  641                 error = EALREADY;
  642                 goto done1;
  643         }
  644 #ifdef KTRACE
  645         if (KTRPOINT(td, KTR_STRUCT))
  646                 ktrsockaddr(sa);
  647 #endif
  648 #ifdef MAC
  649         error = mac_socket_check_connect(td->td_ucred, so, sa);
  650         if (error != 0)
  651                 goto bad;
  652 #endif
  653         if (dirfd == AT_FDCWD)
  654                 error = soconnect(so, sa, td);
  655         else
  656                 error = soconnectat(dirfd, so, sa, td);
  657         if (error != 0)
  658                 goto bad;
  659         if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
  660                 error = EINPROGRESS;
  661                 goto done1;
  662         }
  663         SOCK_LOCK(so);
  664         while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
  665                 error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH,
  666                     "connec", 0);
  667                 if (error != 0) {
  668                         if (error == EINTR || error == ERESTART)
  669                                 interrupted = 1;
  670                         break;
  671                 }
  672         }
  673         if (error == 0) {
  674                 error = so->so_error;
  675                 so->so_error = 0;
  676         }
  677         SOCK_UNLOCK(so);
  678 bad:
  679         if (!interrupted)
  680                 so->so_state &= ~SS_ISCONNECTING;
  681         if (error == ERESTART)
  682                 error = EINTR;
  683 done1:
  684         fdrop(fp, td);
  685         return (error);
  686 }
  687 
  688 int
  689 kern_connect(struct thread *td, int fd, struct sockaddr *sa)
  690 {
  691 
  692         return (kern_connectat(td, AT_FDCWD, fd, sa));
  693 }
  694 
  695 /* ARGSUSED */
  696 int
  697 sys_connectat(td, uap)
  698         struct thread *td;
  699         struct connectat_args /* {
  700                 int     fd;
  701                 int     s;
  702                 caddr_t name;
  703                 int     namelen;
  704         } */ *uap;
  705 {
  706         struct sockaddr *sa;
  707         int error;
  708 
  709         error = getsockaddr(&sa, uap->name, uap->namelen);
  710         if (error == 0) {
  711                 error = kern_connectat(td, uap->fd, uap->s, sa);
  712                 free(sa, M_SONAME);
  713         }
  714         return (error);
  715 }
  716 
  717 int
  718 kern_socketpair(struct thread *td, int domain, int type, int protocol,
  719     int *rsv)
  720 {
  721         struct filedesc *fdp = td->td_proc->p_fd;
  722         struct file *fp1, *fp2;
  723         struct socket *so1, *so2;
  724         int fd, error, oflag, fflag;
  725 
  726         AUDIT_ARG_SOCKET(domain, type, protocol);
  727 
  728         oflag = 0;
  729         fflag = 0;
  730         if ((type & SOCK_CLOEXEC) != 0) {
  731                 type &= ~SOCK_CLOEXEC;
  732                 oflag |= O_CLOEXEC;
  733         }
  734         if ((type & SOCK_NONBLOCK) != 0) {
  735                 type &= ~SOCK_NONBLOCK;
  736                 fflag |= FNONBLOCK;
  737         }
  738 #ifdef MAC
  739         /* We might want to have a separate check for socket pairs. */
  740         error = mac_socket_check_create(td->td_ucred, domain, type,
  741             protocol);
  742         if (error != 0)
  743                 return (error);
  744 #endif
  745         error = socreate(domain, &so1, type, protocol, td->td_ucred, td);
  746         if (error != 0)
  747                 return (error);
  748         error = socreate(domain, &so2, type, protocol, td->td_ucred, td);
  749         if (error != 0)
  750                 goto free1;
  751         /* On success extra reference to `fp1' and 'fp2' is set by falloc. */
  752         error = falloc(td, &fp1, &fd, oflag);
  753         if (error != 0)
  754                 goto free2;
  755         rsv[0] = fd;
  756         fp1->f_data = so1;      /* so1 already has ref count */
  757         error = falloc(td, &fp2, &fd, oflag);
  758         if (error != 0)
  759                 goto free3;
  760         fp2->f_data = so2;      /* so2 already has ref count */
  761         rsv[1] = fd;
  762         error = soconnect2(so1, so2);
  763         if (error != 0)
  764                 goto free4;
  765         if (type == SOCK_DGRAM) {
  766                 /*
  767                  * Datagram socket connection is asymmetric.
  768                  */
  769                  error = soconnect2(so2, so1);
  770                  if (error != 0)
  771                         goto free4;
  772         }
  773         finit(fp1, FREAD | FWRITE | fflag, DTYPE_SOCKET, fp1->f_data,
  774             &socketops);
  775         finit(fp2, FREAD | FWRITE | fflag, DTYPE_SOCKET, fp2->f_data,
  776             &socketops);
  777         if ((fflag & FNONBLOCK) != 0) {
  778                 (void) fo_ioctl(fp1, FIONBIO, &fflag, td->td_ucred, td);
  779                 (void) fo_ioctl(fp2, FIONBIO, &fflag, td->td_ucred, td);
  780         }
  781         fdrop(fp1, td);
  782         fdrop(fp2, td);
  783         return (0);
  784 free4:
  785         fdclose(fdp, fp2, rsv[1], td);
  786         fdrop(fp2, td);
  787 free3:
  788         fdclose(fdp, fp1, rsv[0], td);
  789         fdrop(fp1, td);
  790 free2:
  791         if (so2 != NULL)
  792                 (void)soclose(so2);
  793 free1:
  794         if (so1 != NULL)
  795                 (void)soclose(so1);
  796         return (error);
  797 }
  798 
  799 int
  800 sys_socketpair(struct thread *td, struct socketpair_args *uap)
  801 {
  802         int error, sv[2];
  803 
  804         error = kern_socketpair(td, uap->domain, uap->type,
  805             uap->protocol, sv);
  806         if (error != 0)
  807                 return (error);
  808         error = copyout(sv, uap->rsv, 2 * sizeof(int));
  809         if (error != 0) {
  810                 (void)kern_close(td, sv[0]);
  811                 (void)kern_close(td, sv[1]);
  812         }
  813         return (error);
  814 }
  815 
  816 static int
  817 sendit(td, s, mp, flags)
  818         struct thread *td;
  819         int s;
  820         struct msghdr *mp;
  821         int flags;
  822 {
  823         struct mbuf *control;
  824         struct sockaddr *to;
  825         int error;
  826 
  827 #ifdef CAPABILITY_MODE
  828         if (IN_CAPABILITY_MODE(td) && (mp->msg_name != NULL))
  829                 return (ECAPMODE);
  830 #endif
  831 
  832         if (mp->msg_name != NULL) {
  833                 error = getsockaddr(&to, mp->msg_name, mp->msg_namelen);
  834                 if (error != 0) {
  835                         to = NULL;
  836                         goto bad;
  837                 }
  838                 mp->msg_name = to;
  839         } else {
  840                 to = NULL;
  841         }
  842 
  843         if (mp->msg_control) {
  844                 if (mp->msg_controllen < sizeof(struct cmsghdr)
  845 #ifdef COMPAT_OLDSOCK
  846                     && mp->msg_flags != MSG_COMPAT
  847 #endif
  848                 ) {
  849                         error = EINVAL;
  850                         goto bad;
  851                 }
  852                 error = sockargs(&control, mp->msg_control,
  853                     mp->msg_controllen, MT_CONTROL);
  854                 if (error != 0)
  855                         goto bad;
  856 #ifdef COMPAT_OLDSOCK
  857                 if (mp->msg_flags == MSG_COMPAT) {
  858                         struct cmsghdr *cm;
  859 
  860                         M_PREPEND(control, sizeof(*cm), M_WAITOK);
  861                         cm = mtod(control, struct cmsghdr *);
  862                         cm->cmsg_len = control->m_len;
  863                         cm->cmsg_level = SOL_SOCKET;
  864                         cm->cmsg_type = SCM_RIGHTS;
  865                 }
  866 #endif
  867         } else {
  868                 control = NULL;
  869         }
  870 
  871         error = kern_sendit(td, s, mp, flags, control, UIO_USERSPACE);
  872 
  873 bad:
  874         free(to, M_SONAME);
  875         return (error);
  876 }
  877 
  878 int
  879 kern_sendit(td, s, mp, flags, control, segflg)
  880         struct thread *td;
  881         int s;
  882         struct msghdr *mp;
  883         int flags;
  884         struct mbuf *control;
  885         enum uio_seg segflg;
  886 {
  887         struct file *fp;
  888         struct uio auio;
  889         struct iovec *iov;
  890         struct socket *so;
  891         cap_rights_t rights;
  892 #ifdef KTRACE
  893         struct uio *ktruio = NULL;
  894 #endif
  895         ssize_t len;
  896         int i, error;
  897 
  898         AUDIT_ARG_FD(s);
  899         cap_rights_init(&rights, CAP_SEND);
  900         if (mp->msg_name != NULL) {
  901                 AUDIT_ARG_SOCKADDR(td, AT_FDCWD, mp->msg_name);
  902                 cap_rights_set(&rights, CAP_CONNECT);
  903         }
  904         error = getsock_cap(td->td_proc->p_fd, s, &rights, &fp, NULL);
  905         if (error != 0)
  906                 return (error);
  907         so = (struct socket *)fp->f_data;
  908 
  909 #ifdef KTRACE
  910         if (mp->msg_name != NULL && KTRPOINT(td, KTR_STRUCT))
  911                 ktrsockaddr(mp->msg_name);
  912 #endif
  913 #ifdef MAC
  914         if (mp->msg_name != NULL) {
  915                 error = mac_socket_check_connect(td->td_ucred, so,
  916                     mp->msg_name);
  917                 if (error != 0)
  918                         goto bad;
  919         }
  920         error = mac_socket_check_send(td->td_ucred, so);
  921         if (error != 0)
  922                 goto bad;
  923 #endif
  924 
  925         auio.uio_iov = mp->msg_iov;
  926         auio.uio_iovcnt = mp->msg_iovlen;
  927         auio.uio_segflg = segflg;
  928         auio.uio_rw = UIO_WRITE;
  929         auio.uio_td = td;
  930         auio.uio_offset = 0;                    /* XXX */
  931         auio.uio_resid = 0;
  932         iov = mp->msg_iov;
  933         for (i = 0; i < mp->msg_iovlen; i++, iov++) {
  934                 if ((auio.uio_resid += iov->iov_len) < 0) {
  935                         error = EINVAL;
  936                         goto bad;
  937                 }
  938         }
  939 #ifdef KTRACE
  940         if (KTRPOINT(td, KTR_GENIO))
  941                 ktruio = cloneuio(&auio);
  942 #endif
  943         len = auio.uio_resid;
  944         error = sosend(so, mp->msg_name, &auio, 0, control, flags, td);
  945         if (error != 0) {
  946                 if (auio.uio_resid != len && (error == ERESTART ||
  947                     error == EINTR || error == EWOULDBLOCK))
  948                         error = 0;
  949                 /* Generation of SIGPIPE can be controlled per socket */
  950                 if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
  951                     !(flags & MSG_NOSIGNAL)) {
  952                         PROC_LOCK(td->td_proc);
  953                         tdsignal(td, SIGPIPE);
  954                         PROC_UNLOCK(td->td_proc);
  955                 }
  956         }
  957         if (error == 0)
  958                 td->td_retval[0] = len - auio.uio_resid;
  959 #ifdef KTRACE
  960         if (ktruio != NULL) {
  961                 ktruio->uio_resid = td->td_retval[0];
  962                 ktrgenio(s, UIO_WRITE, ktruio, error);
  963         }
  964 #endif
  965 bad:
  966         fdrop(fp, td);
  967         return (error);
  968 }
  969 
  970 int
  971 sys_sendto(td, uap)
  972         struct thread *td;
  973         struct sendto_args /* {
  974                 int     s;
  975                 caddr_t buf;
  976                 size_t  len;
  977                 int     flags;
  978                 caddr_t to;
  979                 int     tolen;
  980         } */ *uap;
  981 {
  982         struct msghdr msg;
  983         struct iovec aiov;
  984 
  985         msg.msg_name = uap->to;
  986         msg.msg_namelen = uap->tolen;
  987         msg.msg_iov = &aiov;
  988         msg.msg_iovlen = 1;
  989         msg.msg_control = 0;
  990 #ifdef COMPAT_OLDSOCK
  991         msg.msg_flags = 0;
  992 #endif
  993         aiov.iov_base = uap->buf;
  994         aiov.iov_len = uap->len;
  995         return (sendit(td, uap->s, &msg, uap->flags));
  996 }
  997 
  998 #ifdef COMPAT_OLDSOCK
  999 int
 1000 osend(td, uap)
 1001         struct thread *td;
 1002         struct osend_args /* {
 1003                 int     s;
 1004                 caddr_t buf;
 1005                 int     len;
 1006                 int     flags;
 1007         } */ *uap;
 1008 {
 1009         struct msghdr msg;
 1010         struct iovec aiov;
 1011 
 1012         msg.msg_name = 0;
 1013         msg.msg_namelen = 0;
 1014         msg.msg_iov = &aiov;
 1015         msg.msg_iovlen = 1;
 1016         aiov.iov_base = uap->buf;
 1017         aiov.iov_len = uap->len;
 1018         msg.msg_control = 0;
 1019         msg.msg_flags = 0;
 1020         return (sendit(td, uap->s, &msg, uap->flags));
 1021 }
 1022 
 1023 int
 1024 osendmsg(td, uap)
 1025         struct thread *td;
 1026         struct osendmsg_args /* {
 1027                 int     s;
 1028                 caddr_t msg;
 1029                 int     flags;
 1030         } */ *uap;
 1031 {
 1032         struct msghdr msg;
 1033         struct iovec *iov;
 1034         int error;
 1035 
 1036         error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
 1037         if (error != 0)
 1038                 return (error);
 1039         error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
 1040         if (error != 0)
 1041                 return (error);
 1042         msg.msg_iov = iov;
 1043         msg.msg_flags = MSG_COMPAT;
 1044         error = sendit(td, uap->s, &msg, uap->flags);
 1045         free(iov, M_IOV);
 1046         return (error);
 1047 }
 1048 #endif
 1049 
 1050 int
 1051 sys_sendmsg(td, uap)
 1052         struct thread *td;
 1053         struct sendmsg_args /* {
 1054                 int     s;
 1055                 caddr_t msg;
 1056                 int     flags;
 1057         } */ *uap;
 1058 {
 1059         struct msghdr msg;
 1060         struct iovec *iov;
 1061         int error;
 1062 
 1063         error = copyin(uap->msg, &msg, sizeof (msg));
 1064         if (error != 0)
 1065                 return (error);
 1066         error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
 1067         if (error != 0)
 1068                 return (error);
 1069         msg.msg_iov = iov;
 1070 #ifdef COMPAT_OLDSOCK
 1071         msg.msg_flags = 0;
 1072 #endif
 1073         error = sendit(td, uap->s, &msg, uap->flags);
 1074         free(iov, M_IOV);
 1075         return (error);
 1076 }
 1077 
 1078 int
 1079 kern_recvit(td, s, mp, fromseg, controlp)
 1080         struct thread *td;
 1081         int s;
 1082         struct msghdr *mp;
 1083         enum uio_seg fromseg;
 1084         struct mbuf **controlp;
 1085 {
 1086         struct uio auio;
 1087         struct iovec *iov;
 1088         struct mbuf *m, *control = NULL;
 1089         caddr_t ctlbuf;
 1090         struct file *fp;
 1091         struct socket *so;
 1092         struct sockaddr *fromsa = NULL;
 1093         cap_rights_t rights;
 1094 #ifdef KTRACE
 1095         struct uio *ktruio = NULL;
 1096 #endif
 1097         ssize_t len;
 1098         int error, i;
 1099 
 1100         if (controlp != NULL)
 1101                 *controlp = NULL;
 1102 
 1103         AUDIT_ARG_FD(s);
 1104         error = getsock_cap(td->td_proc->p_fd, s,
 1105             cap_rights_init(&rights, CAP_RECV), &fp, NULL);
 1106         if (error != 0)
 1107                 return (error);
 1108         so = fp->f_data;
 1109 
 1110 #ifdef MAC
 1111         error = mac_socket_check_receive(td->td_ucred, so);
 1112         if (error != 0) {
 1113                 fdrop(fp, td);
 1114                 return (error);
 1115         }
 1116 #endif
 1117 
 1118         auio.uio_iov = mp->msg_iov;
 1119         auio.uio_iovcnt = mp->msg_iovlen;
 1120         auio.uio_segflg = UIO_USERSPACE;
 1121         auio.uio_rw = UIO_READ;
 1122         auio.uio_td = td;
 1123         auio.uio_offset = 0;                    /* XXX */
 1124         auio.uio_resid = 0;
 1125         iov = mp->msg_iov;
 1126         for (i = 0; i < mp->msg_iovlen; i++, iov++) {
 1127                 if ((auio.uio_resid += iov->iov_len) < 0) {
 1128                         fdrop(fp, td);
 1129                         return (EINVAL);
 1130                 }
 1131         }
 1132 #ifdef KTRACE
 1133         if (KTRPOINT(td, KTR_GENIO))
 1134                 ktruio = cloneuio(&auio);
 1135 #endif
 1136         len = auio.uio_resid;
 1137         error = soreceive(so, &fromsa, &auio, NULL,
 1138             (mp->msg_control || controlp) ? &control : NULL,
 1139             &mp->msg_flags);
 1140         if (error != 0) {
 1141                 if (auio.uio_resid != len && (error == ERESTART ||
 1142                     error == EINTR || error == EWOULDBLOCK))
 1143                         error = 0;
 1144         }
 1145         if (fromsa != NULL)
 1146                 AUDIT_ARG_SOCKADDR(td, AT_FDCWD, fromsa);
 1147 #ifdef KTRACE
 1148         if (ktruio != NULL) {
 1149                 ktruio->uio_resid = len - auio.uio_resid;
 1150                 ktrgenio(s, UIO_READ, ktruio, error);
 1151         }
 1152 #endif
 1153         if (error != 0)
 1154                 goto out;
 1155         td->td_retval[0] = len - auio.uio_resid;
 1156         if (mp->msg_name) {
 1157                 len = mp->msg_namelen;
 1158                 if (len <= 0 || fromsa == NULL)
 1159                         len = 0;
 1160                 else {
 1161                         /* save sa_len before it is destroyed by MSG_COMPAT */
 1162                         len = MIN(len, fromsa->sa_len);
 1163 #ifdef COMPAT_OLDSOCK
 1164                         if (mp->msg_flags & MSG_COMPAT)
 1165                                 ((struct osockaddr *)fromsa)->sa_family =
 1166                                     fromsa->sa_family;
 1167 #endif
 1168                         if (fromseg == UIO_USERSPACE) {
 1169                                 error = copyout(fromsa, mp->msg_name,
 1170                                     (unsigned)len);
 1171                                 if (error != 0)
 1172                                         goto out;
 1173                         } else
 1174                                 bcopy(fromsa, mp->msg_name, len);
 1175                 }
 1176                 mp->msg_namelen = len;
 1177         }
 1178         if (mp->msg_control && controlp == NULL) {
 1179 #ifdef COMPAT_OLDSOCK
 1180                 /*
 1181                  * We assume that old recvmsg calls won't receive access
 1182                  * rights and other control info, esp. as control info
 1183                  * is always optional and those options didn't exist in 4.3.
 1184                  * If we receive rights, trim the cmsghdr; anything else
 1185                  * is tossed.
 1186                  */
 1187                 if (control && mp->msg_flags & MSG_COMPAT) {
 1188                         if (mtod(control, struct cmsghdr *)->cmsg_level !=
 1189                             SOL_SOCKET ||
 1190                             mtod(control, struct cmsghdr *)->cmsg_type !=
 1191                             SCM_RIGHTS) {
 1192                                 mp->msg_controllen = 0;
 1193                                 goto out;
 1194                         }
 1195                         control->m_len -= sizeof (struct cmsghdr);
 1196                         control->m_data += sizeof (struct cmsghdr);
 1197                 }
 1198 #endif
 1199                 len = mp->msg_controllen;
 1200                 m = control;
 1201                 mp->msg_controllen = 0;
 1202                 ctlbuf = mp->msg_control;
 1203 
 1204                 while (m && len > 0) {
 1205                         unsigned int tocopy;
 1206 
 1207                         if (len >= m->m_len)
 1208                                 tocopy = m->m_len;
 1209                         else {
 1210                                 mp->msg_flags |= MSG_CTRUNC;
 1211                                 tocopy = len;
 1212                         }
 1213 
 1214                         if ((error = copyout(mtod(m, caddr_t),
 1215                                         ctlbuf, tocopy)) != 0)
 1216                                 goto out;
 1217 
 1218                         ctlbuf += tocopy;
 1219                         len -= tocopy;
 1220                         m = m->m_next;
 1221                 }
 1222                 mp->msg_controllen = ctlbuf - (caddr_t)mp->msg_control;
 1223         }
 1224 out:
 1225         fdrop(fp, td);
 1226 #ifdef KTRACE
 1227         if (fromsa && KTRPOINT(td, KTR_STRUCT))
 1228                 ktrsockaddr(fromsa);
 1229 #endif
 1230         free(fromsa, M_SONAME);
 1231 
 1232         if (error == 0 && controlp != NULL)
 1233                 *controlp = control;
 1234         else  if (control)
 1235                 m_freem(control);
 1236 
 1237         return (error);
 1238 }
 1239 
 1240 static int
 1241 recvit(td, s, mp, namelenp)
 1242         struct thread *td;
 1243         int s;
 1244         struct msghdr *mp;
 1245         void *namelenp;
 1246 {
 1247         int error;
 1248 
 1249         error = kern_recvit(td, s, mp, UIO_USERSPACE, NULL);
 1250         if (error != 0)
 1251                 return (error);
 1252         if (namelenp != NULL) {
 1253                 error = copyout(&mp->msg_namelen, namelenp, sizeof (socklen_t));
 1254 #ifdef COMPAT_OLDSOCK
 1255                 if (mp->msg_flags & MSG_COMPAT)
 1256                         error = 0;      /* old recvfrom didn't check */
 1257 #endif
 1258         }
 1259         return (error);
 1260 }
 1261 
 1262 int
 1263 sys_recvfrom(td, uap)
 1264         struct thread *td;
 1265         struct recvfrom_args /* {
 1266                 int     s;
 1267                 caddr_t buf;
 1268                 size_t  len;
 1269                 int     flags;
 1270                 struct sockaddr * __restrict    from;
 1271                 socklen_t * __restrict fromlenaddr;
 1272         } */ *uap;
 1273 {
 1274         struct msghdr msg;
 1275         struct iovec aiov;
 1276         int error;
 1277 
 1278         if (uap->fromlenaddr) {
 1279                 error = copyin(uap->fromlenaddr,
 1280                     &msg.msg_namelen, sizeof (msg.msg_namelen));
 1281                 if (error != 0)
 1282                         goto done2;
 1283         } else {
 1284                 msg.msg_namelen = 0;
 1285         }
 1286         msg.msg_name = uap->from;
 1287         msg.msg_iov = &aiov;
 1288         msg.msg_iovlen = 1;
 1289         aiov.iov_base = uap->buf;
 1290         aiov.iov_len = uap->len;
 1291         msg.msg_control = 0;
 1292         msg.msg_flags = uap->flags;
 1293         error = recvit(td, uap->s, &msg, uap->fromlenaddr);
 1294 done2:
 1295         return (error);
 1296 }
 1297 
 1298 #ifdef COMPAT_OLDSOCK
 1299 int
 1300 orecvfrom(td, uap)
 1301         struct thread *td;
 1302         struct recvfrom_args *uap;
 1303 {
 1304 
 1305         uap->flags |= MSG_COMPAT;
 1306         return (sys_recvfrom(td, uap));
 1307 }
 1308 #endif
 1309 
 1310 #ifdef COMPAT_OLDSOCK
 1311 int
 1312 orecv(td, uap)
 1313         struct thread *td;
 1314         struct orecv_args /* {
 1315                 int     s;
 1316                 caddr_t buf;
 1317                 int     len;
 1318                 int     flags;
 1319         } */ *uap;
 1320 {
 1321         struct msghdr msg;
 1322         struct iovec aiov;
 1323 
 1324         msg.msg_name = 0;
 1325         msg.msg_namelen = 0;
 1326         msg.msg_iov = &aiov;
 1327         msg.msg_iovlen = 1;
 1328         aiov.iov_base = uap->buf;
 1329         aiov.iov_len = uap->len;
 1330         msg.msg_control = 0;
 1331         msg.msg_flags = uap->flags;
 1332         return (recvit(td, uap->s, &msg, NULL));
 1333 }
 1334 
 1335 /*
 1336  * Old recvmsg.  This code takes advantage of the fact that the old msghdr
 1337  * overlays the new one, missing only the flags, and with the (old) access
 1338  * rights where the control fields are now.
 1339  */
 1340 int
 1341 orecvmsg(td, uap)
 1342         struct thread *td;
 1343         struct orecvmsg_args /* {
 1344                 int     s;
 1345                 struct  omsghdr *msg;
 1346                 int     flags;
 1347         } */ *uap;
 1348 {
 1349         struct msghdr msg;
 1350         struct iovec *iov;
 1351         int error;
 1352 
 1353         error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
 1354         if (error != 0)
 1355                 return (error);
 1356         error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
 1357         if (error != 0)
 1358                 return (error);
 1359         msg.msg_flags = uap->flags | MSG_COMPAT;
 1360         msg.msg_iov = iov;
 1361         error = recvit(td, uap->s, &msg, &uap->msg->msg_namelen);
 1362         if (msg.msg_controllen && error == 0)
 1363                 error = copyout(&msg.msg_controllen,
 1364                     &uap->msg->msg_accrightslen, sizeof (int));
 1365         free(iov, M_IOV);
 1366         return (error);
 1367 }
 1368 #endif
 1369 
 1370 int
 1371 sys_recvmsg(td, uap)
 1372         struct thread *td;
 1373         struct recvmsg_args /* {
 1374                 int     s;
 1375                 struct  msghdr *msg;
 1376                 int     flags;
 1377         } */ *uap;
 1378 {
 1379         struct msghdr msg;
 1380         struct iovec *uiov, *iov;
 1381         int error;
 1382 
 1383         error = copyin(uap->msg, &msg, sizeof (msg));
 1384         if (error != 0)
 1385                 return (error);
 1386         error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
 1387         if (error != 0)
 1388                 return (error);
 1389         msg.msg_flags = uap->flags;
 1390 #ifdef COMPAT_OLDSOCK
 1391         msg.msg_flags &= ~MSG_COMPAT;
 1392 #endif
 1393         uiov = msg.msg_iov;
 1394         msg.msg_iov = iov;
 1395         error = recvit(td, uap->s, &msg, NULL);
 1396         if (error == 0) {
 1397                 msg.msg_iov = uiov;
 1398                 error = copyout(&msg, uap->msg, sizeof(msg));
 1399         }
 1400         free(iov, M_IOV);
 1401         return (error);
 1402 }
 1403 
 1404 /* ARGSUSED */
 1405 int
 1406 sys_shutdown(td, uap)
 1407         struct thread *td;
 1408         struct shutdown_args /* {
 1409                 int     s;
 1410                 int     how;
 1411         } */ *uap;
 1412 {
 1413         struct socket *so;
 1414         struct file *fp;
 1415         cap_rights_t rights;
 1416         int error;
 1417 
 1418         AUDIT_ARG_FD(uap->s);
 1419         error = getsock_cap(td->td_proc->p_fd, uap->s,
 1420             cap_rights_init(&rights, CAP_SHUTDOWN), &fp, NULL);
 1421         if (error == 0) {
 1422                 so = fp->f_data;
 1423                 error = soshutdown(so, uap->how);
 1424                 fdrop(fp, td);
 1425         }
 1426         return (error);
 1427 }
 1428 
 1429 /* ARGSUSED */
 1430 int
 1431 sys_setsockopt(td, uap)
 1432         struct thread *td;
 1433         struct setsockopt_args /* {
 1434                 int     s;
 1435                 int     level;
 1436                 int     name;
 1437                 caddr_t val;
 1438                 int     valsize;
 1439         } */ *uap;
 1440 {
 1441 
 1442         return (kern_setsockopt(td, uap->s, uap->level, uap->name,
 1443             uap->val, UIO_USERSPACE, uap->valsize));
 1444 }
 1445 
 1446 int
 1447 kern_setsockopt(td, s, level, name, val, valseg, valsize)
 1448         struct thread *td;
 1449         int s;
 1450         int level;
 1451         int name;
 1452         void *val;
 1453         enum uio_seg valseg;
 1454         socklen_t valsize;
 1455 {
 1456         struct socket *so;
 1457         struct file *fp;
 1458         struct sockopt sopt;
 1459         cap_rights_t rights;
 1460         int error;
 1461 
 1462         if (val == NULL && valsize != 0)
 1463                 return (EFAULT);
 1464         if ((int)valsize < 0)
 1465                 return (EINVAL);
 1466 
 1467         sopt.sopt_dir = SOPT_SET;
 1468         sopt.sopt_level = level;
 1469         sopt.sopt_name = name;
 1470         sopt.sopt_val = val;
 1471         sopt.sopt_valsize = valsize;
 1472         switch (valseg) {
 1473         case UIO_USERSPACE:
 1474                 sopt.sopt_td = td;
 1475                 break;
 1476         case UIO_SYSSPACE:
 1477                 sopt.sopt_td = NULL;
 1478                 break;
 1479         default:
 1480                 panic("kern_setsockopt called with bad valseg");
 1481         }
 1482 
 1483         AUDIT_ARG_FD(s);
 1484         error = getsock_cap(td->td_proc->p_fd, s,
 1485             cap_rights_init(&rights, CAP_SETSOCKOPT), &fp, NULL);
 1486         if (error == 0) {
 1487                 so = fp->f_data;
 1488                 error = sosetopt(so, &sopt);
 1489                 fdrop(fp, td);
 1490         }
 1491         return(error);
 1492 }
 1493 
 1494 /* ARGSUSED */
 1495 int
 1496 sys_getsockopt(td, uap)
 1497         struct thread *td;
 1498         struct getsockopt_args /* {
 1499                 int     s;
 1500                 int     level;
 1501                 int     name;
 1502                 void * __restrict       val;
 1503                 socklen_t * __restrict avalsize;
 1504         } */ *uap;
 1505 {
 1506         socklen_t valsize;
 1507         int error;
 1508 
 1509         if (uap->val) {
 1510                 error = copyin(uap->avalsize, &valsize, sizeof (valsize));
 1511                 if (error != 0)
 1512                         return (error);
 1513         }
 1514 
 1515         error = kern_getsockopt(td, uap->s, uap->level, uap->name,
 1516             uap->val, UIO_USERSPACE, &valsize);
 1517 
 1518         if (error == 0)
 1519                 error = copyout(&valsize, uap->avalsize, sizeof (valsize));
 1520         return (error);
 1521 }
 1522 
 1523 /*
 1524  * Kernel version of getsockopt.
 1525  * optval can be a userland or userspace. optlen is always a kernel pointer.
 1526  */
 1527 int
 1528 kern_getsockopt(td, s, level, name, val, valseg, valsize)
 1529         struct thread *td;
 1530         int s;
 1531         int level;
 1532         int name;
 1533         void *val;
 1534         enum uio_seg valseg;
 1535         socklen_t *valsize;
 1536 {
 1537         struct socket *so;
 1538         struct file *fp;
 1539         struct sockopt sopt;
 1540         cap_rights_t rights;
 1541         int error;
 1542 
 1543         if (val == NULL)
 1544                 *valsize = 0;
 1545         if ((int)*valsize < 0)
 1546                 return (EINVAL);
 1547 
 1548         sopt.sopt_dir = SOPT_GET;
 1549         sopt.sopt_level = level;
 1550         sopt.sopt_name = name;
 1551         sopt.sopt_val = val;
 1552         sopt.sopt_valsize = (size_t)*valsize; /* checked non-negative above */
 1553         switch (valseg) {
 1554         case UIO_USERSPACE:
 1555                 sopt.sopt_td = td;
 1556                 break;
 1557         case UIO_SYSSPACE:
 1558                 sopt.sopt_td = NULL;
 1559                 break;
 1560         default:
 1561                 panic("kern_getsockopt called with bad valseg");
 1562         }
 1563 
 1564         AUDIT_ARG_FD(s);
 1565         error = getsock_cap(td->td_proc->p_fd, s,
 1566             cap_rights_init(&rights, CAP_GETSOCKOPT), &fp, NULL);
 1567         if (error == 0) {
 1568                 so = fp->f_data;
 1569                 error = sogetopt(so, &sopt);
 1570                 *valsize = sopt.sopt_valsize;
 1571                 fdrop(fp, td);
 1572         }
 1573         return (error);
 1574 }
 1575 
 1576 /*
 1577  * getsockname1() - Get socket name.
 1578  */
 1579 /* ARGSUSED */
 1580 static int
 1581 getsockname1(td, uap, compat)
 1582         struct thread *td;
 1583         struct getsockname_args /* {
 1584                 int     fdes;
 1585                 struct sockaddr * __restrict asa;
 1586                 socklen_t * __restrict alen;
 1587         } */ *uap;
 1588         int compat;
 1589 {
 1590         struct sockaddr *sa;
 1591         socklen_t len;
 1592         int error;
 1593 
 1594         error = copyin(uap->alen, &len, sizeof(len));
 1595         if (error != 0)
 1596                 return (error);
 1597 
 1598         error = kern_getsockname(td, uap->fdes, &sa, &len);
 1599         if (error != 0)
 1600                 return (error);
 1601 
 1602         if (len != 0) {
 1603 #ifdef COMPAT_OLDSOCK
 1604                 if (compat)
 1605                         ((struct osockaddr *)sa)->sa_family = sa->sa_family;
 1606 #endif
 1607                 error = copyout(sa, uap->asa, (u_int)len);
 1608         }
 1609         free(sa, M_SONAME);
 1610         if (error == 0)
 1611                 error = copyout(&len, uap->alen, sizeof(len));
 1612         return (error);
 1613 }
 1614 
 1615 int
 1616 kern_getsockname(struct thread *td, int fd, struct sockaddr **sa,
 1617     socklen_t *alen)
 1618 {
 1619         struct socket *so;
 1620         struct file *fp;
 1621         cap_rights_t rights;
 1622         socklen_t len;
 1623         int error;
 1624 
 1625         AUDIT_ARG_FD(fd);
 1626         error = getsock_cap(td->td_proc->p_fd, fd,
 1627             cap_rights_init(&rights, CAP_GETSOCKNAME), &fp, NULL);
 1628         if (error != 0)
 1629                 return (error);
 1630         so = fp->f_data;
 1631         *sa = NULL;
 1632         CURVNET_SET(so->so_vnet);
 1633         error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, sa);
 1634         CURVNET_RESTORE();
 1635         if (error != 0)
 1636                 goto bad;
 1637         if (*sa == NULL)
 1638                 len = 0;
 1639         else
 1640                 len = MIN(*alen, (*sa)->sa_len);
 1641         *alen = len;
 1642 #ifdef KTRACE
 1643         if (KTRPOINT(td, KTR_STRUCT))
 1644                 ktrsockaddr(*sa);
 1645 #endif
 1646 bad:
 1647         fdrop(fp, td);
 1648         if (error != 0 && *sa != NULL) {
 1649                 free(*sa, M_SONAME);
 1650                 *sa = NULL;
 1651         }
 1652         return (error);
 1653 }
 1654 
 1655 int
 1656 sys_getsockname(td, uap)
 1657         struct thread *td;
 1658         struct getsockname_args *uap;
 1659 {
 1660 
 1661         return (getsockname1(td, uap, 0));
 1662 }
 1663 
 1664 #ifdef COMPAT_OLDSOCK
 1665 int
 1666 ogetsockname(td, uap)
 1667         struct thread *td;
 1668         struct getsockname_args *uap;
 1669 {
 1670 
 1671         return (getsockname1(td, uap, 1));
 1672 }
 1673 #endif /* COMPAT_OLDSOCK */
 1674 
 1675 /*
 1676  * getpeername1() - Get name of peer for connected socket.
 1677  */
 1678 /* ARGSUSED */
 1679 static int
 1680 getpeername1(td, uap, compat)
 1681         struct thread *td;
 1682         struct getpeername_args /* {
 1683                 int     fdes;
 1684                 struct sockaddr * __restrict    asa;
 1685                 socklen_t * __restrict  alen;
 1686         } */ *uap;
 1687         int compat;
 1688 {
 1689         struct sockaddr *sa;
 1690         socklen_t len;
 1691         int error;
 1692 
 1693         error = copyin(uap->alen, &len, sizeof (len));
 1694         if (error != 0)
 1695                 return (error);
 1696 
 1697         error = kern_getpeername(td, uap->fdes, &sa, &len);
 1698         if (error != 0)
 1699                 return (error);
 1700 
 1701         if (len != 0) {
 1702 #ifdef COMPAT_OLDSOCK
 1703                 if (compat)
 1704                         ((struct osockaddr *)sa)->sa_family = sa->sa_family;
 1705 #endif
 1706                 error = copyout(sa, uap->asa, (u_int)len);
 1707         }
 1708         free(sa, M_SONAME);
 1709         if (error == 0)
 1710                 error = copyout(&len, uap->alen, sizeof(len));
 1711         return (error);
 1712 }
 1713 
 1714 int
 1715 kern_getpeername(struct thread *td, int fd, struct sockaddr **sa,
 1716     socklen_t *alen)
 1717 {
 1718         struct socket *so;
 1719         struct file *fp;
 1720         cap_rights_t rights;
 1721         socklen_t len;
 1722         int error;
 1723 
 1724         AUDIT_ARG_FD(fd);
 1725         error = getsock_cap(td->td_proc->p_fd, fd,
 1726             cap_rights_init(&rights, CAP_GETPEERNAME), &fp, NULL);
 1727         if (error != 0)
 1728                 return (error);
 1729         so = fp->f_data;
 1730         if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
 1731                 error = ENOTCONN;
 1732                 goto done;
 1733         }
 1734         *sa = NULL;
 1735         CURVNET_SET(so->so_vnet);
 1736         error = (*so->so_proto->pr_usrreqs->pru_peeraddr)(so, sa);
 1737         CURVNET_RESTORE();
 1738         if (error != 0)
 1739                 goto bad;
 1740         if (*sa == NULL)
 1741                 len = 0;
 1742         else
 1743                 len = MIN(*alen, (*sa)->sa_len);
 1744         *alen = len;
 1745 #ifdef KTRACE
 1746         if (KTRPOINT(td, KTR_STRUCT))
 1747                 ktrsockaddr(*sa);
 1748 #endif
 1749 bad:
 1750         if (error != 0 && *sa != NULL) {
 1751                 free(*sa, M_SONAME);
 1752                 *sa = NULL;
 1753         }
 1754 done:
 1755         fdrop(fp, td);
 1756         return (error);
 1757 }
 1758 
 1759 int
 1760 sys_getpeername(td, uap)
 1761         struct thread *td;
 1762         struct getpeername_args *uap;
 1763 {
 1764 
 1765         return (getpeername1(td, uap, 0));
 1766 }
 1767 
 1768 #ifdef COMPAT_OLDSOCK
 1769 int
 1770 ogetpeername(td, uap)
 1771         struct thread *td;
 1772         struct ogetpeername_args *uap;
 1773 {
 1774 
 1775         /* XXX uap should have type `getpeername_args *' to begin with. */
 1776         return (getpeername1(td, (struct getpeername_args *)uap, 1));
 1777 }
 1778 #endif /* COMPAT_OLDSOCK */
 1779 
 1780 int
 1781 sockargs(mp, buf, buflen, type)
 1782         struct mbuf **mp;
 1783         caddr_t buf;
 1784         int buflen, type;
 1785 {
 1786         struct sockaddr *sa;
 1787         struct mbuf *m;
 1788         int error;
 1789 
 1790         if (buflen < 0)
 1791                 return (EINVAL);
 1792 
 1793         if (buflen > MLEN) {
 1794 #ifdef COMPAT_OLDSOCK
 1795                 if (type == MT_SONAME && buflen <= 112)
 1796                         buflen = MLEN;          /* unix domain compat. hack */
 1797                 else
 1798 #endif
 1799                         if (buflen > MCLBYTES)
 1800                                 return (EINVAL);
 1801         }
 1802         m = m_get2(buflen, M_WAITOK, type, 0);
 1803         m->m_len = buflen;
 1804         error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
 1805         if (error != 0)
 1806                 (void) m_free(m);
 1807         else {
 1808                 *mp = m;
 1809                 if (type == MT_SONAME) {
 1810                         sa = mtod(m, struct sockaddr *);
 1811 
 1812 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
 1813                         if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
 1814                                 sa->sa_family = sa->sa_len;
 1815 #endif
 1816                         sa->sa_len = buflen;
 1817                 }
 1818         }
 1819         return (error);
 1820 }
 1821 
 1822 int
 1823 getsockaddr(namp, uaddr, len)
 1824         struct sockaddr **namp;
 1825         caddr_t uaddr;
 1826         size_t len;
 1827 {
 1828         struct sockaddr *sa;
 1829         int error;
 1830 
 1831         if (len > SOCK_MAXADDRLEN)
 1832                 return (ENAMETOOLONG);
 1833         if (len < offsetof(struct sockaddr, sa_data[0]))
 1834                 return (EINVAL);
 1835         sa = malloc(len, M_SONAME, M_WAITOK);
 1836         error = copyin(uaddr, sa, len);
 1837         if (error != 0) {
 1838                 free(sa, M_SONAME);
 1839         } else {
 1840 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
 1841                 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
 1842                         sa->sa_family = sa->sa_len;
 1843 #endif
 1844                 sa->sa_len = len;
 1845                 *namp = sa;
 1846         }
 1847         return (error);
 1848 }
 1849 
 1850 struct sendfile_sync {
 1851         struct mtx      mtx;
 1852         struct cv       cv;
 1853         unsigned        count;
 1854 };
 1855 
 1856 /*
 1857  * Detach mapped page and release resources back to the system.
 1858  */
 1859 int
 1860 sf_buf_mext(struct mbuf *mb, void *addr, void *args)
 1861 {
 1862         vm_page_t m;
 1863         struct sendfile_sync *sfs;
 1864 
 1865         m = sf_buf_page(args);
 1866         sf_buf_free(args);
 1867         vm_page_lock(m);
 1868         vm_page_unwire(m, 0);
 1869         /*
 1870          * Check for the object going away on us. This can
 1871          * happen since we don't hold a reference to it.
 1872          * If so, we're responsible for freeing the page.
 1873          */
 1874         if (m->wire_count == 0 && m->object == NULL)
 1875                 vm_page_free(m);
 1876         vm_page_unlock(m);
 1877         if (addr == NULL)
 1878                 return (EXT_FREE_OK);
 1879         sfs = addr;
 1880         mtx_lock(&sfs->mtx);
 1881         KASSERT(sfs->count> 0, ("Sendfile sync botchup count == 0"));
 1882         if (--sfs->count == 0)
 1883                 cv_signal(&sfs->cv);
 1884         mtx_unlock(&sfs->mtx);
 1885         return (EXT_FREE_OK);
 1886 }
 1887 
 1888 /*
 1889  * sendfile(2)
 1890  *
 1891  * int sendfile(int fd, int s, off_t offset, size_t nbytes,
 1892  *       struct sf_hdtr *hdtr, off_t *sbytes, int flags)
 1893  *
 1894  * Send a file specified by 'fd' and starting at 'offset' to a socket
 1895  * specified by 's'. Send only 'nbytes' of the file or until EOF if nbytes ==
 1896  * 0.  Optionally add a header and/or trailer to the socket output.  If
 1897  * specified, write the total number of bytes sent into *sbytes.
 1898  */
 1899 int
 1900 sys_sendfile(struct thread *td, struct sendfile_args *uap)
 1901 {
 1902 
 1903         return (do_sendfile(td, uap, 0));
 1904 }
 1905 
 1906 static int
 1907 do_sendfile(struct thread *td, struct sendfile_args *uap, int compat)
 1908 {
 1909         struct sf_hdtr hdtr;
 1910         struct uio *hdr_uio, *trl_uio;
 1911         struct file *fp;
 1912         cap_rights_t rights;
 1913         int error;
 1914 
 1915         /*
 1916          * File offset must be positive.  If it goes beyond EOF
 1917          * we send only the header/trailer and no payload data.
 1918          */
 1919         if (uap->offset < 0)
 1920                 return (EINVAL);
 1921 
 1922         hdr_uio = trl_uio = NULL;
 1923 
 1924         if (uap->hdtr != NULL) {
 1925                 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
 1926                 if (error != 0)
 1927                         goto out;
 1928                 if (hdtr.headers != NULL) {
 1929                         error = copyinuio(hdtr.headers, hdtr.hdr_cnt, &hdr_uio);
 1930                         if (error != 0)
 1931                                 goto out;
 1932                 }
 1933                 if (hdtr.trailers != NULL) {
 1934                         error = copyinuio(hdtr.trailers, hdtr.trl_cnt, &trl_uio);
 1935                         if (error != 0)
 1936                                 goto out;
 1937 
 1938                 }
 1939         }
 1940 
 1941         AUDIT_ARG_FD(uap->fd);
 1942 
 1943         /*
 1944          * sendfile(2) can start at any offset within a file so we require
 1945          * CAP_READ+CAP_SEEK = CAP_PREAD.
 1946          */
 1947         if ((error = fget_read(td, uap->fd,
 1948             cap_rights_init(&rights, CAP_PREAD), &fp)) != 0) {
 1949                 goto out;
 1950         }
 1951 
 1952         error = fo_sendfile(fp, uap->s, hdr_uio, trl_uio, uap->offset,
 1953             uap->nbytes, uap->sbytes, uap->flags, compat ? SFK_COMPAT : 0, td);
 1954         fdrop(fp, td);
 1955 
 1956 out:
 1957         free(hdr_uio, M_IOV);
 1958         free(trl_uio, M_IOV);
 1959         return (error);
 1960 }
 1961 
 1962 #ifdef COMPAT_FREEBSD4
 1963 int
 1964 freebsd4_sendfile(struct thread *td, struct freebsd4_sendfile_args *uap)
 1965 {
 1966         struct sendfile_args args;
 1967 
 1968         args.fd = uap->fd;
 1969         args.s = uap->s;
 1970         args.offset = uap->offset;
 1971         args.nbytes = uap->nbytes;
 1972         args.hdtr = uap->hdtr;
 1973         args.sbytes = uap->sbytes;
 1974         args.flags = uap->flags;
 1975 
 1976         return (do_sendfile(td, &args, 1));
 1977 }
 1978 #endif /* COMPAT_FREEBSD4 */
 1979 
 1980 static int
 1981 sendfile_readpage(vm_object_t obj, struct vnode *vp, int nd,
 1982     off_t off, int xfsize, int bsize, struct thread *td, vm_page_t *res)
 1983 {
 1984         vm_page_t m;
 1985         vm_pindex_t pindex;
 1986         ssize_t resid;
 1987         int error, readahead, rv;
 1988 
 1989         pindex = OFF_TO_IDX(off);
 1990         VM_OBJECT_WLOCK(obj);
 1991         m = vm_page_grab(obj, pindex, (vp != NULL ? VM_ALLOC_NOBUSY |
 1992             VM_ALLOC_IGN_SBUSY : 0) | VM_ALLOC_WIRED | VM_ALLOC_NORMAL);
 1993 
 1994         /*
 1995          * Check if page is valid for what we need, otherwise initiate I/O.
 1996          *
 1997          * The non-zero nd argument prevents disk I/O, instead we
 1998          * return the caller what he specified in nd.  In particular,
 1999          * if we already turned some pages into mbufs, nd == EAGAIN
 2000          * and the main function send them the pages before we come
 2001          * here again and block.
 2002          */
 2003         if (m->valid != 0 && vm_page_is_valid(m, off & PAGE_MASK, xfsize)) {
 2004                 if (vp == NULL)
 2005                         vm_page_xunbusy(m);
 2006                 VM_OBJECT_WUNLOCK(obj);
 2007                 *res = m;
 2008                 return (0);
 2009         } else if (nd != 0) {
 2010                 if (vp == NULL)
 2011                         vm_page_xunbusy(m);
 2012                 error = nd;
 2013                 goto free_page;
 2014         }
 2015 
 2016         /*
 2017          * Get the page from backing store.
 2018          */
 2019         error = 0;
 2020         if (vp != NULL) {
 2021                 VM_OBJECT_WUNLOCK(obj);
 2022                 readahead = sfreadahead * MAXBSIZE;
 2023 
 2024                 /*
 2025                  * Use vn_rdwr() instead of the pager interface for
 2026                  * the vnode, to allow the read-ahead.
 2027                  *
 2028                  * XXXMAC: Because we don't have fp->f_cred here, we
 2029                  * pass in NOCRED.  This is probably wrong, but is
 2030                  * consistent with our original implementation.
 2031                  */
 2032                 error = vn_rdwr(UIO_READ, vp, NULL, readahead, trunc_page(off),
 2033                     UIO_NOCOPY, IO_NODELOCKED | IO_VMIO | ((readahead /
 2034                     bsize) << IO_SEQSHIFT), td->td_ucred, NOCRED, &resid, td);
 2035                 SFSTAT_INC(sf_iocnt);
 2036                 VM_OBJECT_WLOCK(obj);
 2037         } else {
 2038                 if (vm_pager_has_page(obj, pindex, NULL, NULL)) {
 2039                         rv = vm_pager_get_pages(obj, &m, 1, 0);
 2040                         SFSTAT_INC(sf_iocnt);
 2041                         m = vm_page_lookup(obj, pindex);
 2042                         if (m == NULL)
 2043                                 error = EIO;
 2044                         else if (rv != VM_PAGER_OK) {
 2045                                 vm_page_lock(m);
 2046                                 vm_page_free(m);
 2047                                 vm_page_unlock(m);
 2048                                 m = NULL;
 2049                                 error = EIO;
 2050                         }
 2051                 } else {
 2052                         pmap_zero_page(m);
 2053                         m->valid = VM_PAGE_BITS_ALL;
 2054                         m->dirty = 0;
 2055                 }
 2056                 if (m != NULL)
 2057                         vm_page_xunbusy(m);
 2058         }
 2059         if (error == 0) {
 2060                 *res = m;
 2061         } else if (m != NULL) {
 2062 free_page:
 2063                 vm_page_lock(m);
 2064                 vm_page_unwire(m, 0);
 2065 
 2066                 /*
 2067                  * See if anyone else might know about this page.  If
 2068                  * not and it is not valid, then free it.
 2069                  */
 2070                 if (m->wire_count == 0 && m->valid == 0 && !vm_page_busied(m))
 2071                         vm_page_free(m);
 2072                 vm_page_unlock(m);
 2073         }
 2074         KASSERT(error != 0 || (m->wire_count > 0 &&
 2075             vm_page_is_valid(m, off & PAGE_MASK, xfsize)),
 2076             ("wrong page state m %p off %#jx xfsize %d", m, (uintmax_t)off,
 2077             xfsize));
 2078         VM_OBJECT_WUNLOCK(obj);
 2079         return (error);
 2080 }
 2081 
 2082 static int
 2083 sendfile_getobj(struct thread *td, struct file *fp, vm_object_t *obj_res,
 2084     struct vnode **vp_res, struct shmfd **shmfd_res, off_t *obj_size,
 2085     int *bsize)
 2086 {
 2087         struct vattr va;
 2088         vm_object_t obj;
 2089         struct vnode *vp;
 2090         struct shmfd *shmfd;
 2091         int error;
 2092 
 2093         vp = *vp_res = NULL;
 2094         obj = NULL;
 2095         shmfd = *shmfd_res = NULL;
 2096         *bsize = 0;
 2097 
 2098         /*
 2099          * The file descriptor must be a regular file and have a
 2100          * backing VM object.
 2101          */
 2102         if (fp->f_type == DTYPE_VNODE) {
 2103                 vp = fp->f_vnode;
 2104                 vn_lock(vp, LK_SHARED | LK_RETRY);
 2105                 if (vp->v_type != VREG) {
 2106                         error = EINVAL;
 2107                         goto out;
 2108                 }
 2109                 *bsize = vp->v_mount->mnt_stat.f_iosize;
 2110                 error = VOP_GETATTR(vp, &va, td->td_ucred);
 2111                 if (error != 0)
 2112                         goto out;
 2113                 *obj_size = va.va_size;
 2114                 obj = vp->v_object;
 2115                 if (obj == NULL) {
 2116                         error = EINVAL;
 2117                         goto out;
 2118                 }
 2119         } else if (fp->f_type == DTYPE_SHM) {
 2120                 shmfd = fp->f_data;
 2121                 obj = shmfd->shm_object;
 2122                 *obj_size = shmfd->shm_size;
 2123         } else {
 2124                 error = EINVAL;
 2125                 goto out;
 2126         }
 2127 
 2128         VM_OBJECT_WLOCK(obj);
 2129         if ((obj->flags & OBJ_DEAD) != 0) {
 2130                 VM_OBJECT_WUNLOCK(obj);
 2131                 error = EBADF;
 2132                 goto out;
 2133         }
 2134 
 2135         /*
 2136          * Temporarily increase the backing VM object's reference
 2137          * count so that a forced reclamation of its vnode does not
 2138          * immediately destroy it.
 2139          */
 2140         vm_object_reference_locked(obj);
 2141         VM_OBJECT_WUNLOCK(obj);
 2142         *obj_res = obj;
 2143         *vp_res = vp;
 2144         *shmfd_res = shmfd;
 2145 
 2146 out:
 2147         if (vp != NULL)
 2148                 VOP_UNLOCK(vp, 0);
 2149         return (error);
 2150 }
 2151 
 2152 static int
 2153 kern_sendfile_getsock(struct thread *td, int s, struct file **sock_fp,
 2154     struct socket **so)
 2155 {
 2156         cap_rights_t rights;
 2157         int error;
 2158 
 2159         *sock_fp = NULL;
 2160         *so = NULL;
 2161 
 2162         /*
 2163          * The socket must be a stream socket and connected.
 2164          */
 2165         error = getsock_cap(td->td_proc->p_fd, s, cap_rights_init(&rights,
 2166             CAP_SEND), sock_fp, NULL);
 2167         if (error != 0)
 2168                 return (error);
 2169         *so = (*sock_fp)->f_data;
 2170         if ((*so)->so_type != SOCK_STREAM)
 2171                 return (EINVAL);
 2172         if (((*so)->so_state & SS_ISCONNECTED) == 0)
 2173                 return (ENOTCONN);
 2174         return (0);
 2175 }
 2176 
 2177 int
 2178 vn_sendfile(struct file *fp, int sockfd, struct uio *hdr_uio,
 2179     struct uio *trl_uio, off_t offset, size_t nbytes, off_t *sent, int flags,
 2180     int kflags, struct thread *td)
 2181 {
 2182         struct file *sock_fp;
 2183         struct vnode *vp;
 2184         struct vm_object *obj;
 2185         struct socket *so;
 2186         struct mbuf *m;
 2187         struct sf_buf *sf;
 2188         struct vm_page *pg;
 2189         struct shmfd *shmfd;
 2190         struct sendfile_sync *sfs;
 2191         struct vattr va;
 2192         off_t off, xfsize, fsbytes, sbytes, rem, obj_size;
 2193         int error, bsize, nd, hdrlen, mnw;
 2194         bool inflight_called;
 2195 
 2196         pg = NULL;
 2197         obj = NULL;
 2198         so = NULL;
 2199         m = NULL;
 2200         sfs = NULL;
 2201         fsbytes = sbytes = 0;
 2202         hdrlen = mnw = 0;
 2203         rem = nbytes;
 2204         obj_size = 0;
 2205         inflight_called = false;
 2206 
 2207         error = sendfile_getobj(td, fp, &obj, &vp, &shmfd, &obj_size, &bsize);
 2208         if (error != 0)
 2209                 return (error);
 2210         if (rem == 0)
 2211                 rem = obj_size;
 2212 
 2213         error = kern_sendfile_getsock(td, sockfd, &sock_fp, &so);
 2214         if (error != 0)
 2215                 goto out;
 2216 
 2217         /*
 2218          * Do not wait on memory allocations but return ENOMEM for
 2219          * caller to retry later.
 2220          * XXX: Experimental.
 2221          */
 2222         if (flags & SF_MNOWAIT)
 2223                 mnw = 1;
 2224 
 2225         if (flags & SF_SYNC) {
 2226                 sfs = malloc(sizeof *sfs, M_TEMP, M_WAITOK | M_ZERO);
 2227                 mtx_init(&sfs->mtx, "sendfile", NULL, MTX_DEF);
 2228                 cv_init(&sfs->cv, "sendfile");
 2229         }
 2230 
 2231 #ifdef MAC
 2232         error = mac_socket_check_send(td->td_ucred, so);
 2233         if (error != 0)
 2234                 goto out;
 2235 #endif
 2236 
 2237         /* If headers are specified copy them into mbufs. */
 2238         if (hdr_uio != NULL) {
 2239                 hdr_uio->uio_td = td;
 2240                 hdr_uio->uio_rw = UIO_WRITE;
 2241                 if (hdr_uio->uio_resid > 0) {
 2242                         /*
 2243                          * In FBSD < 5.0 the nbytes to send also included
 2244                          * the header.  If compat is specified subtract the
 2245                          * header size from nbytes.
 2246                          */
 2247                         if (kflags & SFK_COMPAT) {
 2248                                 if (nbytes > hdr_uio->uio_resid)
 2249                                         nbytes -= hdr_uio->uio_resid;
 2250                                 else
 2251                                         nbytes = 0;
 2252                         }
 2253                         m = m_uiotombuf(hdr_uio, (mnw ? M_NOWAIT : M_WAITOK),
 2254                             0, 0, 0);
 2255                         if (m == NULL) {
 2256                                 error = mnw ? EAGAIN : ENOBUFS;
 2257                                 goto out;
 2258                         }
 2259                         hdrlen = m_length(m, NULL);
 2260                 }
 2261         }
 2262 
 2263         /*
 2264          * Protect against multiple writers to the socket.
 2265          *
 2266          * XXXRW: Historically this has assumed non-interruptibility, so now
 2267          * we implement that, but possibly shouldn't.
 2268          */
 2269         (void)sblock(&so->so_snd, SBL_WAIT | SBL_NOINTR);
 2270 
 2271         /*
 2272          * Loop through the pages of the file, starting with the requested
 2273          * offset. Get a file page (do I/O if necessary), map the file page
 2274          * into an sf_buf, attach an mbuf header to the sf_buf, and queue
 2275          * it on the socket.
 2276          * This is done in two loops.  The inner loop turns as many pages
 2277          * as it can, up to available socket buffer space, without blocking
 2278          * into mbufs to have it bulk delivered into the socket send buffer.
 2279          * The outer loop checks the state and available space of the socket
 2280          * and takes care of the overall progress.
 2281          */
 2282         for (off = offset; ; ) {
 2283                 struct mbuf *mtail;
 2284                 int loopbytes;
 2285                 int space;
 2286                 int done;
 2287 
 2288                 if ((nbytes != 0 && nbytes == fsbytes) ||
 2289                     (nbytes == 0 && obj_size == fsbytes))
 2290                         break;
 2291 
 2292                 mtail = NULL;
 2293                 loopbytes = 0;
 2294                 space = 0;
 2295                 done = 0;
 2296 
 2297                 /*
 2298                  * Check the socket state for ongoing connection,
 2299                  * no errors and space in socket buffer.
 2300                  * If space is low allow for the remainder of the
 2301                  * file to be processed if it fits the socket buffer.
 2302                  * Otherwise block in waiting for sufficient space
 2303                  * to proceed, or if the socket is nonblocking, return
 2304                  * to userland with EAGAIN while reporting how far
 2305                  * we've come.
 2306                  * We wait until the socket buffer has significant free
 2307                  * space to do bulk sends.  This makes good use of file
 2308                  * system read ahead and allows packet segmentation
 2309                  * offloading hardware to take over lots of work.  If
 2310                  * we were not careful here we would send off only one
 2311                  * sfbuf at a time.
 2312                  */
 2313                 SOCKBUF_LOCK(&so->so_snd);
 2314                 if (so->so_snd.sb_lowat < so->so_snd.sb_hiwat / 2)
 2315                         so->so_snd.sb_lowat = so->so_snd.sb_hiwat / 2;
 2316 retry_space:
 2317                 if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
 2318                         error = EPIPE;
 2319                         SOCKBUF_UNLOCK(&so->so_snd);
 2320                         goto done;
 2321                 } else if (so->so_error) {
 2322                         error = so->so_error;
 2323                         so->so_error = 0;
 2324                         SOCKBUF_UNLOCK(&so->so_snd);
 2325                         goto done;
 2326                 }
 2327                 space = sbspace(&so->so_snd);
 2328                 if (space < rem &&
 2329                     (space <= 0 ||
 2330                      space < so->so_snd.sb_lowat)) {
 2331                         if (so->so_state & SS_NBIO) {
 2332                                 SOCKBUF_UNLOCK(&so->so_snd);
 2333                                 error = EAGAIN;
 2334                                 goto done;
 2335                         }
 2336                         /*
 2337                          * sbwait drops the lock while sleeping.
 2338                          * When we loop back to retry_space the
 2339                          * state may have changed and we retest
 2340                          * for it.
 2341                          */
 2342                         error = sbwait(&so->so_snd);
 2343                         /*
 2344                          * An error from sbwait usually indicates that we've
 2345                          * been interrupted by a signal. If we've sent anything
 2346                          * then return bytes sent, otherwise return the error.
 2347                          */
 2348                         if (error != 0) {
 2349                                 SOCKBUF_UNLOCK(&so->so_snd);
 2350                                 goto done;
 2351                         }
 2352                         goto retry_space;
 2353                 }
 2354                 SOCKBUF_UNLOCK(&so->so_snd);
 2355 
 2356                 /*
 2357                  * Reduce space in the socket buffer by the size of
 2358                  * the header mbuf chain.
 2359                  * hdrlen is set to 0 after the first loop.
 2360                  */
 2361                 space -= hdrlen;
 2362 
 2363                 if (vp != NULL) {
 2364                         error = vn_lock(vp, LK_SHARED);
 2365                         if (error != 0)
 2366                                 goto done;
 2367                         error = VOP_GETATTR(vp, &va, td->td_ucred);
 2368                         if (error != 0 || off >= va.va_size) {
 2369                                 VOP_UNLOCK(vp, 0);
 2370                                 goto done;
 2371                         }
 2372                         obj_size = va.va_size;
 2373                 }
 2374 
 2375                 /*
 2376                  * Loop and construct maximum sized mbuf chain to be bulk
 2377                  * dumped into socket buffer.
 2378                  */
 2379                 while (space > loopbytes) {
 2380                         vm_offset_t pgoff;
 2381                         struct mbuf *m0;
 2382 
 2383                         /*
 2384                          * Calculate the amount to transfer.
 2385                          * Not to exceed a page, the EOF,
 2386                          * or the passed in nbytes.
 2387                          */
 2388                         pgoff = (vm_offset_t)(off & PAGE_MASK);
 2389                         rem = obj_size - offset;
 2390                         if (nbytes != 0)
 2391                                 rem = omin(rem, nbytes);
 2392                         rem -= fsbytes + loopbytes;
 2393                         xfsize = omin(PAGE_SIZE - pgoff, rem);
 2394                         xfsize = omin(space - loopbytes, xfsize);
 2395                         if (xfsize <= 0) {
 2396                                 done = 1;               /* all data sent */
 2397                                 break;
 2398                         }
 2399 
 2400                         /*
 2401                          * Attempt to look up the page.  Allocate
 2402                          * if not found or wait and loop if busy.
 2403                          */
 2404                         if (m != NULL)
 2405                                 nd = EAGAIN; /* send what we already got */
 2406                         else if ((flags & SF_NODISKIO) != 0)
 2407                                 nd = EBUSY;
 2408                         else
 2409                                 nd = 0;
 2410                         error = sendfile_readpage(obj, vp, nd, off,
 2411                             xfsize, bsize, td, &pg);
 2412                         if (error != 0) {
 2413                                 if (error == EAGAIN)
 2414                                         error = 0;      /* not a real error */
 2415                                 break;
 2416                         }
 2417 
 2418                         /*
 2419                          * Get a sendfile buf.  When allocating the
 2420                          * first buffer for mbuf chain, we usually
 2421                          * wait as long as necessary, but this wait
 2422                          * can be interrupted.  For consequent
 2423                          * buffers, do not sleep, since several
 2424                          * threads might exhaust the buffers and then
 2425                          * deadlock.
 2426                          */
 2427                         sf = sf_buf_alloc(pg, (mnw || m != NULL) ? SFB_NOWAIT :
 2428                             SFB_CATCH);
 2429                         if (sf == NULL) {
 2430                                 SFSTAT_INC(sf_allocfail);
 2431                                 vm_page_lock(pg);
 2432                                 vm_page_unwire(pg, 0);
 2433                                 KASSERT(pg->object != NULL,
 2434                                     ("%s: object disappeared", __func__));
 2435                                 vm_page_unlock(pg);
 2436                                 if (m == NULL)
 2437                                         error = (mnw ? EAGAIN : EINTR);
 2438                                 break;
 2439                         }
 2440 
 2441                         /*
 2442                          * Get an mbuf and set it up as having
 2443                          * external storage.
 2444                          */
 2445                         m0 = m_get((mnw ? M_NOWAIT : M_WAITOK), MT_DATA);
 2446                         if (m0 == NULL) {
 2447                                 error = (mnw ? EAGAIN : ENOBUFS);
 2448                                 (void)sf_buf_mext(NULL, NULL, sf);
 2449                                 break;
 2450                         }
 2451                         if (m_extadd(m0, (caddr_t )sf_buf_kva(sf), PAGE_SIZE,
 2452                             sf_buf_mext, sfs, sf, M_RDONLY, EXT_SFBUF,
 2453                             (mnw ? M_NOWAIT : M_WAITOK)) != 0) {
 2454                                 error = (mnw ? EAGAIN : ENOBUFS);
 2455                                 (void)sf_buf_mext(NULL, NULL, sf);
 2456                                 m_freem(m0);
 2457                                 break;
 2458                         }
 2459                         m0->m_data = (char *)sf_buf_kva(sf) + pgoff;
 2460                         m0->m_len = xfsize;
 2461 
 2462                         /* Append to mbuf chain. */
 2463                         if (mtail != NULL)
 2464                                 mtail->m_next = m0;
 2465                         else if (m != NULL)
 2466                                 m_last(m)->m_next = m0;
 2467                         else
 2468                                 m = m0;
 2469                         mtail = m0;
 2470 
 2471                         /* Keep track of bits processed. */
 2472                         loopbytes += xfsize;
 2473                         off += xfsize;
 2474 
 2475                         if (sfs != NULL) {
 2476                                 mtx_lock(&sfs->mtx);
 2477                                 sfs->count++;
 2478                                 mtx_unlock(&sfs->mtx);
 2479                         }
 2480                 }
 2481 
 2482                 if (vp != NULL)
 2483                         VOP_UNLOCK(vp, 0);
 2484 
 2485                 /* Add the buffer chain to the socket buffer. */
 2486                 if (m != NULL) {
 2487                         int mlen, err;
 2488 
 2489                         mlen = m_length(m, NULL);
 2490                         SOCKBUF_LOCK(&so->so_snd);
 2491                         if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
 2492                                 error = EPIPE;
 2493                                 SOCKBUF_UNLOCK(&so->so_snd);
 2494                                 goto done;
 2495                         }
 2496                         SOCKBUF_UNLOCK(&so->so_snd);
 2497                         CURVNET_SET(so->so_vnet);
 2498                         /* Avoid error aliasing. */
 2499                         err = (*so->so_proto->pr_usrreqs->pru_send)
 2500                                     (so, 0, m, NULL, NULL, td);
 2501                         CURVNET_RESTORE();
 2502                         if (err == 0) {
 2503                                 /*
 2504                                  * We need two counters to get the
 2505                                  * file offset and nbytes to send
 2506                                  * right:
 2507                                  * - sbytes contains the total amount
 2508                                  *   of bytes sent, including headers.
 2509                                  * - fsbytes contains the total amount
 2510                                  *   of bytes sent from the file.
 2511                                  */
 2512                                 sbytes += mlen;
 2513                                 fsbytes += mlen;
 2514                                 if (hdrlen) {
 2515                                         fsbytes -= hdrlen;
 2516                                         hdrlen = 0;
 2517                                 }
 2518                         } else if (error == 0)
 2519                                 error = err;
 2520                         m = NULL;       /* pru_send always consumes */
 2521                 }
 2522 
 2523                 /* Quit outer loop on error or when we're done. */
 2524                 if (done)
 2525                         break;
 2526                 if (error != 0)
 2527                         goto done;
 2528         }
 2529 
 2530         /*
 2531          * Send trailers. Wimp out and use writev(2).
 2532          */
 2533         if (trl_uio != NULL) {
 2534                 sbunlock(&so->so_snd);
 2535                 error = kern_writev(td, sockfd, trl_uio);
 2536                 if (error == 0)
 2537                         sbytes += td->td_retval[0];
 2538                 goto out;
 2539         }
 2540 
 2541 done:
 2542         sbunlock(&so->so_snd);
 2543 out:
 2544         /*
 2545          * If there was no error we have to clear td->td_retval[0]
 2546          * because it may have been set by writev.
 2547          */
 2548         if (error == 0) {
 2549                 td->td_retval[0] = 0;
 2550         }
 2551         if (sent != NULL) {
 2552                 copyout(&sbytes, sent, sizeof(off_t));
 2553         }
 2554         if (obj != NULL)
 2555                 vm_object_deallocate(obj);
 2556         if (so)
 2557                 fdrop(sock_fp, td);
 2558         if (m)
 2559                 m_freem(m);
 2560 
 2561         if (sfs != NULL) {
 2562                 mtx_lock(&sfs->mtx);
 2563                 if (sfs->count != 0)
 2564                         cv_wait(&sfs->cv, &sfs->mtx);
 2565                 KASSERT(sfs->count == 0, ("sendfile sync still busy"));
 2566                 cv_destroy(&sfs->cv);
 2567                 mtx_destroy(&sfs->mtx);
 2568                 free(sfs, M_TEMP);
 2569         }
 2570 
 2571         if (error == ERESTART)
 2572                 error = EINTR;
 2573 
 2574         return (error);
 2575 }
 2576 
 2577 /*
 2578  * SCTP syscalls.
 2579  * Functionality only compiled in if SCTP is defined in the kernel Makefile,
 2580  * otherwise all return EOPNOTSUPP.
 2581  * XXX: We should make this loadable one day.
 2582  */
 2583 int
 2584 sys_sctp_peeloff(td, uap)
 2585         struct thread *td;
 2586         struct sctp_peeloff_args /* {
 2587                 int     sd;
 2588                 caddr_t name;
 2589         } */ *uap;
 2590 {
 2591 #if (defined(INET) || defined(INET6)) && defined(SCTP)
 2592         struct file *nfp = NULL;
 2593         struct socket *head, *so;
 2594         cap_rights_t rights;
 2595         u_int fflag;
 2596         int error, fd;
 2597 
 2598         AUDIT_ARG_FD(uap->sd);
 2599         error = fgetsock(td, uap->sd, cap_rights_init(&rights, CAP_PEELOFF),
 2600             &head, &fflag);
 2601         if (error != 0)
 2602                 goto done2;
 2603         if (head->so_proto->pr_protocol != IPPROTO_SCTP) {
 2604                 error = EOPNOTSUPP;
 2605                 goto done;
 2606         }
 2607         error = sctp_can_peel_off(head, (sctp_assoc_t)uap->name);
 2608         if (error != 0)
 2609                 goto done;
 2610         /*
 2611          * At this point we know we do have a assoc to pull
 2612          * we proceed to get the fd setup. This may block
 2613          * but that is ok.
 2614          */
 2615 
 2616         error = falloc(td, &nfp, &fd, 0);
 2617         if (error != 0)
 2618                 goto done;
 2619         td->td_retval[0] = fd;
 2620 
 2621         CURVNET_SET(head->so_vnet);
 2622         so = sonewconn(head, SS_ISCONNECTED);
 2623         if (so == NULL) {
 2624                 error = ENOMEM;
 2625                 goto noconnection;
 2626         }
 2627         /*
 2628          * Before changing the flags on the socket, we have to bump the
 2629          * reference count.  Otherwise, if the protocol calls sofree(),
 2630          * the socket will be released due to a zero refcount.
 2631          */
 2632         SOCK_LOCK(so);
 2633         soref(so);                      /* file descriptor reference */
 2634         SOCK_UNLOCK(so);
 2635 
 2636         ACCEPT_LOCK();
 2637 
 2638         TAILQ_REMOVE(&head->so_comp, so, so_list);
 2639         head->so_qlen--;
 2640         so->so_state |= (head->so_state & SS_NBIO);
 2641         so->so_state &= ~SS_NOFDREF;
 2642         so->so_qstate &= ~SQ_COMP;
 2643         so->so_head = NULL;
 2644         ACCEPT_UNLOCK();
 2645         finit(nfp, fflag, DTYPE_SOCKET, so, &socketops);
 2646         error = sctp_do_peeloff(head, so, (sctp_assoc_t)uap->name);
 2647         if (error != 0)
 2648                 goto noconnection;
 2649         if (head->so_sigio != NULL)
 2650                 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
 2651 
 2652 noconnection:
 2653         /*
 2654          * close the new descriptor, assuming someone hasn't ripped it
 2655          * out from under us.
 2656          */
 2657         if (error != 0)
 2658                 fdclose(td->td_proc->p_fd, nfp, fd, td);
 2659 
 2660         /*
 2661          * Release explicitly held references before returning.
 2662          */
 2663         CURVNET_RESTORE();
 2664 done:
 2665         if (nfp != NULL)
 2666                 fdrop(nfp, td);
 2667         fputsock(head);
 2668 done2:
 2669         return (error);
 2670 #else  /* SCTP */
 2671         return (EOPNOTSUPP);
 2672 #endif /* SCTP */
 2673 }
 2674 
 2675 int
 2676 sys_sctp_generic_sendmsg (td, uap)
 2677         struct thread *td;
 2678         struct sctp_generic_sendmsg_args /* {
 2679                 int sd,
 2680                 caddr_t msg,
 2681                 int mlen,
 2682                 caddr_t to,
 2683                 __socklen_t tolen,
 2684                 struct sctp_sndrcvinfo *sinfo,
 2685                 int flags
 2686         } */ *uap;
 2687 {
 2688 #if (defined(INET) || defined(INET6)) && defined(SCTP)
 2689         struct sctp_sndrcvinfo sinfo, *u_sinfo = NULL;
 2690         struct socket *so;
 2691         struct file *fp = NULL;
 2692         struct sockaddr *to = NULL;
 2693 #ifdef KTRACE
 2694         struct uio *ktruio = NULL;
 2695 #endif
 2696         struct uio auio;
 2697         struct iovec iov[1];
 2698         cap_rights_t rights;
 2699         int error = 0, len;
 2700 
 2701         if (uap->sinfo != NULL) {
 2702                 error = copyin(uap->sinfo, &sinfo, sizeof (sinfo));
 2703                 if (error != 0)
 2704                         return (error);
 2705                 u_sinfo = &sinfo;
 2706         }
 2707 
 2708         cap_rights_init(&rights, CAP_SEND);
 2709         if (uap->tolen != 0) {
 2710                 error = getsockaddr(&to, uap->to, uap->tolen);
 2711                 if (error != 0) {
 2712                         to = NULL;
 2713                         goto sctp_bad2;
 2714                 }
 2715                 cap_rights_set(&rights, CAP_CONNECT);
 2716         }
 2717 
 2718         AUDIT_ARG_FD(uap->sd);
 2719         error = getsock_cap(td->td_proc->p_fd, uap->sd, &rights, &fp, NULL);
 2720         if (error != 0)
 2721                 goto sctp_bad;
 2722 #ifdef KTRACE
 2723         if (to && (KTRPOINT(td, KTR_STRUCT)))
 2724                 ktrsockaddr(to);
 2725 #endif
 2726 
 2727         iov[0].iov_base = uap->msg;
 2728         iov[0].iov_len = uap->mlen;
 2729 
 2730         so = (struct socket *)fp->f_data;
 2731         if (so->so_proto->pr_protocol != IPPROTO_SCTP) {
 2732                 error = EOPNOTSUPP;
 2733                 goto sctp_bad;
 2734         }
 2735 #ifdef MAC
 2736         error = mac_socket_check_send(td->td_ucred, so);
 2737         if (error != 0)
 2738                 goto sctp_bad;
 2739 #endif /* MAC */
 2740 
 2741         auio.uio_iov =  iov;
 2742         auio.uio_iovcnt = 1;
 2743         auio.uio_segflg = UIO_USERSPACE;
 2744         auio.uio_rw = UIO_WRITE;
 2745         auio.uio_td = td;
 2746         auio.uio_offset = 0;                    /* XXX */
 2747         auio.uio_resid = 0;
 2748         len = auio.uio_resid = uap->mlen;
 2749         CURVNET_SET(so->so_vnet);
 2750         error = sctp_lower_sosend(so, to, &auio, (struct mbuf *)NULL,
 2751             (struct mbuf *)NULL, uap->flags, u_sinfo, td);
 2752         CURVNET_RESTORE();
 2753         if (error != 0) {
 2754                 if (auio.uio_resid != len && (error == ERESTART ||
 2755                     error == EINTR || error == EWOULDBLOCK))
 2756                         error = 0;
 2757                 /* Generation of SIGPIPE can be controlled per socket. */
 2758                 if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
 2759                     !(uap->flags & MSG_NOSIGNAL)) {
 2760                         PROC_LOCK(td->td_proc);
 2761                         tdsignal(td, SIGPIPE);
 2762                         PROC_UNLOCK(td->td_proc);
 2763                 }
 2764         }
 2765         if (error == 0)
 2766                 td->td_retval[0] = len - auio.uio_resid;
 2767 #ifdef KTRACE
 2768         if (ktruio != NULL) {
 2769                 ktruio->uio_resid = td->td_retval[0];
 2770                 ktrgenio(uap->sd, UIO_WRITE, ktruio, error);
 2771         }
 2772 #endif /* KTRACE */
 2773 sctp_bad:
 2774         if (fp != NULL)
 2775                 fdrop(fp, td);
 2776 sctp_bad2:
 2777         free(to, M_SONAME);
 2778         return (error);
 2779 #else  /* SCTP */
 2780         return (EOPNOTSUPP);
 2781 #endif /* SCTP */
 2782 }
 2783 
 2784 int
 2785 sys_sctp_generic_sendmsg_iov(td, uap)
 2786         struct thread *td;
 2787         struct sctp_generic_sendmsg_iov_args /* {
 2788                 int sd,
 2789                 struct iovec *iov,
 2790                 int iovlen,
 2791                 caddr_t to,
 2792                 __socklen_t tolen,
 2793                 struct sctp_sndrcvinfo *sinfo,
 2794                 int flags
 2795         } */ *uap;
 2796 {
 2797 #if (defined(INET) || defined(INET6)) && defined(SCTP)
 2798         struct sctp_sndrcvinfo sinfo, *u_sinfo = NULL;
 2799         struct socket *so;
 2800         struct file *fp = NULL;
 2801         struct sockaddr *to = NULL;
 2802 #ifdef KTRACE
 2803         struct uio *ktruio = NULL;
 2804 #endif
 2805         struct uio auio;
 2806         struct iovec *iov, *tiov;
 2807         cap_rights_t rights;
 2808         ssize_t len;
 2809         int error, i;
 2810 
 2811         if (uap->sinfo != NULL) {
 2812                 error = copyin(uap->sinfo, &sinfo, sizeof (sinfo));
 2813                 if (error != 0)
 2814                         return (error);
 2815                 u_sinfo = &sinfo;
 2816         }
 2817         cap_rights_init(&rights, CAP_SEND);
 2818         if (uap->tolen != 0) {
 2819                 error = getsockaddr(&to, uap->to, uap->tolen);
 2820                 if (error != 0) {
 2821                         to = NULL;
 2822                         goto sctp_bad2;
 2823                 }
 2824                 cap_rights_set(&rights, CAP_CONNECT);
 2825         }
 2826 
 2827         AUDIT_ARG_FD(uap->sd);
 2828         error = getsock_cap(td->td_proc->p_fd, uap->sd, &rights, &fp, NULL);
 2829         if (error != 0)
 2830                 goto sctp_bad1;
 2831 
 2832 #ifdef COMPAT_FREEBSD32
 2833         if (SV_CURPROC_FLAG(SV_ILP32))
 2834                 error = freebsd32_copyiniov((struct iovec32 *)uap->iov,
 2835                     uap->iovlen, &iov, EMSGSIZE);
 2836         else
 2837 #endif
 2838                 error = copyiniov(uap->iov, uap->iovlen, &iov, EMSGSIZE);
 2839         if (error != 0)
 2840                 goto sctp_bad1;
 2841 #ifdef KTRACE
 2842         if (to && (KTRPOINT(td, KTR_STRUCT)))
 2843                 ktrsockaddr(to);
 2844 #endif
 2845 
 2846         so = (struct socket *)fp->f_data;
 2847         if (so->so_proto->pr_protocol != IPPROTO_SCTP) {
 2848                 error = EOPNOTSUPP;
 2849                 goto sctp_bad;
 2850         }
 2851 #ifdef MAC
 2852         error = mac_socket_check_send(td->td_ucred, so);
 2853         if (error != 0)
 2854                 goto sctp_bad;
 2855 #endif /* MAC */
 2856 
 2857         auio.uio_iov = iov;
 2858         auio.uio_iovcnt = uap->iovlen;
 2859         auio.uio_segflg = UIO_USERSPACE;
 2860         auio.uio_rw = UIO_WRITE;
 2861         auio.uio_td = td;
 2862         auio.uio_offset = 0;                    /* XXX */
 2863         auio.uio_resid = 0;
 2864         tiov = iov;
 2865         for (i = 0; i <uap->iovlen; i++, tiov++) {
 2866                 if ((auio.uio_resid += tiov->iov_len) < 0) {
 2867                         error = EINVAL;
 2868                         goto sctp_bad;
 2869                 }
 2870         }
 2871         len = auio.uio_resid;
 2872         CURVNET_SET(so->so_vnet);
 2873         error = sctp_lower_sosend(so, to, &auio,
 2874                     (struct mbuf *)NULL, (struct mbuf *)NULL,
 2875                     uap->flags, u_sinfo, td);
 2876         CURVNET_RESTORE();
 2877         if (error != 0) {
 2878                 if (auio.uio_resid != len && (error == ERESTART ||
 2879                     error == EINTR || error == EWOULDBLOCK))
 2880                         error = 0;
 2881                 /* Generation of SIGPIPE can be controlled per socket */
 2882                 if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
 2883                     !(uap->flags & MSG_NOSIGNAL)) {
 2884                         PROC_LOCK(td->td_proc);
 2885                         tdsignal(td, SIGPIPE);
 2886                         PROC_UNLOCK(td->td_proc);
 2887                 }
 2888         }
 2889         if (error == 0)
 2890                 td->td_retval[0] = len - auio.uio_resid;
 2891 #ifdef KTRACE
 2892         if (ktruio != NULL) {
 2893                 ktruio->uio_resid = td->td_retval[0];
 2894                 ktrgenio(uap->sd, UIO_WRITE, ktruio, error);
 2895         }
 2896 #endif /* KTRACE */
 2897 sctp_bad:
 2898         free(iov, M_IOV);
 2899 sctp_bad1:
 2900         if (fp != NULL)
 2901                 fdrop(fp, td);
 2902 sctp_bad2:
 2903         free(to, M_SONAME);
 2904         return (error);
 2905 #else  /* SCTP */
 2906         return (EOPNOTSUPP);
 2907 #endif /* SCTP */
 2908 }
 2909 
 2910 int
 2911 sys_sctp_generic_recvmsg(td, uap)
 2912         struct thread *td;
 2913         struct sctp_generic_recvmsg_args /* {
 2914                 int sd,
 2915                 struct iovec *iov,
 2916                 int iovlen,
 2917                 struct sockaddr *from,
 2918                 __socklen_t *fromlenaddr,
 2919                 struct sctp_sndrcvinfo *sinfo,
 2920                 int *msg_flags
 2921         } */ *uap;
 2922 {
 2923 #if (defined(INET) || defined(INET6)) && defined(SCTP)
 2924         uint8_t sockbufstore[256];
 2925         struct uio auio;
 2926         struct iovec *iov, *tiov;
 2927         struct sctp_sndrcvinfo sinfo;
 2928         struct socket *so;
 2929         struct file *fp = NULL;
 2930         struct sockaddr *fromsa;
 2931         cap_rights_t rights;
 2932 #ifdef KTRACE
 2933         struct uio *ktruio = NULL;
 2934 #endif
 2935         ssize_t len;
 2936         int error, fromlen, i, msg_flags;
 2937 
 2938         AUDIT_ARG_FD(uap->sd);
 2939         error = getsock_cap(td->td_proc->p_fd, uap->sd,
 2940             cap_rights_init(&rights, CAP_RECV), &fp, NULL);
 2941         if (error != 0)
 2942                 return (error);
 2943 #ifdef COMPAT_FREEBSD32
 2944         if (SV_CURPROC_FLAG(SV_ILP32))
 2945                 error = freebsd32_copyiniov((struct iovec32 *)uap->iov,
 2946                     uap->iovlen, &iov, EMSGSIZE);
 2947         else
 2948 #endif
 2949                 error = copyiniov(uap->iov, uap->iovlen, &iov, EMSGSIZE);
 2950         if (error != 0)
 2951                 goto out1;
 2952 
 2953         so = fp->f_data;
 2954         if (so->so_proto->pr_protocol != IPPROTO_SCTP) {
 2955                 error = EOPNOTSUPP;
 2956                 goto out;
 2957         }
 2958 #ifdef MAC
 2959         error = mac_socket_check_receive(td->td_ucred, so);
 2960         if (error != 0)
 2961                 goto out;
 2962 #endif /* MAC */
 2963 
 2964         if (uap->fromlenaddr != NULL) {
 2965                 error = copyin(uap->fromlenaddr, &fromlen, sizeof (fromlen));
 2966                 if (error != 0)
 2967                         goto out;
 2968         } else {
 2969                 fromlen = 0;
 2970         }
 2971         if (uap->msg_flags) {
 2972                 error = copyin(uap->msg_flags, &msg_flags, sizeof (int));
 2973                 if (error != 0)
 2974                         goto out;
 2975         } else {
 2976                 msg_flags = 0;
 2977         }
 2978         auio.uio_iov = iov;
 2979         auio.uio_iovcnt = uap->iovlen;
 2980         auio.uio_segflg = UIO_USERSPACE;
 2981         auio.uio_rw = UIO_READ;
 2982         auio.uio_td = td;
 2983         auio.uio_offset = 0;                    /* XXX */
 2984         auio.uio_resid = 0;
 2985         tiov = iov;
 2986         for (i = 0; i <uap->iovlen; i++, tiov++) {
 2987                 if ((auio.uio_resid += tiov->iov_len) < 0) {
 2988                         error = EINVAL;
 2989                         goto out;
 2990                 }
 2991         }
 2992         len = auio.uio_resid;
 2993         fromsa = (struct sockaddr *)sockbufstore;
 2994 
 2995 #ifdef KTRACE
 2996         if (KTRPOINT(td, KTR_GENIO))
 2997                 ktruio = cloneuio(&auio);
 2998 #endif /* KTRACE */
 2999         memset(&sinfo, 0, sizeof(struct sctp_sndrcvinfo));
 3000         CURVNET_SET(so->so_vnet);
 3001         error = sctp_sorecvmsg(so, &auio, (struct mbuf **)NULL,
 3002                     fromsa, fromlen, &msg_flags,
 3003                     (struct sctp_sndrcvinfo *)&sinfo, 1);
 3004         CURVNET_RESTORE();
 3005         if (error != 0) {
 3006                 if (auio.uio_resid != len && (error == ERESTART ||
 3007                     error == EINTR || error == EWOULDBLOCK))
 3008                         error = 0;
 3009         } else {
 3010                 if (uap->sinfo)
 3011                         error = copyout(&sinfo, uap->sinfo, sizeof (sinfo));
 3012         }
 3013 #ifdef KTRACE
 3014         if (ktruio != NULL) {
 3015                 ktruio->uio_resid = len - auio.uio_resid;
 3016                 ktrgenio(uap->sd, UIO_READ, ktruio, error);
 3017         }
 3018 #endif /* KTRACE */
 3019         if (error != 0)
 3020                 goto out;
 3021         td->td_retval[0] = len - auio.uio_resid;
 3022 
 3023         if (fromlen && uap->from) {
 3024                 len = fromlen;
 3025                 if (len <= 0 || fromsa == 0)
 3026                         len = 0;
 3027                 else {
 3028                         len = MIN(len, fromsa->sa_len);
 3029                         error = copyout(fromsa, uap->from, (size_t)len);
 3030                         if (error != 0)
 3031                                 goto out;
 3032                 }
 3033                 error = copyout(&len, uap->fromlenaddr, sizeof (socklen_t));
 3034                 if (error != 0)
 3035                         goto out;
 3036         }
 3037 #ifdef KTRACE
 3038         if (KTRPOINT(td, KTR_STRUCT))
 3039                 ktrsockaddr(fromsa);
 3040 #endif
 3041         if (uap->msg_flags) {
 3042                 error = copyout(&msg_flags, uap->msg_flags, sizeof (int));
 3043                 if (error != 0)
 3044                         goto out;
 3045         }
 3046 out:
 3047         free(iov, M_IOV);
 3048 out1:
 3049         if (fp != NULL)
 3050                 fdrop(fp, td);
 3051 
 3052         return (error);
 3053 #else  /* SCTP */
 3054         return (EOPNOTSUPP);
 3055 #endif /* SCTP */
 3056 }

Cache object: 5134acbd5f5d540db8f098cf81d452da


[ 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.